Proc. Natl. Acad. Sci. USA Vol. 83, pp. 7124-7127, September 1986 Population Biology Structure of evolving populations of : Adaptive changes are frequently associated with sequence alterations involving mobile elements belonging to the Ty family (continuous culture/transposable elements/population genetics) JULIAN ADAMS* AND PAUL W. OELLERt Division of Biological Sciences, University of Michigan, Ann Arbor, MI 48109 Communicated by R. W. Allard, May 15, 1986

ABSTRACT Haploid a and diploid a/a and a/a popula- chemostats and maintained in minimal medium (14) at 30'C in tions of Saccharomyces cerevisawe evolving in laboratory envi- culture vessels of sizes ranging between 150 and 200 ml. ronments for up to 300 generations were analyzed for sequence Glucose was added as a carbon source at a concentration of rearrangements associated with the Ty family of transposable 0.08% (wt/vol). At this concentration glucose is the substrate elements. Ih contrast to results with , evolving limiting growth (14). Dilution rates in the chemostats were populations of yeast exhibit a high frequency of sequence -0.2 hr-'. To detect the occurrence and selection ofadaptive rearrangements associated with mobile genetic elements. In clones, canavanine (haploid strain) or cycloheximide (dip- particular, adaptive shifts in these populations are often loids) resistance was monitored every 12-24 hr. For each associated with such sequence rearrangements. The results are sample, an aliquot of cells was frozen in 15% glycerol at most compatible with the explanation that there is direct -70'C for later analysis. The adaptive clones were identified selection for some ofthe sequence rearrangements. In addition, by fluctuations in the frequency of canavanine or cyclohex- the pattern of changes suggests that the structure of evolving imide resistance. Details of the rationale involved and of the populations may be more complex than expect- procedures and defined media used have been described ed. (14-16). DNA Manipulations and Hybridization Procedures. Cells Mobile genetic elements have now been identified in a wide were streaked from the freezer onto minimal medium with variety'of species (see refs. 1 and 2 and references therein). 0.08% (wt/vol) glucose (15) at 30'C, and five colonies were Although an evolutionary significance for such elements has then picked and grown in YEPD liquid medium (13) overnight frequently been postulated (see refs. 3 and 4 and references at 30'C. About 5 x 108 cells were harvested while the cultures therein), direct evidence for this has been hard to obtain. In were still in logarithmic phase and the DNA was isolated as particular, in asexual populations it is difficult to distinguish described (17). Ten micrograms ofDNA was digested with 30 between direct selective advantage of the transposition units of EcoRI (Boehringer Mannheim) or 25 units of Sal I property and selection of another characteristic ("hitchhik- (Boehringer Mannheim) for 12-16 hr and then the fragments ing"; ref. 5). Chao et al. (6) were able to show a direct were separated by electrophoresis on 0.7% 30-cm agarose selective advantage for TnJO transposition, but only to a "submarine" gels at 40 mA in TBE buffer (0.089 M Tris specific locus in the mapped at 71 minutes (7). A borate/0.089 M boric acid/0.002 M EDTA, pH 7.8; ref. 18) series ofextensive studies by Hartl and his group (8-10) were for 1900 V hr-1. The DNA was then transferred and immo- unable to show a selective advantage associated with trans- bilized on nitrocellulose filters (Schleicher & Schuell, BA85), position of TnS, though presence of this element apparently using the method of Southern (19), and hybridized with increases growth rate. In this communication we report that pJA224 plasmid DNA labeled by nick-translation with adaptive shifts in evolving populations of the yeast Sac- [a-3P]ATP (Amersham) to a specific activity of 6 x 106 charomyces cerevisiae are frequently accompanied by se- cpm/,ug of DNA, using the kit supplied by Bethesda Re- quence alterations involving the mobile elements belonging search Laboratories. pJA224 is pBR322:TyEcoRI/Sal I ob- to the Ty family (11, 12), and there is no apparent unique tained from C. Paquin and contains the 1.25-kilobase (kb) target site. The patterns of sequence alteration in these EcoRI-Sal I fragment of the TyJ element inserted in ADH2- populations are most easily explained by assuming that gc (formerly ADH3-8c; ref. 20) in place of the 0.65-kb sequence alterations bossess a direct selective advantage and EcoRI-Sal I fragment in pBR322 (21). The close similarity of are responsible for 30-50%'of the adaptive changes seen. In the hybridization spectra obtained by using this probe to addition, the patterns of sequence alteration provide impor- those obtained for the closely related strain S288C (13), using tant insight into the structure of evolving populations of an independently constructed probe (22), confirms the spec- . ificity of our probe for Ty sequences. The lengths of the restriction fragments were determined by comparison with MATERIALS AND METHODS comigrating X DNA cleaved with HindIII (Bethesda Re- search Laboratories). Strains, Media, and Growth Conditions. Populations were inoculated with either the a haploid strain CP1AB-1A, the a/a diploid strain CP1AB, or the a/a diploid strain CP1AB- RESULTS 1AA. The genotypes of these strains have been described Haploid a and diploid a/a and a/a populations of S. cerevi- (13). They are all isogeneic to each other except for the siae, initiated from single clones, were grown in glucose- mnating type locus. Continuous cultures were operated as Abbreviation: kb, kilobase(s). The publication costs of this article were defrayed in part by page charge "Present address: Department of Genetics, Washington University, payment. This article must therefore be hereby marked "advertisement" St. Louis, MO 63110. in accordance with 18 U.S.C. §1734 solely to indicate this fact. *To whom reprint requests should be addressed. 7124 Population Biology: Adams and Oeller Proc. Natl. Acad. Sci. USA 83 (1986) 7125 limited continuous cultures for up to 300 generations. Pop- our strains were derived (13). The spectra are quite remark- ulation sizes were large, on the order of4-5 x 109, and tended able as they show differences accompanying the appearance to be higher for the haploid than for the diploid populations of many of the adaptive clones. In some samples, some ofthe (14). Since evolving populations of yeast, grown under our bands, especially in the larger size classes, are darker than conditions, reproduce exclusively asexually, they may be others, suggesting two comigrating DNA fragments contain- regarded as a series of clones over time. Within each clone ing copies of Ty elements. To determine if there were any the frequency of a neutral or quasineutral will further sequence rearrangements undetected by these hy- increase due to recurrent mutation, such that its frequency bridization spectra, DNA preparations from the same sam- will be directly proportional to the time elapsed since the ples of the a haploid population and the a/a population were occurrence and predominance of the clone. Thus, during cleaved with a second restriction enzyme, Sal I, and hybrid- replacement of one clone by another, the frequency of the ization spectra were obtained as before. The majority of neutral mutation will necessarily decrease, due to the lower differences revealed by the EcoRI digests is also revealed by frequency of the neutral mutation in the newly predominating the Sal I digests, as expected for sequence rearrangements clone. The occurrence of adaptive shifts was monitored by involving the Ty elements (results not shown). However, in following the dynamics of the fluctuations of canavanine one case a sequence rearrangement undetected with an resistance in the case of the haploid population and cyclo- EcoRI digest was detected with a Sal I digest. In particular, heximide resistance in the case of the diploid populations. the Sal I hybridization spectrum for the sample isolated after Canavanine resistance is neutral under our conditions and the fourth adaptive shift in the a haploid population revealed there is weak selection against cycloheximide resistance (14). the occurrence of sequence rearrangement, whereas none These are shown in figures 2 and 3 of ref. 14 for the haploid was detected with EcoRI-cleaved DNA. Thus, the hybrid- a population and the a/a diploid population and in figure 1 of ization spectra shown in Fig. 1 may underestimate the true ref. 23 for the diploid a/a population. Further details of the number of rearrangement events. procedures and rationale involved are also presented in these Hybridization spectra obtained from the same samples, papers. Four adaptive changes were identified as having using a probe for yeast sequences with no homology to the Ty occurred in the haploid population, six in the a/a population, family of elements, showed no such differences (data not and seven in the a/a population (14, 23). To detect movement shown), confirming that non-Ty sequences do not show of the Ty elements, restriction endonuclease-cleaved DNA similar changes. Although we cannot rule out the possibility preparations from samples taken after each adaptive shift that a fraction of the differences between the hybridization were hybridized to a probe for the Ty family of elements spectra is the result of the mutational loss or gain of a following the method of Southern (19). This probe consisted restriction site, it is extremely unlikely that more than a very of a pBR322-derived plasmid containing the EcoRI-Sal I small minority of changes can be explained in this way, given fragment isolated from a TyJ element (20). This fragment the concordance between the results for the EcoRI and Sal I hybridizes to TyJ and Ty2 elements but not to delta se- digests, the expected frequency of actual and potential quences. restriction sites, and the expected mutation rate per nucleo- 1 shows the hybridization spectra for five pooled tide. We therefore conclude that the appearance of new Fig. bands in the hybridization spectra results from sequence colonies from the strains used to inoculate the populations, rearrangements associated with Ty elements. together with hybridization spectra for five pooled colonies The pattern of differences in the hybridization spectra isolated after each adaptive shift occurring in the haploid varies substantially between the three populations. Whereas population (Fig. 1 Left), the a/a diploid population (Fig. 1 the diploid a/a population shows no evidence of sequence Center), and the a/a diploid population (Fig. 1 Right). Pooled rearrangement throughout all of the six adaptive shifts, most samples were used to monitor the populations more efficient- of the samples isolated after the adaptive shifts in the haploid ly. The results show that at least 35 copies of the Ty family population and the a/a diploid population show at least one of elements are present in the genome. This is consistent with extra copy of a Ty element. Hybridization spectra obtained earlier published results (22) for the strain S288C from which from two other evolving a/a diploid populations (data not shown) also showed sequence alterations associated with Ty

012 4 6 kb elements, though with a lower frequency than shown by the 0 1234 kb -23.1 1)2l 3n46 i kb population of Fig. 1 Right. -_-1. I In some cases, especially in the later isolated samples, the - adjacent samples differ in more than one band-for example, v- 9.4 . - 9.4 the sample isolated after the third adaptive shift in the a 46ll1 9.4 haploid population-suggesting the existence of polymor- phisms in the population. To test for the existence of a RI "- 6.7 - 6.7 Y. 62.7 polymorphism, DNA was isolated separately from the five * s*- o IfI,I ifltNl colonies pooled for this sample (Fig. 1 Left) and digested with

*| sassiest4.4* EcoRI to give the hybridization spectra shown in Fig. 2. 1-~ - 4.4 - 4.4 These spectra clearly show that three different clones are I-aS3Ion present in this sample, apparently differing from each other O s by one sequence rearrangement. These results also rule out *4* -|I the possibility that the extra bands are artefacts ofincomplete a... digestion of the DNA by the restriction enzymes. Compari- 0*- son of the hybridization spectra between the original strains __0*12 and also with those for the samples isolated after each adaptive shift indicates that, as expected, the original strains show no such FIG. 1. Hybridization spectra of Ty elements in samples isolated polymorphism. after successive adaptive shifts occurring in a haploid a population of strain CP1AB-1A (Left), a diploid a/a population of strain CP1AB DISCUSSION (Center), and a diploid a/a of strain CP1AB-1AA (Right). Lane numbers indicate the number of adaptive shifts. The solid circles Sequence alterations involving Ty elements have been ob- indicate the position of new bands. served after prolonged cultivation by serial dilution in com- 7126 Population Biology: Adams and Qeller Po.Nt.Aa.SiProc. Natl. Acad. Sci. USAS 833(96(1986)

3 transposition rate appears to be remarkably insensitive to a bc dc kb varying environmental conditions. A survey of a number of environmental variables revealed only one, temperature, which affected the transposition rate (refs. 26 and 27; G. Fink, personal communication). The rate of Ty transposition increases significantly at low temperatures (10-20'C), but since all of our experiments were conducted at 30'C, this ews~ effect is not relevant here. Although movement by gene -49.4 conversion occurs with higher frequency (24), on the order of 3 x 10-3, such events typically result in the generation of *! duplications and deletions, and such gross changes would be unlikely to be selectively neutral. Selection against such changes would counteract any hitchhiking effects. We there- fore conclude that the most likely explanation for these results is that Ty-associated sequence rearrangements are the cause of some, but not all, of the adaptive changes in the samples analyzed. However, a definitive test of this hypoth- esis must await the construction of pairs of completely isogeneic strains, differing only in one Ty-associated se- FIG. 2. Hybridization spectra of the five individual clones iso- quence rearrangement. At this time it is not clear if the lated after the third adaptive shift occurring in the a haploid construction of such strains is practical. population of strain CP1AB-1A. Lane 3 in Fig. 1 Left shows the The varying rate of sequence rearrangements observed hybridization spectrum of the pooled DNA from these five colonies. between the populations is consistent with the hypothesis that a fraction is selectively favored. For a small number of plex medium (22). However, to our knowledge, widespread adaptive changes, the involvement of Ty elements in gener- Ty-associated sequence alterations correlated with adaptive ating the adaptive would be expected to be changes in evolving populations have not been reported binomially distributed and vary substantially from population previously. In addition, the results are in sharp contrast to to population. Thus, if movement of Ty elements generates similar studies involving populations of Escherichia coli. 30% of the adaptive changes, the probability of not observing There is apparently selection for TnlO transposition, but only any Ty transposition in a population with six adaptive shifts, to a specific locus in the genome mapped at 71 minutes (6, 7). such as the diploid a/a population, is still quite high, 0.12. Hartl et al. (10) detected only one case of TriS transposition The inability to detect any sequence rearrangements associ- and five cases of IS50 transposition in 98 clones isolated after ated with a/a cells may also be related to the reduced level up to 250 generations of in glucose-limited contin- of Ty transcription (28) and transposition (29) in mating uous culture. Furthermore, from the dynamics of a selected incompetent cells. marker, at least half of these six events were unrelated to The direct involvement of Ty transposition in adaptive selective changes in the populations. Our results show no change under our environmental conditions is certainly specificity and a much higher frequency of changes, between reasonable. Ty elements insert preferentially into regulatory 30% and 50%o of the adaptive changes being associated with regions adjacent to genes (12, 30) and Ty insertions resulting Ty-associated sequence arrangements. in increased, decreased, or constitutive expression have been The contrast between two patterns of results cannot be described for a number of genes (11, 12). We expect that Ty explained by the differences in the rate of Ty and TnS and Tn5 insertions resulting in such changes in gene expression would transposition. Ty transposition into nonhomologous DNA be exactly the type of mutations selected under our condi- sequences has been estimated to occur at a rate of 1 x 10-8 tion's. In comparing our results with those of Hartl et al. (10), to X 10-9 per locus per generation (24), which results in a it is relevant to note that Tn5 apparently shows no such per genome rate of 1 X 10-4 to 1 X 10-i (ref. 24). The lower preferential transposition into regulatory regions. It may also estimated rate of TnS transposition per genome (2 x 10-6 to be significant that the mechanism ofTy transposition appears 5 x 10-6 ; ref. 25) is compensated by a higher population size to be fundamentally different from that of bacterial in the experiments with E. ccli. Thus, the expected total transposons. Ty elements transpose by means of an RNA number of Ty transposition events per generation in the yeast intermediate, and transposition apparently involves reverse populations described here would be between 4 x 104 and 5 transcriptase encoded by the Ty elements themselves (27, x 101, whereas for Tn5 and the E. ccli populations (10), the 31). equivalent figures would be 3 x 104 and 8 x 104. The classical'model of evolution in asexual populations The pattern of Ty-associated sequence rearrangement predicts that adaptive mutations will accumulate sequentially observed in these populations may provide information not and that between adaptive shifts the populations should be only on their role in evolution but also on the general pattern virtually monomorphic, (14, 32, 33). Consequently, we would of evolutionary change in microorganism populations. Two expect to see progressive changes in the hybridization alternate hypotheses may be considered for the sequence spectra, representing Ty insertions or excisions, such that rearrangements seen in these populations: (i) they are direct- samples separated by only one adaptive shift should show a ly responsible for a fraction of the adaptive changes occurring maximum of one difference between the two hybridization in the populations and (ii) there is no direct selection for spectra. Such a pattern would be expected, independent of them-they only increase in the population by virtue of being the role of Ty transposition in adaptive change. Two aspects linked to a selected change in the genome (hitchhiking; ref. 5). of our results suggest that the evolutionary change in these The second hypothesis would require that the frequency of populations is more complex than expected. sequence rearrangement in the population is sufficiently high (i) The analysis of individual clones isolated after the third such that there is a reasonable probability of Ty movement adaptive shift in the a haploid population (Fig. 2) as well as occurring in a clone carrying an adaptive mutation, shortly the appearance of multiple new bands following an adaptive after its origin. However, as noted above, rates of Ty shift (e.g., the second adaptive shift, a/a diploid population, transposition per genome per generation are not high, on the Fig. 1 Left) suggests that the populations may be highly order of 1 x 10-4 to 1 X 10-5 (ref. 24). Furtherinore, Ty polymorphic during most of their history and not, as previ- Population Biology: Adams and Oeller Proc. Natl. Acad. Sci. USA 83 (1986) 7127 ously assumed, polymorphic only during population GM30959. J.A. acknowledges support of a fellowship from the changeovers. Some of the polymorphism observed may be a Alexander-von-Humboldt Foundation. result of the instability of the inserted Ty elements. The frequency of Ty element excision has been reported to be as 1. Calos, M. P. & Miller, J. H. (1980) Cell 20, 579-595. as than the 2. Shapiro, J. A., ed. (1983) Mobile Genetic Elements (Aca- high 4-5 orders of magnitude greater correspond- demic, New York). ing insertion event (25, 34). In this context it is significant that 3. Campbell, A. (1981) Annu. Rev. Microbiol. 35, 55-83. excision of a Ty element leaves behind a single delta 4. Temin, H. M. & Engels, W. (1984) in Evolutionary Theory: sequence in the genome (9). Consequently, Ty excision Paths into the Future, ed. Pollard, J. W. (Wiley, New York), should not result in reversion to the original phenotype, pp. 173-201. although the mutant phenotype may not always remain 5. Maynard Smith, J. & Haigh, J. (1974) Genet. Res. 23, 25-35. unchanged (33, 35). Thus, if maintenance of an additional Ty 6. Chao, L., Vargas, C., Spear, B. B. & Cox, E. C. (1983) element carries an intrinsic selective disadvantage, there Nature (London) 303, 633-635. could be selection for excision of that element, independent 7. Chao, L. & McBroom, S. M. (1985) Mol. Biol. Evol. 2, of any effect caused the 359-369. phenotypic by original transposition 8. Biel, S. W. & Hartl, D. L. (1981) Genetics 97, s1l. event. In addition, many transient polymorphisms, which 9. Biel, S. W. & Hartl, D. L. (1983) Genetics 103, 581-592. represent no more than the replacement of one clone by 10. Hartl, D. L., Dykuizen, D. E., Miller, R. D., Green, L. & de another, will exist in these populations that are undergoing Framond, J. (1983) Cell 35, 503-510. frequent adaptive change such that there is no period of stasis 11. Williamson, V. M. (1983) Int. Rev. Cyt. 83, 1-25. during which no adaptive change is occurring (14, 23). 12. Roeder, G. S. & Fink, G. R. (1983) in Mobile Genetic Ele- Finally, we cannot rule out the possibility that some fraction ments, ed. Shapiro, J. A. (Academic, New York), pp. 300-328. of the polymorphisms may be stable or quasistable for part of 13. Paquin, C. E. & Adams, J. (1982) Curr. Genet. 6, 21-24. the evolutionary history of these populations. 14. Paquin, C. & Adams, J. (1983) Nature (London) 302, 495-500. (ii) The pattern of between adaptive 15. Adams, J. & Hansche, P. E. (1974) Genetics 76, 327-338. differences adjacent 16. Weiss, R. L., Kukora, J. R. & Adams, J. (1975) Proc. Natl. shifts cannot always be explained in terms of a sequential Acad. Sci. USA 72, 794-798. accumulation of adaptive changes. For example, in Fig. 1 17. Denis, C. L. & Young, E. T. (1983) Mol. Cell. Biol. 3, Left, samples isolated after the first and second adaptive 360-370. shifts show two differences with respect to each other but 18. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular only one difference with respect to the original clone. This Cloning: A Laboratory Manual (Cold Spring Harbor Labora- suggests that some of the clones selected during the adaptive tory, Cold Spring Harbor, NY). shifts have arisen directly from the original clone and not, as 19. Southern, E. M. (1975) J. Mol. Biol. 98, 503-517. specified in the classical model, from the predominating clone 20. Williamson, V. M., Young, E. T. & Ciriacy, M. (1981) Cell 23, immediately preceding the adaptive shift. We propose there- 605-614. 21. Sutcliffe, J. G. (1979) Cold Spring Harbor Symp. Quant. Biol. fore that the structure ofevolving microorganism populations 43, 77-90. may consist of a reservoir of adaptive mutations in varying 22. Cameron, J. R., Loh, E. Y. & Davis, R. W. (1979) Cell 16, frequencies, most of which were present in the population 739-751. during the first few generations and which derive directly 23. Adams, J., Paquin, C., Oeller, P. W. & Lee, L. W. (1985) from the original clone. Adaptive shifts in these populations, Genetics 110, 173-185. 24. Roeder, G. S., Smith, M. & Lambie, E. J. (1984) Mol. Cell. manifested in fluctuations in the frequency of independent Biol. 4, 703-711. neutral mutants, will occur due to the predomination of one 25. Berg, D. E., Egner, C., Hirschel, B. J., Howard, J., Johnsrud, ofthe clones. The particular clone predominating will depend L., Jorgensen, R. A. & Tlsty, T. D. (1981) Cold Spring Harbor not only on its selective advantage but also on its initial Symp. Quant. Biol. 45, 115-123. frequency and its interactions with other adaptive clones. As 26. Paquin, C. E. & Williamson, V. M. (1984) Science 226, 53-55. 27. Garfinkel, D. J., Boeke, J. D. & Fink, G. R. (1985) Cell 42, the population evolves, more adaptive mutations will be 507-517. added to the reservoir, and the background in which they 28. Elder, R. T., St. John, T. P., Stinchcomb, D. T. & Davis, occur will depend on the particular clone predominating at R. W. (1983) Cold Spring Harbor Symp. Quant. Biol. 45, the time. Thus, over long periods of evolutionary time, 581-594. 29. Paquin, C. E. & Williamson, V. M. (1986) Mol. Cell. Biol. 6, adaptive mutations may be incorporated sequentially into the 70-79. selected clone, though not in the strict sequential order 30. Eibel, H. & Philippsen, P. (1984) Nature (London) 307, predicted by the classical model. Over shorter periods of 386-388. evolutionary time evolution would proceed by the serial 31. Boeke, J. D., Garfinkel, D. J., Styles, C. A. & Fink, G. R. selection and predominance of adaptive mutations, arising (1985) Cell 40, 491-500. 32. Crow, J. F. & Kimura, M. (1965) Am. Nat. 99, 439-450. mostly in a common genetic background. 33. Paquin, C. E. & Adams, J. (1983) Nature (London) 306, 368-371. We thank C. Paquin for the Ty probe, R. B. Helling for helpful 34. Ciriacy, M. & Williamson, V. M. (1981) Mol. Gen. Genet. 182, discussion, and R. A. Bender for comments on the manuscript. This 159-163. work was supported in part by National Institutes of Health Grant 35. Roeder, G. S. & Fink, G. R. (1980) Cell 21, 239-249.