Interspecific Hybrids of Drosophzla Heteroneura and D

Evolution, 43(2), 1989, pp. 347-361

INTERSPECIFIC HYBRIDS OF DROSOPHZLA HETERONEURA AND D. SZLVESTRZS I. COURTSHIP SUCCESS

JAYNEN. AHEARN' Department of Genetics, University of Hawaii, Honolulu, HI 96822

AND ALANR. TEMPLETON Department of Biology, Washington University, St. Louis, MO 63130

Abstract. -Drosophila heteroneura and D. silvestris are well-defined, sympatric species of the planitibia subgroup of Hawaiian Drosophila. D. silvestris can be subdivided into two allopatric morphotypes that differ in the number of bristle rows on the front tibia (two rows versus three rows). We measured courtship success of intraspecific and interspecifichybrids as the proportion of females inseminated during a two-week period with a single sib male. Proportions were arcsin-transformed so that the values were asymptotically normal in distribution, and tests of homogeneity and of mean differences were performed. Of key importance is the discovery of genetic variation for the proportion of inseminated females within both D. heteroneura and D. silvestris. The interspecific crosses and the D. silvestris intraspecific crosses also provide evidence for a coadapted gene complex with some dominance or heterosis. This coadapted gene complex correlates with the morphotypes of these flies, rather than with the D. heteroneura/D. silvestris contrasts per se. This observation stresses the importance of recognizing both behavioral and morphological components of the mate- recognition system. The incompatible coadaptation that separates the two-row from the three-row forms also supports recent molecular studies which indicate that the three-row form split from the two-row form prior to the split between D. heteroneura and two-row D. silvestris. The observations of intraspecific variability and coadaptation support the predictions of a genetic-transilience model which explains the origin of a new mate-recognition system in terms of sexual selection in the context of a founder-flush event.

Received April 28, 1987. Accepted October 12, 1988

Nowhere on earth does there exist a better studies of Hawaiian Drosophila evolution natural laboratory for evolutionary studies have been concerned with understanding the than the Hawaiian Archipelago. Ofthe myr- genetic processes that lead to speciation. Ex- iad plant and animal groups that show experimental work has focused on newly di- plosive speciation in this unique geological verged species pairs which are sympatric and situation, the drosophilids have been most homosequential (i.e., share a common poly- comprehensively investigated. The total tene-chromosome band sequence). Hawaiian drosophilid fauna is now esti- Drosophila silvestris and Drosophila het- mated at nearly 800 species. Among them, eroneura of the planitibia subgroup are such the picture-wings, so called because of pig- a pair of species endemic to the island of ment markings on the wings, are impres- Hawaii, which is less than 0.5 million-years sively large and long-lived and display com- old (McDougall and Swanson, 1972). They plex behaviors including territorial defense are sympatric everywhere except Kohala (the by males and elaborate courtships. Early oldest volcano of the five comprising the work centered on identifying the species and island), where D. heteroneu-ra is absent. The their relationships to one another and to the two species share a third-chromosome in- ecosystem. Phylogenetic relationships of 103 version polymorphism by which they differ picture-winged Hawaiian Drosophila species from other members of the planitibia have been determined, based upon banding subgroup. Drosophila silvestris is polymor- sequences and inversions in the salivary- phic for ten additional inversions which D. gland polytene chromosomes (Carson and heteroneura does not carry. However, mea- Yoon, 1982; Carson, 1983). More recently, sures of chromosome similarity are quite high (Craddock and Johnson, 1979; Rogers' S = 0.67). A careful analysis of heterochro-

I Present address: 50 1 Hahaione Street, Apartment matin distribution by Chang (1984) using 10-G, Honolulu, HI 96824. C-banding and fluorescent-staining meth- 348 J. N. AHEARN AND A. R. TEMPLETON ods demonstrated that these two species are In all members of the planitibia subgroup identical for this character. of Hawaiian Drosophila, the dorsal tibia1 Genetic similarity based on electropho- bristles of the male foreleg are modified as retic variability for enzyme proteins is also long setae that are used in courtship (Spieth, extremely high. Ten populations of D. sil- 198 1). Heterogeneity between taxa exists in vestris and three populations of D. hetero- the number of bristles found in one row neura were surveyed for 12 loci by Crad- (called 5a) lying between rows 5 and 6. All dock and Johnson (1 979). Mean similarity D. heteroneura populations have at most between the two species, calculated by the one bristle while the north and east (Hilo- method of Rogers (1972), was 0.875. Sene side) D. silvestris populations have 10-45 and Carson (1977) sampled 25 loci in both setae in 5a (and are thus "three-rowed). species at three locations and found genetic The south and west (Kona-side) D. silvestris identity (Net's 7) equal to 0.939. This was populations are similar to D. heteroneura not significantly different from the similar- (as well as all other planitibia-subgroup ities among populations of D. silvestris (I = species), with the "two-rowed" form of this 0.96 1) or among populations of D. hetero- character (Carson and Bryant, 1979; Carson neura (I = 0.949). More recently, Hunt et et al., 1982). al. (198 1) and Hunt and Carson (1983) have In one of the earliest studies ofthis species compared DNA sequences of D. heteroneu- pair, Craddock (1975) discovered that ra and D. silvestris. The mean change in crosses of female D. silvestris x male het- average melting-point temperature (AT,,,) eroneura would give F, progeny that were for this pair was 0.68 Â 0.2OC; interpreted fertile in both sexes. Aheam and Val (1975) as 0.6% difference in DNA sequences be- obtained fertile hybrids from the reciprocal tween them. Finally, DeSalle et al. (1 986b) cross. By morphological comparisons with studied restriction-site variation in the mi- laboratory reared F,'s, Kaneshiro and Val tochondrial DNA (mtDNA) of these two (1977) inferred that about 1Â°/ of the flies species. They found the Nei identity (7) be- from these taxa captured from the Kahuku tween populations ofD. silvestris to be 0.782, Ranch population were F, hybrids. Con- and that between populations of D. heter- trasts of mtDNA haplotypes with mor- oneura to be 0.704. The identities between phology also support the idea that a limited populations of different species were only amount of hybridization occurs in natural slightly lower, ranging from 0.633 to 0.689. populations (DeSalle et al., 1986b). How- Morphologically, these species are strik- ever, in spite of extensive collections over ingly different. D. heteroneura has a laterally many years, both the morphological and broadened, anterior-posteriorly com- mitochondria1 analyses indicate that hy- pressed head, in contrast to the character- bridization only occurs at a few locations istic drosophila shape of the D. silvestris (DeSalle and Templeton, 1987). head. In addition, there are several color- Aheam et al. (1 974) and Kaneshiro (1976) pattern differences, which give the overall showed that strong premating or ethological impression that D. heteroneura is a lighter isolation exists between D. silvestris and D. (almost yellow) fly than D. silvestris (nearly heteroneura. Using the "male-choice" black). Using hybridization methods to be method, isolation indices (the excess in fre- discussed shortly, Val (1 977) and Temple- quency of homogamic matings over heter- ton (1977) performed a genetic analysis of ogamic matings) of 0.92 were obtained for five morphological character differences. both female x male species combinations. They concluded that the head-shape differ- On a gross level, however, the courtship rit- ence is due primarily to a single X-linked, uals of the two species are identical (Spieth, sex-limited locus that is epistatically inter- 198 1). acting with several autosomal loci of minor Details of accumulated work on D. effect. Three additional characters (face col- heteroneura and D. silvestris have been or, wing spotting, and mesopleural pigmen- reviewed by Carson (1978, 1982). This tation) were also controlled by interactions species-pair continues to be of interest. Their of autosomal and X-linked genes. In total, chromosomal, allozyme, and DNA-se- the morphological species differences ana- quence similarities are in sharp contrast to lyzed could be attributed to 15-1 9 loci. the striking morphological differences de- COURTSHIP SUCCESS IN DROSOPHILA HYBRIDS 349

TABLE1. Laboratory stocks used in the hybrid crosses.

Species Hawaii (Big Island) collection site Stock designation Chromosomal inversions - D. heteroneura Olaa Forest Reserve (Hilo side) Q71G12 monomorphic Pauahi (Kona side) R79G5 (no information) Hualalai (Kona side) U51Y40 3m D. silvestris Kilauea Forest Reserve (Hilo side) Q48F5 R59G4 U28T2 Pauahi (Kona side) S91B1 (no information) U21B1 (no information) Hualalai (Kona side) U35B (no information) U5 1B 4k2, 4t, 3m

termined by as few as 15 loci. With the ap- instars are completed in 3-4 weeks; and parent fertility of F, interspecific hybrids, metamorphosis to adults lasts another 3-4 behavioral mechanisms are left as a major weeks. Adults that were to be used in ex- isolating barrier. We have undertaken a perimental matings were separated by sex study of F, and F, hybrids between D. het- at eclosion and aged to sexual maturity. eroneura and D. silvestris to determine The hybridization scheme is shown in whether they are indeed viable, fertile, and Figure 1. For the interspecific hybridiza- able to complete the courtship sequence. tions, the Olaa (Q7 1G 12) population of D. This first paper reports our findings on heteroneura and Kilauea (U28T2) popula- courtship success. tion of D. silvestris (differing in 5a setae number) were used. Intraspecific crosses MATERIALSAND METHODS were made between populations from Hu- Laboratory strains of the two species were alalai and either Kilauea or Olaa for com- begun from individual inseminated fe- parison. Crosses were initiated by single males. Ries were collected at locations on pairs with the exception of the female D. both sides of Hawaii (from Olaa Forest Re- heteroneura x male D. silvestris combina- serve and Kilauea Forest on the Hilo side tion. Craddock (1975) failed to obtain any and from Hualalai and Pauahi on the Kona matings in 36 pairs, but Ahearn et al. (1974) side). Designations for these isofemale lab- discovered that such matings could be ob- oratory stocks are given in Table 1. All of tained when two D. heteroneura females these strains and species are homosequen- were confined with one D. silvestris male. tial in the sense that they share the same Intraspecific pairings were carried on for two polytene-chromosome banding pattern. weeks, while interspecific combinations were However, natural populations of D. heter- maintained for 4-8 weeks in order to obtain oneura are polymorphic for an inversion on matings. Each week, the adults were trans- the third chromosome (3m), and natural ferred to new food vials, and the vacant populations of D. silvestris are polymorphic vials were saved and examined one week for a number of-inversions on the third and later for the presence of larvae. Nineteen fourth chromosomes. Table 1 also gives the inseminations were obtained from 43 fe- inversion polymorphisms found in these male D. silvestris x male D. heteroneura strains (the data were kindly provided by pairs (44.1%). Of 163 D. heteroneura fe- H. L. Carson). Stocks and experimental males that participated in trios with a sil- crosses were reared at 18OC and 60% RH vestris male, 41 were inseminated (25.2%). following standard methods developed for Crosses were designated as follows. For the Hawaiian picture-winged species. The intraspecific matings, the first three char- life cycle in the laboratory is completed in acters from the female and male stock des- three months: adults attain sexual maturity ignations were employed. Hence, female in 3-4 weeks; matings occur within 1-2 Hualalai D. heteroneura x male Olaa D. weeks after this; first-instar larvae hatch heteroneura is designated by U5 1-Q7 1. For from eggs in about five days; the three larval each type of intraspecific, within-strain cross J. N. AHEARN AND A. R. TEMPLETON

HYBRIDIZATIONS

INTRASPECIFIC INTERSPECIFIC

9 Hualalai X fl Kilauea or Olaa $Iheteroneura X d silvestris or Q silvestris X^ d heteroneura 6 SibshipI F; Sibship I 11 I

FIG. 1. Hybridization schemes.

(Q7 1G 12 x Q7 1G 12 [Table 21 and U28T2 the N-numbered series referring to female x U28T2 [Table 41) two sets of crosses were D. heteroneura x male D. silvestris (h x s) carried out at different times (in 1975 and and the J-numbered series referring to the 1979 and in 1977 and 1979, respectively). reciprocal cross (s x h). The crosses are shown in Tables 2 and 4 in Tests of courtship ability (Fig. 2) were their temporal order. Additional numbers made on at least two F, and at least four F2 following these initial characters indicate the sibships for both intra- and interspecific hy- F, and F, sibships that were tested. Thus, brids. Single pairs of sexually mature adults U5 1-Q7 1-24-10 was the F2 progeny of F, were placed in 35-cm3 vials containing pair 10, which were descended from pair 24 Wheeler-Clayton medium (Wheeler and of the cross female Hualalai D. heteroneura Clayton, 1965). Pairs were transferred to x male Olaa D. heteroneura. Interspecific new food vials after one week, and the test crosses are labelled in a similar manner, with was terminated at the end of the second

TEST OF FERTILITY-SINGLE FAIR MATINGS

DAYS -8 0" placed in vial(oged 4 weeks)

dissect Q

9 examine vial for larvae FIG. 2. Test of courtship ability by single-pair matings. COURTSHIP SUCCESS IN DROSOPHILA HYBRIDS 35 1 week. Females and males were dissected in a 0.7% saline solution to examine their reproductive tracts. Wet-mount preparations were viewed under phase contrast at 40 x with a Wild M-20 compound microscope. is asymptotically normal with mean 0 and Only pairs that survived the two-week test variance 1 under the null hypothesis that p, period and had normally developed and = P2- functioning reproductive tracts were scored. A second hypothesis of interest is ho- Courtship success was determined by mogeneity among several different crosses. whether the female was virgin or insemi- Suppose a, and its sampling variance are nated (sperm stored in the seminal recep- determined for each of Z different crosses. tacle or spermathecae). To control for cases Let the null hypothesis be that each cross in which a female was inseminated but used represents a sample with the same under- up the sperm during the two-week test pe- lying proportions of inseminated females; riod, the used food vials were retained and that is, all Z crosses are homogeneous. Con- examined for eggs and larvae. Any female sider now the statistic producing larvae was scored as being inseminated. The data consist of the number of inseminated females out of a certain number of pairs in a cross. Let xi be the number of where inseminated females and ni the number of 7 a I pairs in cross i. Then, the proportion of in- ii=2nlÃ and N=~Ã‡, I= I seminated females in cross i, pi, is estimated I= 1 N by xi/ni. The variance of this estimator is p,(l -pi)/ni and hence depends upon the It follows from the asymptotic normality of parameter to be estimated, pi. In order to the ai7sthat H has a chi-square distribution make the variance independent of pi, the with Z - 1 degrees of freedom under the following transformation is made (Freeman null hypothesis of homogeneity of all Z and Tukey, 1950): crosses. Hence, Equation (3) provides a straightforward test of homogeneity. When the null hypothesis of homogeneity is not rejected, the data from the Z crosses can be pooled to create a pooled estimate of arcsin@)^ with variance l/(4N) that has the same asymptotic expectation and variance as a. The resulting transformed proportion, a,, is A modification of the H statistic can also asymptotically distributed as a normal vari- be used to examine the variance compo- ate with mean arcsin (pi1l2and variance nents present in the Z crosses even when the l/(4ni). Hence, the variance of a, does not null hypothesis of homogeneity is rejected. depend on pi and is completely determined If the only source of variance among the Z by the sample size. Because of these desir- crosses were sampling error, the variance of able statistical properties, the analysis of the a should be 1/(4N). An unbiased estimate data will be performed on these transformed of the variance of a in the Z crosses is given proportions rather than on the proportions by: themselves. The asymptotic normality of the ai's is also useful in constructing various tests. First, consider the problem of testing Indeed, the null hypothesis of homogeneity whether or not two proportions, say pi and tested by Equation (3) is equivalent to test- pi, are different. From the normality of the ing the null hypothesis that s$ = 1/(4N); al's, it follows that the statistic that is, the only source of variance among 352 J. N. AHEARN AND A. R. TEMPLETON the I crosses is the sampling error. If the the significant heterogeneity found within null hypothesis of homogeneity is rejected, the F2's, so the t test [Equation (6)] is used this means that sn2 contains a significant instead. The resulting t value is -0.44 (d.J variance component beyond that induced = 5, ns). by sampling alone. This residual compo- Drosophila silvestris Intraspecific Cross- nent is denoted by V and is simply es. -The intraspecific crosses involving D. silvestris are given in Table 4, and the homogeneity test results are in Table 5. All parental crosses are homogeneous, despite different geographical origins, but the Fl and A final hypothesis to be tested is that a F2 crosses are not homogeneous. The inter- single cross (or pooled group of homoge- pretation of the heterogeneity for the F2 neous crosses) has the same proportion of crosses is complicated by the fact that the inseminated females found in a heteroge- F2's were derived from two, nonhomoge- neous group of crosses; that is, the two neous F; crosses. Accordingly, the F2cross- groups of crosses may differ in their vari- es were subdivided into two groups: three ance but share the same mean. Let a, and crosses derived from the F, cross U35-U28- n, be the transformed proportion of insem- 17 (hereafter abbreviated as Fl-17 and Fa- inated females and the sample size, respec- 17 for the respective Fl and F2 crosses) and tively, within the single or homogeneous two derived from the Fl cross U35-U28- 18 group of crosses, and let a and s2 be the (hereafter Fl-18 and Fa-18). As shown in sample mean and variance of the hetero- Table 5, both groups of F2 crosses were ho- geneous group of crosses, as defined previ- mogeneous. Hence, the F2 heterogeneity is ously. Then, the statistic explicable as a continuation of the Fl heterogeneity. Finally, Table 6 gives the results of the z tests for several contrasts. As expected from the H-tests results, the mean proportion of inseminated females is significantly different in the FI-1 7 and FI-18 has an asymptotic t distribution with I - 1 crosses, and likewise for the respective F2 degrees of freedom under the null hypoth- crosses. Because of this heterogeneity, all esis that p is the same in both groups of other contrasts treat the 17 and 18 series of crosses. crosses separately. Interspecific Crosses. -The interspecific RESULTS crosses between D. heteroneura and D. sil- Drosophila heteroneura Intraspecific vestris are listed in Table 7, and the ho- Crosses.-As shown in Table 2, four paren- mogeneity test results are given in Table 8. tal crosses (including crosses both within The top line of Table 8 is based on the pooled and between geographical populations), data from the parental crosses given in Ta- three F, crosses, and six F2 crosses were bles 2 and 4 and compares the mating suc- made within D. heteroneura. The H tests cess of the two species in intraspecific cross- shown in Table 3 indicate that only the F2 es. As shown in Table 8, all parental lines crosses showed significant heterogeneity. are homogeneous between species and, as Moreover, the H test was applied to all pa- previously shown, within species. Hence, all rental and F, crosses pooled together, yield- parental stocks from both species are pooled ing a nonsignificant result. Hence, all pa- in the remainder of the analysis. The Fl rental and Fl crosses are homogeneous. This crosses with D. heteroneura as the female conclusion is reinforced by the z test. The parent (the h x s F,) are homogeneous, but z test of the pooled parental crosses versus the reciprocal (s x h Fl)crosses are not. As the pooled F, crosses was 0.6 1, which is not shown in Table 8, this heterogeneity is significant at the 5% level. caused by one cross, the J-64 F,, with the The z test cannot be used to test the mean remainder of the F, crosses being homoge- proportion of inseminated females from the neous. The h x s F2 crosses are also ho- F2's versus the parental mean because of mogeneous, but the s x h F2crosses are not, COURTSHIP SUCCESS IN DROSOPHZLA HYBRIDS 353

TABLE2. Intraspecific crosses of Drosophila heteroneura. In crosses, the female parent is always denoted first. Stock designations are as follows: Q71G12 = Olaa; U51Y40 = Hualalai; R79G5 = Pauahi.

Number of insemi- Percentage Type of cross Cross designation Number of pairs nated females insemination Parental Q71G12 x Q71G12 Q71G12 x Q71G12 U51Y40 x Q71G12 R79G5 x Q71G12

FI U51-Q71-4 U51-Q71-12 U5 1-Q7 1-24

F2 U51-Q71-4-11 U5 1-Q7 1-4-48 U51-Q71-12-8 U51-Q71-12-15 U5 1-Q7 1-24-9 U51-071-24-10

even when they are subdivided into those The genetic variance released in the F2 F2's derived from the J-64 Fl versus the can be estimated by noting that the 77-test remainder. Finally, there is evidence for sig- results given in Table 3 indicate that the nificant effects on the means as shown by variance in the transformed proportion of the z and t tests given in Tables 9 and 10, inseminated females is explained entirely by respective1y. the expected sampling variance in all pa- DISCUSSION rental and Fl crosses. Hence, there is no environmental variance beyond that asso- Genetic Interpretations of the Results ciated with sampling variance in this case. Drosophila heteroneura Intraspecific The genetic variance observed in the F2 is Crosses. -The intraspecific D. heteroneura simply the total variance minus the envi- crosses did not differ in mean value for the ronmental/sampling variance (0.0009), proportion of inseminated females, but the which is 0.0032 from Equation (5). Hence, F2 crosses showed significant heterogeneity a substantial amount of additive genetic around this mean that is not explicable variation was released in the F2. through sampling alone (Tables 2,3) These Drosophila silvestris Intraspecific Cross- results indicate that the parental stocks al- es. -The crosses involving Drosophila sil- though phenotypically homogeneous, were vestris also revealed evidence for intraspe- genetically diverse and gave rise to an F2 cific variation, but of a different sort than that had increased phenotypic variance due that revealed by the D. heteroneura crosses. to segregation and assortment. The equality Unlike the D. heteroneura crosses, the D. of the means in this series of crosses implies silvestris F, were heterogeneous, and this that the genetic diversity between the pa- heterogeneity persisted into the F2 (Table rental strains is primarily additive in nature, 5). However, the F2 were homogeneous with no evidence for significant dominance when nested within the F, categories from or epistatic effects. which they were derived. Hence, there is no

TABLE3. Homogeneity tests on some intraspecific D. heteroneura crosses. Data from parental and FI crosses were pooled for testing group labelled "Parental + Fl."

Standard Group a deviation of a H d.f P Inference Parental Stocks 1.018 0.032 1.586 3 ns homogeneous FI 0.987 0.027 0.680 2 ns homogeneous Parental + Fl 1.003 0.0 19 2.604 6 ns homogeneous F2 1.054 0.07 1 27.237 5 <0.01 heterogeneous 354 J. N. AHEARN AND A. R. TEMPLETON

TABLE4. Intraspecific crosses of Drosophila silvestris. The female parent is always denoted first. Stock designations are as follows: Q48F5 = Kilauea, U28T2 = Kilauea, U35B = Hualalai.

Number of msemi- Percentage Type of cross Cross designation Number of pairs nated females insemination Parental Q48F5 x Q48F5 U28T2 x U28T2 U28T2 x U28T2 U28T2 x U35B U35B x U28T2 FI U35-U28-17 U35-U28-18 F2 U35-U28- 17-4 U35-U28- 17-32 U35-U28- 17-55 U35-U28-18-14 U35-U28- 18-48 evidence for a release of additive genetic crease in the mean in the F, is lost in the variance in the F2 of the D. silvestris crosses F2-17, which has a mean not significantly of the sort described above for D. hetero- different from the parental mean but sig- neura. Because the heterogeneity found in nificantly lower than the F,-17 mean. This the F, occurs between replicates of the same decline in the F2 mean is even more dra- cross (U35-U28), the heterogeneity cannot matic in the F2-18 crosses. In the "1 8" se- be explained by fixed genetic differences be- ries, there is no evidence for dominance or tween the strains or maternal effects. In- heterosis, as the Fl-18 mean is not signifi- stead, this heterogeneity is most readily ex- cantly different from the parental mean (z plained by intrastrain segregation of alleles = -0.167, ns; Table 6). But like the "17" affecting the phenotype of insemination series, there is a tremendous decline in the success; that is, one or both of the parental F2 mean, such that the F2-18 mean is sig- laboratory strains of D. silvestris are appar- nificantly lower than either the parental or ently still polymorphic at loci affecting the the Fl-18 mean (Table 6). events leading up to a successful insemi- The decline in the F2 means can be ex- nation. plained by two alternative hypotheses. The There is also evidence for interstrain het- first is inbreeding depression. Under this erogeneity in D. silvestris. Although the pa- hypothesis, one or both of the parental rental H value in Table 5 is not significant strains are segregating for deleterious reces- at the 5% level (P = 0.059), the z test of the sive genes. The Fl mean is high because of parental mean versus the Fl-17 mean is dominance, but since the F2 are the result highly significant. The greatly increased of sib matings of the Fl,they will be inbred proportion of inseminated females seen in with a probability of 0.25 of homozygosity the F, is indicative of interstrain differen- by descent. If some of the Fl were hetero- tiation at loci with nonadditive allele effects, zygous for one or more recessive, deleteri- such as dominance or heterosis. The in- ous alleles, the inbreeding in the F2 would

TABLE5. Homogeneity tests on some intraspecific D. silvestris crosses. F2-17 were derived from the Fl cross U35-U28- 17, and F2- 18 were derived from the Fl cross U35-U28- 18.

Standard Group a deviation of a H d.J P Inference Parental 0.996 0.046 9.077 4 ns homogeneous FI 1.168 0.154 9.847 1 <0.01 heterogeneous F2 0.948 0.068 16.683 4 <0.01 heterogeneous F2- 17 1.032 0.023 0.693 2 ns homogeneous F2-18 0.757 0.070 1.364 1 ns homogeneous COURTSHIP SUCCESS IN DROSOPHZLA HYBRIDS 355 lead to frequent homozygosity of these al- TABLE6. Results of z tests on the mean differences leles, with an attendant decline in the mean. between various pooled groups of intraspecific D. sil- However, there are two reasons for believ- vestns crosses. ing that inbreeding depression is an unlikely Comparison z P explanation for these results. First, the parental strains are long-standing lab lines de- Parental vs. Fl- 17 Parental vs. Fl- 18 scended from single wild-caught females. Fl-17 VS. Fi-18 Hence, these strains are themselves rather Fl-17 vS. F2-17 inbred. Consequently, it is improbable that Fl-18 vs. F2-18 the parental strains would be segregating for F2-1 7 vs. F2-18 Parental vs. F2- 17 the type of genetic variation necessary for Parental vs. F2- 18 this explanation. A way of avoiding this difficulty would be to assume that deleterious alleles at different loci are associated with different inversion polymorphisms. This One can still argue for an inbreeding effect would result in a marginal overdominance of a different sort. Suppose the parental for the inversion polymorphism, and hence strains are fixed, not polymorphic, for some these deleterious alleles would remain seg- recessive, nonepistatic, deleterious genes. regating within the strains as part of the The F, would be relieved from the delete- inversion polymorphism. We know that the rious effects of these genes due to hetero- D. silvestris strains are polymorphic for a zygosity, but the inbreeding in the F, would number of inversions (Table I), so this is a lower the F2 mean. However, this modified real possibility. Moreover, Carson (pers. inbreeding argument still does not fully ex- comm.) has found that inversion polymor- plain the data. This hypothesis predicts an phism~found in these strains are rarely lost, increase in Fl mean relative to that of the despite the bottlenecks and small popula- parental strains. While this was true for the tion sizes associated with laboratory-culture "17" series (aF, = 1.316, aparental= 0.959, z conditions. This observation is consistent = -3.624, P < 0.01), it was not for the "18" with the prediction of marginal overdomi- series (aF, = 1.008, aparental= 0.959, z = nance of the inversions. However, there is -0.49, ns). Moreover, since the F2 would a second difficulty that is more difficult to have levels of heterozygosity at these dele- explain: namely, there was no increase in terious loci that are intermediate between variance demonstrable in the F,. In organ- the parental and Fl levels of heterozygosity, isms such as Drosophila, which have rela- the mean phenotype should likewise be in- tively low chromosome numbers and no re- termediate. This is definitely not the case combination in males, a large variance for the "18" series of crosses; nor is the F, should be induced in the amount of hom- mean significantly greater than the parental ozygosity by descent manifested in individ- mean in the " 17" series. uals sharing the same inbreeding level We therefore turn to our second expla- (Franklin, 1977). Hence, the variance of the nation: coadaptation. Under this explana- F, should increase substantially over that tion, the parental strains have achieved the observed for the parental or F, stocks. This same mean phenotype by genetically di- is particularly true if the deleterious effects verse systems. Moreover, the loci contrib- are associated with inversions, since under uting to these phenotypic systems are char- this hypothesis only a small number of seg- acterized by much epistasis. (Note that, in regating elements are responsible for the ef- this case, epistasis refers not only to non- fects, and the variance would accordingly additive interactions between loci to pro- be expected to be quite high in the F,. As duce the phenotype ofan individual but also shown in Table 5, no increase in F, variance to nonallelic interactions between geno- was detected. Hence, an inbreeding depres- types in different individuals, because our sion caused by deleterious alleles or mar- unit of observation is a male-female pair.) ginally overdominant inversions still seg- This explanation still requires some sort of regating within the parental strains is not dominance or overdominance at some of supported by these data. the loci that differentiate the strains in order 356 J. N. AHEARN AND A. R. TEMPLETON

TABLE7. Interspecific cross between D. heteroneura vere enough such that the F, mean lies be- and D. silvestris. The D. heteroneura were from the low the parental means, as occurred in the Olaa Forest Reserve population, and the D. silvestris "1 8" series. The differences between the F2 were from the Kilauea Forest Reserve population. means in the " 18" and " 17" series could be

Number explained by the same polymorphic system of insemi- Percentage Type of Cross Number nated insemi- responsible for the F1 heterogeneity, which cross designation of pairs females nation may or may not be part of this epistatic heteroneura x silvestris complex. Finally, if a sufficient number of Fl N-24 epistatically interacting loci are involved in N-47 this system, such that virtually all the ga- F2 N-24-39 metes produced by the F, are disrupted, the N-24-40 decline in the F, mean could occur without N-47- 12 N-47- 14 an increase in F2 variance. Interspecific Crosses.-The interspecific silvestris x heteroneura crosses yield evidence for both intra- and FI J-64 J-73 interspecific variation. As with the D. sil- J-79 vestris crosses, significant heterogeneity was F2 J-64-13 found within the Fl, with the J-64 F, being J-64-26 significantly different from the four other F1 J-73- 1 crosses (Table 8). This heterogeneity is not J-73-3 due to maternal effects, because the J-64 J-79- 1 J-79-2 cross (an s x h cross) is significantly different both from the other s x h crosses and from the h x s crosses, which in turn are all homogeneous amongst themselves. to explain the increased insemination rates Hence, just as with the Fl heterogeneity of the Fl. Segregation at such dominant or found within D. silvestris, the simplest ex- overdominant loci would contribute to an planation is polymorphism in one or more F2 decline, but as previously argued, this of the parental strains. Indeed, it is possible explanation is incomplete. However, by that the polymorphic loci responsible for adding epistasis, the F2 data can now be the F, heterogeneity found within D. silves- explained. When the F1reproduce, these co- tris could also be responsible for the inter- adapted systems are disrupted by recom- specific Fl heterogeneity. bination and assortment. Because of the dis- The fact that all the F1 interspecies pairs ruption of these epistatic complexes, the F2 except for J-64 lead to 100% sperm transfer mean declines beyond what is expected from may also indicate intraspecific polymor- segregation of dominant alleles alone. With phism. No other set of crosses had such a sufficient epistasis, this decline could be se- rate of insemination, even when J-64 was

TABLE8. Homogeneity tests between and within parental strains of both species and on some interspecific crosses. "All parental" refers to the pooled data from the parental crosses given in Tables 2 and 4; h x s = female heteroneura x male silvestris; and s x h = female silvestris x male heteroneura.

Standard deviation Group a of a Inference All parental homogeneous hx sF1 homogeneous sXhF1 heterogeneous s x h Fl (J-73 and J-79) homogeneous All Fl excluding s x h J-64 homogeneous hx SF? homogeneous sxhF2 heterogeneous s x h F2 (J-64) heterogeneous s x h F2 (J-73 and J-79) heterogeneous COURTSHIP SUCCESS IN DROSOPHZLA HYBRIDS 357

TABLE9. Results of z tests on the mean differences TABLE10. Results oft tests on the mean differences between various pooled groups of interspecific crosses. between various pooled groups of intraspecific D. silvestris crosses. Groups compared z P Parental vs. J-64 Fl -2.066 <0.05 Groups compared t d.JP Parental vs. J-73 and J-64 Fl vs. J-64 F2 1.006 1 ns

J-79 Fl - 1 1.277 4.01 Parental vs. J-64 FZ 0.082 1 ns h x s Fl vs. h x s F2 5.363 <0.01 J-73 and J-79 Fl vs. Parental vs. h x s F2 -4.470 <0.01 J-73 and J-79 F2 4.565 3 <0.05 Parental vs. .I-73 and averaged in with the others. Recall that only a minority of the interspecific crosses were relative to the F, mean. From Table 9, the successful in producing Fl's (see Materials h x s F2 mean is significantly lower than and Methods). Hence, there is a strong po- that of the h x s Fl mean, but it is also tential for selection on the original parental significantly larger than the parental mean. individuals involved in these interspecific The lowering of the F2 mean could be ex- crosses. If the parental populations were ge- plained by an epistatic gene complex as pre- netically variable in those traits influencing viously explained for the intraspecific D. sil- the chances for interspecific mating, this se- vestris crosses. However, because the F, lection could be effective in altering the mat- mean is intermediate between the parental ing behaviors of the resulting Fl's. The data and F1 means (Table 8), these results could are consistent with the hypothesis that se- also be explained through dominance or lection on the interspecific pairs and trios heterotic effects or inbreeding depression. favored those females that are less discrim- The evidence for coadaptation is more com- inating, and as a consequence, the resulting pelling in the s x h series of crosses. Con- F, females were likewise less discriminat- fining attention first to the J-73 and J-79 ing. Such a situation would cause the F, series, the F, mean is significantly higher insemination rates to increase, as they did than the parental mean (Table 9), but the indeed. F2 mean is significantly lower than the F, There is also evidence for interspecific dif- mean but not significantly different from the ferences in the genes controlling the phe- parental mean (Table 10). This is exactly notypes leading up to successful insemina- the same pattern observed in the " 17" series tion. First, there was a significant increase within D. silvestris (Table 6). The pattern in the variance observed in the s x h F2's observed in the J-64 series also is similar that could not be explained by the hetero- to that obtained with the "17" series, but geneity of the F['S. The estimated genetic the large variance present in the J-64 F2 variance [from Equation (5)] in the J-64 F, prevents the F2 mean from being signifi- is 0.024 (with an environmental variance of cantly different from either the parental or 0.002), and it is 0.007 (with an environ- Fl means. These results are indicative of mental variance of 0.00 1) in the remaining coadaptation, but the alternatives cannot be s x h F2. Thus, there is a significant release so clearly rejected in this case as they were of genetic variation in the interspecific F2's. for the within D. silvestris crosses. The significant effects on the means shown in Tables 9 and 10 also provide evidence General Discussion for interspecific differentiation. All Fl cross- Whether a female has sperm in the stor- es have a significantly higher proportion of age organs after a two-week period with a inseminated females than the parental single male measures the successful com- stocks, thereby indicating interspecific dif- pletion of a very complex chain of stimuli ferentiation coupled with dominance or het- and responses which culminate in copula- erotic effects in the F, or possible sexual tion. Male-female courtship interaction in- selection on intraspecific variation in the volves action patterns supported by mor- parental strains, as previously mentioned. phological structure. Paterson (1978) has There is also a reduction of the F2 mean termed this stimulus-response chain the 358 J. N. AHEARN AND A. R. TEMPLETON

"mate-recognition system." There is no a and Teramoto, 1984). Intrastrain variabil- priori requirement that male and female ity for these bristle traits is indicated by the components be under the control of the same experiments of Carson and Teramoto set of genes. However, the male and female (1984), who selected for high and low num- components of this system must be com- bers of row-5a bristles within the Kilauea patible, and as such, they are under strong population. They obtained a strong and sexual selection. Paterson (1985) has also rapid response to this selection, thereby re- claimed that the mate-recognition system vealing the presence of much intrastrain ge- shows little or no variation within a species; netic variability. These results, in conjunc- indeed, he defines a species as a group of tion with ours, are highly suggestive that organisms that share a single, common morphology and behavior are allied in the mate-recognition system. To the extent that mate-recognition system and that the mate- the phenotype of successful insemination recognition system displays intraspecific measures the mate-recognition system, the variability, both within and between local data we present here do not support the populat'ions. notion that the mate-recognition system is Our results also have some interesting an invariable, species-defining attribute. implications for the status of the three taxa First, the data indicate that there is a con- used in our study; namely, Drosophila het- siderable amount of genetic variability in eroneura, two-row D. silvestris, and three- the phenotypes leading to successful insem- row D. silvestris. In the laboratory, D. ination between the tested strains within D. heteroneura and D. silvestris show strong heteroneura. Likewise, the shifts in the F, sexual isolation (Aheam et al., 1974; Kane- and Fa means within D. silvestris provide shiro, 1976); that is, their respective mate- evidence for interstrain differentiation. recognition systems are quite exclusive. In Moreover, the F, heterogeneity found with- order for us to obtain our hybridizations, in D. silvestris implies intrastrain variability we had to resort to long-term isolation of as well. All these results indicate that con- single pairs or trios, which presumably broke siderable intraspecific variability exists in down the mate-recognition system. The both species for phenotypes involved in molecular and morphological studies on hy- successful insemination. bridization in nature also support the idea This conclusion is consistent with other that interspecific hybridization is very lim- studies on these species. As mentioned ear- ited in occurrence (DeSalle and Templeton, lier, the Hualalai and Kilauea populations 1987). In light of these results, it is not sur- of D. silvestris (the populations involved in prising that we obtained some evidence for our intraspecific crosses) differ for the male incompatible coadaptation of successful in- row-5a tibial-bristle trait, with Hualalai semination between these species. How- being a "two-row" population and Kilauea ever, the evidence for coadaptation is a "three-row" population. Spiess and Car- stronger for the intraspecific crosses of D. son (198 l) demonstrated that, when given silvestris. The intraspecific crosses of D. sil- a choice, three-row females prefer three-row vestris and the interspecific crosses share one males over two-row males. Kaneshiro and common feature: namely, they were all Kurihara (1982) also observed these mate crosses between two-row and three-row taxa. preferences between two-row and three-row Consequently, our results are best sum- forms. It may be that the presence or ab- marized as showing differentiated coadapt- sence of these bristles themselves influence ed gene complexes between two-row flies mate preference. Spieth's (198 1) descrip- and three-row flies, rather than between D. tions of the courtship rituals of D. hetero- silvestris and D. heteroneura per se. neura and D. silvestris include the use of the This conclusion overlays very well upon male's forelegs to stimulate the female's ab- recent molecular studies on the evolution- domen. The difference between the two-row ary relationships between these taxa. De- and three-row forms of D. silvestris is ge- Salle et al. (19866) used restriction-site netically based, depending upon at least one mapping of mtDNA to reconstruct the evo- X-linked and two autosomal segregating lutionary history of this group. Their anal- factors (Carson and Lande, 1983; Carson ysis indicates that the first split among these COURTSHIP SUCCESS IN DROSOPHZLA HYBRIDS 359 taxa was not between D. heteroneura and early, once again favoring the less discrim- D. silvestris, but rather between two-row inating females. All of these alterations in and three-row D. silvestris. D. heteroneura sexual-selection pressures can lead to a dra- later split off from the two-row D. silvestris. matic and rapid alteration of the mate-rec- Thus, we have evidence for partially incom- ognition system in the founding population, patible coadapted complexes between the yielding an asymmetrical pattern of sexual taxa created by the older split in this trio. isolation (DeSalle and Templeton, 1987). Our data, in conjunction with others, can One necessity for this model is the exis- be used to define a parsimonious model of tence of intraspecific variability in the mate- the origins of D. heteroneura, two-row D. recognition system in the ancestral popu- silvestris, and three-row D. silvestris. The lation. In this paper, we have demonstrated ancestor of this Big Island (Hawaii) group genetic variation for the mate-recognition is molecularly most closely related to D. system at the intraspecific level, thereby of- planitibia of Maui (DeSalle et al., 1986a; fering support for this model. Ahearn (1980) DeSalle and Templeton, 1987). An inter- has previously demonstrated that in D. sil- island founder event occurred from the an- vestris the mate-recognition system can be cestral population of D. planitibia on Maui perturbed and altered by new founder to establish a two-row ancestral D. silvestris events. In this paper, we have shown that population on the Big Island volcano of Hu- the F,'s resulting from interspecific crosses alalai (Spieth, 198 1; Kaneshiro and Kuri- (the crosses with the greatest potential for hara, 1982). This founder event resulted in strong sexual selection favoring less dis- major changes in the mate-recognition sys- criminating females) had the highest insem- tem, leading indirectly to strong sexual iso- ination rate, thereby indicating the effec- lation between D. silvestris and D. planitibia tiveness of selection for less discriminating (Kaneshiro, 1976). This alteration of the types, as the model given in DeSalle and mate-recognition system in the founder Templeton (1987) predicts. population could have taken place in the Once this altered two-row D. silvestris context of a genetic transilience (i.e., a rapid ancestor was established on Hualalai. sub- adaptive shift in a previously stable genetic sequent speciation events occurred within system or systems that is triggered by alter- the Big Island (Hawaii). The next event that ations in the genetic environment induced occurred was an intraisland, intervolcano by the founder event, often by the fixation founding event in which a founder popu- or near fixation of a few major genes; Tem- lation was established on Kohala from the pleton, 1980). The mate-recognition system Hualalai populations. All of these taxa are is particularly subject to destabilization dur- high-altitude rainforest species, and the ap- ing a founder event (Templeton, 1979; Gid- propriate habitats on Hualalai and Kohala dings and Templeton, 1983; DeSalle and are separated by about 40 km of dry, lower- Templeton, 1987). The signal-response na- altitude terrain. Thus, this intervolcanic ture of the mate-recognition system yields transfer is of the same order of difficulty as a type of neutral stability; that is, as long as many of the interisland transfers. The the signal invokes the correct response, the founder population at Kohala may have also system works, but there is often nothing in- undergone a genetic transilience in its herently superior about one particular sig- mate-recognition system to establish the nal versus another, as long as the signal and three-row form of D. silvestris. The genetic response are compatible. Hence, random architecture underlying this transilience in- changes in a signal or response induced by volved epistatically interacting genes, which a founder event could trigger a bout of in- resulted in the incompatible coadaptation tense sexual selection that leads to a new, we observe between two-row and three-row compatible signal-response system. More- forms. The three-row Kohala population over, the initial founding population has low subsequently spread down the east side of densities that would favor females that are the island by a series of linear colonizations, less discriminating, and during the popu- as supported both by behavioral (Kaneshiro lation flush that follows the founder event, and Kurihara, 1982) and molecular (De- r-selection would favor females that mate Salle and Templeton, unpubl.) evidence. The 360 J. N. AHEARN AND A. R. TEMPLETON spread of the three-row form on the eastside Studies such as these not only provide tests of the island would be much easier than the of speciation models but also contribute to Hualalai-to-Kohala transfer, because the our understanding of one of the potential major volcanoes on this side of the island contributors to speciation: the evolution of are closer together and are separated only a new mate-recognition system. by high, wet saddles. The two-row D. silvestris form was able to spread to the south and then to the east Many individuals have been indispens- as the still-active volcano of Mauna Loa able to us in this work, and we acknowledge formed the appropriate high-altitude habi- and offer our mahalo to L. Teramoto for tat, including a rather high, wet saddle be- technical assistance, G. N. Nishimura and tween Mauna Loa and Hualalai. Also, G. Shiraki for stock keeping and lab assis- sometime during this period, another spe- tance, K. Kaneshiro, A. Ohta and G. Wat- ciation event occurred that split off D. het- son for field collections, G. Shiraki for il- eroneura from two-row D.silvestris. Unfor- 1ustrations;and H. L. Carson, T. W. Lyttle, tunately, the available evidence offers little E. Shimakawa, and two anonymous review- insight into the nature of this speciation ers for comments on the manuscript. This event, other than that it also resulted in a work was supported by NSF grants substantial alteration of the mate-recogni- GB27597, GB29288, and DEB 74-22532 to tion system, probably including the drastic H. L. Carson and by NIH grant GM3 157 1 change in head shape. The genetic basis of to A.R.T. the head-shape change is a major X-linked segregating unit that epistatically interacts with several minor, autosomal modifiers AHEARN,J. N. 1980. Evolution of behavioral reproductive isolation in a laboratory stock of Drosophila (Templeton, 1977). Once formed, D. het- silvestris. Experientia 36:63-64. eroneura was able to spread to the east side AHEARN,J. N., H. L. CARSON,TH. DOBZHANSKY,AND of the island, except in the Kohala region, K. Y. KANESHIRO.1974. Ethological isolation thereby making it sympatric with the two among three species of the planitibia subgroup of allopatric forms of D. silvestris. Hawaiian Drosophila. Proc. Nat. Acad. Sci. USA 71:901-903. The genetic-transilience model in general, AHEARN,J. N., AND F. C. VAL. 1975. Fertile inter- and its application to sexually selected sys- specific hybrids of two sympatric Hawaiian Dro- tems in particular, has proven to be quite sophila. Genetics 80:s9. controversial (Barton and Charlesworth, BARTON,N., AND B. CHARLESWORTH.1984. Genetic revolutions, founder effects, and speciation. Ann. 1984; Carson and Templeton, 1984; Ehr- Rev. Ecol. Syst. 15: 133-1 64. man and Wasserman, 1987; DeSalle and CARSON,H. L. 1978. Speciation and sexual selection Templeton, 1987). Regardless of whether in Hawaiian Drosophila, pp. 93-107. In P. F. one favors this model or not, one of the Brussard (ed.), Ecological Genetics: The Interface. Springer-Verlag, N.Y. principal strengths of the genetic-transil- . 1982. Evolution of Drosophila on the newer ience model is that it makes specific and Hawaiian volcanoes. Heredity 48:3-25. testable predictions. One such prediction is . 1983. Chromosomal sequences and interis- that there should be genetic polymorphism land colonizations in Hawaiian Drosophila. Ge- within species for phenotypes affecting the netics 102:465^182. CARSON,H. L., AND P. J. BRYANT. 1979. Change in mate-recognition system. We feel that the a secondary sexual character as evidence of incip- data presented in this paper offer strong sup- ient speciation in Drosophila silvestris. Proc. Nat. port for that prediction. Another prediction Acad. Sci. USA 76: 1929-1 932. is that the genetic architecture underlying CARSON,H. L., AND R. LANDE. 1983. Inheritance of alterations in the mate-recognition system a secondary sexual characteristic in Drosophila silvestris. Genetics 104:s 12. should often involve epistasis and major CARSON,H. L., AND A. R. TEMPLETON.1984. Genetic genes. The work presented here offers sup- revolutions in relation to speciation phenomena: port for the prediction of epistasis, and past The founding of new populations. Ann. Rev. Ecol. work (Templeton, 1977) on the genetic basis Syst. 15:97-131. CARSON,H. L., AND L. T. TERAMOTO.1984. Artificial of head-shape differences between D. sil- selection for a secondary sexual character in males vestris and D. heteroneura offers strong evi- of Drosophila silvestris from Hawaii. Proc. Nat. dence for both major genes and epistasis. Acad. Sci. USA 8 1:39 15-39 17. COURTSHIP SUCCESS IN DROSOPHIU HYBRIDS 36 1

CARSON,H. L., F. C. VAL, C. M. SIMON,AND J. W. HUNT,J. A., T. J. HALL,AND R. J. BRITTEN. 1981. ARCHIE.1982. Morphometric evidence for incip- Evolutionary distances in Hawaiian Drosophila as ient speciation in Drosophila silvestris. Evolution measured by DNA reassociation. J. Molec. Evol. 36:132-140. 17:361-367. CARSON,H. L., AND J. S. YOON. 1982. Genetics and KANESHIRO,K. Y. 1976. Ethological isolation and evolution of Hawaiian Drosophila, pp. 297-344. In phylogeny in the planitibia subgroup of Hawaiian M. Ashburner, H. L. Carson, and J. N. Thompson Drosophila. Evolution 30:740-745. (eds.), The Genetics and Biology ofDrosophila, Vol. KANESHIRO,K. Y., AND J. S. KURIHARA.1982. Se- 3b. Academic Press, N.Y. quential differentiation of sexual behavior in pop- CHANG,L. S. 1984. A comparison of heterochro- ulations of Drosophila silvestris. Pac. Sci. 35: 177- matin distribution among five closely related species 183. of Hawaiian Drosophila: The planitibia subgroup KANESHIRO,K. Y., AND F. C. VAL. 1977. Natural M.S. Thesis, Univ. Hawaii, Honolulu. hybridization between a sympatric pair of Hawai- CRADDOCK,E. M. 1975. Reproductive relationships ian Drosophila. Amer. Natur. 11 1:897-902. between homosequential species of Hawaiian Dro- MCDOUGALL,I., AND D. A. SWANSON.1972. Potas- sophila. Evolution 28:593-606. sium-argon ages of lava from the Hawi and Pololu CRADDOCK,E. M., AND W. E. JOHNSON.1979. Ge- volcanic series, Kohala Volcano, Hawaii. Geol. Soc. netic variation in Hawaiian Drosophila. V. Chro- Amer. Bull. 83:3731-3738. mosomal and allozymic diversity in Drosophila sil- PATERSON,H. E. H. 1978. More evidence against vestris and its homosequential species. Evolution speciation by reinforcement. South Afr. J. Sci. 74: 33:137-155. 369-37 1. DESALLE,R., L. V. GIDDINGS,AND K. Y. KANESHIRO. . 1985. The recognition concept of species, pp. 1986a. Mitochondria1 DNA variability in natural 21-29. In E. S. Vrba (ed.), Species and Speciation. populations of Hawaiian Drosophila. 11. Genetic Transvaal Museum, Pretoria, S. Afr. and phylogenetic relationships of natural popula- ROGERS,J. S. 1972. Measures of genetic similarity tions of D. silvestris and D. heteroneura. Heredity and genetic distance. Univ. Texas Publ. Stud. Ge- 56:87-96. net. 7213:145-153. DESALLE,R., L. V. GIDDINGS,AND A. R. TEMPLETON. SENE,F. M., AND H. L. CARSON.1977. Genetic vari- 1986b. Mitochondria1 DNA variability in natural ation in Hawaiian Drosophila. IV. Allozymic sim- populations of Hawaiian Drosophila. I. Methods ilarity between D. silvestrisand D. heteroneura from and levels of variability in D. silvestris and D. het- the island of Hawaii. Genetics 86:187-198. eroneura populations. Heredity 56:75-85. SPIESS,E. B., AND H. L. CARSON.1981. Sexual selec- DESALLE,R., AND A. R. TEMPLETON.1987. Com- tion in Drosophila silvestris of Hawaii. Proc. Nat. ments on "The significance of asymmetrical sexual Acad. Sci. USA 78:3088-3092. isolation." Evol. Biol. 21 :2 1-27. SPIETH,H. T. 198 1. Drosophila heteroneura and Dro- EHRMAN,L., AND M. WASSERMAN.1987. The signif- sophila silvestris: Head shapes, behavior and evo- icance of asymmetrical sexual isolation. Evol. Biol. lution. Evolution 35:921-930. 21:l-20. TEMPLETON,A. R. 1977. Analysis of head shape dif- FRANKLIN,I. R. 1977. The distribution of the pro- ferences between two interfertile species of Hawai- portion of the genome which is homozygous by ian Drosophila. Evolution 3 1:630-64 1. descent in inbred individuals. Theroet. Popul. Biol. . 1979. Once again, why 300 species of Hawai- 11 :60-80. ian Drosophila? Evolution 33:5 13-5 17. FREEMAN,M. F., AND J. W. TUKEY.1950. Transfor- . 1980. The theory of speciation via the found- mations related to the angular and the square root er principle. Genetics 94: 10 11-1038. Ann. Math. Stat. 2 1:607-6 1 1. VAL,F. C. 1977. Genetic analysis of the morpholog- GIDDINGS,L. V., AND A. R. TEMPLETON.1983. Be- ical differences between interfertile species of havioral phylogenies and the direction ofevolution. Hawaiian Drosophila. Evolution 3 1:6 11-629. Science 220:372-378. WHEELER,M. R., AND F. E. CLAYTON.1965. A new HUNT,J. A., AND H. L. CARSON.1983. Evolutionary Drosophila culture technique. Dros. Inf. Serv. 40: relationships of four species of Hawaiian Drosoph- 98. ila as measured by DNA reassociation. Genetics 94:553-564. Corresponding Editor: J. B. Spofford