RECOMBINATION ANALYSIS OF BACTERIAL HEREDITY J. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY Department of Genetics, University of Wisconsin, Madison, Wisconsin I Five years ago, at this Svmposium. E. L. Tatum I , u - , L and J. Lederberg first reported some preliminary experiments on genetic recombination in Esch- erichia coli K-12 (1946). This report will review subsequent developments in this research, with special emphasis on the application of recom- bination analysis in bacteria to problems of general genetic interest. The standard textbooks of bacteriology have emphasized simple fission to the near exclusion of other modes of bacterial reproduction. This is not surprising in view of the confusing aggre- gate of conflicting and unconfirmed claims of morphological observations interpreted as sexual processes (reviewed by Bisset, 1950). More authoritative claims of bacterial fusion have ap- peared (Braun and Elrod, 1946; Stempen and Hutchinson, 19511 but none with the essential \ concurrence of genetic investigation. On the I other hand, as we shall see, there is so far no satisfactory demonstration of the morphological .< basis of genetic interchange in E. coli, but the convergence of these lines of work is imminent. E. coli is a favored organism for genetic work. It is not dangerously pathogenic: it grows rapidly in simple media in highly dispersed form; it is an infrequent aerial contaminant, and is readily identified. It has a widespread habitat, and a great many strains are easily isolated. In ad- dition, we have a considerable background of biochemical information, and the organism is subject to genetic variation in a variety of readily scored characters such as resistance to viruses and chemicals, fermentative reactions, and nu- tritional requirements or auxotrophy. These ad- vantages are not so unique, however, as to justify neglect of other organisms, especially if sweeping generalizations on bacterial heredity are to issue. RECOMBINATION IN Escherichia coli K-12 Owing to their rarity, genetic exchanges in I E. co& populations require selective methods for their demonstration. A preferred technique :I uses auxotrophic mutants, unable to form colonies \ on a defined minimal medium. Small numbers of prototrophic (nutritionally non-exacting) cells (As an alternative method, selection based on specific "drug"- resistances, has also been SIICC~SS- fully applied, as reported by Lederberg, 1950a.) can be selectively isolated in the presence of a preponderance of auxotrophs byinoculating washed cell suspensions into minimal agac: a single plate can be used to screen as many as lo9 cells for the presence of a single prototroph. The latter is relevant to recombination insofar as prototrophs, A+B +, form one of the classes of recombinants expected from the exchange of factors between two distinct auxotroph mutants, viz., A-B+ and A+B-. Recombinant prototrophs are also expected to show various assortments of any additional unselected markers that differ- entiate the auxotrophically distinct parents. This design has been successfully applied to several strains of E. coli, especially "K-12," isolated from human feces in 1922 and maintained since then at Stanford University as a typical strain for laboratory demonstrations (Tatum and Lederberg, 1947; Lederberg, 1947). Techniques for the isolation of auxotrophic mutants have im- proved steadily; many of them, including the use of penicillin as a selective agent in minimal medium, are documented elsewhere (Lederberg, 1950b). The mutants have generally been iso- lated from the survivors of drastic treatments with X-rays, ultraviolet light, nitrogen mustard or other mutagens. To obviate the possible con- fusion between recombinants and those proto- trophs originating by spontaneous reversion, doubly or multiply auxotrophic mutants have been used as the parents in most of the experiments. They were obtained by repeating the mutant- isolation techniques on mutant stocks already established. It may be worthwhile to mention the details of a typical recombination experiment, if only to point out the simplicity of its practice. AS parents, we may take the K-12 derivatives 58-161 and W-1177, which differ as shown in Table 1. The stocks are maintained on ordinary nutrient agar slants, transferred about once every three months. To initiate a cross, each parent is in- oculated into a ten ml broth tube (Difco Penassay or other buffered medium) which is incubated overnight without aeration or agitation. The tur- bid growth is spun down and the supematant de- canted and replaced by sterile saline. It may be advisable to repeat this process, and to rinse the compacted pellet before redispersing it. The [4131 414 BACTERIAL RECOMBINATION TABLE 1. SEGREGATING CHARACTERS IN THE CROSS 58-161 x W-1177 Character Symbol 58-161 w-1177 Biotin' B - + !i 2 Methionine' M + 2 Threonine' T + a 2 Leucine' L + ii Thiamin' Bl + `i Lactose' Lact + Maltose' Mall + -z Mannitol' Mtl m + - tlE d-Xylcse' Li!r XYl + a2 Virus T1; T53 Vl SE S r 7 Streptomycin' s S r `Auxotrophy (+ independent; - dependent) `Fermentation `Sensitivity or resistance The derivation of these stocks is cited in Tatum, 1945 and Lederberg, 1947, 1949. Each character was altered in a single mutational step. final suspension of washed cells should comprise a reduced volume of two to three ml. The two suspensions are combined. Samples of 0.05 to 0.1 ml are then spread over minimal agar plates, or poured with melted agar. The plates are then incubated (at 30o or 37OC). On each plate a few hundred prototroph colonies appear in 24 to 72 hours, while control plates (58-161 or W-1177 suspensions alone) have invariably remained entirely barren. Th e prototrophs are then picked for further tests. It is important to keep in mind the usually invisible background of auxotroph parent cells which may interfere with tests for the unselected markers unless the prototrophs are purified by conventional methods. The media used are not at all critical; the formulas in use at Wisconsin are given elsewhere (Lederberg, 1947, I950b). Fermentation markers are scored by inoculating EMB ( eosin methylene blue peptone) agar containing one per cent sugar. Resistance is scored by cross-streaking a suspension of the bacteria against a loopful streak of the virus or chemical. This cross has been studied extensively in several laboratories besides our own with con- cordant results. Among the prototrophs, each of the possible assortments of the unselected markers may be found if a reasonably large sample is studied. The reassortment of unselected markers is the most cogent evidence that the prototrophs arise by recombination, rather than by any arti- fact. The different combinations do not occur with equal frequencies by any means, pointing to a linkage system which is discussed further below. This simple experiment demonstrates genetic interaction between the parental bacteria, but leaves open several questions as to its mech- anism. If we confine our discussion to effects based upon material exchanges (excluding a priori mitogenetic rays and the like!) two contrasting hypotheses exhaust the possibilities: (A) one or both the parents release chemically definable substances which transform the other into proto- trophs; (B) a f usion of organized elements of cellular origin is involved. Hypothesis (A) pap allels the interpretation given to type transfor- mation experiments with preparations containing polymerized DNA from pneumococci or influenza bacilli (McCarty, Taylor and Avery, 1946; Alex- ander and Leidy, 1951). Hypothesis (B) entails a sexual process with attendant implications of processes analogous to fertilization-reduction cycles. At the outset, it should be made clear that a final decision between these hypotheses must wait upon a positive chemical character- ization or morphological demonstration of the agents of recombination. This is not yet at hand, but there are many genetic and negative physical observations each of which, in our opinion, weighs compellingly in favor of a sexual process in this bacterium. The main difficulty in the way of more direct physical evidence on this question is the relatively low rate of recombination, for the yield of prototrophs is only about lo-* of the auxotroph cells inoculated. Experiments designed to find support for "transformation'? have failed to implicate any unit other than the intact cells of the two parents. The prototroph-forming ("transforming") activity of cultures in broth is quantitatively sedimented with the cells in the centrifuge, and no such activity can be found in the supernatants passed through bacterial filters, nor in preparations prepared from autolysates or other cell-free prep arations. Davis (1950b) has reported that re- peated flushing across a filter of the medium shared by two parents resulted in no transfer of activity. The stratification of the two parents in contiguous layers of minimal agar nearly elim- inates their interaction. Finally, the addition of desoxyribonuclease to interacting cells has no effect, in contrast to its destruction of the trans- forming factors of the pneumococcus and of Hemo- philus. We are led by these experiments to con- I. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 415 elude that the agent of recombination is either cellular, or very labile except in intimate asso- ciation with cells. The genetic properties of the recombination system agree with this conclusion. There is no obvious limit to the number of factors which can be exchanged at one time; in 58-161 x W-1177, there are no less than six unselected markers (7, including B,-on thiamin agar), all 64 com- binations of which occur among the prototrophs. Each of a considerable number of other markers tested in other crosses has segregated in similar fashion. By contrast, the transformation experi- ments so far published have involved one char- acter at a time. Further experiments involving the mixture of three well-marked stocks have shown that recom- bination is limited to factors differentiating single pairs of cells, and there is no pooling of genes from more than two cells at a time (no m&age `a trois). Finally, non-disjunctional exceptions have been found (Lederberg, 1949) in which the parental genomes are associated in heterozygous cells, which segregate at occasional fissions to unmask the recessive components. This inter- pretation is strengthened by Zelle's single cell pedigrees which independently rule out the pos- sibility that the "heterozygotes" are associa- tions of intact cells (Zelle and Lederberg, 1951). The simplest picture which encompasses pres- ent information supposes a life cycle similar to that of Aspergillus or Zygosaccharomyces: a haploid vegetative phase (inferred from segre- gation in the f-l), an infreqnent and usually transient diploid phase following upon fertiliza- tion, and no evidence of heterothallic restrictions on compatibility. For the moment, I am inclined to the view that fertilization is isogamous, and follows the fusion of undifferentiated vegetative cells, but the possibility of gametic specializa- tion is not entirely out of the question. But as stated before, a morphological concordance is needed and E. co& is perhaps not the preferred organism for this purpose. The efficiency of recombination between cells mixed for the first time on minimal agar plates may call for some comment. There is evidently a limited amount of residual growth of the auxo- troph inocula, undoubtedly assisted by syntrophic stimulation. The prototrophs probably result then from the interpenetration of microcolonies rather than individual cell-to-cell contacts. Any analysis of the kinetics or physiology of bacterial recombination should define the proxi- mate conditions of cell contacts more closely than has so far been achieved (e.g., Clark, et al., 1950). For example, the non-specific formatiou of small mixed clumps of cells would be expected to encourage their later sexual association in the face of their immobilization by the physical re- straints imposed by solid medium. This is a reasonable interpretation of Nelson's (1951) findings that the yield of prototrophs increases with the time during which the parents are shaken together in saline suspensions before plating. FORMALGENETICSOFHAPLOIDS Some hint of the genetic structure of E. coli is given by the segregation of markers among pro- totrophs, but the necessity for employing a selec- tive method introduces some difficulties. The exclusion of non-prototrophs precludes the iso- lation of complementary segregants as a class or as multiple products of single meioses. This disallows some necessary checks on the mathe- matical regularity of segregation, and vitiates the random recovery of sexual progeny necessary for statistically mendelian behavior. Furthermore, we have no direct way of knowing what proportion of the zygotes succeed in producing a detectable prototroph recombinant. This difficulty is very well illustrated in attempts to unravel the linkage system that generates the empirical rules for the segregation of the markers in such crosses, as 58-161 x W-1177. When crosses are carried out according to the protocol described in the first section of this paper, the potential zygotes are immobi- lized on the agar, so that their genetic prod- ucts are also localized. With very few excep- tions, the individual prototroph colonies are internally homogeneous with respect to segre- gating markers, and show no signs of subsequent segregation either under ordinary culture or single cell isolation (ZeJle, unpublished; see Lederberg, 1947). However, the different colonies vary from one to the other: segregation occurs when the prototroph colony is initiated. If a reasonable proportion of zygotes produce prototrophs, we could also conclude that the intermediate diploid phase is transient, does not proliferate as such, and resembles the many other haplobiontic thal- lophytes. We might also conclude, but less surely, that diploid fusion nuclei occur individually, despite the cytological evidence of several nuclei per cell. Whether this means anisogamy, or that cell fusion occurs rather more frequently than karyogamy, or something else, we cannot say. The recombinant prototrophs are culturally and genetically indistinguishable (except for their 416 BACTERIAL RECOMBINATION inherited markers) from the typical parental cul- With the data reproduced in Tables 2, 3, and 4, tures. At least three other groups of workers an attempt has been made to map certain factors, working with the cross 58-161 x W-1177 have ob- and especially to determine whether the concept tained detailed results quite consistent with ours of linearity can be unambiguously tested. We (Table 5) on the segregation of the markers among proceed from the assumption that segregation is prototrophs (Cavalli, 1950; Newcombe and Nyholm, regular, but that our observations are confined 1950; Gordon Allen, unpub.). Clearly, the fre- to the prototroph set of recombinants. The dis- quencies with which the different types appear tribution of unselected markers will be then con- are sufficiently stable to justify a search for trolled exclusively by their linkage relationships definite rules of segregation. Further support for to the selected, nutritional markers. Thus we regarding these mutations as indifferent markers infer that the scarcity of one allele of a marker of the genetic mechanism comes from "reverse gene signifies a linkage to the auxotrophic factors crosses." The same differential markers are originally coupled with it in one of the parents. TABLE 2. REVERSE CROSSES INVOLVING PERMUTATIONS OF Lac AND Vt . . Parents* Prototrophs: Percentage Distribution1 B-M-T+L+ B+M+T-L- B+M+T+L+ No. tests Lac vi Lac VI Lac- Vlr Lac- Vls Lac + VI' Lac+VIS + r - s E 42.7 2 23.2 A 32.5 ABC 1.6 2013 + s - r 34.6 42.5 ABC 2.6 A 20.3 696 - r + s A 25.1 ABC 2.5 f: 47.7 C 24.7 518 + r - rt B or C 79.5 0 20.5 0 161 - r - st -134 Lacy Vt not scored,- *RI +/- was also segregatin but has no interaction with other factors. pooled. Adapted from Table 5, tederberg, 1947). The data given here were therefore tAllelism tests. IThe jetters A, B, C, and ABC refer to single and triple crossovers in the regions [BM] -Loo; La-VI',; and VI- 1 TLJ , respectively. introduced, but with the alternate parent: for ex- ample, B-M-Lac+Vlr x T-L-B,-Lac-V,' is comparedwith B-M-Lac-Vlsx T-L-BrLac+ V,`. The segregation frequencies of Lac and VI are inverted by this reversal; in other words, the occurrence of a given allele among the prototrophs is regulated not by the nature of the allele, but by its parental coupling. To apply this test, recurrent mutations in the two parental (BM and TLB,) stocks must be ob- tained and this is not always easy. However, it has been applied successfully to markers at the following loci: Lacl, Malt VI, V6, S. Table 2 illustrates the experimental concordance of bac- terial segregations to a generalized definition of mendelism: "The gametic frequencies are invar- iant in respect of any gene substitution applied systematically to the genie content of an organ- ism and of the gametes it produces" (Fisher, 1947). Here, of course, we must read f-l haploid progeny for gametes. (The following notation is used to avoid confusion between haploid and diploid generations: p-l x p-l (n) -F-l (2n)-" f-l (n). f-l x f-l --+ F-2 (2n) -f-2, etc. These are designated as p-l, f-l crosses, etc.) For example, the relative paucity of VI' among prototrophs (B + M + T + L +) from the cross B- M- Vls x T-L-V2* suggests that VI is linked to T or L. Since these segregate dependently, V, is linked to both, but the order is indeterminate. When the appropriate growth factor is added to the plating medium, individual nutritional loci may be treated as unselected markers in the same way. Thus we may inte ret Table 3 to specify the arrange- ments: B,- B-MI and [T-L]. Insofar as a re- 9 TABLE 3. RELATD'E FREQUENCY OF MONG AUXOTROPH RECOMBINANTS B-M-T+L+Bi+x B+M+T-L-Bl- Monoauxotroph Class B- T- L- B,- Ratio to Prototrophs 0.17 0.24 0.10 9.88 Number of Tests 70 46 56 87 The cross was conducted on minimal agar with single growth factor supplements. The recombinants were then classified as prototrophs or monoauxotrophs, and their relative frequency recorded. M- is not re- corded, as this test cannot be conducted on methionine medium with B- as the only selective marker for the B-M- parent. Adapted from Table 4. Lederberg, 1947. J. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 417 combination between [B-P] and [T-L] is in- dispensable to the occurrence of a detectable prototroph, we can say nothing of their connection so far. Upon these coordinates the data of Table 2 enable us to map the Lac and V, factors. The individual segregation frequencies show: B,- [R-J!] - Lac and Y,- [ T-L] respectively. interaction between Lac and V, (i.e., the marked deficiency in Lac+VIR from RW L + `t"" x TL Lac- Vlr) gives us: B,- fB-& Lm:: VI- CT-L], where the order of factors in [ ] is still indeterminate. So far, we have had no TABLE 4. LINKAGE OF Lac, V1 AND Vs B-hJ--T+L+B1+Lac+VlrV6s x B+M+T-L-Bt-Lot-Vls Vb Prototrophs Lac: - - - - + + + + B+M+T+L+ VI : r s r s r s r s V6:rrs s r r ss % : 43 33 1.7 0 4.6 1.1 15 1.7 Postulated crossover c d abc abd b bed a acd BM V. LlZC V1 TL S + r + + R - s - a b c d As in Table 1, Bfi and Brdata are pooled. Total tests numbered 176. Adapted from Table 6. Lederberg, 1947. test of linearity, for each experimental fact was accommodated by adding an independent element to the scheme. The first opportunity for such a test arose with the placement of V, (resistance to phage %6), as shown in Table 4. The segre- gation ratio of V, places it near [ BM] , and the test of linearity now requires that it be linked either to R, on the left, or Lac to the right. The latter appears to be justified, for there is only about six per cent recombination between Lac and V,. There are, however, an unusual number of complex recombinants that would have to be interpreted as multiple crossovers. Further linkage studies have been somewhat hampered by the paucity of suitable markers, especially in the vicinity of V,. The most concrete difficulty with the scheme developed with attempts to map the MuZ factor, and subsequently, S. In 58-161 x a-1177 the frequency of Alal+ is about I5 per cent, placing it near [M!] (Table 5a). It is not closely linked either to R, or B, for the addition of biotin and thialnin to the medium does not appreciably alter the 1.5 per cent figure. Therefore, if the data v6 Lac VI L T Y . . ..*...... 2!f!mLg FIG. 1. Schematic representation of linkage data. a 58-161 x W-1177. b. 58-161 x kl, W-1177 type. (Compare with Table 5.) This diagram is purely formal and does not imply a true branched chromosome. are to be fitted to a linear scheme we would ex- pect MaZ to appear between [SM] and Luc. However, the segregations of Mu.! and Lac are uncorrelated except that the frequency of MaZ+ is somewhat higher among the Luc+ than the Luc-. Consideration of the other markers of K-II77 multiplies the difficulties. S shows what appears to be a straightforward linkage to Mul, most of the prototrophs being either Ma.!+ 9 or Ma1 - SS (these data were collected in part by M. Dou- doroff). Similarly, Xyl and MtZ appear to show simple linkage to each other. liowever, although the segregation ratio for Mtl+ is also of the order of ten or 15 per cent, it is not clearly linked either to B,, or to Lac, or to Mul, but there are statistically significant interactions, especially with Mul. In a purely formalistic way, these data could be represented in terms of a &armed linkage group, Figure Ia, without supposing for a moment that this must represent the physical situation. `Ihis recalls the branched chromosome represen- tation (fiamlett, 1926) of translocation hetero- zygotes in Drosophila before the cytogenetics of this situation was well understood. Kewcomhe and rJyholm (I950a) have, however, interpreted or described the deviations in linkage behavior as due to "negative interference." These authors place MaZ, S, XyZ, etc., to the left of [ BM] but have not studied the anomalous behavior of 3,. Since we are armed exclusively with genetic data in E. coli, and since the cornerstone of genetic linearity is the lower frequency of mul- tiple compared to single crossovers in small regions, this discussion may be somewhat pre- mature. For this reason, our accumulated data on these segregations have been held in abeyance 418 BACTERIAL RECOMBINATION until they could be correlated with other infor- ratios when crossed either with p-l or f-I B-V- mation to be discussed presently. Re feel that stocks. Table S presents a comparison of crosses the premises on which the mapping from proto- of p-l B-RI- with r-1177 and a phenotypically troph data is founded must be reexamined and similar R-1635, extracted from the heterozygote verified before the problem can be settled. H-267 (R-67 x R-1177), mentioned later. All of Our suspicions concerning the placement of the markers except B, behave in the latter cross Mu1 were deepened by the behavior of nondis- as if they were linked to [ TL], but with the junctional exceptions, the first of which was en- same lack of interaction as before, so that we countered about three years ago. The principal have (again formalistically) a multiarmed con- deviation was the finding that each of the several figuration as in Figure Ib. hundred diploid heterozygotes obtained from The main point is that we have a derived stock crosses of one nondisjunctional line, and almost whose apparent linkage relationships are entirely all others, proved to be haplogenic for MuZ (and S), unique. In any plant or animal this would be TABLE 5. SEGREGATION DATA IN KULTIPOINT TESTS Par Par IA B LOC ---_------ - ----- ++++++++++++++++.% 85 Ma1 I ++++++I-+-- - -------A+ ++++ +++--- -----21 84 S rrrrssssrr r rssasrrrras aarrr r sass'15 72 MtL ++ --++--++ - -++--+ f--+ +--++- -++--j17 73 VI rsrsrsrsrs r s r sr sr srsr s r sr sr srsrsi67 88 A. p-l x p-l 1 3 4 3 22 36 12 1 11 91 111 11 li B. p-l x f-l 1 136 2 11 2 13 4 45 715 3 2 3; The figures represent the numbers of different recombinant types found in 100 tests of randomly isolated proto- trophs from crosses of the form of 58-161 x W-1177. Series A describes just this cross; B shows the deviations found when a phenotypically similar f-l segregant is substituted for r-1177. These data are presented to illus- trate typical patterns. For more detailed linkage studies, rarer subtypes such as Mal+ or !Y are isolated selec- tively for large scale testing (compare Cavalli, 1950; Newcombe and Nyholm, 1950a and bl. Columns A and B summarize the per cent + or s for each marker in the two crosses. although digenic for most other markers. In crosses of the type form B-M-f/et x W-1177, or 58-161 x ff-1177Het (Het is the factor that pro- motes nondisjunction), about four-fifths of the heterozygous prototrophs are MuZ-(i.e., the same as the haploid prototrophs in this or the type cross). This suggests that the segregation fre- quency of Mu1 is determined not by auxotrophic linkage, but by another process which eliminates a segmental homologue encompassing the Ma! and S loci. In this way also, the inconsistencies in the apparent linkage of MuZ- [RVI , and of [BY] to B, and to Luc, may be obviated. The details of this obviously highly aberrant behavior are far from clear. A further indication of struc- tural heterozygosity may be seen in f-I and f-2 crosses, in which haploid segregants from dip- loid nondisjunctions are used as parents. FChen R-M-f-l's were crossed with T-L-B,-p-I stocks carrying the appropriate complementary markers, no alteration of segregation ratios was observed. However, each of four or five T-L-B,- segregants studied in the same way gave highly perturbed prima facie evidence of a chromosome aberration, and this is one working hypothesis. Eowever, we cannot say whether B-1635 or R-1177 (if either) is structurally homologous with 58-161 or the original K-12 Rith some effort, di-auxotrophic recombinants can be recovered (e.g., Table 2 of Tatum and Lederberg, 1947) by conducting the crosses on appropriate media. Crosses of the form M-Z'- x T-L-Luc-MaZ- etc., have been made on minimal medium supplemented either with methionine + leucine, or proline + threonine. A small pre portion of the nutritional selections are M-L- or P-T- recombinations, respectively, most of the colonies being prototroph recombinants, or mon- auxotroph recombinants or reversions. Since the M-L- and P-T- are complements, regular segre- gation would require that the segregation ratios for unselected markers be inverted from one to the other set. In preliminary experiments, bliss Phyllis Fried h as found Luc+/- to conform to this expectation, but both sets were almost uni- formly MaZ- like the prototrophs. If these findings I. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 419 are correct, we can be certain that the segrega- tion of Ma1 is determined otherwise than by linkage to [ BM] . The N--P- stock used in these experi- ments was prepared by Gordon Allen for a some- what similar purpose: to look for the complemen- . tary auxotrophs as segregants from single zygotes. He was unable to find conclusive evidence on the question owing to technical difficulties, but incidentally accumulated segregation data in agreement with those just mentioned. The rule was stated earlier that individual pro- totrophs were generally pure for segregating markers. In addition to nondisjunctional excep tions a second type of exception has been found in what I shall call "twin prototrophs." These are detected most readily as bisectored proto- troph colonies- from crosses on a synthetic in- dicator medium (`EMS") upon which one fermen- tative character of a colony can be read by in- spection. One or two per cent of the prototrophs on EMS-maltose medium are duplex. At first thought, these might be either coincidences of zygotes, or sister segregants from a single zygote following four-strand multiple crossing-over. Both these notions are ruled out by the distribu- r tion of other markers to the twins. In particular, the alleles for Lac and V, are almost always the same in the two gemini, even when the combina- tion is the rare Lac+V,`. V, also folIows Lac; S follows Wal, and Xyl and Mtl show no clear preference. `Ih is result could be interpreted in terms of three chiasmata, two of them (involving f,ac-I',) showing strong negative chromosome and positive chromatid-interference, but this seems rather farfetched. It may be more reasonable that the Lac-V, character of the twins is determined by a single crossover event, but that reduction for MaZ and S was delayed for one fission. This would be the converse of the aberrant nondis- junctions which are reduced for Ma1 and S, but in which reduction for Lac, etc., may be delayed indefinitely. FORMAL GENETICS:~ONDISJUNCTIONAL EXCEPTIONS Throughout the preceding discussion the dip- loid zygote has b een an unseen logical inference deduced from the occurrence and patterns of re- combination, rather than a tangible reality. Many of the assumptions made to provide a framework for further experimentation could be tested much more directly if the zygotes could be collected and characterized as such, and then allowed to segregate. Many early experiments (1947) were carried out to try to induce nondisjunction or persistence of the diploid phase by heat shocks, variations in medium, irradiation, and c-mitotic chemicals, but none were successful or even en- couraging. On one occasion, in 1948, however, one of our strains itself set out on this experi- ment, and the larger part of our attention since then has been directed to the analysis of diploid behavior. The exception was detected in a curious man- ner worth recounting. The genetics of phage resistance was being studied, and a number of mutants resistant to Tl were collected. One of the mutations, designated VIP, was partially re- sistant to TI and T5, and was tested to see if it carried an intermediate allele of V,`, which is immune to both these phages. The V$ mutant occurring in 58-161 (R-J!-) was therefore crossed with a T-L-B,-V,`Lac-stock (Y-64), and 200 pro- totrophs were tested on EMS medium for their reaction to Tl. Of these, I99 segregated VIP and VI'; a single prototroph scored VI: Insofar as the decision between a series of multiple alleles at V, as against two linked loci rested upon this single apparent recombinant, it was retested against Tl on EMB (complete) medium. This time it showed a peculiar reaction, as if it were a mixture of sensitive, partially resistant and resistant. When the culture was streaked out on EMB lactose medium, it became apparent that several types were splitting off, the most obvious being Lac- V, and Lac+ I/# like the parents. Ahen these were purified and tested further, it was found that many of them were auxotrophic for various combinations of B, d!, T, L and R,, although they had been derived directly from colonies growing well on synthetic (EMS) medium. On this medium, there is little overt evidence of segregation, for auxotrophic segregants are sup- pressed. After I5 single colony transfers on EMS lactose, the culture still segregated on EMB medium. From this, and the interaction of Vi' with V,p to give the ViS reaction, the pos- sibility of an extra-cellular association of the two parents was rejected, and either a hetero- karyon or heterozygote was postulated. The occurrence of new combinations among segegated colonies supports the latter, although it is quite likely that a heterokaryon intervenes temporarily between nuclear and cell reduction. The appear- ance of segregating variegated (designated Lam, Mtlu, etc.) colonies on EMB medium is quite characteristic, as shown in Figure 2. Any linger- ing doubt that the variegation might result from the sticking together of cells of the two parents should be dispelled completely by Zelle's single 420 R ACTERIAL RECOMBlNATlON FIG. 2. Segregation of heterozygous diploid cells to form variegated colonies: MaZ+/- on EMB maltose agar. cell pedigrees, in which heterozygous cells were permitted to divide repeatedly under close micro- scopic observation, while the entire pedigree was isolated with the micromanipulator (Zelle and Lederberg, 1951). Two such pedigrees (typical of a score of them) are shown in Figure 3. In 3a, every viable cell produced a variegated colony; ? o ? ? ??? B ,:.:*- /" / \0'" I \ \(" . . /- \<" . . \ . . / <(. \ ,:i I \.. \ <-;.. `.. / \ `X I\ . . /:: \ 0 \0" /\.. `** / 1.. /" \ ????? o ??? ? ? ????? ? ???? ax inviable .- segregant o + Mal- or+ FIG. 3. Cell pedigrees of segregating diploid (H-226). The dichotomy from left to right represents the fis- sions of a clone under microscopic observation. The fission products were separated and the terminal cells isolated with the aid of a micromanipulator (adapted from Zelle and Lederberg, 1951). in 3b, the other pedigree, a segregation occurred during an early fission, giving a clone of eight haploid cells. As a rule, the segregants occurred individually, not as complementary pairs (Zelle and Lederberg, unpublished). This may be at- tributed to a dissynchrony between nuclear and cell-division, and to the postulated haplolethality of the segmental deficiency discussed below. A rather high proportion of inviable cells was ol+ served, but this genetic basis remains to be proven. The first nondisjunction found (H-1) was not very well suited for segregation studies owing to the paucity of easily scored segregating markers, but it quickly became apparent that alternative factors did not occur with equal frequency. Among segregants isolated at random from complete medium, the majority were M+ Lac- V,`T- L - B,-. T!le other parental type, M-Lac+V,p occurred less frequently, but prototrophs and the previously elusive polyauxotroph recombination W- T-L-B,- were also found. Once established, the segre- gants remained perfectly stable, and segregation for different factors occurred concurrently. This, together with the fact that instability was con- fined to the factors which differentiated the parents, leaves little doubt that the genetic in- stability of these cultures results from occasional segregations of hybrid cells. Some of the B-M- and T-L-BI- segregants of 11-I were preserved for use in f-l crosses, par- ticularly A-478 (R -M - Lac +&et) and n-477 . . diploid I. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 421 TABLE 6. NONDISJUNCTION PROTOTROPHS ISOLATED FROM B-M-V6'S'Hetx T-L-B,-L,acI-iMaE,-Xyl-Mtl- A: Mtl u Isolations B: Lac u Isolations (39) (28) No. Lac Vb S Ma1 Xyl No. Mtl Xyl Ma1 19 v* v* s - v 12 v - 6-ss-v 4 r 3 Y vr + + 3-z rt 2 vvs- - 3 v - - 2vvs++ 3++ - 2-vs-v l-v - lvvr+v 1 + v + l+rr+ + 1 - + + 1 s - 1 `ir V + v 1 mucoid: unclassifiable *v refers to beterozygous condition, whether +/-, or r/s. No Malv were recovered in several hundred tests of Mal+ in addition to the prototrophs listed in this table. (T-L - B,- Lac-Het). The possibility of further analysis was widened by the reappearance of un- reduced prototrophs in the F-2 and in f-l x p-l. The initial occurrence in the V,p stock of the nondisjunction-promoting factor was evidently a mutation-like accident, for retests on that stock have given no further sign of Ret. Considerable attention has been given to crosses of the form W-478 x W-1177 because of the wealth of segregating rrarkers. W-1177 type stocks carrying Lac- and Mal- from which spon- taneous reverse mutations (occurring at rates of the order lo'* per division) can be selected by plating on medium containing lactose or maltose as the preferred carbon source have been es- pecially useful. Dispensing with an extensive but concordant sequence of crosses carried out with less satisfactory markers, the data of Table 6 will be used to illustrate the behavior of several hundred similar heterozygotes. To isolate Lac+/- heterozygotes, a cross is conducted on EMS lactose agar. The 70 per cent Lac- prototrophs are disregarded; about ten per cent of the Lac+ prototrophs are unreduced for Lac and give Lacv colonies on EMB lactose. The proposed diploids are purified and maintained in lactose-synthetic medium. In some crosses, but not all, the diploid colonies can be distinguished by inspection, for they tend to grow more slowly and to have a slightly delayed fermentation reaction. A parallel procedure is used to isolate Mtlv, Xylv, etc. The most striking feature of Table 6 is the number of entries needed. Contrary to first naive expectations, the apparent F-l is far from homo- geneous and never uniformly heterozygous. None of the cultures were Malv whether isolated as Lacv or Mtlv, nor has Malv ever been isolated from any Het crosses despite the most extensive trials. bkst of the F-l of W478 x W1177 were Mal-. But this was not the only peculiarity; of 39 Mtlv, 28 were Lacv, but 10 were Lac- and 1 Lac+. Conversely, of 28 Lace, 16 were Mtlv, 7+, 5-. Two interpretations for Lac- and Ma1 -cultures were considered. They might represent a haplo- genie or hemizygous condition of the locus, or they might be homozygous "-." Reverse muta- tion was used as a distinguishing criterion. Lac+ reversions were selected on EMS lactose agar, and purified on the same medium. The re- verted cultures were then streaked out on the various EMB media. Some of the reversions oc- curred in prototroph f-2 giving Lac+ no longer segregating for any marker. A total of 66 inde- pendently occurring reversions were still diploid (i.e., Mtlv, etc.) including selections from eight of the nine Lac- Mtlv cultures of Table 6. With- out exception, each of these cultures had reverted from Lac- to Lacv. From this, it is concluded that the Lac- cultures are homozygous and not haplogenic. Except in consistency the behavior of A!aZ- contrasted with this. illal- is much more stable than Lac-, limiting the number of rever- sions that could be selected in a moderate num- ber of trials. In this series, seven diploid re- versions from four Mal- Mtlv cultures (Table 6) were each pure MaZ+. Similar tests on many Ma1 - diploids from a number of other Het crosses gave the same result. Mal- evidently reflects a haplogenic condition. By inference, the Lac+ and MaZ+ diploids in this series are also assumed to be homo- and hemi-zygous respectively. The only relevant evidence comes from "reverse crosses." As shown inTable 8, B-M- Het Mal-and B-M- Het Lac- were crossed with T-L-B,- respectively, and the issuing heterozygous diploids collected. The transposed prevalence of Lac+ (together with Lacv) and of MaZ+ diploids will be noted, once again supporting the concept of these mutations as inert markers. 16 Lac-reverted diploids from the single Laocultures of the reverse series were each Lacv; conversely, 11 Ma&reverted dip- loids from 7 Ma&cultures were all pure Mal+. If, as supported by the statistical results of the reversed cross, the + products of the original are equivalent to the - of the reversed, we can con- clude definitely that the Lac locus has always appeared in diplogenic condition, whether -/-, 422 BACTERIAL RECOMBINATION +/+ or +/- (~1, whereas the Mal locus has always been haplogenic, whether - or +. The occurrence of homogenic factors is too consistent to allow them to be interpreted as point mutations. Further- more, some linkage relationships, Mal-S, Lac-VI, and Mtl-Xyl persist in the formation of the hetero- zygotes. The explanation we have adopted for homozygosity of Lac is that meiosis intervenes between the time of initial karyogamy and the establishment of the non-reduced prototroph. Such a meiosis must involve a four-strand stage to account for homozygous loci. It is thus pos- sible that the action of Het is not to delay dis- junction of the primary zygote but to stimulate the reunion of a pair of segregating strands. The overall process is reminiscent of somatic segregation in Drosophila (Stern, 1936). The possibility of concurrence of two pairs of such strands to give twin, complementary diploid pro- totrophs has not been realized in our experi- ments, except for two or three instances of twin diploids, Mal+ and Mal-, comparable to the twin haploid prototrophs discussed earlier. The V6-Lac linkage can be applied to a fup ther corroboration of the context of this discus- sion. Many of the Lac -/- of Table 6 are Ve5 (presumably s/s) but in some the postulated crossover has separated Lac and V, giving a LUC -/- V6 S/r. LUC-reVerSiOnS from such a Cd- ture should fall into two classes: Lac + Vss/--r, and + r/-s. A series of 35 independent revep sions was secured. After purification on EMS lactose each reversion was allowed to segregate on EMB lactose. No more than one Lac- and Lac+ was chosen from the progeny of a single Lacv colony, and these were then scored for VS. Table 7 shows how each of the 35 cultures was unmistakably either +r/-s or, +s/-r, numbering 20 and 15 respectively. (X1 for de- viation from 1: 1 is 0.7; p = .4). The result is TABLE 8. NONDISJUNCTION PROTOTROPHS FROM CROSSES REVERSED WITH RESPECT TO TABLE 6. A. B-M-MalrHet x T-L-B,-Lacl-, Lacv selections 11 Mal- 41 h4al+ 1 Mnl+ and Mal- "twin." B. B-M-L acl-Het x T-LyB,-Mtl-Xyl-Mal-, Mtlv selections (26 total) No. Lac XYl Ma1 No. Lac Xyl Ma1 12 v V 1 v + - 6 + V 1 + + + 2 1 1 V + V V + T : v + + 1 + + -, + twin These tables should be compared with Table 6, in particular with respect to the reversal in proportion of MaZ+:- in A., and Lac +:- in B., respectively. an almost trivial consequence of homozygosity, but there are too many unexplained aspects to let any opportunity go by to test the homology of the genetic system of E. coli to that of other forms. Incidentally, this experiment provides another concrete instance of the familiar notion, not often directly tested, that spontaneous mu- tation in a homozygous cell affects the two alleles independently. The numerical bias in segregation from these diploids was mentioned in connection with H-l. In these experiments, it was noted that the -r/+s were readily distinguishable from the +r/-s, for Lac- predominated among the segregants of the latter; Lac+ of the former. (In other words, V6' appeared more frequently than V,`). This type of aberration, which has been noticed repeatedly (in every diploid) may be correlated with the be- havior of Mal. The haplogenic fate of MnZ and S means simply that these "diploids" are in fact aneuploid and it is difficult to escape the conclusion that a TABLE 7. DEMONSTRATION OF THE TWO CLASSES OF Lac REVERSIONS IN LacVs'/Lac-F6S Three to four Lac- and Lac+ segregants from each reverted diploid were tested for V,, and scored as follows: No. of diploids Lac + V6' Segregant Types Lac - I/es Lac + v6 .s Lac-V6' class* 14 3 to 4 3 to 4 0 0 13 0 0 3 to 4 3 to 4 F 4 3 4 1 0 : 2 1 0 4 3 1 0 3 : R 1 2 4 2 0 R? 1 0 1 4 1 S? *Reversion on V6' or V6' chromosome, R or S. Total: ZOR: 15s. J. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, segment (6`S-segment") of the genome, marked by Mal, S and a few other factors, is eliminated during some part of the sexual cycle. As a work- ing hypothesis, we are inclined to retain a chro- mosomal basis for the genome, and to regard the elimination as a superimposition upon its %or- mal" workings. Th e regularity of segmental eli- mination and its extent makes it appear likely that at least a chromosome arm is involved, possibly an entire chromosome. It will also be noticed that, in Table 6, there is a correlation between .Mal+ and Xyl+ or AM+, whereas the Ma1 - are usually either Xyl v or Xyl-. This is most readily interpreted on a crossover basis as illustrated in Figure 4. One especially simple hypothesis for the pseudo-linkages of 58-161 x W-1177 postulated structural heterozygosity of these parents, so that, for example, a quadrivalent translocation ring developed at meiosis. This would account directly for the &armed "linkage map" con- structed earlier, and the "diploids" (including homozygous loci) would be a result of unequal or 3: 1 segregation of the ring such as Burnham (1948) and others have described in maize. But this hypothesis is contradicted by the results of an f-l cross. Viable segregants from the postulated multi-armed configuration would nec- essarily be structurally uniform, although they might carry different genes by crossing-over. When crossed with each other, however, the segregants produced heterozygotes in which the same elimination process was repeated. It seems likely that we have reached a point in this analysis where the genetics has over- reached the essential cytogenetic framework, and possibly an ad hoc statement that elimination occurs is no more objectionable than speculation on specific mechanisms which may be difficult to test for the present. It may be desirable how- ever to conclude this section with a brief general- ization of the possibilities in cytogenetic terms. The occurrence of "twin" haploid or diploid pro- totrophs, the multi-armed linkage pattern and the regularity of S-elimination point to the presence of two, rather than one, chromosomes. On the other hand, some explanation is required for the aberrant segregation ratios of factors which re- main heterozygous, and the S-elimination can be related to this most easily on a single cbro- mosome basis. Perhaps we are relying too strongly on the perfection of specific pairing and synaptic mechanisms, for some of these peculiarities might be interpreted in terms of a high frequency of non-homologous association of PARENTS * Mal-DEFICIENT DIPLOIDS B PRIMARY NON- DISJUNCTIONS c D DERIVED DIPLOlDS E AND E. R. LlVELY 423 S Mal Xyl M Lac LOT s + + +[-I + I 58-161 - - + - L+l W-l,?? . . + + + t - - + . . . + + + J ++ + 0 ++ +- + `I -- + -+ s + + +- + s + + + -+ $ ++ -4. - - + + FIG. 4. Schematic constitution of diploid hetere zygotes. A., B.: Maldeficient diploids, as cited in Tables 6 and 7, and recovered from "Het" crosses. C.: diploids obtained as Lac+ non-disjunctions from crosses of Lact-x L.ac,-. changes from C. E. D., E.: spontaneous is awotrophic and suitable for use in diploid x haploid crosses. the sort invoked by Longley (1945) and Rhoades (1942) to explain false linkage and non-random segregation in maize. As far as we know, how- ever, there is no precedent for dissynchronous reduction of the kind which might be postulated to explain twin prototrophs. The discussion to this point has been confined to the unreduced progeny of crosses involving Het which came from the unique accident of H-l. Once such nondisjunctional types were realized, the possibility of their more general occurrence was reviewed. A few thousand prototropbs had previously been streaked .out on EMB lactose agar, and we were fairly confident that the Lacv character would not have been overlooked. 4 more effective screening technique was therefore devised to look for the occurrence of non-reduced prototrophs by the recurrence of Het, environ- mental effects, or other mechanism. Previously, a lactose-negative mutation at a new locus had been isolated (R-67, B-M-Lac-) and distin- guished from Lac,- (as carried by W-1177) by the symbol Lace. Crosses of Lac,- with Lac,- re- sulted in prototrophs, about 0.1 - .3 per cent of which were lactose-positive, so these factors are very closely linked, but separable. It was anticipated that Lac+ factors generally would be dominant. Consequently, recurrent heterozygotes should be detected more efficiently by testing only the lactose-positive prototrophs in a cross Lac,- X Lac,- (e.g., W-67 x W-1177). The rare 424 BACTERIAL RECOMBINATION lactose-positives can be detected by simple in- spection of prototrophs on the original crossing plates of EMS lactose agar. Unfortunately, R-67 is relatively infertile, but adequate yields can be obtained. These anticipations were soon re- warded, for about a third of the lactose-positive prototrophs from such crosses were Lacv. Thus, standard stocks produce nondisjunctional progeny but some hundred times less frequently than Het. These progeny probably are not a result of new Het mutations, judging from f-l tests, but Het it- and their existence itself shows that the non- occurrence of Malv in the Het progenies is no artefact. Table 9 shows segregation patterns from both types of diploid. Lac,- unfortunately is not easily distinguishable from Lace, and no attempt to map this with the other factors has so far been made. That segregation is always concordant among the Het diploids has already been mentioned. Extensive searches for "partial" segregants e.g., Lacv Xyl-, or auxotrophic Lacv from Lacv TABLE 9. SEGREGANTS FROM HETEROZ YGOUS DlPLOfDS Proportion calculated from combined data* H-212 ("Het" diploid: See Figure 4A) Lac +Xyl+ Mtl+ Lnc-Xyl-Mtl- Lac -Xyl +Mtz+ Lac-Xyl-Mtl+ .18 .61 .15 .05 H-226 t (Primary Non-Disjunction: See Figure 4C) Mal+Xyl+Mti+ Mal-Xyl-Mtl- Mal+Xyl+Mtl+ Mal+Xyl-Mtl- Numbers from: A. `%andom" Lac- , isolates B. Mal+ and -fixed C. Clones in cell pedigrees :z 547 : 0 3 7 10 *Several hundred variegated colonies on EMB lactose or xylose agar were streaked out, and a pure +and a pure - isolated from each one. To rectify the bias unposed by thus fixing the ratio of + to -, data from the two sugars are combined for the computation of the unbiassed frequencies. This does not compensate for perturba- tions due to selective differentials of other loci. +A11 segregants from this balanced heterozygote are Lac- (Lacr or Lac,-), so that may be isolated "ran- domly" as single Lac- from individual Law colonies. The unbalanced ratios under A. are probably a result of differential growth. More precise clonal data, as in C., (Zelle and Lederberg, 1951, and unpublished) cannot be collected on an adequate scale, but do suggest that the "random" statistics are biassed. The low frequency of crossing-over is characteristic of this group of diploid cultures. self has not always been transmitted to the f-2 in comparable crosses. About 90 per cent of these "primary" nondisjunctions have behaved like the Het F-l and F-2, particularly in exhibit- ing hemizygosity for Mal-. Of course, since their constitution is +-/-+ with respect to Lacl and La+ the segregants sre all lactose-negative, barring a negligible frequency of crossing-over. Thus the colonies on EMB lactose tend to a periclinal rather than a sectorial type of varie- gation. For certain studies, it is very useful to be able to use the Lac+ and Lac- appearance as criteria of the diploid or haploid condition of a given colony. In four or five cases the primary nondisjunctions have been Malv exceptions to the rule of Ma1 haplogenicity. These stocks have also shown biased segregations, but the possibility of aneu- ploidy in other segments has not been ruled out. In these diploids, Mal+ is dominant to Mal-, Xylv, etc., were unsuccessful, although one or two exceptions occurred in old cultures under cir- cumstances that did not rule out a repeated SeXU8l generation. The primary Malv nondisjunctions behave somewhat differently: one or two per cent of the colonies that had apparently segregated for Mal (pure maltose-negative) remain Lacv, and conversely. Thus secondary Lacv Mal-, and Lac- Malv cultures are produced. If Y/S" is segregating (H-267) the former are readily de- tected as Lac v colonies on streptomycin medium (which inhibits both the original s/r type, and Ss segregants). One might anticipate that these secondary types would resemble the previous Lacv Mal- cultures, and that the partial segre- gation resulted from segmental elimination such as occurs in the formation of Het diploids. How- ever, all of these secondary Mal- cultures tested by reversion for Ma1 have consistently yielded Malv, not Mal-t. The aspect of Het behavior dup I. LEDERBERG, E. M. LEDERRERG, N. D. ZINDER, AND E. R. LIVELY licated in the secondary types is thus the OCCUP rence of homozygous loci rather than elimination. The distinctive behavior of the secondary Mal- cultures reinforces the previous test for hemi- zygosity. The linkage between Mal and Xyl in the formation and segregation of primary nondis- junctions was noted earlier. However, one Ma1 -/- XyZ +/- allowed a test of the random occur rence of ,a2 reversions similar to that of Table 7 for Lac -/- V, r/s, and with the same result (11 Mal'Xyl +-/-+; 17 ++/--., p = 0.25). The derived Lac - Malv partial segregants are horns- zygous Lac- by a similar test. An especially interesting class of partial seg- regants is that in which auxotroph factors have become pure-(presumably homozygous) while Lac or Mal or both remain heterozygous. Such auxotrophic diploids can be crossed readily with complementary auxotrophs (haploid or diploid), to give large yields of F-2 heterozygotes. Studies in progress indicate that segmental elimination does not occur during these crosses. However, crosses of f-l from primary nondisjunctions show the same trend of S-elimination as the other f-l tested. These 2n x n and 2n x 2n crosses are expected to be especially useful in further work for three reasons: the non-occurrence of S- elimination, the high proportion of heterozygous progeny, and (in principle) that chromosome re- combination rather than specific crossovers is the necessary condition for the formation of de- tectable prototrophs. Since the yield of proto- trophs from such crosses is perhaps tenfold that of haploid crosses, we may reinforce our earlier contention that a fair proportion of zygotes pro- duce prototrophs. The S-elimination was espe- cially troublesome for the problems of construct- ing heterozygotes for dominance tests of the genes involved (e.g., Sr/S'), and a good deal of fruitless effort could have been avoided. Considerable space has been devoted to the present involved status of the formal genetics of E. coli K-12, not because of its intrinsic impor- tance, but because it is a necessary foundation for the evaluation of the application of recom- bination to genetic problems. I think everyone, including the protagonists, would agree that much of the controversy in the genetics of yeast de- volves upon very similar questions on the di4 tribution of chromosomes at meiosis. The foun- dations of yeast genetics are older and broader than those of E. coli, and the advantages of a morphologically well-understood life cycle, the coherence of meiotic products in the ascus, and matings under the control of heterothallism, are not yet shared by the bacterial geneticist. This contrast provides all the more reason for attempt- ing to build a detailed, incontrovertible founda- tion for theoretical construction in this field. It may be pertinent to enquire at this point whether the entire approach to the analysis of E. coli segregations in chromosomal or cyto- genetic terms may not be fallacious-whether there may not be an entirely unique genetic mech- anism involved. To recapitulate, at the risk of redundancy, the primary nondisjunctions are the clearest examples of bacterial hybridity. They are unstable in respect to those genetic factors which delineated the parental bacteria, and in no others. The association of these factors is intracellular, as proven by direct single cell iso- lations., The factors from the individual parents show a strong tendency to separate in the same blocks or combinations, but the linkage is not absolute and all categories of recombinations occur. Every heritable character for which a mu- tation has occurred to permit a test is included in this segregation system. The evidence for a uni-linear association of factors is quite incom- plete, but some groups of factors are most likely organized in this way (e.g., V, - Lac - V, from haploid data). Although the detailed organization of the genetic structures is still obscure, their resemblance to chromosomal systems is incon- trovertible: linkage; high frequency of 2n output in 2n x In crosses; coupling and repulsion phases -by mutation in the diploid; response to muta- genic agents; dominance and over-dominance (vide infia); complementary segregation (qualita- tively); and permutation of segregation ratios in reverse crosses. The anomalies which must be explained include the regular elimination in most crosses of a particular segment of the genome, segregation ratios deviating from the expected 1: 1, and difficulties in a comprehensive linkage analysis. At first sight, an appeal to chromo- somal anomalies may appear to be too special to be justifiable, but we have found no alternative approach which provides a more fruitful working hypothesis. It should be kept in mind that Dro- sophila and maize are popular for genetic work precisely because of the usual regularity of their segregation behavior, and that there are whole orders of arthropods and many plant species (as M. J. D. White, C. R. Metz or R. E. Cleland would testify) whose genetic behavior would seem to show much more profound inconsistencies with a chromosomal interpretation than does that of Escherichia coli K-12. At any rate, we now have good `6reason to believe that the bacterial 426 BACTERIAL RECOMBlNATlON cell contains a special genetic substance or structure, differentiated to perform genetic func- tions.. ." (Taylor, 1949). CORRELATING CYTOLOGICAL AND GENETIC INVESTIGATIONS Other speakers at this symposium will have discussed in more detail the present status of bacterial cytology and its bearings on bacterial genetics. A number of workers have presented convincing evidence for the presence of nuclei in bacterial cells, but their identification as nuclei has hitherto been based only on incomplete morphological and cytochemical evidence, in the absence of any more direct opportunity to locate the genes within them. A most attractive objec- tive would be a documentation of the nuclear events associated with genetic recombination in E. cc& K-12, or any other suitable organism, but this is on the horizon, not at hand. Meanwhile, many investigations of mutagenesis have been predicated on probably fallacious models of bacterial cells as constructively iso- lated genes, despite the contrary cytological evidence for the multi-nucleate condition of most rod-shaped bacteria. Many of the characters used in bacterial mutation research are recessive (e.g., resistance to phage or streptomycin! SO that mutations induced in multinucleate cells could not begin to exert their phenotypic effect until nuclear separation has occurred. In this respect a comparison of vegetative cells with presumably uninucleate endospores might be fruitful. complexity. Very often, there appeared to be two larger, more dispersed "nuclei" per cell, in con- trast to two pairs of more condensed nuclei that often characterize comparable haploid cultures. The structure of the "nuclei" is obscure, for we had been unable to analyse the connections of the granules to determine whether they form a single connected group or several groups. So far, our material, interpretations, or techniques (HCl-Giemsa) have not sufficed to demonstrate clear mitotic figures, but there are many unmis- takable examples of symmetrically placed groups of chromatin both in haploid and diploid cells. The preparations so far studied do not admit of any clear interpretation in terms of doubled chro- mosomes, and it is not yet excluded that the dif- ferences reside principally in a better expansion and resolution of nuclear structure in the diploid cells. In occasional preparations haploid cul- tures have shown nearly the same order of com- plexitv in chromatinic structure as diploid (Figure 61, but to date one of us has consistently been able correctly to classifv stained smears prepared by another, ostensibly by virtue of the nuclear cytology. On two occasions, a cytoloqical de- termination correctly anticipated a later genetic definition of the status of a culture (one was a secondary Lac+ homozyqote; one a haploid cul- ture carrying an unstable gene which simulated the variegation of heterozygosity). The further cytolo$cal analysis may well rest upon tech- nical and conceptual advances of the kind dis- cussed elsewhere in this symposium. The establishment of nondisjunctional or Stempen (1950) and others have reported that "diploid" cultures opened the question of a cyto- nuclei can be visualized in living bacteria by logical comparison of 2n and n for the purposes phase contrast microscopy. This technime has of a bacterial cytogenetics. For some time, remarkable advantages for observing cells as a preparations like that illustrated in Figure 5, whole, but only faint suggestions of the nuclei have encouraged this hope. Diploids often show are apparent in preparations of E. coli K-12. cells of greater uniform length than haploids, and There is considerable fluctuation in the definition with chromatinic structures of greater apparent of the presumed chromatin (which appears light . PLATE I FIG. 5. Haploid parent, W-67 (a), and diploid offspring, H-226 (b). G iemsa hydrolysis with HCl. stain following osmic fixation and FIG. 6. Haploid cells, K-12. Giemsa-osmic-HCl. FIG. 7. Genetic effects of ultra-violet light on a diploid culture, H-226. a. Control plating showing ptepon- derance of balanced lactose-positive colonies (Lac,-/Loca-; see Figure 4C) on EMB lactose agar. b. Cornpap able plating of an aliquot exposed to ultra-violet light. FIG. 8 Phase contrast photomicrographs of microcolonies. cells, H-267, exposed to ultra-violet light. a. Control plating of strain K-12. b. From diploid FIG. 9. Cytological effects of ultra-violet light on a diploid culture, b. Microcolony from treated suspension. Giemsa-osmic-HCl. H-267. a. Microcolony from control sus- pension. FIG. 10. Mutability differences between Lacl-- alleles. a. Mutable, Y-87. b. Stable, W-112. Both on EMB lactose agar, 48-hour plates. FIG. 11. "Large Bodies" from SaZmoneIla filtrates, exposed to antiserum, The bacteria were artificially added to provide a size standard. .I. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 427 BACTERIAL RECOMBINATION in the photographs, Figure 8). Re have not yet succeeded in identifying In versus 2n cells by this method, except by the uncertain criteria of size and growth rate. One question that is intimately concerned with nuclear behavior is the mechanism of bactericide, especially by mutagenic radiations and chemi- cals. The first order kinetics of sterilization often observed with disinfectants has led to the suggestion that lethal mutations of some sort might be implicated. Haploid cells offer no op- portunity to study the possible genetic mechan- isms of their death, but diploid cells should allow of the recovery of the presumed lethals in heterozygous condition. Because of the readiness with which the dip- loid state is detected, Lacv ,I!alv primary non- disjunctional types have been used in this anal- ysis. Owing to their Lac, Lac, +-/-+ constitu- tion, all the haploid segregants are lactose- negative, in contrast to lactose-positive diploids. It was soon found that the W-sterilization re- sponses of the diploid (H-226) and the haploid K-12 were congruent, both showing a target multiplicity of several hundred. The behavior of diploid cells surviving UV is illustrated in Figures 7 and 9 from a genetic and cytological viewpoint, respectively. At first glance, it appeared that an early effect of UV was to induce haploidization, 50 per cent or more of the colonies showing this effect with a SUP- vivorship of 90 per cent, so that selective SUP viva1 is disqualified. This would make for a simple picture if the genome, chromosome, or haploid nucleus were the unit of inactivation, haploidization would be a halfway step to "nulli- ploidy" or death. This view became difficult, however, when it was found that increasing doses of UV did not continue to increase the proportion of apparent "haploid" survivors whose frequency stabilized at about 80 per cent of the total sur- vivors. It became untenable when the plates were incubated for longer periods and inspected more closely. In the center of nearly every "hap- loidized" colony, there developed a lactose- positive spot from which a diploid Lacv could be isolated. The surviving cell evidently had retained the potentiality of transmitting the dip- loid condition. However, if weakly irradiated cells are incubated for a few hours in broth virtually all of the colonies then formed on agar are truly haploid. The surviving cell is so al- tered that haploid cells are segregated and grow at a normal rate, but a residuum remains which recovers only later to form a diploid cell grow- ing normally. Th' IS is shown even more clearly by the following experiment: A drop containing about 10 cells from a diploid culture is streaked along a line at one edge of a plate. Some of the plates are irradiated. After a few hours incubation, a glass spreading rod is firmly stroked once, perpendicular to the line. This disperses the cells of individual micro- colonies in lines across the plate. Most of the progeny of unirradiated cells remain typically diploid. Many haploids are segregated during the first few hours from irradiated cells, leaving other cells producing diploid colonies whose appearance ranges from the typical to the cen- tral spot type. The likely cytological correlate of this be- havior is found in the "snakes" or long fila- mentous bacteria observed by many authors after UV treatment. The cell later forming a "snake" may first divide to form a clone of rapidly divid- ing smaller cells (haploid?) in which the "snake" remains relatively dormant for long periods (dip loid residuum?). The chromatinic material of some of these cells is highly disorganized, ag- gregated, and pycnotic, while the total nuclear material corresponds to that of dozens of ordinary cells closely packed in one filament. It is not clear why haploid segegants should recover preferentially, but it is certain that there is every opportunity for all varieties of genetic dis- and re-organization. This is reflected in the rather high proportion of lactose-positive crossovers (Lacl + Lac,+) found among the segregants, and especially in the very high frequency (50% or more) of partial segregants among residual diploids. Clark et al. (1950) have reported that UV stimulates prototroph formation. If the effect is not an artefact due to increased residual growth in the minimal agar, it may conceivably be re- lated to the centripetal condensation of nuclear material in irradiated cells, but only if cytogamy normally is more frequent than karyogsmy. Their discussion, however, includes the speculation that the irradiation "could cause increased genetic transfer of genetic material between the two strains.. . by recombination by the organisms of desirable genetic properties in an attempt to overcome undesirable side reactions that result from the treatment"-a point of view that we can- not share. The study of UV effects was initially underc taken with the expectation that diploid cells I. LEDERBERG, E. h'. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY would be more resistant than haploid, and that recessive lethals would be detected in diploid survivors by their effect of inhibiting effective segregation. The auxotrophic mutations, which can be regarded as "lethal" on synthetic medium, provide a clear model for this behavior, for seg- regation is very much reduced on this medium owing to the growth requirements of the auxo- trophic segegants. However, no clear cases of balanced lethal survivors have been found, and lethals presumably play a very limited role in bactericide. The development of clones from irradiated cells does not, however, follow an un- complicated course, and it is possible that lethal- carrying genomes are selectively eliminated to some extent. In an earlier, similar study, Atwood (1950a, b) found lethals, probably deletions, in Neurospora heterokaryons, but not in sufficient numbers to account for UV sterilization of conidia. At first sight, th'ese results may appear to con- flict with reports that UV and X-ray sensitivity in yeasts are inversely correlated with their ploidy. Atwood and Norman (1949) found a simi- lar correlation in multinucleate Ne uros pora conidia. The correspondence between the ploidy or number of nuclei and the radiobiological "tar- get multiplicity" has been taken by certain au- thors as evidence for recessive lethals as the specific genetic mechanism of killing. This is, of course, fallacious: the target multiplicity based on lethals would be larger than the ploidy by a large factor depending on the total number of loci. ahen a radiation resistant mutant of E. coli B, B/r, was first reported by ffitkin (19471, the sug- gestion was considered that the resistance might be accounted for by genetic reduplication. It was carried further by Buzzati-Traverso, et al., (19481, who suggested tentatively that certain strains of E. coli might be polyploid on the basis of a correlation between apparent nuclear volume and radiation resistance. At least for B/r, the suggestion should have been settled by the fre- quency of occurrence of recessive mutations, (Demerec and Latarjet, 1946). It is now apparent that a great many environmental variables can influence bactericidal effects of radiation, and we may presume that some of these can be imi- tated by genetypic variation. At best, radio- biological kinetics cannot be relied upon as the principal basis of any genetic interpretations. A mechanism of bactericide by UV not yet considered here is the activation of latent bac- teriophage, as described by Lwoff, et al., (1950). TABLE 10. EFFECTS OF BACTERICIDAL AGENTS ON DIPLOID Escherichia coli Radiomimetic Non-Radiomimetic X-ray Ultra-violet N-Mustard Formaldehyde Hydrogen peroxide Acetic anhydride High temperature Streptomycin Methyl green Urethane Ninhydrin Iodine Iodoacetamide Sodium desoxycholate* Acriflavine* Acetyl chloride Dimethyl sulfate Ethylene oxide *Inconclusive owing to severe clumping orinadequate bactericidal effect. A similar process occurs in lysogenic E. coli K-12, and probably accounts for an appreciable part of UV bactericide under some conditions. The general cytogenetic features of UV-response mentioned previously for lysogenic diploids hold for the irradiation of non-lysogenic cultures as well, but a detailed analysis has yet to be car- ried out to determine the role of this effect under the conditions of our experiments. Insofar as more than 50 per cent of the cells exposed to UV may show genetic responses to very small doses having little bactericidal action, the diploid cell constitutes a very sensitive sys- tem with which to test other agencies for "radio- mimetic" effects. Preliminary studies have been initiated with X-rays. At the lowest dose tested, 1000 r, a survival of about 90 per cent was ob served, and already at least a fourth of the sur- vivors showed the type of colony illustrated in Figure 7b for UV action. `Zith 10,000 r, survival was 30 per cent, and 81 per cent of the survivors were "haploidized." But increasing doses did not increase this proportion, possibly because the criterion for the radiomimetic effect is not absolute, and varies with the time the plates are incubated. A more suitable procedure would count the proportion of pure haploid colonies appearing after a brief time of incubation in broth. This would undoubtedly greatly amplify the apparent effect, but might obscure its complexity. As far as we know, this is the most sensitive radio- biological response observed with microorgan- isms, for which the usual doses are measured in tens and hundreds of kiloroentgens. Although the genetic or physiological basis of death is still not well understood, the induced haploidization can be used to classify chemical bactericides as radiomimetic or otherwise, along the lines of cytopathological studies with plant BACTERIAL RECOMBINATION and animal cells. As might be expected, the ef- fects of UV, X-rays, and nitrogen mustard have been indistinguishable in this system, except that those of UV are reversed by visible light in parallel with the photoreactivation of viability. A goup of other chemicals have been tested also. AS some of these are quite unstable in aqueous solution, the tests have been carried out with five and ten minute exposures, usually in M/IO buffer with or without added calcium carbonate. The cells are washed in the buffer, and the sys- tem is diluted (usually to 10-s or 10-S) to ter- minate the exposure. Table 10 summarizes the tests of these compounds for radiomimetic effects. It will be observed that all of the active com- pounds, except H,O, are effective reagents for substituting labile organic H- with alkyl, acyl, or similar groups. It is quite possible that hydro- gen peroxide may act similarly via the formation of reactive alkyl peroxides. It has been suggested that UV may act indirectly via the formation of hydrogen peroxide in aqueous medium, but the effect of hydrogen peroxide differs from UV in its insusceptibility to photoreactivation. Most of the compounds listed here as radio- mimetic have been reported by other workers to be effective mutagens. But the chemically nearly inert mutagens, caffeine and urethane, are not radiomimetic, and may be presumed to have a different mode of action. Rhatever the intimate mechanisms of action, it appears likely that many disinfectants act in part via effects on the genetic mechanisms. However, other agents in- cluding streptomycin, heat, iodine, and phenol show no such effects in this system. If a model of untreated bacteria as essentially equivalent to suspensions of isolated genes for mutation study is oversimplified, a similar consideration of irradiated bacteria is completely fallacious. Owing to the multinucleate condition of the start- ing material, E. coZi is not entirely suitable for detailed cytogenetic analysis of radiation effects, but the provocative question may once again be raised whether the genetically localized effects of mutagens may not be to some extent secondary concomitants of recovery from what have too often been dismissed as the "physiologicai" effects on nuclei and chromosomes, rather than reactions of individual gene molecules. PHENOGENETICS The most critical application of recombination analysis concerns the identity of mutations as alleles. This question arises frequently in two connections: in comparing different deviations from wild type (heterotypic mimics-such as dif- ferent lactose-negative or drug-resistant mutants), and in testing apparent back mutations to deter mine whether the phenotypic reversal is due to reverse or to homotypic mimic mutation. The methodology of the two tests is very close. For the first, the different mutants are crossed with each other, and the progeny examined for recur- rence of the wild type (Ab x aB = AB, but a x a1 f A); for the second the reversal stock is crossed to type, and the recurrence of the mutant is looked for:- iah! x Am = am, but A' x A #a). If, as is often the case, it is possible to devise a selec- tive or at least an indicator medium to detect the decisive nonparental genotypes by inspection, a very large number of tests may be conducted with reasonably small effort. For example, in inter- crossing Lac- mutants, 400-500 prototroph colonies on a plate can be scored at a glance for the presence of Lac+, so that a single moderate experiment of 40-50 plates will constitute 20,000 tests for crossing-over. In contrast to similar, albeit more laborious, studies in maize and Dro- sophila (Laughnan, 1949; Green and Green, 19491, the chief limitation is the frequency of spon- taneous "mutations" rather than the collection and classification of so many offspring. Crossing- over has been used primarily for the analysis of the grosser structure of the genome and a new type of relationship, pseudoallelism, is being found at the very margin of technical possibilities in higher organisms. The closer relationships of genetic units, of associations which may be represented by linkages of .Ol and ,001 centi- morgans, are an important challenge to the ver- satility and technical plasticity of microbial genetics. The first application of recombination to the differentiation of phenotypically similar mutat tions was the separation of two main types of resistance to the virus TI. Demerec and Fano (1945) found, in E.. coli B, the types B/l and B/1,5(= B/5,1) which were resistant to Tl only, or to both Tl and T5, respectively. They sug gested that these were genotypically distinct effects, but pointed out, rightly that a proper genetic test required a sexual phase to allow . rosesep, then unknown. l Similar phenotypes have `een detected in E. coli K-12 and fortuitously named I',' and V,,`(for "/l,S" and "/l" respec- tively). Intercrosses of a number of V,' stocks have given only this type; similarly for Vlar; but B-M-V,,' x T-L-V%' gave about 15 per cent TI-sensitive (V,' I',,") recombinants. This was the first example of a recombinant which was .I. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY phenotypically more than a reassortment of the overt parental characters. Several authors (see Luria, 1946) have also commented on the (infre- quent)occurrence in E. coli B, of complex resist- ance patterns in a single selective step, which look superficially like the superposition of simple patterns (e.g., B/1,5, 3,4,7), and of associations between auxotrophy and resistance. Efforts to find a genetic rationale for complex resistance have been blocked by our failure to demonstrate such types in E. coli K-12. The genetics of another Tl-(partially&resistant mutant, VIP, was mentioned in the account of Het nondisjunctions. No Tl-sensitive crossovers were found in several hundred additional tests of V,P x VI but the compound VIP/VI' is fully sensitive to Tl and T5. Be are evidently deal- ing with pseudo-alleles of the type discussed by Stephens (1948), wherein a relatively limited test fails to show crossing-over between units whose physiological differentiation is clearly shown by their interaction in heterozygotes. It would be interesting to compare recurrences of V1' of spontaneous and induced origin for their inter- action with VP. Unfortunately, VI5 cannot be scored except by tests on individual colonies, and is therefore not suitable for large scale re- combination tests. Another example of heterotypic mimics is the set of lactose-negative mutations. These are readily detected as light-colored sectors or colonies on EMElactose agar, and occur with a frequency of 1 to 5 x 10'4 in heavily irradiated populations. About 300 independent occurrences of Lac- mutations were isolated from the in- spection of approximately a million colonies during 1948-1949. This program was initiated as a specific test of the "one-gene-one-enzyme" hypothesis which at one time suggested that each enzyme was a specific product of a single gene, and that individual genes probably func- tioned via the activities of a single enzyme. This hypothesis had been supported by previous work by Beadle, Tatum, Horowitz, Bonner and others on Neurospora to the effect that most of the analysable auxotrophic mutants involved specific chemical blocks, including some which at first sight appeared to be more complex (e.g., the isoleucinevalineless). Horowitz has gone to great pains to reply to the criticism voiced here by Delbriick five years ago that many pleio- tropic mutations might be lost if one of the ef- fects were irreparable. It may be questioned whether this question goes to the heart of the matter, although it will doubtless arouse a good deal of discussion. Rhat we should like to see as evidence for the theory is a two-fold demon- stration: a) that all of the mutations directly affecting a single, well defined step are iso- local, and b) that the effects of all mutations at this locus are confined to this step. Aside from the purely technological problem of accumu- lating the necessary numbers of recurrent muta- tions in Neurospora, it is doubtful whether the first criterion could really be satisfied with pres- ent methods. The modalities of enzyme forma- tion and action are so obscure that exceptions to monomorphic action are readily assimilated within our ignorance. But perhaps more crucial is the difficulty of formulating a precise defini- TABLE 11. LACTOSE-NEGATIVE MUTANTS OF Escherichin coli K-12 Locus Lac, Lac, Phenotypic effects Other remarks 5-10% residual lactase; responds optimally to alkyl galactosides. Frequent recurrences to alleles of varying revertibility: may be a complex locus. Very little or no residual lactase. Single occurrence in present series. Lac, Lac4 Lacs Lac4 LM, Glucose-, maltose-negative. No residual activity. Ferments maltose, gluconate POCJT\Y- Ferments galactose slowly. Very little or no residual lactase. Several occurrences with same pleiotropic effect. Frequent recurrences. Very closely linked to Lacl. Several recurrences with same pleiotropic effect. Others Ferment hexoses or all carbohydrates poorly; probably affect intermediary metabolism. 432 BACTERIAL RECOiVBINATlON tion for a single gene, as other speakers at this symposium will have pointed out. The utility of lactose fermentation for such studies ensues from the ease with which mutants can be obtained and intercrossed for genetic pur- poses, and from the technical facility of assay and characterization of the enzyme (lactase or P-galactosidase, Lederberg, 1950~; Cohn and Monad, 1951) that mediates the key step. Most of these 300 mutants have been intercrossed, (not in all possible combinations, but with the estab- lished type stocks), and the results call for a group of at least seven loci, such that crosses of isolocal mutants give only Lac- prototrophs; whereas a heterolocal cross gives a proportion of Lac+, as summarized in Table 11. The only difficulties in genetic differentiation have involved Lac,- and Lac,-. These are so closely linked that several thousand prototrophs must be examined to find Lac+ with assurance; some of these are nondisjunctional heterozygotes, but crossovers are also observed. The Lac,- series has been selected for further investigation because of tbe richness of the avail- able material. A special study has been made of the reverse-mutation potentialities of different Lact-- recurrences (E. Lederberg, 1948; 1950). Some cultures have never been observed to re- vert, (L~c,-~`); others, (I=uc,-~) show numerous reverted papillae in every colony on E'MR lactose agar, (Figure lo), and intermediate rates have also been noted. The reversions have been tested ex- tensively in crosses to Lac+, and found to be bonafide reverse-mutations both phenotypically and genetically. The reversions have always shown the same stability as the original Lac+. It shou!d be pointed out that even the so-called mutable alleles have mutation rates of the order of 10-e or lo-' but selection rather than insta- bility is responsible for their proclivity to re- vert on lactose agar. Inherited shifts to a lower grade of mutability have been found both in un- treated cultures, and especially following UV treatment. I)espite extensive observations, how- ever, no heritable upgrades were noticed. Pap& lating colonies are a favorable system for genetic studies on mutability as such, for each colony on a plate provides a rough but direct score of this attribute. R&out putting undue emphasis on the superficial resemblance between Lac-" and dotted corn, we looked for mutations that might correspond to Dt-dt. Many shifts to lower muta- bility proved to be interallelic, i.e., crosses of the derived Lac-"' with Lac+ gave only the pa- rental types. Some of them, however, are the re- sult of mutations at other loci, for crosses with wild type gave L~c-~ of the original grade, as well as the parental Lac-St and Lac+. Further consideration of the mutations affecting the Lac-"' phenotype has, however, failed to provide con- vincing evidence of a bonafide dotted-like effect on mutability per se. The interactions are prob- ably phenotypic. Once such interaction was predicted a priori: if a second Lac,- mutation occurred in the Lacl-m, it would effect a stabilization of the Lac- pheno- We- Reversions at Lacl would leave the cell Lac,- and vice versa, leaving very little oppor- tunity for phenotypic reversal. This mechanism for stabilizing heterotypic phenotypes has been postulated as playing a role in evolutionary specialization and exemplified as a laboratory curiosity in Nearospora, (Mitchell and Ilitchell, 1950). In E. coli it would be experimentally rather difficult to distinguish this from a dotted- like effect. Lac/ La&" obtained by recom- bination were Lac-st as expected. In one in- stance, the suppression of mutability was asso- ciated with a loss of the ability to ferment butyl galactoside. This is characteristic of other Lac- mutants, and the phenotypic effect itself contrasts the mutation from dotted. In most instances, the second, mutability-suppressing (ms) mutation had an obvious effect on the expression of Lac+ that facilitated the identification and characterization of ms. Some ms simply inhibited glycolysis, (e.g., one owing to a nutritional deficiency for adenine plus thiamine), and directly reduced the selective advantage of Lac+ needed for the expression of reversions as papillae. Others had more unex- pected effects: one ms was identified as a Gal- (galactose) mutation which appears to have a dual effect. As compared with a Gal+ background, the Lac+ is slightly less effective. Probably more important, the residual galactosidase activity of the Lac- is increased, (though not enough to allow Gal- to be classified as a suppressor of homotypical mimic), so that the selective differ- ential between Lac- and Lac+ mutations is di- minished. The effect of the Gal- mutation in enhancing the type function of Lacl--"' is no more perplexing than the fact that butyl galactoside evokes lactase from this mutant, whereas lactose does not. (The residual lactase activity is about 10% of normal at full enzymatic adaptation). This separation of the specificity of the enzyme-forming mechanism from that of the enzyme itself leads to the paradox that lactose-grown cells are much less well adapted to lactose than cells grown on butyl galactoside. 2 J. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY Homotypical mimics of Lac + have been detected in cultures of Lac+ incubated in lactose medium for several days. Their occurrence in Lacrm is presumably overshadowed by the more prominent and earlier developing Lac+. Several loci appear to be involved, but their relationships to other "Lac" loci are not known. None of these mimics equals the original or revert Lac+ in in- tensity of fermentation-possibly a question of the adaptation of a single gene to the genetic background. This may be compared with the incomplete "reversions" of auxotroph mutants reported by Davis (1950a). These have not been studied genetically to assure that they are not mimic mutations. The concept of allelism has been verified ob- jectively in the following ways: a) phenotypically diverse mutations such as Lac,- and Lac,- have always recombined; h) allelic identities have been consistent and unambiguous, except for the closely linked Lac,, Lac,,, and Lac,; c) crosses of a Lac,- mutation (53m) with the same Lac,- allele (extracted from a heterozygote with Lac+) gave no Lac + in over 20,000 tests. For these reasons, we are convinced that these results can be pro- fitably compared with similar work on Drosophila, Neurospora and Zea. This conclusion is empha- sized because some of our crossingover experi- ments point to some complexity of genie structure. The following discussion revolves around three occurrences of Lac-: Lac,-elm; Lac,-Sam and L~c,J*~w At their initial occurrence, the first could be crossed to the other two, and the results of a few thousand tests pointed to their iso-local identity. One or two Lac+ prototrophs were noted, but were at first ascribed to the mutability of the m87 allele. Closer scrutiny showed, how- ever, that Lac+ was a consistent occurrence in crosses of 87m x 112st (6/10,000), but not in 87m x 53m (O/60,000), thus reopening the ques- tion. However, since the EMS lactose crossing medium might select for infrequent Lac+ revep sions occurring in Lac-m prototrophs, a further check on this possibility was desirable. This was afforded by a linkage test simi!ar to that used by Laughnan (1949) in the dissection of the A locus in maize. An m87 V,? stock was prepared and crossed with st-112. Owing to the linkage of V, and Lacl, most of the Lac+ arising from the mutable Lac- should remain V,`. On the other hand, if the sequence of factors were (St-112; m-87; - V6), most of the crossover Lac+ would be V6=. The latter was the case in 11 out of 14 Lac + prototrophs isolated. The St-II2 locus was therefore distinguished as Lacl,-. The question then arose: how would Lac,,- be- have in combination with Lac,- in a heterozygote? The absence of phenotypically Lac+, balanced heterozygotes, in the previous crosses gave a cor- rect hint A V,`Het B-M-m53 stock was avail- able, and crossed with the T-L- Bt-Lacla-Stl~~. Several V, r/s diploid prototrophs were recovered (via the dominance of V,") and characterized as segregating for Lac- Vsr/-st Ves. These hetero- zygotes were phenotypically lactose-negative, although they gave rise to a profusion of Lac+ papillae on further incubation on EMB lactose, probably as a result of segregation and crossing over. On the other hand, the same cross gave the usual low percentage of Lac+ colonies, and now that Het was present a few were Lacv, However, these were segregating +/-, and not balanced, They are presumed to be Lac+ crossovers, which have become involved in nondisjunctions in the usual proportions under the influence of Het. Stringent selection for such crossovers was exer- cised, and their secondary incorporation in a heterozygote was an incidental feature of Het action. The +/- types are probably Lac++/m87- st112-; this is not certain, but at any rate all of them segregated only stable Lac- and Lac +. We have here a position effect which is closely parallel to the lozenge case analysed by Green and Green (1949). The +-/-+ compound is in- effective in contrast to the full effect of the hap- loid ++ and the diploids +-/-I=+; -+/++ and ++/-- (?). Some of the down shifts in mutability in m87- may be due to mutation at Lac,,; one st- derived from m87 lost the potentiality to recombine with st112. A third locus affecting lactose, Lac,, is not far away either, but Lac,- shows comple- mentary action at least with m5.3 to produce bal- anced Lac+ heterozygotes. About 200 other mu- tants belonging to this complex await further study. A second allelic series at the Lac, locus has received more physiological than genetic study. The pleiotropic effect (Lac- MC& Glu-) can be tied to a single locus beyond reasonable doubt: the effect has recurred several times in the same form; all of the Glu- so far isolated proved to be Lac,-; reverse mutations are readily selected on any one sugar, and are homotypic for all three if they involve the LacI locus; an "intermediate," temperature-sensitive allele has occurred twice (Lac,t) which shows a graded effect on all three characters. On the other hand, Lac,t shows best of all that the pleiotropism is not a trivial inter- action in terminal metabolism, for, at different temperatures, the phenotypes Lac- Mal- Glu--; BACTERIAL RECOMBINATION -++, and +++ are found. In addition, the specific enzymatic adaptation of Lac + to the disaccharides provides for cell suspensions with the phenotypes --+, -++, and +-+. Whatever the ultimate basis of the pleiotropic effect may prove to be, it is con- cerned with the formation rather than the action of these enzymes, and as such stands in flat contra- diction to that aspect of the one-gene-one-enzyme generalization which was not at issue in connec- tion with the multigenic control of a single enzyme. However, I think there can be no question but that the most fruitful working hypothesis in any pheno- genetic analysis is that a single primary effect is involved, as is already suggested by the range of the Lac,t mutant. This primary effect cannot be regarded as at the overt enzymatic level here, but rather in or beyond the experimentally al- most inaccessible realm of the ceLZar (i.e., not genie) mechanisms of enzyme synthesis. The point of view expressed above is specif- ically supported by experiments on the interplay of genetic and environmental factors on the forma- tion of lactase. The effect of the Lac,- (and Lucia-) mutations in altering the adaptive respon- siveness of the cell to lactose, rather than the substrate specificity of the lactase, was mentioned before. As a second example, the analogue neo- lactose (altrose-galactoside) also reacts with lactase but does not stimulate its "adaptive" formation by Lac+. Attempts to select for muta- tions which would sensitise the cell to neolactose led instead to one which produced lactase con- stitutively-that is, to at least as large an extent, and perhaps larger, in cells grown on glucose as compared to lactose. A third datum is that Lac,t phenotypes were based on temperature thresholds for formation of the enzymes, rather than their action. Every mutational effect we have been able to analyse, therefore, has had no effect on the action, but only on the formation of lactase. The autonomous action of alleles would be the best criterion for directness of gene action, but so far no examples can be cited in microbial genetics. When Laca- is grown on EMB agar containing Iactose, maltose, or glucose a number of homo- typic mutations appear and are selected for. Most of these are true reverse mutations (by sub- sequent crossing tests) with the phenotype +++. A smal!er proportion, however, show phenotypes with every permutation of +/- for these three sugars, except that --+ so far isolated have all been weak fermenters of glucose. In addition to the "specific suppressors," +-- and -+-, the type ++- can also be selected either on lactose or maltose. One of the +-- types proved to be the mutation for constitutive lactase referred to earlier. The phenotype -+- (lactose-; maltose+; glu- cose-) is especially interesting to the biochemist for it points to a mechanism of maltose metabolism more complex than the usually assumed simple hydrolysis to two moles of glucose. Fortunately, M. Doudoroff and his colleagues at the University of California at Berkeley undertook to !end their skill to the biochemical analysis of this curiosity. Eventually, it was discovered that maltose was metabolized by a dismutative polymerization to starch: n(maltose)-(glucose),, + II glucose, fol- lowing which the starch was phosphorolysed, for the most part bypassing free glucose: (glucose), + (n-l) H,PO, - (n-l) glucose-l-phosphate + (1) d ucose, (Doudoroff et al., 1949). Meanwhile, Monod and Torriani (1950) had come to the same conclusion, without the benefit of a glucose- negative mutant, and named the polymerase "amylo- maltase." The present result is that the mutant has posed a weightier problem than it perhaps helped to solve, for we are left with a neat para- dox in the form of the n glucose molecules in the first equation above. The biochemist's scheme works very well for dried cell preparations, and the n(glucose) accumulates as anticipated. Intact cells, however, utilize maltose completely, with no trace of residual glucose. If glucose is added to a maltose-metabolizing system, however, ad- ded glucose is untouched. A number of untested hypotheses have come to mind, especially that a phosphorylation precedes the dismutative poly- merization of maltose, hut the more immediate question of the basis for the failure to utilize free glucose has not been settled. This sort of impasse is not unique in biochemistry, but it has come to be especially characteristic of the appli- cation of genetic and adaptive enzyme analyses. In other studies, a series of alleles has been established as the basis of streptomycin resist- ance in E. coli K-12 by Newcombe and Nyholm (195Ob) and Demerec (1950). The major allelic states are characterized as type sensitive, Ss; resistant, S'; and Sd, dependent. A number of quantitative variations of resistance or depend- ence have been mentioned, but at the time of this writing, no data have been published to determine whether these are due to modifiers or further allelic changes. If Sd cells are plated on non- streptomycin medium, mutations to types resem- bling Ss and Sr can be selected. These may in- clude both interallelic shifts and modifier muta- tions at other loci. I. LEDERBERG. E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 435 One IOCUS accounts for most or all of the mu- tations leading to streptomycin-resistance. By contrast, resistance to chloromycetin is polygenic, as shown by Cavalli and Maccacaro (1950). Crosses between a stock whose resistance to chloromycetin had been increased to a high level by repeated selection and the type sensitive led to a segregation of a wide range of levels of re- sistance. Furthermore, there was a correlation between the level of resistance of a prototroph and the proportion of unselected markers derived from the resistant parent. The recombinational investigation of these two cases of oligo- and polygenic effects respectively thus accords closely with the concepts of single- and multi- step resistance formulated by Demerec (1948). Two examples of genie instability should be mentioned before we leave the discussion of allelic relationships. Mutation rates as high as 10B2 per division can be detected and distinguished from phenotypic variegation in bacteria,, but relatively few examples have been studied (Zelle, 1942; Bunting, 1946). 0 ne of our mutable stocks, w-716, arose as a phenotypic reversal to Lac+ from a Lacrst plated on EMB lactose. This Lac+ re- peatedly throws a spectrum of types, from Lac-sr and Lac-m of all grades through slow lactose- fermenters to types barely distinguishable from v-716. The La+* typ es reverted at relatively low frequencies to Lac+ showing the same range of instability. The instability makes it difficult to distinguish between mutants and recombinants in critical crosses. However, the occurrence of Lac-at from the derived Lac-m x type makes it likely that the instability involves a suppressor locus. The second example was detected as a varie- gated colony from a UV plating of MuZ+ on EMB maltose. Sectored colonies are very common at the initial occurrence of fermentation mutants, probably due to the presence of several nuclei per bacterial cell, together with the disturbances induced by UV, but the sectoring ordinarily dis- appears after a single streaking-out. In this case, variegated colonies were noted throughout serial streakings. The hfalv behaves like the unstable Lac+ just reviewed, except that a few apparently stable &faZ+ have been isolated. As a rule, the Ma1 slow rJ It cycle is continued: Mal+ the rates \Mal- corresponding to each arrow varying considerably from culture to culture. The curious conclusion derived from these cases is that instability at a locus may persist throughout allelic substitutions. Most of the MaZ+ isolated as reversions from even the more stab1 e UaZ - forms in the series proved to be highly unstable. In many examples of genie in- stability, this attribute is assignable to a specific allele (e.g., dottable in maize), and is lost when the allele shifts to another state, but this type of instability has not been found in our ma- terial. It should be pointed out that almost all of the UV-induced mutations in our studies have in- volved sharp transitions from one state to another, with no evidence of transient or persistent insta- bility. Many of the mutants allow of reverse- mutation and can scarcely be regarded as deletions (most of which should be inviable anyhow in a haploid organism). Comparative studies with X- rays and nitrogen mustard would be desirable. "EXTRANUCLEARHEREDITY" To this point, the genetic changes discussed have been presumably nuclear, and no mention has been made of the possibility of extranuclear agents in bacterial heredity. Experimental prog- ress in cytoplasmic genetics will have been sum- marized by other speakers here. The three ways in which cytoplasmic effects would be most likely to be recognized in material like E. coli K-12 are: a) the physical separation and experimental extra- nuclear transfer of the agent; b) kinetic evidence for induced loss or attenuation; or c) a failure of segregation and of ratio-reversal in reversed crosses. Barring the discovery of specific agents comparable to acriflavine on yeasts cytoplasmic effects, if any, are most likely to be found in crosses between independently isolated strains rather than as the discrete mutations of the pre- vious discussion. Nevertheless, one character of E. coli K-12 can be regarded as subject to extra-nuclear he- reditary control, namely, lysogenicity. The prin- cipal features and the genetic importance of this phenomenon, which consists of a symbiotic, intra- cellular association of bacterium with a virus were pointed out many years ago (Burnet and Lush, 1936). Our first experience of lysogenicity was with SaZmoneZZa typhimurium, where it appears to be nearly ubiquitous (Boyd, 1950). Our interest was accentuated by the accidental discovery that E. coli K-12 was lysogenic. This resulted from the occurrence of a sensitive "mutant" as a survivor of UV treatment. The stock was mixed with type BACTERIAL RECOMBINATION cells for other purposes, and we were surprised to find numerous phage plaques. Contamination with extraneous phage was first suspected, but we soon showed that the phage was carried by all of our stocks except for the unique sensitive strain. The occurrence of the latter was entirely fortuitous, but some of the distinct E. coli strains studied for the purposes described in the next section are also sensitive to this phage, and would have served as independent indicators for its discovery. E. coli K-12 has been studied as a bacteriological type for nearly 30 years with no hint of its lysogenic character-an eloquent commentary on the latency of its symbionts. The exposure of sensitive cells to suspensions of the free phage, which we named "A," by analogy to a killer factor in Paramecium, results in the lysis of a variable proportion of cells. The sur- vivors include sensitives, new lysogenics, and an immune type which is genetically resistant to X, but does not carry it. The genetic relation- ships of these types are under study now. Sen- sitives can be crossed with each other, and the transfer of lambda from a lysogenic to a previously sensitive culture is not associated wit:! altera- tions of any other markers. The speculation that h might be involved in genetic recombination needs no further mention. In addition to mutations of the host bacterium, we have noted mutations of the symbiont to a form which attacks the standard lysogenics, but does not, however, evoke lyso- genicity itself. So far, no marked changes in the phenotype have been found in association with lysogenicity. It is anticipated, however, that the latent virus alters the serological character of the cells at least to tbe extent of the phage-antigens them- selves. In addition, Lwoff has found that, under certain conditions, the latent phage of Bacillus megatheriwn can be activated by UV so as to provoke lysis (Lwoff, et al., 1950). Lwoff and Delbriick (private communication) have extended this observation to K-12, so that infection with X can be regarded as a transformation to high UV- sensitivity. The natural history and phylogeny of the transforming agents of pneumococcus, Hemophihs, and possibly other bacteria are not likely to he uncovered in the near future, but one possibility that deserves close scrutiny is that they are akin to latent viruses whose lytic activity is no longer discernible. As Altenburg (1946) and others have pointed out, the genes of both partners in a symbiosis are available for mutations affecting the adaptation of the complex. The converse, that an adapted symbiont, i.e., a plas- magene, might become virulent has also been postulated. In the absence of any direct evidence against either view, there is no harm in suggesting that both processes take place. Instead of de- bating moot questions on the taxonomy of viruses and plasmagenes, we should encourage the cur- rent trend of emphasis on the extraction of genet- ically useful information from virology and chemo- therapy as well as the more orthodox disciplines of cell behavior. ABILITY OF NEW E. coli STRAINS TO CROSS The recombination studies thus far summarized have all involved mutants derived from a single strain, K-12 of E. ,coZi, and for the most part de- rived from the two polyauxotrophs 58-161 B-M- and Y-10 T-L-Y,-. Every other auxotroph mutant from K-12 (excepting certain pantothenic- less- W. Maas, personal communication) has re- acted in the same way, but this pair was chosen as giving the highest yields of prototrophs, prob- ably owing to linkage relationships. Attempts to find recombination in other strains were at first uniformly unsuccessful, including tests on strains W (B. D. Davis, personal com- munication), B, L-15, and several others. Cavalli and Heslot (1949) then reported that culture "123" of the British National Type Culture Collection could be crossed with WGl (a generic term for all cultures derived from strain K-12). "123" was received as a complex auxotroph whose nutrition could not be satisfactorily analysed either by Cavalli or ourselves, and for this reason, mainly, is not very suitable for more extensive work. In general, the procedure for determining cross- ability by the use of auxotroph mutants is too laborious to be worth applying when the prob ability of success is as low as it proved empiri- cally to be. Fortunately, a simpler screening method has been devised (Lederberg, 1951) that has provided satisfactory solution to one aspect of this problem. The main difficulty had been that each new culture had and still has to be sub- jected to painstaking procedures of isolating two, non-overlapping, diauxotrophic mutants before intra-fertility of the strain could be tested. By a combination of streptomycin resistance and proto- troph selection (SRP), however, new strains can be tested for inter-fertility with WGI with a mini- mum of individual manipulation. For this purpose, the new strains must be streptomycin-sensitive and should he prototrophic, Ss X+ (as most E. coli proves to be). KG1 tester (usually W-1177) is .I. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 437 streptomycin-resistant and polyauxotrophic Sr X'. Thus, neither the tested nor the tester strains will form colonies on a minimal streptomycin medium which selects only the combination S' XC. This genotype will occur as the result of recombination of the two strains or, occasionally, by mutation of Ss to 9. Fortunately, this mutation is one of the least frequent known. The occurrence of re- combination can be verified by comparing the un- selected markers of the SRP with the parents, but the main purpose of the procedure is to screen out the most likely candidates for further study. I n practice, the two strains are inoculated in broth and allowed to grow together overnight. They are then harvested, and about 5 x 10' cells plated on EMS maltose agar to which 100 micro- grams/ml of streptomycin have been added. Tbe strains which consistently produce SRP, espe- cially if they are segregating MoZ+ and -, are retained for closer study. About 650 cultures have been tested in this way, each from a separate individual, mostly hu- man. (We are indebted to the staff of the Wisconsin Public Health Laboratory, to Dr. C. P. Miller and Miss M. Bohnhoff, and especially to II. S. Benham and his staff at the University of Chicago Hospitals for supplying the larger part of these cultures.) About 25 of them have shown signs of recombina- tion with WGl; at least 20 of them almost cer- tainly. Polyauxotroph mutants have been prepared in WGI through -4, and have been used in the demonstration that recombination is successful in all inter- and intra-strain combinations of these strains. Culturally, all of the WGI fertile strains con- form to the description of E. coli, or possibly of intermediates, although a considerable number of aerogenes-type, cellobiose-fermenting cultures have been included in the tests. A variety of somatic antigen serotypes are included (we are indebted to Dr. W. H. Ewing of the USPHS Com- municable Disease Center for carrying out some of these serological determinations). Their genetic diversification has been reflected also in the fer- mentation of lactose and of sucrose, and partic- ularly in patterns of sensitivity to phages, in- cluding h, and to antibiotics produced by various other coli strains, (colicins). WG2 produces a colicin active on most of the others. It appears likely that many potentially compatible combina- tions may fail owing to the suppression of WG-1 by a colicin produced by the other parent. We are preparing for a detailed comparative genetic and serological study of these and additional strains. So far, the main point is that cross- fertile strains do exist in respectable numbers, so that K-12 is not a unique representative of bacterial recombination. Nevertheless, Professor Tatum and I are willing to admit our very good luck in his choice of a fertile strain for the first experiments. One aspect of the problem that is unfortunately not encompassed by the SRP selection method is the existence of other compatibility groups in E. co&. For this objective, there is no alter- native but to continue the isolation of suitable mutants from each culture. However, a distinct recombination system has been found in another distantly related species, Salmonella typhimwium. It should be emphatically stressed at the outset of this discussion that this system has already shown a number of unique features not shared by E. coli WG-1, and that owing to the preliminary character of the work our conclusions may have to be modified as new information is accumulated. GENETIC RECOMBINATION IN Salmonella Evidence for recombination in Salmonella sero- types has been sought in terms of the occurrence of prototrophs in combinations of auxotroph cul- tures. After inconclusive results in a few trials with various species (S. coli, S, poona, S..madelia, S. cholerae-suis) it was decided to make a con- certed survey of a coherent set of Salmonella typhimurium. In order to avoid unnecessary dup- lication, the 22 phage-resistance types described by Lilleengen (1%8) and kindly provided by him, were used as representative of the species. Di- auxotroph mutants have been obtained with the help of the penicillin selection method in 20 of these 22 strains; two were refractory. Of the 200 possible intra- and inter-strain combinations, 99 have been tested, and nine more or less con- sistently produced prototrophs on minimal agar, while control platings of the separate parents were barren. One combination gave exceptionally high yields, and our attention has been focused on this pair: LT-2A, a methionine-histidineless mutant (M-H-) from Lilleengen's type 2, and LT-22A, a two-step mutant from type 22 requiring phenylalanine plus tryosine, and tryptophane. Neither 2A nor 22A has produced any prototrophs even in dense individual platings, but together will produce as many as 10 prototrophs per mil- lion parental ceils inoculated. In most of the experiments a Gal- Xyl- derivative of 22A has been used to provide two unselected markers. The great majority of the prototrophs from 2A x 22A were Gal- Xyl-; however, a very small number of the other combinations have also been found. BACTERIAL RECOMBINATION In one comparison of the Salmonella and E. co& systems, Davis' (1950b) filtration experiment was duplicated. A U-tube, with a sintered glass sterile filter plate in the cross-limb, was filled with broth and the two compartments inoculated with 2A and 22A respectively. From time to time, the liquid was flushed back and forth between the compartments by alternating suction. When the broth was saturated, the cells from each compart- ment were harvested, washed and plated separately on minimal agar. It was repeatedly found that about lo-' prototrophs appeared from the 22A, but none from the 2A culture. Control experiments in which only one compartment was inoculated verified the integrity of the filter. This experiment showed that, in contrast to E. co& combinations, a filtrable agent (FA) was pro- duced by 2A that reacted with 22A to produce prototrophs. However, filtrates prepared directly from 2A were inactive. The paradox was resolved when it was found that the addition of a 22A fil- trate, or of a lysate of 2A originated from a lyso- genie phage secreted by 22A, provoked the for- mation of FA by 2A. FA, then, is not a normal component of 2A, but is produced under the stimu- lus of a latent phage. 9e have not succeeded in extracting significant FA activity from 2A cells heat-killed, dried, or autolysed under conditions which do not destroy FA activity. FA is resistant to a variety of disinfectant treatments that sterilize the bacteria: exposure to 56OC for 30 minutes, precipitation with alcohol, or shaking with chloroform or benzene. It can be concentrated by precipitation with ethanol or ammonium sulfate, or by sedimentation in a high speed centrifuge. A linear assay of FA is avail- able from the yield of prototrophs produced by mixing the sample with 109 - lOlo cells of 22~4 on minimal agar plates, throughout the range of 10 to 1000 prototrophs per plate. The most potent preparations have an activity of about 104 units/ml. (One unit is equivalent to the production of one prototrophic cell.) In a preliminary test, the activity was not altered by DNAse. Our suspicion that FA might be related to the filtrable granules of L-type cultures (reviewed by Klieneberger-Nobel, 1951) led to a test of various agents known to provoke the L-form for their ability to evoke FA. Aging in broth, lithium chloride, crystal violet, and especially penicillin were successfully used in place of the stimulus from 22A to evoke FA from 2A. The concentration of penicillin used, 1 u/ml. does not appreciably inhibit growth, but some abnormal cell forms, including swollen filaments, are seen. The mor- phological details of the origin of FA remain to be worked out. FA can be manifested in several ways. Pro- totrophs are induced from at least five other diauxotrophic cultures, each from a different Salmonella strain, and encompassing requirements for phenylalanine, cystine, leucine, threonine, isoleucine + valine, pyrimidines, and purines. However, platings with mutants of other strains, or with E. coli W-1177, have not given prototrophs. In addition platings of T22A Gal- Xyl- with FA on EMB with galactose plus xylose results in two types of papillate growth which give rise to Gal- Xyl + and Gal + Xyl-, respectively. So far, each genetic factor has tended to change inde- pendently. At any rate, with very few exceptions, only those character alternatives for which selec- tion was exerted have been recovered, and very little concomitant recombination of unselected markers (as in E. coli) has been seen so far. Three possible interpretations may be discussed: 1) FA may induce a non-specific genetic insta- bility in susceptible bacteria, leading to the for- mation of a variety of types including those selec- ted for. 2) FA may consist of a population of specific, distinct, transforming factors, corres- ponding to the genotype of the donor cells, each acting independentiy of the other. 3) FA may be a uniform factor, acting differently in different cells, for example, as a gamete would depending on patterns of recombination and crossing-over after fertilization. The distinction between 1) and 2) will depend largely upon the isolation and use of LT-2 mutants with distinctive markers; between 2) and 3) on further studies of recombination of selected and unselected markers. The possibility must be kept in mind that the treatments required to sterilize FA may result in or select for arte- facts not typical of the normal recombination pattern. As a working hypothesis, we suggest that FA can be correlated with the granular phase of L- type colonies. Kl ieneberger-Nobel reported some years ago that a "pleuropneumonia-like organism" was associated with cultures of Slreptobacillus moniliformis, but regarded this "L.-type" as a symbiont or parasite. How.ever, she has more recently accepted Dienes' vigorously documented proposition that the L form is a stage in the life history of this bacterium. Dienes, meanwhile, has shown that L-forms, in various aspects, can be elicited from a variety of bacterial species, especially with the aid of penicillin. These forms can be propagated on horse-serum agar; according to Tulasne et al., (1950), lactoflavine will re- J. LEDERBERG,E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY 439 place the requirement for serum. In a few cases, especially with Proteus, the cultures will revert to the normal bacterial form. The morphology, and especially the terminology, of the L-cultures are confusing, but in general the picture seems to be that tiny granules, about 0.2 - 0.3 microns in diameter are produced, either directly from bacteria, or via swollen large bodies. The latter are characteristic of L-cultures, but their usual fate is lysis. In some cases, however, they have been reported to become converted into cysts of granules or of bacteria. There has been no agree- ment as to the functions of the L-forms, although there have been persistent suggestions that they may have something to do with sexuality. Several authors describe large bodies as forming from the conjunction of two or more cells, but relevant genetic studies are lacking. The evidence connecting our FA with the L- forms of other workers as hastily summarized above is suggestive rather than conclusive. FA preparations contain tiny "granules," scarcely resolvable with the phase contrast microscope. These granules follow FA in sedimentation. Rabbit O-antiserum against S. typhimurium agglu- 1 tinates the granules together with FA. Incubation of FA with antiserum broth results in the forma- tion of large bodies from the clumps of granules, in an interval of about 4 hours, Figure 11; (bac- terial cells were added to the system to provide a standard of size and form-they have not been recovered from otherwise sterile FA preparations.) The production of large bodies, and the effective- ness of penicillin and of phage in provoking the granules tend to relate this system to those de- scribed by Dienes and others. However, we have not yet succeeded in propagating the granules, or in obtaining growth of FA activity, but we have not yet adequately imitated the design of Dienes' experiments in soft agar. Our FA filtrates have been tested for sterility by prolonged incubation of small and large sam- ples inoculated into Penassay broth, with and without bovine serum, observed for several weeks with no trace of bacterial turbidity. FA-contain- ing filtrates are difficult to sterilize,and we have established the practice of passing the clear supernatants first through a medium, then a fine Mandler filter, and then heating the filtrates to 58 degrees for 30 minutes. Any single one of these treatments is adequate to sterilize an or- dinary broth culture. There is an unmistakable suggestion of "filtrable" eIements which can regenerate bacteria probably along the lines of the reversion of L-type cultures as described by several other workers. The conditions of dormancy and of reversion to bacteria are too poorly under- stood for more detailed discussion. It must be quite obvious that the genetic effects of FA appear, at this time, to duplicate the well- known pneumococcus transformation. While casual published statements concerning the size of the transforming principle may rule out the participa- tion of L-granules in that system, the possibility that they play a part in other transformations should be examined very carefully. It is very likely that criteria of sterility which do not take into account the filtrability, dormancy, and resist- ance of L-granules may lead to artefacts, espe- cially if growing cells of a receptive strain help to provide the obscure conditions which favor the reversion of L-forms to bacteria. The implications of the L-system for any problem in which sterility is a decisive issue-and there are few bacteri- ological problems where it is not-are plainly to be seen. The artefact just mentioned, reversion from L to bacteria, has no special genetic interest, but adds one more reason for studying transformations with more than one marker at a time. In Salmonella, we have not succeeded in reisolating the donor 2A from interactions of FA with 22A under condi- tions which would allow it to be distinguished from the selected type. For example, platings on methionine-histidine agar gave only prototrophs, and no M-H- like the bacterial source, and the same holds for the Cal and Xyl differential markers. There may be a contradiction between associat- ing FA with a gametic phase that can be propagated in some systems or circumstances as a living organism, and the peculiar character of the range of recombination types so far recognized. This will have to be settled by further experiments. The unproven possibility that the filtered and sterilized FA may be degraded has already been mentioned. In addition, recombination may be hindered by structural differentiation between the various strains used in these experiments. In order to establish clearly the unique features of the E. coli system of recombination, we have previously emphasized the very pronounced dif- ferences between it and various interpretations of pneumococcus transformation. Salmonella shows some features in common to both. Vuller (1947) made a very reasonable synthesis in his interpretation of the pneumococcus transformation: "there were, in effect, still viable bacterial chro- mosomes, or parts of chromosomes floating free in the medium used. These might, in my opinion, have penetrated the capsuleless bacteria and in BACTERIAL RECOMBINATION part at least taken root there, perhaps after having undergone a kind of crossing over with the chromo- somes of the host." To the time of this writing, the genetic exchanges in the pneumococcus system have involved single characters at each step. No attempt has been made to ascertain whether more complex exchanges might not occur between intact cells, so that it has been difficult to correlate the different studies. By now, the time has arrived to determine whether the apparently conflicting data from different methods and sources can be assimi- lated into a unified concept of bacterial heredity. ACKNOWLEDGMENTS Financial support of the research summarized in this paper is acknowledged as from the follow- ing sources: the Jane Coffin Childs Vemorial Fund for ?dedical Research (administered by Pro- fessor E. L. Tatum); the Rockefeller Foundation; the Research Committee, Graduate School, Uni- versity of Wisconsin, with funds provided by the Wisconsin Alumni Research Foundation; Division of Research Grants and Fellowships, National Institute of Health, United States Public Health Service (Genetics of Salmonella: RG1445). The work of the second author (E.M.I;.) was carried out in part during the tenure of a Predoctoral Re- search Fellowship, National Cancer Institute, Public Health Service, and a University of Wis consin Fellowship in Genetics. This is Paper No. 466 of the Department of Genetics, College of Agriculture, University of Wisconsin. REFERENCES ALEXANDER, H. E., and LEIDY, G., 1951, Determina- tion of inherited traits of H. influenzae by desoxy- ribonucleic acid fractions isolated from type- specific cells. J. Exp. Med. 93: 345-359. ALTENBURG, E., 1946, The symbiont theory in ex- planation of the apparent cytoplasmic inheritance in Paramecium. Amer. Nat. 80: 661-662. ATWOOD, K. C., 1950a, The role of lethal mutation in the killing of Neurospora conidia by ultra-violet light. Genetics 35: 95-96. 1950b, The homology patterns of induced lethal muta- tions in Neurospora crassa. Biol. Bull. 99: 332. ATWOOD, K. C ., and NORMAN, A., 1949, On the inter- pretation of multi-hit survival curves. Proc. Nat. Acad. Sci. Wash. 12: 696-709. BISSET, K. A., 1950, The Cytology and Life-history of Bacteria. Edinburgh, E. and S. Livingstone. BOYD, J. S. K., 1950, The symbiotic bacteriophages of Salmonella typhi-murium. J. Path. Bact. 52: 501-5 17. BRAUN, A. C., and ELROD, R. P., 1946, Stages in the life history of Phytomonas tumefaciens. J. Bact. 52: 695-702. BUNTING, M. I., 1946, The inheritance of color in bac- teria, with special reference to Serratia marcescens. Cold Spring Harb. Symposium Quant. Biol. 11: 25-32. BURNET, F. M., and LUSH, DORA, 1936, Induced lyso- genicity and mutation of bacteriophage within lyso- genie bacteria. Austr. J. Exp. Biol. Med. Sci. 14: 27-38. BURNHAM, C. R., 1948, Cytogenetic studies of a trans- ocation between chromosomes 1 and 7 in maize. Lenetics 33: 5-21 BUZZATI-TRAVERSO; A., VlSCONTI, N. di M., and CAVALLI, L. L., 1948, Polyploidy in bacteria? Nature, Lond. 162: 295. CAVALLI, L. L., 1950, La sessualita nei batteri. Boll. 1st. sierotera Milan0 29: l-9. CAVALLI, L. L., and HESLOT, H., 1949, Recombina- tion in bacteria: outcrossing Escherichia coli K-12. Nature, Lond. 164: 1057. CAVALLI, L. L., and MACCACARO, G. A., 1950, Chloromycetin resistance in E.. coli, a case of quantitative inheritance in bacteria. Nature, Lond. 166: 991-992. CLARK, J. B., HAAS. F.. STONE, W. S.. and WYSS, O., 1950, The stimulation of gene recombination in Escherichia coli. J. Bact. 59: 375-379. COHN, M., and MONOD. J.. 1951, Purification et Dro- pridtds de la beta-galactosidase (lactase)' d' Escherichia coli. Biochim. Biophys. Acta 7: 153-174. DAVIS, B. D., 1950a, Studies on nutritionally deficient bacterial mutants isolated by means of oenicillin. ' Experientia 6: 41-50. 1950b, Nonfiltrability of the agents of genetic re- combination in Escherichia coli. J. Bact. 60: 507-508. DEMEREC, M., 1948, Origin of bacterial resistance to antibiotics. J. Bact. 56: 63-74. 1950, Reaction of populations of unicellular organisms to extreme changes in environment. Amer. Nat. 84 5-16. 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 Harb. Symposium Quant. Biol. 11: 51-59. DOUDOROFF, M., HASSID, W. Z., PUTMAN, E. W., POTTER. A. L., and LEDERBERG, J., 1949, Direct utilization of maltose by Escherichia coli. I. Biol. Chem. 179: 921-934. FISHER, R. A., 1947, The theory of linkage in poly- somic inheritance. Philos. Trans. 3. 233; 55-87. GREEN, M. M.. and GREEN, K. C., 1949. Crossing- over between alleles at the lozenge locus in Dro- sophila melanogaster. Proc. Nat. Acad. Sci. Wash. 35: 586-591. HAMLETT, G. W. D., 1926, `Ihe linkage disturbance involved in the chromosome translocation I. of Drosophila, and its probable significance. Biol. Bull. Vol. 51: 435-442. KLIENEBERGER-NOBEL, E.. 1951, Filterable forms of bacteria. Bact. Rev. 15: 77-103. LAUGHNAN, J. R., 1949, The action of allelic forms of the gene A in maize. Proc. Nat. Acad. Sci. Wash. 35: 586-591. LEDERBERG, E., 1948, The mutability of several Lac- mutants of Escherichia co& Genetics 33: 617. 1950, Genetic control of mutability in the bacterium Escherichia coli. PhD. thesis. Univ. of Wisconsin. I. LEDERBERG, E. M. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY LEDERBERG, J., 1947, Gene recombination and linked segregations in Escherichia coli. Genetics 32: 505-525. 1949, Aberrant heterozygotes in Escherichia coli. Proc. Nat. Acad. Sci. Wash. 35: 178-184. 1950a, `Ihe selection of genetic recombinations with bacterial growth inhibitors. J. Bact. 59: 211-215. 19%b, Isolation and characterization of biochemical mutants of bacteria. Meth. Med. Res. 3: 5-36. 195Oc, `Ihe Beta-D-galactosidase of Escherichia coli, strain K-12. J. Bact. 60: 381-392. 1951, Prevalence of Escherichia coli strains exhibit- ing genetic recombination. Science 114: 68-69. LEDERBERG, J., and TATUM, E. L., 1946, Novel genotypes in mixed cultures of biochemical mutants of bacteria. Cold Spring Harb. Symposium Quant. Biol. 11: 113-114. LILLEENGEN, K., 1948, Typing of Salmonella trphi muriwn by means of bacteriophage. Acta. Path. Microb. Stand. Suppl. 77. LONGLEY. A. G.. 1945. Abnormal semeaation during megasporogenesis in maize. GenetTcs 30: 100-113. LURIA. S. E.. 1946. Soontaneous bacterial mutations to resistance to a&bacterial agents. Cold Spring Harb. $ymposium Quant. Biol. 11: 130-138. LWOFF, A., SIMINOVI'ICH, L.. and KJELGAARD, N., 1950, Induction de production de phage dans une bact&ie lysogkne. Ann. Inst. Pasteur. 79: 8 15-858. MCCARTY, M., TAYLOR, H. E., and AVERY, 0. T., 1946, Biochemical studies of environmental factors essential in transformation of pneumococcal types. Cold Sorina Harb. Svmposium Onant. Riol. 11: 177-18;. " , . MITCHELL, M. B., and MITCHELL, H. K., 1950, The selective advantage of an adenineless double mu- tant over one of the single mutants involved. Proc. Nat. Acad. Sci. Wash. 36: 115-119. MONOD, J., and TORRIANI, A. M., 1950, De l'Amylo- maltase d, Escherichia coli. Ann. Inst. Past. 78: 65-77. MULLER, H. J., 1947, The gene. Proc. Roy. Sod. Lon. B. 134: l-37. NELSON, T. C., 1951, Kinetics of genetic recombina- tion in Escherichia coli. Genetics 36: 162-175. NEWCOMBE, H. B., and NYHOLM, U. H., 1950a, Ano- malous segregation in crosses of Escherichia coli. Amer. Nat. 84: 457-465. 1950b, The inheritance of streptomycin resistance and dependence in crosses of Escherichia coli. Genetics 35: 603-611. RHOADES, M., 1942, Preferential segregation in maize. Genetics 27: 395-407. STEMPEN, H., 1950, Demonstration of the chromatinic bodies of Escherichia coli and Proteus vulgaris with the aid of the phase contrast microscope. J. Bact. 60: 81-87. STEMPEN. H.. and HUTCHINSON, W. G.. 1951, The forma&on and development of large bodies in Pro- teus vulgaris 0X-19. J. Bact. 61: 321-335: 337-344. STERN, C.,-1936, Somatic crossing-over and segrega- tion in Drosophila melanogas ter. Genetics 21: 625-730. STEPHENS, S. G., 1948, A biochemical basis for the pseudo-allelic anthocyanin series in Gossypium Genetics 33: 191-214. TATUM, E. L., 1945, X-ray induced mutant strains of Escherichia coli. Proc. Nat. Acad. Sci. Wash. 31: 215-219. TATUM, E. L., and LEDERBERG, J., 1947, Gene re- combination in the bacterium Escherichia coli. J. Bact. 53: 673-684. TAYLOR, H. E., 1949, Additive effects of certain trans- forming agents from some variants of pneumococcus. J. Exp. Med. 89: 399-424. TULASNE, R., MINCK, R., and MULLER, L., 1950, Technique pour la culture des formes submicro- scopiques(formes L) du Proteus vulgaris en milieu liquide. C. R. Acad. Sci. 230: 152-154. WITKIN, E. M., 1947, Genetics of resistance to radia- tion in Escherichia coli. Genetics 32: 221-248. ZELLE, M. R., 1942, Genetic constitutions of host and oathoaen in mouse tvohoid. J. Infect. Dis. 71: i31-15u2. e . ZELLE, M. R., and LEDERBERG, J., 1951, Single cell isolations of diploid heterozygous Escherichia coli. J. Bact. 61: 351-355. ATCHLEY: During the past year numerous ex- periments have been performed in our laboratory in an attempt to demonstrate transformation in the colon bacillus. The strains which we have tried to transform were derived from E. coli B, B/R, R', and K12 The markers we have tried to transfer to appro- priately deficient organisms have included (1) the ability to synthesize certain nutrients, and (2) re- sistance to bacteriophage T,. The cultural en- vironments used for these experiments have in- cluded a synthetic broth, filedium A, and two different types of a neopeptone-meat infusion hro th. Beef serum aIhumin has been added in some of the experiments as a source of serum factor. From cultures of organisms showing the char- acter we wished to transfer to receptor organisms we have sought active transforming agents by mak- ing the following types of preparations: (1) a rela- tively pure, lightly polymerized DNA, (2) crude lysates made from cells broken up by repeated freezing and thawing, (3) crude lysates made by treating cells with ultrasonic vibration, and (4) filtrates of cultures in which the presumed "donor" strain has grown. In no case were we able toshow that the preparations we used could increase the rate at which treated strains mutated from auxo- trophism to prototrophism or from T, sensitivity to T, resistance. DELAMATER: I would merely like to note the essential similarity of many of the nuclear figures presented by Dr. Lederberg to those which were described in a previous paper. MULLER: The discussion of Dr. Lederberg's paper reminds me of the time, early in the history of the Drosophila: work-about 1913 to 1921-when we were criticized for trying to present so appar- 442 BACTERIAL RECO~MBINATlON ently mechanical a scheme. It was asserted by Bateson and his group, as well as by others, that we were able to make it fit only by bolstering it up with various accessory ad hoc hypotheses of a special nature, such as differential viability, multiple factors, interference, non-disjunction, etc., a!though these phenomena were in fact of just the sort to be expected if the scheme were true. Fortunately this criticism did not discourage the Drosophila workers from pushing their analyses still further along the same lines, which they saw were yielding results, and so they were enabled to obtain ever more convincing evidence of the validity of the so-called special hypotheses, as well as of the general scheme, and at the same time to uncover deeper-lying problems. In the same way, it seems to me, the present genetic analyses of E. coli by Dr. Lederberg deserve to be pushed still further, until the genetic scheme becomes quite clear, rather than abandoned now in favor of excursions into more mysterious waters. For this methodical procedure will provide both the conceptual and the biological tools whereby the attack on the more recondite problems of gene physiology can finally be pressed home with far greater effectiveness. In regard to one of the puzzling features of the genetic mechanism disclosed by Lederberg's studies, that of a single-branched linkage map resembling one based on a heterozygous trans- location, it would at first sight appear as if this implied the presence of different mating types. For if this really represents a translocation, as the beautiful linkage results indicate, we should have to suppose that the two cells which unite always differ in regard to this translocation. However, in view of Lederberg's evidence that individuals of the same clone can cross, a differ- ent interpretation is needed. * A simple one would be provided by supposing that the haploid contains two non-homologous chromosomes, and that after fusion of haploid cells, at meiosis, the chromosomes derived from the same parent are in each case held together at a given point or region. This may well be a heterochromatic region that includes the centromere, the joint structure form- ing a kind of chromocenter which would resemble that found in certain stages of some other organ- isms, except for the fact that this complex in one genome would remain effectively separate from *The suggestion following the asterisk was not pre- sented at the time of the open discussion immediately following Dr. .Lederberg,s paper, but was proposed the following morning, June 14, during discussion with a smaller group which included Dr. Lederberg. that of the other genome and not have recombina- tion occurring within it. Crossing over would how- ever occur in the four arms, and would give a four-armed map like that of a translocation hetero- zygote. As there would not be interference between crossings over occurring an opposite sides of a centromere, if the relations are like those in Dro- sophila, this would help to explain the abundance of multiple crossovers. It would also harmonize the cytological evidence for the existence of two chromosomes in the haploid, presented by De- Lamater, with the evidence for a single complex 1 inkage map, presented by Lederberg and his associates. POULSON: In listening to this most interesting paper it occurred to me that some of the puzzling aspects of the segregations described might be accounted for if the chromosomes of E. coli possess diffuse kinetochore properties rather than the highly localized type of kinetochore (the centromere) which characterizes those organisms on which most of our present genetic knowledge is based. In so far as I am aware no linkage studies have been carried out in those organisms in which diffuse kinetochores have been demon- strated. The work of the Schraders and others makes it clear that this condition prevails in a number of orders of insects and in scattered other forms. `Ihorougb investigation of segregation and recombination in such organisms ought to be undertaken to learn in how far they follow the rules established in other organisms and in what ways they may differ. The photographs which Dr. DeLamater showed this morning left me with the distinct impression that bacterial chromosomes may very well be pos- sessed of diffuse kinetochores. If this should prove to be so, then your work represents the first thoroughgoing study of linkage in an organism with diffuse kinetochores. The four-armed linkage map certainly suggests, as you have emphasized, the presence of a reciprocal translocation or some mechanism of preferential segregation essentially similar in principle. Since our knowledge of the genetics and cytology of translocation hetero- zygotes has been based on forms with localized kinetochores it is by no means clear how the established rules apply to diffuse forms. The relationships between centromeres, crossing over, and disjunction may very well be completely different for the case of diffuse kinetochores. Perhaps the combination of your techniques with those of DeLamater will provide the answer. I realize that this is only a suggestion, but I hope it will be of value in stimulating study of the J. LEDERBERG, E. ht. LEDERBERG, N. D. ZINDER, AND E. R. LIVELY genetics of organisms with the diffuse type of kinetochore. RESTERGAARD: I want to discuss the problem of the very low recombination values which are obtained in these crosses. One reason may be that you have not yet found the optimal environ- mental conditions for sexuality here. We know both from fungi and from algae, that the sexual phase is evoked only under very special environ- mental conditions, often different from the optimal conditions for growth. Has any work been done to study the influence of the medium on recombination frequencies? The second possibility is that you have a mating type system in these bacteria, which is not yet quite under control. This may also ex- plain why sexuality is confined to rather few strains. rlas it been possible to rule out a mating type system in K-12? LEDERBERG: There can be no doubt of the necessity of following the kinds of problem sug- gested by Dr. Westergaard. Our emphasis on formal analysis is required for just that control of the breeding technique for which Newospora has been appropriately commended. We have done a number of experiments to test the reactivity of the sexual phase to environmental changes, but no marked effects have been found. One reason, perhaps, that we have not done more in this direction is that a number of other workers here present had expressed their interest in that problem, and we were waiting to hear their results. E. coli K-12 is recorded as a homothallic sys- tem, for no preferential compatibilities have been found in recombination experiments involving a wide range of mutants derived from K-12 In par- ticular, no segregation of oppositional compat- ibility factors could be detected from persistent diploids, in contrast to the results expected from mating type mutations as reported in Schizo- saccharomyces pombe. Preferential compatibility would be very useful for further analysis, and is carefully looked for, especially in crosses involving new strains. Unfortunately, no encouragement is yet available from our experiments.