Evolution of overwintering strategies in Eurasian species of the Drosophila obscura species group

The phylogenetic relationship of Eurasian species of the Drosophila obscura species group remains ambiguous in spite of intensive analyses based on morphology, allozymes and DNA sequences. The present analysis based on sequence data for cytochrome oxidase subunit I (COI) and a-glycerophosphate dehydrogenase (Gpdh) suggests that the phylogenetic position of D. alpina is also ambiguous. These ambiguities have been considered to be attributable to rapid phyletic radiation in this group at an early stage of its evolution. Overwintering strategies are diversified among these species: D. alpina and D. subsilvestris pass the winter in pupal diapause, D. bfisciata and D. obscura in reproductive diapause, and D. subobscura and D. guanche without entering diapause. This diversity may also suggest rapid radiation at an early phase of adaptations to temperate climates. On the other hand, adult tolerance of cold was closely related to overwintering strategy and distribution: D. obscura and D. bfmciata with reproductive diapause were very tolerant; D. alpina and D. subsilvestris which pass the winter in pupal diapause were less tolerant; D. subobscura having no diapause was moderately tolerant and D. guanche occurring in the Canary Islands was rather susceptible. Tolerance of high temperature at the preimaginal stages seemed to be also associated with overwintering strategy; i.e. lower in the species with pupal diapause than in those with reproductive diapause or without diapause mechanism.

The Drosophila obscura species group has been the subject of many studies on evolutionary phenomena, for example, reproductive isolation, mating preference, and inversion polymorphism and genetic variation in natural populations (reviewed in Dobzhansky & Powell, 1975;Lakovaara & Saura, 1982;Powell, 1997). The phylogeny of this species group has also been extensively studied on the basis of morphology, chromosomes, allozymes and DNA sequences of several genes. These studies identified two subgroups, a$nis and pseudoobscura, in the North American species, and two clades, obscura and subobscura, plus two other species (0. bfasciata Pomini and D. subsilvestris Hardy & Kaneshiro) with unresolved phylogenetic positions in the Eurasian species (Lakovaara & Saura, 1982;Beckenbach, Wei & Liu, 1993;Pelandakis & Solignac, 1993;RUSSO, Takezaki & Nei, 1995;Gleason et al., 1997;Barrio & Ayala, 1997). It has also been suggested with allozyme data that D. alpina Burla is not appreciably related to any subgroups (Lakovaara et al., 1976), but its phylogenetic position has not been analysed by molecular data. Thus, ambiguities still remain on the phylogenetic relationship among the Eurasian species. Ayala (1997) andGleason et al. (1997) considered that these ambiguities are attributable to rapid phyletic radiation at an early phase of the evolution of this group. Here we report the phylogenetic position of D. alpina based on sequence data of COI (mitochondrial) and Gpdh (nuclear) genes and also add data on overwintering strategies (diapause and temperature tolerance) for the Eurasian species of this group.
Seasonality is an important factor shaping the evolution of temperate and arctic insects (Lumme & Lakovaara, 1983;Tauber, Tauber & Masaki, 1986;Danks, 1987). In fact, the acquisition of overwintering abilities has been suggested to play a key role in the evolution of temperate species in the Drosophila melanogaster species group (Kimura, 1988;Kimura et al., 1994;Ohtsu et al., 1993;Ohtsu, Kimura & Katagri, 1998). In the obscura group, information on overwintering strategies would be also useful to understand its adaptations in temperate regions and resolve its speciation history. It has been reported that the members of the obscura group vary in their  (Begon, 1976;Lumme & Lakovaara, 1983;Toda et al., 1986;Beppu, Yoshida & Kimura, 1996).

Flies
The experimental strains used originated as follows: D. obscura from 10-20 females collected in Tubingen (Germany) in 1993; D. subobscuru from several females collected at Heuberger Tor (near Tubingen) in 1993; D. guanche Monclus from an old strain collected in the 1970s from the Canary Islands; D. subsiluestris from a single female collected at Heuberger Tor in 1992; D. alpina from several females collected in Shiga (central Japan) in 1993; D. b$mciuta from females collected in Sapporo (SP: Japan) and Yakutsk (YK: Siberia, Russia) in 1993. Except for D. guanche, diapause and temperature tolerance were examined within one or two year(s) following the collection of the strains. Cornmeal-malt medium was used for rearing.

Sequencing of COI and Gpdh genesfrom D. alpina
To perform PCR, DNA was extracted from D. ahina according to the method of Goto, Yoshida & Kimura (1998). RNA in the samples was digested with RNase A.
The primers used to amplify COI were designed according to Gleason et al. (1997); F-COI, 5'> CCAGC TGGAG GAGGA GATCC >3'; R-COI, 5'> CCAGT AAATA ATGGG TATCA GTG>3'. In addition, we designed the primer on the basis of the sequence derived from D. tukahashii (Nigro, Solignac & Sharp, 1991) for nested PCR; takl, 5'> GCTTG AGCCG GAATA GTAGG >3'. The first base of takl corresponds to position 1537 in the D. melanogaster Meigen mtDNA sequence (the Genbank accession number is U37541). Primary reaction of nested PCR used 100 ng of DNA, 1 U of AmpliTaq DNA polymerase (Perkin-Elmer), and final concentration of 1.5 mM of MgCl,, 1 x PCR buffer I1 as formulated by Perkin-Elmer, 0.3 pM of takl and R-COI primers, and 0.2 mM of dNTP in a total volume of 50 pl. Amplification was performed with 35 cycles of 30 sec denaturing at 94OC, 30 sec annealing at 50"C, and 45 sec extension at 72°C. Secondary reaction used the same components of the primary reaction, except that the primers were F-COI and R-COI, and the target was 1 p1 of the primary amplified product. PCR conditions were the same as above. The length of amplified products agreed with the one of COI reported by Gleason et al. (1997).
Gpdh fragment was amplified with the primer set of GNL and GNR (Barrio & Ayala, 1997). PCR components were the same as those used for amplification of the COI region, except the primers. Amplification was performed with 35 cycles of 30 sec denaturing at 94"C, 30 sec annealing at 56"C, and 75 sec extension at 72°C.
The amplified fragments were purified from the gel using Prep-A-Gene DNA purification kit (BIO-RAD). The sequences were obtained from an ABI 373A automated sequencer (PE Applied Biosystems) with DNA sequencing kit (Dye Terminator Cycle Sequencing Ready Reaction; PE Applied Biosysterns) according to supplier's instructions. Cycle sequencing was performed using two oligonucleotides (Fand R-COI) and six (GNL, GNR, L4BN, L4E, R5B and R4M; Barrio & Ayala, 1997) for COI and Gpdh fragments, respectively. The obtained nucleotide sequence data of D. abina were deposited in the DDBJ/EMBL/GenBank nucleotide sequence databases with the accession numbers AB016601 and AB016727 for COI and Q d h , rcspecti\,ely.
The species name and the Genbank accession numbers for the COI and Q d h sequences used for the phylogenetic analysis are as follows (two numbers are for the COI and Cpdh sequences, respectively): Drosophila melanogaster (U3754 1 , X14 179); D.

Phylogenetic inference
The sequences were aligned using the CLUSTAL V program (Higgins & Sharp, 1988). The length of the COI sequence was 462 bp and the first base corresponds to position 2167 in the D. melanogaJter mtDNA sequence (the Genbank accession number is U37541). O n the other hand, the first and last bases of the Gpdh sequences correspond to position 3036 and 4230 in the D. melanogaJter gene (the Genbank accession number is X14179), respectively.
For phylogenetic analysis, we used the neighbour-joining (NJ; Saitou & Nei, 1987) and the maximum-parsimony (MP; Swofford & Olsen, 1990) methods. The numbers of nucleotide substitutions per site of COI and Gpdh were estimated by Tamura 8r Nei's (1 993) method and Kimura's (1 980) two-parameter model, respectixrely, by XIEGA 1 .0 (Kumar, Tamura & Nei, 1993). The statistical confidence of a particular cluster of sequences in the NJ trees was evaluated by the bootstrap test (1 000 pseudoreplicates) by MEGA. The h/lP trees and their bootstrap tests (500 pseudoreplicates) were obtained by the programs DNAPARS and SEQBOOT, respectively, implemented in PHYLIP ver. 3.572 (Felsenstein, 1993). Since the introns were so diversified among the species studied, the gaps and missing information in the sequences were aligned with the pairwise-deletion option in MEGA.

Diapause and temperature tolerance
Experimental individuals were reared from eggs and exposed to short (1 0 h-light/ 14 h-dark) and long (15 h-light/9 h-dark) daylengths at 15°C and examined for eclosion. For species which eclosed at these photoperiods, ovarian development was examined 16 days after eclosion; females with undeveloped ovaries were classified a5 bcing in reproductive diapause. In addition, flies were examined to ascertain uhether they were able to complete preimagmal development at 15, 21, 25.3 and EVOLUTION IN THE DROSOPHILI OBSCCRd GROUP 433 29°C (exposed to long daylength or continuous illumination): about 100 eggs were placed in vials containing 10 ml food medium (about 50 eggs per vial) and examined for the rate of pupation (0. alpina and D. subsilvestris) or eclosion (the other species).
To examine tolerance of cold, the experimental species except D. alpina were reared from the egg stage, exposed to short and long daylengths at 15"C, continuous illumination at 21°C and examined for survival at low temperatures 16 days after eclosion. In D. alpina, which entered pupal diapause irrespective of photoperiod, the experimental protocol was as follows: individuals were reared from the egg stage, exposed to a short daylength at 15OC; diapausing pupae thus obtained and aged 2 months after being oviposited were reared exposed to 4°C (constant darkness) for 2.5 months; these cold-treated pupae were exposed to a long daylength at 15°C and adults eclosed from these pupae were examined for cold tolerance 16 days after eclosion. Experimental flies were introduced into vials containing food medium and exposed to low temperature (-1 to -15°C) for 24 h (about 30 individuals at each temperature). Survival was examined after flies were maintained at 15°C for 24 h after cold treatment. Temperatures that killed 25, 50 and 75% of populations (LT2j,jo,7j) were obtained by reading the intercepts after survival was plotted against exposure temperature.

Phylogenetic anabses of COI sequences
The fragment of the COI gene of D. alpina has a high proportion (71.4%) of A + T , especially in third codon positions (95.4%) and at fourfold degenerate sites (100%), as has been reported for the other members of the obscuru species group and also for many other Drosophila species (DeSalle et ul., 1987;Nigro et al., 1991;Tamura, 1992;Beckenbach et ul., 1993;Barrio, Latorre & Moya, 1994;Gleason et al., 1997). Given the existence of substantive composition bias, we estimate nucleotide divergence according to Tamura & Nei's (1 993) method.
The ratio of transitions to transversions (Ti/Tv) between D. alpina and the other species ranges from 0.471 (0. ajhis) to 1.301 (0. guanche) ( Table 1). In the obscuru group, Gleason et al. (1997) reported that the Ti/Tv ratio ranges from 0.29 to 3.68 and is generally high between closely related species and low between distantly related species. This trend has been explained by the fast saturation of transitional substitutions (DeSalle et al., 1987;Gleason et al., 1997). We used only transversions for the phylogenetic analyses, following Gleason et al. (1997).
The number of transversional substitutions per site between D. alpina and the other species ranges from 0.041 1 (0. guanche) to 0.0655 (0. pseudoobscura and some other North American species) (Table 1). Figure 1   from exon 4, 67 bp from intron IV, 154 bp from exon 5, 89 bp from intron V and 53 bp from exon 6. Due to the difficulty of unambiguously aligning the Gpdh intron sequences, we examined separately the exon and the intron sequences. There is little bias in G + C content in the Gpdh exon sequences: the averages of the nucleotide frequencies are 0.242 A, 0.232 C, 0.272 G and 0.254 T . Given the absence of substantive composition bias, we estimate nucleotide divergence according to Kimura's (1 980) two-parameter model (Table 1). Divergence between D. alpina and the other species in all substitutions ranges from 0.0237 (0. obscura) to 0.1522 (0. melanogaster). The topologies of the NJ and MP trees are identical except the position of D. alpina. In the NJ tree, D. alpina is placed at the root of the cluster of the Eurasian species ( Fig. 2A), but at the root of the subobscura subgroup with lower bootstrap value in the MP tree (not shown).
In the NJ tree based only on the intron sequences (which does not include the outgroup D. melanogaster), D. alpina is placed at the root of the D. subsiluestris and D. bijusciata clusters with a lower bootstrap value (Fig. 2B). In the MP tree, D. alpina is diverged from the root of the cluster of the Eurasian species with a bootstrap value of 96 (not shown).

Photoperiodic control of diapause
In D. alpina, almost all individuals which developed to pupae did not eclose in 2 months after being oviposited irrespective of photoperiod (Table 2 and Fig. 3). In D. subsilvestris, most of the pupae did not eclose in the two months after being oviposited and exposed to a short daylength, but about 74% of pupae eclosed with a long daylength (Table 2). When these pupae were maintained at 4OC for 2 4 months and then brought into an environment at 15"C, a number of them eclosed (data not shown), indicating that these pupae were in diapause. When D. alpina pupae were maintained at 15°C further, some of them eclosed 3-6 months after being oviposited with a long daylength, although almost none of them eclosed with a short daylength (Fig. 3).
In D. bfusciata and D. obscura, the percentage of females in reproductive diapause was significantly (XL-test, Pc0.05) higher at a short daylength than at a long daylength (Table 2). O n the other hand, D. subobJcuru and D. guanche showed no sign of photoperiodic diapause.
Half lethal temperature (LTjO) for adult flies of the experimental species is shown in Figure 5. In all species, LT,, was lower in flies reared at 15OC than in those reared at 21°C. In D. bfusciata and D. obscum, flies reared at a short daylength usually had lower LT,, than those reared at a long daylength: the difference in survival rate between individuals reared at short and long daylengths was significant (Xj-test, P<o.oI), at least at one temperature regime. In D. bfasciata, the Yakutsk strain exhibited more tolerance of cold than the Sapporo strain, while D. obscura was as tolerant as the Yakutsk strain. On the other hand, D. suboscura exhibited less tolerance than the above two species, and D. guanche was rather susceptible to cold. The other two species, D. alpina and D. subsiluestriJ, displayed less tolerance, but more than D. guanche.

DISCUSSION
The phylogenetic position of D. alpina Previous molecular studies (Beckenbach et al., 1993;Pelandakis & Solignac, 1993;Russo et al., 1995;Gleason et al., 1997;Barrio & Ayala, 1997) reported that the phylogenetic relationship among the Eurasian members of the obscura species group is ambiguous. The present study also suggested that the phylogenetic position of D.
a&na is ambiguous: its position differed by genes or methods used for the analysis. Gleason et al. ( 1 997) and Barrio & Ayala (1997) considered that these ambiguities arc attributable to rapid phyletic radiation at an early phase of the evolution of this group. They also identified the obscura-tristis-ambkua and subobscura-guanche-madeirensis subgroups plus two other species with unresolved phylogenetic positions, D. bfasciata and D. rubsilvestris, in the Eurasian members of the obscura species group.
The present analysis based on the sequence data of COI and Gpdh revealed that D. alpina does not cluster with any other members of this species group. Lakovaara et al. (1976) also suggested with allozyme data that D. alpina is not appreciably related to any cluster. D. alpina is also distinct in this species group in morphology and chromosomal composition (Lakovaara & Saura, 1982). It is therefore assumed that D. alpina is derived from an early radiation of this species group.

Evolution of overwintering strategies
Throckmorton (1 975) suggested, based on comparison of morphological characters, that the obscura species group is most closely related to the melanogaster species group. This notion is supported by the cladistic analysis of morphological characters (Grimaldi, 1990) and also by molecular data (Pelandakis & Solignac, 1993;Russo et al., 1995). At present, the melanogaster species group shows major radiation in tropical regions (Lemeunier et al., 1983), with the obscura species group predominant in temperate forests (Lakovaara & Saura, 1982). Based on the view that drosophilids are of tropical origin (Throckmorton, 1975), it is assumed that the melanogaster group retains ancestral characteristics and that the ancestor of the obscura group split from the melanogaster group following its colonization of temperate regions.
It is considered that the subobscura subgroup displays an ancestral characteristic in climatic adaptations (lack of diapause), since tropical species of the melanogastei group do not have a diapause mechanism (Kimura, 1988;Kimura et al., 1994). However, there is also the possibility that the subobscura subgroup has lost the diapause mechanism which had once evolved in conjunction with its colonization of temperate regions. Kimura (1 988) reported such a case in D. triaurarza; subtropical populations of this species are assumed to have lost a mechanism of reproductive diapause following secondary colonization of subtropical areas. EVOLUTION IN THE DROSOPHZLA OBSCc'R1 GROUP 439 Overwintering strategies are diversified in the obscura species group: D. alpina and D. subsilvestris pass the winter in pupal diapause, D. bfasciata and D. obscura do so in reproductive diapause, and D. subobscura and D. guanche do so without entering diapause (see also Begon, 1976;Lumme & Lakovaara, 1983). This diversity may also suggest rapid radiation at an early phase of adaptations to temperate climates. It is not clear whether each type of diapause evolved once or repeatedly in this group. However, the distant phylogenetic relationship between the species having the same type of diapause suggests that repeated evolution is likely.
In D. alpina, the photoperiodic response was observed when pupae were maintained for more than 3 months at 15°C. However, this response would be concealed under natural conditions, because this species is univoltine in Shiga (the locality where the experimental strain originated) and usually pupates in mid to late August when daylength is not long (Beppu et al., 1996). The photoperiodic response of this species may be a remnant of past adaptations. At the early phase of adaptations of the obscura group to temperate climates, its distribution would be restricted to relatively warm areas and its members including the ancestor of D. alpina would have multivoltine life cycles and photoperiodically controlled diapause. In D. alpina, the photoperiodic response became to be concealed during the evolution of univoltine life cycle as a result of its adaptation to alpine and boreal climates. Ichijo, Beppu & Kimura (1992) also reported a concealed photoperiodic response in another univoltine species with reproductive diapause, Lkosophila moriwakii.

irmperature tolerance
Among the species studied, adult tolerance of cold is closely related to overwintering strategy and distribution. Adult flies of D. obscura and D. bgasciata that pass the winter in reproductive diapause displayed considerable tolerance of cold. The Yakutsk strain of D. b@sciata revealed more tolerance than the Sapporo strain. Adult D. alpina and D. subsilvestris with pupal diapause were less tolerant, while adult D. subobscura were moderately tolerant and therefore able to overwinter even in England or Germany. D. guanche occurring in the Canary Islands, where mild weather lasts throughout the year, was rather susceptible, although it is highly likely that this species has lost its tolerance over 20 years of laboratory rearing. It is not certain whether this species retains an ancestral characteristic (low tolerance), has lost tolerance consequent to its colonization of the Canary Islands or lost it since the 1970s in the laboratory. In the melanogaster species group, it has also been observed that tolerance is related to its distribution range (Kimura, 1988;Kimura et al., 1994;Ohtsu et al., 1998). Tolerance of cold would be a very important trait to survive the winter in temperate or arctic regions and subject to strong selection. However, in several species of the melanogaster species group, little geographic variation was observed in tolerance of cold (Kimura, 1982(Kimura, , 1988Kimura et al., 1994). It is considered that genes controlling tolerance are highly coadapted or have serious pleiotropic-side effects in these species and that therefore tolerance does not change without drastic genetic reform, for example that leading to speciation.
Among the species studied, D. alpina and D. subsilvestris with pupal diapause had lower tolerance to high temperature in preimaginal development than the other species with reproductive diapause or without diapause. Tolerance to high temperatures may also have evolved in association with overwintering strategies.