Reprinted from Genetics 32: 505-525, September, 1947 “41 GENE RECOMBINATION AND LINKED SEGREGATIONS IN ESCHERICHIA COLI! JOSHUA LEDERBERG? Department of Botany and Microbiology, Osborn Botanical Laboratory, ‘ Yale University, New Haven, Conn. ~ Received August 1, 1947) HE occurrence of factor recombination in the bacterium, Escherichia colt, has been described in previous reports (LEDERBERG and Tatum, 1946 b, c, Tatum and LEDERBERG, 1947). In an attempt to elucidate further the genetic structure of this organism, these studies have been extended to crosses involv- ing several characters, and to the quantitative enumeration of various recom- bination classes. The results described in this paper provide evidence support- ing the sexual basis of factor recombination and of the existence of an organ- ized array of genes comparable to that of higher forms, MATERIALS AND METHODS The parent “wild-type” strain, K-12, of E. coli used in these experiments and the production and behavior of biochemical mutants have been described (Gray and Tatum, 1944, LEDERBERG and TATUM, 1946a, ROEPKr, LIBBY, and SMALL, 1944, TATUM, 1945). Specific requirements, notation, and other data pertinent to the biochemical mutants are summarized in tables 1 and 2. In gen- eral, a biochemical deficiency resulting from mutation is designated by the initial of the substance required (e.g. B~ for biotinless), while. the wild type alternative is written with a “++” sign (e.g. B+ to emphasize the alternative to B-). The term “prototroph” (Ryan and LEDERBERG, 1946) has been devised for strains exhibiting the nutritional behavior of the wild type, which for £. coli implies independence of any specific growth factors. Prototroph is, how- ever, not synonymous with “wild type” since it refers (a) only to the pheno- typic appearance of a culture and (b) only to nutritional and not to other pos- sible mutant characteristics. K-12 as a coliform is capable of fermenting, or producing acid, from a variety of sugars, including glucose, galactose, maltose, lactose and mannitol; how- ever, it ferments glycerol only weakly, and sucrose even less so. Because of the ease of scoring and their biochemical specificity, mutants unable to ferment various sugars have been looked for. Particular attention was paid to the isola- tion of “lactose-negative” or “Zac—” mutants, because of the taxonomic sig- nificance which has beén attached to this character. 1 Abstracted from a dissertation offered in partial fulfillment of requirements for the degree of Doctor of Philosophy at YALE UNIVERSITY, - 2 Fellow of the Jane Corrin Cuztps: MeMoRIAL FUND For MeEpicAL RESEARCH. This work has been supported by the JANE Corrin Curtps MEemoriaL FUND FOR MEpiIcAL RESEARCH. The author’s present address is: Department of Genetics, University of Wisconsin, Madison, Wis GENETICS 32: 505 September 1947 506 JOSHUA LEDERBERG The detection of fermentation mutants is readily accomplished by the use of indicator media. The medium “EMB-lactose” used in routine bacteriological work was found to be highly useful. It consists of the following (in g/l): pep- tone (or “N-Z-Case”) 10, yeast extract 1, lactose 10, agar 15, eosin Y 0.4, methylene blue 0.06, sodium chloride 5, dipotassium phosphate 2. On this medium, colonies of bacteria which can ferment lactose (or any other sugar added in its place) rapidly turn a deep purple color, while colonies of non- fermenting organisms remain white or pink but may slowly turn light blue. Lac” mutations have been recovered in two instances. Among 15,000 colonie, TABLE 1 Symbols used for various loci. 1. Nutritional requirements. Allele for requirement of a given substance is designated by the superscript “~”; independence by “+”. E.G., B7 is biotinless; B+ is biotin-independent. B biotin L leucine ' Pa phenylalanine (¢ was used previously, B, thiamin M = methionine but has been modified for typographical C cystine P proline reasons) T threonine 2. “Sugar” fermentations. The ability to ferment is designated “+”; the inability “—”. Lac lactose Gly glycerol 3. Bacteriophage resistance. Resistance is designated by the superscript “"”; sensitivity by “*”. E.G. Vy. V, resistant to Tz, T5 Vie resistant to Tr; sensitive to T'5 Vy» resistant to Tr; mucoid colonies Ve __ resistant to T6. 4. Resistance to chemical agents. Resistance and sensitivity “r” and “s” respectively, as Cla’. Cla sodium chloroacetate A sodium azide of strain Y-ro0 (T~L~B,-) obtained by spreading a culture previously treated with ultraviolet light on EMB-lactose agar, a single pink colony was noted. It proved to be the same, nutritionally, as Y-10 and was therefore regarded as 4 Lac~ mutant; this stock is labelled Y-53. Among 30,000 colonies of Y-40 (B-M-Vy’) a single Lac~ was recovered following treatment with nitrogen- mustard (TATUM, 1946), and was designated as Y-87. Tests showing that these independent mutations are probably allelic will be described in a later section (see table 5). Strains Y-53 and Y-87 differ in the rate at which the Lac~ char- acter reverts to the Lec* condition, but whether this is due to different allelic states or to differences at other loci, cannot be definitively asserted. Attempts to obtain maltose, mannitol, and galactose-negative mutants were not successful, presumably because the populations tested were too small. A glycerol-negative strain has been obtained, but the wild type ferments this polyalcohol so poorly to begin with that accurate scoring is difficult; studies on this character will not be further reported here, - 508 JOSHUA LEDERBERG been reported by ANDERSON (1946) for the corresponding mutants of E. colt B. In this paper, the designation Vi" will be used for the more frequent Ty-resist- ant mutant, which is also resistant to Ts. The symbol Vid? is reserved for the T5-sensitive, Tr-resistant mutant, but the evidence that distinct loci are in- volved will be presented 7 extenso in another place. In addition to Vi" and Vi", just mentioned, a third type of “secondary colony” has been found among populations treated with the virus 77. This type, Vie" is characterized by an exceedingly slimy or mucoid colony confor- oo FicureE 1.—The phenotypes of the four combinations of Lac and V are illustrated. In order they are: LactV; LactV ie; Lac Vy; Lae’ V 3. An EMB-lactose agar plate was first streaked vertically with the virus Tx, Subsequently, each of the bacterial types was streaked, from left to right, perpendicularly across the virus streak. After 16 hours incubation, both the Lac and Vy phenotypes are well developed. Developing in the zone where Lac~V 18 has been lysed can be seen two colonies of resistant mutants: Lac V1’. mation. Recombination studies on this mutant are complicated by its genetic instability; V0" rapidly reverts to the wild type, and in addition may also be strongly selected against in competition with Vy’. However, the locus of Vw can be distinguished from the locus of the other V; mutants by the demon- stration of a different recombination frequency with Lac. These data are sum- marized in order to emphasize the importance. of genetic tests to insure the allelic identity of phenotypically similar mutants. It is particularly fortunate that resistance tests can be conducted on EMB agar, since this allows the characterization of a strain with respect to virus- resistance and to lactose fermentation with a single streaking (see fig. 1). Mutants resistant to sodium chloroacetate (Cla’) were obtained by streaking a large number (about 107) of bacteria on putrient agar to which filter-sterilized chloroacetate has been added to make a final concentration of 2 mg/ml. At this concentration, the wild type is substantially inhibited, while resistant mutants grow luxuriously. This mutation is accompanied by deficiencies in SEGREGATIONS IN ESCHERICHIA COLI 509 the metabolism of pyruvic and acetic acids, which will be described in more detail elsewhere, Independent mutations to other inhibitors, including iodo- acetate, azide, streptomycin, streptothricin, mercuric chloride, and Brilliant Green, can be secured in a similar fashion, but genetic analysis of these muta- tions has not been completed. Morphological variation has occasionally been noted (exceedingly rough or very mucoid colonial form) but is relatively unsuitable for genetic work be- cause the presumably random choice of prototroph recombinants may be influenced. In addition to the EMB agar already described, a number of other natural or “complete” media have been used. The Difco product “Penassay Broth” has been used most extensively, and is satisfactory for the preparation of inocula, except that it must be supplemented with cystine for the growth of cystineless organisms, such as strain Y-24. Other satisfactory media include a broth con- sisting of: peptone 5, glucose 5, yeast extract 3) g/l, as well as Difco Nutrient Broth, and diverse concoctions containing peptone or casein hydrolysates and meat or yeast extract. The synthetic or minimal medium contains, in g/l: NH,Cl s, NH.NO; 1, Na,SO, 2, K2HPO, 3, KH»PO, 1, glucose 5, asparagine 1.5, MgSQ, 0.1, trace elements (Gray and Tatum 1944), and CaCh, a trace. The medium is made solid by the addition of agar in a concentration of 1. 5 percent. To avoid flocculation when used with agar, the glucose and agar in solution should be autoclaved separately, and mixed with the other components just before using. Unwashed agar (Difco) is sufficiently free of the growth factors under consideration to be satisfactory for many experiments; the use of washed agar, however, is recommended for the cleanest results. The detection of recombinants is based upon the inability of biochemical mutant bacteria to proliferate in the absence of their specific growth sub- stances. Plating in minimal agar, therefore, has the effect of a sieve for proto- troph cells. To insure against contamination with prototrophs derived by re- verse mutation, which has been noticed at certain loci, it has been desirable to use multiple biochemical mutants as the parental stocks in recombination studies. Coincidental reversion at two or more loci is theoretically improbable, and experimentally undemonstrable (RYAN, 1946, Tatum and LEDERBERG, 1947). Forexample, plating either B- ~T+L+Byt or BtMt+T ~L~ By separately into minimal agar did not lead to the appearance of prototrophs, B+M+T+L+. By*. When, however, a mixture of these cell types was so “sieved,” one proto- troph was found for about each 107 cells inoculated. These have been assumed to arise from the recombination of “+” alleles to form the prototroph. In previous experiments, the two multiple mutants were inoculated together into a complete medium and allowed to grow in mixed culture before plating into minimal agar. This method is not satisfactory for present Purposes be- cause it allows possible selective differentials to alter the relative frequencies of different recombination classes. A modified procedure has been developed, which will now be described in detail. The mutant stocks are maintained on “complete” agar slants, transferred 510 JOSHUA LEDERBERG at intervals of 6-8 weeks. They are inoculated separately into test-tubes con- taining about ten ml of liquid comy. lete medium and incubated overnight at 30°C with gentle shaking. The following morning, an additional ten ml of the same medium is added to each culture, and the tubes are incubated in the same manner for an additional three to five hours. These cultures contain from 1-4X 108 cells per ml. They are then washed in the following manner: the cotton ° plugs are replaced with sterile corks which have been kept in gg percent alcohol and the alcohol flamed off just before using. The cultures are then centrifuged at about 2500 r.p.m. for 20 minutes, which suffices to pack the cells in the bottom of the test tubes. The supernatant medium is carefully poured off, and the tube is rinsed with about ro ml sterile distilled water, care being taken not to disturb the pellet. The cells are then resuspended in an additional 15-20 ml sterile water, and recentrifuged. The Supernatant wash water is decanted and replaced with an equal volume of fresh sterile water, in which the cells are suspended. In the meantime, minimal agar plates are prepared. A bottom layer of about 15 ml minimal agar is poured into each Petri plate and allowed to solidify. Cell suspensions of different mutant stocks are mixed at this time and measured quantities (usually about 10%~10° cells) are pipetted onto the agar surface. At this time also, one may add such growth factor supplements as are desired to permit the growth of recombination types other than prototrophs. The cell suspensions are then mixed into a layer of about ten ml molten mini- mal agar (at 45~50°C) which is poured onto the plates. After the agar hardens, the plates are incubated at 30°C for a period of 48 hours. At this time proto- troph colonies will be found distributed throughout the plate, many of them at or near the surface and accessible to picking for further characterization. The procedure may be varied in several ways. It is important however that the inoculum consist of “young” cells, since cultures of 24 hours or older have given quite inconsistent results. It is possible to store the inoculum in distilled water for at least twenty-four hours without appreciably affecting the yield, which suggests that the aggregation of genetic types leading to the recombi- nation process occurs in the molten or the solidified agar. This occurrence must, however, take place within a few hours, since the recombinant proto- trophs are not appreciably slower to appear than wild type cells in a similar physiological state which may be streaked on the surface of the plates. Pre- sumably, therefore, one could increase the yield of prototrophs by making conditions more favorable for the free contact of the cells, as by packing them together in a centrifuge tube in minimal liquid medium. However the compli- cation of proliferation of prototrophs already formed would interfere with the interpretation of such an experiment. Many physiological factors may inter- fere with the recombination process, and, for example, the yield may be re- duced markedly by inoculating too heavily, or by omitting an under-layer of agar into which, presumably, deleterious metabolic products may diffuse. In- stead of mixing the cells in semisolid agar, it is possible to streak the mixture on the surface of slightly dried minimal agar plates. Under these conditions, how- ever, the prototroph colonies are likely to be more heavily contaminated with the residual parental mutant types. SEGREGATIONS IN ESCHERICHIA COLI 511 For most purposes, however, this contamination may be ignored, as will be shown in a later section. Prototroph colonies are then fished and streaked di- rectly on EMB plates, or otherwise tested, to classify them with respect to other factors that may be segregating. RESULTS AND CONCLUSIONS In most organisms inheritance is studied by the examination cf zygotes carrying the gene alternatives determining a character. The segregants are chosen at random, and factor linkage is recognized by deviations in the fre- quency of parental and new couplings of a series of characters. In the absence of a random method of separating zygotes in E. coli, one is limited here to the members of specific recombination classes, namely the prototrophs. It is how- ever, possible to introduce other factor differences into the biochemical mu- tants from which prototrophs are obtained, and to determine how such factors segregate into this recombination class. It. was hoped in this way to obtain in- formation concerning the haploid or diploid condition of the bacterial cell, and to determine whether factors segregated at random, or according to specific, perhaps linear chromosomal laws. The first factor pair to which this approach was applied was V,"/V;3 (LEDER- BERG and TatuM, 1946b). In the cross B-M~P+T+V\ x B+M+P-T-V;, ten BtM+*P*+T* were isolated. Eight proved to be Vy’ while two were V;*. This at once suggested that the vegetative cell of £. cold is haploid, since segregation could be observed in the first filial generation clone. It was noted also at that time that the “reversed” cross: B>-M~P+T+Vy XX B+M+P-T-Vy' gave quite a different ratio of r/s in the prototrophs, riamely 3:7. Results on so small a sample are of doubtful significance, but they suggested the technique by which the basis of this character “segregation” could be elucidated. For this reason, the study of “reversed” crosses was extended to include numerically more data, using various combinations of mutants, and involving in addition to Vy /Vi,Lact/Lac-.'The information which was obtained is summarized in tables 3 and 5. The data show clearly that neither of the factor alternatives Vi"/ Vi! or Lact/Lac~ segregates at random into the prototroph recombination class, However, the occurrence of all factor combinations, albeit with different frequencies, is evident, at least with respect to Lac and V;. It seemed clear that there are only two alternative explanations for the unequal frequencies with which alternative alleles are manifested in the prototrophs: (a) that the alleles were characterized by some differential physiological property, such as dominance, or preferential segregation, or (b) that the nonrandom segregation was due purely to the mechanics of factor recombination, which is to say a linkage system. The results of “reversed crosses” have a distinct bearing on this problem. If nonrandom segregation into protctrophs were due to some physiological prop- erty of the allele concerned, its particular coupling in the parent in which it is introduced should have no great effect on the segregation frequency; if on the other hand, the effect were purely mechanical, the segregation would reflect entirely the couplings of the parents, and the substitution of one allele for 512 JOSHUA LEDERBERG another in the parents (as in reversed crosses) should lead to a corresponding inversion in the ratios with which that allele is found in the prototrophs. The tables cited show that in every case there is no agreement between the ratios found in reversed crosses, unless the comparison is made with one of the ratios inverted, in which case there is reasonably good agreement. This result is in accord with the hypothesis that the genes in F. coli are arranged in one or more linkage groups, and is in disagreement with the postulation of a diploid con- TABLE 3 Comparisons of V1" segregations when introduced with alternative parents.* protorrorus [BtAd+PatCtT+L +B tPt] PARENTS Vy Vy Vy B Pa C-TtPt BtPatCtT-P- eee Vy x eae Vy 76 6 92 “Vy x 3 Vy 30 107 22 B-Pa-CT+L*B,t BtPatCtT-L-B,- soe Vir x see Vy 80 23 77 oe Vy x see Vie 53 133 28 B-M-TtP* BYM+T-P- eee Vy x coe Vy 49 8 86 ++ Ve x +e Vier 5 19 21 * See LEDERBERG (1947) for a statistical analysis of tables 3, 5, and 6. dition, or with a state of indefinite “ploidy” which would be characteristic of a system of cytoplasmic inheritance. The results of these experiments seemed sufficiently secure that one could adopt the existence of a linkage system as a working hypothesis and on this foundation, an attempt has been initiated to “map” a number of markers in E. coli. It was hoped at first that there might be found linkage groups which would be independent of one another, so that recombination between bio- chemical markers in one group could be used to detect recombinants, yet not interfere with the segregations in the other group(s). There was, however, no immediate prospect that these relationships could be found initially, so it was decided to study linkage relationships in a single pair of mutant stocks, and their derivatives. The stocks which were selected for this study were 58-161 (B-M~) and Y¥-53 (7~L~B,-Lac~) and their Vi" mutants. Since Lac and V; could be so readily scored, using only a single streak from each prototroph colony which appeared, it was hoped that the collection of an adequate volume of data could be accomplished with greater facility than if biochemical markers only were used. It was, however, necessary to determine the relationships of the biochemical mutant loci of which at least four must be used to obtain recombinants, Mix- tures were, therefore, plated into minimal medium supplemented with a single SEGREGATIONS IN ESCHERICHIA COLI 513 nutritional requirement, i.e., either biotin, methionine, threonine, leucine, or thiamin, allowing the proliferation of the corresponding single mutant as well as the prototrophic type. Colonies were then picked at random and scored ac- cording to their nutritional requirements. The results are summarized in table 4. Unfortunately, it was found that the addition of methionine to the minimal medium allowed excessive growth of B~M-, presumably because of a degree of contamination of the methionine with biotin. This datum is, however, not essential for the argument. In general, it will be seen that the + classes are markedly and significantly more frequent than the single mutant types, with TABLE 4 Relative frequency of various biochemical recombination classes tn the cross. BUM-T*+L*B\+X BtM+tT-L-B,-* FROM NUMBER RECOMBINATION CLASSES FOUND PLATES OF RATIO x? SUPPLEMENTED COLONIES TYPE NUMBER TYPE NUMBER WITH TESTED Biotin 70 B- 10 Bt 60 0.17 36 Threonine 46 T- 9 Tt 37 0.24 17 Leucine 56 L 5 rt 51 0.096 38 Thiamin 87 By 79 Byt 8 9.88 56 * Cells of the parental types were mixed and plated into agar supplemented with the growth factor indicated. On this medium, the two recombination classes indicated on each line of the table could form colonies. Contrasting alleles only are specified; other loci, unless otherwise speci- fied, have the “+” configuration. The x? for the ratio of single biochemically deficient types to prototrophs is calculated for a comparison with the 1:1 expectation of a random segregation. As can be seen from the x? values, the probability that the deviations are due solely to chance is, in each case, less than .oo1. the exception of B;~ which is nearly ten times as frequent as By+. Writing the cross as BTM~T+L*B,+ X B*M+T-L~B,-, these results may be interpreted as follows: 1. BtM+ T+L*+B,+ more frequent than B-M+. Therefore B and M are linked. 2. T+L* B+M*B,* more frequent than either T-L+ or T+L~. Therefore T and £ are linked. 3. Br B*M+t T+L* more frequent than B,+B+M+. Therefore B, is linked to B and M, but probably not between them. One may therefore map these five loci onto not more than two linkage groups, according to the scheme in fig. 2a. In all that follows, the [B-M] and [T-L] combinations will be regarded as single units, since conclusive information as to their relative order has not been obtained. These data so far do not allow any conclusion to be drawn as to whether the regions B, [BM] and [TJ] are linked or are independent of each other, since a recombination between them is a necessary requirement for a detectable type. 514 JOSHUA LEDERBERG TABLE 5 Segregation of Lac and V, into prototrophs issuing from various parental combinations.* PARENTS RECOMBINATIONS BOM-TtL*B,+ BtM+T-L~B,- BtM+*T+Lt Lac Vy Lac" Vy LactV yt LactVye » By 602 203 387 22 [45.8] [23.1] [29.4] { 1.7] 1. BP org 8 8 ° LactV Lac V} [45] [28] [28] [o] 70 Br** 244 157 150 Io [42.8] [27.5] [27.9] {1.9] (D) (E) (C) (triple) - By 107 145 9 61 LactVy Lac Vir [33.2] [45.0] [ 2.8] [19.0] > Brt* 134 151 9 80 [35.8] [40.4] [ 2.4] [21.4] (E) (D) (triple) (C) + Byt 28 6 46 37 [23.9] [ 5.1] [39-3] [3x.6] LacVy LactV i" -By** 102 7 201 gI [25.4] [ x.7] [50.3] [22.7] (C) (triple) (D) (E) tLactV," Lac Vy - By ** 128 ° 33 ° tLac Vir LacV +++By** 134 Lac~; not scored for V1 (P-87) (¥-53) * Cell mixtures of the indicated composition were plated into minimal agar plates or into plates supplemented with thiamin. B,*+ types refer to scores of prototrophs picked at random from minimal plates. ** B,- refers to colonies picked at random from thiamin supplemented plates. Although pre- dominantly B,- they contain B,t colonies in the proportion 1:10.as may be seen from table 4. *** BL. In this series, colonies were scored as to By, and only the By are recorded. The letters (C), (D), (E), refer to crossover types corresponding to the regions [B M{]-Lac; Lac— V1; and V,—[T L] respectively, according to the map of Fig. ad. ft Test for allelism. On the basis of table 5, the factors Vi and Lac may be brought into the argu- ment. In addition to the joint segregations of these factors, the effect of the B, segregation was studied in the following way. It would be uneconomical, in view of the relative paucity of B,*+ types, to separate these from the B,— by nutritional testing of colonies which appear on thiamin supplemented agar. Instead, the entire sample was regarded as B,~ with the proviso that it might be contaminated to the extent of ten percent with B,+. However, it has been found that the distribution of Lec and V on colonies picked from thiamin SEGREGATIONS IN ESCHERICHIA COLI 515 supplemented agar is homogeneous with the distribution in prototrophs, so that the segregation of these factors is not influenced by the A, segregation. The data in table 5 show that Lac is inclined not to separate from BM, and is therefore regarded as linked to it, while there is a similar linkage of V; to TL. Since the recombination of Lac with BM is not influenced by the inter- change between B, and BM, they are on opposite sides of BM as suggested by map 2b. Finally, a scrutiny of the interaction between the Lac and V segre- gations shows that these are not independent of each other, particularly be- cause of the rarity of the least frequent class. This suggests, then, that the two linkage groups of fig. 2b be combined to give the map of fig. 2c. (The locus of V,_ on this map is obtained from additional data.) According to this interpre- tation, the rarity of the least frequent Lec-V combination stems from the fact that a triple-crossover is necessary for its production. In fig. 2d, the cross Y-40 X Y-53 is interpreted according to the map, with a table citing the regions in which interchange rust take place to yield the given types. That the first seven factors to be investigated should fall in the same linkage group leads to the inference that there is only a single chromosome in E#. coli. This inference is supported by incomplete analyses of the segregations of 8 other markers referred to in table 1. None of these factors has been found to segregate independently of the factors which have already been described as belonging to a single linkage group. The possibility that segregation interac- tions may, in some cases, be based upon an inter-chromosomal type of inter- ference (compare STEINBERG and FRASER, 1944), has not been ruled out, however. The distances recorded in fig. 2c are derived from the recombination totals in tables 5 and 6. However, the distance between [BM] and [TL] cannot be estimated directly, but only the partition of that distance among the regions BM-Lac, Lac-V;, and Vi-TL. The relative frequency of the “triple-inter- change” type can be used to estimate the absolute map distances, if it is as- sumed that there is no interference. This frequency, about 2.1 percent, is readily calculated to be consistent with a map length of between 75 and 80 units altogether either in a two-strand or a four-strand system (LEDERBERG, 1947). These values must be regarded as rough approximations, because they are extremely sensitive to error in the estimation of the proportion of. the “triple” types. Linearity In constructing a map, and calculating distances, it has been taken for granted that there is in Z. coli a system of linear linkage, such as has been demonstrated quite conclusively in Drosophila, and inferred in all higher or- ganisms. What direct evidence may one bring to bear on this question? The method which one is forced to em; loy in hybridizing this bacterium introduces certain complications. The classical proof of linearity is based on the additive character of distances, expressed in morgans, between loci occur- ring within the same linkage group. The determination of map distances is based upon a comparison between parental and new combinations of linked 516 JOSHUA LEDERBERG genes, as determined in the progeny of zygotes selected at random. In E. coli, on the other hand, one is limited to the recovery of that recombination class in which there has necessarily been an interchange between certain biochemical loci, in the cases here discussed, betwen [BM] and [TL]. For this reason, it is not possible to obtain a direct measure of the absolute distance between factors which are located within this critical region, and any argument in favor of linearity which is based on the segregations of such factors may have the a.—t + + | + 4 4 ir Z) b—5-4 { 4 _ } 4 ' Bla 3) Lee 4 [7 4 o4 9 , ; 17 5. 33 20 A lM 3) Leo v ir) Oy a b c d e f + - = + a—t= > + f+ + Crossovers Recoverable type (c, d, or e) Bi, -M+B4T4+L14+ (a) (c, d, or e) By+M+B4+T+ I+ (2) @) BytAM+B-T+L+ (2) (f) BytM+B+T-L+ (2) (f) (e, 4, or e) By+M+B4+T41— (a) ©) By+AM+B4+T4+L+4 +++ LactVyt (a) @) see Lae Vir r+ Lacy OOOO | ++ Lact" Ficure 2.—a, b, and ¢. Mapping of genetic factors. d. The cross Bjt+Mf-B-LactV;"T+L+ X B,~ M*B*Lac-VT-L~ and some of the recoverable crossover classes. (See table 5.) flavor of circular reasoning. It would be preferable to study the segregations of factors which are assigned to loci distal to the biochemical factors whose re- combination is the basis of the detection of sexual offspring. The stocks with which this might be accomplished are not yet available, but it is hoped that they will be for future work. That there does exist some sort of linkage system is made highly credible by the results of the “reverse crosses” tabulated in tables 3 and 5. The chief diffi- culty in proving that this system is linear has been to formulate the feasible alternatives, so that critical experiments, the results of which could discrimi- nate between linearity and a given alternative, might be set up. Certain types of “linkage” can be disqualified by the data already at hand. For example, one might postulate that genes of bacteria are embedded in a two-dimensional SEGREGATIONS IN ESCHERICHIA COLI 517 matrix, and there occasionally occurs a gene-for-gene interchange. This is equivalent to the “Konversion” theory once proposed by WINKLER ( 1932), to account for interchanges in Drosophila. While this type of arrangement would account for a tendency to preserve the parental configuration, it fails to ex- plain either quantitative linkage intensities, or the interaction of segregations which is revealed by the data on Lac and V in table 5. Naturally, one could further modify the “Konversion” theory to take these exigencies into account, but in so doing one would be elaborating an exceedingly complicated theory which would, in fact, be a re-expression of a mechanical theory of linkage. The interaction of the Lac and V segregations is perhaps the most critical datum with which a genetic system for E. coli can be formulated. The inter- action may be expressed as follows: the frequency of interchanges between [BM] and Lac is dependent upon the interchanges between [BM] and V. Spe- cifically, in the cross B+M+T-L-B,- Lac-V 3X B-M-T+tL*B,+ LactVy, one finds in the B+M+T+L+B,+ the following distribution of classes: Lac-Vy" 23 percent, LactV," 2g percent (for the parental combinations) and Lac-Vy" 46 percent, LactVy 2 percent (for the new combinations). With reference to [B+*M*+], Lacm is the parental, Lact the interchange type. The proportion of Vy" (representing an interchange between V; and [BM]) is different in the Lac— and Lact segregations: namely 46:23=2:1 and 2932=14.5:1 respectively. This interaction between interchanges is most simply explained by the assump- tion that factors are located on a linear segment, so that interchanges between proximal factors also lead to the crossing over of more distal factors, barring the occurrence of additional interchanges. Additional support for the theory of linear arrangement has been found in the segregation of Vs, summarized in table 6. It will be noted that the segrega- tions of Lac, Vi, and V¢ are quite congruous in the B;~ and By*+ classes. In the totals, one finds the ratios, for each factor separately, of Lac~ 78 percent; Vo" 82 percent; V1" 36 percent; indicating that the first two are both linked to [BM] while the latter is linked to [TL]. Vs cannot, however, be to the left of {BM} because it does not interact with B,. If, therefore, there is a linear order of genes, Vs must be to the right of [BM], and because of its greater linkage intensity, nearer [BM] than is Lac. This arrangement is indicated in the map in table 6, and in fig. 2c. The agreement of the data with the hypothesis can be examined at several points. In the first place, the single exchange types, as indicated in the table, should be the most frequent. Secondly, barring multiple exchanges, an interchange between Vs and Lac should lead also to an inter- change between V, and V,. That is to say, the LactV," class should be more often V;" than V;*. Finally, in view of the similarity in linkage intensities to [BM], Lac and Vg must be closely linked. Although the “triple-interchange” types would seem to be rather frequent, reference to the table may suggest that these conditions are fulfilled. In particular, it will be noted that among the Lac”, the ratio of Ve": V6" is 94:3, or 31:1, while among the Lact, this same ratio is 10:29, or 1:3. This difference is interpreted to mean that Lac and Ve are linked to each other, as demanded by the theory of linearity. It is not, of course, proven that the gene order is not branched at some other 518 JOSHUA LEDERBERG point. The most economical hypothesis at this time, however, is that there is a single unbranched chromosome as the physical basis of inheritance in E. coli. Attempts to Induce Aberrations Using a chromosomal theory as a working hypothesis, it was hoped that some verification could be found by the study of types in which the normal order of genes was disturbed. Since there is only one chromcsome (from the TABLE 6 Segregation of Lac, V; and V5. BOM -TtL*By LactViVeXBtMtT-L By Lac VyvV g BYMtTt+Lt Lac: - _ - — + + + + Vi: r Ss r Ss Y Ss s TOTAL Ve: r r s s r r s s + Byt . 24 16 I ° 2 I 10 2 56 ‘By 52 42 2 ° 6 I 16 I 120 Total 76 58 3 ° 8 2 26 3 176 % 4333 «1-7 TST Crossover region e tT cde cdf d def ¢ ced B, see M ++ B see Vg cee Lac 68 e Vy eee T ted a b 6 d é tf + - - 5 + r + + - + + r _ s — — ** See footnote to table 4. genetic evidence), the only types of rearrangements would be changes leading to a series of inversion-transposition types. It was thought that such types might be detected by genetical procedures by virtue of their effect on crossing over. In particular, the occurrence of an inversion in the region B, - - - [MB] would be expected to have the effect of eliminating the recombination classes involving interchanges in this region. In the cross B~M~T+LtB,t XBtM+T-L-B, this would be equivalent to the suppression of prototroph recombinants; By- types, however, would be recoverable, and allow the in- vestigation of the extent of the changes. Preliminary attempts to find such aberration types have, te date, been un- successful. The procedure was as follows: Following treatment with nitrogen mustard (Tatum, 1946) or 20,000 r of X-rays, cells of Y-40 and of Y-53 were incubated separately for 24 hours, to allow the separation of cells or nuclei that might have been associated at the time of treatment. The cultures were then streaked out on nutrient agar SEGREGATIONS IN ESCHERICHIA COLI 519 plates. Single colonies of Y-40 were picked and streaked across a nutrient agar plate. Streaks of similarly treated Y-53 colonies were made from the opposite direction, so that in the center of the plate, cells of the two types were mixed, treated colony by treated colony. The plates were incubated for 24 hours, the mixed growth scraped from the plates, suspended in sterile water and plated into minimal agar. The occurrence of colonies which would not interact to produce prototrophs, as detected by plating into minimal medium, would be an indicator that the combination was heterogeneous for an aberration. Since in these experiments, both “parents” were exposed to treatment, each plating was equivalent to the testing of two chromosomes for the occurrence of an aberration. No marked variation in the yield of prototrophs was noted in tests involving 121 mustard- and 28 x-ray-treated chromosomes. This can scarcely be regarded as an adequate sample in view of the stringent selection imposed by the technique, which might be expected to eliminate any aberra- tion types which are even slightly less vigorous than the normal. This con- sideration is especially relevant in view of the “hemizygous” condition of any aberrations in the probably haploid vegetative cells. These studies will be continued. How Many Segregants per Zygote? In the experiments detailed in this paper, recombinants were obtained from different cell types which were exposed to each other in an agar medium. Therefore each prototroph recombinant colony seen by the experimenter marks the site of formation of a zygote. The question may immediately be raised whether there are at that site other recombination classes which, by virtue of their biochemical deficiencies, remain dormant within the prototroph colony on the minimal selective medium. This is equivalent to inquiring whether there is but a single viable product of meiosis (as in megasporogenesis in many higher plants) or more than one, as in the ascomycetes. The solution to this problem would be of special interest in relation to the possible occurrence of four-strand crossing over. In addition, if an appreciable proportion of rototroph colonies consisted of two distinct segregation types, it would be necessary to isolate these types for the collection of segregation data. There are at least three ways in which a zygote might yield more than one haploid recombinant. Firstly, the zygote might be capable of proliferation in the diplophase (or sporophyte), leading to the concurrence of several diploid cells, each of which might undergo meiosis independently, and by chance yield several segregation types. Secondly, a single zygote might produce, after meiosis, in addition to the prototroph, the complementary multiple mutant class. Thirdly, in a system of four-strand crossing-over, there might be two supplementary prototroph recombinants differing in the segregation of factors such as Lac and V; for which the diploid was heterozygous. Obviously, the proper investigation of these possibilities requires that one stringently avoid contamination of one colony with another. For this reason, the cell suspensions used were diluted so as to yield only about five to ten recombination colonies per plate. 520 JOSHUA LEDERBERG Crosses were made between Y-4o and Y-53 (B-M-T+tLtBytLact Vy" X Bt- M*+T~L-B,-Lac-V°) on By-containing minimal agar medium. As already noted, about go percent of the colonies from such a cross are BtM+T+L+By. The theoretical complementary class would be B-M-T-L7B,*. Because of its nutritional deficiencies, it could not be expected to proliferate on the minimal medium even had it been produced after meiosis. The possibility remains, how- ever, that a few cells of this constitution might still be present among the 103 or so By cells of the predominant type in a colony. By plating such colonies into medium lacking B, but containing biotin, methionine, threonine and leucine, the By cells would be suppressed, while the postulated multiple mu- tant type could form colonies and be recovered. The experiment just described was carried out, testing 52 colonies for their content of other cell types. In general, a thiaminless colony could be shown to contain from 10-100 cells capable of forming colonies on the B, M, 7, L medium. However, in each case investigated these have been shown to be in- distinguishable from the Y-40 parental B-M7— type, and must be presumed tc arise from a surprisingly low degree of contaminaticn of the colony with these cells from the heavily seeded plate. A few colonies were found which could be characterized as reversions from B,~ to B,+. These experiments are then, in- conclusive with respect to the occurrence of complementary genotypes in the same colony. With appropriate stocks, not as yet available, it should eventu- ally be possible to manipulate the situation so that the complementary type could be recovered selectively, excluding both parents and the predominant recombination class. A search for supplementary types was conducted with the same crosses, except that colonies appearing on By agar were streaked out directly on EMB- lactose agar to determine whether any of them were heterogeneous for Lac. In some cases, a number of isolated colonies from each EMB-test plate were then also tested for homogeneity with respect to 7z-resistance. About 90 colonies were so tested; only one colony was found containing both Lact and Lac cells. It is impossible to be certain that, with this low frequency, the single colony which was picked was not actually derived from two distinct zygotes. These experiments cannot be considered as bearing critically on the question of the occurrence of two- or four-strand crossing over because of the -absence of information concerning (a) the viability of more than one meiotic product and (b) chiasma interference. The results do, however, justify the technique of picking the prototroph colonies directly, and testing them without further purification for the collection of segregation data. A Comparison of Sexual Recombination and Transformation The occurrence of recombination types has been interpreted by us (LEDER- BERG and Tatum 1946c, TATUM and LEDERBERG 1947) as a consequence of cell fusion, “karyogamy” and meiosis with crossing over. This is, however, not the only allowable interpretation of the general phenomenon of the occurrence of new character combinations. By analogy with the systems which have been described in pneumococci (AVERY, MacLrop and McCarty 1944) and other SEGREGATIONS IN ESCHERICHIA COLI 521 strains of E. coli (Borvin and VENDRELEY 1946) one might postulate that genotypically distinct cells interact not through cell fusion, but through the release of “transforming substances” diffusing through the medium. Such transforming substances would have the property of inducing or directing mu- tational changes in the cell receiving them so as to lead to what appear to be recombination types. Our inability to separate such postulated transforming substances from the cells themselves is not proof of their absence but could be due to their lability in our hands. In previous publications, certain reasons were given for the rejection of the transformation hypothesis in favor of a picture of cell fusion, and so forth. It was not our intention thereby to state, with clairvoyant insight, that no in- vestigator will be able to duplicate the results which we have reported, using instead of living cells extracts specially prepared. It is, rather, our view that since we have been able to demonstrate no appreciable point of difference be- tween the features of gene exchange in this strain of Z. coli and in the classical materials of Mendelian experimentation, the most economical conclusion is that the mechanisms involved are also similar. In the absence of more detailed information on the behavior of transforming systems, a critic would be free to impute to such systems all of the properties which have been found to charac- terize the genetic system of E. colt, K-12. While this would he tailoring the cloth to suit the customer, it cannot be disputed that the only conclusive method by which it could be shown that cell fusion underlies gene recombina- tion would be a direct cytological demonstration. The rarity with which the — presumed zygote occurs, however (as indicated by the low frequency of effec- tive recombination types) is very discouraging to attempts to find and char- acterize the “fusion-cell,” at least in the present material. Certain genetic experiments were performed in an attempt to characterize further the behavior of this system. On the transformation hypothesis, one must attribute the rarity of the imputed transformations primarily to re- stricted conditions for susceptibility to the transforming factors released into the milieu. Otherwise, one would expect to find “transformations” for single factors much more frequent than those involving more than two factors. A glance at tables 4 and 6 illustrates that certain “multiple transformed” types are much more frequent than singly transformed classes. Under these con- ditions, one might also anticipate that genetic materials from two different kinds of cells could mix in the medium and together transform a third. In a mixture of three cell types then, one should find cases where genes from all three have combined. Using Lac and V; as markers, this type of experiment was set up in several different ways, as summarized in table 7. Pairwise, proto- trophs can be formed only from biochemically distinct and nonoverlapping parents. Combinations of B-M~- and of T~L~B,~ were arranged so that taken two at a time they were heterozygous either for Lac or for V; but not both. For example, a mixture of B-M-Lac-V1", T~-L~B,~Lac- Vy and T~L~B,~-Lact+ Vy" was plated. Prototrophs could be formed by recombination between either of the two latter and the former types. In one case, only Vi would be heterozy- gous, and the expected types would be Lac~V;' and Lac~V,°. In the other, Lac §22 JOSHUA LEDERBERG would be heterozygous, and prototrophs carrying the markers LactV; and Lac-V;" could be produced. The type LactV,* would not be expected unless, indeed, genetic material from all three types could combine in a sort of ménage 4 trois. As recorded in table 7, no instance of such a three-way combination was found in 628 tests, a different class being vacant, as anticipated, in each of the four parts of the experiment. It may be concluded that genetic factors from different cells are not freely miscible, as would be demanded by the most economical version of the interpretation of transformations. From all the experiments so far cited, it must be concluded that if trans- TABLE 7 Pairwise occurrence of recombination in mixtures of three components. RECOMBINANT PROTOTROPHS* PARENTAL TYPES BYtMtT+L*B,* or By- B-M-TtL*Byt BtM+T-L-B,- LacVy LacVy* LactVy LactV\* TOTAL Lac Vy" eel 173 49 4 ° 226 LactV" on ve 16 ° 7 28 51 LactV\" re re ° 136 37 40 213 ey Lac Vy 65 48 ° 25 138 Total 628 *. * Mixtures of the three types indicated in each experiment were plated into thiamin-contain ing agar. The prototrophs are therefore a mixture of B,- and B,+ types, as indicated in table 4, footnote. forming factors are operating in this system, the diverse factors (or genes) are not independent of one another, but are grouped in separate and immiscible parcels. Such parcels would also be potentially capable of transmitting all of the genetic factors of a cell, so that there seems to be no compelling reason why such a parcel, speaking purely genetically, could not be regarded as a gamete. MULLER (1947) has interpreted the pneumococcus transformation in terms of “still viable bacterial chromosomes or parts of chromosomes floating free in the medium . . . these have penetrated the capsuleless bacteria and in part at least, taken root there, perhaps after having undergone a kind of crossing over with the chromosomes of the host.” It remains to be seen whether this interpretation will be upheld by further studies on factor interaction in bona fide transforming systems. Several attempts were made to determine whether “transforming activity” could be separated from the living cell under conditions comparable to the . An SEGREGATIONS IN ESCHERICHIA COLI 523 platings in minimal agar medium, or after extraction of cells by Borvin’s method (Borvin ef. al., 1946). No activity was found in the supernatant of a suspension of Y-40 and Y-53 together or separately in the same minima] liquid medium to which agar is added for plating experiments. The only manipulation involved here consists of the removal of most of the bacteria by ordinary centrifugation. It could thus be shown that the “activity” was associated with the cells. Equally negative results characterized attempts to reveal transform- ing activity on culture filtrates and cell autolysates prepared as crude fractions according to Boivry’s procedure. Finally, the addition of desoxyribonuclease in a final concentration of .o5 mg/ml to the mixing and plating medium had no effect on the number of prototrophs which appeared in the cross of Y-40 and Y-53. Tests for the destruction of enzymatic activity under these conditions were, however, not done. The conclusions which we draw from these experiments are (a) that the existence of transforming factors is exceedingly unlikely and (b) it would be not worthwhile to go to extreme trouble to attempt to isolate such factors from this system until the study of bona fide transforming systems has progressed sufficiently that the genetical criteria already discussed might be applied. DISCUSSION Regardless of the stand that one takes on the issue of invisible zygotes versus non-extractable transforming factors, it can be asserted that EZ. coli K-12 pro- vides a useful tool for genetic analysis. The use of biochemical mutants as parents allows crosses which are nearly as well controlled as in Neurospora. The segregational behavior of mutant factors seems to be closely analogous to that of higher forms, and seems to compel their admission into the same arena as the genes of Drosophila. However, it would be premature to transfer these conclusions to other genetic characters of other microorganisms, each of which must be examined on its own merits. It may be wondered that the apparent recombination rate is so low. How- ever, this is possibly not to be attributed to any sexual imperfections of E. coli, but to the method of enumeration. It seems likely that an analogous com- parison of the number of somatic and generative cells in an organism like the oak-tree, or man (especially the female of the species) would give ratios similar to those prevailing in Z. coli. It is also possible that the optimal conditions for zygote formation or germination have not yet been achieved and that by special procedures the rate of zygote-formation may be accelerated to the level where there might be some hope of finding it in the field of the microscope. Attempts to detect recombination in two other strains of E. coli, B (DEMEREC and Fano, 1945) and L-15 (RoEPKE, Lipsy, and SMALL, 1944) by analogous methods have been unsuccessful (Lurtra, 1947, Tatum and LEDER- BERG, 1947). At least two strains then must be classified with the “Fungi Im- perfecti.” This dismal conclusion is, however, illuminated by the fact that many heterothallic species have been eliminated from the Fungi Imperfecti with the discovery of the appropriate opposite mating-types. At the present time, one scarcely knows where to begin to look for the bacterial analogy. The 524 JOSHUA LEDERBERG application of genetic techniques to the elucidation of unusual life-cycles in diverse bacteria (BRAUN and ELRop, 1946, DIENES, 1946) cannot fail, how- ever, to be most fruitful. The evolutionary significance of gene recombination has been made so widely familiar by DopzHansky’s book (1941), and adequately discussed, more re- cently, by MULLER (1947), that it would be impertinent to do more than simply refer to these papers. SUMMARY The recombination of genetic factors and their segregation into prototroph recombinants of Escherichia coli have been studied. It was found that genetic markers behaved as if they were part of a system of linked genes. Some evi- dence for linear order of genes was obtained. Each of 15 factors studied fell into the same linkage group. Data are given in detail for the segregation of factors involved in the biosyntheses of biotin, methionine, threonine, leucine, or thi- amin; in the fermentation of lactose, and in resistance to bacterial viruses Tr and 76. On the basis of these data a tentative 8-point genetic map of the chro- mosome of E. coli is presented. ACKNOWLEDGEMENTS The author is deeply indebted to many of his colleagues and friends, too numerous to mention, for stimulating discussions of the problems discussed in this paper. They are numbered, by and large, among the participants and discussants of the 1946 Cold Spring Harbor Symposium on Quantitative Biology. He is indebted to Dr. Mactyn McCarry for a generous sample of purified desoxyribonuclease, and to Drs. S. E. Lurra and M. Demerec for cultures of the bacteriophages used in this investigation. He owes much of the genetic analysis to the criticism of Dr. K. Maruer of the Joun Innes Hortt- CULTURAL INstITUTION. Above all, he is indebted to Proressor E. L. Tatum for the opportunity and much of the stimulus to do these experiments. LITERATURE CITED Anperson, E. H., 1946 Growth requirements of virus-resistance mutants of Escherichia coli strain “B.” Proc. Nat. Acad. Sci. 32: 120-128. Avery, O. T., C. M. MacLeop, and M. McCarry, 1944 Studies on the chemical nature of the substance inducing transformation of pneumococcal types. J. Exp. Med. 79: 137-158. Borvin, A., and R. VENDRELEY, 1946 Role de V’acide désoxy-ribonucléique hautement poly- mérisé dans le déterminisme des caractéres héréditaires des bactéries. Signification pour la biochemie générale de l’hérédité. Hely. Chim. Acta 29: 1338-1344. Bravy, A. C., and R. P. Errop, 1946 Stages in the life history of Phytomonas tumefaciens. J. Bact. 52: 695-702. “ DeEMEREC, M., and U. Fano, 1945 Bacteriophage-resistant mutants in Escherichia coli. Genetics 30: 119-1306. DreEneEs, L., 1946 Complex reproductive processes in bacteria. Cold Spring Harbor, Symposia on Quantitative Biology 11: 51-59. Doszuanskxy, Th., 1941 Genetics and the Origin of Species. Rev. Ed. xviiit+446 pp. New York: Columbia Univ. Press. Gray, C. H., and E. L. Tarum, 1944 X-ray induced growth factor deficiencies in bacteria. Proc, Nat. Acad. Sci. 30: 404-410. SEGREGATIONS IN ESCHERICHIA COLI 525 LEDERBERG, G., 1947 Genetic recombination in Escherichia colt. Dissertation; Yale University. LEDERBERG, J., and E. L. Tatum, 1946a Detection of biochemical mutants of microorganisms. J. Biol. Chem. 165: 381-382. 1946b Novel genotypes in mixed cultures of biochemical mutants of bacteria. Cold Spring Harbor, Symposia on Quantitative Biology 11: 113-114. 1946c Gene recombination in Escherichia coli. Nature 158: 558. LuriA, 8. E., 1947 Recent advances in bacterial genetics. Bact. Rev. 11: 1-40. Mutter, H. J., 1947 The gene. Proc. Roy. Soc., London B134: 1-37. Roepke, R. R., R. L. Lippy, and M. H. Smart, 1944 Mutation or variation of Escherichia coli with respect to growth requirements. J. Bact. 48: 401-412. Ryan, F. J., 1946 Back-mutation and adaptation of nutritional mutants. Cold Spring Harbor, Symposia on Quantitative Biology 11: 215-226. Ryan, F. J., and J. Leperserc, 1946 Reverse-mutation and adaptation in leucineless Neuro- spora. Proc. Nat. Acad. Sci. 32: 163-173. STEINBERG, A. G., and F. C. Fraser, 1944 Studies on the effect of X chromosome inversions on crossing over in the third chromosome of Drosophila melanogaster. Genetics 29: 83-103. Tarum, E. L., 1945 X-ray induced mutant strains of Escherichia coli. Proc, Nat. Acad. Sci. 31: 215-219. 1946 Induced biochemical mutations in bacteria. Cold Spring Harbor, Symposia on Quan- titative Biology 11: 278-284. Tatum, E. L., and J. LEDERBERG, 1947 Gene recombination in the bacterium, Escherichia coli. J. Bact. 53: 673-684. WINKLER, H., 1932 Konversions-Theorie und Austausch-Theorie. Biol. Zentralbl. 52: 163-189