Phenotypic Plasticity through Transcriptional Regulation of the Evolutionary Hotspot Gene tan in Drosophila melanogaster

Phenotypic plasticity is the ability of a given genotype to produce different phenotypes in response to distinct environmental conditions. Phenotypic plasticity can be adaptive. Furthermore, it is thought to facilitate evolution. Although phenotypic plasticity is a widespread phenomenon, its molecular mechanisms are only beginning to be unravelled. Environmental conditions can affect gene expression through modification of chromatin structure, mainly via histone modifications, nucleosome remodelling or DNA methylation, suggesting that phenotypic plasticity might partly be due to chromatin plasticity. As a model of phenotypic plasticity, we study abdominal pigmentation of Drosophila melanogaster females, which is temperature sensitive. Abdominal pigmentation is indeed darker in females grown at 18°C than at 29°C. This phenomenon is thought to be adaptive as the dark pigmentation produced at lower temperature increases body temperature. We show here that temperature modulates the expression of tan (t), a pigmentation gene involved in melanin production. t is expressed 7 times more at 18°C than at 29°C in female abdominal epidermis. Genetic experiments show that modulation of t expression by temperature is essential for female abdominal pigmentation plasticity. Temperature modulates the activity of an enhancer of t without modifying compaction of its chromatin or level of the active histone mark H3K27ac. By contrast, the active mark H3K4me3 on the t promoter is strongly modulated by temperature. The H3K4 methyl-transferase involved in this process is likely Trithorax, as we show that it regulates t expression and the H3K4me3 level on the t promoter and also participates in female pigmentation and its plasticity. Interestingly, t was previously shown to be involved in inter-individual variation of female abdominal pigmentation in Drosophila melanogaster, and in abdominal pigmentation divergence between Drosophila species. Sensitivity of t expression to environmental conditions might therefore give more substrate for selection, explaining why this gene has frequently been involved in evolution of pigmentation.


Introduction
Phenotypic plasticity, "the property of a given genotype to produce different phenotypes in response to distinct environmental conditions" [1], is a widespread phenomenon. Phenotypic plasticity can be adaptive if different but optimal phenotypes are produced by a given genotype in distinct environments [2]. Furthermore, phenotypic plasticity could facilitate evolution [3][4][5][6]. In particular, Conrad Waddington showed that changes in environmental conditions can reveal cryptic genetic variation that can be selected, allowing to fix a phenotype initially observed only in particular environmental conditions [7,8]. Waddington called this process "genetic assimilation". Analysis of phenotypic plasticity and morphological complexity in an evolutionary framework supports indeed the idea that phenotypic plasticity increases evolutionary potential. For example, a recent study on feeding structure evolution in nematods revealed that phenotypic plasticity correlates with morphological diversification [9]. The question then arises whether the same genes are involved in phenotypic plasticity and in phenotypic variation within and between species. To address this question, the molecular mechanisms underlying phenotypic plasticity need to be identified. Several examples show that environmental factors can strongly affect the transcriptome [10] through modification of chromatin structure by DNA methylation [11], histone mark apposition [12] or nucleosome remodelling [13]. In Drosophila melanogaster, female abdominal pigmentation is a plastic trait as it is darker in females grown at 18°C than at 29°C [14]. As low temperature leads to darker pigmentation, which increases body temperature, the thermal plasticity of female abdominal pigmentation is thought to be adaptive [14]. Abdominal pigmentation in drosophilids is a particularly appropriate model to study phenotypic plasticity, as the genes involved in abdominal pigmentation are well known. Indeed, abdominal pigmentation has been used as a model to dissect the genetic bases of sexual dimorphism and of variation within or between species [15][16][17][18][19][20][21][22][23].
In none of these studies, which focussed on genetic factors and were performed in standard conditions (usually at 25°C), was the effect of the environment taken into account. However, Drosophila melanogaster can develop between 12°C and 30°C [24]. As temperature varies spatially and temporally in the wild, taking it into account is paramount to understand the development and evolution of abdominal pigmentation. Using mainly genetics approaches, we previously showed that temperature acts on melanin production by modulating a chromatin regulator network, but we did not further dissect the underlying molecular mechanisms [25]. Here, we identify the pigmentation gene tan (t) as the major structural gene involved in female abdominal pigmentation plasticity and we show that chromatin structure at this locus is modulated by temperature. Temperature dramatically modulates t expression in the female abdominal epidermis and this modulation plays a major role in female abdominal pigmentation plasticity. Temperature modulates the activity of an enhancer of t, t_MSE [17], but had no detectable effect on its chromatin structure. By contrast, the active histone mark H3K4me3 is strongly enriched on the t promoter at low temperature. The H3K4 methyl-transferase responsible for this effect is likely Trithorax (Trx). Indeed, we show that Trx regulates t expression and the level of H3K4me3 on the t promoter, and is involved in abdominal pigmentation as well as in its plasticity. As t has been linked to pigmentation divergence within or between Drosophila species [17,19,20,26], t is listed among hotspot loci of evolution [27]. Our study therefore suggests that the sensitivity of particular genes to environmental changes could turn them into evolutionary hotspots by giving more substrate for selection.

Results
Temperature modulates the expression of the pigmentation gene tan in the posterior abdominal epidermis of females To focus on the effect of temperature, we quantified abdominal pigmentation in females from an inbred w 1118 line, the wild-type stock commonly used in our laboratory for molecular experiments (Fig 1). As previously described for other D. melanogaster lines [14], flies raised at 18°C were darker than flies raised at 25°C or 29°C (Fig 1A). Female pigmentation plasticity was observed in the whole abdomen but was particularly pronounced in posterior abdominal segments A5, A6 and A7 (Fig 1B, A5: p<0.001; A6: p = 0.001; A7: p<0.001). Furthermore, statistical analyses revealed that temperature accounted for most of the variation of pigmentation (Eta-squared, A5: 0.91; A6: 0.93; A7: 0.95).
Cuticle pigmentation is a complex trait that involves the coordinated expression of many pigmentation enzyme coding genes, expressed from the second half of pupal life to the beginning of adulthood depending on the gene [28,29] (Fig 2A). To test whether the expression of these genes was modulated by temperature, we performed RT-qPCR experiments on epidermes of A5, A6 and A7 segments from w 1118 females grown at 18°C or 29°C and collected at late pupal stage (pharates, Fig 2B left), or within two hours after eclosion, i.e. when cuticle tanning occurs (young adults, Fig 2B right). In pharates, the expression of tan (t), ebony (e), Dopa Decarboxylase (DDC), yellow (y) and black (b) was moderately modulated by temperature (less than 2 times). In young adults, among all genes tested, only t showed a significant modulation of expression by temperature. This modulation was very strong as t was expressed 7 times more at 18°C than at 29°C ( Fig  2B, p<0.01). We therefore focused on t and we analysed its spatial expression by in situ hybridization in D. melanogaster female abdominal epidermis (line w 1118 ) (Fig 2C and 2D). t was strongly expressed in the posterior abdomen of females grown at 18°C, as previously shown for D. yakuba females whose abdomen is darkly pigmented [30]. However, in D. melanogaster, t expression was strongly reduced at 29°C, which correlates with the lighter pigmentation of adult females. As t activity increases melanin production ( [31] and Fig 2A), its changing expression with temperature might be directly linked to abdominal pigmentation plasticity of females.
Temperature modulation of tan expression is essential for abdominal pigmentation plasticity in females If modulation of t expression by temperature were necessary and sufficient for thermal plasticity of female abdominal pigmentation, then manipulating t expression should counteract the effect of temperature. To test this hypothesis, we down-regulated or over-expressed t throughout development using the pannier-Gal4 driver [32] (pnr-Gal4) combined with a UAS-RNAi-t ( [33]) or a UAS-t ( [31]) transgene ( Fig 3A). As pnr is expressed only in the dorsal region of the body [32], the lateral regions serve as internal controls. t down-regulation at 18°C was sufficient to reduce pigmentation, which shows that high t expression at low temperature is required for dark pigmentation. Conversely, t over-expression at 29°C was sufficient to increase pigmentation, proving that at high temperature the lower level of t expression is limiting for melanin production. Similar results were obtained with yellow-wb-Gal4 (y-Gal4), a driver expressed in wing and body epidermes at the late pupal stage (Fig 3B). These results show that modulation of t expression by temperature plays a major role in thermal plasticity of female abdominal pigmentation.
In the pigment synthesis pathway, e encodes the enzyme that synthesizes the substrate of Tan (Fig 2A). We thus wondered whether a functional e gene was required to observe the effect  Cuticle pigment synthesis pathway [28]. Enzymes are indicated in red. (B) Quantification of pigmentation gene expression in posterior abdomen epidermes (segments A5, A6 and A7) from female w 1118 pharates (left) and young w 1118 adult females (right) grown at 18°C or 29°C (pools of 50 epidermes for pharates and 30 epidermes for adults, n = 3, error bars: standard deviations; gene expression at 18°C has been normalized on gene of t modulation on pigmentation. To test this, we manipulated t expression in an e loss-offunction mutant background (e 1 allele). t mis-regulation had no phenotypic consequence on pigmentation in this background (S1 Fig), showing that e is epistatic over t. Hence, a functional e gene is required to observe the phenotypic effect of t expression modulation. This result again points towards t as the major effector of pigmentation thermal plasticity.
Involvement of a gene in thermal plasticity is quantified by the effect of the interaction between genotype and temperature. To further establish the role of t in thermal plasticity of female abdominal pigmentation, we compared the reaction norms [pigmentation = f(temperature)] of control flies and of t loss-of-function mutant flies (t d07784 allele) (Fig 4, S2 Fig). We observed a very strong effect of temperature (T, p<0.001; Eta-squared = 0.38) and of genotype (G, p<0.001; Etasquared = 0.49) alone. As t is involved in abdominal pigmentation [31], this result was expected. In addition, the effect of the interaction between genotype and temperature was also very strong (GxT, p<0.001; Eta-squared = 0.08). Hence, t d07784 females are less plastic than wild type females, thus corroborating the role of t in thermal plasticity of abdominal pigmentation.

Regulation of tan expression by temperature implicates an abdominal epidermis enhancer
The effect of temperature on t expression could be mediated by its cis-regulatory sequences. An enhancer essential for driving t expression in the epidermis of abdominal segments A5 and A6 in males, t_MSE, was previously mapped upstream of t, between the genes CG15370 and Gr8a [17] (Fig 5A). We analysed the activity of a t_MSE-nEGFP reporter transgene [17] in young females grown at 18°C and 29°C. Quantification of nEGFP in segments A5, A6 and A7 showed that this enhancer was also active in female abdominal epidermes. Furthermore, its activity was modulated by temperature, as nEGFP was between 1.3 and 2 times more expressed at 18°C than at 29°C, depending on the segment (Fig 5B and 5C, p<0.001). When using an ebony-nEGFP transgene in which nEGFP is under control of the regulatory sequences of ebony [34], a pigmentation gene not modulated by temperature in the posterior abdominal epidermis of young females (Fig 2B and S3A Fig), we observed no higher nEGFP expression at 18°C as compared to 29°C (S3B and S3C Fig). This indicates that transcription of nEGFP and not stability of the nEGFP protein was responsible for the effect observed with the t_MSE-nEGFP transgene. Interestingly, the fold change observed with the t_MSE-nEGFP transgene between 18°C and 29°C was lower than that of t expression ( Fig 2B). This could be due to the genetic background. Alternatively, additional regulatory sequences of t may be important to mediate the effect of temperature. In conclusion, these results show that the effect of temperature on t expression is mediated, at least partly, by t_MSE.

Temperature affects chromatin configuration in tan region
Modulation of t_MSE activity by temperature prompted us to analyse the chromatin structure of this enhancer in epidermes of female abdominal segments A5, A6 and A7 at 18°C and 29°C ( Fig 6). As nucleosome depletion characterizes active regulatory chromatin regions [35,36], we performed Formaldehyde Assisted Isolation of Regulatory Elements (FAIRE)-qPCR expression at 29°C). The expression of tan, ebony, DDC, yellow and black is moderately modulated by temperature in pharates, whereas only tan is dramatically modulated in young adults (t-test: *: p<0.05; **: p<0.01). The expression of Tyrosine Hydroxylase (TH) and Laccase 2 is modulated neither in pharates nor in adults. (C) Analysis of tan expression in abdominal epidermes from young w 1118 adult females grown at 18°C or 29°C. Note that tan is more strongly expressed in the posterior abdominal epidermis at 18°C than at 29°C. (D) Adult cuticle (left) and tan expression in abdominal epidermis (right) from females in which tan was down-regulated using the pnr-Gal4 driver and a UAS-RNAi-t transgene. The dashed line marks the limit between the pnr driver expression domain (a dorsal strip) and the lateral region used as an internal control. Note the loss of pigmentation (left panel) and the strong decrease in tan expression (right panel) in the dorsal region, showing specificity of tan antisense probe.
doi:10.1371/journal.pgen.1006218.g002  [37,38]. FAIRE experiments have previously shown that the VG01 enhancer of vestigial (vg), which recapitulates vg expression in wing and haltere imaginal discs, was specifically open in these tissues, but not in leg imaginal discs where vg is not expressed [38]. As vg is not expressed in the abdominal epidermis either (S4A Fig), we used VG01 as a negative control. FAIRE signal was significantly higher on t_MSE, showing that t_MSE was less compact than VG01 at 18°C and 29°C (18°C: p<0.01; 29°C: p<0.05) (Fig 6A). However, compaction of t_MSE was similar at 18°C and 29°C. Similar conclusions were drawn from analysis of total histone 3 (panH3) enrichment by chromatin immunoprecipitation experiments (ChIP-qPCR), which showed a higher nucleosome concentration on the VG01 enhancer than on t_MSE, but no difference between 18°C and 29°C for both enhancers (S5A Fig). We then analysed the enrichment of t_MSE in H3K27ac, a histone mark characteristic of active enhancers [39]. t_MSE was enriched in H3K27ac compared to VG01 enhancer at both 18°C (p<0.001) and 29°C (p<0.05). However, we detected no significant H3K27ac enrichment on t_MSE at 18°C compared to 29°C (Fig 6B and S5B Fig). This result indicates that t_MSE is active at 18°C and at 29°C. Furthermore, temperature affects neither the compaction of t_MSE nor the apposition of H3K27ac. However, other histone marks on t_MSE might be modulated by temperature. Alternatively, the effect of temperature on chromatin structure could target another region of t, for example its promoter. We thus studied chromatin compaction and the H3K4me3 active mark at the t promoter. The 500 base pair region upstream of the transcription start site (TSS) of active genes, which includes the promoter, is known to be depleted in nucleosomes [40]. FAIRE-qPCR experiments showed that chromatin upstream the t TSS (t-TSS-up, -253 to -151 bp) tended to be less compact at 18°C than at 29°C (Fig 6A, p = 0.087), which correlated with the higher expression of t at 18°C compared to 29°C. No such difference was observed for CG12119 (Fig 6A, CG12119-TSS-up, -266 to -200 bp), a gene nearby t (Fig 5A) that was expressed at the same level at 18°C and 29°C (S4B Fig), or for an untranscribed region between CG12119 and t (Fig 6A, NC). Highly transcribed genes are enriched in H3K4me3, with a Such a difference between 18°C and 29°C, which correlates with higher t expression at 18°C, was not observed for CG12119-TSS-down (204 to 256 bp) or for NC (Fig 6C and S5C Fig).
In conclusion, our results show that temperature modulates chromatin compaction and H3K4me3 enrichment on the t promoter in the posterior abdominal epidermis of females.
trithorax participates in tan regulation and thermal plasticity of pigmentation As temperature modulates deposition of the H3K4me3 active mark on t, we addressed the role of genes involved in H3K4 methylation in pigmentation and its plasticity. In D. melanogaster, H3K4 mono-, di-and tri-methylations are catalysed by three complexes of the COMPASS family called Trithorax (Trx), Trithorax-related (Trr) and Set1. These complexes are characterised by their histone methyl-transferase subunit encoded by the genes trx, trr and Set1, respectively [42]. The histone methyl-transferase Trx was also purified previously from another complex, TAC1 [43]. Whereas Trr is involved in H3K4 mono-methylation [44], Set1 is responsible for the bulk of H3K4 di-and tri-methylation [45]. Independent studies indicate a role for Trx in H3K4 mono-and tri-methylation [45][46][47].
We first down-regulated trx, trr or Set1 using UAS-RNAi transgenes and the late pupal driver y-Gal4 to analyse their implication in abdominal pigmentation (Fig 7A and 7B). Downregulation of trr and Set1 using two different RNAi lines for each gene induced no changes in pigmentation. By contrast, trx down-regulation induced strong depigmentation of all abdominal segments. A similar phenotype was obtained by inducing trx down-regulation at late-pupal life with the pnr-Gal4 driver combined with Gal80 ts (Fig 7C). These results show that Trx, but neither Trr nor Set1, participates in the late steps of female abdominal pigmentation establishment.
As the level of H3K4me3 on the t promoter was modulated by temperature, we wondered whether Trx participates in t regulation. We thus quantified t expression in abdominal epidermes of y-Gal4>UAS-RNAi-trx females raised at 18°C, a temperature at which loss of pigmentation induced by trx down-regulation was very strong (Fig 8A). trx down-regulation induced a significant decrease in t expression (Fig 8B, 2.1 fold down, p<0.05), showing that trx is required for the strong expression of t in abdominal epidermes at 18°C. In addition, down-regulation of trx in abdominal epidermes significantly reduced H3K4me3 on the t promoter (Fig 8C, 3.7 fold down, p<0.05), which suggests that Trx participates in H3K4me3 deposition on t.
Interestingly, trx RNAi females exhibited stronger loss of melanin than t d07784 mutants suggesting that Trx controls the expression of other pigmentation genes. Therefore, we analysed the expression of pigmentation genes in the abdominal epidermis of y-Gal4>UAS-RNAi-trx females raised at 18°C (S6 Fig). In addition to t, TH, DDC and b were down-regulated showing that Trx also participates in their regulation.
We then investigated the involvement of trx in thermal plasticity of pigmentation. As we could not use a trx UAS-RNAi transgene since the UAS/Gal4 system is temperature-sensitive, we established the pigmentation reaction norms of trx j14A6 heterozygous mutant females ( Fig  8D and S7 Fig). The effect of this allele on pigmentation was not as strong as the one of the trx UAS-RNAi transgene, probably because only heterozygous females could be studied (the trx j14A6 allele is lethal homozygous). Nevertheless, the interaction between genotype and temperature was highly significant (Fig 8D, GxT, p<0.01), indicating that trx is involved in thermal plasticity of pigmentation.

Discussion
We show here for the first time that thermal plasticity of female abdominal pigmentation in D. melanogaster involves strong modulation of the expression of the pigmentation gene t. Furthermore, our results demonstrate that this modulation plays a major role in female abdominal pigmentation plasticity. Interestingly, a previous study analysing thermal plasticity of gene expression in the whole body of three days old D. melanogaster females showed that t expression diminishes when temperature increases [48]. However, as the abdominal pigmentation pattern is already established at this stage, it is likely that, in these experiments, other tissues contribute to the variation of t expression. As t is expressed in photoreceptors and plays a role in vision [31], it would be interesting to test whether its expression varies with temperature in adult eyes.
In young adults, t is the only pigmentation gene among those tested which is significantly modulated by temperature. However, we observed a trend towards a weaker e expression at  18°C than at 29°C, although not statistically significant (Fig 2B, p = 0.06). In pharates, several pigmentation genes, including t and e, are moderately modulated by temperature. In addition, we observed a weaker expression of e-nEGFP in A6 and A7 at 18°C than at 29°C (S3 Fig). These findings agree with our previous data showing the qualitative analysis of e expression at different temperatures using an e-lacZ transgene [25]. In this previous publication we showed that e mutants remain dark at all temperatures and concluded that "a functional e gene is required for the plasticity of pigmentation". Our present data complete this conclusion. Indeed, we show here that e is epistatic over t. This explains why e mutants lose abdominal pigmentation plasticity, as a functional e gene is required to observe plasticity induced by modulation of t expression.
Furthermore, our data show that the expression of e, DDC, y and b is modulated by temperature in pharates. This could explain the residual pigmentation plasticity observed in t mutants.
Lastly, spatial analysis of e expression by in situ hybridization reveals a stronger expression at 29°C than at 18°C in anterior abdominal segments. This observation suggests that the reduced but observable plasticity of these anterior segments might be due to e temperature sensitive expression.
The effect of temperature on t expression is mediated, at least partly, by the t_MSE enhancer. Thus, this enhancer may have particular properties making it temperature sensitive. Indeed, recent data showed that the number of redundant binding sites for a particular transcription factor in an enhancer could influence its temperature sensitivity [49]. Another, nonexclusive, explanation could be that temperature affects the expression or the activity of regulatory factors upstream of t. We detected no chromatin modification of t_MSE at different temperatures, possibly because this enhancer is active, although at different levels, at the temperatures we tested. The level of H3K27ac could therefore be saturated and the chromatin on t_MSE decompacted at both temperatures. By contrast, the effect of temperature on t expression is correlated with the modulation of H3K4me3 deposition on the t promoter. As this histone mark correlates with active transcription [50], the strong accumulation of t transcripts at 18°C is more likely caused by a transcriptional response to temperature than by modulation of a post-transcriptional mechanism that would stabilize them. Interestingly, deposition of H3K4me3 can also be modulated by environmental conditions such as diet in mouse liver [51], drought stress in plants [52,53] or chemical stress in yeast [54]. This histone mark emerges therefore as a general mediator of environmental impact on the genome.
We show that the H3K4 methyl-transferase Trx is involved in t regulation, but also in the regulation of other pigmentation genes. As the level of H3K4me3 on the t promoter decreases when trx is inactivated, it is tempting to speculate that Trx directly regulates t. However, Trx might also indirectly control t expression through the regulation of genes upstream t. Furthermore, as Trx has no intrinsic DNA binding activity, its recruitment on t or on upstream regulators must depend on specific transcription factors. Thus, it would be interesting to identify the epidermes (segments A5, A6 and A7) from young y-Gal4>UAS-RNAi-trx and y-Gal4>UAS-RNAi-GFP females grown at 18°C (pools of 30 epidermes, n = 3, error bars: standard deviations; t expression in y-Gal4>UAS-RNAi-trx females was normalized on t expression in y-Gal4>UAS-RNAi-GFP females). trx down-regulation induced a decrease in tan expression (t-test; *: p<0.05). (C) Analysis by chromatin immuno-precipitation of the H3K4me3 mark in abdominal epidermes of young y-Gal4>UAS-RNAi-trx and y-Gal4>UAS-RNAi-GFP females grown at 18°C. trx is required for the high level of H3K4me3 downstream the TSS of tan (t-TSS-down), on tan exon 2 (t-ex2) and downstream the TSS of CG12119 (CG12119-TSS-down), a gene near tan expressed in the abdominal epidermis. NC: untranscribed region between tan and CG12119. The graph represents the mean of three independent experiments, the error bars correspond to standard deviations. (D) Pigmentation reaction norms (n = 30 per genotype and per temperature) showed that thermal plasticity is significantly different between w 1118 and trx j14A6 /+ females. Statistical test: two-way ANOVA. **: p<0.01; ***: p<0.001. T: effect of temperature; G: effect of genotype; GxT: effect of the interaction between genotype and temperature. Pigmentation corresponds to the first principal component (PC1) extracted from pigmentation in segments A5, A6 and A7 that captures more than 95% of total variance. doi:10.1371/journal.pgen.1006218.g008 upstream regulators of t controlled by Trx as well as the transcription factors recruiting Trx on t or on its upstream regulators.
Trx also participates in the thermal plasticity of female abdominal pigmentation. This confers to Trx a very specific role as compared to other H3K4 methyl-transferases. Indeed, Set1 has been described as the main H3K4 di-and tri-methyl-transferase during Drosophila development [45]. However, our results demonstrate for the first time that Trx is involved in the thermal plasticity of female abdominal pigmentation.
Modulation of pigmentation by environmental conditions is observed in many insects [55,56]. Interestingly, t expression is strongly modulated by environmental conditions in the developing wings of Junonia coenia, a butterfly with contrasting seasonal morphs [57]. The involvement of t in pigmentation plasticity might therefore be widespread in insects.
Several studies have also linked t to pigmentation variation within or between Drosophila species. Modulation of t expression through modification of t cis-regulatory sequences has been implicated in evolution of abdominal pigmentation between species [17,19,26]. Remarkably, in D. santomea, independent mutations in t_MSE have generated three distinct loss-of-function alleles involved in the reduced pigmentation of this species [17]. Furthermore, SNPs associated with variation of abdominal pigmentation in D. melanogaster females have been identified in t_MSE [20]. Interestingly, abdominal pigmentation dimorphism in female Drosophila erecta was recently shown to be caused by sequence variation in t_MSE maintained by balancing selection [58]. The recurrent implication of t in pigmentation evolution has led to list this gene among hotspots of evolution [27]. In other organisms, genes sensitive to environment and involved in phenotypic plasticity are also responsible for differences within or between species. For example, in Brassicaceae, the reduced complexity locus (RCO) that participates in leaf margin dissection is modulated by temperature and has been repeatedly involved in leaf shape evolution through cis-regulatory sequence variation or gene loss [59]. Therefore, sensitivity of particular genes to environmental conditions might turn them into evolutionary hotspots. Indeed, this broadens the range of phenotypes produced by a particular allele, providing more substrate for natural selection.

Fly stocks
We used a w 1118 inbred line as wild-type. The UAS-t line was a gift from Dr. Nicolas Gompel, whereas t_MSE-nEGFP was from Dr. Sean Carroll's lab. The ebony-nEGFP line (ebony-(ABC+intron)-nEGFP) was from Dr. Mark Rebeiz. The UAS-RNAi-t (GD18124) and UAS-R-NAi-Set1 (GD40683) lines were from the VDRC Stock Center. The pnr-Gal4 (BL3039), y-Gal4 (BL44267), P(XP)t d07784 (BL19282), e 1 (BL1658), trx j14A6 (BL12137), as well as the VALIUM UAS-RNAi lines (Transgenic RNAi Project at Harvard Medical School) against trx (BL33703), trr (BL29563 and BL36916), Set1 (BL33704) and GFP (BL41556) were from the Bloomington Stock Center. The homozygous lethal trx j14A6 allele that corresponds to an insertion of a w+ P transposon was used in this study. This allowed us to introgress this allele in the w 1118 background (ten generations), so that the mutation is in the same genetic background as the control. Complementation test with a well characterized trx loss-of function allele (trx E2 [60]) indicated that trx j14A6 is a genuine loss-of-function allele of trx. To control the expression of RNAi transgenes during development, we combined the pnr-Gal4 driver with the tub-Gal80 ts transgene from the Bloomington Stock Center (BL7019). Gal80 inactivation was performed by shifting the progeny at late pupal stage from 18°C to 29°C. We tested that all lines allowing trx, trr or Set1 down-regulation induced lethality with the ubiquitous daughtherless-Gal4 (da-Gal4) driver. Efficiencies of BL29563 (UAS-RNAi-trr) and GD40683 (UAS-RNAi-Set1) were previously published [42,45]. For the UAS-RNAi-trx line (BL33703), quantification of trx expression level in da-Gal4>UAS-RNAi-trx embryos showed a 1.5 fold down regulation as compared to control embryos, thus proving its efficiency.

Cuticle preparations
Adult females between 3 and 5 days old were stored for 10 days in ethanol 75% before dissection. Abdominal cuticles were cut just beyond the dorsal midline, which was therefore entirely included in each preparation. After dissection, cuticles were dehydrated 5 minutes in ethanol 100% and mounted in Euparal (Roth). For nEGFP observations, abdomens were dissected in PBS, fixed 20 minutes in 3.7% paraformaldehyde in PBS, washed twice 10 minutes in PBS and mounted in Mowiol.

In situ hybridizations
Fragments of cDNAs from t (611 bp) and e (639 bp) were amplified by PCR (primer sequences are listed in S1 Table) and cloned by Topo-Cloning and LR-Recombination (Gateway) in pBlueScript vector (Invitrogen). Sense and antisense DIG-labelled RNA probes were synthesized using the appropriate RNA polymerase. In situ hybridizations were performed according to the Carroll's lab protocol (http://carroll.molbio.wisc.edu). Specificity of the antisense probe was assessed by comparison with signal from the sense probe. For t, we also performed in situ hybridization with the t antisense probe on UAS-RNAi-t/pnrGal4 females and observed a strong decrease of the signal in the pnr domain ( Fig 2D).

Image acquisitions and quantifications
Adult cuticles and abdominal in situ hybridizations were imaged with a binocular equipped with a Leica DC480 digital camera using the Leica IM50 Image Manager software. They were imaged using identical settings and an annular lamp to ensure homogeneous lighting. To quantify pigmentation, each entire hemi-segment was circled by hand. For A5 and A6, the melanic line at the dorsal limit of each hemi-segment (i.e the dorsal midline) separates the two hemisegments. Cuticle pigmentation in hemi-tergites A5, A6 or A7 was measured as mean grey value using ImageJ. This value was subtracted from 255 to get a final pigmentation value comprised between 0 (white) and 255 (black).

RT-qPCR experiments
RNA was extracted from pools of dissected female posterior abdominal epidermes (A5, A6 and A7) with the RNAeasy Mini kit (Qiagen) (50 abdominal epidermes for pharates, 30 for young adults). We could not use developmental time to stage pharates as it is temperature sensitive. We therefore used morphological markers (wing colour, abdominal bristles, localisation of the meconium in anterior abdomen) to collect pharates grown at 18°C or 29°C at a similar developmental stage. This stage corresponds to the stage P12(i) described by Bainbridge and Bownes [61]. For each experiment three independent replicates were analysed for each genotype and each temperature except for S6 Fig (two replicates). After treatment of RNA with Turbo DNAse (Ambion), cDNA were synthesized with the SuperScript II Reverse transcriptase kit (Invitrogen) using random primers. RT-qPCR experiments were carried out in a CFX96 system (Biorad) using SsoFast EvaGreen Supermix (Biorad). Expression levels were quantified with the Pfaffl method [62]. The geometric mean of two reference genes (Fig 2B and S4 Fig: rp49 and Act5C; Fig 8B: rp49 and eIF2,) was used for normalization [63]. Primers used are listed in S1 Table.

FAIRE experiments
75 posterior abdominal epidermes (A5, A6 and A7) of females between 0 and 2h after hatching were used for each FAIRE experiment. Fixation and lysis protocols were similar to those used for ChIP except that fixation was performed for 5 minutes at room temperature in PBS-1% paraformaldehyde with gentle shaking. Chromatin sonication was performed in 300μl in a Bioruptor sonifier (Diagenode) with 8 cycles of 30'' ON, 30'' OFF, High power, allowing to obtain chromatin fragments between 300-400 bp. 100μl of chromatin preparation was kept as the input (total chromatin). The rest (200μl) was submitted to phenol-chloroform extraction and the aqueous phase containing the decompacted chromatin (FAIRE chromatin) was kept. Input and FAIRE DNA were purified with the Ipure kit (Diagenode) in 150μl of water and 4μl were used per qPCR reaction. qPCR experiments were carried out in a CFX96 system (Biorad) using SsoFast EvaGreen Supermix (Biorad). Primers used are listed in S1 Table. Data were normalized against input chromatin. Results present the mean of three independent experiments.

Statistical analyses
To analyse the effect of temperature on A5, A6 and A7 pigmentation, we performed a one-way ANOVA (or Welch's ANOVA when variances were heterogeneous) with temperature as factor. To analyse the effect of t (Fig 4) or trx (Fig 8D) on pigmentation plasticity, we used a twoway ANOVA with genotype and temperature as factors. The variable analysed was the first component of a Principal Component Analysis of pigmentation in A5, A6 and A7 conducted on correlations, which captures more than 95% of total variation in both cases. ANOVAs and Welch's ANOVA were performed using the OpenStat software (W.G. Miller, http:// statprogramsplus.com/OpenStatMain.htm). Normality of the residual distributions was checked with a Shapiro-Wilk test (Anastats; http://anastats.fr).
For t-tests, we checked first homogeneity of variance using a Levene Test (Anastats; http:// anastats.fr) and then used the appropriate option of t-test.
Supporting Information S1 Fig. ebony is epistatic over tan. Down-regulation or up-regulation of tan in the abdominal dorsal domain using the pnr-Gal4 driver and UAS-RNAi-t or UAS-t transgene, respectively, in a wild-type (above) or ebony (e 1 , below) mutant background at 25°C. In a wild-type background, tan down-regulation strongly reduced pigmentation in the 6 th abdominal segment ( Ã ), whereas tan over-expression increased melanin production in all segments. In contrast, the modulation of tan expression had no effect on pigmentation in an ebony mutant background.