L. L. CAVALLI SFORZA and J. LEDERBERG Genetics of resistance to Bacterial inhibitors SYMPOSIUM INIBITOR] DI CRESCITA E CHEMIOIERAPIA GROWTH INHIBITION AND CHEMOTHERAPY 7-11 SETTEMBRE 1953 ISTITUTO SUPERIORE DI SANITA ROMA - VIALE REGINA MARGHERITA, 299 - ANNO 1953 L. L. CAVALLI-SFORZA ISTITUTO SIEROTERAPIGO MILANESE & ISTITUTO DI MICROBIOLOGIA DELL’ UNIVERSITA, MILANO and J. LEDERBERG DEPT. OF GENETICS OF THE UNIVERSITY OF WISCONSIN, MADISON, wIs. (*) GENETICS OF RESISTANCE TO BACTERIAL INHIBITORS. I. - INTRODUCTION. The development of new antibacterial agents has brought to the fore- ground a new aspect of evolution in microorganisms, viz. drug adapta- tion. The practical importance of this phenomenon is so great, that much study has been dedicated to the problem of its origin, at a time when genetic analysis of microorganisms is in its infancy and the prejudice that their genetic systems are thoroughly different from those of the more fully investigated higher organisms is still widespread. Thus it is perhaps not too surprising, that the issue has from the beginning been a highly controversial one, and that the discussion has been centered around motives which are essentially similar to those prevailing in the last century on theories of evolution in general. It may be added that present knowledge allows one to discriminate much of the current prejudice on microbial genetics and thus to settle a fairly large portion of the controversy on drug adaptation. Most microorganisms can, in fact, show at least some degree of adaptation to most of the drugs, to whose action they are subjected. Rarely adaptation reaches a very high degree (i.e., to high concentration of the drug) after one or few exposures to the drug of a bacterial popula- tion. More commonly, adaptation develops gradually, but even so it may often result, at the end of a long process, in a high level of resistance being attained in the adapted strain. Whether adaptation occurs gra- (*) The work at Madison was supported in part by a research grant (E72-cq4) from the National Microbiological Institute, National Institute of Health, United States Public Health Service, by the Rockefeller Foundation, and by the Research Committee, Graduate School, University of Wisconsin with funds provided by the Wisconsin Alumr Research Foundation. This constitutes Paper No 517 in the series of the Genetics Departn::at. — 109 — dually or suddenly depends mostly on the drug, and presumably there- fore on the mechanisms of its action; too little of which is usually known, however, to permit further speculation along these lines. Complete single step adaptation is observed with streptomycin and, perhaps less through- ly, with isoniazide; gradual adaptation was first described in genetical terms for penicillin by Demerec (1943, 1949) and hence called by this author the «penicillin pattern» of drug-adaptation as contrasted with the «streptomycin pattern». Another angle from which the facts concerning drug-resistance can be viewed is the variability of the resistance level gained by a given type of microorganisms in a single exposure to the drug. Survivors of a bac- terial population plated in the presence of a suitable concentration of the drug may show, on successive subculture, an imercascd resistance, as compared with that of the original strain; the measure of the resistance level thus acquir 'd by one exposure (often called first-step-resistance) may show differences between independent isolates. The variability of first-step-resistance is of course likely to be highest for drugs of the streptomycin pattern, these being defined as permitting full resistance to be developed with the first step. However, a variation of the degree of first-step-resistance is observed, even if less impressively, for drugs to which only gradual adaptation is possible, and, apart from variability of the level of resistance acquired, morphological, biochemical, genetical variation can be detected among first-step-resistants to a given drug. Only with one agent, and this perhaps because of insufficient analysis, is no important variability of first-step-resistance encountered, namely azide-resistance in E. coli. A technical detail of some importance is whether drug adaptation is carried out in solid or in liquid media, containing the drug at suitable concentrations. Adaptation in liquid medium for drugs of the «penicillin pattern», by means of successive subcultures in increasingly higher con- centrations of the drug may result in the bacterial population showing an almost continuous rise in the resistance level, although stepwise in- creases are also observed. Increase in steps is forcefully observed when adaptation to such drugs is carried out in solid media, since every suc- cessive transfer on a drug-agar plate with a suitably high inoculum and drug concentration is likely to give rise to a definite increase in resistance of some of the survivors. . Finally, observations may bear on the stability of resistance thus acquired, and here a full range of stability degrees may be observed, though carefully isolated resistant survivors which show some degree of — 110 — resistance on a successive subculture tend to keep this resistance more or less unaltered over a number of transfers. This oversimplified summary of the main facts may serve as an in- troduction to the explanation which have been offered for the behaviour exhibited by microorganisms in the presence of antimicrobial drugs. First let us consider the explanation which is provided by standard genetical theory for adaptive changes in the reaction of an individual or a population to a given environment. Two mechanisms which are not mutually exclusive may come into play: A) Physiological adaptation: every individual] of a given genotype can adapt to a certain range of environmental variation. The new en- vironment often directly induces the changes, as in enzyine adaptation; on return to the old environment the change usually disappears both rapidly and entirely. Adaptation and deadaptation leave the hereditary background unaltered even if carried out repeatedly. B) Genefie adaptation. Individuals of a given genotype are unable to cope effectively with the situation created by a new environment. Mutation or other genctic changes provide the basis of new heritable variation. If one or more mutants can propagate themselves in the new environment, they will initiate clones which will eventually be sub- stituted for the old population. Thus the environment acts by selecting fitter variants, endowed with a hereditary capacity to withstand the new conditions. Genetic adaptation is essentially preadaptation, in contrast with the physiological kind (A) which is postadaptive. The change is not readily reversible as in the preceding instance, because reversion can o¢- cur only by back mutation or mutation at another locus with modifying effects on the primary mutation, and these kinds of genetic changes prove to be generally very rare. These txo mechanisms of adaptive change have been fairly well ascertained in a number of instances and some models have been examin- ed in detail (Monod, 1952; Ryan, 1952. to quote an example of each type). Formal genetic analysis can ascertain which of these two mechanisms operates by an analysis of individual behaviour from homogeneous pop- ulations, i.e. essentially clones of cells. Clones of sufficient size for muta- tion to have occurred in one or more of its members will form an inci- piently heterogeneous population on which selection may act. Deliberate mixture of genetically different lines will provide means of analysis by genetic transfer in order to analyse the genetic nature of the heritable changes. The effort of the geneticist may therefore be directed to the detection of the relative importance of the two mechanisms in the deter- — 111 — mination of drug-resistance, i.e. to the control of pre- and post-adaptive changes, by the use of such techniques. A variety of alternative theories of drug adaptation have been pro- posed; just to summarise the most important ones: C) Direct induction of genetic change by the drug (Linz 1980, others). The change to resistance is a truly genetic one, i.e. heritable etc. but is induced by the drug itself, and is hence of a postadaptive nature, which has no parallel in the above scheme. D) Self-adjustment of enzyme-balances to permit growth in the new conditions without nuclear change or selection (Hinshelwood 1945). Al- though this theory is not substantially different from what has been call- ed physiological adaptation, this author has pushed its application so far as to cancel or minimise the role of selection or mutation. It may be noted that this expectation may represent an oversimplification of the genetic theory, as it tends to draw no distinction between the individual and a population. This theory could therefore be formulated in terms which would include the preadaptation theory. If the population is studied as a whole, and no attempt made to resolve its elements, the enzyme balance of the mass will doubtless be found in the long run to show adaptive adjustments of the kind subsumed under this hypo- thesis. For an understanding of the genetic basis of such changes, how- ever, we must examine the adaptive responses of individual cells, and we shall find that the best studied cases of apparently directed, post adaptive changes of whole populations can be interpretcd in terms of pre-adaptive changes of small numbers of its constituent cells. Both these theories demand a post-adaptive change and ignore pre- adaptation. An attempt to compromise was recently advanced by Yud- kin, under the name of «clonal variation with selection», where it is (implicitly) assumed that phenotype and genotype of a cell are identical, and variation originates from the distribution in unequal parts at cell division of the systems responsible for resistance. Actual cell resistance determines the average resistance of the progeny (phenotype = genotype); selection acts in the same way as in genetic adaptation. Here changes would be preadaptive. It may be added already at this stage that the knowledge on bacterial genetics is strongly at variance with such as- sumptions. Again, however, Yudkin’s approach may be considered as an aversimplification of the genetic theory, assuming an inordinately extreme polygenic picture. ~- 442 —__-_.._. op - Models of drug resistanc Fig. 1. — 118 — Tl - THE ANALYSIS OF INDIVIDUAL CHANGES. A visual aid to the formal analysis of the various interpretations is given in figure 1. We can be satisfied, at this stage, with the analysis of two, or better two pairs of models taken to explain the origin of drug resistance. The first (1 A) shows strict genetic preadaptation. The change occurs prior fo addition of the drug, and is also phenotypically expressed before it. The second model (1 B) differs from the first in that mutation determines only the potential capacity to withstand the drug; the neces- sary enzymatic svstems will however, be manufactured or adjusted only in the presence of the antibacterial agent. Here, genetic adaptation is coupled with physiological adaptation. The third and fourth models (2 A and 2 B) show two varieties of «physiological adaptation» with low and high survival of the mass of the population respectively. In the third scheme, cach cell has the same probability of giving rise to a resistant clone following the addition of the drug, or if the probabilities differ from cell to cell. there is no simple relationship between the probability of survival of different cells and their common ancestry. In both models however. it is assumed that heritable resistance is postadaptive. The essential difference between models 1 A, 1 B on one side and 2 A, 2B on the other is whether or not preadaptation plays a role. I it does, two verifiable consequences follow from this: 1) resistants will appear in clones; 2) it will be possible to accomplish selection of resistant mutants in the absence of the drug. Both these methods have been used experimentally, with results to be summarised now. It may be added that the difference between schemes 1 A and 4 B will demand physiolog- ical investigations subsidiary to the major issue; and that even if the difference between models 2 A and 2 B, as well as between 2 B and the remainder may seem obvious on the basis of the ratio of output of resistants to input of sensitives, the analysis of this ratio may de- mand a more careful scrutiny than one may on first sight be led to believe. Thus situations in which even 100% of the cells seem to adapt may be entirely deceptive, as we shall see later, with incompletely bac- teriostatic drugs, or in other conditions. If it is ascertained that this ratio is low, as is usually the case, discussion may be confined to distin- euishing between models 1 A (or 1 B) and 2 A. CLONAL APPEARANCE OF RESISTANT VARIANTS. Pedigree analysis by isolation of cells after each division has been attempted by Zelle (1942) in the case of the smooth to rough variation — 114 — in Salmonella typhimurium and showed that the mutants appeared in clones. Unfortunately there is a complete bar to this direct approach in drug-resistance of bacteria, owing to the extremely low frequency of mutation for resistance, so that millions of isolation of cells imme- diately after division would be necessary. It is not surprising, there- fore, if that various statistical devices have been elaborated to meet this deadlock. 1) A detailed analysis was first given by Luria and Delbriick (1943). They considered that, if there is preadaptation and therefore re- sistant mutants are to be found in clones, the number of resistant mu- tants in parallel cultures each started from a single sensitive cell and grown under identical conditions should be highly variable and their distribution should indeed be predictable, given certain easy assumptions on the probability of mutation per cell, and the growth rate and lag of mutants. The method was extensively used by a number of authors in a crude way, by comparing the variance of observed numbers of resistant mu- tants in parallel cultures with that expected under model 2 B, assuming this to give rise to a Poisson distribution, as would indeed be the case if, in model 2 B, every cell had the same probability of adapting, inde- pendently of the ancestral history of the cell, In a Poisson distribution the variance should be equal to the mean; in all but three scries of experimental data observed variances of numbers of resistant mutants have instead been much higher than the respective means. This was taken to imply that model 2 B was untenable, and therefore to be dis- carded in favor of models 1 A or 4B. This crude analysis is often re- ferred to under the name of «fluctuation test», The three exceptions quoted are: streptomycin dependence in E. coli (Newcombe and Hawirko 1949) where clones of dependents obviously cannot form in absence of the drug, so that the test is inapplicable; streptomycin-resistance in Staph. aureus observed by Welsch (4952), who also showed the ex- istence of a strong selection against the resistant mutants, so that this case approaches that of streptomycin-dependence: finally, penicillin- resistance in Staphylococcus aureus (Eriksen 1949) where among other difficulties the use of small samples from the cultures made the statistical tests inefficient. These exceptions cannot therefore he taken as proofs of model 2 A. {t must be stressed that the analysis offered by the socalled «fluctua- tion test» is open to two sources of error. One, which was pointed out by Eriksen (4949) and by Dean and Hinshelwood (1932), is based on the fact that no check is available of uncontrollable differences of en- — 145 — vironmental origin between parallel cultures. We shall see later how this objection can be met. For a more satisfactory support of model 1 (A or B) a comparison of the observed distribution of resistant colonies with that expected un- der one of these models would therefore seem essential. The mathema- tical theory was provided by Lea and Coulson (1949), and extended by Armitage (1952) to cover certain necessary side assumptions (phenotypic delay, back mutation, unequal growth rate of mutant and parent). Ana- lysis of data published by Luria and Delbriick, and by Newcombe (4948) on phage resistance showed a tolerably good fit with the theory; a good fit was obtained by Ryan (1952) for mutation to lactose-adaptability (mo- del 2 B) in E. col. As to drug-resistance proper, only one example (New- combe and Hawirko, 1949, streptomycin-resistance in E. colt) was given, and showed a good fit with model 1. 2) A method, which may be called the method of statistical clones was proposed by Newcombe (1949). It consists of the comparison be- tween the number of resistant colonies arising from plates on which bacteria have been allowed to grow into microcolonies prior to the appli- cation of the antibacterial agent (and this is administered in such way as to keep all the cells in their original location) with the number of resistant colonies arising in parallel plates in which microcolonies had been spread just before addition of the agent. Clones of resistant cells would thus be spread over the plate, and while each clone would give rise no more than a resistant colony in the «unspread» plates, its mem- bers would all give rise to independent resistant colonies in the «spread» rise to no more than a resistant colony in the cunspread» plates, its members would all give rise to independent resistant colonies in the «spread» plates. Assuming equal size of the sensitive microcolonies, and calling: r = No. of microcolonies per plate containing at least one resistant mutant, i.e. equal to No. of resistant colonies per «unspread» plate; n = Total No. of microcolonies per plate; m = Mean number of mutations per microcolony: the value of m could be calculated from - =1—en giving m==— In (1 —=) This is approximately _— n ~ ~ | D / WMe=-n when r/n is small. In such conditions, therefore, the mean number of mu- tations per plate would be n m=r. Thus if the total number of resistant colonies per «spread» plate is called R. the ratio R/r will give the aver- age number of individuals per clone. Tt must be higher than unity if models 1 A or B are correct. In Newcombe’s experiment the antibacterial agent used was phage, — 116 — which was nebulized on the preincubated plates, both spread and un- spread. Bornschein et al. tested chloramphenicol resistance in E. coli by this method, using membrane filters as a tool for transfer of the bac- terial incculum from the preincubation plate to the drug-agar-plate (4954). Both experiments yielded high ratios R/r, qualitatively at least in agree- ment with the theory of genetic adaptation. This result does not suffer from the objections raised to the fluctuation test, though it may be claim- ed that model 2 A is not excluded by this criterion if it is assumed that in «spread» plates, where cells are not packed in microcolonies as in «cunspread» plates, resistance may arise more easily. 3) Another approach makes use of the secondary differences be- tween independent mutations for resistance to the same drug which can he brought under observation by further analysis of the resistant colonies. Such differences, no matter whether of a morphological, bio- chemical, genetical character, or simply of degree of resistance should be found, if resistants appear in clones, only between members of diffe- rent clones and not between members of the same clone. Since resistants from parallel cultures will have arisen from inde- pendent mutations, among which some difference may be detected, while resistants from the same culture are likely to be members of the same clone, the comparison of variability for some suitable character between resistants from independent cultures with that between resistants from the same culture will permit a test of the clonal origin of mutants, under the form of a test of the correlation between relatives. If the character tested is quantitative in nature, (e.g., degree of resistance) the test will take the statistical form of a study of intraclass correlation and thus may be performed as a straight analysis of variance by comparing the variance between cultures with that within cultures. This method has been applied to chloramphenicol-resistance in E. coli (Cavalli 1952) where a high correlation between relatives was found using degree of resistance as the differential character. The same approach was used by Mitchison, and pursued further by him in the course of study of streptomycin- resistance in F. colt (1983); here, using various differential characters it was possible to recognize clones within cultures and show that the num- ber of clones was distributed at random among the various cultures. This test is exposed to the objection of confusion between environ- mental and genetic variability in a way somewhat different from that to which the fluctuation test is subjected, if it is assumed that uncon- trolable environmental variation between cultures may favor one or other type. form or degree of resistance, varying with the culture. — ANT — 4) Although we may now be asking too much of the potentialities of environmental variation if we want to explain all the above results by model 2 A plus environmental variation between cultures, and to consider the agreement with the genetical model as purely coincidental, a test of the environmental variation may be useful at this point. Such tests can be devised along lines such as those tried for chloramphenicol- resistance in E. colt, in which both the number of resistant colonies and the average resistance of a sample of them from each of a series of parall.] cultures were tested and the parallel cultures were found to differ significantly both for number of resistants (fluctuation test) and for average resistance (correlation between relatives). If such differ- ences were due to environmental variation between cultures, it would be expected that factors favoring the appearance of a high number of mutants in one culture should also increase their average degree of resistance and hence these two variables should be correlated. Instead a correlation of zero was found, (Cavalli 1982), showing no evidence of environmental variation. This example encourages one to believe that variation of environmental origin between parallel cultures may be ne- gligible, although tests for this source of error should be provided when- ever possible. a) A fechnical device known as replica-plating permits one to show the clonal origin of resistant mutants in a more direct way than any ol the above methods. This method also allows one to obtain the second kind of proof postulated above, namely the establishment of re- sistant strains in the absence of the drug (by sib-selection). It will there- fore be described under the following section. INDIRECT SELECTION. Let us consider, hypothetically, the types of experiment that are capable of separating the environmental and genetical components of variance. To do this, only a simple extension of the design of the fluc- tuation experiments is required. Let us, as before, consider groups of cultures, prepared from small inocula, in a uniform medium. The initial series will be considered as a parental group, whose descendants will be compared in order to demonstrate the hereditary component of vari- ance. For simplicity, let us consider only the two extreme cultures of the parental series, those from which samples give the highest and lowest assays, respectively. It has been urged that such extremes, instead of heing the expected result of pre-adaptive mutation and clonal growth of resistant cells, in the absence of any drug, may instead reflect uncon- trolled physiological differences between the cultures, which influence — 11s — the likelihood of an adaptive response to exposure to the drug (Dean and Hinshelwood, 1952). On this physiological hypothesis, a second cycle of growth must be expected to smoothe out any transient differences in inocula taken from the extreme cultures, so that whatever their own fluctuation, «high» and «ow» filial series should show no correlation with their parents in the average assay of resistant cells. On the genetic hypothesis, however, the result will depend on the size of the inoculum which is taken from the parental «high» and «low» respectively, and distributed to the two series of filial cultures. In particular, the filial ingh serics should again display the saine high average assay, and with relatively little further fluctuation, provided cach sample inoculated was large enough to contain at least one resistant (as calculated from the assay of the parental high). On the other hand, if the inoculum sample is made so small as to exclude the likelihood of introducing a resistant cell into the Glial cullure, the genetic component of variance could not be propagated, and the parental and two filial series should show the same average assays, with a high variance from culture to culture. In principle (though this would be quite tedious in fact) one could adjust the inoculum size so that. on the genetic hypothesis, a resistant cell is moculated into only an occasional filial tube. which is then detected by its high assay. Repetition of this procedure would accomplish a gradual concentration of the resistant cells. relative to the sensitive, and ultim- ately a purely indirect selection and purification of the resistant mutant. To our knowledge, these experiments have not yet been executed in just this fashion. although the design should provide a definite test, without serious departures in the program of experimentation, of the criticisms of the genetic inferpretation of Muctuation analysis. These hypothetical experiments are discussed here in order to introduce a radically different approach which is. however, based on the same theoretical foundation for the separation of environmental and genetic variability. ‘Phe hypo- thetical experiments involving the tedious manipulations of many indi- vidual tube cultures can be executed much more expeditiously on agar plates by means of a fechnieal artifice, replica plating. REPLICA PLATING, The most serious limitation to the treatment of these problems ob- viously concerns the purely technical difficulty of handling very large numbers of cells and families under adequate control. By previous me- thods, the clones of resistant bacteria have been identifiable only as statistical entities, or else by the destruction of their preponderant sensi- tive neighbours. The technique of replica plating greatly mitigates this — 149 — difficulty. Samples from innumerable clones, immobilized on an agar surface, can be trasferred or replicated to a series of copy plates without seriously disturbing their spatial arrangement. This is accomplished by using a sheet of velveteen, with its dense pile of vertical fibers as the vector for the imoculation. In practisc, the initial agar plate, with its pattern of surface bacterial growth, is inverted with slight pressure onto a sheet of velveteen, mounted on a cylindrical support. The initial plate is then removed, and replaced by a series of fresh plates. Each of these will pick up an impression of bacteria from the velvet which subsequ- ently develop into a remarkably faithful rendition of the original pattern of bacterial growth. The accurate replication of this pattern is most obvious when it con- sists of discrete colonies, but the principle applies equally well when the growth on the initial plate is confluent owing to the character of its incculum. The original plate is left sufficiently undisturbed by taking a velvet replica that it can be used repeatedly for the same or other purposes. Two adaptive responses have been studied in detail by means of replica plating technique, resistance to streptomycin and resistance to the phage T, in Escherichia coli strain K-42 (Lederberg & Lederberg, 1952), with concordant results. Plain nutrient agar plates were spread with several thousands of sensitive bacteria, so that a smooth, confluent growth developed after 24 hours incubation, while no resistants were in- cluded in the inoculum. A series of replicas was then made to plates of sticptomycin agar. In order to ensure objective registration of the plates, pins were included in the support blocks to mark the agar surfaces. After 24 hours, resistant colonies would often develop on the replica plates. More than half of these were observed to oceupy superimposable posi- tions on the successive replicas! This could be explained only by the derivation of these resistant colonies from localized clusters of cells on the original plain agar plate. These clusters are, of course, the clones subsumed by genetic theory; the physiological theory can offer no plaus- ible explanation for such clusters in a uniform film of growth, Rather than argue this point more fully, however, we proceed to the next and more conclusive stage of the experiments. It will be recalled, and must be emphasized here, that the original plate has not been damaged, except for the removal of a sample of its growth, nor has it had any knowledge of the streptomycin to which only the sample on the velvet has been exposed. Nevertheless, the location of many vesistant clones has been revealed by the congruent patterns of colony development on the replicas on streptomycin-agar, Let us -— 120 — now assume that the conditions of growth on the original plate have been adjusted so as to permit the development of just one clone per plate. on the average. Let us assume also that the registration of the plates and the technique of transferring inocula permits us to pick an area one-hundredth of the total film of growth (i. e. about 60 mm?) and still include the entire resistant clone. It is evident that by thus retaining all of the resistant cells on a plate, and discarding 99% of the sensitives we could accomplish a 100-fold enrichment for the resi- stant relative to the sensitive bacteria (provided of course that, as on the genetic hypothesis, these bacteria are actually present prior to treatment with any drug). This is, in fact, precisely what was observed, as told from the assays of cultures grown from such indirectly selected inocula. The practical enrichment factor is actually closer to 41: 1000 (6 mm?) than 1:100, but this is counterbalanced by the difficulty of controlling conditions to permit just one detectable clone per plate. However, the advantage of this procedure and one which obviates the necessity for quoting any numerical details is that it can be reiterated, by plating appropriate dilutions of the first enrichment culture, locat- ing the resistant clones by replica plating, and picking the correspond- ing sites for a further cycle of enrichment. It is evident that even with the very low incidence of approximately one streptomycin-resistant cell per billion bacteria, it requires (and actually did require) only three cycles of indirect selection to increase the proportion to one per thous- and. At this point, the inoculum for the next stage may be diluted to the point where discrete colonies will appear on the plain agar, and yet one or more of them will be resistant. This colony is quickly identified by a final round of replica plating and having been located, yields to conventional methods of purification. In this way, streptomycin and virus-resistant cultures, respectively, were isolated by purely indirect se- lection. The replica plating technique was also put to work to facilitate tests of the unifermity and stability of these resistant cultures after they were once purified. The one detail of the foregoing account that must be kept in mind is that the indirectly selected bacteria (or rather their ancestors) have never been exposed to the drug. The drug was used exclusively to locate the resistant clones by tests of their sibs, transferred via velvet. In fact. the manipulations to which the selection line has been exposed consist only of several cycles of plating on plain agar, the removal of a sample by a sheet of velveteen, and after an appropriate site has been revealed by comparison of the replica plates, picking with a platinum loop to plain — 1214 — broth. The same manipulations lead, in one case to a streptomycin-re- sistant pure culture, in the other to a virus-resistant, depending on the auxiliary procedures which do not directly affect the main selection linc. There is no opportunity whatsoever for any chemical reaction of the drug to play a role in the establishment of the resistant mutant. We do not address ourselves to the question whether individual resistant bac- teria may not in fact be only uniformly but potentially resistant until they actually encounter the drug. We note, however, that each cell of the indirectly selected culture promptly develops into a colony on the streptomycin agar, which destroys all but a minute fraction of the pa- rental culture. In addition, this potential is inherited for an indefinite number of generations to all of the progeny, prior to any exposure of these cells to the drug. When taken together with the evidence of clo- nal occurrence from the first stage of the experiment, the success of indirect selection provides incontrovertible evidence for the genetic theory jor those systems to which it has been applied. As in previous sections, we must give some consideration to the possibilities of a spurious negative result. The resistant mutant must not be drastically selected against on plain agar —- we have no hope, for example, of isolating a streptoniycin dependent mutant by this method. It is also ossential that single resistant cells from the initial plate (as- suming the genetic hypothesis) be unequivocally identifiable on the selcc- five replica plate, and these not be confused with mutants that might ultimately develop from populations growing on the latter, where selec- tion is insufficiently drastic. If indirect selection should fail when ap- plied to an untreated population, but sueceed in extracting added resi- stant bacteria, originally isolated by direct selection, it would be neces- sary to concede that direct selection is relevant to the development of resistance. Among the most obvious possibilities that would have to be excluded by further analysis is sequential selection. in two stages, for cases when only the first step of resistance is likely to be present as such in unselected population. However, the positive results already obtained are quite unambiguous, and further discussion of spurious negative re- sults may be deferred until specific details are submitted for adjudication. Drug-resistant variants have alsa been secured in the absence of the drug by selection for correlated characters (English & Mc Coy. 41981. Gale, 1949; Goldstein, 1952). This approach also confirms the notion of preadaptation, but is not a rigorous proof of it, as the unusual phy- siological conditions could be regarded as the specific inciters. — 122 — TH. - DIRECT GENETIC EFFECTS OF DRUGS The proven role of mutation and selection in the development. of drug-resistance in bacteria, and the vague contrary evidence that has been offered serve only to accentuate interest in a few clearcut cases of specific direct genetic effects of antibacterial agents on microorganisms. We do not refer here to the possibility that some antibiotics may have some incidental mutagenic activity of a nonspecific character (Newcombe & Mc Gregor, 1954). Streptomycin exerts a most remarkable influence on the unicellular alga, Euglena gracilis (Provasoli, Hutner and Pintner, 1951; cf. Pringsheim, 1952) as well as on some higher plants. Euglena cultures exposed to streptomycin suffer a rapid and irreversible bleach- ing, which results from the destruction of the chloroplasts. Fortunately, Euglena can be cultured heterotrophically, but the apochlorosis is perma- nent, and the chloroplasts are never restored. The organisms which have been found to respond to streptomycin in this way are very limited in number, perhaps for incidental reasons, and none are quite suitable for zenetic analysis, so that it cannot be asserted where the streptomycin acts. It seems most likely however that this is a direct destruction of the chloroplasts, which have been believed to possess considerable genetic autonomy from other (not entirely conclusive) experiments. In a sense, the apochlorosis is a result of chemotherapy, the «curing» of the plant of its plastids (sec discussion in Lederberg, 1952), In somewhat analogous fashion, the exposure of yeast to acriflavine induces the development of sinall colony variants, which have evidently lost part of their (mito- chondrial) respiratory apparatus. In this case, morphological analysis is incapable of demonstrating the direct locus of effect of the acriflavine. Which may be more complex than the simple destruction of the target organelle, (Ephrussi, 1954). The end effect, like that of streptomycin, is believed to be loss of a genetically autonomous particle, in this case popu- jation analyses have confirmed that the chemical effects are in fact induc- tive, and not selective. This has been confirmed for the yeast by beautiful single cell experiments, in which individual buds engendered in the presence of acriflavine were shown to develop almost consistently into the small colony variant. A third example concerns the parabasal body of hemoflagellates: Jong known to disappear during the development ot chemotherapeutic resistance to trypaflavine. Circumstantial evidence fa- vors the conclusion that the aparabasal variants are induced by the drug. perhaps by interference with the division of the parabasal granule dur- ing cell fission (Piekarski, 1949), but the problem has not been closely studied from a genetic viewpoint. There is no evidence in any of these — 123 — examples that the induced variant enjoys any selective advantage over the normal wild type, a circumstance that contributes to the clar ity of the proof that they are induced. It is therefore doubtful whether these variants can be regarded as adaptive or drug-resistant, and their discus. sion is therefore not directly pertinent to our paper. A final example lies on the semantically confused boundary of para- sitology and genetics (Lederberg, 1982). Kappa, a determinant of the killer trait in Paramecium aurelia shares many attributes of a cyto- plasmic hereditary determinant on the one hand, and of an intracellular symbiotic microorganism on the other (Sonneborn, 1950). It has been found that kappa is destroyed by cultivating paramecia carrying them in the presence of chloromycetin or of purine analogues (Brown, 1980; Jacobson, 1952). This is no more decisive for the « genetic quality » of kappa, than is the effect of streptomycin for the chloroplasts. The four examples share the principle that an apparently autonomous. but some- how dispensable, hereditary particle is lost under the influence of the drug, and illustrate a surprising application of chemotherapeutic princi- ples to the problems of genetics. The importance of such cases provides all the more reason why possible examples in the bacteria should he supported by ample proof. It is difficult to quote any at the present time which are not beclouded by the undisqualified possibilities of mu- tative, selected change. Three which perhaps deserve closer attention are the attenuation of virulence of the crown-gall bacterium in the pre- sence of amino acids (van Lanen et al., 1952); the development of small colony variants of E. coli in the presence of naphthaquinone and other chemicals (Colwell, 1946) and antigenic changes in streptococci grown in the absence of acetate (McIlroy, Axelrod & Mellon. 1948). For want of more information, Voureka’s findings may also be sum- marized at this point. Her first finding was that a penicillin resistant strain of Staphylococcus, became sensitive when exposed to DNA ex- tracts of sensitive bacteria. At the time, this was thought to be possi- bly analogous to the pneumococcus transformation. It was later found, however, that extracts of the resistant bacteria had the same effect. Most recently, Voureka (4952) has reported a single cell analysis of staphylo- cocci exposed to various injurious agents such as chloramphenicol, peni- cillin, nitrogen mustard, and hydrogen peroxide, each of which seems to induce the same syndrome, which includes increased sensitivity to peni- cillin and streptomycin, and a loss of coagulase formation, mannitol fermentation, and virulence. The mechanism of these changes is ob- scure, but in view of the non-specificity of the inciters is clearly not adaptive. Voureka notes that the cultures have also lost their phage- — 124 — specificity and this, together with some of the circumstances noted for the development of colonies from single treated cells suggests that the phenomenon may be related to the induction of lysis by latent bacte- riophage (Lwoff, 1952). The report of morphological changes induced by treatment with chloramphenicol + antiserum was confirmed by Man- ten and Rowley working on E. coli K-12 (1953). Again these changes had no adaptive value. IV. - THE ROLE OF PHYSIOLOGICAL ADAPTATION The experiments which were referred to in section II give strong experimental support to the belief that genetic adaptation plays a major role in the origin of drug resistance. They were not, however, designed to test effects predicted by schemes of the kind postulated under 2B, because they deal essentially with cases where only a small minority of cells survive the action of the drug or at least develop, im its presence, into a visible colony. which has acquired some stable degree of resistance as shown by subculturing in the absence of the drug. Such experiments can discriminate between models 1 (A or B) and 2 A, in the sense that the first two give rise to a precise hypothesis which can be accurately tested, and which has not been contradicted by any of the available, critically collected data. Complete exclusion of model 2A would demand a perfect fit of model 1 and for this more work may be necessary, at least for some systems. It is of importance, however, that none of the analy- ses sofar reported are capable of throwing doubt on the existence of mechanisms of type 1. Model 2 A may arise either as a consequence of direct induction of the genetic change by the drug, or as an extension of model 2B under consideration of simultaneous partial killing. Model 2B has the advan- tage that it can be more easily put to a straight experimental test than model 2A. It is a classical basis for the description of enzyme adapta- tion to carbon sources, it should be noted, however, that in all cases analysed the rate of deadaptation would seem to be of the same order of magnitude as the adaptation rate. A brilliant experimental confirma- tion that cells adapt has been recently given by Benzer (1983) for the adaptive synthesis of beta galactosidase in FE. coli ML. Experiments showing that this is possible for drug resistance are scanty. Baskett (1952) was able to increase resistance to proflavine in B. lactis aerogenes by adding repeatedly small sublethal amounts of the drug to a growing culture. Cells could thus be induced to grow at a drug concentration which, when applied to unprimed cultures would cause a killing of — 1258 — L- 10°) cells, i. e. a level at which one may hope to select some re- sistant mutants, if any were available. Adaptation was rapidly lost in subculture on drug free medium, This could be a good example of stan- dard physiological adaptation. Eagle, Fleischman and Levy (1952) found somewhat more stable changes in resistance by selecting survivors in plates containing suble- thal concentrations, i.e. just above the threshold for some observable inhibitory action. The change in resistance of the survivors seemed to be correlated with the level of drug used for selection. Heritability of such changes should be tested before conclusions on their nature can be reached. This result is, incidentally, not entirely in agreement with observations on similar lines by Hughes (1952), Demerec (1945), New- combe and Hawirko (1949), It may be useful to note, at this point, that when incompletely bac- teriostatic drugs, or drug concentrations, are employed, the applicability of model 2B may be erroneously claimed. Subvisible growth of the cells seeded may result in an actual input of sensitive cells far higher than that assumed on the basis of the inoculum employed. One clear cut case of adaptive response to drugs is penicillinase for- mation in B. cereus (Pollock). In this case, it is obvious that addition of subminimal amounts of penicillin would protect a culture from the later addition of greater amounts of the drug. Knox and Collard were able to show that growth at higher temperatures (42°) which blocks penicillinase adaptation is far more easily inhibited by penicillin than growth at a temperature permitting adaptation (at 37°); preadaptation of the culture greatly increased resistance at 42°. On the basis of the early work on drug-resistance in Paramecia it has also been claimed that stable changes in resistance without the in- tervention of nuclear changes are possible (Dauermodifikationen). The recent work on induced antigenic changes in Paramecium aurelia (Son- neborn, Beale), and other Protozoa (for a review, Harrison) is in formal agreement with model 2B, although here the change induced environ- mentally, and cytoplasmic in nature, is not, or not clearly, adaptive. On the other hand, extrapolation to bacteria of the conclusions reached in Paramecium may be dangerous. The limited knowledge available for both types of organisms indicates differences in the immediacy of nu- clear control over cytoplasmic activity. It would not be too surprising if the stability of cytoplasmic systems and the latitude of variation under environmental influence and outside nuclear control were, on average, different in the two categories of unicellular organisms, though the pre- sent evidence is insufficient to warrant such a generalization. — 126 — Summing up this scanty evidence, bona fide physiological adapta- tion to drugs can hardly be expected to be missing, althoug convincing evidence that physiological adaptation of a high degree can be reached is lacking. Conditions usually employed for selecting resistant mutants, as for instance high drug concentrations with large inocula are of course of the worst type both for detecting physiological variation, and for al- lowing it to play freely. However one finds, especially at lower drug concentrations, a spread of individual variation in the ability to grow visible colonies, the range of this variation depending on the particular drugs and organisms employed. Part of this variation may be stocha- Stic, i. e. the variation in the total amount of residual growth of a cell in the presence of a drug may be due to random events of a physiological nature; part may be due to the variation in the initial physiological con- ditions of genetically homogeneous cells; part to small genetic differen- ces, which are not easily measured. Two facts should be stressed: 4) that no genetical theory could reject the effects of environmental varia- tion, or previous history of the cells (as in penicillinase adaptation); 2) that the existence of physiological variation, both immediate and « po- tential » (adaptive) does not conflict with the fact that high and stable levels of resistance can be reached, and have been proved to be reached, by mutation and selection, VY. - THE ANALYSIS OF GENETIC INTERACTION BETWEEN INDI- VIDUALS (Genetic transfers from one individual to another, i.e. the genetical consequences of sexuality in higher organisms. has been clearly de- monstrated in a variety of microorganisms. While in Fungi and Pro- tozoa, in which strains with a sexual cycle are available, the process parallels closely sexuality of higher organisms, in several bacterial spe- cies a new kind of genetical interaction has been disclosed, in which one strain can effect genetic transfers to another Qvith a relatively low fre- quency, but at any rate higher than spontaneous mutation in the recei- ver) without direct interaction between the cells of the donor and receiver. Aay character difference tested js transmissible, although as far as ,-ve- sent experience goes, the transfer seems to affect only one genetic charac- ter at time. This kind of transfer has been named transduelion (Leder- berg, 1952). In Salmonella typhimurium the agent of transfer, which acts purely as a passive carrier, has been found to be a phage. The term transduction is also meant to include an earlier described type of gene- tic interaction, carrying traditionally the name of transformation, which — 127 — has reproducibly been obtained in two bacterial species, Pneumococeus (Griffith 1928; Avery, McLeod, McCarthy 1943) and Hemophilus (Alexan- der and Leidy 1983). In this particular type of transduction no special carrier appears to be involved other than what is believed to be the specific inducer of the genetic change, i. e. a highly polymerised DNA. Another type of genetic transfer has been demonstrated in bacteria, and specifically in E. colt (Lederberg and Tatum, 1946): namely recombi- nation. Here direct interaction between the cells of the two strains is necessary. Both transduction and recombination occur as a rule with low frequency, so that the progeny, i.e. the result of the interaction must usually be selected from a mass of unchanged parents. The main formal difference between transduction and recombination is that. while in transduction only the character specifically selected for is recovered, in combination the transfer of the selected character is usually accompa- nied by the simultaneous transfer of other unselected markers according to a complex, but well reproducible pattern of reciprocal linkages. At present the genetic system of the more completely analysed strain of E. coli is best interpreted as formed by twe chromosomes, recombination being the result of mating between compatible cells, followed by a pola- rised segregation which tends to eliminate preferentially segments from one mating type. Heterozygous individuals segregating for various mark- ers have been recovered, as well as apparently fully diploid individuals. The diploid state is usually unstable, and the normal condition of the cell is believed to be haploid (though often multinucleate). (Lederberg et al., 1952; Cavalli, Lederberg and Lederberg, 1953). While the type of analysis permitted by recombination in E. coli k-12 closely approximates mendelian analysis in higher organisms, trans- duction offers opportunities for another, more restricted type of genetic analysis which may, however, show certain advantages over the more strictly mendelian one. MENDELIAN ANALYSIS. It is important to distinguish between what may be expected, from a mendelian analysis, of drug-resistances which show the streptomycin pattern, and those showing the penicillin pattern. Or, more exactly, the distinction must be drawn between inheritance of first-step-resistance Gf truly such) and that of multi-step-resistance (for drugs to which gra- dual adaptation is the rule). In the first instance we expect, if the theory of genetic adaptation is correct, that the genetic change to resistance is inherited as an all-or-none character, i.e. the progeny of first-step-resi- stants crossed with sensitive should give rise only to sensitives and resist- — 128 — ants, the progeny rembling closely either one or the other parent in respect of drug-resistance. Crosses of resistant x resistant will give rise entirely to resistant progeny, unless of course the changes in resistance represented in the two parents have originated through independent mutation at different loci, i.e. the two changes are not allelic. In the case of multistep-resistance, successive adaptation must have been due to the accumulation of successive mutations, presumably at a number of different loci. This polygenic picture was early advanced by Demerec to explain penicillin resistance in staphilococci; the later availability of crossing techniques in E. coli allowed one io test this hypothesis in the latter organism. At least five cases of drug-resistances have been analysed in some detail, two of them following the streptomycin pattern and three the penicillin pattern. Streplomycin. — Various types of changes to streptoumycin-resistance (or dependence) are possible in E. coli. all appearing as first-step muta- tions. Very low degrees of resistance, which are considered adaptive by tibson and Gibson, have not been followed genetically, except in an incidental observation by Newcombe and Nyholm, on a fully resistant- strain which was presumably a second-step mutant. Differences are detec- table between strains endowed with high resistance, and Mitchison (1983) distinguished four types on the basis of growth rate, stability and beha- viour at high concentrations. Demerec (1950 a, b) found a variety of changes mostly of a nutritional type accompanying resistance (Sr) and dependence (Sd). It may be noted that changes correlated with the character primarily selected may be accidental, i. e. due to concomitant mutation; in which case primary and secondary character might be sepa- rated by recombination. Unless such an analysis is available, the estima- tion of the number of different alleles or loci cannot be considered as valid, but is likely to be excessive. Smith, Oginski and Umbreit (1949) reported that Sr mutants showed several physiological deviations, includ- ing a failure to show improvement of growth under aerobic conditions. As this is reminiscent of the small-colony variants of yeast, this effect was looked for in our own Sr cultures. All of the Sr mutants which were examined, in our own stocks as well as in the initial culture of the cited authors have displayed a typical response to aeration (Lederberg, un- published). It is possible that the non-aerobie behaviour of the Smith- Oginski-Umbreit culture resulted from a coincidental mutation, The pyruvic acid metabolism of our Sr strains, a defect of which is claimed to be a more consistent concomitant of streptomycin-resistance has not been studied. Mendelian analysis of changes to full resistance, performe:) by se- — 129 — veral authors (Newcombe and Nyholm 1950, Demerec, 1950, 1984, etc.); showed: 1) clear cut segregation into sensitives and resistants in the cross sensitive x resistant; the frequency of segregation, which depends on the mating-type of the parents (polarity of the cross) and on linkage of the resistance markers with the markers used for selecting recombi- nants need not concern us here; 2) perfect reversal of the ratio sensitive / resistant in the progeny of crosses in which the resistance marker is swit- ched over from one parent to the other, keeping polarity unaltered (New- combe and Nyholm); this is an important requirement of the mendelian theory; 3) allelism of all the various St mutations tested (Demerec); 4) al- lelism of Sr and Sd, although some room for pseudoallelism may still remain (Newcombe and Nyholm, Demerec); 5) linkage with the methio- nine marker, which has recently been confirmed inspite of the difficulties met in mapping this locus (Fried and Lederberg, 1952); 6) streptomycin resistance is completely recessive to sensitivity (Lederberg, 1981); at- tempts to secure heterozygotes of Sd with Ss or Sr have failed, perhaps because of physiological incompatibility between these alleles. The mutation rate to streptomycin resistance has been determined at about 107° per bacterial generation in a number of bacterial strains and species. H. P. Treffers (personal communication) observed that a bio- tin-phenylalanine requiring mutant of E. coli K-12 displayed a much higher rate of mutation to Sr (about 10° per generation), Crossing expe- ronents have shown that f) this high rate is unrelated to the nutritional mutations im the stock: 2) the mutation occurs at the same S-locus; 3) the high rate is governed by a mutability modifier, separable from the S lo- cus by recombination /P. D. skaar and Lederberg, unpublished). A more detailed study of these effects is being made in Prof. Treffer’s laboratory. Aside, — E, coli k-t2 can mutate in a single step to a small degree of resistance to azide. This mutation was used by Lederberg (1950) as a selective marker for recombination, in association with streptomycin. Mutation-rate to azide resistance (about 107) is fairly high. and the effi- ciency of this crossing technique can be increased by using T, phage resistance in coupling with Az; the Sr parent can thus be fully elimi- nated (Cavalli, unpublished). This association was suggested by the close linkage of Azt and Tr. Az having been mapped between Tr and the threonine- leucine markers. Only one Azt locus has been found: also here, the switching over of Az™ from one parent to the other, keep- ing polarity unaltered gives results in agreement with mendelian expec- tations (Cavalli, 41952), Chloramphenicol. — Resistance to this drug in E. coli K-12, obtained through gradual adaptation was analysed with mendelian techniques by — 130 — Cavalli and Maccacaro (1951. 1952). Crossing first-step resistants to sen- sitives gave a segregation into sensitives and resistants, without inter- mediate degrees of resistance in the progeny. Crossing first-step-resi- stants between themselves resulted practically always in the reappea- rance of a proportion of sensitives in the progeny, showing that first- step resistance is here due to mutation at a number of different loci four or five at least). Some crosser between first-steps gave rise to progeny Which showed resistance higher than that of either parent, i.e, recombi- nation mediated the accumulation of the resistance acquired by inde. pendent mutations at different loci. In a more fully analysed cross of this kind. doubly resistant mutants arose in approximately the expected proportion, calculated from the proportion of resistants in the crosses of the parental first-steps to the origina] sensitives. Crossing a second Step to a sensitive, intermediate resistance between that of either parent appeared. On crossing multi-step resistants, obtained by prolonged « training » in liquid medium, with sensitixes, the progeny was found to exhibit degrees of resistance varying over almost the full range from full sensitivity to the full resistance of the resistant parent. Linkage of resistance with the ordinary chromosome markers could be detected in this progeny, as would be expected if several loci were involved. Al- though there was evidence for at least some activity in a number of chro- mosome regions, the maximum concentration of genetic activity was near Lac. Crossing multi-step-resistants between themselves did nol show positive interactions in the sense of increased resistance above that of cither parent, perhaps because the physiological limit to resistance had already been reached by selection. There was instead a spread towards the low valucs of resistance. and some fully sensitive indivi- duals were recovered among this progeny. This shows 4) that full or at Jeast high resistance can be acquired in gradual adaptation through se- veral independent paths; 2) that interaction between independent muta- tions is not necessarily additive. Indeed, after the first resistance- step. al] successive steps acquired by selection may well be, in theory at least. due to modifiers, i. e. genes having no direct effect on resistance by themselves. but interacting with other loci to give augmented resistance. The possiblity of direct accumulation of independently obtained resistan- ces has been observed, however. Terramycin. — This drug is known to show a fairly high degree of cross-resistance to chloramphenicol. In the analysis by recombination of two strains of E. coli K-12 which had adapted to high concentrations of terramycin after a long « training » period, resistance of the progeny ‘to both drugs was tested. Correlation was still apparent in the progeny. — 131 — but it was evident that while some changes had affected only, or mostly, resistance to one drug, others had affected resistance to both. This shows at once the existence of receptors within the cell which are « hit » pre- ferentially, or exclusively, bq one or other of the two drugs, as well as of receptors « hit» by both approximately in the same way (Ca- valli, 1982). Furadroxyl (NF 67). —- T. C. Nelson has analysed, in unpublished experiments, resistance to furadroxyl (8-nitro-2-furaldehyde-2-(2 hydrox- yethyl)-semicarbazone) developed by E. coli K-12. At least four steps can be recognized: the inhibitory level for the parent is 4 pez/ml, and for successive steps 20-30, 40-50, 90-100, 160-200 ng/ml. The survival curves are remarkably sharp, the threshold covering only a few % of the dose. The mutation rates for each step are co. 10° - 10°, uncorrected for plate mutations. A satellite effect was prominent around resistant colonies. No mutants have been found to develop a two-step level of resistance in a single plating. Crosses of Fx* x 1-step Fr showed a linkage of this particular factor to TL, with a sharp segregation into the parental categories. Different first-step mutants when crossed, yielded occasional sensitive as well as two-stepw level recombinants, along with the parental types. It is apparent that multistep resistance is here also controlled by a polygenic system. The incidence of recombination is so low, however, as to sunggest very close linkage of some of the factors. Other drugs have been analysed to a smaller extent. Nitrogen- mustard resistance in &. Coli K-12 is not a single-step-resistance as in E. colt B, and is not correlated with UV or X-ray resistance; it shows an apparently polygenic behaviour on crossing (Cavalli, unpublished), but the instability of the drug prevents a detailed analysis. Valine has a marked antibacterial effect on E. coli K-12 when tested in minimal medium but none in nutrient agar, due to antagonism with isoleucine (Tatum, unpublished; Rowley, 1983). In minimal medium, strains end- owed with a fairly high degree of resistance can be isolated in one step and, on crossing show linkage with M (Cavalli and Bolzoni, unpublish- ed). This marker can be used to select recombinants in minimal. Colicine- resistance has been analysed genetically by Fredericq and Betz. Resistance to eight types of colicines was shown to segregate in a mendelian fashion. Some discrepancy is observed in the segregations given by reversed crosses, probably due to viability differences. One of the colicines studied, K, showed cross-resistance with phage T,, and the resistant locus showed, on mapping, to fall in a position comparable to that ascertained for phage T, resistance by Lederberg (4947). — 132 — TRANSDUCTION ANALYSIS. In Salmonella typhimurium Lederberg and Zinder (1982) obtained the transfer of a number of genetic markers, ranging from nutritional deficiencies of various types and sugar fermentations, to drug-resistance; the only one tested here has been streptomycin-resistance. As no more than one character at a time can be transmitted in these conditions, proof of the transfer is given by the fact that the change to, say, strepto- mycin-resistance is significantly higher than in untreated controls. For streptomycin resistance, transduction shows an increase of 100-1000 times in the number of resistant colonies as compared with the number due to spontaneous mutation in the untreated, sensitive controls. The transfer can occur through a filter (which would bar any genetic exchange in E. coli K-42) and, when the transducing agent is purified, the carrier, i.e. phage, is recovered. Sr is here apparently a single-step-resistance, as could be inferred from the method of selecting the mutant. In Hemophilus influenzae, Alexander and Leidy (4953) could obtain transformations of a kind similar to those classically described for Pneu- mococcus. The spontaneous mutation rate to streptomycin-resistance is stated to be ca. 107’ per cell per generation, and the tolerance level is increased from 3 to 1000 ug/ml in one step. DNA extracts prepared from resistant cells were used for treating sensitive cells for 10 minutes, then DNA-ase was applied. If plating on streptomycin was carried out im- mediately at this stage, transformation of sensitives to resistants could be observed only with a low concentration of streptomycin (10 y2/m)); but the resistants thus obtained showed the same degree of resistance as the DNA-donor. If treated cells were incubated for two hours before plating on streptomycin, the rate of transformation was identical in pla- foes with high and low concentration of streptomycin.