Abstract
The C. elegans insulin-like signaling network supports homeostasis and developmental plasticity. The genome encodes 40 insulin-like peptides and one receptor. Feedback regulation has been reported, but the extent of feedback and its effect on signaling dynamics during a state transition has not been determined. We measured mRNA expression for each insulin-like peptide, the receptor daf-2, components of the PI3K pathway, and its transcriptional effectors daf-16/FoxO and skn-1/Nrf at high temporal resolution during transition from a starved, quiescent state to a fed, growing state in wild type and mutants affecting daf-2/InsR and daf-16/FoxO. We also analyzed the effect of temperature on insulin-like gene expression. We found that numerous PI3K pathway components and insulin-like peptides are affected by signaling activity, revealing pervasive positive and negative feedback regulation. Reporter gene analysis demonstrated that the daf-2/InsR agonist daf-28 positively regulates its own expression and that other agonists cross-regulate daf-28 transcription through feedback. Our results show that feedback regulation of insulin-like signaling is widespread, suggesting a critical role of feedback in signaling dynamics in this endocrine network and likely others.
Introduction
Insulin-like signaling maintains homeostasis by responding to fluctuations in nutrient availability and altering gene expression. Work in C. elegans has shown that insulin-like signaling also allows developmental plasticity. For example, insulin-like signaling regulates whether larvae become reproductive or arrest as dauer larvae, a developmental diapause that occurs in unfavorable conditions (Hu, 2007). Insulin-like signaling also contributes to continuous variations in phenotype, for example in regulation of aging and growth rate (Murphy and Hu, 2013). However, it is unclear how signaling dynamics are regulated such that the pathway can maintain a phenotypic steady-state (homeostasis) or promote developmental plasticity, depending on conditions.
Insulin-like signaling is regulated by feedback in diverse animals. Pancreatic β-cell-specific insulin receptor-knockout mice are poor at glucose sensing, have a diminished insulin secretory response, and tend to develop age-dependent diabetes (Otani et al, 2004). In addition, the full effect of glucose on pancreatic β-cells grown in vitro requires the insulin receptor (Assmann et al, 2009). FoxO transcription factors, effectors of insulin signaling, activate transcription of insulin receptors in Drosophila and mammalian cells (Puig and Tjian, 2005), suggesting a relatively direct, cell-autonomous mechanism for feedback regulation. However, evidence for such direct feedback regulation has not been found in C. elegans (Kimura et al, 2011).
Insulin-like signaling regulates the expression of insulin-like peptides in C. elegans, suggesting a relatively indirect, cell-nonautonomous mechanism for feedback regulation. The C. elegans genome encodes a family of 40 insulin-like peptides that can function as either agonists or antagonists of the sole insulin-like receptor daf-2 (Pierce et al, 2001). Systematic analyses of insulin-like peptide expression and function suggest substantial functional specificity rather than global redundancy (Fernandes de Abreu et al, 2014; Ritter et al, 2013). daf-2/InsR signals through a conserved phosphoinositide 3-kinase (PI3K) pathway to antagonize the FoxO transcription factor daf-16 (Fig. 1A; Murphy and Hu, 2013). daf-16/FoxO represses transcription of the daf-2 agonist ins-7, creating positive feedback (Murphy et al, 2003). This positive feedback results in “FoxO-to-FoxO” signaling, which has been proposed to coordinate the physiological state of different tissues in the animal (Alic et al, 2014; Murphy et al, 2007; Zhang et al, 2013). daf-16 also activates transcription of the daf-2 antagonist ins-18, again producing positive feedback (Matsunaga et al, 2012a; Murphy et al, 2003). Insulin-like peptide function has been reported to affect insulin-like peptide expression (Fernandes de Abreu et al, 2014; Ritter et al, 2013), consistent with feedback regulation. To the best of our knowledge, negative feedback regulation has not been reported, despite the fact that homeostasis generally relies on it (Cannon, 1929). Furthermore, the extent of feedback regulation, and whether it is positive or negative with respect to pathway activity, is unknown.
We sought to determine the extent of feedback regulation in insulin-like signaling in C. elegans. C. elegans larvae that hatch in the absence of food arrest development in the first larval stage (“L1 arrest” or “L1 diapause”), and insulin-like signaling regulates L1 arrest and development (Baugh, 2013). We performed a genetic analysis of gene expression, measuring expression of all 40 insulin-like peptides as well as components of the PI3K pathway in daf-2/InsR and daf-16/FoxO mutants, which have perturbed signaling activity. We analyzed larvae in L1 arrest and over time after feeding, as they transition from quiescence to growth. The rationale is that by identifying genes whose expression is affected by insulin-like signaling that themselves affect signaling activity we can infer feedback regulation. We report extensive feedback, both positive and negative, acting relatively directly at the level of the PI3K pathway and also indirectly via regulation of peptide expression. This work suggests that feedback regulation of insulin-like signaling is pervasive and that this feedback functions to stabilize signaling activity during constant conditions while allowing rapid responses to new conditions.
Results
daf-2/InsR acts through daf-16/FoxO to affect gene expression
We used the NanoString nCounter platform to measure expression of genes related to insulin-like signaling in fed and starved L1 larvae at high temporal resolution during the transition between developmental arrest and growth (Malkov et al, 2009). Total RNA was prepared from whole worms and hybridized to a codeset containing probes for all 40 insulin-like genes as well as components of the PI3K pathway and sod-3, a known DAF-16/FoxO target. In addition to wild type (WT), we analyzed mutations affecting daf-2/InsR and daf-16/FoxO to ascertain the effects of insulin-like signaling activity on expression. We used the reference allele of daf-2, e1370, as well as a stronger allele, e979 (Gems et al, 1998). We used a null allele of daf-16, mgDf47, as well as a daf-16(mgDf47); daf-2(e1370) double mutant to analyze epistasis. Mutations affecting daf-2 are generally temperature sensitive, and insulin-like signaling responds to temperature. We therefore measured expression during L1 starvation at three different temperatures. We also fed bacteria to starved L1 larvae of each of the five genotypes and measured gene expression over time during recovery from arrest in a highly synchronous population (Fig. 1B). This experimental design enabled us to measure the effects of temperature, nutrient availability, and insulin-like signaling activity on genes related to insulin-like signaling itself during a critical physiological state transition.
daf-16/FoxO mediates the effects of daf-2/InsR on expression of genes involved in insulin-like signaling. daf-16 is required for canonical effects of daf-2, such as dauer formation and lifespan extension (Hu, 2007; Murphy and Hu, 2013). However, daf-2 also acts through other effector genes of the PI3K pathway, such as skn-1/Nrf (Tullet et al, 2008), as well as other signaling pathways, such as RAS (Nanji et al, 2005). In addition, genome-wide expression analyses of daf-16 have mostly been performed in a daf-2 mutant background (daf-2 vs. daf-16; daf-2) without analysis of WT and/or daf-16 single mutants (Tepper et al, 2013), making analysis of epistasis between daf-2 and daf-16 with gene expression as a phenotype impossible. Since epistasis was not analyzed, these studies could not determine whether daf-16 mediated all of the effects of daf-2 on gene expression or if other effectors made a significant contribution. A correlation matrix between genotypes over all conditions tested indicates that mutating daf-2 affected expression, with a stronger effect of the e979 allele than e1370, as expected (Fig. 1C). daf-16 also had a clear effect, and it was epistatic to daf-2. That is, the expression profile of the double mutant is similar to that of the daf-16 single mutant but not daf-2. Statistical analysis of individual genes together with examination of expression patterns across genotypes corroborated the results of correlation analysis, failing to identify genes with significant effects of daf-2 not meditated by daf-16. These results show that daf-2 affects expression of genes involved in insulin-like signaling and that these effects are mediated exclusively by daf-16, consistent with feedback regulation.
daf-16/FoxO affects expression of multiple PI3K pathway genes
We analyzed expression of several components of the PI3K pathway, as well as daf-2/InsR and its transcriptional effectors daf-16/FoxO and skn-1/Nrf (Lin et al, 1997; Ogg et al, 1997; Tullet et al, 2008). The known direct target of DAF-16, sod-3/SOD (Oh et al, 2006), was up-regulated in daf-2 mutants and down-regulated in the daf-16 mutant, with daf-16 epistatic to daf-2, in both starved and fed larvae (Fig. 2, S1 and Table 1). The exemplary behavior of this positive control demonstrates the power of our experimental design. Notably, daf-16 expression drops to background levels in the daf-16 deletion mutant (Fig. 2 and S1), as expected. We previously reported that daf-2 is up-regulated during L1 arrest (Chen and Baugh, 2014). We see here that daf-2 is actually repressed by daf-16 (Fig. 2 and S1). Given that daf-2 is up-regulated during starvation, when daf-16 is active, this result may be considered paradoxical. Our interpretation is that daf-2 expression is independently regulated by nutrient availability and daf-16 in opposing ways, illustrating regulatory complexity of the system. Nonetheless, since DAF-2 antagonizes DAF-16 activity via the PI3K pathway, these results indicate positive feedback between the sole insulin-like receptor and its FoxO transcriptional effector (Table 1). Likewise, age-1/PI3K, which transduces daf-2 signaling activity, was repressed by daf-16, also suggesting positive feedback. However, pdk-1/PDK, akt-1/Akt and akt-2/Akt, downstream components of the PI3K pathway, were each activated by daf-16, albeit with relatively complex dynamics, suggesting negative feedback. Likewise, daf-16 expression is reduced in daf-2 mutants (Fig. 2), where its activity is increased, suggesting it represses its own transcription to produce negative feedback (Table 1). skn-1/Nrf expression was also reduced in daf-2 mutants and increased in daf-16 mutants, suggesting that insulin-like signaling positively regulates expression of both of its transcriptional effectors. Notably, the effects described here for each gene were consistent for fed and starved larvae (Fig. 2, S1 and Supp. Data File 2). In summary, insulin-like signaling acts through daf-16/FoxO to regulate multiple critical components of the pathway itself, consistent with a combination of positive and negative cell-autonomous feedback regulation.
daf-16/FoxO affects expression of most insulin-like peptides
Insulin-like genes display complex dynamics in response to different levels of insulin-like signaling activity. Our codeset contained probes for all 40 insulin-like genes, and we reliably detected expression for 28 of them. Similar to what we saw with components of the PI3K pathway (Fig. 2, S1 and Table 1), daf-16 appears to function as an activator in some cases and a repressor in others (Fig. 3, S2 and Table 1), but its function with respect to each gene affected was again consistent between fed and starved conditions (Fig. 3, S2 and Supp. Data File 2). For example, expression of daf-28, perhaps the most studied insulin-like peptide in C. elegans (Chen and Baugh, 2014; Cornils et al, 2011; Fernandes de Abreu et al, 2014; Hung et al, 2014; Li et al, 2003; Patel et al, 2008), was up-regulated in daf-16 mutants and down-regulated in daf-2 in starved and fed larvae (Fig. 3, S2 and Table 1), suggesting it is repressed by daf-16. Remarkably, all but three of the 28 reliably detected insulin-like genes were significantly affected by daf-16 (Table 1). Mutation of daf-16 caused up-regulation of twelve insulin-like genes and down-regulation of thirteen, suggesting that daf-16 directly or indirectly regulates transcription of most insulin-like genes.
Inference of feedback as positive or negative is complicated by the fact that individual insulin-like peptides function as either agonists or antagonists of daf-2/InsR (Pierce et al, 2001). Biochemical data and structural modeling suggest that function as an agonist or antagonist is a property of the peptide (Matsunaga et al, 2018), as opposed to the context in which it is expressed. To infer whether the net effect of feedback regulation is positive or negative with respect to insulin-like signaling activity (daf-2/InsR activity), we took into account whether daf-16 appears to activate or repress the insulin-like gene and whether that gene encodes a putative agonist or antagonist. DAF-2 antagonizes DAF-16 activity, and so daf-16 repression or activation of an agonist or antagonist, respectively, would hypothetically result in positive feedback. daf-16 repression or activation of an antagonist or agonist, respectively, would hypothetically result in negative feedback. For example, daf-28 was originally identified on the basis of its constitutive dauer-formation phenotype. daf-28 is up-regulated in rich conditions and it promotes dauer bypass (reproductive development), similar to daf-2/InsR, consistent with function as an agonist of daf-2 (Li et al, 2003). daf-16 repression of daf-28 expression therefore suggests positive feedback in this case (Table 1).
A number of studies have performed genetic analysis of insulin-like peptide function, determining whether individual insulin-like genes have similar or opposite loss-of-function phenotypes to daf-2, and thus whether they presumably function as agonists or antagonists, respectively (Chen and Baugh, 2014; Cornils et al, 2011; Fernandes de Abreu et al, 2014; Hung et al, 2014; Kawano et al, 2006; Li et al, 2003; Matsunaga et al, 2012a; Matsunaga et al, 2012b; Michaelson et al, 2010; Patel et al, 2008; Pierce et al, 2001). When we previously analyzed expression of insulin-like peptides in starved and fed L1 larvae, we found remarkable concordance between function (agonist or antagonist) and expression (positive or negative effect of food, respectively) (Chen and Baugh, 2014). Out of thirteen insulin-like peptides consistently found to function as putative agonists or antagonists based on genetic analysis, we classified all thirteen the same way based on expression, while classifying eight additional peptides as well. This classification relied on separate time-series analyses of starved and fed larvae (Chen and Baugh, 2014), and inspection of the fed time series here did not reveal discrepancies between the two studies. We therefore included our previous putative functional classifications based on nutrient-dependent expression in Table 1, which tentatively assigns function to all but two of the 25 genes affected by daf-16. As explained above, putative agonists repressed by daf-16, like daf-28, hypothetically result in positive feedback, since daf-2 signaling antagonizes daf-16. We identified seven genes like this in addition to daf-28. Conversely, activation of a putative antagonist should also produce positive feedback, which we infer in five cases, while activation of an agonist should produce negative feedback, which we infer in six cases. Finally, repression of a putative antagonist should produce negative feedback, which we infer in four cases. In summary, activation and repression of putative agonists and antagonists by daf-16 is common, with positive and negative feedback hypothetically resulting from each different regulatory combination in multiple instances.
Temperature affects insulin-like gene expression
We analyzed expression of insulin-like genes at 15, 20 and 25°C during L1 starvation. daf-2 mutants are generally temperature-sensitive (Gems et al, 1998), daf-16 is localized to the nucleus at high temperatures (Henderson and Johnson, 2001), and daf-2 mutants are heat-resistant (Munoz and Riddle, 2003). These observations suggest that insulin-like signaling responds to temperature. We hypothesized that temperature sensitivity results from temperature-dependent regulation of insulin-like peptide expression. Consistent with daf-16 being active at elevated temperature, expression of its direct target sod-3 was positively affected by temperature (Fig. S2 and Table S1). In support of our hypothesis, temperature affected mRNA expression of 21 out of 28 reliably detected insulin-like genes (Fig. S2 and Table S1). daf-28 expression was lower at higher temperatures, consistent with its role in promoting dauer bypass (Li et al, 2003), and confirmed in a recent publication (O’Donnell et al, 2018). Expression of twelve insulin-like genes was lower at higher temperatures and nine were expressed higher at higher temperatures. However, there is no apparent correlation between putative function as agonist or antagonist and positive or negative regulation in response to higher temperature. Notably, although most insulin-like genes displayed significant temperature-dependent expression, the effect of temperature on expression was minor compared to nutrient availability.
Feedback mediates cross-regulation among insulin-like genes
Reporter gene analysis validated the effect of daf-16/FoxO on daf-28 expression. We previously used quantitative RT-PCR to validate the nCounter approach to measuring insulin-like gene expression in C. elegans (Baugh et al, 2011), and we used transcriptional reporter genes to confirm positive regulation of several putative agonists in fed larvae, including daf-28 (Chen and Baugh, 2014). A Pdaf-28::GFP transcriptional reporter gene again confirmed up-regulation in response to feeding (Fig 4A). Expression was evident but faint in anterior neurons and posterior intestine of starved L1 larvae, and it was brighter after being fed for 6 hr. Quantification of whole-animal fluorescence with the COPAS BioSorter provided robust statistical support for qualitative observations (Fig. 4B). Note that the statistics for this analysis were performed on the means of individual biological replicates, as opposed to each individual in a replicate. Thus, statistical significance is due to reproducibility despite relatively small effect sizes. Critically, expression appeared elevated in daf-16 mutants compared to WT, in both starved and fed larvae (Fig. 4A). However, we did not observe a difference in the anatomical expression pattern in daf-16 compared to WT. Quantification showed that the effect of daf-16 is statistically significant (Fig. 4B). Notably, the effect of food was larger than that of daf-16, as expected based on nCounter results (Fig. 3). In addition, the effects of food and daf-16 are independent, suggesting that up-regulation of daf-28 in response to feeding is not simply due to inhibition of daf-16 leading to de-repression of daf-28. These results support the conclusion that daf-16 represses daf-28 transcription, consistent with feedback regulation.
Widespread feedback regulation of insulin-like signaling via transcriptional control of insulin-like peptides suggests that activity of individual insulin-like genes should affect expression of themselves and others. We analyzed expression of Pdaf-28::GFP in insulin mutants to test this hypothesis. Pdaf-28::GFP transgene expression was significantly reduced in a daf-28 mutant (Fig. 4A,B). This result suggests that positive feedback mediated by daf-16 repression of daf-28, daf-28 agonism of daf-2/InsR, and daf-2 inhibition of daf-16 results in daf-28 effectively promoting its own expression. daf-28, ins-4 and ins-6 coordinately regulate dauer entry and exit (Cornils et al, 2011), and they redundantly promote L1 development in response to feeding (Chen and Baugh, 2014). ins-4, 5 and 6 are in a chromosomal cluster, so we analyzed a deletion allele that removes all three (Hung et al, 2014). Pdaf-28::GFP expression was significantly reduced in fed larvae of the ins-4, 5, 6 mutant compared to WT (Fig. 4A,B). This result suggests that feedback regulation results in cross-regulation among insulin-like peptides such that the function of one peptide affects the expression of others. Compound mutants affecting ins-4, 5, 6 and daf-28 grow slowly as fed L1 larvae and display starvation resistance during L1 arrest (Chen and Baugh, 2014), and Pdaf-28::GFP expression was also reduced consistent with these phenotypes (Fig. 4B). In summary, reporter gene analysis suggests physiological significance of feedback regulation, consistent with function of individual insulin-like peptides affecting expression of others.
Discussion
We determined the extent of feedback regulation of insulin-like signaling in C. elegans in starved and fed L1 larvae. We show that mRNA expression of nearly all detectable insulin-like genes is affected by insulin-like signaling activity, revealing pervasive feedback regulation. We also show that several components of the PI3K pathway, including daf-2/InsR and daf-16/FoxO, are affected by signaling activity. Together these results suggest that feedback occurs inter- and intra-cellularly (Fig. 4C). Furthermore, we show that feedback is positive and negative at both levels of regulation. Finally, we demonstrate that feedback regulation results in auto- and cross-regulation of insulin-like gene expression.
We detected substantially more regulation of insulin-like genes by daf-16/FoxO than previously reported in genome-wide expression analyses. We also detected extensive effects of temperature on insulin-like gene expression. In contrast to other expression analyses, our analysis employed highly synchronous populations of larvae, improving sensitivity. Sensitivity was also likely improved by focusing on proximal effects of nutrient availability, which has robust effects on insulin-like signaling. In addition, the nCounter assay conditions used are optimized for sensitivity and precision (Baugh et al, 2011), improving power to detect differential expression. We also analyzed the effects of daf-16 mutation in a WT background as well as a daf-2 mutant background, in fed and starved larvae, producing four independent opportunities to detect an effect of daf-16. Finally, we sampled extensively, not only with biological replicates, but also with three different temperatures during L1 arrest as well as nine time points after feeding. Taken together, these features likely explain why we detected such extensive effects.
Other nutrient-dependent pathways also regulate expression of insulin-like genes and PI3K pathway components. That is, insulin-like signaling does not account for all of the observed effects of nutrient availability on gene expression (Fig. 4C). For example, we show that daf-28 expression is up-regulated in response to feeding and that it is repressed by daf-16/FoxO. Since DAF-16 is nuclear and active during starvation and is excluded from the nucleus in response to feeding (Henderson and Johnson, 2001), it is conceivable that up-regulation of daf-28 in response to feeding is due to inactivation of DAF-16 and de-repression of daf-28. However, this model predicts that daf-28 expression should be equivalent in starved and fed daf-16 mutant larvae, but it is not. To the contrary, induction of daf-28 in fed larvae occurs with similar magnitude in each genotype tested. This was true with mRNA expression analysis by nCounter as well as transcriptional reporter gene analysis. Despite numerous examples of daf-2 and daf-16 affecting expression, the effects of nutrient availability are generally evident in all genotypes, indicating the influence of other nutrient-dependent pathways (Fig. 4C).
We provide evidence that daf-16/FoxO activity leads to activation and repression of genes involved in insulin-like signaling. Both modes of regulation were observed for putative daf-2/InsR agonists and antagonists, supporting the conclusion agonists and antagonists both contribute to positive and negative feedback regulation. However, we used genetic and not biochemical analysis, so we do not know if DAF-16 regulation is direct or indirect. DAF-16 is thought to function primarily as an activator (Riedel et al, 2013; Schuster et al, 2010), with repression (“class II” targets) occurring indirectly via its antagonism of the transcriptional activator PQM-1 (Tepper et al, 2013). However, a role of pqm-1 in L1 arrest and recovery has not been investigated. Nonetheless, akt-1/Akt, akt-2/Akt, skn-1/Nrf and daf-16/FoxO were each included on a list of 65 high-confidence direct DAF-16 targets (Schuster et al, 2010). We found each of these to be regulated by daf-16, with skn-1 and daf-16 being repressed, consistent with direct repression independent of PQM-1. Mechanistic details aside, this work reveals extensive positive and negative feedback regulation of insulin-like signaling.
Insulin-like peptide function regulates expression of insulin-like genes. We used reporter gene analysis to show that function of daf-28, a daf-2 agonist repressed by daf-16, affects its own transcription. Furthermore, we showed that function of other agonists cross-regulate daf-28 transcription. These results are consistent with reports of insulin-like peptides affecting expression of insulin-like genes (Fernandes de Abreu et al, 2014; Ritter et al, 2013), though in this case we demonstrate an intermediary effect of daf-16/FoxO. Given that we found most insulin-like genes to be regulated by insulin-like signaling, cross regulation among insulin-like peptides is likely common.
We believe the physiological significance of feedback regulation is to stabilize signaling activity in variable environments. Negative feedback supports homeostasis, returning the system to a stable steady state (Cannon, 1929). In contrast, positive feedback supports rapid responses and switch-like behavior (Ingolia and Murray, 2007). We speculate that by combining negative and positive feedback, the insulin-like signaling system is able to maintain homeostasis at different set points of signaling activity. That is, in constant conditions negative feedback stabilizes signaling activity, but when conditions change (e.g., differences in nutrient availability) positive feedback allows signaling activity to respond rapidly and negative feedback helps it settle to a new steady state rather than displaying runaway dynamics. In addition, signaling occurs in the context of a multicellular animal, with tissues and organs that presumably vary in their energetic and metabolic demands. Consequently, FoxO-to-FoxO signaling resulting from feedback may be relatively positive or negative in different anatomical regions, governed by the peptides involved, serving to coordinate the animal’s physiology appropriately (Kaplan and Baugh, 2016; McMillen et al, 2002). In any case, the extent of feedback suggests that it is a very important means of regulation. We imagine that insulin-like signaling in other animals and other endocrine signaling systems are also rife with feedback, and that it is critical to system dynamics.
Materials and Methods
Nematode culture and sample collection
The following C. elegans strains were used for gene expression analysis on the NanoString nCounter platform: N2 (wild type), PS5150 (daf-16(mgDf47)), CB1370 (daf-2(e1370)), DR1942 (daf-2(e979)), GR1309 (daf-16(mgDf47); daf-2(e1370)). Strains were maintained on NGM agar plates with E. coli OP50 as food at 15°C (DR1942) or 20°C (all others). Liquid culture was used to obtain sufficiently large populations for time-series analysis with microgram-quantities of total RNA. Larvae were washed from clean, starved plates with S-complete and used to inoculate liquid cultures (Lewis, 1995). A single 6 cm plate was typically used, except with CB1370 and DR1942, for which two and three plates were used, respectively. Liquid cultures were comprised of S-complete and 40 mg/ml E. coli HB101. These cultures were incubated at 180 rpm and 15°C for four days (with the exception of DR1942, which was incubated for five days), and eggs were prepared by standard hypochlorite treatment, yielding in excess of 100,000 eggs each. These eggs were used to set up another liquid culture again consisting of S-complete and 40 mg/ml HB101 but with a defined density of 5,000 eggs/ml. These cultures were incubated at 180 rpm and 15°C for five days (N2, PS5150 and GR1309), six days (CB1370) or seven days (DR1942), and eggs were prepared by hypochlorite treatment with yields in excess of one million eggs per culture. These eggs were cultured in S-complete without food at a density of 5,000 eggs/ml at 180 rpm so they hatch and enter L1 arrest. For starved samples at 20°C and 25°C, they were cultured for 24 hr and collected, and for 15°C they were cultured for 48 hr. Fed samples were cultured for 24 hr at 20°C, and then 25 mg/ml HB101 was added to initiate recovery by feeding. Fed samples were collected at the time points indicated. Upon collection, larvae were quickly pelleted by spinning at 3,000 rpm for 10 sec, washed with S-basal and spun three times, transferred by Pasteur pipet to a 1.5 ml plastic tube in 100 µl or less, and flash frozen in liquid nitrogen. Samples were collected in at least two but typically three independent biological replicates where the entire culture and collection process was repeated.
RNA preparation and hybridization
Total RNA was prepared using 1 ml TRIzol (Invitrogen) according to the manufacturer’s instructions. 3 µg total RNA was used for hybridization by NanoString, Inc (Seattle, WA USA), as described (Chen and Baugh, 2014). The codeset used included the same probes for all insulin-like genes as in Chen, 2014 with the exception of ins-13, which was replaced here. The codeset also included probes for additional genes not included in Chen, 2014 (for a complete list of genes targeted see Supplementary Data File 1) as well as standard positive and negative control probes.
Data analysis
nCounter results were normalized in a two-step procedure. First, counts for positive control probes (for which transcripts were spiked into the hybridization at known copy numbers) were used to normalize the total number of counts across all samples. Second, the total number of counts for all targeted genes except daf-16 (the deletion mutant used did not produce signal above background) was normalized across all samples. Insulin genes with a normalized count of less than 5,000 were excluded from further analysis because they displayed a cross-hybridization pattern indicating that they were not reliably detected. The complete normalized data set is available in Supplementary Data File 1.
Statistical analysis was used to assess the effects of daf-16 (in fed and starved samples) and temperature (starved samples only). For the effect of daf-16 in fed samples, two tests were used: a non-parametric ANCOVA with the null hypothesis that loess lines connecting the points of the daf-16 single mutant (or the daf-16; daf-2 double mutant) and wild type (or daf-2(e1370)) are overlapping. This test was implemented using the R package “sm” (Bowman and Azzalini, 1997). For the effect of daf-16 in starved samples, two tests were used: a bootstrap test was used with the null hypothesis that the daf-16 single mutant (or the daf-16; daf-2 double mutant) has the same mean expression level as wild type (or daf-2(e1370)) for all temperatures. The effect size of genotype is calculated within each temperature, so it controls for temperature. 10,000 permutations of genotype were calculated to get the p-value. For the effect of temperature during starvation, a chi-squared goodness of fit test was used to ask whether temperature explained additional variance in gene expression after controlling for genotype. Benjamani-Hochberg was used to calculate the ‘q-value’ (Benjamini and Hochberg, 1995), and these q-values were used to identify genes affected by daf-16 or temperature at a false-discovery rate of 5%. The complete results of statistical analysis is available in Supplementary Data File 2.
Reporter gene analysis
The mgIs40 [Pdaf-28::GFP] reporter (Li et al, 2003) was analyzed using the following genetic backgrounds: wild type (N2), daf-16(mu86), daf-28(tm2308) and ins-4, 5, 6(hpDf761). Strains were maintained on NGM agar plates with E. coli OP50 as food at 20°C. Eggs were prepared by standard hypochlorite treatment. These eggs were used to set up a liquid culture consisting of S-basal without ethanol or cholesterol with a defined density of 1,000 eggs/ml. After 18 hours to allow for hatching, E. coli HB101 was added at 25 mg/ml to the fed samples. 6 hours post food addition, the samples were washed three times with 10 ml S-basal and then run through the COPAS BioSorter measuring GFP fluorescence. Analysis of the COPAS data was performed in R. Tukey fences were used to remove outliers. Data points were also removed if they were determined to be debris by size or lack of fluorescent signal. This cleanup left a total of almost 165,000 data points. Fluorescence data was normalized by worm density. The Bartlett test of homogeneity of variances rejected the null hypothesis that the samples had equal variance. Therefore, unpaired t-tests with unequal variance were used to determine the significance of condition and genotype on mean normalized fluorescence. There were three biological replicates for the insulin-like peptide mutants and seven biological replicates for wild type and daf-16 mutants.
For imaging, the samples were prepared in the same way then paralyzed with 3.75 mM sodium azide and placed on an agarose pad on a microscope slide. Images were taken on a compound fluorescent microscope.
Data Availability
The complete normalized data set is available in Supplementary Data File 1. Complete results of statistical analysis is available in Supplementary Data File 2. Raw data and strains used here are available upon request.
Author Contributions
LRB conceived of the study and provided funding. LRB, REWK and NKC performed the experiments. CSM and REWK analyzed the data. LRB, CSM and REWK prepared the manuscript.
Conflict of Interest
The authors have no conflicts of interest to declare.
Acknowledgements
We would like to thank NanoString, Inc. for performing nCounter hybridizations for us. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). The National Science Foundation (IOS-1120206) and the National Institutes of Health (R01GM117408) funded this work.