Evolution, 56(11), 2002, pp. 2262±2277

GENETICS OF A DIFFERENCE IN PIGMENTATION BETWEEN DROSOPHILA YAKUBA AND DROSOPHILA SANTOMEA

ANA LLOPART,1,2 SUSANNAH ELWYN,1,3 DANIEL LACHAISE,4,5 AND JERRY A. COYNE1,6 1Department of Ecology and Evolution, The University of Chicago, 1101 East 57 Street, Chicago, Illinois 60637 2E-mail: [email protected] 3E-mail: [email protected] 4Laboratoire Populations, GeÂneÂtique et Evolution, CNRS, 91198 Gif-sur-Yvette, France 5E-mail: [email protected] 6E-mail: [email protected]

Abstract. Drosophila yakuba is a species widespread in Africa, whereas D. santomea, its newly discovered sister species, is endemic to the volcanic island of SaÄo Tome in the Gulf of Guinea. Drosophila santomea probably formed after colonization of the island by its common ancestor with D. yakuba. The two species differ strikingly in pigmentation: D. santomea, unlike the other eight species in the D. melanogaster subgroup, almost completely lacks dark abdominal pigmentation. D. yakuba shows the sexually dimorphic pigmentation typical of the group: both sexes have melanic patterns on the abdomen, but males are much darker than females. A genetic analysis of this species difference using morphological markers shows that the X chromosome accounts for nearly 90% of the species difference in the area of abdomen that is pigmented and that at least three genes (one on each major chromosome) are involved in each sex. The order of chromosome effects on pigmentation area are the same in males and females, suggesting that loss of pigmentation in D. santomea may have involved the same genes in both sexes. Further genetic analysis of the interspeci®c difference between males in pigmentation area and intensity using molecular markers shows that at least ®ve genes are responsible, with no single locus having an overwhelming effect on the trait. The species difference is thus oligogenic or polygenic. Different chromosomal regions from each of the two species in¯uenced pigmentation in the same direction, suggesting that the species difference (at least in males) is due to natural or sexual selection and not genetic drift. Measurements of sexual isolation between the species in both light and dark conditions show no difference, suggesting that the pigmentation difference is not an important cue for interspeci®c mate discrimination. Using DNA sequence differences in nine noncoding regions, we estimate that D. santomea and D. yakuba diverged about 400,000 years ago, a time similar to the divergences between two other well-studied pair of species in the subgroup, both of which also involved island colonization.

Key words. Drosophila, genetics, pigmentation, reproductive isolation, speciation.

Received February 22, 2002. Accepted August 15, 2002.

Recent genetic studies of reproductive isolation are begin- chromosomes? Rice (1984), for example, posits that species ning to reveal patterns that may help us understand speciation differences that evolved by antagonistic coevolution between (Coyne and Orr 1999). However, there are still relatively few males and females may be caused by genes preferentially genetic studies of species differencesÐthose morphological, located on the X chromosome, as will genes ®xed by natural physiological, and behavioral traits that distinguish closely selection whose favorable effects are partially recessive related species but do not necessarily cause reproductive iso- (Charlesworth et al. 1987). lation. A recent survey (Orr 2001) describes only 13 genetic Orr (2001) describes other ways that the genetic basis of studies of species differences, six in the genus Drosophila species differences can help us understand evolution. Such and four in the plant genus Mimulus. analyses are likely to become more important with the in- Such work, however, is important in answering long-stand- creasing use of quantitative-trait-locus (QTL) mapping, ing questions of evolutionary genetics. For example, are spe- which can be used to locate and estimate the effects of chro- cies differences caused by natural selection or other evolu- mosome regions affecting traits between any pair of crossable tionary forces such as genetic drift? Theoretical work shows species that meet two criteria: (1) hybrids are somewhat fer- that if alleles from one species act in a consistent direction tile; and (2) one can construct a molecular map based on on a trait, then the species difference is likely to have evolved DNA differences (for an exemplar of the use of QTL mapping by natural or sexual selection (Orr 1998b). Are such differ- in evolution see Bradshaw et al. 1998; Schemske and Brad- ences due to many genes of small effect, as Fisher (1930) shaw 1999). posited, or are fewer genes of larger effect involved? For Here we present a genetic analysis of a striking character example if species differences in male-limited traits such as differenceÐthe degree of abdominal pigmentationÐbetween plumage color prove to be highly polygenic, this would imply two closely related species of Drosophila, D. yakuba and D. that the trait evolved by gradual coevolution of males and santomea. Drosophila yakuba is widespread across western females (Coyne and Orr 1999). Theoretical studies show, Africa, but D. santomea, discovered in 1998, is endemic to however, that evolution by natural selection toward a ®xed the island of SaÄo TomeÂ, an 860-km2 volcanic island 320 km optimum should lead to an exponential distribution of gene west of Gabon (Lachaise et al. 2000). Drosophila yakuba also effects with many factors having small effects, but also with inhabits SaÄo TomeÂ, but tentative molecular evidence points some factors having relatively large effects (Orr 1998a). Do to D. santomea originating allopatrically after a colonization genes for species differences tend to be located on particular event by the ancestor of modern D. yakuba, with D. yakuba 2262 ᭧ 2002 The Society for the Study of Evolution. All rights reserved. PIGMENTATION DIFFERENCES IN DROSOPHILA 2263 subsequently invading the island a second time (Cariou et al. F1 hybrids, the relative effects of individual chromosome 2001). On SaÄo TomeÂ, D. yakuba is limited to lower elevations, segments could not be discerned because a lack of genetic whereas D. santomea lives in the mist forests at higher ele- markers in these species necessitated a biometrical analysis vations. A hybrid zone occurs between 1150 m and 1450 m (Hollocher et al. 2000b). elevation on the volcano, where one ®nds a low frequency Kopp et al. (2000) suggest that, in Drosophila, differences of hybrids (about 1%; Lachaise et al. 2000). in sexual dimorphism of pigmentation among species may Molecular phylogenetic analysis indicates that D. yakuba be based on changes at the bric-aÁ-brac (bab) locus, which and D. santomea are sister species within the eight species contains two adjacent genes (bab1 and bab2) whose wild- constituting the monophyletic D. melanogaster subgroup (La- type alleles repress male-speci®c pigmentation in females of chaise et al. 2000; Cariou et al. 2001). This pair thus rep- D. melanogaster. Their hypothesis rests on a correlation: most resents a speciation event independent of the well-studied species with darkly pigmented males show no expression of speciation event separating D. simulans and D. melanogaster bab in males (thus permitting the sex-speci®c pigmentation), and of the two speciation events in which the ancestor of D. whereas species lacking male pigmentation show expression simulans produced two island endemics, D. sechellia and D. of bab in males. Some species, however, do not obey this mauritiana (Lachaise et al. 1988). Molecular evidence puts generalization. Kopp et al. (2000) also posit that other genes, the divergence between D. yakuba and D. santomea at about including Abdominal-B and doublesex, may regulate bab, and 450,000 years ago (Cariou et al. 2001), a divergence time there are clearly many other steps in the melanin-synthesis similar to that separating D. simulans from each of its two and segment-identity pathways that might affect abdominal sister species (ϳ260,000±410,000 year; Kliman et al. 2000). pigmentation (Wright 1987; Hopkins and Kramer 1992; True The D. melanogaster subgroup thus includes three episodes et al. 1999). Only direct genetic analysis can determine the of speciation following island colonization, all occurring at number, location, and effects of genes causing a species dif- roughly the same time. ference in pigmentation. The most striking aspect of D. santomea is that it is the Despite considerable sexual isolation between D. santomea only species in the D. melanogaster subgroup lacking pro- and D. yakuba and the sterility of hybrid males (Lachaise et nounced abdominal pigmentation in both sexes. Drosophila al. 2000; Coyne et al. 2002), one can perform genetic analysis yakuba and the seven other species share a single sexually by making backcrosses using the fertile F1 females. Here we dimorphic form of pigmentation: males have thin black report the results of two genetic studies of the interspeci®c stripes along the posterior portions of tergites 2, 3, and 4 difference in abdominal pigmentation. One study uses mor- (tergites are the sclerotized dorsal plates), whereas tergites phological mutants as markers to examine the effects of three 5±7 are completely black (Fig. 1c). Females have stripes chromosome regions on pigmentation. The other uses mo- along the posterior portions of all tergites, and tergites 5±7 lecular markers as tools and is limited to studying the pig- show substantial (but not complete) black pigmentation (Fig. mentation difference in males. Our goal is to provide a pre- 1d). In D. santomea males have virtually no pigmentation liminary analysis of this trait difference that will answer the (Fig. 1a), and females show very light striping on the pos- following questions: Does the species difference involve only terior parts of tergites 2±5 and no pigmentation on tergites a single gene, or is it more polygenic? Do certain chromo- 6 and 7 (Fig. 1b). Given the similarity in pigmentation among somes or chromosome regions have more effects than others all other species in the subgroup (including the outgroup on the trait? Do the differences in pigmentation between species and the putative ancestor D. yakuba), the absence of males and between females of the two species involve the melanic pigmentation in D. santomea is clearly a derived trait. same chromosome regions? Do genes affecting pigmentation The genetics of pigmentation has been studied in a few act in the same direction (i.e., do all alleles from D. santomea species of Drosophila. At least four species, D. polymorpha, tend to reduce pigmentation)? Such consistent directionality D. rufa, D. erecta, and D. kikkawai, show discrete intraspeci®c would imply that the species differences evolved by natural polymorphisms for abdominal pigmentation that, as expect- selection (Orr 1998b). ed, are controlled by segregation at a single locus (Oshima 1952; Heed and Blake 1963; Payant 1988; Gibert et al. 1999). MATERIALS AND METHODS The difference in the width of the light-colored abdominal Drosophila Stocks stripe between D. novamexicana and D. virilis, an index of general pigmentation, is apparently oligogenic, with all ®ve Flies were raised on standard cornmeal-yeast-agar medium chromosomes carrying at least one gene affecting the trait, at 24ЊC with a 12-h light-dark cycle. All stocks of D. yakuba but with chromosome 2 explaining at least half of the species and D. santomea were founded from single females captured difference (Spicer 1991). Members of the Caribbean D. dunni from the wild. The D. santomea STO.4 stock was collected subgroup vary strikingly in abdominal pigmentation (Hol- on March 1998 in the Obo Natural Reserve on SaÄo Tome locher et al. 2000a). Genetic analysis of hybrids between two Island in the zone of sympatry with D. yakuba (Lachaise et of these, a lightly pigmented, sexually dimorphic species (D. al. 2000). arawakana) and a dark, sexually monomorphic species (D. Two D. yakuba stocks, Taõ¨ 18 and wor; no; se, were used nigrodunni), showed that the species difference involved both in the analysis. Taõ¨ 18 was derived from a female collected X chromosomes and autosomes, but also maternal and pa- by D. Lachaise in 1983 in the Taõ¨ rainforest on the border ternal effects, depending on which of the three abdominal between Liberia and the Ivory Coast. areas was examined. Except for the effect of the X chro- The D. yakuba multiple-mutant stock wor; no; se, contain- mosome in males, which was estimated from the reciprocal ing a morphological mutation on each of the three major 2264 ANA LLOPART ET AL. PIGMENTATION DIFFERENCES IN DROSOPHILA 2265 chromosomes, was constructed using mutants found in three yakuba Taõ¨ 18 strains was checked by inspecting orcein- isofemale lines from the Ivory Coast and Gabon. White-or- stained preparations of salivary gland polytene chromosomes or ange (w ), a recessive X-linked mutation producing light in female F1 hybrid larvae (Ashburner 1989). The number of orange eyes, was tentatively veri®ed as an allele of the white independent preparations checked were 26, 19, 20, 13, and locus: more than 100 crosses between D. mauritiana white 26 for the X, 2R, 2L, 3R, and 3L chromosomes or chromo- males and the D. yakuba orange-eyed females produced a somal arms, respectively. We found only one inversion in single female offspring (clearly a hybrid because of its nu- the D. yakuba strain: 2Rn, which covers approximately 40% merous phenotypic abnormalities) having light orange eyes, of the right arm of chromosome 2 (Lemeunier and Ashburner as expected if the mutation producing orange eyes is an allele 1976). This inversion was seen in 12 of the 19 independent of white. preparations and is obviously polymorphic in Taõ¨ 18. The The recessive mutation notch (no), which produces nicked proximal and distal breakpoints of this inversion have been wing tips, is located on chromosome 2. It is not identical to mapped to bands 48F and 58D respectively (F. Lemeunier, Notch of D. melanogaster, which is X-linked. The approxi- pers. comm.). We have no information about the karyotype mate chromosomal location of notch was assessed using mo- of the wor; no; se strain of D. yakuba; however, no inversions lecular markers: F1 hybrid females (ϩ/no) obtained from have been reported on the X chromosome of this species and crossing the D. yakuba notch females to D. santomea STO.4 inversions on 3L are extremely rare (Lemeunier and Ash- males, were backcrossed to D. yakuba (no/no) males and the burner 1976). It is thus possible that notch and sepia are no offspring analyzed using polymerase chain reaction (PCR) linked to inversions and so would be nonrandomly associated with the twinstar primers speci®c to D. santomea STO.4 (see with more than 50 cM of their respective chromosomes. below). The observed frequency of recombinants between notch and twinstar was 3/35, indicating that notch in D. yak- Molecular Markers uba is located on the right arm of the second chromosome, Genetic analysis of pigmentation using molecular markers about 10 cM from the tip. The penetrance of no in the D. requires strain-speci®c markers whose association with the yakuba wor; no; se stock is 0.905 in males and 0.805 in fe- degree of pigmentation can be assessed in backcrosses. We males at 24ЊC. Thus some notch ¯ies will have been scored developed eight molecular markers, spread among all four as wild type in the ®rst genetic analysis described below. chromosomes, that differentiated the two strains selected for The autosomal recessive mutation sepia (se)inD. yakuba molecular-marker analysis, D. santomea STO.4 and D. yak- was shown to be identical to the chromosome 3 mutation uba Taõ¨ 18. Seven of the eight markers were chosen from sepia of D. melanogaster (located in that species at 3±26.0; nucleotide sequences of noncoding regions (introns and un- cytological position 66D13) through hybridization with D. translated segments) of the following loci: yellow (y), ver- mauritiana se, which was shown through hybridization to be milion (v), Annexin X (Ann X), bric-aÁ-brac 1 (bab1), Abdom- identical to D. melanogaster sepia (D. yakuba does not pro- inal B (Abd-B), twinstar (tsr), and Phospholipase C at 21C duce hybrids with D. melanogaster). Figure 2 gives the pre- (Plc21C). The eighth marker was developed based on the sumed chromosomal and map locations of these mutant mark- coding sequence for the chromosome 4 locus cubitus inter- ers based on the cytological maps of Lemeunier and Ash- ruptus (ci). The use of noncoding regions maximizes the burner (1976) and Lindsley and Zimm (1992) and the mo- probability of ®nding sections of DNA showing enough nu- lecular map of D. yakuba by Takano-Shimizu (2001). We cleotide differences between D. santomea and D. yakuba to assume, based on studies by Takano-Shimizu (2001) on re- facilitate designing strain-speci®c primers. Because of the combination in the X and second chromosomes, that map large effect of the X chromosome found in an initial genetic distances in D. yakuba are about 1.5 times larger than those analysis (see below), we used three markers on the X chro- in D. melanogaster. We have no information on map distances mosome (roughly evenly distributed along its length), two in D. santomea, but assume they are close to those of D. on chromosome 2 (one on each arm), two on chromosome 3 yakuba. Because of these gaps in our knowledge, all marker (one on each arm), and one on chromosome 4. positions should be viewed as tentative. There are potential problems with using D. melanogaster to design molecular markers distinguishing D. santomea from Chromosome Arrangements D. yakuba, because extensive chromosomal rearrangements Drosophila santomea chromosomes are homosequential occurred after the divergence between the ancestor of D. me- (have identical banding sequence) with those of D. yakuba, lanogaster and of D. yakuba/D. santomea (Lemeunier and and the analysis of 12 different D. santomea isofemale lines Ashburner 1976). However, most of these reorganizations did not reveal any polymorphic chromosomal arrangements have not traversed chromosome arms, and at least one num- (Lachaise et al. 2000). Drosophila yakuba, in contrast, is bered section (using the map of Bridges, which divides the polymorphic for inversions (Lemeunier and Ashburner 1976). major autosomes into 100 sections; Lindsley and Zimm 1992) The presence of different (polymorphic or ®xed) chromo- is conserved at both ends of each major chromosome. Our somal arrangements between the D. santomea STO.4 and D. molecular markers were chosen because of their cytological

←

FIG. 1. Pigmentation of the pure species and of their F1 hybrids. (a) Drosophila santomea STO.4 male; (b) D. santomea STO.4 female; or or (c) D. yakuba w ; no; se male; (d) D. yakuba w ; no; se female; (e) F1 hybrid male with D. santomea mother; (f) F1 hybrid female with D. santomea mother; (g) F1 hybrid male with D. yakuba mother; (h) F1 hybrid female with D. yakuba mother. 2266 ANA LLOPART ET AL.

FIG. 2. Approximate cytological and map positions in Drosophila yakuba of the eight molecular markers (above lines) and three morphological markers (below lines) used in the two genetic analyses. Estimated breakpoints of the 2Rn inversion segregating in the D. yakuba Taõ¨ 18 stock are indicated by arrows. Black ovals mark the position of centromeres. position in these chromosomally conserved regions and/or to design primers in conserved regions, thus maximizing the because of their proposed role in Drosophila pigmentation probability that the primers would work in D. yakuba/D. san- or melanin synthesis (True et al. 1999; Kopp et al. 2000). tomea. PCR ampli®cations were performed using approxi- The positions of these markers are given in Table 3 and shown mately 25 ng of genomic DNA extracted from single-¯y prep- in Figure 2. arations (Ashburner 1989). PCR products, ranging approxi- For the markers in the X chromosome, the estimated ge- mately from 0.6 kb to 1.4 kb in length, were directly se- netic distance in D. yakuba is 48.2 cM and 50.8 cM for the quenced with a 377 ABI Prism automated sequencer (Applied y-v and v-Ann X regions, respectively (Takano-Shimizu Biosystems, Inc., Foster City, CA) after puri®cation using 2001). For the autosomal markers, the estimated genetic dis- the QIAquick system (Qiagen, Valencia, CA). Sequences tance in D. melanogaster is approximately 107 cM and 58 were aligned with ClustalX (Thompson et al. 1997). Newly cM (Lindsley and Zimm 1992) between members of the pairs reported sequences have been deposited in GenBank, EMBL, Plc21C-tsr (chromosome 2) and bab1-Abd-B (chromosome and DDBJ database libraries under accession numbers 3), respectively. In D. yakuba, Plc21C, tsr, and bab1 have AF533212±AF533229. maintained the same chromosomal location as in D. melan- Strain-speci®c primers were designed for six of the eight ogaster. In contrast, Abd-B in D. yakuba has moved to a more markers using the newly obtained sequences. In the cases of distal position on the 3R arm (Lemeunier and Ashburner Ann X and ci, we were able to use the same pair of primers 1976), approximately 68 cM from bab1. However, these ge- in both strains because the strain-speci®c PCR products differ netic distances are almost certainly higher in D. santomea substantially in size (Ann X) or contain a nucleotide difference and D. yakuba, for empirical estimates of recombination fre- affecting the recognition sequence for the endonuclease HpaI quency are about 1.5 times higher in D. yakuba than in D. (ci). Strain-speci®c primer sequences are available upon re- melanogaster (Takano-Shimizu 2001). The map positions of quest. our markers have hence been adjusted to re¯ect this increase. These map positions should be regarded as tentative, but they Genotyping of Molecular Markers are probably not far off the true values. To obtain the sequence of these regions in the D. santomea PCR reactions using 12.5±25.0 ng of genomic DNA were and D. yakuba strains, we initially designed PCR primers performed using annealing temperatures between 50ЊC and based on the D. melanogaster sequence for six of the eight 62ЊC and cycle numbers between 30 and 45. PCR products markers. In the case of the v marker, one of the primers was were run in agarose gels containing ethidium bromide. All designed using the D. yakuba sequence present in GenBank eight markers were assayed with genomic DNA extracted (accession number AF255325). For ci, we used the D. yakuba from single-¯y preparations. A total of 382 males were ge- sequence AB005797. In all cases, when sequences from other notyped. For the X-linked markers y and v, ¯ies were ge- Drosophila species were available, we used this information notyped using one pair of primers, and any individuals in PIGMENTATION DIFFERENCES IN DROSOPHILA 2267 which there was no ampli®cation were reassayed with the of David et al. (1985; see also Hollocher et al. 2000b). This other pair. In this way, we could estimate the number of ¯ies involves examining only the three posterior tergites (numbers that do not amplify with any of our two pairs of primers 5, 6, and 7). Each tergite is given a score from 0 to 10 based (false negative results; these were fewer than 2%). We could on the proportion of the tergite area that was pigmented as not use this approach for the autosomal markers Plc21C, tsr, estimated by eye. A score of 0 corresponds to 0% pigmen- bab1, and Abd-B because of the presence of the chromosome tation, 10 to 100% pigmentation; nearly all scores were in- from the male parent. For these markers ampli®cations with tegers between 0 and 10 except for those ¯ies with very little negative results were repeated a second time with the same pigmentation, which were scored as 0.5. The pigmentation pair of primers for con®rmation. Each pair of primers was areas of the three tergites are highly correlated (see Results), tested for false positive results (primers that amplify the and hence we analyzed total pigmentation scores. These wrong regions) and false negative results (correct primers scores were calculated by adding the scores from all three that fail to amplify the correct region). For false positive tergites, so that scores for a single ¯y could range from 0 controls, single-¯y genomic DNA extractions from nine fe- (no pigmentation on any of the three tergites) to 30 (all ter- males isolated from each of the parental lines were checked gites fully pigmented). Our scoring procedure was checked with the pair of primers speci®c for the other parental strain. for accuracy by P. Gibert, who used this method extensively For the false negative controls, each pair of primers was used (e.g., Gibert et al. 1997, 1999). A single person (S. Elwyn) for amplifying genomic DNA extracted from 18 F1 hybrid scored the pigmentation of all ¯ies in this analysis. females. We used only those primers giving no false positive or negative results. For ci, PCR products were puri®ed using Analysis 2 the PCR clean-up kit (Millipore, Bedford, MA), and approx- Because of the labor involved in using molecular instead imately 200 ng were digested overnight at 37ЊC with 2.5 units of morphological markers and the greater variation in male of the endonuclease HpaI. than female pigmentation in backcrosses (there is very little obvious variation in pigmentation among female offspring Methods of Genetic Analysis from the backcross to D. santomea), we limited the molecular We conducted two different sets of genetic analyses, both mapping of pigmentation to males. This analysis used the involving backcrosses between F1 hybrid females and males method of selective genotyping (Lynch and Walsh 1998, p. of one or both parental species. The analyses differ in the 401), in which genotyping in backcrosses is performed not markers used, which sexes were scored, the direction of cross- on randomly selected individuals but on individuals having es, and the method of scoring pigmentation. extreme phenotypes. Compared to using randomly selected individuals, selective genotyping gives additional power to Analysis 1 detect quantitative trait loci (QTLs) of small effect, although it is less useful for estimating the effects of these QTLs This analysis involved a conventional genetic dissection (Lynch and Walsh 1998). We chose this method because we using one mutant marker per major chromosome, a method were interested more in gene number than in gene effects pioneered by Dobzhansky (1936). Although crude, this meth- and because we were able to estimate the effect of at least od enables us to look at the phenotypic effects of gene seg- the X-linked genes by comparing the pigmentation of F ments in both sexes, as well as the interaction among gene 1 males from the two reciprocal hybridizations. segments from different chromosomes. In this analysis, we produced F females by crossing D. Initial crosses were made between D. yakuba wor; no; se 1 yakuba Taõ¨ 18 females to D. santomea STO.4 males. These and D. santomea STO.4 (wild type); these crosses were made hybrid females were then used in two backcrosses, each to in both directions so that F males and females could be 1 males from one parental species. Thus, we produced two sets assessed for the effect of the X chromosome (in males) and of male offspring: backcross to yakuba (BY) and backcross possible maternal effects (in females). For the genetic anal- to santomea (BS). From each backcross, we chose for ge- ysis of chromosome segments, D. yakuba wor; no; se females notyping 98 males with the darkest pigmentation (designated were crossed to D. santomea STO.4 males, and the F hybrid 1 ``D''; closest in phenotype to D. yakuba males) and 98 males females were backcrossed to D. yakuba wor; no; se males. with the lightest pigmentation (designated ``L''; D. santo- This backcross produces eight classes of genotypes, involv- mea-like). The frequency of ¯ies with these extreme phe- ing all possible combinations of markers segregating against notypes in each backcross was roughly 2±4% of the total a haploid genetic background from D. yakuba. It should be offspring. Our four sets of offspring used for molecular map- remembered that each marker is linked only to a region of ping are thus designated BYD, BYL, BSD, and BSL and maximum length 50 cM on each side (longer if the marker comprised 392 ¯ies (4 ϫ 98). Each ¯y was genotyped to is within or close to an inversion), and thus this cannot iden- determine whether it carried the D. santomea or the D. yakuba tify pigmentation genes farther away on the chromosome. allele for all eight molecular markers. Unfortunately, all three available mutants are near the ends of chromosomes (Fig. 2), so each marker gives information Molecular Estimates of Species Divergence only about pigmentation genes occupying no more than one chromosome arm. To estimate the age of the D. yakuba/D. santomea species All ¯ies were collected as virgins and aged for four days. split, we calculated an estimate of DNA divergence to which Individuals were then etherized and scored under a dissecting we could apply a molecular clock (see Results). The number microscope. Pigmentation was measured using the method of nucleotide substitutions per noncoding site (K) between 2268 ANA LLOPART ET AL.

TABLE 1. Mean total pigmentation scores for males and females of pure species, F1 hybrids, and backcross ¯ies. In the two crosses, the female parent is shown ®rst. Backcross genotypes are designated by which recessive marker from D. yakuba was visible; ϩ indicates the D. santomea allele was present at that marker locus.

Males Females Genotype N Mean SE N Mean SE D. yakuba TaõÈ18 50 28.83 0.09 50 19.83 0.18 D. yakuba wor; no; se 50 28.26 0.12 50 18.38 0.15 D. santomea STO.4 50 0.00 0.00 50 0.17 0.03 or F1: STO.4 ϫ w ; se; no 100 0.05 0.02 100 5.03 0.17 or F1: w ; se; no ϫ STO.4 100 24.96 0.15 100 4.78 0.20 Backcross progeny wor; no; se 200 23.20 0.43 150 15.99 0.24 ϩ; no; se 200 21.64 0.49 150 15.78 0.17 wor; no; ϩ 200 24.07 0.36 150 15.51 0.21 wor; ϩ; se 200 19.52 0.69 150 13.57 0.35 ϩ; no; ϩ 200 21.95 0.50 150 14.72 0.24 ϩ; ϩ; se 200 15.52 0.75 150 11.59 0.40 wor; ϩ; ϩ 200 17.44 0.75 150 13.22 0.34 ϩ; ϩ; ϩ 200 13.52 0.78 150 10.25 0.38

D. santomea and D. yakuba was estimated for nine different tency (the time elapsing between when males and females regions using the program K-estimator version 5.5 (Comeron were put together until they were ®rst seen copulating). 1999). Divergence was calculated for all seven noncoding regions described in the Molecular markers section above RESULTS and also for two additional intronic regions of Decapenta- plegic and twinstar respectively. The sequence of each of Genetic Analysis 1 these gene segments was concatenated in each species to estimate the overall genetic divergence. Pigmentation of pure species and F1 hybrids The heavy pigmentation of D. yakuba is similar to that of Mating Tests seven of the other eight species in the D. melanogaster sub- group. In males of the D. yakuba wor; no; se stock, tergite 7 Kopp et al. (2000) suggested that some Drosophila species (the most posterior tergite) is entirely pigmented, tergite 6 is may use pigmentation as a mating cue. Although their dem- roughly 90±100% pigmented, and tergite 5 is 80±90% pig- onstrations of this phenomenon in D. melanogaster were not mented (Fig. 1c), giving these males an average pigmentation replicated in later work (Llopart et al. 2002), it is possible score of 28.26 of a maximum of 30 (Table 1). In females of that sexual isolation between D. yakuba and D. santomea may this strain, tergite 7 is 90±100% pigmented, tergite 6 is 60± be based in part on their differences in pigmentation. To test 70% pigmented, and tergite 5 is 20±30% pigmented (Fig. this, we determined the degree of sexual isolation of ¯ies 1d), giving them an average pigmentation score of 18.38 mating in the light and in the dark (this, of course, tests only (Table 1). Scores for males and females of the wild-type D. whether sexual isolation depends on any visual cue, not just yakuba Taõ¨ 18 stock, used in the molecular mapping exper- pigmentation). We performed simultaneous sets of sexual- iment, are similar (Table 1). isolation tests, each involving no-choice trials in which a In contrast, D. santomea species have virtually no pig- single male and female are observed over a period of time mentation (Fig. 1a,b; Table 1). Males from the D. santomea to determine whether they copulate. Each test involved 40 STO.4 strain have no pigmentation on tergites 5±7, yielding vials divided into four sets of 10: D. santomea male with D. an average pigmentation score of 0.00. Most females from santomea female, D. santomea male with D. yakuba female, this strain are also unpigmented on the posterior three ter- D. yakuba male with D. santomea female, and D. yakuba male gites, although occasional individuals have slight striping with D. yakuba female. Tests were conducted between 0900 (score 0.5) on tergite 5, yielding an average pigmentation h and 1000 h under a constant temperature regime (21±23ЊC), score in our sample of 0.17. with each observation period lasting 45 min. Light and dark F1 hybrids. Reciprocal-cross F1 hybrids were generated tests were run simultaneously, for a total of 12 pairs of tests. by crossing D. santomea STO.4 with the D. yakuba wor; no; Matings in the dark were set up under red-®ltered light, be- se stock in both directions. Pigmentation scores of reciprocal cause Drosophila cannot see in this spectrum (Frank and F1 females are intermediate between those of the parents but Zimmerman 1969). After males and females were placed in closer to those of D. santomea (Table 1, Fig. 1f, h). Female vials in red light, the room was completely darkened, with offspring from the two reciprocal crosses do not differ sig- the red light turned on only brie¯y every 5 min to determine ni®cantly in pigmentation score (t198 ϭ 0.91, P ϭ 0.36), so which pairs were copulating (nearly all matings, intra- or there is no evidence of a maternal effect acting in these ge- interspeci®c, last at least 25 min). We recorded the presence netically identical females. The females' scores do, however, or absence of copulation, and, if copulation occurred, its la- show a marked deviation from additivity. Because the two PIGMENTATION DIFFERENCES IN DROSOPHILA 2269

TABLE 2. Analysis of variance of marker effects on pigmentation data classes of F1 female do not differ signi®cantly, we can com- bine them into one population (mean ϭ 4.91, SD ϭ 1.90, n from Table 1. Asterisks give probabilities that are signi®cant using the sequential Bonferroni correction (Rice 1989). ϭ 200) and compare this population with the average value of the two parental classes (9.28) using the Z-statistic (Hays (A) Males 1963). This comparison yields a Z of 32.51, with an attendant Sum of Mean Chromosome df squares squares FP probability of 0.0009. Obviously, F1 females are signi®cantly lighter than expected from the average of parental females; X(or) 1 3294.8 3294.8 43.80 Ͻ0.0001* in female hybrids the D. yakuba pigmentation area is partially 2(notch) 1 15345.0 15345.0 204.02 Ͻ0.0001* recessive. 3(sepia) 1 196.0 196.0 2.61 0.11 X ϫ 2 1 472.0 472.0 6.28 0.012 The reciprocal F1 males, however, differ strongly, with X ϫ 3 1 8.41 8.41 0.11 0.74 each class of males having a phenotype and pigmentation 2 ϫ 3 1 716.9 716.9 9.53 0.0021* score similar to that of males from the maternal species (Fig. X ϫ 2 ϫ 3 1 13.9 13.9 0.18 0.67 1e,g; Table 1). Because females show no evidence for ma- Residual 1592 119735 75.2 ternal effects, we can assume that this difference re¯ects a (B) Females very large effect of the X chromosome on the pigmentation Sum of Mean area of males, a difference con®rmed by statistical compar- Chromosome df squares squares FP ison of the two classes of males (t198 ϭ 168, P Ͻ 0.0001). X(or) 1 663.2 663.2 48.26 Ͻ0.0001* The magnitude of this X-effect relative to the total difference 2(notch) 1 3353.7 3353.7 244.03 Ͻ0.0001* between the species is simply the difference between the 3(sepia) 1 193.5 193.5 14.08 0.0002* X ϫ 2 1 291.9 291.9 21.24 Ͻ0.0001* pigmentation scores of reciprocal F1 males divided by the X ϫ 3 1 45.7 45.7 3.32 0.068 difference between scores from males of the parental species 2 ϫ 3 1 0.48 0.48 0.04 0.85 (Hollocher et al. 2000b). Genes on the X chromosome thus X ϫ 2 ϫ 3 1 3.0 3.0 0.22 0.64 account for 88.2% (24.92/28.26) of the total difference in Residual 1192 16381 13.7 pigmentation area between the two species. The X-effect is much larger than the relative size of this chromosome, which constitutes roughly 16% of the haploid genome and suggests stitution of the D. yakuba se marker in genotypes heterozy- either that the X chromosome carries a disproportionate num- gous for the chromosome 2 segment (ϩ/no) produces a large ber of genes affecting pigmentation or that X-linked genes increase in pigmentation area, but this substitution has only have disproportionately large effects. Of course, these ex- a slight negative effect in a background that is no/no. planations are not mutually exclusive. In females, the main effects of all three chromosome mark- ers are highly signi®cant, again with all D. yakuba segments Backcross hybrids increasing the area of pigmentation. The order of main effects is 2 Ͼ X Ͼ 3. There is also a highly signi®cant interaction Female hybrids from a cross between D. yakuba wor; no; between the X and chromosome 2, resulting from the fact se females and D. santomea STO.4 males were backcrossed that adding the wor marker causes a much larger increase in to D. yakuba wor; no; se males, and pigmentation scores de- pigmentation on a ϩ/no genetic background than on a no/no termined for each of the eight genotypic classes of backcross genetic background. males and females (200 males and 150 females were scored We considered the effect of including sex as an in¯uencing per genotype). Table 1 gives the total pigmentation areas factor on pigmentation by performing an ANOVA on the total (sum of three tergites) of the pure species, F1 hybrids, and dataset, with gender being one of the ®xed variables. This backcross offspring; Table 2 shows the analysis of variance analysis shows that all three chromosomes have signi®cant for the main effects of chromosome markers and their inter- main effects on pigmentation, with chromosome 2 having the actions on pigmentation area. As noted above, each chro- largest effect (F1,2784 ϭ 318.0, P Ͻ 0.0001) followed by the mosome marker is apparently located near the end of the X(F1,2784 ϭ 66.6, P Յ 0.00001) and chromosome 3 (F1,2784 chromosome, so marker effects are limited roughly to pig- ϭ 7.92, P ϭ 0.005). There is a signi®cant interaction between mentation genes lying with 50 cM of chromosome tips. Be- sex and both the X chromosome (F1,2784 ϭ 6.71, P ϭ 0.0096) cause of the backcross used, the effects of markers and in- and chromosome 2 (F1,2784 ϭ 28.48, P Ͻ 0.0001), suggesting teractions are gauged against a genotype in which half of the that the magnitude of chromosome effects differ between the chromosomes in female hybrids are always pure D. yakuba. sexes. This is expected because the size of the interspeci®c In male hybrids, half of the autosomes are pure D. yakuba, difference in pigmentation is much larger in males than in whereas the X chromosome region associated with wor is females. hemizygous and can come from either species. In sum, each of the three chromosomes carries genes in- In males, the markers on both the X and second chro- ¯uencing pigmentation area in both sexes, although in males mosomes have signi®cant effects on pigmentation, and in the chromosome 3 segment has a signi®cant effect only both cases segments carrying the D. yakuba markers increase through its interaction with chromosome 2. This is the max- the area of pigmentation. The effect of the second-chromo- imum number of genes detectable by this analysis, so the some segment exceeds that of the X. The main effect of the true number of loci responsible for the interspeci®c difference chromosome 3 marker se is in the same direction, but is not in pigmentation is likely to be larger, especially because each signi®cant. However, there is a signi®cant interaction be- autosomal marker is linked to only one of the two chromo- tween the two autosomal segments; this is because the sub- some arms, whereas the or marker is linked to only about 2270 ANA LLOPART ET AL.

TABLE 3. Number of males from the backcross to D. santomea (BS) and 6 is 0.900 in males and 0.359 in females, between tergites and D. yakuba (BY) bearing the D. santomea genomic region for each 6 and 7 is 0.807 in males and 0.808 in females, and between molecular marker. In each of the four classes collected (BS light, BS dark, BY light, and BY dark), 98 ¯ies were genotyped at each marker tergites 5 and 7 is 0.830 in males and 0.303 in females (all locus to determine whether they carried the D. santomea or D. yakuba probabilities Ͻ 0.0001). Likewise the correlations among ter- allele. gites within each of the eight backcross genotypes are high: of the 48 correlations (eight genotypes ϫ 2 sexes ϫ 3 ter- BS BY Chromo- Cytological gites), all were positive and all but ®ve had probabilities of some band* Light Dark Light Dark 0.0001 or less. All 48 correlations were signi®cant using the yellow X 1A4-5 69 35 62 33 sequential Bonferroni correction (Rice 1989). We conclude vermilion X 10A1 75 32 72 26 that genes affecting the pigmentation difference among spe- Annexin X X 19C1 87 4 88 24 cies affect the three posterior tergites in similar directions. Plc21C 2L 21B8-C1 51 45 52 46 twinstar 2R 60B2 56 17 76 38 bric-aÁ-brac 1 3L 61F1 53 38 56 41 Genetic Analysis 2 Abdominal B 3R 89E4 64 10 90 31 ci 4 102A1 44 47 43 43 As described above, we determined the genotypes of 98 ¯ies from each of the two extreme phenotypic classes from *InD. melanogaster. each backcross (BSL, BSD, BYL, and BYD). To determine how these four phenotypes extremes ranked in pigmentation 50% of the X. This underestimate is supported by the results area, we scored 50 ¯ies from each extreme using the method of Analysis 2 (see below). Moreover, in both males and fe- employed in Analysis 1. The average pigmentation score was males the order of chromosome effects is 2 Ͼ X Ͼ 3, with 0.05 Ϯ 0.04 (SE) for light individuals in the backcross to D. the second chromosome having roughly twice the effect as santomea (BSL) and 24.58 Ϯ 0.20 for the dark males in the the X. This parallelism suggestsÐbut of course does not same backcross (BSD). The values for the light and dark proveÐthat the same loci are responsible for pigmentation males in the backcross to D. yakuba were 4.80 Ϯ 0.53 (BYL) differences in males and females. and 27.04 Ϯ 0.11 (BYD), respectively. The mean pigmen- Given the observation from F1 males that genes on the X tation score for the BSL males is not statistically different chromosome account for nearly 90% of the species differ- from the mean value of the pure D. santomea STO.4 males ences, it may seem puzzling that in the backcross analysis shown in Table 1 (Z ϭ 1.21, P ϭ 0.67). In contrast, the the chromosome 2 segment has a much larger effect than the pigmentation score of BYD individuals is signi®cantly lower X-chromosome segment. This may be explained if the mul- than that of pure D. yakuba males (t98 ϭ 12.75, P Ͻ 0.00001). tiple-marker stock of D. yakuba contained (as did our D. As expected, ¯ies in the dark category are lighter if they are yakuba Taõ¨ 18 stock) a chromosomal inversion in the right sired by D. santomea males than by D. yakuba males (t98 ϭ arm of chromosome 2. If such an inversion contained more 10.84, P Ͻ 0.00001), and ¯ies in the light category are darker than one factor affecting pigmentation, 2R would have a larg- when fathered by D. yakuba males than by D. santomea males er effect on pigmentation because such inversions suppress (t98 ϭ 8.87, P Ͻ 0.00001). recombination. In addition, Analysis 2 (which used additional Tables 3 and 4 give the results of the genetic analyses of genetic markers) shows that this disparity is probably explained 392 hybrid males. Table 3 shows the number of individuals by X-linked genes that were undetected in Analysis 1. in each class (of 98 examined) that carried the D. santomea Finally, two pieces of evidence suggest that the genes af- marker, and Table 4 gives the results of statistical tests of fecting total pigmentation area act in a similar way in each association between the phenotypes (dark or light) and ge- of the three posterior tergites. First, the analyses of variance notypes (presence of the D. santomea or D. yakuba genomic for each of the individual tergites (not shown) give results markers). Each association was determined separately for nearly identical to that of their sum (Table 2). Second, the each backcross using a 2 ϫ 2 test of independence. For ex- correlations between the pigmentation-area scores of differ- ample, in the backcross to D. santomea, the light class con- ent tergites are highly signi®cant, although somewhat lower tained 69 individuals having the D. santomea yellow allele in females than in males. The correlation between tergites 5 (Table 3) and 29 (98±69) individuals with the D. yakuba

TABLE 4. Degree of association between genotype (D. santomea/D. yakuba allele) and phenotype (light/dark class) in backcrosses to D. yakuba (BY) and to D. santomea (BS) males. See text for explanation of how statistics are calculated. Asterisks indicate signi®cant probabilities of association using the sequential Bonferroni test.

BS BY Combined data GPGP ␹2 P yellow 24.19 Ͻ0.00001 17.44 0.00003 43.85 Ͻ0.00001* vermilion 39.45 Ͻ0.00001 44.93 Ͻ0.00001 46.05 Ͻ0.00001* Annexin X 168.46 Ͻ0.00001 94.00 Ͻ0.00001 46.05 Ͻ0.00001* Plc21C 0.74 0.39 0.74 0.39 3.75 0.44 twinstar 34.54 Ͻ0.00001 31.21 Ͻ0.00001 46.05 Ͻ0.00001* bric-aÁ-brac 1 4.63 0.031 4.61 0.032 13.82 0.0079* Abdominal B 68.72 Ͻ0.00001 83.08 Ͻ0.00001 46.05 Ͻ0.00001* ci 0.18 0.67 0 1.00 0.81 0.94 PIGMENTATION DIFFERENCES IN DROSOPHILA 2271 yellow allele. Likewise, the dark class contained only 35 in- genes (higher probabilities of association could mean either dividuals with the D. santomea allele and 63 individuals with tight linkage to a gene of small effect, looser linkage to a the D. yakuba allele. Thus, for yellow one obtains a 2 ϫ 2 gene of large effect, or linkage to several genes, each of table of pigmentation versus species marker containing the moderate effect), we can use the average degree of association cell values 69, 29, 35, and 63. This gives a G-value (the to estimate maximum recombinational distance between a measure of association between marker and pigmentation) of molecular marker and a candidate gene with a very large 24.2, with P Ͻ 0.00001 (Table 4; all G-values have one effect on pigmentation. For example, the percentage asso- degree of freedom). One can construct a similar table for the ciation between the Annexin X marker and extreme pigmen- backcross to D. yakuba, obtaining a G of 17.4, P ϭ 0.00003. tation (averaged across BYD, BYL, BSD, BSL crosses) is To gauge the overall degree of association, we combined the 87.5% (343/392). If one assumes that this association is due probabilities from the two backcrosses using Fisher's test to a linked pigmentation allele of very large effect (so that (Sokal and Rohlf 1995). Applying this to the yellow marker, the extreme class always contains the proper allele at this we obtain a chi-square value (1 df) of 43.8, P Ͻ 0.00001. In locus), the estimated frequency of observed recombinants cases where the exact P-value for a G-test was not obtainable between the marker Annexin X and the candidate gene is 0.125 from our software (i.e., P Ͻ 0.00001), we used this maximum (49/392). Using the mapping function of Kosambi (1944), value as the probability in the Fisher test. The statistics for we estimate the maximum genetic distance between Annexin each backcross and the combined probability for both back- X marker and a large-effect candidate gene at 12.8 Ϯ 3.3 cM. crosses are given in Table 4 for the eight markers (raw data This method cannot, however, be used to estimate positions shown in Table 3). Statistical signi®cance in the last column of genes with smaller effects. of Table 4 is assessed using the sequential Bonferroni cor- Using haplotype information from the ®ve genetically in- rection. Identical results were obtained using single-marker dependent markers signi®cantly associated with pigmenta- regressions performed with the QTL cartographer suite of tion, we determined the percentage of ¯ies harboring a given programs (Basten et al. 2002). number of markers from each species that fall into each pig- For six of the eight markers (Plc21C and ci are the ex- mentation class in the two backcrosses. As shown in Figure ceptions), the probability of being classi®ed as a dark or light 3, the presence of at least two D. yakuba markers in the BS ¯y was nonrandomly associated with the presence or absence cross and one D. yakuba gene in the BY cross is required for of the D. santomea or D. yakuba allele. Except for the ci membership in the dark class, but in both crosses a ¯y needs marker, all of the associations, signi®cant or not, were in the at least three D. yakuba markers to have a substantial prob- expected direction, that is, individuals in the light classes had ability of being included in this class. Conversely, in both a higher association with the D. santomea marker than with backcrosses, three markers from D. santomea are required to the D. yakuba marker, and vice versa for individuals in the give a reasonable chance that a haplotype will fall into the dark classes. The highest degree of association was seen using light extreme class. In both the BS and BY backcrosses, the the Annexin X marker, located near the base of the X chro- chance of a ¯y falling into an extreme class rises markedly mosome. Surprisingly, yellow, a gene strongly in¯uencing when its genotype goes from carrying two to carrying three pigmentation within species (Lindsley and Zimm 1992) had markers of the appropriate type. Examination of individual the lowest association with pigmentation of all X-linked haplotypes shows that no single marker gene was necessary genes. (As we describe below, a direct genetic test indicates or suf®cient for inclusion in a given class. that yellow may play a small role in the pigmentation dif- Our molecular mapping explains the disparity in Analysis ference.) Among the autosomal markers, Abd-B shows the 1 between the large effect of the X chromosome on pigmen- highest association, followed by tsr and bab1. The Abd-B and tation seen in reciprocal F1 males and its smaller effect seen tsr markers appear to be more strongly associated with pig- in the backcross. First, the wor marker in Analysis 1 is un- mentation in the BY than in the BS cross, suggesting that linked to the Annexin X marker, which itself has the highest these markers interact epistatically with the genetic back- association with pigmentation of any molecular marker. Ob- ground. viously, wor is unlinked to some other genes on the X chro- With the exception of vermilion, our markers are geneti- mosome affecting pigmentation. Second, of the two sections cally independent, that is, located on different chromosomes of chromosome 2 studied in Analysis 2, the segment linked or on the same chromosome but more than 100 cM apart. to notch (which had a larger effect than wor in Analysis 1) Vermilion is located about halfway between yellow and An- has a very strong association with pigmentation, whereas the nexin X (roughly 50 cM from each) and associations between marker at the other end of chromosome 2 (Plc21C) has no vermilion and pigmentation may re¯ect only genes already signi®cant association. Thus, the conclusion from Analysis detected using yellow and Annexin X, both of which have 1 that the chromosome 2 marker had a larger effect than the signi®cant associations with pigmentation. Of the seven in- X-chromosome marker applies only to those linked segments dependent markers, ®ve (yellow, Annexin X, twinstar, bric-aÁ- and not to the chromosomes as a whole. As the comparison brac 1, and Abdominal B) are signi®cantly associated with of F1 hybrids shows, most of the pigmentation difference pigmentation, so we conclude that at least ®ve genes are between males of these species derives from X-linked genes. responsible for the difference in male pigmentation between Finally, it was possible to directly test the possibility that D. yakuba and D. santomea, with at least two genes in the the yellow gene was involved in the pigmentation difference, X chromosome, one on the right arm of chromosome 2, and in particular that the loss of pigmentation in D. santomea was one on each of the two arms of chromosome 3. caused at least partially by a mutation of yellow.Ayellow- Although we cannot estimate the relative effects of these like male was detected in an inbred line of D. yakuba derived 2272 ANA LLOPART ET AL.

FIG. 3. The relationship in the two backcrosses in Analysis 2 between the number of Drosophila yakuba markers in a haplotype and the proportion of such haplotypes that fall into the dark extreme class (this proportion is represented by the height of the shaded bar). Only the ®ve genetically independent markers having a signi®cant association with pigmentation are used in this analysis. Data are averaged across all haplotypes containing a given number of D. yakuba markers. BS are individuals from the backcross to D. santomea, BY from the backcross to D. yakuba. The numbers above each bar are the number of haplotypes occurring in that class. The graphs can also be read in reverse, as the association between the number of D. santomea markers in a haplotype and the proportion of those individuals in the light class; one does this simply by subtracting the number of D. yakuba markers from ®ve and noting the height of the unshaded portion of each bar.

from the Taõ¨ 18 stock. This mutation was con®rmed to be n ϭ 50; t98 ϭϪ1.12, P ϭ 0.27), indicating that the yellow an allele of the yellow locus by crossing females carrying mutation is completely recessive for pigmentation in D. yak- this mutation to D. mauritiana yellow males; all female off- uba. (Indeed, the yellow heterozygotes were slightly but in- spring were yellow. This D. yakuba yellow stock was used signi®cantly darker than homozygous wild-type ¯ies.) The to perform a genetic complementation test with the D. san- difference between these two crosses is highly signi®cant ϩ tomea STO.4 strain. D. yakuba Taõ¨ 18 (yyak ) and D. yakuba (t196 ϭ 2.78, P ϭ 0.006), showing that the effect of hetero- Taõ¨ 18 yellow (yyak) males were crossed separately to D. san- zygosity for yellow occurs in the interspeci®c but not in the ϩϩ tomea STO.4 females (yysan / san ). The pigmentation score of intraspeci®c cross. interspeci®c hybrids heterozygous for yellow (mean ϭ 7.79, These data suggest, but do not prove, that the yellow locus SE ϭ 0.40, n ϭ 50) was signi®cantly lower than that of may play a role (albeit a small one) in the pigmentation hybrids not heterozygous for yellow (mean ϭ 9.10, SE ϭ difference between D. yakuba and D. santomea. The results 0.29, n ϭ 50; t98 ϭ 2.67, P ϭ 0.009), suggesting that the from these crosses might also be explained, however, if the wild-type allele of D. santomea yellow locus does not fully yellow locus shows some loss of recessivity in a hybrid but complement the D. yakuba yellow mutation. not in a pure-species genetic background. To determine whether this difference between genotypes might re¯ect only the decreased expression of yellow when Molecular Estimates of Species Divergence heterozygous, we performed a control cross, similar to that described above but using the D. yakuba Taõ¨ 18 strain instead We estimated nucleotide divergence between D. santomea of the D. santomea strain. Pure D. yakuba females hetero- and D. yakuba using sequences of nine noncoding regions. zygous for yellow (mean score ϭ 21.22, SE ϭ 0.24, n ϭ 50) The comparison of this value with estimates obtained for were not signi®cantly lighter than D. yakuba females ho- other pairs of species in the D. melanogaster subgroup allows mozygous for the wild-type allele (mean ϭ 20.79, SE ϭ 0.30, us to infer their relative and absolute divergence times. We PIGMENTATION DIFFERENCES IN DROSOPHILA 2273

TABLE 5. Test of sexual isolation under light and dark conditions. regime) seen in of each of the four possible pairings between Table gives the number of copulations (matings) of 120 trials for each these species. Under both light regimes, sexual isolation in of the pairs, as well as the latency of copulations that did occur. S, D. santomea STO.4 line; Y, D. yakuba TaõÈ 18 line. For each mating, the this species is asymmetrical: interspeci®c matings between female parent is given ®rst. D. yakuba females and D. santomea males are far more fre- quent than the reciprocal mating. Each of the four types of Matings matings was signi®cantly less frequent in dark than in light Regime S ϫ SSϫ YYϫ SYϫ Y conditions (P Ͻ 0.05 under expectation of equal frequency), Light Matings 79 13 72 83 and for all but one mating (D. santomea female ϫ D. yakuba Latency (min) 13.69 23.52 21.47 19.44 male), the onset of copulation (copulation latency) is signif- (SE) (1.12) (3.54) (1.42) (1.25) icantly shorter in the light than in the dark (for this mating, Dark Matings 55 2 33 38 t13 ϭ 0.13, P ϭ 0.13; for the other three matings, P Ͻ 0.005). Latency (min) 22.64 38.25 28.34 28.65 However, sexual isolation between the species is not great- (SE) (1.39) (0.90) (1.70) (1.69) er in the light than in the dark, as one might expect if pig- mentation is a cue for conspeci®c mating. The 2 ϫ 4 table of mating type versus light regime (Table 5) shows no sig- were especially interested in comparing the D. yakuba/san- ni®cant heterogeneity (␹2 ϭ 6.48, df ϭ 3, P ϭ 0.09). If tomea divergence with that between D. simulans and its clos- anything, sexual isolation is slightly increased in the dark, est relatives, the island endemics D. mauritiana and D. se- as the proportion among all dark matings of the D. santomea chellia, for we know a great deal about the genetics of re- female ϫ D. yakuba mating is only about one-third the fre- productive isolation and character differences between the quency seen in the light, whereas the other three matings last three species (Coyne and Orr 1999). Like D. santomea, occur in similar proportions in light versus dark. The very D. sechellia and D. mauritiana arose after colonization of slight increase in sexual isolation in the dark is con®rmed islands by a mainland ancestor, and comparing the three is- using the chi-square index of Gilbert and Starmer (1985). land speciation events may reveal regularities in how repro- This index ranges from Ϫ1 (complete disassortative mating) ductive isolation evolves after colonization. to 1 (complete assortative mating), and for D. yakuba and D. Cariou et al. (2001) estimate the divergence between D. santomea it is 0.37 for matings in the light and 0.49 for santomea and D. yakuba at 450,000 years using a single se- matings in the dark. Obviously, both species mate less fre- quence of amylase from each species. However, single se- quently in the dark than in the light, but there is no evidence quences may yield inaccurate divergence times for two rea- that sexual isolation is reduced in the absence of visual cues. sons. First, such estimates of divergence have large variances. Second, when speciation events are recent, ``species differ- ences'' may re¯ect only ancestral polymorphisms still seg- DISCUSSION regating in the newly derived species (Kliman et al. 2000). Our main conclusion is that the loss of pigmentation in D. We acquired polymorphism data in D. yakuba for the bab1 santomea involved evolutionary changes occurring at several noncoding region (average number of nucleotide differences genes. Analysis 1, using both sexes, implies that at least three per site [k] was 7/539), and hence we can estimate the fraction genes in males and femalesÐone on each major chromo- of the apparent divergence between D. yakuba and D. san- someÐaffect the pigmented area of the posterior three ter- tomea that is actually due to variation within D. yakuba, gites, and the comparison of reciprocal F1 hybrid males in- assuming that all noncoding regions have the same levels of dicates that genes on the X chromosome are responsible for variation. After this correction, the net divergence between nearly 90% of the character difference. Because the markers D. yakuba and D. santomea turns out to be 0.0154 noncoding used were linked to only one chromosome arm (and to only substitutions per site, approximately one-eighth of the esti- about half of the X chromosome), this is a minimum estimate mate of synonymous divergence between D. melanogaster of the number of genes responsible for the pigmentation dif- and D. simulans (Ks ϭ 0.1176; Swanson et al. 2001). If the ference. As noted in the introduction, oligogeny or polygeny split between D. melanogaster and D. simulans occurred ap- of interspeci®c differences in pigmentation has been seen in proximately 3 million years ago, as estimated by Hey and two other studies of Drosophila (Spicer 1991; Hollocher et Kliman (1993), and we assume D. santomea and D. yakuba al. 2000b). follow the same molecular clock as do D. melanogaster/si- The effects of the three tested chromosome arms are in the mulans, then the split between D. yakuba and D. santomea same order in males as in females, suggesting that changes occurred roughly 393,000 years ago. Our estimate of diver- at the same loci have reduced pigmentation in both males gence time is thus close to that calculated by Cariou et al. and females. Moreover, the positive correlation of pigmen- (2001). Moreover, our estimate is close to divergence times tation scores among the three tergites suggests that evolution estimated between D. simulans and D. mauritiana and D. has ®xed alleles affecting the general posterior region of the simulans and D. sechellia: 263,000 and 413,000 years ago, abdomen rather than alleles whose phenotype is limited to respectively (Kliman et al. 2000). All three island speciation particular posterior tergites (i.e., the genetic factors revealed events appear to be about the same age. in this study do not modify positional cues but change the effect of pre-existing cues on pigmentation). Mating Tests Analysis 2, involving the association between molecular Table 5 gives the number of matings and the copulation markers and extreme phenotypes in backcrosses, reveals that latency (of a total of 120 pairs observed under each light at least ®ve genes are responsible for the pigmentation dif- 2274 ANA LLOPART ET AL. ferences between D. yakuba and D. santomea males; that is, ®ed by the small number of segregating genetic units in our ®ve loci are signi®cantly associated with the probability of cross (roughly the number of chromosome arms), one would a ¯y having either a D. santomea-like or D. yakuba-like pat- not expect the parental phenotypes to be so frequent if there tern of pigmentation. The trait of belonging to an extreme were, say, 25 genes in¯uencing pigmentation. Again, this class is not identical to the pigmentation area measured in question can be resolved only through QTL analysis. Analysis 1, as Analysis 2 involved assessing pigmentation The only other genetic study of a species difference in intensity as well as area. Nevertheless, the traits are strongly pigmentation is Hollocher et al.'s (2000b) analysis of pig- related, because, however darkly pigmented, a ¯y was not mentation in two Caribbean species in the Drosophila cardini included in an extreme class unless the area of pigmentation group: D. arawakana (lighter abdomen and sexually dimor- was close to that of one parental species. It is safe to say, phic, with males darker than females) and D. nigrodunni (ab- then, that the loss of pigmentation in D. santomea involved domen almost completely pigmented and sexually mono- changes in at least ®ve genes in males and three in females. morphic). Considering the posterior region of the abdomen As noted above, Analysis 2 is unable to estimate effects (``area 3'' as shown in Hollocher et al. 2000a, ®g. 2), which of individual QTLs on pigmentation. Nevertheless, it seems is roughly equivalent to the area scored in our analysis, we unlikely that the association of these QTLs with pigmenta- ®nd little similarity between the genetic basis of pigmentation tion, some of them extremely strong, re¯ects single genes of in the two pair of species. Hollocher et al. (2000b) found very small effect, and it is more reasonable to suppose that only a minor effect of the X chromosome in males, and this the very strong associations re¯ect either single genes of was in the direction opposite to ours (i.e., reciprocal F1 males relatively large effect or several to many genes of smaller in the D. arawakana ϫ D. nigrodunni cross have pigmentation effect. Analysis 2 con®rms the large X-effect seen in the closer to that of males from the paternal species, indicating reciprocal F1 hybrid males of Analysis 1, as there is strong a paternal effect but no discernible effect of the X chromo- statistical association between pigmentation and all three X- some in males). Moreover, reciprocal F1 females showed a linked chromosome segments, while at least one of the two maternal effect, whereas we found no evidence of this in our markers on each autosome has either a weaker association study. We cannot make further comparisons between our re- or no signi®cant association. Whether the large effect of the sults and those of Hollocher et al. (2000b) because the effects X chromosomes in males resides in its possession of a few of individual chromosomes could not be assessed in the latter genes of large effect, many genes of small effect, a combi- analysis. Nevertheless, the location and effects of genes af- nation of these two circumstances, or simply the recessivity fecting pigmentation clearly differ between the two studies. of genes affecting pigmentation, awaits a more re®ned genetic Of course, unless only a small number of genes can poten- analysis in both sexes, which we are attempting with QTL tially be involved in the evolution of pigmentation, there is mapping using many molecular markers. no reason to expect a similarity in the genes associated with A large X-effect in males is predicted by some theories of pigmentation differences between distantly related pairs of evolution, including those involving traits based on genes species. that have antagonistic ®tness effects on males and females, Three genes are worth discussing in more detail: yellow genes whose ®tness advantages during ®xation were partially (y), bric-aÁ-brac1 (bab1), and bric-aÁ-brac2 (bab2). A corre- recessive, or genes affected by ¯uctuating sexual selection lation between the amount of the yellow protein in the epi- (Rice 1984; Charlesworth et al. 1987; Reinhold 1998). How- dermis and the intensity of melanization has been proposed ever, such effects are not commonly seen for species differ- in D. melanogaster (Walter et al. 1991). Moreover, Hollocher ences involving other sexually monomorphic or dimorphic et al. (2000b) suggest that pigmentation differences in the traits in Drosophila. Male-limited traits such as differences Drosophila dunni subgroup may involve mutations at the yel- in genital morphology or sex-comb tooth number are rarely low locus. Our complementation tests also suggest that yellow based on genes that show disproportionately large effects of may play at best a small role in the pigmentation difference the X chromosome (Coyne and Orr 1999; Orr 2001). A no- between D. santomea and D. yakuba. This suggestion is con- table exception is Lepidoptera, in which many species dif- sonant with another observation in our study: at least four ferences, whether or not they involve sexually dimorphic markers show a stronger association with pigmentation than characters, are X-linked, even though the X chromosome con- does yellow. Indeed, among the three X-linked markers ex- stitutes a very small proportion of the genome (Prowell 1998). amined, yellow shows the weakest association with pigmen- This taxon-speci®c occurrence of large X-effects remains a tation, with a far smaller association than Annexin X. This mystery. does not rule out an effect of the yellow locus in the species Although our interspeci®c difference in pigmentation is difference, but the results of complementation tests with the caused by at least several genes, it is unlikely to rest on a D. yakuba yellow mutation may re¯ect only the semidomi- very large number of loci spread throughout the genome. nance of a normally recessive gene in a hybrid D. santomea/ First, at least one chromosome arm (2L) has no signi®cant D. yakuba genetic background. Thus, the involvement of yel- association with pigmentation. Second, in both the BYD and low in the D. santomea/D. yakuba pigmentation difference BSL classes we recovered many individuals with pigmen- must be viewed as tentative at best. tation scores identical to those of pure D. yakuba and D. In Drosophila, three lines of evidence suggest that bab1/ santomea males, respectively. A signi®cant fraction of pa- bab2 act as general repressors of pigmentation (Kopp et al. rental phenotypes appearing in backcrosses implies that the 2000). First, in sexually dichromatic species in which the number of genes involved in the trait difference is not ex- posterior three tergites of males but not females are darkly tremely large. Although this conclusion is somewhat quali- pigmented, bab1 and bab2 proteins can be detected in pos- PIGMENTATION DIFFERENCES IN DROSOPHILA 2275 terior female but not male tergites of midstage pupae. Second, pigmentation yields a probability of 0.045 (six of six) that overexpression of bab represses non-sex-speci®c pigmenta- this result would occur by chance; this is a more conservative tion stripes in the anterior tergites of D. melanogaster. Third, estimate than a simple two-tailed sign test (which yields P loss of bab1 and bab2 genes in D. melanogaster produces ϭ 0.06) because it is conditioned on the fact that the majority female ¯ies with heavy pigmentation of all but one abdominal of pigmentation alleles must reside in the D. yakuba line. If segment. Segment shape and patterns of abdominal bristles we apply the same test but using only the ®ve loci having a and trichomes are also regulated by bab. signi®cant effect on pigmentation (®ve of ®ve) test result is The idea that bab represses male-speci®c pigmentation in not signi®cant (P ϭ 0.17). Based on these results, we cannot females led Kopp et al. (2000) to examine the correlation strongly reject the hypothesis that random drift may have between bab expression and pigmentation across a diverse been the main force responsible for this species difference. group of Drosophila species. In general, those species mono- The discovery of additional QTLs will provide a better un- morphic for pigmentation also show similar distribution of derstanding of the relative roles of selection (natural or sex- the bab1 and bab2 proteins in males and females, whereas ual) and drift on this trait. dimorphic species with darkly pigmented males show no bab The light and dark mating tests provide no evidence that expression in the posterior segments of males. This corre- pigmentation in¯uences sexual isolation in these species. Al- lation (although not perfect) led Kopp et al. (2000) to suggest though the frequency of mating was reduced and the latency that species differences in sex-speci®c pigmentation (the de- of copulation increased in dark as compared to light condi- rived evolutionary state) resulted from changes in the reg- tions, we found no evidence that sexual isolation was relaxed ulatory regions of bab1 and bab2 that enabled these loci to in the dark. There was certainly courtship in the dark (both respond to signals from the controlling genes Abdominal B inter- and intraspeci®c), for there were numerous matings and doublesex. It is thus possible that the loss of pigmentation and males were frequently seen courting females when the in D. santomea may re¯ect overexpression of bab in both red light was turned on at 5-min intervals. The general re- sexes, eliminating not only sex-speci®c pigmentation, but duction of mating activity may be due to two factors: either nearly all pigmentation. ¯ies fail to encounter each other in the dark (visual recog- For several reasons, however, this suggestion seems un- nition of another individual may be importantÐbut not re- likely to explain the pigmentation difference between D. yak- quiredÐfor initiating courtship), or mating might be in¯u- uba and D. santomea. First, this difference is based on at enced by some visual cue, such as male wing-extension or least three genes in females and ®ve in males. In Analysis circling behavior, that does not differ between the species. 2, bab1 (closely linked to bab2) was used directly as a mo- The fact that sexual isolation remains as strong in the dark lecular marker but had the smallest association with male as in the light suggests that nonvisual clues, such as differ- pigmentation among all markers that showed signi®cant as- ences in wing-vibration song (e.g., Doi et al. 2001) or pher- sociation. In Analysis 1, bab1 and bab2 are closely linked omonal hydrocarbons (e.g., Coyne et al. 1994), are the pri- (ϳ9 cM) to sepia, the marker with the smallest effect on mary factors involved in sexual isolation of these species. pigmentation area in both males and females. This analysis However, gas chromatography of cuticular hydrocarbons in also showed that genes on the X chromosome account for 13 lines of D. yakuba and seven lines of D. santomea from about 90% of the species difference in the area pigmented various populations shows that their main hydrocarbons are in tergites 5±7 (bab1 and bab2 are on chromosome 3). These identical. Both males and females of each species show a results rule out the notion that the D. santomea/yakuba pig- predominance of 7-tricosene, the most common hydrocarbon mentation difference is due solely or even primarily to mu- pro®le in the D. melanogaster subgroup (Jallon and David tations in the regulatory region of the two bab genes. 1987). It is therefore unlikely that cuticular hydrocarbons It is of course still possible that the pigmentation difference play a role in the sexual isolation of D. yakuba and D. san- could be due to mutations of unlinked genes that affect the tomea. It is clear from observations of these species, in both expression of bab. In such a case, the bab product would be light and dark, that sexual isolation is based primarily on the proximate cause of the interspeci®c pigmentation differ- female rejection of courting heterospeci®c males and not ence, but bab itself (and its controlling region) would not males' refusal to court heterospeci®c females. Males of either differ among the species. Moreover, as proposed by Kopp et species court females of the other avidly, but failure to cop- al. (2000), differences in bab expression explain differences ulate occurs via female refusal, particularly the refusal of in sexually dimorphic pigmentation. The evolutionary change courting D. yakuba males by D. santomea females. of D. santomea, in contrast, has been the loss of pigmentation Although we have no direct evidence for this suggestion, in both sexes, and need not involve genes normally involved we propose that natural rather than sexual selection was the in sexual dimorphism. force responsible for the loss of pigmentation in D. santomea. Seven of the eight tested markers are genetically indepen- First, this species is not sexually dimorphic for pigmentation, dent and, of these, six show a consistent effect on pigmen- yet sexual selection often (but not always) produces sexual tation in both backcrosses (i.e., D. yakuba alleles were as- dimorphism. Second, the mating tests show no evidence that sociated with darker pigmentation, D. santomea alleles with either D. yakuba or D. santomea uses pigmentation (or any lighter pigmentation), but for only ®ve of the six were the visual signal) as a cue for mate discrimination. One might effects signi®cant. This consistency suggests that the species expect, although again this need not always be the case, that difference may have resulted from selection rather than drift a character undergoing divergent sexual selection in two re- (Orr 1998b). Using Orr's (1998a, eq. 6) equal-effects sign lated species should ultimately be involved in their sexual test on chromosome regions having a consistent effect on isolation. 2276 ANA LLOPART ET AL.

If natural selection has caused the species difference, the lopatric or sympatric populations of D. yakuba using paralogous basis of this selection is unclear. Some species of Drosophila amylase lines and migration scenarios along the Cameroon vol- canic line. Mol. Ecol. 10:649±660. show clinal geographic variation of thoracic and abdominal Charlesworth, B., J. A. Coyne, and N. Barton. 1987. The relative pigmentation that is based on genetic change, implying spa- rates of evolution of sex chromosomes and autosomes. Am. Nat. tially varying selection. Generally, pigmentation is darker in 130:113±146. ¯ies from cooler climes and higher altitudes. Because ¯ies Comeron, J. 1999. K-Estimator: Calculation of the number of nu- cleotide substitutions per site and the con®dence intervals. of a given strain also become darker at cooler rearing tem- Bioinformatics 15:763±764. perature, paralleling the evolutionary production of darker Coyne, J. A., and H. A. Orr. 1999. The evolutionary genetics of ¯ies at higher altitudes and latitudes, temperature is often speciation. Pp. 1±36 in A. E. Magurran and R. M. May, eds. suggested as the selective pressure for divergent evolution Evolution of biological diversity. Oxford Univ. Press, Oxford, of pigmentation within species (David et al. 1985; Capy et U.K. Coyne, J. A., K. Mah, and A. Crittenden. 1994. Genetics of a pher- al. 1988; Gibert et al. 1998a,b). Drosophila santomea and D. omonal difference contributing to reproductive isolation in Dro- yakuba, however, violate this rule because the lighter species sophila. Science 265:1461±1464. lives at higher altitudes. Factors that have been invoked in Coyne, J. A., S. Y. Kim, A. S. Chang, D. Lachaise, and S. Elwyn. the evolution of insect pigmentation include radiant heat up- 2002. Sexual isolation between two sibling species with over- lapping ranges: Drosophila santomea and D. yakuba. Evolution take, thermal tolerance, UV tolerance, and resistance to des- 56: In press. iccation. (David et al. 1985; Crill et al. 1996; Gibert et al. Crill, W. D., R. B. Huey, and G. W. Gilchrist. 1996. Within- and 1998a). Until we know much more about the ecology of D. between-generation effects of temperature on the morphology santomea and D. yakuba, we will not understand what type and physiology of Drosophila melanogaster. Evolution 50: of selection, if any, has shaped their striking difference in 1205±1218. David, J. R., P. Capy, V. Payant, and S. Tsakas. 1985. Thoracic color. trident pigmentation in Drosophila melanogaster: differentiation Finally, our estimate of 393,000 years as the age of di- of geographical populations. Genet. Sel. Evol. 17:211±224. vergence of these species puts them at roughly the same stage Dobzhansky, T. H. 1936. Studies in hybrid sterility. II. Localization of temporal divergence as two other pairs in the D. melan- of sterility factors in Drosophila pseudoobscura hybrids. Ge- netics 21:113±135. ogaster subgroup: D. simulans and D. sechellia (263,000 year) Doi, M., M. Matsuda, M. Tomaru, H. Matsubayashi, and Y. Oguma. and D. simulans and D. mauritiana (413,000 year). It is likely 2001. A locus for female discrimination behavior causing sexual that both D. sechellia and D. mauritiana resulted from in- isolation in Drosophila. Proc. Natl. Acad. Sci. USA 98: dependent colonization of islands by a D. simulans-like an- 6714±6719. cestor (Kliman et al. 2000), and so the D. melanogaster sub- Fisher, R. A. 1930. The genetical theory of natural selection. Clar- endon Press, Oxford, U.K. group includes three species that arose after island coloni- Frank, K. D., and W. F. Zimmerman. 1969. Action spectra for phase zation. There are already parallels between these three pairs shifts of a circadian rhythm in Drosophila. Science 163:688±689. of species that imply common evolutionary processes (e.g., Gibert, P., B. Moreteau, J.-C. Moreteau, and J. R. David. 1997. in all cases sexual isolation is stronger between island-species Growth temperature and adult pigmentation in two Drosophila sibling species: an adaptive convergence of reaction norms in females and mainland-species males than vice versa). The sympatric populations? Evolution 50:2346±2353. discovery of further parallels may imply that, at least in this Gibert, P., B. Moreteau, J. C. Moreteau, R. Parkash, and J. R. David. group of ¯ies, speciation on islands involves repeatable 1998a. Light body pigmentation in Indian Drosophila melano- events. gaster: a likely adaptation to a hot and arid climate. J. Genet. 77:13±20. Gibert, P., B. Moreteau, S. M. Scheiner, and J. R. David. 1998b. ACKNOWLEDGMENTS Phenotypic plasticity of body pigmentation in Drosophila: cor- We thank O. Betz for helping with the photographs and related variations between segments. Genet. Sel. Evol. 30: 181±194. F. Lemeunier for checking our chromosome preparations. Gibert, P., B. Moreteau, A. Munjal, and J. R. David. 1999. Phe- This work was supported by National Institutes of Health notypic plasticity of abdominal pigmentation in Drosophila kik- grant GM58260 to JAC. kawai: multiple interactions between a major gene, sex, abdomen segment, and growth temperature. Genetica 105:165±176. Gilbert, D. G., and W. T. Starmer. 1985. Statistics of sexual iso- LITERATURE CITED lation. Evolution 39:1380±1383. Ashburner, M. 1989. Drosophila: a laboratory manual. Cold Spring Hays, W. L. 1963. Statistics for psychologists. Holt Reinhart and Harbor Press, Cold Spring Harbor, NY. Winston, New York. Basten, C. J., B. S. Weir, and Z. B. Zeng. 2002. QTL cartographer. Heed, W. B. and N. B. Krishnamurthy. 1959. Genetic studies on Ver. 1.16. Department of Statistics, North Carolina State Uni- the cardini group of Drosophila in the West Indies. Univ. Texas versity, Raleigh, NC. Pub. 5914:155±179. Begun, D., and P. Whitley. 2000. Reduced X-linked nucleotide Hey, J., and R. M. Kliman. 1993. Population genetics and phylo- polymorphism in Drosophila simulans. Proc. Natl. Acad. Sci. genetics of DNA sequence variation at multiple loci within the USA 97:5960±5965. Drosophila melanogaster species complex. Mol. Biol. Evol. 10: Bradshaw, H. D., S. M. Wilbert, K. G. Otto, and D. W. Schemske. 804±822. 1998. Quantitative trait loci affecting differences in ¯oral mor- Hollocher, H., J. L. Hatcher, and E. G. Dyreson. 2000a. Evolution phology between two species of monkey¯ower (Mimulus). Ge- of abdominal pigmentation differences across species in the Dro- netics 149:367±382. sophila dunni subgroup. Evolution 54:2046±2056. Capy, P., J. R. David, and A. Robertson. 1988. Thoracic trident ÐÐÐ. 2000b. Genetic and developmental analysis of abdominal pigmentation in natural populations of Drosophila simulans:a pigmentation differences across species in the Drosophila dunni comparison with D. melanogaster. Heredity 61:263±368. subgroup. Evolution 54:2057±2071. Cariou, M.-L., J.-F. Silvain, V. Daubin, J.-L. DaLage, and D. La- Hopkins, T. L., and K. J. Kramer. 1992. Insect cuticle sclerotization. chaise. 2001. Divergence between Drosophila santomea and al- Annu. Rev. Entomol. 37:273±302. PIGMENTATION DIFFERENCES IN DROSOPHILA 2277

Kliman, R. M., P. Andolfatto, J. A. Coyne, F. Depaulis, M. Kreit- Prowell, D. P. 1998. Sex linkage and speciation in Lepidoptera. Pp. man, A. J. Berry, M. McCarter, J. Wakeley, and J. Hey. 2000. 309±319 in D. J. Howard and S. H. Berlocher, eds. Endless The population genetics of the origin and divergence of the forms: species and speciation. Oxford Univ. Press, Oxford, U.K. Drosophila simulans complex species. Genetics 156:1913±1931. Reinhold, K. 1998. Sex linkage among genes controlling sexually Kopp, A., I. Duncan, D. Godt, and S. B. Carroll. 2000. Genetic selected traits. Behav. Ecol. Sociobiol. 44:1±7. control and evolution of sexually dimorphic characters. Nature Rice, W. R. 1984. Sex chromosomes and the evolution of sexual 408:553±559. dimorphism. Evolution 38:735±742. Kosambi, D. D. 1944. The estimation of map distances from re- ÐÐÐ. 1989. Analyzing tables of statistical tests. Evolution 43: combination values. Ann. Eugen. 12:172±175. 223±225. Lachaise, D., M.-L. Cariou, J. R. David, F. Lemeunier, L. Tsacas, Schemske, D. W., and H. D. Bradshaw. 1999. Pollinator preference and M. Ashburner. 1988. Historical biogeography of the Dro- and the evolution of ¯oral traits in monkey¯owers (Mimulus). sophila melanogaster species subgroup. Evol. Biol. 22:159±225. Proc. Natl. Acad. Sci. USA 96:11910±11915. Lachaise, D., M. Harry, M. Solignac, F. Lemeunier, V. BeÂnassi, Sokal, R. R., and F. J. Rohlf. 1995. Biometry. 3rd ed. W. H. Free- and M.-L. Cariou. 2000. Evolutionary novelties in islands: Dro- man, New York. sophila santomea a new melanogaster sister species from SaÄo Spicer, G. S. 1991. The genetic basis of a species-speci®c character TomeÂ. Proc. R. Soc. Lond. B 267:1487±1495. in the Drosophila virilis species subgroup. Genetics 128: Lemeunier, F., and M. Ashburner. 1976. Relationship within the 331±337. melanogaster subgroup of the genus Drosophila (Sophophora). Swanson, W. J., A. G. Clark, H. M. Waldrip-Dail, M. F. Wolfner, II. Phylogenetic relationships between six species based upon and C. F. Aquadro. 2001. Evolutionary EST analysis identi®es polytene chromosome banding sequences. Proc. R. Soc. Lond. rapidly evolving male reproductive proteins in Drosophila. Proc. B 193:275±294. Natl. Acad. Sci. USA 98:7375±7379. Lindsley, D. L., and G. G. Zimm. 1992. The genome of Drosophila Takano-Shimizu, T. 2001. Local changes in GC/AT substitution melanogaster. Academic Press, San Diego, CA. biases and in crossover frequencies on Drosophila chromosomes. Llopart, A., S. Elwyn, and J. A. Coyne. 2002. Pigmentation and Mol. Biol. Evol. 18:606±619. mate choice in Drosophila. Nature 419:360. Thompson, J. D., T. J. Gibson, F. Plewniak, F. Jeanmougin, and Lynch, M., and B. Walsh. 1998. Genetics and analysis of quanti- D. G. Higgins. 1997. The CLUSTALX windows interface: ¯ex- tative traits. Sinauer Associates, Sunderland, MA. ible strategies for multiple sequence alignment aided by quality Orr, H. A. 1998a. The population genetics of adaptation: the dis- analysis tools. Nucleic Acids Res. 25:4876±4882. tribution of factors ®xed during adaptive evolution. Evolution True, J. R., K. A. Edwards, D. Yamamoto, and S. B. Carroll. 1999. 52:935±949. Drosophila wing melanin patterns form by vein-dependent elab- ÐÐÐ. 1998b. Testing natural selection vs. genetic drift in phe- oration of enzymatic prepatterns. Curr. Biol. 9:1382±1391. notypic evolution using quantitative trait locus data. Genetics Walter, M. F., B. C. Black, G. Afshar, A. Y. Kermabon, and T. R. 149:2099±2014. Wright. 1991. Temporal and spatial expression of the yellow ÐÐÐ. 2001. The genetics of species differences. Trends Ecol. gene in correlation with cuticle formation and DOPA decarbox- Evol. 16:343±350. ylase activity in Drosophila development. Dev. Biol. 147:32±45. Oshima, C. 1952. Genetic analysis of the dimorphism of color pat- Wright, T. R. F. 1987. The genetics of biogenic amine metabolism, tern in D. rufa. Drosophila Inf. Serv. 26:116. sclerotization, and melanization in Drosophila melanogaster. Payant, V. 1988. Le polymorphisme de coloration abdominale de Adv. Genet. 24:127±222. Drosophila erecta est-il gouverne par un geÁne seÂlectivement neu- tre. Genet. Sel. Evol. 20:37±50. Corresponding Editor: J. Fry