Copyright 0 1996 by the Genetics Society of America

The history ofDrosophila Pseudoobscura and Close Relatives: Inferences from DNA Sequence Variation at the Period Locus

Rong-Lin Wang' and Jody Hey Department of Biological Sciences, Rutgers University, Piscataway, NewJersey 08855-1 059 Manuscript received June 3, 1996 Accepted for publication August 3, 1996

ABSTRACT Thirty-five period locus sequences from pseudoobscura and its siblings , D. p. bogotana, D. persirnilis, and D.rniranda, were studied. A largeamount of variation was found within D. pseudoobscura and D.persirnilis, consistent with histories of large effective population sizes. D. p. bogotana, however, has a severe reduction in diversity. Combined analysis of perwith two other loci, in both D.p. bogotana and D. pseudoobscura, strongly suggest this reduction is due to recent directional selection at or near per within D.p. bogotana. Since D. p. bogotana is highly variable and shares variation with D.pseudoobscura at other loci, the low level of variation at perwithin D. p. bogotana can not be explained by a small effective population size or by speciation via founder event. Both D. pseudoobscura and D. persirnilis have consider- able intraspecific gene flow. A large portion of one D.persirnilis sequence appears to have arisen via introgression from D. pseudoobscura. The time of this event appears to be well after the initial separation of these two species. The estimated times since speciation are one mya for D. pseudoobscura and D. persirnilis and 2 mya since the formation of D. miranda.

VOLUTIONARY biologists areconcerned with One of the best studied species groups includes Dro- E both the population genetic mechanisms behind sophila pseudoobscura and its close relatives:D. persirnilis, evolutionary processes and the reconstruction of phy- D.miranda, and subspecies D.p. bogotana. D.pseudoobsc- logeny. The study of speciation is at the interface of ura and D. persirnilis were once thought to be a single these two areas of investigation. Knowledge of specia- species but were later differentiated from each other tion is fundamental to our understanding of how the based primarily on different chromosome arrange- biological landscape grows and changes at a macrolevel. ments and nearly complete sexual isolation (DOBZHAN- Yet it is the forces at the microlevel, including mutation, SKYand EPLING1944). Females ofboth species discrimi- migration, random genetic drift, and natural selection, nate against heterospecific maleswhile males court that lead to speciation. females of either species (MAYR 1946). Crossed under Speciation, as a genetic process, can be studied from laboratory conditions in either direction,they generate two different empirical perspectives. First are studies of sterile males but fertile females. There is some chromo- the genetic basis of species differencesand of reproduc- some inversion (DOBZHANSKY1973) and mitochondrial tive barriers among species. These types of studies are DNA data (POWELL 1983; HALEand BECKENBACH1985) focused on the number and identity of loci at which that suggest some degree of gene flow between these divergence may have, in some sense,contributed to speci- two species. ation (Wu and PALOPOLI1994). Whilecritically im- Later came the discovery of D. miranda (DOBZHANSKY portant, these analyses do not inform on the historical 1935) and D. pseudoobscura bogotana (DOBZHANSKYet al. microlevel forces associated with speciation. A different 1963), anisolated population in Bogota, Columbia. Re- perspective can be gained by studying patterns of DNA productive isolation between D. miranda and its sibling sequence variation, withinand among closely related spe- species is essentially complete (DOBZHANSKYand cies, at loci that may or may not have been associated EPLING1944) but that between D. pseudoobscura and D. with species formation. This approach is essentially an p. bogotana is not: female D.p. bogotana crossed to male extension of traditional population genetic questions to D. pseudoobscura giverise to sterile males and fertile distant time periods, and it is made possible by a recent females (PRAKASH1972). synergism between genealogicalpopulation genetic mod- D.pseudoobscura has the widest range of geographic els and studies on DNA sequence variation (see e.g, AVISE distribution within the group, extending from British and BALL 1990; HEY1994; TEMPLETON 1994). Columbia, along thewestern third of the North Ameri- can continent, to Mexico and Guatemala. D. persirnilis Cmespondzng author:Jody Hey, Department of Biological Sciences, and miranda are sympatric to pseudoobscura and Rutgers University, Nelson Labs, P.O. Box 1059, Piscataway, NJ 08855 D. D. 1059. E-mail: [email protected] found in the mountain ranges along the US. Pacific 'Present address: Department of Plant Biology, University of Minne- Coast. Morphological, chromosomal, and molecular ev- sota, 1445 Gortner Ave., St. Paul, MN 55108-1020. idence indicate that D.pseudoobscura is most closely re-

Genetics 144: 1113-1126 (November, 1996) 1114 R.-L. Wang and J. Hey

TABLE 1 A list of samples and their geographic origins

Species NDSRC" lines Sample numbers Locations D. pseudoobscura 0121.0 1 Tucson, A2 0121.1 2 Death Valley, CA 0121.33 3 Kelowna, British Columbia 0121.34 4 Cuernavaca (Mexico) 0121.38 5 Mather, CA 0121.41 6 Mather, CA 0121.37 7 Mather, CA 0121.32 8 Zimapan, Hidalgo (Mexico) 0121.78 9 Dilley, TX 0121.79 10 Port Coquitlam, British Columbia, Canada 0121.83 11 Spray, OR D. miranda 0101.4 22 Port Coquitlam, British Columbia 0101.3 23 Port Coquitlam, British Columbia 0101.5 24 Port Coquitlam, British Columbia 0101.7 25 Port Coquitlam, British Columbia D. persirnilis 0111.0 40 Cold Creek, CA 0111.2 41 Mt. San Jacinto, CA 0111.17 42 Port Townsend, Washington 01 11.23 43 Victoria, British Columbia 0111.24 44 McDonald Ranch, CA 0111.1 45 Quesnal, Canada 0111.18 46 McDonald Ranch, CA 0111.27 47 Mt. San Jacinto, CA 0111.29 48 Mt. San Jacinto, CA 0111.35 49 Mt. San Jacinto, CA 0111.38 50 Victoria, British Columbia D. p. bogotana 0121.35 60 Bogota, Columbia 0121.68 61 Bogota 0121.69 62 Bogota 0121.71 63 Bogota Oicata 9A 67 Bogota Sutatausa 4 70 Bogota Chiquinquira A 73 Bogota A Cacientes 4C 74 Bogota EL Recreo 77 Bogota National Drosophila Species Resources Center (Bowling Green, OH) lated to D.p. bogotana, and together, they share a com- POWELL1992). To date there has not been a nuclear mon ancestor with D. persirnilis (GODDARDet al. 1990). gene DNA sequence study on variation in D. persirnilis. D. rniranda is an outgroup to all other three species. Our study has focused on a 1.5-kb portion of the Together, D. pseudoobscura and its close relatives have period locus (per). Per is an essential component of the beenthe subject of vigorous population genetic re- biological clock in D. rnelanogaster (KONOPKAand BEN- search for several decades (LEWONTINet al. 1981). At ZER 1971; ROSBASHand HALL 1989; EDERYet al. 1994) the DNA level, much of the focus has been on D.pseudc- and found to occur as a single copy gene on the X obscura and onthe relationship between D. pseudoobscura chromosome. Per is also involved in the determination and D. p. bogotana. D. pseudoobscura is clearly a species of male courtship songs (Kmcou and HALL 1980, with a large historical effective population sizewith 1989; Kmcou 1990) which in turn affect the female ample gene flow across much of its range (RILEY et al. response to male mating attempts (KYRIACOUand HALL 1992; SCHAEFFERand MILLER1992; VEUILLEand KING 1982, 1984; K~COUet al. 1992). Per thus appears to 1995). There is also strong evidence for the action of be a candidate for evolutionary changes in mating be- natural selection acting on third chromosome inversion havior that contributeto and spe- polymorphisms (AQUADROet al. 1991; POWELL1992). ciation. However, in the rnelanogaster species complex, D. pseudoobscura and D. p. bogotana share considerable there is no evidence that variation at perhas contributed DNA sequence variation at the Alcohol dehydrogenase to reproductive isolation or to species divergence (KLI- locus (Adh either because of ongoing gene flow or re- MAN and HEY 1993;RITCHIE and KYRIACOU1994). Re- cent divergence or both (SCHAEFFERand MILLER1991; gardless of whether evolution at per contributes to re- Speciation 1 D. psevdoohscura Speciation 115 exon II 111 IV V VI VI1 Vlll

"threonine - glycine"- like repeats

j/ * _,. FIGURE1.-Diagram of the period locus! D. "z psmdoohscurn, based on COI.OT rt al. 1988. bp total 4 4763 b

-1.5 kb sequenced productive isolation, a study of sequence variation estimate of the population recombination parameterwas ob- within and amongclosely related species can yield valu- tained using the method of HUDSON(1987). Theestimate of able information on species divergence. divergence times among the species is based on the substitu- tion rates at synonymous sites (SHARP andLr 1989). The primary objective of this studyis to gain a genea- Outlier HKA test: The HKA test of natural selection is simi- logical perspectiveof the population genetic forces asso- lar to a chi-square test and asks whether the relative levels of ciated with the formation of the two sibling species, D. intraspecific polymorphism and interspecific divergence for pseudoohscura and D. pmirnilk, as well as the geographi- a locus, are consistent across loci (HUDSONel al. 1987). The first step in applying the test is to complete a table that has, cally isolated subspecies, D. p. boptana. Assessments of for example, as many rows as there are loci in the data set DNA sequence variation can reveal relative historical ef- and three columns: one for variation in one species, one for fective population sizes, and the patterns of shared varia- variation in the second species, and one for the divergence tion may be used to aSsess historical levels of gene flow between species. Like a chi-square test, the test proceeds by within and between species (HUDSONet al. 1992). Under generating a companion table of values that are expected some circumstances, thistype of data can also reveal under a null model and then a third table of standardized discrepancies between observations and expectations (HUD the action of positivedirectional and balancing selection SON rl al. 1987). Rejection of the null hypothesis requires that (HUDSONet al. 1987). In addition, we hope also to esti- the sum of standardized discrepancies between observations mate the divergence times among the species. and expectations be greater than expected by chance. How- ever, this may not be thenull hypothesis of the greatest inter- est. If selection has occurred at just onelocus in one species, MATERIALSAND METHODS then only one cell of the table of standardized discrepancies Species and strains: A total of 35 strains were studied (Ta- may have a large value, and the overall test may not detect ble l). Most were acquired from the NationalDrosophila the discrepancy. Like the chi-square test with multiple rows Species Resource Center (NDSRC, Bowling Green, OH). The and columns, the test is most powerful for departures from rest were kindly provided by DANIEI.M'EINREICH. the null model that affect multiple cells in the data matrix. Sequencing: DNA from individual male was extracted We have modified the HKA test to one that is sensitive to according to protocol 48 of ASHRURNER(1989). From each single cell departures from thenull hypothesis. In this outlier sample of genomic DNA, a section of pm (either between test, the test statistic is the maximum standardized discrepancy positions 2849-4932 or 2993-4932, based on Col.oT rt al. (MSD) observed for any polymorphism value. If one of the 1988) was PCR-amplified. Additional DNA preparation and species has experienced a recent selective sweep or a long sequencing followed the method of KIMAN and HEY(1993). standing balanced polymorphism atjust oneof the loci, then Both strands were sequenced for eachstrain. A total of seven the standardizeddiscrepancy for thatobservation will be high. 20-bp long sequencing primers, spaced -200 bp apart, were This test statistic is then compared with a neutral distribution used on cach strand. The final length of sequenced portion of the MSD that is generated by 5000 independent HKA coa- was -1.5 kb, flanking the D. psmdoobscurn equivalent of the lescent simulations (HUDSONet al. 1987; HILTONet al. 1994). Thr-Gly rcpcat and covering the exon5-7 and intron 57 (Fig- For each simulation, the MSD observed for any locus, for ure 1). All sequences have been submitted toGenBank (acces- polymorphism in either species, is the MSD measure for that sion numbers L81261-L81295). simulation and is recorded. If the observed MSD value is Data analysis: DNA sequences were assembled and aligned >95% of the entire frequency distribution of simulated MSD visually. Intraspecific diversity was estimated by average pair- values, then the outlier test is significant. wise differences per base pair (T) and by thenumber of polymorphic sites within a species. Interspecific divergence was assessed by the number of fixed and shared polymor- RESULTS phisms between species and the net average painvise diver- gence (NEI 1987). Several tests were conducted to assess the Polymorphisms anddivergence: Polymorphisms in history of natural selection at pm (HUDSONrt nl. 1987; MCDON- 35 sequences are summarized in Figure 2. Drosophila AI.D and KREITMAN1991; TAIIMA1989). Neighbor-joining pseudoobscura is the most variable and D. p. bogotana trees(SAITOU and NEI 1987) and bootstrap estimates were the least (Table 2). As expected under a hypothesis of constructed using theprograms DNADIST, NEIGHBOR, SEQBOOT and CONSENSE in Phylip3.5 (FEI-SENSTEIN1993). reduced constraint, thereis greater variation in synony- The minimum numberof recombination events was assessed mous sites than replacement sites and greater variation with themethod of HUDSONand KAPIAN (1985),and an in introns than in exons for D. pseudoobscura and D. 1116 R.-L. Wang and J. Hey

position 1111 1122233333 3334444555 5555566666 6667777888 88889 position 3466890012 5727822355 6780125001 1244601135 7884457034 44560 position 9068661432 8358336805 1665682142 5158717884 7690680292 48069 S/R/I/D RRRSSRSSSS SSRSRSSSSR RRSRSSSRRS SSSSRRSRRR SSSSSSSRII 11111 amino acids NNDDGSPPLS FHTGGGNQLS GAAPTAAAATPPFZNAAQAIA PFYPPNLY” ----- replacement GGN..A...... A.A....N AV.S...SP. ....PG.TMT ...... F...... Consensus AAGCGTGCCG CTACGCCGCG GCCCCCAGGGTGCCGCGGAG GCCCCTCACC CTGCA BOGOTA6 1 .....G...... T...... G...... A.. ...T...... G. BOGOTA6 0 .....G...... T...... NN.. A... ..T...... G. BOGOTA6 2 .....G...... T...... G...... A.....T...... G. BOGOTA6 3 .....G...... T...... G...... A.....T...... G. BOGOTA6 7 .....G...... T...... A.. A.....T...... G. BOGOTA70 .....G...... T...... A.....T...... G. BOGOTA7 3 .....G...... T...... A.. A.....T...... G. BOGOTA7 4 .....G...... T...... A... ..T...... G. BOGOTA7 7 .....G...... T...... G...... A.....T...... G. PSEUDO1 ...... T...... T...... PSEUDO2 .....G...... TT...... T...... T..... A. ... PSEUDO3 .....G...... T...... T...... T..T ..C.. PSEUDO4 ...... T...... T...... T.. A.... PSEUDO5 ..A...... T..A ...T...... T...... T. T.... PSEUDO6 ..A....T ...... T...... T...... NN...... T.. ... A.... PSEUDO7 ...... T...... T...... T...... T..... A.... PSEUDO8 .....G...... A...... T.. ... A ...... A.....T...... G. PSEUDO9 ...... A ...... T..T .A... PSEUDO10 .....G...... T..T. .NT...... A.... PSEUDO11 .....G...... T. T.... MI RANDAZ GG..C..... 2 TCG ...... T....T.C. .AA.CG. .G...... C..T. ....T MIRANDA23 GG..C....A TCGGC...... T...... C. .AA.CG. .G.... T.C..T...... MIRANDA24 GG..C....A TCGGC...... T...... C. .AA.CG. .G.... T.C..T...... MIRANDA2 5GG..C....A TCGGC...... T...... C. .AA.CG. .G...... C..T...... PERSIMI4O GG.TC...... C...... T...... T. T.... PERSIMI4l GG.TC...... C...... T...... A.. . ..T.....T. T.... PERSIMI42 GG.TC...... C...... T...... A ...... T. T.... PERSIMI43 ....c...... C...... T..T...... T. T.... PERSIMI44 ....c...... C...... T...... T. T.... PERSIMI45 GG.TC...... C...... C.....T ...... A...... T. T.... PERSIMI46 GG.TC...... C ...... TT...... T. T.... PERSIMI47 GG.TC.A... .C...... T...... A...... T. T.... PERSIMI48 GG.TC. ..T..C...... A ...... T...... A...... T. T.... PERSIMI49 GG.TC ...T. .C...... A ...... T...... T. T.... PERSIMI50 GG.TC...... C ...... C .....T...... T. T.... FIGURE2.-Pol~1norphic sites of 35 samples from D. pseudoobscuru (pseudo), D. persirnilis (persimi), D. p. bogotana (bogota), and D. miranda (miranda). S/R/I/D, synonymous/replacment/insertion/deletion.Amino acids: A, alanine; D, aspartic acid; E, glutamic acid; F, phenylalanine; G, glycine; H, histidine; I, isoleucine; K, lysine; L, leucine; M, methionine; N, asparagine; P, proline; Q glutamine; S, serine; T, threonine; V, valine; Y, tyrosine.

persirnilis. Measured by T,the level of intraspecific varia- most closely related taxa. However, these taxa differ by tion at per in D. pseudoobscura is lower than that found six fixed differences and share just a single polymor- at othergenes in the same species (Table 3).D. persirnilis phism in our sample. In contrast, there are only two is somewhat less variable than D. pseudoobscura at per. D. fixed differences but six shared polymorphisms be- rniranda has less and D. p. bogotana has very little. Per tween D. pseudoobscura and D. persirnilis, a species pair locus for D. pseudoobscura and D. persirnilis is comparable that exhibits greater net pairwise divergence. There ap with perlocus estimates from some of the D. rnelanogaster pear to be two reasons why D. pseudoobscura is more species complex. The estimate of T at perfor D. p. bogo- closely related to D. persirnilis from the perspective of tanais equivalent to that for D. sechellia, an island species shared and fixed differences yet closer to D. p. bogotana in the D. rnelanogastercomplex that is thought to have a from the viewpoint of net divergence. First, is that D. very reduced effective population size (HEYand KLIMAN p. bogotana has little variation at per, and this has the 1993). Most striking, however, is the finding that D. p. effect of elevatingfixed differences, and decreasing bogotana diversity at per is about oneorder of magnitude amounts of shared polymorphism. Second, is that most lower than those of other loci from the same subspecies. of the shared polymorphisms between D. pseudoobscura In contrast to the situation with D. sechellia in which and D. persirnilis are due to one particular sequence low variation has been found at multiple loci (HEY and in the sample. The sequence from strain persimilis-40 KLIMAN 1993; HILTONet al. 1994), low variation at per accounts for five out of the six shared polymorphisms in D. p. bogotana does not reflect genome-wide levelsof observed between the sibling species (Figure 2). When variation for this species. this strain is excluded from the comparison, the net Table 4 shows measures of interspecific divergence divergence between the two species becomes 60% for each pair of species. Based on net divergence, it greater (Table 4). The patternof variation suggest that appears that D. p. bogotana and D.pseudoobscura are the persimilis-40 is a hybrid sequence: the first926 bp D.pseudoobscura Speciation 1117

position 111111111111 1111111111 1111111111 1111111111 1111111111 111111 position 999000000000001 1111111222 3333333333 3333444444 4444444444 444555 position 299012235555772 3667899248 1555666678 8899011112 2224557888 999000 position 704578910367235 1264807376 0057234950 6934067890 1239791467 356458 S/R/I/D ISRSSSSRSSSRRRS SIIIIIISSS SIIIIIIIII IIIIIIDDDD DDDIIIIIII 111111 minoacids -SQSEWSNKKppN A-""-GST D-"""-- ______replacement ..K....A...ESS...... Consensus AGCGGACTCCAACCT ACACCAAATT TCGGTGAGCCCCTTTC------TCTGGCGTCCCGT BOGOTA61 ...... T...... G ...... G...... A.. ..C ...... C...... BOGOTA60 ...... T...... G...... G...... A.. ..C ...... C...... BOGOTA62 ...... T...... G ...... G...... A.. ..C ...... C...... BOGOTA63 ...... T...... G...... G...... A.. ..C ...... C...... BOGOTA67 ...... T...... G ...... G...... A.. ..C ...... C...... T.. BOGOTA70 ...... T...... G...... G...... A.. ..C ...... C...... BOGOTA73 ...... T...... G...... G...... A.. ..C...... C...... T.. BOGOTA74 ...... T...... G...... G...... A.. ..C ...... C...... BOGOTA77 ...... T...... G ...... G...... A.. ..C ...... C...... PSEUDO1 G...... AG...... T.. PSEUDO2AG.TC. G...... PSEUDO3 G.. AG. G..PSEUDO3 ...... PSEUDO4 ...... TC....AG ...... A...... PSEUDO5 G .....TC ....A G...... PSEUDO6 G .....TC ....A G...... PSEUDO7 G. ....TC.. ..AG...... PSEUDO8 ...... TC....AG ....A..C... .A ...... A ...... PSEUDO9 G... ..TC.TG...... A C...... TT...... C... C...NN PSEUDO10 ...... TC....AG ...... A...... NN PSEUDO11 G... ..TC.. ..AG..T ...... A...... C..T...... N MIRANDAS2 G..A ..TC..G A .... C...... C.ATAAG.GT ...C..TTTA AGC.A..A..C ..... MIRANDA23 G.A ..GTC..GA .... C ...... C.ATAAG.GT ...C..TTTA AGC.A..A..C ..... MIRANDA24 G.A..GTC ..G.... C ...... A C.ATAAG.GT ...C..TTTA AGC.A..A...... C MIRANDA25 G.A.. GTC..G .... C ...... A C.ATAAG.GT ...C..TTTA AGC.A..A...... C PERSIMI40 G..... TC.TG ...C C.T...... PERSIMI4l .C.. A...... G...... AA.C. PERSIMI42 .C.. A...... G.GT ....G.TTTA AGC.A..A...... C PERSIMI43 .C ..A...... AT...... T. ....G.TTTA ACC.A..A...... PERSIMI44 .C ..A...... AT...... T. ....G.TTTA ACC.A..A...... PERSIMI45 .C ..A...... G.GT ....G.TTTA AGC.A..A...... C PERSIMI46 .C..A...... G.GT ....G.TTTA AGC.A..A...... C PERSIMI47 .C.. A...... G.GT ....G.TTTA AGC.A..A...... C PERSIMI48 .C ..A...... G.TTTA AGC.A..A...... PERSIMI49 .C..A...... G.GT ....G.TTTA AGC...... PERSIMI50 .C..A...... G...... G.GT ....G.TTTA AGC.A..A...... C FIGURE2. - Continued closely resemble other D. persirnilis sequences, while the natural selection) or neutral (KIMURA 1983). TheHKA remainderare more similar to D. pseudoobscura se- test is similar to a chi-square test and asks whether the quences (Figure 2). This is graphically demonstrated in relative levels of intraspecific polymorphism and inter- a sliding-window plot of the average painvise divergence specific divergences fora given locus are consistent between persimilis-40 and the othersequences from D. across loci (HUDSONet al. 1987).We employed the HKA pseudoobscura and D.persirnilis (Figure 3). Thepersimilis- test by comparing the per data with sequence datafrom 40 sequence has less divergence from D. persirnilis se- the alcohol dehydrogenase locus (SCHAEFFERand quences than it does from D. pseudoobscura sequences, MILLER1991) and the small ribosomal RNA locus of in the first 1000 bp of the region. This pattern is re- mitochondria (SrRNA; T. M. JENKINS, C. J. BASTEN,and versed in the remainder of the sequence. In both re- W. W. ANDERSON, GenBank accession numbers X'74849- gions, there are sections where persimilis-40 is equally 75012; Table 5). In particularwe focused on D. pseudoob- divergent from both sibling species. As expected, neigh- scura and D. p. bogotana, the one species pair with data bor-joining trees constructed for each of the two re- available from multiple loci (Table 3). We conducted gions have different topologies (data not shown). For tests with different numbers of loci and with and with- the first 926 bp, persimilis-40 is clustered with the D. out the inclusion of intraspecific variation from both persirnilis group. For the latter portion, however, it is species (Table 5). Tests that include per locus data and with D. pseudoobscura. just theD. p. bogotanaintraspecific variation (but notthe Natural selection: In recent years, several methods D. pseudoobscura intraspecific variation) are significant at have been developed to test whether a pattern of DNA the 5% level. Thus the pattern of low variation within sequence variation has been shapedby natural selection D. p. bogotanaat perbut not at otherloci is not consistent (HUDSONet al. 1987; TAJIMA1989; MCDONALD and with the neutral model. However, the HKA tests that KFWTMAN 1991). All of these tests employ, as part of include polymorphism data from just D. pseudoobscura, the null hypothesis, the neutral model of molecular or from both D. pseudoobscura and D. p. bogotana, are evolution in which mutations are either deleterious (in not significant (Table 5). which case they are removed from the population by We also conducted outlier HKA tests (see MATERIALS 1118 R.-L. Wang and J. Hey

TABLE 2 The number of polymorphic sites within species

Sample Species Syn" size Repb Intron' Total Od TAJIMALY

Drosophila p. bogotana 9 1 1 1 3 1 0.60 D. pseudoobscura 11 22 7 20 49 16 -1.2 D. miranda 4 6 2 1 9 5 -0.49 D.persirnilis 11 12 9 15 36 12 -0.76

a Synonymous changes in exons, total exon length is 1131 bp. Replacement change in exons. 'Total intron length is 345 bp. 0 is estimated by S/X( 1/1), I = 1, . . . , N, -1, S, total number of polymorphic sites, n, sample size. (WATTERSON1975). eT~~~~~(1989).

AND METHODS) with data from D. pseudoobscura and D. Yet another kind of test asks whether the ratio of p. bogotana. In a test that included data from per and synonymous to replacementvariation appears to be the Adh, the MSD score occurred for per variation within same for both intraspecific polymorphisms and inter- D. p. bogotana. The observed value was 2.55, and the specific fixed differences (MCDONALDand KREITMAN probability of observing a higher value was estimated 1991). For most species pairs in this study, the MCDON- by simulation to be 0.024. In an outlier test using per, ALD-KREITMAN test can not be carried out because of Adh, and SrRNA, the MSD also occurred for pervariation an absence of fixed replacement differences between within D. p. bogotana. In this case the value was 2.513, species. For contrasts involving D. miranda, the MCDON- and the probability of observing a higher value was ALD-KREITMANtest revealed no significant departures estimated to be 0.034 (Table 6). Ina test with just Adh from the neutral model (data not shown). and SrRNA, the MSD score was 0.36 for SrRNA in D. p. Recombination,gene tree estimates, and gene bogotana, and the probability of observing a higher value flow: We applied the counting methodof HUDSONand was estimated to be 0.429. KAPLAN (1985) and found evidence of a minimum of Table 2 shows values of TAJIMA'SD (TAJIMA1989). If four recombination events at per in D. pseudoobscura and population size has been changing or natural selection three in D. persimilis.We also applied the method of has alteredthe frequency distribution of polymor- HUDSON(1987) to estimate the population recombina- phisms from that expected under the null model, then tion parameter, 4Nc (3Nc for sex-linked loci; c, the re- D will be different from 0. None of the observed values combination rate per generation for the sequenced re- of D are significantly different from 0 (Table 2). gion). These estimates are 37.9 for D. pseudoobscura and

TABLE 3 Comparison of the period locus of D. pseudoobscura and the sibling species with some previous studies

A dh XdhPen'od Adh Est-5B SrRNA Species sex-linkedSpecies autosomal autosomal sex-linked mitochondrial

D. pseudoobscura 8.4 11.7" 22.6d12.Ob 12.2' D. persimilis 5.8" - - - - D.8.4"p. bogotana 0.9 - - 9.4d D. miranda 3.2 - - - - D. melanogaster8.1g 6.2l - - - D. simulans 1l.V 16.4h - - - D. mauritiana 11.g 16.6' - - - D. sechellia 0.9f - - - -

T (per base pair, X 1000) Observedaverage pairwise differences among DNA sequences or nucleotide diversity (NEI 1987). a SCHAEFFERand MILLER(1991). RILEY et al. (1992). 'VEUILLEand KING (1995). T. M. JENKINS, C. J. BASTEN,and W. W. ANDERSON, GenBank accesson numbers X74849-75012. e Sample persimilis-40 excluded. 'KLIMAN and HEY(1993). KREITMAN (1983). BODMERand ASHBURNER (1984). COHNet al. (1984). D. pseudoobscura Speciation 1119

TABLE 4 portions of the locus have different gene trees, and an Number of fixed polymorphisms and analysis that imposes a single tree estimate cannot be net average painvise divergence taken as an estimate of the true genealogy. In effect much of the tree in Figure 4, especially those portions involving sequences from D.pseudoobscura and D.persim- ilis, must be considered only a summarization of the 1. D. p. bogotana 6 (1) 43 (0) 16 (0) as 2. D. pseudoobscura 5.0 26 (1) 2/4" (6/1") pattern of observed divergences among sequences and 3. D. rniranda 30.0 23.5 18 (0) not as an estimate of the genealogy. 4. D. persirnilis 12.3 9.7/15.6" 19.8 Among samples from D. miranda and D.p. bogotana, the sequences do cluster with their respective groups, a Persimilis-40 is excluded. Number of shared polymorphisms are shown in parenthe- and each of these groups are exclusive of sequences ses. Net distance between D. p. bogotana and the other species from other groups (Figure 4). However, apart from are corrected for the low diversityof D. p. bogotana NEI(1987). these groupings, the tree diagram does not suggest a Divergence values are in the lower triangle, X 1000. sister taxa relationship among species. In particular, D. p. bogotana sequences do not form a sister group to D. 23.7 for D. persimilis. However, these arerough esti- pseudoobscura but rather appear to be more closely re- mates, as the methodis not expected to bevery accurate lated to some sequences of D. pseudoobscura than to for sample sizes of <20 sequences (HUDSON 1987). others. This pattern is consistent with a speciation A neighbor-joining distance tree (SAITOUand NEI model in which D.p. bogotana was recently formed by 1987) was constructed for the 35 samples and the con- a small isolate from a much larger population of D. fidence levels for various clades are indicated by boot- pseudoobscura that still segregates old lineages. However, strapping values (Figures 4 and 5).With the noteworthy it is also consistent with a recent selective sweep, at or exceptions of the branch leading to the D.p. bogotana near per, that was limited to D.p. bogotana. The relation- sequences and the branch leading to the D. miranda ship between D.pseudoobscura and D.persimilis is compli- sequences, most of the branches on the gene treeesti- cated by the D.persimili.+40 sequence. In Figure 4, per- mate havelow bootstrapping values (Figure 5). This similis-40 appears as a sister group to agroup that pattern is expected if the sequences of D.pseudoobscura includes all other sequences (excluding those of D. mi- and D. persirnilis have histories that include multiple randa and collectively the sequences of D.persimilis do recombination events. When the genealogical history not form a monophyletic group. However, persimilis- for a locus has included recombination, then different 40 may be a hybrid sequence formed by gene flow and

lo 0 d

FIGURE3.-A sliding window plot of averagepainvise diver- gence of D. persirnilis (excluding the sample persimilis4O) and D. pseudoobscura frompersimilis-40. Window size is 200 nucleotides.

nucleotide positions 1120 R.-L. Wang and J. Hey

TABLE 5 The HKA test of natural selection for D. pseudoobscum and D. p. bogotanu

Species Period Adh“ SrRNA Diversity and divergence” 0 within D. p. bogotana 0.75 7.5 7.6 0 within D. pseudoobscura 11.2 11.9 24.6 interspecific divergence‘ 13.4 12.7 17.3

Test combinations Loci included x2 d.f. Probability D.p. bogotana only period, Adh 4.85 1 0.028” paiod, Adh, SrRNA 6.15 2 0.046* D.pseudoobscura only period, Adh 0.04 1 0.845 period, Adh, SrRNA 0.12 2 0.943 D. p. bogotana f D. pseudoobscura period, Adh 3.39 2 0.184 period, Adh, SrRNA 6.63 4 0.157 ’Estimated from SCHAEFFERand MILLER (1991). * Per basepair X 1000. ‘Average pairwise differences between species. recombination. With persimilis-40 excluded, the tree is using FYI. and the average painvise differences within more consistent with the established species phylogeny and between populations (WRIGHT1951; SLATKIN1991; (GODDARDet al. 1990). HUDSONet al. 1992). The estimate of Nm between D. The tree estimate also suggestslarge amounts of gene pseudoobscura and D.p. bogotana is 0.13 and thatbetween flow within D. pseudoobscura and D. persirnilis (Table 1, D. pseudoobscura and D. persirnilis 0.20 (0.17 without per- Figure 4) as there does not appear to be a tendency similis-40). For a rough estimate of intraspecific gene for samples to cluster by their geographic origin. The flow, the samples from D. pseudoobscura and D. persimilis amount of gene flow, both inter- and intraspecific, can were both split into a “northern” and a “southern” be estimated under a model of population subdivision group. For D. pseudoobscura, the Nm between the north- ern (samples 3, 10, 11) and the southern (the rest) is TABLE 6 17.6. And for D. persimilis, the Nm for the northern Three locus HKA test with D. pseudoobscura (samples 42, 43, 45, 50) and the southern (the rest) is and D. p. bogotana 16.1. In a very large study of the Adh locus in multiple populations of D. pseudoobscura, estimates of Nm varied D. D. p. from one to 12, depending onwhich populations were pseudoobscura bogotana compared (SCHAEFFERand MILLER1992). In general, a Locus S n S n Divergence Nm value 2 1 leads to considerable homogeneity among populations while populationstructuring and diver- period gence can occur when Nm is <1 (WRIGHT1940). These Observed 48 11 3 9 20 estimates of Nm also assume an equilibrium between a Expected“ 41.3 16.6 12.8 Deviation* 0.129 2.513” 0.413 steady level of gene flow and divergence. This assump- Adh tion may be appropriatefor the within species esti- Observed’ 114 10 66 8 37 mates. For the interspecific estimates, however, current Expected 127.5 50.6 38.7 levels of shared variation may be remnants from the Deviation 0.056 0.379 0.003 time of speciation, in which casethe Nm estimates could SrRNA be much greater than current levels of gene flow be- Observedd 43 15 10 7 9 tween species. Expected 36.2 11.8 14.3 Deviation 0.200 0.070 0.32 Time of introgression: One sequence, persimilis-40, contains a stretch of 586 bases (from position 927 to S, number of polymorphic sites; n, number of DNA se- the 3’ end, Figures 2 and 3) that seems typical of D. quences in the sample. * Statistically significant in an outlier pseudoobscura sequences and is divergent with respect to HKA test, P = 0.034; see text. Expected values are calculated using expressions (1) and D. persirnilis sequences. To estimate the time of intro- (3) of HUDSONet al. (1987). gression, we assume that the patternarose via introgres- ’ Deviations are calculated using expressions (2) and (4) sion of a single sequence from D. pseudoobscura into and the expression for x‘ in HUDSONet al. (1987). persirnilis, followed by one or more recombination ‘ Data from SCHAEFFERand MILLER(1991). D. “Data fromT. M. JENKINS, C.J. BASTEN,and W. W. ANDER- events. Two different approachesto estimating the time SON, GenBank accesson numbers X74849-75012. of introgression were considered. In the following: I is D. pseudoobscura Speciation 1121

PERSIM140-CA PERSIM141 -CA PERSIM142-WA PERSIM145-Canada PERSIMISO-BC PERSIM147-CA

PERSIM148-CA PERSIMl43-BC PERSIM144-CA FIGURE4.-A neighbor- PSEUDO9-TX joining distance tree. The PSEUD03-BC sample numbersare fol- PSEUD08-Mexico lowed by their geographic BOGOTA63 locations: CA, California; BOGOTA62 WA, Washington; BC, Brit- BOGOTA77 ish Columbia; TX, Texas; BOGOTA61 AZ,Arizona; OR, Oregon. The scale and the units of BOGOTA67 individual branch lengths BOGOTA73 can be assessed by compar- BOGOTA74 ison with the branch that IBOGOTA70 leads to PERSIMI40, which PSEUDO1 -AZ corresponds to 0.0035 sub- PSEUD02-CA stitutions per base pair. PSEUDO6-CA

PSEUDO1 0-BC

MIRANDA23 MIRANDA24 MIRANDA25 IMlRANDA22 the time of introgression in units of generations; N, is attention has fallen on the largest nonrecombining the effective population size of D. persirnilis; and N, is fragment of the introgressed sequence (a conservative the effective population size of D. pseudoobscura. assumption, since it is possible that a larger portion The first method is based on the divergence of the may have occurred in a larger sample of sequences), 586-bp region in persimilis-40 from D.pseudoobscuru se- then use can be made of an expression in FELLER(1971, quences. Let D be the average number of differences expression 9.9, p. 28) for the probability distribution between persimilis-40 and D. pseudoobscuru sequences of the length, a, of the longest unbroken portion of that is expected to be observed for the introgressed an interval that is randomly divided into a number of portion. Under a simple neutral model, in which the subintervals. The derivative of FELLER’S(1971) expres- amount of variation within D.pseudoobscuru at the time sion (9.9), taken with respect to a, is the probability of introgression is the same as is currently present, density of a and is a functionof the numberof random breaks. In the present context, setting u equal to 0.387 D = 2Iu + 3Nu. (1) (586/1512) yields a maximum likelihood estimate of In other words, D is equal to the divergence that has five recombination events since the time of introgres- accumulated since introgression plus divergence that sion. Thus the interval between the expected times for was present at the time of introgression. For the 586- five and six recombination events can be used as an bp region, D is estimated to be 8.82, and 3N,u is esti- estimate of the time since introgression. The popula- mated to be 4.87. Then from (l), luis equal to 1.975. tion recombination rate for D.persirnilis, 3N,c, was esti- Since 4.87 mutations accumulate in 3N, generations, mated to be 23.7 (see Recombination, gene tree esti- the estimated time of introgression is 1.2Nsgenerations mates, and gene flow). If persimilis-40 is excluded from (i.e., 1.975 * 3Ns/4.87). the calculations for this estimate, the value rises to 28. Introgression time can also be estimated using a This is an estimate of the number of recombination model of recombination. Given that a region of 586 events per 3Nr generations per 1512 bases of per. If we bases (out of 1512) has been left intact following one take the time of introgression as the midpoint between or more recombination events, it is possible to generate the expectedtime for five and six recombination events, a maximum likelihood estimate of the number of re- then the time since introgression is estimated to be combination events since the time of the introgression O.6Nr generations (5.5 * 3Nr/28). of the D. pseudoobscuru sequence. Assuming that our The two methods yield estimates of 1.2Nsgenerations 1122 and R.-L. Wang J. Hey

PERSIM140-CA 106 PERSIM150-BC 101 __c PERSIMl45-Canada 189 - PERSIM147-CA - PERSIMIISCA 123- PERSIM142-WA 134 PERSIM149-CA PERSIM148-CA 132 PERSIMW1 -CA I 2001 PERSIM144-CA L PERSIM143-BC BOGOTA77 BOGOTA61 BOGOTA62 BOGOTA63 BOGOTA73 BOGOTA67 113 BOGOTA60 1 108 20- BOGOTA70 BOGOTA74 PSEUDO8-Mexico PSEUD03-BC PSEUDO1-A2 158 PSEUD02-CA PSEUD07-CA PSEUDO6-CA PSEUDO4-Mexico PSEUDOIO-BC PSEUDO9-TX PSEUDO11 -OR PSEUDOS-CA MIRANDA23 200 P-iX MIRANDA24 MIRANDA22 MIRANDA25 FIGURE5.-A neighbor-joining consensus tree representing200 bootstraps. The nodes with values

TABLE 7 Net average painvise divergence atsynonymous sites and estimated divergence time between species

Species 1 2 3 4 1. D. p. bogotana 3.18 10.33 16.27 (0.012) (0.062) (0.039) (0.062) (0.012) 2. D. pseudoobscura 0.5 6.29/6.83‘ 11.35 (0.043) (0.024/0.026a) 3. D. miranda 2.6 1.8 +- 1.22 12.54 (0.048) 4. D. persirnilis 1.61.0 5 1.0 (1.1 2 0.93n) 2.0 2 1.0

~ ~ ~~~~~~~~~~~~ Pairwise divergences are stated as divergence per synonymous site in the upper (right) triangle; divergence times are means 2 SE in million years ago, lower (left) triangle. Divergence time is estimated based on a rate of 1.2 X lo-’ substitutions per synonymous site per year (SHARP and LI 1989) and the presence of 265 synonymous positions within the sequenced region. The standard error was calculated following TAKAHATA and NEI (1985) for net average pairwise divergence at synonymoussites. The divergence betweenD. p. bogotana and the other sDecies was corrected for thelow diversity within D.p. bogotana and the standard error was not calculated for these values. a Sequence persimilis40 excluded. regions that have been sequenced and that are sub it seems possible that there has been a small degree of jected to an HKA test, then the variance assumed for ongoing gene flow between the two siblings since their each locus will tend to be too high.effect, In recombina- speciation. tion causes loci to act like multiple loci, so that a mea- Speciation via founder event probably did not occur sure of variation that sums across a region with a history in the divergence of D. pseudoobscura and D. persirnilis. of recombination will have a variance that is reduced Both species reveal evidence of large long-term effective by the presence of recombination (HUDSON1983). population sizes, and the pattern of introgression sug- Given what is known of the function of the per locus gests that they may have never been completely isolated in D.rnelanoguster, it seems possible that the gene could from each other, geographically or genetically. It re- be a candidate for natural selection at times of specia- mains possible that these species arose via sympatric tion. The perlocus is known to be involved in the control speciation, a history that is supported by the geographic of interspecific differences in male courtship songs (KY- distribution and the pattern of chromosome inversion RIACOU and HALL 1980, 1989; KYRIACOU1990). The polymorphisms. The range of D. persirnilis is entirely region we have sequenced is adjacent to the Thr-Gly- included within that of D.pseudoobscura, and both spe- like repeats, and this and the downstream flanking re- cies segregate multiple inversion polymorphisms on the gion have been explicitly implicated in such control third chromosome (reviewed in POWELL1992). How- (Yu et al. 1987; WHEELERet at. 1991). However, D. p. ever, only one arrangement, “Standard”, is shared by bogotana does exhibit polymorphisms in the region be- both species. The Standard arrangement is relatively tween the Thr-Gly-like repeats and the region that is recent among the inversions in D. pseudoobscura but it most divergent (between positions 618 and 1504) from appears to be the oldest arrangement in D. persirnilis. D.pseudoobscura. It is possible that divergence at per may The implication is that D.persirnilis third chromosomes be associated with behavioral differences and contrib- are derived from the Standard arrangement, an argu- ute to reproductive isolation between D. pseudoobscura ment strongly favored by both allozyme and molecular and D. p. bogotana. To our knowledge, however, there data (PRAKASHand LEWONTIN1968; AQUADROet al. has been no experimental demonstration of interspe- 1991). In D. pseudoobscura, the historical development cific differentiation in male courtship songs between of inversion polymorphisms is closely correlated with them. their geographic distributions (POWELL1992), with the The divergence betweenD. pseudoobscura and D. per- most ancestral type found dominatingin southern Mex- sirnilis. One sequence, persimilis-40, appears to be a ico and Guatemala and the morerecent ones gradually hybrid between D. persirnilis and D. pseudoobscura se- extend northward in thewestern third of the US., form- quences. Two methods of estimating the time of the ing a fairly good temporal gradient along the north- introgression, by divergence data and by the estimated south axis. This pattern may not have been disrupted, number of associated recombination events, suggest since no glaciation occurred in this region during the that the introgression occurred well after the initial di- Pleistocene era (MIELKE 1989), and it is possible that vergence of D. persirnilis and D. pseudoobscura. This find- the geographic range of D.pseudoobscura has changed ing is consistent with evidence of limited gene flow from little in the past two million years. Today, the Standard chromosomal (DOBZHANSKY1973) and mitochondrial arrangement of D.pseudoobscura is found in the highest data (POWELL1983; HALE and BECKENBACH1985), and frequency along the Pacific Coast, from southern Cali- D. pseudoobscura Speciation 1125 fornia to British Columbia, exactly matching the distri- COHN,V. H., M.A. THOMPSONand G. P. MOORE, 1984 Nucleotide sequence comparison of the Adh gene in three Drosophilids. J. bution of D. persirnilis, the inversions of whichare either Mol. Evol. 20: 31-37. the Standard or its direct derivatives (POWELL1992). COI.OT,H. V., J C. HALLand M. ROSBASH,1988 Interspecific com- Evidence that introgression occurred at per after the parison of the period gene of Drosophila reveals large blocks of nonconserved coding DNA. EMBO J. 7: 3929-3737. separation of D. pseudoobscura and D. persirnilis argues DOBZHANSKY,TH., 1935 Drosophila miranda, a new species. Genetics against a direct contributionby perto their reproductive 20: 377-391. isolation. If reduced fitness of hybrids were due in part DOBZHANShT, TH., 1973 Is there gene exchange between Drosophila psrudoobscura and Drosophila persirnilis in their natural habitats? to per locus differences, then selection would prevent Am. Nat. 107: 312-314. the introgression of per locus sequences. However, this DOBZHANSW, TH., and C. EPLING,1944 Contributions to the genet- argument does not preclude amechanism where other ics, , and ecology of Drosophilapseudoobscura and its relatives. Carnegie Inst. Washington Publ. 554 1-183. sex-linked loci play a dominant role in the initial devel- DOBZHANSKY,TH., A. S. HUNTER,0. PAVI.OVSKY,B. SPASSKYand B. opment of reproductive isolation. WALIACE,1963 Genetics of an isolated margmal population of Time of divergence: Estimates of the age of diver- Drosophila pseudoobscura. Genetics 48: 91 - 103. EDERY,I., J. E. RLTTIIAand M. ROSBASH,1994 Phase shifting of gence between D.pseudoobscura and D. p. bogotana range the circadian clock by induction of the Drosophila prriod protein. from millions of years (AYALAand DOBZHANSKY1974) Science 263: 237-240. to decades (PRAKASH1972). Our estimate is 0.5 mya, FELLER,W. 1971 An Introduction toProbability Theory and its Applica- tions, Vol 11, Ed. 2. John Wiley and Sons, New York. assuming a substitution rate of 1.2 x lops per synony- FEUENSTEIN,J., 1993 Phylip. PhylogenyInference Package, version3.5. mous site per year. However, this estimate is problem- GODDARD,K., A. CACCONEand J. R. POWEI.L,1990 Evolutionary atic because of the suspected directional selection or implications of DNA divergence in the Drosophila obscura group. Evolution 44: 1656-1670. selective sweep at the per locus of D. p. bogotana, which HALE,L. R., and A. T. BECKENBACH,1985 Mitochondrial DNA varia- will significantly alter the number of segregating sites tion in Drosophilapseudoobscura and related species in pacific and the divergence between D.p. bogotana and the oth- Northwest populations. Can. J. Genet. Cytol. 27: 357-364. HEY,J. 1994 Bridging phylogenetics and population genetics with ers. SCHAEFFERand MILLER (1991) estimated a diver- gene tree models, pp. 435-449 in Molecular Approaches to Ecology gence time of 0.15 mya from the pattern of divergence and Evolution, edited by B. SCHIERWATER,B. STREIT, G. WAGNER at Adh. Our estimates of divergence times among D. and R. DESALLE.Birkhiuser-Verlag, Basel. HEY,J., and R. KI.IMAN, 1993 Population genetics and phylogenetics rniranda, D. pseudoobscura, and D. persirnilis are not sub- of DNA sequence variation at multiple loci within the Drosophila ject to this difficulty and may therefore be more accu- melanogaster species complex. Mol. Biol. Evol. 10: 804-822. rate. These estimates are also similar to independent HILTON,H., R. KI.IMAN and J. HEY, 1994 Using hitchhiking genes to study adaptation and divergence during speciation within the Dm time estimates based on inversion polymorphisms. Of sophila mlanogaster species complex. Evolution 48 1900-1913. the 36 types of inversions on the third chromosome in HUDSON,R. R., 1983 Properties of a neutral allele model with intra- D. pseudoobscura and 17 in D.persirnilis, Standard is the genic recombination. Theor. Popul. Biol. 23: 183-201. HLJDSON,R. R., 1987 Estimating the recombination parameter of a only one shared by both species, therefore predating finite population model without selection. Genet. Res. 50: 245- their speciation (POWELL1992). Molecular data suggest 250. this inversion arises 1.7 mya (AQUADROet al. 1991). HUDSON,R. R., and N. L. KAPIAN, 1985 Statistical properties of the - number of recombination events in the histoly of a sample of Our divergence figures, 2 mya between D. rniranda and DNA sequences. Genetics 111: 147-164. the others and 1 mya between D.pseudoobscura and D. HUDSON,R. R., and N.L. KAFTAN, 1988 The coalescent process in persirnilis, are therefore compatible with the age of inver- models with selection and recombination. Genetics 120 819- 829. sion polymorphisms. HUDSON,R. R., M. hElTMAN and M. AGUAD&,1987 A test of neutral molecular evolution based on nucleotide data. Genetics 116: We thank DANIEL. WEINREICHfor providing us several strains of D. 153-159. p. bogotana. We are grateful toJOHN WAKELEYand HOLLYHILTON for HUDSON,R. R., M. SIATKINand W. P. MADDISON,1992 Estimation helpful discussions. This study was funded in part by a grant from of levels of gene flow from DNA sequence data. Genetics 132 the National Science Foundation to J.H. (DEB-9306625) and a post- 583-589. doctorate fellowship to R.L.W. by the Bureau of Biological Research, KIMuRA, M. 1983 ?‘he Neutral Theory of Molecular Evolution. Cam- Rutgers University. bridge University Press, Cambridge. KI.ImN, R. M., and J. HEY, 1993 DNA sequence variation atthe period locus within and amongspecies of the Drosophila melanogas- LITERATURE CITED ter complex. Genetics 133: 375-387. KONOPIW,R. J., and S. BENZKR,1971 Clock mutants of’ Drosophila AQUADRO, C.F., A. L. WEAVER,S. W. SCHAEFFERand W. W. ANDERSON, melanogaster. Proc. Natl. Acad. Sci. USA 68: 2112-2116. 1991 Molecular evolution of inversions in Drosophila pseudoobsc- KREITMAN, M., 1983 Nucleotide polymorphism at the alcohol dehy- ura: the amylasegene region. Proc. Natl. Acad. Sci. USA 88: 305- drogenase locus of Drosophila wlmogaster. Nature 304 412-417 309. KYRIACOU,C. P., 1990 The molecular ethology of the period gene ASHBURNER, M., 1989 Protocol 48:preparation of DNA from single in Drosophila. Behav. Genet. 20: 191-211. flies, pp. 108- 109 in Drosophila: A Labmatory Manual. Cold Spring KYRIACOU, C.P., and J. C. HALL,1980 mutations Harbor Laboratory, Cold Spring Harbor, NY. in Drosophilamelanogaster affect short term fluctuations in the AWE, J. C., and R. M. BN.I.,JR., 1990 Gene genealogies and the male’s courtship songs. Proc. Natl. Acad. Sci. USA 77: 6729- coalescent process. Oxf. Surv. Evol. Biol. 7: 45-67. 2733. AYU, F. J., andTH. DOBZHANSKY,1974 A new subspecies ofDrosoph- KYRIA(:OU, C. P., and J. C. HAM., 1982 The function of courtship ila D.pseudoobscura. Pan Pacific Entomol. 50: 211-220. song rhythms in Drosophila. him. Behav. 30 794-801. BODMER,M., and M. ASHBURNER, 1984 Conservation and change in KYRIACOU,C. P., andJ. C. HALL,1984 Learning and memory muta- the DNA sequences coding foralcohol dehydrogenase in sibling tions impair acoustic priming of mating behavior in Drosophila. species of Drosophila. Nature 309: 425-430. Nature 308 62-65. 1126 R.-L. Wang and J. Hey

KYRIACOU,C. P., and J. C. HALL, 1989 Spectral analysis of Drosophila SCHAEFFER,S. W., and E. L. MILLER,1991 Nucleotide sequence anal- courtship songs. Anim. Behav. 37: 850-859. ysis of Adh genes estimates the time of geographic isolation of KYRIACOU,C. P., M. L. GREENACRE,M. G. RITCHIE and B. C. BYRNE, the Bogota population of Drosophila D. pseudoobscura. Proc. Natl. 1992 Genetic and molecular analysis of the love song prefer- Acad. Sci. USA 88 6097-6101. ences of Drosophila females. Am. Zool. 32: 31-39. SCHAEFFER,S. W., and E. L. MILLER,1992 Estimates of gene flow in LEWONTIN,R. C., J. A. MOORE,W. B. PROVINEand B. WALLACE,1981 Drosophilapseudmbscura determined fromnucleotide sequence anal- Dobzhansky 's Genetics of Natural Populations. Columbia University ysis of the alcohol dehydrogenase region. Genetics 132: 471-480. Press, New York. SCHAEFFER,S. W., C. F. AQUADROand W. W. ANDERSON, 1987 Restric- MAYNARDSMITH, J., and J. HAIGH,1974 The hitchhiking effect of a tion-map variation in the alcohol dehydrogenase region of Dro- favorable gene. Genet. Res. 23: 23-35. sophila pseudoobscura. Mol. Biol. Evol. 4: 254-265. MAYR,E., 1946 Experiments on sexual isolation in Drosophila. VII. SHAFT, P. M., and W. H. LI, 1989 On the rate of DNA sequence The nature of the isolating mechanisms between Drosophila pseu- evolution in Drosophila. J. Mol. Evol. 28 398-402. doobscura and . Proc. Natl. Acad. Sci. USA 32: SIATKIN,M., 1991 Inbreeding coefficients and coalescencetime. 128-137. Genet. Res. 58: 167-175. MCDONALD,J. H., and M. KREITMAN,1991 Adaptive protein evolu- TAJIMA,F., 1989 Statistical methods for testing the neutral mutation tion at the Adh locus in Drosophila. Nature 351: 652-654. hypothesis by DNA polymorphism. Genetics 123: 585-595. MIELKE,H. W., 1989 Patterns of Lqe--Biogeography of a Changang TAKAHATA,N., and M. NEI, 1985 Gene genealogy and variance of World, Unwin Hyman, Boston. interpopulational nucleotide differences. Genetics 100 325-344. NEI, M., 1987 Molecular Evolutionaq Genetics. Columbia University TEMPLETON,A. R. 1994 The role of molecular genetics in speciation Press, New York. studies, pp. 455-477 in Molecular Approaches to Ecology and Evolu- POWELL,J. R., 1983 Interspecific cytoplasmic gene flow in the ab tion, edited byB. SCHIERWATER,B. STREIT,G. WAGNERand R. sence of nuclear geneflow: evidence from Drosophila. Proc. Natl. DESALLE.Birkhauser-Verlag, Basel. Acad. Sci. USA 80: 492-495. THROCKMORTON,L. H., 1975 The phylogeny, ecology and geogra- POWELL,J. R., 1992 Inversion polymorphism in Drosophilapseudoobsc- phy of Drosophila, pp. 421-469 in Handbook of Genetics, Vol. 3, ura and Drosophila persirnilis, pp. 73-126 in Drosophila Inversion edited by R. C. KING. Plenum Press, New York. Polymolghism, edited by C. B. KRIMBAS and J. R. POWELL.CRC VEUILLE,M., and L. M. KING,1995 Molecular basis of polymorphism Press, Boca Raton, FL. at the esterase-5B locus in Drosophila pseudoobscura. Genetics 141: PWH,S., 1972 Origin of reproductive isolation in the absence 255-262. of apparent genic differentiation in a geographic isolate of Dro- WATTERSON,G. A,, 1975 Onthe number of segregating sites in sophila pseudoobscura. Genetics 72: 143-155. genetical models without recombination. Theor. Popul. Biol. 7: PRAKASH,S., and R. C. LEWONTIN,1968 A molecular approach to the 256-276. WHEELER,D. A,, C. P. IOU,M. GREENACRE,Q. Yu, J. RWLA et aL, study of genic heterozygosity in natural populations. 111. Direct 1991 Molecular transfer of a speciesspecificbehavior from Drosoph- evidence of coadaptation in gene arrangements of Drosophila. ila sirnulam to hs@hila melanogastm. Science 251: 1082-1085. Proc. Natl. Acad. Sci. USA 59: 398-405. WRIGHT,S., 1940 Breeding structure of populations in relation to RILEY, M. A,, S. R. KAFIAN and M. VEUILLE,1992 Nucleotide poly- speciation. Am. Nat. 74: 232-248. morphism at theXanthine Dehydrogenase locus in Drosophila pseudo- WRIGHT,S., 1951 The genetical structure of populations. Ann. Eu- obscura. Mol. Biol. Evol. 9: 56-69. genet. 15 323-354. RITCHIE,M. G., and C.P. KYRIACOU,1994 Reproductive isolation Wu, C.-I., and M. F. PALOPOI.I,1994 Genetics of postmating repro- and the period gene of Drosophila. Mol. Ecol. 3: 595-599. ductive isolation in . Annu. Rev. Genet. 28 283-308. ROSBASH, M.,and J. C. HALL, 1989 The molecular biology of circa- Yu, Q., H. V. COLOT, C. P. KYRIACOU,J. C. HALLand M. ROSBASH,1987 dian rhythms. Neuron 3: 387-398. Behavior modification by in vitro mutagenesis of a variable region SAITOU,N., and M. NEI, 1987 The neighbor-joining method: a new within the @nod gene of Drosophila. Nature 326 765-769. method for reconstructing phylogenetic trees. Mol. Bio. Evol. 4: 406-425. Communicating editor: W.-H. LI