papers in

MICROBIAL GENETICS

bacteria and bacterial viruses

selected by JOSHUA LEDERBERG

Marine Biological Laboratory

P_ vf H Aug, 1958

Accession No.

66750

University of "Wisconsin Press
Uven by Madison, Wisconsin

Place

^^^^

nj

WHOI

LO

nj

D

CO •

o

i-=)

SSSSt

o

BBSS

m

;

o

□

— — —

^=

MICROBIAL GENETICS

papers in

MICROBIAL GENETICS

bacteria and bacterial viruses

selected by JOSHUA LEDERBERG

UNIVERSITY OF WISCONSIN PRESS • MADISON — 1952

COPYRIGHT 1951 BY THE
REGENTS OF THE UNIVERSITY OF WISCONSIN

SECOND PRINTING, 1952

PRINTED IN THE UNITED STATES OF AMERICA BY
CUSHING-MALLOY, INC. — ANN ARBOR, MICHIGAN

PREFACE

The first purpose of this book is to supplement a lecture course on microbial
genetics with original readings in a form most readily available to the student.
At most universities, the papers included in this collection will be represented
by single copies of the original journals. Thus, it is usually difficult for any
number of students to read the papers at all, and impossible for anyone (includ-
ing the instructor) to read and perhaps annotate them at leisure. This volume
is not intended to include all the reading on this subject which should be
required for a course. I hope, however, that it will make available most of
the papers requiring the most general or the most intensive study.

The scope of the collection must obviously have some limits in size and
cost. These have led to the restriction to bacteria and bacterial viruses, although
consideration of fungi, protozoa, algae, and viruses is essential to a well-rounded
course, even to one organized, as at Wisconsin, primarily for students majoring
in bacteriology. Within this field, there are a great many important papers
which have necessarily been omitted. No apologies need be offered for a selection
which must be largely arbitrary. The papers selected were those which seemed
to illustrate best certain concepts and methodologies for the benefit of a course
in microbial genetics. In some choices, weight had to be given to other
circumstances such as the length and availability of alternative papers. It must
be emphasized that these selections do not reflect historical priority or scientific
worth as such. A bibliographic list is included to help remedy the unavoidable
deficiencies of the collection.

This book could not have been produced without the magnanimous coopera-
tion of the publishers and copyright holders of the original papers, to whom
specific acknowledgment is made elsewhere. Special thanks are due to the
authors who have been very helpful, especially in providing scarce reprints for
use as copy. I must also record an obligation to my colleagues who did their
best to advise me on the proper scope and contents of this collection.

j. L.
Madison, April i, 1951

CONTENTS

Introduction by Joshua Lederberg ix

Mutations of Bacteria from Virus Sensitivity to Virus Resistance, by
S. E. Luria and M. Delbruck 3

The Distribution of the Numbers of Mutants in Bacterial

Populations, by D. E. Lea and C. A. Coulson 24

Origin of Bacterial Variants, by Howard B. Newcombe 46

Delayed Phenotypic Expression of Spontaneous Mutations in Escherichia

coli, by Howard B. Newcombe 49

Origin of Bacterial Resistance to Antibiotics, by M. Demerec 79

Chloromycetin Resistance in E. coli, a Case of Quantitative Inheritance

in Bacteria, by L. L. Cavalli and G. A. Maccacaro 91

Studies on Nutritionally Deficient Bacterial Mutants Isolated by Means

of Penicillin, by Bernard D. Davis 94

The Demonstration of Non-specific Components in Salmonella paratyphi

A by Induced Variation, by D. W. Bruner and P. R. Edwards 104

The Natural Occurrence of Phase 2 of Salmonella paratyphi A,

by P. R. Edwards, L. A. Barnes, and Mary C. Babcock i 16

Changes Induced in the O Antigens of Salmonella, by D. W. Bruner

and P. R. Edwards 118

Mutations in Escherichia coli Induced by Chemical Agents,

by Evelyn M. Witkin 119

The Effect of Metabolites upon Growth and Variation of Brucella
abortus, by Robert J. Goodlow, Leonard A. Mika, and Werner
Braun 133

Gene Recombination and Linked Segregations in Escherichia coli,

by Joshua Lederberg 143

Aberrant Heterozygotes in Escherichia coli, by Joshua Lederberg 164

CONTENTS

Cytological Observations on Bad. coli, Proteus vulgaris and Various
Aerobic Spore-Forming Bacteria with Special Reference to the Nuclear
Structures, by C. F. Robinow 17 2

Studies on the Chemical Nature of the Substance Inducing Transforma-
tion of Pneumococcal Types, by Oswald T. Avery, Colin M.
MacLeod, and Maclyn McCarty 187

Induced Lysogenicity and Mutation of Bacteriophage within Lysogenic

Bacteria, by F. M. Burnet and Dora Lush 210

Mutations of Bacterial Viruses Affecting Their Host Range,

by S. E. Luria 222

Genetic Recombination between Host-Range and Plaque-Type Mutants
of Bacteriophage in Single Bacterial Cells, by A. D. Hershey and
Raquel Rotman 238

Genetic Recombinations Leading to Production of Active Bacteriophage
from Ultraviolet Inactivated Bacteriophage Particles, by S. E. Luria

AND R. DULBECCO 266

Bibliography 299

INTRODUCTION

Mutation and adaptation. — The problem of organic adaptation in bacteria
has been resolved into two points of view which are closely analogous to "neo
Darwinian" and "Lamarckian" concepts of organic evolution, respectively.
The first holds that adaptive changes occur spontaneously, as sporadic mutations,
not in any specific relationship to environmental conditions, and that natural
selection functions to fix the best adapted genotypes. The Lamarckian viewpoint
suggests that the adaptive mutation is itself directed by the environment, and
that natural selection plays a subsidiary role. Many of the papers reprinted
show experiments which support the spontaneous mutation hypothesis ( i, 3, 5) .
One (6) deals with the end result of bacterial adaptation to Chloromycetin
and also argues for spontaneous mutations with accumulative effect by
separating the components by recombination. There is no representative here
of the alternative point of view which is defended strongly by Hinshelwood and
his associates (28, 67, 70). In considering this problem it is essential to keep
in mind physiological adaptations which are nonheritable, but whose interplay
with heritable changes may trap the unwary (97, 120, 45-Lwoff). General
discussions of this problem are included in most of the review articles (32, 54,
81, 82, 87, isi, 129).

Mutations for resistance to bacteriophage had been considered some years
before Luria and Delbriick's paper (1), notably by Burnet (34). This author
drew the same conclusions concerning the spontaneity of the resistance muta-
tions, mainly through his ability to pick out resistant mutants by their colonial
morphology without the use of phage as a selective agent. Much of the current
work on E. coli phages (43, 44, 21) involves a set of 7 virus strains described
as Ti through T7. Mutation rates for resistance to these phages have been
recorded, together with useful information on cross-resistance patterns (48).

The physiological study of mutant characters is known as phenogenetics,
in contrast to formal or cryptogenetics which deals directly with the mechanism
of hereditary transmission. One would reasonably expect that a gene mutation
would require a period of time to work its effects on the phenotype or outward
behavior of the organism. This lag in bacterial mutation effects (phenotypic
or phenomic lag) was first noticed directly with phage-resistance mutations

The citations in the Introduction refer to the Bibliography, pp. 299-303.

JOSHUA LEDERBERG

induced by radiations (45-Demerec). Later, Newcombe postulated a similar
lag for spontaneous mutations to account for discrepancies in the estimation of
mutation rates by different methods, according to whether they counted
mutant individuals or mutant clones (4). In both of these instances, however,
the interpretation is complicated by the cytological structure of the bacterial
cell (15) which speaks for the concurrence of several nuclei (and therefore of
several replicate gene sets) in each cell. Recessive mutations such as phage
or drug resistance would require for their phenotypic expression that mutant
and nonmutant nuclei be segregated, so that there would also be a segregation
lag in the genesis of the mutant phenotype (103, 14, 47). The description of
phenomic lag in the development of nutritionally sufficient "back-mutations"
induced by ultra-violet light (UV) (7) avoids this difficulty, for these mutations
are presumably dominant, and would not require segregation for their expres-
sion. A final caution which must be entered is that we cannot be entirely certain
that the mutation-producing event within the cell is coincidental with the
experimental irradiation (mutation delay hypothesis) (26, 49).

Mutant characters. — Something is known of the phenogenetics of phage
resistance. The extraction of sensitive and resistant cultures results in prepara-
tions differing in the ability to neutralize the phage "in vitro." This argues
strongly for the concept that phage sensitivity is correlated with the presence
of a specific receptor substance (44, 61 ) . Another metabolic correlate of phage
resistance is less explicable. In certain specific combinations of bacterium and
phage, the resistance mutation simultaneously leads to new nutritional require-
ments, (often for tryptophane or proline). Unfortunately, this correlation has
not been found in strains which can be studied by recombination, so that it
is not certain whether a simple genetic change causes the two effects, (23, 139,
45-Luria) .

Phage resistant mutations play such a large part in genetic study because
of the ease with which they can be selected out of large populations, simply by
adding suspensions of the virus under appropriate conditions. This facility is
shared by antibiotics and other antibacterials as selective agents. The advantage
that the "drug" is not another biological system whose variability must be
watched is counterbalanced by the fact that resistance to most chemicals is a
quantitative rather than an all-or-none phenomenon, and that a great many
genes may be involved. The probable reason for this is that antibiotics interfere
with key metabolic steps that are linked to a wide diversity of other reactions,
so that any one of a number of biochemical changes may influence the growth
response to an inhibitor. The initial reaction between phages and cells is, on the
other hand, directly associated with the presence of a receptor substance with a
high order of specificity.

Current interest in drug-resistance mutations is largely motivated by their
importance in limiting the effectiveness of chemotherapy (95), but they have
provided interesting material for more theoretical genetic studies as well. In

INTRODUCTION

most cases, resistance may be developed in a series of individually small steps,
speaking for the cumulative action of a number of gene mutations as postulated
above. The evidence for this comes primarily from kinetic studies of increasing
resistance under selection (5, 107, 135) . It has been confirmed by direct crossing
experiments involving Chloromycetin resistant mutants of E. coli (6). Progeny
of crosses between a sensitive and a fully-adapted (many-step) resistant parent
showed a segregation of many intermediate grades of resistance, representing
the reassortment of the sensitive and resistant genes in a variety of combinations.

Streptomycin-resistance shows some unique features in contrast to the
agents just summarized. In many species, a mutation conferring full resistance
occurs at a rate, about io~ 10 per fission, which overshadows the smaller step
mutations characteristic of resistance to other agents (5, 28, 100, 135). This
low rate is perhaps the smallest mutation rate to be accurately measured in
any organism. A more striking oddity is the mutation which over-adapts the cell
to streptomycin so that the resistant mutant is dependent upon streptomycin
for growth (95, 100). The function of streptomycin is possibly to regulate an
over-expanded enzyme system, analogous to the precedent of a sulfonamide-
dependent mutant of Neurospora (141).

Amino acids and vitamins are not commonly thought of as antibiotics but
it has long been known that they sometimes interfere with, rather than promote,
bacterial growth (78, in, 117-Snell, 140). The very ubiquity of these com-
pounds imparts a special interest to them as possible natural regulators. Of equal
interest is the correlation between sensitivity to amino acids and virulence in
Salmonella and Brucella (12, 109) which may reflect a hitherto unsuspected
general principle.

It is likely that a biochemical basis will be found for the effects of all types
of mutations (29), but microorganisms have been especially prolific in the
production of mutants with overt effects on metabolism. For this reason, micro-
bial and biochemical genetics are intimately associated and often confused.
Gene mutations affecting anabolic processes are usually detected as nutritional
or auxotrophic mutants, whose growth depends upon an external supply of
the missing metabolite. Auxotrophs have been used for the exploration of many
biosynthetic pathways, which are remarkably similar in bacteria, molds, and
mammals. Davis' paper (7) outlines this methodology, which is based on the
pioneer investigations of Tatum, Beadle, and other workers on the production
and characterization of auxotrophic mutants in fungi (30, 69, 117-Tatum)
and in bacteria (63, 126, 127, 112). Mutations leading to catabolic defects
have been especially useful in bacterial work, both as genetic markers and
in the analysis of fermentation pathways (53) .

Mutations to auxotrophy do not lend themselves to quantitative estimation
of rates, despite more or less efficient selective methods for their isolation (88) .
On the other hand, mutations from auxotrophy to the nutritionally wild type
or prototrophic state are suitable for selective counting methods, but due care

JOSHUA LEDERBERG

must be taken to minimize residual growth of auxotrophic inocula at the
expense of growth factors carried over with the cells or resident as impurities in
the test medium (114, 116).

Serological variation is of key importance in medical bacteriology, but its
genetic study is barely under way. The work of Kauffmann and of Edwards
and Bruner on Salmonella illustrates the provocative information now at hand
(8, 9, 10, 71 ). A compendium of the serological variation of bacteria generally
would be inappropriate here, but the subject has been treated exhaustively in
excellent reviews and books (66, 36, 54, 32).

Given two alternative forms of a gene — say a and A, each of which mutates
into the other at a definite rate — it is easy to show that an equilibrium will
eventually be established such that the ratio of a : A will be equal to the ratio
of the mutation rates to the respective conditions (42). Most mutation rates
are so low (of the order of one per million or billion cell divisions) that muta-
tional equilibria would take too long for human observation, even if the neces-
sary constancy of the environment were possible. Mutations involving the
antigenic structure of the flagella of Salmonella have, however, been found
to have unusually high rates (accounting for the readiness with which they
have been found), and Stocker (124) has described mutational equilibrium
as approached from inocula of either form.

Induced mutations. — Since Muller's announcement in 1928 that X-rays
would induce mutations in fruitflies, an extensive segment of genetic research
has concerned the discovery of mutagenic agents and the conditions of their
effect. Higher organisms like Drosophila and maize are indispensable in the
finer analysis of the cytological basis of induced genetic alterations, but micro-
organisms are very useful tools in the screening of new agents for mutagenic
activity, and in the study of the gross quantitative aspects of such activity.
The same types of mutants already mentioned as best enumerated on a selective
basis are particularly useful here. Dose-response data have been published
for mutations induced by X-rays (113, 45) and by UV (49, 104) but their
interpretation, especially for UV, is far from simple.

It has been thought that radiations induced mutations by direct photo-
chemical processes, i.e., that the gene itself might be activated by the absorption
of a quantum of UV, or by collision with a secondary electron following an
X-ray quantum absorption, (60, 80). Some revision of this concept is now
necessary on the basis of recent research. The effectiveness of X-rays is
potentiated by the presence of oxygen, and there may be a tenfold difference
between the doses required for a given effect in oxygen as against an inert atmos-
phere (47). This argues for a radiochemical intermediate, possibly some free
radical (peroxide?) which depends upon oxygen for its production under the
influence of X-radiation.

Owing to the powerful penetrability of X-radiation, it has been used in
Drosophila studies more extensively than UV, which penetrates through living
material so poorly that there are serious experimental difficulties in its ap-

INTRODUCTION

plication in the genetics of plants and animals. The small size of bacterial cells
ideally suits them for experiments with this agent, and UV is probably the
most convenient and widely used mutagenic treatment in microbial experiments.
The analysis of UV-effects has been stimulated by the discovery of photore-
activation (72, 55). A number of workers have found that a treatment of
various types of cells (actinomycete spores, bacteria, bacteriophage, mold
spores, Arbacia eggs) with visible light partially cancels both the lethal and
the mutation-inducing effects of a previous dose of UV (56, 73, 104, 105, 134).
This might suggest that a light-sensitive substance is produced by UV, but the
possible nature of this substance and its locus within the cell are obscure. Bac-
teriophage inactivated with UV is photoreactivated by visible light only
following its absorption into sensitive cells. The effect of UV in inhibiting
enzyme synthesis is also subject to photoreactivation (125) but no systems
simpler than intact cells have been shown to give such an effect.

Chemical mutagens. — A byproduct of research on chemical warfare agents
during World War II was the realization of the possibility of mutagenic activity
of chemicals. The nitrogen and sulfur mustards (jg-chloro-alkylamines and
sulfides) have been studied especially extensively, and found to be potent
mutagens for all organisms studied. In general, their effects are similar to those
of X-rays and UV, but there are differences in details (26). The similarity
of effects is made the more pronounced by the fact that mutants that are
relatively resistant to "mustard" can be selected in E. coli, strain B, and these
mutants also show augmented resistance to UV and X-ray. (33, 138).

Active programs are under way in several laboratories to screen compounds
for mutagenic activity, with bacteria prominent among the test organisms.
The methodology of such a program, and some of the precautions needed
to justify a positive conclusion, are illustrated in Witkin's paper (11). There
is a wealth of further literature on this subject, and it is likely to remain an
active field. There is so far no rational basis which can be used to predict
the activity of new compounds, and substances as diverse as formaldehyde,
acriflavine, urethane, caffeine, hydrogen peroxide, and manganous ion are
credibly reported as active in one or another system (26, 49, 58, 65, 47). In
general, reagents with a high reactivity for labile organic H groups, e.g., for-
maldehyde, organic peroxides, acyl halides and sulfates, ethylene oxide, acetic
anhydride, and diazomethane are probably mutagenic (86) but this does
not account for the mutagenicity of such chemically inactive compounds as
caffeine or urethane.

Much more careful work is needed in this field, but so far there is no
convincing evidence for any appreciable specificity in the mutagenic capacities
of any of these chemicals or physical treatments. Each of them induces a wide
diversity of mutations (as far as this has been investigated), and in general,
the results of, e.g., X-ray treatment, would not be readily distinguishable from
that of formaldehyde. On the other hand, it is possible that different gene
forms may differ in the frequency with which they will respond to different

1

JOSHUA LEDERBERG
mutagenic treatments; but if there is any specificity, it has so far been of a
second order. In this respect, induced resemble spontaneous mutations, whence
some genes may mutate more frequently than others, but not in such a way
that the environment can be said to direct the mutation of a specific gene pre-
ferentially to the exclusion of others.

Spontaneous mutation. — We may return at this point to the mechanism of
"spontaneous mutation," keeping in mind that the study of experimentally
controlled variables on mutation erases the distinction between spontaneous
and induced. So long as the concept connoted by these terms is kept clearly
in mind as one distinct from that of directed mutation, there need be no
confusion.

Evidence bearing on the relationship between growth and mutation is
especially paradoxical, for mutations to phage resistance, for instance, ap-
parently do not accumulate in a resting culture ( i ) . On the other hand, cultures
whose growth is regulated in a special steady-state, controlled-flow culture
vessel ( "chemostat" ) mutate at nearly constant rates per unit time, whether
the cells are proliferating slowly or rapidly (106). One interpretation is that
spontaneous mutations are due not to intramolecular accidents or reproductive
errors, but rather to the action of intracellular chemical mutagens formed by
metabolic processes. In this connection it is worth noting that formaldehyde
and hydrogen peroxide are both fairly common metabolic intermediates, and
that two other mutagens, caffeine and allyl isothiocyanate (mustard oil), are
also natural products. Information on the effect of temperature changes on
mutation rate would be especially valuable if it could be dissected from effects
on growth or metabolism. Under conditions of steady growth a temperature
increment of io° accelerated mutations in E. coli by a factor of about 2, and a
similar increment is reported in other systems (85) .

Bacterial populations. — The necessity for thinking of bacterial cultures
always in terms of populations, which may have genotypically diverse com-
ponents, can scarcely be over-emphasized. The process reviewed to this point,
mutation, is the fundamental source of genetic variation, but in view of the
smallness of spontaneous mutation rates, it is obvious that the occasional
change of a cell from one genetic condition to another can make little impression
upon the composition of bacterial populations. The forces that determine which
genetic types will predominate in bacterial cultures are the subject of population
dynamics. In diploid sexual organisms, population genetics is greatly compli-
cated by recombination and by the concealment of genetic variation in the
heterozygous condition, so that the most drastic culling may have to be carried
out for a great many generations to have a marked effect on the relative
frequency of different gene forms. Selection in bacteria is, as a rule, more
straightforward, as shown, for example, by the quantitative isolation of phage-
resistant mutants by a single application of the virus. The physiological inter-
actions of bacteria in dense cultures lead to less trivial problems in population
dynamics. Such interaction may involve an obvious competition for nutrients,

INTRODUCTION

or the production of auto- or trans-inhibitory metabolites. An especially clearcut
example of the latter in an economically important organism is reprinted here
(12). The intricate interrelationships of the production of and sensitivity
to species-specific antibiotics ("colicins") have been thoroughly analyzed for
E. coli by Fredericq (59). One of the most exhaustive analyses of selection
dynamics in bacteria will serve best to illustrate the complexity of populational
interactions in which several distinct effects may be superimposed (115). A
number of populational interactions have been described with a less complete
analysis (138, 136, 85).

Population complexities may also arise when more than one mutation
occurs, so that the population consists of several categories of genotypes. One
example has been cited already in the stepwise development of resistance by
the cumulative effect of serial mutations. A second has been described so
recently that its implications are still under discussion. In independent experi-
mental work, three groups of investigators noticed a perplexing sequence of
cycles in bacterial populations under conditions of continuous or reiterated
culture (25, 106, 124). "Marker" mutant cells increased in proportion as
mutations accumulated, but instead of increasing indefinitely, their ratio was
subject to sporadic downward shifts. The same interpretation was independently
formulated for each case: an adaptive mutation increasing the fit of the
bacteria to their rather artificial in vitro environment. Owing to the overwhelm-
ing preponderance of the cells not carrying the marker, the adaptive mutation
will usually occur in an unmarked cell, the descendants of which will then
displace the rest of the population, markers and all. After the changeover,
marker mutations accumulate .again until a possible second changeover takes
place to complete another cycle. It is not predicted that this process could
alter the ultimate equilibrium of cell types under mutation "pressure," but
that it would alter the short-term course of cultures in which marker mutation
pressures outweigh selection effects is apparent. Such adaptive mutations un-
doubtedly serve as models for evolutionary specialization ; as the streptomycin-
dependent mutation cited earlier shows most strikingly, genetic adaptations
are often quite specific for the immediate environment. This type of specializa-
tion undoubtedly accounts for the often noted loss of virulence encountered
frequently among pathogenic bacteria maintained on artificial culture media.

Inter clonal variation: Sexual recombination. — The preceding discussion
of bacterial populations has supposed that each bacterial cell is genetically
isolated from its partners, i.e., that reproduction is exclusively vegetative or
clonal. A large body of evidence is now at hand, however, which shows that
this picture is incomplete, and that some account must be taken of interclonal
processes. One such process was discovered in 1946 by Tatum and Lederberg
(45, 84, 128) and described as genetic recombination, for the experiments are
based on the selective isolation of genetic factor combinations from mixtures
of different kinds of mutant cells. The conditions under which these genetic
exchanges take place, and the patterns in which they result, led these authors

JOSHUA LEDERBERG

to conclude that recombination in E. coli, strain K-12, results from a sexual
process (13). This conclusion is based upon evidence which is, to date, entirely
negative, indirect, or genetic. It is substantiated, however, by single-cell studies
on heterozygous diploid cultures (14, 142), and by a direct demonstration of
the nonfiltrability of the agents of recombination (40). A variety of selective
techniques can be used for the isolation of recombinants from strain K-12 of
E. coli. In addition to the selection of prototrophs from mixtures of auxotrophs
which is most commonly used, one may select for dually resistant recombinants
from mixtures of cells each resistant to one inhibitor ( 83 ) . A combination of
these methods has been profitably employed to detect new strains crossable to
K-12; a number we're found, some with many properties differentiating them
from each other. Developments in the author's laboratory to the date of this
writing have been summarized (47-Lederberg) .

Recombination in strain K-12 has been applied by a number of other
workers for various problems ( 102) . The genetics of resistance to Chloromycetin
(6) and to streptomycin have been contrasted (46, 101). The very occurrence
of factorial segregation shows, of course, that bacterial enzymes are not to be
identified with the genes required for their formation (98). In fact, it is
controversial whether the patterns of gene-enzyme interrelationships in bacteria
support the "one gene-one enzyme" proposed as a generalization from the
properties of auxotrophs in Neurospora (69, 47-Lederberg, Horowitz, Bonner).
Transformations. — The history of "type transformation" is often regarded as
dating from Griffith's classical experiment on pneumococcal types in 1928 (64),
but this work is antedated by a confusing array of studies which date at least
as far back as the controversy over the etiology of typhus fever. When the
rickettsial etiology of this disease was finally established, many workers sug-
gested that the serological reaction of Proteus OX- 19 with rickettsial antibodies
(Weil-Felix reaction) was due to a transformation of non-reactive Proteus
by rickettsial products, and gave the name "paragglutination" to induced
serological variations such as were supposed to be involved here. In support
of this hypothesis, (to which no credit is now given) many workers reported
that the serological reaction of E. coli and other enteric bacteria could be
modified by cultivation in filtrates, extracts, or lysates of serologically distinct
cultures (reviewed in 81 ) . Griffith's demonstration was, however, the first which
proved to be readily reproducible by other workers, and remains a proper
starting point for modern discussions of transformation. His paper is unfortun-
ately too lengthy to be appropriate for reprinting here, but his observations, and
those of other workers leading to the isolation and characterization of the
active principle responsible for the pneumococcus transformation are sum-
marized in the paper reprinted here ( 16) .

The genetic interpretations of this transformation are currently a subject of
lively discussion. Its description as a specific induced mutation is probably less
fruitful than as a transfer of hereditary material from one cell to another
(47-Taylor) . The main questions which still have to be answered include

INTRODUCTION

a) the further physical and chemical characterization of the agent, b) its
genetic complexity (that is, whether it involves single characters, the full
hereditary material of the pneumococcus, or something intermediate), and c)
its cytological relationships to nuclear or extranuclear structures in these cells.
The current approaches to these problems are summarized in a number of
papers and reviews, (27, 91, 93, 130, 131).

In addition to the volume of older literature which must be supposed
to carry some grain among the chaff, there are a number of more recent reports
of transformations in various organisms. These include E. coli (31 ) , Hemophilus
influenzae (22), Shigella paradysenteriae (133), Alkaligenes radiobacter (39),
and staphylococcus (38). Discussion of these and other transformations should
take into account complications which might arise from bacterial life cycles
more complex than usually regarded.

This presentation has emphasized the contrasting features rather than
the similarities of the phenomena described as genetic recombination vs. trans-
formation. Both phenomena tend to the same genetic result: the elaboration
of the cells whose hereditary traits are derived from more than one parent.
When more is known of the morphological basis of "sexual" genetic recombina-
tion on one hand, and of the genetic properties of transformations, on the other,
a more profitable synthesis of these contrasting concepts may issue.

The recrudescence of interest in bacterial cytology has been largely inde-
pendent of, though contemporary with the development of bacterial genetics
documented in this book (77). The geneticist notes at least two fields where
cytological information is indispensable to him: a) the form and behavior of
the bacterial nucleus, and b) the possible existence of complex "life cycles."
Reliable information on both these subjects is relatively meagre, but the
evidence of at least the existence of nuclei in bacterial cells is relatively convinc-
ing. Robinow's paper reprinted here (15) may be taken as one point of de-
parture for the more recent work on this subject (41, 74) . With the improvement
of techniques, we may look forward to rapid progress in the establishment of
the details of the nuclear cycle at cell division, segregation, etc. (99, 118, 92,
122, 47).

The expression "life cycle" has come to carry many connotations in
bacteriology which hinder a careful discussion of the often conflicting observa-
tions of generations of bacteriologists. At least one type of cycle is indisputable,
the formation of highly resistant endospores by many bacteria. Whether
endosporogenesis has any genetic significance is controversial. At the least,
they may be supposed to represent the stage at which the presence of only one
nucleus per cell (in contrast to the two, four, or more, characteristic of most
rods) is the most likely (see 54), and on this basis may be particularly useful
for certain types of genetic experiments. More than one student of the mode of
nuclear segregation into the endospore has proposed a sequence of nuclear
fusion and meiotic reduction (i.e., autogamy). However, the cytological figures
are not easily interpreted, especially in view of the pitfalls of logical recon-

JOSHUA LEDERBERG

struction of fixed cells, and in this form autogamy is a blind alley without the
possibility of genetic novelty, and therefore not amenable to genetic test, (74).

The genetic significance of another cycle, the "L-forms," is equally obscure.
It is expected that this problem will be clarified in the very near future to an
extent that makes a detailed discussion of its present status here premature. The
reader is referred to provocative discussions published elsewhere (50, 51, 52,
75, 76, 132, 123) These studies reopen the possibility of the existence of
diminutive forms of many bacteria which may be capable of passing through
ordinary bacterial filters. These forms may also be highly resistant (like the
endospores) to heat and antiseptics. Since they also may have special nutritional
requirements, they may remain dormant in ordinary bacteriological media
which support the normal bacterial form. If these observations are confirmed,
the filtrable L-forms might well play a role in some of the transformation
experiments cited earlier. However, the transforming agent would consist not
of a single genetic factor transferred from one cell to another, but of a diminutive
form of one cell which germinates under the influence of another. The analogy
between the diminutive forms and the gametes of other plants and animals
is obvious on a physical plane; the possibility that they may play a similar
function in bacterial biology is a provocative one which will require careful
study.

One of the reasons for including papers on viruses and bacteria in the
same volume is the necessity for considering both the host and the "parasite"
in any physiological studies of their association. Furthermore, it is becoming
apparent that a great many bacterial cultures carry bacteriophages in a symbio-
tic relationship ( "lysogenicity" ) so that what are taken to be bacteriological
investigations have an unavoidable virological component. A second reason is
that the phage-bacterium complex may be profitably regarded as a unit for
comparison with other cellular systems which carry extranuclear hereditary
components (57, 119). Burnet's paper illustrates this very clearly (17).

It might be worth pointing out also the formal analogy between the transfer
of a latent virus from a lysogenic to a sensitive bacterium and the transfer
of the capsular attributes from smooth to rough bacteria in the pneumococcus
transformation. Whether this analogy is more than superficial only time will
tell. It is becoming clear, however, that there is a continuous gradation of pro-
perties between systems of cytoplasmic heredity such as the chloroplasts of green
plants, and the metabolic granules of yeast, and infective associations such as
the "killer" factors in Paramecium, rickettsia in the Arthropoda, and viruses in
plants and mammals. In many instances at both extremes, "disinfection" or cure
is possible by the use of appropriate drugs (streptomycin for chloroplasts, acri-
flavine for yeast granules) (47, 57).

Burnet's paper also illustrates one of the earliest and clearest cases of muta-
tion in a bacteriophage. In a later study, Luria described host range mutations
in phages which restore the virus' ability to attack bacteria which had mutated
to resistance to the previous form of the virus ( 18) . This process of compensatory

INTRODUCTION

mutations on the part of bacterium and virus evidently can sometimes be
repeated over a great many cycles, and is of some economic importance in the
acetone fermentation industry. These mutations in bacteriophage have been
used as the basis for further genetic studies which have revealed the occurrence
of genetic recombination among viruses (as well as in bacteria) (19, 20). Two
lines of attack are exemplified in the papers here. One uses patent markers,
such as host range and plaque morphology. For the other, recombination among
lethal mutations induced by UV is used to explain multiplicity reactivation,
or the cooperation of several particles damaged by UV to initiate phage growth,
in contrast to the ineffectiveness of individual damaged particles.

This work has naturally stimulated attempts to show similar phenomena in
animal viruses. At present these are represented by reports by Burnet and
Lind (37) of probable recombination in the influenza virus, and similar investi-
gations are under way in other laboratories. Plant and animal viruses are likely
to be more difficult to study from this point of view, chiefly because of technical
difficulties in initiating infections with single virus particles (35, 62, 79).

Genetic study of bacteria and viruses is closely interwoven with the most
general problems of their biology; this is not surprising, for the same has hap-
pened in other areas of biology. But it is to be hoped that genetics will be re-
garded not as a unique or isolated part of bacteriological study, but as an ele-
ment of all teaching and research in microbiology.

THE PAPERS

MUTATIONS OF BACTERIA FROM VIRUS SENSITIVITY
TO VIRUS RESISTANCE 12

S. E. LURIA» and M. DELBRtJCK

Indiana University, Bloomington, Indiana, and
Vanderbilt University, Nashville, Tennessee

Received May 29, 1943
INTRODUCTION

WHEN a pure bacterial culture is attacked by a bacterial virus, the cul-
ture will clear after a few hours due to destruction of the sensitive cells
by the virus. However, after further incubation for a few hours, or sometimes
days, the culture will often become turbid again, due to the growth of a bac-
terial variant which is resistant to the action of the virus. This variant can be
isolated and freed from the virus and will in many cases retain its resistance
to the action of the virus even if subcultured through many generations in the
absence of the virus. While the sensitive strain adsorbed the virus readily, the
resistant variant will generally not show any affinity to it.

The resistant bacterial variants appear readily in cultures grown from a
single cell. They were, therefore, certainly not present when the culture was
started. Their resistance is generally rather specific. It does not extend to
viruses that are found to differ by other criteria from the strain in whose pres-
ence the resistant culture developed. The variant may differ from the original
strain in morphological or metabolic characteristics, or in serological type or in
colony type. Most often, however, no such correlated changes are apparent,
and the variant may be distinguished from the original strain only by its re-
sistance to the inciting strain of virus.

The nature of these variants and the manner in which they originate have
been discussed by many authors, and numerous attempts have been made to
correlate the phenomenon with other instances of bacterial variation.

The net effect of the addition of virus consists of the appearance of a vari-
ant strain, characterized by a new stable character— namely, resistance to the
inciting virus. The situation has often been expressed by saying that bacterial
viruses are powerful "dissociating agents." While this expression summarizes
adequately the net effect, it must not be taken to imply anything about the
mechanism by which the result is brought about. A moment's reflection will
show that there are greatly differing mechanisms which might produce the
same end result.

D'Herelle (1926) and many other investigators believed that the virus
by direct action induced the resistant variants. Gratia (1921), Burnet (1929),
and others, on the other hand, believed that the resistant bacterial variants
are produced by mutation in the culture prior to the addition of virus. The

1 Theory by M. D., experiments by S. E. L.

* Aided by grants from the Dazian Foundation for Medical Research and from the
Rockefeller Foundation.

1 Fellow of the Guggenheim Foundation.

[Reprinted by permission from Genetics 28: 491-511, November, 1943]

492 S. E. LURIA AND M. DELBRtlCK

virus merely brings the variants into prominence by eliminating all sensitive
bacteria.

Neither of these views seems to have been rigorously proved in any single
instance. Burnet's (1929) work on isolations of colonies, morphologically
distinguishable prior to the addition of virus, which proved resistant to the
virus comes nearest to this goal. His results appear to support the mutation
hypothesis for colony variants. It may seem peculiar that this simple and im-
portant question should not have been settled long ago, but a close analysis of
the problem in hand will show that a decision can only be reached by a more
subtle quantitative study than has hitherto been applied in this field of re-
search.

Let us begin by restating the basic experimental finding.

A bacterial culture is grown from a single cell. At a certain moment the
culture is plated with virus in excess. Upon incubation, one finds that a very
small fraction of the bacteria survived the attack of the virus, as indicated by
the development of a small number of resistant colonies, consisting of bacteria
which do not even adsorb the virus.

Let us focus our attention on the first generation of the resistant variant —
that is, on those bacteria which survive immediately after the virus has been
added. These survivors we may call the "original variants." We know that
these bacteria and their offspring are resistant to the virus. We may formulate
three alternative hypotheses regarding them.

a. Hypothesis of mutation to immunity. The original variants were resistant
before the virus was added, and, like their offspring, did not even adsorb it. On
this hypothesis the virus did not interact at all with the original variants,
the origin of which must be ascribed to "mutations" that occur quite inde-
pendently of the virus. Naming such hereditary changes "mutations" of course
does not imply a detailed similarity with any of the classes of mutations that
have been analyzed in terms of genes for higher organisms. The similarity
may be merely a formal one.

b. Hypothesis of acquired immunity. The original variants interacted with
the virus, but survived the attack. We may then inquire into the predisposing
cause which effected the survival of these bacteria in contradistinction to the
succumbing ones. The predisposing cause may be hereditary or random. Ac-
cordingly we arrive at two alternative hypotheses — namely,

bi. Hypothesis of acquired immunity of hereditarily predisposed individuals.
The original variants originated by mutations occurring independently of the
presence of virus. When the virus is added, the variants will interact with it,
but they will survive the interaction, just as there may be families which are
hereditarily predisposed to survive an otherwise fatal virus infection. Since we
know that the offspring of the original variants do not adsorb the virus, we
must further assume that the infection caused this additional hereditary
change.

b2. Hypothesis of acquired immunity — hereditary after infection. The original
variants are predisposed to survival by random physiological variations in
size, age, etc. of the bacteria, or maybe even by random variations in the

MUTATIONS OF BACTERIA 493

point of attack of the virus on the bacterium. After survival of such random
individuals, however, we must assume that their offspring are hereditarily
immune, since they do not even adsorb the virus.

These alternative hypotheses may be grouped by first considering the origin
of the hereditary difference. Do the original variants trace back to mutations
which occur independently of the virus, such that these bacteria belong to a
few clones, or do they represent a random sample of the entire bacterial popu-
lation? The first alternative may then be subdivided further, according to
whether the original variants do or do not interact with the virus. Disregarding
for the moment this subdivision, we may formulate two hypotheses:

1. First hypothesis {mutation): There is a finite probability for any bac-
terium to mutate during its life time from "sensitive" to "resistant." Every
offspring of such a mutant will be resistant, unless reverse mutation occurs.
The term "resistant" means here that the bacterium will not be killed if ex-
posed to virus, and the possibility of its interaction with virus is left open.

2. Second hypothesis {acquired hereditary immunity)'. There is a small finite
probability for any bacterium to survive an attack by the virus. Survival of
an infection confers immunity not only to the individual but also to its off-
spring. The probability of survival in the first instance does not run in clones.
If we find that a bacterium survives an attack, we cannot from this information
infer that close relatives of it, other than descendants, are likely to survive the
attack.

The last statement contains the essential difference between the two hy-
potheses. On the mutation hypothesis, the mutation to resistance may occur
any time prior to the addition of virus. The culture therefore will contain
"clones of resistant bacteria" of various sizes, whereas on the hypothesis of
acquired immunity the bacteria which survive an attack by the virus will be
a random sample of the culture.

For the discussion of the experimental possibility of distinction between
these two hypotheses, it is important to keep in mind that the offspring of a
tested bacterium which survives is resistant on either hypothesis. Repeated
tests on a bacterium at different times, or on a bacterium and on its offspring,
could therefore give no information of help in deciding the present issue. Thus,
one has to resort to less direct methods. Two main differences may be derived
from the hypotheses:

First, if the individual cells of a very large number of microcolonies, each
containing only a few bacteria, were examined for resistance, a pronounced
correlation between the types found in a single colony would be expected on
the mutation hypothesis, while a random distribution of resistants would be
expected on the hypothesis of acquired hereditary immunity. This experiment,
however, is not practicable, both on account of the difficulty of manipulation
and on account of the small proportion of resistant bacteria.

Second, on the hypothesis of resistance due to mutation, the proportion of
resistant bacteria should increase with time, in a growing culture, as new
mutants constantly add to their ranks.

494 S. E. LURIA AND M. DELBRtlCK

In contrast to this increase in the proportion of resistants on the mutation
hypothesis, a constant proportion of resistants may be expected on the hy-
pothesis of acquired hereditary immunity, as long as the physiological condi-
tions of the culture do not change. To test this point, accurate determinations of
the proportion of resistant bacteria in a growing culture and in successive sub-
cultures are required. In the attempt to determine accurately the proportion
of resistant bacteria, great variations of the proportions were found, and results
did not seem to be reproducible from day to day.

Eventually, it was realized that these fluctuations were not due to any un-
controlled conditions of our experiments, but that, on the contrary, large
fluctuations are a necessary consequence of the mutation hypothesis and that
the quantitative study of the fluctuations may serve to test the hypothesis.

The present paper will be concerned with the theoretical analysis of the
probability distribution of the number of resistant bacteria to be expected on
either hypothesis and with experiments from which this distribution may be
inferred.

While the theory is here applied to a very special case, it will be apparent
that the problem is a general one, encountered in any case of mutation in uni-
parental populations. It is the belief of the authors that the quantitative study
of bacterial variation, which until now has made such little progress, has been
hampered by the apparent lack of reproducibility of results, which, as we shall
show, lies in the very nature of the problem and is an essential element for its
analysis. It is our hope that this study may encourage the resumption of quan-
titative work on other problems of bacterial variation.

The aim of the theory is the analysis of the probability distributions of the
number of resistant bacteria to be expected on the hypothesis of acquired
immunity and on the hypothesis of mutation.

The basic assumption of the hypothesis of acquired hereditary immunity
is the assumption of a fixed small chance for each bacterium to survive an at-
tack by the virus. In this case we may therefore expect a binomial distribution
of the number of resistant bacteria, or, in cases where the chance of survival
is small, a Poisson distribution.

The basic assumption of the mutation hypothesis is the assumption of a
fixed small chance per time unit for each bacterium to undergo a mutation to
resistance. The assumption of a fixed chance per time unit is reasonable only
for bacteria in an identical state. Actually the chance may vary in some manner
during the life cycle of each bacterium and may also vary when the physio-
logical conditions of the culture vary, particularly when growth slows down on
account of crowding of the culture. With regard to the first of these variations,
the assumed chance represents the average chance per time unit, averaged
over the life cycle of a bacterium. With regard to the second variation, it
seems reasonable to assume that the chance is proportional to the growth rate
of the bacteria. We will then obtain the same results as on the simple assump-

MUTATIONS OF BACTERIA 495

tion of a fixed chance per time unit, if we agree to measure time in units of divi-
sion cycles of the bacteria, or any proportional unit.

We shall choose as time unit the average division time of the bacteria, di-
vided by In 2, so that the number N t of bacteria in a growing culture as func-
tion of time t follows the equations

(i) dNt/dt = N t , and N t = N e 4 .

We may then define the chance of mutation for each bacterium during the
time element dt as

(2) adt,

so that a is the chance of mutation per bacterium per time unit, or the "muta-
tion rate."

If a bacterium is capable of different mutations, each of which results in
resistance, the mutation rate here considered will be the sum of the mutation
rates associated with each of the different mutations.

The number dm of mutations which occur in a growing culture during a
time interval dt is then equal to this chance (2) multiplied by the number
of bacteria, 4 or

(3) dm = adtNt,

and from this equation the number m of mutations which occur during any
finite time interval may be found by integration to be

(4) m = a(N t - No)

or, in words, to be equal to the chance of mutation per bacterium per time unit
multiplied by the increase in the number of bacteria.

The bacteria which mutate during any time element dt form a random
sample of the bacteria present at that time. For small mutation rates, their
number will therefore be distributed according to Poisson's law. Since the
mutations occuring in different time intervals are quite independent from each
other, the distribution of all mutations will also be according to Poisson's law.

This prediction cannot be verified directly, because what we observe, when
we count the number of resistant bacteria in a culture, is not the number of
mutations which have occurred, but the number of resistant bacteria which
have arisen by multiplication of those which mutated, the amount of multipli-
cation depending on how far back the mutation occurred.

If, however, the premise of the mutation hypothesis can be proved by other
means, the prediction of a Poisson distribution of the number of mutations

* We assume that the number of resistant bacteria is at all times small in comparison with the
total number of bacteria. If this condition is not fulfilled, the total number of bacteria in this
equation has to be replaced by the number of sensitive bacteria. The subsequent theoretical de-
velopments will then become a little more complicated. For the case studied in the experimental
part of this paper the condition is fulfilled.

496 S. E. LURIA AND M. DELBRUCK

may be used to determine the mutation rate. It is only necessary to determine
the fraction of cultures showing no mutation in a large series of similar cultures.
This fraction p , according to theory, should be:

(5) Po = e~ m .

From this equation the average number m of mutations may be calculated,
and hence the mutation rate a from equation (4).

Let us now turn to the discussion of the distribution of the number of
resistant bacteria.

The average number of resistant bacteria is easily obtained by noting that
this number increases on two accounts — namely, first on account of new muta-
tions, second on account of the growth of resistant bacteria from previous
mutations. During a time element dt the increase on the first account will be,
by equation (3) : adtN t . N t , the number of bacteria present at time t, is
given by equation (1). The increase on the second account will depend on the
growth rate of the resistant bacteria. In the simple case, which we shall
treat here, this growth rate is the same as that of the sensitive bacteria, and
the increment on this account is p dt, where p is the average number of re-
sistant bacteria present at time t. We have then as the total rate of increase
of the average number of resistant bacteria dp/dt = aN t +p and upon integra-
tion

(6) p = taNt

if we assume that at time zero the culture contained no resistant bacteria.

It will be seen that the average number of resistant bacteria increases more
rapidly than the total number of bacteria. Indeed the fraction of resistant
bacteria in the culture increases proportionally to time. This, as pointed out
in the introduction, is a distinguishing feature of the mutation hypothesis
but unfortunately, as will be seen in the sequel, is not susceptible to experi-
mental verification due to statistical fluctuations.

The resistant bacteria in any culture may be grouped, for the purpose of
this analysis, into clones, taking together all those which derive from the
same mutation. We may say that the culture contains clones of various age
and size, calling "age" of a clone the time since its parent mutation occurred
and "size" of a clone the number of bacteria in a clone at the time of observa-
tion. It is clear that size and age of a clone determine each other. If, in par-
ticular, we make the simplifying hypothesis that the resistant bacteria grow
as fast as the normal sensitive strain, the relation between size and age will be
expressed by equation (1), with appropriate meaning given to the symbols.

The relation implies that the size of a clone increases exponentially with its
age. On the other hand, the frequency with which clones of different ages may
be encountered in any culture must decrease exponentially with age, according
to equations (3) and (1).

Combining these two results — namely, that clone size increases exponen-
tially with clone age and that frequency of clones of different age decreases
exponentially with clone age — we see that the two factors cancel when the

MUTATIONS OF BACTERIA 497

average number of bacteria belonging to clones of one age group is considered.
In other words, at the time of observation we shall have, on the average, as
many resistant bacteria stemming from mutations which occurred during the
first generation after the culture was started as stemming from mutations
which occurred during the last generation before observation, or during any
other single generation.

On the other hand, for small mutation rates it is very improbable that any
mutation will occur during the early generations of a single or of a limited num-
ber of experimental cultures. It follows that the average number of resistant
bacteria derived from a limited number of experimental cultures will, probably,
be considerably smaller than the theoretical value given by equation (6), and,
improbably, the experimental value will be much larger than the theoretical
value. The situation is similar to the operation of a (fair) slot machine, where
the average return from a limited number of plays is probably considerably
less than the input, and improbably, when the jackpot is hit, the return is
much bigger than the input.

This result characterizes the distribution of the number of resistant bacteria
as a distribution with a long and significant tail of rare cases of high numbers of
resistant bacteria, and therefore as a distribution with an abnormally high vari-
ance. This variance will be calculated below.

For such distributions the averages derived from limited numbers of samples
yield very poor estimates of the true averages. Somewhat better estimates of
the averages may in such cases be obtained by omitting, in the calculation of
the theoretical averages, the contribution to these averages of those events
which probably will not occur in any of our limited number of samples. We
may do this, in the integration leading to equation (6), by putting the lower
limit of integration not at time zero, when the cultures were started, but at a
certain time t , prior to which mutations were not likely to occur in any of our
experimental cultures. We then obtain as a likely average r of the number of
resistant bacteria in a limited number of samples, instead of equation (6),

(6a) r = (t - t )aN t .

It now remains to choose an appropriate value for the time interval t — to.

For this purpose we return to equation (4), in which it was stated that the
average number of mutations which occur in a culture is equal to the mutation
rate multiplied by the increase of the number of bacteria. Let us then choose
to such that up to that time just one mutation occurred, on the average, in a
group of C similar cultures, or

1 = aC(N t0 - No).

In this equation we may neglect No, the number of bacteria in each inoculum,
in comparison with Nt , the number of bacteria in each culture at the critical
time t . We may also express N to in terms of N t , the number of bacteria at the
time of observation, applying equation (1):

N t e-

(t-to)

498 S. E. LURIA AND M. DELBRtlCK

We thus obtain

(7) t - to = ln(NtCa).

Equations (6a) and (7) may be combined to eliminate t — to and to yield
a relation between the observable quantities r and N t on the one hand and
the mutation rate a on the other hand, to be determined by this equation:

(8) r = aNJn (N t Ca).

This simple transcendental equation determining a may be solved by any
standard numerical method. In figure 1, the relation between r and aN t is
plotted for several values of C.

I .2 .5 12 5

Figure i. — The value of aN t as a function of r for various values of C. The upper left hand part
of the figure gives the curves for low values of aNt and of r on a larger scale. See text.

Estimates of a obtained from equation (8) will be too high if in any of the
experimental cultures a mutation happened to occur prior to time to. From the
definition of to it will be seen that this can be expected to happen in little more
than half of the cases.

While we have thus obtained a relation permitting an estimate of the muta-
tion rate from the observation of a limited number of cultures, this relation is
in no way a test of the correctness of the underlying assumptions and, in par-
ticular, is not a test of the mutation hypothesis itself. In order to find such
tests of the correctness of the assumption we must derive further quantitative
relations concerning the distribution of the number of resistant bacteria and
compare them with experimental results.

MUTATIONS OF BACTERIA 499

Since we have seen that the mutation hypothesis, in contrast to the hypothe-
sis of acquired immunity, predicts a distribution of the number of resistant
bacteria with a long tail of high numbers of resistant bacteria, the determina-
tion of the variance of the distribution should be helpful in differentiating be-
tween the two hypotheses. We may here again determine first the true vari-
ance — that is, the variance of the complete distribution — and second the likely
variance in a limited number of cultures, by omitting those cases which are
not likely to occur in a limited number of cultures.

The variance may be calculated in a simple manner by considering sep-
arately the variances of the partial distributions of resistant bacteria, each
partial distribution comprising the resistant bacteria belonging to clones of
one age group. The distribution of the total number of resistant bacteria is
the resultant of the superposition of these independent partial distributions.

Each partial distribution is due to the mutations which occurred during a
certain time interval dr, extending from (t — r) to (t — r+dr). The average
number of mutations which occurred during this interval is, according to
equation (3),

(9) dm = aN T dr = aN t e _T dT.

These mutations will be distributed according to Poisson's law, so that the
variance of each of these distributions is equal to the mean of the distribution.
We are however not interested in the distribution of the number of mutations
but in the distribution of the number of resistant bacteria which stem from
these mutations at the time of observation — that is, after the time interval t.
Each original mutant has then grown into a clone of size e T . The distribution
of the resistant bacteria stemming from mutations occurred in the time inter-
val dr has therefore an average value which is e T times greater than the average
number of mutations, and a variance which is e 2r times greater than the vari-
ance of the number of mutations. Thus we find for the average number of
resistant bacteria:

dp = aN t dr,

and for the variance of this number

var dp = aNtCdr.

From this variance of the partial distribution, the variance of the distribution
of all resistant bacteria may be found simply by integrating over the appro-
priate time interval — that is, either from time t to time o (r from o to t), if
the true variance is wanted, or from time t to time t (r from o to t — 1 ), if the
likely variance in a limited number of cultures is wanted. In the first case we
obtain:

(10) var p = aNtCe* — 1).
In the second case we obtain:

(10a) var r = aN t [e (t-to) — 1 J.

500 S. E. LURIA AND M. DELBRUCK

Substituting here the previously found value of (t — to) and neglecting the
second term in the brackets, we obtain:

(n) var, = Ca 2 N t 2 .

Comparing this value of the likely variance with the value of the likely
average, from equation (8), we see that the ratio of the standard deviation
to the average is:

(12) Vvar~r/r = VC/ln (N t Ca).

It is seen that this ratio depends on the logarithm of the mutation rate and
will consequently be only a little smaller for mutation rates many thousand
times greater than those considered in the experiments reported in this paper.
In the beginning of this theoretical discussion we pointed out that the
hypothesis of acquired immunity leads to the prediction of a distribution of
the number of resistant bacteria according to Poisson's law, and therefore to
the prediction of a variance equal to the average. On the other hand, if we com-
pare the average, equation (8), with the variance, equation (n), (not, as above,
with the square root of the variance), we obtain

(12a) var r = rN t Ca/ln (N t Ca).

Equation (12a) shows that the likely ratio between variance and average is
much greater than unity on the hypothesis of mutation, if (N t Ca), the total
number of mutations which occurred in our cultures, is large compared to
unity. 5

It is possible to carry the analysis still further and to evaluate the higher
moments of the distribution function of the number of resistant bacteria, or
even the distribution function itself. The moments are comparatively easy to
obtain, while the calculation of the distribution function involves considerable

5 In some of the experiments reported in the present paper we did not determine the tota'
number of resistant bacteria in each culture, but the number contained in a small sample from
each culture. In these cases the variance of the distribution of the number of resistant bacteria
will be slightly increased by the sampling error. The proper procedure is here first to find the
average number of resistant bacteria per culture by multiplying the average per sample by the
ratio

volume of culture
volume of sample*

second, to evaluate the mutation rate with the help of equation (8); third, to figure the likely
variance for the cultures by equation (n); fourth, to divide this variance by the square of the
ratio (13) to obtain that part of the variance in the samples which is due to the chance distribution
of the mutations. The experimental variance should be greater than this value, on account of
the sampling variance. The sampling variance is in all our cases only a small correction to the
total variance, and it is sufficient to use its upper limit, that of the Poisson distribution, in our
calculations. Consequently, when comparing the experimental with the calculated values, we
first subtract from the experimental value the sampling variance, which we take to be equal to
the average number of resistant bacteria.

MUTATIONS OF BACTERIA 501

mathematical difficulties. An approximation to the beginning of the distribu-
tion function — that is, to its values for small numbers of resistant bacteria —
may be obtained by grouping mutations according to the bacterial generation
during which they occurred. For instance, the probability of obtaining seven
resistant bacteria may be broken down into the sum of the following alterna-
tive events: (a) seven mutations during the last generation; (b) three mutations
during the last generation and two mutations one generation back; (c) three
mutations during the last generation and one mutation two generations back;
(d) one mutation during the last generation and three mutations one genera-
tion back; (e) one mutation during the last generation, one mutation one gen-
eration back and one mutation two generations back.

The probability of each of these events depends only on the mutation rate
and on the final number of bacteria.

The grouping of mutations according to the bacterial generation during
which they occurred, and the assumption that the bacteria increase in simple
geometric progression, simplify the calculation sufficiently to permit numerical
computation. On the other hand, the classes with two, four, eight, etc., mu-
tants are artificially favored by this procedure, so that a somewhat uneven
distribution results, with too high values for two, four, eight, etc., resistant
bacteria (see fig. 2).

MATERIAL AND METHODS

The material used for our experimental study consisted of a bacterial virus
a and of its host, Escherichia coli B (Delbruck and Luria 1942). Secondary
cultures after apparently complete lysis of B by virus a show up within a few
hours from the time of clearing. They consist of cells which are resistant to
the action of virus a, but sensitive to a series of other viruses active on B.
The resistant cells breed true and can be established easily as pure cultures.
No trace of virus could be found in any pure culture of the resistant bacteria
studied in this paper. The resistant strains are therefore to be considered as
non-lysogenic.

Tests were made to see whether the resistance to virus a was a stable char-
acter of the resistant strains. In the first place, it was found that virus a is
not appreciably adsorbed by any of the resistant strains. In the second place,
when a certain amount of virus a is mixed with a growing culture of a resistant
strain, no measurable increase of the titer of virus a occurs over a period of
several hours. This is a very sensitive test for the occurrence of sensitive bac-
teria, and its negative result for all resistant strains shows that reversion to
sensitivity must be a very rare event.

Morphologically at least two types of colonies of resistant bacteria may be
distinguished. The first type of colony is similar to the type produced by the
sensitive strain both in size and in the character of the surface and of the
edge. The second type of colony is much smaller and translucent. The differ-
ence in colony type is maintained in subcultures. Microscopically the bacteria
from these two types of colonies are indistinguishable. They also do not differ

5 o2 S. E. LURIA AND M. DELBRUCK

from each other or from the sensitive strain in their fermentation reactions on
common sugars and in the characteristics of their growth curves in nutrient
broth. In particular, the lag periods, the division times during the logarithmic
phase of growth, and the maximum titers attained are identical for the sensi-
tive strain and for the two variants. Both variants, therefore, fulfill the require-
ments for the applicability of the theory developed above.

In the presentation of our experimental results we have lumped the counts
of the two types of colonies together, because: (i) theoretically, this is equiva-
lent to summing the corresponding mutation rates; (2) experimentally, we
are not certain whether each of these types does not actually comprise a diver-
sity of variants; (3) experimentally, no correlation appeared to exist between
the occurrence of these variants, which shows the independence of the causes
of their occurrence.

Cultures of B were grown either in nutrient broth (containing .5 percent
NaCl) or in an asparagin-glucose synthetic medium. In the latter, the division
time during the logarithmic phase of growth was 35 minutes, as compared
with 19 minutes in broth. In synthetic medium, the acidity increased during
the time of incubation from pH 7 to pH 5.

In cultures of strain B, between io~ 8 and io~ 6 of the bacteria are found
usually to give colonies resistant to the action of virus a when samples of such
cultures are plated with large amounts of virus. In order to be reasonably cer-
tain that the resistant bacteria found in the test had not been introduced into
the test culture with the initial inoculum, the test cultures were always started
with very small inocula, containing between 50 and 500 bacteria from a grow-
ing culture. Thus any resistant bacterium found at the moment of testing
(when the culture contains between io 8 and 5X10 9 bacteria/cc) must be an
offspring of one of the sensitive bacteria of the inoculum.

All platings were made on nutrient agar plates. The plating experiments for
counting the number of resistant bacteria in a liquid culture of the sensitive
strain were done by plating either a portion or the entire culture with a large
amount of virus a. The virus was plated first, and spread over the entire sur-
face of the agar. A few minutes later the bacterial suspension to be tested was
spread over the central part of the plate, leaving a margin of at least one cen-
timeter. Thus all bacteria were surrounded by large numbers of virus par-
ticles.

Microscopic examination of plates seeded in this manner showed that lysis
takes place very quickly; only bacteria which at the time of plating were in
the process of division may sometimes complete the division. The resistant
colonies which appear after incubation are therefore due to resistant bacterial
cells present at the time of plating.

The total number of bacteria present in the culture to be tested was deter-
mined by colony counts in the usual manner.

The resistant colonies of the large type appear after 12-16 hours of incuba-
tion, the colonies of the small type appear after 18-24 hours, and never reach
half the size of the former ones. Counts were usually made after 24 and 48
hours.

MUTATIONS OF BACTERIA 503

EXPERIMENTAL

A Test of the Reliability of the Plating Method

In our experiments we wanted to study the fluctuations of the numbers of
resistant bacteria found in cultures of sensitive bacteria. It was therefore
necessary to show first that the method of testing did not involve any unrecog-
nized variables, which caused the number of resistant colonies to vary from
plate to plate or from sample to sample.

Therefore, parallel platings were made using a series of samples from the
same bacterial culture. If our plating method is reliable, fluctuations should
in this arrangement be due to random sampling only, and the variance from a
series of such samples should be equal to the mean.

Table 1 gives the results of three such experiments. It .will be seen that in

Table i

The number of resistant bacteria in different samples from the same culture.

EXP.

, NO.

10a

EXP

. no. 11a

EXP

. NO. 3

SAMPLE NO.

RESISTANT COLONIES

RESISTANT COLONIES

RESISTANT COLONIES

I

14

46

4

2

IS

56

2

3

13

52

2

4

21

48

I

5

IS

65

5

6

14

44

2

7

26

49

4

8

16

Si

2

9

20

56

4

10

13

47

7

mean

16.

7

51.4

3-3

variance

IS

27

3-8

x 2

9

5-3

12

p

4

.8

.2

all three cases variance and mean agree as well as may be expected. There is
therefore no reason to assume that the method of sampling or plating intro-
duces any fluctuations into our results besides the sampling error.

Fluctuations of the Number of Resistant Bacteria in Samples from a
Series of Similar Cultures

As pointed out in the introduction and in the theoretical part, the hy-
pothesis of acquired immunity and the hypothesis of mutation lead to radi-
cally different predictions regarding the distribution of the number of resistant
bacteria in a series of similar cultures. The hypothesis of acquired immunity
predicts a variance equal to the average, as in sampling, while the mutation
hypothesis predicts a much greater variance.

Series of five to 100 cultures were set up in parallel with small equal inocula,
and were grown until maximum titer was reached. Three kinds of cultures

504 S. E. LURIA AND M. DELBRUCK

were used — namely: (i) io.o cc aerated broth cultures; (2) .2 cc broth cultures;
(3) .2 cc synthetic medium cultures.

The results of all tests for the number of resistant bacteria are summarized
in table 2 and table 3.

Table 2

The number of resistant bacteria in series of similar cultures.

EXPERIMENT NO.

1

10

ir

IS

16

17

21a

2lb

Number of cultures

9

8

10

10

20

12

19

s

Volume of cultures, cc

10.0

10.0

10.0

10.

.2*

.2*

.2

IO.O

Volume of samples, cc

•05

• 05

• 05

•°S

.08

.08

•os

• 05

38

28

3S

107
S 14 31 183 24 o 8 13

6

10

29

30

6

18

41

10

5

25

17

40

10

10

20

45

8

14

31

183

24

27

30

12

13

3

7

173

165

17

17

23

IS

17

57

6

51

10

Average per sample

26.8

23.8

62

26. 2

ii-3S

3°

3.8

48.2

Variance (corrected for

sampling)

1217

84

3498

2178

694

6620

40.8

1171

Average per culture

S360

4760

12400

5240

28.4

75

IS-I

8440

Bacteria per culture

3.4X10U

4 Xio"

4 Xio>o

2.QXI0 10

5-6Xio»

5 X108

i.iXio'

3.2X10"

Mutation rate

i.8Xio-»

1. 4X10-8

4.1X10-8

2.IXIO-8

1.1X10-8

3-oXio-«

3.3X10-8

3.0X10-8

Standard deviation fexp.

1.3

•39

•95

1.8

2-3

2-7

1-7

■ 71

Average [calc.

■35

■3i

•33

• 37

• 94

.67

1.04

.26

* Cultures in synthetic medium.

It will be seen that in every experiment the fluctuation of the numbers of
resistant bacteria is tremendously higher than could be accounted for by the
sampling errors, in striking contrast to the results of plating from the same
culture (see table 1) and in conflict with the expectations from the hypothesis
of acquired immunity.

We want to see next whether these results fit the expectations from the
hypothesis of mutation. We must therefore compare the experimental results
with the relations developed in the theoretical part, keeping in mind that the
theory con tains several simplifying assumptions.

First we can compare, according to equation (12), the experimental and the
calculated values of the ratio between the standard deviation and the average
of the numbers of resistant bacteria. These ratios are included in tables 2 and
3. It is seen that the experimental and theoretical values are reasonably close.

MUTATIONS OF BACTERIA 505

However, in all but one case the experimental ratio is greater than the value
calculated from the theory — that is, the variability is even greater than pre-
dicted.

Table 3
Distribution of the numbers of resistant bacteria in series of similar cultures.

EXPERIMENT NO. 2 2 2\

Number of cultures
Volume of cultures, cc
Volume of samples, cc

87

Resistant
bacteria
o
1
2
3
4
5
6- 10
11- 20
21- 50
51- 100
101- 200
201- 500
501-1000

Number of

cultures

57

20

5
2

3

1
7
2
2
o

Resistant
bacteria

o

1

2

3

4

5

6- 10

11- 20

21- 50

51- 100

101- 200

201- 500

501-1000

Number of
cultures
29
17
4
3
3
2

5
6
7
5

Average per sample

10.12

28.6

Variance (corrected for sampling)

6270

6431

Average per culture

40.48

28.6

Bacteria per culture

2.8X10 8

2.4 Xio 8

Mutation rate

2.3X10- 8

2.37X10- 8

Standard deviationfexp.

7-8

2.8

Average \calc.

i-5

1.5

* Cultures in synthetic medium.

A part of this discrepancy may be accounted for by the fact that the
time t , mutations occurring prior to which were disregarded by the theory,
was chosen in such a manner that on the average one mutation would occur
prior to time t . This mutation, if it occurs, will of course tend to increase the
variance, and in some of the experiments the high value of the experimental
variance can be traced directly to one exceptional culture in which a mutation
had evidently occurred several generations prior to time to. Unfortunately,
there is no general criterion by which one might eliminate such cultures from
the statistical analysis, because, in a culture with an exceptionally high count
of resistant bacteria, these do not necessarily stem from one exceptionally
early mutation, but may also be due to an exceptionally large number of
mutations after time to.

There may also be other reasons why the observed variances are higher than
the expected ones. First of all, the simplifying assumption that the mutation

506

S. E. LURIA AND M. DELBRtJCK

rate per bacterial generation is independent of the physiological state of the
bacteria may be too simple. If the mutation rate is higher for actively growing
bacteria than for bacteria near the saturation limit of the cultures, early muta-
tions and big clone sizes will be favored, and therefore higher variations of the
numbers of resistant bacteria can be expected. Second, the assumption of a
sudden transition from sensitivity to resistance may also be too simple. It is
conceivable that the character "resistance to virus" may not fully develop in
the bacterial cell in which the mutation occurs, but only in its offspring, after
one or more generations. However, if this were the case, cultures with only one
or two resistant bacteria should be relatively rare. The last experiment listed
in table 3, in which the entire cultures were plated, shows a rather high propor-
tion of cultures with only one resistant bacterium. This seems to show that the

Figure 2. — Experimental (Experiment No. 23) and calculated distributions of the numbers
of resistant bacteria in a series of similar cultures. Solid columns: experimental. Cross-hatched
columns: calculated.

character "resistance to virus" in general does come to expression in the bac-
terial cell in which the corresponding mutation occurred, as assumed by the
theory.

Another way of comparing the experimental results with the theory is to
compare the experimental distribution of resistant bacteria with the approxi-
mate distribution calculated by the method outlined at the end of the theo-
retical part. The theoretical distribution has to be calculated from the aver-
age number of mutations per culture given by equation (5). Only experi-
ments where the whole culture is tested can therefore be used for such a
comparison. This method tests the fitting of the expectations for small numbers
of resistant bacteria, in contrast to the comparison of the standard deviations,
which involves predominantly the cultures with high numbers of resistant
bacteria.

Figure 2 shows the experimental and calculated distributions for Experi-
ment No. 23; the cultures with more than nine resistant bacteria are lumped
together in one class, since the distribution has not been calculated for values
higher than nine.

It is seen that the fitting for small values is satisfactory. In particular, the

MUTATIONS OF BACTERIA 507

number of cultures with one resistant bacterium very closely fits the expecta-
tion. The classes with two, four, eight, etc., resistant bacteria are bound to
be favored in the theoretical distribution, as explained in the theoretical part.

The results shown in figure 2 also confirm the assumption that the dis-
crepancy between experimental and calculated standard deviations must be
due to an excess of cultures with large numbers of resistant bacteria.

Summing up the evidence, we may say that the experiments show clearly
that the resistant bacteria appear in similar cultures not as random samples
but in groups of varying sizes, indicating a correlating cause for such grouping,
and that the assumption of genetic relatedness of the bacteria of such groups
offers the simplest explanation for them.

Mutation Rate

As pointed out in the theoretical part of this paper, mutation rates may be
estimated from the experiments by two essentially different methods. The
first method makes use of the fact that the number of mutations in a series of
similar cultures should be distributed in accordance with Poisson's law; the
average number of mutations per culture is calculated from the proportion
of cultures containing no resistant bacteria at the moment of the test, accord-
ing to equation (5).

There are two technical difficulties involved in the application of this
method. In the first place, rather large numbers of cultures have to be handled
and conditions have to be chosen so that the proportion of resistant bacteria
is neither too small nor too large. In the second place, the entire cultures have
to be tested, which means, in our method of testing, that cultures of rather
small volume have to be used and great care must be taken to plate as nearly
as possible the entire culture.

Experiment No. 23 (see table 3) permits an estimate of the mutation rate
by this method. Out of 87 cultures, no resistant bacteria were found in 29
cultures, a proportion of .33. From equation (5) we calculate therefore that the
average number of mutations per culture in this experiment was 1.10. Since
the total number of bacteria per culture was 2.4X10 8 , we obtain as the mu-
tation rate, from equation (4),

a = .47 X io~ 8 mutations per bacterium per time unit
= .32 X io~ 8 mutations per bacterium per division cycle.

This calculation makes use exclusively of the proportion of cultures contain-
ing no resistant bacteria. It is therefore inefficient in its use of the information
gathered in the experiment.

The second method makes use of the average number of resistant bacteria
per culture. The relation of this average number with the mutation rate was
discussed in the theoretical part of this paper and was found to be expressed
by equation (8). The mutation rates calculated by this method for each experi-
ment are collected in table 4.

5 o8

S.

E. LURIA

AND M.
Table 4

DELBRUCK

Values

of mutation rate from different experiments.

EXPERIMENT NO.

NUMBER OF CULTURES VOLUME OF CULTURES

MUTATION RATE

Mutations per bacterium

cc

per time unit

I

9

10.

1.8X10- 8

IO

8

10.

1.4X10- 8

II

10

10.

4.1X10- 8

IS

10

10.

2.1X10-"

16

20

.2*

i.iXio-'

17

12

.2*

3.0X10- 8

21a

19

.2

3-3Xio- 8

21b

5

10.

3.0X10- 8

22

100

.2*

2.3X10- 8

23

87

.2*

2.4X10- 8

Average

2.45X10- 8

* Cultures in synthetic medium.

It will be seen that the values of the mutation rate obtained by the second
method are all higher than the value found by the first method. This dis-
crepancy may be traced back to the same cause as the discrepancy between
the calculated and observed values of the standard deviation of the numbers
of resistant bacteria. This, we found, was due to an excess of early mutations,
giving rise to big clones of resistant bacteria. These big clones do not affect
the mutation rate calculated by the first method, but they do affect the results
of the second method, which is based on the average number of resistant bac-
teria.

One sees in table 4 that the mutation rate calculated by the second method
does not vary greatly from experiment to experiment. In particular, it will be
noted that there is no significant difference between the values obtained from
cultures in broth and from cultures in synthetic medium, notwithstanding the
considerable difference of metabolic activity and of growth rate of the bacteria
in these two media. This shows that the simple assumption of a fixed small
chance of mutation per physiological time unit is vindicated by the results.
It may also be noted in table 4 that there is no significant difference between
the mutation rates obtained from 10 cc cultures and those obtained from .2 cc
cultures, or between the experiments with many and those with few cultures.
The variability of the value of the mutation rate seems to be solely due to the
peculiar probability distribution of the number of resistant bacteria in series
of similar cultures predicted by the mutation theory.

At this point an experiment may be mentioned by which it was desired to
find out whether or not mutations occur in a culture after the bacteria have
ceased growing. A culture was grown to saturation and was then tested re-
peatedly for resistant bacteria and for total number of bacteria over several

MUTATIONS OF BACTERIA 509

days. The proportion of resistant bacteria did not change, even when the
sensitive bacteria began to die, showing that the resistant bacteria have the
same death rate in aging cultures as the sensitive bacteria.

DISCUSSION

We consider the above results as proof that in our case the resistance to
virus is due to a heritable change of the bacterial cell which occurs independ-
ently of the action of the virus. It remains to be seen whether or not this is
the general rule. There is reason to suspect that the mechanism is more com-
plex in cases where the resistant culture develops only several days after lysis
of the sensitive bacteria.

The proportion of mutant organisms in a culture and the mutation rate are
far smaller in our case than in other studied cases of heritable bacterial varia-
tion. The possibility of investigation of such rare mutations is in our case
merely the result of the method of detecting the mutant organisms. In other
cases, the variants are detected by changes in the colony type which is pro-
duced by the mutant organism, either in the pigmentation or in the character
of the surface or the edge of the colony. Often, colonies of intermediate charac-
ter occur, and it is difficult to decide whether they are mixed colonies or stem
from bacteria with intermediate character. This is particularly true of cases
where the mutation rate is high and where reverse mutation occurs. Fairly
high mutation rates, however, are a prerequisite of any study of colony vari-
ants, since the number of colonies that can be examined is limited by practical
reasons.

The study of mutations causing virus resistance is free of these difficulties.
The segregation of the mutant from the normal organisms occurs in the one-
cell stage by elimination of the normal individuals, and the character of the
colony which develops from a mutant organism is of secondary importance.
Owing to the total elimination of the normal individuals, the number of organ-
isms which may be examined is very much higher than for any other method;
more than io 8 bacteria may be tested on a single plate. Since the mutations
to virus resistance are often associated with other significant characters, the
method may well assume importance with regard to the general problems of
bacterial variation.

It must not be supposed that the peculiar statistical difficulties encountered
in our case are restricted to cases of very low mutation rates. The essential
condition for the occurrence of the peculiar distribution studied in the theoreti-
cal part of this paper is the following: the initial number of bacteria in a culture
must be so small that the number of mutations which occur during the first division
cycle of the bacteria is a small number. This will always be true, however great
the mutation rate, if one studies cultures containing initially a small number of
organisms.

In a series of very interesting studies of the color variants of Serratia mar-
cescens, Bunting (1940a, 1940b, 1942; Bunting and Ingraham 1942) suc-
ceeded to some extent in obviating the statistical difficulties by always using

510 S. E. LURIA AND M. DELBRtJCK

inocula of about 100,000 bacteria. In some of her cases this number was suffi-
ciently high to result in numerous mutations during the first division cycle
of the bacteria. In other cases the number was apparently not high enough,
since the author reports troublesome variations of the fractions of variants in
successive subcultures. In those cases where the size of the inocula was high
enough, the author succeeded in deriving reproducible values for the mutation
rates from the study of single cultures, followed through numerous subcultures.
In these cases it is sufficient to apply the equations of the theory referring to
the average numbers of mutants as a function of time. It is clear, however,
that this method is applicable only in cases of mutation rates of at least io~ 4
per bacterium per division cycle.

In our case, as in many others, the virus resistant variants do not exhibit
any striking correlated physiological changes. There is therefore little oppor-
tunity for an inquiry into the nature of the physiological changes responsible
for the resistance to virus. Since the offspring of the mutant bacteria, when iso-
lated after the test, are unable to synthesize the surface elements to which the
virus is specifically adsorbed in the sensitive strain, one might suppose that
this loss is a direct effect of the mutation. However, it is also conceivable that
the loss occurs upon contact with virus, since it is detected only after such
contact (hypothesis bi). In some of the cases studied by Burnet (1929),
where the mutational change to resistance is correlated with a change of phase,
from smooth to rough or vice versa, the change of the surface structure must
be a direct result of the mutation, since the mutant colonies may be picked up
prior to the resistance test and, when tested, exhibit the typical change of
affinity of the surface structure. These findings make it more probable that
the loss of surface affinity to virus is a direct effect of the mutation.

The alteration of specific surface structures due to genetic change is a phe-
nomenon of the widest occurrence. The genetic factors determining the anti-
genic properties of erythrocytes are well known. There is evidence (Webster
1937; Holmes 1938; Stevenson, Schultz, and Clark 1939) that resistance
or sensitivity to virus in plants and animals is correlated with, or even de-
pendent on, genetic changes, possibly affecting the antigenic make-up of the
cellular surface. The proof that resistance to a bacterial virus may be traced
to a specific genetic change may assume importance, therefore, with regard to
the general problems of virus sensitivity and virus resistance.

summary

The distribution of the numbers of virus resistant bacteria in series of similar
cultures of a virus-sensitive strain has been analyzed theoretically on the basis
of two current hypotheses concerning the origin of the resistant bacteria.

The distribution has been studied experimentally and has been found to
conform with the conclusions drawn from the hypothesis that the resistant
bacteria arise by mutations of sensitive cells independently of the action of
virus.

The mutation rate has been determined experimentally.

MUTATIONS OF BACTERIA 511

LITERATURE CITED

Bunting, M. I., 1940a A description of some color variants produced by Serratia marcescens,

strain 274. J. Bact. 40: 57-68.

1940b The production of stable populations of color variants of Serratia marcescens #274

in rapidly growing cultures. J. Bact. 40: 69-81.

1942 Factors affecting the distribution of color variants in aging broth cultures of Serratia

marcescens #274. J. Bact. 43: 593-606.
Bunting, M. I., and L. J. Ingraham, 1942 The distribution of color variants in aging broth

cultures of Serratia marcescens #274. J. Bact. 43: 585-591.
Burnet, F. M., 1929 Smooth-rough variation in bacteria in its relation to bacteriophage.

J. Path. Bact. 32: 15-42.
Delbruck, M., and S. E. Luria, 1942 Interference between bacterial viruses. I. Arch. Biochem.

1: in-141.
Gratia, A., 1921 Studies on the d'Herelle phenomenon. J. Exp. Med. 34: 115-131.
d'Herelle, F., 1926 The Bacteriophage and Its Behavior. Baltimore: Williams and Wilkins.
Holmes, F. O., 1938 Inheritance of resistance to tobacco-mosaic disease in tobacco. Phyto-
pathology 28: 553-561;
Stevenson, F. J., E. S. Schultz, and C. F. Clark, 1939 Inheritance of immunity from virus X

(latent mosaic) in the potato. Phytopathology 29: 362-365.
Webster, L. T., 1937 Inheritance of resistance of mice to enteric bacterial and neurotropic virus

infections. J. Exp. Med. 65: 261-286.

THE DISTRIBUTION OF THE NUMBERS OF MUTANTS
IN BACTERIAL POPULATIONS

By D. E. LEA,* Department of Radiotherapeutics and Strangeivays Laboratory, Cambridge
and C. A. COULSON, Wheatstone Physics Laboratory, King's College, London

Introduction
Luria & Delbruck (1943) have shown that if a culture of some hundreds or thousands of
millions of Bacterium coli, grown from a single cell, is plated out on a nutrient medium
impregnated with a bacteriophage to which the strain of coli is sensitive, the vast majority
of the bacteria are lysed, but a few give rise to colonies. These colonies contain only
bacteria resistant to the bacteriophage, and give rise only to resistant bacteria on further
subcultivation. Evidently hereditary variations or mutations can occur in bacteria.
Numerous other examples are known of mutations in bacteria, affecting fermentation
reactions (e.g. Lewis, 1934), resistance to chemicals (e.g. Stewart, 1947), to antibiotics
(e.g. Demerec, 1945), or to radiation (Witkin, 1946).

The demonstration of phage-resistant mutants necessarily involves the exposing of the
bacteria to the phage, and it is not immediately obvious whether the mutation to phage
resistance occurs spontaneously during the growth of the culture, and is merely made
apparent by subsequently testing with phage, or whether the mutation is induced by the
phage and does not occur until the bacteria are brought into contact with phage. Most
experiments on bacterial variation have left open the two alternatives of spontaneous
mutation on the one hand, and induced mutation or adaptation on the other, and the
interpretation adopted has usually been determined by the previous training of the
individual worker rather than by any compelling evidence provided by the experiments.

Luria and Delbruck, however, in their paper, described a method by which a decision
between the two alternative explanations may be reached, and concluded that the
acquirement of resistance to phage is a spontaneous mutation which occurs during the
growth of the culture and prior to its treatment with phage. Demerec (1945) and Witkin
(1946) have applied the same method to mutants resistant to penicillin and to X-rays
respectively, and have concluded that these changes also are spontaneous mutations
occurring independently of the penicillin or of the radiation respectively.

The principle of Luria and Delbruck's test is as follows. A culture of (say) 10 9 bacteria
is divided into (say) ten equal portions which are separately tested for phage-resistant
organisms by plating out on a phage-impregnated medium. A small number is found in
each of the ten portions, and the numbers are found experimentally to be distributed with
a variance approximately equal to the mean. This result is not surprising on either hypo-
thesis. On the spontaneous mutation theory, we suppose that mutations to phage
resistance occurred from time to time during the growth of the culture. All the bacteria
produced by subsequent divisions of a mutant bacterium were similarly phage resistant.
Thus the culture of 10 9 bacteria contained a certain number of phage-resistant bacteria,

* [Note by C. A. C. A few days before Dr Lea's untimely death in June 1947, the manuscript and the calcula-
tions reported here had just been completed. It was Dr Lea's intention to make further experiments more
suitable to a test of the theory outlined in this paper. These experiments cannot now be made, but it has been
thought wise to publish the theory and numerical tables because of their value to other investigators.]

[Reprinted l>y permission of the Cambridge University Press from Journal of Genetics
49 : (3) 264-285, December, 1949]

2 4

D. E. Lea and C. A. Coulson 265

being either bacteria which had recently undergone mutation, or bacteria derived from
the division of mutants which arose earlier in the growth of the culture. When the culture
was divided into ten equal portions the phage-resistant organisms were distributed
at random between the ten portions. We may expect, therefore, the numbers in the
different portions to fall in a multinomial distribution with variance nearly equal to the
mean.

On the adaptation or induced mutation theory, it is supposed that no phage-resistant
bacteria arose during the growth of the culture. The ten portions, at the time of plating
out, each contained 10 8 normal bacteria and no resistant bacteria. On being brought into
contact with the phage most were lysed, but a few were able to adapt themselves to the
phage (or the phage-induced mutations in them). The probability of this process is very
small, but was presumably the same for all the bacteria. On this theory, therefore, we
expect the number of resistant colonies on the ten parallel plates to be distributed in a
Poisson distribution with variance equal to the mean. Either theory is thus capable of
accounting for the experimental variance, and this experiment alone does not make
possible a decision between the two theories.

A second experiment is now made in which (say) ten cultures, of (say) 10 8 bacteria are
tested for phage-resistant organisms. On the adaptation or induced-mutation theory this
experiment is not essentially different from the preceding one, and we again expect the
numbers of phage-resistant colonies on the ten test plates to be distributed in a Poisson
distribution with variance equal to the mean. For, on this theory, the phage-resistant
mutants do not appear until the bacteria are plated out on the phage-impregnated medium,
and there can be no relevant difference between a culture of 10 9 bacteria divided into ten
equal portions, and ten separately grown cultures of 10 8 bacteria.

In practice a very different result is obtained : the distribution obtained is much wider
than in the former experiment, and has a variance many times — perhaps fifty times —
the mean.

On the spontaneous mutation hypothesis a very wide distribution of the number of
phage-resistant bacteria in parallel cultures is to be expected. The reason is that not only
do the parallel cultures differ in the numbers of mutations which have occurred, but also,
and much more importantly, they differ in the stages at which the mutations occurred.
If a mutation occurs towards the end of the growth of a culture, it will give rise to one
phage-resistant organism, but if it occurs early in the growth, say when the culture is only
one-hundredth of its final size, it will give rise to a large number of phage-resistant
organisms. Thus even in cultures in which equal numbers of mutations have occurred, the
numbers of phage-resistant organisms will usually be widely different.

It is evident, therefore, that the hypothesis that spontaneous mutation to phage
resistance occurs during the growth of the culture before it is brought into contact with
the phage is in qualitative agreement with the experimental result, while the alternative
hypothesis of mutation induced by the phage, or adaptation of the bacterium to the phage,
is not. Luria and Delbruck's method thus provides, for the first time, a clear means of
distinguishing between the two hypotheses.

As left by Luria and Delbruck, the method is a qualitative one, since they do not derive
the shape of the distribution to be expected on the spontaneous mutation theory. They do
derive expressions for the mean and variance of the distribution, but as they point out, on
account of the extreme skewness of the distribution, the mean and variance are very

25

266 Distribution of numbers of mutants in bacterial populations

inefficient statistics for estimating the parameters of the distribution from experimental
results, or for testing the agreement of experiment and theory.

The purpose of the present paper is to extend Luria and Delbruck's method by calculating
the form of the distribution of numbers of mutants in parallel cultures to be expected on
the spontaneous mutation theory, so making the test of the applicability of the spontaneous
mutation theory a quantitative test. Statistically efficient methods of deducing the
mutation rate from experimental observations are also discussed.

The distribution
First method
During the active growth of a culture, the number of organisms increases as an exponential
function of the time, and may be represented as

n = eP*, (1)

there being one organism at time t = 0.

Thus dn = pndt. (2)

If a is the mutation rate, defined by the relation that a dt is the probability that an
individual phage-sensitive organism shall undergo mutation to phage resistance in time dt,
n<x.dt is the mean number of mutations which occur in time dt. (Strictly, since n is the total
number of organisms, we should subtract from n in this formula the number of resistant
organisms, but in practice the number of mutant organisms in a culture is a minute
fraction of the total number.) Hence the mean number (m) of mutations which will have
occurred in the culture by the time it has grown to size n at time t is

noidt = - 5 \ dn = -(n-l).
Jo pji R

Since at all relevant times n much exceeds unity, we may write

m = -n (3)

for the mean number of mutations in the culture by the time it has attained size n. The
mean number of mutations which occur while the culture grows from n x to n 2 organisms
is evidently

(4)

P?)

If it should happen that the mutation rate a and the growth rate fi are equally affected
by factors such as nutritional conditions and density of population which affect j3, then
the mutation rate per generation, though not per unit time, will be independent of these
factors, and a//3, which may be regarded as the probability of mutation per division, will
be constant even though a and /S are not. Under these conditions equations (3) and (4)
can be derived without the assumption of exponential growth.

We must distinguish between the number of mutations which have occurred, and the
number of mutants, the latter being derived not only by mutation of the normal bacteria
but also by division of bacteria which have suffered mutation earlier. If r (r ^ m) is the
number of mutants in the culture at a given time, and if we assume that the division rate of
the mutant is the same as that of the normal bacteria, then rfidt = rdn/n is the probability

26

D. E. Lea and C. A. Coulson 267

that one of the mutants shall divide in time dt. Similarly, (r-1) dnjn is the probability
that one of the mutants shall divide in time dt in a culture containing (r-1) mutants.

Let p r (a function of n) be the probability that a culture of n bacteria grown from
a single bacterium at t = shall have r mutants (i.e. of a large number of cultures of
n bacteria, a proportion p r of cultures will have r mutants).

Consider the proportion Pr + ^dn of cultures which, at time t + dt, when the culture

size is n + dn, have r mutants. These will be derived from:

(a) cultures which, at time t, had r-1 mutants and in which a mutation occurred in the

interval dt;
(6) cultures which, at time t, had r — 1 mutants and in which a mutant divided in the

interval dt;
(c) cultures which, at time t, had r mutants and in which neither mutation nor division
of a mutant occurred in the interval dt ;
providing that the interval dt is small enough. For (<x/]8) dn and rdn/n to be much less than
unity, the possibility of more than one of the rare events mutation and division of a
mutant in the interval dt can be neglected. We see, therefore, that

dp r a r /a r — 1\

sothat sV + » Pr= M£ + ~r

Making, from (3), the substitution m = (a//3) n, we have

dm rT m \ m J

Multiplying by the integrating factor e m we have

^ r +-?r = ?r-i( 1 + r ). where q r = e m p r .

dm m \ m J

Now m is the mean number of mutations which have occurred in a culture by the time it con-
tains n bacteria. Therefore e~ m , the first term of the Poisson distribution, is the probability
p that no mutation shall have occurred. Thus q Q = e m p = 1 for all values of ra. Evidently
initially, when m ■= 0, q r = for all r > 0.

Starting from q = l we can calculate q lt q 2 , q 3 , etc. in succession from the differential
equation (6). Thus:

-^ + —0, = !, whence o, = Am ;

dtn m

(5)

(6)

dm m

dm

H — <7 2 = \m 1 1 + — ) , whence q 2 = hn + hn 2 ;
m \ m)

3 / 2\

+ — q z = (\m + \m 2 ) 11+—), whence 5 , 3 = Y^w+Y^m 2 4-^

Evidently q r is a polynomial in m, with powers ranging from 1 to r.

Writing ?r =2C,, r ^ = C 1)r m + C 2ir | [ %...+C r , r ^, (7)

268 Distribution of numbers of mutants in bacterial populations

do r yy\) ~

we have j- L = I! C jr —. — — . (8)

dm j=l hr (j-l)\

Inserting (7) and (8) in (6), and equating coefficients of mP^fj !, we have

ti+r) C j>r =jC,-i,r-i+(r-l) <W (9)

With this recurrence relation a table of C jr may be drawn up. Such a table, for values
of r ^ 10, is given in the Appendix.

Evidently C ^ = 7^T)> C r,r = 2 ~ r - ( 10 )

From (6) and (7), p = e- m , and for r^ 1

fr=:£C, >r (^^ (11)

A generating function for p r

oo

Define a function f (x, m) = q + q 1 x + q 2 x 2 + ... = 2 q r x r . (12)

r=0

We have %=I,raf- 1 q r and J^ = Zaf ^ . (13)

ox dm dm

Multiplying equation (6) by x r and summing for all r we have

Saf 4 1 + ~ 2> x r ~ x q r = x Saf- 1 a,, + - 2(r - 1 ) a;'- 2 ? r _ ,,
dm m 2r a 1 m

9/" a; 3f a; 2 3/*

or, using (13), -*- +-J- = x f+- -f,

cm mox mox

whence -?- + — (1 -a;) -^- = x, (14)

dm m dx

where* </> = log/. This equation is satisfied by

<f> (x, m) = mifj (x),
providing ip + x (l—x) tfj' = x,

i.e. + / . =- .

x(l—x) l—x

Multiplying by x/( l—x) and integrating

rb^iV 108 * 1 -*'- 1 -

the integration constant —1 being introduced since when a: = 0, f=q = l, so that ^> =
and so i/r = 0. Thus

0=1+ __i og(1 _ x)= _ + _ + __ + ...,

whence /= e w, A = e m ( 1 - x) m &-*te,

so that p r = e~ m q r is the coefficient of x r in the expansion in ascending powers of x of

of e -.«exp[,„( I ^ + -^ + ...)

(l-i)" 11 -* or

* log moans natural logarithm to base e = 2-718... throughout this paper.
28

(15)

D. E. Lea and C. A. Coulson 269

/ x x 2 \ m ( x x 2 \ 2 m m 2 ,._.

,. e . of ^ + (_ + _ + ..J^_ + (_ + _ + ...j«--_ + .... (16)

Comparing with (11), it is evident that C jr is the coefficient of x r in the expansion of

(JL+^L+.Y. (17)

\1.2 2.3 /

It will be shown later that Cj r is the probability that a culture in which exactly./ muta-
tions have occurred shall contain r mutants. If we define

A>= s c itTt

r=j

Dj r is the probability that a culture in which exactly j mutations have occurred shall
have ^r mutants. Summing equation (9) over all r between ^ and r leads to the following
recurrence relation between the D i<r ,

(r+fiC^JiDj.^-D^), (18)

or (r+jjD^rD^+jD^^. ^ (19)

Equation (18) is useful as an arithmetical check during the computation of the C jr by
means of equation (9).

It is sometimes convenient to discuss P r denned as

P r =Z Pr , (20)

r=0

P r being the probability that a culture shall contain 0, 1, 2, . . ., or r mutants, i.e. any number
of mutants up to r. P = e~ m ; P r =l for all r at m = 0. The following equations can be
readily deduced by summing (5) and (11) over all r between and r:

p ^ + iM^j)-

(22)

Second method

An alternative method of calculating p r , the probability of a culture having r mutants,
is instructive. A mutant appearing by mutation any time after the culture has passed
the size \n will not have time to divide by the time the culture size reaches n, and will
therefore contribute 1 to the final complement of mutants. A mutant appearing by
mutation during the period in which the culture grows from \n to \n will have time to
divide once only and will therefore contribute 2 mutants to the final complement. A
mutant appearing by mutation in the period in which the culture grows from \n to \n will
have time to divide twice only and will contribute 4 mutants to the final complement,
and so on.

Confining attention for the moment to all those cultures in which exactly one mutation
occurs during the growth of the culture from one organism to n organisms, in one-half of
the cultures the mutation will occur while the culture is growing from \n to n, in one-quarter
of the cultures the mutation will occur while the culture is growing from \n to \n, and so
on (compare equation (4)). Or, in other words, in one-half of these cultures the mutation
occurs at such a time that it gives rise finally to 1 mutant, in one-quarter of the cultures

29

270 Distribution of numbers of mutants in bacterial populations

the mutation occurs at such a time that it gives rise finally to 2 mutants, in one-eighth at
such a time that it gives rise to 4 mutants, and so on.

According to this argument, a mutation is necessarily represented, by the time the
culture has reached size n, by a clone of 1 or 2 or 4 or 8, etc., organisms, there being no
intermediate numbers. This would be so if divisions in a clone were synchronous. It is
probably true that clones of 3, 5 or 7 cells will be less common than clones of 2, 4 or 8 cells
(cp. Adolph & Bayne-Jones, 1932), but rather than make the extreme assumption that
only integral powers of 2 are to be considered it is probably preferable to neglect this fact
and to assume that the frequency of clones of different sizes is a smooth function of
clone size.

A clone which, by time t, contains v mutants will have originated when the number of
bacteria in the culture was about njv. We thus replace the subdivision of the growth of the
culture into generations by subdivision into intervals in which the population increased
from \n to n, from \n to \n, from \n to \n, and so on, and suppose that a mutation which
occurred while the population increased from n/(v + 1) to njv is, by the time the population
has grown to n, represented by a clone of v mutants. Now, of those cultures in which
exactly one mutation has occurred, the proportion in which the mutation occurred while

the culture grew from n/(v + 1) to njv is I I =— — (cp. equation (4)). This may

be represented as the coefficient of x v in the generating function

1.2 2.3 * v(v + l)

Considering now all those cultures in which exactly j independent mutations occurred,
the fraction of cultures in which the final number of mutants is v is, from (23), evidently
the coefficient of x v in the expansion of

t_x_ tf_ a?_ y
\l.2 + 2.3 + 3.4 + "'7 '

Now if m is the mean number of mutations per culture, the proportion of cultures in
which exactly j mutations occurs is e~ m — . Thus the proportion of all cultures in which the
final number of mutants is r is the coefficient of x r in the expansion of
f° / x x 2 a; 3 V m mi

Thus p r , the probability of a culture having r mutants, m being the mean number of
mutations per culture, is the coefficient of x r in the expansion of

e-exp[m( I ^ + ^+...)]

in agreement with equation (15).

Arithmetical procedure
By means of the recurrence relation (9) and the boundary values (10) a table of values
of Cj r has been computed for all (integral) values of j from 1 to 36 and of r from 1 to 64,
subject to r^j. Equation (18) was employed as a check on the arithmetic at ^=63. Cj r

30

D. E. Lea and C. A. Coulson 271

is the probability that a culture in which exactly j mutations have occurred shall contain
r mutants. In practice when comparing the theoretical and experimental distributions the
distributions will always be grouped. For economy of space, therefore, we do not publish
the full table of C jr but give in Table 1 grouped values for r = l, 2, 3, 4, 5-8, 9-16,
17-32, 33-64 and >64.

Table 1. C Ur

Cj T is the probability that a culture in which exactly j mutations have occurred shall have r mutants (r^j).
For values of r greater than 2, the values of Cj r have been grouped. Thus the numbers in the column headed

r=32

' 17-32' are values of S C.- r , See also the Appendix for certain other values of C j>r .

r=17 '

/V

1

2

3-4

5-8

9-16

17-32

33-64

>64

i

0-5000

0-1667

01333

0-0889

00523

00285

00149

00154

2

0-2500

0-2778

0-2118

0-1272

00671

00335

00325

3

0-2500

0-3100

0-2161

01161

00563

00515

4

—

—

00625

0-3093

0-2986

01735

00834

00726

5

.

0-2060

0-3479

0-2353

01149

00960

6

—

—

0-0885

0-3445

0-2949

0-1504

01216

7

.

—

0-0260

0-2907

0-3441

0-1894

01497

8

—

—

—

0-0039

0-2100

0-3753

0-2305

0-1804

9

01312

0-3830

0-2721

0-2137

10

—

0-0718

0-3665

0-3122

0-2495

11

.

—

00342

0-3294

0-3486

0-2878

12

_

—

—

00138

0-2790

0-3789

0-3284

13

0-0045

0-2234

0-4010

0-3711

14

—

0-0011

01698

0-4135

0-4157

15

. .

.

—

00002

0-1228

0-4155

0-4616

16

—

—

—

—

00000

00846

0-4071

0-5084

17

,

00553

0-3891

0-5555

18

.

—

—

00343

0-3633

0-6024

19

—

—

—

00200

0-3315

0-6485

20

—

—

—

—

—

0-0109

0-2961

0-6930

21

,

.

00055

0-2591

0-7354

22

—

00026

0-2222

0-7752

23

.

—

0-0011

01871

0-8118

24

.

_

—

—

—

00004

01546

0-8450

25

00001

01254

0-8745

26

—

00000

0-0998

0-9001

27

—

—

— .

0-0779

0-9221

28

—

—

—

—

—

0-0596

0-9404

29

0-0447

0-9553

30

_

—

00327

0-9673

31

—

—

00234

0-9766

32

—

—

—

—

—

00164

0-9836

33

00111

0-9889

34

.

.

.

_

0-0074

0-9926

3a

—

00048

0-9952

36

_

—

—

—

0-0030

0-9970

p r is the probability that a culture of such a size that the mean number of mutations is

r / m?\

\ shall contain r mutants. p = e~ m and for r^l, p r = 2 C jr \e~ m ^-A. For any given

j = \ \ J • I
alue of m, p r can be calculated with the aid of the table of C jr , for the same groupings

iir

of r. The calculation is facilitated if a table of Poisson coefficients e~ m -rr is available

(Molina, 1942). In Table 2 values of p r are given for a number of values of m from 0-05
to 15.

The figures in Tables 1, 2 and 4 are liable to occasional rounding errors of one unit in
the last decimal place.

Journ. of Genetics 49 19

3*

272 Distribution of numbers of mutants in bacterial populations

Limiting form of distribution for large numbers
Table 2 provides the means of testing the agreement between theoretical and experimental
distributions in experiments in which the mean number of mutations per culture is 15 or
fewer, and in which a minority of the cultures have more than 64 mutants. To extend
Tables 1 and 2 by use of the recurrence relation (9) to cover experiments in which the
mean number of mutations per culture considerably exceeds 15, and to subdivide the class

Table 2. p r

p r is the probability that a culture shall have r mutants, the average number of mutations which have occurred
per culture being m. For values of r greater than 2, the values of p r have been grouped. Thus the numbers

r=32

in the column headed ' 17-32' are values of S p r .

m\r

1

2

r=17

3-4

5-8

9-16

17-32

33-64

>64

0-05

0-9512

00283

0-0082

00067

00045

0-0026

00014

0-0008

00008

0-10

0-9048

00452

0-0162

0-0134

0-0090

00053

0-0029

0-0015

00015

015

0-8607

00646

00239

00200

00137

0-0081

0-0044

00023

00023

0-20

0-8187

0-0819

00314

00267

00184

0-0109

0-0059

00031

0-0031

0-25

0-7788

00974

0-0385

00332

00231

0-0138

0-0074

00038

00039

0-30

0-7408

01111

00454

0-0397

00279

00167

00090

0-0046

00047

0-35

0-7047

01233

00519

0-0462

0-0328

00196

0-0106

00055

0-0055

0-40

0-6703

01341

0-0581

00525

00376

00226

00122

00063

00063

0-45

0-6376

01435

00640

0-0587

00425

0-0257

00139

00071

00071

0-50

0-6065

01516

00695

0-0648

0-0475

0-0288

0-0155

0-0079

0-0079

0-55

0-5769

0-1587

0-0747

0-0707

0-0524

00319

00172

0-0088

0-0087

0-60

0-5488

01646

0-0796

00765

00573

00351

00189

00096

0-0096

0-65

0-5220

0-1697

0-0841

00821

0-0622

00383

0-0207

00105

00104

0-70

0-4966

0-1738

0-0884

0-0876

0-0672

0-0415

0-0224

00114

00112

0-75

0-4724

01771

00923

0-0928

00721

0-0448

00242

00123

00120

0-80

0-4493

0-1797

0-0959

0-0979

00769

0-0482

0-0260

00132

00129

0-85

0-4274

0-1817

0-0992

01028

0-0818

00515

0-0279

00141

00137

0-90

0-4066

01830

0-1022

0-1076

0-0866

0-0549

0-0297

0-0150

0-0146

0-95

0-3867

01837

01049

0-1121

0-0914

0-0583

00316

0-0159

0-0154

10

0-3679

01839

01073

01164

0-0961

0-0617

00335

0-0168

0-0163

1-2

0-3012

0-1807

01145

01317

01144

00757

00414

0-0207

00198

1-4

0-2466

0-1726

0-1180

0-1438

01316

0-0899

00496

0-0247

00233

1-6

0-2019

01615

01184

01528

01473

0-1041

0-0581

0-0288

00270

1-8

01653

0-1488

0-1165

0-1587

01615

01184

0-0670

0332

0-0307

20

01353

01353

01128

01620

0-1738

01324

0-0761

00377

00345

2-2

0-1108

01219

0-1077

01629

0-1844

01462

0-0855

00423

00383

2-4

0-0907

0-1089

01016

01617

01930

01595

00951

0-0471

0-0423

2-6

00743

0-0966

0-0949

01587

0-1998

01723

01049

00521

00464

2-8

0-0608

00851

0-0880

01543

0-2048

01844

01149

0-0573

00505

30

0-0498

00747

0-0809

0-1487

0-2080

01958

01249

00626

0-0547

3-2

0-0408

0-0652

00739

0-1421

0-2095

0-2063

01351

0-0680

0-0590

3-4

00334

0-0567

0-0671

01350

0-2095

0-2159

01452

00736

00634

3-6

00273

00492

0-0607

01274

0-2081

0-2246

01554

00794

00679

3-8

00224

00425

00545

01 195

0-2054

0-2324

01655

00853

00725

40

00183

00366

0-0488

01115

0-2016

0-2391

01755

00913

0-0771

4-2

00150

00315

0-0436

01036

0-1969

0-2447

0-1854

0-0975

0-0819

4-4

00123

00270

0-0387

0-0958

0-1912

0-2493

0-1951

01038

0-0867

4-6

0-0101

00231

0-0343

0-0882

01849

0-2529

0-2046

0-1102

0-0917

4-8

0-0082

00198

00303

00809

0-1780

0-2555

0-2139

01167

00967

5

00067

0-0168

0-0267

00739

0-1707

0-2571

0-2228

01234

0-1018

00025

00074

00136

00451

01311

0-2516

0-2620

01577

01289

7

00009

00032

0-0066

00258

00936

0-2285

0-2898

01932

0-1584

8

0-0003

00013

0-0031

00141

0-0630

01953

0-3043

0-2283

01903

9

00001

00006

00014

00074

00404

0-1586

0-3056

0-2615

0-2244

10

00000

00002

00006

0-0038

00249

01233

0-2950

0-2913

0-2608

11

00000

00001

0-0003

00019

00148

00923

0-2751

0-3165

0-2991

12

0-0000

0-0000

0-0001

00009

0-0086

00668

0-2486

0-3359

0-3391

13

0-0000

00000

0-0001

00004

0-0048

00469

0-2185

0-3489

0-3805

14

00000

0-0000

00000

00002

00027

00320

0-1871

0-3552

0-4228

15

00000

0-0000

00000

00001

00014

00214

01564

0-3549

0-4657

32

D. E. Lea and C. A. Coulson

273

>64 mutants into further classes, e.g. 65-128, 129-256, 257-512 mutants, etc., would
involve an impracticable amount of arithmetic. An attempt was therefore made to find
asymptotic formulae for p r or P r valid for large values of m. We have not succeeded in
finding explicit formulae, but have obtained some information on the form of the function.
If we consider P r as a continuous function of the two variables r and ra, then for values
of r>l we have approximately P r — P r _ 1 = dP r /dr. Thus equation (21) approximates to

dm \ ml dr

which is satisfied by

K)
■"I 1

= 0,

(24)

•).

(25)

where F is any function.

In Fig. 1 we have plotted P r (derived from Table 2, i.e. based on the recurrence relation)
against (r/m — log m) for r = 8, 16, 32 and 64, using five values of m (viz. 4, 6, 8, 13, 15)

10

1 1

1 1 1 1 1

1 1 1 1 1 I 1

1 1 1

0-8

-

0-6

/

• m=4

-

A m = 6

0-4

-

X m = 8

-

-

m=13

-

0-2

-

+ m = 15

-

*JTi

i i i i i

1 l 1 1 I I 1

1 1 1

10

r/m -log m

Fig. 1. P r , for different r and m, is a function of rjm -log m. The points are plotted for
r = 8, 16, 32 and 64, and with wi=4, 6, 8, 13 and 15.

selected so that the twenty points are conveniently spaced. It is seen that the points lie
quite well on a single curve, showing that these values of r are large enough for equa-
tion (24) to be a satisfactorily close approximation to equation (21). The smooth curve
in Fig. 1 is thus a graph of the function F which enters into equation (25).

For any given value of m, {r/m — log m) is evidently distributed in a skew distribution
about a median 1-24. We have found by trial that the derived variate (r/m -log m + 4-5)- 1
is distributed in a distribution rather closely approximating to a Gaussian distribution of
standard deviation 0-086. This is shown by the closeness with which the points in Fig. 2
lie on a straight line. The points in Fig. 2 are derived from those of Fig. 1 by transforming

the ordinates to probits, defined by the relation

V(2

L_ [ v ~

e~ iyi dy = P r , where y is the

probit corresponding to P r . (Tables of probits are given in Fisher & Yates, 1938.) Also,

19-2

33

274 Distribution of numbers of mutants in bacterial populations

the abscissae are transformed to values of (r/m — log m + 4-5) -1 . Fig. 2 shows that,
approximately,

x= I- ~ — -0-174) /o-086 = (- — ~ 6 — — -2-02)

\r/m — log m + 4-5 / / \r]m - log m + 4-5 /

(26)

is a normal deviate.

We conclude, in this semi-empirical manner, that when the spontaneous mutation
theory is to be compared with experiments falling outside the scope of Table 2 (i.e. experi-
ments in which cultures containing more than 64 mutants are frequent), it will be satis-
factory for practical purposes to suppose a?= I — ^ -——2-02 J to be normally

7

5 -

4 -

•

/n*4

A

m=6

X

m = 8

m=13

+

m=15

0-5

r/m -log m+4-5

Fig. 2. l/(r/m - log m+4-5) is distributed in an approximately normal distribution. The
points are plotted for r = 8, 1(5, 32 and 64, and with m=4, 6, 8, 13 and 15.

distributed with unit variance about the value 0. r is the number of mutants in an
individual culture, m is the mean number of mutations per culture in the batch of parallel
cultures.

The median of the distribution satisfies the relation
11-6

2-02 = 0, i.e. at the median r/m — log m = 1-24.

(27)

r/m — log m + 4-5

This equation provides a means of making a first estimate of m from the count of the
number (r) of mutants in the median culture of the batch.

The quartiles of the distribution (i.e. values of r making P r = 0-25 or 0-75) satisfy the
relations :

at quartiles : r/m - log m = - 0-2 and r/m — log m = 4- 1 , (28)

34

D. E. Lea and C. A. Coulson 275

which relations may be used as a first test of whether the spread of an experimental
distribution is comparable with the theoretical spread.

All the relations in this section are approximations, to be used only when dealing with
experiments which lie outside the scope of Table 2. The approximation should not be used
for the extreme ends of the distribution, e.g. for values of P r exceeding 0-95 or less than 0-05.

The estimation of mutation rate from experimental observations
m from the mean number of mutants per culture
As shown by Luria and Delbruck, the mean and variance of the distribution can be simply
calculated, without knowing the distribution p r , as follows:

While the culture grows from n x to n 1 + dn 1 , the mean number of mutations will be
(m/n) dn x (cp. equation (4)), the actual number being distributed in a Poisson distribution
about this mean with variance also (m/n) dn x (since the variance of a Poisson distribution
is equal to the mean). The contribution to the final number of mutants (when the culture
size is n) will be n/n x mutants for each mutation. Thus the contribution to the final number

of mutants will be distributed about a mean dn, with a variance I — ) — dn,. Thus

n x n L \n x J n

the mean of the required distribution is

f=\ dn,=m log n, (29)

J l n x n

and the variance of an individual determination of r will be

oZ^ri-Y-d^mn* (30)

J i W n

We can confirm that our distribution p r yields the same mean and variance. The mean is

r = Xrp r .

Since j9 r is the coefficient of x r in the expansion of e _w exp m I r— + .y^ + ••• I (equation (15)),
e
S ?r af = exp[,„( 1 ^ + 2 4 + ...)]. (31)

Differentiating

^f-^m g+f+f +~) exp [m (o+0 + -)]- W

Inserting x=l, and putting - + - + ... +- = log n and 7-^+^+ ...4=1, we have

r = T,rp r = e~ m Y*rq r = m log n.
Again, multiplying (32) by x 2 and differentiating,

Sr(r + l)z' ?r = m (x + ^ + £3+...)+m^

Inserting x = 1 , Sr(r + l)j r = me m (n + m log 2 n)

T,r(r + l)p r = mn + m 2 log 2 n.

* But see Appendix: the correct value is a 2 = 2mn.

and since q r = z e m p r , we have

35

276 Distribution of numbers of mutants in bacterial 'populations
Now the variance cr 2 = 2(r — f) 2 p r =Hr 2 p r — 2r"Zrp r +r 2 Hp r

= Er (r + 1) p r — r — r 2 .

Thus o 2 = mn — m\ogn, i.e. o 2 = mn since npl.

Since f = mlogw, a possible method of determining m (and hence the mutation rate)
experimentally would be to divide by log n the mean number of mutants per culture
experimentally determined in a batch of N parallel cultures. However, on examination
it appears that the precision of the estimate of m given by this method does not increase
with increase of N. For it is evident that the total numbers of mutants in batches of
N parallel cultures each of size n will be distributed (from batch to batch) in much the
same way as the numbers of mutants in parallel cultures of size nN. The mean number of
mutations in a culture of size nN will be mN (since the mean number of mutations is
proportional to the size of the culture, cp. equation (3)), and hence by application of (30)
the variance of the number of mutations in cultures of size nN is mN .nN = mnN 2 . Thus
the total number of mutants in a batch of N cultures of size n is distributed from batch to
batch with a variance mnN 2 . A fraction 1/N of this total number (i.e. the mean number
per culture derived from a count of N cultures) is therefore distributed with variance mn.

Thus we see that the variance of the mean number r of mutants in N cultures is no smaller
than the variance of the number of mutants in an individual culture, which shows that
however many cultures are averaged, no improvement in precision is obtained over the
use of a single culture selected at random. Consequently, the mean number of mutants per
culture is an extremely inefficient statistic from which to calculate the mutation rate. If,
nevertheless, this method of estimating m is employed, the variance ( a 2 m ) of the estimate
of m will (from (29) and (30)) be

mn m)

(ioi^ 2 ' (33)

independent of the number N of cultures averaged.

m from proportion of cultures without mutants
In view of the unsuitability of r as a means of estimating m from numerical data, Luria
and Delbruck proposed its estimation by equating e~ m to the proportion of cultures
experimentally determined to be without mutants. In a batch of N parallel cultures, in
which the mean number of mutations is m per culture, the expected number of cultures
without mutation is Ne~ m , the actual number being distributed about this mean in a
binomial distribution having a variance Ne~ m (1 — e~ m ). Thus the variance of the estimate

e~ m is e~ m (1 —e~ m )/N. Since , = —e m , the corresponding estimate of m has a variance

(ct^) which is e 2m times as great, i.e.

•t-V- (34)

Thus the standard error (a m ) in the estimate of m is given by

o m jm thus varies with m. It has a minimal value when m = 1-594, when a fraction 0-2032
of cultures have no mutants. At m = 1-594, (a m /m) 2 takes the value 1-544/2V. At small or

36

D. E. Lea and C. A. Coulson 277

large values of m (e.g. when the fraction of cultures without mutants exceeds 0-9 or is less

than 0-01), the value of (ar m /m) is much increased. Fig. 2>A shows graphically — sjN as

m

a function of m.

The low precision at small values of m is to be attributed simply to the fact that an

experiment in which the great majority of the cultures have no mutants does not provide

much precise information about the mutation rate. The reduced precision at high values

of m is, however, to be ascribed to the fact that this method of determining m does not

1000

Fig. 3. The precision of the estimate of m derived by various methods: A, the method of the proportion of
cultures without mutants; B, the method of the median; C, the method of S[x]-0; D, the method of
maximal likelihood.

make full use of the experimental data, and in these cases more suitable methods, which
we shall describe, enable a more precise estimate of m to be made from the same data.

mfrom the median
When the mutation rate is to be deduced from an experiment in which all, or nearly all,
the cultures had mutants, so that the method just discussed is inapplicable, a very
convenient method is to deduce m from the median of the distribution. The counts of the
numbers of mutants in N parallel cultures are arranged in ascending order, and the middle
one selected. The count in this culture is an estimate of r , the median of the distribution
of r . Since we know that (approximately) the derived variate

4

Jm — log m + b

with a = 11-6, 6 = 4-5, c = 2-02

(36)

37

278 Distribution of numbers of mutants in bacterial populations
is normally distributed about median 0, it follows that

^logm=- -6 = 1-24. (37)

This equation enables an estimate of m to be made from an experimentally determined
value of r . With its aid Table 3 has been constructed, which enables m to be obtained for
any value of r up to 4400. While the derivation of m from the median is not the most
efficient way of utilizing the experimental data from a statistical standpoint, it is the
quickest satisfactory method, and is useful for making a preliminary estimate even if
a more elaborate method is to be employed in making the final estimate.

Table 3. Preliminary estimation of mfrom median value of r

Thus if the middle culture of the series has 50 mutants, interpolation in the table between r =49-2 and
r =55-8 gives /•„/>« =3-81, so that m = 50/3-81 =13-1. This is the mean number of mutations per culture.

r

rjm

r o

r jm

r

rjm

r o

r \m

1-4

1-3

15-3

2-9

117

4-5

787

61

lti

1-4

17-4

30

132

4-6

884

6-2

1-9

1-5

19-9

31

150

4-7

993

6-3

2-3

1-6

22-7

3-2

169

4-8

1115

6-4

2-7

1-7

25-9

3-3

190

4-9

1251

6-5

3-2

1-8

29-5

3-4

215

5-0

1404

6-6

3-7

1-9

33-5

3-5

242

51

1575

6-7

4-3

20

381

3-6

273

5-2

1767

6-8

5-0

21

43-3

3-7

307

5-3

1981

6-9

5-7

■>.•>

49-2

3-8

346

5-4

2221

7-0

6-6

2-3

55-8

3-9

389

5-5

2490

71

7-7

2-4

03-2

4-0

438

5-6

2791

7-2

8-8

2-5

71-6

41

493

5-7

3127

7-3

101

20

811

4-2

554

5-8

3503

7-4

11-6

2-7

91-7

4-3

623

5-9

3924

7-5

13-3

2-8

104

4-4

700

6-0

4395

7-6

The precision of an estimate of m made in this way from counts of N cultures may be
determined by calculating a m jm. We shall make use of the approximate result that x is
distributed in a normal distribution with unit variance. The probability of .z lying between

1 2 1 C x

x and x + dx is -..- - <H r2 dx. The probability of its lying between and x is , e~ ix2 dx,

vw' T J yl^Jo

and for observations in the neighbourhood of the median (^ 2 <^1) we may write this as

j~= -. Thus |-+ llct . 1 is the probability of getting an observation ^x, and I-— t^—A

V(2tt) \2 V(27r)/ j h \2 v '(2tt>;

is the probability of getting an observation ^ x. Thus the probability that, of N = 2s + 1
observations, s shall be ^.r, s shall be ^x, and one shall be between x and x + dx is
(providing x is in the neighbourhood of the median)

(2x4-1)1/1

(2s + l)!

(U x —\ 8 ( l - r V s ,h ' = (2 * +1)! /i %*\' < lr

\2 s l(2n)f [■> j(2n)J 4(2n) (2°s\f\ ir ) ~Jfrr)

for s^> 1. Thus the median value of a set of A' = 2,s + 1 values of x is distributed about x =

.,, 77 77

with variance — = — .

4s ' 2A

3»

D. E. Lea and C. A. Coulson

279

It follows that if equation (37) is used to deduce from an experimentally determined
median value r an estimate of m, then this estimate will be subject to a variance

-^/i*—\ > tne suffix denoting evaluation at the median. Differentiating (36),
2N/ \omjo

dx (x + c) 2 (1 -6 + log m) x + c

S—L + .

= (x + c)''

d x + c

(39)

where d = (l —6 + log m)/a. Hence at the median x = 0, 1^— I =— (cd+l). Thus the
variance a 2 m of the estimate of m derived from the median is given by

\m) N (l+ajc-b + \ogm) 2 N c 2 (cd + l)*'
or, inserting the values of a, b, c from (36),

\mj iV(2-24+loj

mf

(40)

(41)

make an

Fig. 35 is a plot of ( — I *JN against m as given by (41). Having used Table 3 to

estimate of m from the observation of the median value of r, Fig. 3 B is consulted to obtain
the standard deviation to be ascribed to the estimate of m.

m from S [x] =
An alternative method of estimating m from experiments in which all or nearly all of
the cultures have mutants is the following. Since x is distributed approximately normally
about the value x — 0, the mean value of x is zero. An estimate of m from a set of N
observations can therefore be made by finding that value of m which makes

S[x]=S

rjm — log m + b

0,

(42)

the summation being over the N experimental observations. In using this method a first
estimate of m is made by the median method. Inserting this value of m into (36), each
experimental value of r is converted into a value of x, and the sum S [x] formed. A series
of adjacent values of m are then tried, and the value of m which makes S[x]=0 found
(e.g. by plotting S [x] against m).

The estimate of m obtained in this way is a little more precise than that based on the
median. The mean $ [x]/N of a batch of N independent values of x will be distributed
(from batch to batch) with variance 1/iV about a mean zero. Suppose that its value for
a particular batch is 8, so that

S[x] = m.

If m + 8 m is the estimate of m derived from this particular batch (S m being the deviation
between the estimated and true values of m),

h!N=°-

39

280 Distribution of numbers of mutants in bacterial populations

Thus S m — ^Jr — - = — S or approximately 8 m E L— = — 8, where we have replaced the

=—\ of ^— .

Now from (39) we have

^=-{x 2 d + x(2cd+l) + c 2 d + c\,
dm m

and x being normally distributed with unit variance about mean zero,

E[x 2 ] = l and E [x] = 0.

Thus E

?\r

-{d(l+c 2 ) + c},
m

and so S,

m d(l+c 2 ) + c'

The variance of S from batch to batch being l/N, we obtain for the variance( a m 2 ) of m the
relation

fasV-i 1- (43)

UJ iV{d(l+c 2 ) + c} 2 ' <**'

with a = 11-6, 6 = 4-5, c = 2-02, d = (l -6 + log m)/a.

A plot of — ^iV against m as computed by this formula is given in Fig. 3 C. Having

derived m by the method described in this section, the standard deviation to be ascribed
to it is read from Fig. 3 C.

Maximal likelihood method: large counts

None of the methods we have so far described is fully efficient statistically. At the
expense of somewhat more laborious computation a fully efficient estimate of the mutation
rate may be made by employing the method of maximal likelihood. We give two solutions:
one for experiments which fall within the range of Tables 1 and 2, i.e. in which most of the
cultures have fewer than 64 mutants, which is set out in the next section, and one for
experiments falling outside the range of Tables 1 and 2, and for which the approximation
that £ is a normal deviate is employed, which is set out in the present section.

The probability that the number of mutants shall lie between r and r+dr is given
approximately (for r not too small) as

i^*-^)^?*-/* (Say) ' (44)

(45)
(46)

d 1 df _ dx d 2 x Idx

Thus a^ l0g ^Ja^ = ~ X d^i + drJmlTr'

Now x = (-, r^ 7-c) with a = 11-6, 6 = 4-5, c = 2-02,

\rjm - log m + b J

and by differentiating we find

dx . , 9 d x + c d 2 x Idx d 1

dm m m omor\ or mm

40

D. E. Lea and C. A. Coulson 281

where d = (l-b + \og m)ja. (48)

Thus \M-=-{- x (x + c) 2 d-x (x + c) + 2 (x + c) d + 1}. (49)

fdm m K

Now L, the log likelihood, is (apart from irrelevant terms) S [log/], the summation being
for the N observations of r, and the maximal likelihood condition is

oAs

"If

fdm
i.e. S[x(x + c) 2 d + x(x + c)-2(x + c)d-l] = 0. (50)

The routine for applying this method is as follows. Employing the preliminary estimate
of m given by the median method, (48) is used to calculate d, and then (46) is used to
calculate a value of x from each of the N experimental observations of r. For each of these
N values of x the expression

x (x + c) 2 d + x (x + c) -2 (x + c) d-l

is evaluated and the N quantities added. The sum is similarly evaluated for several
adjacent values of m, and by plotting against m (or otherwise) the value of m which satisfies
(50) is deduced.

The variance to be attached to the maximal likelihood estimate of a parameter m is
given by Fisher's formula (cp. e.g. Fisher, 1938)

Ni'

(51)

where i = E

(1 df \ 2 ] /l df\ 2

-rj—\ is the expectation of (->-/-) • Hence, using (49),

im 2 = E[{xH + x 2 (2cd+\) + x(c 2 d + c-2d)-(2cd + \)} 2 }
= d 2 E [x 6 ] + (&c 2 d 2 + Gcd-4:d 2 + l)E [z 4 ]
+ (c*d 2 + 2cH - 12c 2 d 2 -\2cd + U 2 + c 2 -2)E [x 2 ]
+ (ic 2 d 2 + ±cd + l)
+ terms involving odd powers of a;.

Now it is readily shown that x being distributed normally about zero with unit variance,
E [x n ] vanishes for odd n, and

E [x°] = E [x 2 ] =1, E O 4 ] = 3, E [x*] = 15.

Inserting these values in (52) we obtain

im 2 = d 2 (c* + 10c 2 + 7)+d(2c 3 + 10c) + (c 2 + 2),

1

J 2 (c 4 + 10c 2 + 7)+d(2c 3 + 10c) + (c 2 + 2) '

with a = 11-6, 6 = 4-5, c = 2-02, d = (l-b + \ogm)/a.

The part of Fig. 3D to the right of ra = 10 is a plot of I — I JN against m. Having

determined the maximal likelihood estimate of m, as described in this section, the standard
deviation to attach to it is read off from Fig. 3 D.

sothat ft) a 4,-

4i

282 Distribution of numbers of mutants in bacterial populations

Maximal likelihood method: smaller counts

In this section we describe the method of arriving at the maximal likelihood estimate of
m from an experiment falling within the scope of Tables 1 and 2; i.e. one in which the
majority of cultures have fewer than 64 mutants.

p r is the probability of a culture having r mutants. The log likelihood of a set of N values
of r is (apart from irrelevant terms)

L=S [log p r ],

S denoting summation over the N experimental values of r.
The maximal likelihood value of m is that satisfying

am

'^dpr
p r dm

Now from equation (11)

Pr = £ C j>r e-

3 = 1

j\' dm j-_

ZC j>r e
= i

(m 7-1 m>\

SO that ^IrJrZSr m

where t r = Zc,, r (^' { ~j ■ (55)

Thus the maximal likelihood estimate of m is that satisfying

S P-^ -0, (56)

L Pr -

t r has been computed for a range of values of m exactly as described earlier for p r , and
in Table 4 values of {t r —p r )/p r are listed for a range of values of m and for the same
grouped ranges of r as were used previously.

The method of estimating m is therefore the following. A preliminary estimate of m is
obtained either by the median method or by equating e~ m to the proportion of cultures
without mutants. Table 4 is entered at the value of m nearest to this preliminary estimate,
and a value of {t r —p r )/p r read off for each of the N experimental values of r. The N values
are summed. The procedure is repeated for several adjacent values of m, and thence

(graphically or otherwise) the value of m inferred which would make S - — — =0.

The variance of this maximal likelihood estimate of m is given by the relation

2 l

m Ni'

where i- £ X frY = S <^>'

r=0pr\dmj p r

7T3p
Pr

S here means summation over all values of /• from to infinity, and is to be distinguished
from S, meaning summation over the N experimental observations.

D. E. Lea and C. A. Coulson 283

Table 4. ±-& and ^N

This table is used in estimating mutation rate by the maximal likelihood method.
-° — i-5 = - 1 for all values of m.

?n\r

1

2

3-4

5-8

9-16

17-32

33-64

>64

o m JNIm

005

19000

19-723

20-019

20179

20-211

20-176

20124

20-057

4-527

010

9000

9-698

9-998

10-166

10-206

10174

10124

10057

3-239

015

5-667

6-341

6-644

6-821

6-868

6-839

6-790

6-723

2-674

0-20

4000

4-652

4-957

5142

5196

5- 171

5123

5-057

2-341

0-25

3000

3-632

3-939

4130

4191

4170

4123

4057

2116

0-30

2-333

2-946

3-254

3-452

3-519

3-501

3-456

3-390

1-951

0-35

1-857

2-451

2-760

2-964

3038

3023

2-979

2-914

1-824

0-40

1-500

2-077

2-386

2-596

2-676

2-665

2-622

2-557

1-722

0-45

1-222

1-783

2-092

2-307

2-393

2-386

2-344

2-279

1-638

0-50

1-000

1-545

1-855

2-075

2166

2162

2-121

2-057

1-568

0-55

0-818

1-349

1-658

1-882

1-979

1-978

1-939

1-875

1-507

0-60

0-667

1184

1-492

1-720

1-822

1-825

1-787

1-723

1-455

0-65

0-538

1043

1-350

1-582

1-689

1-695

1-658

1-595

1-409

0-70

0-429

0-920

1-227

1-462

1-574

1-584

1-548

1-485

1-368

0-75

0-333

0-813

1119

1-357

1-474

1-487

1-452

1-390

1-332

0-80

0-250

0-719

1023

1-264

1-386

1-402

1-369

1-307

1-299

0-85

0176

0-634

0-937

1181

1-308

1-327

1-295

1-233

1-269

0-90

0111

0-559

0-860

1-107

1-237

1-259

1-229

1-168

1-242

0-95

0053

0-491

0-790

1039

1174

1199

1170

1-109

1-217

10

0000

0-429

0-727

0-978

1117

1145

1117

1057

1194

1-2

-0167

0-228

0-520

0-777

0-931

0-971

0-948

0-890

1-118

1-4

-0-286

0-080

0-364

0-627

0-793

0-845

0-828

0-771

1-059

1-6

-0-375

-0034

0-243

0-508

0-686

0-749

0-736

0-682

1-012

1-8

- 0-444

-0125

0144

0-411

0-599

0-672

0-665

0-612

0-973

2-0

- 0-500

-0-200

0063

0-330

0-526

0-609

0-608

0-557

0-940

2-2

-0-545

-0-262

-0-007

0-260

0-464

0-556

0-560

0-511

0-913

2-4

-0-583

-0-315

- 0066

0-200

0-409

0-510

0-520

0-473

0-888

2-6

-0-615

-0-361

-0119

0147

0-362

0-471

0-486

0-441

0-867

2-8

-0-643

-0-401

-0164

0099

0-319

0-436

0-456

0-413

0-848

30

-0-667

-0-436

-0-205

0-057

0-280

0-404

0-430

0-390

0-831

3-2

-0-688

-0-467

- 0-242

0018

0-245

0-376

0-407

0-369

0-816

3-4

-0-706

- 0-495

-0-275

-0017 .

0-213

0-350

0-386

0-350

0-802

3-6

-0-722

- 0-520

-0-305

- 0-050

0183

0-326

0-367

0-334

0-789

3-8

-0-737

-0-542

-0-333

-0080

0155

0-304

0-350

0-319

0-777

40

-0-750

- 0-563

-0-358

-0107

0129

0-284

0-334

0-305

0-766

4-2.

-0-762

-0-581

-0-381

-0133

0105

0-264

0-320

0-293

0-756

4-4

-0-773

- 0-598

-0-402

-0-157

0-082

0-246

0-306

0-282

0-747

4-6

-0-783

-0-614

-0-422

-0179

0061

0-229

0-293

0-272

0-738

4-8

-0-792

-0-629

-0-441

-0-200

0041

0-213

0-282

0-263

0-730

50

-0-800

-0-642

- 0-458

-0-220

0-021

0197

0-271

0-254

0-722

6

-0-833

- 0-697

-0-530

-0-304

- 0062

0130

0-223

0-219

0-690

7

-0-857

-0-737

-0-584

- 0-369

-0128

0073

0184

0194

0-665

8

-0-875

-0-768

-0-626

-0-422

-0184

0025

0151

0174

0-644

9

-0-889

-0-792

-0-660

- 0-465

-0-231

-0016

0121

0157

0-627

10

- 0-900

-0-812

-0-689

-0-502

-0-272

-0053

0095

0143

0-613

11

-0-909

-0-828

-0-712

-0-534

-0-307

-0086

0071

0131

0-600

12

-0-917

-0-842

-0-733

-0-561

-0-339

-0116

0-048

0120

0-589

13

-0-923

- 0-853

-0-750

- 0-586

-0-367

-0143

0028

0110

0-580

14

-0-929

-0-863

- 0-765

- 0-607

-0-393

-0-167

0-008

0101

0-571

15

-0-933

-0-872

-0-779

-0-626

-0-416

-0190

-0010

0092

0-564

43

284 Distribution of numbers of mutants in bacterial populations
In the final column of Table 4 we have tabulated

1

Ji*&

JN. (58)

Having determined the maximal likelihood estimate of m as just described, the value of
(j m ^N/m is read off from the last column of Table 4. These values of a^N/m have been
used in plotting the part of Fig. 3D to the left of m = 10. Between m = 3 and m= 15, the
values of a m ^jN/m calculated from (58) and from (53) agree satisfactorily.

Summary
Statistical calculations are made of the distribution numbers of mutants in a culture of
bacteria in which the number of mutants increases on account both of new mutations and
of division of old mutants. In this way the largely qualitative conclusions of Luria and
Delbruck are extended and placed on a firm quantitative basis. The results of these
calculations, which enable the mutation rate to be inferred from experiments with parallel
cultures, are presented in the form of tables. Statistically efficient methods of using these
tables are discussed.

REFERENCES
Adolph, E. F. & Bayne- Jones, S. (1932). J. Cell. Comp. Physiol. 1, 409.
Demerec, M. (1945). Proc. Nat. Acad. Sci., Wash., 31, 16.

Fisher, R. A. (1938). Statistical Theory of Estimations. Calcutta University Readership lectures.
Fisher, R. A. & Yates, F. (1938). Statistical Tables for Biologists, Agriculturalists and Medical Research.

Edinburgh : Oliver and Boyd.
Lewis, I. M. (1934). J. Bad. 28, 619.
Luria, S. E. & Delbruck, M. (1943). Genetics, 28, 491.

Molina, E. C. D. (1942). Poisson's Exponential Binomial Limit Tables. New York: Van Nostrand.
Stewart, F. M. (1947). J. Hyg., Camb., 45, 28.
Witkin, E. M. (1946). Proc. Nat. Acad. Sci., Wash., 32, 59.

44

D. E. Lea and C. A. Coulson 285

APPENDIX. (ByC.A.C.)

(1) It has been suggested that a table of the individual coefficients C jr introduced in
equation (7), and which give the expansion of q r in powers of m, might be useful. Such
a table, for r^ 10, is shown below.

Table of C^

r mi*

According to equation (7), q r 2 C jr —
j=i ' J ■
r\j 12 3 4 5 6 789 10

1 I

2 i |

q _1_ i 'A

O 12 6 8

V A _!_ A i -A_

* 20 8 8 18

5 To e"o 4¥ n 3~2

ft J_ , «1 , 181 . _S_ -5- _i_

u 42 720 2180 12 96 84

7 -JL. 109 9 7 41 3 5 1 1

' 58 2520 1440 540 578 32 128

8 1 853 55 1 17 3 107 1 _7 1

72 25200 10080 2692 1728 24 384 256

Q -A- -AS- 13579 1 313 307 203 _7_ _1_ 1

y 90 700 302400 22880 5184 4320 256 96 512

(2) It should perhaps be pointed out that the replacement in (3) of n — 1 by n is an
approximation whose effect is quite negligible provided that r<^n, as occurs in all
experiments. In fact, even for r of the order of n-, the values of q r are seriously in error.
As a result of this, and of the fact that it allows r to exceed n (which is manifestly
impossible since r is the number of mutants and n is the total number of bacteria), the
generating function (15) actually gives an infinite value for all the moments. These two
difficulties have been removed in a development of this theory, to be published by Mr
D. G. Kendall, of Oxford. But unfortunately his more strictly correct generating function
cannot be expanded with any ease to determine the q r . Except for large r or small n,
however, it differs insignificantly from our (15).

(3) Mr Kendall has kindly pointed out to me that the argument in (31) and (32),
which was copied from Luria and Delbruck, is not quite valid. For in (32) the complete

series - + - + — + . . . is not convergent when x = 1 , and in order to get an expression for

Z O 4:

the mean value and the variance it was necessary artificially to curtail this series by
truncating it at its term x n /n. This device is not a valid procedure, and it appears that
although there is no change in the mean r , the variance a 2 of an individual determination
of r requires to be multiplied by 2, so that the correct relation a 2 = 2mn.

45

Origin of Bacterial Variants

Numerous bacterial variants are known which will
grow in environments unfavourable to the parent
strain, and to explain their occurrence two conflicting
hypotheses have been advanced. The first assumes
that the particular environment produces the observed
change in some of the bacteria exposed to it, whereas
the second assumes that the variants arise spontan-
eously during growth under normal conditions, tho
part played by the adverse environment being purely
selective. These are known respectively as the
'adaptation' and the 'spontaneous mutation' hypo-
theses.

In order to discriminate between them an experi-
mental approach (the 'fluctuation test') was developed
by Luria and Delbriick in 1943 l . This test has been
applied to a number of variants, some of them in
widely separated strains of bacteria (see Table ] ) .
and in each case the conclusion reached has been
that the variant arose by spontaneous mutation. The
validity of the fluctuation test has not been challenged,
at least so far as I am aware ; but on the other hand

Bacterial variations shown by means of the fluctuation
test to arise through spontaneous mutation

Organism

Variation to

Reference

E. coli

phage Tl resistance

1, 6, 5

phage TS-T7 resistance

6

radiation resistance

7

histidine independence

8

streptomycin resistance
and dependence

9, 10

Staph, aureus

penicillin resistance

11

Bulphathiazole resistance

12

streptomycin resistance

13

Clostridium septicum

uracil independence

14

Haemophilus influenzal

streptomycin resistance

15

EbertheUa typhosa

tryptophan independence

16

See also data from :

E. coli-mutabile

lactose fermentation

17

Salmonella

fermentation variants

18

it has gained only limited recognition (see ref. 2).
In part this may be due to the statistical and essen-
tially indirect nature of the argument on which it
is based, and if so the more direct experimental
evidence described below would seem to be of value.
Bacteria of Escherichia coli strain Bjr susceptible
to phage Tl were plated on agar and incubated until
a limited population increase had taken place. On
alternate plates the bacteria were redistributed over
the surface of the agar by spreading with 0-1 c.c.
of sterile saline. All were then sprayed with phage T\ ,
and counts made of the colonies of resistant survivors

[Reprinted by permission of MacMillan & Co. Ltd. from Nature 164 : 150, July 23, 1949]

4 6

which developed after further incubation. (For details
of the spray technique, see refs. 3, 4, 5.)

On the adaptation hypothesis, the bacteria present
at the end of the initial growth-period would all be
phage-susceptible, and spreading would serve only
to redistribute the members of a homogeneous popu-
lation. No striking differences in colony count be-
tween spread and unspread plates would therefore be
expected.

On the alternative (spontaneous mutation) hypo-
thesis, both susceptible and resistant cells would be
present at the end of the initial growth-period
wherever a sufficient end-population had been reached.
Further, mutations taking place a generation or more
before the cessation of growth would each be repre-
sented by a minute cluster of resistant cells all de-
scended from the one original mutant. Where the
arrangement of the bacteria is left undisturbed, a
cluster would give rise to a single resistant colony
after the application of phage ; but where the bacteria
are redistributed over the surface of the agar by
spreading, a colony would develop from each resistant
cell. Higher counts would thus be expected from
the spread than from the unspread plates. This is,
in fact, what was found (see Table 2), the difference
being as much as fifty-fold whore the end population
was highest.

Table 2. Resistant colonies developing after spraying agar plate
cultures of B. coli with phage 21, showing the effect of redistribution
by spreading prior to spraying

Incub. (hr.)

3

4

6

6

Bact. plated
End No. bact.*
Factor Increase

51 x 10*

1-7 x 10*

33

5 1 X 10*

2-3 x 10 7

480

5 1 x 10*

2-6 x 10»

5,100

6 1 x 10 4

2-8 x 10»

54,900

Resist, colonies
replicate test 1
2
3
4
6
6

unsp. sp.

1

1

unsp. sp.

3

6
2

1

2 2

unsp. sp.

5 194

3 14

4 16
8 13
2 4

6 112

unsp. sp.
46 2,254

25 1,434
45 3,294
49 3,710

26 1,538
49 399

Total

2

8 8

28 353

240 12,638

• Estimated by washing and assay ; averages of three independent
determinations, unap. — unspread; ap. — spread.

It might be suggested that the bacteria are less
likely to become 'adapted' when crowded together
in the developing colonies. If this were true, tho
proportion of bacteria becoming 'adapted' should bo
least where microcolony size is greatest. Table 2
shows that the average number of bacteria per micro -
colony at the time of spraying rose with increasing
period of incubation from 33 up to 54,900, but that
there was no corresponding decline in the proportion

of resistant colonies to bacteria sprayed. Thus,
crowding cannot account for the lower colony counts
from unspread plates.

This experiment, therefore, confirms the conclusion
drawn from the fluctuation test, namely, that phage-
resistant variants arise by spontaneous change prior
to contact with phage.

Howard B. Newcombe
Atomic Energy Project,
National Research Council (Canada),

Chalk River, Ontario. Jan. 7.
1 Luria, 3. E., and Delbrfick, M., Genetics, 28, 491 (1943).

I Hlnshelwood, C. N., "The Chemical Kinetics of the Bacterial Cell'

(Oxford University Press, London, 1946).

• Demerec, M., Proc. Nat. Acad. Sci., U.S., 32, 36 (1940).

• Beale, G. H., J. Gen. Microbiol., 2, 131 (1948).
5 Newcombe, H. B., Genetics, 33, 447 (1948).

• Demerec, M., and Fano, U., Genetics, 30, 119 (1945).
' Witkin, E. M., Genetics, 32, 221 (1947).

s Ryan, F. J., Cold Spring Harbor Symposia on Quantitative Biology,
11, 215 (1946) ; Proc. Nat. Acad. Sci., U.S., 34, 426 (1948).

• Scott, G. W. (in preparation).

10 Newcombe, H. B., and Hawirko, R. (in preparation).

II Demerec, M., Proc. Nat. Acad. Sci., U.S., 31, 16 (1945); Ann.

Missouri Botan. Garden, 32, 131 (1945).
" Oakberg, E. F., and Luria, S. E., Genetics, 32, 249 (1947).
" Demerec, M., J. Bad., 56, 63 (1948).

•« Ryan, F. J., et al., Proc. Nat. Acad. Sci., U.S., 32, 261 (1946).
18 Alexander, H. E., and Leidy, G. J. Exp. Med., 86, 607 (1947).
•• Curcho, M. de la G., J. Bad., 66, 374 (1948).
" Lewis, I. M., J. Bad., 28, 619 (1934).
18 Kristensen, M., Acta Path. Microbiol., Scand., 17, 193 (1940).

4 8

DELAYED PHENOTYPIC EXPRESSION OF SPONTANEOUS
MUTATIONS IN ESCHERICHIA COLI*

HOWARD B. NEWCOMBE 1

Carnegie Institution, Cold Spring Harbor, N. Y.

Received April 24, 1948
INTRODUCTION

THE quantitative study of mutations occurring at low rates requires an
organism that can be grown conveniently in large numbers, and mutations
that can be readily detected. Furthermore, where information is desired con-
cerning the effect of a gene change shortly after its occurrence, it is necessary
to be able to examine the phenotype of the organism immediately following the
change and at intervals thereafter.

Bacteria are superior to higher organisms for these purposes, since (1) large
populations can be handled, (2) there are numerous mutants that can be readily
detected and counted, and (3) individual organisms result from each cell
division, enabling the phenotype to be determined at any time after a gene
change. It was the purpose of this investigation to discover the rate of spon-
taneous mutation of the bacterium Escherichia coli, strain B/r, from sensi-
tivity to resistance to the phage 77, and also, by indirect means, to determine
the interval between time of occurrence of the mutation and phenotypic ex-
pression.

The reasons for investigating spontaneous mutation rate and time of pheno-
typic expression are as follows:

Two previously developed methods of estimating mutation rate in bacteria
have yielded discrepant results (Luria and Delbruck 1943). This discrepancy
has been ascribed to an error in the assumptions on which one of the methods
is based; but, which assumption and which method is in error is not known.

One of the methods rests on the assumption that a gene mutation expresses
itself immediately in the individual cell in which it occurs; and one of the
possible interpretations of the discrepancy is that this is not the case, but that
on the contrary one or more generations of growth are required before the
mutation is expressed.

Delayed phenotypic expression, or "cytoplasmic lag" as it has been termed,
has been observed in Paramecium (see Sonneborn 1947); and if something of
this nature occurs also in bacteria it is important from the standpoint of under-
standing gene action.

* The cost of the accompanying tables has been paid by the Galton and Mendel Memorial
Fund.

1 Present address: Biological and Medical Research Branch, National Research Council,
Atomic Energy Project, Chalk River, Ontario, Canada.

[Reprinted by permission from Genetics 33:447-476, September, 1948]

49

448 HOWARD B. NEWCOMBE

Methods of Previous Workers

The early experiments of Luria and Delbruck (1943) referred to above
will now be considered in detail.

To find out the number of phage-resistant mutants in a phage sensitive
liquid culture, the whole culture if its population size is small, or a sample of it
if it is large, is spread on agar together with the particular phage under con-
sideration. On incubation, all bacteria that are sensitive to the phage are
lysed, leaving only the resistant mutants. Each such mutant will eventually
form a separate colony; and from the number of colonies the number of mu-
tants in the liquid culture, or the sample, can be estimated.

By using this technique it is possible to determine the rate of mutation to
phage resistance. As stated, two methods have been devised by Luria and
Delbruck (1943). The experimental procedure in both of the methods is to
grow a series of similar liquid cultures from small inocula, and to determine
the numbers of resistant bacteria in each, as well as in the average population.
The two estimates are derived from these primary data, the first using the
proportion of cultures in which resistant mutants have appeared, and the sec-
ond using the average number of resistant mutants per culture.

These methods are of course only strictly applicable where it is possible to
eliminate the original type without affecting the mutant type, and where
there are no selective differentials between mutant and original types.

In their experiments these test cultures were started with small inocula (50
to 500 bacteria from a growing culture of E. coli, strain B) and were grown to
saturation either in broth or in synthetic medium. (The volume of the cultures
was 10 cc in some experiments and 0.2 cc in others, the final numbers of bac-
teria being of the order of 3X10 10 and 3X10 8 respectively.) At the end of
growth, samples of the cultures — or in some cases whole cultures — were tested
to determine the numbers of bacteria resistant to the phage 77.

The cultures from which the inocula were taken contained between 1 and
1,000 resistant individuals per 10 8 total bacteria. Thus, the chance of intro-
ducing a resistant bacterium into the test cultures via the inoculum was small
and in the rare event of one being introduced the fact would be indicated by an
excessive proportion of resistant bacteria in the fully grown test culture. In
practice, any resistant bacteria found at the time of testing — that is, after the
cultures are fully grown — -will therefore be the mutant offspring of one of the
sensitive bacteria in the inoculum.

An estimate of mutation rate per bacterium per division cycle can be ob-
tained if the numbers of mutations, and the average number of cell
divisions, occurring in a series of cultures are known. The latter may be cal-
culated from the final population in the series and the former from the propor-
tion of cultures containing no mutants. The greatest accuracy is obtained when
this proportion is neither too large nor too small, and the method cannot be
used if every culture contains a mutant. Since the proportion of cultures having
no mutants is a function of the number of cell devisions in a culture, it may
be adjusted by altering the volume of medium, 0.2 cc being the amount used

5°

DELAYED EXPRESSION OF MUTATIONS 449

in the experiments under consideration. To determine this proportion it is of
course necessary to test whole cultures, as distinct from samples.

This method will be known as method 1 throughout the present paper. It
should be noted that it is based upon an estimate of the number of resistant
clones developing in the series (this estimate being obtained from the propor-
tion of cultures in which no resistant mutants have developed), and that
it takes no account of the numbers of individuals in these clones at the end of
the growth. For the purpose of this paper, the term "mutant clone" will
refer to those individuals carrying genetic factors for phage resistance which
have a common origin in a single mutation. Within a mutant clone individuals
which are phenotypically resistant to phage will be collectively termed a
"resistant clone." A culture may contain one or more mutant clones of varying
age.

The second method of Luria and Delbruck uses the average number of
resistant bacteria in a series of similar cultures and calculates mutation rate
from this value, the average population, and the number of cultures. The
number of mutants arising during the growth of a culture is of course, on the
average, a function of the mutation rate. But there are very large variations in
the number of mutants present in different cultures grown under identical
conditions, these being due to chance variations in the time of occurrence of
the mutations. Thus the occasional occurrence of a mutation early in the
growth of a culture, at a time when the population is small, willjesult in a
much higher than average number of mutants in that culture. Because of
these statistical fluctuations, mutation rate cannot be calculated from the
number of mutants in a single culture started from a small inoculum. It can,
however, be estimated from the average number of mutants in a series of
similar cultures; and the mathematical details of the method have been
worked out by Luria and Delbruck (1943). This method will be known as
method 2 throughout the present paper. It differs from method 1, which utilizes
the number of resistant clones occurring in liquid cultures, in that it takes into
consideration the number of resistant individuals. Furthermore, in the event
of any change in the mutation rate during growth, method 2 would give an
average of the mutation rates obtaining in each of the generations during which
mutations had occurred — equal weight being given to the early generations
when the population and the number of mutations occurring were small, and
to the later generations when both these values had increased. In contrast to
this, in method 1 changes in the mutation rate during growth would give an
estimate strongly biased in favor of the rate obtaining during the later period
when the population and the absolute number of mutations occurring was
large. Moreover, a delay in the phenotypic appearance of a mutation would re-
duce the rate as estimated by method 1, because recent mutations would not
be detectable. The rate as estimated by method 2 would be affected less,
since early mutations, which have a greater number of generations in which to
become phenotypically resistant, are represented by larger numbers of de-
scendants than are the later mutations. The possibility that mutation rate is

450 HOWARD B. NEWCOMBE

not constant throughout growth, and the possibility that phenotypic resistance
does not appear for one or more generations after mutation to resistance
has taken place, will now be considered.

A striking and unexpected finding of the Luria and Delbruck experi-
ments was that the rates estimated by these two methods differed by a con-
siderable factor, that from method 1 (utilizing the number of resistant clones
developing in liquid cultures) being lower than that from method 2 (utilizing
the average number of resistant individuals per culture). The averages of their
estimates are .32X10 -8 and 2.4 X10~ 8 per bacterium per division cycle, re-
spectively. This difference has since been confirmed by Demerec and Fano
(1945), who have in addition shown that it is not peculiar to experiments
using Tl but is also true of rates of mutation to resistance to other phages
(T3, T4, T5, T6, and T7) when estimated by methods 1 and 2.

A statistical bias in method 2 (which gives the high estimate) may contrib-
ute to the discrepancy, but the work of Luria and Delbruck suggests
that its contribution is small. (For a discussion of this point the reader is re-
ferred to the original publication. Also, a method of estimating mutation rate,
which avoids this source of error will be considered in a later section of the
present paper.)

Since the two methods will give the same estimate if, and only if, (1) the
rate of mutation is constant throughout growth and (2) the occurrence of a
mutation gives rise to a phenotypic mutant without delay, it was concluded
that one of these two conditions did not obtain. There was no critical evidence
to indicate which one, however, since the discrepant estimates could be ex-
plained by assuming either a high mutation rate during the greater part of the
growth period, dropping during the last few divisions, or a delay of one or
more generations between mutation and phenotypic expression.

Luria and Delbruck did not favor the latter interpretation, since a fixed
delay of one or more generations before the development of phenotypic re-
sistance would mean that mutant clones would number two or more individ-
uals at the time when phenotypic resistance appeared. It was therefore as-
sumed that cultures with just one resistant individual would be rare if there
were a delay; and these, instead of being rare, had been observed in consider-
able numbers (see Luria 1946). It has been pointed out, on the other hand,
that if some lines of descent within the clone were to develop resistance earlier
than others, mutation plus delay could give rise to cultures having only one
phenotypically resistant individual (Sonneborn 1946). Thus the assumption
of a delay is permissible provided it is also assumed that phenotypic expression
is earlier in some lines of descent than in others within the same mutant clone.

Two alternative possibilities therefore exist: (1) that of a relatively high mu-
tation rate during all but the last few generations, and (2) that of a variable
delay in phenotypic expression. The possible significance of the second of these
two alternatives should be considered. If there is a delay in phenotypic expres-
sion, then some cultures which showed no phenotypically resistant bacteria
would contain mutants that could not be detected. Also, the end number of
resistant bacteria in a culture would represent only part of the genetic mutants

52

DELAYED EXPRESSION OF MUTATIONS 451

present, some not having developed resistance by the time growth stopped.
Thus the mutation rates calculated by methods 1 and 2 would both be under-
estimated, and the extent of the underestimate would depend upon the magni-
tude of the delay.

The information available so far sets no limit on the suspected delay or its
variability within clones of mutants, and it is even possible that an extreme
situation exists in which both are considerable. If this is true, mutation rate is
greatly underestimated by the methods outlined.

It should be noted at this point that a delay similar to that suspected in
the case of spontaneous mutations does in fact occur in irradiated material
(Demerec 1946; and Demerec and Latarjet 1946). Although there is at
present no certainty that spontaneous and induced mutations behave in pre-
cisely the same manner, it is of interest to consider the nature of the delay in
the one case in which it has been established, that is, in induced mutations.

To determine the time of appearance of induced mutations, Demerec ir-
radiated bacteria in liquid suspension, using ultraviolet radiation in some ex-
periments and X-rays in others. These treated bacteria were spread on agar,
incubated for varying periods of time to permit cell reproduction, sprayed
with phage Tl, and incubated again until colonies appeared. The spraying
caused all susceptible bacteria to be infected and lysed; but where a mutation
to resistance had occurred and had been expressed phenotypically, the result-
ing bacteria continued to grow after phaging, giving rise to one visible colony
for each such mutation.

Irradiation caused an enormous increase in the number of resistant clones
that appeared during growth, over the number that appeared during the same
number of generations in untreated bacteria. The delay between irradiation
and phenotypic change was such that less than one percent of the induced
mutants appeared prior to the first division, 50 percent appeared after about
five divisions, and some did not appear until after 11 or 12 divisions.

These observations give some support to the possibility that there is also a
delay in the expression of spontaneous mutations.

The experiments described here were designed to distinguish between this
and the alternative possibility of a change in the rate of mutation during
growth. They show that there is in fact a delay, and that it is the cause of the
discrepant estimates of rate obtained by using the Luria and Debruck
methods 1 and 2. Some indication of the extent and variability of the delay
has been sought, and an attempt has been made to obtain a more accurate es-
timate of mutation rate.

materials

The bacterium Escherichia coli strain B/r was used in these investigations.
This is a mutant derived from strain B (Witkin 1946, 1947), and is more re-
sistant than B to the action of radiations.

The mutations studied are those resulting in resistance to phage Tl. There
are at least two different categories of mutant: those resistant to Tl but
not to any other of the known phages, and those resistant to phages Tl and

53

452 HOWARD B. NEWCOMBE

T5. These mutant categories are designated B/r/1 and B/r/1,5 respectively.
Within each there occurs a number of morphologically distinguishable colony
forms, and it is possible that these represent a number of mutations of dis-
similar origin; but in this study no attempt has been made to distinguish
between the various types of mutation that give rise to resistance to Tl.

DISCREPANT ESTIMATES OF MUTATION RATE FROM NUMBERS OF RESISTANT
CLONES (METHOD 1) AND OF RESISTANT INDIVIDUALS (METHOD 2)

In view of the possibility that B/r may differ from B in the rate with which it
mutates, rates for B/r were determined by each of the Luria and Delbruck
methods, using phage Tl.

Eight separate experiments were carried out, and for each experiment 25
broth cultures of 0.2 cc were grown. Small inocula were used, and the cultures
were incubated for 18 hours, by which time growth had stopped.

The inocula contained approximately ten bacteria per culture in four of the
experiments, and approximately 10 4 in the other four. These numbers were
small enough so that the chance carry-over of a mutant in the inoculum would
be readily detected.

Method 1 was used to calculate mutation rate, a, from the proportion of
cultures having no resistant bacteria, Po, and the average number of bacteria
at the end of growth, N, using the formula:

a = - (In 2) (In P )/N. (1)

The above formula is derived from formulas (4) and (5) of Luria and Del-
bruck (1943), In being the natural logarithm.

Method 2 was used to calculate mutation rate, a, from the average number
of resistant bacteria per culture, r, the average number of bacteria at the end
of growth, N, and the number of cultures, C, using the formula:

r = (aN/ln 2) In (CaN/ln 2). (2)

This is derived from formula (8) of Luria and Delbruck.

The natural logarithm of 2 appears in these formulas because the mutation
rate refers to the rate per bacterium per division cycle, as distinct from the
rate per bacterial division. The significance of this distinction is best visualized
by using a concrete example. If a population of 10 8 bacteria passes through
one division cycle and one mutation takes place, the number of bacterial di-
visions is 10 8 and the mutation rate per bacterial division is 1 X 10~ 8 . The
mean population throughout the cycle, however, is 10 8 /ln 2, so that the rate
per bacterium per division cycle is In 2X10 -8 , which is .693 X10 -8 .

The first of these two methods of expressing mutation rate would be appli-
cable if mutation took place only at the time of cell division, and affected
just one of the offspring. The second would be applicable if mutability were
continuous throughout the division cycle. In the absence of information on
this point the choice is arbitrary, and since the latter method has been used
by previous authors its use is continued in this paper to facilitate comparisons.

54

DELAYED EXPRESSION OF MUTATIONS 453

These considerations are of course based on the assumption that each bacterium
divides, an assumption which will be discussed later in the paper.

Methods 1 and 2 have been used with strain B/r in order to determine (a)
whether, as with strain B, estimates obtained by method 1 are lower than
those obtained by method 2, and (b) whether the estimates from these two
methods are the same for B/r as for B.

The data from these experiments and the estimated mutation rates are
given in table 1. Those obtained using method 1 average .40X10 -8 and, those
using method 2 average 3.6 X10~ 8 .

It will be seen from table 9 — in which the results of previous workers, using
strain B, have been quoted — that the discrepancy between the estimates of
mutation rate given by the two methods as applied to B/r is similar to the
discrepancy using strain B. It is also evident that the estimate of mutation
rate for strains B/r and B are similar.

THE ELIMINATION OF A POSSIBLE UPWARD BIAS IN METHOD 2
BY THE USE OF LARGE INOCULA (METHOD 3)

The formula for calculating mutation rate from the average number of
mutants per culture (method 2) disregards the early divisions, when the popu-
lation is small and it is unlikely that a mutation will occur. The divisions
which enter into the calculation are those occurring after an arbitrary time, this
time being chosen so that on the average one mutation will occur prior to it
in the whole series of cultures. Luria and Delbruck point out that the chance
occurrence of this early mutation might account for part of the discrepancy
between the estimates of rate obtained with the two methods. For a detailed
discussion of this point the reader is referred to their paper.

It was therefore necessary to arrive at an estimate which, like that obtained
by method 2, would utilize the number of resistant bacteria arising during
growth in liquid culture, but which would not be biased by the chance occur-
rence of early mutations. This was done by growing the test cultures from in-
ocula of sufficient size to ensure that an appreciable number of mutations would
take place during the first division. Since much of the statistical fluctuation in
end numbers of resistant bacteria is thus eliminated, mutation rates may be
estimated from single cultures. (An experiment similar to this has been pro-
posed by Shapiro 1946.)

The method can be used only if the proportion of resistant bacteria in the
inoculum is small, since otherwise the relatively small increase due to mutation
during growth could not be accurately determined. To serve as inocula, there-
fore, cultures containing very small proportions of resistant bacteria were
chosen.

Five 50 cc and five 300 cc aerated cultures were grown from inocula of
2.6 X10 9 and 2.1 X10 8 bacteria, respectively. Synthetic medium, the M-9 of
Anderson (1946), was used because the bacteria can be grown to a higher
number per unit volume in it than in broth. In two separate sets of experi-
ments growth resulted in increases in the numbers of individuals of approxi-
mately a hundredfold and three thousandfold, respectively.

55

454

HOWARD B. NEWCOMBE
Table 1

Estimates of mutation rate o/B/r to resistance to phage Tl, using the methods of LURIA and DEL-
BRtJCK (1943) (methods 1 and 2 in the present paper) , calculating from the number of cultures
with no resistant bacteria, and from the average number of resistant bacteria per culture, respectively,
in series of similar cultures started from small inocula.

EXPERIMENT

A

B

C

D

E

F

G

H

Inoculum (no. of bact.)

10

10

10

10

10*

10 4

10 4

10*

Number of cultures

25

25

25

25

25

25

25

25

Vol. of cultures, cc

0.2

0.2

0.2

0.2

0.2

0.2

0.2

0.2

Culture Number

Number of Resistant Bacteria

1

140

2

2

1

1

1

134

4

2

41

57

5

9

21

18

1

3

45

1

1

2

34

34

4

43

11

6

4

287

380

6

5

60

8

1

4

242

46

3

149

6

13

1

1

71

2

33

15

176

7

48

30

1

5

12

23

2

31

8

3

1

2

12

3

6

1

9

4

1

14

1

2

102

4

10

36

1

2

84

2

48

11

55

11

3

49

11

148

5

12

447

139

9

29

18

45

13

1

1

40

17

120

3

14

9

7

142

42

10

33

2

15

27

1

1

1

7

4

725

134

16

14

4

154

1

15

1

25

17

30

6

3

60

14

5

34

18

160

15

4

37

45

13

14

19

231

9

1

1

110

131

20

35

158

7

32

376

118

22

21

37

1

5

44

1

1

60

22

21

1

18

32

133

15

23

1

8

4

158

24

8

27

2

12

36

42

151

6

25

3

221

79

4

9

263

Method 1

Cult, with no resist, b.

3

4

3

3

4

4

4

4

Bact. per cult., X 10 8

3.1

4.6

2.5

2.8

4.2

3.7

3.2

3.8

Mutation rate, X 10 -8

.42

.28

.59

.53

.31

.34

.40

.33

Method 2

Av. resist, b. per cult.

56.1

18.5

29.5

16.9

26.8

59.5

68.6

52.6

Mutation rate, X10~ 8

5.1

1.5

3.8

2.1

2.1

4.5

5.8

3.9

Assays were made of total bacteria and of numbers of resistant individuals
in the inocula and in the fully grown test cultures. These four values are desig-
nated Ni, ti, N2 and r 2 , respectively. For the sake of accuracy, five or ten inde-
pendent assays were made in each case. The number of generations of growth,
g, was determined for each of the cultures from the values N x and N2.

56

DELAYED EXPRESSION OF MUTATIONS 455

Where the mutation rate per bacterium per division cycle is a, and the rate
per bacterial division is a/(ln 2), the proportion of mutant bacteria in a cul-
ture will rise during growth by a fixed increment of a/2(ln2) per generation,
provided the inoculum is of sufficient size so that there are no appreciable
statistical fluctuations in the numbers of mutations occurring in the first divi-
sion. Thus mutation rate can be obtained from the formula

a = 2(ln 2)(r 2 /N 2 - r 1 /N 1 )/g.

(3)

Table 2

Mutation rates of B/r to resistance to phage Tl, calculated from the increase in the proportion of
resistant bacteria in cultures grown from large inocula {method 3 of this paper). Short grouth period.

TEST CULTURE A B C D E INOCULUM

10

7.0
1.3

Replicate assays

10

10

10

10

10

Inoculum, bact., X10 8

26

26

26

26

26

Vol. of culture, cc

50

50

50

50

50

Incr. in no. of bact.

114X

75X

112X

135X

46X

Resist, bact., X10~ 8

19.6

20.8

24.4

20.9

15.4

Standard deviation

3.7

4.7

2.6

3.5

2.7

Generations growth

6.8

6.2

6.8

7.1

5.5

Incr. resist, b., X10 -8

12.6

13.8

17.4

13.9

8.5

—

Mutation rate, X10" 8

2.6

3.1

3.5

2.7

2.1

—

Table 3

Mutation rates of B/r to resistance to phage Tl, calculated from the increase in the proportion of
resistant bacteria in cultures grown from large inocula {method 3 of this paper). Long growth period.

TEST CULTURE F G H I J INOCULUM

Replicate assays

5

5

5

5

5

10

Inoculum, bact., X10 8

2.1

2.1

2.1

2.1

2.1

—

Vol. of culture, cc

300

300

300

300

300

—

Incr. in no. of bact.

3600 X

2910X

2780X

2770X

3470 X

—

Resist, bact. X10~ 8

37.3

30.3

29 8

42.4

24.7

4.4

Standard deviation

6.9

3.9

3.5

6.2

5.6

1.1

Generations growth
Incr. resist. b„ X 10~ 8
Mutation rate, X10" 8

11.8

32.9

3.9

11.5

25.9

3.1

11.5

25.4

3.0

11.5

38.0

4.6

11.8

20.3

2.4

—

The data from two sets of experiments (involving 6 generations and 11 genera-
tions of growth) are given in tables 2 and 3, together with the mutation rates
obtained. These rates average 2.8X10~ 8 and 3.4X10~ 8 , respectively— values
which are not appreciably different from each other, or from those obtained
by method 2, using the average number of resistant bacteria per culture in
cultures started with small inocula.

Since there is no question of a statistical bias in calculating mutation rate
by method 3, the agreement may be interpreted as confirming the higher

57

456 HOWARD B. NEWCOMBE

value obtained by means of method 2. This is important, inasmuch as there has
been no estimate of the magnitude of the bias in method 2, or of the extent to
which the discrepancy between methods 1 and 2 is a product of it.

The above experiments demonstrate that this discrepancy is due to biologi-
cal rather than to statistical causes.

There are two reasons for measuring mutation rate over periods of 6 and
of 11 generations. First, had it been measured over the shorter period only, it
would be possible for the results to be biased by a high mutation rate during
the first generation or so. The obtaining of similar estimates of rate from both
a short and a long period, however, eliminates this as an appreciable source
of error.

Second, formula (3) assumes that the mutants multiply at the same rate
as the parent strain. That this assumption is approximately true is indicated
by certain experiments of of Luria and Delbruck and later of Demerec and
Fano. Somewhat more critical evidence however is obtainable from experi-
ments such as the above in which close agreement between estimates of muta-
tion rate from short and from long periods of growth indicate that there is no
appreciable bias arising from differential increase during logarithmic growth.
A mathematical demonstration of this will be found in a paper by Shapiro
(1946).

For present purposes it is sufficient to state the argument in general terms.
From formula (3) it is evident that mutation acting alone causes the propor-
tion of mutants r/N to rise arithmetically with each successive generation.
If, however, the mutants were to increase more (or less) rapidly than the
parent strain there would be superimposed upon this an exponential increase
(or decrease) in r/N, and estimates of rate obtained using formula (3) would
tend to rise (or fall) correspondingly with increasing periods of exponential
growth. Thus it is clear that close agreement between estimates of mutation
rate over periods of 6 and 11 generations constitutes evidence that the high
estimates obtained using methods 2 and 3 are not the result of a differential
favoring the mutant during periods of logarithmic growth.

It would be quite possible to supplement the above evidence on the relative
rates of increase of mutant and non-mutant strains, by preparing mixed
cultures and determining the change in proportion which takes place as the
result of competitive growth, as in the above mentioned experiments of Luria
and Delbruck, and Demerec and Fano. However the results of such experi-
ments are not entirely critical in the case of the spontaneously occurring mu-
tants, since the mutant strain must first be selected by growing in the presence
of phage, and then must be freed from all phage particles by suspending a
resistant colony in liquid, streaking the suspension on agar, and incubating
until visible colonies are formed, the process being repeated a number of times.
Such prolonged growth offers considerable opportunity for further mutation
and selection, and it is impossible to be certain that the strain which is finally
obtained will not have changed with regard to its ability to compete with the
non-mutant.

This difficulty may be avoided, however, by the use of radiation induced

58

DELAYED EXPRESSION OF MUTATIONS 457

mutants of a similar kind, since these occur much more frequently, and cultures
containing mutants and non-mutants in suitable proportions for competitive
growth experiments can be obtained without the necessity for first isolating
the mutant strain. Evidence has been obtained in this manner in connection
with a separate study (Newcombe and Scott) which will be published later.
Six independent mutant clones were tested over prolonged periods of growth
(18 to 20 generations) in competition with the corresponding non-mutant
strain. In some cases the proportion of mutants was found to remain unchanged
and in others to have declined slightly, the factors of change ranging from
1.20±40 down to 0.16+ .05. In no case did the proportion of mutants increase
significantly during growth.

Thus, provided one assumes that the radiation induced mutants are identi-
cal to those occurring spontaneously, evidence from this source is in agreement
with the above conclusion that the high estimates of mutation rate from meth-
ods 2 and 3 cannot be due to a differential favoring the mutants during the
period of rapid growth.

These experiments do not eliminate, however, the possibility that during
the approach to saturation, when the environment has altered considerably,
there may be conditions favoring the mutant strain. An increase in the propor-
tion of mutants during this period of approximately six times, would be suffi-
cient to produce the observed high estimates of mutation rate arrived at from
the numbers of individual resistant bacteria.

In order to test this possibility, an experiment was designed in which the
conditions associated with the approach to saturation could act repeatedly
on a bacterial population passing through approximately five to six generations,
thus accentuating any selection.

Five replicate test cultures, each containing in the first instance one cc of
broth, were incubated over a period of three days, during which time the
amount of liquid medium was doubled at regular intervals and finally brought
up to fifty cc. If the approach to saturation in the test cultures used in connec-
tion with methods 2 and 3 is accompanied by a six-fold increase in the propor-
tion of mutants, one would expect from the above treatment an increase of
much more than six times, in addition to any mutants which arose from spon-
taneous mutation. Also, estimates of mutation rate obtained from these cul-
tures using method 3, should be very much greater than the corresponding
estimates from normal test cultures.

Such estimates of mutation rate from the above experiment are presented
in table 4. It will be seen that these are not appreciably increased by keeping
the cultures under conditions approaching saturation throughout the whole
of the growth period.

In evaluating these data it will be noted that all assays are of numbers of
viable bacteria, and that if there is appreciable death due to the conditions of
growth, the number of cell generations will be underestimated. This problem
is considered in detail in a later section, where it will be shown that the effect
of undetected cell death is to increase the estimate of mutation rate obtained,
to a value above that of the true rate. Thus the experiment is weighted against

59

458 HOWARD B. NEW COM BE

the argument, and the results can be considered as critical evidence that there
are no appreciable differentials favoring the mutants during the latter part
of the growth of a culture.

The conclusions from the experiments discussed above are applicable to
differentials, both of division and of survival, and eliminate all possibility
that the widely different estimates of mutation rate obtained using method 1

Table 4

Mutation rates of B/r to resistance to phage Tl, calculated from the increase in the proportions of
resistant bacteria in cultures grown from large inocula {method 3 of this paper). Entire growth under
conditions approaching saturation.

TEST CULTURE

K

L

M

N

O

INOCULUM

Replicate assays
Inoculum, bact., X10 8

5
29.1

5
29.1

5
29.1

5
29.1

5
29.1

5

Initial vol. of cult., cc

1

1

1

1

1

—

Final vol. of cult., cc

50

50

50

50

50

—

Incr. in no. of bact.

33X

24 X

48 X

60 X

46X

—

Resist, bact., XlO" 8

15.9

15.5

17.2

18.7

21.6

5.0

Standard deviation

2.0

2.9

1.3

1.2

1.8

0.7

Generations growth

5.1

4.6

5.6

5.9

5.5

Incr. resist, b., XlO -8

10.9

10.5

12.2

13.7

16.6

—

Mutation rate, X 1(T 8

4.3

4.6

4.4

4.6

6.5

—

on the one hand, and methods 2 and 3 on the other, are due to such differen-
tials.

Method 3 thus provides critical confirmation of the estimates obtained by
method 2.

VARIATIONS IN MUTATION RATE WHICH WOULD BE REQUIRED TO EXPLAIN
THE DISCREPANT ESTIMATES FROM METHODS 1 AND 2

It thus seems certain that the high estimates of mutation rate from the
numbers of resistant individuals are not the product of an upward bias, and
we may now turn to the possibility that the low estimates from the numbers
of resistant clones are due to a downward bias. Two alternative possibilities
have been suggested which would account for these lower values: (1) a change
to a low rate of mutation during the later part of growth, and (2) a delay in the
phenotypic expression of a mutation.

The problem of distinguishing between these will be simplified if we con-
sider the first and determine the time at which the supposed transition
would have to occur, and whether this time is related to the number of genera-
tions from resting stage or to the approach to saturation.

To determine whether the supposed transition would be a function of the
number of generations from resting stage the data of table 1 may be used.
Half the cultures were inoculated with 10 and half with 10,000 bacteria, and

6o

DELAYED EXPRESSION OF MUTATIONS 459

these passed through 25 and 15 generations respectively (the relevant data
are given in table 5). The mutation rates calculated by methods 1 and 2

Table 5

Generations between resting stage and the approximate time of the first mutation in series of liquid
cidtures started from widely different inocula showing that mutation rate is independent of this vari-
able over the range of 6 to 18 generations. {Data from table 1.)

EXPERIMENT

A

B

C

D

E

F

G

H

Inoculum (no. of bact.)
End no. of bact. X 10 8
Generations growth

10
3.1
24.9

10
4.6
25.5

10
2.4
24.6

10
2.8
24.7

10 4
4.2
15.4

10 4
3.7
15.2

10*
3.2
15.0

10*
3.8
15.8

Max. resist, b. in series
Gen. after first mutation

447
8.8

221
7.8

158

7.3

79
6.4

242
8.0

376
8.6

725
9.5

263
8.1

Gen. to first mutation

Mut. rate (method 1), XlO -8

Mut. rate (method 2), XlO -8

16.1

.42
5.1

17.7
.28
1.5

17.3
.59
3.8

18.3
.53
2.1

7.4
.31
2.1

6.6
.34
4.5

5.5

.40
5.8

7.7
.33
3.9

showed no effect due to the difference in number of generations from resting
stage. Thus the time of the supposed transitions would have to be a function
of the approach to saturation, and not of the number of generations from rest-
ing stage. This means that mutation rate would have to be of the order of
3 XlO -8 during the whole of the logarithmic growth phase, dropping to some-
thing like .4 X10 -8 during the last few divisions.

It will be shown later that there are apparent variations in mutation rate
during early growth, but that they are associated with the first few divisions
after resting stage and occur too early to have any bearing on the immediate
problem.

It is thus evident that the possible effect of a low mutation rate during the
later part of growth can be eliminated by confining one's tests to the period of
rapid growth; and critical evidence for or against a possible delay can be ob-
tained from estimates of mutation rate based on the number of resistant clones
appearing during rapid growth.

EVIDENCE FOR A DELAY BETWEEN MUTATION AND PHENOTYPIC
EXPRESSION (METHOD 4)

It now remains to determine the rate of appearance of resistant clones dur-
ing rapid growth, when no approach to saturation is involved.

If the rate of appearance of resistant clones is high (3 X10 -8 ), the discrep-
ancy between methods 1 and 2 can be interpreted without assuming a delay
between mutation and phenotypic expression. If, on the other hand, rate of
appearance of resistant clones is low (.4X10~ 8 ), then from the previous evi-
dence the number of individuals in an average resistant clone must be greater
than expected on the basis of the number of generations passed through after

6i

460 HOWARD B. NEWCOMBE

its first appearance. This would be interpreted as indicating that the mutant
clone had its origin one or more divisions prior to its becoming phenotypically
detectable.

One obvious alternative to this interpretation should be mentioned, namely
that the excess numbers of mutants are due to more r^pid division in these
than in the parent strain. This is rendered unlikely however by the evidence
of Demerec and Fano (1945) that these mutants do not divide more rapidly,
and in addition more critical evidence against the possibility has been obtained
from the experiments described under method 3.

The experimental procedure was essentially that used by Demerec (1946)
in his work on mutation rates in E. coli following irradiation. Bacteria are grown
on agar for varying periods of time, sprayed with phage, and incubated until
colonies appear. Mutations occurring during growth, and gaining phenotypic
expression, will give rise to resistant clones. Since individual bacteria cannot
move about on the agar, the members of a clone are confined to a particular
locality. These resistant clones survive the application of phage, and eventually
form colonies of visible size. Thus each mutation which gives rise to phage
resistance is in the end represented by one colony.

An estimate of mutation rate is obtained by dividing the number of resistant
clones appearing in a given period by the number of bacterial divisions times
1/ln 2. Thus, if Ri and R 2 are the numbers of resistant clones present at times
1 and 2, the number of resistant clones arising during the interval between
times 1 and 2 is R 2 — Ri. Similarly, if Ni and N 2 are the numbers of bacteria
present at times 1 and 2 respectively, the increase during the interval will be
N 2 — Ni. Since each division of a bacterium increases the total number by
one, this value is equal to the number of bacterial divisions during the period.
Mutation rate per bacterium per division cycle, a, will therefore be obtained
from the formula:

a = (In 2)(R 2 - R,)/(N, - N x ). (4)

Since the values of R 1? R 2 , Ni and N 2 represent viable cells only, an assump-
tion is involved, namely that all bacteria divide. It will be shown later that this
assumption is approximately correct for the early stages of logarithmic growth,
and it is assumed that no appreciable increase in the proportion of cells which
fail to divide, takes place until the phase of declining growth rate is approached.
In these experiments precautions were taken to ensure that growth is limited
to the period of exponential increase.

It will also be shown that the effect of the presence of cells which do not di-
vide further, will be to increase the estimated rate of mutation. The present
experiments can therefore be considered critical if the estimates of mutation
rate obtained using method 4 are found to be low relative to those obtained
using methods 2 and 3.

Estimates of the values of Ri, R 2 , Ni, and N 2 are obtained as follows. Four
plates are inoculated with a suitable number of bacteria, two being incubated
until time 1 and two until time 2. One plate from each incubation period is
sprayed with phage and then incubated further. The numbers of colonies

62

DELAYED EXPRESSION OF MUTATIONS 461

developing on these plates represent the numbers of resistant clones present
at the time of spraying, that is, Ri and R 2 , respectively. The remaining plate
from each incubation period is washed with ten cc of normal saline and the
numbers of bacteria present (Ni and N 2 , respectively) determined by colony
counts.

Where time 1 is the time of plating the bacteria, and no divisions can have
taken place, Ni is determined in a more direct manner. Instead of plating and
then washing off the bacteria plated, an equivalent quantity of the culture
from which the inoculum was taken is diluted and colony counts made.

An estimate of mutation rate is thus obtained from four plates. In all ex-
periments these four plates were replicated several times, and a corresponding
number of independent estimates of rate obtained. These independent esti-
mates have been averaged, and the standard deviations calculated.

In all experiments in which growth was determined by the use of duplicate
plates, care was taken to ensure the same amount of growth on both plates.
All plates were warmed in the incubator before plating the bacteria, and when
removed for plating were kept warm on a thermostatically controlled warm
table until returned to the incubator. All platings and removals from the
incubator followed an accurately timed schedule. The temperature in the incu-
bator was kept as uniform as possible by circulating the air rapidly with fans.
At the end of incubation, growth was stopped abruptly by chilling plates in
contact with the cold metal of a refrigerator freezing unit. This chilling did
not affect the survival of the bacteria or the phenotypic expression of the
mutants.

Additional precautions were required with respect to (1) the choice of cul-
tures from which to inoculate the plates, (2) the number of bacteria plated,
and (3) the amount of phage applied by spraying.

(1) When relatively large numbers (of the order of 10 8 ) of bacteria are
plated, there will be a certain number of resistant cells in the inoculum. These
resistant cells have occurred by mutation during the growth of the culture
from which the inoculum is taken. Cultures vary widely in the number of
mutants present at the end of growth, and in these experiments the number
present in the inoculum (Ri) was determined by spraying with phage immedi-
ately after the bacteria had been plated. Where the number is excessive it is
apt to obscure the increase in resistant clones resulting from growth, or to ren-
der the determination of the increase less accurate. For this reason, cultures
having an excessive number of resistant cells were not used asinocula. The cul-
tures that were used contained from 5 to 50 resistant bacteria per 10 8 sensitive.

There is no evidence that this selection biased the results, since mutation
rates obtained using these inocula were the same regardless of the number of
resistant bacteria present.

(2) The size of the inoculum was adjusted so that the end number of bac-
teria on the plate would be approximately 2X10 9 . With end numbers of less
than this the number of resistant colonies was reduced, and at the same time
the accuracy of the method. With excessively large end numbers of bacteria
there is a reduction in the apparent mutation rate. The precise interpretation

63

462 HOWARD B. NEWCOMBE

of the phenomenon is uncertain. In the absence of evidence to the contrary
it has been assumed that mutations do occur at the normal rate, but that they
fail to develop visible colonies owing to the presence of large numbers of sensi-
tive bacteria that are not lysed by the phage — a situation which occurs if phage
is applied after bacterial growth has passed the logarithmic phase.

Tests showed that this apparent reduction in mutation rate occurred only
if the end number of bacteria exceeded 5 X 10 9 .

(3) It is known that bacteria which are infected during rapid growth have a
latent period of 13 minutes — at the end of which time they burst, liberating
on the average 180 phage particles (Delbruck and Luria 1942). If some of
the bacteria fail to be infected at the time of spraying, it is unlikely that they
will escape infection once lysis of the others starts. Any uninfected individuals
would, on the average, pass through somewhat less than one division during the
13-minute latent period. Thus, with a large proportion of the bacteria unin-
fected at the time of spraying, a somewhat less than twofold increase in popu-
lation would be expected before all the bacteria became infected, and the ap-
parent mutation rate from such an experiment would be increased propor-
tionally. Where all but a small proportion of the bacteria are infected at the
time of spraying, the apparent mutation rate would not be appreciably greater
than in the case of 100 percent infection.

In the present experiments the number of phage particles applied was equal
to the number of bacteria, or slightly in excess. Large excesses were not used
since these involve long periods of spraying, with resultant wetting of the sur-
face of the agar and a tendency for the bacteria to be moved about by the mois-
ture. Larger numbers of phage particles have been used, however, by Beale
(1948), who concentrated the phage by centrifuging. The estimates of muta-
tion rate that he obtained in this manner do not differ appreciably from those
obtained in the present experiments, and it may be assumed that the quantities
used in the latter were adequate.

By the method described in this section it was possible to determine the
rate with which resistant clones appear during the period of rapid growth
from the resting stage onward.

The results of four separate experiments, each with eight independent repli-
cates, are given in table 6. In these experiments the bacteria were grown
over a period of approximately eight generations; and the mutation rate
obtained is that for the whole growth period.

It will be seen that the average of estimates of rate from all experiments is
low (.59 X 10~ 8 ) and is in close agreement with the low estimates from method 1
(average .40X10 -8 ). As pointed out earlier, this constitutes evidence that re-
sistant clones arise from mutations occurring one or more generations prior to
their first becoming detectable.

It should be mentioned at this point that subsequent experiments, described
in the next section and summarized in figure 1, have enabled mutation rate to
be calculated from a tenfold increase in bacterial titer onward. This eliminates
the contribution of the first few divisions, during which an excessive number of

64

DELAYED EXPRESSION OF MUTATIONS
Table 6

463

Mutation rateof B/r to resistance to phage Tl, estimated from the number of resistant clones appearing
during bacterial multiplication on agar (method 4 of this paper).

EXPERIMENT

A

B

C

D

Inoculum (no. of bact.), X10 8

.115

.119

.106

.055

Resist, b. in inoc, X 10 -8

30.2

30.2

10.0

38.0

Incr. in no. of bact.

264X

250X

345 X

317X

Generations of growth

8.1

8.0

8.5

8.3

Replicates

Mutation

Rate, X10~ 8 *

1

.49

.54

.48

.53

2

.47

.61

.69

.38

3

.56

.70

.57

.49

4

.42

.49

.59

.71

5

.70

.54

.89

.50

6

.69

.46

.44

.52

7

.45

.51

.69

.63

8

.53

.66

1.04

.89

Av. mutation rate, X 10 -8

.54

.57

.67

.58

Standard deviation

.098

.078

.205

.156

Mut. rate from 10 X increase

.45

.48

.60

.52

onward, X10~ 8 t

* Each replicate mutation rate is calculated from an independent single-plate estimate of each
of the following four values: (1) number of bacteria in the inoculum, (2) end number of bacteria,
(3) number of resistant bacteria in the inoculum, and (4) end number of resistant clones, using
formula 4 of this paper.

t This mutation rate is calculated from a tenfold increase onward using the information in
figure 1 and the method outlined in the section on method 4. It is a more accurate estimate of the
rate of formation of resistant clones during logarithmic growth, since the bias from the high early
rate of appearance of resistant clones is removed.

resistant clones appears, and reduces the average rate obtained from these
experiments from .59X10" 8 to .51X10- 8 .

RATE OF APPEARANCE OF RESISTANT MICROCOLONIES DURING THE EARLY
DIVISIONS ON A SOLID MEDIUM

Where there is a delay between mutation and phenotypic expression such
that the first phenotypically resistant individual appears n generations after
the mutation has occurred, there will be 2 n bacteria in the mutant clone when
it first becomes detectable. When a mutation occurs, a delay of n generations
in the time of its appearance will correspond to a 2 n -fold increase in the popu-
lation. The apparent number of bacterial divisions in which the mutation has
occurred will then be 2 n times the true number, and when an estimate is made
from the number of detectable mutant clones, the apparent rate will be reduced
to

6s

464

HOWARD B. NEWCOMBE

From the interpretation here adopted, it follows that a sample from a culture
will contain some mutant individuals which are not yet phenotypically re-
sistant. When such a sample is spread on agar and allowed to grow (as in
method 4) these hidden mutants develop phenotypic resistance during the
first few divisions and thus give rise to resistant microcolonies. Since a mutant
clone may contain a number of these hidden mutants, and since these are dis-
persed over the surface of the agar, a number of resistant microcolonies can

GROWTH

Fig. 1.— Numbers of mutations to phage (Tl) resistance arising during bacterial multiplication, as

estimated from the numbers of resistant microcolonies present after varying periods of growth.

result from a single mutation prior to plating. This is of course not true for
mutations occurring after plating, because the products of these are confined
to one locality.

Thus the rate of appearance of resistant microcolonies during the early
divisions would be expected to be high, approaching the true mutation rate;
and the rate of appearance during later divisions would be expected to be low,
approaching 2 n times the true mutation rate.

Evidence has been obtained on this point, using method 4 and plotting
the numbers of resistant clones arising during varying periods of incubation
(R 2 — Ri, where time 1 is the time of plating) against growth in terms of factor
increase in the number of bacteria (N 2 /Ni). The absolute increase in resistant
clones with growth depends, of course, upon the original number of bacteria
plated, and where data are obtained from a number of separate experiments
they are comparable only if expressed in terms of the number of individuals in

66

DELAYED EXPRESSION OF MUTATIONS 465

the inoculum. It is therefore convenient, when plotting a curve using data
from a number of experiments, to express the increase in terms of resistant
clones per 10 8 bacteria in the inoculum. This has been done in figure 1, where
(R 2 — Ri)/NiXl0 8 is plotted against growth (N 2 /Ni), time 1 being the time of
plating. Each point shown in figure 1 was obtained by averaging at least eight
independent estimates of increase in number of resistant clones and of growth.

Two experimental procedures for estimating growth were used in this case.
One was that mentioned earlier, in which bacteria are washed from the plate
and the number determined by dilution and assay. The other was by direct
count of the numbers of bacteria in the developing microcolonies.

For the latter purpose small numbers of bacteria were plated on agar, and
the plates were incubated at the same time and for the same period as those
that were to be sprayed with phage. These growth assay plates were then
chilled to stop division and examined under a high-power dry objective. The
numbers of bacteria in 50 microcolonies were counted and averaged. The
method gave accurate results up to an increase in number of bacteria of ap-
proximately 64 times.

In figure 1, points obtained by calculating growth by means of assay of
bacteria washed from duplicate plates are shown as hollow dots. Those for
which growth was calculated from the average number of bacteria per micro-
colony are shown as solid dots.

Something similar to the expected high early rate of appearance of resistant
clones is observed in these results. In figure 1 the numbers of resistant clones
appearing during growth are plotted on a logarithmic grid against growth ex-
pressed in terms of factor increase in numbers of bacteria. For comparison, a
curve is drawn showing the expected numbers — assuming a rate of .55X10 -8
and immediate phenotypic expression. The experimental curve does show a
high early rate, declining as growth proceeds. This is qualitatively what would
be expected where there is a delay, though the rate is actually higher than ex-
pected.

The increase in resistant clones when material is sprayed with phage after
a twofold increase in population is approximately 12 per 10 8 bacteria plated,
which represents an apparent mutation rate of 8.3 X10 -8 ; whereas the rate
expected during the first division is the same as that obtained from liquid cul-
tures, that is, something between 2.8 and 3.4X10 -8 .

Furthermore, by extrapolating the curve backward it would appear that
about seven per 10 8 bacteria become resistant during the lag phase before any
division has taken place.

This point has been studied by incubating bacteria with a known lag phase
of 70 minutes for 30, 60, and 90 minutes on agar and then spraying with phage.
The results are given in table 7, and show an appreciable increase in resistant
colonies from plates sprayed just before the onset of the first division and dur-
ing the very early part of the division.

The points just considered are not directly related to the main issue. The
important contribution of these experiments is to show that the rate of appear-
ance of resistant clones declines during the first few divisions on agar.

67

10

10

10

11.0

11.0

11.0

10.1

10.1

10.1

30

60

90

1.0X

l.OX

1.3X

0.1

5.8

10.7

1.89

1.87

2.05

466 HOWARD B. NEWCOMBE

A METHOD OF REDUCING THE DOWNWARD BIAS IN METHOD 2 DUE TO THE
DELAY IN PHENOTYPIC EXPRESSION (METHOD 5)

As mentioned in the introduction, a delay between mutation and phenotypic
expression must be variable within a mutant clone, expression occurring in
some lines of descent earlier than in others. This leaves open the questions of
the extent of the delay, the nature of the variation, and the rate of gene muta-

Table 7

Increase in number of bacteria resistant to phage Tl during early growth, using strain B/r grown
on agar, and spraying with phage during the lag phase and early part of the first division.

EXPERIMENT ABC

Replicate plates
Bact. in inoculum, X10 8
Resist, bact. in inoc. r per 10*
Incubation, minutes

Incr. in no. of bact.
New resist, bact., per 10*
Standard deviation

tion, the last being to a greater or smaller degree underestimated by the meth-
ods dealt with so far. Method 5 is designed to obtain a less biased estimate of
mutation rate, and, from this, some idea of the extent of the delay.

With a variable delay it would be expected that mutant clones arising from
mutations early in the growth of a culture would contain a higher proportion
of phenotypically resistant cells than would younger mutant clones, and that a
higher and less biased estimate of mutation rate would be obtained if it could
be calculated from these older clones alone.

To do this the following method has been devised. In a series of similar test
cultures a few of the cultures contain many times the average number of re-
sistant cells, because of the chance occurrence in these cultures of an early
mutation. In such cases the precise numbers of resistant bacteria from the
earliest mutation and from subsequent mutations cannot be determined di-
rectly. The probable number from later mutations, however, is approximately
equal to the mean number in the whole series. This is, incidentally, a very
slight overestimate, and a more precise approach will be considered later.

Using this method, the probable number of resistant bacteria descended from
the earliest mutation in a series of cultures is obtained by subtracting the
mean number of resistant bacteria per culture, r, from the highest number
occurring in any one of the cultures, h. At the probable time of occurrence of
this mutation the population in the culture would have been N/(h — r), where
N is the end population; and the population in the whole series of cultures
would have been CN/(h — r), where C is the number of cultures in the series.
Thus the first mutation in the series occurred when the population was
CN/(h — r), and since the inocula were small the number of bacterial divisions

68

DELAYED EXPRESSION OF MUTATIONS 467

giving rise to this population would also be'CN/(h — r). The mutation rate, a,
when one mutation occurs in this number of bacterial divisions is of course
given by the formula:

a=(ln2)(h-r)/CN. (5)

Mutation rate is very slightly underestimated by this formula, since r is an
overcorrection for the probable number of resistant cells from mutations sub-
sequent to the first. A more precise correction would be obtained by averaging
the number of resistant cells per culture, exclusive of those from the first mu-
tation in the series. This complication has not been introduced, however, since
the gain in accuracy would not be appreciable.

The rates calculated in this manner may be considered as approaching the
true rate if most of the members of these older clones have become pheno-
typically resistant. If only a small proportion have become resistant, this rate
is still an underestimate, although it would be closer to the true rate of gene
mutation than the estimates using method 2. This method will be known as
method 5.

The rates obtained in this manner, using the data from table 1, are presented
in table 8, together with the values required for the calculations. Data of

Table 8

Mutation rate of B/r to resistance to phage Tl, calculated from the maximum number of resistan
bacteria in any one culture of a series, using method 5 of this paper, and the data in table 1.

experiment A B C D E F G H

Number of cultures
Bact. per cult., X 10*
Av. resist, b. per cult.

Max. no. resist, bact.
Mutation rate, XlO -8

Luria and Delbruck, and Demerec and Fano, obtained using phage Tl,
have been treated similarly, and a detailed comparison is made in table 9 of
rates from methods 1, 2, and 5, using information from all sources. For the
sake of convenience, the rates obtained by methods 2 and 5 have been ex-
pressed in the last two columns of the table in terms of the smaller value from
method 1, as ratios— rate (2)/rate (1), and rate (5)/rate (1).

Taking into consideration all available information on mutation to resistance
to phage Tl, rates from methods 2 and 5 differ only slightly and are between
four, and nine times the rates from method 1.

The application of method 5 has also been extended to the data obtained
by Demerec and Fano using phages other than Tl; and averages of mutation
rates from methods 1, 2, and 5, and of the ratios rate (2)/rate (1), and rate
(5)/rate (1), have been worked out for all available data on strains B and B/r
and phages Tl, T3, T4, T5, T6, and T7. They are given in table 9.

25

25

25

25

25

25

25

25

3.1

4.6

2.5

2.8

4.2

3.7

3.2

3.8

56.1

18.5

29.5

16.9

26.8

59.5

68.6

52.6

447

221

158

79

242

287

725

263

3.6

1.3

1.3

.83

1.5

2.3

5.7

1.5

6o

468

HOWARD B. NEWCOMBE

The purpose of these calculations was to determine whether mutation rate
when estimated from the older mutant clones only is higher than when esti-
mated from the average numbers of resistant bacteria per culture. If the delay
is sufficiently variable and of sufficient magnitude, it would be expected that

Table 9

Mutation rates of strains B and B/r to resistance to phage Tljrom series of similar liquid cultures
started with small inocula, using methods 1, 2, and 5 and calculating from data of Luria and Del-
bruck (1943, table 3, experiment 23), Demerec and Fano (1945, table 4), and the present paper
(tables 1 and <?). All the experiments which are suitable for the application of the three methods have
been used.

EXPER.

STRAIN

CULTURES

RATE (1)

X10- 8

RATE (2)

xio- 8

RATE (5)

XIO" 8

RATIO

(2)/(D

RATIO

(5)/(D

Luria and Delbruck

A

B

87

.32

2.4

1.5

7.5

4.4

Demerec and Fano

A

B

53

.63

1.4

1.8

2.2

2.9

B

53

.33

2.3

6.1

7.0

18.5

C

17

1.15

3.9

9.2

3.4

7.9

D

25

.71

3.1

4.1

4.4

5.8

E

19

.66

3.2

3.0

4.8

4.5

F

38

1.26

6.1

7.8

4.7

6.2

G

45

.24

0.8

1.3

3.3

5.4

H

44

.63

3.2

6.0

5.7

9.6

I

45

.31

2.0

2.9

6.4

9.4

J

45
Average

.92

1.3

1.3

1.4

1.4

.68

2.7

4.4

4.3

7.2

Present investigation

A

B/r

25

.42

5.1

3.6

12.2

8.6

B

25

.28

1.5

. 1.3

5.4

4.4

C

25

.59

3.8

1.3

6.6

2.1

D

25

.53

2.1

8.3

4.0

1.6

E

25

.31

2.1

1.5

6.8

4.7

F

25

.34

4.5

2.3

13.2

6.7

G

25

.40

5.8

5.7

14.5

13.8

H

25
Average

.33

3.9

1.5

11.8

4.6

.40

3.6

2.3

9.3

5.8

in these older clones a higher proportion of the mutants would have become
phenotypically resistant.

Not all experiments show a difference between estimates of rate from the
average resistant bacteria (method 2) and estimates from the highest number
of resistant bacteria, of which the majority are presumably descended from
an early mutation (method 5). Examining data from all sources, the estimates
from method 5 and from method 2 do not differ significantly.

70

DELAYED EXPRESSION OF MUTATIONS 469

With the exception of the mutations to B/3, 4 and B/3, 4, 7, estimates from
method 5 are of the order of six to seven times those from method 1, and the
lower ratios in the case of these exceptions are in all probability due to the
high proportion of slow-growing B/3, 4, and B/3, 4, 7 mutants — Demerec
and Fano having shown in their experiments on competitive growth that these
mutants grow much more slowly than the parent strain B.

A crude average of the ratio rate (5)/rate (1) from mutations other than
those to B/3, 4, and B/3, 4, 7 shows that estimates obtained by method 5

Table 10

Mean mutation rates of strains B and B/r to phage resistance, estimated by methods 1, 2, and 5;
data (7/Luria and Delbruck, Demerec and Fano, and the present investigation {table 9), all from
series of similar liquid cultures started with small inocula.

PHAGE

STRAIN

RESISTANT
BACTERIA

NO. OF
EXPER.
AVER-
AGED

AVER.
RATE

(1)

xio-»

AVER.
RATE

(2)

xio-«

AVER.
RATE

(5)
XIO" 8

AVER.
RATIO

(2)/(D

AVER.
RATIO

(5)/(D

SOURCE

Tl

B/r

B/r/ 1,5; B/r/ 1

8

.40

3.6

2.2

9.3

5.8

table 9

Tl

B

B/l, 5; B/l

1

.32

2.4

1.5

7.5

4.4

L. &D.

Tl

B

B/1,5;B/1

10

.68

2.7

4.4

4.3

7.1

D. &F.

T5

B

B/l, 5

5*

.38

1.5

2.6

4.3

7.5

D. &F.

T6

B

B/6

3

5.0

30.8

42.0

5.3

7.1

D. &F.

T3

B

B/3, 4, 7, -B/3, 4

7

5.1

10.0

12.5

2.3

2.9

D. &F.

T4

B

B/3, 4, 7; B/3, 4

3

5.5

12.9

34.8

2.7

5.8

D. &F.

T7

B

B/3, 4, 7

3

4.4

12.2

9.6

2.8

2.4

D. &F.

* Note: One experiment containing an exceptional culture with 20,000 resistant bacteria has
been omitted from these averages.

are approximately 6.4 times those obtained by method 1. The significance of
this ratio in relation to the extent and variability of the delay between muta-
tion and phenotypic expression will be considered in the discussion.

THE INFLUENCE OF DIVISION MORTALITY UPON ESTIMATES
OF MUTATION RATE

All the methods of estimating mutation rate are subject to a small error
because some of the bacteria resulting from a division fail to divide again
(Wilson 1922). The number of mutations occurring during a division is thus
underestimated, and the number of divisions, where calculated from numbers
of viable bacteria, is also underestimated.

It might be assumed that the apparent variations in the mutation rate
during the early part of the growth cycle are due to corresponding changes in
the mortality associated with division. That this is not the case will be demon-
strated; and corrections to be applied to mutation rates which have been based
upon the assumption of unity survival will be considered.

Where growth has been estimated from the numbers of viable bacteria, a
correction may be obtained in the following manner. Let S be the chance of

71

470 HOWARD B. NEWCOMBE

survival of one of the bacteria resulting from a division. In a population of N
viable bacteria, one division will raise the number from N to 2NS viable in-
dividuals, and, where A is the true rate of mutation, will result in the produc-
tion of ANS/ln 2 viable mutant individuals. If the increase in population is
determined from assays of viable bacteria, the apparent mutation rate, a,
during this division will be (In 2) (ANS)/(ln 2) (2NS-N), which reduces to
AS/(2S— 1). The true mutation rate may thus be obtained from the apparent
rate by the formula:

A = a(2S - 1)/S. (6)

The relationship between the true mutation rate and an apparent rate, ai,
obtained by utilizing the increase in total population as distinct from viable
population, may be derived in a similar manner and is expressed by the for-
mula:

A = ai /S. (7)

Where estimates of mutation rate have been made in both ways, that is,
using both viable population increase and total population increase, the value
of S may be determined from the apparent mutation rates a and ai. The for-
mula for this as obtained from formulas (6) and (7) by simultaneous solution
and is:

S = (a + ai)/2a. (8)

The true mutation rate also may be obtained from the values a and ai,
using equation (8) and substituting for S. This yields the formula:

A = 2a-ai(a + ai). (9)

These formulae may be applied to the data from figure 1 in order to deter-
mine whether the high apparent mutation rate during the first few divisions
could be due to a low survival. It will be noted that points on figure 1 obtained
by estimating growth from viable numbers of bacteria do not differ greatly
from those obtained by estimating growth from total bacteria. This suggests
that mortality due to division is not excessive, and we may calculate the chance
of survival of the individual products of one division, by using formula (8).
Thus, over the period from a twofold to a tenfold increase in population, the
rates estimated from the two curves would be approximately 2.6 X10 -8 and
1.4X10~ 8 , respectively. From these two values the calculated survival is 0.75.
Therefore the high rate of appearance of resistant clones during the early
divisions on a solid medium discussed earlier cannot be due to an excessive di-
vision mortality.

It should also be noted that the estimate of rate which would have been
obtained over this period of growth by either method, had survival been unity,
is given by formula (9) and is 1.9X10 -8 . This latter figure, however, has little
bearing on the present discussion, as it is in any case intermediate between the
estimates of rate obtained from numbers of resistant clones and from numbers
of resistant individuals.

72

DELAYED EXPRESSION OF MUTATIONS 471

It is realized that this early low division mortality may not persist through-
out the whole of the growth cycle but may rise as the population in a culture
approaches the saturation density. In order to minimize the possibility of this
effect, the end population of cultures grown on agar (method 4) was not al-
lowed to reach that period at which growth rate begins to decline (approxi-
mately 5X10 9 bacteria on a 100 millimeter plate of broth agar). If, despite
these precautions, there was an increase in division mortality, then as pointed
out earlier, the estimates of mutation rate obtained would be somewhat greater
than the true rate. It will be noted that this would tend to weight the data ob-
tained by method 4 against the interpretation arrived at. We may therefore
conclude that, errors from this source do not affect the main conclusions.

DISCUSSION

From the experiments described, it is apparent that those methods which
take into account only the number of resistant clones developing in a given
period, and not the number of resistant individuals in these clones, give a low
estimate of the mutation rate (see table 11 for summary).

Method 1 falls in this category, since it is based upon the proportion of cul-
tures in which no phenotypic mutants have appeared — this being a function
of the total number of phenotypically resistant clones in the series of cultures —
and is independent of the size of the clone.

Method 4 also falls in this category, since the members of a mutant clone
are confined to one spot on the agar and after one member has become re-
sistant others may change in the same manner without being detected.
Method 4, which uses the number of resistant clones appearing during rapid
growth, provides a necessary check on method 1, which uses the number of
resistant clones developing during growth to saturation, and which would be
strongly biased if resistant clones appeared at a different rate during the few
divisions prior to the cessation of growth. The estimates of rate obtained by
these two methods are, within the limits of experimental error, similar, and in
the present experiments — using resistance of strain B/r to the phage Tl —
average approximately 0.5 X 10~ 8 .

Methods that are based on the increase in numbers of resistant individuals
developing during growth, however, as distinct from the numbers of resistant
clones, produce much higher estimates of mutation rate. The following fall in
this category: method 2, which uses the average number of resistant bacteria
in a series of similar cultures started with small inocula; method 3, which uses
the number of resistant bacteria in a single culture started with a large in-
oculum; and method 5, which uses the highest number of resistant bacteria in
any one culture of a series.

Method 3, which is based on a relatively direct calculation, provides a neces-
sary check on method 2, which is subject to a statistical bias. Similarly,
method 5, which is based on the early mutations, provides a check on method
2, which would be biased if the delay in developing phenotypic resistance were
excessive in an appreciable proportion of the lines of descent within a mutant
clone. Method 3 is also of use in this connection in the one experiment in which

472

HOWARD B. NEWCOMBE
Table 11

Averages of mutation rates for strains B and B/r to resistance to phage Tl obtained by methods
1 to 5. Summary of calculations using data, or mutation rates already calculated, from the present in-
vestigation, Luria and Delbruck (1943), Demerec and Fano {1945), and Beale (1948).

MUTATION RATE, XI 0~ 8
METHOD FORMULA '"

Present L. & D. D. & F. Beale

Estimates based on numbers
of resistant clones
(1) Series of liquid cult.,

no resist . b. a = - (In 2) (In P ) /N

(4) Solid medium, resist.

clones a = (ln2)(R 2 -R 1 )/(N 2 -Ni)

0.40

0.32 0.68

0.51 —

0.49*

Estimates based on numbers
of resistant individuals

(2) Series of liquid cult.,

av. res. b. r = (aN/ln2) ln(CaN/ln2) 3.6 2.4

(3) Liquid cult, from

large inocula a = (ln2)(r 2 /N 2 — n/NO/g 3.1 —

(5) Series of liquid cult.,

max. res. b. a = (ln 2)(h-r)/CN 2.2 1.5

2.7 —

Early growth

(4a) Solid med., res.

clones, first div. a=(ln 2)(R 2 -R 1 )/(N 2 -N 1 ) 8.3 —

Symbols:

a, mutation rate per bacterium per division cycle.

r, average number of resistant bacteria per culture.

h, highest number of resistant bacteria in any one culture of a series.

N, average number of bacteria per culture; Ni and N 2 , average number at times 1 and 2

respectively.
C, number of cultures.
R, average number of resistant clones per plate; Ri and R 2 , average numbers at times 1 and

2 respectively,
g, number of cell generations.

Po, proportion of test cultures in which there are no resistant mutants.
* Note: the value .7X10~ 8 bacterial divisions obtained by Beale has been converted to the
rate per bacterium per division cycle by multiplying by In 2 ( = .693).

the growth period is long, and in which the majority of the resistant individuals
are from mutations occurring many generations previous to the cessation of
growth.

The estimates of rate obtained by these methods are all considerably greater
than those which took into consideration only the numbers of resistant clones
and not the numbers of resistant individuals, and differ only slightly among
themselves. In the present experiments they average approximately 3X10 -8 .

It is thus evident that the numbers of resistant individuals in a mutant
clone must be greater than would be expected on the basis of the time of pheno-
typic appearance of that clone. One simple interpretation of this — namely,

74

DELAYED EXPRESSION OF MUTATIONS 473

that the growth rate is higher in the mutant clones than in the parent strain —
is ruled out by the results obtained using method 3. The only plausible in-
terpretation, therefore, seems to be that a mutation to resistance is not pheno-
typically expressed until after a number of divisions have taken place.

Two possibilities exist with regard to such a delay: (1) that after a fixed
period of one or more generations all of the offspring of the original mutant be-
come phenotypically resistant at the same time, and (2) that the delay is vari-
able within the mutant clone, some lines of descent becoming resistant earlier
than others. As pointed out by Luria (1946), the presence of cultures having
only one resistant bacterium eliminates the first of these. Thus the delay must
be variable within the mutant clones, and a single individual may develop
phenotypic resistance in the first instance, being followed later by other mem-
bers of the clone.

As pointed out, the presence of phenotypically susceptible individuals in the
younger mutant clones, and the occurrence of changes in these to phenotypic
resistance, results in a high apparent mutation rate when the members of the
clone are dispersed and the changes are thus rendered detectable in each in-
dividual, and in a low apparent mutation rate when the members of a clone
are grouped together and only the first change to phenotypic resistance is
detectable. In the early growth on agar a situation exists in which the members
of mutant clones in the inoculum are dispersed over the plate but the products
of subsequent divisions of individual bacteria are grouped together. In such
circumstances the apparent mutation rate would be expected to be high during
the first division, as in liquid cultures, and to decline during subsequent genera-
tions to the apparent rate obtained when only the numbers of resistant clones
are considered. This, in fact, has been observed and lends support to the con-
clusions reached.

In this connection, however, it should be noted that the rate of appearance
of resistant bacteria during the first division on agar following a resting stage
is higher than expected, by a factor of approximately three or four. This does
not affect the main line of reasoning, although it is of interest in itself and ap-
pears to be well substantiated by the data.

Attempts to interpret this as due to the bacteria's failing to adsorb phage
during the early stages of growth after a resting stage have not been successful,
since one would have to suppose complete failure of adsorption over a period of
at. least two generations.

Two possibilities remain: (1) that of a genuine high rate of mutation during
this division, together with physiological conditions favoring immediate pheno-
typic expression, since otherwise the products of these mutations would not be
detectable; and (2) that of physiological conditions favoring phenotypic ex-
pression of mutants which had not hitherto become resistant. So far, no
methods have been devised to distinguish between these two possibilities.

Thus the rate of appearance of resistant clones during early growth on agar
is in agreement with the concept of a delayed phenotypic expression, although
an additional phenomenon appears to be involved.

The size of a mutant clone at the time of appearance of the first resistant in-

75

474 HOWARD B. NEWCOMBE

dividual may be obtained by comparing the estimates of mutation rate from
method 1, utilizing the number of resistant clones developing in liquid cultures,
with those from method 5, utilizing the number of resistant individuals in old
mutant clones. Table 9 may be used for this purpose, and the ratio rate (5)/
rate (1) considered. If one omits the data on mutations to B/3, 4 and B/3, 4,
7, because of the bias resulting from the known low growth rate of these mu-
tants (Demerec and Fano 1945, table 7), the ratios in the cases of the remain-
ing mutations appear to be similar. A crude average of these is 6.4. This in-
dicates that there are approximately six or seven individuals in a mutant clone
at the time of appearance of the first resistant individual; that is, that the
delay between mutation and phenotypic expression in at least one member of
the resulting clone is of the order of 2.7 generations.

The extent of the delay in individual lines of descent within a mutant clone
other than the first one to become resistant, cannot be determined directly.
However, the absence of any striking difference between the estimates of
mutation rate obtained by methods 2 and 5 suggests that the delay is not ex-
cessive; and the similar values, 2.8X10~ 8 and 3.4X10 -8 , obtained by method
3 with growth periods of six and eleven generations respectively, indicate that
no very large proportion of a mutant clone changes to phenotypic resistance
after six generations from the time of the mutation.

The possibility that the delay in expression of irradiation-induced changes
is of the same origin as that observed in spontaneous mutations should be con-
sidered at this point, although a positive answer cannot be given, since strictly
comparable studies of the induced changes have not yet been carried out, and,
in particular, nothing is known of the numbers of resistant individuals develop-
ing in the late-appearing resistant clones resulting from irradiation.

If irradiation results in the immediate induction of gene changes similar to
those occurring spontaneously, and the delay is one of phenotypic expression
similar to that in untreated material, then the numbers of resistant individuals
in the late-appearing resistant clones should rise rapidly after the clones first
become detectable, because of susceptible members becoming phenotypically
resistant, and should eventually approach the numbers in the resistant clones
that appeared soon after treatment. Variation in the time of occurrence of
phenotypic expression in different lines of the same mutant clone would also
be expected in irradiated material.

Until information of this nature is obtained it is not possible to say with
certainty that the observed delays in appearance of induced and spontaneous
mutations are of the same origin. In the absence of more detailed information,
one can only speculate on the basis of the variation between clones. Since in
untreated material there is no striking difference between the apparent muta-
tion rates as determined over periods of six and eleven generations (method 3),
it would seem that six generations is adequate both for phenotypic expression
in most of the mutants within a clone and also for expression in at least one
individual in most mutant clones. In irradiated material an appreciable pro-
portion of the resistant clones do not appear until after the sixth generation;

76

DELAYED EXPRESSION OF MUTATIONS 475

and it is therefore tentatively suggested that the two delays are not of the same
origin.

It should also be noted that a delay in phenotypic expression has been ob-
served in Paramecium following hybridization (Kimball 1937; Sonneborn
1943; Sonneborn and Lynch 1934) and has been termed "cytoplasmic lag."
This phenomenon is at least superficially similar to that described in the pres-
ent work, in that "phenotypic expression . . . commonly appears first at
slightly different times in different lines of descent within the clone" (Sonne-
born 1947). The extent of the delay appears to vary in Paramecium with the
nature of the change, and in two instances of alteration in type there is no
detectable delay. (These are the alterations from mating type I to mating type
II in variety 1, and from mating type V to mating type VI in variety 3. For a
discussion of this, see Sonneborn 1947.) It will therefore be of interest to
determine whether a similar diversity in the delay exists in bacteria, since the
apparent differences in the mutations so far studied, all of which mutations
are to resistance to one or another of the phages, are relatively small and can
be explained as due to known differences in the growth rates of the mutants.
Until more information of this nature is available it is not profitable to specu-
late further about whether "cytoplasmic lag" in protozoa and "delayed pheno-
typic expression" in bacteria have a common basis.

conclusions

Evidence of a delay in the phenotypic expression of spontaneous bacterial
mutations to phage resistance, using E. coli, strain B/r, has been obtained. This
evidence comes from a comparison of the rate of appearance of phenotypically
resistant clones during bacterial multiplication with the rate of gene mutation,
the latter being estimated from the numbers of resistant individuals arising
during similar periods of growth.

The rate of appearance of resistant clones per bacterium per division cycle
is approximately 0.5 X10~ 8 , whereas the numbers of resistant individuals cor-
respond to a much higher rate of mutation, a rate of approximately 3X10 -8 .

Since it is known that the mutants do not divide more rapidly than the
parent strain, the excess number of individuals composing a mutant clone in-
dicates that the clone was formed sometime prior to its becoming phenotypi-
cally detectable. Thus there appears to be a delay between mutation and pheno-
typic expression.

The possibilities of alternative interpretations of the data of previous
workers on the basis of a statistical bias in one of the methods, and of a change
in the mutation rate during the later part of growth, have been eliminated by
using methods which are not susceptible to these sources of error.

A delay in phenotypic expression would be expected to result in a high rate
of appearance of resistant clones during the first few divisions on agar, declin-
ing to the value 0.5 X10 -8 with further growth. This has been observed, al-
though an additional phenomenon appears to be present which results in the
appearance of a higher-than-expected number of resistant clones during the

77

476 HOWARD B. NEWCOMBE

first division. This phenomenon, although of interest in itself, does not affect

the main conclusions.

It is estimated that phenotypic expression occurs in at least one member of a
mutant clone approximately two or three generations after the mutation, and
that in many cases it affects at first only one individual, others of the mutants
becoming resistant during subsequent divisions.

Expression does not appear to be delayed in any appreciable proportion of a
clone beyond approximately six generations after the mutation.

ACKNOWLEDGEMENTS

I wish to thank Dr. M. Demerec for much helpful advice and criticism,
Dr. B. McClintock for valuable discussion of the work during its progress
and Dr. S. E. Luria for a number of very useful comments on the manuscript.

LITERATURE CITED

Anderson, E. H., 1946 Growth requirements of virus resistant mutants of Escherichia coli

strain U B." Proc. Nat. Acad. Sci. 32: 120-128.
Beale, G. H., 1948 J. Gen. Microbiol, in press.
Delbruck, M., and Luria, S. E., 1942 Interference between bacterial viruses I. Arch.Biochem.

1: 111-141.
Demerec, M., 1946 Induced mutations and possible mechanisms of transmission of heredity in

Escherichia coli. Proc. Nat. Acad. Sci. 32: 36-46.
Demerec, M., and Fano, U., 1945 Bacteriophage-resistant mutants in Escherichia coli. Genetics

30: 119-136.
Demerec, M., and Latarjet, R., 1946 Mutations in bacteria induced by radiation. Cold Spring

Harbor Symp. Quant. Biol. 11: 38-50.
Kimball, R. F., 1937 The inheritance of sex at apomixis in Paramecium aurelia. Proc. Nat.

Acad. Sci. 23:469-474.
Luria, S. E., 1946 Spontaneous bacterial mutations to resistance to antibacterial agents. Cold

Spring Harbor Symp. Quant. Biol. 11: 130-137.
Luria, S. E., and Delbruck, M., 1943 Mutations of bacteria from virus sensitivity to virus re-
sistance. Genetics 28: 491-511.
Shapiro, A., 1946 The kinetics of growth and mutation in bacteria. Cold Spring Harbor Symp.

Quart. Biol. 11:228-235.
Sonneborn, T. M., 1943 Development and inheritance of serological characters in variety one

of Paramecium aurelia. Genetics 28: 90.

1946 Discussion of paper by Luria, in Cold Spring Harbor Symp. Quant. Biol. 11: 138.

1947 Recent advances in the genetics of Paramecium and Euplotes. Advances in Genetics 1 :
263-358.

Sonneborn, T. M., and Lynch, R. S., 1934 Hybridization and segregation in Paramecium

aurelia. J. Exp. Zool. 67: 1-72.
Wilson, G. S., 1922 The proportion of viable bacteria in young cultures with especial reference

to the technique employed in counting. J. Bact. 7: 405-446.
Witkin, E. M., 1946 Inherited differences in sensitivity to radiation in Escherichia coli. Proc.

Nat. Acad. Sci. 32: 59-68.

1947 Genetics of resistance to radiation in Escherichia coli. Genetics 32: 221-248.

78

ORIGIN OF BACTERIAL RESISTANCE TO ANTIBIOTICS 1

M. DEMEREC
Department of Genetics, Carnegie Institution of Washington, Cold Spring Harbor, New York

Received for publication March 29, 1948

In this brief review of the problem of the genetic aspects of the origin of bac-
terial resistance to antibiotics, I intend to discuss mainly work done in my lab-
oratory. I shall (1) offer evidence that bacterial resistance to penicillin and
streptomycin is not induced by these compounds but originates spontaneously
through genetic changes comparable to gene mutations; (2) describe the resist-
ance patterns observed in experiments with penicillin and streptomycin; and
(3) outline a possible mechanism responsible for resistance, and for the differ-
ences between the resistance patterns observed with penicillin and those observed
with streptomycin.

All experiments were done in vitro. Penicillin was tested with Staphylococcus
aureus, strain NRRL 313 (Demerec, 1945a), and streptomycin was tested with
the same strain of S. aureus and with Escherichia coli, strain B. The strep-
tomycin was obtained from Chas. Pfizer & Co., New York, at a time when this
compound was still very scarce, and I wish here to acknowledge their generosity.
In all the experiments bacteria were first grown in broth cultures without any
penicillin or streptomycin ; these were used only in the tests for resistance, made
by growing bacteria on broth agar plates containing various concentrations of
one or the other antibiotic.

SURVIVAL CURVES

Figures 1 and 2 show the behavior of our strain of S. aureus when plated on
nutrient agar plates to which specified amounts of penicillin (figure 1) or strep-
tomycin (figure 2) had been added. It is evident from these curves that on low
concentrations of either antibiotic all bacteria survived and formed colonies.
Threshold concentrations are indicated on the curves by sharp breaks, the num-
bers of survivors decreasing very rapidly with increase of concentrations. The
slope of the curves is very steep at concentrations near the threshold, but levels
out as concentrations become increasingly higher. These curves show that we
were dealing with mixed populations of types sensitive to antibiotics and types
more or less resistant. Sensitive bacteria made up by far the largest portion of
the populations.

The two curves are very similar in the region of lower concentrations. A
striking difference is evident, however, in the region of higher concentrations;
on the medium containing streptomycin, survivors continued to appear even at
the highest concentration used in the experiments.

1 Substance of a paper delivered at the Symposium on Antibiotics at the Chicago meet-
ing of Section N of the American Association for the Advancement of Science, December 30,
1947.

63

[Reprinted by permission of The Williams & Wilkins Company from Journal of
Bacteriology 56 : (1) 63-74, July, 1948]

79

C4

M. DEMEREC

[vol. 56

Tests of survivors confirmed the conclusion indicated by the survival curves.
Strains isolated from the colonies of plates having high concentrations of either
penicillin or streptomycin proved to be more resistant to the respective compound
and grew on concentrations at least as high as those of the plates from which they

I0 8

^WL 1 1 1

1 1 1 1

1 1 1

I0 7
I0 6

CO

|.0«

>
a:

CO

u. io 4
o

a:

UJ

2 10

z

lr\\\

55

IO 2
10

\m\

\A

W?^^

■

1 1 1 1

1 1 1 1

1 1 1

' .01 .02 .03 .04 .05 .06 .07 .08 .09 .10 .11 .12 .13
PENICILLIN CONCENTRATION' OXFORD UNITS PER ML

Figure 1. Survival curves for Staphylococcus aureus plated on nutrient agar containing
various concentrations of penicillin. The six light curves represent results of six independ-
ent experiments, and the heavy curve represents the average of these experiments.

had been isolated. It was found that resistance to one of these antibiotics is
independent of resistance to the other; that is to say, strains with increased re-
sistance to penicillin are still sensitive to streptomycin, and vice versa. It seems
unnecessary to discuss in detail the by now well-established fact that strains that
once become resistant as a rule continue so.

8o

1948]

ORIGIN OF BACTERIAL RESISTANCE TO ANTIBIOTICS

65

ORIGIN OF RESISTANCE

The numerous experiments made in gathering the data for the survival curves
showed clearly that in large populations of bacteria there were always some indi-
viduals more resistant to the antibiotics than others. Since we used very small

4 6 8 10 12 14 16 18 20f-40 50
CONCENTRATION' UNITS PER ML

Figure 2. Survival curves for Staphylococcus aureus plated on nutrient agar containing
various concentrations of streptomycin.

inocula (50 to 300 bacteria), the proportion of resistant bacteria was too small
to account for their presence by assuming that they came about through division
of one or more resistant individuals that may have been present in the inoculum.
Therefore the resistant individuals must have originated in the experimental
cultures. Two alternative possibilities were considered with respect to the

8i

66 M. DEMEREC [VOL. 56

mechanism of this origin: (1) that resistance was induced by some interaction
between the antibiotic and the bacteria when they were together on the plate;
and (2) that it originated independently of the antibiotic, by mutation, the anti-
biotic acting only as a selective agent in the isolation of mutants by destruction
of sensitive bacteria.

A relatively simple method was available for distinguishing experimentally
between these two possibilities. It was devised by Luria and Delbruck (1943)
in a study of the origin of bacterial resistance to bacteriophages, and was adapted
for work with antibiotics in our study of the origin of resistance to penicillin
(Demerec, 1945a). Following is a brief description of the method as used in
our experiments.

From a single culture of bacteria, small inocula (50 to 300 bacteria) were taken
and used to start 21 or more independent broth cultures. These were incubated
for 24 hours, or until growth had reached the saturation point. During incu-
bation the number of bacteria (in experiments using E. coli or S. aureus) in-
creased to about 2 X 10 8 per ml. From the same culture that served as the
source of the inocula, 10 samples of bacteria of the same size as the inocula were
plated on a culture medium containing the same concentration of antibiotic as
was used in the tests, in order to determine if any resistant bacteria were present
in these samples. The concentration of the antibiotic had to be high enough so
that there were no survivors among the small number of bacteria plated — in
other words, that there were no bacteria resistant to that concentration in the
inocula used to start cultures. It followed, then, that all resistant bacteria found
in full-grown cultures had necessarily originated in these cultures during the
period when the number of bacteria increased from 50 to 300 to about 2 X 10 8
per ml.

Next, from 20 of the broth cultures, samples of 0.1 ml were taken and plated
on petri dishes containing the same concentration of the antibiotic (0.064 units
of penicillin, or 5 units of streptomycin per ml of medium). Fifteen 0.1 -ml samples
were plated from the twenty-first tube. Thus two sets of plates were obtained:
one set of 20 in which each plate had bacteria from a different culture, and another
set of 15 all having bacteria from the same culture. The plates were incubated
for 24 hours, or for a longer period if the growth of colonies was slow. After in-
cubation the number of colonies on each plate was determined. These colonies
represented resistant bacteria that had been present in the sample plated.

On both sets of plates in this test the experimental conditions were similar,
like numbers of bacteria (about 2 X 10 7 ) having been plated onto nutrient agar
containing identical concentrations of antibiotic. Therefore, if resistance were
induced through interaction between the bacteria and the antibiotic when they
were in contact with each other, approximately similar numbers of resistant
bacteria would presumably be obtained on all the plates, regardless of the origin
of the bacterial samples ; the variation between plates should not exceed random
variability. In the event that the origin of resistance is mutational, on the other
hand, similar numbers of resistant colonies would be obtained only among the
platings taken from the same culture, since these represented repeated tests of

82

1948]

ORIGIN OF BACTERIAL RESISTANCE TO ANTIBIOTICS

07

the same mixture of resistant and sensitive bacteria. Among the samples from
separate cultures, if mutations occur at random, a large number of resistant
colonies would be obtained from cultures in which mutation happened to occur
early in the growth of the culture, and a small number of resistant colonies from
cultures in which mutation happened to occur late, provided the growth rate of
resistant bacteria is not appreciably different from that of normal ones. If
resistance originates by mutation, then, the variation in numbers of resistant
bacteria would be much greater between samples taken from separate cultures
than between samples taken from the same culture.

TABLE 1

Number of bacteria (E. coli) resistant to a concentration of 5 units of streptomycin per ml of

agar medium in samples taken from a series of independent cultures and similar samples

taken from a single culture which assayed l.S X 10* bacteria per ml

SAMPLES FROM INDEPENDENT CULTURES

Culture no.

No. of
resistant
bacteria

Culture no.

1

67

11

2

159

12

3

135

13

4

291

14

5

75

15

6

117

16

7

73

17

8

129

18

9

86

19

10

101

20

No. of resistant
bacteria

56
91
123
97
48
52
54
89
111
164

Average 105.9

Variance 2913.9

Chi-square 550.3

P much less than 0.001

SAMPLES FROM SINGLE CULTURE

Sample no.

No. of
resistant
bacteria

Sample no.

1

142

11

2

155

12

3

132

13

4

123

14

5

140

15

6

146

7

141

8

137

9

128

10

121

No. of resistant
bacteria

110
125
135
121
112

Average 131.2

Variance 151 . 1

Chi-square 17.3

P 0.26

Table 1 shows the results of such an experiment with E. coli and streptomycin.
In addition to the tests represented in the table, the concentration of bacteria was
determined in 11 cultures, including the one from which the 15 samples were
taken. The average number of bacteria in 10 cultures was 2.2 X 10 8 per ml,
with extreme variants of 1.9 and 2.3, and the average number in the eleventh was
2.1 X 10 8 per ml. Thus the variation in numbers of bacteria among the differ-
ent cultures was so small that it could have introduced only negligible differ-
ences between the numbers of resistant colonies observed on different plates.
It is evident from table 1 that the variation in number of resistant colonies was
considerably greater among platings from independent cultures than among
platings from a single culture. The extreme variants of independent cultures
were 48 and 291, the average 106, the variance 2,914, chi-square 550, and the
probability that this variation was due to chance is insignificant. On the other

83

€8 M. DEMEREC [VOL. 56

hand, the variation in number of resistant colonies among platings of samples
taken from one culture was very small; and the probability that this variation
was due to chance is 26 per 100 trials. Very similar results were obtained in
experiments using S. aureus and penicillin (Demerec, 1945a).

These results, then, favor the assumption that resistance to certain concentra-
tions of penicillin or streptomycin originates through mutation, and that resistant
bacteria may be found in any large population, the proportion depending on the
mutation rate.

Oakberg and Luria (1947) reached identical conclusions after experimenting
with S. aureus and sodium sulfathiazole. This suggests that mutations may be
generally responsible for the origin of resistance that is transmitted to the off-
spring of the individuals that acquire it.

RESISTANCE STEPS

A very interesting feature of bacterial resistance to antibiotics is the stepwise
increase in degree of resistance that can be brought about by selection. This
feature is particularly well expressed in penicillin resistance. Figure 3 reproduces
curves from an earlier paper (Demerec, 1945a) showing the effect of selection on
the increase in resistance of S. aureus to penicillin. The first is the survival curve
of the stock culture. At a concentration of 0.15 units per ml there were no sur-
vivors, but at a concentration of 0.12 units about 4 per 10 8 bacteria lived. First-
step resistant strains were isolated from stock culture bacteria surviving sub-
lethal concentrations. The second curve of figure 3 is a typical survival curve
of such first-step resistant strains. Some individuals of these strains survived
concentrations up to about 0.2 units. When first-step resistant strains were
grown on sublethal concentrations of penicillin, second-step resistant strains
were isolated from the survivors. A typical second-step survival curve is shown
third on figure 3. Third- and fourth-step resistant strains were obtained in
similar manner.

It is of interest to note that the building up of resistance is more rapid with each
selection step. Thus, with our strain of S. aureus, a concentration of 0.15 units
was sufficient to eliminate all bacteria of the original strain, but a concentration
of about 0.2 units was required to eliminate all bacteria of the first-step resistant
strain, and concentrations of about 0.4 units for the second-step, 1 unit for the
third-step, and 7 units for the fourth-step. The fifth-step strain was for all
practical purposes completely resistant to penicillin, since it was not affected by
a concentration of 250 units per ml. With each step the increase in resistance
appeared to be exponential.

Whereas the building up of resistance to penicillin followed a definite pattern,
resistance to streptomycin showed a considerable degree of variability. Among
first-step resistant strains — that is, among strains isolated from colonies of the
original strain that survived sublethal doses of streptomycin — there were some
that were only slightly more resistant than the original strain, some that were
almost completely resistant, and some that fell between these two extremes
(figure 4). It has been found that the variability in degree of resistance among

84

1948]

ORIGIN OF BACTERIAL RESISTANCE TO ANTIBIOTICS

69

first-step penicillin-resistant strains (Demerec, 19456) is slight as compared with
the variability observed among first-step streptomycin-resistant strains.

10'

10

10

CO

or
> 10

>
cr

3
CO

cr

UJ

CD

§ io :

z

10'

10

.1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 U 12 13 1.4 15
PENICILLIN CONCENTRATION' OXFORD UNITS PER ML

Figure S. Stepwise build-up of resistance to penicillin in S. aureus. Fir6t from left,
survival curve of stock culture; second, survival curve of first-step resistant strain isolated
from a colony of the stock culture growing on the concentration indicated by the arrow.
The other curves are of second-, third-, and fourth-6tep resistant strains isolated from
colonies growing on the concentrations indicated by the arrows.

In the build-up of resistance to streptomycin, resistant strains of the second
step, third step, etc., showed behavior similar to the first-step strains; that is,
they exhibited a wide range of variability in degree of resistance. By plating
bacteria of a low-resistance first-step strain on a concentration of streptomycin
that is sublethal for that strain, one can isolate second- step resistant strains that

k* /J 1 1 1

1^-

i i i i

i i i i

r

i^V

i

\

j_

\

i

►

i

"i

i i i

iiit

i i i i

85

70

M. DEMEREC

[VOL. 56

vary in degree of resistance from only slightly more resistant than the original
strain to very resistant. In fact, there is no difference in degree of resistance
between the most resistant strains of the first step and those of the second, third,

20 30 40 50 60 70 80 90 150 200

CONCENTRATION' UNITS PER ML

Figure 4. Survival curves of first-step strains of S. aureus resistant to streptomycin,
showing the great variability in degree of resistance. Each strain was isolated from a
colony of the stock culture growing on the concentration of streptomycin indicated by the

and higher steps. Consequently, strains that are highly resistant to strep-
tomycin may be obtained either in one step, by selection of survivors of very high
concentrations, or in several steps, by repeated selection of survivors from
bacteria grown on increasingly higher concentrations.

86

1948] ORIGIN OF BACTERIAL RESISTANCE TO ANTIBIOTICS 71

POSSIBLE MECHANISM OF ORIGIN OF HIGH-DEGREE RESISTANCE

The experimental evidence available at present indicates that resistance to
penicillin and resistance to streptomycin are independent of each other; that such
resistance is a heritable property induced by genetic changes comparable to
mutations; that first-step penicillin-resistant strains are fairly uniform in their
degree of resistance, a highly resistant strain being built up by selection through
several steps ; and that first-step streptomycin-resistant strains show a great deal
of variability in degree of resistance, highly resistant strains being produced either
in one step by selection from among first-step resistant mutants or in several
steps by repeated selection of strains having higher and higher degrees of
resistance.

What mechanism is responsible for the stepwise build-up of resistance, and for
the difference between penicillin resistance and streptomycin resistance in this
regard? Evidence accumulated by several investigators in recent genetical re-
search with bacteria makes it appear reasonably certain that mutations in bac-
teria are caused by changes in genes. Granting this assumption, the complexity
of behavior observed in studies of resistance to antibiotics indicates that several
genes must be involved. Such an inference is not new. Several years ago
Demerec and Fano (1945) suggested that the complex situation observed in their
study of resistance of E. coli to 7 phages pointed to the presence of about 20
distinct mutant types in their material. Since many more than 7 phages affect
the strain of coli investigated, and since it is reasonable to assume that extension
of the study to these phages would reveal additional mutants, it is evident that
the genetic background of resistance to phages is very complex indeed, and that
it involves a considerable number of genes.

If a like situation exists in respect to resistance to antibiotics, then it can be
assumed that many genes are instrumental in determining resistance to the two
antibiotics used in these experiments, and that the genes affecting resistance to
penicillin are different from those affecting resistance to streptomycin. If any
one of these genes should mutate, the bacterium in which such a mutation oc-
curred and the strain developed from that bacterium would be more resistant to
the respective antibiotic than was the original parent strain. Such a strain
would be what we have called a "first-step resistant strain."

The fact that first-step penicillin-resistant strains are fairly uniform in degree
of resistance (Demerec, 19456, figure 2) is consistent with the assumption that all
genes affecting resistance to penicillin have a similar potency, so that the effect
of mutation is the same regardless of which of the genes happens to mutate.
According to this hypothesis, there is still present in a first-step resistant strain a
number of unmutated genes that affect resistance. Mutation of any of these
produces a second-step resistant strain, which possesses a higher degree of resist-
ance than the first-step strain. Similarly, by mutation of another gene in a
second-step resistant strain, a still higher degree of resistance is attained, charac-
teristic of the third-step resistant strain. From this, by further mutations,
highly resistant fourth- and fifth-step strains may be obtained. The curves in
figure 3 indicate that the increase in degree of resistance with each step is ex-

87

72 M. DEMEREC [VOL. 56

ponential. This means that the effect of two or more mutants together is con-
siderably greater than would be expected from the added values of the effects of
single mutants.

No attempt has been made to determine the frequency with which genes affect-
ing resistance to penicillin mutate. It may be estimated from survival curves,
however, that the mutation rate is low, in the neighborhood of 1 X 10 -8 . With
such a low mutation rate, it is evident that the increase in resistance must occur
in successive steps, and that the chance that one step will be skipped is very
slight, the chance that two steps will be skipped practically nil. (A step would
be skipped when mutations in two genes occurred simultaneously in the same
bacterium; and two steps would be skipped if three mutations occurred simul-
taneously in a single cell. The chance of two simultaneous mutations is 10 -8 X
10-s = 1 q-i6 ) an( j f three simultaneous mutations, 10" 8 X 10~ 8 X 10~ 8 = 10~ 24 .
Since the volume of an S. aureus cell is about one cubic micron, it would be ex-
pected that one double mutant, on the average, would be found in ten liters of
bacteria, and one triple mutant in one million cubic meters.)

The observed behavior of resistance to streptomycin also can be explained by
assuming the existence of several genes determining such resistance. Unlike the
genes for penicillin resistance, however, these differ greatly from one another in
potency. If a gene of low potency mutates, the first-step resistant strain will
have a low degree of resistance, but if mutation occurs in a highly potent gene,
the first-step resistant strain will be highly resistant. Consequently, consider-
able variation in degree of resistance is to be expected between first-step strains;
and for the same reason a highly resistant strain may be obtained either in one
step, by selection of a highly resistant first-step mutant, or in several steps, by
selection of mutants of low resistance values.

CLINICAL CONSIDERATIONS

A major consideration in the clinical use of antibiotics is how to avoid the
development of resistant strains, since the usefulness of an antibiotic is closely
related to the number of resistant pathogens and to their incidence in infections.
For this reason, analysis of the mechanism of origin of resistance to penicillin and
streptomycin has an important bearing on the clinical application of these
antibiotics.

From the clinical standpoint, the situation in regard to penicillin is relatively
simple and well defined. Since resistance develops in steps, and it is very un-
likely that a step will be skipped in the process, the clinician can avoid develop-
ment of resistant pathogens by using initial doses that are adequate for the
elimination of first-step resistant individuals. Fortunately, most of the common
pathogenic strains that have been investigated (North and Christie, 1945; Meads
et al., 1945) are very sensitive to penicillin, so that large doses are not required in
clinical use. It is equally important for the clinician to maintain the effective con-
centration in treatment as long as the infection persists, because decrease of the
concentration below the effective level will permit the accumulation of first-step
resistant bacteria, which may increase to a point that will allow the occurrence of

88

1948] ORIGIN OF BACTERIAL RESISTANCE TO ANTIBIOTICS 73

second-step mutants. These are difficult to control. If there is any suspicion
that a pathogen may be more resistant to penicillin than the usual strains, it is
advisable to determine the degree of resistance before starting treatment, and to
adjust the concentration accordingly. If adequate precautions are taken against
the development of second-step resistant bacteria, there should be no danger from
resistant pathogens in penicillin treatment. It is particularly important to avoid
the indiscriminate use of penicillin, however, especially for applications where it
can scarcely be of any help (for example, as a mouth wash), because there is
positive danger that such use may stimulate the development of resistant strains.

In the clinical use of streptomycin, the situation can be controlled to a much
smaller extent. Since highly resistant bacteria are found among the first-step
mutants, treatment with high concentrations is not effective in eliminating the
whole population of bacteria present in an infection. What it does accomplish
is a reduction of the number of bacteria to a level with which the organism is
capable of dealing. If for some reason the organism cannot do this, the chances
for the development of a resistant strain are exceedingly good. Therefore, it
must be expected that pathogenic strains resistant to streptomycin will frequently
develop, in the course of time replacing sensitive strains in communities where
streptomycin is used and rendering this antibiotic ineffective. This eventuality
can be postponed by restricting the use of streptomycin to serious infections
which cannot be controlled in any other way.

It may be appropriate to mention here the most effective way, theoretically,
of preventing the origin of resistant strains of bacteria. This is the use in clinical
treatment of a mixture of two antibiotics, when such are available, that affect the
same pathogen but are independent in their actions. The evidence of independ-
ence is that bacterial strains that have developed resistance to one antibiotic are
still sensitive to the other, and vice versa. If such a mixture of two antibiotics
is used, then only bacteria that are resistant to both can survive the treatment
and form first-step resistant strains. Such bacteria would be exceedingly rare.
For example, if first-step resistant bacteria for each of two antibiotics should be
found in a large population with a frequency of 1 X 10 -7 , then the expected
frequency of bacteria resistant to both these antibiotics would be 1 X 10~ u .

SUMMARY

A method is described that has been used to determine whether resistance to
streptomycin is induced by interaction of the compound with bacteria or origi-
nates by gene mutation. Data are presented indicating that mutations are
responsible for the origin of streptomycin resistance in Staphylococcus aureus.
These agree with previously published data regarding the origin of penicillin
resistance in the same organism.

The stepwise increase of resistance to penicillin by selection is explained by
assuming that mutations in several equally potent genes are effective in inducing
resistance, and that the slight degree of resistance characteristic of the first step
is due to a mutation in one of these genes, the higher degrees of resistance of
subsequent steps to successive mutations in other genes.

89

74 M. DEMEREC [VOL. 56

The increase in resistance to streptomycin also can be explained by the as-
sumption that several genes are instrumental in the process. These genes vary
greatly in their potency, however, and consequently a mutation in a highly potent
gene will be responsible for a high degree of resistance, a mutation in a less potent
gene for a low degree of resistance.

From the knowledge gained concerning the mechanism of origin of resistance,
it is concluded that in treatment with penicillin the development of highly re-
sistant strains can be avoided by application of the penicillin in doses sufficiently
large to prevent survival of first-step resistant mutants. In treatment with
streptomycin, however, the development of highly resistant strains cannot be
prevented; effective treatment does not eliminate all bacteria, but it probably
reduces their number to a level at which the organism is able to eliminate them.

REFERENCES

Demerec, M. 1945a Production of Staphylococcus strains resistant to various concen-
trations of penicillin. Proc. Natl. Acad. Sci. U. S., 31, 16-24.

Demerec, M. 19456 Genetic aspects of changes in Staphylococcus aureus producing
strains resistant to various concentrations of penicillin. Ann. Missouri Botan. Garden,
32, 131-138.

Demerec, M., and Fano, U. 1945 Bacteriophage-resistant mutants in Escherichia coli.
Genetics, 30, 119-136.

Luria, S. E., and Delbruck, M. 1943 Mutations in bacteria from virus sensitivity to
virus resistance. Genetics, 28, 491-511.

Meads, M., Ory, E. M., Wilcox, C., and Finland, M. 1945 Penicillin sensitivity of
strains of six common pathogens and of Hemophilus hemolyticus. J. Lab. Clin. Med.,
30, 725-729.

North, E. A., and Christie, R. 1945 Observations on sensitivity of staphylococci to
penicillin. Med. J. Australia, 2, 44-45.

Oakberq, E. F., and Luria, S. E. 1947 Mutations to sulfonamide resistance in Staphy-
lococcus aureus. Genetics, 32, 249-261.

90

Chloromycetin Resistance in £. coli, a Case
of Quantitative Inheritance in Bacteria

Drug resistance in bacteria has been investigated
genetically in several cases, since Demerec began
with penicillin resistance in Staphylococcus aureus 1 .
The proposal to distinguish two classes of drug
resistance — those in which the top level of resistance
can be obtained in a single step, and those in which
it can be acquired only gradually — has clearly a
practical importance ; its theoretical meaning has
not been entirely % elucidated, since genetical analysis
has several limitations in bacteria. In fact, while
mutational patterns are open to investigation in most
cases, crossings can only be regularly undertaken,
as in Lederberg and Tatum's original discovery 2 ,
in a few strains of E. coll. In one of these, E. coli
K-12, some work on the genetics of drug resistance
has been carried out, the most extensive investigation
relating to streptomycin. Full resistance in E. coli
is developed in a single step 3 - 4 .

Besides resistance, dependence is known, whereby
some resistant mutants fail to grow in absence of
streptomycin. Full streptomycin-resistance and
dependence 4 - 8 seem to be due to a single locus (or
closely linked loci). Further, Lederberg introduced
resistance to azide 6 as a genetic marker ; here a
single step brings about a moderate degree of resist-
ance, and the determining factor has been mapped,
not only in this laboratory but also independently
in two other laboratories, and found to be about
midway between V x and TL on the 'chromosome
map' of E. coli K-12. Nitrogen mustard resistance
was found to be gradual or abrupt in increase in
different experiments, only a moderate degree of
resistance being acquired, which made a detailed
analysis difficult. In E. coli K-12 nitrogen mustard
resistance is not accompanied by higher resistance to
radiations, as in the case of E. coli B.

Completely 'gradual' increase of resistance has
been found in the case of Chloromycetin in E. coli
K-12. Selecting for Chloromycetin resistance by
serial transfers in liquid cultures containing increasing
concentrations of the drug (on a geometric scale), a
rapid and relatively smooth increase of resistance is
observed. Starting from a tolerated level of 5-10
[jgm./ml., as is usual with strains derived from the
original E. coli K-12, levels of resistance as high as
1,280 ijgm./ml. (which is not far from the solubility
limit of Chloromycetin) can be obtained. The increase

[Reprinted by permission of MacMillan & Co. Ltd. from Nature 166:991, December 9,
1950]

91

in resistance is twofold or lower, on the average,
at every transfer. Resistance is relatively stable
on repeated subculture in the absence of Chloro-
mycetin.

The smoothness of increase of resistance might
constitute a prima facie case for non-genetical
adaptation ; if, on the other hand, adaptation is
genetical, a relatively large number of genes should
be found affecting Chloromycetin resistance, assuming
that it is to be accounted for by known genetical
mechanisms.

A cross was made between a W 677 strain
(T-L-B^Lac.-V/ Mal x - Gal x - Xyl- Ara- Mil-)
which had been grown with 640 |i.gm./ml. Chloro-
mycetin in broth, and a fully sensitive 58-161 strain
(B-M-). Crosses were made with the usual technique
introduced by Lederberg and Tatum 2 . By this
technique, recombinants are selected out of a mix-
ture of parental cells using 'fixed' recombination
markers, in this case capacity of growth in absence
of given growth factors. Strain 58-161 can grow in
the absence of threonine, leucine, vitamin 3 lt which
are needed by W 677 for growth ; strain W 677 can
grow in absence of biotin and methionine which are
needed for growth by 58-161. Plating the mixture
of the two strains on minimal medium, where such
growth factors are not supplied, parental cells will
not develop, while recombinants of the right type
can develop into visible colonies. The two strains
differ by a number of other markers, lactose fermenta-
tion (absent in W 677, present in 58-161), reaction
to virus Ti ( VS indicating resistance in W 677, while
58-161 is sensitive), etc., as indicated by the other
symbols.

Such differences can be used as 'free' recombination
markers in that they are not directly selected for
or against in the minimal agar plates, and their
segregation in the recombined colonies is used to
build a 'chromosome map' 7 . The relative order of
the genes which have been 'mapped' satisfactorily,
and which are relevant for this and other crosses
conducted during the present work, is as follows :
fi 1 -( J B,il/)-F 8 -Lac 1 -F 1 -^z-(T,L), where V, refers to
virus T t resistance, Az to azide resistance.

Recombinants between the chloromycetin-resistant
W 677 and sensitive 58-161 showed a scatter from
full sensitivity to nearly full resistance, the resistance
of each recombinant being roughly proportional, on
the average, to the length of the chromosome segment
which — as could be inferred from recombination
markers — given recombinants got from the resistant
parent. A closer analysis shows one and possibly
two loci for resistance in the segment between BM
and Lac ; one locus at least between Lac and V x —

9 2

there is evidence that other loci, too, must take part
in building full resistance. These data are therefore
consistent with the hypothesis of many loci with
cumulative action.

First-step resistants, obtained by both plate and
test-tube selection methods, were analysed in crosses
of sensitive with resistant and resistant with resistant,
and they showed a clear-cut segregation of sensitivity
or resistance, within experimental error. Mapping
loci of first-step resistance, the loc;:s most frequently
met so far was closely linked to F,. In this way, a
cross with four free markers between BM and TL
(fixed recombination markers) was made, involving
Chloromycetin resistance, V t , Lac and V lt giving a
relatively satisfactory test of linearity of the chromo-
some region between BM and TL. First -step resist-
ants were also found which gave no resistant recom-
binants when crossed to sensitive, probably because
of close linkage with fixed recombination markers
(BM in this case). The interactions in resistance
between different loci are under investigation at
present. In fact, this case seems to offer un-
usual possibilities of a detailed analysis of the inter-
actions between genes in quantitative inheritance.
The stability of the drug is a factor of no little
importance in facilitating quantitative work with
Chloromycetin.

The K-12 strains used in this research have been
kindly sent by Dr. J. Lederberg. One of us (G. A. M.)
has been supported during this work by a grant from
the Rockefeller Foundation.

L. L. Cavalli
G. A. Macoaoaro

Department of Genetics,

Whittingehame Lodge,

44 Storey's Way,

Cambridge.

Aug. 15.

1 Demerec, M., Ann. Mist. Hot. Oard., 82, 131 (1946).

• Lederberg, J., and Tatum, E. L., Nature, 168, 558 (1946).

1 Newcombe, H. B., and Hawirko, R., J. Bad., 67, 565 (1949).
4 Demerec, M., Amer. Nat., 84, 5 (1950).
' Newcombe, H. B. (unpublished).

• Lederberg, J., J. Bad., 69, 211 (1950).
' Lederberg J., Genetics, 32, 505 (1947).

93

Studies on Nutritionally Deficient Bacterial Mutants
Isolated by Means of Penicillin 1

By Bernard D. Davis 2 , New York, N.Y.

Studies on the mode of action of chemotherapeutic
agents have not shown any direct effect on the known
catabolic processes of bacteria. There is good reason,
by exclusion, to suppose that they act by means of
interfering with anabolic reactions, about which much
less is known. In the last decade, Beadle and his
school have made striking contributions to this area in
biochemistry, as well as to genetics, by the use of
nutritionally deficient ("biochemical") mutants of the
mold, Neurospora -mutants which have lost one or
another specific enzymic activity 3 . More recently,
bacteria have been found to yield mutants of the same
sort 4 . These microbial mutants offer a promising
approach to the major theoretical problems of chemo-
therapy: the mechanism of chemotherapeutic action
and the biochemical alterations responsible for drug-
resistance in microorganisms. Furthermore, the pheno-
menon of drug-resistance presents a direct problem in
microbial genetics. Finally, just as the nutritional
requirements of various wild-type microorganisms have
been of tremendous value in isolating the known
vitamins, so the extension of this approach to the
induction of artificial nutritional requirements offers
promise as a method for detecting metabolites peculiar
to microorganisms -metabolites of special promise as
models for the construction of inhibitory analogues.
For these several reasons this laboratory, interested in
chemotherapy, has felt warranted in undertaking an
excursion into microbial genetics.

It is a simple matter to isolate, from huge microbial
populations, rare mutants that can survive or prolifer-
ate in a medium which suppresses the parent strain.
Examples include mutations to drug-resistance, to
bacteriophage-resistance, or to decreased nutritional
requirements. Comparable techniques have not been
available for the nutritionally deficient mutants,
though these are in some respects more interesting. In

1 Based on material delivered in a symposium on The Relation
of Genetics to Biochemistry at the annual meeting of the American
Chemical Society in San Francisco, Cal., March 30, 1949.

2 U. S. Public Health Service, Tuberculosis Research Laboratory,
( ornell I iniversity Medical College, New York -21, X. Y. - This paper
is in the hands of the editor since [une 16th, 1949,

3 G. W. Beadle, Physiol. Rev 2S, 643 (1945); Chem. Rev.
:i7, 15 (1945).

4 E.L.TATUM,ColdSpringHarborSymp.Quant.Biol.lJ,278(1946).

the early investigations, deficient mutants were isol-
ated by testing of spores or colonies selected at random--
a tedious process, since the total frequencv of recogniz-
able mutants, even after optimal irradiation, rarely
exceeds 1 or 2 per cent. Improved techniques of isola-
tion have recently been introduced for Neurospora 1
and other molds 2 , and for bacteria 3 . In the case of the
bacteria, the techniques depended upon delayed or
limited enrichment of the medium, with consequent
production of small colonies of mutants which could be
distinguished from the larger colonies of the parent
strain. Since the colonies must not be too crowded,
these techniques were limited to a few hundred colonies
per Petri dish.

The penicillin method

For the isolation of rare mutants, a method permit-
ting selection from much larger populations would be
desirable. The possibility of such a method suggested
itself on the basis of the report that penicillin has the
remarkable property of sterilizing ("killing") bacteria
only under conditions which permit growth 4 . We have
confirmed and extended this conclusion. In a minimal
medium containing glucose, lactate, ammonia, and
sulfate as sole sources of carbon, nitrogen, and sulfur,
various amino acid-requiring mutants of E. coli are
completely resistant to the bactericidal action of
penicillin. The addition of the required amino acid,
however, renders the mutant as sensitive to penicillin
as its parent wild-type strain. In applying this method
to the isolation of new mutants, a suspension of E. coli
was irradiated with ultraviolet light, further cultivated
in an enriched medium, washed, and then incubated,
in inocula of suitable size, with penicillin in. minimal
medium for 24 hours. Up to 100 p. c. of the large num-
ber of survivors, recovered by plating in agar media
supplemented with casein hydrolysate or yeast extract,
were found to be nutritionally deficient mutants. The
penicillin method has been independently developed in

1 J. Lein, H. K. Mitchell, and M. B. Houlahan, Proc. Nat.
Acad. Sci. 34, 435 (1948).

2 N. Fries, Nature 159, 199 (1917).

3 J. Lederherg, and E. L. Tatim, J. Hid. Chem. 165, 3S1
(1946). - B. I). Davis, Arch. Biochem. 20, 166 (1949).

4 ('.. I.. Hobby, K. Meyer, and E. Chaffee, Proc. Soc. Exp.
Biol. Med. 50, 281 (1942).

[Reprinted by permission of Vcrlag Birkhauser, Basel, from Expi rii ntia 6 : (2) 41-50, 1950]

94

42

IS. D. Davis: Studios on Nutritionally Deficient Bacterial Muta

this laboratory 1 and by Lederbekg and Zinder 2 .
After isolation, the growth requirement of a mutant
can be simply determined by distributing tiny drops
of solutions of nutrilites on the surface of a heavily
seeded pour plate 3 (Beijerinck's "auxanography").

Since the publication of this method, a modification
has been introduced which substantially improves its
efficiency. Instead of inoculating the washed organisms
directly into minimal medium containing penicillin,
the inocula are placed in a medium lacking nitrogen.
After 4 to b hours of incubation, during which dissimil-
ation of glucose promotes the exhaustion of stored
metabolites, the minimal medium is completed by
addition of ammonium sulfate. Penicillin is added, and
the procedure is continued in the usual way. The period
of dissimilation particularly improves the recovery of
certain vitamin-requiring mutants.

Although the penicillin method provides a marked
improvement in the efficiency of isolating mutants, it
nevertheless has serious limitations. The population
density cannot be indefinitely large, as is possible with
the drug-resistant mutants, for at population densities
above 10 5 to 10 7 cells per ml, the non-mutants release
enough of various metabolites ("syntrophism" ; cf. 4 ) to
permit certain mutants to grow slightly and hence be
sterilized by penicillin. In addition, even at low popula-
tion densities the recovery of mutants is not quantit-
ative. Finally, the use of this method for estimating
mutation rates is limited by the requirement of a stage
of intermediate cultivation between the irradiation
and the exposure to penicillin, in order to permit pheno-
typic expression of the induced mutation. During this
cultivation, the distribution of the population will
undoubtedly be distorted.

Delayed phenotypic expression

The requirement of a certain amount of growth for
the phenotypic expression of an induced mutation is
itself an interesting point, and has also been reported
for phage-resistance 5 . The explanation, at least with
the nutritionally deficient mutants, appears to be that
the cell must undergo some growth, possibly several
generations, before the premutational products of the
mutated gene (enzymes; possible intermediates be-
tween genes and enzymes) are exhausted. Only then
will the pattern of enzymes in the cell correspond to
the new pattern of genes. For this particular mechanism
of delayed phenotypic expression we have proposed
the term "phenomic delay" 6 , the "phenome" being

1 i; I). Davis, J. Amer. Chein. Soc. 70, 1267 (1948); Proc. Nat.
Acid. Sri. 35, 1 (1949).

2 J. I.i in Kiuui., and N. J. Zinder, Amer. Cheni. Soc. 70, 1267

3 J. Lederberg, J. Bact. 52, 503 (1946). G. Pontecorvo, J.
Gen. Microbiol. -3, 122 (1949).

4 J. Lederberg, J. Bart. 52, S03 (1946).

5 M. Dkmerec, and R. Latarjet, Cold Spring Harbor Symp.
Quant, Biol. 11, 38 (1946).

6 B. I). Davis, Proc. Nat. Acad. Sci. 35, 1 (1949).

defined as the total non-self-reproducing part of the
cell, under the control of the self-reproducing genes.
The occurrence of the same phenomenon in mutations
in the reverse direction will be discussed later.

Other possible explanations of the requirement of
intermediate cultivation in the penicillin method in-
clude segregation of mutant and non-mutant nuclei
from a multinucleate cell, and a syntrophic effect of
the non-viable irradiated bacteria, which would pro-
mote sterilization of mutants by penicillin. Evidence
will be published elsewhere that neither of these
mechanisms furnishes an adequate explanation, while
the phenomic delay accounts for all the available facts.
A similar conclusion was reached by Newcombe in a
thorough analysis of delayed phenotypic expression of
phage resistance 1 .

Biochemical advantages of bacteria

Bacteria are less easily studied genetically than
molds such as Neurospora, which can be made to
multiply sexually or asexually at will. While genetic
recombination in bacteria has recently been demon-
strated by Tatum and Lederberg 2 , it apparently
occurs in only a few strains ami a tiny proportion of the
population. For biochemical investigation, however,
bacteria appear to have several advantages. Not only
can a variety of mutants be isolated relatively quickly,
but it is possible to demonstrate very simply, by the
syntrophic interaction of adjacent streaks on solid
media, the instances in which one mutant accumulates
a metabolic intermediate which is utilized as a nutrilite
bv another mutant. Metabolite accumulations, when
present, are extremely useful in analysing biosynthetic
pathways 3 .

A further advantage of bacteria is the uniformly
dispersed growth of certain species in liquid media.
which permits simple and precise quantitative experi-
ments, using colony counts for low population densities
and turbidimetry for high densities. On solid media,
the production of uniform colonies has made possible a
variety of experiments involving prolonged cultivation.
without risk of confusion from back-mutants, which
are readily distinguished from the rest of the population.
In addition, slight variations in colony size and syntro-
phism have made it possible to recognize unexpected
phenomena which might easily have gone unnoticed
in a mycelial mat. Finally, in relation to chemotherapy,
the metabolism of bacteria is of particular interest.
Although some of the problems to be described here
are still under investigation, it seems desirable to
illustrate at this time the types of phenomena that can
be revealed by these primitive techniques, especially
bv the test for syntrophism. This effect has been long

1 H. B, Newcombe, Genetics 33, 4 17 (1948).
- E. I.. Tatum, and J. Lederberg, J. Bact. 53, 673 (1947).
J. Lederberg, Genetics 32, 505 (1947).

3 N. H. Horowitz, J. Biol. Chem. '«-', 413 11946).

95

15. D. Davis: Studi

Nutritionally Deficient Bacterial Mutants Isolated by Means of Penicillin

known in microbiology as the satellite phenomenon;
systematically employed, it has been our vade-mecum.

Mutants obtained

Mutants of E. colt ("Waksman" strain 1 , Amer. Type
Culture Collection 9637) have been obtained with re-
quirements for all the naturally occurring amino acids
except alanine, aspartic acid, and hydroxyproline. A
number of mutants respond to either serine or glycine;
thus far, none of our strains has been specific for either
of these interconvertible amino acids. The sulfur-
deficient mutants, which are a very common class,
respond to cystine or less rapidly to methionine, and
are blocked in the reduction of sulfate to sulfite or of
sulfite to thiosulfate or sulfide. In addition, there are
mutants, unresponsive to thiosulfate or sulfide, with
specific requirements for cystine and others for
methionine. Besides specific proline-requiring mutants,
there are others which respond to either proline or
glutamic acid or a-ketoglutaric acid (but not ornithine) .

Many of these types of mutants have already been
isolated from E. coli by earlier techniques 2 . In addition,
we have isolated strains with more complex require-
ments which should throw light on certain metabolic
relationships. Alternative requirements exist for lysine
or threonine, and, in another mutant, for a-amino
butyric acid or isoleucine (or, curiously, D-threonine
but not L-threonine). One peculiar mutant responds
either to methionine or to thiamine: to methionine in
the concentrations of several micrograms per ml usual
for amino acid mutants; to thiamine or its pyrimidine
in the concentrations, one thousandth as great, re-
quired by other thiamine mutants. Another mutant
similarly requires, under special conditions, either
methionine or vitamin B-12. Finally, there are several
mutants with a multiple requirement apparently due
to a single genetic block: isoleucine plus valine;
phenylalanine plus tyrosine; phenylalanine plus
tyrosine plus tryptophan; and these three aromatic
amino acids plus />-amino benzoic acid. No peptide-
requiring mutants have been obtained; although much
of our isolation work has been done with tryptic casein
hydrolysate, which contains many peptides, all the
mutants isolated from this enrichment have grown on
a mixture of known amino acids. One serine or glycine
mutant and one methionine mutant, like several re-
ported Neurospora mutants, are temperature sensitive,
with an absolute requirement at 37°C, and none at
25°C.

With yeast extract or hydrolysed yeast nucleic
acid, mutants have been obtained with requirements

1 The initiation of this work with the "Waksman" strain has
In en accidental. For any new program, it would undoubtedly be
preferable to use the K-12 strain of E, coli, with which genetic
rei ombination can be studied 3 .

- E.L. Tatum, Cold Spring Harbor Symp. Quant. Biol. n, 278
(1946). - B. I). Davis, Arch, liiochem. 20, 166 (19491.

:l ]■;. I.. Tatum, and .]. Li: of rrf.ro, J. Bact. S3, 673 (1947).

for purines or pyrimidines. The purine mutants,
however, have all responded to adenine or guanine or
hypoxanthine or their ribosides or nucleotides, while
several have also responded to xanthine ; the pyrimidine
mutant has responded to cytosine or thymine or uracil
or their ribosides or nucleotides. Because of this non-
specificity, this group has not been further studied.

Among the vitamins, mutants have been obtained
with individual requirements for thiamine, nicotin-
amide or nicotinic acid, pyridoxin or its amine or
aldehyde, />-amino benzoic acid (PABA), pantothenic
acid, and biotin. In spite of a number of attempts, none
have been obtained with requirements for riboflavin,
inositol, choline, or hemin. The reason for these
failures is not apparent. One mutant requiring an
unknown factor in yeast extract has been obtained
twice.

Microbiological assay

We have not engaged in extensive studies on the use
of these mutants for microbiological assay. Since the
medium employed is so simple, mutants might be
expected to have some advantage over the wild-type
species with complex requirements which are generally
used. In addition, turbidimetric assay of bacterial
growth is more convenient than the measurement of
dry weight of pellicle which is required with Neuro-
spora mutants 1 . The advantage of the mutants would
be largely lost, however, if a heavily enriched medium
were required in order to prevent other substances
present in the material under assay from altering the
quantitative response to the required nutrilite. Since
secondary effects of other substances have been
observed in many instances with wild-type organisms,
it would be necessary with each mutant to determine
the conditions for assay. Furthermore, the instability
of most of the mutants requires caution. With several
mutants of ordinary stability (phenylalanine, tyrosine)
we have been able to obtain satisfactory growth curves,
without significant appearance of reversions, at 24
hours, but by 48 hours irregular increases in turbidity
due to reversions had appeared.

Syntrophism

The syntrophic technique is illustrated with three
arginine-requiring mutants which are shown in Fig. 1
to be blocked at different stages in the well-known
Krebs-Henseleit scheme: ornithine-x:itrulline->-argi-
nine. Sets of mutants of both Neurospora and E. coli,
blocked at these stages, had previously been reported 2 ,
having been recognized by their response to these
precursors of arginine, In the case of growth factors
whose precursors are unknown, however, and which

1 F. J. Ryan, Feder. Proc. S, 366 (1916).

8 A. M. Srb, and N. H. Horowitz, J. Biol. Chem. 154, 129 (1944).
R. R. Roepke, quoted by E. L. Tatum, Cold Spring Harbor Symp.
Quant. Biol. 11, 278 (1946).

9 6

44

B. D. Davis: Studies on Nutritionally Deficient Bacterial Mutants Isolated by Means of Penicillin [Experientia Vol. VI/2]

are therefore the most interesting, this technique of
substituting precursors cannot be used, except by
guesswork, to recognize differences in the site of
genetic blocks. For this reason, the alternative tech-
nique of syntrophism is particularly useful. This

Fig. 1.- Growth of mutants and wild-type (S stock) on minimal
medium enriched with ornithine, citrulline, or arginine. It is seen
that one mutant responds to any of the three related compounds,
another to either citrulline or arginine, the third to arginine only.

technique is shown in Fig. 2, in which the mutant
specifically requiring arginine is seen to excrete a
factor, presumably citrulline, which stimulates the
growth of the two mutants blocked earlier ; while the
citrulline mutant in turn feeds the ornithine-requiring
mutant. The gradient of observed growth reflects the
gradient of diffusion through the agar.

Similar relationships, involving unknown precursors,
have been observed among proline and among histidine-
requiring mutants and among certain mutants requir-
ing aromatic amino acids. Syntrophic accumulation of
precursors appears to be quite widespread with E. coli
and has been observed among sets of mutants (e. g.
arginine) which fail to show accumulation in Neuro-
spora. Its absence between two mutants with a com-
mon requirement, however, does not prove that they
are blocked at the same enzymic site, since precursors
may fail to accumulate because of instability, diversion
along an alternative path, or the inability of the cell
to build up a concentration adequate for excretion.

Causes of syntrophism

Since syntrophism leads to recognition of the ac-
cumulation of a precursor whose subsequent isolation
and identification will contribute to the analysis of a
biosynthetic chain, it becomes important to be aware
of other possible sources of this phenomenon. Several
have so far been observed.

(1) Accumulation of the precursor of a genetically
blocked reaction. This mechanism is the one discussed
above.

(2) Excretion of metabolites by wild-type E. coli. It is
known that bacteria not only remove nutrilites from
the medium, but contribute metabolic products to it.
In order to find whether these include growth factors
for any available mutants, the wild-type strain was
streaked on minimal medium adjacent to mutants with
single requirements for each of the factors listed above.
After 48 hours, three of the mutants were so heavily
fed as to produce maximal growth. These were the
strains requiring biotin, PABA, and pantothenic acid.
In addition, the nicotinamide-less mutant was moder-
ately fed. The other available vitamin-requiring
mutants (thiamine and pyridoxin), as well as a purine,
a pyrimidine, and all the amino acid-requiring mutants,
showed that the amounts of their required factors
excreted by wild-type were negligible, supporting at
most only microscopic growth after 2 to 5 days. It is
concluded that this strain of E. coli is economical in its
synthesis of all the growth factors that could be tested
except four vitamins. It is of interest to note, however,
that the amounts of the other factors excreted, at

Fig.. 2. - Syntrophism among arginine-requiring mutants of Fig. 1.
Mutant responds to ornithine or citrulline or arginine; C responds
to citrulline or arginine; A responds to arginine onlv. 48 hours of
growth (37° C) on medium enriched with very small amount of casein
hydrolysate ("NZ Case").

97

15. II. 1050]

Deficient Bacterial Mutants Isolated

least in the presence of penicillin are sufficient to
limit the permissible population density in the penicillin
method of mutant isolation ; this procedure is appar-
ently much more sensitive to traces of syntrophism
than is the technique of adjacent streaks.

(3) Conversion of a precursor. It has been observed
that wild-type, growing on an excess of the keto acid
precursor of isoleucine, will feed a mutant that responds
to isoleucine but not to its keto acid. The same con-
version and excretion is carried out by a mutant,
blocked earlier, which can use this precursor as well as
isoleucine itself. It has also been observed with wild-
type acting on certain other precursors. In some
reactions, however, such as the conversion of ornithine
or citrulline to arginine, it has not been possible with
an excess of either precursor to stimulate the excretion
of arginine by wild-tvpe or by a mutant. It is evident
that the organism possesses more than one type of
mechanism for determining the rate of synthesis of an
amino acid ; in the case of isoleucine, the capacity for
amination of the keto acid exceeds the requirement,
and hence cannot be the rate-governing mechanism.

A +- »■ specific thiazole .
B -\— ► specific pyrimidine '

C -] — ► pantonine -\— >• pantoic acid,
D -|— »• /S-alanine

pantothenic acid

Fig. 3. - Synthesis of thiamine and pantothenic acid. Blocked arrows
represent sites of genetic blocks.

(4) Excretion of a lone conjugant. Thiamine is
composed of two moieties, a substituted thiazole and a
substituted pyrimidine (Fig. 3). Thiamine-less mutants
were obtained which respond respectively to the
pyrimidine, to the thiazole, and to neither (implying a
deficiency in the conjugation of the two) ; a fourth type,
whose site of genetic deficiency is not readily inter-
preted, responds to the specific thiazole plus pyrimi-
dine, as well as to thiamine itself, but not to either
moiety alone. Tests for syntrophism showed that the
thiazole-less mutant feeds the pyrimidine-less mutant,
but not vice versa. Mutants with similar blocks and
accumulations have been observed with Neurospora 1 .

Another vitamin, pantothenic acid, also consists of
two components, pantoic acid and /J-alanine (Fig. 3).
W. K. Maas in this laboratory has isolated various
pantothenic acid-less mutants which respond respect-
ively to pantoic acid or pantonine; to /S-alanine; to
pantothenic acid only; and to a mixture of pantoic
acid and /5-alanine. The pantoic-less mutant feeds the
/3-alanine-less mutant, but not vice versa.

In both these instances of syntrophism, the cause of
the accumulation of the intermediate is not absence of

the enzyme concerned with its further conversion, but
inability of that enzyme to perform the conversion
in the absence of the second conjueant.

(5) Release of physiological brake on a synthetic
process. One of the most interesting observations to
turn up is mutual syntrophism between certain
tyrosine-less and phenylalanine-less mutants. Paper
chromatography confirmed the inference that the
block in phenylalanine synthesis resulted in excretion
of tyrosine itself, rather than a precursor, while a
block in tyrosine synthesis caused excretion of phenyl-
alanine. Feeding of a tyrosine-less mutant by culture
filtrates of a phenylalanine-less mutant has also been
observed with another strain of E. coli 1 .

From the presence of mutants with a double re-
quirement for these two compounds, as well as others
with triple and quadruple requirements for these plus
other aromatic compounds, all resulting from single
mutations, it had been inferred that phenylalanine
and tyrosine arose from a common precursor (Fig. 4).
A possible explanation for the mutual syntrophism
therefore appeared to be the diversion of this precursor
in one rather than the normal two directions, with
resultant excretion of the excess of phenylalanine or
tyrosine. Further study, however, showed this explan-
ation to be inadequate, as the amount of phenylalanine
(or tyrosine) excreted was much larger than the amount
of tyrosine (or phenylalanine) consumed. Since the
requirements of mutants of E. coli for these two
compounds are of the same order of magnitude, simple
diversion could not account for the large production.

An alternative explanation has been developed,
stimulated largely by the interesting work of Bonner 2 .
He showed that a single block in isoleucine synthesis,
between the a-keto and the amino acid, accounted for
the double requirement of a Neurospora mutant for
isoleucine and valine. Apparently the accumulated
isoleucine precursor competes as a structural analogue
with the corresponding compound in the valine-chain,
causing a requirement for this amino acid as well:

\ — ► ketoisoleucine »- isoleucine

Recently Adelberg 3 has found that the compound
accumulated by this mutant is not ketoisoleucine, but
is rather the a-/?-dihydroxy acid. The principle of
internal inhibition by a normal metabolite, however
remains unchanged.

This evidence of inhibition of a normal reaction by
increased concentration of a normal metabolite has
seemed to us to point to the possibility of a general
mechanism of integration of various parallel sequences

Biol. CI
g, perso

46

tionallv DehYimt B.i< t.-nal Mutants

of biosynthesis. If the accumulation of normal meta-
bolite A in excessive concentration completely blocks
a certain enzymic reaction in the production of com-
pound B, then the normal concentration of metabolite
A might exert a governing effect on that reaction,
while a complete absence of metabolite A, resulting
from a genetic block, might permit excessive synthesis
of compound B, out of proportion to the remainder of
the metabolites being synthesized, with resulting ex-
cretion of compound B.

To apply this concept to the present problem, we
would postulate a tyrosine precursor which interferes
with phenylalanine synthesis, and vice versa. The ex-
cretion of phenylalanine would therefore be due to a
block early enough in the synthetic chain of tyrosine to
cause absence of that tyrosine precursor which normally
governs the rate of phenylalanine synthesis, and the
same consideration would apply to the phenylalanine
mutant which secretes tyrosine. On the other hand,
there is one tyrosine mutant which fails to feed phenyl-
alanine; its block would occur after the governing
intermediate (E in Fig. 4). This scheme is strongly
supported by the fact that the excretion of phenyl-
alanine by the tyrosine mutant is prevented in the
presence of an excessive concentration of tyrosine ; the
same is true of the phenylalanine mutant whose ex-
cretion of tyrosine is inhibited by an excess of phenyl-
alanine or phenylpyruvic acid, which can be sub-
stituted for phenylalanine as a growth factor for this
mutant. Presumably the excess of the amino acid
causes reversal of the normal processes of synthesis,
restoring from without the governing compound whose
synthesis is genetically blocked.

This scheme for explaining the output of phenyl-
alanine and tyrosine is on speculative grounds, since
the intermediates in the synthesis of these amino acids
(except for phenylpyruvic acid) are as yet unknown.
The concept of normal physiological interaction among
separate biosynthetic paths, however, is further
strengthened by returning to the better documented
isoleucine-plus-valine case. We have isolated a mutant
of this type with E. coli whose mechanism appears to
be identical with that of the similar Neurospora
mutant. In addition, another E. coli mutant has been
isolated which responds equally well to isoleucine, its

a-keto acid, or y.-amino butyric acid. The double-
requiring mutant, which cannot use ketoisoleucine, it
known to accumulate an inhibitor of valine synthesis;
the single-requiring mutant, which can use ketoiso-
leucine, and which is fed by the double, must be block-
ed earlier ; it would therefore follow, from the reasoning
outlined above, that the single-requiring mutant
should not form the isoleucine intermediate that gov-
erns valine synthesis, and hence might be expected to
excrete valine. On testing, this mutant was indeed
found to feed a valine mutant heavily, and paper
chromatography on the culture filtrate showed a
dense spot corresponding to valine. While in this case,
as in those of tyrosine and phenylalanine, neither
microbiological assay nor paper chromatography alone
furnishes complete proof that the compound excreted
is the amino acid rather than a related compound, it
seems unlikely that the non-specificitv of these two
techniques would overlap to give a false answer when
used together.

In several instances, therefore, interference with
synthesis of one amino acid leads to excessive synthesis
of another. This evidence encourages us to feel that
the mutants lend themselves not only to determining
the normal steps in biosynthesis, but also to unravelling
the mechanisms of integration by which the normal
cell determines that its metabolites should be synthes-
ized in proper proportions and not wasted. Recent ex-
periments have shown that this economy of the
organism is upset in diverse ways by a single mutation.
A tyrosine mutant, for example, not only excretes
phenylalanine, but also, to a smaller degree, feeds the
mutants requiring tryptophan, lysine, valine, and
leucine. This type of study has just begun, but it is
already clear that the interrelations between various
biosynthetic paths are complex (Fig. 4).

Precursors of aromatic compounds

The mechanism of biosynthesis of aromatic com-
pounds has long been a subject of speculation. The
availability of mutants with multiple aromatic re-
quirements (Fig. 4) offers an opportunity to obtain
definite information on this matter. A wide variety of
aromatic compounds, with single and multiple sub-

s»

*S

phenylpyruvic acid — > phenylal,

|— ► tryptophan

99

]'.. D. Davis: Studies r.n Nutritionally Deficient Bacterial Muta

47

stitutions of carboxyl, amino, and phenolic groups,
were tested for their capacity to substitute for the
multiple requirements of the aromatic-less mutants.
None of these was effective. Similarly cyclohexane
carboxylic acid, cyclohexanol, and inositol were not
utilized. Shikimic acid, however (suggested and fur-
nished by R. Stanier) was used by aromatic-less
mutants with quadruple, but not with triple or double
requirements.

C
HC CH 2

HOHC CHOH

\/

CHOH

shikimic acid

It therefore appears that the precursor of the benzene
ring is at least partly saturated. These mutants do not
use quinic acid, in which the double bond of shikimic
acid is hydrated.

It will be noted that Fig. 4 contains an unknown
"end-product", presumably aromatic, labelled Y. The
evidence for the existence of this postulated compound
is as follows, (a) The aromatic-less mutants with
quadruple requirements require more PABA than does
the PABA mutant, and their growth is accelerated by
high concentrations of PABA, suggesting that the
synthesis of PABA may be reversed, to yield a precursor
common to PABA and an unknown compound, (b)
Growth of the aromatic-less mutants on large amounts
of the four aromatic compounds is slow; similarly,
growth on shikimic acid alone is slow. But shikimic
acid plus the three amino acids yields much faster
growth, practically as fast as wild-type. These results
suggest that shikimic acid may occupy position B,
rather than A, serving rapidly as a precursor of Y plus
PABA (for which PABA alone serves only slowly), but
serving only slowly, by reversal of the normal process,
as precursor for the amino acids.

The scheme of Fig. 4 must be considered quite
tentative. Certain phenomena are difficult to explain,
such as the fact that mutants with a double aromatic
requirement (tyrosine plus phenylalanine) heavily
feed the quadruples, while the triple neither feeds the
quadruples nor is fed by the doubles. One must there-
fore consider more seriously the possibility, always
theoretically present, that some of these multiple re-
quirements may depend on internal inhibition, as with
isoleucine and valine, rather than on block at an early
stage of synthesis.

Partial back-mutants

Almost all the mutants are detectably unstable;
that is to say, spontaneous reversions to nutritional

independence (prototrophs) occur with a high enough
frequency (10- 7 -10 -8 ) to give rise after several days to
a few large colonies of back-mutants in a streak on a
medium which is sufficiently enriched to permit limited
growth (cf. Fig. 2). For this reason practically all of
our experiments, even on quantitative response, are
carried out on solid media ; in tubes of liquid media the
greater precision of measurement by turbidimetry is
accompanied by greater difficulty in distinguishing
growth of back-mutants from that of parent mutants.
Maintaining transfer cultures of mutants in liquid
media, however, has caused no difficulty provided
selection of back-mutants is avoided by the presence
of an excess of the nutrient requirement.

Our study of back-mutants had a casual origin which
is pointed out here since it illustrates the possibility,
based on the uniform colonial growth of bacteria, of
encountering interesting phenomena by simple observa-
tions on plates. Two presumptive back-mutant colonies
from a PABA-less mutant were isolated and streaked
on minimal medium, along with wild-type, to verify
their nutritional independence. One of these grew as
rapidly as wild-type ; the other was by chance observed
at an early time (18 hrs.) to form slightly smaller
colonies than wild-type. Further study showed that
the slightly slow prototroph grew as rapidly as wild-
type in a medium supplemented with PABA; it ap-
parently had recovered the capacity to synthesize
PABA, but not rapidly enough to permit optimal
growth.

Since PABA is one of the factors excreted by wild-
type in large amounts, this hypothesis was easily
tested. The two PABA back-mutants were compared
with wild-type for their output of PABA on minimal
medium, by pouring plates of minimal medium con-
taining few cells of the prototroph and many PABA-
requiring cells. (This technique is more sensitive to
slight differences in syntrophism than is the technique
of adjacent streaks.) As expected, the wild-type
colonies were surrounded by an extensive halo of
satellites, while the slow prototroph had none. This
fact confirmed the earlier conclusion that its growth
rate was limited by its rate of synthesis of PABA ; it
had none to spare. Unexpectedly, however, the other
back-mutant, previously indistinguishable from wild-
type, was surrounded by a smaller halo of satellites
than was wild-type. Apparently its recovered capacity
to synthesize PABA was not as great as that of wild-
type.

Following this a dozen different back-mutant
strains, spontaneous and ultraviolet induced, were
isolated from a PABA mutant ; among these 6 different
rates of growth in the absence of PABA were recogniz-
ed. In addition, all were more susceptible than wild-
type to sulfonamide inhibition, which is not surprising
since all synthesized PABA at a lower rate than wild-
type. Some of the slower strains had their growth rate

48

B. D. Davis: Studies on Nutritionally Deficient Bacterial Mutants Isolated by Means of Penicillin [Experientia Vol. VI/2]

restored to optimal by the addition of PABA; others
did not. Similar studies were carried out with mutants
requiring 6 amino acids, purines, biotin, and panto-
thenic acid. In all cases but one amino acid, a variety
of degrees of back-mutation were observed; in only
a fraction of the strains were slow growth rates restored
to optimal by addition of the parent's growth require-
ment.

In the absence of the required genetic techniques it
has not been possible to demonstrate whether these
various degrees of restoration of a nutritional deficiency
represent quantitative alleles of the same gene, or
mutations in other genes which modify the deficiency.
In the cases where the back-mutant is slow either with
or without the growth factor, an allelic change seems
much less likely than does a mutation of another gene
which imposes a rate-limiting alteration of metabolism
at the same time that it restores the deficiency. In any
event, it is clear that back-mutants, which have lent
themselves to quantitative genetic studies, represent a
genetically very heterogeneous class. Indeed, it is not
evident that any of the back-mutants are truly
identical with wild-type. Perhaps they should be
understood as "backward" mutants rather than true
back-mutants or reversions, having mutated in the
direction, but not necessarily to the precise position,
of wild-type. One study of back-mutants of Neuro-
spora 1 has failed to reveal such frequent and varied
partial restorations of a deficiency. Incidentally, in this
connection, it should be noted that mutations from
wild-type to partial deficiencies of various factors are
frequent, with Neurospora as well as with bacteria ;
they are not often mentioned in the literature since
they have not seemed to lend themselves to genetic or
biochemical analysis as well as mutants with absolute
requirements.

Reversion of multiple requirements

Another genetic question, more essential for bio-
chemical investigation, is whether certain multiple
requirements arise from a single mutation and hence
presumably from a change in a single enzyme. While
we have not tried to solve this problem directly by
recombination techniques, it has been possible to
circumvent it by use of back-mutations 2 . In several
instances it has been easy to obtain spontaneous
prototrophs, with no growth requirements, from
mutants with multiple requirements (e. g. quadruple
and double aromatic), and has not been possible to
obtain split reversions, with loss of only part of the
requirements. From these results it is inferred that a
single mutation has occurred and then reverted. In
another instance, requiring histidine and the three

1 N. H. Giles, Jr., and E. Z. Lederberg, Ainer. J. Bot. 35, 150
(1948).

8 R. R. Roepke, quoted by E. L. Tatum, Cold Spring Harbor
Symp. Quant. Biol. 11, 278 (1946).

aromatic amino acids, it has been possible, in appro-
priate media (limited histidine, excess of the other re-
quirements) to lose the histidine requirement alone,
but it has not been possible to lose all four require-
ments in one step. It is therefore concluded that this
mutant had been altered in two separate steps. Non-
allelic reversions from mutant to apparent wild-type
in higher organisms have been described in which a
mutation of a second gene at a different locus (modifier
gene) restores the normal condition. In the absence of
a test for allelism it cannot be decided whether our
reverse mutations involve a change of the same gene
or of another one. It is conceivable that a mutation of
a modifier gene might suppress two independent
growth requirements. But whatever the genetic
mechanism of the reversions may be, one would expect
with a double mutant to recover the single back-
mutants as well as the prototrophs. Complete failure
to isolate the single reversions, together with isolation
of the total reversions, is therefore considered excellent
evidence for a single mutation producing a multiple
requirement. On the other hand, it should be pointed
out that the recovery of split reversions is not conclusive
evidence for the presence of independent mutations,
unless it is accompanied by failure to isolate any total
reversions. With the isoleucine-plus-valine mutant, for
example, one back-mutant was obtained with a relative
requirement for valine (slow growth without it), but
no requirement for isoleucine. Since many completely
prototrophic back-mutants were also obtained from
this strain, a probable mechanism for the split reversion
would be incomplete restoration of the isoleucine
enzyme, leaving enough residual accumulation of the
intermediate to inhibit valine synthesis.

This stress on the importance of negative as well as
positive results is warranted only because of the great
efficiency, essentially 100%, with which reversions to
nutritional independence can be isolated from huge
bacterial populations.

Delayed phenotypic expression of back-mutation

The number of spontaneous back-mutant (proto-
troph) colonies observed on a minimal medium plate is
proportional to the size of the inoculated population.
In a medium with limited enrichment, in contrast, the
inoculum grows until the population reaches a size
which is limited by the amount of enrichment; the
mutants observed are therefore largely "plate-
mutants" (i.e., those arising during generations
occurring on the plate), and their number is, as a first
approximation, a function of the enrichment rather
than the inoculum size. Following ultraviolet irradia-
tion, however, an entirely different situation was en-
countered.

Ultraviolet irradiation increases the frequency of
back-mutants among the survivors by as much as many

IOI

4 f >

thousandfold. Under these circumstances, the number
of plate-mutants in .1 heavily inoculated plate with
limited enrichment should I"- only a negligibli frai tion
of the number of mutants inoculated. The number of
ultraviolet induced prototroph colonies observed would
therefore be expected to be relatively independent of
the enrichment. To our surprise, with certain mutants
the dependence was found to be extreme. With a
washed inoculum of a mutant, almost no ultraviolet-
induced prototrophs could be detected unless the
medium was enriched with a trace of the factor re-
quired by tin parent strain. To look at one instance:
an inoculum of 10 8 cells of a tryptophan-requiring
mutant yielded 1 to 3 visible back-mutant colonies
after three days on minimal agar, and 1 to 6 colonies
on agar slightly enriched with 0-01 to 0-5 y ml of
tryptophan. The suspension was irradiated until about
2 and 50% of the cells survived, and the same volumes
were inoculated. On minimal agar, to 2o colonies
developed; on slightly enriched agar, 2(>(> to 600! Of
these, 25 were pieked at random, ranging from the
large to the microscopic. All were prototrophs, growing
on minimal agar, though many of the smallest grew
very slowly. Qualitatively similar observations have
been made with several amino acid, purine, and vita-
min requiring mutants.

It appears that a certain amount of growth by the
parent strains is necessary to permit the ultraviolet-
induced back-mutants to get started. It is known,
however, that small inocula of various bacteria often
fail to initiate growth in a medium which is adequate
for larger inocula; the small inocula are presumably
unable to accumulate enough C0 2 or other essential
metabolites. To rule out the possibility that the
growth of the parent strain was promoting tin- ap-
pearance of back-mutant colonies by some such non-
specific mechanism, variously enriched plates were
inoculated with an irradiated strain, an unirradiated
strain with a different requirement, and a mixture of
the two. Prototrophs appeared in huge numbers only
from the irradiated strain, and when the medium was
supplemented with its growth fat tor rather than that
of the companion unirradiated strain.

We are therefore brought back to the consideration
of the phenomic lag, introduced into this work at the
outset by the initial failures of the penii illin method. It
now appears that a lag occurs not only in exhausting
preroutational enzymes, but also in building up new
enzymes which the mutated cell is ( apableof construct-
ing; to build up the new enzyme the cell requires a
complete set of building blocks, including the product
of the previously deficient enzyme. In other words, to
1. nt the cycle, the pump must first be pruned. This
phenomenon stands in contrast to the behavior of
certain adaptive fermentative enzymes, which are re-
ported to be foi tned by bacteria or yeasts in a nitrogen
free medium, in which no net growth can take place.

I'se 0/ mutants in studying mechanisms <>/ bacterial
inhibition

D-Serine exerts a marked inhibitory effect on the
growth of our wild-type strain of /:. coli; many amino
acids are able to antagonize the inhibition, but aspartic
acid, in contrast, enhances the effect, although by itself
it has no influence on growth 1 . Werner Maas, further
studying this problem in this laboratory, found that
the inhibition is oven ome by extremely low 1 oncentra-
tions of pantothenic acid, and by somewhat higher
concentrations of /^-alanine. The antagonistic action
was apparently non-competitive with pantothenate
and competitive with ^-alanine. The results suggested
that /^-alanine is the substrate and pantothenate the
product of the inhibited reaction. This type of inhibi-
tion analysis lias been widely used to determine
whether a given antagonist serves .is substrate, pre-
cursor, or product of the inhibited reaction. For several
reasons, however, which are dis< ussed elsewhere-, it
seems difficult to draw rigorous conclusions as to the
site of the inhibition, especially in those cases where
only a narrow range of concentration of the inhibitor is
possible before other reactions become affected. With
appropriate mutants, on the other hand, it is possible
to dissect out the system under investigation ami
produce more direct evidence of the site of inhibition.
In addition, mutants show whet her an apparent product
of the inhibited reaction, acting non-competitively on
wild-type, has done so in "physiological" concentra-
tions.

Mutants blocked at various stages in the synthesis
of pantothenic acid were therefore isolated (Fig. 3). A
mutant blocked in the synthesis of ^-alanine showed a
competitive relation between /^-alanine and D-serine;
a mutant blocked in the synthesis of pantoic acid
showed inhibition by D-serine when grown on pantoic
acid, but none when grown on pantothenic acid; and a
mutant blocked in the formation of pantothenic acid
grew in proportion to the amount of pantothenic acid
present, regardless of the presence or absence of D-
serine. It has therefore 1 been demonstrated, more con-
clusively than would be possible with wild-type alone,
that D-serine (or a product of it) interferes with the
conversion of ^-alanine to pantothenic acid'-. Similarly,
salicylic acid, which is known to be antagonized bv
pantothenic acid 3 , has been found 4 to interfere' with
the synthesis rather than the utilization of pantoic
acid. Furthermore, several of the amino acids which
antagonize D-serine inhibition of wild-type were tested
against D-serine inhibition of a mutant unable to
synthesize /5-alanine. No antagonism of the inhibition

102

was observed. This result indicates that these amino
acids antagonize D-serine only indirectly, acting via
the final common path of yS-alanine synthesis. Finally,
serine-resistant and salicylate-resistant mutants have
been isolated and arc under investigation. We feel that
the pitfalls in interpreting the mode of action of anti-
bacterial agents can sometimes be avoided by the
combination of biochemical studies with the use of
appropriate mutants.

In connection with the studies on pantothenic acid.
certain pantothenate-requiring mutants have been
observed to respond to either pantoic acid or to
pantonine (the sc-amino analogue of the x-hydroxy
acid, pantoic acid); others respond to pantoic acid
only (Fig. 3). It may be inferred that pantonine is
either a normal precursor of pantoic acid, or is con-
vertible to such a precursor.

A number of the problems described above are as
yet only partly solved; yet so closely intertwined are
the various metabolic sequences that each step toward
a solution has revealed unexpected new problems.
Herein lies much of the fascination of this field. It has
been the main purpose of the present paper to illustrate
the ease with which a variety of problems can be re-
vealed and solved, up to a point, by simple methods:
the isolation of desired bacterial mutants by the
penicillin method, followed by direct observation of
their behavior, especially with respect to syntrophism
and back-mutation, on solid media. Some of the studies
described in this paper will be published elsewhere in
more detail.

It is a pleasure to acknowledge the expert technical assistance of
Mrs. Harlean Cort and Mrs. Elizabeth Mingioi.i. We are grateful
to W". Shin and S. Moore, R. D. Hotchkiss, V. duVigni u d,
1). Bonner, X. 11. Horowitz, K. Stanier, to T. Jukes of Lederle
Laboratories, and to K. Folkers, M. Tishter, and E. E. How i ,.1
Merck and Co. for generous gifts of chemicals.

Zusammenfassung

In dieser Arbeit wird cine Methode beschrieben, die
es gestattet, in groBerer Zalil Stoffwechsel-Minus-
mutanten von Baktcrien zu isolieren. Sie beruht auf
der bemerkenswerten Eigenschaft des Penicillins, aus-

schlieBlieh wachsendo und si. h vermehrende Baktcrien
abzutoten. Wird demzufolge cine groBe Bakterien-
kultur, die verschiedene, durch vorangegangene (Jltra-
violettbestrahlung lierbeigefiihrte Mutanten entha.lt, in
einem eben geniigenden Mahrmedium der Wirkung von
Penicillin ausgesetzt, so wird die Mehrzahl der Kultur,
die aus den normalen Bakterien des Stammes besteht,
abgetotet, wahrend die Minusmutanten, die nicht zu
wachsen vermogen, selektiv iiberleben.

Mit Hilfe dieser Technik wurden Mutanten von E.
coli nut individuell verschiedenen Anspriichen auf
nahezu samtliche Aminosauren, Vitamine, Purine und
Pyrimidine isoliert. Einzelne Mutanten nut alter-
uativen Oder multiplen Stoffwechselbedurfm>seii er-
laubten Einblicke in den Gang -von Biosynthesen. So
wachsen eine Anzahl von «ringfreien» Mutanten, die
gleichzeitig Tyrosin, Phenylalanin, Tryptophan und
Paraaminobenzoesaure benotigen, auch in alleinigei
Gegenwart von Shikimisaure, einem teihveise dehj
drierten Derivat der Cyclohexancarbonsaure. Dies laBt
vermuten, da!3 im Organismus aromatische Verbin-
dungen aus vvenigstens teihveise gesattigten Ringver-
bindungen entstehen konnen. Ferner ist bekannt, dab
der Gan^; einer Biosynthese durch die Isolierung von
Zwischenprodukten analysiert werden kann, die sou
gewissen Mutanten angestapelt werden. Derartige An
sammlungen lassen sich technisch mit Hilfe von Bak-
terien leicht nachweisen, vvenn verschiedene Mutanten,
die sich gegenseitig ernahren, nebeneinander ausge-
strichen werden. Nicht nur werden so Stoffwechsel-
vprstufen (precursors) angestapelt; in gewissen Fallen
fuhrt das Unvermpgen, eine bestimmte Aminosaure
(■/-. B. Tyrosin, Phenylalanin, Isoleucin) zu syntheti-
sieren. zu Anhaufungen anderer Aminosauren. Diese
Beobachtung leitet zur Annahme, daB normale Zwi-
schenstufen einer Verbindung einen regulierenden Ein-
flufi auf die Synthese anderer Stoffe ausiiben.

Ferner lassen sich diese Mutanten gut verwenden,
um die Wirkungsweise bestimmter antibakterieller
Wirkstoffe endgultig festzulegen. Es konnte nachge-
wiesen werden, daB die Umwandlung von /J-Alanin in
Pantothensaure durch o-Serin verhindert wird, wah-
rend Salicylsaure auf die Bildung d(^r Pantoinsaure
storend einwirkt.

Durch Ultraviolettbestrahlung hervorgerufene Mu-
tanten, sowohl Stoff\vechsel-Minus\arianten als Riick-
mutierungen zur Norm, brauchen eine bestimmte Zeit,
bis sie phanotypisch erkennbar sind, was darauf
schlieBen laBt, daB sich die Fermentzusammensetzung
erst nach einer Einstellungsphase der neuen Gen-
zusammensetzung angleicht (phenomic lag).

103

THE DEMONSTRATION OF NON-SPECIFIC COM-
PONENTS IN SALMONELLA PARATYPHI A
BY INDUCED VARIATION 1

D. W. BRUNER and P. R. EDWARDS

Department of Animal Pathology, Kentucky Agricultural Experiment Station,
Lexington, Kentucky

Received for publication February 21, 1941

Phases induced by growth in immune serum have been demon-
strated in a number of enteric bacilli. Kauffmann (1936) found
a second phase in Eberthella typhosa after the organism was cul-
tivated in agglutinating serum derived from the Muenchen type.
While Kauffmann found this variant to be irreversible, Edwards
and Bruner (1939) reverted similar forms to the original phase by
cultivation in serum derived from the variants. Induced phases
were described by Kauffmann and Tesdal (1937) in the Schleiss-
heim type and by Gard (1938) in Salmonella abortus-canis. Both
the latter types are monophasic under ordinary conditions of
culture. Gnosspelius (1939) induced variants with altered flagel-
lar antigens in the naturally diphasic types Stanley and Hvitting-
foss. These forms also were irreversible even when cultivated
in homologous immune serums. It is significant that all the
phases described above were "artificial" phases; that is, they pos-
sessed no antigenic relationships to the naturally occurring anti-
gens of the genus Salmonella.

The cultivation of Salmonella strains in immune serum does
not always lead to the isolation of artificial phases, as witness
the results obtained in the study of the "totally and permanently
non-specific" Salmonella types. Thus, through cultivation in
immune serum, Scott (1926) isolated a specific phase from a non-
specific culture of the Thompson type and Gard isolated specific

1 The investigation reported in this paper is in connection with a project of
the Kentucky Agricultural Experiment Station and is published by permission
of the Director.

467

[Reprinted by permission of The Williams & Wilkins Company from Journal of Bacteri-
ology 42: (4) 467-478, October, 1941]

IO4

468 D. W. BRUNER AND P. R. EDWARDS

phases identical with that of Salmonella choUrae-suis from cul-
tures of S. cholerae-suis var. kunzendorf. Bruner and Edwards
(1939) isolated specific phases from Salmonella typhi-murium var.
binns, Salmonella typhi-suis var. voldagsen, S. cholerae-suis var.
kunzendorf and the Berlin and Puerto Rico types by cultivation
in immune serums. In each instance the specific phases so iso-
lated were identical with those that occur naturally in the diphasic
counterparts of the variants. Edwards and Bruner (1939a) work-
ing with the naturally monophasic Salmonella abortus-equi iso-
lated two additional phases by cultivation of the organism in
various immune serums. One of these was an "artificial" phase
closely related to the induced phase of the Schleissheim type
described by Kauffman and Tesdal. The second induced phase,
on the contrary, was the antigen a of the Kauffmann-White
classification and was identical with the flocculating antigen of
Salmonella paratyphi A.

The purpose of the present work is to describe the isolation of
phases closely related to the naturally occurring non-specific
phases of the genus Salmonella from S. paratyphi A. It is realized
that the serology of the Salmonellas is already sufficiently com-
plicated and that work such as that presented here may seem on
first thought only to confuse the classification of the bacilli.
However, the demonstration of inapparent components of the
bacilli adds to our knowledge of the genetic relationships and
hereditary tendencies in the genus.

MATERIALS AND METHODS

Sixteen cultures of S. paratyphi A were used in the study.
The sources and designations of the strains were as follows:

1015, Marta, HA1, HA6, Durazzo— From Dr. F. Kauffmann,
International Salmonella Center, Copenhagen.

38250, 37407— From Dr. Ruth Gilbert, New York State Depart-
ment of Health.

Fried, Boyd, 17, W, WB39— From Dr. Ralph Muckenfuss, New
York City Department of Health.

228, HR, GV— From Dr. L. F. Rettger, Yale University.

Cal. 49— From Dr. W. R. Hinshaw. Stock culture from Medical
School, University of California.

105

NON-SPECIFIC COMPONENTS IN S. PARATYPHI A 469

All the cultures possessed the biochemical and serological prop-
erties generally attributed to S. paratyphi A with the following
exceptions: The strain Durazzo lacked antigen I of the Kauff-
mann-White classification and produced large amounts of hydro-
gen sulphide. In these respects it conformed to the description
of Kauffmann (1937). The remainder of the strains produced
smaller amounts of hydrogen sulphide, as evidenced by the black-
ening of lead acetate papers suspended over cultures in 2 per cent
Bacto-peptone water. In this connection it should be stated that
the observations of the writers confirm the conclusions of Hunter
and Crecelius (1938) that differences observed in hydrogen sul-
phide production by different members of the genus Salmonella
are quantitative and not qualitative. It is probable that all
Salmonella strains produce hydrogen sulphide and that the results
obtained in testing for its production depend upon the methods
employed for the detection of the substance.

The methods employed to induce variation were largely the
same as those used by Edwards and Bruner (1939a) in the study
of S. abortus-equi. The organisms were grown in semi-solid agar
to which was added sufficient agglutinating serum to immobilize
the bacilli. The medium was inoculated by stabbing at one side
of the tube. Outgrowths from the line of stab occurred only
when the flagellar antigens were altered. The spreading growth
was transferred successively in tubes of the same medium until
the serological reactions indicated that the induced phase was
pure. The culture was then plated and isolations made from
single colonies for further study. Serum derived from phase 1
of the Bispebjerg type was used to induce variations in the normal
phase of the cultures. As additional phases were isolated agglu-
tinating serums were prepared from them. These were used in
the study of induced phases and in forcing the bacilli to produce
further alterations in the flagellar antigens. Serum derived from
> S. cholerae-suis var. kunzendorf was used in the reversion of non-
specific phases. In many instances, it was found necessary to
use combinations of serums to force variation in the desired direc-
tion or to revert induced phases to the original phase. When the
serums contained or H agglutinins which might interfere with

106

470 D. W. BRUNER AND P. R. EDWARDS

the migration of the organisms, such agglutinins were removed
by absorption with appropriate bacilli. The serums were pre-
served by the addition of small amounts of chloroform. The
chloroform effectively sterilized the serums and did not affect
the growth or motility of the bacilli.

RESULTS

It should be emphasized that the changes described below in-
volved only the H or flagellar antigens; that is, the antigens which
are affected in the phase variation of normally diphasic types.
The heat-stable antigens remained unchanged throughout the
experiments. On the whole the cultures were quite stable and it
was more difficult to induce variation in S. paratyphi A than in
Eberthella typhosa, S. abortus-equi or the "totally and permanently"
non-specific types. In some instances it was necessary to trans-
fer the organisms repeatedly in the presence of agglutinating
serum before variation was observed.

From cultures GV and W no induced phases were isolated.
Strains 17 and 37407 each yielded two induced phases (phases 3
and 4) neither of which was closely related to any of the naturally
occurring antigens of the genus. All the remaining cultures
yielded three induced phases. The normal phase of the bacilli
(antigen a of the Kauffmann-White classification) was designated
as phase 1. The induced antigens were denoted as phases 2, 3
and 4. Phase 2 is closely related to the non-specific phases of
the diphasic Salmonella types. Phases 3 and 4 are "artificial"
phases which closely resemble none of the natural antigens. The
relationships of the phases to each other and to the antigens of
other species are given in table 1. Only one culture of S. para-
typhi A is included in the table since the phases isolated from
other strains reacted similarly. Phase 2 is agglutinated in high
dilution by Kunzendorf and non-specific Sendai serums. Like-
wise serum derived from phase 2 agglutinates Kunzendorf and
the non-specific phase of Sendai to the titre of the serum. Phase
3 is slightly related to the normal phase 1, but has little relation-
ship to the other phases or to the natural antigens of the genus.
It is agglutinated in low dilution by all the serum of all types that

107

NON-SPECIFIC COMPONENTS IN S. PARATYPHI A

471

contain antigen a. This agglutination probably represents a
slight residue of the normal flagellar antigens of the species.
Phase 4 shows evidence of a slight serological relationship to
phase 3 and to the non-specific phases of other types, otherwise
it is unrelated to the known antigens of the Salmonella.

The phases were further examined by agglutinin absorption.
The results of these tests confirmed the agglutination tests and
need not be given in detail. When absorbed serums were used
the cross reactions between the four phases were no longer evident.

Agglutination tests

TABLE 1

with phases of

S. paratyphi A

SERUMS

ANTIGENS

Para-
typhi
A 228
Phase 1

Bispe-
bjerg
Phase 1

Para-
typhi
A 228
Phase 2

. Para-
typhi
A 228

Phase 3

Para-
typhi

A 228
Phase 4

Sendai
Phase 2

Kunzen-
dorf

Paratyphi A 228:

Phase 1

20,000

500

20,000

20,000

40,000

500

40,000

40,000

40,000

1,000

40,000
40,000

2,000

10,000

1,000

1,000

1,000

200

20,000

500

500

10,000

10,000

10,000

Phase 2

Phase 3

Phase 4

Abortus equi — Phase 1 —

Bispebjerg — Phase 1

Sendai — Phase 2

Kunzendorf

20,000

20,000
20,000

Figures indicate highest dilution at which agglutination occurred.
"0" indicates no agglutination at dilution of 1 to 200.

Furthermore, the absorption of serum derived from one phase by
organisms of another did not result in an appreciable reduction
of the titre of the serum for the homologous phase. Absorption
of serum derived from phase 3 of one strain by phase 3 organisms
of a second strain resulted in a complete removal of agglutinins
from the serum. This was also true of phase 4. The non-specific
components (phase 2) derived from all the cultures were closely
related but not identical. This will be discussed below.

While the variants isolated from the various cultures of S.
paratyphi A were similar, the variational tendencies of the cultures
differed. When the organisms were inoculated into semi-solid

:o8

472

D. W. BRUNER AND P. R. EDWARDS

agar which contained serum derived from phase 1 of the Bispeb-
jerg type, culture HR yielded phase 2, 228 yielded a mixture of
phases 2 and 4, while the remainder of the organisms yielded phase
3 or a mixture of phases 3 and 4. For this reason mixtures of
serums were used to force variation in the desired direction.
The variations induced in one culture, strain 228, are given in

Phase

1

B+4 > Phase 2

*+•* > Phase 1

B±* >. Phase 3

_I » Phase l — ^ * Phase 2

^ Phase 3

Kt3 > P* 1 *" 1

B+i.

Fig. 1. Changes Induced in S. paratyphi A 228
Symbols on arrows indicate what serums were added to medium.
B, serum derived from phase 1 of Bispebjerg type.
K, serum derived from S. cholerae suis var. kunzendorf.

3, serum derived from S. paratyphi A, phase 3. The serum was absorbed with
ohase 1 before use. „ . , , ...

4, serum derived from S. paratyphi A, phase 4. The serum was absorbed with
phase 3 before use.

figure 1. The changes produced in other cultures differed slightly
but they were all quite similar.

When once isolated the induced phases of S. paratyphi A are
quite stable. No changes in the phases were noted during the
two years they were maintained in the laboratory. In this respect
they resemble the induced phases of E. typhosa and S. abortus-equi
and differ from the normal phases of the diphasic Salmonella

109

NON-SPECIFIC COMPONENTS IN S. PARATYPHI A

473

types. The latter display phase variation under normal condi-
tions of culture. As shown in figure 1, it is possible to produce
changes in the induced phases of S. paratyphi A by cultivation in
appropriate immune serums. The reversibility of the phases is
illustrated in figure 2. Phases 1, 2 and 3 were completely inter-
changeable and reversible. The changes illustrated in these
phases were accomplished repeatedly with cultures isolated from
single colonies. Phase 4, on the contrary, was more stable. It
was readily converted to phase 2 but not to phase 1 or phase 3.
Efforts to convert phase 4 to phase 1 or phase 3 resulted in the

Pbaae t

Phase 3

Fig. 2. Reversibility of Phases of S. paratyphi A 228
Arrows indicate direction in which variation occurred

production of a series of ill-defined, serologically-related variants.
This observation confirms the work of Gnosspelius (1939), who
concluded that through proper manipulation it was possible to
isolate an endless number of antigenic components from
Salmonella strains.

It is known that "artificial" phases that display little or no
relationships to the normal antigens of the genus can be isolated
from a number of Salmonella types by induced variation. The
isolation of components closely related to the naturally occurring
antigens is much more unusual. The isolation of non-specific

474

D. W. BRUNER AND P. R. EDWARDS

components from a normally monophasic-specific species has
not been reported previously. For that reason the character-
istics of the non-specific components (phase 2) of S. paratyphi A
are given in more detail than are the characteristics of the "arti-
ficial" phases 3 and 4. The agglutinative relationships between
the non-specific components of S. paratyphi A and the non-specific
phases of diphasic types are given in table 2.

It is apparent that the non-specific phases of S. paratyphi A
are more closely related to the 1,5... phases than to the 1,2...,
1, 6. . . or 1, 7. . . phases. The organisms were tested with ab-

TABLE 2

Agglutination of non-specific phases of S. paratyphi A by non-specific serums of

other types

Paratyphi B (1, 2. .
Newport (1, 2. ..)..
Kunzendorf (1, 5 .
Paratyphi A 228. . .
Paratyphi A WB39.
Sendai (1, 5...)....
Anatum (1, 6. . .). .
Nyborg (1,7...)...

Para-
typhi B
(1,2.)

10,000
5,000
2,000
2,000
2,000
2,000
2,000
5,000

Newport
(1,2...)

10,000
40,000
10,000
2,000
2,000
5,000
2,000
5,000

Kunzen-
dorf
(1,5...)

5,000
5,000
20,000
10,000
10,000
10,000
5,000
2,000

Para-
typhi
A 228
Phase 2

1,000
1,000
40,000
40,000
40,000
40,000
10,000
200

Para-
typhi
AWB39
Phase 2

500

1,000
40,000
20,000
40,000
20,000

5,000
200

Anatum
(1.6...)

2,000
2,000
5,000
5,000
5,000
5,000
10,000
2,000

Nyborg
(1,7...)

2,000
2,000
2,000
1,000
1,000
1,000
1,000
10,000

Figures indicate highest dilution at which agglutination occurred.

sorbed serums containing agglutinins for non-specific factors 2,
3, 5, 6 and 7, respectively. The preparation of these serums
was described by Bruner and Edwards (1941). The non-specific
components of S. paratyphi A were flocculated only by serums
containing agglutinins for factors 3 and 5. In this respect they
resembled the Kunzendorf, Berlin and Sendai types. Phase 2
of S. paratyphi A, therefore, may be expressed as 1, 5. . ..

As mentioned above, the non-specific phases derived from cul-
tures of S. paratyphi A were not identical. The differences in
the phases were discernible only in agglutinin absorption tests.
It was possible to divide the phases into two groups whose non-

iu

NON-SPECIFIC COMPONENTS IN S. PARATYPHI A 475

specific components were identical. In one group were cultures
228, HR and Fried; while 38250, Marta, Cal. 49, 1015, HA1,
HA6, WB39, Boyd and Durazzo were included in the second.
The differences between these two groups were evidenced only
by a slight residue of agglutinins remaining when the serum
derived from a member of the second group was absorbed with
a member of the first group. These residues of agglutinins did
not exceed 2 per cent of the original titres of the serums. The
differences in the non-specific components of S. paratyphi A
were no more pronounced than those that exist in the non-specific
phases of different diphasic species in which the second phase is
expressed as 1, 5.... The phases denoted as 1, 5... in the
Kauffmann-White classification are quite complex and their
exact relationships cannot be expressed without the use of further
symbols. In this connection it should be remembered that the
classification was designed by Kauffmann as a diagnostic schema
and that it does not give a complete antigenic delineation of
the bacilli.

The relationships of the non-specific components of S. para-
typhi A to the naturally occurring 1, 5. . . phases were studied
by agglutinin absorption tests. It was found that serums derived
from 228, HR or Fried were completely exhausted of agglutinins
by absorption with the non-specific components of 38250, Marta,
Cal. 49, 1015, HA1, HA6, WB39, Boyd, Durazzo or Sendai.
The serums of the latter group were not completely absorbed by
228, HR, Fried or Sendai. None of the non-specific phases of
S. paratyphi A removed all the agglutinins from Sendai serum.
The relationships of the non-specific components of S. paratyphi A
and Sendai were appreciably closer than were their relationships
with the non-specific phases of any other type.

DISCUSSION

While earlier workers realized that the Salmonella group was
composed of a mosaic of interrelated antigens, Bruce White
(1926) was the first to identify these antigens by symbols and
thus express the relationship and divergences of different species
in graphic form. In addition he was the first to discuss the

112

476 D. W. BRUNER AND P. R. EDWARDS

genetic relations of the various forms with reference to antigenic
composition and to propose a theory of Salmonella phylogeny.
White believed that the Salmonella species as we know them
today arose from a primitive diphasic ancestral stock by varia-
tion in O and H antigens. He regarded the monophasic state,
as seen in S. paratyphi A, S. dbortus-equi and E. typhosa, as an
acquired characteristic which developed through loss variation.
The demonstration by Edwards and Bruner that antigen a could
be isolated from S. abortus-equi revealed that the organism was
more closely related to diphasic forms having the antigens a-enx
than was previously realized. They suggested that in mono-
phasic types, the antigens present in the original diphasic state
were not lost but merely suppressed. The isolation of non-specific
components from S. paratyphi A confirms this view. Both ob-
servations strengthen the theory of White that monophasic
types have evolved from a diphasic ancestry.

White emphasized the close relationship that existed between
S. paratyphi A and Sendai, the only two types then known which
possessed antigen a. He used these two types to illustrate the
identity of the H antigens of monophasic types and the specific
antigens of diphasic types. The isolation of non-specific phases
from S. paratyphi A which are all but identical with the non-
specific phase of Sendai reinforces the relationship between the
specific phases of the two types and demonstrates the keen ap-
preciation of White of the evolutionary tendencies in the genus.

With the exception of the recovery of specific phases from
"totally and permanently" non-specific types, the isolation of
antigen a from S. abortus-equi and of non-specific components
from S. paratyphi A constitute the only instances in which anti-
gens resembling the naturally occurring phases have been induced
in monophasic cultures. The stability of these induced antigens
has been mentioned. It is in direct contrast to the instability of
the phases of the diphasic types. These results indicate that
once an organism has lost the power of phase variation, it cannot
resume this function, even when suppressed components become
dominant. Only under the influence of a stimulant, such as
specific immune serum, does phase variation occur.

NON-SPECIFIC COMPONENTS IN S. PARATYPHI A 477

It has been stated that the third and fourth phases induced in
S. paratyphi A showed little relationship to the natural antigens
of the genus. However, they do display relationships to the
induced phases of several other Salmonella types. As mentioned
by Edwards and Bruner (1939a) these induced phases follow a
recurring pattern, just as do the naturally occurring antigens.
Whether they are artifacts produced under the influence of im-
mune serum or whether they represent components that have
been suppressed in the evolution of known species is not clear at
present. The writers are inclined to the latter view.

SUMMARY

1. Through cultivation in agglutinating serums three induced
phases were isolated from cultures of Salmonella paratyphi A.
One of the phases was closely related to the non-specific phases
of diphasic types. The other two bore little resemblance to the
normal Salmonella antigens.

2: Non-specific components were isolated from 12 of 16 cultures
studied. These non-specific antigens were expressed as 1, 5. . .
and were more closely related to the non-specific phase of the
Sendai type than to any of the other non-specific phases.

3. The induced phases were stable under ordinary conditions
of culture and could be reverted only by the addition of ap-
propriate serums to the medium.

4. The isolation of non-specific antigens from S. paratyphi A
was cited as additional evidence that monophasic Salmonella
types are derived from multiphasic types through suppression of
phases.

REFERENCES

Bruner, D. W., and Edwards, P. R. 1939 A note on the monophasic non-
specific Salmonella types. J. Bact., 37, 365-370.

Bruner, D. W., and Edwards, P. R. 1941 Microorganisms of group E of the
genus Salmonella with special reference to a new Salmonella type.
Am. J. Hyg., In press.

Edwards, P. R., and Bruner, D. W. 1939 Reversibility of the alpha and beta
phases of Salmonella typhi. Proc. Soc. Exptl. Biol. Med., 41, 223-224.

Edwards, P. R., and Bruner, D. W. 1939a The demonstration of phase varia-
tion in Salmonella abortus-equi. J. Bact., 38, 63-72.

114

478 D. W. BRTJNER AND P. R. EDWARDS

Gard, S. 1938 Ein neuer Salmonella-Typ (S. abortus-canis) . Z. Hyg. In-

fektionskrankh., 121, 139-141.
Gnosspelius, A. 1939 Ueber kiinstliche Veranderungen des H-Antigens in

der Salmonella-Gruppe. Z. Hyg. Infektionskrankh., 121, 529-532.
Hunter, C. A., and Crecelius, H. G. 1938 Hydrogen sulfide studies. I.

Detection of hydrogen sulfide in cultures. J. Bact., 35, 185-196.
Kauffmann, F. 1936 Ueber die diphasische Natur der Typhusbacillen. Z.

Hyg. Infektionskrankh., 119, 104-118.
Kauffmann, F. 1937 Salmonella Probleme. Z. Hyg. Infektionskranhk., 120,

177-197.
Kauffmann, F., and Tesdal, M. 1937 Ueber zwei neue Salmonellatypen mit

a-/S-Phasenwechsel. Z. Hyg. Infektionskrankh., 120, 168-176.
Scott, W. M. 1926 The "Thompson" type of Salmonella. J. Hyg., 25, 398-405.
White, P. B. 1926 Further studies of the Salmonella group. Med. Research

Council (Brit.) Special Rept. Series No. 103.

'5

THE NATURAL OCCURRENCE OF PHASE 2 OF SALMONELLA
PARATYPHI A 1

P. R. EDWARDS

Laboratory Division, Communicable Disease Center, Public Health Service, Federal Security
Agency, Atlanta, Georgia,

AND

L. A. BARNES and MARY C. BABCOCK

Bacteriology Division, Naval Medical Research Institute, Bethesda, Maryland

Received for publication October 24, 1949

By a modification of the Gard technique, Bruner and Edwards (J. Bact.,
42, 467, 1941) obtained from typical cultures of Salmonella paratyphi A
(I, II, XII: a) a variant having the formula I, II, XII: 1, 5. This experimental
evidence supported the hypothesis of White (Med. Research Council, Brit.,
Special Rept. Series, No. 103, 1926) that S. paratyphi A, as well as other mono-
phasic Salmonella types, was a loss variant of an originally diphasic ancestral
form. In the isolation of the 1,5 phase and its reversion to the original form,
two "artificial" phases that did not correspond to any of the known naturally
occurring antigens of the genus were produced. These were designated as z 5
and zn, respectively. The natural occurrence of phase 2 (I, II, XII :1 ,5) has not
been reported heretofore.

In 1948 a culture isolated by the United States Naval Medical Research Unit
No. 3 in Egypt from the urine of a patient affected with enteric fever of 49 days'
duration was forwarded to the Naval Medical Research Institute and subse-
quently to the Communicable Disease Center for study. Examination revealed
that the bacterium was a motile rod that possessed the tinctorial and cultural
properties of Salmonella.

The organism reduced nitrates but failed to utilize citrate or D-tartrate.
Acid and gas were formed in the butt of triple-sugar iron agar slants; the medium
was not blackened. Indole was not formed in peptone water. Glucose, mannitol,
arabinose, rhamnose, maltose, and trehalose were fermented with the production
of acid and gas within 24 hours; dulcitol was fermented in 48 hours. Xylose,
cellobiose, lactose, sucrose, raffinose, inositol, and salicin were not attacked.
The biochemical properties described above are typical of S. paratyphi A.

Upon serological examination it was found that the organism was agglutinated
strongly by O serum derived from S. paratyphi A and possessed somatic antigens
I , II , XII. In absorption tests it removed all somatic agglutinins from two serums
derived from S. paratyphi A. When the H antigens were examined, it was found

1 The opinions or assertions contained herein are the private ones of the writers and are
not to be construed as official or reflecting the views of the Navy Department or the Naval
Service at large.

135

[Reprinted by permission of The Williams & Wilkins Company from Journal of
Bacteriology 59: (1) 135-136, January. 1950]

u6

136 NOTE [vol. 59

that the culture was not flocculated by S. paratyphi A serum (a) but reacted
with serum derived from the "nonspecific" phases of the genus (1,2: 1,5: etc.).
When tested for single factors 2, 3, 5, 6, and 7, it was agglutinated only by 5
serum. In absorption tests it removed all agglutinins from serums prepared from
1,5 phases of S. paratyphi A. Thus the culture was identical with the 1,5 phases
previously induced by growing S. paratyphi A in homologous serum.

The culture was monophasic and was immobilized when placed in semisolid
medium to which had been added various 1,5 serums. After repeated transfers,
bulbs extending from the site of inoculation appeared. When these bulbs were
transferred to additional tubes of the same medium, the culture migrated rapidly
through the agar. From the spreading growth was isolated a form that was ag-
glutinated to 20 per cent of the titer of H serum of S. paratyphi A and that
reacted to approximately the same degree with serums produced from the
"artificial" phases z 6 and z n . When the culture was grown in semisolid medium
to which z 5 and Zn serums as well as 1,5 serum had been added, no spreading
occurred although the serums were carefully absorbed to remove interfering
factors and the organism was carried through a number of transplants extending
over a period of 6 months. It should be emphasized that it was very difficult to
cause reversion of induced 1 , 5 forms of S. paratyphi A to the original form and
that attempts to cause reversion failed in most instances.

The observations outlined above provide the first report of the natural occur-
rence of phase 2 of S. paratyphi A having the antigenic formula I, II, XII: 1 ,5.
Such a variant is not, therefore, merely a curiosity produced by artificial manip-
ulation, although it was biochemically and antigenically identical with strains
developed by cultivation of typical cultures in agglutinating serums. The findings
emphasize further the importance of the study of phylogenetic and evolutionary
trends of the Enterobacteriaceae in the identification of aberrant cultures.

17

CHANGES INDUCED IN THE O ANTIGENS OF SALMONELLA 1

D. W. BRUNER and P. R. EDWARDS

Department of Animal Pathology, Kentucky Agricultural Experiment Station,
Lexington, Kentucky

Received for publication December 11, 1947

It is possible to induce profound changes in the flagellar antigens of Salmonella
through growth in media containing H antisera. These changes were recently
summarized by Edwards and Moran (Proc. Soc. Exptl. Biol. Med., 61, 242).
Hitherto, the only changes reported in the O antigens were those associated with
variation between smooth and rough forms and those occurring in natural form
variation. In this laboratory attempts to transform O antigens by the method
of Boivin et al. (Experientia, 2, 139) and by growth in various combinations of
S and R antisera and S vaccines have been unsuccessful. However, by a modi-
fication of the method of Gard (Z. Hyg., 120, 615) in which absorbed O anti-
serums were used in high concentration, it was possible to bring about certain
changes.

When S. anatum (III,X,XXVI: e,h-l,6) was cultivated in semisolid medium
containing III,X,XXVI serum that had been absorbed with a type having O
antigens III,XV, the organisms gradually spread through the medium. From
the spreading growth was isolated a form that was indistinguishable from S.
newinglon (III,XV:e,h-l,6) by agglutination and absorption tests. Absorption
of the III,X,XXVI serum by S. newington, S. Cambridge (III,XV: e,h-l,w), or S.
new-brunswick (III,XV: l,v-l,7) gave the same results. The induced III,XV:
e,h-l,6 form was then reverted to a typical S. anatum strain by cultivation in
absorbed III,XY serum. Similarly, S. meleagridis (III,X,XXVI: e,h-l,w) was
changed to a form indistinguishable from S. Cambridge (III,XV: e,h-l,w).

Although filtrates similar to those employed by Boivin were not used in the
experiments, it must be remembered that the serums were absorbed with very
large doses of bacteria and probably contained dissolved antigens as well as
metabolic products. Thus the principle that induced the changes may be the
same as that involved in Boivin 's work with Escherichia coli. Further, attention
should be called to the fact that the changes described are only transformations
between subgroups of the same O group. Whether similar experiments will
permit transformation between distinct O groups or whether they will lead only
to rough variation remains to be determined.

1 The investigation reported in this paper is connected with a project of the Kentucky-
Agricultural Experiment Station and is published by permission of the Director. It was
supported in part by a research grant from the U.S. Public Health Service.

449

[Reprinted by permission of The Williams & Wilkins Company from Journal of
Bacteriology 55 : 449, 1948]

Il8

MUTATIONS IN ESCHERICHIA COLI INDUCED

BY CHEMICAL AGENTS'

EVELYN M. WITKIN

For over two decades geneticists have been in-
terested in the possibility of inducing mutations
with chemicals. Particularly, it has been hoped that
mutagenic compounds might be discovered which,
through their specificity of action, would lead to
some understanding of the chemical basis of muta-
tion, and ultimately of the structure and organiza-
tion of the gene.

The first clearly successful attempt to induce
genetic changes chemically was described by Auer-
bach and Robson (1944), who produced mutations
and chromosomal abberrations in Drosophila by
exposing the flies to mustard gas and related com-
pounds. More recently, Demerec (1947 and un-
pub.) has shown that certain carcinogenic hydro-
carbons (1, 2, 5, 6-dibenzanthracene, methylcho-
lanthrene, beta naphthylamine, benzpyrene) are also
effective in inducing mutations in Drosophila. These
agents, like radiations, appear to be entirely non-
specific in the sense that the affected loci are dis-
tributed at random along the treated chromo-
somes. Nitrogen mustard has been shown to induce
genetic alterations in Neurospora (Tatum, unpub.;
Horowitz, Houlahan, Hungate and Wright, 1946),
and in bacteria (Tatum, 1946; Bryson, unpub.),
and at least one of the carcinogens, methylcholan-
threne, is effective in Neurospora (Tatum, unpub.).
These successful results seem to be the opening guns
in what Muller called, a few years ago, "the com-
ing chemical attack on the nature of the gene"
(Muller, 1947). They suggest the need for a system-
atic survey to determine the distribution of muta-
genic compounds among various chemical groups,
and to lay the groundwork for subsequent analysis
of their mode of action. This paper will deal with
preliminary results obtained in tests of 4 substances,
the first of a series to be investigated in an exten-
sive survey.

Two methodological factors are of critical im-
portance in an attempt to examine large numbers
of compounds for mutagenic activity: the basis upon
which the chemicals are selected for test, and the
choice of biological material. Concerning the method
of selecting chemicals, one sober if somewhat un-
imaginative approach is an indiscriminate raid on
the nearest chemical shelf, which has the advantage
of objectivity and avoidance of the hazards of pre-
mature preconceptions. On the other hand, it is

'This work was done under an American Cancer Society
fellowship recommended by the Committee on Growth of the
National Research Council.

far more tempting to extend oneself on the basis
of present ideas concerning the possible organiza-
tion of genie material, and to select chemicals which
might reasonably be expected to affect, or fail to
affect, the projected hereditary units. The ap-
proach used in these experiments has been to
assume that nucleoproteins are somehow centrally
involved in the genetic system, and, as a starting
point, to investigate chemicals known to have some
more or less well-defined chemical or physical effect
on nucleoproteins or nucleic acids. It must be em-
phasized that this approach has no greater justifica-
tion on a priori grounds than many others, and that
the basis of selection may prove to be entirely
spurious, since none of the chemicals tested thus
far is specific in its action on nucleoproteins or
nucleic acids.

The choice of biological material is obviously
very important in this type of investigation. The
primary requirements are 1) the availability of
techniques for treating the organism with chemicals
so as to be reasonably certain that they will reach
and penetrate the critical sites, and 2) the avail-
ability of clear-cut genetic methods for detecting
induced mutations. The penetration problem has
been the most serious difficulty in the use of Dro-
sophila for chemical induction, and although im-
proved methods of treatment have been developed,
the possibility remains that negative results may
be due to the failure of some chemicals to penetrate
the germ cells in sufficient concentration. The
genetic techniques for detecting induced mutations
in Drosophila are unparalleled in many respects,
but for purposes of an extensive survey of the mu-
tagenic action of chemicals, they are extremely
laborious and slow. The problem of penetration is
much less serious in microorganisms. Until recently,
however, genetic methods analogous to the C1B
and similar techniques in Drosophila have not been
available for bacteria. At the present time, Escher-
ichia coli provides promising material for a survey
of the mutagenic activity of chemicals, and for de-
tailed analysis of certain aspects of their mode of
action.

Luria and Delbriick (1943) described mutants
of strain B of E. coli which are resistant to one or
more bacteriophages to which the parent strain is
sensitive. These mutants arise spontaneously in cul-
tures of the B strain at a rate of about 10 -8 muta-
tions per bacterium per generation, and can be de-
tected easily by plating out samples of the culture
in the presence of an excess of bacteriophage. The

[256]

[Reprinted from Cold Spring Harbor Symposia on Quantitative Biology 12 : 256-269,
published by the Long Island Biological Assn.]

IIQ

INDUCED MUTATIONS IN E. COLI

257

sensitive bacteria are quickly lysed, while mutants
resistant to the particular bacteriophage applied ap-
pear as distinct colonies after suitable incubation.
Demerec (1946) and Demerec and Latarjet
(1946) studied the effect of ultraviolet and X-rays
in inducing mutations to resistance to one bacterio-
phage, Tl, and the techniques used by them have
been adapted in these experiments to the investiga-
tion of chemical mutagens. These authors showed
that a certain proportion of the mutations induced
by radiations are expressed immediately, before the
treated bacteria undergo division. These mutations
were called "zero points," as opposed to the "end-
point" mutations which are also induced by the
radiation, but which are expressed phenotypically
only after a number of cell divisions. In the in-
vestigations to be discussed here, the effect of four
chemicals on the induction of zero point mutations
to resistance to bacteriophage Tl was studied.

Material

Strain B/r of E. coli, a radiation-resistant mutant
of the B strain (Witkin, 1947), was used exclu-
sively throughout these experiments. This strain
was employed by Demerec (1946) and by Demerec
and Latarjet (1946) in their studies of radiation-
induced mutations, and it seemed desirable to use
the same material in these investigations, in order
to simplify comparison of the mutagenic action of
radiation and chemicals.

The bacteriophage used to isolate resistant mu-
tants was Tl, sometimes known also as alpha, or
P28. Mutants of strain B/r resistant to phage Tl
are referred to as B/r/1.

Difco nutrient agar was used for all platings,
and cultures were grown initially in a synthetic
medium known as M-9, having the following com-
position per 1,000 ml. of distilled water:

KH 2 P0 4 3 g. NH 4 C1 1 g.

MgS0 4 0.2 g. Na 2 HP0 4 -12H 2 IS g.

NaCl 0.5 g. Dextrose 4 g.

Methods

The method used to determine the number of
mutants resistant to bacteriophage Tl, in both con-
trol and experimental cultures, was the standard
procedure of coating the surface of agar plates with
a suspension containing about 10 9 particles of Tl,
and then spreading 0.1 ml. of the undiluted bac-
terial culture on the phage-coated plate with a
sterile glass rod. The plates are incubated for 48
hours at 37° C, after which time the colonies of
resistant mutants are counted. The total number of
bacteria per ml. of the culture is determined by
plating suitable dilutions on agar, and making
colony counts. The frequency of B/r/1 mutants is
expressed throughout as the number per 10 8
bacteria.

The procedure for testing the mutagenic activity

of chemicals has been standardized as far as pos-
sible. Variations in technique are required to allow
for the peculiar properties of each chemical, and
will be described in connection with the specific
experiments. The basic procedure used in prelimi-
nary tests may be outlined as follows:

(1) The preparation of suitable bacterial cultures
Ten to 20 cultures are usually started at one time,
each with a small inoculum (about 100 cells) from
a stock slant of strain B/r. The cultures are grown
for 48 hours at 37° C, with aeration, in a volume
of 40 ml. of the synthetic medium, M-9, described
above. Each culture is assayed to determine the
number of bacteria per ml., and the number of
B/r/1 mutants per 10® bacteria. The number of
mutants per 10 8 bacteria in a fully grown, untreated
culture is called the "background" number, and
must be subtracted from the number of mutants
per 10 8 survivors in the same culture after treat-
ment with a possible mutagenic chemical, to deter-
mine the number of induced mutations. It is desir-
able, therefore, to use cultures having the lowest
possible background number, since a high back-
ground may obscure a positive effect, particularly
where the mutagenic activity is weak. Cultures
found to contain more than 10 background mutants
per 10 8 bacteria are discarded, and the remaining
cultures are combined to form a stock pool, which
is stored at 5° C, and serves to provide samples for
experiments for a period of about a week.

(2) Toxicity tests

The optimum concentration of a chemical to be
tested depends largely upon its toxicity for the
bacteria. In general, for preliminary tests, a com-
bination of concentration and time of exposure was
used so as to kill about 99% of the bacteria. The
high killing is an indication that the substance is
penetrating the cell and reacting with its com-
ponents, and 1% survival is usually just enough to
permit the detection of mutants among the survivors
even if no mutations are induced. The concentra-
tion or time of exposure may be adjusted later, de-
pending upon the results of the preliminary tests.
For convenience, an arbitrary time, usually 2 or 3
hours, was chosen, and the concentration of the
chemical that would kill 99% of the bacteria in
this period of time was determined. The following
method was used.

A concentrated solution of the chemical in dis-
tilled water or a suitable buffer was prepared, and
a series of widely spaced dilutions of the stock solu-
tion were made. A number of centrifuge tubes, each
containing the same volume of bacterial culture, was
set up and centrifuged for 20 minutes at 4 y 000 RPM.
The clear supernatant was decanted from each tube
to eliminate the nutrient medium. The bacterial
pellets were then resuspended, so as to have one
tube for each dilution of the chemical, and one

258

EVELYN M. WITK1N

tube in buffer or distilled water to serve as the
control. The volume of liquid in each tube was the
same. The tubes were incubated for the arbitrarily
chosen time, usually 2 or 3 hours, at 37° C, and
then assayed to determine the number of bacteria
per ml. in each of the tubes. Sometimes a second
toxicity test was required, using dilutions between
two of the original steps, to determine the proper
concentration more precisely.

(3) Test for mutagenic activity
Knowing the concentration of the chemical that
will kill 99% of the bacteria in 2 or 3 hours, a pre-
liminary test is run to determine the effectiveness
of the chemical under these conditions in inducing
zero point mutations. Since sample experiments
will be described in detail in the experimental sec-
tion, only a general outline of the procedure will be
given here:

A sample of a stock low-background culture is
divided into two equal parts, centrifuged, and the
nutrient medium decanted. One pellet is resuspended
in the proper concentration of the chemical, and the
other in distilled water or buffer as a control. The
tubes are incubated for the length of time required
to kill 99% of the bacteria in the experimental tube,
and then the cultures are assayed to determine the
number of viable bacteria per ml., and the number
of B/r/1 mutants per 10 8 bacteria. The number of
mutants per 10 8 bacteria in the control, the back-
ground number, is subtracted from the number of
mutants per 10 8 survivors in the treated culture. As
the background number of a given culture is ex-
tremely constant in independent determinations,
seldom differing by more than 2 or 3 mutants per
10 8 , an increase of 10 mutants per 10 8 over the
background number is considered to be an indica-
tion of a positive effect.

Since the effect of chemicals on zero point muta-
tions is under consideration, it is important to
establish conditions under which division of bac-
teria during treatment will not occur. The use of a
non-nutrient medium, and a concentration of bac-
teria at least as great as, and often 10 to 20 times
greater than, the concentration reached after maxi-
mal growth, were among the precautions taken to
prevent division during treatment. In addition, the
exposure time was usually well below the normal
lag, even for small inocula in fresh nutrient medium.
Microscopic total counts revealed no measurable
increase in cell number under the conditions de-
scribed, over periods as long as 72 hours.

(4) Differential survival test
Whether the preliminary test for mutagenic
activity is positive or negative, it is important to
eliminate the possibility of a selective action of the
chemical. If B/r/1 mutants are, for some reason,
less sensitive to the toxic effects of the chemical, the
proportion of mutants among the survivors of the

treated culture will be higher than in the control,
and a mutagenic action might be erroneously
ascribed to the compound. If the mutants should be
more sensitive than the nonmutants, it is possible
that a positive mutagenic effect could be masked by
the differential killing of mutants. Thus, to be cer-
tain that selection for or against the mutants is not
responsible for an apparent positive or negative
effect, it is necessary to compare the sensitivity of
mutants and nonmutants to each of the chemicals
tested. The standard procedure used in these experi-
ments involved the following steps:

A stock of B/r/1 was established by inoculating a
small amount of growth from about ten representa-
tive B/r/1 colonies on a control plate into a tube of
M-9. A new stock was made up for each experiment,
and the colonies which were isolated always came
from a plating of the control tube of an experi-
ment designed to test the mutagenic action of the
chemical in question. Thus, the B/r/1 stock iso-
lated in this way was representative of the mutants
present in an experimental culture at the start of
the exposure to the chemical. The culture derived
from these colonies was streaked on agar to elimi-
nate contaminating particles of bacteriophage, and
another culture was started with an inoculum from
at least ten colonies. This culture was used as a
source of B/r/1 for the selective killing test.

To compare the sensitivity of B/r and B/r/1 to a
given chemical, 48-hour aerated M-9 cultures of
the two strains were grown, and equal volumes of
the two cultures were mixed together, to give a cul-
ture containing approximately half mutants and
half nonmutants. Two tubes were set up, each con-
taining the same volume of the mixed culture, cen-
trifuged, and the nutrient medium poured off. One
pellet was resuspended in the standard concentra-
tion of the chemical, and the other in buffer or dis-
tilled water. The tubes were incubated for the
standard time, and assayed to determine the pro-
portion of mutants in each culture. If mutants and
nonmutants are killed at the same rate, there
should be no difference in the proportion of mutants
between the control and experimental tubes. A sam-
ple experiment of this type will be described below.

(5) Test of mutant colonies
In all cases where a positive effect was obtained,
samples of the mutant colonies obtained after treat-
ment with the chemical were isolated, and tested
carefully to establish the fact that they were true
B/r/1 mutants. In experiments where the total
number of colonies obtained was small, and where
a few contaminants could distort the results, every
colony was isolated and tested. The colonies were
examined for resistance to Tl, and also to another
phage, T2, to which strain B/r is sensitive. Colonies
showing resistance to Tl and sensitivity to T2 were
regarded as B/r/1 mutants. Colonies showing re-
sistance to phages were regarded as contaminants.

:2i

INDUCED MUTATIONS IN E. COLI

259

Experimental Results
Since the effect of chemicals in inducing zero
point mutations is to be investigated, it is important
to know to what extent mutations occur spon-
taneously, if at all, in resting cells. Luria and
Delbriick (1943) have shown that mutations to
resistance to Tl do not occur spontaneously in

other compounds tested in these experiments. It is
a highly reactive compound, forming addition com-
pounds known as choleates with a wide variety of
organic substances, including fatty acids, ether,
xylol and certain carcinogenic hydrocarbons ( Wie-
land and Sorge, 1916; Fieser and Newman, 1935).
Alloway (1933) used sodium desoxycholate in his

Table 1. Number of B/r/1 Mutants in Resting Culture or B/r Suspended in Distilled Water for 24 Hours at 37° C.

Time after suspension in distilled H 2

5hrs.

12 hrs.

24 hrs.

Colony Counts .05 ml. of 10 6 dilution on each of

259

234

255

264

4 plates

276

238
247

268
240
279

254
218
260

270
231
208

No. bacteria per ml.

5.1X10»

5.1X10 9

4.9X10'

4.9X10 9

No. B/r/1 mutants in ten 0.1 ml. samples of

33

41

39

23

61 37

45 50

undiluted culture

27

44

50

52

48 31

28 61

28

50

31

40

32 54

37 32

39

36

37

39

39 51

37 45

19

40

24

41

41 45

35 36

No. B/r/1 mutants per 10 8 bacteria

7.0

7.3

8.9

8.3

resting bacteria, and our own observations have
confirmed this. Table 1 shows the results of an
experiment in which resting bacteria, suspended in
distilled water and incubated for a period of 24
hours at 37° C, were periodically assayed to deter-
mine the number of B/r/1 mutants per 10 8 cells.
The number of mutants per 10 8 bacteria remains
extremely constant under these conditions, and any
mutations induced in resting cells must be compared
with a baseline of zero.

1. Sodium Desoxycholate

Sodium desoxycholate is a salt of the bile acid
desoxycholic acid, the molecular structure of which
is shown in Fig. 1, along with the structures of the

^ H ,CH e CM 2 -COONa

(I) SODIUM DESOXYCHOLATE

:h/

,-o

o

l/l H

(3) METHYL GREEN

(2) PYRONIN Y

CH 3 .C\

(4) NEUTRAL ACRIFLAVINE

Fig. 1. Molecular structure of Sodium Desoxycholate,
Pyronin; Methyl Green and Acriflavine.

early work on type transformation, and it is used
by Avery and his coworkers in the preparation of
transforming principle (1944), where its function is
to dissolve the desoxyribose nucleoprotein complex
of the pneumococcus. Mirsky and Pollister (1946)
have shown that a .5% solution of sodium desoxy-
cholate dissolves their preparations of thymus
chromosin fibers, and can be used to extract the
desoxyribose nucleoprotein complex from minced
thymus. While this solvent action of desoxycholate
on nucleoproteins served as a basis for selecting
this compound for a test of its mutagenic activity,
there is no doubt that it is a nonspecific effect.

Sodium desoxycholate dissolves easily in water
above pH 6.5, but solutions at pH 6.5 to 7.5 form a
gel at high concentrations. Since the reaction of
desoxycholate in distilled water is about 7.7, un-
buffered solutions of the compound in distilled
water were used. Fresh solutions were made up for
each experiment, and sterilized by immersion in a
boiling water bath for 15 minutes.

Toxicity. Preliminary toxicity tests showed that
99% of the exposed bacteria were killed in 3 hours
by a 5% solution, or in 48 hours by a 0.3% solu-
tion. All of the experiments reported here were
done using the 5% solution.

Test for mutagenic activity. To illustrate in de-
tail the method used in testing chemicals for ability
to induce mutations, a typical experiment done with
sodium desoxycholate will be described.

Ten-ml. aliquots of a stock low-background cul-
ture of B/r were pipetted into each of 4 centrifuge
tubes, which were centrifuged for 20 minutes at
4,000 RPM. The supernatant was then decanted to

122

260

EVELYN M. WITKIN

eliminate the nutrient buffered medium. The bac-
terial pellets in two of the tubes were resuspended
in 2 ml. of a 5% solution of sodium desoxycholate,
and the bacteria in the remaining two tubes were
resuspended in 2 ml. of distilled water to serve as
controls. In some experiments, a phosphate buffer
of pH 7.7 was used for the control, but distilled
water was found to be equally satisfactory. In
neither case was there any change in the number of
viable bacteria or in the background number during
the course of any experiment. The initial centrifu-
gation, in addition to eliminating the nutrient

mutants and 50% nonmutants. Ten ml. of the
mixed culture was pipetted into each of two centri-
fuge tubes, centrifuged in the usual way, and the
pellets resuspended in 2 ml. of liquid, 5% desoxy-
cholate for the experimental tube and distilled water
for the control. The tubes were incubated for three
hours at 37° C. Thus, these cultures received the
same treatment throughout as the cultures used to
test for mutagenic activity. At the end of the 3-hour
incubation, the tubes were assayed to determine the
proportion of mutants in the control and experi-
mental cultures. This was done by plating the final

Table 2. Zero-Point Mutations Induced by 3-Hour Exposure to 5% Sodium Desoxycholate
1 and 2 are duplicate control tubes. 3 and 4 are duplicate experimental tubes, a and b are independent dilutions of each tube.

Experimental

Assay
Sample per plate
Colony Counts

No. bacteria per ml.

No. B/r/1 mutants in samples of 0.1 ml. of
undiluted culture (8 samples per tube)

No. B/r/1 mutants per 10 s living bacteria

.05 ml. of 10 7 dilution
#1 #2

122 a. 101

86 89

b.

95

b.

94

97

105

1.97X10 10

#1

n

71

59

68

83

78

67

82

71

67

79

74

72

74

97

90

60

.05 ml.
#3
a. 79
88

of 10 5 dilution

a. 76
73

b. 101

82

b. 87
90

1.69X10"

10
9
14
6

11 6

7 18

8 4

12 12

Total — <50.2
Induced — 56.4

8
10
13
15

medium, served to concentrate the bacteria by a
factor of 5 (in some experiments by as much as a
factor of 20), thus facilitating the detection of
mutants among the survivors.

The control and experimental tubes were in-
cubated for 3 hours at 37° C, and were then as-
sayed to determine the number of viable bacteria
per ml., and the number of B/r/1 mutants per 10 8 .
The results are shown in Table 2. These results
indicate the proportion of mutants among the sur-
vivors of the exposure to desoxycholate was in-
creased from 3.8 per 10 8 , the background number,
to 60.2 per 10 8 . Subtracting the background, the
increase in mutants per 10 s survivors is 56.4.

Differential Survival test. In order to test the
possibility that the observed increase in the propor-
tion of mutants among the survivors of desoxycho-
late-treated bacteria is due to simple selection, the
following experiment was done.

Cultures of B/r and B/r/1 were grown with
aeration in M-9 for 48 hours, and equal volumes of
the two cultures were combined in a single tube, to
give a mixed culture containing approximately 50%

dilution of each culture on two series of plates, one
consisting of ordinary agar plates, and one consist-
ing of plates which had been coated previously with
a heavy suspension of bacteriophage Tl. The ratio
of the colony counts on phaged plates to that on
unphaged plates gives the proportion of mutants in
the mixture. The results are given in Table 3. In
the control culture, the number of colonies develop-
ing on phaged plates was 48.9% of the number
appearing on unphaged plates. In the experimental
culture, in which about 99% of the bacteria were
killed, the proportion of mutants among the sur-
vivors was 51.6%. Thus, it appears that mutants
and nonmutants are equally sensitive to the toxic
effects of sodium desoxycholate. These tests were
made on mixed cultures rather than on pure cultures
of the two strains in view of the possibility that
competition phenomena might be involved in selec-
tive killing.

Tests of Mutant colonies. Ninety-four colonies
appearing on phaged plates after treatment with
desoxycholate were isolated and tested for re-
sistance to Tl and T2. All but three proved to be

123

INDUCED MUTATIONS IN E. COLI

261

resistant to Tl and sensitive to T2, and were there-
fore considered to be true B/r/1 mutants. Two
colonies were resistant to both phage strains, and
were regarded as contaminants (in both cases the
colony morphology alone rendered them suspect).
The third colony was sensitive to both Tl and T2,

Table 3. Test for Differential Killing of Strain B/r

and B/r/1 by 3-Hour Exposure to 5%

Sodium Desoxycholate

Control = mixed culture in distilled water

Experimental mixed culture in 5%

Sodium Desoxycholate

tially the same methods as those described in the
sample experiment above. Wherever possible, a
single experimental tube was incubated and sampled
over the entire 10-hour period, since the use of

Control

Experimental

Sample per plate

.05 ml. of
5X10 6 dilution

.05 ml. of
5 X10 4 dilution

Colony Counts

plated
without
phage

plated

with

excess of

phage

plated
without
phage

plated

with

excess of

phage

462
377
468
408

192
226
198

224

441
443
482
493

233
240
246
239

No. bacteria

per ml.

B/r+B/r/1

B/r/1

4.29X10"

2.1 X10 10

4.65X10 8

2.4X10"

% B/r/1

48.9

51.6

and since this colony appeared on the edge of the
plate, it is likely that it arose from a sensitive
bacterium which escaped lysis due to faulty spread-
ing of the phage suspension.

Relation between the number of induced muta-
tions and time of exposure to 5% desoxycholate.
Demerec and Latarjet (i Q 46), in their studies of
radiation-induced mutations to phage-resistance,
investigated the relation between dosage of radia-
tion and the number of mutations induced. They
found that the number of zero point mutations to
B/r/1 is directly proportional to X-ray dose, and
bears a more complicated exponential relation to
ultraviolet dose. In the present investigations, ex-
periments were conducted to determine the
analogous relation between the number of induced
mutations and time of exposure to a 5% solution
of sodium desoxycholate.

Fig. 2 shows a survival curve obtained with
sodium desoxycholate, in which the percentage of
surviving bacteria is plotted against time of ex-
posure to the standard concentration. Fig. 3 shows
the number of zero point mutations, corrected for
the background number, as a function of time of
exposure to 5% desoxycholate. The data upon
which this curve is based were obtained by essen-

2 4 6 8 10

HOURS EXPOSED TO 5%
SODIUM DESOXYCHOLATE

Fig. 2. Survival curve of resting bacteria exposed to 5%
Sodium Desoxycholate.

separate tubes for each determination gave some-
what more variable results. The linear relation be-
tween induced mutations and exposure time can be
compared directly with the X-ray dosage-effect

<£ X-RAY DOSE IN ROENTGEN UNITS XIO" 4
2.5 5 7.5 10

2 4 6 8 10 12

HOURS EXPOSED TO 5% SODIUM
DESOXYCHOLATE

Fig. 3. Relation between the number of zero point muta-
tions induced and time of exposure to 5% sodium desoxy-
cholate. The upper abscissa represents X-ray dose in roent-
gen units, on a scale required to make this curve coincide
with the curve obtained by Demerec and Latarjet (1946)
for the relation between X-ray dose and number of zero
point mutations.

124

262

EVELYN M. WITKIN

curve obtained by Demerec and Latarjet, and dif-
fers fundamentally from their analogous ultraviolet
curve. The upper abscissa in Fig. 3 represents X-ray
dose in r units, on a scale required to make the
desoxycholate curve coincide with the X-ray curve
of Demerec and Latarjet. It will be noted that
100,000 r units correspond, in terms of the number
of induced zero point mutations, to an 8-hour ex-
posure to 5% sodium desoxycholate.

On the basis of these positive results obtained
with desoxycholate as a mutagenic agent in bacteria,
Demerec (unpub.) has tested the ability of this
compound to induce lethal mutations in Drosophila.
The proportion of lethals among the tested sperm
was about 1.5%, as compared with 0.24% in con-
trols, an effect of the same order of magnitude as
that obtained with nitrogen mustard by the same
technique. These results with Drosophila suggest
that sodium desoxycholate may prove to be a non-
specific mutagen, like radiations and mustard. No
tests have been made to investigate the possibility
of specific effects on the particular mutation used
as an index of mutagenic activity, namely, re-
sistance to bacteriophage Tl. For this purpose it
will be necessary to investigate the action of de-
soxycholate on other bacterial mutations.

2. Pyronin and Methyl Green

Basic dyes, which stain chromatin by virtue of
their affinity for nucleic acids, were considered as a
class of compounds worthy of investigation for
possible mutagenic activity. Pyronin and methyl
green, the components of the Unna-Pappenheim
mixture currently of interest in cytochemical re-
search, were the first dyes to be tested. Brachet
(1940) has described the usefulness of this mixture
as a means of differentiating cytochemically be-
tween ribose and desoxyribose nucleic acids, by
virtue of the selective affinity of pyronin for the
ribose type of nucleic acid. Although there is good
evidence for the specificity of the pyronin-methyl
green mixture under certain conditions, it must be
pointed out that the treatment of living bacteria,
without recourse to the procedures of fixation and
differentiation which are standard in cytological
work, may very well present entirely different con-
ditions. Thus pyronin may be taken up by fatty
acids, or other basophilic elements ordinarily re-
moved in cytological preparations. There is no
basis, therefore, for assuming that these dyes are
specific, under the conditions of these experiments,
in acting exclusively upon nucleic acids, or in dif-
ferentiating between the two types of nucleic acid.

Pyronin was tested by essentially the same tech-
niques described in connection with desoxycholate.
One difficulty encountered, however, was the great
variability of the toxic effects of a given concentra-
tion in different experiments, or in different tubes
in the same experiment. One possible source of the
variability was thought to be the precipitate which

forms in the presence of bacteria at high concentra-
tions of the dye. Low concentrations, in which
there is no precipitate, were tried, and in some
experiments bacteria were spread on the surface of

Table 4. Zero Point Mutations Induced by Pyronin Y

Cone, of Time Method

Pyronin Y of of

(%) Exposure Treatment

No. In-
Survival duced Mu-
(%) tations/10 8
Survivors

0.5 20min. in agar 43.0 10.6

0.01 5hr. in liquid 32.2 22.4

0.075 1 hr. in liquid 18.6 14.0

0.75 lhr. in agar 18.1 22.8

0.5 2§hr. in agar 12.5 33.7

0.05 3hr. in liquid 3.9 36.0

2.0 30min. in agar 2.5 41.3

0.5 15min. in agar 0.16 62.7

0.75 2hr. in agar 0.11 150.0

0.5 2hr. in agar 0.03 100.7

0.025 4hr. in liquid 0.01 1080

agar containing pyronin, were washed from the
surface, and a concentrate of the wash was assayed.
The variability of the results was not overcome by
these modifications in technique, and the basis of
the difficulty is not yet understood. The results
obtained with pyronin, at various concentrations,
and under various conditions of treatment, are sum-

Table 5. Number of B/r/1 Mutants in Cultures
Exposed to 1% Methyl Green

Time of Experiment #1

Cul- Ex- Sur- No.

ture posure vival Mut.

to 1% (%) per 10"

Experiment #2
Sur- No.

vival Mut.

(%) per 10*

Control — —

Exp. 1 hr. 40

Exp. 2hr. 11

Exp. 3hr. 1.1

Exp. 4hr. 0.2

7.7

10.2

5.8 34 10.1

8.1 7 8.6

7.2 1.1 11.2

6.9 0.1 9.9

marized in Table 4. All results obtained were con-
sistent in indicating a higher proportion of mutants
among the survivors of treated bacteria than among
controls. Tests for selective killing of mutants
showed no difference in the sensitivity of B/r and
B/r/1 to pyronin. Although these results are to be
regarded as preliminary and tentative, they suggest
that pyronin is active as a mutagenic agent. The
problem of variability will have to be solved before
any detailed quantitative analysis can be made.

Pyronin has not yet been tested on Drosophila,
and there is no evidence as to its specificity in
inducing mutations.

Toxicity tests with methyl green showed that

:2 5

INDUCED MUTATIONS IN E. COLI

263

99% of the exposed bacteria were killed in about
three hours by a 1% solution in distilled water.
Table 5 gives the results of two experiments with
methyl green, in which the survival was as low as
0.2%, with no increase in the proportion of mutants
among the treated bacteria. These results were con-
firmed by repeated tests. Within the limits of the
sensitivity of the technique, which permits the
detection of induced mutants under conditions re-
sulting in the destruction of 99.9% of the treated
cells, methyl green seems to show no mutagenic
activity.

3. Acriflavine
Neutral acriflavine, or euflavine, is an acridine
dye used therapeutically as a bacteriostatic agent.
Mcllwain (1941) has shown that the inhibition of
bacterial division caused by acriflavine can be re-
versed by adding polymerized yeast or thymus
nucleic acids. He also showed, in vitro experi-

Table 6. Zero Point Mutations Induced by Acriflavin

Cone.

of
Acriflavin

Time

of

Exposure

Survival

No. of
Induced Mu-

(%)

tations/10 8
Survivors

.02

3hr.

50.4

8.0

.01

2|hr.

12.8

16.5

.01

4hr.

6.5

54.1

.05

2hr.

2.5

24.7

.05

2hr.

1.1

60.6

.05

3hr.

0.4

51.3

.05

4hr.

0.08

400

.05

4i hr.

0.02

420

.05

4hr.

0.005

1540

.05

4hr.

0.0007

4000

ments, that stable complex salts are formed by
acriflavine and nucleates. There is again no basis
for assuming specificity in these effects, since a
number of other compounds, including certain
amino acids, were also found to be effective in
reversing the inhibition brought about by acri-
flavine.

Table 6 gives a summary of results obtained with
acriflavine. Although reproducible results can be
obtained if all factors are carefully standardized,
acriflavine presents certain difficulties which stand
in the way of quantitative investigation. In the
presence of large numbers of bacteria, some of the
dye precipitates at high concentrations, and the
precipitate goes gradually kto solution as the tubes
are incubated. Thus it is likely that the effective
concentration changes during the exposure. This
difficulty can be overcome to some extent by using
low concentrations and longer periods of exposure.
Results with acriflavine consistently indicate an
increase in the proportion of mutants among sur-
vivors of treated cultures, and suggest that this com-
pound is active in inducing mutations.

Demerec et al (1946) tested the action of acri-
flavine on Drosophila, prior to the experiments re-
ported here, and obtained negative results. More
recently, Demerec (unpub.) has retested acriflavine,
using an improved method of administering the
chemical, and has found that it is active in inducing
lethal mutations in Drosophila.

Discussion

The experiments described above have indicated
that exposure to three out of the four chemicals
tested, at concentrations sufficiently toxic to kill
all but a small fraction of the treated bacteria,
results in a heightened proportion of mutants among
the survivors. Since B/r/1 was found to be no more
resistant than the nonmutant strain B/r to each of
the chemicals, it has been concluded that simple
selection is not responsible for this effect. Sodium
desoxycholate, pyronin and acriflavine are there-
fore regarded as mutagenic chemicals, although the
results for the latter two compounds are only pre-
liminary. Methyl green is apparently unable to
induce mutations to phage-resistance.

Although a single specific phenotype, resistance
to bacteriophage Tl, was the genetic character used
in these experiments, it is likely that at least two
separate mutations were involved, since B/r/1 mu-
tants are known to fall into two distinct classes,
differentiated by such secondary characters as
colony size, cross resistance to another bacterio-
phage and growth factor requirements. No attempt
was made to separate the two mutant types in these
experiments, and the frequencies observed are prob-
ably the sums of the two independent mutations.

Because of the lack of specificity in the chemical
action of the compounds tested, no attempt can
be made at present to relate their effectiveness as
mutagenic agents to the properties for which they
were selected. The high incidence of positive results
obtained, in 3 out of 4 compounds examined, is also
difficult to interpret at the present time. Further
experiments are required to determine whether these
results are due to a particularly fruitful or fortunate
basis of selecting chemicals, or, as appears more
likely, whether mutagenic action may prove to be
more common among biologically active compounds
than has hitherto been believed.

Although bacteria are becoming increasingly use-
ful as material for genetic investigations, it is still
necessary to be cautious in generalizing results
obtained exclusively in bacterial studies. Thus, the
fact that sodium desoxycholate and acriflavine ap-
pear to induce mutations in Drosophila as well as in
E. colt is an important contribution toward the
validation of the techniques used in these experi-
ments. In addition to confirmation provided by tests
on other organisms, the use of other mutations in
E. coli, entirely independent of the phage-resistance
system, would constitute a valuable check on results
obtained with nonspecific mutagens. Experiments to

126

264

EVELYN M. W1TKIN

determine the effects of chemicals on the frequency
of reverse mutations from biochemically deficient
mutants of E. coli are being planned with this need
in mind.

The use of phage-resistance in E. coli as material
for the investigation of chemically induced muta-
tions offers certain unique possibilities, in addition
to its value as a screening test. The analysis of
dosage-effect relations, as well as the quantitative
investigation of the delayed expression of induced
mutations characteristic of radiations and mustard
gas, can be carried out with this material. It may
also be possible to approach another interesting
aspect of induced mutations, the comparison of the
effectiveness of mutagenic chemicals on resting and
dividing cells. It must be remembered, however, that
bacteria are relative newcomers to the laboratory of
the geneticist, and that a longer and more intimate
acquaintance may be required to establish the re-
liability of these organisms as tools for the study of
the broader problems of heredity and mutation.

Summary

Four compounds were tested for mutagenic
activity in E. coli. The techniques used involved
suspending resting bacteria in solutions of the
chemicals, under conditions resulting in the death
of about 99% of the treated cells. The number of
mutants resistant to a bacteriophage, Tl, per 10 8
survivors of a treated culture was compared with
the number per 10 8 untreated bacteria from the
same culture. Only mutations expressed phenotypi-
cally before division of the exposed individuals
were detected.

The number of mutants per 10 8 survivors was
found to be higher in cultures treated with sodium
desoxycholate, pyronin and acriflavine than in un-
treated samples of the same cultures. No such in-
crease was obtained with methyl green. Since mu-
tants and nonmutants were shown to be equally
sensitive to the toxic action of each of the com-
pounds, it has been concluded that selection is not
responsible for these results. Sodium desoxycholate,
pyronin and acriflavine are considered, therefore, to
ue mutagenic, while methyl green is not.

The number of mutations induced by sodium
desoxycholate is directly proportional to the time
of exposure to a 5% solution of this compound.

Acknowledgments
The author wishes to acknowledge the helpful
suggestions of Dr. M. Demerec, and the efficient
assistance of Miss Marion Crippen in many of these
experiments.

References

Alloway, J. L., 1933, Further observations on the use of
pneumococcus extracts in effecting transformation of
types in vitro. J. exp. Med. 57: 265-278.

Auerbach, C, and Robson, J. M., 1944, Production of mu-
tations by allyl isothiocyanate. Nature, Lond. 154: 81-82.

Avery, O. T., MacLeod, C. M., and McCarty, M., 1944,
Studies on the chemical nature of the substance inducing
transformation of pneumococcal types. Induction of
transformation by a desoxyribonucleic acid fraction iso-
lated from pneumococcus Type III. J. exp. Med. 79:
137-158.

Brachet, J., 1940, La detection histochimique des acides
pentosesnucleiques. C. R. Soc. Biol., Paris 133: 88-90.

Demerec, M., 1946, Induced mutations and possible mechan-
isms of the transmission of heredity in Escherichia
coli. Proc. Nat. Acad. Sci., Wash. 32: 36-46.
1943, Mutations in Drosophila induced by a carcinogen.
Nature, Lond. 159: 604.

Demerec, M., and Latarjet, R., 1946, Mutations in bacteria
induced by radiation. Cold Spring Harbor Symp. Quant.
Biol. 11: 38-50.

Demerec, M., Latarjet, R., Luria, S. E., Oakberc, E. F.,
and Witkin, E. M., 1946, The gene. Yearb. Cam. Inst.
45: 143-157.

Fieser, L. F., and Newman, M. S., 1935, The choleic acids
of certain carcinogenic hydrocarbons. J. Amer. Chem.
Soc. 57: 1602-1604.

Horowitz, N. H., Houlahan, M. B., Huncate, M. V., and
Wricht, B., 1946, Mustard gas mutations in Neurospora.
Science 104: 233-234.

Luria, S. E., and Delbruck, M., 1943, Mutations of bacteria
from virus sensitivity to virus resistance. Genetics 28:
491-511.

McIiavain, H., 1941, A nutritional investigation of the anti-
bacterial action of acriflavin. Bio-chem. J. 35: 1311-
1319.

Mtrsky, A. E., and Pollister, A. W., 1946, Chromosin, a
desoxyribose nucleoprotein complex of the cell nucleus.
J. gen. Physiol. JO: 117-148.

Muller, H. J., 1947, The gene. (Pilgrim Trust Lecture).
Proc. Roy. Soc. Lon. (B) 134: 1-37.

Tatum, E. L., 1946, Induced biochemical mutations in bac-
teria. Cold Spring Harbor Symp. Quant. Biol. 11: 278-
284.

Wteland, H., and Sorce, H., 1916, Untersuchungen iiber die
Gallensauren. II Mitteilung. Zur Kenntnis der Cholein-
saure. Hoppe-Seyl. Z. 97: 1-27.

Witkin, E. M., 1947, Genetics of resistance to radiation in
Escherichia coli. Genetics 32: 221-248.

Discussion
Hotchkiss: In examining the most interesting
results of Dr. Witkin, I have been struck by the
fact that the actual recovery of mutants from a fixed
number of original cells, which is the count actually
observed in the laboratory, is maximal in the high
survival range and decreases markedly as the per-
centage of total survivors drops. Since the number
of mutants does not however decrease to the extent
that the total population does, it tends to rise
rather considerably, if calculated over to the basis of
survivors. It has been recognized by the author that
to view this latter calculated value as a "mutation
rate" involves the assumption that the phage-
resistant mutants do not have increased survival
value in solutions of the chemical mutagens used.
Indeed for this reason, trouble was taken to demon-
strate that stock phage-resistant mutants do not
survive longer than non-mutants in desoxycholate.

!27

INDUCED MUTATIONS IN E. COLI

265

It appears, however, that the actual mutants isolated
after desoxycholate, pyronin or acriflavine treat-
ment were not tested for resistance to these chemi-
cals. It should be pointed out that only a modest
degree of chemical-resistance in only a portion of the
isolated mutant clones would allow the observed
increased proportion of mutants among the sur-
vivors.

To bring out this point more clearly I may be
permitted perhaps to recalculate one of the tables
from the lantern slide dealing with acriflavine-
treated cultures. The other tables would give very
similar results. From the figures given it is pos-
sible to recalculate the actual yield of mutants ob-
served from each 10 8 bacteria of the original cul-
ture. While this column has to be calculated from
another in this instance the point must be stressed
that it was originally an observed part of the ex-
perimental data and tabulating it involves no par-
ticular assumptions.

the individual cells affected within a population or
to the characters, genetic or somatic, affected within
a given cell. Once a particular cell, by virtue of its
permeability, physiological state, internal pH, or
other inherent or transient property, has accumu-
lated in its interior sufficient chemical mutagen to
suffer a mutation at one locus, we would indeed
expect that many other changes might become
manifest. This picture would furnish one plausible
basis for the conservative viewpoint that only a
moderate amount of chemically-induced mutation
has taken place, and that this has been to some
degree associated with temporary or inheritable re-
sistance to the chemicals, a finding not to be ex-
pected to any great extent in spontaneous or irradia-
tion-induced mutants.

Witkin: Dr. Hotchkiss has emphasized the fact
that the absolute recovery of mutants falls off, ex-
cept in the range of high survival, as the exposure
to the chemical is extended. This is true not only

Corrected for

Yield

of mutants

Per cent of

Per cent

Mutants

per

estimated

per

10 s original

Yield expressed as

survival

10 8 survivors

"background"

cells

original mutants

100

5

5

(100)

50

8

13

6.5

ca.130

6.5

51

56

3.6

72

1

61

66

0.7

14

0.02

420

425

0.085

1.7

o.oos

1540

1545

0.077

1.5

0.0007

4000

4000

0.028

0.6

It will be noted that, considered from this point of
view, the most striking production of mutants was
in the low-killing range where an increase of some
30% was observed. Detected as it presumably was
in counts of some one or two hundred colonies, this
increase represents at present the only positive in-
dication, free of assumptions, that new mutants have
been produced. The particular figure is of course
rather dependent upon the value taken for the
background content of mutants; what was under-
stood to be an average figure was used here. In any
case, where the killing has been greater, the "yield"
of mutants has also decreased well below the back-
ground value.

Although several theories or explanations come
to mind to account for the findings, it should be
pointed out that these comments arise as it were
from experimental considerations only and are sug-
gested mainly as a more conservative way of ex-
pressing the data. When it has been demonstrated to
what degree, if any, representative mutants recovered
from the chemical mutagen solutions are resistant to
that mutagen, then it will be more safe to calculate
the actual yield of mutants, dead and alive, produced
from 10 8 original cells.

Finally, it may be pointed out that there is no
reason to suppose that mutations induced by chemi-
cals need show randomness with repect either to

for the chemicals studied in these experiments, but
for radiations as well. An absolute increase of mu-
tants is to be expected only in the case of nontoxic
mutagens, or in the case of agents whose mutagenic
activity is relatively much greater than its toxicity.
No such case is known as yet, and it seems unlikely,
at least for mutagens acting in an aspecific manner,
that such cases are theoretically possible, since in-
duced lethal mutations may be expected to exceed,
as a class, any single phenotype under consideration.
Nevertheless, the fact that absolute increases are
not observed, except where survival is high, does
create the necessity for caution in interpreting the
results.

First of all, it can be pointed out that, for each
of the three chemicals considered to be mutagenic,
absolute increases in the recovery of mutants were
observed under special circumstances. In the case
of acriflavin and pyronin, the actual number of
mutants recovered from a given number of original
bacteria increases significantly when the killing does
not exceed 50% to 60%. In the case of sodium
desoxycholate, it is possible to use low concentra-
tions where the maximal killing is reached by 24
hours, and there is no further killing between 24 and
48 hours of exposure. The number of mutants, how-
ever, does increase during this period, resulting in
actually higher mutant counts on the experimental

128

266

EVELYN M. WITKIN

plates at 48 hours, starting with a sample of the same
size. Thus, for each chemical it is possible to show
that new mutants do actually appear.

Dr. Hotchkiss suggests that the induced mutants
may differ from the spontaneous mutants in re-
sistance to the chemical used to induce them. In
the case of desoxycholate, where there is a linear
relation between the number of induced mutations
and the time of exposure to the chemical, in spite
of the fact that the rate of killing changes markedly
during the period of exposure, it is difficult to see
how selection can be involved. If the increase in
number of mutants were due in large measure to
the selective survival of a few mutants induced early
in the treatment, surely the rate of increase in the
proportion of mutants with time could not be total-
ly independent of the survival curve.

Dr. Hotchkiss suggests that the mutant colonies
actually obtained after treatment with the chemical
should be compared with the nonmutant stock with
respect to resistance to the chemical. It seems to
me that this would not constitute a critical test
of the hypothesis Dr. Hotchkiss has raised. First
of all, a certain amount of the chemical is carried
over onto the plates, and then mutant colonies de-
velop in the presence of a very low concentration
of the chemical, negligible in terms of toxic or
mutagenic activity, but possbily sufficient to permit
selection during the growth of the colony of spon-
taneous variants better able to withstand the action
of the substance. Thus, to find that the mutant
colonies obtained after exposure to the chemical are
more resistant to its toxic action than the untreated
nonmutant stock, is not a fair test of the situation
before the development of the colony on the plate.
Should the mutant colonies prove no more resistant
to the action of the chemical than the untreated
non-mutant stock, the possibility of a temporary
resistance, as Dr. Hotchkiss has suggested, can be
raised. I cannot at present see any way of testing
this possibility critically, since it is based on the
state of a recently induced mutant, during treatment
with the chemical, and a transient resistance to the
chemical can be postulated. One possible way to
avoid considerations of this kind might be to start
out by developing a strain which is maximally re-
sistant to the chemical to begin with, and use this
strain throughout.

The best evidence that new mutations are in fact
appearing throughout the exposure, to my mind,
will be based upon comparison of survival curves
and curves of induced mutations as a function of
exposure time. If, as has been found for desoxy-
cholate, the proportion of mutants increases at a
constant rate, regardless of changes in killing rate,
the possibility that recently induced mutants sur-
vive selectively will become quite remote.

Kxjrnick: Dr. Witkin has expressed some sur-
prise at the fact that 3 out of 4 tested chemicals
proved mutagenic in her experiments and suggested

that it will be found that mutagenicity is a com-
mon property of chemical reagents. This appears
to be a not unlikely prognostication. While not
wishing to deny the possibility of direct chemical ac-
tion to induce mutations, I should like to present
an alternative hypothesis here.

Dr. Witkin searches only for a specific mutation-
resistance to Tl phage — not random mutations. She
observes that this mutation is present spontaneously
in her controls in 5 to 10 per 10 8 bacilli. It is not im-
probable that this mutant contains a specific muta-
genic substance similar to that isolated by Avery for
pneumococcic strains and by Boivin for a colon
bacillus strain. Dr. Witkin has observed that the
mutation effect of a given chemical corresponds
more nearly to its killing effect than to any other
factor, such as concentration of the chemical or
the time of exposure. It is obvious that if, as she
has shown, the mutant and the original strains
are equally sensitive to the lethal effect of the drugs
used, the higher the percentage of lethality, the
greater will be the chance of destruction of some of
the 5 to 10 spontaneous mutants per 10 8 bacteria.
The autolysis of the killed mutant (or penetration
through its wall whose permeability is apt to differ
from the living cell) will free the specific mutator
substance (for Tl phage resistance) into the
medium where it may cause the specific, directed
mutation of other organisms. The nearer the killing
rate approaches 100%, the higher will be the con-
centration of this mutator substance, and so the
higher the mutation rate.

Dr. Witkin has observed that after about 99.9%
of the bacteria have been killed, the killing rate
approaches zero, but the production of mutations
continues. She suggests that this may mean that the
lethal and mutagenic properties of her chemicals
are unrelated, and that the 0.1% of surviving cells,
while no longer susceptible to the lethal effect are
still susceptible to the mutagenic effect of the drug.
If the hypothesis presented herein were correct, we
could readily account for the continued occurrence
of mutations in the surviving bacteria by the fact
that they are continuously exposed to the liberated
mutator of the killed spontaneous mutants and so
continue to undergo directed mutation.

Of the 4 chemicals studied by Dr. Witkin, only
methyl green was inert as a "mutagen," although
its lethality in the concentrations used was compar-
able to that of the others. Now if, in fact, as is
indicated by Boivin and Avery, desoxyribonucleic
acid is an essential component (perhaps the only
component, but this argument is not germaine to
this discussion), a chemical which combined spe-
cifically with this nucleic acid might inactivate
the mutator substance. Methyl green is such a com-
pound (as opposed to pyronin, for example). Thus,
we may account for the lethality of this compound
without the expected production of mutations.

On the other hand, some chemicals, which might

129

INDUCED MUTATIONS IN E. COLI

267

themselves not inactivate the mutator substance,
might nevertheless render it ineffective by altering
the cell surface of the bacteria so as to deny access
of the mutator substance to the interior of the cell.
Such a reagent may be chloroform, which Dr. Beale
has found ineffective as a mutagen despite its lethal
effect.

Thus, the induction of mutations by chemicals
which do not themselves directly interact with the
gene would be a function of their lethality, the de-
gree of interaction between the chemical reagent
and the liberated mutator substance, and the degree
to which the permeability of the cell wall is altered.
In this manner we may account for differences in
"mutagenicity" of compounds with identical killing
rates.

This hypothesis is susceptible of experimental
test in several ways. I shall suggest two such experi-
ments. One would be to test (a) the effect of lethal
reagents which do not induce mutations (particu-
larly methyl green) on the mutator substances of
Boivin and Avery before addition to the culture
substrates to determine whether inactivation does in
fact occur, and (b) the effect of adding the chemical
(particularly chloroform) to the appropriate cul-
ture before adding Boivin's or Avery's substance to
determine whether a reagent which does not itself
inactivate the mutator substance but nevertheless
fails to produce mutations, prevents the effect of
the mutator substance by its action in the otherwise
susceptible cells. Another would be to centrifuge
off the surviving bacteria after 99.9% had been
killed, wash them several times in water and then
resuspend them in a medium containing the chemi-
cal under investigation (let us say pyronin). Con-
trols would consist of bacteria left in the original
medium and bacteria resuspended in water. If the
bacteria resuspended in a fresh solution of the rea-
gent and those resuspended in water show a signifi-
cantly lower rate of continued mutation than those
permitted to remain in the original suspension (pro-
vided the killing rate remained near zero in all
samples), one could conclude that the mutagenic
factor was not the chemical reagent, but rather
some material which was present in the first super-
nate alone — this could only be derived from the
cultures themselves. If, on the other hand, muta-
tions continued to occur in all three test samples at
the same rate, the conclusion would again be that
the mutagenicity was probably not a direct func-
tion of the chemical (since one suspension is re-
suspended in water), but that the mutator sub-
stance was adsorbed to the washed bacteria or had a
significant lag time between its action on the bac-
teria and the emergence of the mutant. The third
result which might occur, namely that the sample
resuspended in a fresh solution of the chemical
showed a continued mutation rate equal to that of
the original suspension and greater than that of the
suspension in water, would disprove the hypothesis

I have proposed above and demonstrate the capacity
of the chemicals themselves to produce mutations,
presumably by direct attack on the gene.

I do not by any means, I repeat, wish to imply
that chemical mutations due to the direct action of
a chemical reagent on the gene do not occur. Quite
the contrary, there is good reason to believe that
such effects are possible. I wish only to interject a
word of caution in interpreting the induction of
mutations by a considerable number of reagents,
particularly when working with such material as
bacteria where the opportunity for interaction be-
tween large numbers of organisms exists. This pit-
fall may be avoided by suitable controls, as sug-
gested above, to distinguish between direct inter-
action with the gene and the indirect effect due to
autolysis.

Witkin: It is, of course, possible that phage-
resistant mutants of strain B/r contain specific
mutator substances, similar to the transforming fac-
tors of Avery and Boivin. The addition of filtrates
of heat-killed cultures of B/r/1, however, does not
increase the number of resistant mutants in cultures
of B/r. Numerous other attempts to detect evidence
for transformation in this system have failed. This,
in addition to the fact that the number of induced
mutants obtained is independent of the background
number, over a range of one to fifty per 10 8 bac-
teria, seems to me to render Dr. Kurnick's hy-
pothesis unlikely.

Bryson: In the event that we continue to find
numerous and unrelated chemicals that are able to
induce mutation, it may be of particular interpretive
value to study with care those substances that are
not mutagenic. It would also be desirable that more
simple chemical substances such as inorganic salts
be surveyed as mutagens, even as Drs. Greenstein,
Carter and Chalkley have surveyed them for effect
on enzymatic degradation of nucleic acid. One might
then bring to bear what is known of the effects of ions
on cells, and compare mutagenesis with what has
been established about permeability and relative
toxicity of various materials whose fate as reactants
in living systems has already been the subject of
extensive biochemical investigations. The problem of
toxicity itself presents a difficulty in the classification
of chemical mutagens since toxicity may act as a
limiting factor in exploring mutagenic potential. For
example, we have found that the induction of mu-
tations of phage resistance in E. colt by bis-beta-
chloroethylamine hydrochloride is most readily per-
formed on a strain of cells that has been through
twelve consecutive exposures to the inducing agent
with repeated selection of survivors for resistance to
the chemical. The number of zero point mutations
induced by nitrogen mustard in stock B 12 /M at a
survival level of 0.007% is 240 per 10 8 . It is always
possible that a chemical like methyl green that Dr.
Witkin has described as negatively mutagenic could
be made to induce mutations in a strain of cells

130

268

EVELYN M. WITKIN

selected for resistance to its toxic effects and there-
fore capable of surviving in relatively high concen-
trations.

No attempt has been made to approach a compa-
rable killing value with another mutagenic agent
(Zephiran chloride) because this quaternary am-
monium compound shows a selective toxic effect
against B/l at high concentrations, a process which
incidentally would tend to make the detection of
mutations a more difficult process. An experiment
using 1:14,000 aq. Zephiran chloride (a non-selec-
tive concentration) is shown in Table 1. A definite
mutagenic effect is observed if periodic sampling of
the same culture is performed.

Table 1. Periodic Assay of Cells Incubated with

Zephiran Chloride at 37° C. and Tested for

Point Mutations to Phage Resistance

Time of Bacteria B/l per 10" Per cent
Assay per cc. bacteria survival

Zero
point
mutants

5.6X10 10

4.2/10 8

—

—

3 min.

2.2X10'°

8.1/10 8

39%

3.9/10 8 *

10 min.

2.0X10 10

8.6/10 8

36%

4.4/10 8

30 min.

1.1X10 10

13.3/10 8

20%

9.1/10 8

100 min.

6.4X10'

21.1/108

11%

16.9/10*

♦8.1/10 8

-4.2/10 8 .

In another experiment yielding 19% survival the
zero point mutations numbered 24 per 10 8 .

The intergeneric variation in sensitivity of micro-
organisms to toxic agents is perhaps of more gen-
eral importance than experimentally induced inter-
strain differences as a factor to be kept in mind
when screening of possible mutagens is contem-
plated. Unless one adopts the relatively extreme
position that cell death in the presence of chemicals
is the consequence of induced lethals it is not un-
duly speculative to assume that cells highly re-
sistant to deleterious chemical effects on extragenic
processes will be best suited for a demonstration of
mutagenesis. If analysis of toxicity is to be per-
formed as a differential survival test it may also be
necessary with some chemicals to extend the tests
over the killing range included within limits of the
experiment.

As Dr. Witkin has observed, the span of our ex-
perience with induced phage resistance as a genetic
tool has not been of sufficient length to evaluate its
place in the general scheme of experimentally in-
duced mutations in more complex biological sys-
tems. Investigators using the method may find
themselves in the paradoxical situation that each
new success contributes to a failure, that is, to the
widening of a gap between mutagenesis of micro-
organisms and of higher forms of life. This would
inevitably decrease the value of bacterial studies in
interpreting mutation among higher organisms. Un-

til other mutations in bacteria have been studied
by similar methods and until the growing list of
chemical mutagens has been used on more familiar
genetic material and placed within some kind of
quantitative and qualitative limits, it will be im-
possible to judge the real significance of Dr. Wit-
kin's most stimulating and capable study.

Herskowitz: The work of Dr. Witkin demon-
strates a facile bacterial method for identifying cer-
tain groups of chemicals as mutagens. She has al-
ready pointed out that by using this technique nega-
tive mutagenic action by a chemical is not con-
clusively demonstrated. However, it may well be
that those chemicals screened out by the bacterial
technique as being non-mutagenic will be im-
portant in the long run. Therefore, it seems ad-
visable to supplement this technique with others
which may permit a more sensitive test for mu-
tagenic action.

For the detection of the action of a chemical on
nucleic acids and nucleoproteins the use of Droso-
phila sperm seems most suitable. There are several
reasons for this. In spermatozoa there are a mini-
mum of cytoplasmic substances which might inter-
fere with a chemical affecting genie nucleoprotein.
Moreover, even though a chemical is not specific
for nucleoprotein it may yield positive results more
readily in sperm than in other cells. There is ex-
cellent evidence from genetic and irradiation experi-
ments that genie nucleoprotein may be drastically
changed without killing the sperm cell; this permits
the utilization of such sperm for fertilization with
the subsequent detection of lethal genes and
chromosome rearrangements in addition to other
types of inherited changes. In Drosophila it is also
possible accurately to localize inherited changes in
the chromosomes. Dr. Demerec no doubt had these
advantages in mind when he developed the aerosol
technique.

As Dr. Witkin has mentioned, there is no con-
clusive proof that aerosols always reach the sper-
matozoa in the testes. Again, negative mutagenic
results with a chemical may be misleading. There-
fore, a technique which would directly treat
Drosophila sperm with chemicals is highly desirable.
Accordingly, an investigation was initiated to dis-
cover if chemicals could be injected into the vagina
of adult females, which would then copulate, and
thereby expose the sperm to the respective
chemicals.

A vaginal douche technic was successfully worked
out, and is described in detail in a recent published
note (Herskowitz, Evolution 1: 111-112, 1947).
The method was tried with methyl bis amine hydro-
chloride at concentrations of 0.2% and 10.0%, with
positive results.

The data establish the practicality of the vaginal
douche technique for the detection of chemical
mutagens. There are two advantages of this method.
Low concentrations of mutagens may be effective

131

INDUCED MUTATIONS IN E. COLI

269

because of the direct contact of the chemical with
the sperm; high concentrations of chemicals can
be used without killing the organism, since a lo-
calized part of the body rather than the whole indi-
vidual is treated. These advantages point to the
possibility of a chemical analysis of the processes
leading to gene changes as well as the analysis of
the processes involved in direct changes in genie
nucleoprotein. With such objectives in mind, the
search for and study of the mutagenic activity for
many compounds would most efficiently be investi-
gated by using the bacterial or fungus method first,
then the use of the aerosol method, and finally, if
necessary, the vaginal douche technic.

Zamenhof: I wonder if Dr. Witkin would be
willing to comment on qualitative differences be-
tween spontaneous mutations and mutations in-
duced by chemicals, in addition to the quantitative
ones. As Dr. Witkin has pointed out, the spon-
taneous mutations do not seem to occur in the
non-dividing cells (see also Zamenhof, Genet. Soc.
Rec. 13: 41-42, 1944). The mutation ratio in this
case can therefore be defined as the number of
mutations per cell division. On the other hand, the
mutations induced by chemicals seem to take place
even in the absence of cell divisions; in this case the
mutation ratio can be defined as the number of

mutations per cell per unit of time. Thus the two
phenomena seem to be different. What is your
opinion on this subject?

Witkin : The fact that mutations can be induced
in resting bacteria by radiations and chemicals,
whereas spontaneous mutations seem to occur only
in dividing cells, certainly suggests an important
difference between induced and spontaneous muta-
tions, although the end product of the two processes,
the mutated genes, need not necessarily differ in a
fundamental way. Chemical or mechanical errors of
duplication, or slips in some metabolic cycle inactive
in resting cells, may be the only natural circum-
stances capable of bringing about mutations. Thus,
the failure of spontaneous mutation to occur in rest-
ing cells may result from the relatively inert and
quiescent state of the genie material, and of the
metabolic activities associated with cell division,
rather than to a greater degree of inherent stability.
Zero point mutations may be due to the direct
action of chemicals or radiations on the gene, or
more indirectly, to the stimulation of cellular
activities normally involved in the production of
spontaneous mutations. A careful comparison of the
properties of mutants arising spontaneously and by
induction may throw some light on this important
question.

THE EFFECT OF METABOLITES UPON GROWTH AND VARIATION
OF BRUCELLA ABORTUS

ROBERT J. GOODLOW, LEONARD A. MIKA, and WERNER BRAUN

Camp Detrick, Frederick, Maryland

Received for publication June 22, 1950

In previous work with Brucella abortus (Braun, 1946) it has been demonstrated
how mutation and selection can account for the progressive establishment of
nonsmooth types in smooth broth cultures frequently observed during prolonged
periods of growth. Evidence was presented that indicated that such population
changes, common to most bacterial species and often referred to as dissociation
(Braun, 1947a), involve the spontaneous occurrence of a small number of un-
directed variants (mutants) and their subsequent establishment within a popu-
lation under the control of inherent and environmental factors governing popu-
lation dynamics. It was also illustrated how competition between spontaneously
arising mutants with different selective values can produce the appearance of
apparently cyclic, successive, orderly changes (Braun, 19476).

In the earlier studies it was recognized that variants do not establish them-
selves within growing populations until competitive conditions exist. Such com-
petitive conditions presumably occur when the number of viable cells (plate
counts) reaches a maximum, whereas the total number of cells (direct cell
counts) continues to increase. However, the factors responsible for the limitation
of the size of the viable population (i.e., the number of cells that retained the
ability to multiply), or the "M concentration" of Bail (1929), have remained
obscure. A paucity of information also exists regarding the possible influence
of metabolites produced by one cell type upon the establishment of variant
types and the role that such "association factors" (Braun, 1947a) play in popu-
lation changes. Zamenhof (1946) and especially Ryan and Schneider (1949)
obtained data in their studies with Escherichia coli that suggested an important
role for such postulated factors in bacterial variation. Etinger-Tulczynska (1932),
Neufeld and Kuhn (1934), and Mohr (1934) are among the earlier workers who
assumed such effects but were unable to demonstrate their metabolic nature.
Similarly, in studies with complex media, Braun (1946) was unable to demon-
strate any effect of old culture filtrates upon the establishment of variants in
B. abortus cultures. In contrast, in recent studies with B. abortus, in which paper
partition chromatography and a simple synthetic medium containing DL-aspar-
agine as the only amino acid source were utilized, the effect of specific cell
metabolites of one type upon the establishment of variant types became clearly
evident, and the manner in which such metabolites restrict the "M concentra-
tion" of the cells that produce them became obvious. The following data will
describe these observations.

291

[Reprinted by permission of The Williams & Wilkins Company from Journal of
Bactkriolocy 60: (3) 291-300, September, 1950]

133

292 R. J. GOODLOW, L. A. MIKA, AND W. BRAUN [VOL. 60

METHODS AND MATERIALS

Smooth and rough clones isolated from the C0 2 -requiring strain 6232 of B.
abortus were used throughout. For the preparation of inocula the growth from
a 24-hour modified tryptose agar slant culture (tryptose agar supplemented
with glucose, thiamine, and Fe — McCullough et at., 1947) was washed off with
sterile Gerhardt and Wilson (1948) synthetic medium (G-W medium) and ad-
justed in the Coleman spectrophotometer to 70 per cent light transmittance.
To each of a series of flasks containing 100 ml of G-W medium was added 1.0
ml of the adjusted suspension of organisms, and the cultures were then incubated
at 37 C under approximately 10 per cent C0 2 . At regular intervals throughout
the periods of observation samples for the determination of viable cell counts
were removed from the cultures, and serial dilutions, prepared in tryptose saline
(0.1 per cent tryptose, 0.5 per cent NaCl), were mixed with modified tryptose
agar. In addition, samples were streaked on "2-1" agar plates 1 for the deter-
mination of the percentage of nonsmooth variants (Braun, 1946).

In the determination of the effects of the addition of old culture filtrates to
freshly inoculated smooth cultures, Seitz filtrates were employed. Ten ml of
such filtrates were added to 90 ml of G-W medium, inoculated with 1.0 ml of a
tryptose-saline suspension of smooth Brucella abortus, and incubated at 37 C
under 10 per cent C0 2 .

One-dimensional ascending paper chromatograms of Seitz filtrates were pre-
pared according to the technique used by Home and Pollard (1948). The strips
were suspended in 82 per cent distilled phenol for 24 hours and then developed
with 0.1 per cent ninhydrin solution in 85 per cent butanol.

RESULTS

The growth of smooth B. abortus, 6232, in G-W synthetic medium (figure 1)
was characterized by a steady increase in viable cell counts until the fourth
day of incubation, after which the number of viable organisms steadily decreased
for a period of 6 to 7 days. After the tenth day of incubation a second rise in
viable cells occurred. This increase reached its peak on the twentieth day. The
second peak usually exceeded that of the initial maximum viable level. Examina-
tion of 2-1 agar plates revealed that during the first 10 days of incubation the
population was smooth. Coincidentally with the second rise in viable cells, there
was a steady increase of nonsmooth variants (predominantly rough types),
which, at the time of the second peak of increase in viable cells, frequently con-
stituted over 90 per cent of the population (table 1).

The effects of adding to freshly inoculated smooth cultures Seitz filtrates of
24-day-old cultures of B. abortus, in which the original smooth population had
been replaced by nonsmooth variants, were then investigated. The addition of
10 per cent nitrate was found to produce a marked effect on the subsequent

1 Composition of 2-1 agar: 2.5 per cent Difco purified agar; 1 per cent Difco peptone;.
0.5 per cent NaCl ; 0.5 per cent Difco beef extract ; 2 per cent glycerol ; and 1 per cent glucose .
Adjusted to pH 7.4 before autoclaving.

134

1950]

GROWTH AND VARIATION OF BRUCELLA ABORTUS

293

5x10

I x 10

1x10

4x10*

92% ROUGH-

ALANINE
ASPARAGINE
ASPARTIC ACID

INITIAL INOCULUM:

SMOOTH ORGANISMS

ROUGH ORGANISMS

2 4 6 8 10 12 14 16 18 20
DAYS
Figure 1. Growth (viable cell counts) of smooth and rough B. abortus in synthetic me
dium.

TABLE 1

Growth and variation of smooth B. abortus in synthetic medium

DAY OF GROWTH

VIABLE CELL COUNT*

% NONSMOOTH VARIANTSf

8.4 X

10 6

1

6.9 X

10 6

2

2.8 X

10 7

3

11.3 X

10 7

4

13.5 X

10 7

5

11.8 X

10 7

6

11.9 X

10 7

7

10.2 X

10 7

8

8.6 X

10 7

9

6.0 X

10 7

11

—

2

14

9.1 X

10

6

19

12.4 X

10

76

23

19.6 X

10

92

* Averages of 6 plates.

t Predominantly rough type.

135

294

R. J. GOODLOW, L. A. MIKA, AND W. BRAUN

[VOL. 60

growth and variation of smooth cultures. The increase of the viable smooth
organisms was repressed, and an earlier and more rapid establishment of non-
smooth variants occurred (table 2). When such filtrates were added to freshly-
inoculated rough B. abortus cultures, only a slight inhibitory effect on the growth
of the rough organisms was observed. Subsequent experiments demonstrated
that the smooth-suppressing factor in filtrates was dialyzable and heat-resistant
(100 C for 5 minutes).

Paper chromatograms were then prepared daily on filtrates of actively grow-
ing smooth cultures in G-W medium. Until the eighth or tenth day of growth
these chromatograms showed no spots, other than those of asparagine and aspar-

TABLE 2

Growth and variation of smooth B. abortus in synthetic medium after addition of old culture

filtrate

DAY OF GROWTH

VIABLE CELL COUNT*

% NONSMOOTH VARIANTS!

8.8 X

10 8

1

9.2 X

10 8

2

5.1 X

10 7

3

5.3 X

10 7

4

6.3 X

10 7

<.01

5

6.7 X

10 7

6

6.5 X

10 7

1

7

5.8 X

10 7

2

8

5.0 X

10 7

2

9

6.8 X

10 7

10

11

—

64

14

16.2 X

10 7

91

19

16.7 X

10 7

91

23

16.4 X

10 7

90

* Averages of 6 plates.

t Predominantly rough type.

tic acid. After that period another spot with the Rf value of glutamine, thre-
onine, or alanine made its appearance. The use of collidine-lutidine 2 solvent
subsequently indicated that this spot was probably due to alanine. Further obser-
vations showed that this spot became more intense during the subsequent 10
days of growth. After this time no further differences in intensity were visible.
Because of the interesting correlation between the appearance in the culture
medium of this amino acid and the simultaneous increase in the establishment
of nonsmooth variants, the effects of adding glutamine, threonine, or alanine to
freshly inoculated smooth cultures were tested. Neither the addition of 0.5 mg
per ml glutamine nor 0.015 mg per ml of DL-threonine resulted in any appreci-
able change in growth and variation when compared with control cultures. The
addition of 0.5 mg per ml DL-alanine resulted, however, in a marked suppression

2 Thirty -five per cent collidine (mixed), 35 per cent 2,6-lutidine, and 30 per cent distilled
water.

136

1950]

GROWTH AND VARIATION OF BRUCELLA ABORTUS

295

of viable smooth organisms and enhanced the earlier and more rapid establish-
ment of nonsmooth variants. The curves of increase in viable cells of cultures

txlO

IxlO 8

9 1 % ROUGH . 92 % ROUGH ^^

— ~"\ ' *" —

^v s ^ — '

z
o

Is"

"*--< 2% ROUGH
^2% ROUGH

o

_J

Ui

o

/i

l

1 /

UI

_)

CO

LEGEND

>

NO FILTRATE ADDED (CONTROL)

IxlO 7

',_/

GROWTH WITH ADDED FILTRATE

•i «ir> 6

-

, ,

,,■■■■ i . i i < i i i < >

12 3

5 6 7 8 9 10 II 12 13 14 15 16 17 18 19 20 21 22 23 24
DAYS
Figure 2. Growth (viable cell counts) of smooth B. abortus in synthetic medium supple-
mented with filtrates from old cultures.

4 x 10°

>
IxlO 7

3x10'

■

' / m2$/

/ /

^ LEGEND

CONTROL
100 jj,q

NO dl-olonine
dl-olanine/ml

-1

-X-,-

200 (iq

dl-alanine/ml

i

500 /ig

dl-olonine/ml

■

— ■ — i —

1000 ft,q

dl-olonine/ml

10 II

13

12 3 4 5 6 7 8

DAYS

Figure 8. Growth (viable cell counts) of smooth B. abortus in the presence of different
concentrations of DL-alanine.

to which alanine had been added were strikingly similar to those of cultures to
which filtrates had been added (figures 2 and 3).

In order to determine a possible differential resistance of smooth and rough

137

296

R. J. GOODLOW, L. A. MIKA, AND W. BRAUN

[VOL. 60

types of B. abortus to the toxicity of alanine a series of flasks of G-W medium
(100 ml) were inoculated with smooth or rough organisms. To each culture
was added sufficient DL-alanine to yield final concentrations of 100 ng, 200 ng,

4x10*

I x t0°

-

_„.-»-o — o_o -_„_._

;/CZ>^

/ LEGEND

/ CONTROL -

NO dl-alanine

/ _.-o- ioo fiq

dl-olanine/ml

:/ — »— *- 200 fiq

dl-olanine/ml

- 500 fiq

dl-alonine/ml

7 -.-.- 1000 fiq

dl-alanine/ml

IxlO'

3xl0 6

I 2 3 4 5 6 7 8 9 10 II 12 13 14 15

DAYS
Figure 4- Growth (viable cell counts) of rough B. abortus in the presence of different con-
centrations of DL-alanine.

TABLE 3

Growth and variation of smooth B. abortus in synthetic medium containing different
concentrations of DL-alanine

1

CONTROL

(no alanine)

100 mG dl-
alanine/ml

200 MG dl-
alanine/ml

500 mg dl-
alanine/ml

1,000 \xG dl-
alanine/ml

DAY

Viable-
count
(X10«)

% non-
smooth
vari-
ants!

Viable

count
(X 106)

% non-
smooth
vari-
ants

Viable
count
(X 106)

% non-
smooth
vari-
ants

Viable
count
(X 106)

% non-
smooth
vari-
ants

Viable
count
(X 10=)

% non-
smooth
vari-
ants

2
4
6
8
10
15

6.2
60.0
79.0
20.0
22.0
44.0
117.0

3
11
22
42

6.3
61.0
57.0
13.0
22.0
72.0
120.0

1

6
14
43
57

6.7

54.0
24.0
13.0
63.0
70.0
121.0

2
19
36
83
86

6.7
58.0
14.0
17.0
101.0
80.0

3

61

63

6.5
45.0
19.0
13.0
69.0
65.0
71.0

5
54
57
88
82

* Averages of 6 plates.

t Predominantly rough type.

500 Mg, or 1,000 Mg per ml of medium. Upon repeated sampling of these cultures
during incubation it was found that alanine in a concentration of 100 ng per
ml or more exhibited a significant growth-depressing effect on the smooth type,
whereas the rough variant was unaffected until a concentration of 1,000 Mg of

138

1950]

GROWTH AND VARIATION OF BRUCELLA ABORTUS

297

alanine per ml of medium was reached (tables 3 and 4, figure 4). The resistance
of the rough variant to the toxicity of alanine was, therefore, approximately 10
times that of the smooth type.

It was also observed that during prolonged incubation of cultures in which
rough variants had become established alanine accumulation continued that
led to an eventual second decline of the viable count. This decline was then
followed, approximately 50 days after the start of the originally smooth cultures,
by a renewed increase in viable cells and a simultaneous establishment of another
smooth variant. This highly alanine-resistant variant, labeled S', proved to be
similar antigenically to the original smooth type (acriflavine test and aglutinin
absorption test), but when S' was plated in mixtures with the original S a distinct
difference in colony morphology was detectable.

Growth of rough B. abortus

TABLE 4
in synthetic medium containing different concentrations
DL-alanine*

of

CONTROL

(no alanine)

100 /iG DL-ALA-

NINE/ML

200 mg dl-ala-
nine/ml

500 HG DL-ALA-

nine/ml

1,000 flG DL-ALA-

nine/ml

Viable countt
(X10«)

(Viable count
(X10«)

Viable count
(X 10«)

Viable count
(X 10«)

Viable count
(X 10«)

3.7

3.3

3.3

3.1

3.6

2

83.0

83.0

67.0

67.0

62.0

4

146.0

168.0

140.0

100.0

52.0

6

145.0

118.0

108.0

42.0

8

132.0

130.0

114.0

107.0

63.0

10

96.0

111.0

83.0

65.0

56.0

15

66.0

51.0

54.0

62.0

44.0

* No variants were observed in these cultures up to 15 days,
t Averages of 6 plates.

DISCUSSION

These data demonstrate that the accumulation of a specific metabolite, alanine,
limits the increase in viable smooth cells and at the same time creates an en-
vironment favorable for the progressive establishment of nonsmooth (e.g., R)
types with greater resistance to this toxic metabolite. It is interesting to note
that prior to population changes in a closed system (test tube) the parent type
creates conditions that may be termed suicidal and thereby sets up specific
environmental effects that will favor the establishment of metabolite-resistant
mutants. In a way, then, the parent type actually influences the direction in
which population changes may occur since its metabolic products may determine
which of the many possible spontaneously arising mutants will have the greatest
survival value at any given time of a culture's growth. It is easy to visualize
how the continual quantitative and qualitative alteration of metabolite con-
centration may create a constantly changing environment that successively
favors different mutants and thereby apparently causes progressive population
changes as long as reproduction continues.

139

298 R. J. GOODLOW, L. A. MIKA, AND W. BRAUN [VOL. 60

The reasons for past failures to detect this phenomenon in nonsynthetic media
are still obscure. That the complexity of such media may be responsible is in-
dicated by the recent results of Schuhardt, Rode, and Oglesby (1949), who
demonstrated that many amino acids, including alanine, are toxic to Brucella,
whereas other amino acids may neutralize their effects. Whether such inter-
actions may have masked the influence of filtrates when complex media were
used remains questionable since population changes similar to those in simple
media regularly occur in complex media, even though they remain unaffected
by the addition of filtrates from old cultures.

A neutralization of the toxic effects of alanine accumulation in Brucella cul-
tures, possibly with the help of specific antagonists, should be capable of pre-
venting population changes in smooth cultures during prolonged periods of
growth in liquid media. Actually, such interference with the accumulation of
metabolites favoring the establishment of mutants may have been practiced
unwittingly in prior experiments that were concerned with specific selective
agents that suppressed the establishment of nonsmooth types in smooth popula-
tions. Thus it has been demonstrated (Braun, 1949) that the addition of small
amounts of normal serum from Z?rwceZZa-susceptible hosts to smooth broth cul-
tures will prevent the establishment of nonsmooth types. Preliminary data sug-
gest that in such serum cultures the accumulation of alanine is actually reduced
or that in addition to alanine a different amino acid, possibly antagonistic to
alanine, may accumulate. Similarly, the prevention of population changes in
smooth Brucella cultures containing 0.03 m sodium pyrophosphate (Braun, 1950)
may involve an interference with enzymatic processes leading to alanine produc-
tion through the removal of trace metals that may act as catalysts for these
reactions. This possibility is further supported by the fact that the addition of
Mn++ or Mg++ to smooth cultures results in a considerable enhancement of the
establishment of nonsmooth types (Cole and Braun, 1950).

Attempts are being made to determine the pathways that may be involved
in alanine production. At least three mechanisms may be suggested, namely:
(1) There may be transamination between aspartic and pyruvic acids yielding
oxalacetic acid and alanine. The pyruvic acid may be formed as an end product
in the oxidation of lactic acid initially present in the medium, whereas the hy-
drolysis of asparagine, also initially present, would yield aspartic acid and free
ammonia. (2) There may be direct amination of pyruvic acid with the free am-
monia to yield alanine. (3) There may be a combination of pathways (1) and
(2).

It has been ascertained that the accumulation of alanine in smooth cultures
and the simultaneous establishment of alanine-resistant nonsmooth mutants
described for B. abortus occurs in cultures of B. suis, B. melitensis, and strain
19 of B. abortus as well. This phenomenon, therefore, appears to be of general
significance for brucellae. To what extent it may also apply to other bacterial
species remains to be determined. However, there are a number of observations,
reported in the recent literature, which indicate the metabolic accumulation of
specific amino acids and their toxicity for certain cell types. If tested by methods

140

1950] GROWTH AND VARIATION OF BRUCELLA ABORTUS 299

analogous to those used in the present study, these metabolites may prove to
have similar effects on population changes. For example, Dubos (1949) has re-
ported that virulent types of Mycobacterium tuberculosis are highly susceptible
to DL-alanine and serine, whereas avirulent variants are more resistant to these
amino acids; Dagley, Dawes, and Morrison (1950) determined with the help of
chromatography that certain amino acids, including alanine, accumulated when
E. coli was grown in a simple synthetic medium; Linggood and Woiwod (1949)
have described parallel increases between alanine accumulation and toxin pro-
duction in cultures of Corijnebacteriumdiphtheriae; and Gordon and Gordon
(1947) noted the development of alanine- and glycine-resistant strains of Shigella
dysenteriae.

Finally, these results suggest the possibility of controlling population changes
of pathogenic bacteria within infected hosts, such as enhancing the establishment
of less virulent nonsmooth types in vivo through the administration of alanine,
which in turn may result in therapeutic effects. Studies in this direction are
now under way.

SUMMARY

The application of paper chromatography to studies of the growth and varia-
tion of smooth Brucella abortus in Gerhardt and Wilson's synthetic medium
revealed a striking correlation between the accumulation of certain amino acids
in the medium and the appearance of nonsmooth variants.

The role of one of the amino acids, alanine, in favoring the establishment of
nonsmooth variants, was verified by the addition of filtrates of old cultures or
of alanine alone to freshly inoculated smooth cultures in synthetic medium.
Under both conditions, a more rapid and enhanced establishment of nonsmooth
variants was observed, and it was found that alanine markedly suppressed the
viable count of smooth cells but failed to exhibit a similar marked effect on
nonsmooth types.

It thus appears that the accumulation of alanine as a metabolite of smooth
cells creates an environment favorable for the establishment of spontaneously
occurring nonsmooth variants.

The possible metabolic pathways involved and the relation of these data to
general phenomena of population changes have been discussed.

REFERENCES

Bail, O. 1929 Ergebnisse experimenteller Populationsforschung. Z. Immunitats., 60,

1-22.
Bkaun, W. 1946 Dissociation in Brucella abortus : a demonstration of the role of inherent

and environmental factors in bacterial variation. J. Bact., 61, 327-349.
Braun, W. 1947a Bacterial dissociation. Bact. Revs., 11, 75-114.
Braun,W. 19476 The production of apparent cycles in bacterial variation. J. Bact., 53,

250-251.
Braun, W. 1949 Studies on bacterial variation and selective environments. II. The

effects of sera from Bruce^a-infected animals and from normal animals of different

species upon the variation of Brucella abortus. J. Bact., 58, 299-305.

141

300 R. J. GOODLOW, L. A. MIKA, AND W. BRAUN [VOL. 60

Braun, W. 1950 Variation in the genus Brucella. In Brucellosis. AAAS Symposia, 10,
26-36.

Cole, L., and Braun, W. 1950 The effect of ionic Mn and Mg on the variation of Brucella
abortus. J. Bact., 60, 283-289.

Dagley, S., Dawes, E. A., and Morrison, G. A. 1950 Production of amino-acids in syn-
thetic media by Escherichia coli and Aerobacter aerogenes. Nature, 165, 437^438.

Dubos, R. J. 1949 Toxic effects of dl-ser'me on virulent human tubercle bacilli. Amer.
Rev. Tuberc.,60, 385.

Etinger-Tulcznyska, R. 1932 tlber bakterien Antagonismus. Z. Hyg. Infektion-
skrankh., 113, 762-780.

Gerhardt, P., and Wilson, J. B. 1948 The nutrition of brucellae: growth in simple
chemically defined media. J. Bact., 66, 17-24.

Gordon, J., and Gordon, M. 1947 Development of resistance of Shigella shigae to gly-
cine . J . Path . Bact . , 59 , 445-451 .

Horne, R. E., and Pollard, A. L. 1948 The identification of streptomycin on paper
chromatograms. J. Bact., 55, 231-234.

Linggood, F. V., and Woiwod, A. J. 1949 The application of paper partition chromatog-
raphy to the production of diphtheria toxin; two-dimensional chromatography. Brit.
J. Exptl. Path., 30, 93-100.

McCullough, W. G., Mills, R. C., Herbst, E. J., Roessler, W. G., and Brewer, C. R.
1947 Studies on the nutritional requirements of Brucella suis. J. Bact., 53, 5-15.

Mohr, W. 1934 Untersuchungen liber antagonistische Vorgange zwischen Varianten
desselben Stammes. Z. Hyg. Infektionskrankh., 116, 288-294.

Neufeld, F., and Ktjhn, H. 1934 Untersuchungen tiber "direkten" Bacterienantagonis-
mus. Z. Hyg. Infektionskrankh., 116, 95-110.

Ryan, F. J., and Schneider, L. K. 1949 The consequences of mutation during the
growth of biochemical mutants of Escherichia coli. J. Bact., 58, 201-213.

Schtjhardt, V. T., Rode, L. J., and Oglesby, G. 1949 The toxicity of certain amino acids
for brucellae. J. Bact., 58, 665-674.

Zamenhof, S. 1946 Studies on bacterial mutability: the time of appearance of the mu-
tant of Escherichia coli. J. Bact., 51, 351-361.

I 4 2

GENE RECOMBINATION AND LINKED SEGREGATIONS
IN ESCHERICHIA COLI 1

JOSHUA LEDERBERG 2

Department of Botany and Microbiology, Osborn Botanical Laboratory,
Yale University, New Haven, Conn.

Received August i, 1947

THE occurrence of factor recombination in the bacterium, Escherichia coli,
has been described in previous reports (Lederberg and Tatum, 1946 b, c,
Tatum and Lederberg, 1947). In an attempt to elucidate further the genetic
structure of this organism, these studies have been extended to crosses involv-
ing several characters, and to the quantitative enumeration of various recom-
bination classes. The results described in this paper provide evidence support-
ing the sexual basis of factor recombination and of the existence of an organ-
ized array of genes comparable to that of higher forms.

MATERIALS AND METHODS

The parent "wild-type" strain, K-12, of E. coli used in these experiments
and the production and behavior of biochemical mutants have been described
(GRAY^nd Tatum, 1944, Lederberg and Tatum, 1946a, Roepke, Libby, and
Small, 1944, Tatum, 1945). Specific requirements, notation, and other data
pertinent to the biochemical mutants are summarized in tables 1 and 2. In gen-
eral, a biochemical deficiency resulting from mutation is designated by the
initial of the substance required (e.g. B~ for biotinless), while the wild type
alternative is written with a " + " sign (e.g. B + to emphasize the alternative to
B~). The term "prototroph" (Ryan and Lederberg, 1946) has been devised
for strains exhibiting the nutritional behavior of the wild type, which for E.
coli implies independence of any specific growth factors. Prototroph is, how-
ever, not synonymous with "wild type" since it refers (a) only to the pheno-
typic appearance of a culture and (b) only to nutritional and not to other pos-
sible mutant characteristics.

K-i 2 as a coliform is capable of fermenting, or producing acid, from a variety
of sugars, including glucose, galactose, maltose, lactose and mannitol; how-
ever, it ferments glycerol only weakly, and sucrose even less so. Because of the
ease of scoring and their biochemical specificity, mutants unable to ferment
various sugars have been looked for. Particular attention was paid to the isola-
tion of "lactose-negative" or "Zac - " mutants, because of the taxonomic sig-
nificance which has been attached to this character.

1 Abstracted from a dissertation offered in partial fulfillment of requirements for the degree
of Doctor of Philosophy at Yale University.

2 Fellow of the Jane Coffin Childs Memorial Fund for Medical Research. This work
has been supported by the Jane Coffin Childs Memorial Fund for Medical Research. The
author's present address is: Department of Genetics, University of Wisconsin, Madison, Wis

[Reprinted by permission from Genetics 32:505-525, September, 1947]

M3

S o6 JOSHUA LEDERBERG

The detection of fermentation mutants is readily accomplished by the use of
indicator media. The medium "EMB-lactose" used in routine bacteriological
work was found to be highly useful. It consists of the following (in g/1) : pep-
tone (or "N-Z-Case") 10, yeast extract i, lactose 10, agar 15, eosin Y 0.4,
methylene blue 0.06, sodium chloride 5, dipotassium phosphate 2. On this
medium, colonies of bacteria which can ferment lactose (or any other sugar
added in its place) rapidly turn a deep purple color, while colonies of non-
fermenting organisms remain white or pink but may slowly turn light blue.

Lac mutations have been recovered in two instances. Among 15,000 colonie s

Table i
Symbols used for various loci.

1. Nutritional requirements. Allele for requirement of a given substance is designated by the
superscript "~"; independence by " + ". E.G., B~ is biotinless; B + is biotin-independent.

B biotin L leucine Pa phenylalanine (<£ was used previously,

Bi thiamin M methionine but has been modified for typographical

C cystine P proline reasons)

T threonine

2. "Sugar" fermentations. The ability to ferment is designated " + "; the inability "~".
Lac lactose Gly glycerol

3. Bacteriophage resistance. Resistance is designated by the superscript " r "; sensitivity by "•".
E.G. Vf.

Vi resistant to Ti, T5

Via resistant to Ti; sensitive to T5

V lb resistant to Ti; mucoid colonies

V t resistant to T6.

4. Resistance to chemical agents. Resistance and sensitivity V and u s" respectively, as Cla*.
Cla sodium chloroacetate

A sodium azide

of strain Y-10 (T~L~Bi~) obtained by spreading a culture previously treated
with ultraviolet light on EMB-lactose agar, a single pink colony was noted. It
proved to be the same, nutritionally, as Y-10 and was therefore regarded as a
Lac~ mutant; this stock is labelled Y-53. Among 30,000 colonies of Y-40
(B~M~Vi r ) a single Lac~ was recovered following treatment with nitrogen-
mustard (Tatum, 1946), and was designated as Y-87. Tests showing that these
independent mutations are probably allelic will be described in a later section
(see table 5). Strains Y-53 an d Y-87 differ m the rate at which the Lac~ char-
acter reverts to the Lac+ condition, but whether this is due to different allelic
states or to differences at other loci, cannot be definitively asserted.

Attempts to obtain maltose, mannitol, and galactose-negative mutants
were not successful, presumably because the populations tested were too small.
A glycerol-negative strain has been obtained, but the wild type ferments this
polyalcohol so poorly to begin with that accurate scoring is difficult; studies
on this character will not be further reported here.

144

SEGREGATIONS IN ESCHERICHIA COLI 507

Mutations for resistance to specific bacteriophages or bacterial viruses
have proven to be exceedingly useful. They are readily obtained as spontaneous
mutants by plating a large number of sensitive bacteria with the particular
virus in question; only resistant mutants escape lysis and may be recovered as
"secondary" colonies (fig. 1). Resistant mutants are readily freed from residual
virus by serial single colony isolation. Resistance to a given virus may be

Table 2
A summary of the mutants used.

STRAIN NO.

GENOTYPE

ORIGIN

GENOTYPE

AGENT

K-I2

prototroph.

Original wild
strain

58

B~

K-12

B +

X-ray

58-161

B-M-

58

B-M+

X-ray

58-278

B-Pa~

58

B~Pa+

X-ray

Y-24

B-Pa-C-

58-278

B~Pa~C+

ultra-violet

679

T~

K-12

T+

X-ray

679-680

T~Lr

679

T~L+

X-ray

Y-io

T-L-Br

679-680

T-L-Bi+

X-ray

Y-46

T-L~Br\\ r

Y-10

T-L-BcVx"

selection

Y-S3

T-L-BrLaC

Y-io

T-L~BrLac +

ultra-violet

Y-64

T-L-BrLac-Vf

Y-53

T-L-BrLaC-Vx'

selection

Y-40

B-M~Vi r

58-161

B-M-]\>

selection

Y-87

*B-M-VfLac*-

Y-40

B-M-VfLac +

nitrogen mustard

Y-24- IV

B-ParC-Vf

Y-24

B-Pa-C-\\>

selection

679-183

T~P-

679

T~P+

X-ray

Y-88

T-L'BcLac~Cla T

Y-53

T-L-BrLac-Cla*

selection

Y-80

B-M-Vf Gly-

Y-40

B-\[-Vf Gly+

nitrogen mustard

Y-91

B-M-V? Cla r

Y-40

B-M-V? Cla»

selection

Y-92

B-M~Vi r Az r

Y-40

B-M-Vi T Az>

selection

Y-94

T-L-BrLac-V?

Y-53

T-L-BrLac-Vt*

selection

Y-100

T-L-BcLac~\\ a r

Y-53

T-L-BrLac~V la >

selection

Y-86

T-IrBrLac-Vif

Y-53

T~L~Bi~Lac~Y ib*

selection

* Lac 2 in mutant Y-87 differs from Lac in mutant Y-53 and its derivatives in the greater
reverse-mutability of the latter. Lac" and Lac 2 ~ are otherwise similar, and allelic.

scored by streaking a loopful of bacteria on an EMB or nutrient agar plate at
right angles to a previous streak of the virus suspension (Demerec and Fano,
1945, see fig. 1 of the present report.)

It was found, however, that mutations for resistance to a given virus are
not entirely specific, but that resistant mutants display "cross-resistance,"
i.e., are also resistant to other viruses. For example, most Tz-resistant types
are also resistant to T5. (For the nomenclature of the bacterial viruses used in
this investigation, and a detailed account of the cross-resistance patterns of
another strain, E. coli B, see Demerec and Fano, 1945). The cross-resistance
patterns of K-12 are similar to those of E. coli B with the exception that Ti-
resistant mutants which are sensitive to T5 are not tryptophaneless, as has

145

5 o8 JOSHUA LEDERBERG

been reported by Anderson (1946) for the corresponding mutants of E. coli B.
In this paper, the designation V x r will be used for the more frequent TVresist-
ant mutant, which is also resistant to T5. The symbol \\ a r is reserved for the
r5-sensitive, Tz-resistant mutant, but the evidence that distinct loci are in-
volved will be presented in extenso in another place.

In addition to Vf and V u r , just mentioned, a third type of "secondary
colony" has been found among populations treated with the virus Ti. This
type, Vib r is characterized by an exceedingly slimy or mucoid colony confor-

Figuee i. — The phenotypes of the four combinations of Lac and V are illustrated. In order
the}' are: Lac + Vi r ; Lac + Vi s ; Lac'Vf; Lac~V\ s . An EMB-lactose agar plate was first streaked
vertically with the virus Ti. Subsequently, each of the bacterial types was streaked, from left to
right, perpendicularly across the virus streak. After 16 hours incubation, both the Lac and Vj
phenotypes are well developed. Developing in the zone where Lac'V V has been lysed can be seen
two colonies of resistant mutants: Lac~V{.

mation. Recombination studies on this mutant are complicated by its genetic
instability; V\h T rapidly reverts to the wild type, and in addition may also be
strongly selected against in competition with V^*. However, the locus of
Fib can be distinguished from the locus of the other V\ mutants by the demon-
stration of a different recombination frequency with Lac. These data are sum-
marized in order to emphasize the importance of genetic tests to insure the
allelic identity of phenotypically similar mutants.

It is particularly fortunate that resistance tests can be conducted on EMB
agar, since this allows the characterization of a strain with respect to virus-
resistance and to lactose fermentation with a single streaking (see fig. 1).

Mutants resistant to sodium chloroacetate (Cla r ) were obtained by streaking
a large number (about io 7 ) of bacteria on nutrient agar to which filter-sterilized
chloroacetate has been added to make a final concentration of 2 mg/ml At
this concentration, the wild type is substantially inhibited, while resistant
mutants grow luxuriously. This mutation is accompanied by deficiencies in

146

SEGREGATIONS IN ESCHERICHIA COLI 509

the metabolism of pyruvic and acetic acids, which will be described in more
detail elsewhere. Independent mutations to other inhibitors, including iodo-
acetate, azide, streptomycin, streptothricin, mercuric chloride, and Brilliant
Green, can be secured in a similar fashion, but genetic analysis of these muta-
tions has not been completed.

Morphological variation has occasionally been noted (exceedingly rough or
very mucoid colonial form) but is relatively unsuitable for genetic work be-
cause the presumably random choice of prototroph recombinants may be
influenced.

In addition to the EMB agar already described, a number of other natural or
"complete" media have been used. The Difco product "Penassay Broth" has
been used most extensively, and is satisfactory for the preparation of inocula,
except that it must be supplemented with cystine for the growth of cystineless
organisms, such as strain Y-24. Other satisfactory media include a broth con-
sisting of: peptone 5, glucose 5, yeast extract 3, g/1, as well as Difco Nutrient
Broth, and diverse concoctions containing peptone or casein hydrolysates and
meat or yeast extract.

The synthetic or minimal medium contains, in g/1: NH 4 C1 5, NH4NO3 1,
Na 2 S0 4 2, K2HPO4 3, KH2PO4 i, glucose 5, asparagine 1.5, MgS0 4 0.1, trace
elements (Gray and Tatum 1944), and CaCl 2 , a trace. The medium is made
solid by the addition of agar in a concentration of 1.5 percent.

To avoid flocculation when used with agar, the glucose and agar in solution
should be autoclaved separately, and mixed with the other components just
before using. Unwashed agar (Difco) is sufficiently free of the growth factors
under consideration to be satisfactory for many experiments; the use of washed
agar, however, is recommended for the cleanest results.

The detection of recombinants is based upon the inability of biochemical
mutant bacteria to proliferate in the absence of their specific growth sub-
stances. Plating in minimal agar, therefore, has the effect of a sieve for proto-
troph cells. To insure against contamination with prototrophs derived by re-
verse mutation, which has been noticed at certain loci, it has been desirable to
use multiple biochemical mutants as the parental stocks in recombination
studies. Coincidental reversion at two or more loci is theoretically improbable,
and experimentally undemonstrable (Ryan, 1946, Tatum and Lederberg,
1947) . For example, plating either B'M-T+L+BS or B + M + T~L- Brsepa.ra.tely
into minimal agar did not lead to the appearance of prototrophs, B+M+T+L+-
Bi + . When, however, a mixture of these cell types was so "sieved," one proto-
troph was found for about each io 7 cells inoculated. These have been assumed
to arise from the recombination of "+" alleles to form the prototroph.

In previous experiments, the two multiple mutants were inoculated together
into a complete medium and allowed to grow in mixed culture before plating
into minimal agar. This method is not satisfactory for present purposes be-
cause it allows possible selective differentials to alter the relative frequencies
of different recombination classes. A modified procedure has been developed,
which will now be described in detail.

The mutant stocks are maintained on "complete" agar slants, transferred

M7

510 JOSHUA LEDERBERG

at intervals of 6-8 weeks. They are inoculated separately into test-tubes con-
taining about ten ml of liquid comj lete medium and incubated overnight at
30°C with gentle shaking. The following morning, an additional ten ml of the
same medium is added to each culture, and the tubes are incubated in the same
manner for an additional three to five hours. These cultures contain from 1-4X
io 9 cells per ml. They are then washed in the following manner: the cotton
plugs are replaced with sterile corks which have been kept in 95 percent alcohol
and the alcohol flamed off just before using. The cultures are then centrifuged
at about 2500 r.p.m. for 20 minutes, which suffices to jack the cells in the
bottom of the test tubes. The supernatant medium is carefully poured off, and
the tube is rinsed with about 10 ml sterile distilled water, care being taken not
to disturb the pellet. The cells are then resuspended in an additional 15-20 ml
sterile water, and recentrifuged. The supernatant wash water is decanted and
replaced with an equal volume of fresh sterile water, in which the cells are
suspended. In the meantime, minimal agar plates are prepared. A bottom layer
of about 15 ml minimal agar is poured into each Petri plate and allowed to
solidify. Cell suspensions of different mutant stocks are mixed at this time and
measured quantities (usually about io 8 -io 9 cells) are pipetted onto the agar
surface. At this time also, one may add such growth factor supplements as are
desired to permit the growth of recombination types other than prototrophs.
The cell suspensions are then mixed into a layer of about- ten ml molten mini-
mal agar (at 45-5o°C) which is poured onto the plates. After the agar hardens,
the plates are incubated at 30°C for a period of 48 hours. At this time proto-
troph colonies will be found distributed throughout the plate, many of them
at or near the surface and accessible to picking for further characterization.

The procedure may be varied in several ways. It is important however that
the inoculum consist of "young" cells, since cultures of 24 hours or older have
given quite inconsistent results. It is possible to store the inoculum in distilled
water for at least twenty-four hours without appreciably affecting the yield,
which suggests that the aggregation of genetic types leading to the recombi-
nation process occurs in the molten or the solidified agar. This occurrence
must, however, take place within a few hours, since the recombinant proto-
trophs are not appreciably slower to appear than wild type cells in a similar
physiological state which may be streaked on the surface of the plates. Pre-
sumably, therefore, one could increase the yield of prototrophs by making
conditions more favorable for the free contact of the cells, as by packing them
together in a centrifuge tube in minimal liquid medium. However the compli-
cation of proliferation of prototrophs already formed would interfere with the
interpretation of such an experiment. Many physiological factors may inter-
fere with the recombination process, and, for example, the yield may be re-
duced markedly by inoculating too heavily, or by omitting an under-layer of
agar into which, presumably, deleterious metabolic products may diffuse. In-
stead of mixing the cells in semisolid agar, it is possible to streak the mixture on
the surface of slightly dried minimal agar plates. Under these conditions, how-
ever, the prototroph colonies are likely to be more heavily contaminated with
the residual parental mutant types.

SEGREGATIONS IN ESCHERICHIA COLI 511

For most purposes, however, this contamination may be ignored, as will be
shown in a later section. Prototroph colonies are then fished and streaked di-
rectly on EMB plates, or otherwise tested, to classify them with respect to
other factors that may be segregating.

RESULTS AND CONCLUSIONS

In most organisms inheritance is studied by the examination of zygotes
carrying the gene alternatives determining a character. The segregants are
chosen at random, and factor linkage is recognized by deviations in the fre-
quency of parental and new couplings of a series of characters. In the absence
of a random method of separating zygotes in E. coli, one is limited here to the
members of specific recombination classes, namely the prototrophs. It is how-
ever, possible to introduce other factor differences into the biochemical mu-
tants from which prototrophs are obtained, and to determine how such factors
segregate into this recombination class. It was hoped in this way to obtain in-
formation concerning the haploid or diploid condition of the bacterial cell, and
to determine whether factors segregated at random, or according to specific,
perhaps linear chromosomal laws.

The first factor pair to which this approach was applied was Vi r /V\* (Leder-
berg and Tatum, 1946b). In the cross B-M-P+T+VSXB+M+P-T-V^, ten
B+M+P+T+ were isolated. Eight proved to be V{ while two were V x \ This at
once suggested that the vegetative cell of E. coli is haploid, since segregation
could be observed in the first filial generation clone. It was noted also at that
time that the "reversed" cross: B-M-P + T+V{XB+M + P-T-Vi T gave quite a
different ratio of r/s in the prototrophs, namely 3:7. Results on so small a
sample are of doubtful significance, but they suggested the technique by which
the basis of this character "segregation" could be elucidated. For this reason,
the study of "reversed" crosses was extended to include numerically more
data, using various combinations of mutants, and involving in addition to
V! r /Vi\Lac + /Lac~. The information which was obtained is summarized in
tables 3 and 5. The data show clearly that neither of the factor alternatives
V\ r /V\* or Lac + /Lac~ segregates at random into the prototroph recombination
class. However, the occurrence of all factor combinations, albeit with different
frequencies, is evident, at least with respect to Lac and V\. It seemed clear
that there are only two alternative explanations for the unequal frequencies
with which alternative alleles are manifested in the prototrophs: (a) that the
alleles were characterized by some differential physiological property, such as
dominance, or preferential segregation, or (b) that the nonrandom segregation
was due purely to the mechanics of factor recombination, which is to say a
linkage system.

The results of "reversed crosses" have a distinct bearing on this problem. If
nonrandom segregation into protctrophs were due to some physiological prop-
erty of the allele concerned, its particular coupling in the parent in which it is
introduced should have no great effect on the segregation frequency; if on the
other hand, the effect were purely mechanical, the segregation would reflect
entirely the couplings of the parents, and the substitution of one allele for

149

512 JOSHUA LEDERBERG

another in the parents (as in reversed crosses) should lead to a corresponding
inversion in the ratios with which that allele is found in the prototrophs. The
tables cited show that in every case there is no agreement between the ratios
found in reversed crosses, unless the comparison is made with one of the ratios
inverted, in which case there is reasonably good agreement. This result is in
accord with the hypothesis that the genes in E. coli are arranged in one or more
linkage groups, and is in disagreement with the postulation of a diploid con-

Table .3
Comparisons of IV segregations when introduced with alternative parents.*

PROTOTROPHS

[B + M + Pa + C + T+L+B 1 + P + ]

PARENTS

IV

iv

%IV

B~Pa-C-T + P +

B+Pa + C + T~P-

... Vl r

X

• ■ ■ Vi'

7 6

6

92

■ • • TV

X

• ■ • IV

30

107

22

B-Pa-C-T+L+Bi

B+Pa+C+T-L-Br

• • • V?

X

■ ■ ■ IV

80

23

77

■ ■ ■ IV

X

. • • TV

53

133

28

B-M-T+P+

B+M+T-P-

■ • • IV

X

■ • • IV

49

8

86

. . . PV

X

■ • • Vf

5

19

21

* See Lederberg (1947) for a statistical analysis of tables 3, 5, and 6.

dition, or with a state of indefinite "ploidy" which would be characteristic of a
system of cytoplasmic inheritance.

The results of these experiments seemed sufficiently secure that one could
adopt the existence of a linkage system as a working hypothesis and on this
foundation, an attempt has been initiated to "map" a number of markers in
E. coli. It was hoped at first that there might be found linkage groups which
would be independent of one another, so that recombination between bio-
chemical markers in one group could be used to detect recombinants, yet not
interfere with the segregations in the other group(s). There was, however, no
immediate prospect that these relationships could be found initially, so it was
decided to study linkage relationships in a single pair of mutant stocks, and
their derivatives. The stocks which were selected for this study were 58-161
(B~M~) and Y-53 (T~ L~ Br Lac~) and their IV mutants. Since Lac and Vi
could be so readily scored, using only a single streak from each prototroph
colony which appeared, it was hoped that the collection of an adequate volume
of data could be accomplished with greater facility than if biochemical markers
only were used.

It was, however, necessary to determine the relationships of the biochemical
mutant loci of which at least four must be used to obtain recombinants. Mix-
tures were, therefore, plated into minimal medium supplemented with a single

150

SEGREGATIONS IN ESCHERICHIA COLI 513

nutritional requirement, i.e., either biotin, methionine, threonine, leucine, or
thiamin, allowing the proliferation of the corresponding single mutant as well
as the prototrophic type. Colonies were then picked at random and scored ac-
cording to their nutritional requirements. The results are summarized in table
4. Unfortunately, it was found that the addition of methionine to the minimal
medium allowed excessive growth of B~M~, presumably because of a degree of
contamination of the methionine with biotin. This datum is, however, not
essential for the argument. In general, it will be seen that the + classes are
markedly and significantly more frequent than the single mutant types, with

Table 4

Relative frequency of various biochemical recombination
classes in the cross.

B-M'T+L+Bi+XB+M+T-L-Br*

FROM

NUMBER

RECOMBINATION CLASSES FOUND

SUPPLEMENTED
WITH

COLONIES
TESTED

TYPE

NUMBER

TYPE

NUMBER

RATIO

X 2

Biotin

70

B~

IO

B+

60

O.17

36

Threonine

46

T~

9

T +

37

O.24

17

Leucine

56

lr

5

L+

51

O.O96

38

Thiamin

87

Br

79

Bi +

8

9.88

56

* Cells of the parental types were mixed and plated into agar supplemented with the growth
factor indicated. On this medium, the two recombination classes indicated on each line of the
table could form colonies. Contrasting alleles only are specified; other loci, unless otherwise speci-
fied, have the " + " configuration. The x 2 for the ratio of single biochemically deficient types to
prototrophs is calculated for a comparison with the 1 : 1 expectation of a random segregation.
As can be seen from the x 2 values, the probability that the deviations are due solely to chance is,
in each case, less than .001 .

the exception of B-r which is nearly ten times as frequent as Bi + . Writing the
cross as B-M'T+L+B^XB+M+T-L-Br, these results may be interpreted
as follows:

1. B + M+ T+L+Bi+ more frequent than B~M + . Therefore B and M are
linked.

2. T+L+ B + M+Bi+ more frequent than either T~L+ or T+L~. Therefore T
and L are linked.

3. BrB+M+ T+L+ more frequent than B 1 +B+M+. Therefore B x is linked
to B and M, but probably not between them.

One may therefore map these five loci onto not more than two linkage groups,
according to the scheme in fig. 2a. In all that follows, the [B-M] and [T-L]
combinations will be regarded as single units, since conclusive information as
to their relative order has not been obtained. These data so far do not allow
any conclusion to be drawn as to whether the regions B X ~[BM] and [TL] are
linked or are independent cf each other, since a recombination between them
is a necessary requirement for a detectable type.

151

5M

JOSHUA LEDERBERG

Table 5

Segregation of Lac and Vi into prototrophs issuing from
various parental combinations*

RECOMBINATIONS

B-M-T+L+Bf B+M+T-L-Br B+M+T+L+ Lac~V{ Lac~V ',« Lac+V? Lac + \\

Lac + V,

Lac Vi

Bi + 602 303

[45-8] [23.:

Br*** 13 s

[45] [28]

Br** 244 157

[42.8] [27.5]

(D) (E)

387 22

[29-4] [1.7]

8 o

[28] [o]

159 10
[27.9] [i.Q]

(C) (triple)

Lac+Vf

Lac' V{

Bi + 107 145 9 61

[33-2] [45-o] [ 2.8] [19.0]

Br** 134 151 9 80

[35.8] [40.4] [ 2.4] [21.4]

(E) (D) (triple) (C)

Lac-V?

Lac+Vs

Br 28 6 46 37

[23-9] [ S-i] [39-3] [31-6]

JBi - ** 102 7 201 91

[25-4] [1.7] [5o.i] [22. 7I

(C) (triple) (D) (E)

\Lac + Vf Lac'V?

• Br**

128

33

\Lac~Vf Lac-Vs
(F-87)

(Y-53)

• Br**

134 Lac~

; not scored for V\

* Cell mixtures of the indicated composition were plated into minimal agar plates or into
plates supplemented with thiamin. B± + types refer to scores of prototrophs picked at random from
minimal plates.

** Br refers to colonies picked at random from thiamin supplemented plates. Although pre-
dominantly Br they contain B x + colonies in the proportion 1 : 10 as may be seen from table 4.

*** Br- In this series, colonies were scored as to B\, and only the Bi~ are recorded.

The letters (C), (D), (E), refer to crossover types corresponding to the regions [B M)-Lac;
Lac—V\\ and V t — [T L] respectively, according to the map of Fig. 2d.

t Test for allelism.

On the basis of table 5, the factors V\ and Lac may be brought into the argu-
ment. In addition to the joint segregations of these factors, the effect of the B\
segregation was studied in the following way. It would be uneconomical, in
view of the relative paucity of Bi + types, to separate these from the Br~ by
nutritional testing of colonies which appear on thiamin supplemented agar.
Instead, the entire sample was regarded as Br~ with the proviso that it might
be contaminated to the extent of ten percent with Bi + . However, it has been
found that the distribution of Lac and V on colonies picked from thiamin

152

SEGREGATIONS IN ESCHERICHIA COLI 515

supplemented agar is homogeneous with the distribution in prototrophs, so
that the segregation of these factors is not influenced by the Bi segregation.

The data in table 5 show that Lac is inclined not to separate from BM, and
is therefore regarded as linked to it, while there is a similar linkage of \\ to
TL. Since the recombination of Lac with BM is not influenced by the inter-
change between Bi and BM, they are on opposite sides of BM as suggested
by map 2b. Finally, a scrutiny of the interaction between the Lac and V segre-
gations shows that these are not independent of, each other, particularly be-
cause of the rarity of the least frequent class. This suggests, then, that the
two linkage groups of fig. 2b be combined to give the map of fig. 2c. (The locus
of V& on this map is obtained from additional data.) According to this interpre-
tation, the rarity of the least frequent Lac-V combination stems from the fact
that a triple-crossover is necessary for its production. In fig. 2d, the cross
Y-40X Y-53 is interpreted according to the map, with a table citing the regions
in which interchange must take place to yield the given types.

That the first seven factors to be investigated should fall in the same linkage
group leads to the inference that there is only a single chromosome in E. coli.
This inference is supported by incomplete analyses of the segregations of 8
other markers referred to in table 1. None of these factors has been found to
segregate independently of the factors which have already been described as
belonging to a single linkage group. The possibility that segregation interac-
tions may, in some cases, be based upon an inter-chromosomal type of inter-
ference (compare Steinberg and Fraser, 1944), has not been ruled out,
however.

The distances recorded in fig. 2c are derived from the recombination totals
in tables 5 and 6. However, the distance between [BM] and [TL] cannot be
estimated directly, but only the partition of that distance among the regions
BM-Lac, Lac-V u and Vi-TL. The relative frequency of the "triple-inter-
change" type can be used to estimate the absolute map distances, if it is as-
sumed that there is no interference. This frequency, about 2.1 percent, is
readily calculated to be consistent with a map length of between 75 and 80
units altogether either in a two-strand or a four-strand system (Lederberg,
1947). These values must be regarded as rough approximations, because they
are extremely sensitive to error in the estimation of the proportion of the
"triple" types.

Linearity

In constructing a map, and calculating distances, it has been taken for
granted that there is in E. coli a system of linear linkage, such as has been
demonstrated quite conclusively in Drosophila, and inferred in all higher or-
ganisms. What direct evidence may one bring to bear on this question?

The method which one is forced to em\ loy in hybridizing this bacterium
introduces certain complications. The classical proof of linearity is based on
the additive character of distances, expressed in morgans, between loci occur-
ring within the same linkage group. The determination of map distances is
based upon a comparison between parental and new combinations of linked

'53

516 JOSHUA LEDERBERG

genes, as determined in the progeny of zygotes selected at random. In E. coli,
on the other hand, one is limited to the recovery of that recombination class
in which there has necessarily been an interchange between certain biochemical
loci, in the cases here discussed, betwen [BM] and [TL]. For this reason, it is
not possible to obtain a direct measure of the absolute distance between factors
which are located within this critical region, and any argument in favor of
linearity which is based on the segregations of such factors may have the

a.— +

A k A [J" L]

b.-H=-nt

B, \tf £} L.c if [r

9

17 5

38

20

C W

f\ i Lc

d

b [T
e

1}

a

b c

+

/

4-

a - +

+

5

—

Crossovers Recoverable type

(c,d, ore) B l -M+B+T+L+

(a) (c, d, or e) B 1 - r M+B+T+L+

(a) (b) B 1 +M+B-T+L+

(a) (/) B 1 +M+B+T-L+

(a) (/) (c, d, or e) B.+M+B+T+L-

(a) (c) B 1 +M+B+T+L+ ■ ■ ■ Lac+Vf

(a) (d) • • • Lac-Vf

• • • Lac-Vf

(a) (c) (d) (e) ••• Lac + \\>

Figure 2. — a, b, and c. Mapping of genetic factors, d. The cross B{* M~B~ Lac' h V \ r T + L + XB{
M^B + Lac~Vi'T~L~ and some of the recoverable crossover classes. (See table 5.)

flavor of circular reasoning. It would be preferable to study the segregations of
factors which are assigned to loci distal to the biochemical factors whose re-
combination is the basis of the detection of sexual offspring. The stocks with
which this might be accomplished are not yet available, but it is hoped that
they will be for future work.

That there does exist some sort of linkage system is made highly credible by
the results of the "reverse crosses" tabulated in tables 3 and 5. The chief diffi-
culty in proving that this system is linear has been to formulate the feasible
alternatives, so that critical experiments, the results of which could discrimi-
nate between linearity and a given alternative, might be set up. Certain types
of "linkage" can be disqualified by the data already at hand. For example, one
might postulate that genes of bacteria are embedded in a two-dimensional

154

SEGREGATIONS IN ESCHERICHIA COLI 517

matrix, and there occasionally occurs a gene-for-gene interchange. This is
equivalent to the "Konversion" theory once proposed by Winkler (1932), to
account for interchanges in Drosophila. While this type of arrangement would
account for a tendency to preserve the parental configuration, it fails to ex-
plain either quantitative linkage intensities, or the interaction of segregations
which is revealed by the data on Lac and V in table 5. Naturally, one could
further modify the "Konversion" theory to take these exigencies into account,
but in so doing one would be elaborating an exceedingly complicated theory
which would, in fact, be a re-expression of a mechanical theory of linkage.

The interaction of the Lac and V segregations is perhaps the most critical
datum with which a genetic system for E. coli can be formulated. The inter-
action may be expressed as follows: the frequency of interchanges between
[BM] and Lac is dependent upon the interchanges between [BM] and V. Spe-
cifically, in the cross B+M+T'L-Br Lac-V 1 S XB-M~T+L+B 1 + Lac+VS, one
finds in the B + M + T + L+Bi + the following distribution of classes: Lac~Vi 9 23
percent, Lac + Vi r 29 percent (for the parental combinations) and Lac~V{ 46
percent, Lac + Vy s 2 percent (for the new combinations). With reference to
[B+M+], Lac~ is the parental, Lac + the interchange type. The proportion of
Vi T (representing an interchange between V\ and [BM]) is different in the Lac~
and Lac + segregations: namely 46:23 = 2:1 and 29:2=14.5:1 respectively.
This interaction between interchanges is most simply explained by the assump-
tion that factors are located on a linear segment, so that interchanges between
proximal factors also lead to the crossing over of more distal factors, barring
the occurrence of additional interchanges.

Additional support for the theory of linear arrangement has been found in
the segregation of V 6 , summarized in table 6. It will be noted that the segrega-
tions of Lac, Vi, and V 6 are quite congruous in the Bi~ and Bi + classes. In the
totals, one finds the ratios, for each factor separately, of Lac~ 78 percent;
PV 82 percent; Vi" 36 percent; indicating that the first two are both linked to
[BM] while the latter is linked to [TL]. V 6 cannot, however, be to the left of
[BM] because it does not interact with B\. If, therefore, there is a linear order
of genes, V 6 must be to the right of [BM], and because of its greater linkage
intensity, nearer [BM] than is Lac. This arrangement is indicated in the map
in table 6, and in fig. 2c. The agreement of the data with the hypothesis can
be examined at several points. In the first place, the single exchange types, as
indicated in the table, should be the most frequent. Secondly, barring multiple
exchanges, an interchange between V 6 and Lac should lead also to an inter-
change between V 6 and V\. That is to say, the Lac + Vf class should be more
often Vi r than V\*. Finally, in view of the similarity in linkage intensities to
[BM], Lac and V 6 must be closely linked. Although the "triple-interchange"
types would seem to be rather frequent, reference to the table may suggest that
these conditions are fulfilled. In particular, it will be noted that among the
Lac~, the ratio of V 6 r : V 6 * is 94:3, or 31:1, while among the Lac+, this same
ratio is 10:29, or 1:3. This difference is interpreted to mean that Lac and V 6
are linked to each other, as demanded by the theory of linearity.

It is not, of course, proven that the gene order is not branched at some other

155

51 8 JOSHUA LEDERBERG

point. The most economical hypothesis at this time, however, is that there is
a single unbranched chromosome as the physical basis of inheritance in E. coli.

Attempts to Induce Aberrations

Using a chromosomal theory as a working hypothesis, it was hoped that
some verification could be found by the study of types in which the normal
order of genes was disturbed. Since there is only one chromcsome (from the

Table 6

Segregation of Lac, Vi and V 6 .

B-M-T+L+Bi+LactVfVfXB+M+r-L-BrLdc-Vi'V,'

B+M+T+L+

Lac:

_

_

_

_

+

+

+

-f

Vi :

r

5

r

s

r

s

r

s

TOTAL

V% :

r

r

s

5

r

r

s

s

• • • B,+

24

16

1

2

1

10

2

56

• • • Br**

52

42

2

6

1

16

I

I20

Total

76

58

3

8

2

26

3

176

%

43

33

r-7

4.6

1.1

15

1-7

Crossover region

e

f

cde

cdf

d

def

c

cef

B, ••

■M

• • ■ B

■v.

Lac

■ ■ Vi

• T

• • L

a

b

c

d

e

f

+

-

-

s

+

r

+

+

-

+

+

r

-

s

-

-

** See footnote to table 4.

genetic evidence), the only types of rearrangements would be changes leading
to a series of inversion-transposition types. It was thought that such types
might be detected by genetical procedures by virtue of their effect on crossing
over. In particular, the occurrence of an inversion in the region Bi • ■ • [MB]
would be expected to have the effect of eliminating the recombination
classes involving interchanges in this region. In the cross B~M~T + L + Bi +
X B + M + T~L~Bi~ this would be equivalent to the suppression of prototroph
recombinants; Bi~ types, however, would be recoverable, and allow the in-
vestigation of the extent of the changes.

Preliminary attempts to find such aberration types have, to date, been un-
successful. The procedure was as follows:

Following treatment with nitrogen mustard (Tatum, 1946) or 20,000 r of
X-rays, cells of Y-40 and of Y-53 were incubated separately for 24 hours, to
allow the separation of cells or nuclei that might have been associated at the
time of treatment. The cultures were then streaked out on nutrient agar

156

SEGREGATIONS IN ESCHERICHIA COLI 519

plates. Single colonies of Y-40 were picked and streaked across a nutrient agar
plate. Streaks of similarly treated Y-53 colonies were made from the opposite
direction, so that in the center of the plate, cells of the two types were mixed,
treated colony by treated colony. The plates were incubated for 24 hours, the
mixed growth scraped from the plates, suspended in sterile water and plated
into minimal agar. The occurrence of colonies which would not interact to
produce prototrophs, as detected by plating into minimal medium, would be
an indicator that the combination was heterogeneous for an aberration. Since
in these experiments, both "parents" were exposed to treatment, each plating
was equivalent to the testing of two chromosomes for the occurrence of an
aberration. No marked variation in the yield of prototrophs was noted in
tests involving 121 mustard- and 28 x-ray-treated chromosomes. This can
scarcely be regarded as an adequate sample in view of the stringent selection
imposed by the technique, which might be expected to eliminate any aberra-
tion types which are even slightly less vigorous than the normal. This con-
sideration is especially relevant in view of the "hemizygous" condition of any
aberrations in the probably haploid vegetative cells. These studies will be
continued.

How Many Segregants per Zygote?

In the experiments detailed in this paper, recombinants were obtained from
different cell types which were exposed to each other in an agar medium.
Therefore each prototroph recombinant colony seen by the experimenter marks
the site of formation of a zygote. The question may immediately be raised
whether there are at that site other recombination classes which, by virtue of
their biochemical deficiencies, remain dormant within the prototroph colony on
the minimal selective medium. This is equivalent to inquiring whether there is
but a single viable product of meiosis (as in megasp orogenesis in many higher
plants) or more than one, as in the ascomycetes. The solution to this problem
would be of special interest in relation to the possible occurrence of four-strand
crossing over. In addition, if an appreciable proportion of prototroph colonies
consisted of two distinct segregation types, it would be necessary to isolate
these types for the collection of segregation data.

There are at least three ways in which a zygote might yield more than one
haploid recombinant. Firstly, the zygote might be capable of proliferation in
the diplophase (or sporophyte), leading to the concurrence of several diploid
cells, each of which might undergo meiosis independently, and by chance yield
several segregation types. Secondly, a single zygote might produce, after
meiosis, in addition to the prototroph, the complementary multiple mutant
class. Thirdly, in a system of four-strand crossing-over, there might be two
supplementary prototroph recombinants differing in the segregation of factors
such as Lac and V\ for which the diploid was heterozygous.

Obviously, the proper investigation of these possibilities requires that one
stringently avoid contamination of one colony with another. For this reason,
the cell suspensions used were diluted so as to yield only about five to ten
recombination colonies per plate.

157

520 JOSHUA LEDERBERG

Crosses were made between Y-40 and Y-53 (B-M-T + L + B 1 + Lac+Vi T XB + -
M + T~L~Bi~Lac~Vi 8 ) on ^-containing minimal agar medium. As already
noted, about 90 percent of the colonies from such a cross are B + M + T + L + Bi~.
The theoretical complementary class would be B~M~T~LrBi + . Because of its
nutritional deficiencies, it could not be expected to proliferate on the minimal
medium even had it been produced after meiosis. The possibility remains, how-
ever, that a few cells of this constitution might still be present among the io 8
or so Bi~ cells of the predominant type in a colony. By plating such colonies
into medium lacking B\ but containing biotin, methionine, threonine and
leucine, the B{~ cells would be suppressed, while the postulated multiple mu-
tant type could form colonies and be recovered.

The experiment just described was carried out, testing 52 colonies for their
content of other cell types. In general, a thiaminless colony could be shown to
contain from 10-100 cells capable of forming colonies on the B, M, T, L
medium. However, in each case investigated these have been shown to be in-
distinguishable from the Y-40 parental B~M~ type, and must be presumed tc
arise from a surprisingly low degree of contamination of the colony with these
cells from the heavily seeded plate. A few colonies were found which could be
characterized as reversions from B\~ to B\ + . These experiments are then, in-
conclusive with respect to the occurrence of complementary genotypes in the
same colony. With appropriate stocks, not as yet available, it should eventu-
ally be possible to manipulate the situation so that the complementary type
could be recovered selectively, excluding both parents and the predominant
recombination class.

A search for supplementary types was conducted with the same crosses,
except that colonies appearing on Bi agar were streaked out directly on EMB-
lactose agar to determine whether any of them were heterogeneous for Lac.
In some cases, a number of isolated colonies from each EMB-test plate were
then also tested for homogeneity with respect to Tj-resistance. About 90
colonies were so tested; only one colony was found containing both Lac + and
Lac~ cells. It is impossible to be certain that, with this low frequency, the
single colony which was picked was not actually derived from two distinct
zygotes. These experiments cannot be considered as bearing critically on the
question of the occurrence of two- or four-strand crossing over because of the
absence of information concerning (a) the viability of more than one meiotic
product and (b) chiasma interference. The results do, however, justify the
technique of picking the prototroph colonies directly, and testing them without
further purification for the collection of segregation data.

A Comparison of Sexual Recombination and Transformation

The occurrence of recombination types has been interpreted by us (Leder-
berg and Tatum 1946c, Tatum and Lederberg 1947) as a consequence of cell
fusion, "karyogamy" and meiosis with crossing over. This is, however, not the
only allowable interpretation of the general phenomenon of the occurrence of
new character combinations. By analogy with the systems which have been
described in pneumococci (Avery, MacLeod and McCarty 1944) and other

158

SEGREGATIONS IN ESCHERICHIA COLI 521

strains of E. coli (Boivin and Vendreley 1946) one might postulate that
genotypically distinct cells interact not through cell fusion, but through the
release of "transforming substances" diffusing through the medium. Such
transforming substances would have the property of inducing or directing mu-
tational changes in the cell receiving them so as to lead to what appear to be
recombination types. Our inability to separate such postulated transforming
substances from the cells themselves is not proof of their absence but could be
due to their lability in our hands.

In previous publications, certain reasons were given for the rejection of the
transformation hypothesis in favor of a picture of cell fusion, and so forth. It
was not our intention thereby to state, with clairvoyant insight, that no in-
vestigator will be able to duplicate the results which we have reported, using
instead of living cells extracts specially prepared. It is, rather, our view that
since we have been able to demonstrate no appreciable point of difference be-
tween the features of gene exchange in this strain of E. coli and in the classical
materials of Mendelian experimentation, the most economical conclusion is
that the mechanisms involved are also similar. In the absence of more detailed
information on the behavior of transforming systems, a critic would be free to
impute to such systems all of the properties which have been found to charac-
terize the genetic system of E. coli, K-12. While this would be tailoring the
cloth to suit the customer, it cannot be disputed that the only conclusive
method by which it could be shown that cell fusion underlies gene recombina-
tion would be a direct cytological demonstration. The rarity with which the
presumed zygote occurs, however (as indicated by the low frequency of effec-
tive recombination types) is very discouraging to attempts to find and char-
acterize the "fusion-cell," at least in the present material.

Certain genetic experiments were performed in an attempt to characterize
further the behavior of this system. On the transformation hypothesis, one
must attribute the rarity of the imputed transformations primarily to re-
stricted conditions for susceptibility to the transforming factors released into
the milieu. Otherwise, one would expect to find "transformations" for single
factors much more frequent than those involving more than two factors. A
glance at tables 4 and 6 illustrates that certain "multiple transformed" types
are much more frequent than singly transformed classes. Under these con-
ditions, one might also anticipate that genetic materials from two different
kinds of cells could mix in the medium and together transform a third. In a
mixture of three cell types then, one should find cases where genes from all
three have combined. Using Lac and V\ as markers, this type of experiment
was set up in several different ways, as summarized in table 7. Pairwise, proto-
trophs can be formed only from biochemically distinct and nonover lapping
parents. Combinations of B~M~ and of T-L~Br were arranged so that taken
two at a time they were heterozygous either for Lac or for V\ but not both. For
example, a mixture of B-M-Loc-Vf, T-L~BrLac-Vi a and T-LSrLac+Vf
was plated. Prototrophs could be formed by recombination between either