sign in sign up
<<
Home , Allopatric speciation, Drosophila mojavensis, Drosophila persimilis, Sympatry

The evolutionary of

Jerry A. Coyne1 and H. Allen Orr2 1Department of and , University of Chicago, 1101 E. 57th Street, Chicago, IL 60637, USA 2Department of , University of Rochester, Rochester, NY 11794, USA

The last decade has brought renewed interest in the genetics of speciation, yielding a number of new models and empirical results. De¢ning speciation as `the origin of between two taxa', we review recent theoretical studies and relevant data, emphasizing the regular patterns seen among genetic analyses. Finally, we point out some important and tractable questions about speciation that have been neglected. Keywords: inviability; hybrid sterility; reproductive isolation; sexual isolation;

evolutionists, and includes those acting before fertilization 1. INTRODUCTION (`prezygotic' mechanisms, such as mate discrimination When we last reviewed the evolution of reproductive isola- and gametic incompatibility), and after fertilization tion (Coyne & Orr 1989a), we complained that workers on (`postzygotic' mechanisms, including hybrid inviability speciation were considered poor cousins in the family of and sterility). evolutionists, mired in endless and untestable speculations Like most evolutionists, we adopt the BSC as the most about a process that no one could witness. Since then, the useful , and our discussion of the genetics study of speciation has grown increasingly respectable, of speciation will accordingly be limited to the genetics of recruiting ever more experimentalists and theorists. A reproductive isolation. We recognize that this view of number of new phenomena have been uncovered and speciation is not universal: systematists in particular new theories o¡ered to explain them. Here we summarize often reject the BSC in favour of concepts involving recent progress on the genetics of speciation, highlighting diagnostic characters (Cracraft 1989; Baum & Shaw areas where important and tractable questions remain 1995; Zink & McKitrick 1995). We have argued against unanswered. these concepts elsewhere (Coyne et al. 1988; Coyne 1992a, 1993a, 1994) and will not repeat our contentions here. We note only that the recent burst of work on 2. WHAT ARE SPECIES? speciation re£ects almost entirely the e¡orts of those Any discussion of the genetics of speciation must begin adhering to the BSC. In fact, every recent study on the with the observation that species are real entities in `genetics of speciation' is an analysis of reproductive nature, not subjective human divisions of what is really a isolation. continuum among . We have previously summarized the evidence for this view and counter- arguments by dissenters (Coyne 1994). The strongest 3. WHY ARE THERE SPECIES? evidence for the reality of species is the existence of One of the most important but neglected questions distinct groups living in (separated by genetic about speciation is why organisms fall into many discrete and phenotypic gaps) that are recognized consistently by groups instead of constituting a few extremely variable independent observers. To a geneticist, these disjunct `types'. The answer to the question of why species exist groups suggest a species concept based on gene £ow. As may not be the same as the answer to `how do species Dobzhansky (1935, p. 281) noted: `Any discussion of these arise?' The only coherent discussion of this problem is problems should have as its logical starting point a that of Maynard Smith & Szathma¨ ry (1995, pp. 163^ consideration of the fact that no discrete groups of - 167) who give three possible reasons for the existence of isms di¡ering in more than a single gene can maintain discrete species: (i) species might represent stable, discon- their identity unless they are prevented from inter- tinuous states of matter; (ii) species might be adapted to breeding with other groups. Hence, the existence of discontinuous ecological niches; and (iii) reproductive discrete groups of any size constitutes evidence that some isolation (which can arise only in sexual taxa) might mechanisms prevent their interbreeding, and thus isolate create gaps between taxa by allowing them to evolve them.' This conclusion inspired Dobzhansky (1935) and independently. Mayr's (1942) biological species concept (BSC), which The second and third hypotheses seem most plausible. considered species to be groups of repro- There are several ways to distinguish between them. ductively isolated from other such groups by `isolating One is to determine if asexually reproducing groups form mechanisms', genetically based traits that prevent gene sympatric taxa just as distinct as do sexually reproducing exchange. The list of such mechanisms is familiar to all groups. Although such comparisons are hampered by the

Phil. Trans. R. Soc. Lond. B (1998) 353, 287^305 287 & 1998 The Royal Society 288 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation rarity of large asexual groups, bacteria seem reasonable 4. STUDYING REPRODUCTIVE ISOLATION candidates. Although little work has been published, two studies (Roberts & Cohan 1995; Roberts et al. 1996) show (a) What is novel about speciation? that forms of Bacillus subtilis from the American desert fall Some of our colleagues have suggested that speciation is into discrete clusters in sympatry. Moreover, coalescence not a distinct ¢eld of study becauseöas a by-product of models (Cohan 1998) show that a combination of new conventional evolutionary forces like selection and conferring ecological di¡erence, periodic selec- driftöthe origin of species is simply an epiphenomenon tion and limited gene £ow can produce distinct taxa of of normal -genetic processes. But even if specia- bacteria living sympatrically. tion is an epiphenomenon, it does not follow that the The only empirical work on clustering in asexual eukar- mathematics or genetics of speciation can be inferred yotes is that of Holman (1987), who, studying successive from traditional models of evolution in single lineages. revisions of taxonomic monographs, determined that the Under the BSC, the origin of species involves reproductive nomenclature of bdelloid rotifers (not known to have a isolation, a character that is unique because it requires the sexual phase) was more stable than that of sexually repro- joint consideration of two species, and usually an interac- ducing relatives. From this he concluded that because they tion between the of two species (Coyne 1994). are recognized more consistently, asexual rotifers are actu- The distinctive feature of the genetics of speciation is, ally more distinct than their sexual counterparts. While therefore, epistasis. This is necessarily true for all forms of intriguing, this result is hardly conclusive, and we badly postzygotic isolation, in which an that yields a need similar studies based not on nomenclature but on normal in its own species causes hybrid invia- genetic and phenotypic cluster analyses. We hasten to add bility or sterility on the genetic background of another (see that although studies of clustering in asexuals are worth- below). Epistasis also occurs in many forms of prezygotic while, they must not be overinterpreted. Such studies isolation. Sexual isolation, for example, usually requires might show that ecological specialization can produce the of male traits and female preferences, so discrete asexual forms, but it does not follow that such that the ¢tness of a male trait depends on whether the specialization explains clustering in sexuals:the mainte- choosing female is conspeci¢c or heterospeci¢c. nance of discrete forms by in sexuals is These complex interactions between the genomes of two far more di¤cult and could be far rarer. species guarantee that the mathematics of speciation will We suggest two other approaches not discussed by di¡er from that describing evolutionary change within Maynard Smith & Szathma¨ ry (1995). First, if distinct species, and suggest that speciation may show emergent niches alone (and not reproductive isolation) can explain properties not seen in traditional models. Indeed, such the existence of species, then should properties have already been identi¢ed for postzygotic be common. The whole premise that geographic isolation isolation (e.g. see discussion of the `snowball e¡ect' below). is essential for speciation rests on Dobzhansky's (1935) Two motives usually underlie genetic analyses of specia- and Mayr's (1942) idea that the swamping e¡ect of gene tion. First, just as with quantitative-trait- (QTL) £ow prevents the evolution of reproductive isolation. But analyses of `ordinary' characters, we would like to under- if reproductive isolation is not important in explaining stand the genetic basis of . That is, we would the existence of species, to distinct niches like to know the number of genes involved in reproductive could occur in sympatry, and phylogenetic analyses isolation, the distribution of their phenotypic e¡ects and should often reveal that the most recently evolved pairs their location in the . Second, we expect genetic of species are sympatric. Although there is some evidence analyses of reproductive isolation to shed light on the for sympatric speciation based on niche use in ¢sh (e.g. process of speciation, as di¡erent evolutionary processes Schliewen et al. 1994), we know of no other studies should leave di¡erent genetic signatures. The observation featuring similarly rigorous phylogenetic analyses. Pertur- of more genes causing hybrid male than hybrid female bation experiments could also address this problem. For sterility (see below) has suggested, for example, that these example, sexual isolation between sympatric species critical substitutions were driven by (Wu could be overcome by hybridization in the laboratory, & Davis 1993; True et al. 1996). and the resulting (fertile) hybrids reintroduced into their original . If the `ecological-niche' explanation is (b) What traits should we study? correct, the hybrids should revert to the parental types Because speciation is complete when reproductive isola- (or to similar, but distinct, types). But if reproductive tion stops gene £ow in sympatry, the `genetics of isolation helps maintain species distinctness, the hybrids speciation' properly involves the study of only those should either revert to a single parental type or remain isolating mechanisms evolving up to that moment. The a hybrid swarm. Such a study would, however, probably further evolution of reproductive isolation, although inter- require an unrealistic amount of time. esting, is irrelevant to speciation. This point is widely Of course, the existence of species in sexual taxa could recognized but understandably often ignored in practice. well depend on both distinct adaptive peaks and reproduc- If speciation is allopatric, and several isolating mechan- tive isolation. But it is hard to believe that ecological niches isms evolve simultaneously, it is hard to know which will alone can explain distinct species, if for no other reason be important in preventing gene £ow when the taxa than such species, lacking reproductive isolating mechan- become sympatric. simulans and D. mauritiana, isms, would hybridize. If they were then to remain for example, are allopatric, and in the laboratory show distinct, hybrids would have to su¡er a ¢tness disadvantage sexual isolation, sterility of F1 hybrid males, and inviability due to their inability to ¢nd a suitable niche, and this disad- of both male and female backcross hybrids. We have no vantage is a form of postzygotic isolation. idea which of these factors would be most important in

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 289

1

0.8

0.6

0.4

Prezygotic Isolation 0.2 (a) 0

1 Figure 2. Total reproductive isolation in Drosophila 0.8 (combining both pre- and postzygotic isolation) plotted against Nei's (1972) genetic distance. Each point represents a single pair of species. See Coyne & Orr (1989b, 1997) for further 0.6 details.

0.4 comparing the rates of evolution of di¡erent forms of

Postzygotic Isolation 0.2 reproductive isolation is Coyne & Orr's (1989b, 1997) analysis of pre- versus postzygotic isolation in Drosophila. (b) In these studies, prezygotic isolation (mate discrimination) between species is on average a stronger barrier to gene 0.5 1 1.5 2 0 £ow than is postzygotic isolation (¢gure 1). This disparity, Genetic Distance (D) however, is due entirely to much faster evolution of sexual Figure 1. Strength of reproductive isolation in Drosophila isolation in sympatric than allopatric species pairs, plotted against Nei's (1972) electrophoretic genetic distance suggestingöas we discuss belowöthe possibility of direct (an index of divergence time). Each point represents the selection for sexual isolation in sympatry. Among allopa- average among pairs within a species group, so points are tric taxa, pre- and postzygotic isolation arise at similar evolutionarily independent. (a) Prezygotic (sexual) isolation; rates. When both pre- and postzygotic forms of reproduc- (b) postzygotic isolation. See Coyne & Orr (1989b; 1997) for tive isolation are considered simultaneously, the total further details. strength of reproductive isolation increases quickly with time (¢gure 2). We are unaware of any analogous data on preventing gene exchange in sympatry, or if other unstu- the rate at which total reproductive isolation increases in died factorsösuch as ecological di¡erencesöwould play other taxa. It would certainly be worth obtaining such a role. information, as it would allow one to see if the `speciation It seems likely, in fact, that several isolating mechanisms clock' ticks at the same rate in di¡erent groups. evolve simultaneously in allopatry and act together to In other taxa, the data are far more impressionistic. In both prevent gene £ow in sympatry and allow coexistence. many groups of plants, such as orchids and the genus (Although reproductive isolation is su¤cient for specia- Mimulus, temporal or pollinator isolation sometimes tion, di¡erent species must coexist in sympatry in order evolve faster than postzygotic isolation, because related to be seen.) There are two reasons why multiple isolating species produce fertile o¡spring when forcibly crossed in mechanisms seem likely. In theory, no single isolating the greenhouse but fail to hybridize in nature (Grant mechanism except for distinct ecological niches or some 1981). Students of bird evolution have noted that prezy- types of temporal divergence can at the same time gotic isolation often seems to evolve well before hybrid completely prevent gene £ow and allow coexistence in sterility and inviability (Prager & Wilson 1975; Grant & sympatry. Two species solely isolated by hybrid sterility, Grant 1996). for example, cannot coexist: one will become extinct But these conclusions must be seen as preliminary. We through excessive hybridization or ecological . require additional and more systematic studies in which Species subject only to sexual isolation are ecologically di¡erent forms of reproductive isolation are assessed unstable because they occupy identical niches. Second, among pairs of species that diverged at about the same direct observation often shows that complete reproductive time. Such work is especially practicable in plants, as isolation in nature often involves several isolating mechan- ecological di¡erences can be studied in the greenhouse, isms. Schluter (1997), for example, describes several species pollinator isolation can be studied in situ, and postzygotic pairs having incomplete prezygotic isolation. When isolation can be studied through forced crossing. hybrids are formed, however, they are ecologically unsuited for the parental , and do not thrive. (c) A summary of genetic studies We know little about the temporal order in which repro- Because there are relatively few studies of the genetics of ductive isolating mechanisms appear. The only study speciation, we have summarized them all in table 1. We

Phil. Trans. R. Soc. Lond. B (1998) 290 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation

Table 1. Summary of existing genetic analyses of reproductive isolation between closely related species (The reference to each study includes a summary of the results, emphasizing any interesting features. The text gives our criteria for including studies in this table.)

reference species pair trait number of genes (see Appendix 1)

Drosophila heteroneura^D. silvestris head shape 9 1 ^D. simulans hybrid inviability 59 2 female pheromones 55 3 Drosophila mauritiana^D. simulans hybrid male sterility 515 4 hybrid female sterility 54 5 hybrid inviability 55 5 male sexual isolation 52 6 female sexual isolation 53 6 genital morphology 59 7 shortened copulation 53 8 Drosophila mauritiana^D. sechellia female pheromones 56 9 ^D. sechellia hybrid male sterility 56 10 hybrid inviability 52 11 female sexual isolation 52 12 Drosophila mojavensis^D. arizonae hybrid male sterility 53 13 male sexual isolation 52 14 female sexual isolation 52 14 ^D. persimilis hybrid male sterility 59 15 hybrid female sterility 53 15 sexual isolation 53 16 Drosophila pseudoobscura U.S.A.^Bogota hybrid male sterility 55 17 Drosophila buzatti^D. koepferae hybrid male inviability 54 18 hybrid male sterility 57 19 Drosophila subobscura^D. madeirensis hybrid male sterility 56 20 Drosophila virilis^D. littoralis hybrid female viability 55 21 Drosophila virilis^D. lummei male courtship song 54 22 hybrid male sterility 56 23 Drosophila hydei^D. neohydei hybrid male sterility 55 24 hybrid female sterility 52 24 hybrid inviability 54 25 Drosophila montana^D. texana hybrid female inviability 52 26 Drosophila virilis^D. texana hybrid male sterility 53 27 Drosophila auraria^D. biauraria male courtship song 52 28 Ostrina nubialis, Z and E races female pheromones 1 29 male perception of pheromones 2 29 Laupala paranigra^L. kohalensis song pulse rate 58 30 Spodoptera latifascia^S. descoinsi pheromone blend 1 31 Xiphophorus helleri^X. maculatus hybrid inviability 2 32 Mimulus lewisii^M. cardinalis 8 £oral traits (see note) 33 Mimulus guttatus^M. micranthus bud growth rate 8 34 duration of bud development 10 34 Mimulus, four taxa (see note) (see note) 35 Mimulus guttatus pops. hybrid inviability 2 (system 1) 36 52 (system 2) 36 Mimulus guttatus^M. cupriphilus £ower size 3^7 37 Helianthus annuus^H. petiolarus pollen viability 514 38

used two criteria for including a study in this table. First, indicated that an isolating mechanism was due to changes the character studied must be known to cause reproductive at a single locus; or (iii) biometric analyses, in which isolation between species in either nature or the labora- measurement of character means and variances in back-

tory, or be plausibly involved in such isolation. Second, crosses or F2s yielded a rough estimate of gene number. the genetic analysis must have been fairly rigorous, using Table 1 also gives the actual or minimum number of genes one of three methods: (i) classical-genetic analyses, in involved for each isolating mechanism. The notes in which species di¡ering in molecular or visible mutant Appendix 1 give more detail about the type of genetic markers are crossed, and the segregation of reproductive analysis, whether the species pair was sympatric or allopa- isolation with the markers examined. Here we included tric, and a brief summary and critique of the results. only those studies in which markers were distributed We must add several caveats. First, despite our attempts among all major ; (ii) simple Mendelian to comprehensively comb the literature, we have surely

analyses in which segregation ratios in backcrosses or F2s missed some studies. Second, the quality of the analyses is

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 291 uneven: some classical-genetic studies involved detailed Nonetheless, consideration of the role of sexual selection mapping experiments with many markers, whereas others in speciation was remarkably late in coming. The earliest relied on only one marker per . Third, most paper even mentioning such a possibility appears to be studies have underestimated the number of genes causing that of Haskins & Haskins (1949), who posited that a single reproductive isolating mechanism. This may sexual selection in guppies might lead to female recogni- re£ect a limited number of markers, failure to test all tion of only conspeci¢c males as mates, which would possible interactions between chromosomes, or the use of serve as an isolating mechanism among sympatric biometric approaches, which nearly always underestimate species. This line of thought then vanished from the litera- true gene number. Finally, data are given for single ture until Nei (1976) made a model of sexual selection in isolating mechanisms, but several mechanisms may often which one gene controlled the male trait and another the operate together to impede gene £ow in nature. female preference. Exploring the conditions that could The most striking feature of table 1 is the imbalance of lead to the joint evolution of these traits, Nei noted that both species and isolating mechanisms. Roughly 75% of such a process could cause reproductive isolation. More all the studies involve Drosophila, with only seven other recently, Ringo (1977) proposed that sexual selection pairs of taxa, mostly plants from the genus Mimulus. More- might explain both the and the strong over, nearly two-thirds of the Drosophila work is on hybrid sexual dimorphism of Hawaiian Drosophila. sterility and inviability. There is no published genetic study The most extensive theoretical work on this problem is of ecological isolation (but see below). Obviously we must that of Lande (1981, 1982). In 1981, he showed that sexual extend such studies to other groups and other forms of isolation could be the by-product of sexual selection if reproductive isolation. random in small populations triggered the For convenience, we discuss pre- and postzygotic isola- `runaway' process suggested by Fisher (1930). Later tion separately. workers (e.g. Kirkpatrick 1996) have shown, however, that such instability is unlikely if natural selection acts on female preference. Lande's 1982 model, an explicit quanti- 5. PREZYGOTIC ISOLATION tative-genetic treatment of clinal speciation via sexual Although there are many forms of prezygotic isolation, selection, is more realistic. Here he showed that adaptive including di¡erences in ecology, behaviour, time of repro- geographic di¡erentiation in male traits could be ampli- duction, gametic incompatibility and (in plants) ¢ed by the evolution of female preferences, resulting in di¡erences in pollinators, only one formösexual isola- reproductively isolated populations along a . Data tionöhas been the subject of much theory and that may support this model come from the guppy Poecilia experiment. reticulata, in which local populations have di¡erentiated so that females prefer to mate with local rather than foreign (a) Sexual isolation males (Endler & Houde 1995). Many evolutionists have noted that closely-related There is only one theoretical analysis of sympatric species, particularly those involved in adaptive speciation resulting from sexual selection. Turner & radiations, seem to di¡er most obviously in sexually Burrows (1995) constructed a genetic model of a male dimorphic traits. This impression has been con¢rmed by trait a¡ected by four loci and a female preference for that two phylogenetic studies of birds. Barraclough et al. (1995) trait a¡ected by a single locus with two . Under found a positive correlation between the speciosity of some conditions, this model produced sympatric taxa groups and their degree of sexual dimorphism, suggesting showing complete sexual isolation; but these results may a link between sexual selection and speciation. (It is not be highly dependent on the assumptions. (One such likely that species are simply recognized more easily in assumption was that the preference locus showed complete strongly dimorphic groups, for the authors note that dominance, and more realistic assumptions of inter- females of such species can also be distinguished easily.) mediate dominance and additional loci would almost Mitra et al. (1996) found that taxa with promiscuous certainly reduce the probability of speciation.) mating systems contain more species than their non- The only theoretical study of promiscuous sister taxa. Promiscuously mating species resulting from sexual isolation is that of Iwasa & are, of course, more likely to experience strong sexual Pomiankowski (1995). Their quantitative-genetic model of selection. sexual selection on a male trait and on female preference The connection between sexual selection and sexual assumes a ¢tness cost to increased female preference. This isolation may seem obvious. After all, it is a tenet of neo- cost results in a cyclical £uctuation of both trait and that reproductive isolation is a by-product of preference. Di¡erent populations undergoing the same evolutionary change occurring within populations. Two selection regime could then fall out of synchrony, causing populations undergoing sexual selection may readily sexual isolation. Kirkpatrick & Barton (1995), however, diverge in both male traits and female preferences, and note that this model neglects the possibility of stabilizing the natural outcome of this would be sexual isolation selection acting on the male trait around its optimum between the populations. This idea is much easier to value for survival, and that such selection (even if very grasp than, say, the notion that adaptation among isolated weak) could eliminate the cycling. Moreover, related populations would cause sterility or inviability of their species do not usually show di¡erent stages of elaboration hybrids; under sexual selection, the pleiotropic e¡ect of or diminution of the same male-limited trait, but instead the diverging genes (sexual isolation) is obviously the exaggeration of completely di¡erent traits. (Some connected to their primary e¡ect (exaggeration of male species of bowerbirds, for example, have di¡erent forms traits or female preferences). of elaborate male plumage, while others lack such

Phil. Trans. R. Soc. Lond. B (1998) 292 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation plumage but have males who build elaborate bowers; tions that have evolved some postzygotic isolation in Gilliard 1956.) A useful extension of Iwasa & Pomian- allopatry undergo selection for increased sexual isolation kowski's (1995) model might incorporate either novel when they later become sympatric. Reinforcement was environments or genetic drift that could launch popula- introduced and popularized by Dobzhansky (1937), who tions on di¡erent trajectories of sexual selection. Schluter apparently considered it the necessary last step of specia- & Price (1993) also suggest that, if the secondary sexual tion. Its wide popularity may have re£ected its appealing traits of males re£ect their ¢tness or physiological condi- assumption of a creative (and not an incidental) role for tion, di¡erences among habitats that change the female's natural selection in speciation (Coyne 1994). Although ability to detect di¡erent traits could lead to sexual isola- theoretical studies of reinforcement appeared only tion among populations. recently, two analyses of Drosophila (Ehrman 1965; Sexual isolation is often asymmetric, i.e. species or Wasserman & Koepfer 1977) supported the idea: in populations show strong isolation in only one direction of species pairs with overlapping ranges, sexual isolation was the hybridization. This is a common phenomenon in stronger when populations derived from areas of sympatry Drosophila (Watanabe & Kawanishi 1979; Kaneshiro than from allopatry. 1980), is also seen in salamanders (Arnold et al. 1996), and In the 1980s, however, several critiques eroded the may be ubiquitous in other . Kaneshiro (1980) popularity of reinforcement. First, Templeton (1981) o¡ered an explanation of this pattern based on biogeo- pointed out that the pattern of stronger sexual isolation graphy, but newer data contradict his theory (Cobb et al. among sympatric than allopatric populations could be 1989; David et al. 1974). Arnold et al. (1996) proposed that caused not by reinforcement but by `di¡erential fusion', in mating asymmetry is a transitory phenomenon that decays which species could persist in sympatry only if they had rapidly as populations diverge, but such asymmetry is evolved su¤ciently strong sexual isolation in allopatry. often seen among even distantly related species of Thus, stronger isolation in sympatry might re£ect not Drosophila (Coyne & Orr 1989b). (Of course asymmetry direct selection, but the hybridization and fusion of must eventually disappear, because sexual isolation will, weakly isolated populations. Moreover, it became clear with time, become complete in both directions.) There that some of the data o¡ered in support of reinforcement are likely to be other explanations for mating asymmetry, was £awed (Butlin 1987). Finally, the ¢rst serious theore- such as the combination of open-ended female preferences tical treatment of reinforcement (Spencer et al. 1986) and sexual selection operating in only one of two showed that even under favourable conditions (e.g. populations. If asymmetry does prove ubiquitous, it may complete sterility of hybrids), of populations provide important clues about how sexual selection causes occurred more often than reinforcement. speciation. Recently, however, a combination of empirical and Unfortunately, there are too few data about the genetics theoretical work has resurrected the popularity of reinfor- of sexual isolation to con¢rm or motivate any theory. The cement. In an analysis of 171 pairs of Drosophila species, few studies listed in table 1 show only that sexual isolation Coyne & Orr (1989b, 1997) found that recently diverged may sometimes have a simple basis, as in races of Ostrina pairs show far more sexual isolation when sympatric than nubialis, where complete sexual isolation is apparently allopatric (¢gure 3). (An independent analysis of these based on changes at only three loci (one each for di¡er- data by Noor (1997a), making less restrictive assumptions, ences in female pheromones, male perception and male arrived at similar conclusions.) Although in 1989 there was attraction), or can be more complicated, as in Drosophila no theoretical work showing that reinforcement was mauritiana/D. sechellia, where di¡erences in female phero- feasible, two such studies have appeared recently. Liou & mones alone involve at least ¢ve loci. One conclusion that Price (1994) and Kelly & Noor (1996) showed that reinfor- seems reasonable, based as it is on three independent cement can occur frequently even if hybrids have only studies (table 1), is that for a given pair of species, the moderate postzygotic isolation. The important di¡erence genes causing sexual isolation of males di¡er from those between these models and that of Spencer et al. (1986) is causing sexual isolation in females. This is not surprising, that the former explicitly allow for sexual selection, which as there is no obvious reason why genes a¡ecting male greatly enhances the evolution of sexual isolation. traits should be identical to those a¡ecting female percep- Newer data also support reinforcement. These include a tion. (It is possible that sexual selection could cause male reanalysis of the earlier literature, ¢nding many more and female genes to be closely linked by selecting for possible examples (Howard 1993), and two new studies of genetic correlations between trait and preference, but species pairs with partially overlapping ranges (Noor 1995; there is no theory addressing this possibility.) This lack of Saetre et al. 1997). The work of Saetre et al. (1997) on two correlation is worth verifying, however, as its absence is species of European £ycatchers is especially interesting, as assumed in many models of sexual isolation (e.g. Spencer the reduced hybridization in sympatry is caused by the et al. 1986). divergence of male plumage occurring in that area, a Given the likely importance of sexual selection in di¡erence presumably caused by sexual selection. animal speciation, more genetic analyses are clearly Although the pattern of stronger isolation in sympatry needed. Fortunately, the advent of QTL analysis has shown in ¢gure 3 might in principle be explained by made such studies feasible in any pair of species that several processes, including reinforcement and di¡erential produces fertile hybrids. fusion (Coyne & Orr 1989b), reinforcement seems most likely for several reasons. First, di¡erential fusion posits (b) Reinforcement that cases of strong sexual isolation in sympatry form a One of the greatest controversies in speciation concerns subset of the levels of isolation seen in allopatry. Although reinforcement: the process whereby two allopatric popula- allopatric taxa might show either strong or weak isolation,

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 293

1 views about speciation over the last decade. Future work must determine whether reinforcement is rare or ubiquitous across animal taxa, and whether it occurs in 0.8 plants.

0.6 (c) Ecological isolation Ecological isolation must play a role in maintaining 0.4 biological diversity asöeven if it is not a primary isolating mechanismöecological di¡erences are required for the

Prezygotic Isolation sympatric coexistence of taxa. Consider, for instance, the 0.2 fate of a newly arisen polyploid plant species. Although all (a) auto- or allopolyploids are automatically postzygotically 0 isolated from their parental species (hybridization produces mostly sterile triploid hybrids), any polyploid that does not di¡er ecologically from its ancestral species 1 will be quickly driven extinct through either competition or the production of sterile hybrids with its ancestors, 0.8 rendering it unavailable for study. `Ecological isolation'actually subsumes three phenomena: 0.6 1. Individuals of di¡erent species live in the same region 0.4 and may encounter each other, but con¢ne mating and/or reproduction to di¡erent habitats so that

Prezygotic Isolation hybrids are not formed (e.g. the sympatric, host-speci¢c 0.2 Drosophila that breed on di¡erent cacti; Ruiz & Heed (b) 1988). This is a form of prezygotic reproductive isola- tion, and the only type of reproductive isolation that 0 0.5 1 1.5 2 can by itself both cause complete speciation and allow Genetic Distance (D) persistence of species in sympatry. This form of isolation Figure 3. Prezygotic isolation in Drosophila plotted against may be the most common result of sympatric speciation Nei's (1972) electrophoretic genetic distance. Each point (Rice & Hostert 1993). represents the average among pairs within a species group. (a) 2. Species live in di¡erent subniches of the same area and Allopatric taxa; (b) sympatric taxa. See Coyne & Orr (1989b, rarely, if ever, come into contact (e.g. the spadefoot 1997) for further details. toads Scaphiopus holbrooki hurteri and S. couchi). Although their ranges overlap extensively, heterospeci¢c toads the latter cases disappear by fusion on geographic contact, almost never meet because they are restricted to leaving us with a pre-existing set of strongly isolated taxa. di¡erent soil types (Wasserman 1957). Although this This is not, however, the pattern seen in Drosophila. Figure situation corresponds to Dobzhansky's (1937, p. 231) 3 shows that no recently diverged allopatric taxa have de¢nition of ecological isolation, such species are e¡ec- sexual isolation as strong as that seen among sympatric tively allopatric. taxa of the same age. Furthermore, di¡erential fusion 3. Species live in di¡erent subhabitats of the same area predicts that postzygotic as well as prezygotic isolation and sometimes come into contact with one another will be stronger in sympatry, as the probability of fusion and mate, forming hybrids that are not well-adapted should decrease with any form of reproductive isolation. to available habitat. Such cases may be common, and The Drosophila data also fail to support this prediction: have been studied extensively in stickleback ¢sh although prezygotic isolation is much stronger in (Schluter 1996). They constitute examples of postzygotic sympatry than allopatry, postzygotic isolation is virtually isolation that, while technically a form of hybrid invia- identical in the two groups (Coyne & Orr 1997). bility, depend on ecological details of the environment Finally, it should be noted that reinforcement does not and not on inherent problems of development. Like any necessarily require the pre-existence of hybrid sterility or other form of postzygotic isolation, `ecological invia- inviability, but might also result when allopatric popula- bility' could trigger reinforcement between sympatric tions evolve some behavioural or ecological di¡erence species (Coyne & Orr 1989b). Unlike other forms of that leaves hybrids behaviourally or ecologically mala- postzygotic isolation, however, ecological inviability dapted. Stratton & Uetz (1986), for example, observed mayödepending on how population size is regulated that hybrids between two species of spiders were in the di¡erent habitatsöallow two species to coexist `behaviourally sterile', so that hybrid males were rejected without prezygotic isolation. In addition, postzygotic by females of both species, and hybrid females refused to isolation based on ecological divergence need not mate with any male. Similarly, Davies et al. (1997) involve complementary gene interactions (epistasis) observed that hybrid females between two species of between alleles of two species. Instead, the genetics of butter£ies su¡er a reduced tendency to mate. this type of ecological isolation would presumably The data and theory reviewed above strongly suggest resemble the genetics of ordinary adaptation (about that reinforcement of sexual isolation can occur. This which we unfortunately know little; Orr & Coyne conclusion represents one of the most radical changes of 1992).

Phil. Trans. R. Soc. Lond. B (1998) 294 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation

Schluter (1996, 1997) makes a strong case for the impor- nated by both conspeci¢c and heterospeci¢c males tance of ecological isolation (especially type 3) in the produce very few hybrids. Such reproductive isolation origin and persistence of species. Unfortunately, there has thus depends on competition between conspeci¢c and been only one study, not yet published, of the genetics of heterospeci¢c sperm, and resembles interspeci¢c pollen ecological isolation. The island endemic Drosophila sechellia competition in some plants, in which heterospeci¢c pollen breeds exclusively on the normally toxic fruit of Morinda tubes grow more slowly than conspeci¢c tubes, yielding citrifolia. All of D. sechellia's relatives, including its reproductive isolation detectable only after double pollina- presumed ancestor D. simulans, succumb to Morinda's tion (Rieseberg et al. 1995). The only genetic analysis primary toxin, octanoic acid. Although D. sechellia and among these studies is that of Price (1997), who showed D. simulans were originally allopatric, they have recently that, among females, `conspeci¢c sperm precedence' was become sympatric and breed on separate hosts. Recent dominant, so that F1 female hybrids between D. simulans work (Jones 1998) shows that D. sechellia's resistance to and D. mauritiana induce the same sperm preference as do octanoic acid involves at least ¢ve mostly dominant D. simulans females. Gametic incompatibilities may be alleles, distributed over all three of the major chromo- among the earliest-evolving forms of reproductive isola- somes, with the largest e¡ect mapping to chromosome tion, and are worthy of more attention. three.

(d) Pollinator isolation 6. POSTZYGOTIC ISOLATION Pollinator isolation probably represents a common form Postzygotic isolation occurs when hybrids are un¢t. of reproductive isolation in plants (Grant 1981). A variant Evolutionists have, historically, pointed to three types of of this is the isolation of -pollinated plants from self- genetic di¡erences as causes of these ¢tness problems: compatible species, a mechanism described in Mimulus. species may have di¡erent chromosome arrangements, The few data at hand (table 1) show that the di¡erence di¡erent ploidy levels, or di¡erent alleles that do not func- between outcrossing and inbreeding is due to several tion properly when brought together in hybrids. Although genes, but that di¡erences in £ower shape, colour or each of these modes of speciation has enjoyed its advo- nectar reward, that attract di¡erent pollinators, may be cates, it is now clear that the latter two are by far the due to one or a few major genes (Prazmo 1965; Bradshaw most important. et al. 1995). This latter observation, if common, might Speciation by auto- and allopolyploidy is clearly suggest a rapid form of speciation. common in plants, as about 60% of angiosperms are of polyploid origin (Masterson 1994). We will not, however, (e) Postmating, prezygotic isolation discuss polyploid speciation here as it has been thoroughly Biologists have begun to appreciate that sexual selection reviewed elsewhere (see, for example, Grant 1981; Ramsey is not limited to obvious behavioural and morphological & Schemske 1998). Instead, we concentrate on the ques- traits that act before copulation, but can include `cryptic' tion of whether postzygotic isolation in animals is based characters acting between copulation and fertilization. on chromosomal or genic di¡erences. We will conclude Such selection (as well as sexually antagonistic selection; that genes play a far more important role in hybrid steri- Rice 1996) can lead to female control of sperm usage, lity and inviability than do structural di¡erences in male^male sperm competition within multiply insemi- chromosomes. nated females and the mediation of such competition by the female. Selection acting between copulation and ferti- (a) Chromosomal speciation lization has also been invoked to explain the striking The notion that chromosome rearrangements are a diversity of male genitalia among animal species, bizarre major cause of hybrid sterility was once very popular conformations of female reproductive tracts and postcopu- (White 1969, 1978), and still enjoys a few adherents (e.g. lation `courtship' behaviour by males (Eberhard 1985, King 1993). The idea of chromosomal speciation, 1996). Moreover, the relatively rapid evolution of proteins however, su¡ers from both theoretical and empirical di¤- involved in reproduction (Coulthart & Singh 1988; Metz culties. The theoretical problems have been discussed & Palumbi 1996; Tsaur & Wu 1997) is also consistent with elsewhere (Lande 1979; Walsh 1982; Barton & Charlesworth sexual selection. 1984), but the empirical problems have not been widely Just as sexual selection acting on male plumage or recognized. courtship behaviour can pleiotropically produce sexual The ¢rst such di¤culty is that many species producing isolation, so postcopulation, prezygotic sexual selection sterile hybrids are homosequential, that is, they do not can produce `cryptic' sexual isolation detectable only after di¡er in chromosome arrangement. White (1969, p. 77), fertilization. Such isolation may take the form of either the greatest champion of chromosomal speciation, argued blocked heterospeci¢c fertilization, such as the `insemina- that such cases `represent only a small fraction of the total tion reaction' of Drosophila (Patterson 1946) or the number of species complexes that have been extensively preferential use of conspeci¢c sperm when a female is studied;'. But this argument is potentially misleading: sequentially inseminated by heterospeci¢c and conspeci¢c many of the taxa showing ¢xed chromosomal di¡erences males. This latter phenomenon has recently been are old, and may have accumulated manyöif not allöof described in grasshoppers (Bella et al. 1992), crickets their chromosomal di¡erences after the actual speciation (Gregory & Howard 1994), £our beetles (Wade et al. event. Resolving this issue requires systematic compari- 1994) and Drosophila (Price 1997). In three species of Droso- sons of the extent of chromosomal divergence with both phila, single heterospeci¢c inseminations produce large the age of taxa (as determined by molecular data) and numbers of hybrid o¡spring, but females doubly insemi- the strength of postzygotic isolation. In addition, although

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 295 sterile hybrids often su¡er meiotic pairing problems, this so is sterile. Using the metaphor of adaptive landscapes, it sterility is often limited to the (see is hard to see how two related species can come to reside below). This is not expected if sterility results from the on di¡erent adaptive peaks unless one lineage passed disruption of chromosome pairing, which in most cases through an adaptive valley. should a¥ict both males and females. Indeed, in some This problem was ¢nally solved by Bateson (1909), species, such as Drosophila, intraspeci¢c chromosomal rear- Dobzhansky (1937) and Muller (1942), who noted that, if rangements typically sterilize females only (male postzygotic isolation is based on incompatibilities between Drosophila have no recombination), but hybrid sterility is two or more genes, hybrid sterility and inviability can far more severe in males (Ashburner 1989). evolve unimpeded by natural selection. If, for example, Third, even in species hybrids whose chromosomes fail the ancestral species had aabb, a new to pair during meiosis, we do not know whether this at one locus (allele A) could be ¢xed by selection or drift failure is caused by di¡erences in chromosome arrange- in one isolated population because the Aabb and AAbb ment or di¡erences in genes. As Dobzhansky (1937) are perfectly ¢t. Similarly, a new allele (B) at emphasized, both rearrangements and gene mutations the other locus could be ¢xed in a di¡erent population can disrupt meiosis within species and so, presumably, since aaBb and aaBB genotypes are also ¢t. But it is within species hybrids. The best attempt to disentangle entirely possible that when the AAbb and aaBB populations these causes remains the ¢rst: Drosophila pseudoobscura and come into contact, the resulting AaBb hybrids could be D. persimilis, which di¡er by at least four inversions, sterile or inviable. The A and B alleles have never been produce sterile hybrid males (Dobzhansky 1937). Meiotic `tested' together within a genome, and so may not function pairing in hybrids is abnormal and univalents are properly in hybrids. common. Dobzhansky (1933) showed that islands of tetra- Alleles showing this pattern of epistasis are called ploid spermatocytes are often found in hybrid testes. All `complementary genes'. Such genes need not, of course, chromosomes in these 4N cells have a homologous have drastic e¡ects on hybrid ¢tness; any particular pairing partner, and hence should show improved pairing incompatibility might lower hybrid ¢tness by only a small if pairing problems in hybrid cells re£ect structural di¡er- amount. It should also be noted that the Dobzhansky^ ences between chromosomes. However, univalents are just Muller model is agnostic about the evolutionary causes of as common in the hybrid tetraploid as in diploid sperma- substitutions that ultimately produce hybrid sterility or tocytes. The hybrid meiotic problems must therefore have inviability; purely adaptive or purely neutral evolution a genetic and not a chromosomal basis. This test has within populations can give rise to complementary genes apparently not been repeated in any other hybridization. and thus to postzygotic isolation. A further problem with chromosomal speciation is that The Dobzhansky^Muller model is the basis for almost it depends critically on the semisterility of hybrids who are all modern work in the genetics of postzygotic isolation. heterozygous for chromosome rearrangements. It is not There is now overwhelming evidence that hybrid sterility widely appreciated, however, that heterozygous rearran- and inviability do indeed result from such between-locus gements theoretically expected to be deleterious (e.g. incompatibilities (reviewed in Orr 1997). Curiously, there fusions and pericentric inversions) in reality often enjoy have been few theoretical studies of the Dobzhansky^ normal ¢tness, probably because segregation is regular or Muller model. Recent analyses, however, predict that the recombination is prevented (see discussion in Coyne et al. evolution of postzygotic isolation should show several (1997)). Any putative case of chromosomal speciation regularities. requires proof that di¡erent rearrangements actually cause semisterility in heterozygotes, and almost no studies (c) Patterns in the genetics of postzygotic isolation have met this standard. Long before any formal studies of the Dobzhansky^ Finally, direct genetic analyses over the last decade have Muller model, Muller (1942) predicted that the alleles shown conclusively that postzygotic isolation in animals is causing postzygotic isolation must act asymmetrically. To typically caused by genes, not by large chromosome rear- see this, consider the two-locus scenario sketched above. rangements. Many of these genes have been well mapped, Although the A and B alleles might be incompatible in and several have been genetically characterized or even hybrids, their allelomorphs a and b must be compatible. cloned (Wittbrodt et al. 1989; Orr 1992; Perez et al. 1993). This is because the aabb genotype must represent an ances- Our main task, therefore, is to understand how the evolu- tral state in the evolution of the two species. tion of genic di¡erences can produce hybrid sterility and There is now good evidence that genic incompatibilities inviability. do in fact act asymmetrically. The best data come fromWu & Beckenbach's (1983) study of male sterility in Drosophila (b) The Dobzhansky^Muller model pseudoobscura^D. persimilis hybrids. When an X-linked Understanding the evolution of postzygotic isolation is region from one species caused sterility upon introgression di¤cult, because the we are hoping to into the other species' genome, they found that the reci- explainöthe inviability and sterility of hybridsöseem procal introgression had no such e¡ect. This observation maladaptive. The di¤culty is best seen by considering the has now been con¢rmed in other Drosophila hybridizations simplest possible model for the evolution of postzygotic (e.g. Orr & Coyne 1989). isolation: change at a single gene. One species has geno- The second pattern expected under the Dobzhansky^ type AA and the other aa, and Aa hybrids are completely Muller model was pointed out more recently. If hybrid sterile. Regardless of whether the common ancestor was sterility and inviability are caused by the accumulation of AA or aa, ¢xation of the alternative allele cannot occur complementary genes, the severity of postzygotic isolation, because the ¢rst mutant individual has genotype Aa and as well as the number of genes involved, should `snowball'

Phil. Trans. R. Soc. Lond. B (1998) 296 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation much faster than linearly with time (Orr 1995; see, also, Orr & Orr (1996) showed that if the substitutions ulti- Menotti-Raymond et al. 1997). This follows from the fact mately causing Dobzhansky^Muller incompatibilities are that any new substitution in one species is potentially driven by natural selectionöas seems likely (Christie & incompatible with the alleles at every locus that has Macnair 1984)öthe waiting time to speciation grows previously diverged in the other species. (In our discussion longer as a species of a given size is splintered into ever above, the B substitution is potentially incompatible with smaller populations. If, however, the substitutions causing the previous A substitution in the other species.) Later hybrid problems are originally neutral, population subdi- substitutions are therefore more likely to cause hybrid vision has little e¡ect on the time to speciation. In no case incompatibilities than earlier ones. Consequently, the is the accumulation of hybrid incompatibilities greatly cumulative number of hybrid incompatibilities increases accelerated by small population size. Unfortunately, we much faster than linearly with the number of substitutions, have little empirical data bearing on this issue. Although K. If all incompatibilities involve pairs of loci, the it might seem that the e¡ect of population size on specia- expected number of Dobzhansky^Muller incompatibil- tion rates could be estimated by comparing the rate of ities increases as K 2, or (assuming a rough of postzygotic isolation on islands versus conti- clock) as the square of the time since species diverged nents, this comparison is confounded by the likelihood of (Orr 1995). If incompatibilities sometimes involve interac- stronger selection in novel island habitats, which itself tions between more than two loci (see below), the number might drive rapid speciation. of hybrid incompatibilities will rise even faster. This snow- Finally, much of the theoretical work on the evolution of balling e¡ect requires that we interpret genetic studies of postzygotic isolation has been devoted to explaining one of postzygotic isolation with caution. Because the genetics of the most striking patterns characterizing the evolution of hybrid sterility and inviability will quickly grow compli- hybrid sterility and inviability, Haldane's rule. Because cated as species diverge, it is easy to overestimate the this large and confusing literature has recently been number of genes required to cause strong reproductive reviewed elsewhere (e.g. Wu et al. 1996; Orr 1997), we will isolation (see Orr (1995) and below). not attempt a thorough discussion here. Instead, we brie£y The limited data we possess are consistent with a snow- consider the data from nature and sketch the leading balling e¡ect, although they do not prove it (see Orr hypotheses o¡ered to explain them. (1995) for a discussion). The simplest prediction of the snowballing hypothesis is that the number of mapped (d) Haldane's rule genes causing hybrid sterility or inviability should increase In 1922, Haldane noted that, if only one hybrid sex is quickly with molecular genetic distance between species. sterile or inviable, it is nearly always the heterogametic But given the enormous di¤culties inherent in accurately (XY) sex. More recent and far more extensive reviews mapping and counting `speciation genes', it may be some (Coyne 1992c) show that `Haldane's rule' is obeyed in all time before such direct contrasts are possible. animal groups that have been surveyed, e.g. Drosophila, Third, while we have assumed that hybrid incompatibil- mammals, Orthoptera, birds and Lepidoptera (the latter ities involve pairs of genes, analysis of the Dobzhansky^ two groups have heterogametic females). Indeed, it is Muller model shows that more complex hybrid incompat- likely that Haldane's rule characterizes postzygotic ibilities, involving interactions among three or more of isolation in all animals having chromosomal sex determi- genes, should be common. The reason is not intuitively nation. Moreover, these surveys show that Haldane's rule obvious, but is easily demonstrated mathematically is consistently obeyed. In Drosophila, for instance, 114 (Cabot et al. 1994; Orr 1995). Certain paths to the evolu- species crosses produce hybrids that are sterile in only one tion of new species are barred because they would require gender. In 112 of these, it is the males who are sterile passing through intermediate genotypes that are sterile or (Coyne 1992c). Comparative work in Drosophila also shows inviable. It is easy to show, however, that the proportion of that Haldane's rule represents an early stage in the evolu- all imaginable paths to speciation allowed by selection tion of postzygotic isolation: hybrid male sterility or increases with the complexity of hybrid incompatibilities inviability arises quite quickly, whereas female e¡ects (Orr 1995). Thus, for the same reason that two-gene appear only much later (Coyne & Orr 1989b, 1997). speciation is `easier' than single-gene speciation, so three- For obvious reasons, Haldane's rule has received a great gene is easier than two-gene, and so on. deal of attention: it represents one of the strongest patterns The evidence for complex incompatibilities is now over- in , and perhaps the only pattern whelming. They have been described in the Drosophila characterizing speciation. In addition, the rule implies obscura group (Muller 1942; H. A. Orr, unpublished data), that there is some fundamental similarity in the genetic the Drosophila virilis group (Orr & Coyne 1989), the Droso- events causing speciation in all animals. phila repleta group (Carvajal et al. 1996), and the Drosophila Although many hypotheses have been o¡ered to explain melanogaster group (Cabot et al. 1994; Davis et al. 1994). Haldane's rule, most have been falsi¢ed. We will not Indeed, such interactions could prove more common than consider these failed explanations here (for reviews, see the two-locus interactions discussed at length by Bateson, Orr 1997; Wu et al. 1996; Coyne et al. 1991; Coyne 1992c). Dobzhansky and Muller. Instead, we brie£y review the three explanations of Theoretical analysis of the Dobzhansky^Muller model Haldane's rule that remain viable. There is strong evidence also has yielded counterintuitive results about the e¡ect of for two of these, and suggestive evidence for the third. In a population subdivision on speciation. Many evolutionists ¢eld that has historically been rife with disagreement, a have maintained, for example, that speciation is most surprisingly good consensus has emerged that some combi- likely in taxa subdivided into small populations. At least nation of these hypotheses explains Haldane's rule (Orr for postzygotic isolation, this idea is demonstrably false. 1997; Wu et al. 1996; True et al. 1996).

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 297

The ¢rst hypothesis, the `dominance theory', posits that sexual selection involve males per se, whereas Haldane's Haldane's rule re£ects the recessivity of X-linked genes rule pertains to heterogametic hybrids, male or not. causing hybrid problems. This idea was ¢rst suggested by Second, the faster-male theory cannot account for Muller (1942), and his verbal theory was later formalized Haldane's rule for inviability in any taxa. Because there is by Orr (1993b) and Turelli & Orr (1995). The mathema- strong evidence that genes causing lethality are almost tical work shows that heterogametic hybrids su¡er greater always expressed in both sexes (reviewed in Orr 1997), it sterility and inviability than homogametic hybrids when- seems unlikely that hybrid `male lethals' can evolve faster ever the alleles causing hybrid incompatibilities are, on than `female lethals'. Finally, it is not obvious that sexual average, partially recessive d < 1; this parameter incorpo- selection would inevitably lead to faster substitution of 2 rates both the e¡ects of dominance per se and any `male' than `female' alleles. One can easily imagine, for correlation between dominance and severity of hybrid instance, forms of sexual selection in which each substitu- e¡ects (see Turelli & Orr 1995). The reason is straight- tion a¡ecting a male character is matched by a forward. Although XY hybrids su¡er the full hemizygous substitution a¡ecting female preference for that character. e¡ect of all X-linked alleles causing hybrid problems If sexual selection causes faster evolution of male sterility, (dominant and recessive), XX hybrids su¡er twice as one may need to consider processes like male^male many X-linked incompatibilities (as they carry twice as competition in addition to male^female coevolution. many Xs). These two forces balance when d ˆ 1. But if Despite these caveats, there is now good evidence at 2 ö d < 1 the expression of recessives in XY hybrids outweighs least in Drosophila that alleles causing sterility of hybrid 2 ö the greater number of incompatibilities in XX hybrids, males accumulate much faster than those a¡ecting females and Haldane's rule results. Obviously, the dominance (True et al. 1996; Hollocher & Wu 1996). (Unfortunately, theory can account not only for Haldane's rule, but also both of these studies analysed the same species pair; for the well-known large e¡ect of the X chromosome on analogous data from other species are badly needed.) hybrid sterility and inviability (Wu & Davis 1993; Turelli Although we cannot be sure of the mechanism involved, & Orr 1995). it certainly appears that `faster-male' evolution plays an There is now strong evidence that dominance explains important role in Haldane's rule for sterility in taxa with Haldane's rule for hybrid inviability. In particular, the heterogametic males. dominance theory predicts that, in Drosophila hybridiza- The last hypothesis, the `faster-X' theory, posits that tions obeying Haldane's rule for inviability, hybrid Haldane's rule re£ects the more rapid divergence of X- females who are forced to be homozygous for their X chro- linked than autosomal loci (Charlesworth et al. 1987; mosome should be as inviable as F1 hybrid males. (Such Coyne & Orr 1989b). Charlesworth et al. (1987) showed `unbalanced' females possess an F1 male-like genotype in that, if the alleles ultimately causing postzygotic isolation which all recessive X-linked genes are fully expressed.) In were originally ¢xed by natural selection, X-linked genes both of the species crosses in which this test has been will evolve faster than autosomal genes if favourable muta- performed, unbalanced females are, as expected, comple- tions are on average partially recessive (h < 1 ). (It must be 2 tely inviable (Orr 1993a; Wu & Davis 1993). Similarly, emphasized that this theory requires only that the favour- there is evidence from haplodiploid species that hybrid able e¡ects of mutations on their `normal' conspeci¢c backcross males (who are haploid) su¡er more severe genetic background are partially recessive; nothing is inviability than their diploid sisters (Breeuwer & Werren assumed about the dominance of alleles in hybrids. 1995). Conversely, the dominance theory requires only that the There is also weaker indirect evidence that the alleles alleles causing hybrid problems act as partial recessives in causing hybrid sterility act as partial recessives; Hollocher hybrids; nothing is assumed about the dominance of these & Wu (1996) and True et al. (1996) found that although alleles on their normal conspeci¢c genetic background.) most heterozygous introgressions from one Drosophila Under various scenarios, this faster evolution of X-linked species into another are reasonably fertile, many homozy- genes can indirectly give rise to Haldane's rule (Orr 1997). gous introgressions are sterile. It therefore seems likely There is some evidence that X-linked hybrid steriles and that dominance contributes to Haldane's rule for both lethals do in fact evolve faster than their autosomal ana- hybrid inviability and hybrid sterility. Last, it is worth logues. In their genome-wide survey of speciation genes in noting that the dominance theoryöunlike several alter- the Drosophila simulans^D. mauritiana hybridization, True et nativesöshould hold in all animal taxa, regardless of al. (1996) found a signi¢cantly higher density of hybrid which sex is heterogametic (Orr & Turelli 1996). male steriles on the X chromosome than on the autosomes. The second hypothesis posits that Haldane's rule re£ects Hollocher & Wu (1996), however, found no such di¡erence the faster evolution of genes ultimately causing hybrid in a much smaller experiment. Thus, while faster-X evolu- male than female sterility (Wu & Davis 1993; Wu et al. tion may contribute to Haldane's rule, the evidence for it is 1996). Wu and his colleagues o¡er two explanations for considerably weaker than that for both the dominance and this `faster male' evolution: (i) in hybrids, spermatogenesis faster-male theories. (The faster-X theory also su¡ers may be disrupted far more easily than oogenesis, and (ii) several other shortcomings described elsewhere; Orr 1997.) sexual selection may cause genes expressed in males to De¢nitive tests of the faster-X theory must await new evolve faster than those expressed in females. Although experiments that, following True et al. (1996), allow direct there is now good evidence for faster male evolution (see comparison of the number of hybrid steriles and lethals on below), this theory cannot be the sole explanation of the X versus autosomes. Haldane's rule. First, it cannot explain Haldane's rule for In summary, there is now strong evidence for both the sterility in those taxa having heterogametic females, e.g. dominance and faster-male theories of Haldane's rule. birds and butter£ies. After all, spermatogenesis and Future work must include better estimates of the

Phil. Trans. R. Soc. Lond. B (1998) 298 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation dominance of hybrid steriles, better tests of the faster-X namely, failure to condense chromosomes during mitosis. theory, and, most important, genetic analyses of Haldane's (Such a suggestion may seem contradicted by recent intro- rule in taxa having heterogametic females. Although both gression experiments showing that many di¡erent the dominance and faster-male theories make clear chromosome regions cause postzygotic isolation when predictions about the genetics of postzygotic isolation in introgressed from one Drosophila species into another these groups (Orr 1997), we have virtually no direct (True et al. 1996; Hollocher & Wu 1996). But the over- genetic data from these critical taxa. Last, it is important whelming majority of these introgressions have to determine if Haldane's rule extends beyond animals, discernible e¡ects only when homozygous and, so, would particularly to species of plants with heteromorphic sex not play any role in F1 hybrid inviability or sterility chromosomes. (Hollocher & Wu 1996).) The discovery of hybrid rescue genes may also provide (e) Hybrid rescue mutations an important short-cut to the cloning and characterization No discussion of postzygotic isolation would be of speciation genes. It is possible that rescue mutations are complete without mentioning a recent and remarkable alleles of the genes that normally cause hybrid inviability discovery,`hybrid rescue' mutations. These are alleles that, or sterility (Hutter & Ashburner 1987). If so, character- when introduced singly into Drosophila hybrids, rescue the ization of these mutations might quickly lead to the viability or fertility of normally inviable or sterile indivi- molecular isolation of speciation genes. Alternatively, duals (Watanabe 1979; Takamura & Watanabe 1980; rescue mutations might be second-site suppressors, i.e. Hutter & Ashburner 1987; Hutter et al. 1990; Sawamura et mutations at some second set of loci that suppress the al. 1993a,b,c; Davis et al. 1996). All rescue mutations problems caused by a di¡erent, primary set of genes. studied to date involve the D. melanogaster^D. simulans hybridization. In one direction of this cross, hybrid males (f) The role of endosymbionts in speciation die as late larvae; in the other, hybrid females die as Recent work on postzygotic isolation has pointed to a embryos. All surviving hybrids are completely sterile novel cause of hybrid incompatibilities, cytoplasmic (Sturtevant 1920). Several mutations are known that incompatibility (CI) resulting from infection by cellular rescue the larval inviability: Lethal hybrid rescue (Lhr) from endosymbionts. Although CI has now been found in at D. simulans (Watanabe 1979), and Hybrid male rescue (Hmr) least ¢ve orders of (Stevens & Wade 1990; and In(1)AB from D. melanogaster (Hutter & Ashburner Ho¡mann & Turelli 1997), most work has focused on two 1987; Hutter et al. 1990). Despite some early confusion, it model systems, the £y, Drosophila simulans (Ho¡mann et al. now appears that these rescue mutations have no e¡ect on 1986; Turelli & Ho¡mann 1995), and the small parasitic hybrid embryonic inviability, which is instead rescued by a wasp, Nasonia (Perrot-Minnot et al. 1996; Werren 1997). di¡erent set of mutations: Zygotic hybrid rescue (Zhr) from In both systems, hybrid embryonic lethality results D. melanogaster (Sawamura et al. 1993c), and maternal hybrid when males from infected populations or species are rescue (mhr) from D. simulans (Sawamura et al. 1993a). The crossed to females from uninfected populations or species. fact that larval and embryonic lethality are rescued by Moreover, in both cases the infective agent is the rickettsia- non-overlapping sets of mutations strongly suggests that like endosymbiont Wolbachia (Turelli & Ho¡mann 1995; these forms of isolation have di¡erent developmental Werren 1997). Remarkably, antibiotic treatment of infected bases (Sawamura et al. 1993b). lines cures Wolbachia infections, allowing fully compatible Recently, Davis et al. (1996) described a mutation that crosses. Because infected-by-infected crosses are compa- rescues, albeit weakly, the fertility of D. melanogaster^ tibleöwhile the same crosses using genetically identical D. simulans hybrid females. Although little is known about but cured females are incompatibleöthe presence of this allele, its discovery suggests that it may be possible to Wolbachia in females must confer immunity to the e¡ects bring all of the genetic and molecular technology available of fertilization by sperm from infected males. The basis of in D. melanogaster to bear on speciation. (This species has this immunity is unknown. previously been of limited use in the genetics of speciation Because CI only occurs when `naive' uninfected cyto- because it produces no fertile progeny when crossed to any plasm is fertilized by sperm from infected males, CI is other species.) typically unidirectional. Recently, however, cases have The discovery of hybrid rescue genes has several impor- been found in both Drosophila (O'Neill & Karr 1990), and tant implications. First, it suggests that postzygotic Nasonia (Breeuwer & Werren 1995; Perrot-Minnot et al. isolation may have a simple genetic basis. It seems quite 1996), in which two di¡erent populations or species are unlikely that a single mutation could restore the viability infected by di¡erent varieties of Wolbachia. Crosses of hybrids if lethality were caused many di¡erent develop- between individuals carrying these di¡erent types of mental problems. But a simple developmental basis implies Wolbachia are bidirectionally incompatible, so that post- in turn a simple genetic basis; if many genes were involved, zygotic isolation is complete. This represents a remarkable it seems unlikely that they all would a¡ect the same devel- mode of speciation, for no changes are required in the opmental pathway. These inferences have been roughly host's genome. con¢rmed by more recent work on the developmental Although Wolbachia infections will surely prove basis of larval inviability in D. melanogaster^D. simulans commonöand speciation workers should routinely test for group hybrids, which shows that lethal hybrids su¡er a themöthere are reasons for questioning whether CI will profound mitotic defect that can be reversed by introduc- prove an important cause of speciation. First, Wolbachia tion of hybrid rescue mutations (Orr et al. 1997). This cannot explain Haldane's rule, which is ubiquitous among suggests (but does not prove) that hybrid inviability in animals and characterizes (at least in Drosophila) an early this case results from a single developmental defect, and nearly-obligate step in the evolution of postzygotic

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 299 isolation (Coyne & Orr 1989b, 1997). Although CI results in Packard Foundation to H.A.O. We thank M. Turelli for helpful dead embryos, we know of no cases in whichWolbachia causes comments and discussion of the `snowball e¡ect', and A. Magur- lethality of one sex only. Similarly, Wolbachia seems unlikely ran for calling our attention to the Haskins & Haskins paper. to be a common cause of hybrid sterility; among the many genetic analyses of hybrid sterility, only a single case involves APPENDIX 1. NOTES FOR TABLE 1 an endosymbiont (Somerson et al. 1984). 1. (Templeton 1977; Val 1977.) Sympatric species, biometric analysis. Head shape di¡erence conjectured 7. CONCLUSIONS (but not known) to be involved in sexual isolation The increasing prominence of work on speciation between the two species. Templeton (1977, p. 636) re£ects real progress: a number of important questions concludes that shape di¡erence `is a polygenic trait have been resolved by experiment and a number of determined by alleles with predominantly additive patterns explained by theory. This progress, in turn, e¡ects'. No con¢dence limits are given for either of derives from several fundamental but rarely recognized the two estimates of gene number (nine in one study changes in our approach to speciation. First, the ¢eld has and ten in the other). grown increasingly genetical. As a consequence, a large 2. (Pontecorvo 1943.) Sympatric species, genetic analysis. body of grand but notoriously slippery questions (How (Both species are cosmopolitan human commensals.) important are peak shifts in speciation? Is sympatric Gene number is a minimum estimate and includes speciation common?), have been replaced by a collection factors from both species. Results based on analysis of simpler questions (Is the Dobzhansky^Muller model of `pseudobackcross' hybrids between marked triploid correct? What is the cause of Haldane's rule?). Although D. melanogaster females and irradiated D. simulans it would be fatuous to claim that these new questions are males. more important than the old, there is no doubt that they 3. (Coyne 1996a.) Sympatric species, genetic analysis. are more tractable. Second, the connection between (Both species are cosmopolitan human commensals.) theory and experiment has grown increasingly close. Genes a¡ect the ratio of two female cuticular Whereas speciation once seemed riddled with amorphous hydrocarbons that appear to be involved in sexual and untestable verbal theories, the last decade of work has isolation. produced a body of mathematical theory yielding clear 4. (True et al. 1996; Wu et al. 1996.) Allopatric species, and testable predictions about the basis of reproductive genetic analysis. True et al.'s (1996) analysis implicates isolation. Last, but most important, many of these predic- at least 14 regions of the D. mauritiana genome that tions have been tested. cause male sterility when introgressed into a D. simulans Despite this progress, many questions about speciation background. Wu et al. (1996) report 15 genes on the remain unanswered. Throughout this review we have D. simulans X chromosome causing hybrid sterility; by tried to highlight those questions that seem to us both extrapolating to the entire genome, they estimate at important and tractable. Most fall into two broad sets. least 120 loci causing hybrid male sterility. This latter The ¢rst concerns speciation in taxa that have been rela- ¢gure may, however, be an overestimate, as some of tively ignored: does reinforcement occur in plants? Do the X-linked regions studied were known in advance pre- and postzygotic isolation evolve at about the same to have large e¡ects on sterility. See also Coyne & rate in most taxa as in Drosophila? Do plants with hetero- Charlesworth (1986), Coyne (1989), Davis & Wu morphic sex chromosomes obey Haldane's rule? Do (1996). hybrid male steriles still evolve faster than female steriles 5. (True et al. 1996.) Allopatric species, genetic analysis. in taxa having heterogametic females? How distinct are Estimate based on homozygous introgression of D. asexual taxa in sympatry? mauritiana segments into D. simulans; we have given our The second set of questions concerns the evolution of minimum estimate of gene number based on the size of prezygotic isolation, which has received less attention introgressions. Regions causing inviability invariably than postzygotic isolation: how complex is the genetic a¡ect both males and females. No large e¡ect of the X basis of sexual isolation? How common is reinforcement? chromosome for either character. There could be many Why is sexual isolation so often asymmetric? What is more factors a¡ecting both female sterility and invia- the connection between adaptive radiation and sexual bility, as 19 sites out of 87 cytological positions isolation? produced hybrid inviability and 12 sites caused female It may seem that trading yesterday's grand verbal spec- sterility. Density of male steriles is probably higher ulations for today's smaller, more technical studies risks a than that of female steriles: 65 out of the 185 regions permanent neglect of the larger questions about specia- caused male sterility. See also Davis et al. (1994), tion. We believe, however, that more focused pursuits of Hollocher & Wu (1996). tractable questions will ultimately produce better answers 6. (Coyne 1989a, 1992b, 1996b.) Allopatric species, genetic to these bigger questions. Just as no mature theory of analysis. Character studied was sexual isolation was possible until we understood the between D. mauritiana females and D. simulans males. mechanism of inheritance, so no mature view of speciation The major chromosomes had di¡erent e¡ects in the seems possible until we understand the origins and two sexes, implying that di¡erent genes are involved in mechanics of reproductive isolation. female discrimination versus the male character that females discriminate against. Our work is supported by grants from the National Institutes of 7. (True et al. 1997.) Allopatric species, genetic analysis. Health to J.A.C. and to H.A.O. and from the David and Lucile Character measured was the shape of the posterior lobe

Phil. Trans. R. Soc. Lond. B (1998) 300 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation

of the male genitalia.The interspeci¢c di¡erence in this koepferae. A total of two X-linked loci from D. character is thought, but not proven, to be involved in koepferae cause male lethality on a D. buzzatii shortened copulation time between D. mauritiana males genetic background; this inviability can be rescued and D. simulans females, which itself causes reduced by co-introgression of two autosomal segments from progeny number in hybrid matings (see note 8 below, D. koepferae. Therefore, at least four loci are and Coyne 1983). involved. 8. (Coyne 1993b.) Allopatric species, genetic analysis. 19. (Naveira & Fontdevila 1986, 1991.) Sympatric species, Shortened copulation time between D. simulans genetic analysis. This is probably a considerable under- females and D. mauritiana males contributes to the estimate of gene number, particularly on the X, as reduced number of progeny in interspeci¢c crosses. heterospeci¢c X chromosomal (but not autosomal) 9. (Coyne & Charlesworth 1997.) Allopatric species, segments of any size cause complete male sterility. genetic analysis. Ratio of two female cuticular hydro- Marin (1996), however, argues that Naveira & carbons that appears to a¡ect sexual isolation between Fontdevila over-estimated the number of autosomal these species. factors causing hybrid male sterility. 10. (Coyne & Kreitman 1986; Coyne & Charlesworth 20. (Khadem & Krimbas 1991.) Sympatric species, 1989; Cabot et al. 1994; Hollocher & Wu 1996.) genetic analysis. Male sterility quanti¢ed as testis size. Allopatric species, genetic analysis. At least three Both X-linked and autosomal factors a¡ect testis size genes on the X, two on the second, and one on the in hybrids; the X chromosome has the largest e¡ect. third chromosome cause sterility in hybrid males. 21. (Mitrofanov & Sidorova 1981.) Sympatric species 11. (Hollocher & Wu 1996.) Allopatric species, genetic (D. virilis is a cosmopolitan species associated with analysis. A total of two out of three large non-overlap- humans), genetic analysis. Character measured was ping regions on the D. sechellia second chromosome reduced viability of o¡spring from backcross females. (the only chromosome analysed) cause inviability 22. (Hoikkala & Lumme 1984.) Sympatric species, genetic when introgressed as homozygotes into D. simulans. analysis. A total of four factors (at least one on each 12. (Coyne 1992.) Allopatric species, genetic analysis. autosome) a¡ect the interspeci¢c di¡erence in Only two of three major chromosomes are involved in number of pulses in pulse train of male courtship female traits leading to reduced mating propensity of song. No apparent e¡ect of X. Character is possibly D. sechellia females with D. simulans males. but not yet de¢nitely known to be involved in sexual 13. (Vigneault & Zouros 1986; Pantazidis et al. 1993.) isolation between these species. Sympatric species, genetic analysis. Trait studied was 23. (Heikkinen & Lumme 1991.) Sympatric species, sperm motility in hybrid males. The Y chromosome genetic analysis. F1 males are weakly sterile. Backcross and two autosomes a¡ect the character, whereas two analysis uncovered X-2, X-4, X-5 and Y-3, Y-4, Y-5 other chromosomes do not. The X chromosome was incompatibilities. Slight female sterility was also not studied. detected in backcross. 14. (Zouros 1973, 1981.) Sympatric species, genetic 24. (SchÌfer 1978.) Sympatric species (D. hydei is a cosmo- analysis. Di¡erent pairs of chromosomes (and hence politan species associated with humans), genetic genes) are involved in sexual isolation among males analysis. F1 hybrids are fertile, but backcross hybrids versus females (see note 6). show severe sterility. Female sterility involves a 3^4 15. (Orr 1987, 1989b.) Sympatric species, genetic analysis. incompatibility; some hint of X^A incompatibilities Large X-e¡ect was observed in both male and female were also suggested, but not proven. Male sterility hybrid sterility. Genes causing male and female steri- involves 3^4, X^3, X^4,Y^3,Y^4, and Y^5 incompat- lity are probably di¡erent given the di¡erent locations ibilities. and e¡ects. See, also, Wu & Beckenbach (1983). 25. (SchÌfer 1979.) Genetic analysis, see note 24 for 16. (Noor 1997b.) Sympatric species, genetic analysis. In geographic distribution. Backcross hybrid inviability each of the two backcrosses, one X-linked and one involves chromosomes X, 2, 3, and 4. Chromosomes 5 autosomal gene a¡ect the male traits discriminated and 6 play no role. Thus, whereas a modest number of against by heterospeci¢c females. Because the two X- genes are involved, lethality is not truly polygenic. linked genes map to di¡erent locations, the minimum 26. (Patterson & Stone 1952.) Sympatric species, genetic estimate of between the species is analysis. F1 females having D. montana mothers are three loci. See, also, Tan (1946). inviable. Analysis showed that inviability results from 17. (Orr 1989a,b; H. A. Orr, unpublished data.) Genetic an incompatibility between a dominant X-linked analysis between USA populations and an allopatric factor from D. texana and recessive maternally acting isolate from Bogota, Colombia. Hybrid male sterility factor(s) from D. montana. Whereas the D. texana factor is mostly caused by X-autosomal incompatibilities. appears to be a single gene (mapped near the echinus Recent work shows that at least three genes are locus), nothing is known about the number or location involved on the Bogota X chromosome and two on of the maternally acting factor(s). the USA autosomes (H. A. Orr, unpublished data). 27. (Lamnissou et al. 1996.) Sympatric species (D. virilis is a Nonetheless, large regions of genome have no discern- cosmopolitan species associated with humans), genetic ible e¡ect on hybrid sterility (e.g. chromosome 4, half analysis. Male sterility appears only in backcrosses. of chromosome 2, Ychromosome, etc.). Sterility is primarily caused by a Ytex ^5vir incompat- 18. (Carvajal et al. 1996.) Sympatric species, genetic ibility, although milder Ytex ^2^3vir incompatibility analysis. Male inviability in backcrosses is caused also occurs. (2^3 is fused in D. texana; thus, 2^3 by at least two factors on the X chromosome of D. behaves as single linkage group in backcrosses.)

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 301

28. (Tomaru & Oguma 1994; Tomaru et al. 1995.) Sympa- 34. (Fenster et al. 1995.) Sympatric species, biometric tric species, genetic analysis. These species show strong analysis. Bud-growth rate and duration re£ect di¡er- sexual isolation in the laboratory and di¡er in male ence in £ower size between these species: M. courtship songs. Genetic analysis showed that both micranthus is largely sel¢ng and apparently derived major autosomes carried at least one gene a¡ecting from the outcrossing M. guttatus. It is not known the interpulse interval (IPI) of the male mating whether this di¡erence causes reproductive isolation songs, but the X chromosome had no e¡ect (Tomaro in nature, but this seems likely given the decrease in & Oguma 1994). In a later study, Tomaru et al. (1995) gene £ow caused by sel¢ng. No evidence for factors of tested D. biauraria females with conspeci¢c males large e¡ect. Lower bound of 95% con¢dence interval whose wings and antennae had been removed, and for gene number is 3.2 for bud growth rate and 4.6 for played synthetic courtship song during these encoun- duration of bud development. ters. Conspeci¢c matings occurred more frequently 35. (Fenster & Ritland 1994.) Biometric analysis. A total and female rejection behaviours less frequently, when of four taxa of controversial status, named M. arti¢cial songs contained the conspeci¢c IPI than guttatus, M. nasutus, (both outcrossing), and M. when they contained the longer IPI characteristic of micranthus and M. laciniatus (predominantly sel¢ng). D. auraria. This indicates that the song di¡erences The latter three taxa are sometimes classi¢ed as probably contribute to sexual isolation. of M. guttatus. Traits studied included 29. (Roelofs et al. 1987; Lofstedt et al. 1989.) Genetic di¡erences in £owering time, corolla length, corolla analysis, sympatric races of European corn borers in width, stamen level, pistil length, stigma^anther New York State, USA. Characters studied were female separation. Sel¢ng species have shorter and narrower pheromone blend (ratios of two long-chain acetates, corollas, shorter stamens and pistils, and less whose interspeci¢c di¡erence is apparently caused by stigma^anther separation than outcrossers. Flower a single autosomal locus). The electrophysiological characters are therefore associated with breeding response to these pheromones by male antennae systems. Minimum number of genes for character sensilla is caused by another, unlinked autosomal di¡erences averaged across all taxa varied between locus, and the behavioural response of male to female 5 and 13 per trait. Standard errors are large. Our pheromones is caused by a single X-linked factor. impression is that these phenotypic di¡erences 30. (Shaw 1996.) Allopatric species, biometric analysis. involve several to many genes, with no single locus Trait studied was male song pulse rate of two species causing most of the di¡erence in any character of Hawaiian crickets, which may be involved in between any two taxa. Within a cross, positive sexual isolation. Reciprocal F1 crosses indicate no genetic correlations were often seen between many disproportionate e¡ect of the X chromosome. traits, so that genes causing these character di¡er- 31. (Monti et al. 1997.) Sympatric species, genetic analysis. ences are not necessarily independent. Trait studied was ratio of two pheromones, apparently 36. (Macnair & Christie 1983; Christie & Macnair 1984, caused by a single factor. The timing of female emis- 1987.) Allopatric populations, genetic analysis. sion of the pheromones, which di¡ers between the Complementary lethal loci are polymorphic (one in species, appears to be polygenic, and the authors give each population) in two North American populations no estimate of number of factors. Male perception of of Mimulus guttatus. There are two separate systems of traits must, of course, also di¡er if there is to be inviability. The ¢rst involves only two genes, both reproductive isolation, so that, as in corn borers (see polymorphic for complementary lethals. The other note 28), sexual isolation must be caused by changes involves one locus (possibly the gene involved in in at least two genes. copper tolerance) that interacts with an unknown 32. (Wittbrodt et al. 1989.) Sympatric species, genetic number of genes in non-tolerant population. analysis. Hybrid inviability between X. helleri (swordtail) 37. (Macnair & Cumbes 1989.) Sympatric species, and X. maculatus (platy¢sh) is caused by appearance of biometric analysis. A total of seven £ower-size charac- malignant melanomas, which are often fatal. These ters studied, some of which (e.g. height, width, pistil cancers are caused by the interaction between a domi- length) are probably related to di¡erence in breeding nant, X-linked oncogene encoding a receptor systems between these species. M. cupriphilus selfs much tyrosinane kinase, and an autosomal suppressor that is more often than does M. guttatus, and this di¡erence in either missing in the swordtail or dominant in the platy- breeding systems may contribute to reproductive isola- ¢sh. Some backcross hybrids inherit the oncogene tion. Stamen^pistil length ratio is also important in without the suppressor, yielding melanomas. reproductive isolation, but it was not studied genetically. 33. (Bradshaw et al. 1995.) Sympatric species, genetic Each character di¡erence was caused by between three analysis. Traits studied by QTL analysis include and seven genes. High genetic correlations were £ower colour, corolla and petal width, nectar volume observed between many characters, so that traits are and concentration, and stamen and pistil length. All of not genetically independent. these traits a¡ect whether a £ower is pollinated by 38. (Rieseberg 1998.) Sympatric species, genetic analysis. bumblebees (M. lewisii) or hummingbirds (M. cardinalis). Within the colinear portions of the genome of these Species are at least partly reproductively isolated by two sun£ower species, Rieseberg estimates that at pollinator di¡erence. For most traits, the species di¡er- least 14 chromosomal segments are responsible for ence involved at least one chromosome region of large inviability of hybrid pollen. Thus, approximately 40 e¡ect. The di¡erence in carotenoids in petal lobes was genes are involved if one assumes a similar density of governed by a single gene. factors in the rearranged portions of the genome.

Phil. Trans. R. Soc. Lond. B (1998) 302 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation

Coyne, J. A. 1993a Recognizing species. Nature 364, 298. REFERENCES Coyne, J. A. 1993b The genetics of an isolating mechanism Arnold, S. J., Verrell, P. A. & Tilley, S. G. 1996 The evolution of between two sibling species of Drosophila. Evolution 47, asymmetry in sexual isolation: a model and a test case. 778^788. Evolution 50, 1024^1033. Coyne, J. A. 1994 and the origin of species. Evolution Ashburner, M. 1989 Drosophila: a laboratory handbook. New York: 48, 19^30. Cold Spring Harbor Laboratory Press. Coyne, J. A. 1996a Genetics of di¡erences in pheromonal hydro- Barraclough,T. G., Harvey, P. H. & Nee, S. 1995 Sexual selection carbons between Drosophila melanogaster and D. simulans. Genetics and taxonomic diversity in birds. Proc. R. Soc. Lond. B 143, 353^364. 259, 211^215. Coyne, J. A. 1996b Genetics of sexual isolation in male hybrids of Barton, N. H. & Charlesworth, B. 1984 Genetic revolutions, Drosophila simulans and D. mauritiana. Genet. Res. 68, 211^220. founder e¡ects, and speciation. A. Rev. Ecol. Syst. 15, 133^164. Coyne, J. A. & Charlesworth, B. 1986 Location of an X-linked Bateson, W. 1909 Heredity and variation in modern lights. In factor causing male sterility in hybrids of Drosophila simulans Darwin and modern science (ed. A. C. Seward), pp. 85^101. and D. mauritiana. Heredity 57, 243^246. Cambridge University Press. Coyne, J. A. & Charlesworth, B. 1989 Genetic analysis of X- Baum, D. & Shaw, K. L. 1995 Genealogical perspectives on the linked sterility in hybrids between three sibling species of species problem. Monogr. Syst. Bot. 53, 289^303. Drosophila. Heredity 62, 97^106. Bella, J. L., Butlin, R. K., Ferris, C. & Hewitt, G. M. 1992 Coyne, J. & Charlesworth, B. 1997 Genetics of a pheromonal Asymmetrical homogamy and unequal sex ratio from reci- di¡erence a¡ecting sexual isolation between Drosophila maur- procal mating-order crosses between Chorthippus parallelus itiana and D. sechellia. Genetics 145, 1015^1030. subspecies. Heredity 68, 345^352. Coyne, J. A. & Kreitman, M. 1986 Evolutionary genetics of two Bradshaw, H. D., Wilbert, S. M., Otto, K. G. & Schemske, D. W. sibling species, Drosophila simulans and D. sechellia. Evolution 40, 1995 Genetic mapping of £oral traits associated with 673^691. reproductive isolation in monkey£owers (Mimulus). Nature 376, Coyne, J. A. & Orr, H. A. 1989a Two rules of speciation. In 762^765. Speciation and its consequences (ed. D. Otte & J. Endler.), Breeuwer, J. A. J. & Werren, J. H. 1995 Hybrid breakdown pp. 180^207. Sunderland, MA: Sinauer Associates. between two haplodiploid species: the role of nuclear and Coyne, J. A. & Orr, H. A. 1989b Patterns of speciation in cytoplasmic genes. Evolution 49, 705^717. Drosophila. Evolution 43, 362^381. Butlin, R. 1987 Speciation by reinforcement. Trends Ecol. Evol. 2, Coyne, J. A. & Orr, H. A. 1997 Patterns of speciation in 8^13. Drosophila revisited. Evolution 51, 295^303. Cabot, E. L., Davis, A. W., Johnson, N. A. & Wu, C.-I. 1994 Coyne, J. A., Orr, H. A. & Futuyma, D. J. 1988 Do we need a Genetics of reproductive isolation in the Drosophila simulans new species concept? Syst. Zool. 37, 190^200. : complex epistasis underlying hybrid male sterility. Coyne, J. A., Charlesworth, B. & Orr, H. A. 1991 Haldane's rule Genetics 137, 175^189. revisited. Evolution 45, 1710^1713. Carvajal, A. R., Gandarela, M. R. & Naveira, H. F. 1996 A Coyne, J. A., Barton, N. H. & Turelli, M. 1997 A critique of three-locus system of interspeci¢c incompatibility underlies Sewall Wright's shifting balance theory of evolution. Evolution male inviability in hybrids betwen Drosophila buzzattii and 51, 643^671. D. koepferi. Genetica 98, 1^19. Cracraft, J. 1989 Speciation and its ontology: the empirical Charlesworth, B., Coyne, J. A. & Barton, N. 1987 The relative consequences of alternative species concepts for understanding rates of evolution of sex chromosomes and autosomes. Am. Nat. patterns and processes of di¡erentiation. In Speciation and its 130, 113^146. consequences (ed. D. Otte & J. A. Endler), pp. 28^59. Christie, P. & Macnair, M. R. 1984 Complementary lethal Sunderland, MA: Sinauer Associates. factors in two North American populations of the yellow David, J., Bocquet, C., Lemeunier, F. & Tsacas, L. 1974 monkey £ower. J. Hered. 75, 510^511. Hybridation d'une nouvelle espe¨ ce Drosophila mauritiana avec Christie, P. & Macnair, M. R. 1987 The distribution of post- D. melanogaster et D. simulans. Ann. Gene¨ t. 17, 235^241. mating reproductive isolating genes in populations of the Davies, N., Aiello, A., Mallet, J., Pomiankowski, A. & yellow monkey £ower, Mimulus guttatus. Evolution 41, 571^578. Silberglied, R. E. 1997 Speciation in two neotropical butter- Cobb, M., Burnet, B., Blizard, R. & Jallon, J.-M. 1989 Courtship £ies: extending Haldane's rule. Proc. R. Soc. Lond. B 264, in Drosophila sechellia: its structure, functional aspects, and rela- 845^851. tionship to those of other members of the Drosophila melanogaster Davis, A. W. & Wu, C.-I. 1996 The broom of the sorcerer's species subgroup. J. Insect Behav. 2, 63^89. apprentice: the ¢ne structure of a chromosomal region Cohan, F. M. 1998 Genetic structure of bacterial populations. In causing reproductive isolation between two sibling species of Evolutionary genetics from molecules to morphology (ed. R. Singh & Drosophila. Genetics 143, 1287^1298. C. Krimbas). Cambridge University Press. (In the press.) Davis, A. W., Noonburg, E. G. & Wu, C. I. 1994 Evidence for Coulthart, M. D. & Singh, R. S. 1988 High level of divergence of complex genetic interactions between conspeci¢c chromo- male reproductive tract proteins between Drosophila somes underlying hybrid female sterility in the Drosophila melanogaster and its sibling species, D. simulans. Molec. Biol. simulans clade. Genetics 137, 191^199. Evol. 5, 182^191. Davis, A. W., Roote, J., Morley,T., Sawamura, K., Herrmann, S. Coyne, J. A. 1983 Genetic basis of di¡erences in genital & Ashburner, M. 1996 Rescue of hybrid sterility in crosses morphology among three sibling species of Drosophila. between D. melanogaster and D. simulans. Nature 380, 157^159. Evolution 37, 1101^1118. Dobzhansky, T. 1933 On the sterility of the interracial hybrids in Coyne, J. A. 1989 The genetics of sexual isolation between two Drosophila pseudoobscura. Proc. Natn. Acad. Sci. USA 19, 397^403. sibling species, Drosophila simulans and Drosophila mauritiana. Dobzhansky, T. 1935 A critique of the species concept in biology. Proc. Natn. Acad. Sci. USA 86, 5464^5468. Phil. Sci. 2, 344^345. Coyne, J. A. 1992a Much ado about species. Nature 357, 289^290. Dobzhansky, T. 1937 Genetics and the origin of species. New York: Coyne, J. A. 1992b Genetics of sexual isolation in females of the Columbia University Press. Drosophila simulans . Genet. Res. Camb. 60, 25^31. Eberhard, W. G. 1985 Sexual selection and animal genitalia. Coyne, J. A. 1992c Genetics and speciation. Nature 355, 511^515. Cambridge, MA: Harvard University Press.

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 303

Eberhard, W. G. 1996 Female control: sexual selection by cryptic female King, M. 1993 Species evolution. Cambridge University Press. choice. Princeton University Press. Kirkpatrick, M. 1996 Good genes and direct selection in the Ehrman, L. 1965 Direct observation of sexual isolation between evolution of mating preferences. Evolution 50, 2125^2140. allopatric and between sympatric strains of the di¡erent Kirkpatrick, M. & Barton, N. 1995 De¨ ja© vu all over again. Nature Drosophila paulistorum races. Evolution 19, 459^464. 377, 388^389. Endler, J. A. & Houde, A. E. 1995 Geographic variation in Lamnissou, K., M. Loukas, M. & Zouros, E. 1996 Incom- female preferences for male traits in Poecilia reticulata. Evolution patibilities between Ychromosome and autosomes are respon- 49, 456^468. sible for male hybrid sterility in crosses between Drosophila Fenster, C. B. & Ritland, K. 1994 Quantitative genetics of mating virilis and Drosophila texana. Heredity 76, 603^609. system divergence in the yellow monkey£ower species Lande, R. 1979 E¡ective sizes during long-term evolution complex. Heredity 73, 422^435. estimated from rates of chromosomal rearrangement. Evolution Fenster, C. B., Diggle, P. K., Barrett, S. C. H. & Ritland, K. 1995 33, 234^251. The genetics of £oral development di¡erentiating two species Lande, R. 1981 Models of speciation by sexual selection on poly- of Mimulus (Scrophulariaceae). Heredity 74, 258^266. genic traits. Proc. Natn. Acad. Sci. USA 78, 3721^3725. Fisher, R. A. 1930 The genetical theory of natural selection. Oxford Lande, R. 1982 Rapid origin of sexual isolation and character University Press. divergence in a cline. Evolution 36, 213^223. Gilliard, E. T. 1956 Bower ornamentation versus plumage char- Liou, L. W. & Price, T. D. 1994 Speciation by reinforcement of acters in bower-birds. Auk 73, 450^451. prezygotic isolation. Evolution 48, 1451^1459. Grant, B. R. & Grant, P. R. 1996 Cultural inheritance of song Lofstedt, C., Hansson, B. S., Roelofs,W. & Bengtsson, B. O. 1989 and its role in the evolution of Darwin's ¢nches. Evolution 50, No linkage between genes controlling female pheromone 2471^2487. production and male pheromone response in the European Grant, V. 1981 Plant speciation, 2nd edn. New York: Columbia corn borer, Ostrinia nubialis HÏbner (Lepidoptera; Pyralideae). University Press. Genetics 123, 553^556. Gregory, P. G. & Howard, D. J. 1994 A postinsemination barrier Macnair, M. R. & Christie, P. 1983 Reproductive isolation as a to fertilization isolates two closely related ground crickets. pleiotropic e¡ect of copper tolerance in Mimulus guttatus. Evolution 48, 705^710. Heredity 50, 295^302. Haldane, J. B. S. 1922 Sex-ratio and unisexual sterility in hybrid Macnair, M. R. & Cumbes, Q. J. 1989 The genetic architecture animals. J. Genet. 12, 101^109. of interspeci¢c variation in Mimulus. Genetics 122, 211^222. Haskins, C. P. & Haskins, E. F. 1949 The role of sexual selection Marin, I. 1996 Genetic architecture of autosome-mediated as an isolating mechanism in three species of poeciliid ¢shes. hybrid male sterility in Drosophila. Genetics 142, 1169^1180. Evolution 3, 160^169. Masterson, J. 1994 Stomatal size in fossil plants: evidence Heikkinen, E. & Lumme, J. 1991 Sterility of male and female for in majority of angiosperms. Science 264, hybrids of Drosophila virilis and Drosophila lummei. Heredity 67, 421^424. 1^11. Maynard Smith, J. & Szathma¨ ry, E. 1995 The major transitions in Ho¡mann, A. A. & Turelli, M. 1997 Cytoplasmic incompatibility evolution. Oxford: W. H. Freeman/Spektrum. in insects. In In£uential passengers: inherited microorganisms and Mayr, E. 1942 and the origin of species. New York: reproduction (ed. S. L. O'Neill, J. H. Werren & A. A. Columbia University Press. Ho¡mann), pp. 42^80. Oxford University Press. Menotti-Raymond, M., David,V. A. & O'Brien, S. J. 1997 Pet cat Ho¡mann, A. A., Turelli, M. & Simmons, G. M. 1986 implicates murder suspect. Nature 386, 774. Unidirectional incompatibility between populations of Metz, E. C. & Palumbi, S. R. 1996 Positive selection and Drosophila simulans. Evolution 40, 692^701. sequence rearrangements generate extensive in Hoikkala, A. & Lumme, J. 1984 Genetic control of the di¡erence the gamete recognition protein binding. Molec. Biol. Evol. 13, in male courtship sound between D. virilis and D. lummei. Behav. 397^406. Genet. 14, 827^845. Mitra, S., Landel, H. & Pruett-Jones, S. 1996 Species richness Hollocher, H. & Wu, C.-I. 1996 The genetics of reproductive covaries with mating system in birds. Auk 113, 544^551. isolation in the Drosophila simulans clade: X versus autosomal Mitrofanov,V. G. & Sidorova, N.V. 1981 Genetics of the sex ratio e¡ects and male vs. female e¡ects. Genetics 143, 1243^1255. anomaly in Drosophila hybrids of the virilis group. Theor. Appl. Holman, E. W. 1987 Recognizability of sexual and asexual Genet. 59, 17^22. species of rotifers. Syst. Zool. 36, 381^386. Monti, L., Ge¨ nermont, J., Malosse, C. & Lalanne-Cassou, B. Howard, D. J. 1993 Reinforcement: origin, dynamics, and fate of 1997 A genetic analysis of some components of reproductive an evolutionary hypothesis. Hybrid zones and the evolutionary process isolation between two closely related species, Spodoptera (ed. R. G. Harrison), pp. 46^69. Oxford University Press. latifascia (Walker) & S. descoinsi (Lalanne-Cassou and Silvain) Hutter, P. & Ashburner, M. 1987 Genetic rescue of inviable (Lepidoptera: Noctuidae.) J. Evol. Biol. 10, 121^143. hybrids between Drosophila melanogaster and its sibling species. Muller, H. J. 1942 Isolating mechanisms, evolution, and Nature 327, 331^333. temperature. Biol. Symp. 6, 71^125. Hutter, P., Roote, J. & Ashburner, M. 1990 A genetic basis for Naveira, H. & Fontdevila, A. 1986 The evolutionary history of the inviability of hybrids between sibling species of Drosophila. Drosophila buzzatii. The genetic basis of sterility in hybrids Genetics 124, 909^920. between D. buzzattii and its sibling D. serido from Argentina. Iwasa, Y. & Pomiankowski, A. 1995 Continual change in mate Genetics 114, 841^857. preferences. Nature 377, 420^422. Naveira, H. & Fontdevila, A. 1991 The evolutionary history of Kaneshiro, K.Y. 1980 Sexual isolation, speciation, and the direc- Drosophila buzzatii. XXI. Cumulative action of multiple steri- tion of evolution. Evolution 34, 437^444. lity factors on spermatogenesis in hybrids of D. buzzatii and Kelly, J. K. & Noor, M. A. F. 1996 Speciation by reinforcement: D. koepferae. Heredity 67, 57^72. a model derived from studies of Drosophila. Genetics 143, Nei, M. 1972 Genetic distances between populations. Am. Nat. 1485^1497. 106, 282^292. Khadem, M. & Krimbas, C. B. 1991 Studies of the species barrier Nei, M. 1976 Mathematical models of speciation and genetic between Drosophila subobscura and D. madeirensis. 1. The genetics distance. In Population genetics and ecology (ed. S. Karlin & of male hybrid sterility. Heredity 67, 157^165. E. Nevo), pp. 723^766. New York: Academic Press.

Phil. Trans. R. Soc. Lond. B (1998) 304 J. A. Coyne and H. A. Orr Evolutionary genetics ofspeciation

Noor, M. 1995 Speciation driven by natural selection in Ramsey, J. M. & Schemske, D. W. 1998 The dynamics of poly- Drosophila. Nature 375, 674^675. ploid formation and establishment in £owering plants. A. Rev. Noor, M. A. F. 1997a How often does sympatry a¡ect sexual Ecol. Syst. (In the press.) isolation in Drosophila? Am. Nat. 149, 1156^1163. Rice,W. R. 1996 Sexually antagonistic male adaptation triggered Noor, M. A. F. 1997b Genetics of sexual isolation and courtship by experimental arrest of female evolution. Nature 381, dysfunction in male hybrids of Drosophila pseudoobscura and 232^234. D. persimilis. Evolution 51, 809^815. Rice, W. R. & Hostert, E. E. 1993 Laboratory experiments on O'Neill, S. L. & Karr, T. L. 1990 Bidirectional incompatibility speciation: what have we learned in 40 years? Evolution 47, between conspeci¢c populations of Drosophila simulans. Nature 1637^1653. 348, 178^180. Rieseberg, L. H. 1998 Genetic mapping as a tool for studying Orr, H. A. 1987 Genetics of male and female sterility in hybrids speciation. In Molecular systematics of plants, 2nd edn (ed. D. E. of Drosophila pseudoobscura and D. persimilis. Genetics 116, Soltis, P. S. Soltis & J. J. Doyle). New York: Chapman & Hall. 555^563. (In the press.) Orr, H. A. 1989a Genetics of sterility in hybrids between two Rieseberg, L. H., Desrochers, A. M. & Youn, S. J. 1995 subspecies of Drosophila. Evolution 43, 180^189. Interspeci¢c pollen competition as a reproductive barrier Orr, H. A. 1989b Localization of genes causing postzygotic isola- between sympatric species of Helianthus (Asteraceae). Am. J. tion in two hybridizations involving Drosophila pseudoobscura. Bot. 82, 515^519. Heredity 63, 231^237. Ringo, J. M. 1977 Why 300 species of Hawaiian Drosophila? The Orr, H. A. 1992 Mapping and characterization of a `speciation sexual selection hypothesis. Evolution 31, 694^696. gene' in Drosophila. Genet. Res. 59, 73^80. Roberts, M. S. & Cohan, F. M. 1995 Recombination and migra- Orr, H. A. 1993a Haldane's rule has multiple genetic causes. tion rates in natural populations of Bacillus subtilis and Bacillus Nature 361, 532^533. mojavensis. Evolution 49, 1081^1094. Orr, H. A. 1993b A mathematical model of Haldane's rule. Roberts, M. S., Nakamura, K. L. & Cohan, F. M. 1996 Bacillus Evolution 47, 1606^1611. vallismortis sp. nov., a close relative of Bacillus subtilis, isolated Orr, H. A. 1995 The population genetics of speciation: the evolu- from soil in Valley, California. Int. J. Syst. Bacteriol. 46, tion of hybrid incompatibilities. Genetics 139, 1805^1813. 470^475. Orr, H. A. 1997 Haldane's rule. A. Rev. Ecol. Syst. 28, 195^218. Roelofs, W., Glover, T., Tang, X.-H., Srent, I., Robbins, P., Orr, H. A. & Coyne, J. A. 1989 The genetics of postzygotic isola- Eckenrode, C., Lofstedt, C., Hansson, B. S. & Bengtson, B. tion in the Drosophila virilis group. Genetics 121, 527^537. 1987 Sex pheromone production and perception in European Orr, H. A. & Coyne, J. A. 1992 The genetics of adaptation: a corn borer moths is determined by both autosomal and sex- reassessment. Am. Nat. 140, 725^742. linked genes. Proc. Natn. Acad. Sci. USA 84, 7585^7589. Orr, H. A. & Orr, L. H. 1996 Waiting for speciation: the e¡ect of Ruiz, A. & Heed, W. B. 1988 Host-plant speci¢city in the population subdivision on the time to speciation. Evolution 50, cactophilic Drosophila mulleri species complex. J. Anim. Ecol. 1742^1749. 57, 237^249. Orr, H. A. & Turelli, M. 1996 Dominance and Haldane's rule. Saetre, G.-P., Moum, T., Bures, S., Kra¨ l, M., Adamjan, M. & Genetics 143, 613^616. Moreno, J. 1997 A sexually selected in Orr, H. A., Madden, L. D., Coyne, J. A., Goodwin, R. & £ycatchers reinforces premating isolation. Nature 387, 589^592. Hawley, R. S. 1997 The developmental genetics of hybrid Sawamura, K., Taira, T. & Watanabe, T. K. 1993a Hybrid lethal inviability: a mitotic defect in Drosophila hybrids. Genetics 145, systems in the Drosophila melanogaster species complex. I. The 1031^1040. maternal hybrid rescue (mhr) gene of Drosophila simulans. Genetics Pantazidis, A. C., Galanopoulos, V. K. & Zouros, E. 1993 An 133, 299^305. autosomal factor from Drosophila arizonae restores normal sper- Sawamura, K., Watanabe, T. K. & Yamamoto, M.-T. 1993b matogenesis in Drosophila mojavensis males carrying the Hybrid lethal systems in the Drosophila melanogaster species D. arizonae Ychromosome. Genetics 134, 309^318. complex. Genetica 88, 175^185. Patterson, J. T. 1946 A new type of isolating mechanism in Sawamura, K., Yamamoto, M.-T. & Watanabe, T. K. 1993c Drosophila. Proc. Natn. Acad. Sci. USA 32, 202^208. Hybrid lethal systems in the Drosophila melanogaster Perez, D. E., Wu, C.-I., Johnson, N. A. & Wu, M.-L. 1993 species complex. II. The Zygotic hybrid rescue (Zhr) gene of Genetics of reproductive isolation in the Drosophila simulans D. melanogaster. Genetics 133, 307^313. clade: DNA-marker assisted mapping and characterization of SchÌfer, U. 1978 Sterility in Drosophila hydeiÂD. neohydei hybrids. a hybrid-male sterility gene, Odysseus (Ods). Genetics 134, Genetica 49, 205^214. 261^275. SchÌfer, U. 1979 Viability in Drosophila hydeiÂD. neohydei hybrids Perrot-Minnot, M.-J., Guo, L. R. & Werren, J. H. 1996 Single and its regulation by genes located in the sex heterochromatin. and double infections with Wolbachia in the parasitic wasp Biol. Zentralblatt 98, 153^161. Nasonia vitripennis: e¡ects on compatibility. Genetics 143, Schliewen, U. K., Tautz, D. & PÌÌbo, S. 1994 Sympatric specia- 961^972. tion suggested by of crater lake . Nature 368, Pontecorvo, G. 1943 Viability interactions between chromosomes 629^632. of Drosophila melanogaster and Drosophila simulans. J. Genet. 45, Schluter, D. 1996 Ecological causes of adaptive radiation. Am. 51^66. Nat. 148, S40^S64. Prager, E. R. & Wilson, A. C. 1975 Slow evolutionary loss of the Schluter, D. 1998 Ecological causes of speciation. In Endless forms: potential for interspeci¢c hybridization in birds: a manifesta- species and speciation (ed. D. J. Howard & S. H. Berlocher). tion of slow regulatory evolution. Proc. Natn. Acad. Sci. USA 72, Oxford University Press. (In the press.) 200^204. Schluter, D. & Price, T. 1993 Honesty, perception, and popula- Prazmo, W. 1965 Cytogenetic studies on the genus Aquilegia. III. tion divergence in sexually selected traits. Proc. R. Soc. Lond. B Inheritance of the traits distinguishing di¡erent complexes in 253, 117^122. the genus Aquilegia. Acta Soc. Bot. Poloniae 34, 403^437. Shaw, K. L. 1996 Polygenic inheritance of a behavioral pheno- Price, C. S. C. 1997 Conspeci¢c sperm precedence in Drosophila. type: interspeci¢c genetics of song in the Hawaiian cricket Nature 388, 663^666. genus Laupala. Evolution 50, 256^266.

Phil. Trans. R. Soc. Lond. B (1998) Evolutionary genetics ofspeciation J. A. Coyne and H. A. Orr 305

Somerson, N. L., Ehrman, L., Kocka, J. P. & Gottlieb, F. J. 1984 Val, F. C. 1977 Genetic analysis of the morphological di¡erences Streptococcal L- forms isolated from Drosophila paulistorum between two interfertile species of Hawaiian Drosophila. semispecies cause sterility in male progeny. Proc. Natn. Acad. Evolution 31, 611^629. Sci. USA 81, 282^285. Vigneault, G. & Zouros, E. 1986 The genetics of asymmetrical Spencer, H. G., McArdle, B. H. & Lambert, D. M. 1986 A theo- male sterility in Drosophila mojavensis and Drosophila arizonensis retical investigation of speciation by reinforcement. Am. Nat. hybrids: interaction between the Y chromosome and auto- 128, 241^262. somes. Evolution 40, 1160^1170. Stratton, G. E. & Uetz, G. W. 1986 The inheritance of courtship Wade, M. J., Patterson, H., Chang, N. W. & Johnson, N. 1994 behavior and its role as a reproductive isolating mechanism in Postcopulatory, prezygotic isolation in £our beetles. Heredity two species of Schizocosa wolf spiders (Araneae: Lycosidea). 72, 163^167. Evolution 40, 129^141. Walsh, J. B. 1982 Rate of accumulation of reproductive isolation Stevens, L. & Wade, M. J. 1990 Cytoplasmically inherited repro- by chromosome rearrangements. Am. Nat. 120, 510^532. ductive incompatibility in Tribolium £our beetles: the rate of Wasserman, A. O. 1957 Factors a¡ecting interbreeding in sympa- spread and e¡ect on population size. Genetics 124, 367^372. tric species of spadefoots (genus Scaphiopus). Evolution 11, Sturtevant, A. H. 1920 Genetic studies on Drosophila simulans. I. 320^338. Introduction. Hybrids with Drosophila melanogaster. Genetics 5, Wasserman, M. & Koepfer, H. R. 1977 Character displacement 488^500. for sexual isolation between Drosophila mojavensis and Drosophila Takamura, T. & Watanabe, T. K. 1980 Further studies on the arizonensis. Evolution 31, 812^823. lethal hybrid rescue (Lhr) gene of Drosophila simulans. Jap. J. Watanabe, T. K. 1979 A gene that rescues the lethal hybrids Genet. 55, 405^408. between Drosophila melanogaster and D. simulans. Jap. J. Genet. Tan, C. C. 1946 Genetics of sexual isolation between Drosophila 54, 325^331. pseudoobscura and Drosophila persimilis. Genetics 31, 558^573. Watanabe, T. K. & Kawanishi, M. 1979 Mating preference and Templeton, A. R. 1977 Analysis of head shape di¡erences the direction of evolution in Drosophila. Science 205, 906^907. between two interfertile species of Hawaiian Drosophila. Werren, J. H. 1997 Wolbachia and speciation. In Endless forms: Evolution 31, 630^641. species and speciation (ed. D. Howard & S. Berlocher). Oxford Templeton, A. R. 1981 Mechanisms of speciationöa population University Press. genetics approach. A. Rev. Ecol. Syst. 12, 23^48. White, M. J. D. 1969 Chromosomal rearrangements and specia- Tomaru, M. & Oguma, Y. 1994 Genetic basis and evolution of tion in animals. A. Rev. Genet. 3, 75^98. species-speci¢c courtship song in the Drosophila auraria White, M. J. D. 1978 Modes of speciation. San Francisco: W. H. complex. Genet. Res. Camb. 63, 11^17. Freeman & Co. Tomaru, M., Matsubayashi, H. & Oguma,Y. 1995 Heterospeci¢c Wittbrodt, J., Adam, D., Malitschek, B., Mqaueler, W., Raulf, F., inter-pulse intervals of courtship song elicit female rejection in Telling, A., Robertson, S. M. & Schartl, M. 1989 Novel Drosophila biauraria. Anim. Behav. 50, 905^914. putative receptor tyrsosine kinase encoded by the melanoma- True, J. R., B. S., Weir & Laurie, C. C. 1996 A genome-wide inducingTu locus in Xiphophorus. Nature 341, 415^421. survey of hybrid incompatibility factors by the introgression Wu, C.-I. & Beckenbach, A. T. 1983 Evidence for extensive of marked segments of Drosophila mauritiana chromosomes into genetic di¡erentiation between the sex-ratio and the standard Drosophila simulans. Genetics 144, 819^837. arrangement of Drosophila pseudoobscura and D. persimilis and True, J. R., Liu, J., Stam, L. F., Zeng, Z.-B. & Laurie, C. C. 1997 identi¢cation of hybrid sterility factors. Genetics 105, 71^86. Quantitative genetic analysis of divergence in male secondary Wu, C.-I. & Davis, A. W. 1993 Evolution of postmating repro- sexual traits between Drosophila simulans and D. mauritiana. ductive isolation: the composite nature of Haldane's rule and Evolution 51, 816^832. its genetic bases. Am. Nat. 142, 187^212. Tsaur, S.-C. & Wu, C.-I. 1997 Positive selection and the mole- Wu, C.-I., Johnson, N. & Palopali, M. F. 1996 Haldane's rule and cular evolution of a gene of male reproduction, Acp26Aa of its legacy: why are there so many sterile males? Trends Ecol. Drosophila. Molec. Biol. Evol. 14, 544^549. Evol. 11, 281^284. Turelli, M. & Ho¡mann, A. A. 1995 Cytoplasmic incompat- Zink, R. M. & McKitrick, M. C. 1995 The debate over species ibility in Drosophila simulans: dynamics and parameter concepts and its implications for ornithology. Auk 112, 701^719. estimates from natural populations. Genetics 140, 1319^1338. Zouros, E. 1973 Genic di¡erentiation associated with the early Turelli, M. & Orr, H. A. 1995 The dominance theory of stages of speciation in the mulleri subgroup of Drosophila. Haldane's rule. Genetics 140, 389^402. Evolution 27, 601^621. Turner, G. F. & Burrows, M. T. 1995 A model of sympatric Zouros, E. 1981 The chromosomal basis of sexual isolation in two speciation by sexual selection. Proc. R. Soc. Lond. B 260, sibling species of Drosophila: D. arizonensis and D. mojavensis. 287^292. Genetics 97, 703^718.

Phil. Trans. R. Soc. Lond. B (1998)