Introduction

Invasion and exploitation of novel habitats by species often require the co-evolution of genes encoding the molecular mechanisms that enable the invading organisms to withstand the novel environment. Although this process is increasingly well understood in real time for prokaryotes (Elena and Lenski 2003), for complex eukaryotes its occurrence is typically inferred from patterns of genomic variation remaining from genetic events of long ago (Li 1997). In complex eukaryotes, documenting molecular adaptation as it happens is still comparatively rare (but see for example Huey et al. 2000; Lerman et al. 2003). Here we describe the changes in complex loci containing various numbers of hsp70 gene copies that have accompanied the evolution of two closely related species, Drosophila lummei and Drosophila virilis. Both are members of the virilis group of Drosophila, which comprises 12 species divisible into two phylads: virilis and montana (Patterson and Stone 1952; Throckmorton 1982). According to many investigators (Patterson and Stone 1952; Evgen’ev et al. 1982; Throckmorton 1982; Spicer 1991, 1992; Nurminsky et al. 1996) D. virilis is the most primitive species of the virilis phylad and is probably ancestral to it if not to the entire virilis group. Its distribution throughout the Northern Hemisphere is primarily below 40°N latitude. D. lummei, considered the closest relative of D. virilis, occurs from just above 40° to just above 65°N latitude and from Sweden west to the Pacific coast of Asia. The most parsimonious interpretation of the above information is that D. lummei descended from D. virilis or a common ancestor approximately 5 MYA (Spicer 1992; Spicer and Bell 2002) and invaded higher latitudes than its ancestor.

Given the correlation of latitude and temperature, our prior work (Garbuz et al. 2002, 2003) focused on the phenotypic aspects of temperature tolerance, which show countergradient variation with the latitudinal ranges of the two species. In the cold, D. lummei is far more tolerant than D. virilis, and also undergoes a diapause absent in all other virilis group species (Lumme 1982). By contrast, D. virilis has greater basal and inducible high temperature tolerance than does D. lummei (Mitrofanov and Blanter 1975; Garbuz et al. 2002, 2003). In Drosophila, the DnaK-Hsp70 superfamily of molecular chaperones provides a key molecular mechanism underlying high temperature tolerance (Feder and Krebs 1998; Zatsepina et al. 2001). Corresponding to their differing temperature tolerances, D. virilis expresses both more diverse electromorphs of Hsp70-family members and several-fold more Hsp70-family protein and mRNA than does D. lummei, especially at high temperatures (Garbuz et al. 2002, 2003). Different strains or populations of D. virilis also vary in Hsp70 levels (Garbuz et al. 2002, 2003).

Here we examine the genetic basis for these inter-specific and intra-specific differences in thermotolerance and Hsp70-family proteins. All Diptera that have been studied share an inverted pair of Hsp70-encoding genes at a single chromosomal locus; in mosquitoes (Benedict et al. 1993) and the relatively basal Drosophila species Drosophila pseudoobscura, (B. Bettencourt, pers. commun.) these are the only known hsp70 genes. The hsp70 genes have undergone various duplications in more derived species, however. In the montium subgroup of the melanogaster species group of Drosophila, for example, one of these paired genes has undergone a tandem duplication to yield three hsp70 genes (Konstantopoulou et al. 1998). In some other subgroups of this species group, the entire locus has undergone duplication to yield four hsp70 genes; in Drosophila melanogaster, the derived locus (at 87C1) has undergone tandem duplication to yield five (in various wild or wild-type strains) or six (in a single laboratory strain) hsp70 genes total (Leigh-Brown and Ish-Horowicz 1981; Bettencourt and Feder 2001; Maside et al. 2002). Given that the virilis group is nearly basal to the Drosophila lineage, our expectation was that D. virilis and D. lummei should have only the inverted pair of hsp70 genes at a single locus, which would be consistent with earlier cytogenetic localizations of hsp70 to a single locus (29C) in D. virilis (Evgen’ev et al. 1978; Peters et al. 1980). To the contrary, here we report that while the hsp70 locus is singular, the hsp70 genes have undergone multiple tandem duplications to yield differing hsp70 copy numbers both within and between D. virilis and D. lummei. These differences may explain in part the differing thermotolerances of these populations and species.

Materials and methods

Drosophila strains and collection dates

Drosophila virilis strains 160 (b, gp, cd, pe, gl), 9 (Batumi, Caucasus), 1433 (Leeds, England), T53 and T40 (collected in Tashkent 30 years apart) and D. lummei strains 200 (Moscow region, ca. 1970), 202 (Krasnodar, Southern Russia) and 1102 (Finland) were obtained from the stock center of the Institute of Developmental Biology, Moscow. The following D. virilis strains were obtained from Dr. Anneli Hoikkala, Oulu University, Finland: A11 (Matsuyama, Ehime, Japan, 1973), V-ZZP-01 (Zeziping, Hunan Province, China, 2001), V-WW-08 (Wuwei, Gansu Province, China, 2002), SBB (Sapporo, Hokkaido, Japan, 1986), V-Hunan (Hunan Province, China), and V-Nanjing (Nanjing, Jiangsu Province, China). The latter strains were collected within the past 2 years. All flies were reared on a yeast, cornmeal, molasses, and agar medium at 25°C.

Cytological analysis

Larvae were grown at 18°C on medium supplemented with live yeast solution for 2 days before dissection. Salivary glands from third instar larvae were dissected in 45% acetic acid and squashed (Lim 1993). Procedures and labeling of DNA probes for in situ hybridization were as described (Lim 1993).

DNA manipulations and Southern analysis

Southern blot analysis of D. virilis and D. lummei genomic DNA was performed (Evgen’ev et al. 2000a). Five micrograms of each DNA sample was digested with restriction endonucleases. After agarose gel electrophoresis, each gel was treated for 15 min in 0.25 M HCl and then incubated twice in denaturing buffer (1.5 M NaCl, 0.5 M NaOH) for 30 min. After 30 min incubation in neutralization buffer, gels were capillary-blotted onto nylon membranes and fixed by UV cross-linking using the UV Stratalinker 2400 (Stratagene) protocol. Standard high-stringency hybridization and wash conditions were used for Southern blot analysis. To detect Hsp70 sequences in Southern blots and genomic libraries, fragments of cloned D. lummei, D. virilis, or D. melanogaster hsp70 genes were labeled by random priming and used as probes. The original screen of all libraries and Southern blots was performed with 5′ specific and 3′ specific probes generated from the ClaI–SalI fragment containing an hsp70 gene of D. melanogaster (McGarry and Lindquist 1985). The 5′-specific probe represents the ClaI–BamHI portion of the clone indicated, while the 3′-specific probe represents the BamHI–SalI portion.

Genomic libraries from D. virilis strain 160 and D. lummei strain 200 were prepared by partial Sau3A digestion with subsequent ligation into the BamHI site of lambda Dash phage arms (Stratagene). Isolated phages containing hsp70 genes were identified by restriction analysis and hybridization (see above), and subcloned into KS-Bluescript for subsequent sequencing (with one exception, see below). Clones were sequenced with Sequenase (Amersham) and ABI 377 sequencers. Sequences were assembled manually and aligned using CLUSTAL X (Jeanmougin et al. 1998). Relevant sequence information has been deposited in GenBank.

When the above procedure failed to yield reliable sequence for hsp70c and the hsp70c–d intergenic region in D. lummei, genomic DNA isolated from the D. lummei 200 strain was digested with XbaI, and fragments varying between 2.5 kb and 3.5 kb ligated into pBluescriptSK. Presumptive positive clones were detected by colony hybridization to the XbaI–SacI fragment of D. lummei hsp70d, and sequenced. Primers were designed complementary to the 3′ end of this clone (5′-AATAATAAGAGCTAGAGC-3′) and to the 5′ end of clone 17 (5′-ATACCAGAACGTGATCAGAA-3′), which contains hsp70d. These were used to amplify the hsp70c–d intergenic region from genomic DNA with the following conditions: 2.0 mM MgCl2; 10 pg of each primer; 5 min at 95°C, 30 cycles of 94°C for 1 min denaturation, 58°C for 30 s, and 70°C for 1.5 min, and finally 70°C for 7 min. The resultant 2.8 kb product was cloned into pGEM T-Easy Vector (Promega) and sequenced from both ends as described above.

For densitometric estimation of hsp70 gene copy numbers, X-ray film images of genomic Southern blots were analyzed with a Molecular Dynamics 300A Computing Densitometer and Origin 6.0.

For densitometric estimation of hsp70 gene copy numbers, genomic DNA of selected D. virilis and D. lummei strains was digested with either XbaI or XbaI + AccI before Southern blotting. Blots were hybridized with the XbaI–SacI fragment of the lambda phage 17 clone, which contains the corresponding sequence of the D. virilis hsp70g gene (Accession no. AY445084). X-ray film images of blots were analyzed with a Molecular Dynamics 300A Computing Densitometer and Origin 6.0. XbaI and XbaI + AccI digestion yielded an 11 kb and a 5.5 kb fragment, respectively, containing a single hsp70 gene; for each strain, the density of other fragments relative to this fragment was calculated. Five such independent estimates of density for each selected strain were averaged to calculate the putative hsp70 copy number.

Results

Chromosomal localization and puffing of hsp70 gene loci in the genomes of D. virilis and D. lummei

Putative hsp70-containing phages isolated from genomic libraries of D. virilis and D. lummei, identified by hybridization to a D. melanogaster hsp70 probe, hybridized strongly with a single region of Chromosome II (Fig. 1a). The D. melanogaster hsp70 probe also strongly hybridized to this region (data not shown). Previously the hsp70 locus was localized cytogenetically to 29C in Chromosome II in D. virilis (see Introduction).

Fig. 1a–d
figure 1

In situ hybridization to specific loci and heat-shock puffs in chromosomes of Drosophila virilis and D. virilis × D. lummei strain 200 (with 3 complete hsp70 copies) hybrids. a A D. virilis hsp70 probe hybridizes to a single locus in D. virilis strain 160. b In D. virilis strain 160 (with at least 7 hsp70 copies) × D. lummei hybrids, the heat-induced puff at 29C is assymetric, corresponding to the differing hsp70 copy number in the two parent species. Specific probes were not used to label these chromosomes. c,d In situ hybridization of a D. lummei probe to chromosomes of D. virilis × D. lummei hybrids. The probe, lambda phage 17, includes both a complete copy of hsp70 (hsp70d) and flanking sequence, which also contains a repetitive element specific to D. lummei. Thus this probe labels both a heat-induced puff in the 29C region of each homeolog (solid arrows) and the repetitive element throughout the D. lummei homeolog (e.g., open arrows). The D. virilis parent was strain 160 in c and strain ZZP (with 4 putative hsp70 copies) in d. Note the differing size of labeled puffs, consistent with the differing hsp70 copy numbers

Full size image

As expected, the 29C region forms large puffs after temperature elevation in both species and in inter-specific hybrids. Characteristically, puffs in the 29C region in inter-specific hybrids were often asynapsed, with the homeologous regions differing in size (Fig. 1b–d). Several features unambiguously assign each homeologous region to its parental species. One probe used in in situ labeling, D. lummei phage 17 DNA, contains an unidentified transposable element present in D. lummei but absent in D. virilis (unpublished data) in addition to the hsp70d gene. Conversely, other hybridizations (data not shown) were with probes that contained Penelope and Ulysses, transposable elements present in D. virilis but absent in D. lummei (Zelentsova et al. 1999) (data not shown). In all cases (e.g., Fig. 1c,d) the D. lummei homeolog was the smaller component in heat-induced puffs at 29C from D. virilis × D. lummei hybrids.

Inter-specific and intra-specific variation in hsp70 gene copy number

Southern blot hybridizations indicate that D. virilis and D. lummei differ strikingly in the size of genomic DNA fragments generated by restriction enzyme digestion (Fig. 2a). Although such differences are consistent with variation in restriction sites, hsp70 copy number, or both, we present evidence below that these differences reflect intra-specific and inter-specific variation in hsp70 copy number, at least in part. Densitometry of genomic Southern hybridizations for D. virilis (Fig. 2c,d) is consistent with seven copies of hsp70 genes in strain 160, six copies in strains 9, 1433 and T40, and four to five copies in strains T53, V-ZZP-01, V-WW-08, V-Hunan (data not shown), V-Nanjing (data not shown), and SBB (data not shown), and only three copies in strain A11. Two different restriction digests yield nearly identical data. Densitometry for the D. lummei strains studied is likewise consistent with intra-specific variation in copy number (four to five copies of the hsp70 gene depending on strain; quantitative data not shown), although the maximum copy number (five) is less than in D. virilis.

Fig. 2a–d
figure 2

Restriction fragment variation in the hsp70 gene clusters of Drosophila virilis and D. lummei. a,b Variation in genomic DNA from D. virilis and D. lummei strains hybridized to a D. virilis hsp70 probe: a XbaI digest; bXbaI + AccI digest. Note that the first eight strains and the second six strains were on separate blots. c,d Relative density of hsp70-hybridizing restriction fragments in selected D. virilis and D. lummei strains. Means and standard errors are plotted, and represent summary data from multiple hybridization experiments. XbaI digestion yielded an 11 kb fragment containing a single hsp70 gene; for each strain, the density of other fragments is standardized according to this fragment in c. XbaI/AccI digestion yielded a 5.5 kb fragment containing a single hsp70 gene; for each strain, the density of other fragments is standardized according to this fragment in d

Full size image

XbaI and XbaI/AccI restriction digests of D. lummei strain 200 genomic DNA were hybridized to both a complete D. lummei hsp70 sequence and a probe (the BamHI–SalI portion of D. melanogaster hsp70) specific for the 3′ region of hsp70 coding sequence (Fig. 3). Whereas the former probe hybridized to all fragments, the latter probe failed to hybridize to the 2.8 kb XbaI fragment and the 1.6 kb XbaI/AccI fragment, suggesting that this sequence (which below we name hsp70c) does not encode a complete Hsp70 protein. This truncation is present in all D. lummei strains studied so far (data not shown).

Fig. 3
figure 3

Hybridization of Drosophila lummei DNA restriction fragments with full-length D. lummei hsp70 sequence and the 3′ portion (BamHI–SalI fragment) of D. melanogaster hsp70

Full size image

Structure and sequence of the hsp70 gene cluster in D. virilis and D. lummei

The detailed organization of the hsp70 genes in D. virilis strain 160 and D. lummei strain 200 emerges from the subcloning and sequencing of overlapping lambda clones isolated from the corresponding genomic libraries (Fig. 4). In D. virilis strain 160 there are six copies in tandem orientation approximately 4.8 kb apart (hsp70b–g), and a seventh copy (hsp70a) in inverse orientation. Although genes c–f are identically spaced, they are distinguishable by the location of restriction sites (e.g., an AccI site adjacent to the 3′ end of gene f, HindIII sites in gene e and d) and 3′ flanking sequence. By contrast, in D. lummei strain 200 there are a pair of copies (hsp70a and hsp70b) in inverse orientation as in D. virilis, a third copy (hsp70c) 5.5 kb from hsp70b, and a fourth copy (hsp70d). Failure of a 3′ region-specific probe to hybridize to its corresponding restriction fragment (see above) suggests that hsp70c of D. lummei is a pseudogene. In both species, the hsp70a–hsp70b intergenic region is approximately 0.8 kb. As described above, the lambda clones yielding these data all hybridize to the 29C region of Chromosome II.

Fig. 4
figure 4

Arrangement of hsp70 copies at chromosomal locus 29C in Drosophila virilis (top) and D. lummei (bottom). The purple arrow indicates a putative pseudogene. Note scale in legend. Grey bars indicate size of XbaI fragments in kilobases. Numbers in boxes indicate specific lambda clones from which the arrangement (top row for each species) was deduced; in these arrangements (top rows), broken horizontal lines indicate hypothesized regions for which no contiguous lambda clones or sequences were obtained. Restriction sites include those represented in Fig. 2 and others (see key)

Full size image

The complete coding sequence was obtained for six of the seven hsp70 genes in D. virilis strain 160 (Table 1, Genbank Accession numbers AY445084–7 and AY445090–1). The six genes are highly conserved, varying at only five of 1923 sites. Of these varying sites, three are silent and two result in changes in amino acid sequence (998, GGG Glycine in one of six genes for GTG Valine; 1472, GTA Valine in one of six genes for GCA Alanine). In D. lummei strain 200, sequence was obtained for the coding region of the hsp70 genes and pseudogene (Table 1, Genbank Accession numbers AY4450888–9 and AY445092–3). Compared with the D. virilis genes, the complete D. lummei hsp70 genes exhibit a 9 bp deletion beginning at the counterpart of site 1456 in the D. virilis hsp70 genes. With these exceptions, the complete D. lummei hsp70 genes consistently differ from D. virilis hsp70 genes a–c and e–g at only 21 of 1923 sites, of which 19 are silent differences and two result in a single replacement (1666–8, TCC Serine in D. lummei and CCT Phenylalanine in D. virilis). The three complete D. lummei genes vary among themselves at ten sites (five silent substitutions and five replacements), and hsp70b lacks an entire codon (at 1873–5) present in the other two genes and in the D. virilis hsp70 genes.

Table 1 Nucleotide divergence in the hsp70 coding and 5′ flanking sequences of Drosophila lummei and D. virilis. Each column provides the site number (e.g., −0408 in the first column for 5′ flanking sequence) and the corresponding nucleotide for each gene included. The first base of the start codon is numbered + 1; the flanking nucleotide just 5′ to that is numbered − 1. For coding sequence, synonymous and replacement differences are noted and labeled S and R, respectively. For replacement differences, the corresponding amino acids are also listed. Nucleotides identical to the first sequence are indicated by dots; deletions by dashes. Blank entries indicate lack of sequence for the indicated nucleotide position. This analysis does not include hsp70d of D. virilis, for which only partial sequence was obtained, and hsp70c of D. lummei, which is a putative pseudogene
Full size table

For all of the above genes, at least 408 bp of sequence was obtained 5′ to the start codons. These D. lummei and D. virilis flanking sequences differ consistently at 17 sites, and all six D. virilis hsp70s exhibit a 15 bp deletion beginning 86 bp 5′ to their start codons. With one exception (371 nucleotides 5′ to coding sequence), these differences are at positions whose functional significance has not been established experimentally or is negligible; position −371 is in a putative GAGA element. Within each species, the hsp70 genes are highly conserved, varying at three to four sites in each species. The 3′ flanking sequence is likewise highly conserved among the D. lummei and D. virilis hsp70 genes. In particular, the 3′ flanking sequence of D. virilis hsp70b, hsp70e, and hsp70f is near-identical for more than 1600 bp.

The D. lummei pseudogene hsp70c lacks the first 300 nucleotides of coding sequence and the 13 nucleotides immediately 5′ to those. It then contains the next 809 nucleotides of D. lummei hsp70 genes, but lacks the remaining 813 nucleotides of D. lummei hsp70 coding sequence.

Discussion

From the foregoing data, we infer that multiple tandem duplications of hsp70 genes accompanied the evolution of the D. virilis group, and that changes in hsp70 copy number accompanied changes in thermotolerance both among D. virilis populations and between D. virilis and D. lummei. Multiple lines of evidence are all consistent with this inference: in situ hybridization indicating that the hsp70 locus itself is singular: chromosome puff size corresponding to putative hsp70 copy number; genomic Southern hybridizations and quantitative densitometry of restriction fragment polymorphisms; and the coding and flanking sequences of the genes themselves. Moreover, the data are consistent with earlier chromosomal localizations of these genes (Evgen’ev et al. 1978; Peters et al. 1980).

Several distinctive features of D. lummei and D. virilis have facilitated this diversity of approaches. First, decades of study uniformly indicate that the virilis subgroup is a (if not the) basal member of the Drosophila clade, that D. virilis is itself a (if not the) basal member of the group, and that D. lummei is recently (<5 MYA) derived from D. virilis or a common ancestor (see references cited in Introduction). Second, unlike many species pairs, these can interbreed to yield viable and partially fertile hybrids. Indeed, this partial fertility and multiple markers on each chromosome enable inter-specific genetics (Garbuz et al. 2002, 2003). Third, transposable elements present in each but not both species (Zelentsova et al. 1999) permit unambiguous identification of the parentage of any homeolog in hybrids.

According to diverse genetic and biochemical evidence, the inducible Hsp70 proteins play key roles in inducible tolerance of extreme temperatures, and both the magnitude and threshold of Hsp70 expression are correlated with the natural thermal regime in many species in many taxa (Feder and Hofmann 1999). Indeed, as we have shown (Garbuz et al. 2002, 2003), the low-latitude species D. virilis has both greater thermotolerance and greater Hsp70 levels than the high-latitude species D. lummei; in D. virilis populations, thermotolerance is usually correlated with Hsp70 level; and D. virilis × D. lummei hybrids are intermediate to the parental species in these respects. The coding and flanking sequence for the hsp70 genes, in those rare cases where it differs within and among species, provides little explanation for the inter-specific differences in thermotolerance and Hsp70 protein levels (Table 1). By contrast, as noted, hsp70 copy number differs dramatically. Although evolutionary loss or duplication of hsp70 genes and loci is not the only mechanism of achieving such evolutionary variation in Hsp70 level and thermotolerance, it is an important mechanism that has occurred repeatedly in evolution. The present consensus is that a single locus bearing two copies in inverted orientation is the primitive state of hsp70 in Diptera (see Introduction). Most likely hsp70a and hsp70b are orthologs of these two copies. Because D. pseudoobscura, intermediate in derivation to the virilis group and other Drosophila, also has a single locus bearing two copies in inverted orientation (B. Bettencourt, pers. commun.), parsimony suggests that an ancestor of the virilis group (rather than a common ancestor of all Drosophila) gained additional copies by tandem duplications. D. virilis strain A11 is from Southeast Asia, where the species probably originated (Throckmorton 1982), and has at most three hsp70 copies. By contrast, all other strains examined have four to seven copies, suggesting that tandem duplication of hsp70 accompanied the geographic range expansion of this species. The maximal copy number (seven), however, is in a strain (160) with multiple marker mutations and numerous transposable elements (Evgen’ev et al. 1997), and may thus be unrepresentative of natural variation. In D. virilis, these tandem duplicates are approximately 4.8 kb apart from one another and the ancestral pair member hsp70b (Fig. 4a), whereas the first tandem duplicate in D. lummei (hsp70c) is approximately 5 kb from hsp70b. We suggest, therefore, that one or more of the internal tandem duplicates were lost during or after the divergence of D. virilis and D. lummei to result in the lesser copy number but greater hsp70b–hsp70c spacing in D. lummei. Indeed, hsp70c in D. lummei is a putative pseudogene. Additional support for this scenario is that the sequences of both hsp70g in D. virilis and hsp70d in D. lummei diverge from that for the other D. virilis genes 202 bp 3′ of the end of the coding sequence. In any event, D. virilis clearly represents a remarkable example of gene proliferation by rampant gene duplication. D. melanogaster has evolved comparable hsp70 copy numbers, but apparently by a dissimilar route: duplication of the primitive hsp70 gene cluster and tandem duplication of one gene in the derived gene cluster, with the primitive gene cluster retaining its original copy number (Leigh-Brown and Ish-Horowicz 1981; Bettencourt and Feder 2001; Maside et al. 2002). Evidently in Drosophila the hsp70 copy number has been remarkably malleable during evolution (Fig. 5).

Fig. 5
figure 5

Mapping of hsp70 clusters and their arrangement onto a phylogeny of selected Drosophila and other Diptera (Ashburner 1989; Bettencourt and Feder 2001). A heavy bar underlies each hsp70 cluster. Five events are inferred by parsimony or from the cited references: (1) tandem duplication of a member of the ancestral inverted pair (present study); (2) loss of one or more tandem duplicates (present study); (3) tandem duplication of a member of the ancestral inverted pair (Konstantopoulou et al. 1998); (4) duplication of the ancestral inverted pair via retroposition (Leigh-Brown and Ish-Horowicz 1981; Bettencourt and Feder 2001; Maside et al. 2002); (5) tandem duplication of a member of the duplicated ancestral pair (Leigh-Brown and Ish-Horowicz 1981; Bettencourt and Feder 2001; Maside et al. 2002). The cluster marked by an asterisk contains at most three copies, and is for D. virilis recently collected from Southeast Asia. Ψ in D. lummei indicates a pseudogene lacking 3′ sequence of hsp70. Note that characterizations of hsp70 clusters for subgroups are based on one or a few species in each subgroup

Full size image

One possible explanation for this malleability is the abundance of transposable elements in the Drosophila genome; these facilitate retrotransposition, ectopic recombination, and other modes of gene duplication or rearrangement (Bushman 2002). Moreover, the hsp70 genes may be especially susceptible to the insertion of transposable elements owing to their constitutively open chromatin conformation (Zatsepina et al. 2001; Lerman et al. 2003). Indeed, not only do the sequences flanking the D. virilis hsp70 genes bear micro-satellites and PDV sequences (Zelentsova et al. 1986), the 3′ flanking sequence includes a portion of an ancient, now non-functional LTR-containing retrotransposon, JEM (J. Vieira, pers. commun.). These sequences collectively are abundant throughout the virilis group genomes (Biessmann et al. 2000; Evgen’ev et al. 2000b), and may explain the hybridization of hsp70-containing lambda clones to diverse chromosomal loci.

Once having duplicated, the hsp70 genes have remained remarkably similar in both D. virilis and D. lummei, and in both coding and flanking sequence. This similarity is not due to inadvertent multiple cloning of the same gene because the various genes in both D. virilis and D. lummei are distinguishable by restriction sites, fragment sizes, and 3′ flanking regions (Fig. 3). Potential explanations for this similarity include strong positive selection on the hsp70 loci, selective sweeps of linked loci, and concerted evolution via gene conversion. Because we have sequence for only a single line or population of each species, we cannot differentiate among these explanations directly. Nonetheless, as in D. melanogaster, the hsp70a and hsp70b genes in each species have likely been distinct for >100 MY (Bettencourt and Feder 2002). Despite this time, these paralogs are identical in both coding and flanking sequence in D. virilis, even at silent sites. We therefore hypothesize that, as in D. melanogaster and other members of its subgroup (Bettencourt and Feder 2002), gene conversion has been especially effective in homogenizing this sequence. The discovery of the ancient transposable element JEM in the 3′ flanking region of most hsp70 copies of D. virilis and D. lummei favors this hypothesis because transposable elements have been implicated in the concerted evolution of tandemly repetitious DNA (Thompson-Stewart et al. 1994). Interestingly, the Adh gene in three virilis group species, as Nurminsky et al. (1996) describe, has many parallels with the present study, including multiple tandem duplications, repeated elements in flanking sequence, and extensive similarity of both coding and flanking sequence between the ancestral and duplicated genes (Nurminsky et al. 1996). As do we, they attribute this similarity to extensive gene conversion.