The Insulin-Like Proteins dILPs-2/5 Determine Diapause Inducibility in Drosophila

Diapause is an actively induced dormancy that has evolved in Metazoa to resist environmental stresses. In temperate regions, many diapausing insects overwinter at low temperatures by blocking embryonic, larval or adult development. Despite its Afro-tropical origin, Drosophila melanogaster migrated to temperate regions of Asia and Europe where females overwinter as adults by arresting gonadal development (reproductive diapause) at temperatures <13°C. Recent work in D. melanogaster has implicated the developmental hormones dILPs-2 and/or dILP3, and dILP5, homologues of vertebrate insulin/insulin-like growth factors (IGFs), in reproductive arrest. However, polymorphisms in timeless (tim) and couch potato (cpo) dramatically affect diapause inducibility and these dILP experiments could not exclude this common genetic variation contributing to the diapause phenotype. Here, we apply an extensive genetic dissection of the insulin signaling pathway which allows us to see both enhancements and reductions in egg development that are independent of tim and cpo variations. We show that a number of manipulations dramatically enhance diapause to ~100%. These include ablating, or reducing the excitability of the insulin-producing cells (IPCs) that express dILPs-2,3,5 employing the dilp2,3,5-/- triple mutant, desensitizing insulin signaling using a chico mutation, or inhibiting dILP2 and 5 in the hemolymph by over-expressing Imaginal Morphogenesis Protein-Late 2 (Imp-L2). In addition, triple mutant dilp2,3,5-/- females maintain high levels of diapause even when temperatures are raised in adulthood to 19°C. However at 22°C, these females all show egg development revealing that the effects are conditional on temperature and not a general female sterility. In contrast, over-expression of dilps-2/5 or enhancing IPC excitability, led to levels of ovarian arrest that approached zero, underscoring dILPs-2 and 5 as key antagonists of diapause.


Introduction
In temperate regions, several holometabolous insects overwinter in a state of diapause, an actively induced dormancy that blocks developmental growth at species-specific stages of their life cycles and enhances cold tolerance [1]. Drosophila melanogaster, despite its Afro-tropical origins, has evolved a diapause response in temperate regions where adult females induce reproductive dormancy by arresting gonadal growth in response to cold environmental conditions below 13°C in combination with photoperiodic stimuli [2][3][4]. A fine-tuned regulation of diapause timing is important for synchronizing Drosophila's life-cycle with environmental changes and several genes have been implicated in reproductive arrest under colder temperatures and shorter photoperiods.
The s-tim/ls-tim natural polymorphism in the circadian clock gene timeless generates a substantial difference in diapause inducibility as revealed in several natural populations and confirmed with transgenic lines, where the recently derived ls-tim mutation enhances diapause at all photoperiods [2,5]. Two closely linked polymorphisms in couch potato (cpo), cpo A356V and cpo I462K (now reannotated as cpo A347V and as intronic polymorphism SNP 48034(A/T)- [6]) show latitudinal clines in North America, have also been suggested to modulate diapause induction [7] but no relationship between these polymorphisms and diapause is present in Australian populations [8]. cpo lies between the proximal and distal breakpoints of the cosmopolitan Payne inversion, In(3R)P, which shows a cline in frequency on both continents complicating interpretation of the cpo cline [9,10]. Nevertheless, complementation analysis reveals that these two natural cpo variants as well as artificially generated mutations of cpo, significantly affect diapause with higher levels of diapause inversely correlated with cpo expression levels [7].
In addition, natural variation in another gene that lies within In(3R)P and which encodes insulin-regulated phosphatidylinositol 3 kinase (PI3K), dp110, appears to affect diapause levels and may also contribute to a latitudinal cline in diapause induction that is observed in North America, but the molecular basis for the polymorphism remains obscure [11,12]. The insulinlike receptor (InR) gene also lies within In(3R)P and encodes considerable polymorphism, with one allele showing a similar latitudinal cline in frequency in both North America (standard chromosome only) and Australia (inverted chromosome only) [13]. Heteroallelic mutational combinations in InR have long been known to generate female sterility as the eggs remain previtellogenic at 25°C but whether these polymorphisms affect diapause in Drosophila has yet to be studied [14,15]. Thus, until recently, there was only circumstantial evidence that insulin signaling might be involved in conditional ovarian developmental arrest at low temperatures in D. melanogaster.
However, in a recent study of D. melanogaster diapause, females maintained in diapause for several weeks showed elevated gene expression for Drosophila insulin-like peptides, dILPs 2-6 [16]. If there exists a causal relationship between diapause and dilp levels, then mutations in some of these dilp genes might be expected to reduce levels of diapause. In contrast, mutations in dilp2-3 and dilp5 appeared to enhance the depth of diapause as measured by ovarian development [16], possibly caused by compensatory up-regulation of other dILPs [17]. One potential problem with this experiment, apart from the small numbers of female flies that were analysed, was that the wild-type and dilp2-3 -/and dilp5 -/mutants were not assessed for either their cpo or tim genotypes which could conceivably have generated their different diapause profiles.
Given the available evidence that suggests insulin signaling involvement in the Drosophila overwintering response, we sought to further dissect the role of these dilps in diapause induction by using an extensive set of genetic manipulations on known tim and cpo genetic backgrounds to clarify the role of insulin signaling in this important seasonal adaptation.

Results/Discussion
We screened all our lines for the tim [2] and cpo [7] natural alleles (see Materials and Methods). Among our lines, the single dilp2 -/-, dilp5 -/-, dilp3 -/-, chico KG00032 mutants, UAS-dilp2 RNAi, UAS-dilp5RNAi, UAS-sNPF;UAS-sNPF, akh-Gal4 and Lk6 DJ634 -Gal4 carry the ls-tim allele so we used a corresponding ls-tim as a control. For cpo, the upstream C/T (cpo A347V ) substitution showed some variation among strains whereas the downstream SNP 48034(A/T) substitution was represented as A in all strains except for UAS-hid,rpr which carried a T and is reported to promote reduced levels of diapause [6,7]. S1 Table carries the summary of all the diapause results and the background genotypes of all strains tested.
Strikingly, IPCs ablation (dilp2>hid,rpr and Insp3>hid,rpr) promotes a near complete diapause response (97.6 ± 2.9% and 97.3 ± 1.7%, respectively) compared to controls (Fig 1A). This is not female sterility because at 23°C these females all produce late stage eggs (S2 Table and S1 Fig, also shown by others [29]). If the ablations were simply causing a slowing down in maturation of the gonads rather than an arrest, we should observe that after a longer period in diapause inducing conditions, the experimental genotypes should 'catch up' and give similar levels of diapause to controls. We therefore maintained our females for 28 days at 12°C and observed that diapause levels fell to 0-1% in dilp2>+ and UAS-hid,rpr controls, and 18% in InsP3>+ but were maintained at 66% and 50% in dilp2>hid,rpr and InsP3>hid,rpr respectively ( Fig 1A). These results reveal a key role for neural IPCs and their functions in modulating and maintaining Drosophila diapause at colder temperatures.
To further understand the role of IPCs on reproductive dormancy, we disrupted the neuronal physiology of IPCs (and, in turn, the release of dILPs) by over-expressing a potassium (UAS-dOrkΔ-C, designated as UAS-Ork1) or a sodium channel (UAS-NaChBac) which reduce or enhance neuronal excitability, respectively [30,31]. Ork1ΔC over-expression from early larval life (dilp2(p)>Ork1) promotes very high levels of reproductive diapause in a similar manner to IPCs ablation (91.9 ± 2.8%) compared to its dilp2(p)>dORKΔ-NC (NOrk1) non-conducting control (36.8 ± 5.0%) whereas NaChBac over-expression (dilp2(p)>NaChBac) inhibits ovarian dormancy and induces gonadal growth in the cold (Fig 1B), reinforcing the notion that dILPs have a regulatory role on dormancy. Again Ork1ΔC over-expression females produced stage 14 eggs at 23°C and after 28 days, dilp2(p)>Ork1 females still showed >30% diapause compared to their corresponding driver and UAS controls that showed almost zero diapause (S1 Table  and Fig 1B).
The higher levels of diapause in the experimental flies illustrated in Fig 1 are generated even though these females carry the low diapause s-tim allele and were maintained in the longer summer LD16:8 photoperiod. Furthermore, the most frequent cpo genotype in these lines was C/C A/A (cpo 347Ala , cpo 48034A , S1 Table). Zonato et al (submitted) reveals that the cpo 48034A/T polymorphism does not play any role in diapause at either 12 or 28 days in the s-tim background. However the cpo A347V polymorphism was observed to have a significant effect with cpo Val showing higher levels of diapause than cpo Ala . S1 Table shows the relevant cpo A347V genotypes for all the lines, and some are heterozygous, while the majority are homozygous for cpo Ala . If we compare the control strains we observe at 11 and 28 days, diapause levels lie between 32-47% and 0-18% respectively, but there is no correlation with cpo A347V genotype, suggesting that the heterozygous combination of alleles does not enhance diapause compared with cpo Ala homozygotes. Consequently, cpo A347V genotypes in the s-tim background do not appear to play any significant role in the large changes we observe in diapause induction compared to the dramatic effects we observe when we manipulate the IPCs.
Similarly, the two genotypes that were triply mutant, dilp2,3,5 -/- [17] and Df(3L)dilp1-5/ dilp2,3,5induced maximum diapause which in the latter genotype was maintained for 28 days at 84% (Fig 2A).  We also investigated hypomorphic chico mutants (chico KG00032 ) [23], that disrupt dILPs sensitivity, and observed that they too exhibited very high levels of diapause (89% falling to 46% after 28 days, Fig 2A) Similarly the weakly hypomorphic InR EY00681 mutant [32] also showed very high levels of diapause (90.3 ± 6.2%) after 11 days, and over-expressing Imaginal Morphogenesis Protein-Late 2 (Imp-L2), using the neuroendocrine cell driver c929> [33,34], provoked 100% diapause, consistent with the function of Imp-L2 in inhibiting dILPs in the haemolymph [35]. All of these experiments were carried out at LD16:8 and all experimental genotypes carried the s-tim background that favors low diapause levels (except Imp-L2, where s/ls were segregating, Fig 2A). In contrast, none of the single null mutants (dilp2 -/-, dilp3 -/or dilp5 -/-) caused strong induction of diapause, compared to Df(3L)dilp1-5 -/-, although individually, both dilp2 -/and dilp5 -/modestly enhanced the frequency of diapause (59.8 ± 7.0% and 57.5 ± 3.7%, respectively, Fig 2A) compared to controls sharing the same ls-tim backgrounds. As dilp3 -/was homozygous for the diapause promoting cpo Val allele, whereas dilp5 -/was polymorphic at this site, any enhancement of diapause in the latter compared to dilp3 -/is possibly underestimated. These results, coupled with the suggested autocrine regulatory role of dILP3 [17] lead us to suggest dILP2 and 5 as the key IPCs-released dILPs for diapause suppression. A similarly modest but significant enhancement was observed on knocking down dilp2 (65.5 ± 8.1%) or dilp5 (74.2 ± 7.4%) with RNAi using the dilp2 promoter, compared to the corresponding UAS-RNAi controls which both gave just under 50% diapause (Fig 2B). The knocked down flies were heterozygous for cpo Val whereas the controls were homozygous for cpo Ala , which could possibly explain why these effects were a little more dramatic than the single mutants (S1 Table).
Furthermore, we also induced ectopic dilps-2/5 over-expression with a number of promoters for neuroendocrine and endocrine cells (c929, [33,34]), larval (cg, [22]) or adult (Lk6 DJ634 , [36,37]), fat-body, corpus allatum (hmgcr Di-11 [38,39]) and corpus cardium (akh) [40]. All of these manipulations significantly reduced diapause inducibility in a similar manner to dILPs-2/5 overexpression in MNCs (Fig 3), suggesting that irrespective of the tissue from which dILPs are expressed and the timing of their release, it is their presence in the hemolymph that is the critical factor. Taken together, these results suggest that dILPs-2/5 are the key antagonists of diapause and that they lie at the core of the genetic mechanism underlying induction of ovarian dormancy.
Wild-type diapausing flies can switch from diapause to full fecundity if they re-encounter a favorable environment [16]. If dILPs-2/5 levels are critical for modulating the diapause response, then we expect that Df(3L)dilps1-5 -/mutants should not transition out of diapause as quickly if they experience a shift to favorable conditions. We therefore exposed Df(3L) dilps1-5 -/and Df(3L)dilp1-5/dilp2,3,5null mutants to diapause-inducing conditions (12°C) for 11 days before switching them for an additional 5 days to non-diapausing conditions (15°C, 19°C, or 22°C). We observed that even at 19°C, the null mutants maintained high frequencies of diapause compared to the controls (Time x Genotype F 6,42 , 18.2 p<0.01, Fig 4).
Only at 23°C did the nulls start to mature their gonads, suggesting that dILPs are the limiting signals determining diapause in colder conditions. These results are important as they show clearly that these mutants are conditional sterile at low temperature and reinforce the view that we are examining diapause, a reversible phenomenon, not simply female sterility.
If our proposal is correct, then dILPs-2/5 signaling should be perturbed by environmental conditions that induce reproductive diapause. In particular, we predict that if dILPs-2/5 inhibit reproductive diapause, then dILPs-2/5 signaling should be repressed during dormancy induction. To address this issue, we used a FoxO response element-luciferase reporter (FoxO.RE-Luc) [41] to estimate the levels of insulin/FoxO signaling in diapausing flies (reared at 12°C, LD8:16) versus "non-diapausing" ones (reared at 23°C). Under conditions of reduced insulin signaling FoxO should enter the nucleus and up-regulate luciferase activity. Consistent with the notion that reduction in dILPs-2/5 signaling promotes reproductive dormancy, diapausing flies exhibited a significant (p<0.05) 6-fold up-regulation of FoxO activity in isolated abdomen (containing the gonads, direct downstream targets of brain dILPs [42]) (Fig 5A).
In other cases of insulin signaling down-regulation, there is often an accompanying compensatory enhancement of dilps-2/5 gene transcription in the IPCs. For example, increased mRNA levels of dilps-2/5 are observed in a SH2B -/mutant (SH2B is an intracellular adaptor of InR) [23] or when Imp-L2 is over-expressed [43]. In dilp2 -/mutants there is up-regulation of the dilp3 and dilp5 transcripts, while in dilp5 -/mutants, dilp3 is up-regulated, with dilp2-3 -/double mutant also enhancing dilp5 expression [17]. As dilp3 -/mutants also reduce dilp2 and dilp5 expression, there is both positive and negative cross-regulation among these dilps [17]. Similar up-regulation of dilps-2/5 transcripts have also been reported upon Imp-L2 over-   expression [43] and ablation of the germ line [44] which are direct targets of brain dILPs [42]. Thus, paradoxically, we predicted that diapausing flies might exhibit up-regulation of dilps-2/5 genes as a compensatory response. Strikingly, the expression levels of dilps-2/5 genes were both up-regulated in reproductively diapausing flies (reared at 12°C) versus non-diapausing flies (reared at 23°C), in both whole body and head-only samples of two different 'wild-type' strains, dilp2>+ and FoxO.RE-Luciferase (Fig 5B), both of which showed normal levels of diapause (40.3 ± 3.7% and 49.5 ± 4.8% in LD8:16 and 33.2 ± 4.6% and 36.2 ± 3.4% in LD16:8, respectively). A similar result was obtained with diapausing Canton-S flies by Kubrak et al [16]. Although this feedback mechanism remains to be clarified, these results support the notion that dILPs-2/5 signaling fails under the perturbing effects of an adverse environment and this is the key event for diapause induction. We note that the up-regulation of dilps-2/5 mRNA levels is the opposite to that reported in the monarch butterfly [45] and mosquito [46] where diapause correlates with down-regulation of insulin/IGFs expression. This is a curious difference, but variation in hormonal regulation among insect species, especially with respect to diapause or metamorphosis regulation, is quite common (reviewed in [1]). For example, Juvenile hormone prevents metamorphosis by specifying larval molts in Lepidoptera but it does not in Drosophila, [47,48]). Independently of the feedback mechanism that regulates the compensatory phenotype, the key aspect of our results is that, in Drosophila, reproductively diapausing flies exhibit a disruption of the normal systemic dILPs-2/5 signaling, which is consistent with the shutdown of this hormonal signaling pathway during diapause of other insect species [1,45,46] and the role of insulin-like peptides as diapause antagonists.
Our findings provide direct genetic evidence that dILPs-2/5 signaling is a central regulator of reproductive diapause in Drosophila that is independent of tim and cpo genetic backgrounds and supports and significantly extends the work on dilp2-3 -/and dilp5 -/mutants of Kubrak et al [16]. Our primary data demonstrate that loss of dILPs-2/5 signaling promotes, whereas the over-expression of dilps-2/5 prevents ovarian dormancy. The level of change between these two genetic alternatives (~100%) far outstrips any changes in diapause levels that are determined by s-tim/ls-tim or by variants in cpo. We can now reconsider the results of Kubrak et al [16] who observed significantly reduced levels of ovarian development in dilp5 -/and dilp2-3 -/double mutants. The former single mutant is in a diapause promoting ls-tim background and has a polymorphic cpo A347V genotype so the effects may have been over-estimated by Kubrak et al [16], depending on whatever were the background genotypes of their controls. In a controlled genetic background, any effect on diapause of the single dilp5 -/or dilp2 -/mutants is modest in our hands while dilp3 -/has no effect. However, combining dilps2,3,5 mutations generates 100% diapause, demonstrating the epistatic nature of these mutations which will largely bypass the mRNA compensation phenotypes observed in each of the single mutants by Grönke et al [17].
It is further worth pointing out that the changes we have observed by manipulating dILPs signaling are conditional on temperature, so, for example, dilps2-3,5 -/null mutants are not sterile, even though they remain in ovarian arrest at temperatures up to 19°C. Indeed all the variants we have used are fertile at 23°C, even the most severe, Df(3L)dilp1-5 -/-, which as a homozygote has about 10% fertility and 50% viability at this temperature [25]. At 12°C this variant showed 100% diapause, so even partially fertile females show a temperature-sensitive phenotype. Furthermore, by extending some of our key observations to 28 days we avoid basing our interpretations solely on the first two weeks of diapause which can show some fluctuation in the dynamics of egg development [16]. We conclude that our extensive genetic manipulations of insulin-like signaling in D. melanogaster reveal dILPs-2/5 to be the key regulators and antagonists of seasonal diapause.

cpo sequencing
For cpo genotyping, the genomic DNA of several individuals from each strain used in the diapause assay was extracted, and the cpo region encompassing the nucleotide polymorphisms encoding the cpo A347V and cpo 48034(A/T) was amplified via PCR. Primers used were: forward primer 5'-aaca tccgttgctgctgtc-3' reverse primer 5'-ccccaagctgtcacttttgt-3' Templates were purified through minicolumns (Wizard1 SV Gel and PCR Clean-Up System) and subsequently sequenced.

Luciferase Reporter Gene Assay
The FoxO.RE-Luciferase transgene contains a firefly luciferase reporter gene under the control of 8 consecutive FoxO-response elements (FRE) [41]. FoxO.RE-Luc larvae were reared as for the diapause assay. Females were dissected in dry ice and, then, frozen at -80°C until processing. Luciferase extraction was performed using the Promega Luciferase Assay System (Firefly Luciferase). 150μL lysis buffer (LB) was added to 40 abdomens (9 biological replicates). Samples were frozen in liquid nitrogen and thawed in a 37°C water bath three times and then centrifuged to remove debris. This process was repeated and the two resulting supernatants were combined and stored at -70°C. Luciferase activity of adult protein extracts was measured using a Berthold Technologies Luminometer microplate scintillation and counter. 100μL of Promega Luciferase Assay Reagent was added to 20μL of protein extract and light production was measured in relative light units (RLU) emitted over a 10 s time period. Final Luciferase values (Lv) were normalized to the protein concentration (RLU/μg of protein). Protein concentration was determined using the Pierce BCA Protein Assay Kit-Reducing Agent Compatible. Data from two temperatures (12 and 23°C) were compared using t-test. qPCR and Gene expression mRNA was extracted from 50 isolated heads (3 biological replicates) or 30 whole bodies (5 biological replicates) of adult females by using TRIzol (Invitrogen) and RNeasy Mini Kit and RNase-Free DNase Set (QIAGEN), respectively. The first-strand cDNA was synthesized by using the Invitrogen SuperScript II First-Strand Synthesis SuperMix. qPCR was performed in LightCycler DNA Master SYBR Green I (Roche) on LightCycler 480 System (Roche). Primers are following: dilp2 (heads), F: 5'-gtatggtgtgcgaggagtat, R: 5'-tgagtacacccccaagatag; dilp5 (heads), F: 5'-agttctcctgttcctgatcc, R: 5'-cagtgagttcatgtggtgag; rp49 (heads), F: 5'-agggtatcgacaacagagtg, R: 5'-caccaggaacttcttgaatc; dilp2 (body), F: 5'-acgaggtgctgagtatggtgtgcg, R: 5'-cacttcgcagcggttccgatatcg; dilp5 (body), F: 5'-tgttcgccaaacgaggcaccttgg, R: 5'-cacgatttgcggcaacaggagtcg; rpL23 (body), F: 5'-ga caacaccggagccaagaacc, R: 5'-gtttgcgctgccgaataaccac. For each dilps transcript, we normalized message levels relative to rpL23 (whole body) or rp49 (isolated heads) housekeeping genes by using the 2 -ΔΔCT . Data were analyzed using t-test.