Copyright 0 1983 by the Genetics Society of America

ROLE OF BIASED CONVERSION IN ONE-LOCUS NEUTRAL THEORY AND GENOME

JAMES BRUCE WALSH’

Department of Genetics, SK-50, University of Washington, Seattle, Washington 98195 Manuscript received May 12, 1983 Revised copy accepted July 5, 1983

ABSTRACT The implications of biased acting on selectively neutral alleles are investigated for a single diallelic locus in a finite population. Even a very slight conversion bias can significantly alter fixation probabilities. We argue that most newly arising mutants will be at a conversion disadvantage, resulting in a potentially greatly decreased substitution rate of new alleles com- pared with predictions from strict neutral theory. Thus, conversion bias poten- tial allows for conservation of particular alleles without having to invoke selec- tion. Conversely, we also show that bias can be important in the maintenance of repeated gene families without altering the substitution rate at other loci that experience the same amount of conversion bias, provided that the number of in the family is sufficiently large. Bias can, therefore, be important at the genomic level and yet be unimportant at the populational level. Finally, we discuss the role of biased gene conversion in events, concluding that this type of molecular turnover acting independently at many individual loci is very unlikely to decrease the time required for two allopatric populations to speciate.

HE evolutionary implications of molecular turnover processes within the T genome such as gene conversion, and transposition have recently received much attention. The potential for these processes to give rise to “selfish” DNA (DOOLITTLEand SAPIENZA1980; ORCELand CRICK 1980) as well as the role of these forces in speciation events (DOVER1982; ROSEand DOOLITTLE1983) has been the subject of much discussion. Popula- tion genetics models of molecular turnover have examined the evolution of multigene families (OHTA 1981; NAGYLAKIand PETES 1982), the amount of selfish DNA in the genome (OHTA and KIMURA1981; OHTA 1983) and de- terministic single-locus models of gene conversion and transposition opposed by selection (HICKEY1982; LAMBand HELMI 1982). None of these models specifically addresses the consequences of molecular turnover processes for a single-locus neutral theory. Here, we examine the implications of biased gene conversion acting on a single diallelic locus in a finite population, with special attention to the effects of biased gene conversion on selectively neutral alelles. We also examine the

’ Present address: Department of Biophysics and Theoretical Biology, The University of Chicago, 920 East 58th Street, Chicago, Illinois 60637.

Genetics 105: 461-468 October, 1983. 462 J. B. WALSH role of biased gene conversion in speciation events and investigate under what conditions bias in conversion can be important in the evolution of multigene families and yet be unimportant in the evolution of unique single loci which experience the same levels of biased conversion. The first section reviews molecular models of conversion bias. In the second, we use standard diffusion results to determine the fixation probabilities and substitution rates of selec- tively neutral alleles with a conversion bias. The third section examines under what conditions bias in conversion is important at the genomic level (changing the composition of multigene families) and yet unimportant at the population level (not altering the substitution dynamics of alleles at unique single loci). The fourth section deals with the speciation implications of single-locus biased conversion acting independently at many loci as a model system of speciation by molecular turnover mechanisms.

MOLECULAR MODELS OF BIASED CONVERSION Gene conversion is the nonreciprocal transfer of genetic information from one variant to another and usually results in a departure from normal Men- delian segregation. Conversion is said to be biased if one variant is preferen- tially produced over another during conversion events. Existing molecular models of conversion imply that conversion bias is not at all unexpected, and data from fungal systems (LAMBand HELMI1982; NAGYLAKIand PETES 1982) support this view. Current models of gene conversion comprise two classes: single strand break models and double strand break models. Single strand break models were introduced by HOLLIDAY(1 964) with subsequent refine- ments by MESELSON and RADDING(1975). Conversion is initiated by a single strand break in one of the DNA duplexes of an interacting chromatid pair. This single strand invades the other duplex, forming a region of heteroduplex DNA through strand displacement. Conversion occurs when mismatch repair of the heteroduplex DNA results in a change of the allelic composition of the chromatid pair. Bias in conversion results if one strand of the heteroduplex DNA preferentially serves as the template for mismatch repair. Recent exper- imental results (SAVAGand HASTINGS1981; FOCELet al. 1978) suggest that, for single strand break models, mismatch repair must use the invading strand for the template. Thus, an unequal frequency of single strand breaks and invasions between two different variants can result in conversion bias. Double strand break conversion models (SZOSTAKet al. 1983) proceed from a double strand break in one of the DNA duplexes of a chromatid pair which is subsequently enlarged into a double strand gap by exonuclease action. Gap repair occurs by a double strand transfer of information from the other duplex without the direct involvement of heteroduplex DNA formation. In this model, biased conversion results from unequal formation of double strand nicks on chromatids carrying different alleles. It is interesting to note that in the single strand break model sequences that are preferentially nicked are at a conversion advantage, whereas such sequences are at a conversion disadvantage in double strand break models. BIASED GENE CONVERSION 463

FIXATION PROBABILITIES AND SUBSTITUTION RATES Theoretical models of biased gene conversion acting at a single locus are formally equivalent to meiotic drive (GUTZand LESILE1976; LAMBand HELM 1982), because both result in departures from normal 1:l Mendelian segre- gation in heterozygotes. For a diallelic system with alleles A and a, we can define the drive strength, d = 2k- 1, where the segregation ratio of A:a is k:l - k from Aa heterozygotes. d = 0 if normal 1:l segregation occurs (e.g., k = lh); otherwise, d ranges from 1 (only A-bearing gametes are produced by heterozygotes) to -1 (only a-bearing gametes are produced). To compute d for gene conversion, let be the probability of an unequal conversion event and /3 be the conditional probability that allele a is converted to A given that an unequal conversion event occurs. Hence, (1 - y) of the segregation events produce equal numbers of A- and a-bearing gametes, whereas y/3 of the seg- regation events produce only A-bearing gametes, implying k = (1 - y)(l/z) + y/3 and d = y(28 - 1). If the gene conversion is unbiased (i.e., neither allele is favored in an unequal conversion event) /3 = '1'2 and d = 0, independent of the frequency of unequal conversion y. The most general one-locus diallelic model of biased gene conversion as- sumes that the genotypes AA:Aa:aa have fitnesses 1: 1 + h: 1 + s. Provided that d, h and s are small enough to ignore terms of second and higher order, an allele A with the above fitnesses which has a conversion bias d behaves dynam- ically like an allele with no conversion bias and fitnesses 1 + 2d:l + d + h:l + s (GUTZand LESILE 1976; WALSH 1982; NACYLAKI1983a). It immediately follows that a selectively neutral (h = s = 0) allele with a conversion bias is equivalent to an unbiased allele with additive fitness d. Unless otherwise stated, we restrict our attention to such alleles in one-locus dialleic models for the remainder of this paper. Standard results from the theory for additive selection apply to our problem using this equivalence. For an infinite, randomly mating population, an allele at a conversion disadvantage (d < 0) is lost, whereas an allele at a conversion advantage (d > 0) is fixed. The dynamics of multiallelic systems in infinite populations are considerably more complex (NAGYLAKI1983b). Focusing on finite populations with random mating, KIMURA'S (1957) clas- sical results for loci with additive fitnesses can be used to obtain fixation probabilities and substitution rates for the diallelic case with no selection. NA- GYLAKI (1 983a) investigates the more general case of multiple alleles with conversion bias, selection and , obtaining the diffusion equation this processes satisfies. Equilibrium properties of such multiple allele systems can be examined by the methods employed by LI (1978), but, here, we wish to focus only on the properties of a simple diallelic locus. Given that Ne, the variance effective population size, is large and Id I is small, the probability (U[p]) that allele A is fixed given an initial frequency p is qp]x (1 - e-4fiN9/(1- (1) We are particularly interested in v[ 1/(2Na)], the probability of fixation given 464 J. B. WALSH that A began as a single copy in a population of iV, diploids. Since I d I << 1, and 1/(2N,) << 1, equation (1) gives (KIMURA1964) U[1 @No)] = 1/(2No) for 4NeI d I << 1 (24 U[1 /( 2N,)] = 2dN,/N, for 4Ned >> 1 (2b) U[1/(2N,)] = (-2d~V,/N~)e~"*~for 4N,d << -1 (2c) For a strictly neutral allele (no conversion bias or selection), U[1/(2N,)] = 1/ (2N,), implying that conversion is overcome by drift when 4N,I d I << 1. Re- marks by DOVER(1982) and LAMBand HELMI(1982) that fixation by biased gene conversion proceeds as easily in small as in large populations are, there- fore, incorrect, at least in the context of conversion acting on a single locus. However, when 4Ne1d I >> 1, bias in conversion overcomes and can be a significant evolutionary force. In these cases newly arising mutants at a conversion disadvantage (d < 0) are fixed only very rarely compared with strictly neutral alleles, whereas alleles at a conversion advantage (d > 0) have a much higher relative chance of fixation compared to strictly neutral alleles. What little information is available on (dI suggests that some populations are large enough for biased conversion to be a significant evolutionary force. Fungal systems are the best characterized, with a mean value of Id 1 ranging from 3 x to 4 x lo-* in a variety of species (LAMBand HELMI1982). Preliminary evidence from corn (NELSON1975) and Drosophila (HILLIKERand CHOVNICK1981) suggests that I d I may be considerably lower in higher eu- karyotes (I d] < Taking 4 X as an estimate of the average value of Id I in implies that effective population sizes in excess of lo4 are required for biased gene conversion to be an important evolutionary force when acting on a single locus. If the average value of Id I in eukaryotes is considerably smaller than 10-4, larger effective population sizes are required and vice versa. Intimately coupled with fixation probabilities is the per generation substitu- tion rate R of new alleles at a specific locus. Let p be the per generation mutation rate of the type allele to new alleles at the locus in question. We assume that all mutants have the same conversion parameter d relative to the type allele and further assume that, if biased gene conversion occurs between mutant alleles, it is sufficiently small relative to d to be ignored. Under these assumptions, the probability of fixation of each new mutant is the same and is given by U[1 /(21Vn)]. Since 2N0p new mutants arise on average each generation, R (2Nnp)U[1/(21Vc,)](KIMURA and OHTA 1971). R measures the rate of allelic substitutions, and given that new alleles are often created by mutational events which involves more than single nucleotide changes, it is very difficult to accurately determine R from comparative sequence data (see NEI 1975, pp. 101-102). R = p for strictly neutral alleles (KIMURA1968; GUESSand EWENS 1972), so from (2) it follows that (1) if d >> 1/(4Ne), gene conversion increases the substitution rate compared with strictly neutral alleles; (2) if d << -1/(4Ne), conversion exponentially decreases the substitution rate; and (3) if - 1/(4N,) < d 1/(41V?),conversion does not alter the substitution rate appreciably. At BIASED GENE CONVERSION 465 those loci that experience gene conversion, these substitution rates suggest that existing selectively neutral alleles most likely have undergone several rounds of conversion. These surviving alleles will have been “selected” to be at a conversion advantage relative against a wide spectrum of competing alleles. NAGYLAKIand PETES (1 982), addressing the role of intrachromosomal gene conversion in the homogenization of gene families, come to a similar conclu- sion regarding existing repeats in a multigene family. Thus, for selectively neutral alleles at loci experiencing gene conversion we expect most newly arising mutants to be at a conversion disadvantage relative to the standard allele, potentially resulting in a greatly decreased substitution rate at these loci compared with predictions from strict neutral theory. Given the equivalence between additive selection and biased conversion most newly arising mutants at these loci have the same properties as slightly deleterious alleles. OHTA(1976) reviews the role of slightly deleterious alleles in , demonstrating that the presence of such alleles can account for the excess of rare alleles seen in some populations over predictions from strict neutral theory. Biased gene conversion can, therefore, also account for these deviations from strict neutral theory without having to invoke selection.

EFFECTIVENESS OF BIAS AT THE GENOMIC AND POPULATIONAL LEVELS In the previous section we examined the consequences of biased gene con- version acting at single loci. This is “classical” gene conversion, conversion between different alleles at the same locus on homologous (we will refer to this as interchromosomal gene conversion). More generally, con- version can occur between different loci, which has very important conse- quences for genome evolution. One scheme for such conversions is intra- chromosomal gene conversion, in which the conversion occurs between different loci on the same . OHTA(1977) and NAGYLAKIand PETES(1982) showed that intrachromosomal conversion results in the homogenization of repeated gene families. Thus, we can view conversion as operating on two different evolutionary levels: populational and genomic. The populational level refers to the effects of conversion at single unique loci (i.e., not members of repeated gene families), and as was shown bias in conversion can have an important effect at the populational level. The genomic level refers to the organization and composition of repeated sequences within the genome. Unlike at the populational level, both biased and unbiased conversion are potentially important at the genomic level, with bias being particularly important in gene families with a large number of members (NAGYLAKIand PETES 1982). Although the rates of interchromosomal and intrachromosomal conversion are likely to be very different, the amount of bias given a conversion event occurs conceivably is quite similar. Therefore, we wish to ask whether it is possible for bias in conversion to be important at the genomic level and yet be unimportant at the populational level. That is, are substitution rates at certain loci constrained when bias is important at the genomic level? Equations 13 and 17 in NAGYLAKIand PETES (1982) give a sufficient con- 466 J. B. WALSH dition for biased conversion to behave significantly different from unbiased conversion in gene family evolution provided small amounts of bias (@close, but not equal, to Vz). This condition can be expressed as GI 2P - 1 I/@ >> 1, where g is the number of members of the gene family and /3 (as defined earlier) is our measure of bias. If this condition is satisfied, those variants at a conver- sion advantage (P > l/z) have a much higher relative probability of fixation compared with unbiased conversion, and the appearance of such variants can result in a change in the composition of a gene family. Those newly arising variants that are at a conversion disadvantage (/3 C Y2), however, have an extremely low relative probability of fixation compared with unbiased conver- sion, and in these cases the effect of the bias is to conserve the existing repeats in the family. The rate of intrachromosomal gene conversion does not enter into the condition because NACYLAKIand PETESfollow only a single chromo- somal lineage, rescaling time so that one conversion event occurs each gener- ation. Although this rescaling effects the time to fixation, it does not affect the probability of fixation, and, hence, this condition is independent of the actual rate of intrachromosomal conversion, provided that it is positive. Combining our conditions for drift overcoming conversion at a single locus (4N,yl2/3 - 1 I < 1) with NACYLAKIand PETES’ condition gives G >> PI1 2P - 1 I > 4NeyP (3) as a sufficient condition that allows for biased conversion to greatly alter fix- ation probabilities in gene families (compared with unbiased conversion) and still be ineffective at the population level (for those individual loci with the same conversion bias parameter P). A necessary condition for equation (3) to be satisfied is that the number of members in the gene family must be suffi- ciently large (G >> 4Ney/3). Values for y from fungal systems (LAMBand HELMI 1982) are approximately lo-‘, so if Ne = 1000 and /3 close (but not equal) to 0.5, G >> 20. For these parameters, if intrachromosomal conversion bias is important in gene families with fewer than 20 members, then we expect in- terchromosomal gene conversion to reduce the substitution rate of selectively neutral alleles at those other loci that have the same conversion bias 8. Like- wise, if the number of genes in the gene family is sufficiently large, very slight amounts of bias which would be unimportant at the populational level are nevertheless quite important at the genomic level.

BIASED CONVERSION, MOLECULAR DRIVE AND SPECIATION DOVER(1 982) has proposed that molecular turnover mechanisms (which he terms “molecular drive”) can potentially result in rapid speciation between geographically isolated populations. Gene conversion is an example of such a turnover mechanism. In light of recent criticisms of molecular drive (ROSEand DOOLITTLE1983), it is important to consider how molecular turnover effects the rates of speciation in particular population genetic models of . One scheme allowing for accumulation of reproductive isolation between isolated populations is the fixation of different underdominant alleles at many loci. What effect, if any, does biased gene conversion acting inde- BIASED GENE CONVERSION 467 pendently at many single loci have on the speciation rate of isolated populations if we assume speciation by fixation of underdominant alleles? WALSH (1 982) examined speciation rates in finite populations, assuming fixation of underdom- inant chromosome rearrangements, and addressed the consequences of meiotic drive. If d is the drive strength, and the fitnesses of genotypes AA:Aa:aa are 1:l - h:l, then, provided that Nd >> 1, genes with h < d are likely to become fixed. If NJ 1, genes with h > 1.3/Neare very unlikely to become fixed. Thus, meiotic drive increases the speciation rate by allowing fixation of genes that would not be fixed without drive (viz., those genes for which 1.3/Ne< h C d). Likewise, gene conversion can potentially increase speciation rates but only in populations with fairly large effective population sizes, e.g., N, >> l/d. Even in these populations, reproductive isolation accumulates very slowly be- cause only those underdominant alleles with h C d are likely to become fixed, and since d is expected to be small many fixation events are required for even a small increase in reproductive isolation. Furthermore, we would expect the occurrence of underdominant alleles with a conversion advantage to be very infrequent. Finally, as we have shown, gene conversion is not a powerful force in small populations, but there is widespread conviction that population bottlenecks are critical to species formation (CARSON1975). If such founding events are important in speciation, gene conversion plays little role in this type of speciation process. Therefore, it seems unlikely that gene con- version acting independently at many single loci could appreciably increase speciation rates, except in very unusual circumstances. ROSE and DOOLITTLE (1 983) suggest on biological grounds that gene conversion acting to homoge- nize a multigene family is also unlikely to be effective in speciation events. Thus, it appears from the present state of knowledge of genome organization that gene conversion does not play an important role in changing the rate at which allopatric populations speciate.

I thank L. SANDLER,C. LAIRD,M. SLATKINand T. OHTAfor useful discussions. I especially wish to thank TOMNAGYLAKI for helpful comments and communications. This research was supported in part by a National Science Foundation predoctoral fellowship, by National Institutes of Health training grant GM07748, and in part by test agreement DE-AT06-76EV71005 of contract DE-AM06-76L02225 between the United States Department of Energy and the University of Washington. This paper is dedicated to HERSCHELROMAN, a pioneer in the study of gene conversion, for his outstanding commitment to graduate education. \ LITERATURE CITED

CARSON,H. L., 1975 The genetics of speciation at the diploid level. Am. Nat. 109 83-92. DOOLITTLE,W. F. and C. SAPIENZA,1980 Selfish genes, the phenotype paradigm and genome evolution. Nature 284: 601-603. DOVER,G., 1982 Molecular drive: a cohesive mode of species evolution. Nature 299: 11 1-1 17. FOGEL,S., R. K. MORTIMER,K. LUSNAKand F. TAVARES,1978 Meiotic gene conversion: a signal of the basic recombination event in yeast. Cold Spring Harbor Symp. Quant. Biol. 43: 1325- 1341. GUM, H. A. and W. J. EWENS,1972 Theoretical and simulation results relating to the neutral allele theory. Theor. Pop. Biol. 3: 434-447. 4 68 J. B. WALSH

GUTZ, H. and H. F. LESILE, 1976 Gene conversion: a hitherto overlooked parameter in popula- tion genetics. Genetics 83: 861-866. HICKEY,D. H., 1982 Selfish DNA: a sexually transmitted nuclear parasite. Genetics 101: 519- 531. HILLIKER,A. J. and A. CHOVNICK,1981 Further observations on intragenic recombination in Drosophila melanogaster. Genet. Res. 38: 28 1-296. HOLLIDAY,R., 1964 A mechanism for some conversion in fungi. Genet. Res. 5: 282-304. KIMURA,M., 1957 Some problems of stochastic processes in genetics. Ann. Math. Stat. 28: 882- 901. KIMURA,M., 1964 Diffusion models in population genetics. J. Appl. Prob. 1: 177-232. KIMURA,M., 1968 Evolutionary rate at the molecular level. Nature 217: 624-626 KIMURA,M. and T. OHTA,1971 On the rate of molecular evolution. J. Mol. Evol. 1: 1-17. LAMB,B. C. and S. HELMI, 1982 The extent to which gene conversion can change allele fre- quencies in populations. Genet. Res. 39: 199-217. LI, W.-H., 1978 Maintenance of genetic variability under the joint effect of mutation, selection, and random drift. Genetics 90: 349-382. MESELSON,M. S. and C. M. RADDING,1975 A general model for genetic recombination. Proc. Natl. Acad. Sci. USA 72: 358-361. NAGYLAKI,T., 1983a Evolution of a finite population under gene conversion. Proc. Natl. Acad. Sci. USA In press. NAGYLAKI,T., 1983b Evolution of a large population under gene conversion. Proc. Natl. Acad. Sci. USA In press. NAGYLAKI,T. and T. D. PETES, 1982 lntrachromosomal gene conversion and the maintance of sequence homogeneity among repeated genes. Genetics 100 3 15-337. NEI, M., 1975 Molecular Population Genetics and Evolution. American Elsevier, New York. NELSON,0. E., 1975 The Waxy locus in maize. 111. Effect of structural heterozygosity on intra- genic recombination and flanking marker assortment. Genetics 79: 3 1-44. OHTA,T., 1976 Role of very slightly deleterious in molecular evolution and poly- morphism. Theor. Pop. Biol. 10: 254-275. OHTA, T., 1977 On the gene conversion model as a mechanism of homogeneity in systems with multiple genomes. Genet. Res. 30 89-91. OHTA,T., 1981 Evolution and Variation in Multigene Families. Springer-Verlag, Berlin. OHTA,T., 1983 Theoretical study on the accumulation of selfish DNA. Genet. Res. 41: 1-15. OHTA,T. and M. KIMURA,1981 Some calculations on the amount of selfish DNA. Proc. Natl. Acad. Sci. USA 78: 1129-1 132. ORCEL,L., and F. H. C. CRICK,1980 Selfish DNA: the ultimate parasite. Nature 284: 604-607. ROSE, R. M. and F. W. DOOLITTLE,1983 Molecular biological mechanisms of speciation. Science 220: 157-162. SAVAGE,E. A. and P. J. HASTINGS,1981 Marker effects and the nature of the recombination event at the his1 locus of Saccharoinjces cerevisiae. Curr. Genet. 3: 37-47. SZOSTAK,J., T. L. ORR-WEAVER,R. J. ROTHSTEINand F. W. STAHL, 1983 The double-strand- break repair model for recombination. Cell 33: 25-35. WALSH,J. B., 1982 Rate of accumulation of reproductive isolation by chromosome rearrange- ments. Am. Nat. 120 510-532.

Corresponding editor: M. NEI