JNK signaling regulates oviposition in the malaria vector Anopheles gambiae

The reproductive fitness of the Anopheles gambiae mosquito represents a promising target to prevent malaria transmission. The ecdysteroid hormone 20-hydroxyecdysone (20E), transferred from male to female during copulation, is key to An. gambiae reproductive success as it licenses females to oviposit eggs developed after blood feeding. Here we show that 20E-triggered oviposition in these mosquitoes is regulated by the stress- and immune-responsive c-Jun N-terminal kinase (JNK). The heads of mated females exhibit a transcriptional signature reminiscent of a JNK-dependent wounding response, while mating—or injection of virgins with exogenous 20E—selectively activates JNK in the same tissue. RNAi-mediated depletion of JNK pathway components inhibits oviposition in mated females, whereas JNK activation by silencing the JNK phosphatase puckered induces egg laying in virgins. Together, these data identify JNK as a potential conduit linking stress responses and reproductive success in the most important vector of malaria.

www.nature.com/scientificreports/ a temporary refractoriness to further copulation and increased oviposition. Interestingly, while 20E in D. melanogaster does have an important role in regulating courtship behavior and specifically in the consolidation of long-term courtship memory in males 14,15 , post-mating responses in fruit fly females are not driven by 20E but by small male accessory gland peptides (Acps) transferred to the female during mating. These include ovulin, which is implicated in the control of oviposition 16 , and most importantly Acp70A, also known as Sex Peptide (SP) 17 . SP is necessary and sufficient to induce the post-mating switch: injection of exogenous SP induces both refractoriness to mating and oviposition 17 while loss of the G protein-coupled Sex Peptide Receptor (SPR) in females largely blocks these responses 18 . Consistent with its role in modulating female post-mating behavior, SPR is found in neurons innervating the reproductive tract as well as the brain and ventral nerve chord 18,19 , while ovulin acts through octopamine-dependent neurons 16 . The importance of the brain in controlling female responses after copulation in Drosophila is also demonstrated by the fact that in genetic SPR mutants post-mating responses can be rescued by introducing a mutation that yields a leaky blood-brain barrier phenotype 20 , identifying the entry of mating factors into the brain as a potentially crucial step in inducing post-mating changes. Consistent with these findings, the Drosophila brain exhibits a robust transcriptional program following mating or injection of exogenous SP 21 .
Here we show that 20E-induced oviposition behavior in An. gambiae is partially regulated by c-Jun N-terminal kinase (JNK) signaling in the female head. We detect a strong, mating-induced transcriptional signature in female heads, enriched in immune genes and reminiscent of a JNK-dependent wound-healing response. Silencing multiple components of JNK signaling reduces oviposition rates of mated females, as well as those of virgin females injected with 20E. Conversely, JNK activation by depletion of the negative regulator puckered increases oviposition rates in virgin females. Our results unveil an unexpected link between an important mosquito reproductive behavior and the activation of JNK, a pathway classically associated with stress resistance and longevity 22,23 but which is also pivotal to anti-plasmodium immunity 24,25 .

Results
A transcriptional signature of wounding response is found in the head after mating. To gain insight into the molecular basis of the mating response in An. gambiae, we performed transcriptional analysis of the heads of mated and age-matched virgin females at 3 and 24 h post mating (hpm). Microarray analysis identified a strong immune signature triggered by mating in the head, the like of which had not been detected in similar analyses of other An. gambiae female tissues 12,13,26 . As summarized in Table 1, 23 genes were differentially regulated after mating at the two time points under analysis, 22 of which were upregulated at either 3 hpm (11 genes) or 24 hpm (11 genes). A single gene, an acyltransferase (AGAP007078), was down regulated in the head 24 hpm. Functionally, 6 of the 22 upregulated genes were common to a group of genes previously linked to the wounding response in An. gambiae 27 , while 16 were implicated in the melanization pathway, which has been studied in this species predominantly in the context of Plasmodium infection 28,29 but is also strongly linked to wound healing in other insects including Drosophila 30,31 . In addition, 13 of the upregulated genes were previously found to be preferentially expressed in hemocytes 32 , cells related to mammalian macrophages and central to the repair and regeneration of damaged cells 33 and to the wounding-induced transcriptional response in Drosophila 34 .
At the earlier 3 hpm time point we found mating-induced upregulation of transglutaminase 2 (TGase2), a member of an enzyme family involved in chemical crosslinking of proteins in the hemocel that is implicated in the coagulation responses that follow infection or trauma 31 and was previously shown to be involved in JNKdependent wounding responses in An. gambiae 27 (Table 1). Also upregulated at 3 hpm were two members of the CLIP protease family, CLIPs B1 and 11E-like, serine proteases which initiate proteolytic cascades leading to the activation of prophenol oxidases (PPOs) that mediate both melanization 35 and coagulation 31 . We also identified the CLIP inhibitor, Serpin 17, a member of the serine protease inhibitor (Serpin) family which block CLIP-mediated proteolytic processing of PPO enzymes 36 . Finally, we detected the up-regulation of L-dopachrome tautomerase (also known as Yellow F2), an enzyme involved in melanin biosynthesis 35 (Table 1).
Among the genes upregulated at 24 hpm we identified three PPOs (PPO2, PPO5, and PPO6) and three additional CLIP proteases; CLIP15B, CLIPC7, and gastrulation defective. Beyond the melanization pathway we also found multiple thioester-containing proteins (TEP1, TEP4, TEP8 and TEP19) which are linked to the mosquito complement-like response 37 , along with APL1C (Anopheles-Plasmodium-responsive leucine-rich repeat protein 1, isoform C), a member of the leucine-rich repeat (LRR) immune protein (LRIM) family. APL1C associates with another LRIM family member (LRIM1, not identified here), and TEP1 in a complex that retains TEP1 in a stably active form that can then form thioester bonds with surface-exposed proteins of invading pathogens including Plasmodium parasites [38][39][40] , targeting them for lysis. Taken together, these data suggest mating induces a strong wounding response in the head of An. gambiae females.
JNK is activated in the head after mating. The preponderance of wounding-related transcripts in the head after mating prompted us to ask what signaling pathways might lead to such a response. The importance of the JNK signaling pathway in the insect wounding response has been highlighted in both An. gambiae 27 and Drosophila 41 . Since JNK is well known to be activated post-transcriptionally by phosphorylation, we asked whether mating might increase levels of active (phosphorylated) JNK (pJNK) in the female's head. An. gambiae females were dissected around the onset of detectable transcriptional changes (1-6 hpm), and heads were analyzed by Western blot using an antibody specifically recognizing pJNK. Mating induced a marked increase in the levels of pJNK in the head compared to virgin controls, an effect that was already detectable at 1 hpm and was still evident at 4 hpm across multiple experiments (Fig. 1A,B). In mated females we detected a strong band at 46 kDa and a much weaker band at 52 kDa, similar to reports in An. stephensi 42  www.nature.com/scientificreports/ response was tissue-specific as we did not detect it in other tissues including the reproductive tract (ovaries, atrium and spermatheca) or the rest of the body (Fig. 1C, Supplementary Fig. S1). Moreover, while levels of pJNK increased in the heads of mated females (one sample t test, p = 0.0496, n = 4), levels of the phosphorylated (active) form of extra-cellular signal-regulated kinase (pERK), a second member of the MAP kinase family to www.nature.com/scientificreports/ which JNK belongs, albeit one traditionally associated with growth stimuli 43 , were reduced in mated heads relative to virgin controls (one sample t test, p = 0.02, n = 4) ( Supplementary Fig. S1), suggesting a level of pathway specificity.

JNK1 is required for mating-induced oviposition.
To determine the contribution of the JNK pathway to post-mating responses, we silenced elements of the pathway by RNAi and examined the effect on oviposition, a physiological response that is induced in mated females once eggs are fully developed after blood feeding. Given the annotation of two distinct An. gambiae JNK genes, JNK1 (AGAP029555) and JNK3 (AGAP009460), we performed preliminary experiments to assess the relative expression of each transcript and found that JNK1 transcript levels were 2-3 orders of magnitude more abundant than JNK3 in all tissues measured (Supplementary Fig. S2) leading us to focus on the JNK1 gene product. Thus, virgin females were injected with dsRNAs targeting JNK1 (dsJNK1, Supplementary Fig. S3) or its two transcription factor targets, Jun (dsJun) and Fos (dsFos), using dsGFP as control. Using a previously established protocol 12 , injected females were then blood-fed and mated, and oviposition rates were measured. JNK1 silencing inhibited the mating-induced increase pJNK in the head ( Supplementary Fig. S4) and significantly reduced oviposition compared to control females (logistic regression model, p < 0.0001, Fig. 1D). The failure to oviposit after mating was 5.6-fold more likely in dsJNK1treated females than in dsGFP-treated controls (odds ratio [OR] = 5.6; p < 0.0001). Similarly increased rates of oviposition failure were observed with dsFos (OR = 4.03, p = 0.0003) and dsJun (OR = 6.2, p = 0.0002) (Fig. 1D).
We observed no significant effect on the total number of eggs developed or the number of eggs oviposited (Supplementary Fig. S5) in any group, and similarly, gene silencing did not affect the ability of females from any of the groups to mate, as assessed by the detection of mating plugs in the atrium ( Fig. S6). Together, these www.nature.com/scientificreports/ data suggest the involvement of the JNK pathway in the mating-induced cascades leading to oviposition in An. gambiae.

Depletion of the JNK-phosphatase puckered is sufficient to induce oviposition in virgins.
Having established a role for the JNK pathway in mating-induced oviposition, we went on to determine whether an increase in the level of pJNK in the head might be sufficient, per se, to induce oviposition in blood-fed virgin females. The activation of JNK is regulated by the dual phosphorylation of its TxY motif by the MAP kinase kinase (MAP2K) Hemipterous, and is prevented by dephosphorylation of the same motif by the dual specificity phosphatase Puckered (Puc), also known as MAP kinase phosphatase 5 (MKP5). Because depletion of puc in Drosophila leads to the spontaneous activation of JNK-responsive genes 44 , we reasoned that silencing of this phosphatase might mimic the activation of JNK noted after mating. Indeed, puc silencing ( Supplementary  Fig. S3) increased pJNK levels relative to controls in the female head across multiple experiments ( Fig. 2A,B), while no noticeable effects were observed in the reproductive tract of the same females suggesting tissue-specific activation (Fig. 2C). Moreover, a significant proportion of virgin females (logistic regression model, p < 0.0001) injected with dspuc laid eggs after blood feeding (Fig. 2D). Using the logistic regression model applied above, we determined that dspuc virgins were approximately 25-fold more likely to oviposit than controls (OR 24.6, p < 0.0028). These results are consistent with our findings that JNK activation in the head after mating induces oviposition. While in Drosophila Puc acts selectively to regulate the JNK pathway 44 , a recent study in An. stephensi suggested that the same phosphatase can impact MAP kinases other than JNK 42 . To address the role of JNK1 in the oviposition induced by dspuc, we performed a double knock down of puc and JNK1. Consistent with a predominant role for JNK signaling in regulating this behavior, coinjection of dspuc with dsJNK1 significantly reduced the frequency of oviposition induced by dspuc alone (OR relative to dspuc alone 3.7, p = 0.021) without affecting the efficiency of puc knock down ( Supplementary Fig. S3). These data suggest that activation of JNK in the head of a virgin female is sufficient to induce oviposition. www.nature.com/scientificreports/ The JNK pathway is required for oviposition induced by exogenous 20E. Since thoracic delivery of exogenous 20E is sufficient to induce oviposition in blood-fed virgin females 12 , we next assessed whether injected 20E-like mating (Fig. 1A,B)-might also induce an increase in pJNK levels in the head. When compared to controls, a robust increase in pJNK in the head was observed across multiple experiments 1-2 h post injection (hpi) (Fig. 3A,B), while the reproductive tract again showed no JNK activation at these time points (Fig. 3C).
To determine if 20E-induced oviposition occurs via JNK signaling, we silenced JNK1, Jun or Fos and allowed females to take a blood meal. After completion of egg development, females were injected with either 20E or a solvent control, and oviposition rates were measured. The number of females failing to oviposit after 20E-injection increased markedly after treatment with dsJNK1 (logistic regression model, p < 0.0001), compared to dsGFPtreated control females (OR of failed oviposition compared to dsGFP control = 5.4, p < 0.0001). Similar results were obtained after silencing of Jun (OR 5.8, p < 0.0001) and Fos (OR 4.1, p = 0.0006) (Fig. 3D). These data support the involvement of the JNK pathway in oviposition induced by sexual transfer of the male steroid hormone 20E. Injection of a solvent control failed to induce oviposition in any dsRNA group (94-100% failed oviposition, see Supplementary Table S1), as expected 12 .

Discussion
The transfer of the ecdysteroid hormone 20E from An. gambiae males to females during copulation is linked to a life-long license to oviposit eggs developed across multiple gonotrophic cycles 10,12 . Here, we identify the JNK pathway, traditionally associated with responses to diverse environmental stressors including wounding 27 and Plasmodium infection 24,25 , as a key component of the downstream events linking sexual transfer of 20E to oviposition. We document a rapid increase in the amount of active pJNK in the heads of mated females that is recapitulated by injection of exogenous 20E. The activation of JNK in the head after mating appears pathway-specific www.nature.com/scientificreports/ since ERK, an independently-regulated, growth-associated member of the MAP kinase family to which JNK belongs, is not activated in the same tissue. Moreover, JNK itself is not activated by mating in other tissues studied (reproductive tract or rest of body). The lack of a detectable JNK response to mating in the reproductive tract tallies with the absence of an obvious immune-or wounding-related gene signature in our previous microarray studies of mating-regulated genes in that tissue 12,13,26 , in contrast to the striking immune signature and copulatory wounding reported in Drosophila reproductive tract after mating 45,46 . We find that the activation of the JNK pathway in the head is both necessary and sufficient for at least one element of the post-mating response-the oviposition of developed eggs. RNAi-mediated depletion of JNK1 inhibited the mating-induced pJNK signal (Supplementary Fig. S4) and reduced oviposition induced either by mating (Fig. 1D) or by the injection of exogenous 20E (Fig. 3D). The link between this reduced egg laying phenotype and JNK function is highlighted by the fact it was phenocopied by depletion of Jun or Fos, transcription factor targets of pJNK. Moreover, activation of the JNK pathway-by RNAi-induced depletion of puc-was sufficient to both increase pJNK levels in the head ( Fig. 2A,B) and trigger oviposition in blood-fed-virgin females (Fig. 2D). Importantly, while recent data have shown that in some settings Puc is capable of modulating MAP kinase pathways other than JNK 42 , our data show that the induction of oviposition by depletion of puc is largely reversed by concomitant depletion of JNK1, identifying this MAP kinase as the dominant Puc target in regulating this phenotype (Fig. 2D).
Despite the 100-1,000-fold greater transcript abundance of JNK1 ( Supplementary Fig. S2), we cannot exclude a contribution to the oviposition phenotype from the JNK3 gene product, given the 70% identity between the two over the mRNA sequence targeted by dsJNK1. Similarly, the origin and importance of the two protein bands observed using the pJNK antibody remain unclear. They could reflect the products of the JNK1 and JNK3 genes as suggested by others 42 but might equally represent alternative post-translationally modified forms of the same gene product. Future studies using specific antibodies or gene targeting strategies will help address these questions.
Given the partial nature of the effects observed, our data also highlight the likely existence of JNK-independent pathways controlling oviposition, although we cannot exclude that these effects are due to incomplete gene silencing. Importantly, in the absence of tissue-specific gene knockout studies, the systemic nature of RNAi makes it difficult to exclude the possibility that JNK signaling outside the head may contribute to the regulation of oviposition, notwithstanding the head-specific activation of JNK observed after mating, 20E injection or puc depletion discussed above. The effects of the JNK pathway appear to affect oviposition rather than oogenesis, since neither the number of eggs developed or laid are affected by JNK pathway depletion ( Supplementary  Fig. S5). Also, we observed no effects on the mating ability of JNK-depleted females, although it remains to be determined whether this pathway may affect female remating frequencies.
While, to our knowledge, this is the first demonstration that the stress-responsive JNK pathway is involved in an important reproductive behavior such as oviposition, egg laying has been previously linked to stress responses in mosquitoes. Stressful stimuli including heat, dessication, starvation and infection have all been shown to impact on the timing of oviposition [47][48][49] , whereas confinement stress has been hypothesized to increase the frequency of oviposition in various anophelines 50 . Links between steroid hormones and stress responses have also been documented [51][52][53] . Of particular relevance, 20E titers have been shown to increase following stressful social interactions such as courtship 15 . This has led to the proposal that-akin to related sex steroids in mammals 54 -20E functions in insects to consolidate stress-associated memory and to drive pathways of neuronal remodeling underpinning the development of appropriate adaptive behaviors 14 .
Although the mating-mediated mechanisms controlling egg laying remain largely elusive, the data presented here provide clear evidence of collaboration between 20E and the JNK pathway to induce oviposition. In fact, such collaboration is well documented during larval/pupal metamorphosis in Drosophila, where 20E and the JNK pathway together drive waves of apoptosis which control the reshaping or destruction of obsolete larval tissues 55 as well as the remodeling of neuronal connections in the larval brain by axon and dendrite pruning 56 . It is plausible that an irreversible process of this type might regulate the lifelong behavioral changes induced by mating in An. gambiae. In future studies, it will be important to examine whether other elements of the post mating response, such as mating refractoriness, show a similar requirement for the JNK pathway. In addition, given the importance of the JNK pathway in the An. gambiae immune response to some Plasmodium infections 24,25 and the centrality of 20E to mosquito physiology and parasite development 57 , it will be informative to examine whether 20E signaling in other contexts (e.g. after blood feeding) also engages with the JNK pathway.

Methods
Rearing of Anopheles gambiae mosquitoes. An. gambiae mosquitoes of the G3 line were maintained at 28 °C, 70% relative humidity with a 12 h light/ dark diurnal cycle and water and 10% glucose solution ad libitum and fed weekly on human blood obtained from the blood bank of the Servizio Immunotrasfusionale, Ospedale Santa Maria della Misericordia di Perugia. For mating experiments, males and females were maintained as virgins by separation of male and female pupae by microscopic examination of the terminalia and kept in separate cages until sexual maturity (3 days post eclosion) or as required for mating experiments (see below). oviposition experiments. Three-day old virgin females were injected (Nanoject II, Drummond Scientific/Olinto Martelli Srl, Florence, Italy) with 0.69 μg (138 nl, 5 μg/μl double stranded RNAs [dsRNAs, see below]) targeting JNK pathway components (JNK1, Jun, Fos, or puckered [puc]) or GFP, a gene not expressed in these mosquitoes, as a negative control. In experiments in which two genes were targeted simultaneously (e.g. Fig. 2), a mixture of the two dsRNAs, in which the concentration of each was 5 μg/μl, was prepared and injected in the same volume (138 nl) as a single dsRNA treatment. Assignment of mosquitoes to different groups was random but groups were not blinded. Using oviposition protocols described previously 12  www.nature.com/scientificreports/ 3 days post-injection, and 2 days later (upon completion of egg development), were induced to oviposit developed eggs either by mating (see below) or by injection of exogenous 20E (Sigma, Milan, Italy; see below). For mating-induced oviposition, females were added in groups of 10-20 to cages of 150-200 age-matched males, captured in copula and successful mating was certified using fluorescent microscopy to verify the presence of an auto-fluorescent mating plug. Next day, mated females were placed in individual oviposition cups as described previously 12 . For oviposition induced by exogenous 20E, two days post blood feeding, females were injected (138 nl) with 20E (38 mM, equivalent to 2.5 ng) dissolved in H 2 O containing 10%EtOH and 5% DMSO or with a solvent control (the same diluent minus 20E) then next day placed in individual oviposition cups. Since oviposition, in our hands, routinely takes place 2 days after mating or 20E injection 12 , females were checked every day for four days and oviposition was deemed to have occurred if any eggs were detected in the oviposition cup. Females who died before ovipositing and any females who failed to develop eggs were excluded on the assumption that they had not taken a blood meal. Differences between groups in the frequency of oviposition were assessed using a generalized linear model based on the binary outcome 'oviposition or no oviposition' by day 4 post mating or 20E injection or day 5 post blood feeding in the case of dspuc treatment. 'Experiment' was included as a variable in the model and experiments statistically distinguishable from the others were removed as outliers. In the case of mating-induced oviposition this led to the removal of 2 of 11 experiments while in the case of dspuc-or 20E-induced oviposition no outliers were identified. The model was used to calculate the odds ratios (OR) of the relative likelihood of oviposition, or failed oviposition, in each group as well as the statistical significance of inter-group differences.
Microarray experiments and analysis. Using protocols described previously 12 heads from 15 mated or age-matched virgin control females were dissected in ice-cold PBS at 3 h and 24 h post mating and immediately transferred to Tri Reagent. Four independent biological replicates were performed on different generations of the same mosquito line (G3). Total RNA was recovered by phenol/chloroform extraction, DNase-treated to remove genomic DNA and quantified using a NanoDrop spectrophotometer (Thermofisher). Using a one-color labelling strategy, RNA (100 ng) from each of the four replicates was labelled using a Low Input Quick Amp Labeling kit (Agilent, Stockport, UK) following protocol G4140-90040. Labelled RNA was hybridised to 44-K An. gambiae whole genome microarrays (Design ID G2519F-020449). Labelling, hybridization and scanning were performed by the Institute of Genetics and Molecular and Cellular Biology (Illkirch, France). Microarray datasets were analyzed using the R statistical software environment (version 2.15.0) running the Linear Models for Microarray Data (Limma) package (version 3.14.4) 58 . Dye signals were background-corrected using the normexp method with an offset of 16 and normalized across microarrays using the quantile method. Multiple probes for the same transcript identifier were collapsed to individual genes, and average fold-change results were generated for each unique array identifier. Package ArrayQualityMetrics 59 was used for quality control of all microarrays before significance was estimated by fitting a linear model to each gene across replicated arrays, applying a contrast matrix of comparisons of interest in the mating-array experiment, and determining an empirical Bayes moderated t statistic. The decideTests function for multiple testing across genes and contrasts ("global strategy") was also used to classify the related t statistics as up, down, or not significantly regulated. P values then were corrected for multiple testing by the Benjamini-Hochberg method 60 . Results were exported as Microsoft Excel files, and transcript identifiers [Ensembl Gene ID (AGAP) numbers or ESTs] with adjusted P < 0.05 were selected for further analysis. When possible, ESTs were identified by using manual submissions of Blastn 61 to the An. gambiae PEST strain genome or were classified as unknown. Gene function annotation was assigned via VectorBase gene description, AnoXcel summary 62 , and/or orthology.
Gene expression analysis by qRT-PCR. Three-day-old females were injected with dsGFP or dsJNK and two days later mated with age-matched males, as indicated in the figure legend and described above (oviposition experiments). Gene expression was assessed by qRT-PCR as described previously 13 . Briefly, dissected tissues were recovered directly to 10 μl of RNA-later (Ambion), immediately supplemented with 250 μl Tri-reagent and homogenized using a motorized pestle. RNA in clarified supernatants (14,000×g, 15 min, 4 °C) was extracted, DNase digested and eluted using Direct-zol RNA miniprep columns (Zymo Research/Euroclone, Milan, Italy) according to the manufacturer's instructions. Some of this material (0.5-1 μg) was reverse transcribed to cDNA as described in detail elsewhere 12 . Triplicate 5 μl aliquots were analysed using Fast SybrGreen master Mix (Thermofisher) and the forward and reverse primers listed in Supplementary Table S2. Reactions were run on a QuantStudio 3 thermocycler (Thermofisher). For each sample, the mean number of cycles to cross threshold (CT) obtained for a the gene of interest was expressed relative to that of a reference gene (Ribosomal protein L19, Rpl19; AGAP004422) whose expression is insensitive to mating 13 and blood feeding 65 (delta CT). Delta CT values were used to calculate the relative expression of the gene of interest using the formula: relative expression = 2 (−delta CT) . On occasion the relative expression in one group (e.g. virgin females) was expressed relative to that in another group (e.g. mated females) as fold-change in expression (delta delta CT). Effects of RNAi-mediated JNK1 depletion on mating-induced expression changes were analyzed by two-way ANOVA using relative expression data (2 -delta CT ). Efficiency of knock down of target genes was calculated using delta delta CT values from dstarget vs dsGFP controls and analyzed using a one sample t-test for deviation from a hypothetical value of 100%.
Statistical analysis. All Statistical analyses were performed using Prism 8.0 (GraphPad, La Jolla, USA) except logistic regression analyses which were performed using JMP version 14 (SAS, Cary, USA). Non-parametric methods were used unless the normality of all groups analyzed could be demonstrated using a Kolmogorov-Smirnov test. In all tests, a significance cut-off of p = 0.05 was applied. Comparisons of multiple treatment groups to a single 'control' group were corrected using Dunnett's (for parametric tests) or Dunn's (for nonparametric tests) multiple comparison correction. If all possible intergroup comparisons were made, Tukey's correction was used. On occasion (Supplementary Figs. S1, S2) the effect of a manipulation was assessed using a one sample t-test. In these cases, the normality of all groups was verified using a Kolmogorov-Smirnov test ( Supplementary Fig. S2) or a Shapiro-Wilk test ( Supplementary Fig. S1).