RNA-binding protein GLD-1/quaking genetically interacts with the mir-35 and the let-7 miRNA pathways in Caenorhabditis elegans

Messenger RNA translation is regulated by RNA-binding proteins and small non-coding RNAs called microRNAs. Even though we know the majority of RNA-binding proteins and microRNAs that regulate messenger RNA expression, evidence of interactions between the two remain elusive. The role of the RNA-binding protein GLD-1 as a translational repressor is well studied during Caenorhabditis elegans germline development and maintenance. Possible functions of GLD-1 during somatic development and the mechanism of how GLD-1 acts as a translational repressor are not known. Its human homologue, quaking (QKI), is essential for embryonic development. Here, we report that the RNA-binding protein GLD-1 in C. elegans affects multiple microRNA pathways and interacts with proteins required for microRNA function. Using genome-wide RNAi screening, we found that nhl-2 and vig-1, two known modulators of miRNA function, genetically interact with GLD-1. gld-1 mutations enhance multiple phenotypes conferred by mir-35 and let-7 family mutants during somatic development. We used stable isotope labelling with amino acids in cell culture to globally analyse the changes in the proteome conferred by let-7 and gld-1 during animal development. We identified the histone mRNA-binding protein CDL-1 to be, in part, responsible for the phenotypes observed in let-7 and gld-1 mutants. The link between GLD-1 and miRNA-mediated gene regulation is further supported by its biochemical interaction with ALG-1, CGH-1 and PAB-1, proteins implicated in miRNA regulation. Overall, we have uncovered genetic and biochemical interactions between GLD-1 and miRNA pathways.

the 3 0 UTR of target mRNAs. Over the past decade, they have been implicated in many biological processes and are now considered to be major modulators of gene expression [1,2]. After biogenesis, miRNAs and Argonaute proteins (AGO; ALG-1/ALG-2 in C. elegans) form the miRNAinduced silencing complex (miRISC) together with GW182 proteins (AIN-1/AIN-2 in C. elegans) [3]. VIG-1 is thought to function in the miRNA pathway and was identified as an interactor of miRNA/Argonaute complexes by coimmunoprecipitation [4]. Mammalian TRIM32 and C. elegans NHL-2 also interact with AGO and promote miRNA activity [5,6].
Deleting the vast majority of known C. elegans miRNAs individually does not result in obvious overt phenotypes [19]. Phenotypes tend to arise when several members of a miRNA family are deleted [20]. Alternatively, mutating miRNA pathway genes also generate sensitized system that helps us to unravel miRNA function [21]. Such synthetic phenotypes point towards the existence of extensive redundancy in miRNA-mediated gene regulation. Caenorhabditis elegans genetics allows for using 'sensitized' genetic backgrounds to study subtle phenotypes associated with redundant mechanisms of miRNA-mediated gene regulation.
Initially aiming to identify genes required for GLD-1mediated translational regulation, we performed a genomewide RNAi screen for enhancers of the gld-1(op236) hypomorphic allele. This screen identified vig-1 and nhl-2, both of which are modulators of miRNA function, thus suggesting that GLD-1 might affect general miRNA-mediated gene regulation. We indeed found that gld-1 enhances multiple let-7 and mir-35 family miRNA phenotypes affecting somatic development. Using stable isotope labelling with amino acids in cell culture (SILAC)-based proteomics, we show that the upregulation of the histone mRNA-binding protein CDL-1 is partially responsible for the genetic interactions between GLD-1 and let-7 miRNA. A role for GLD-1 in miRNA-mediated gene regulation is further supported by the interaction of GLD-1 with ALG-1, CGH-1 and PAB-1, proteins previously implicated in miRNA-mediated gene regulation.
Caenorhabditis elegans larvae were grown on Escherichia coli strain OP50 at 208C unless otherwise stated. let-60(ga89) mutants were grown at 208C and their progeny was shifted to 258C at the L1 larval stage and scored at adult stage for the presence of multi-vulva phenotype. Microscopic analysis of the animals was carried out by anaesthetizing with levamisole and observed using a Zeiss axioscope.

RNAi in Caenorhabditis elegans
We used the whole-genome RNAi library generated by Ahringer Laboratory [22,23], and a modified version of the RNAi feeding protocol described in [24]. RNAi-expressing bacteria were grown from frozen stocks overnight at 378C in LB medium containing 50 mg ml 21 ampicillin and 10 mg ml 21 tetracycline in 96-well plates. A fresh culture seeded from the overnight culture was incubated at 378C in 96-deep-well plates until OD 600 nm was 0.6-1 and then induced with 1 mM IPTG for 2 h at 208C. L1 larvae were dispensed into 96-well plates (10 worms well 21 ) in 100 ml M9 medium supplemented with 50 mg ml 21 ampicillin, 10 mg ml 21 tetracycline, 10 mg ml 21 cholesterol, 0.1 mg ml 21 fungizone and 1 mM IPTG. Induced bacteria (50 ml) were also dispensed into each well, and worms were grown at 208C with constant shaking at rsob.royalsocietypublishing.org Open Biol 3: 130151 180 r.p.m. The presence of progeny was scored after 4-5 days. To validate candidates from the RNAi screen, 500-1000 L1 larva were treated with RNAi in 50 ml falcon tubes. Volumes of M9 medium and bacterial culture were scaled up accordingly. RNAi of GLD-1 protein interactors and glp-1 was performed in a similar manner in 50 ml falcon tubes, and worms were transferred to plates seeded with the RNAi bacteria at L2-L3 stage. Number of assayed animals is presented on related figures.

SILAC in nematodes
SILAC growth conditions, subcellular fractionations, peptide preparations and mass spectrometry were carried out as described in [34][35][36]. The number of proteins detectable by mass spectrometry was increased by subcellular fractionation [34]. For determining B to A and C to A ratios, we considered proteins that were only detected in one experiment if a minimum of two peptides were detected (the detection of two or more independent peptides all passing though MAXQUANT adding to the statistical significance [37]). We also considered proteins detected in two independent experiments but excluded those where the standard error of the mean based on multiple detected peptides from the same protein was bigger than the mean. In the B to A comparison, 2708 proteins passed those criteria, whereas 2927 passed in the C to A comparison. We focused on the 2179 proteins that were reliably detected in both datasets (see electronic supplementary material, table S1).

4.1.
A genetic screen identifies vig-1 and nhl-2 as enhancers of gld-1(op236) We performed a genetic screen to search for genes whose inactivation enhances the gld-1(op236) defect. gld-1(op236) animals have wild-type germlines at 208C but are sterile at 258C [13]. In a whole-genome RNAi feeding screen, we looked for genes whose depletion by RNAi causes sterility at 208C specifically in gld-1(op236) mutants but not in wildtype animals (figure 1a). Out of over a hundred initial candidates (data not shown), 20 reproducibly showed reduced progeny in the gld-1(op236) background (figure 1b). Of these, deletions were available for seven candidates, and double mutants of five of these with gld-1 confirmed the reduced progeny or sterility phenotype observed in the RNAi screen. A detailed analysis of synthetic phenotypes with hecd-1, eel-1 and larp-1 will be described elsewhere.
Unexpectedly, 29% of the surviving gld-1(op236); vig-1(ok2536) larvae have arrested development and abnormal somatic morphology (figure 1c,d). 'Extruding bumps' are readily observed at various positions along the body axis of these animals. Using the col-19::GFP hypodermal cell marker, we found that hypodermal cells accumulate in these protrusions rather than being regularly spaced along the body axis (figure 1c). In addition, double-mutant animals appear shorter and sick. Sterility and embryonic lethality phenotypes have already been described for dcr-1 (dicer endonuclease) mutants as well as for alg-1/alg-2 Argonaute double mutants [38][39][40]. Given this similarity, and as both VIG-1 and NHL-2 modulate miRNA function, we decided to investigate possible genetic interactions between gld-1 and miRNAs.

gld-1 genetically interacts with mir-35 family miRNAs
To investigate an interaction between gld-1 and miRNAs, we started by making double mutants of gld-1(op236) and mir-35 family miRNAs. The mir-35 family of miRNAs comprises eight members (miR-35-42), which are highly enriched in oocytes [20] and are required for embryonic development. While mir-35 family mutants do not individually exhibit an observable phenotype, combined mutation of either all or most mir-35 family members causes severe embryonic and larval lethal phenotypes [20], similar to those we observed in the gld-1(op236); vig-1(ok2536) double mutant. The expression pattern and phenotypes of mir-35 family miRNAs make them suitable to investigate possible genetic interactions with gld-1.
At the end of the L4 larval moult, lateral seam cells fuse and form extracellular structures called alae [47]. The timing of seam cell fusion and alae formation is controlled by let-7 family miRNAs [43]. Similarly, mutations in C. elegans genes encoding AGO (alg-1, alg-2) and GW182 proteins (ain-1, ain-2) also have seam cell fusion and alae formation defects [18,38]. We assayed seam cell fusion using the AJM-1::GFP junction marker as indicated (figure 3f ). Analysis of differential interference contrast (DIC) images and the AJM-1::GFP junction marker indicate defects in alae formation and seam cell fusions in gld-1(op236), let-7(mg279) and gld-1(op236); let-7(mg279) animals (figure 3e). We next quantified the extent of seam cell fusion defects and found that the incidence of seam cell fusion defects is higher in gld-1(op236); let-7(mg279) double-mutant animals than in single mutants (figure 3c). In summary, our combined data suggest that gld-1 affects hypodermal development in let-7 mutant background, either by acting through let-7 or through a parallel pathway.

GLD-1 affects let-60 signalling
To check whether the genetic interactions of gld-1 with the let-7 miRNA family are restricted to the hypodermal development, we looked into the let-60/RAS pathway that functions during vulva formation [48]. mir-84 and let-7 antagonize let-60/RAS signalling in vulval precursor cells that are not destined to form the vulva. Such regulation can be assessed in vivo using let-60/RAS gain-of-function alleles that induce

Germline expression of gld-1 is not necessary for somatic phenotypes
GLD-1 protein expression has previously been reported in the germline [7]. We therefore wished to determine whether gld-1 expression also occurs in somatic tissues, and whether the genetic interactions we observe during somatic development depend on a germline. To determine whether a germline is required for seam cell fusion defects in gld-1(op236) and gld-1(op236); let-7(mg279) animals, we used RNAi to inactivate glp-1, which is essential for germline development [49]. We observed that the defective seam cell fusion phenotype occurs to a similar extent in animals lacking a germline as in those with a germline (figure 3c). Furthermore, somatic expression of an mCherry::H2B transcriptional reporter (the histone fusion used to focus diffuse, low level cytoplasmic GFP expression to the nucleus) under the control of the gld-1 promoter and the gld-1 3 0 UTR was mainly localized to the head, tail and ventral side of the animals (see electronic supplementary material, figure S3). DIC microscopy analysis indicates that most of the positive cells are neuronal cells in the head and tail ganglia and the ventral nerve cord. Interestingly, let-7 and at least one let-7 target, hbl-1, is also reported to have a similar expression pattern [50], although it is not known to what extent let-7 miRNA phenotypes require hypodermal or neuronal expression. It is likely that high levels of background fluorescence masked low levels of gld-1 reporter expression in other cell types. We could not confirm this localization pattern by antibody staining owing to background problems. However, we generated transgenic animals expressing full-length gld-1 fused to histone::GFP by an operon linker, generated by single copy insertion using the MosSCI technique [29,51,52]. This transgene shows the same pattern of somatic expression, shows expression in embryos and rescues gld-1(null) in the germline (electronic supplementary material, figure S4). Even though these reporter constructs might not exactly represent endogenous GLD-1 expression, together with our genetic results, they suggest somatic roles for gld-1.

4.7.
Overexpressing a lin-41 3 0 UTR construct acts as a 'sponge' to sequester let-7 miRNA and provides a sensitized system to assay GLD-1 activity One of the targets of let-7 miRNA during larval development is the lin-41 mRNA [45]. Hypodermal defects in gld-1(op236); let-7(mg279) could be the result of mis-regulation of lin-41 mRNA. To investigate this possibility, we generated a single copy MosSCI insertion of GFP::lin-41-3 0 UTR fusion construct under the control of the col-10 promoter that ensures expression in hypodermal tissues. We did not observe any phenotype (figure 5a) in wild-type or in gld-1(op236) animals. However, let-7(mg279) mutants showed a low penetrance bursting through the vulva phenotype reminiscent to let-7(null) phenotype (figure 5a). We likened this observation to a sponge-like effect of the GFP::lin-41-3 0 UTR towards let-7 miRNA. miRNA sponges are complementary target sequences that can sequester the miRNA from its endogenous target [53]. In C. elegans, such a sponge was used for the lin-4 miRNA [54]. The penetrance of the vulva-bursting phenotype is dramatically enhanced in gld-1(op236); let-7(mg279) double mutants Thus, gld-1(op236) specifically enhances the let-7-dependent phenotypes, and the extent of genetic interactions between gld-1 and the let-7 miRNA pathway becomes more evident when the let-7 miRNA pathway is further compromised. Expression of a let-7 sponge with a deletion of the 3 let-7 binding sites or expression of the unrelated unc-54 3 0 UTR did not cause any bursting phenotype in let-7(mg279) and in gld-1(op236); let-7(mg279) animals supporting the specificity of the let-7 sponge and the interactions between gld-1 and let-7 miRNA (figure 5a). We next expressed gld-1 under the control of the col-10 promoter in the hypodermis to investigate whether such expression of gld-1 might cause any phenotype associated with the loss of let-7 targets. col-10::GLD-1 expression in the hypodermis caused a loss of adult-stage alae (figure 5b) and a dumpy phenotype (shortened size; figure 5c,d). A dumpy phenotype also occurs following mutation of let-7 targets such as lin-41 [45]. Co-expression of the let-7 sponge partially rescues the dumpy and loss of alae phenotypes (figure 5b-d). Based on these results, we cannot exclude the possibility that gld-1 and let-7 miRNA function in parallel pathways during the hypodermal development. Equally, GLD-1 expression in the hypodermis might cause unrelated phenotypes. However, another likely interpretation of these experiments is that GLD-1 and let-7 act in conjunction to excessively repress target mRNAs possibly in the same pathway, and that reducing the 'dose' of let-7 using the sponge alleviates target gene repression.

SILAC in nematodes identifies proteome wide changes in gld-1 and let-7 mutants
Our results suggest that GLD-1 and let-7 synergistically affect animal development. We wanted to determine whether the   rsob.royalsocietypublishing.org Open Biol 3: 130151 phenotypes we observed are due to the mis-regulation of either a single or small number of genes, as opposed to affecting multiple target genes. In addition, we wanted to determine whether GLD-1 and the let-7 miRNA regulate distinct or same targets. To address this question, we used nematode SILAC to systematically quantify protein levels [34,36]. The strongest genetic interaction between gld-1 and let-7 occurs in the let-7 sponge system (figure 5a). Thus, to assess changes in the proteome of the respective strains, we used the following experimental set-up. For the SILAC experiment synchronized L1 larvae of three strains, namely (A) [let-7 sponge], (B) let-7(mg279); [let-7 sponge] and (C) gld-1(op236); let-7(mg279); [let-7 sponge] were grown up to the young-adult stage until the bursting phenotype just becomes visible and subjected to quantitative mass spectrometry (figure 6a). As the animals expressing let-7 sponge alone do not display any phenotype, we considered them as the baseline similar to using wildtype. Thus, by comparing the animals with a weak phenotype (B) to animals with a strong phenotype (C), we aimed to identify proteins whose expression change might be responsible for the bursting through the vulva phenotype and help explain the interaction between gld-1 and the let-7 miRNA. In a SILAC experiment, the relative ratios of two sets of proteins can be measured by differential isotope labelling. We then determined proteins differentially expressed when B was compared with A. At the same time, we also determined differentially expressed proteins comparing C with A. As a control, we analysed the levels of GFP expression under the control of lin-41 3 0 UTR sponge by Western blotting. GFP levels are higher in the let-7(mg279); let-7 sponge animals, compared with the let-7 sponge only (B to A; electronic supplementary material, figure S5a). We identified a similar rise in GFP levels in SILAC experiments, and the level of GFP was further increased in gld-1(op236); let-7(mg279); let-7 sponge animals (C to A; electronic supplementary material, figure S5b). This result confirms the sensitivity of our SILAC-based approach. In the B/A comparison, 2708 proteins passed our specificity criteria, whereas 2927 passed in the C/A comparison (see Materials and methods). We focused on comparing the abundance of the 2179 proteins that were reliably detected in both datasets. The ratios of (C) gld-1(op236); let-7(mg279); [let-7 sponge] protein to (A) [let-7 sponge] protein shown on the y-axis are compared with the (B) let-7 (mg279); [let-7 sponge], with (A) [let-7 sponge] ratios depicted on the x-axis (figure 6a). Among these 2179 proteins, 252 overlap with the 1084 previously described GLD-1 targets (figure 6a, coloured in blue) [55,56]. Proteins (239) overlap with 1322 predicted let-7 targets (figure 6a, coloured in red, mirWIP database [57]). Fifty-four proteins are predicted to be both GLD-1 and let-7 targets (figure 6a, coloured in purple). The relative abundance of the majority of suspected GLD-1 and let-7 co-targets do not change when the C/A ratios are compared with the B/A ratios. However, three such candidate proteins, namely cdl-1, dnj-2 and B0303.3 were more prominently upregulated in C/A.
We tested whether the depletion of these three proteins suppresses the vulva-bursting phenotype of the gld-1(op236); let-7(mg279); [let-7 sponge] strain (figure 6b). Targeting the [let-7sponge]whichisacol-10::GFP::lin413 0 UTRconstructbyGFP RNAi lead to a complete suppression of the vulva-bursting phenotype and thus served as a positive control. The depletion of one of the candidates, namely cdl-1 lead to a reduced vulva-bursting phenotype consistent with the notion that the upregulation of CDL-1 in the gld-1(op236); let-7(mg279) background might contribute to the vulva-bursting phenotype. Indeed, cdl-1 3 0 UTR harbours a GLD-1 and a let-7 binding site (electronic supplementary material, figure S7). CDL-1 is a histone mRNA hairpin binding protein required for expression of histones [58][59][60][61]. It is expressed in all somatic cells with strong expression in proliferating cells such as hypodermal cells, intestinal cells and the germ cells [59]. CDL-1 is required for larval development and affects vulva morphology [59,61]. However, the suppression of the vulva-bursting phenotype by CDL-1 RNAi in gld-1(op236); let-7(mg279); [let-7 sponge] animals is not complete. This indicates that other factors also contribute to the vulva-bursting phenotype.  [55,56] are coloured blue, mirWIP database let-7 target predictions [57] are coloured red, and the possible GLD-1 and let-7 co-targets based on these lists are coloured purple. The remaining proteins are coloured in grey. CDL-1, DNJ-2 and B0303.3 are possible GLD-1 and let-7 targets that are upregulated more than 1.2-fold (arrows). (b) RNAi-mediated knockdown of 9 genes upregulated in the gld-1(op236); let-7(mg279); [let-7 sponge] animals (RNAi is done in the same strain). We picked six genes upregulated more than twofold and are GLD-1 or predicted let-7 targets (red and blue spots above the twofold line) and three genes upregulated more than 1.2-fold that are GLD-1 and predicted let-7 targets (purple spots above the 1.2-fold line). Empty vector RNAi and GFP RNAi are used as negative and positive controls respectively. B0303.3 RNAi results are omitted due to the early larval lethality in these animals (error bars ¼ s.e.m. of triplicate, p-values calculated using Fisher's exact test). rsob.royalsocietypublishing.org Open Biol 3: 130151

Identification of GLD-1-containing complexes
Our data are consistent with GLD-1 either directly or indirectly interacting with the miRNA pathway. To investigate this genetic interaction at biochemical level, we next aimed to purify proteins that interact with GLD-1. We used GLD-1 antibodies to immunoprecipitate GLD-1 complexes in wild-type animals and used mass spectrometry to identify GLD-1 interactors. To verify the specificity of GLD-1 interactions, we also pulled down GLD-1 using a GFP-binder from animals expressing a GLD-1::GFP fusion protein and focused our subsequent analysis on proteins specifically pulled down in both purification approaches (figure 7a). Among the GLD-1 interactors identified, many are associated with RNA and some are involved in miRNA-mediated gene regulation (figure 7a). We identified the CGH-1 helicase, CAR-1 and the Y-box domain proteins CEY-2, CEY-3 and CEY-4, all previously reported to be components of a complex localized to RNP granules in the C. elegans germline, and that are likely to function in translational repression [62]. Importantly, CGH-1 is proposed to act in miRNA-mediated gene expression. Depletion of cgh-1 enhances the defects of let-7 family mutants and CGH-1 biochemically interacts with ALG-1, AIN-1 and NHL-2 [5]. Indeed, ALG-1 also co-purified with GLD-1 in our experiments (figure 7a). Other proteins identified as GLD-1 interactors include PAB-1 and SQD-1. PAB-1 is a C. elegans poly(A) binding protein and also a component of AIN-1, AIN-2 and CGH-1 complexes, suggesting a role in translational regulation and miRNA-mediated repression [18,63,64]. SQD-1 is the C. elegans orthologue of the Drosophila squid hnRNP protein, previously identified as an AIN-2 and mir-35 miRNA-associated protein [18,65]. PAB-1, CGH-1, CAR-1, CEY-1-4 and SQD-1 were also identified as GLD-1 protein interactors in a recent study [66]. In the same study, interaction of CGH-1 with GLD-1 was shown to depend on the presence of RNA. Considering all the GLD-1 interactors have RNA-binding domains, the presence of RNA might be essential in all these interactions. In summary, GLD-1 interactors identified in this and previous studies associate with proteins known or suspected to be involved in miRNA-mediated gene repression.

Discussion
The starting point of our work was the identification of two miRNA effector genes, vig-1 and nhl-2 as genetic enhancers of gld-1(op236) (figure 1). Both genes had previously been shown to affect miRNA function [4,5], although not in the germline. It is possible that the germline phenotypes we observed are related to a miRNA-associated function, although it is equally likely that VIG-1 and NHL-2 may, either directly or indirectly, affect GLD-1 function, or else may act in conjunction with GLD-1 to mediate translational repression of GLD-1 targets in the germline. Different phenotypes observed in the gld-1(op236); vig-1(ok2536) and the gld-1(op236); nhl-2(ok818) double mutants suggest that vig-1 and nhl-2 function in different pathways in the germline. We will investigate these possibilities in future studies.
Intrigued by the genetic interactions with those miRNA effector genes, we investigated possible roles of gld-1 in multiple miRNA-mediated pathways and uncovered novel roles for gld-1 during the somatic development of C. elegans. Given that gld-1 phenotypes and expression were previously only described for the germline, we were surprised to find GLD-1 phenotypes associated with somatic development. Enhancement of mir-35 family embryonic and larval lethal phenotypes may be explained by perturbation of maternal mRNA pools derived from gld-1(op236) germlines that may enhance mir-35 phenotypes. However, it is unlikely that such a model can explain the genetic interactions we observed between gld-1 and let-7 family miRNAs. let-7-related  rsob.royalsocietypublishing.org Open Biol 3: 130151 phenotypes arise much later during development, making a mechanism involving the maternal contribution of miRNAs unlikely. Indeed, we show that gld-1(op236) mþ z-animals have a comparable phenotype with gld-1(op236) m-z-animals ( figure 2). Furthermore, we observed that let-7 phenotypes are enhanced by gld-1 even when glp-1 RNAi animals lacking a germline were analysed. Finally, recent data published by the modENCODE consortium indicate that gld-1 is transcribed in glp-1 mutants that lack a germline [67]. Consistent with this, using two independently generated reporters we observed gld-1 reporter expression in somatic tissues. Unfortunately, GLD-1 antibody staining was not successful owing to unspecific background in somatic tissues. We focused on the let-7 miRNA pathway, whose components have been much more characterized in C. elegans. By using already established tools, we could show that gld-1 affects multiple let-7 miRNA regulated pathways (figures 2-4). Importantly, we generated a sensitized system using a let-7 sponge and showed that gld-1(op236) specifically enhances let-7 loss-offunction phenotypes (figure 5). Our let-7 sponge system confirms the notion that the miRNA pathways are highly redundant. let-7 miRNA levels in wild-type animals are sufficient to regulate both the endogenous targets and also an additional transgene target (let-7 sponge). However, when the let-7 miRNA levels are limiting, such as in the hypomorphic let-7(mg279) mutants, endogenous targets are not efficiently dealt with when the let-7 sponge is present (figure 5a). Only in this 'very sensitive' situation, a role for gld-1 in the let-7 miRNA pathway becomes apparent. This finding can be explained by the robustness and redundancy of the let-7 miRNAs and also by the target genes whose mis-regulation is well tolerated. When we expressed GLD-1 specifically in the hypodermis, we observed defects in alae formation and dumpy animals. Even though these phenotypes might be unspecific, as we do not know whether GLD-1 is expressed in this tissue, co-expressing the let-7 sponge partially suppresses these phenotypes. This further supports the involvement of gld-1 either in the let-7 pathway or in a parallel pathway. This hypothesis is further supported by the functions of the GLD-1 interactors we identified (figure 7): ALG-1 is a core component of the miRNA machinery [38], and CGH-1 can interact with ALG-1 to modulate miRNA-mediated gene regulation [5]. PAB-1 interacts with AIN-1/2, core miRISC components [18,63], and PAB-1 homologues are required for miRNAmediated gene regulation in mammalian cells [68]. We have shown that the GLD-1 interactors CGH-1 and PAB-1 affect let-7 miRNA function ( figure 7).
Hundreds of genes are regulated by GLD-1, and it is likely that the expression of an even higher number of genes is modulated by miRNAs. In this context, how can we explain the genetic and biochemical interactions we observe? To address this question, we took a state of the art proteomics approach based on SILAC in nematodes. With this approach, we were able to detect the differential expression of more than 2000 proteins, and we were able to compare the relative abundance of these proteins in animals with weak and strong phenotypes. When we tested some of the upregulated proteins for suppression of the vulva-bursting phenotype associated with the gld-1(op236); let-7 (mg279); [let-7 sponge] strain, we detected a strong suppression with the RNAi-mediated knockdown of cdl-1 gene. cdl-1 is a predicted let-7 miRNA target and it was identified as a GLD-1 target [55][56][57]. Thus, cdl-1 is a strong candidate to be co-regulated by both let-7 and GLD-1. However, we cannot rule out the possibility that the cdl-1 upregulation is not directly controlled by let-7 or GLD-1 and it may arise owing to secondary effects. Either way, we can conclude that CDL-1 upregulation in a let-7 and GLD-1-dependent manner is in part responsible for the vulva-bursting phenotype.
Overall, we suggest two possible mechanisms for GLD-1 and miRNA interactions that are not mutually exclusive and may act in parallel. First, mutation of both GLD-1 and miRNAs could mis-regulate several targets within a single pathway, leading to the observed phenotypes. For instance, moulting defects in let-7 mutants are partly due to mis-regulation of the nuclear hormone receptors nhr-23 and nhr-25 [41], and nhr-23 is a predicted GLD-1 target [55,56]. We could not detect NHR-23 in our SILAC experiments. Similarly, some of the phenotypes we observed (figure 5b) might be explained by negative regulation of LIN-28 by GLD-1 as it was reported for the germline [55]. However, we did not detect changes in LIN-28 protein levels, which is why we did not follow up on this lead.
A second possibility is that GLD-1 and miRNAs may coregulate a subset of mRNAs. The interactions we observed between GLD-1, CGH-1, PAB-1 and ALG-1 in GLD-1 complexes support such a mechanism. Consistent with this model, GLD-1 appears to associate with several 3 0 UTRs that are known or predicted miRNA targets in C. elegans. For example, the potential mir-35 targets, lin-23 and gld-1 [69], have also been identified as GLD-1 targets [55]. Similarly, lin-28 and ztf-7 are let-7 miRNA targets [70,71] and these genes have also been identified as GLD-1 targets [55].
We are aware that biochemical interactions reflect total animal extracts and that most GLD-1 protein expression is likely to occur in the germline. Thus, the association of GLD-1 with 3 0 UTRs in somatic cells may be underrepresented in biochemical experiments aimed to determine GLD-1 targets [55]. GLD-1 has been extensively studied in the germline and consensus-binding motifs have been defined [55,72]. We also know that GLD-1 appears to bind RNA as a dimer and that GLD-1 RNA binding is enhanced when multiple binding sites are available [72]. Still, we know next to nothing regarding how GLD-1 actually confers translational repression. It is intriguing that several proteins previously associated with miRNA regulation biochemically interact with GLD-1. CGH-1 interacts with the miRISC complex and GLD-1 in an RNA-dependent manner [5,66], and therefore raises the possibility of being the mediator between the two translational repression mechanisms. In a recent study, one of the C. elegans poly(A) binding proteins, PABP-2, was shown to antagonize let-7 miRNA function [73]. In our study, we show that PAB-1 is required for proper let-7 function and this is in line with the interactions between PAB-1 and AIN-1 [63]. These results indicate that different poly(A) binding proteins may affect miRNA function in opposite ways and perhaps these effects depend on other RNA-binding proteins like GLD-1. We therefore speculate that the mechanisms of GLD-1-mediated translational repression and miRNA-mediated translational repression may overlap. It will be interesting to investigate this potentially-shared mechanism in the future.
By taking advantage of multiple genetically sensitized miRNA pathways that act during C. elegans development to reliably assess subtle changes in gene expression, we could implicate GLD-1 in miRNA-mediated gene regulation. The mammalian GLD-1 homologue QKI has been reported to colocalize and interact with ALG2 [74], and recently QKI was shown to directly interact with and stabilize miR-20a, revealing a role for QKI in tumour suppression [75]. In a more recent study, two QKI isoforms were shown to regulate miR-7 expression in glial cells [76]. Studying the interactions between GLD-1-and miRNA-mediated gene regulation can thus reveal important regulatory interactions occurring during animal development.