miR-58 family and TGF-β pathways regulate each other in Caenorhabditis elegans

Despite the fact that microRNAs (miRNAs) modulate the expression of around 60% of protein-coding genes, it is often hard to elucidate their precise role and target genes. Studying miRNA families as opposed to single miRNAs alone increases our chances of observing not only mutant phenotypes but also changes in the expression of target genes. Here we ask whether the TGF-β signalling pathways, which control many animal processes, might be modulated by miRNAs in Caenorhabditis elegans. Using a mutant for four members of the mir-58 family, we show that both TGF-β Sma/Mab (controlling body size) and TGF-β Dauer (regulating dauer, a stress-resistant larval stage) are upregulated. Thus, mir-58 family directly inhibits the expression of dbl-1 (ligand), daf-1, daf-4 and sma-6 (receptors) of TGF-β pathways. Epistasis experiments reveal that whereas the small body phenotype of the mir-58 family mutant must invoke unknown targets independent from TGF-β Sma/Mab, its dauer defectiveness can be rescued by DAF-1 depletion. Additionally, we found a negative feedback loop between TGF-β Sma/Mab and mir-58 and the related mir-80. Our results suggest that the interaction between mir-58 family and TGF-β genes is key on decisions about animal growth and stress resistance in C. elegans and perhaps other organisms.


INTRODUCTION
Although around 60% of protein-coding genes are estimated to be under the control of microRNAs (or miRNAs) and despite the general assumption that miRNAs are able to co-ordinate the expression of multiple mRNAs, the path to elucidate miRNA functions and/or interactions with target genes remains a challenge (1)(2)(3). For instance, in Caenorhabditis elegans, only 10 out of 434 mature miRNAs have been experimentally validated as regulators of particular target genes (miRecords April 2013; miRBase Release 21). This slow progress is caused, in part, by the fact that only a small minority of miRNAs shows an obvious phenotype when mutated (4). Hence, one way forward may be to study strains with loss-of-function mutations at various redundant miRNAs, as opposed to mutants that only affect single miRNAs. This approach not only increases our chances of observing phenotypic anomalies that we can relate to other mutants, but also of identifying target genes by observing their altered expression patterns.
In this way, Alvarez-Saavedra & Horvitz identified three miRNA-family mutants of C. elegans with obvious abnormalities that were mostly absent in their respective single miRNA mutants (5). One of the aberrant family strains was a quadruple mutant for the cel-mir-58 family, which had small body size and was dauer defective. It is known that both, body size and dauer formation in worms, depend on two analogous signalling pathways, Transforming Growth Factor (TGF)-␤ Sma/Mab and TGF-␤ Dauer, respectively (6). We here ask whether the miR-58 family could regulate one or both pathways in C. elegans.
TGF-␤ pathways are key in the proliferation and differentiation of animal cells, and their core signalling pathway is conserved across metazoa (7). Upon extracellular binding of a dimeric ligand (TGF-␤) to two pairs of transmembrane receptors (type I and type II), the resulting hexameric complex becomes phosphorylated at type I receptors. Then, a cascade of intracellular phosphorylations at various transcriptions factors known as Smads is triggered, which, depending on the cellular context, induces or represses specific sets of target genes. In worms, the number of proteins and pathways involved are much smaller (e.g. five ligands, three receptors) than in mammals, which simplifies its study. The two known TGF-␤ pathways in C. elegans are the already mentioned TGF-␤ Sma/Mab and TGF-␤ Dauer.
TGF-␤ Sma/Mab is the best-known pathway controlling body size in C. elegans. Ligand DBL-1 (TGF-␤), which is expressed in the ventral cord (8,9), promotes growth by its effects on hypodermis (10,11), where SMA-6 (type I) and DAF-4 (type II) receptors are present (12). Worms homozygous for null mutations at several genes of this pathway show a Sma phenotype, that is, they are dwarfed, similar in length to worms defective in the mir-58 family (6).
TGF-␤ Dauer is one of the signalling cascades that regulates dauer formation. Dauer is an alternative L2-L3 larval stage that is resistant to harsh environmental conditions, such as low food, high population density, or high temperatures. When worms sense those challenges, daf-7 (TGF-␤) transcription, carried out in a pair of head sensory neurons known as ASI, is silenced. In turn, this downregulation leads to a dauer phenotype (13,14). DAF-7 receptors are DAF-1 and DAF-4 (this last one shared with TGF-␤ Sma/Mab), and they are broadly expressed (12). Downstream Smad include activators DAF-8 and DAF-14, and inhibitors DAF-3 and DAF-5 (15)(16)(17).
What else do we know about the mir-58 family apart from the fact that its absence leads to a small and dauer-defective mutant? This family is made of five members, mir-58 (chromosome IV), mir-80 (III), mir-81 and mir-82 (approximately 4 kb apart from each other in chromosome X), and mir-1834, although this last one has not been fully validated as a functional miRNA (chromosome IV; >3 Mb apart from mir-58) (18). The strain family mutant, MT15563, referred also as mir-58f(-) hereafter, holds three deletions covering the first four miRNAs. mir-58f(-) was described as sluggish, small, and egg-laying and dauer defective (5). In contrast, single or double mir-58-family deletions do not result in obvious developmental defects. In consequence, mir-58(-) apparently develops normal and only shows a 20% shorter life span (19), which is surprising considering that miR-58 is the miRNA with the highest (20), or one of the highest levels of expression (21), at every developmental stage and across multiple tissues, with the significant exception of the nervous system (22). A lower, but still broad expression has been described for miR-80 and less so for miR-82 (20). However, these two are found in neurons, and the later is only expressed from L4 onwards. In contrast to the previous three miRNAs, miR-81 has only been weakly detected in head neurons from L1 (22,23). These tissue-specificity patterns of expression suggest that cel-miR-58 members could have redundant as well as divergent functions.
There are mir-58 orthologs in other invertebrates, like Drosophila, where it is known as bantam (24,25). According to some reports several human mir-58 orthologs also exist, although this is not firmly established (21,26,27). Interestingly, bantam is among the few mutated miRNAs with an obvious phenotype in Drosophila, and like in nematodes, those fruitflies are small (28,29). A number of biochemical pathways have been related to bantam, both upstream and downstream of it (30,31). Drosophila's DPP/TGF-␤ is known to be one of the pathways upstream of bantam (32)(33)(34).
In this work, we primarily focus on the relationship between the mir-58 family and TGF-␤, Sma/Mab and Dauer, in C. elegans. We find that various genes from both TGF-␤ pathways are controlled by the mir-58 family. We also find a positive regulation of mir-58 transcription by TGF-␤.

Epistasis analyses
To study body size, synchronized worms grown individually on 5 cm NGM Petri dishes were measured at 24 h intervals since around 100 h post-hatching until they reached plateau sizes. Final lengths were calculated on averages of two to three consecutive days, when the worms had reached a plateau size. Images were captured using a video camera (KY-F550, JVC Professional Europe Ltd.) attached to a dissecting microscope (x50, Leica MZ7.5), and images were analysed with ImageJ software (1.46r; National Institutes of Health). A minimum of two independent experiments was carried out for each strain and treatment, and a range of 20-139 worms was measured per strain.
For epistatic dauer assays, we made worms deficient in both daf-1 and the mir-58 family by repeatedly backcrossing DR609 males to MT15563 hermaphrodites until we obtained a compound mutant from both. We followed the same protocol to create daf-2(e1370);mir-58f(-) mutants. Dauer assays were performed as follows: young adult hermaphrodites of N2, MT15563, DR609, CB1370, daf-1(m213);mir-58f(-) and daf-2(e1370);mir-58f(-) mutants were let to lay eggs for 4 h at 20 o C. After removing the hermaphrodites, NGM plates were incubated at 20 o C or 25.5 o C up to 3 days and dauers were scored. Animals were anesthetized with 25 mM NaN 3 in M9 and pictured with a confocal microscope (Leica DMI 6000) under 63x objective (for pharynx images) and Zeiss Axio Imager.A1 microscope with 100x objective under Nomarski conditions (for alae).
Statistical pairwise comparisons were run on R version 2.15.1. We first checked whether each data set had a normal distribution, by using the Shapiro-Wilk normality test. In case normality was rejected, data would be logtransformed. Then, pairwise comparisons were performed with a Welch Two Sample t-test. The same procedure was also used for obtaining P-values throughout this report, with the exception of the qPCRs for which we used a specific methodology (see below).

Quantitative real-time PCR assay (qPCR)
We synchronized worms and isolated their total RNA with miRNeasy Kit (Qiagen), then used to quantify both mRNA and miRNA expression.
For quantification of mRNAs, we synthesized cDNAs with SuperScript III First-Strand Synthesis System (Invitrogen). Real-time PCR of cDNAs was run on an Applied Biosystems 7500 Fast thermocycle, using TaqMan Assays (Applied Biosystems). We tested each sample in triplicate, and analysed relative changes of transcripts by the 2-CT method (36), using levels of crn-1 and cey-1 transcripts as control. P-values were calculated with REST software (37).
For miRNA assessment, we conducted reverse transcription and real-time quantification for each miRNA by using TaqMan MicroRNA Assays (Applied Biosystems) and the same thermocycle than above. The normalization of miRNA real-time PCRs was performed with geNorm (38) and NormFinder (39) algorithms, whereby the two most stable miRNAs across samples were miR-47 and miR-81 for TGF-␤ Sma/Mab genes, and miR-81 and miR-82 for the TGF-␤ Dauer pathway.

Luciferase assays
We grew HeLa cells in Dulbecco's modified Eagle's medium (Lonza) with 10% FBS in 5% CO 2 at 37 o C and seeded them at a density of 1 × 10 4 cells per well into 96-well plates. Cells were transfected in triplicate 24/48 h later with Lipofectamine 2000 (Invitrogen), 150 ng of a 3 UTR luciferase vector (see below), and 50 nmol of test miRNA mimic (miR-58, miR-80, miR-81 and miR-1834; miRIDIAN, Dharmacon) or the standard control miRNA mimic miR-67 provided by the manufacturer. The 3 UTRs (wild-type and mutant) of sma-6, dbl-1, daf-1, daf-7 and daf-4 were cloned after the renilla reporter gene in the vector psiCHECK-2 (Promega). For obtaining the 3 UTR mutants, we mutated the second, third and fourth positions of the predicted mir-58 family binding sites in 3 UTRs by PCR (primers in Supplementary  Table S3). Dual-luciferase reporter assay were performed 48 h after transfection using Dual-Luciferease Reporter Assay System (Promega) to detect firefly and renilla luciferase activity, and luminescence was measured with an Infinite M200 TECAN luminometer (TECAN Group Ltd). Renilla luciferase activity was first normalized using the firefly luciferase activity as intraplasmid transfection reporter. Resulting values for miRNA-3 UTR co-expression were further normalized to those from the same 3 UTRs but incubated with control mimic miRNA (miR-67).

In vivo 3 UTR activity assays
To test in vivo 3 UTR activity, we made sensor mCherry constructs with either the wild-type 3 UTR of each target gene or a mutated version as described in the luciferase assays and Supplementary Table S3. To obtain reliable mCherry quantifications when comparing the transgenic lines generated in the wild-type N2 or mir-58f(-) MT15563 backgrounds we did two things. First, each transgenic line generated in the N2 background were outcrossed to the MT15563 strain, in order to obtain worms in a mir-58f(-) background with the same arrays, or as similar as possible, than their N2 counterparts. That required backcrosses through various generations and subsequent genotype confirmation. Secondly, we included a P sca-1 ::gfp construct in all our arrays, so that we could use GFP levels to normalize the expression of mCherry in every worm. Thus, mCherry constructs were mixed with P sca-1 ::gfp (pGK10 plasmid, (40)), hygromycin resistant vector pHygroSfi (a gift from Dr J. Pérez-Martín) and a 1 kb DNA ladder (Invitrogen; see Supplementary Table S4 for concentrations of each component). For generating transgenic lines, we used standard microinjection techniques (41). Most often, to calculate reporter expression, we used two or more transgenic lines from each microinjected mixture. Animals were mounted in agar pads and pictured at specific stages under a Leica stereomicroscope (M165FC) attached to a camera (Leica DFC360FX). Pictures were analysed with ImageJ by measuring the whole-body area of worms.

RNA interference
Synchronized L4 worms were fed HT115 (DE3) bacteria expressing sma-6 or daf-4 dsRNA on NGM plates. Adult body size of the progeny was measured (as described in 'Epistasis analyses') every 24 h, up to around 140 h, when they had already reached their final body size. The sma-6 RNAi plasmid included a genomic fragment of sma-6, obtained by PCR with forward 5 -CAGCTGTACACGGAACTGG and reverse 5 -TCAACTTTACGCTGCGATTG oligonucleotides. This fragment was inserted into L4440 vector (Addgene) and transformed into HT115 (DE3) bacteria. The daf-4 RNAi plasmid was purchased from Source BioScience (C05D2.1/III-2P04). Clones were cultured overnight in LB with 200 g/ml ampicillin and 12.5 g/ml tetracycline. We spread 120 l of this bacterial culture onto each NGM plate supplemented with 1 mM IPTG and 25 g/ml carbenicillin. Plates were left overnight to induce dsRNA before adding the worms. We measured at least 20 worms per strain in each of two independent experiments.

Quantification of hyp 7 nuclei
For comparing the number of hyp 7 nuclei between N2 and MT15563, we microinjected both strains with a hypodermal marker, P dpy-7 ::4xNLS::gfp, together with pHy-groSfi and an mCherry microinjection marker (p374, a gift from Dr A. Miranda-Vizuete (42); Supplementary Table  S4). Thus, GFP expression was under the control of dpy-7 promoter (43) and, because of its nuclear-localization sequence (NLS), it was only localized at nuclei. To make this construct, we replaced the rpl-28 promoter of L4455 vector (Addgene) by a 616 bp fragment encompassing the dpy-7 promoter (see Supplementary Table S2). We used a Leica stereomicroscope (M165FC) to count hyp 7 nuclei in 18 animals of each strain, which were kept at approximately 5 o C to keep them immobile during nuclei counting. We only counted hypodermal nuclei from one side of each worm, excluding dorsal and ventral nuclei.

TGF-␤ Sma/Mab reporter activity
We chose sma-6 promoter as reporter of the Sma/Mab pathway, as it is known that such pathway positively controls sma-6 transcription (44). Our P sma-6 plasmid construct was very similar to previously described P sma-6 ::mCherry::3 UTR sma-6WT (see 'In vivo 3 UTR activity assays'). The only difference is that here the 3 UTR was the original unc-54 present in pPD95.77. We did not include 3 UTR sma-6 to limit all expression differences to promoter control, excluding differences based on miRNAs. Transgenic worms were generated by microinjection of N2 with P sma-6 ::mCherry::3 UTR unc-54 , together with a GFP marker (pGK10) and the other DNAs already mentioned ('In vivo 3 UTR activity assays'; Supplementary Table  S4). Several independent transgenic lines generated in the wild-type background were repeatedly backcrossed to MT15563, to test whether the TGF-␤ Sma/Mab signalling is up or down in the mir-58 family mutant with respect to N2. As control, we also crossed the transgenic lines with either NU3 dbl-1(nk3) or BW1940 dbl-1(++), for which Sma/Mab is known to be down-and up-regulated, respectively (8,9). Pictures and analysis of L4 worms, again referring mCherry reporter levels to GFP control expression of the extrachromosomal array, were performed as previously explained ('In vivo 3 UTR activity assays').

TGF-␤ Dauer pathway reporter activity
For estimating TGF-␤ Dauer pathway activity in mir-58 mutants, we used daf-7 and daf-8 that are known to respond positively to this pathway (45). For quantification of transcriptional activation of both genes, we performed qPCR in mixed stage populations of a variety of genetic backgrounds using specific daf-7 and daf-8 TaqMan probes (Applied Biosystems).

Transcriptional regulation of P miR-58
EUB0032 P mir-58 ::gfp was crossed with NU3 dbl-1(nk3), RB1739 sma-10(ok2224), RB2589 daf-3(ok3610), DR609 daf-1(m213) and DR63 daf-4(m63) to test whether the inhibition of the TGF-␤ pathways may regulate the activity of P mir-58 . To check how the increase of Sma/Mab signalling affects P mir-58 regulation, we microinjected the fosmid WRM0624CB02, which carries the dbl-1 gene, together with other DNAs (see Supplementary Table S4 for transgene details and concentrations) into EUB0032. As a control of this last experiment, we microinjected the strain EUB0032 with the same DNA mixture except for the dbl-1 fosmid. In all these tests we analysed GFP expression in L4 worms as previously described.

RESULTS
Our hypothesis was that the mir-58 family could regulate body size and dauer response through TGF-␤ Sma/Mab and TGF-␤ Dauer pathways, respectively. Then we first carried out a computational search looking for miR-58 targets related to these two pathways. Our quest in eight different databases (Microcosm targets, microrna.org, Target Scan, PICTAR, mirWip, Diana Lab, RNA22 and mirSom) rendered three putative miR-58 target genes belonging to the TGF-␤ Sma/Mab pathway: dbl-1, daf-4 and sma-6. Genes daf-1 and daf-7 -apart from daf-4-all of the TGF-␤ Dauer pathway, also became predicted targets of miR-58. The predicted binding sites for each of these genes are summarized in Supplementary Table S5.

mRNA levels of dbl-1, sma-6, daf-4, daf-1 and daf-7 are upregulated in the mir-58 family mutant
We performed quantitative real-time PCR (qPCR) on mir-58f(-), the C. elegans strain missing four of the five miRNAs of the mir-58 family (see Materials and Methods). As shown in Figure 1 (mixed stage, bottom panel), the expression of dbl-1, daf-4, daf-1, daf-7 and most notably sma-6 (sixfold), was increased in mir-58f(-) with respect to N2. We also separately analysed the mRNA levels of those TGF-␤ genes at L1, L2 and L4 stages ( Figure 1). We chose those stages because of our interest in the TGF-␤ Dauer and the Sma/Mab pathways, which are supposed to be key in L1-L2 and L4, respectively. The highest mRNA induction was presented by sma-6, at any stage. daf-1 and daf-4 showed moderate although clearly statistically significant increments at all stages. In relation to dbl-1, L1 and L4 but not L2, showed a significant but moderate rise. Surprisingly, in mixed stage we observed the highest dbl-1 mRNA induction (threefold). Finally, daf-7 showed a variable stage-dependent increase ranging from no significant change at L2 to fivefold increment at mixed stage. In the cases where daf-7 is upregulated, the data showed high standard deviations, indicating large variability between the four independent experimental replicas.

Luciferase reporter assays suggest that miR-58 family members directly regulate TGF-␤ genes
qPCR experiments can not discriminate between direct and indirect genetic interactions. That is why we ligated the Figure 1. dbl-1, sma-6, daf-4, daf-1 and daf-7 mRNA levels are upregulated in the mir-58 family mutant. mRNA levels of the above genes were measured in larval stages L1, L2, L4 and a mixed stage population of wild-type strain N2 and MT15563 (mir-58f(-)), which lacks four miRNAs of the mir-58 family. Measurements were carried out by quantitative Real-Time PCR (qPCR) with gene-specific TaqMan probes. mRNA levels of MT15563 (grey bars) were normalized to those of N2 (white bars). Each value represents the average from four independent experiments. Error bars indicate standard deviations. Significant statistical differences between N2 and MT15563 are indicated as *(P < 0.01) or **(P < 0.001).

Figure 2.
Luciferase reporter assays show that miR-58 family members effectively repress gene expression through the 3 UTRs of TGF-␤ genes. Human HeLa cells were transiently transfected with psiCHECK-2 vector containing either wild-type (white) or mutated (grey) 3 UTRs from TGF-␤ genes dbl-1, sma-6, daf-4, daf-1 and daf-7, along with miR-58 family mimics of miR-58, miR-80, miR-81 and miR-1834, or the unrelated miR-67 as negative control. The luciferase activity for each mimic was normalized to the value obtained with miR-67 using the same 3 UTR (dashed line). Data shown are representative of, at least, two independent experiments. Error bars indicate standard deviations. *P < 0.005, comparing the luciferase activity corresponding to each inhibitory miRNA and miR-67.
-80, -81 or -1834. All these miRNAs belong to the mir-58 family, having small differences among themselves (miR-81 and -82 differ only in one nucleotide, and therefore we only used miR-81). mir-1834 is just known for its presence in the C. elegans genome, but there are no mutants available or any other functional feature known about it.
With the exception of daf-7 3 UTR, we confirmed that all the members of miR-58 family (including the uncharacterized miR-1834), but not our negative control miR-67, strongly reduced the expression of the luciferase when placed in front of dbl-1, sma-6, daf-4 and daf-1 3 UTRs (white bars in Figure 2). This translational repression Nucleic Acids Research, 2015, Vol. 43, No. 20 9983 was dependent on the predicted miRNA binding sites in 3 UTRs, because when these sites were mutated at only three nucleotides the luciferase repression was abolished without exception (grey bars in Figure 2; Supplementary  Table S3). Apparently, the 3 UTRs of daf-1 and daf-4 are the most efficiently mediators of genetic repression by the miR-58 family, as their means were the lowest for any tested miRNA. The 3 UTR of sma-6 could also inhibit luciferase expression with any of the miR-58 family members tested. In contrast, dbl-1 3 UTR activity seemed more restricted. We did not observe luciferase repression in the case of miR-58. Finally, the 3 UTR of daf-7 seemed unable to inhibit luciferase gene expression through any of the miRNAs (Figure 2), with the possible exception of miR-58, because of a small but statistically significant difference with respect to control miR-67.

In vivo assays of 3 UTR inhibitory activity confirm that genes of the TGF-␤ Sma/Mab and Dauer pathways are regulated by miR-58 family
With the possible exception of daf-7, all the other assayed TGF-␤ genes (dbl-1, sma-6, daf-4 and daf-1) have shown, according to qPCR and luciferase experiments, some degree of downregulation by mir-58 or other members of its family. In order to confirm a direct regulation of those four genes by the miR-58 family, we performed in vivo assays. Thus, transgenic animals carrying an mCherry reporter ligated to every correspondent promoter and 3 UTR were tested in N2 versus a mir58f(-) background.
As explained in Materials and Methods, to account for spurious oscillations in mCherry intensity, due for instance to differences in extrachromosomal array copies, each transgenic isolate generated in N2 was outcrossed with mir-58f(-) worms, to transfer each extrachromosomal array between both backgrounds. Additionally, we coinjected mCherry constructs together with a P sca-1 ::gfp DNA in all our assays, so that we could use GFP levels as a control of normalization of the mCherry expression ( Figure 3).
All the four wild-type 3 UTR tested, dbl-1, sma-6, daf-4 and daf-1, rendered more mCherry activity in a mir-58f(-) background than in N2, suggesting that miR-58-family members effectively bind to TGF-␤ 3 UTRs ( Figure 3A and C). However, whereas the repression observed in N2 with the 3 UTRs of dbl-1 and daf-1 was moderate (2-4 fold), such repression was much more pronounced with sma-6 and daf-4 3 UTRs (12-15 fold). In relation to sma-6 we initially observed a massive 90-fold difference (second row from the top of Figure 3). However, it is known that the sma-6 promoter is positively regulated by the TGF-␤ Sma/Mab pathway (44,46) and, as we will show below, TGF-␤ Sma/Mab is in fact more active in mir-58f(-) than in N2 worms ( Figure  4). Accordingly, to assess the inhibitory activity of the sma-6 3 UTR independently from its own promoter, we repeated the same experiment but with the daf-4 promoter instead. Then, in fact, the difference in mCherry intensity between the N2 and mir-58f(-) was not as striking as before (now around 12-fold, third row from the top, Figure 3).
In all cases, when 3 UTRs were mutated in their predicted miR-58f binding sites, the mCherry fluorescence difference between N2 and mir-58f(-) disappeared or was largely abolished ( Figure 3B and C).
For the assessment of the activity of the TGF-␤ Dauer pathway we used daf-8 and daf-7 mRNA levels. Both are not only components of the pathway but may also be taken as TGF-␤ Dauer sensors, as they are transcriptionally repressed by downstream Smad transcription factor DAF-3, which is itself antagonized by the TGF-␤ Dauer pathway (15). The more active this pathway is, the higher is the activation of both daf-8 and daf-7, which leads to an inhibition of the dauer phenotype (45). Do mir-58f(-) worms, unable to enter dauer (5), have higher daf-8 and daf-7 mRNA content than N2? We measured daf-8 and daf-7 transcripts in mixed-stage by qPCR. As positive control of the pathway activity, we included daf-3(ok3610), a DAF-3 defective worm known to overexpress daf-8 and daf-7 mRNAs (45). We also included daf-4(++) (47) to find out whether the sole overexpression of the DAF-4 receptor was enough to increase the signalling, similarly as mir-58f(-) may do. Figure 5 shows that both, daf-8 and daf-7 mRNA levels (especially the second) were significantly elevated in mir-58f(-), and in daf-4(++) and daf-3(ok3610), compared to N2. These results are consistent not only with a higher expression of daf-4 and daf-1 in mir-58f(-) worms, but also with its dauer-defective phenotype.
We conclude that both TGF-␤ pathways, Sma/Mab and Dauer, are upregulated in the absence of miR-58 family members.

miR-58 family influences body size both dependent and independently of TGF-␤ Sma/Mab signalling pathway
As described by Alvarez-Saavedra and Horvitz (5), the mir-58 family mutant presents a highly reduced body size, similar in length to TGF-␤ Sma/Mab depleted animals. To tackle this issue we first aimed to study how each of the mir-58 family members contributes to adult body size. Therefore, we measured the length of worms carrying each of the three null mutations present in mir-58f(-) separately, and also the corresponding three double mutants. We observed that neither mir-80(nDf53), nor mir-81&mir-82(nDf54), or a combination of both, showed any significant reduction in length with respect to N2 ( Table  Figure 3. miR-58 family reduces the expression of dbl-1, sma-6, daf-4 and daf-1 through their 3 UTR binding sites in vivo. Transgenic worms expressing sensor mCherry constructs (see diagrams to the left) with either the wild-type (A) or mutated version of the 3 UTR of each target gene (B), namely dbl-1, sma-6, daf-4 and daf-1, were generated in N2. Two to three independent transgenic lines were crossed into MT15563 (mir-58f(-)) to compare mCherry expression between wild-type and mir-58f(-) backgrounds. Mean mCherry signal was referred to the mean expression of a coinjected P sca-1 ::gfp, to account for varying copy numbers of extrachromosomal arrays. (C) Quantification of the normalized mCherry expression using approximately 20 animals per construct and genetic background. Error bars indicate 95% confidence intervals. **P < 0.001, compares normalized mCherry averages between wild-type (blue) and mir-58f(-) (red) backgrounds for the same constructs.  We used daf-8 and daf-7 expression as a sensor of the activity of TGF-␤ Dauer (45). daf-8 and daf-7 mRNA levels were measured by qPCR with specific TaqMan probes in synchronized mixed stage worms of shown mutants compared to wild-type N2 (white bars). Error bars indicate standard deviations, and *P < 0.001 significant differences between each strain and N2. Each value represents the average of four independent experiments. 1). However, mir-58(n4640) showed such a reduction (P < 0.001), which was even more pronounced in the company of mir-80(nDf53) (P < 0.001) but not of mir-81&mir-82(nDf54) (P = 0.2). When the three deletions were present the length reduction was even larger (P < 0.001 with respect to mir-58(n4640) alone or together with any of the other two mutations). We conclude that miR-58 is the miRNA that contributes the most on body size, followed by miR-80 and the tandem miR-81/-82 (in that order).
We then asked whether double mutants for the TGF-␤ Sma/Mab pathway and the miR-58 family, both of a similar length on their own, would reach a similar size or become even smaller. Our results showed that the double mutant mir-58(n4640);dbl-1(nk3) was significantly shorter than worms with either mutation alone (P < 0.001; Table 1). mir-58(n4640);ctIs40(dbl-1++) animals presented an intermediate size between the shorter mir-58(n4640) and the longer DBL-1 overexpressing worm ctIs40(dbl-1++). As an alternative approach, we inhibited two Sma/Mab genes by RNAi, sma-6 and daf-4, in a mir-58f(-) background. The depletion of any of those two genes was able to further diminish the length of mir-58f(-) ( Table 1). We conclude that the effect of the miR-58 family on body size is at least partly independent of TGF-␤ Sma/Mab pathway.
Next, we tested and quantified the presumptive inhibitory effect of mir-58f on growth through the TGF-␤ Sma/Mab pathway, without the opposite (i.e. positive) effects on growth via other genes or signalling pathways. For that purpose, a sma-6 construct, that includes its 3 UTR either wildtype or mutated in its putative mir-58f binding sites, was microinjected in sma-6(wk7) defective worms. Overexpression of sma-6 with the mutated 3 UTR makes worms 12% longer than N2 (P < 0.001), whereas the same construct but with the wild-type sma-6 3 UTR results in smaller worms (only 5% longer than N2; P < 0.001) ( Figure 6 and Table  1). Importantly, the 7% difference in favour of the worms overexpressing sma-6 with mutated 3 UTR, compared to those with a wild-type 3 UTR, is statistically significant (P < 0.001). As negative control, both sma-6 constructs were microinjected in mir-58f(-) mutants. We expected to detect no difference in length between worms carrying the wildtype 3 UTR extrachromosomal array and those expressing the mutant 3 UTR construct. Indeed no significant difference was found between both 3 UTRs ( Figure 6 and Table  1). These results confirm that, independently of its promoting effects on growth through other pathways, mir-58f also inhibits growth through TGF-␤ Sma/Mab.  Figure 6. Inactivation of miR-58f binding sites on sma-6 3 UTR makes C. elegans longer. Adult worm length was estimated in animals that overexpressed genomic sma-6 with either wild-type or mutated 3 UTR in sma-6(wk7) defective background (A) or mir-58f(-) worms (B). Data represent the average of two independent experiments. Error bars indicate 95% confidence intervals and significant statistical differences between strain average length are indicated as **(P < 0.001).

miR-58 family mutants do not enter dauer due to upregulation of TGF-␤ Dauer pathway
Another interesting phenotype of mir-58f(-) is its inability to enter the dauer stage (5). Could that be related to our finding that the TGF-␤ Dauer pathway, which is dauer inhibitory, is upregulated in mir-58f(-)? We blocked the TGF-␤ Dauer pathway in mir-58f(-) worms by impairing daf-1, and ask whether dauer formation could be rescued. Figure 7A shows that mir-58f(-);daf-1(m213) mutants produced an even higher percentage of dauers than daf-1(m213) at 20 o C, whereas both genotypes gave a 100% of dauers at 25.5 o C, fully overturning mir-58f(-)'s dauer deficiency. We did not observe additional phenotypes, such as embryo or larval lethality. To verify the presence of true dauers in mir-58f(-);daf-1(m213) plates, we first looked at their sodium dodecyl sulphate (SDS) resistance, which confirmed them as dauers. Secondly, we looked for alae structures in animals looking like dauers. As seen in Figure 7E, mir-58f(-);daf-1(m213) exhibited alae very similar to those of daf-1(m213) dauers ( Figure 7D), whereas, as described, they were missing from mir-58f(-) (5). Thirdly, we compared the pharynx morphology of mir-58f(-);daf-1(m213) and daf-1(m213) and found that both had constricted pharynges and sealed mouths, as expected for dauers ( Figure  7C and B, respectively). In summary, our results indicate that mir-58f(-) is unable to form dauers because of an hyperactivation of the TGF-␤ Dauer pathway. Could other molecular routes be involved as well? Although we have just showed that an upregulation of TGF-␤ Dauer is enough to explain the dauer-deficient phenotype of mir-58f(-) (Figure 7), we also checked the insulin/IGF-1 signalling (IIS) pathway. First, we gauged the activity of IIS in mir-58f(-) animals using sod-3 as reporter. Because IIS is dauer inhibitory and sod-3 is repressed by IIS, if IIS was responsible of the dauer deficiency of mir-58f(-), we would expect IIS to be upregulated and sod-3 mRNA to be downregulated in mir-58f(-). In contrast, we found a nearly threefold increment in sod-3 transcript levels in mir-58f(-) L1 worms compared to N2 (Figure 8). Additionally, we found daf-16, ins-1 and ins-17 also upregulated in mir-58f(-) (Figure 8). Because the dauer stage depends on the presence of DAF-16 (48), and INS-1 and INS-17 are insulin antagonists (49,50), our results clearly show a downregulation of IIS activity in mir-58f(-), which again is inconsistent with a responsibility of IIS in the dauer deficiency of mir-58f(-). Furthermore, we produced a mir-58f(-);daf-2(e1370) mutant (DAF-2 is the receptor of IIS), and observed that it had a proportion of arrested larvae which resembled dauers (Supplementary Table S6). However, after close inspection, we observed that they were able to pump, that they were unable to survive 1% SDS treatments, that they lacked the characteristic cuticle alae of true dauers, and that their mouths were open and their pharynges unconstricted (Supplementary Figure S1). We conclude that mir- 58f(-);daf-2(e1370) worms do not enter dauer and that IIS does not mediate mir-58f(-) dauer defective phenotype.
Using the same two previous approaches, we assessed the transcriptional levels of mir-58 in the context of TGF-␤ Dauer pathway. We found that the downregulation of In the Sma/Mab branch, this inhibition goes through ligand DBL-1 and receptor proteins SMA-6 and DAF-4. Therefore, miR-58 family activity is growth-inhibitory through Sma/Mab. In parallel, it also acts as a larger growth-promoting signal, by repressing unknown growth inhibitors. Additionally, TGF-␤ Sma/Mab positively regulates the transcription of mir-58 and mir-80, thus creating a negative feedback loop. In the TGF-␤ Dauer pathway, the miR-58 family targets receptors DAF-1 and DAF-4. This inhibition shall inactive DAF-14/DAF-8 dimers, which will allow the activation of downstream antagonistic DAF-3. Through DAF-3, DAF-8 and DAF-7 will be transcriptionally downregulated and therefore dauer entry will be promoted. In summary, the miR-58 family modulates both TGF-␤ routes, acting as a new hub that may coordinate growth and dauer decisions.

DISCUSSION
One way to facilitate the identification of miRNA targets is to study miRNA-family mutants as opposed to single mutants. In this way, we are more likely to observe abnormal phenotypes as well as interactions between those miRNAs and other loci. This is what we have done for the mir-58 family (or mir-58f), by using the strain MT15563 (mir-58f(-)), a C. elegans strain missing four members of that family: mir-58, -80, -81 and -82 (5). Our results are summarized in the model of Figure 10.

Control of TGF-␤ genes by the mir-58 family
Our gene reporter assays strongly suggest that mir-58f directly downregulates four genes (sma-6, daf-4, daf-1 and Nucleic Acids Research, 2015, Vol. 43, No. 20 9989 dbl-1) from the TGF-␤ signalling pathways, Sma/Mab and Dauer, in C. elegans (Figures 2 and 3). Two recent papers (51,52), using high-throughput genomic methods, support our conclusion that miR-58f repress the mRNA levels of those genes through direct interactions with their corresponding 3 UTRs. In the more recent approach (51), they cross-link, ligate and sequence miRNA-mRNA hybrids to 'unambiguously identify miRNA:target site interactions'. Searching their data set, we found that miR-58, -80, -81 and -82, they all bind to a region in chromosome III that corresponds to the daf-4 3 UTR, and that perfectly match two of the three sites that we identified as target regions of miR-58 family (Supplementary Table S5). Using a second method (51,52), they sequenced thousands of 3 UTR fragments associated to Argonaute (ALG-1), and for this reason likely targets of miRNA regulation. In this way, they found three different 3 UTR fragments of sma-6 (57 bp the smallest) that contain our predicted miR-58f-binding sites. They also found two sequences of 47 bp and 162 bp, from the dbl-1 and daf-1 3 UTRs, respectively, again containing our predicted miR-58f-binding sites (Supplementary Table S5). In summary, recent high-throughput results provide supporting evidence for a direct physical association between miR-58f and the 3 UTRs of sma-6, daf-4, daf-1 and dbl-1. Of course, our results go beyond the mere description of those physical interactions and present some of their unfolding consequences in relation to gene expression, cell signalling and phenotypes.
It is interesting that for both, TGF-␤ Sma/Mab and Dauer, we find that miR-58f is acting at the beginning of their signalling cascades, at the ligand/receptor level (Figures 2 and 3). Perhaps that has been a feature evolutionarily selected on the basis of higher biochemical efficiency. However, our luciferase assays do not really support TGF-␤ Dauer ligand mRNA, daf-7, as direct target of miR-58f ( Figure 2). We think that the overexpression of TGF-␤ Dauer receptor DAF-1/DAF-4, in mir-58f(-) animals, might trigger a positive feedback loop that would eventually increase the levels of DAF-7. In this way, an overexpression of TGF-␤ Dauer receptor would activate downstream Smads DAF-14/DAF-8, which in turn would inhibit another Smad, DAF-3, that is a transcriptional repressor of daf-7 and daf-8 (45). In consequence, perhaps an even small increment of daf-4 and daf-1 transcripts could lead to a circular self-reinforcement of the TGF-␤ Dauer pathway, giving rise in the process to increasing levels of daf-7 mRNA. Two of our results especially support this hypothesis. First, that the TGF-␤ Dauer pathway appears more active in mir-58f(-) than in N2, and with high levels of daf-7 mRNAs (Figures 1 and 5). Second, that the experimental overexpression of daf-4 alone suffices to trigger an activation of the TGF-␤ Dauer pathway, again involving higher expression of the endogenous daf-7 gene, similarly to what happens in mir-58f(-) ( Figure 5).
sma-6 is the most overexpressed mRNA at any developmental stage (Figure 1). We think that this is due to a combination of direct and indirect effects on sma-6 by mir-58f. Thus, it is known that the sma-6 promoter is positively regulated by its own TGF-␤ Sma/Mab pathway (44,46). In agreement with that, when we substituted P sma-6 by P daf-4 , in mCherry::3 UTR sma-6WT construct in our in vivo reporter assays, there was an 7-fold reduction in mCherry overexpression in mir-58f(-) worms. In spite of that reduction, the intensity of P daf-4 ::mCherry::3 UTR sma-6WT was still 15 times higher in mir-58f(-) than in N2 (Figure 3). Therefore, both the 5 -flanking region of sma-6 and its 3 UTR, must be responsible for upregulating sma-6 mRNA in the quadruple mir-58f(-) mutant (Figure 1). The overexpression of the other ligand in mir-58f(-), daf-7, is irregular but generally high nonetheless ( Figure 1). As previously discussed, that is most likely due to transcriptional activation (45).

Control of adult growth by mir-58 family
The most striking and recognizable feature of mir-58f (-) worms is their small body size (5). However, overexpression of DBL-1 results in longer, not shorter, animals (8,9). Therefore, it appears that the upregulation of the TGF-␤ Sma/Mab pathway in mir-58f(-) worms can not explain their small body size. In agreement with that, our epistasis experiments show that the effects of mir-58f and TGF-␤ Sma/Mab on body size are at least partly independent from each other ( Table 1), suggesting that other pathways affecting growth are regulated by mir-58f.
The fact that the worms with sma-6 3 UTRs altered in key miR-58f recognition sites were 7% longer than those with wild-type 3 UTRs leaves little doubt that miR-58f inhibits worm length through sma-6 (and presumably daf-4 and dbl-1 as well; Figure 6 and Table 1). Therefore, independently of its stimulating effect on body size through unknown growth-inhibitory genes, mir-58f also has an inhibitory influence on growth derived from its interaction with TGF-␤ Sma/Mab. Since mir-58f(-) worms are much smaller than N2, of course these two antagonistic actions are not equivalent. But perhaps subtle spatiotemporal differences in gene expression of either the target genes, and/or the mir-58f members, can modulate those two opposing effects, at times giving way to a predominantly inhibitory role to miR-58f.
In relation to the overall effect on growth of each miR-58 family member, we have shown that their contribution to body length is highest for miR-58, intermediate for miR-80 and lowest for miR-81 & -82 (Table 1). One possibility is that this is due to their relative abundances, which generally fits well with their expression levels (20,21). However, these miRNAs also differ on their expression patterns, and then we should expect some degree of specialization in relation to body size and other functions (22,23). As a first attempt to understand the relative importance of the various organs in relation to body size we performed rescue experiments using promoters specific of two of the tissues where miR-58 is present, gut and hypodermis, in mir-58f(-) worms. We observed an approximate 25% length recovery for each promoter (Supplementary Figure S3). Therefore, we suspect that miR-58 acts on body size through a combination of tissues.
We may speculate that an additional tissue involved could be the germline. We know that the germline inhibits body growth because when that tissue is ablated in larvae, adults become larger (53,54). Interestingly, it has been recently shown that mir-58f motifs are underrepresented in the germline transcriptome, suggesting a strong regulation by mir-58f at this tissue (55).
It is also interesting to compare worms with fruitflies in relation to miR-58f and growth. As previously mentioned, bantam is the sole mir-58 Drosophila ortholog, and the correspondent loss-of-function flies are small because of a reduction in cell number but not cell size (24,28). Then, we checked whether mir-58(-) nematodes also have a proliferation defect, but actually these mutants contain the same number of hypodermal nuclei at adulthood as N2 do (Supplementary Figure S4). We looked at the hypodermis because it is largely responsible for adult growth (10). In spite of those differences, Drosophila might also guide us to look for miR-58f target genes in miR-58f's growth-promoting task. It is possible that some of the miRNA targets were shared across far-related species, as it is the case for other miRNAs, like let-7 and two of its targets, RAS (56) and lin-41, present in Drosophila and vertebrates (57). Five targets of bantam have been experimentally validated so far, i.e. mei-P26, hid, enabled, capicua and Socs36E (58). ncl-1 is C. elegans' counterpart of mei-P26, and also a predicted mir-58f target by specialized software. Mei-P26 / NCL-1 is a zinc finger protein that is involved in cell proliferation and growth through the regulation of dMyc activity in Drosophila, whereas in C. elegans is thought to repress ribosome synthesis (59). That is consistent with the small phenotype of mir-58f(-), because then NCL-1 would be expected to inhibit growth.

Role of mir-58 family in dauer formation
Our finding that the TGF-␤ Dauer pathway, which represses dauer formation, is upregulated in mir-58f(-) opens up the question of whether it is that upregulation what makes the mir-58f(-) strain dauer defective. Our epistasis analyses support this view, because the quintuple mutant mir-58f(-);daf-1(m213) is dauer constitutive, as daf-1(m213) is, in consequence placing daf-1 downstream from mir-58f. In fact, the percentage of dauers at 20 o C is even larger in mir-58f(-);daf-1(m213) than in daf-1(m213), i.e. 93% versus 70%, respectively (Figure 7). One possibility why this might be so is that not all the four members of mir-58f behaved as dauer facilitators. Thus, one of them could be dauer repressor acting through a different signalling pathway, although masked by the action of the facilitators. If that was correct, when the TGF-␤ Dauer signalling pathway becomes inactive, like in a daf-1(m213) background, the only miRNA working independently from TGF-␤ Dauer pathway would be the dauerrepressor miRNA. As a result, there would be more dauers in mir-58f(-);daf-1(m213) than in daf-1(m213), as we observe. In accordance with this hypothesis, a recent paper by Than et al. suggests that mir-81, in contrast with the rest of its family, is a dauer repressor acting through the cyclic guanosine monophosphate signalling (cGMP) (60). On the other hand, miR-81 is much less expressed than miR-58 and -80 (21), and then this may restrict its potential to act as a dauer inhibitor in the company of the other members of the family. Also, in support of the above, we know that miR-81, in contrast to miR-58, -80 and -82, is the only one that is not able to rescue the dauer defective phenotype of mir-58f(-) when expressed by itself (5).
Although we can explain mir-58f(-) dauer defectiveness on the basis of daf-7, daf-1 and daf-4 overexpression, it would be possible that other mir-58f target genes may also be implicated in dauer regulation, just because individual miRNAs, let alone families, are supposed to control multiple genes (61). We investigated whether the insulin/IGF-1 signalling (IIS) pathway might be necessary for the dauerdefective phenotype of mir-58f(-), and we found that it is not. First, it is downregulated in mir-58f(-) (Figure 8) and that alone should favour dauer formation (i.e. IIS is dauer inhibitory). Second, the suppression of the insulin pathway in mir-58f(-) worms does not elicit the development of true dauers. Instead, mir-58f(-);daf-2(e1370) worms at 25 o C generate arrested larvae often resembling dauers but without their key features (i.e. cuticle alae, sealed buccal cavity, constricted pharynx; Supplementary Figure S1, Table  S6). Rather, it seems that mir-58f(-);daf-2(e1370) show an intermediate phenotype with respect to mir-58f(-) and daf-2(e1370). This gives support to the idea of a parallel signalling for miR-58f / TGF-␤ Dauer on the one hand and IIS on the other, with respect to dauer, which is in agreement with the existing literature as far as the TGF-␤ Dauer and IIS pathways are concerned (62). Recently Vora et al. observed that mir-80(-) worms are healthier and live longer than N2, and that this phenotype is daf-16 dependent (27). That is consistent with our results because the activities of DAF-16 and IIS antagonize each other (63), and therefore one could expect that mir-80(-) presented a somewhat downregulated IIS, which is what we observe in mir-58f(-) (Figure 8).

Autoregulatory feedback loops
We have shown that not only mir-58f negatively controls TGF-␤ Sma/Mab, but also that this pathway stimulates the transcription of mir-58 and mir-80. Therefore, a negative feedback loop is established between both molecular sets. However, the two less abundant members of mir-58f, miR-81 and -82 (21), do not significantly change their expression depending on the TGF-␤ Sma/Mab pathway (Figure 9A). Nor have we observed substantial miRNAs upregulation if the TGF-␤ Sma/Mab pathway is overexpressed, only a consistent miRNA fall when TGF-␤ Sma/Mab is disrupted ( Figure 9). Perhaps the overexpression of dbl-1(++) that we used exceeds what may be physiological in the wildtype, and that is why the miR-58/-80 promoters do not have a natural response for it.
The control over mir-58 and mir-80 could be transcriptional or posttranscriptional. In mammals there are examples of both forms of regulation between miRNAs and TGF-␤ signalling (64). Our system appears transcriptional because of our results with a P mir-58 ::gfp reporter ( Figure  9B). Moreover, the miRNA posttranscriptional control involves the presence of Smad binding elements (SBE) in the premature sequence of miRNAs, and we have not found any SBE in the corresponding sequences of the mir-58 family.
Are there any examples of autoregulatory loops involving miRNAs and TGF-␤? In mammals there are many cases of either miRNAs controlling TGF-␤ cell signalling or vice versa, and there are even some instances of autoregulatory loops running in both directions (64). However, as far as we know, no autoregulatory loop involving orthologs of mir-58f has been identified in any system. In Drosophila, Dpp/TGF-␤ positively modulates bantam expression through the Smad Mad, which together with Yorkie, bind directly to bantam and promote its transcription (33). And in humans, differentiation of smooth muscle cells is stimulated by TGF-␤1's direct control over miR-143/145 transcription (65), which is a putative homologue of miR-58 (21). What our results suggest is that perhaps these other systems may conserve still undiscovered mechanisms of TGF-␤ regulation by the corresponding miR-58 orthologs.
What should be the functional consequences of a negative feedback loop between TGF-␤ Sma/Mab and miR-58 in C. elegans? Autoregulatory negative feedback loops contribute to biological homeostasis by limiting the range of outcoming gene expression within certain borders for a given genetic network (66). Thus, they are thought to favour stable responses (biochemical or phenotypic) to genetic or environmental alterations. On the other hand, it has been suggested that miRNAs, rather than promoting drastic variation in protein expression, buffer perturbations and confer robustness (67,68). It is possible that negative feedback loops are a way for miRNAs to acquire those properties. Therefore, in our system, the identified negative feedback loop might help to bring the TGF-␤ Sma/Mab signalling into a convenient equilibrium. And in doing so, the loop presumably helps nematodes to attain optimal body sizes or other fitting phenotypes, maybe also as part of a responding mechanism to the always-challenging environmental conditions (27,69,70).
It is interesting to compare such a suggested homeostatic process to the positive feedback loop that we discussed in relation to the TGF-␤ Dauer signalling pathway ( Figure 10). This second kind of genetic loop, in contrast to the previous, enhances or amplifies the final molecular outcome, and for that reason it should ensure a particular physiological response. The decision whether to enter the dauer stage or not must be unambiguous. No in-between phenotype would be adaptive at all. Therefore, we would find the positive feedback loop at the TGF-␤ Dauer pathway rather useful for a developing C. elegans larva when having to take an unequivocal dauer decision.