Natural Selection Favors a Newly Derived timeless Allele in Drosophila melanogaster

See allHide authors and affiliations

Science  29 Jun 2007:
Vol. 316, Issue 5833, pp. 1895-1898
DOI: 10.1126/science.1138412


Circadian and other natural clock-like endogenous rhythms may have evolved to anticipate regular temporal changes in the environment. We report that a mutation in the circadian clock gene timeless in Drosophila melanogaster has arisen and spread by natural selection relatively recently in Europe. We found that, when introduced into different genetic backgrounds, natural and artificial alleles of the timeless gene affect the incidence of diapause in response to changes in light and temperature. The natural mutant allele alters an important life history trait that may enhance the fly's adaptation to seasonal conditions.

Although polymorphism at a single locus may sustain adaptive variation in nature for both behavioral and morphological phenotypes, there are few well-documented examples where a new, naturally arising adaptive mutation has spread through a population (1, 2). In D. melanogaster, circadian behavior is generated by regulatory interactions among a number of canonical clock genes (3). One of these genes, timeless (tim), encodes a light-responsive component that has two allelic forms, ls-tim and s-tim (4). The ls-tim allele generates both full-length L-TIM1421 and truncated S-TIM1398 products from an upstream initiating methionine codon and a second ATG 23 codons downstream (Fig. 1). In s-tim, deletion of the G nucleotide at position 294 interrupts the upstream reading frame with a stop codon, generating S-TIM1398, from the downstream ATG (4, 5) (Fig. 1). These variants were identified initially in laboratory strains, so we sought to investigate whether this polymorphism was present in nature.

Fig. 1.

Alternative ATG start codons of ls-tim and s-tim. The N-terminal coding sequences for both alleles are shown together with their corresponding protein translations. The G insertion/deletion (position 294, GenBank U37018) in ls-tim allows it to generate both the L-TIM1421 and S-TIM1398 isoforms, whereas the s-tim allele may also generate a 19-residue peptide from the upstream ATG (4, 5).

Drosophila melanogaster isofemale lines were established from natural populations collected from southern Italy to Sweden (table S1). A polymerase chain reaction–based strategy identified the status of the two 5′ tim haplotypes in flies from isofemale lines. The frequency of ls-tim was plotted against latitude (Fig. 2A and table S1), and regression analysis (F1,11 = 43.7, P < 0.00005, R2 = 0.80) and subsequent spatial autocorrelation statistics (Moran's I = 0.28, P < 0.05) revealed a significant latitudinal cline, with high frequencies of ls-tim in southern Europe. Phylogenetic analyses of tim alleles showed that all ls-tim haplotypes, irrespective of geographical location, clustered at the top of the trees, which suggests that this is the derived allele produced by the insertion of the G nucleotide (fig. S1). Assuming that the split between D. melanogaster and D. simulans occurred 2 million to 2.5 million years ago (6, 7), we calculated that the ls-tim allele originated ∼8000 to 10,000 years ago, coinciding with the postglacial period and subsequent colonization of the Eurasian continent by D. melanogaster (7).

Fig. 2.

Frequency of ls-tim in European natural populations. Open circles, 13 populations collected in 1997; solid square, Heraklion, Crete, solid triangle, Haifa, Israel, both collected 2002; solid circles, Bitteto and Salice Salento, Italy, and Houten, Netherlands, collected 2004. (A) ls-tim versus latitude. Regression line is fitted through the data for the 1997 collections only. (B) ls-tim versus direct distance from Novoli, Italy. Regression line is fitted through 1997 populations plus Heraklion and Haifa. (C) ls-tim versus overland distances from Novoli. Regression line is fitted as in (B) (20).

We next examined the frequencies of the derived ls-tim allele southward from the putative site of origin, Novoli, Italy, which has the highest ls-tim frequency. Isofemale lines were established from populations in Crete, Israel, and Africa (Kenya and Zimbabwe) (table S1). The frequency of ls-tim was 0.138 in Crete, 0.318 in Israel, and zero in sub-Saharan Africa (Fig. 2A and table S1). When the Cretan and Israeli populations were added to our analyses, the data did not conform to the latitudinal cline [F1,13 = 1.94, not significant (n.s.), R2 = 0.13, Fig. 2A]. However, when we replotted all the ls-tim frequencies against direct distance from Novoli, the regression was highly significant (F1,13 = 16.35, P = 0.0014, R2 = 0.56; Fig. 2B and table S1) and was further enhanced when realistic and predominantly land-based distances between Novoli and all locations were used (F1,13 = 40.61, P < 0.00001, R2 = 0.76; Fig. 2C and table S1).

To test whether the pattern of tim polymorphism is consistent with selection or reflects historical or demographic processes, we applied Tajima's D (8) and Fu andLi's statistics (9) to the polymorphic region and to two intergenic regions downstream of the polymorphic fragment: 3′A and 3′B, located 2.5 kb and 14 kb downstream of the tim transcription unit, respectively. The results were highly significantly negative for the polymorphic ls/s-tim region, reflecting an excess of mutations that appeared only once as singletons, which suggests directional selection (Table 1). The 3′A fragment gave marginal significance, whereas 3′B gave nonsignificant results for all tests, revealing a consistent change in evolutionary dynamics with increasing distance from the tim ls/s polymorphic site (Table 1). The HKA test was also used to compare the relative amounts of polymorphism and divergence among these three sites (10). Comparing the N-terminal tim ls/s fragment with each of the two downstream flanking sequences (3′A and 3′B) gave highly significant deviations from neutral expectations, whereas comparing the two downstream flanking regions did not (Table 2). We also observed that one haplotype dominated the ls-tim allelic class; 16 of 21 alleles were monomorphic and the others were singletons. Using a haplotype test that can detect very recent selective events (11), we rejected the neutral hypothesis for ls-tim (P < 0.05), but not for the s-tim class, in which the most common haplotype was represented by 10 of 24 sequences (P = 0.2).

Table 1.

Results of neutrality tests: intraspecific analyses. The ls/s tim polymorphic site was compared to two downstream sequences, 3′A and 3′B. *P <0.05, **P < 0.01, ***P < 0.001. Sequences are described in (20).

View this table:
Table 2.

Results of neutrality tests: interspecific analyses. The ls/s tim polymorphic site was compared to two downstream sequences, 3′A and 3′B. In these analyses, HKA comparisons of ls/s tim with the downstream sequences to tim were highly significant, except for the HKA test comparing 3′A with 3′B2 = 0.09, P = 0.76, n.s.). ***P < 0.001, ****P < 0.0001. Sequences are described in (20).

View this table:

Our results support the view that the derived ls-tim allele arose in southern Italy about 8000 to 10,000 years ago and has spread, perhaps quite recently, in all directions as a result of directional selection. Alternatively, a recent selective sweep might not have allowed enough time for the accumulation of genetic variation around the polymorphic tim site, and consequently balancing selection might be difficult to detect with neutrality tests. Under such a balancing scenario, ls-tim would be particularly well adapted to southern Italy but would be less advantageous farther north or farther south.

To investigate phenotypes that might provide the substrate for selection, we examined whether temperature compensation—the ability of the clock to maintain a circadian 24-hour period during fluctuations in temperature—is driving the observed directional selection. Polymorphism in another clock gene, period, ismaintained by balancing selection (12), possibly by differential circadian temperature compensation (13), and shows a robust latitudinal distribution in Europe (14). However, replicate homozygous natural lines of s-tim and ls-tim and two laboratory strains carrying ls-tim showed similar temperature compensation (fig. S2). To avoid complications with genetic background and to study the effect of the L-TIM isoform, we generated four independent transgenic lines for two tim transgenes, P[L-tim] and P[S-tim], which generated one or the other isoform, respectively, with the available tim promoter sequences (15). All lines rescued circadian locomotor rhythms in arrhythmic tim01 hosts (16), with no effect of genotype on temperature compensation (table S2 and fig. S2).

D. melanogaster survive unfavorable seasons by entering a reproductive adult diapause that is mediated in part by a response to short days and long nights at low temperatures (1719). This combined response can be diagnosed in individual females by the lack of eggs in their ovaries caused by an arrest in oogenesis (18). We established isofemale lines from recently captured natural populations, two from southern Italy (Bitetto and Salice Salento) and one from the Netherlands (Houten) (table S1). Analysis of diapause in homozygous ls-tim and s-tim females within these populations revealed highly significant effects for population (F2,83 = 18.9, P <10–7), genotype (F1,83 = 103.8, P < 10–8), photoperiod (F5,83 = 11.0, P <10–8), and population × genotype interaction (F2,83 = 5.03, P <0.01) (Fig. 3, A to C). The ls-tim females showed reproductive arrest more readily than s-tim females in all three populations (Fig. 3, A to C). The photoperiodic curves for the two genotypes were largely parallel for the Salice and Houten populations, where even at the longest photoperiod, ls-tim females were more prone than s-tim females to diapause (Fig. 3, B and C). In contrast, for Bitteto, significant genotype differences emerged only as photoperiods grew shorter [14 hours light/10 hours dark (LD14:10), P = 0.005, Duncan's test; genotype × photoperiod interaction, F5,36 = 2.13, P = 0.08] (20). Latitude also had a significant effect, with northern s-tim females showing significantly higher levels of diapause than southern s-tim females (Houten versus Salice, P = 0.01, Houten versus Bitteto, P = 0.0002, Duncan's test), whereas for ls-tim, Houten was significantly different from Bitetto only (P = 0.0001) (Fig. 3, A to C). These results reveal that the ovarian diapause of European D. melanogaster is enhanced at shorter day lengths, at northern latitudes, and by the derived ls-tim allele relative to the ancestral s-tim variant.

Fig. 3.

Ovarian diapause of tim variants at 13°C (mean proportions, arcsin ± SEM). Photoperiod denotes hours of light per 24 hours. (A to C) Natural populations: Bitetto, Salice Salento (both Italy), and Houten (Netherlands). Squares, ls-tim; triangles, s-tim (N = 5555) (20). (D) tim transformants. Squares, P[LS-tim]; triangles, P[S-tim] (four lines with no significant line effect, F3,63 = 2.43, n.s.); diamonds, P[L-tim] (two lines, F1,35 = 0.05, n.s.). P[L-tim] and P[S-tim] transformant females were also tested at LD20:4 (N = 3404). (E) Cantonized wild-type (squares) and tim01 (circles, N = 1444).

We also examined diapause in the tim01 hosts transformed with P[S-tim], P[L-tim],and P[LS-tim], a corresponding transformant line that carries the ls-tim sequence (15). Highly significant genotype (F2,124 = 7.06, P = 0.00012) and photoperiod (F3,124 = 47.4, P < 0.00001) effects were observed, with an enhancement in the diapause responses of females carrying the P[L-tim] and P[LS-tim] transgenes relative to those carrying P[S-tim] (F1,118 = 5.24, P = 0.024, and F4,118 = 40.0, P = < 0.0001, respectively). In addition, an unexpectedly low diapause response was observed at the shortest 10-hour photoperiod (LD10:14), in contrast to the data from natural strains (Fig. 3, A to C). Similar results have been observed in shorter photoperiods with long-standing laboratory stocks (21), so this may reflect the genetic background on which the transgenes are expressed, or limitations of the 5′ tim promoter used to drive the transgenes. Nonetheless, these results reveal that irrespective of genetic background, ls-tim females show a significantly higher level of diapause than s-tim females, and that this is due to the tim locus itself. If we extrapolate these findings to nature, ls-tim (and P[L-tim] and P[LS-tim]) females might be expected to enter diapause earlier than s-tim (and P[S-tim]) females in response to the oncoming European winter.

We also investigated whether the circadian arrhythmic tim01null mutant would affect the ovarian phenotype. We compared tim01 to a wild type (ls-tim) after minimizing differences in genetic background, and significantly higher levels of diapause were observed in the mutant (F1,29 = 14.81, P = 0.0006), but without a significant response to photoperiod (Fig. 3E). A similar result was obtained with hemizygous P[LS-tim] transformants compared to tim01 at two photoperiods (LD8:16 and LD16:8) on a different genetic background (F = 6.3, P = 0.027; fig. S3). Consequently, variation in tim itself, not genetic background, is responsible for these changes in the incidence of diapause.

A latitudinal cline in the incidence of diapause was observed in natural D. melanogaster populations in the eastern United States, with a higher incidence at northern latitudes (18). Within a single temperate population, genotypes that show higher levels of diapause are stress-resistant and have enhanced fitness under such unfavorable conditions, demonstrating that in temperate habitats with strong seasonality, enhanced diapause in D. melanogaster has adaptive value (1719). The higher levels of diapause observed with ls-tim genotypes may have similar adaptive value in the European environment. Our natural strains also show a higher incidence of diapause in carriers of the ancestral s-tim allele in northern populations than in southern populations (Fig. 3, A to C); this result is consistent with findings that in arthropods, the higher the latitude or altitude, the more readily diapause is induced (18, 22, 23). Consequently, there is genetic variation other than in tim that is causing this latitudinal change within the s-tim genotype. A candidate locus is the insulin-regulated PI3 kinase gene, which determines diapause levels in two D. melanogaster populations from North America (24). If diapause contributes to the enhanced adaptive value of ls-tim, it is difficult to envisage a balancing scenario where ls-tim would be highly favored in a small region of southern Italy but less favored farther north or south. We propose that an origin of the derived ls-tim allele in southern Europe, followed by its subsequent spread by directional selection, provides—counterintuitively—a more compelling model for understanding the elevated frequencies of ls-tim in this geographical region.

Supporting Online Material

Materials and Methods

Tables S1 and S2

Figs. S1 to S3


References and Notes

View Abstract

Navigate This Article