LIN28 binding poorly predicts mRNA level changes

To identify functional targets of C. elegans LIN28, we performed RNA immunoprecipitation followed by sequencing (RIP-seq) by immunoprecipitating endogenously tagged LIN28. (We will refer to the relevant strain as lin-28::large_tag in the following to indicate that, in addition to the GFP epitope used in this experiment, it contains additional tags (Fig 1A). Note that although the animals appear superficially wild-type, we show below evidence that tagging subtly impairs the activity of the LIN28 protein.) Examining mid-L1 stage animals (6 h at 25 °C after plating larvae hatched in the absence of food), we found 581 transcripts enriched for LIN28 binding (≥2-fold, Fig 1B). Notably, this set overlapped significantly with targets identified by [24] through HITS-CLIP (110/1015; 3.7-fold over expectation, p = 1.7 × 10⁻³⁴ hypergeometric test), despite differences in animal stage, genetic background, and experimental procedure.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. let-7 and lin-46 stand out among LIN28 candidate targets.

A: Schematic representation of the lin-28 genomic locus with two lin-28 isoforms. Mutant alleles and endogenously tagged alleles used in this study are indicated. Allele numbers refer to modifications in otherwise wild-type backgrounds except the lin-28b isoform-specific exon deletion, xe344, which was introduced into an endogenously FLAG::HA-tagged background (xe59). The “large tag” xe197 allele corresponds to a C-terminal insertion of linker::AID::3xFLAG::TEV::GFP::linker::HiBit sequence. B: Volcano plot showing transcript enrichment in LIN28::large_tag IP compared to untagged control IP (N2). Immunoprecipitation was performed with anti-GFP antibody on lysates from synchronized early L1-stage animals collected at 6 h of development at 25 °C (n = 3 IP replicates). Transcripts significantly enriched (FDR ≤ 0.01, log₂(FC) > 1) are indicated in red. Data available at GEO under SuperSeries accession number GSE315957 (sample IDs GSM9440832 - GSM9440843). C: Volcano plot showing differential gene expression in lin-28(0) animals relative to wild type. Log₂ fold change in gene expression was quantified by RNA-seq of synchronized animals collected hourly from 3 to 6 h at 25 °C in three biological replicates. The x-axis represents the average expression of each gene across the three biological replicates, with different time point samples pooled for each replicate. Three genes with significant expression changes (adjusted p-value < 0.01, threshold indicated by horizontal dashed line) are labeled by name. Genes whose transcripts are bound by LIN28 (from (B)) are indicated in red. F02E9.3 expression was undetectable in both input and IP samples of LIN28::large_tag RIP-seq (shown in (B)). Data available at GEO under SuperSeries accession number GSE315957 (sample IDs GSM9440875 - GSM9440878, GSM9440920 - GSM9440923, GSM9441623 - GSM9441630 and GSM9441631 - GSM9441638). D: Expression levels of lin-46 mRNA in wild-type and lin-28(0) animals as determined by mRNA-seq. Animals were grown at 25 °C with hourly sampling. Data available at GEO under SuperSeries accession number GSE315957 (sample IDs GSM9440872 - GSM9440961). E: Plot showing log₂-transformed read counts of pri-let-7 in input and IP samples from LIN28::large_tag animals and untagged control animals (N2). Data are from the same LIN28::large_tag RIP-seq experiment as shown in (B). Red horizontal lines indicate the mean of three technical IP replicates. Significant differences were determined by an unpaired two-sided t test (ns: not significant; ***p < 0.001). Data available at GEO under SuperSeries accession number GSE315957 (sample IDs GSM9440832 - GSM9440843). F: Levels of mature let-7 miRNA in wild-type and lin-28(0) animals as determined by small RNA-seq. The two dashed lines indicate the window of time points pooled for differential gene expression analysis in (G). Samples are identical to those in (D). Data available at GEO under SuperSeries accession number GSE315957 (sample IDs GSM9441669 - GSM9441710) and GSE245904 (sample IDs GSM7850636 - GSM7850683). G: Scatterplot comparing miRNA expression in wild-type and lin-28(0) animals. miRNA expression levels were quantified by small RNA-seq of lysates from synchronized animals collected hourly from 3 to 6 h at 25 °C. This corresponds to a subset of samples from the series shown in (F) but sequenced to greater depth and published previously in [33]. Samples were treated as pseudoreplicates for statistical testing, with the average miRNA expression across the four time points plotted. Two miRNAs with significant expression changes are colored in red (FDR < 0.01). let-7 sisters are labeled in gray. Data available at GEO under SuperSeries accession number GSE245904 (sample IDs GSM7850639 - GSM7850642 and GSM7850663 - GSM7850666).

https://doi.org/10.1371/journal.pbio.3003892.g001

The role of LIN28 and its molecular mechanisms have previously been studied using lin-28(n719), an allele isolated after chemical mutagenesis that carries a point mutation in the 5′ splice donor of the second exon and that is considered null [4, 5]. To investigate the consequences of a complete loss of LIN28 in a genetically clean background, we used CRISPR-Cas9-mediated genome editing to introduce a ~3.2 kb long deletion spanning the whole lin-28 locus (lin-28(xe192), henceforth referred to as lin-28(0), Fig 1A). As we show below, lin-28(0) mutant animals recapitulate published lin-28(n719) mutant phenotypes [4, 34], yet the two strains shared surprisingly few dysregulated genes (S1 Fig). Specifically, when compared to the wild-type N2 strain, many more genes were dysregulated in lin-28(n719) than in lin-28(0), suggesting differences in strain background or additional mutations in the former.

To identify genes with robust changes in expression, we collected samples of several time points and in replicates for lin-28(0) and wild-type animals. We focused on the early first larval stage (eL1; 3–6 hours of development on food at 25 °C), a time window when LIN28 levels are high [5, 35] and potentially confounding rhythmic mRNA expression has not yet begun [36, 37]. For one of these experiments, sampling was extended to cover the entire course of larval development to allow assessment of mRNA trajectories over time. For those samples, we also sequenced small RNAs (see below).

Strikingly, besides lin-28 mRNA itself, the levels of only two mRNAs changed significantly in lin-28(0) animals (Fig 1C). The change in the first one, F02E9.3, is likely an artifact: It is not bound by LIN28 and its locus is immediately downstream of lin-28, suggesting that deletion of the lin-28 coding sequence placed this gene under the control of the lin-28 promoter. Consistent with this notion, F02E9.3 levels were unaffected in lin-28(n719) animals (S1A Fig).

By contrast, the second transcript, lin-46, is not only bound by LIN28, but is also among the few transcripts whose levels were upregulated in both lin-28(0) and lin-28(n719) animals (S1A Fig). Moreover, regulation of lin-46 by LIN28 had previously been reported and was proposed to occur at the level of translation [22, 23]. The long timecourse revealed that the levels of lin-46 mRNA are dynamic in wild-type worms, with a notable increase during mid/late L1 (~10 h) and again at ~18 hours (~mid-L2; Fig 1D). In lin-28(0) animals, lin-46 mRNA levels are upregulated during the first 18 hours of development (Fig 1D), i.e., the time when LIN28 protein is abundant in wild-type animals [5, 35]. These observations suggest that LIN28 represses lin-46, at least in part, by destabilizing its transcript.

let-7 is the only significantly upregulated miRNA in lin-28 mutant animals

The best-characterized LIN28 target in C. elegans as elsewhere is the let-7 miRNA [13, 16, 24, 33], and our RIP-seq analysis confirmed binding of the primary let-7 transcript to LIN28 (Fig 1E). (Note that Fig 1B displays only mRNAs.) We had previously reported small RNA sequencing of the first 23 h of the long timecourse, which had revealed extensive upregulation of let-7 miRNA levels in lin-28(0) animals [33]. Interestingly, when we extended this analysis to the whole timecourse, we observed that the upregulation was specific to early larval development, whereas from ~25 h onward, let-7 levels were identical in wild-type and lin-28 mutant animals (Fig 1F). This temporal pattern is consistent with the known restriction of LIN28’s activity to the early larval stages [5, 35].

To examine whether LIN28 regulates the expression of other miRNAs, we determined differentially expressed miRNAs in the time window of 3–6 hours investigated also for mRNAs above. Using the four time points as pseudoreplicates for statistical testing, only let-7 and one other miRNA, miR-229-3p, changed significantly. In contrast to let-7, miR-229-3p levels were decreased, not increased, in lin-28(0) animals (Figs 1G and S2). We expressly note that the let-7 sister miRNAs, miR-48, miR-84, and miR-241 were unaffected (Figs 1G and S2).

Taken together, although LIN28 binds hundreds of transcripts, only a few of these binding events yield substantial changes in transcript levels. We consider three possible explanations: First, LIN28 may act preferentially through regulatory mechanisms other than RNA destabilization, e.g., through translational control. Second, most instances of binding may not result in actual regulation and could be biologically inconsequential. Third, many of the binding events may be false positives. To distinguish among these scenarios, we sought to investigate to what extent regulation of only let-7 and lin-46 can account for the developmental functions of LIN28.

lin-46 but not let-7 is required for the protruding vulva phenotype of lin-28(0) animals

Animals lacking LIN28 exhibit numerous heterochronic phenotypes, such as a protruding vulva due to precocious vulval precursor cell division, a precocious synthesis of the adult-specific skin structure called alae and a precocious cessation of the molting cycle after three instead of the wild-type four molts [4, 34]. We hypothesized that if let-7 and lin-46 were the only two developmentally relevant LIN28 targets, we should be able to explain these phenotypes by misregulation of just these two genes. In other words, mutating one or both target genes in the background of the lin-28(0) mutation should fully suppress the precocious lin-28 mutant phenotypes. Moreover, the phenotypes resulting from their combined deletion should not be further modulated by loss of LIN28, their upstream repressor; i.e., lin-28; lin-46; let-7 triple mutant animals should recapitulate the retarded lin-46; let-7 double mutant phenotypes.

To test these two predictions, we used a previously published lin-46(0) allele, lin-46(ma385) (a ~1.6 kb deletion spanning all lin-46 exons [38]), and a newly generated let-7 allele, let-7(xe150), which carries an ~1 kb deletion of the promoter region (S3A Fig). Considering the massive effect on let-7 synthesis (S3B and S3C Fig) and the full recapitulation of previously described let-7 mutant phenotypes as described below, let-7(xe150) is a strong hypomorph and we will refer to it as let-7(−) in the following.

First, we examined vulval development. lin-28(0) animals exhibit a fully penetrant protruding vulva (Pvl) phenotype, which we find to be fully suppressed by the lin-46(0) allele (Fig 2A), as reported earlier [23]. By contrast, this phenotype is not suppressed by a let-7(−) mutation (Fig 2A). Instead, lin-28(0) suppresses a vulval phenotype associated with let-7(−): let-7(−) mutant animals have a functionally deficient vulva through which they burst as young adults (Fig 2B and [20])—a phenotype that is absent from lin-28(0); let-7(−) double mutant animals. This finding can either be interpreted as let-7 functioning upstream of lin-28 or as evidence for lin-28 regulating additional targets that act in parallel to let-7. Consistent with the latter scenario, we found that lin-28(0); lin-46(0); let-7(−) triple mutant animals phenocopied let-7(−) single mutant animals, exhibiting fully penetrant vulval bursting (Fig 2B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. lin-46 but not let-7 is required for the lin-28(0) mutant protruding vulva (Pvl) phenotype.

A: Example DIC images of vulvae in the indicated genetic backgrounds. The animals were grown at 20 °C and examined shortly after exit from the final molt, i.e., the fourth molt in all cases except lin-28(0) animals, which execute only three molts and were grown at 25 °C. Arrow indicates the pathological vulva protrusion (Pvl). The percentage of animals exhibiting the Pvl phenotype is indicated. Scale bars: 10 µm. B: Quantification of vulva bursting of adult animals of indicated genotypes. Animals were grown at 25 °C and examined at 48 h of development. N = 3 technical replicates, with n = 240 animals scored per genotype. Error bars represent standard deviation. Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133).

https://doi.org/10.1371/journal.pbio.3003892.g002

Taken together, our findings are consistent with let-7 and lin-46 acting in parallel pathways as targets of LIN28 with a partially redundant function in vulval development. Dysregulation of lin-46 appears to chiefly account for the Pvl phenotype of lin-28(0) animals.

Both let-7 and lin-46 are required for the precocious alae synthesis in lin-28(0) animals

Second, we evaluated the formation of alae, an adult-specific cuticular structure. In wild-type animals, alae are generated during the fourth molt. In contrast, lin-28(0) animals form alae precociously with 98% and 81% of animals having at least some alae already after the third molt when grown at 20 °C and 25 °C, respectively (Figs 3A and S4). In fact, and consistent with earlier work [4], nearly half of the lin-28(0) animals exhibit alae as early as immediately after the second molt, but these alae are weak and incomplete (Figs 3A, 3B, and S4). The precocious alae synthesis is strongly suppressed by loss of either lin-46 or let-7, with no or very rare alae in lin-28(0); lin-46(0) and lin-28(0); let-7(−) double mutant animals after the third molt, respectively (Fig 3A).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. The precocious alae phenotype of lin-28(0) animals is suppressed by loss of either let-7 or lin-46.

A: Quantification of the presence of alae in animals of indicated genotypes. Molting animals were singled and examined shortly after they exited the respective molt as indicated by ecdysis. Phenotypes of lin-28(0) and lin-28(0); let-7(−) animals were quantified twice on different days and the results were pooled in the presented table. Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). B: Example DIC images of alae in the indicated genetic backgrounds. The animals were grown at 20 °C and examined shortly after the exit from the indicated molt, with the exception of the “lin-28(0) M2” condition in which the animals were grown at 25 °C. Arrowheads point to the alae. The dashed line indicates the alae gap. Scale bars: 10 µm.

https://doi.org/10.1371/journal.pbio.3003892.g003

Conversely, the lin-28(0) allele suppresses the alae phenotypes of the individual target mutations. Upon exit from the fourth molt, lin-46(0) animals have either complete (67%) or gapped (33%) alae, while all lin-28(0); lin-46(0) animals display complete, gap-free alae (Figs 3A, 3B, and S4). At the same developmental stage, let-7(−) animals lack alae entirely, whereas around half of lin-28(0); let-7(−) animals exhibit weak partial alae (Figs 3A and S4). These findings again argue against a simple linear pathway where LIN28 controls alae formation through a single target. To determine whether regulation of let-7 and lin-46 is sufficient, we compared alae formation in lin-46(0); let-7(−) double and lin-28(0); lin-46(0); let-7(−) triple mutant animals. As in let-7(−) single mutant animals, neither strain formed any alae even after the fourth molt (Fig 3A), supporting the conclusion that let-7 and lin-46 are likely the sole targets mediating LIN28’s role in the temporal control of alae synthesis. Moreover, unlike for the Pvl phenotype, dysregulation of both lin-46 and let-7 is required for precocious alae formation in lin-28(0) mutant animals.

Both let-7 and lin-46 are required for the precocious exit from the molting cycle in lin-28(0) animals

Animals lacking LIN28 were previously reported to execute only three instead of the wild-type four larval molts [4]. To examine the contributions of let-7 and lin-46, we employed a high-throughput assay allowing us to quantify the number of molts based on the alternating periods of feeding and lethargus in developing larvae [37, 39]. Briefly, animals expressing a luciferase transgene are provided with a medium containing luciferin, the luciferase substrate. As the animal grows and feeds, luciferin is ingested and, consequently, luminescence signal is emitted. When the animal enters a molt, it stops feeding and the signal decreases steeply, generating a characteristic luminescence intensity pattern (Fig 4A).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. The precocious exit from the molting cycle of lin-28(0) animals is suppressed by both let-7(−) and lin-46(0).

A: Examples of luminescence intensity traces. Vertical blue bars indicate molts. Red arrow points to the signal increase indicative of vulva bursting. Blue arrow points to a sudden signal increase, the 4M* phenotype. Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). B: Quantification of the number of molts in animals of the indicated genotypes. The number at the bottom of each bar indicates the number of examined animals (n). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). C: Example DIC images of the head (top row) and the vulva (bottom row) regions of lin-28(0); lin-46(0) animals that underwent a normal (left, 4M) or aberrant (middle and right, 4M*) fourth molt. The animals were extracted from the luciferase assay plate after 52 h, classified according to their luminescence intensity pattern and examined under the microscope. The 4M* animals exhibited a defect in pharyngeal cuticle shedding ranging from a part of the old cuticle remaining attached to the animal through the mouth plug (middle, white arrow) to a ripped-out pharynx (right, pharynx traced with a red dashed line, head traced with a white dashed line). All 4M* animals exhibited egg-laying defect as demonstrated by the retained embryos of advanced developmental stages. Scale bars: 10 μm. D: Quantification of the number of molts in animals of the indicated genotypes grown on either empty-vector-containing (control) or cdk-1 RNAi bacteria. Color key as in (B). The wild-type and lin-28(0) conditions are replotted from (B). The number at the bottom of each bar indicates the number of examined animals (n). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133).

https://doi.org/10.1371/journal.pbio.3003892.g004

The analysis revealed the expected four molts in wild-type and three in lin-28(0) animals (Fig 4A and 4B). The precocious exit from the molting cycle observed upon LIN28 loss requires the activities of both let-7 and lin-46, as both lin-28(0); let-7(−) and lin-28(0); lin-46(0) double mutant animals undergo more than three molts (Fig 4A and 4B). However, in the case of lin-28(0); lin-46(0) double mutants, suppression was only partial: nearly half of the animals exhibit a defect during the fourth molt (henceforth referred to as the 4M* phenotype) (Fig 4B): After completing three properly executed molts, a steep increase in the luminescence intensity occurred at a time where wild-type animals would exhibit the signal drop typical of feeding cessation (Fig 4A). This unusual signal reflects a severe defect in shedding of the pharyngeal cuticle, with the anterior part of the old cuticle remaining attached via the mouth plug (Fig 4C, middle). In extreme cases, the pharynx appears to be partially ripped out, likely due to mechanical forces applied during failed cuticle shedding (Fig 4C, right). Additionally, late-stage embryos accumulate in the affected animals, indicating an egg-laying defect (Fig 4C), possibly a consequence of embryo retention due to starvation [40]. We note that the 4M* phenotype is specific to liquid culture, as lin-28(0); lin-46(0) animals appear wild-type when grown on agar plates.

lin-46(0) mutant animals undergo the wild-type number of four molts. By contrast, if let-7(−) animals are prevented from vulval bursting, they execute additional molts. Thus, 100% of let-7(−) animals depleted of CDK-1 to prevent vulva formation [41] undergo extra molts (Fig 4D). The lin-28(0) mutation not only fully suppressed vulval bursting, but also partially suppressed the extra molt phenotype with 24% of lin-28(0); let-7(−) double mutant animals undergoing the normal number of molts, four (Fig 4A and 4B). This suppression depended on lin-46: Like let-7(−) single and lin-46(0); let-7(−) double mutant animals, all lin-28(0); lin-46(0); let-7(−) triple mutant animals executed supernumerary molts when grown on cdk-1 RNAi (Fig 4C). We conclude that let-7 and lin-46 are the essential targets of LIN28 for controlling molt number. As for the alae phenotype, precocious termination of the molting cycle in lin-28(0) mutant animals appears to require dysregulation of both lin-46 and let-7.

Mutation of the putative LIN28 binding site in the primary let-7 transcript does not alter let-7 levels

With the suppression analysis, we tested whether the two LIN28 targets, let-7 and lin-46, are necessary for lin-28 mutant phenotypes. Next, we wanted to test whether the precocious expression of these targets is sufficient to induce those phenotypes. Thus, we aimed to use genome editing to uncouple the two targets from LIN28-imposed repression.

In humans, LIN28 recognizes a GGAG motif located in the apical part of the let-7 hairpin [42–44]. However, the let-7 hairpin in C. elegans lacks this motif, and instead, LIN28 has been proposed to bind a downstream region outside of the actual hairpin ([24]; S3A Fig). This region contains four GGAG motifs, which have been implicated in LIN28 binding in vitro [24]. To uncouple let-7 from LIN28-mediated repression in vivo, we mutated all four motifs in the endogenous let-7 locus, changing them from GGAG to CTCC (let-7(xe336), henceforth referred to as let-7(ΔBS) for mutation of the putative Binding Sites) (S3A Fig). However, this mutation did not result in appreciable let-7 derepression, as mature let-7 levels remained largely unaffected (S3D Fig), suggesting that the four GGAG motifs in the putative LIN28 binding site are not required for let-7 repression in vivo, possibly due to redundancy with other binding sites or repressive mechanisms. (We have not determined whether LIN28 binding was perturbed by these mutations.)

Mutation of the putative LIN28 binding site in lin-46 5′UTR results in weakly precocious LIN-46 accumulation

The lin-46 transcript is bound by LIN28 (Fig 1B and [24]) and the lin-46 5′UTR has been shown to prevent precocious accumulation of LIN-46 protein during development [22]. To uncouple lin-46 from LIN28, we mutated the single GGAG motif in the short endogenous lin-46 5′UTR to CTCC (lin-46(xe370), henceforth referred to as lin-46(++)). To evaluate the levels of lin-46 mRNA, we collected samples of synchronized animals hourly from 5 to 14 hours of development (i.e., from early/mid-L1 to early/mid-L2) and performed mRNA-seq. As shown in Fig 5A, lin-46 mRNA levels were upregulated to an intermediate level in lin-46(++) animals, with the characteristic increase at ~8–9 hours observed in wild-type animals preserved (Fig 5A, compare with Fig 1D).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Mutation of the GGAG motif in lin-46 5′ UTR recapitulates the precocious LIN-46 accumulation seen in lin-28(0) animals only partially.

A: Levels of lin-46 mRNA as determined by mRNA-seq of synchronized animals collected hourly from 5 to 14 h of development at 25 °C. Data available at GEO under SuperSeries accession number GSE315957 (sample IDs GSM9441639 - GSM9441668). B: Confocal images of endogenously split-GFP-tagged LIN-46 in the epidermal seam cells of wild-type, lin-46(++) and lin-28(0) animals. Animals were grown at 25 °C and examined at the indicated time points. Arrowheads indicate seam cells with detectable GFP signal. An arrow points to the predominantly nuclear accumulation of LIN-46 in lin-28(0) animals. Scale bars: 10 μm. C: Confocal images of endogenously split-GFP-tagged LIN-46 in the epidermal seam cells of lin-28(0) and lin-28(0); lin-46(++) animals. Animals were grown at 25 °C and examined at the indicated time points. Arrowheads indicate seam cells with detectable GFP signal. Scale bars: 10 μm.

https://doi.org/10.1371/journal.pbio.3003892.g005

To determine whether this mild increase in lin-46 mRNA levels also manifests at the protein level, we examined endogenous LIN-46 protein C-terminally tagged with a 4xsplit-GFP (lin-46(xe347)). In agreement with previously published observations using a different fluorophore [22], we detected LIN-46 in wild-type animals in the epidermal seam cells (Fig 5B), as well as in the vulval precursor cells, tail and rectum (S5 Fig). Focusing on the seam cells, we observed LIN-46 accumulation starting in late L3 (24–26 h; Fig 5B). In lin-28(0) mutant animals, LIN-46 accumulated two stages earlier, at late L1, from 12 hours onwards. This was true regardless of whether the lin-46 5′UTR was mutated (Fig 5B and 5C). By contrast, and consistent with the RNA-seq data, in animals with the GGAG→CTCC mutation in the lin-46 5′UTR (lin-46(xe351xe347)), LIN-46 is detected at an intermediate time, in late L2 (from ~18 hours onwards), one stage earlier than in wild-type but one stage later than in lin-28(0) (Fig 5B). We noticed, but did not investigate further, that LIN-46 protein accumulates in the nucleus in lin-28(0) animals at ~26 hours (Fig 5B, white arrow).

Taken together, lin-46 repression is only partially dependent on the single GGAG motif within its 5′UTR. The fact that some repression remains in lin-46(++) animals and is still LIN28-dependent suggests either that LIN28 has an additional binding site in the lin-46 transcript, or that it additionally represses lin-46 through an indirect mechanism.

Enhanced expression of let-7 and lin-46 in two independent sensitized backgrounds emulates lin-28 mutant phenotypes

Since the manipulation of putative LIN28 binding sites achieved only a partial decoupling of lin-46 from LIN28-mediated regulation, insufficient to elicit heterochronic phenotypes (see below) and no decoupling of let-7, we performed further sufficiency testing in a sensitized genetic background. Specifically, LIN28 occurs in two isoforms, LIN28a and LIN28b, encoded in the same locus (Fig 1A). The two proteins share the C-terminal part containing the RNA-binding domains but differ in their N-terminal sequences encoded by isoform-specific first exons. A deletion of the LIN28b-specific first exon in a lin-28::flag::ha background results in the complete loss of detectable LIN28b protein (S6A Fig; lin-28(xe344xe59), henceforth referred to as lin-28(Δb)) and an intermediate upregulation of both let-7 miRNA and lin-46 mRNA throughout the examined time window (12–17 hours on food at 25 °C; S6B Fig). Despite these gene expression changes, the animals exhibit no signs of a protruding vulva (Fig 6A) or precocious alae synthesis (Fig 6B). However, in the luciferase assay, lin-28(Δb) animals exhibit a fully penetrant 4M* phenotype (i.e., a defect in pharyngeal cuticle shedding during the fourth molt; see above; Fig 6C and 6D).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Enhanced expression of let-7 and lin-46 is sufficient to induce lin-28 mutant-like phenotypes in a lin-28(Δb) background.

A: Quantification of the protruding vulva (Pvl) phenotype of adult animals of indicated genotypes. Animals were grown at 20 °C and examined at 52 h of development, when they had reached adulthood. N = 4 technical replicates, with n > 230 animals scored per genotype. Error bars represent standard deviation. Significant differences to wild type were determined by a Dunnett’s test (***p < 0.001). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). B: Quantification of the presence of alae in animals of indicated genotypes. Molting animals were singled and examined shortly after they exited the respective molt as indicated by ecdysis. The wild-type and lin-28(0) conditions are from Fig 3A and included for comparison. Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). C: Quantification of the number of molts in animals of the indicated genotypes. The number at the bottom of each bar indicates the number of examined animals (n). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). D: Examples of luminescence intensity traces. Molt phenotypes are labeled as 3M for animals undergoing three molts, 4M for four molts, and 4M* for animals exhibiting an aberrant increase in luminescence intensity at the time of the expected fourth molt (indicated by a blue arrow). The two distinct outcomes of enhanced let-7 and lin-46 expression in the lin-28(Δb) background are illustrated. Note that in these animals, 4M involves a delayed and not properly executed fourth molt (4M), characterized by elevated luminescence intensity during this molt (arrowhead). Vertical blue bars indicate molts. The percentages indicate the penetrance of each phenotype (as shown in (C)). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). *: LIN28 is C-terminally FLAG::HA-tagged.

https://doi.org/10.1371/journal.pbio.3003892.g006

In the lin-28(Δb) background, we further enhanced the expression of let-7 and/or lin-46, by introducing an additional let-7 gene copy (hereafter let-7(++)), and by mutating the lin-46 5′UTR as in the lin-46(++) allele, respectively. Enhanced expression of lin-46 in the lin-28(Δb) background resulted in partially penetrant protruding vulva (71%) and precocious alae (47%) phenotypes (lin-28(Δb); lin-46(++) in Fig 6A and 6B). The penetrance of these phenotypes increased further when we additionally enhanced let-7 expression (from 71% to 87% and from 47% to 100%, respectively; lin-28(Δb); let-7(++) lin-46(++) in Fig 6A and 6B).

Upon individual enhanced expression of either let-7 or lin-46 in the lin-28(Δb) background, the animals still underwent the aberrant fourth molt characteristic of the 4M* phenotype (Fig 6C). However, when we jointly enhanced the expression of both genes in this background, 40% of animals failed to undergo a fourth molt entirely, while the remaining animals underwent three proper molts followed by an incomplete drop in the luminescence signal, possibly reflecting an aberrant molt (Fig 6C and 6D). We note that this latter molting pattern was distinct from the 4M* pattern, in which the signal steeply increases rather than decreases at the time of the expected fourth molt (Fig 6D). In repeat experiments, we observed an even higher frequency of animals undergoing only three molts, namely 63% (n = 27) and 67% (n = 45).

These findings were qualitatively recapitulated using a second sensitized background, lin-28::large_tag (S7 Fig). These animals look overtly wild-type, but, as noted also by others [45], GFP-tagging subtly impairs LIN28 function, causing an increase in let-7 miRNA levels comparable to lin-28(Δb) animals (S6B Fig).

LIN28 controls not only the number of larval stages but also their duration

C. elegans larval development is temporally segmented such that cells assume specific identities characteristic of a given larval stage [46]. Heterochronic genes were defined by their ability to specify these temporal fates, and their key role thus appears to be the maintenance of temporal order [4]. Other, clock-type mechanisms, involving oscillatory gene expression, are thought to set the tempo at which larval stages are executed [47, 48]. Surprisingly then, although lin-28(0) mutant animals execute only three instead of four larval stages, we found that the average duration from hatching to the exit from the final molt was comparable between wild-type and lin-28(0) animals (Fig 7A): the final, third molt in lin-28(0) animals occurs only shortly before the fourth molt in wild-type (Fig 7B). This is due to a substantial extension of larval stages in lin-28(0) animals, with the third larval stage being most strongly affected (Fig 7C). This extended L3 stage phenotype depends on the expression of both lin-46 and let-7 (Figs 7D and S8) and can be induced by their enhanced expression in sensitized genetic backgrounds with reduced LIN28 function (Fig 7E and 7F).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Lack of LIN28 results in extension of larval stage durations.

A: Quantification of total duration of larval development from hatching until exit from the last executed molt. Number of animals n = 36, 27 and 27 for wild type, lin-28(0) and lin-28(n719), respectively. The horizontal line represents the mean. Significant differences were determined by an unpaired two-sided t test (***p < 0.001; *p < 0.05). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). B: Average duration of developmental stages in wild-type, lin-28(0) and lin-28(n719) animals as determined by the luciferase assay (same experiment as shown in (A)). Rectangles indicate the molts, horizontal lines represent the intermolts. C: Quantification of larval stage durations of wild-type, lin-28(0) and lin-28(n719) animals (same experiment as shown in (A)). The horizontal line represents the median, hinges extend to first and third quartiles, and the whiskers extend up to the most extreme value within 1.5*IQR (interquartile range) of the median. Significant differences relative to wild type for each larval stage are indicated. P‐values were determined by a non‐parametric Wilcoxon rank sum test. ***p < 0.001. Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). D: Quantification of the L3 stage duration of the same animals as in Fig 4B. The horizontal line represents the median, hinges extend to first and third quartiles, and the whiskers extend up to the most extreme value within 1.5*IQR (interquartile range) of the median. Significant differences were determined by a Dunn’s test with Holm correction for multiple testing. The statistical significance of selected comparisons is indicated (***p < 0.001; *p < 0.05; ns, not significant). For statistical significance of all comparisons, see S8 Fig. Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). E: Quantification of the L3 stage duration of the same animals as in Fig 6C, examining the effects of lin-46 and let-7 expression modulation in a lin-28(∆b) mutant background. The horizontal line represents the median, hinges extend to first and third quartiles, and the whiskers extend up to the most extreme value within 1.5*IQR (interquartile range) of the median. Significant differences to wild type were determined by a Dunn’s test with Holm correction for multiple testing (***p < 0.001; no label, not significant). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). F: Quantification of the L3 stage duration of the same animals as in S7C Fig, examining the effects of lin-46 and let-7 expression modulation in a lin-28::large_tag background. The horizontal line represents the median, hinges extend to first and third quartiles, and the whiskers extend up to the most extreme value within 1.5*IQR (interquartile range) of the median. Significant differences to wild type were determined by a Dunn’s test with Holm correction for multiple testing (***p < 0.001; no label, not significant). Raw data available on Zenodo (https://doi.org/10.5281/zenodo.20787133). *: LIN28 is C-terminally FLAG::HA-tagged.

https://doi.org/10.1371/journal.pbio.3003892.g007

In wild-type animals, the expression of thousands of genes is rhythmic, peaking once per larval stage (Fig 8A and 8B and [36, 37]). Consistent with the reduced number of molts accompanied by the gradually increasing extension of larval stages, oscillating genes peaked only three times in lin-28(0) animals over the course of the experiment (Fig 8A and 8B), with an increased period (Fig 8C and 8D).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Oscillatory gene expression reflects the perturbed molting cycle in lin-28(0) animals.

A: Heatmap showing mean-normalized log₂-transformed expression levels of oscillating genes (n = 2,004; as defined in [37], with log₂ amplitude ≥ 1 in that study) over time in wild-type (left) and lin-28(0) animals (right). Genes are ordered according to the published peak phase. B: Expression profile of a representative oscillating gene, lin-42, across development in wild-type and lin-28(0) animals. C: Distribution of instantaneous periods of all oscillating genes (n = 3,680) in wild-type and lin-28(0) animals. Color bar represents number of genes per bin at each time point. Average period is shown as a dashed line. The instantaneous period for each gene was computed using the Hilbert Transform on the data after processing it with a Butterworth filter. D: Comparison between the average periods replotted from (C). The shadowed area corresponds to the values between the 20th and 80th percentiles, from the distributions shown in (C). Data available at GEO under SuperSeries accession number GSE315957 (sample IDs GSM9440872 - GSM9440961).

https://doi.org/10.1371/journal.pbio.3003892.g008

Taken together, our findings show that LIN28 controls not only the identity of larval stages but also their duration. Whether this extends to other heterochronic genes and through what molecular mechanism remains to be elucidated.

Limitations of the study

The RIP-seq method that we used to identify mRNAs bound by LIN28 may also identify non-physiological binding partners, that emerged from post-lysis associations. However, since the overlap to previously published CLIP-seq data [24] was substantial and significant, and lin-46, the only dysregulated gene, was also LIN28-bound, false positive binding events are not expected to impact the key conclusions.

As we were unable to effectively uncouple let-7 and lin-46 from regulation by LIN28, we had to rely on a sensitized background for sufficiency testing. Although similar results were obtained with two different sensitized backgrounds, it remains possible that this sensitization involves upregulation of additional LIN28 targets not captured by mRNA-seq. Future work, achieving complete decoupling of the two targets from LIN28-mediated repression without perturbation of lin-28 itself, will be able to rule in or out this possibility. Irrespectively, when viewed together with the analysis of the suppression analysis, it is evident that lin-46 and let-7 are the functionally essential LIN28 targets.

C. elegans handling

The wild-type strain used in this study was Bristol N2. Transgenic strains used in this study are listed in S1 Table. To synchronize animals, arrested L1 stage larvae were obtained by extracting embryos from gravid adults using a bleaching solution (30% (v/v) sodium hypochlorite 4%–6% reagent (Sigma-Aldrich; 419550010), 750 mM KOH) and hatching overnight in the absence of food, at room temperature in M9 buffer (42 mM Na₂HPO₄, 22 mM KH₂PO₄, 86 mM NaCl, 1 mM MgSO₄). Experiments were performed at 25 °C unless explicitly stated otherwise.

Generation of transgenic CRISPR strains

Alleles generated in this study are listed in S2 Table. Sequences of oligos used for CRISPR strain generation are listed in S3 Table.

The let-7(xe150) and lin-28(xe192) alleles were generated following an adapted “dpy-10 conversion protocol” [61, 62]. sgRNAs were cloned into a NotI-digested pIK198 plasmid using Gibson cloning. The injection mix contained: 50 ng/μl pIK155 (Cas9 expression vector), 100 ng/μl pIK208 (sgRNA expression vector targeting dpy-10), 20 ng/μl AF-ZF-827 (homology repair oligo for dpy-10) and 100 ng/μl of each pIK198 with a cloned sgRNA targeting the gene of interest.

The let-7(xe336), lin-28(xe344xe59), lin-46(xe347), lin-46(xe351xe347), lin-46(xe370) and lin-28(xe197) alleles were generated following published protocols [63, 64]. crRNAs targeting the genes of interest were ordered from IDT (Alt-R CRISPR-Cas9 crRNA, 2 nmol, standard desalting). A pre-injection mix with 0.5 μl Alt-R S.p. Cas9 Nuclease V3 (10 μg/μl; IDT, 1081058), 5 μl tracrRNA (0.4 μg/μl; IDT) and 2.8 μl crRNA (0.4 μg/μl) (or 1.4 μl of each crRNA if two were used) was incubated for 15 min at 37 °C before adding 2 μl pGH8 (100 ng/μl), 2 μl pRF4 (400 ng/μl) and either 500 ng of melted dsDNA donor or 2.2 μl single-stranded oligo donor (1 μg/μl; ordered as 4 nmole Ultramer DNA Oligo Standard Desalting from IDT).

To test whether LIN28-mediated lin-46 repression depends on the single GGAG motif present in lin-46 5′UTR, we changed it from GGAG to CTCC, generating the lin-46(xe370) allele. lin-46(xe370) animals did not exhibit a protruding vulva phenotype (Results) unlike other lin-46 5′UTR mutant lines previously described [22]. To address whether these differences reflected differences in the mutation or strain background, as the previously reported lines also contained the maIs105 [col-19p::gfp] V multicopy array integrated in the same chromosome as lin-46, we sought to outcross the maIs105 array from the lin-46(ma467) allele, a 12 nucleotide long deletion removing the first two G’s of the GGAG motif and 10 nucleotides upstream (strain VT3855) [22]. Despite screening over 300 F2 progeny from three separate outcross attempts, we did not detect any recombination events between the two loci. As an alternative approach, we introduced a deletion identical to ma467 in either wild-type or lin-46::3xflag::gfp11 backgrounds using CRISPR/Cas9. This did not cause precocious LIN-46 accumulation or a protruding vulva phenotype, and the maIs105 array could readily be crossed into that strain, suggesting that the inability to cross maIs105 out of VT3855 is not due to a small genetic distance to lin-46(ma467) or crossover suppression but due to some other strain background effect. As we will detail elsewhere, genome resequencing revealed the existence of a ~7.6-kb deletion near lin-46 in VT3855. Hence, we conclude that strain VT3855 in addition to ma467 has several aberrations from the wild type N2 so that the Pvl phenotype cannot be confidently attributed to ma467.

Experiments to quantify gene expression

To quantify gene expression using sequencing techniques or RT-qPCR, animals were sampled either hourly over a developmental timecourse (denoted as “TC” in S4 Table), or only at selected time points (denoted as “Exp” in S4 Table). Samples were collected as described below. Prior to RNA extraction, animals were lysed by five cycles of thawing (42 °C) and snap-freezing in liquid nitrogen. RNA was extracted either using Direct-zol RNA Microprep kit (Zymo Research; R2062; DNase treatment included) according to the manufacturer’s protocol or performing phenol-chloroform extraction according to the Tri Reagent (Molecular Research Center, TR-118) manufacturer’s protocol followed by DNase treatment (RNase-Free DNase I kit (#25710 NORGEN); Single Cell RNA Purification Kit (#51800 NORGEN)).

LIN28 RNA-immunoprecipitation followed by sequencing (RIP-seq)

Luciferase assay

To assess the molting cycle progression of individual animals, we employed a luciferase assay as described previously [37, 39]. Gravid adults were bleached and single embryos expressing luciferase from a constitutive and ubiquitous promoter (transgene xeSi312) were transferred by pipetting to a white, flat-bottom, 384-well plate (Berthold Technologies, 32505). Embryos were allowed to develop in 90 μl S-Basal medium containing E. coli OP50 bacteria (OD600 = 1) and 100 μM Firefly D-Luciferin (p.j.k., 102111). Plates were sealed with Breathe Easier sealing membrane (Diversified Biotech, BERM‐2000). Luminescence was measured every 10 min for 0.5 s for a total duration of 52–96 h (depending on the purpose of the experiment) using a luminometer (Berthold Technologies, Centro XS3 LB 960). The luminometer was placed in a temperature-controlled incubator set to 18.5 °C, which translated to an effective temperature of 22.5 °C experienced by the animals inside the luminometer.

Luminescence intensity traces of individual animals were further analyzed using a Python-based custom code (L.J.M.M., unpublished, code available at https://doi.org/10.5281/zenodo.20742641). Hatch, as well as both molt entries and exits, were automatically annotated using a small convolutional neural network and further refined manually. Molts were defined as periods of stable low signal, preceded and followed by a sudden drop and rise in luminescence intensity, respectively. Significant differences in total durations of larval development (Fig 7A) were determined by an unpaired two-sided t test. P-values of all pair-wise comparisons of L3 durations were determined by a non-parametric Dunn’s test with Holm correction for multiple testing (Figs 7D, 7E, 7F, and S8).

Examination of balanced animals carrying the lethal let-7(xe150) allele

Several strains examined in this work (S1 Table) carry the lethal let-7(xe150) allele, which causes vulva bursting shortly after exit from the fourth molt. To maintain these strains, they were balanced with an extrachromosomal array [51], which contains a let-7 transgene and several mCherry reporters (xeEx365[let-7p::let-7::SL1_operon_GFP, unc-119(+), rab-3p::mCherry, myo-2p::mCherry, myo-3p::mCherry]). This array segregates randomly during meiosis, resulting in incomplete inheritance, with only a subset of progeny carrying the array and thus surviving into adulthood. In order to examine the molting cycle of animals with lethal genotypes, animals were retrieved from the luciferase assay plate after the assay was finished. They were visually inspected for the presence of the mCherry-marked balancer array using a fluorescent binocular microscope and only the luminescence intensity traces of animals not carrying the balancer array were included in the analysis.

Examination of the 4M* phenotype in lin-28(0); lin-46(0) animals

To examine the partially penetrant and liquid culture-specific aberrant fourth molt phenotype of lin-28(0); lin-46(0) animals (HW3504), synchronized L1s were loaded to the luciferase assay plate. The reason for using synchronized L1s instead of embryos was to achieve a synchronously developing population to enrich for animals at the desired developmental stage at the time of examination. The luciferase assay was run for 52 hours. Afterwards, the luminescence intensity traces of individual animals were examined and classified into two groups—animals that underwent a normal fourth molt and animals that exhibited the 4M* phenotype (a steep increase in luminescence intensity instead of the typical signal drop at the time of the expected fourth molt). Fifteen animals from each group were retrieved from the plate, mounted onto two separate 2% (w/v) agarose pads and immobilized in 10 mM levamisole. Differential Interference Contrast (DIC) images were acquired with a Zeiss Axio Observer Z1 microscope using the Zen 2 software. Selections of regions and image processing were performed with Fiji [66].

Confocal imaging

To assess the time of LIN-46 protein accumulation, animals were grown on 2% NGM agar OP50 plates at 25 °C for the indicated time before mounted on a 2% (w/v) agarose pad with a drop of 10 mM levamisole solution. Animals were imaged on an upright microscope Axio Imager M2 (Zeiss) equipped with CSU W1 with Dual T2 spinning disk confocal scanning unit (Yokogawa) driven by Visiview 6.0 (Visitron). DIC and fluorescent images were acquired with a 40×/1.3 oil immersion objective (EC Plan Neofluar). Using the Fiji software, images were processed after selecting representative regions.

Western blotting

Synchronized L1s were plated on 10 cm 2% NGM agar OP50 plates, grown at 25 °C and harvested at 3 and 6 h (20,000 animals per time point) in M9 buffer, washed twice with M9 buffer, resuspended in ~1 ml of 0.01% Tween-20 (Sigma-Aldrich, P9416) in M9 buffer and transferred to 1.5 ml Eppendorf tubes. Samples were spun down and dry pellets were snap-frozen in liquid nitrogen. Lysates were prepared by boiling (5 min, 95 °C) and sonication (Bioruptor Plus, 13 cycles, 30 s on/60 s off at 4 °C) in SDS lysis buffer (63 mM Tris-HCl (pH 6.8), 5 mM DTT, 2% SDS, 5% sucrose) and cleared by centrifugation. Proteins were separated by SDS-PAGE on a 4%–12% Bis-Tris gel (Thermo Fisher Scientific, NP0335BOX) in MOPS running buffer (Thermo Fisher Scientific, NP0001). Samples were run at 80 V for ~30 min and then at 150 V until the desired protein separation was achieved. Forty micrograms of protein extract was loaded per well and PageRuler Prestained Protein Ladder (10–180 kDa; Thermo Fisher Scientific, 26616) was used as a molecular weight reference. Proteins were transferred to a PVDF membrane (Bio-Rad, 1620177) by semi-dry blotting at 20 V for 1 hour (transfer buffer: 250 mM Tris, 1.9 M glycine, 0.5% SDS + 20% methanol). The membrane was blocked with 5% (w/v) milk in TBST (0.1% Tween-20 in TBS) for 1 h shaking (150 rpm). LIN28 was detected using polyclonal rabbit anti-LIN28 antibody (custom-made by GenScript, R13880; raised against full-length recombinant His6::TEV::LIN28a; 1:1,000), followed by polyclonal anti-rabbit HRP-conjugated antibody (Santa Cruz Biotechnology, sc-2004; 1:5,000). Antibodies were diluted in 0.5% (w/v) milk in TBST and incubated with the membrane for 1 h shaking at 150 rpm at room temperature. ECL Detection Reagent (Cytiva, RPN2232) and Amersham Imager 680 were used for detection. After LIN28 detection, the membrane was stripped (Restore PLUS Western Blot Stripping Buffer; Thermo Fisher Scientific, 46430), washed with TBST, blocked as described above and incubated with monoclonal mouse anti-Actin, clone C4 (MilliporeSigma, MAB1501; 1:7,500) and anti-mouse HRP-conjugated antibody (Cytiva, NXA931; 1:5,000) in 0.5% (w/v) milk in TBST for 1 h shaking at 150 rpm at room temperature, followed by Actin detection as described above.

Examination of precocious alae phenotypes upon molt exit

To assess the presence of alae, synchronized L1s of indicated genotypes were grown on 2% NGM agar OP50 plates at either 20 or 25 °C and inspected with a binocular microscope. When ~50% of the animals entered the respective molt (S5 Table), as indicated by the lack of movement and pharynx pumping, ~35 animals were individually transferred to small plates and inspected at regular intervals (every ~20 min) for signs of molt completion, as indicated by resumed movement, pharynx pumping and the presence of the old shed cuticle on the plate. The first 10 animals that exited the molt were mounted onto a 2% (w/v) agarose pad, immobilized in 10 mM levamisole and the phenotypes were examined. Afterwards, 10 more animals that exited the molt in the meantime were processed in the same way. DIC images were acquired with a Zeiss Axio Observer Z1 microscope using the Zen 2 software. Selections of regions and image processing were performed with Fiji.

Quantification of the protruding vulva phenotype

Synchronized L1s were plated on 2% NGM agar OP50 plates and grown at 20 °C for 52 h, before animals exhibiting either normal or protruding vulva were counted using a binocular microscope. Four plates per condition were examined (N = 4), with > 230 animals scored per condition. Significant differences to wild type were determined by a Dunnett’s test.

Quantification of the vulva bursting phenotype

Synchronized L1s were plated on 2% NGM agar OP50 plates and grown at 25 °C for 48 h, before animals exhibiting either normal or burst vulva were counted using a binocular microscope. Three plates per condition were examined (N = 3), with 240 animals scored per condition (n = 240).