ABSTRACT
Finding the specific nucleotides that underlie adaptive variation is a major goal in evolutionary biology, but polygenic traits pose a challenge because the complex genotype-phenotype relationship can obscure the effects of individual alleles. However, natural selection working in large wild populations can shift allele frequencies and indicate functional regions of the genome. Previously, we showed that the two most common alleles of a complex amino acid insertion-deletion polymorphism in the Drosophila insulin receptor show independent, parallel clines in frequency across the North American and Australian continents. Here, we report that the cline is stable over at least a five-year period and that the polymorphism also demonstrates temporal shifts in allele frequency concurrent with seasonal change. We tested the alleles for effects on levels of insulin signaling, fecundity, development time, body size, stress tolerance, and lifespan. We find that the alleles are associated with predictable differences in these traits, consistent with patterns of Drosophila life history variation across geography that likely reflect adaptation to the heterogeneous climatic environment. These results implicate insulin signaling as a major mediator of life history adaptation in Drosophila, and suggest that life history tradeoffs can be explained by extensive pleiotropy at a single locus.
INTRODUCTION
Many organisms display intraspecific variation in life history traits, including differences in reproductive timing and allocation, development, body size, stress tolerance, and lifespan. Alternative life history strategies have long been hypothesized to represent adaptive responses to variable selection (Stearns 1992), but examples with known causal nucleotides are rare (Flatt & Heyland 2011). However, sequence polymorphisms have been identified recently for flowering time in Arabidopsis thaliana (Méndez-Vigo et al. 2013) and adaptation to freshwater environments by marine stickleback fish, which includes changes in life history as well as in physiology and morphology (Jones et al. 2012). One reason that identification of molecular targets of selection is valuable is because it can illuminate mechanisms of multi-trait correlation. For example, the gene Catsup in Drosophila melanogaster affects longevity, locomotor behavior, and bristle number, but individual polymorphisms within the locus act on the traits independently and do not show pleiotropic effects (Carbone et al. 2006). Likewise, coat color in deer mice is a multi-phenotype trait associated with the Agouti locus, but mutations within the gene appear to have been targeted independently by selection with minimal pleiotropy (Linnen et al. 2013). These examples suggest that fitness-related traits, even those affected by the same gene, may be functionally independent such that natural populations might achieve any genetic and phenotypic combination. Alternatively, life history traits routinely exhibit strong genetic correlations, including negative associations between lifespan and reproduction and positive associations between lifespan and stress resistance (Reznick 1985; Stearns 1991; Partridge et al. 2005; Vermeulen & Loeschcke 2006; Harshman & Zera 2007; Toivonen & Partridge 2009). These associations mirror pleiotropic effects of lab-derived mutations and can be difficult to break in artificial selection experiments (Leroi et al. 2005; Patridge et al. 1999; Anderson et al. 2011; but see Khazaeli & Curtsinger 2010; Khazaeli & Curtsinger 2013). Consequently, pleiotropy, and specifically antagonistic pleiotropy, in which a genetic element encodes both positive- and negative-fitness phenotypes, remains an important component in the discussion of how life histories evolve (Williams 1957; Flatt & Heyland 2011).
D. melanogaster are distributed across environments that range from temperate to tropical, and exhibit variation in life history that appears adaptive. At high latitudes, populations exhibit higher incidence of reproductive diapause, larger body size, higher cold stress tolerance, and longer lifespan, but lower fecundity, relative to low latitude populations (Capy et al. 1993; Mitrovski & Hoffmann 2001; de Jong & Bochdanovits 2003; Schmidt et al. 2005a; Trotta et al. 2006). There is substantial genetic variance for these traits and pervasive genetic correlations among them, indicating that selection in the local environment may act on some phenotypes but drive expression of others through tradeoffs (David 1975; Anderson et al. 2003; de Jong & Bochdanovits 2003; Schmidt et al. 2005b; Rako et al. 2007; Schmidt & Paaby 2008). Tolerance to environmental stress may be especially important, and the ability to resist desiccation, starvation and temperature stress correlates with climatic environment (Hoffmann & Harshman 1999; Hoffmann et al. 2001; Hoffmann et al. 2005; Hoffmann et al. 2007). This framework suggests a hypothetical selection regime: at high latitude, cold winters impose seasonal stress and favor genotypes that confer stress tolerance and overwintering ability; correlated traits such as larger body size, longer lifespan, slower development and lower fecundity may evolve as co-adapted responses to the same selection regime or by indirect selection via pleiotropy (Paaby & Schmidt 2009).
Few loci have been found to explain the observed genetic variance for D. melanogaster life history (De Luca et al. 2003; Carbone et al. 2006; Paaby & Schmidt 2008; Schmidt et al. 2008; Bergland et al. 2012; Remolina et al. 2012, Sgrò et al. 2013). Genes in the Insulin/Insulin-like Signaling (IIS) pathway are good candidates because experimental manipulation of IIS mirrors the life history tradeoffs observed in natural populations (Partridge & Gems 2002; Tatar et al. 2003; Giannakou & Partridge 2007; Grönke 2010). We previously evaluated sequence variation at the Insulin-like Receptor (InR) in wild populations of D. melanogaster and observed striking reciprocal clines in the northern and southern hemispheres for the common alleles of a complex amino acid indel polymorphism (Paaby et al. 2010). The polymorphism disrupts a region of glutamine-histidine repeats in the first exon (Guirao-Rico & Aguadé 2009). In both North America and Australia, the most common allele is at low frequency in tropical and subtropical populations and increases in frequency with latitude, showing highest frequency in temperate populations; the second most common allele exhibits the inverse cline. These patterns appear to be responses to similar but independent selection pressures between the continents: fly populations in North America and Australia were founded at different times and from different source populations (Bock & Parsons 1981; David & Capy 1988), and nucleotide variants on either side of the polymorphism show neutral patterns across geography.
Initial tests showed that the two common alleles, which differ in length by two amino acids and are herein designated InRshort and InRlong, demonstrated significant effects on fitness traits. InRshort, the allele common at high latitudes, was associated with greater stress tolerance; InRlong, which is common at low latitudes, was associated with greater fecundity (Paaby et al. 2010). Here we show that the allele frequency cline in North America persists five years later and that temporal changes in allele frequency within a single population correlate with fluctuations in the seasonal environment, which mimic the temperature and resource availability differences associated with latitudinal climate. We also find that the InR locus in general exhibits elevated patterns of clinality and seasonality relative to the rest of the genome. We more thoroughly tested the functional effects of the complex polymorphism by measuring levels of IIS and a spectrum of life history phenotypes in InRshort and InRlong lines with randomized genetic backgrounds. We find that the alleles associate predictably with the tested traits, suggesting that the complex amino acid indel polymorphism at InR is an important target of contemporaneous selection in wild D. melanogaster populations, is highly pleiotropic, and contributes to observed tradeoffs in D. melanogaster life history.
METHODS
Allele frequency estimates. As described in Bergland et al. (In Review), between 50-200 isofemale lines were established from six populations spanning Florida to Maine along the North American east coast (latitudes 25.5°N, 30.9°N, 33.0°N, 35.5°N, 39.9°N, 44.1°N). The population from Linvilla Orchard in Media, PA (39.9°N, 74.4°W) was additionally sampled in the spring and fall in 2009, 2010, and 2011. Within five generations a single male from each line was pooled into population samples, which were sequenced on an Illumina HiSeq2000 using 100-bp paired-end reads. Reads were aligned to the D. melanogaster reference genome v5.39 using bwa v0.5.9-r16 (Li & Durbin 2009). The indel polymorphism at InR was called using UnifiedGenotyper from GATK v3.1-1 (McKenna et al. 2010), which identified multiple discrete polymorphisms within the complex indel haplotypes that we previously characterized (Paaby et al. 2010). That characterization identified many haplotypes of different lengths, two of which were common (herein InRshort and InRlong); the reference genome to which the new pooled samples were mapped has a haplotype of intermediate length (previously referred to as InR251). Thus, the complex haplotypes, previously identified by their lengths, were decomposed into sets of shared or distinct alleles at discrete polymorphic sites.
We used our prior Sanger sequence data (see GenBank accessions GQ927177–GQ927244) to assign alleles of the discrete polymorphisms to the InRshort and InRlong allele classes. Here we report those classifications (polymorphisms are denoted by their position in the D. melanogaster reference genome v5.48). Three discrete polymorphisms in the pooled sequencing data discriminate unambiguously between InRshort and InRlong: 17405631-7, 17405651-4 and 17405784. In each case, the reference genome allele matches InRshort. Only one of the discrete polymorphisms, SNP 17405614, was variable within an allele class; the alternate and reference alleles are both associated with InRshort such that the association between the discrete allele identities and the larger haplotypes in which they are embedded (InRshort and InRlong) are approaching equilibrium (R2=0.45). SNP 17405619 segregates at low frequency and was invariant in our Sanger sequencing data such that the InRshort and InRlong haplotypes we observed shared the same discrete allele. Both the alternate allele of SNP 17405631 and the alternate allele of 17405634-7 match both InRshort and InRlong. For polymorphism 17405637, two alternate alleles were called; the 6-nt insertion matches InRlong, but the 3-nt insertion is confounded with the alternate allele of 17405634-7 (which is called consistently). Thus, we ignored the 3-nt insertion allele at 17405637.
Allele frequencies of SNPs across the entire InR locus and across other sites on chromosome 3R were called with CRISP (Bansal 2010) as described in Bergland et al. (In Review). To test whether the lnR locus (3R:17385970-17455043) demonstrates elevated clinality or seasonality relative to the rest of the genome, we calculated the proportion of SNPs within InR that were identified to be clinal (with associated clinal FDR < 0.01) or to vary repeatedly among seasons over the course of three years (with associated seasonal FDR < 0.5) using estimates of clinality and seasonality reported in Bergland et al (In Review). We note that the high FDR for seasonal SNPs was chosen to increase power for detecting fluctuations at InR outside of the InRlong/short polymorphism, which is not included in this scan or in the dataset evaluated by Bergland et al. (In Review). We contrasted the proportion of clinal and seasonal SNPs within the InR locus to 1000 random chromosomal regions matched for length, chromosome, and inversion status for In(3R)Payne.
Genotyping. To genotype flies for the experimental assays, the identity of the complex InR indel was determined by measuring its amplified fragment length following polymerase chain reaction (PCR) with a fluorescent-tagged primer on an Applied Biosystems 3100 capillary sequencer. PCR conditions, including primer sequences, are described in Paaby et al. (2010). That paper reported the distribution of alleles across geography and referred to the alleles by their PCR fragment lengths, which we erroneously stated as 248 bases for the high-latitude allele and 254 bases for the low-latitude allele. The actual fragment lengths using the cited primers are 5 bases longer. To avoid confusion, in this paper we refer to these two alleles as InRshort (previously InR248) and InRlong (previously InR254). It should be noted that although fragment-length genotyping does not discriminate between the SNP alleles we characterized in or near the complex indel polymorphism (described above), only one of those SNPs (17405614) appears to be independent of the InRshort and InRlong allele classes. All evidence to date indicates that fragment-length genotyping unambiguously discriminates between consistent, but complex, haplotype classes (see Figure 1A-B).
Fly stocks. To determine whether InRshort and InRlong alleles have different effects on phenotype, we generated stocks with InRshort or InRlong on the third chromosome, and for which the X and second chromosomes were replaced so that they were isogenic within and across lines. The genome regions surrounding InR on the third chromosome were randomized, from lines originally genotyped as InRshort and InRlong, across lines. To generate the stocks, we selected InRshort and InRlong parental strains, neither of which carried the In(3R)Payne inversion, and for which the X and second chromosome were already replaced (for the X chromosome, stock 2475; for the second chromosome, stock 6326 from the Bloomington Stock Center; originally described in Paaby et al. 2010). The parental strains were crossed and the offspring permitted to recombine for four generations, so that genetically variable regions from the parental third chromosomes were distributed as randomized blocks across lines. Fourteen isogenic stocks were subsequently established per allele by homozygosing the third chromosome via balancer extraction and genotyping the InR indel polymorphism. This entire scheme was performed twice, using lines derived from two independent populations: Mount Sinai, NY (40.95°N, 72.84°W) and Bowdoinham, ME (44.01°N latitude, 69.90°W longitude).
InRshort and InRlong alleles were evaluated as homozygotes and heterozygotes, as well as with InR mutant and wild-type alleles derived from a laboratory strain. The InR mutant was the hypomorphic InRp5545 allele generated by P-element insertion in exon 1 (Tatar et al. 2001) balanced over the TM3 chromosome (stock 11661 from the Bloomington Stock Center, ry506 P{PZ}InRp5545/TM3, ryRK Sb1 Ser1). The TM3 balancer contains a wild-type InR allele (herein InRTM3) that is intermediate in length between InRshort and InRlong: it is three nucleotides longer than InRshort, corresponding to an additional histidine, and three nucleotides shorter than InRlong, which contains an additional glutamine.
Flies for assays. Once the stocks were established, we used three methods to generate flies for the phenotype assays (derived from one or both of the independent populations). The first two tested the effects of InRshort and InRlong in a wild-type background; the third tested InRshort and InRlong effects while paired with either the InR hypomorphic allele or a wild-type allele in the TM3 balancer chromosome. In the first method, we exerted nominal control over the background variation in the stocks by crossing 14 InRshort and 14 InRlong stocks in a round-robin design to generate a total of 42 lines from which we collected flies for phenotyping (14 each of InRshort/InRshort, InRlong/InRlongand InRshort/InRlong). The crosses were conducted in standard media vials with 7 virgin females (2-5 days old) and 4 young males (3-8 days old). This design produced non-independent replication within each genotypic class, and because test stocks were never mated to themselves, also ensured that no individuals were strictly isogenic at the third chromosome. In the second method, we combined stocks (minimum of five) carrying the same allele (InRshort or InRlong) but otherwise randomized variation on the third chromosome in bottles, permitting the stocks to continue recombining freely. To rear flies for the assays, 40 virgin females (2-5 days old) and 20 young males (3-8 days old) were collected and mated in fresh bottles, either within or across genotype class, to generate the two homozygous and the heterozygous genotypes. Unless otherwise specified, the phenotype assays under this method were replicated via 10 mating bottles for each genotypic class. In the third method, we crossed 12 InRshort stocks and 12 InRlong stocks to the hypomorphic InRp5545/TM3 stock (in vials as in method one). From these 24 lines, we collected progeny carrying both the InR hypomorphic allele and the wild-type allele in the balancer chromosome for phenotyping.
Phenotype assays. For all assays, flies were reared and assays were performed on standard cornmeal-molasses media at room temperature and subject to ambient light cycles. Larval density was kept low to limit overcrowding. We tested InRshort and InRlong alleles derived from Mount Sinai in all assays; we tested InRshort and InRlong alleles from Bowdoinham in a subset of the assays. All InRshort and InRlong alleles tested in the presence of the InR hypomorph (InRp5545) or wild-type balancer (InRTM3) were from Mount Sinai. All statistics were performed using JMPv7 (SAS Institute, Cary, NC). Within each analysis, we used planned comparisons to test for significant differences between InRshort and InRlong genotypes.
Lifespan and lifetime fecundity. Bottle cages were populated with 30 virgin females and 30 males, collected over 24 hours, from lines generated by method one from the Mount Sinai population. A total of 42 bottles, comprising 14 of each genotypic class, were maintained by changing standard media egg-laying plates every day (days 1-26) or every other day (days 28 onward) until all flies were dead. Bottles were inverted, so that dead flies collected on the media plates. Dead flies and eggs laid were scored at every plate change. Lifespan was analyzed by a proportional hazards model, and cumulative fecundity was analyzed by ANOVA.
Early fecundity. Virgin females and males were collected over 6 hours, separated into single-sex vials of 5 females or 3 males each, and aged for 3 days on standard media with topical yeast. Flies were transferred in mixed-sex batches (five females and three males) to media-free vials and starved for 4 hours, then transferred again to fresh media vials and permitted to lay eggs. The number of eggs laid was recorded after 12 hours and analyzed by ANOVA. Flies were generated by method two from both populations, and also with the InR hypomorph via method three.
Development time. Young (3-6 days old) males and females (5 flies of each sex) were transferred from standard media vials to media-free vials and starved for four hours, then transferred again to fresh media vials and permitted to lay eggs, at densities of ≤30 eggs per vial, for up to four hours before being removed. Flies derived from Mount Sinai were generated via method one; flies from Bowdoinham via method two, and with the hypomorph via method three. The number of adults per vial was recorded daily at 8 AM, 12 PM and 8 PM for all emerging adults, and eclosion time was analyzed by ANOVA.
Body weight, lipid content and body size. Using methods and replication schemes identical to those described for development time, eggs were reared and permitted to eclose as adults. After eclosion, adults were transferred to fresh media vials containing topical yeast and aged for 24 hours, then frozen at −80°C. To measure dry weight, frozen flies were dried for 24 hours at 55°C and weighed to 0.0001 g in single-sex batches of 7 individuals on a Sartorius microbalance (1-2 batches per original vial cross). The batches were transferred to 1.5 mL tubes and lipids were extracted by adding 400 uL diethyl ether and gently rocking the tubes for 24 hours. Flies were dried overnight at 55°C and batches re-weighed; lipid content was computed by subtraction. To estimate body size, left wings were removed from frozen flies and affixed to an index card using clear tape. Pictures of wings from a DC 300 camera (Leica) on a dissecting microscope were measured for L2 and L3 vein length using landmarks 6, 12 and 13 (Gidaszewski et al. 2009) on 3-8 males and 3-8 females from each vial cross, using tpsDigv2.12 (Rohlf 2008). By the same method thorax width was measured for the Bowdoinham and hypomorph crosses. Measurements were analyzed by ANOVA.
Chill coma recovery. Freshly eclosed flies were collected over 24 hours from flies generated by method one from the Mount Sinai population, and sorted into standard media vials in batches of 5 males and 5 females each. After 3 days, males were sorted into fresh vials and aged another day. To induce chill coma, vials were covered in ice and placed at 4°C for three hours, then restored to room temperature. Time to recovery (transition to the upright position) was recorded using a video camera and analyzed by ANOVA.
Cold shock and starvation. Freshly eclosed flies were collected over 24 hours from flies generated by method one from the Mount Sinai population and method three with the hypomorph, sorted into standard media vials in batches of 5 males and 5 females, and aged for 3 days. The 42 vials generated by method one were replicated twice. To induce cold shock, vials were inverted and exposed to −20°C (air temperature) for 25 minutes. After 24 hours, individuals were scored as dead or alive. To measure starvation resistance, flies were transferred into media-free vials containing a cotton ball saturated with 2 mL of water. At 51 hours, males were scored as dead or alive; at 64 hours, females were scored as dead or alive. Both assays were analyzed by nominal logistic regression modeling the log odds of mortality/survivorship.
Heat shock. Freshly eclosed flies were collected over 24 hours and sorted into mixed-sex standard media vials of 5 males and 5 females each. Flies were generated by method two from the Mount Sinai population (15-20 replicate vials per genotypic class) and by method three. Flies were aged for 4 days and then transferred to fresh media vials. Vials were inverted, placed at 25°C, and warmed to 37°C; after reaching 37°C (approximately 20 minutes), flies were kept at 37°C for 25 minutes and then removed to room temperature to recover. After 24 hours, flies were scored as dead or alive and analyzed by nominal logistic regression modeling the log odds of mortality/survivorship.
Quantitative PCR. To test whether InR alleles affect IIS, we used quantitative polymerase chain reaction (qPCR) to determine the relative abundance of 7 transcriptional targets of dFOXO, a central transcription factor within the IIS pathway (Jünger et al. 2003; Puig et al. 2003; Wang et al. 2005; Casas-Tinto et al. 2007; Flatt et al. 2008; Vihervaara & Puig 2008; Matilla et al. 2009). Total RNA was prepared from flies generated by method two from flies derived from Mount Sinai. Thirty virgin females from each of the three genotypic classes were collected over 24 hours, aged on standard media, and snap frozen. RNA was extracted using RNeasy (Qiagen) and reverse transcribed using iScript cDNA synthesis kit (Bio-Rad), following the manufacturer's protocol. Relative abundance of transcript levels was determined using an ABI 7500 Fast Real-Time PCR machine and SYBR Green PCR Master Mix (Applied Biosystems) by the DDCT relative quantitation method. Four technical replicates were used for each sample and relative abundance was normalized by using GAPDH2 as an endogenous control. Other work has demonstrated that GAPDH2 is an appropriate control for measuring relative levels of IIS (Hwangbo et al. 2004, Flatt et al. 2008). Primer sequences for the qPCR reaction are: GAPDH2 forward, GCGGTAGAATGGGGTGAGAC; GAPDH2 reverse, TGAAGAGCGAAAACAGTAGC; 4E-BP forward, GAAGGTTGTCATCTCGGATC; 4E-BP reverse, ATGAAAGCCCGCTCGTAGA; l(2)efl forward, AGGGACGATGTGACCGTGTC; l(2)efl reverse, CGAAGCAGACGCGTTTATCC; ac76e forward, CAGGATGAATGACGCCCTTTCGG; ac76e reverse, ATGGACACAACACATGCCAGCAGC; dLip4 forward, GATAGCAATGTGCGGTTGGA; dLip4 reverse, TCATCCGTCTCCAAGGTGTG; InR T1 forward, CACAAGCTGGAAAGA AAGTGC; InR T1 reverse, CAAACACGTTTCGATAATATTTTTCT; InR T2 forward, GCCTCGCACTTTGCTTATGT; InR T2 reverse, AAAAACAACGACAGCGACAA; InR T3 forward, TTACGCCACTGCATTCGTTC; InR T3 reverse, ATGGCCTCTCTCTCCGTCTC.
RESULTS
Clinal and seasonal estimates of allele frequency. To test whether InR allele frequencies change in association with the climatic environment, we identified multiple discrete polymorphisms within and near the complex indel polymorphism from pooled population samples. We identified all variants (SNPs or small indels) that comprise the complex indel itself and three SNPs −11, −16 and +130 nucleotides from the indel that segregate at frequencies above 5% (Figure 1A-B). Four of these discrete polymorphisms show clear seasonality, with repeating fluctuations in allele frequency of approximately 20% between spring and fall time-points over three consecutive years (Figure 1C). The other four discrete polymorphisms are suggestive of seasonality: they demonstrate either comparable changes in amplitude but a deviation in the frequency pattern at a single time-point, or a repetitive pattern at low amplitude. The discrete polymorphisms also show clines in allele frequency across latitude (Figure 1D). We were able to unambiguously assign allele identities for three of the discrete polymorphisms to the InRshort or InRlong allele classes (Table S1), which exhibited strong allele frequency clines five years previously (Paaby et al. 2010). In each of these cases, changes in allele frequency across latitude replicated our earlier observations, and seasonal fluctuations in frequency for these discrete polymorphisms support the hypothesis that alleles favored at high latitude are favored during the winter.
Compared to the rest of the genome, the InR locus exhibits elevated patterns of allele frequency change across latitude and seasonal time. The region surrounding InR contains significantly more clinal polymorphisms relative to other regions of chromosome 3R matched for size and inversion status (42% vs. 20%, respectively; Pr(%InR > %control) = 0.986). Relative to matched control regions, the InR locus also contains slightly more seasonal polymorphisms (6% vs. 4.7%; Pr(%InR > %control) = 0.8). Previous examination of linkage disequilibrium across InR (Paaby et al. 2010) indicates that the complex indel polymorphism is largely independent of variation elsewhere in the gene, which implies that many sites within InR may be targets of spatially and temporally varying selection.
Allele effects. We evaluated the effects of InR alleles on levels of IIS and multiple phenotypes and found that individuals carrying the InRshort allele, which is which is prevalent in high-latitude, temperate environments, exhibited lower levels of IIS, better survived cold temperature stresses and starvation, and showed sex-specific evidence for longer lifespan. Flies carrying the InRlong allele, which is common in low latitude, warm environments, were associated with increased signaling and exhibited higher rates of fecundity, larger body mass, faster development time and better survival of heat shock. These phenotypic differences are consistent with experimental reduction of IIS as well as life-history tradeoffs observed in natural populations, and suggest that a single locus can act pleiotropically on a suite of fitness-related traits. We replicated some of the assays by using InRshort and InRlong alleles derived from two independent populations (Bowdoinham and Mount Sinai); the results are reported separately.
Levels of insulin signaling. To test whether the observed phenotypic differences between InRshort and InRlong are mediated by different levels of IIS, we used qPCR to measure relative abundance of seven transcriptional targets of dFOXO, a central transcription factor in the IIS pathway that is repressed by InR activity. Decreased InR activity increases target abundance, and transcripts for five of these (4E-BP, l(2)efl, ac76e, dLip4 and InR T1) were highest in InRshort, intermediate in the heterozygote, and lowest in InRlong; all but 4E-BP showed statistical significance (p<0.05) between InRshort and InRlong (Figure 2). This pattern is robust considering that the InRshort genotype exhibited at most twofold higher levels relative to InRlong: a difference that may be biologically meaningful, and realistic for wild-type genotypes, but close to the limit of detection by qPCR methods. Two other targets, transcripts of InR itself, show no difference in abundance level between the homozygote genotypes, but significantly higher abundance in the heterozygote.
Fecundity. Surprisingly, the homozygous InR genotypes showed no significant differences in total fecundity: a fly carrying the InRlong allele laid on average only 5.3 more eggs over her lifetime than a fly carrying InRshort (p=0.6949, Table 1). This was unexpected because we routinely observed substantially greater population numbers in our InRlong bottle cultures, and previous results showed that the InRlong genotype is more fecund (Paaby et al. 2010). Examination of the eggs laid per day, however, revealed that the InRlong genotype lays more eggs early in life (Figure 3A). We performed another assay to explicitly measure early fecundity, or how many eggs young mated females laid upon first access to fresh food. Here, we found that the InRlong genotype is nearly six times more fecund than InRshort in the first 12 hours of egg-laying (p<0.0001, Table 1, Figure 3B). This result was again observed in a replicate assay using flies derived from the second population (p=0.0005) (Table 1, Figure 3C).
Development time. From egg to emerging adult, flies carrying the InRlong allele developed faster than flies carrying InRshort. This result was replicated in flies derived from two populations: in the first, males and females carrying InRlong developed an average of 6.0 hours faster than those carrying InRshort (p<0.0001); in the second population, development time was not significantly different among male genotypes, but InRlong females emerged an average of 18.2 hours ahead of InRshort females (p=0.0029) (Table 1, Figure 4).
Body weight and size. Because reduction in IIS reduces body size (Clancy et al. 2001; Tatar et al. 2001), we predicted that individuals carrying the InRlong allele would be larger than individuals carrying InRshort. Indeed, the average dry body weights of InRlong flies were heavier than InRshort flies. This was true for alleles tested from both populations: InRlong females and males derived from Bowdoinham were 26.1% and 18.6% heavier, respectively (p<0.0001 for both), while InRlong females and males derived from Mount Sinai were 15.5% and 10.9% heavier (p=0.0002 and p=0.0004) (Table 1, Figures 5, S1). The higher mass associated with the InRlong genotype is partially explained by lipid content. For both populations, the
InRlong allele was associated with greater lipid mass than InRshort (Figures 5, S1); however, only alleles derived from Bowdoinham demonstrated a statistically significant difference when non-lipid mass was included as a covariate (p=0.0066 for females, p<0.0001 for males) (Table 1). Wings of InRlong flies were also larger than those for InRshort, at least in the Bowdoinham population. In that comparison, the average L2 wing vein length was 0.24 mm longer in InRlong females (p<0.0001) and 0.12 mm longer in InRlong males (p<0.0001) (Table 1, Figure 5). In flies derived from Mount Sinai, InRshort females had the larger wings, but only by 0.02 mm (p=0.0029), and InRshort and InRlong males were not different from each other (p=0.5999) (Table 1, Figure S1).
An association between large size and InRlong, the low-latitude allele, is predictable in terms of insulin signaling, but unpredictable given the observation that Drosophila are larger at higher latitude (Huey et al. 2000; De Jong & Bochdanovits 2003). Others have shown that proportionally larger wings relative to thorax size are associated with better flying in Drosophila (Hoffmann et al. 2007) and that the wing:thorax ratio increases with latitude (Karan et al. 1998), possibly because muscles in colder temperatures generate less power and proportionally larger wings can compensate for this (Gilchrist & Huey 2004). We measured thorax width to see if bigger wing:thorax ratios were associated with the high-latitude InRshort allele, but they were not. InRshort flies averaged smaller wing:thorax ratios relative to InRlong flies (p<0.0001 for females; p=0.0218 for males) (Table 1, Figure 5). For all samples in which we measured the L2 wing vein, we also measured the L3 vein, and observed qualitatively identical results (data not shown).
Stress tolerance. Reduction in IIS confers increased resistance to stress (Giannakou & Partridge 2007), and results for three of our four stress assays accord with these findings and our hypothesis that the high-latitude InRshort allele mediates increased stress tolerance via reduced signaling (Figure 6). Females carrying the InRshort allele recovered from chill coma on average 89 seconds faster than females carrying InRlong; InRshort males recovered 123 seconds faster (although this comparison was only statistically significant for males; p=0.1448 and p<0.0001, respectively) (Table 1, Figure 6). Likewise, InRshort females and males exhibited 3.8 and 2.4 times the odds of surviving cold shock and 1.3 and 1.8 times the odds of surviving starvation, respectively, compared to InRlong flies. However, the InRshort allele did not correlate with higher tolerance of heat shock: instead, InRshort flies showed 15.3 times the odds of dying in the assay compared to InRlong flies (Table 2, Figure 6). All four of these outcomes are consistent with presumed selection pressures in natural populations, as flies in high latitude, temperate environments, where the InRshort allele is common, likely face cold temperatures and scarce food availability during the winter, while flies in low latitude, semi-tropical environments, where InRlong is common, should face thermal stress from higher temperatures. However, if levels of IIS mediate these stress phenotypes, the heat shock outcome provides a unique example of correlation reversal between signaling and stress. Nevertheless, the phenotypic correlations across these traits, and the correlations between the phenotypes and latitude, are patterns consistent with other examples of reciprocal tolerance for environmentally imposed thermal stresses (Hoffmann et al. 2001; Hoffmann et al. 2005).
Lifespan. Insulin signaling has been established as a major regulator of longevity in many animals, including Drosophila: hypomorphic alleles of InR increase lifespan relative to wild-type controls (Tatar et al. 2001), as do other manipulations that reduce IIS (Clancy et al. 2001; Hwangbo et al. 2004; Grönke et al. 2010). Consequently, we hypothesized that InRshort might increase lifespan relative to InRlong. Flies carrying the InRshort and InRlong alleles showed little difference in rates of aging over the majority of lifespan, though InRshort males exhibited reduced mortality late in life (Figure 7). The lifetime survivorship of InRshort males was significantly longer than that of InRlong males, with a mortality risk ratio of 1.3 for InRlong over InRshort (p=0.0095). The female mortality risk ratio for InRlong over InRshort was only 1.01, with no significant difference in longevity between these genotypes (p=0.8909). Interestingly, the heterozygote of both sexes lived significantly longer (p<0.0001) than either of the homozygote genotypes (Table 2). We hypothesize that the heterozygote longevity is a result of heterosis across the extended InR-embedded locus, a consequence of the fact that the homozygotes carried identical alleles with haplotypes extending beyond InR. Any general increase in fitness associated with greater heterozygosity was probably not mediated by IIS, as the heterozygotes were associated with intermediate levels of gene expression for the majority of IIS targets.
Allele effects in an InR-background. To examine how the InRshort and InRlong alleles performed in a genetic background with reduced InR activity, we crossed them into a stock with a hypomorphic allele (InRp5545) over a wild-type copy of InR on the TM3 balancer chromosome (InRTM3), generating the four possible genotypic combinations. Previous work with this hypomorphic allele showed that a single copy of InRp5545 reduces InR kinase activity by approximately half, and not as much as a double mutant (Tatar et al. 2001). The genotypic combinations allowed us to compare single copies of InRshort to InRlong, and wild-type flies to flies carrying a single copy of the hypomorphic InR mutant allele.
Consistent with the expectation that InR function increases fecundity, flies carrying the hypomorphic allele laid fewer eggs in the first 12 hours after mating than wild-type flies (p=0.0109). When paired with the hypomorphic allele, InRlong females laid on average 38% more eggs than InRshort females, which is consistent with the direct allele tests but was not statistically significant in this assay, and there was no difference in egg-laying between InRshort and InRlong in the wild-type background (Table S2, Figure 8). As in the direct allele tests on Bowdoinham flies, here patterns of development time were sex-specific. Females carrying the InR hypomorph developed 9.1 hours slower than InR wild-type flies (p<0.0001), but females carrying InRshort and InRlong were not different from each other. Males with InRshort developed 1.3 hours slower than InRlong males (p=0.0064), but males with or without the hypomorphic allele were not different from each other (Table S2, Figure 8). These patterns are consistent with the expected association of high InR activity and faster development, and the difference between the wild-type InRshort and InRlong alleles in males is much subtler than the difference between the wild-type and mutant alleles in females.
Dry weight of wild-type flies was greater than those with the InR hypomorph (p=0.0061 for females; p=0.0763 for males), consistent with the prediction that increased InR activity results in larger mass, but there were no differences between InRshort and InRlong (Table S2, Figure 9). However, single copies of InRshort and InRlong were weakly associated with lesser and greater lipid levels, as we saw in the direct allele tests (p=0.0729 for females; p=0.0624 for males), but there were no differences between InR hypomorphic and wild-type alleles (Table S2, Figure 9). The observations of InRshort with lower lipid level suggest that minor reduction in InR activity can reduce lipids even though the more general relationship between IIS and lipids is a negative one. This finding is in fact consistent with earlier characterization of InR mutant alleles, in which single copies of hypomorphic alleles were associated with reduced lipids (though not in our data), and double mutants with increased lipid content, relative to the wild-type (Tatar et al. 2001). Female wing sizes were not different from each other across any of the genotypic comparisons, but wings of InRlong males were bigger than wings of InRshort males (p=0.0179), as we saw in the direct allele tests (Table S2, Figure 10). The InRshort and InRlong alleles had no effect on the wing:thorax ratio, but both sexes had bigger ratios in the mutant background relative to the wild-type background (p<0.0001 for females; p=0.0001 for males) (Table S2, Figure 10). The correlation between the hypomorph and larger ratio suggests that decreased InR activity may increase the wing:thorax size relationship, which is what we observed in the direct allele tests as well.
DISCUSSION
Our results indicate that the complex indel polymorphism in the first exon of InR is likely a direct target of natural selection in wild populations of D. melanogaster. Allele frequencies exhibit persistent latitudinal clines, change rapidly and cyclically over seasonal timescales, and the two common alleles are pleiotropically associated with multiple traits that correspond to expected pressures in distinct climatic environments. The alleles are also associated with predicted up- and down-regulation of IIS, strongly suggesting that they mediate signaling behavior to effect differential trait expression. The fact that the alleles encode different amino acid sequences suggests that this occurs through changes in protein function. Overall our findings provide an example of life history tradeoffs influenced by a complex, strongly pleiotropic polymorphism.
While the phenotypes associated with InRshort and InRlong were specific and predictive, our assays did not precisely test the effects of the glutamine and histidine changes that account for the length differences in the complex indel polymorphism. This is for two reasons. First, the InRshort and InRlong alleles themselves contain several other SNPs, each of which exhibits nonrandom patterns of allele frequency across latitude and seasonal time (Figure 1). Only one of these (a non-synonymous SNP at position 17405614) has the potential to confound our functional results, however, as the other SNPs did not vary between our InRshort and InRlong lines. Allele frequencies at 17405614 exhibited substantial seasonal and clinal patterns, though not as strongly as several of the discrete indel variants that comprise the complex indel polymorphism (Figure 1). Second, genome regions flanking the complex indel polymorphsim were not fully randomized due to the limits of experimentally imposed recombination. Consequently, in our test lines, the InRshort and InRlong alleles were embedded in haplotypes that extended beyond the InR locus. Based on prior estimates of linkage disequilibrium, however, flanking variants should be unlinked to the focal indel polymorphism (Paaby et al. 2010). Although these other variants may be targets of selection—a conclusion supported by our observation that the entire InR locus exhibits elevated clinality and seasonality relative to the rest of the genome—the fact that the indel polymorphism exhibits such strong changes in allele frequency across climatic environments provides complementary evidence for allele functionality. This accords with other reports of high recombination and rapid decay of linkage disequilibrium (Mackay et al. 2012), and evidence that selection precisely targets functional loci, including InR (Fabian et al. 2012), in natural populations of D. melanogaster.
Two of the discrete polymorphisms associated with the complex indel polymorphism, 17405631 and 17405634-7, do not discriminate between InRshort and InRlong but rather discriminate between the published reference genome genotype (which is of intermediate amino acid length) and InRshort or InRlong. Both of these discrete polymorphisms show strong seasonal patterns of allele frequency, which suggests that the intermediate allele may be a target of selection as well. The discrete 17405631 and 17405634-7 alleles segregate at substantial frequencies in all populations, and in both North America and Australia the intermediate-length haplotype is the third most common allele class (Paaby et al. 2010). In our previous survey, the intermediate allele showed no cline in North America and a weak cline in Australia. Here, the discrete 17405631 and 17405634-7 alleles again show little evidence of clinality in North America (Figure 1D, Table S1). These observations support the conclusion that seasonal and geographical environments may impose different evolutionary forces, including aspects of demography and selection, even in the face of generally concordant responses to shared climatic selection pressure (Bergland et al. In Review).
Since flies derived from high latitudes live longer than those from low latitudes (Schmidt et al. 2005a), and reduction in IIS extends lifespan (Clancy et al. 2001; Tatar et al. 2001; Hwangbo et al. 2004; Grönke et al. 2010), we predicted that the InRshort allele might be associated with longer lifespan than the InRlong allele. InRshort males did live significantly longer than InRlong males, but the magnitude of difference was small compared to other traits and there were no differences between the alleles in females. The comparatively weak longevity effect may reflect decoupling between lifespan and reproduction that has been observed elsewhere (Khazaeli & Curtsinger 2010)—yet it may also represent true functional variation. Laboratory-derived InR mutations have shown a dramatic effect on lifespan, but only the heterozygous combination of two hypomorphic alleles, of all genotypes tested, significantly extended lifespan (Tatar et al. 2001). This suggests that IIS-mediated traits may be determined by precise titration of IIS levels.
In addition to longer male lifespan, the InRshort allele was also associated with better cold and starvation tolerance and lower fecundity in females, a classic example of antagonistic pleiotropy (Williams 1957). The differences in fecundity were limited to early life, and represent variation in egg-laying rate, because total lifetime fecundity was not significantly different between the alleles. Modest differences in egg-laying in early life could have massive fitness consequences in the wild, particularly on patchy substrates with larval competition, yet the competitive advantage may be undetectable in the lab if fecundity is measured over more than one day. Consequently, we emphasize that estimates of reproduction and other quantitative fitness traits should be examined carefully, particularly in the context of decoupling correlated traits. We also note that although the high-latitude InRshort allele is associated with smaller body size, in general high-latitude flies are bigger (Huey et al. 2000; De Jong & Bochdanovits 2003). If InR contributes to selection-mediated differences in body size, it either acts epistatically with other body-size loci or suffers antagonistic selection pressures across multiple fitness axes. This finding, like observations of the pleiotropic polymorphism at neurofibromin 1 in which the high-latitude haplotype is negatively associated with wing size (Lee et al. 2013), demonstrates the complexity of selection dynamics in natural populations.
One hypothesis for strong genetic correlations among life history traits, including the tradeoff between lifespan and reproduction, is the presence of pleiotropic alleles. (Williams 1957; Flatt & Heyland 2011). If many genetic variants for life history are pleiotropic, then correlations among traits can never be completely dissolved. Our results implicate extensive pleiotropy, but the complexity of the InR indel polymorphism limits our ability to map phenotype to specific nucleotides. The two alleles we tested, InRshort and InRlong, differ at four amino acids across a span of 16 residues (Paaby et al. 2010). Even if in isolation these sites could break pleiotropy by functioning independently, over short timescales the distinction between pleiotropy and close linkage may be inconsequential (Paaby & Rockman 2013). However, this region at InR clearly experiences sufficient recombination (or mutation) over longer timescales to generate a hyper-variable set of alleles on which selection may act. Tests on additional InR alleles would help to resolve the functional roles of individual nucleotides. Moreover, our findings in support of pleiotropy may not represent the majority of life history alleles segregating in natural populations. Other work has shown that recombination can generate genotypes with positively correlated effects for both lifespan and reproduction, which suggests that life history variation in nature is likely determined by both pleiotropic and recombining non-pleiotropic loci (Khazaeli & Curtsinger 2013). Sampling genotypes from nature may represent blocks of alleles that produce the commonly observed correlations, but which can be broken under the right circumstances. Finally, although the two major alleles of the InR indel polymorphism showed canonical tradeoffs in the lab, the role of these alleles in the natural environment is unknown. Whether and how they effect trait expression might vary across different environments (Fournier-Level et al. 2013), with the possibility that under some conditions, pleiotropy may present as conditional neutrality (Anderson et al. 2011). Thus, true understanding of the role of this polymorphism in adaptive response, including fitness consequences and degree of pleiotropy, requires investigation rooted in the natural environment.
ACKNOWLEDGEMENTS
We thank Katherine O'Brien for assistance in measuring the fly wings and Luke Noble for lending expertise on the pooled sample analyses. We also thank Li Yang and Nancy Bonini for sharing expertise and resources for qPCR. This work was supported by NSF DEB 0921307, NIH F32 GM097837 and the Charles H. Revson Foundation.