A functional study of all 40 Caenorhabditis elegans insulin-like peptides

The human genome encodes 10 insulin-like genes, whereas the Caenorhabditis elegans genome remarkably encodes 40 insulin-like genes. Knockout strategies to determine the roles of all the insulin/insulin-like peptide ligands (INS) in C. elegans has been challenging due to functional redundancy. Here, we individually overexpressed each of the 40 ins genes pan-neuronally, and monitored multiple phenotypes including: L1 arrest life span, neuroblast divisions under L1 arrest, dauer formation, and fat accumulation, as readouts to characterize the functions of each INS in vivo. Of the 40 INS peptides, we found functions for 35 INS peptides and functionally categorized each as agonists, antagonists, or of pleiotropic function. In particular, we found that 9 of 16 agonistic INS peptides shortened L1 arrest life span and promoted neuroblast divisions during L1 arrest. Our study revealed that a subset of β-class INS peptides that contain a distinct F peptide sequence are agonists. Our work is the first to categorize the structures of INS peptides and relate these structures to the functions of all 40 INS peptides in vivo. Our findings will promote the study of insulin function on development, metabolism, and aging-related diseases.

The Caenorhabditis elegans insulin/insulin-like growth factor signaling (IIS) 2 pathway has been extensively studied and the IIS pathway components are evolutionary conserved in metazoans (1). Insulin-like (INS) peptides bind to and activate cell-surface receptors with intrinsic tyrosine kinase activity (2, 3). Autophosphorylation of the receptors promote the recruitment and activation of downstream components to initiate their biological effects (4). The insulin superfamily genes are ubiquitous and have been identified in all animals (5,6). Compared with human and the fruit fly, Drosophila melanogaster, which have 10 and eight INS peptides, respectively, the C. elegans genome encodes 40 INS genes (5,7), suggesting that there may be more functional diversity among the C. elegans INS peptides. Strikingly, there is only one INS receptor, DAF-2/INSR, which the 40 INS peptides are thought to bind as ligands. To date, no single loss of function mutation can fully recapitulate the phenotypes associated with loss of the daf-2 insulin receptor (8). The lack of loss-of-function phenotypes for many of the INS peptides suggests that some act redundantly in C. elegans, and this makes addressing INS functions by knockout strategies challenging and limited (8,9).
In mammals (including humans), INS peptides and IIS signaling control glucose levels, hormone homeostasis, and metabolism (10). In C. elegans, IIS signaling controls aging, development, behavior, dauer formation, as well as fat accumulation (11). By using dauer formation as a readout phenotype, only INS-4, -6, and -9 and DAF-28 were identified as potential agonists, whereas INS-1, -17, and -18 were identified as potential antagonists (7,9,(12)(13)(14). Some INS peptides were suggested to be potential agonists or antagonists based on mRNA expression dynamics between fed and starved conditions (15) and the INS peptides have been shown to regulate each other transcriptionally (9). However, the roles and functions of all 40 INS peptides remain unclear. INS peptides are expressed primarily in the nervous system (7). As such, we individually overexpressed each of the 40 INS peptides in all neurons using a pan-neuronal promoter to direct gene expression (16). Overexpression (oe) lines were then characterized based on phenotypes associated with abnormal IIS signaling in C. elegans. As an example, in the absence of food, C. elegans can arrest development in the first larval stage (L1 arrest), preventing further growth and development (17). The level of INS signaling can shorten or lengthen the life span of L1-arrested animals. Here, we assayed the contribution of individual INS peptides on phenotypes associated with IIS signaling that include alterations in L1 arrest life span, Q neuroblast divisions, dauer formation, and fat metabolism. Based on these assays, seven INS peptides (INS-3, -4, -6, -9, -19, and -32 and DAF-28) were categorized as strong agonists and three INS peptides  were strong antagonists of the IIS receptor in vivo. Nine

Overexpressed INS affect L1 arrest life span
In the absence of food, newly hatched C. elegans larva (L1 stage) undergo a developmental quiescence called L1 arrest. Previously, we and others have shown that down-regulation of the IIS pathway is critical for L1 arrest survival (18,19). WT L1 arrested worms live for a maximum of 21 days with a mean life span of 13 days when grown at 20°C. Manipulation of IIS signaling can alter the normal 21 day survival period. DAF-18 is the worm orthologue of the human PTEN tumor suppressor. DAF-18/PTEN functions to inhibit the IIS pathway. Enhanced IIS caused by the loss of daf-18 resulted in shortened life span during L1 arrest (Fig. 1A). Blocking IIS by loss of the daf-2/INSR resulted in lengthened life span in L1-arrested worms (Fig. 1A).

Agonistic INS peptides cause Q cell divisions during L1 arrest
During L1 arrest all cell divisions are halted due, in part, to the shutdown of the IIS pathway. When the IIS pathway is activated during L1 arrest, we showed that the Q cell lineage undergoes cell divisions and movements. 3 We asked whether INS (oe) could cause Q cell divisions during L1 arrest. To aid with the scoring of the Q cell divisions in L1 arrest, we scored the presence of the Q cell neuroblast descendants AVM and PVM. Normally, L1-arrested worms only have the embryonic mechanosensory neurons ALMs and PLMs ( Fig. 2A). Our previous work found that AVM and PVM were present in L1-arrested daf-18 mutants (Fig. 2B), as loss of daf-18 enhances insulin signaling. 3 Therefore in L1 arrest, if A/PVM are present it tells us the Q cell lineage has undergone its terminal divisions and can be used as a new readout to analyze the functions of the INS ligands.

Functions of C. elegans insulin peptides
Functions of C. elegans insulin peptides these INS peptides are strong IIS agonists. To provide evidence whether these INS peptides have a physiologically relevant role in L1 arrest Q cell divisions, we predicted that a loss of function mutation in these agonistic INS may suppress the daf-18 mutant L1 arrest Q cell divisions. We tested an ins-4 ins-6 double mutant (13) and indeed this double mutant did show suppression (Fig. S3). The suppression was not complete suggesting the other INS may also have roles for L1 arrest Q cell divisions. The Q cell divisions during L1 arrest is an excellent readout for INS peptides that activate the receptor. However, this phenotype on its own is not sufficient to determine INS peptides that are inhibitory to DAF-2 as the loss of DAF-2 is the same as WT (i.e. no Q cell divisions during L1 arrest). To examine INS peptides found to be inhibitory based on L1 arrest life span, we used daf-18 mutant worms (i.e. increased IIS signaling) and asked whether the inhibitory INS peptides could suppress the daf-18 Q L1 arrest cell divisions. As predicted, we found that all the ins (oe) strains, which could make L1-arrested worms live longer (Fig. 1D) could also significantly suppress the daf-18 Q cell divisions (Fig. 2E). These results confirm that the eight INS peptides (INS-12, -14, -17, -22, -28, -34, -37, and -39) function as IIS antagonists in both L1 arrest life span and Q cell divisions. All INS peptides that could induce L1 arrest Q cell divisions also shortened L1 arrest lifespan. However, not all the ins genes, which shorten L1 arrest life span can induce L1 arrest Q cell divisions (Fig. 1C), suggesting that these INS peptides have specific activities in controlling L1 arrest life span and cell divisions.
To ensure that INS peptides overexpressed in the nervous system were dependent on normal peptide processing, we tested whether the proprotein convertase-deficient animal (egl-3 mutant) (13) could suppress the function of agonistic INS peptides on L1 arrest Q cell divisions. We found that L1-arrested Q cell divisions in two strong ins (oe) worms (INS-3 and INS-4) were almost completely suppressed by egl-3 ( Fig. S4A) suggesting that these insulin peptides need to be processed properly for their function. In addition, if these agonists INS peptides bind to and activate the DAF-2/INSR then daf-2 mutants should suppress the ins (oe) L1 arrest Q cell divisions. We found that each ins (oe) strain that induced L1-arrested Q cell divisions was completely suppressed in the daf-2 (e1370) mutant background (Fig. S4B). Previous work showed that the activated IIS pathway also induced germ cell and M cell divisions in L1-arrested worms (15,20), therefore we determined whether these overexpressed INS peptides would induce L1 arrest germ cell divisions and/or M cell divisions. Indeed, the overexpressed agonistic INS peptides were able to act cell nonautonomously, resulting in M cell (INS-4) and germ cell divisions (INS-3, -4, and -9) during L1 arrest (Fig. 3).  Table S3.

Functions of C. elegans insulin peptides INS peptide function on dauer formation
In C. elegans, animals in the second larval stage can enter a dauer diapause phase under adverse environmental conditions. Mutants that reduced IIS (such as daf-2 INSR (lf) or age-1 PI3K (lf)) have a dauer-constitutive (Daf-c) phenotype. Yet, individual INS knockout mutants only show a very weak dauer phenotype, which may imply functional redundancy (9). Functional redundancy was supported by creating a fully penetrant Daf-c phenotype by simultaneous removal of ins-4, ins-6, and daf-28 (13). However, the functions of the 40 ins genes on dauer formation are still not well-addressed. Here, we tested the function of individual pan-neuronal INS on dauer formation. Previous work showed that WT C. elegans can go into dauer under high temperatures even in the presence of food or noncrowding conditions (21). At   Table S4.

Functions of C. elegans insulin peptides
reduced dauer formation, suggesting that these 10 INS can activate IIS consistently when scored by different phenotypes. Notably, not all the INS peptides, which extended L1 arrest life span enhanced dauer formation, suggesting that these INS peptides have different functions to control L1 arrest and dauer. For example, INS-12, -14, -22, -28, and -34 acted like IIS antagonists as they increased L1 life span, whereas they acted like agonists or had no function in dauer formation. INS-7, -16, -18, -25, -30, and -31 acted as IIS agonists as they reduced L1 arrest life span, but acted as IIS antagonists by increasing dauer formation.

INS peptide function on fat accumulation in adult worms
The IIS pathway plays an important role in controlling fat accumulation (22,23). The daf-2/INSR (lf) causes fat accumulation in adults (24 -27). Food cues are sensed by an olfactory receptor in the amphidal sensory neurons and this, in turn, is relayed to the IIS pathway to control fat metabolism (28,29).  (Fig. 5B). Our results suggest that most of the INS that act as agonists had reduced fat staining, whereas most of the antagonists had increased fat staining. Of all the 13 INS (oe) that could induce L1 arrest Q cell divisions (Fig. 2D), most behaved as agonists for fat accumulation with the exception of INS-1, -2, -7, and -31, which had no effect, and INS-8, which acted as an antagonist for fat accumulation. Only INS-21 appeared to be specific for a role in fat accumulation exhibiting no effects on the other three phenotypes scored (Fig. 5B).

An F peptide and the ␤ class INS act as agonists
According to our results, it is apparent that INS peptides that were structurally characterized as the ␤ class and contain a sequence known as the F peptide are activators of the IIS (Fig.  6A). Nine of the ␤ class INS peptides contain the F peptide (Fig.  6A). All of the ␤ class INS behaved as agonists of IIS in our L1 arrest Q cell division assay, except for INS-10, which does not have an F peptide. We hypothesized that the F peptide contributes to the C. elegans INS activation in L1 arrest Q cell divisions. To test this hypothesis, we pan-neuronally expressed a strong agonist, INS-4 with a deletion of the F-peptide region or the F peptide alone and found that both failed to induce Q cell divisions during L1 arrest (Fig. 6, B and C). Pan-neuronal co-expression of the F peptide and the F peptide-lacking INS-4 resulted in the induction of Q cell divisions during L1 arrest (Fig. 6, B and C), showing that the F peptide can act in trans and is necessary for INS-4 activation of DAF-2/INSR. We then asked if co-expression of the F peptide with the ␤ class INS-10, whose overexpression does not affect Q cell division, could induce L1 arrest Q cell divisions. The F peptide and INS-10 co-expression in trans failed to induce Q cell divisions during L1 arrest (Fig. 6C). This suggests that the F peptide functionally complements INS-4 (minus F peptide) activity in a peptide sequence-specific manner and/or that the pool of F peptides released upon processing of F peptide INS may not functionally complement INS peptides that lack an embedded F peptide sequence.  Table S5.

Discussion
In this study, we created independent worm lines that overexpressed each of the 40 C. elegans INS peptides and assayed for IIS phenotypes to assign in vivo roles. Because this is an overexpression study, INS are not at their normal physiological levels (Table S1 and  Our study is the first to provide the functional data for all 40 INS on L1 arrest life span, Q cell divisions, heat-stress induced dauer formation, and fat accumulation. Our results are summarized in Table 1 and Fig. 7. Mutants with reduced IIS signaling have both a Daf-c phenotype and extended L1 arrest survival. We show that the INS-17, -37, and -39 are IIS antagonists exhibiting increased L1 arrest survival and increased dauer formation. INS-17 was previously reported to work as a IIS antagonist for dauer regulation (12), but this work is the first to assign INS-37 and -39 as antagonists. Our work also demonstrated that select INS peptides are IIS antagonists in controlling L1 arrest survival but are pleiotropic in their action in controlling dauer formation. For example, INS-12, -14, -28, and -34, which act as antagonists and can extend L1 arrest survival, can also act as agonists and have significantly lower dauer formation than WT worms. On the other hand, INS-7, -16, -18, -25, -30, -31, act as agonists in L1 arrest, but can have the opposite role in dauer, acting as an antagonist by increasing dauer formation. Our results suggest that INS function in L1-arrested worms may be different from that of controlling dauer formation because it is an alternative L3 development stage, and thus INS peptides have spatiotemporal

Functions of C. elegans insulin peptides
compartmentalization with respect to their function. Our finding is consistent with studies that show that dauer arrest and adult lifespan regulation by IIS are also decoupled (30 -33).
Pan-neuronal INS overexpression that caused L1 arrest Q cell divisions identified 13 INS peptides (INS-1, -2, -3, -4, -6, -7, -8, -9, -19, -25, -31, -32 and DAF-28) that act as agonists for the IIS. All 13 INS peptides also have short L1 arrest life span, suggesting that all 13 INS peptides behave as potential IIS agonists. Previous studies, based on differing assays assigned INS-3, -4, -6, -9, DAF-28 as potential agonist INS peptides (13)(14)(15)34). These studies are consistent with our findings and support the reliability of our L1 arrest Q cell division readout as a means of categorizing the ins genes. INS-5 has been suggested to be an agonist (15), however, we did not find INS-5 to have any function in all of our assays, which is consistent with another report (13). INS-1 was shown to be an antagonistic peptide based on dauer formation (7,13,34). Overexpression of ins-1, enhances dauer arrest in weak daf-2 mutants, suggesting that INS-1 antagonizes insulin-like signaling. Also, INS-1 is antagonistic to DAF-2 for behavior (35). However, in our assays we found INS-1 to have weak activation properties. INS-1 may be a complex peptide as INS-1 acts as an agonist for IIS in salt chemotaxis learning (36).
We identified eight antagonistic INS peptides that could significantly extend L1 arrest life span and when overexpressed from the nervous system could suppress the daf-18/pten L1 arrest Q cell divisions. Thus, these INS peptides acted as therapeutic peptides for daf-18/pten worms. In humans, insulin and IGFs are thought to work as agonists and do not have antagonistic properties. Our work showed that C. elegans INS-6 is a strong agonist and INS-6 has been shown to bind and activate the human insulin receptor (37). It would be interesting to know whether the antagonistic INS peptides we have identified in this study can bind to and inhibit the human insulin or IGF-1 receptor, if so, these C. elegans INS peptides could be used as future therapeutics for hyperinsulinemia.
Our study revealed that INS-8 behaves as an agonist of IIS, because ins-8 (oe) shortens the life span of L1-arrested worms, has low penetrance dauer formation, and promotes L1 arrest Q cell divisions. A previous study suggested that INS-8 may work as an agonist (8). However, ins-8 (oe) worms behaved as an antagonist of IIS exhibiting higher fat accumulation. One study showed that ins-8 (oe) enhances ins-7 mutant life span, which would suggest that INS-8 is an antagonist (8). We suggest that the neuronal ins-8 (oe) is sufficient to work as an agonist to activate the IIS pathway, which in turn controls the L1 arrest life span and dauer formation, but in adult worms, it may work as an antagonist. This result with INS-8 is consistent with our finding that many INS peptides have distinct roles in mediating fat accumulation that is developmentally separate from its effects on dauer and L1 arrest life span. Insulin signaling temporally and in varying tissues of the body contributes differently to fat content (31).
Of   (7). ␥-Insulins have the arrangement of three disulfide bonds as found in vertebrates, whereas ␣and ␤-insulins contain an additional intra-chain disulfide bond (red). ␣-Insulins lack the common intra-chain bond in the A chain, which is substituted by the interaction of aromatic amino acid side chains. INS-31 constitutes its own additional class with three repeats of B and A peptide chains. In our study we classified the 40 insulin ligands into 6 functional groups: strong agonist/antagonist: activity consistence within all tested phenotypes; weak agonist/antagonist: activity consistence within most tested phenotypes, but have no significant activity in other phenotypes; diverse, can have both agonist and antagonistic roles, and neutral ligands, no significant activity in all tested IIS assays.

TABLE 1 Data summary of functions of C. elegans INS peptides
Data are compared to wild type. ϩ/Ϫ: with/without L1 arrest Q cell divisions. N: normal; L: low; H; high. See details in Supporting information for raw data. Color code: green indicates strong agonist; brown indicates agonist; red indicates strong antagonist; dark red indicates weak antagonist; and blue indacates neutral.

Functions of C. elegans insulin peptides
To understand what makes an INS an activator we focused on the L1 arrest Q cell divisions as this assay determined with certainty which INS peptides acted as IIS activators. Our study revealed that the ␤ class INS peptides, which contains the three canonical disulfide bonds as well as an additional inter-chain disulfide bond are good predictors of an INS peptide agonist. INS-1 to INS-10 and DAF-28 fall into this class (Fig. 7). Nine of the ␤ class INS contain an F peptide (7), the exception is INS-10. The F peptide is processed at the N terminus by the signal peptidase cleavage site and at the C terminus by either the proprotein convertase enzymes EGL-3/PC2-like with cleavage sequence (RR or KR) or a KPC-1/PC1-like site (R-X-X-R) (13) (Fig. S1). INS-10 does have activation properties and reduces L1 arrest life span and dauer formation, but could not induce L1 arrest Q cell divisions. INS-5 was predicted to contain an F peptide (7), but upon further examination, INS-5 does not have a proprotein convertase site that would release the F peptide, but instead would be incorporated as the B chain (Fig. 7, Fig.  S1). Thus, our results reveal a striking revelation that all INS peptides that are predicted to contain an F peptide should behave as agonists of the IIS (Fig. 7). We showed that the F peptide is indeed required for INS-4 to induce L1 arrest Q cell divisions and the F peptide can be added back in trans to restore INS-4 (minus F peptide) function. Note that the F peptide is not an absolute requirement for an INS to induce L1 arrest Q cell divisions as INS-1, -19, -25, -31, and -32 could induce L1 arrest Q cell divisions (albeit not as strong as other, e.g. INS-4). Interestingly, the predicted signal sequences for INS-32 was longer than average INS peptides and therefore may produce an F peptide. This prompted us to look more closely at the predicted peptides and using the SignalP 4.1 (38), we identified INS-32 as having potential F peptide (Fig. S1). In addition, we also propose that INS-19 has an F peptide as it has a potential proprotein convertase-cleavage site (Fig. S1).
Human insulin has a C peptide, and human IGF-1 and IGF-2 have E peptides that are cleaved during processing analogous to the F peptides identified in C. elegans INS peptides. Our work on the F peptide should stimulate closer examination of peptides released upon processing of human insulin and IGF. For instance, IGF-1 is one of the key molecules in cancer biology, however, little is known about the role of the E peptide. The E peptide is thought to have functional properties as the release from IGF-1 is thought to induce cellular proliferation in the human prostate cancer (39). The C peptide of proinsulin is important in the processing of mature insulin and may have biological activity as a report suggests that it binds to a G protein-coupled surface receptor and activates Ca 2ϩ -dependent intracellular signaling pathways (40). Because we have provided evidence in C. elegans that the F peptide can work in trans with the INS-4 lacking an F peptide, the F peptide serves as a modulator of INS-4 to induce L1 arrest Q cell divisions.
Finally, of the 40 INS (oe) strains tested, INS-5, -23, -26, -27, and -33 were not functional in the selected assays. These INS peptides may have specific roles that have not been uncovered through the assays selected. The INS peptides are also thought to function in a combinatorial fashion and perhaps these single INS peptides have no function on their own and may participate with the other INS to exert their function (9). Alterna-tively, these INS may bind to receptors other than DAF-2/ INSR. A report has suggested that additional insulin-like receptors have been identified in the C. elegans genome (41).
In conclusion, our work systematically tested the functions of each of 40 INS on dauer formation, L1 arrest life span, L1 arrest Q cell divisions, and fat accumulation phenotypes (Table 1). By using these IIS phenotypes as readouts of insulin peptide activity, we found that seven INS peptides (INS-3

Transgenic strains
For the ins overexpression strains, the insulin genomic sequences were amplified from a N2 genomic DNA and placed under control of the pan-neuronal promoter Prgef-1 by standard cloning procedures (16). A plasmid with the injection marker odr-1::rfp was injected into Pmec-4::GFP(zdIs5) (43) worms using standard microinjection methods (44). For the F peptide experiment: Q5 mutagenesis (New England Biolabs) using primers was used to delete the F peptide sequence from the Prgef-1::INS-4 plasmid. For these extrachromosomal arrays, each injected plasmid we established and scored at least 3 independent lines. For a list of strains and primer sequences please see Tables S6 and S7.

L1 arrest Q Cell divisions
Nonstarved well-maintained mixed stage worms were collected to prepare embryos, as described (45). In brief, embryos were maintained and hatched in sterile M9 and incubated at 20°C with low speed rocking to initiate L1 arrest. The final Q cell descendants (A/PVM) were observed under an Axioplan fluorescent microscope (Zeiss, Germany) after 2 days or more in L1 arrest. 50 -100 l of M9 containing greater than 50 L1-arrested worms were removed from the culture. The total number of worms and the worms with A/PVM cell divisions were

Functions of C. elegans insulin peptides
counted. For transgenic strains, only the worms with the injection marker were counted and analyzed. Similarly, M cell divisions were analyzed by using ayIs6 strains (46).

Antibody staining
Antibody staining was performed as previously described (47). Worms were fixed by using 1ϫ witches brew and paraformaldehyde (2.5% final). To detect germline cells, rabbit anti-PGL-1 (P-granule component) (1:20,000) (a gift from Dr. Susan Strome) was used as the primary antibody, which was diluted in PBST-A. Worms were incubated at room temperature overnight. Detection was with a FITC-labeled goat anti-rabbit secondary antibody (1:100). For transgenic strains, only the worms with the injection marker were counted and analyzed. The total number of worms and the worms with germ-cell divisions were counted. Analysis of worms was using an Axioplan fluorescent microscope (Zeiss, Germany).

L1 arrest life span assays
Life span was assessed in liquid medium (19). L1 worms were cultured in 1 ml of M9, 50 -100 l was taken to ensure the sample size was larger than 50, and the worms were scored every day. We scored survival by counting the number of worms that were moving (alive) and then dividing that number by the total number of worms in the aliquot. To compare the survival rates between strains, the L1 arrests were carried out in triplicate with at least 100 L1s and the mean survival rate calculated by the Kaplan-Meier method (48), which is the fraction of living animals over a time course. The significance of difference in overall survival rate is performed using the log-rank test (49).

Fat staining
Synchronized eggs were cultured on OP50 plates with 25 ng/ml of Nile Red for 3 days at 20°C, and then washed 3 times with M9, cultured on normal OP50 plates for 1 more day at 20°C to eliminate the Nile Red OP50 background in the intestine. Worms were collected and washed in M9 3 times, then fixed in 40% isopropyl alcohol for 3 min. At least 30 animals were imaged in at least three separate experiments using a Zeiss Axioplan. The fluorescent intensity was captured by using the TurboRFP (573 nm) channel and was quantified by using ImageJ. The nontransgenic (WT) siblings' intensity was set to "1" and the transgenic worms were calculated as "fold-increased or decreased" compared with WT animals. Nile Red has been a controversial as a stain for fat accumulation (27), so we also tested some of INS overexpressing strains with Oil Red O and they were consistent with the results obtained with Nile Red (see Fig. S5).

Dauer formation at high temperature
We analyzed L2 dauer formation at 29°C as synchronized zdIs5 worm eggs hatched at 29°C presented a higher percentage dauer phenotype. The dauer, dauer-like, and adult worms with injection marker were counted. The dauer percentages were calculated. Three independent trials were performed for each strain, each sample size was greater than 50.
For Q cell divisions, dauer formation, and fat staining experiments, nontransgenic siblings were used as WT controls. All the scatter plot data presented in the figures were made using GraphPad Prism 7.