A biological clock system controls development timing in C. elegans

The complex nature of the alleles, each with at least one breakpoint in a noncoding region, precludes a straightforward attribution of the phenotypes to specific features of the LIN-42 protein. Therefore, we decided to generate precise deletions of the two conserved regions, PAS and SYQ-LT (Fig. 1A). Henceforth, and based on the functional characterization that we describe below, we will refer to SYQ-LT as CK1BD for CK1-Binding Domain. To our surprise, lin-42(wrd67[∆PAS]) mutant animals exhibited essentially wild-type molting patterns in the luciferase assay, resembling neither lin-42(n1089) nor lin-42(ok2385) mutant animals. By contrast, lin-42(wrd63[∆CK1BD]) mutant animals exhibited increasing loss of molting synchrony over time (Fig. 1B). However, unlike lin-42(ok2385), lin-42∆CK1BD mutant animals executed four detectable molts during the assay (Figs. 1B and EV1D).

To get a better understanding of the developmental relevance of each domain, we examined additional phenotypes (Fig. 1D–F; Appendix Table S1). We found that all four alleles affected brood sizes, albeit to varying degrees, in the order ok2385»∆CK1BD ≈ n1089 > ∆PAS of decreasing severity (Fig. 1E; Appendix Table S1). Only ok2385 and n1089 caused highly penetrant egg-laying (Egl) and precocious alae defects (Edelman et al, 2016; Monsalve et al, 2011; Tennessen et al, 2006), whereas ∆PAS and ∆CK1BD mutants had completely wild-type egg laying and only a low penetrance of precocious alae (Fig. 1C,D; Appendix Table S1). Finally, both ok2385 and, to a lesser extent, ∆CK1BD caused larval arrest and lethality, a phenotype interestingly not observed in the luciferase assay, suggesting an environmental dependency (Fig. 1F; Appendix Table S1). Taken together, these results indicate surprisingly mild phenotypes upon loss of the LIN-42 PAS region and highlight an important function of the CK1BD for rhythmic molting.

The mRNA levels of both lin-42 and kin-20 oscillate (Hendriks et al, 2014; Meeuse et al, 2020). Further analysis of the published sequencing data revealed that the peak phases of the two RNAs differ, but that there are periods where both transcripts accumulate (Fig. 2B). Periods of co-expression were also observed for the proteins: we found by Western blotting that the levels of endogenously tagged KIN-20 protein were largely invariant over time, whereas those of endogenously tagged LIN-42 protein changed rhythmically (Fig. 2C). Both KIN-20 and LIN-42 occur in several isoforms that partially overlap in size (Fig. 2C; Appendix Fig. S2). We did not pursue the isoform identity of each band further. Finally, we confirmed a physical interaction with a reciprocal anti-HA pulldown from lin-42(xe321[3xflag::lin-42]); kin-20(xe328[3xflag::ha::kin-20]) animal lysates, again using endogenously tagged proteins (Fig. 2D). Taken together, we conclude that LIN-42 and KIN-20 form a complex in vivo.

We produced recombinant proteins to test this possibility. We were unable to express soluble KIN-20 from a bacterial or insect system, so we used recombinant human CK1δ kinase domain (hereafter referred to as CK1), similar to previous studies of KIN-20 (LaBella et al, 2020). In vitro pulldown assays with biotinylated LIN-42 protein containing the SYQ/CK1BD-A and LT/CK1BD-B domains as well as C-terminal tail (residues 402–589) revealed that CK1 bound to wild-type LIN-42 protein. Deletion of either the SYQ/CK1BD-A or LT/CK1BD-B motifs reduced this interaction (Fig. 3B).

To test whether the CK1BD is sufficient to bind CK1, we purified a truncated fragment of LIN-42 lacking the C-terminal tail (CK1BD) and performed bio-layer interferometry (BLI) assays with biotinylated CK1 and titrating LIN-42 fragments. This optical technique measures macromolecular interactions in real-time, allowing the extraction of kinetic parameters (k_on and k_off) from model fits of the association and dissociation phases across a range of analyte concentrations. These parameters are then used to calculate the equilibrium dissociation constant (K_D, also known as an affinity constant). The amplitude of the BLI response is proportional to the mass of the bound complex and is therefore reflective of the molecular weight of the analyte. Accordingly, the CK1BD fragment produced a lower response signal compared to the CK1BD + Tail protein (residues 402–598; same construct used in Fig. 3B), reflecting the smaller size of the CK1BD alone rather than a difference in affinity. Global fitting across concentrations revealed that both constructs bound to biotinylated CK1 with comparable nanomolar affinities (K_D = 303 ± 107 nM (CK1BD) and 309 ± 153 nM (CK1BD + Tail) (Fig. 3C–E).

We also purified an N-terminal LIN-42 construct that contains the single canonical PAS-B domain and the adjacent structured but unclassified element (N-terminus; residues 1–315) to confirm that this region, as in the mammalian system, is not involved in binding to CK1. We titrated LIN-42 N-terminus as high as 200 µM and, as expected, we did not observe any evidence of binding to biotinylated CK1 (Appendix Fig. ). Taken together, these results demonstrate that the LIN-42 SYQ and LT motifs encompass a minimal CK1BD. S3

In mammals, the CK1BD region of PER1/2 is composed of two structured helical motifs (CK1BD-A and CK1BD-B) that facilitate stable binding to CK1. To confirm that the LIN-42 SYQ and LT regions function in a similar binding capacity as the CK1BD-A and CK1BD-B motifs, we performed BLI experiments with LIN-42 CK1BD + Tail protein in which the SYQ/CK1BD-A and LT/CK1BD-B motifs were removed individually and in tandem. LIN-42-binding to CK1 was not detectable when CK1BD-B was deleted, either singly (CK1BD-∆B) or in combination with CK1BD-A (CK1BD-∆A/B) even at higher concentrations (Fig. 3C,F,G). In contrast, deletion of the CK1BD-A domain reduced the affinity of LIN-42 for CK1 approximately 10-fold (K_D = ~3 µM) (Fig. 3C,H), necessitating higher concentrations of the CK1BD-∆A + Tail protein to determine an accurate K_D (Fig. 3H). These data indicate that CK1BD-A and B both contribute to high-affinity binding of CK1 in vitro, while revealing a more important contribution of the CK1BD-B motif, consistent with recent work that demonstrated an essential role of the human PER2 CK1BD-B motif for stable association with CK1 in a mouse model and mammalian cells (An et al, 2022). Collectively, our data establish that the LIN-42 CK1BD is functionally conserved and mediates stable binding to the human CK1 kinase domain.

To explore the kinetics of CK1-mediated phosphorylation of the LIN-42 extended C-terminus, we performed substrate titration experiments using an ADP-Glo enzymatic assay. Here, we observed a decrease in kinase activity with high levels of wild-type LIN-42 CK1BD + Tail (Fig. 4D). This result mirrors the ability of phosphorylated PER as well as the apoptosis substrate p63, to inhibit the activity of CK1 (DBT in Drosophila) in both mammalian and Drosophila systems through product inhibition (Gebel et al, 2020).

Closer examination revealed that deletion of the LIN-42 CK1BD-A motif relieved feedback inhibition (Fig. 4D), explaining why the ³²P-ATP data had shown little decrease in the overall phosphorylation level (Fig. 4A–C) despite reduced affinity for the kinase (Fig. 3C). These data suggest to us that the SYQ/CK1BD-A motif contributes to an inhibitory mechanism, possibly by stabilizing a non-productive complex between CK1 and phosphorylated substrate, thereby limiting product turnover. Without SYQ/CK1BD-A, product release may be facilitated, enabling more sustained CK1 activity despite weaker binding. Such a mechanism would be consistent with recent studies identifying a conserved mode of CK1 product inhibition (Gebel et al, 2020; Philpott et al, 2023; Harold et al, 2024). Conversely, deletion of the B motif either in isolation or with the A motif, caused a significant reduction in overall phosphorylation of CK1BD + Tail (Fig. 4D). No serine and only 4 threonine residues are removed in this deletion, so this effect was likely not due to loss of phosphosites. The catalytic efficiency (k_cat/K_m) of CK1 for each LIN-42 substrate decreased significantly with deletion of the CK1BD motifs relative to wild-type, with activity most severely impacted upon loss of the CK1BD-B motif (Fig. 4D,E). Taken together, these results suggest that while both motifs are involved in binding to CK1, deletion of the CK1BD-A region predominantly relieves feedback inhibition without fully disrupting the anchoring interaction seen with the deletion of the CK1BD-B domain. Moreover, and similar to the mammalian PER–CK1 complex (Marzoll et al, 2022), loss of the anchoring interaction compromises, but does not fully abrogate phosphorylation.

To determine potential CK1-dependent LIN-42 phosphorylation sites, we performed in vitro kinase reactions followed by phosphoenrichment and mass spectrometry to identify phosphopeptides. There are a total of 26 serine and 21 threonine residues in the CK1BD + Tail construct. Of these, 12 serine and 3 threonine sites in the C-terminus were phosphorylated upon incubation of LIN-42 CK1BD + Tail with CK1 at least once in three replicates in vitro (Fig. 4F). Although we detected low levels of phosphorylation on LIN-42 constructs lacking the tail in the ³²P-ATP assay (Fig. 4A), no serine and only 1 threonine phosphorylation within the CK1BD was detected in vitro, reflecting possible limitations in the assays.

To test whether LIN-42 was phosphorylated in vivo, we immunoprecipitated endogenously tagged 3xFLAG::LIN-42 at various time points from a population of synchronized L2 stage larvae and subjected precipitates to mass spectrometry. We identified 15 serine residues, 1 threonine residue, and 1 tyrosine residue that were phosphorylated on LIN-42. All but four of these phosphosites (all serine residues) are located on the CK1BD + Tail (Fig. 4F). Among these, 6 serine residues overlap with the 13 phosphoserines identified after the in vitro reaction while no threonine or tyrosine residues were phosphorylated in both the in vitro and in vivo datasets (Fig. 4F). The overlap of in vitro and in vivo results is consistent with the possibility that LIN-42 is a CK1 substrate in vivo. Whether incomplete overlap reflects the activity of additional kinases, phosphatases, and/or specificity factors in vivo, or technical differences in the experimental approaches remains to be determined.

Deletion of a major phosphorylation target, the LIN-42 tail, did not recapitulate lin-42∆CK1BD mutant molting timing defects. This prompted the question whether KIN-20 was at all required for rhythmic molting. Hence, we monitored molt timing for kin-20(ok505) null mutant animals (kin-20(0)). kin-20 loss recapitulated both the slow development and arrhythmic molt phenotypes of lin-42(ok2835) and lin-42∆CK1BD mutations, but with more pronounced arrhythmicity (Figs. 1B, 5A, and EV3A–C). Like lin-42(ok2835) and lin-42∆CK1BD mutations, the kin-20(0) mutation also caused larval arrest when animals were grown on the plate but not in the liquid culture luciferase assay (Figs. 1F and 5E). Additionally, we found that kin-20(0) mutant animals resembled lin-42(ok2835) mutants in their severe egg-laying defects and brood size reduction, phenotypes that are weaker or even absent from lin-42∆CK1BD animals (Figs. 1D,E and 5B,C). Finally, like lin-42∆CK1BD animals but unlike lin-42(ok2385) mutants, kin-20 null mutants did not develop precocious alae (Figs. 1C and 5D), consistent with earlier work (Banerjee et al, 2005).

To assess the relevance of KIN-20's enzymatic activity for rhythmic molting, we engineered a D310A point mutation into the endogenous kin-20 locus to abrogate catalytic activity. This mutation disrupts the HRD motif in the catalytic loop that coordinates phosphorylated residues (Fig. EV2) (Johnson et al, 1996; Nolen et al, 2004). The resulting kin-20(xe355[D310A]) animals recapitulated the kin-20(0) mutant arrhythmic molting phenotype, although KIN-20 levels were not decreased (Figs. 5A and EV3A–C; Appendix Fig. S4A–C). Indeed, we observed a trend towards increased accumulation of the mutant KIN-20 protein, possibly indicating autoregulation but not further pursued by us (Appendix Fig. S4A). We conclude that KIN-20 and especially its enzymatic activity are required for rhythmic molting.

The timing of increased nuclear KIN-20 in the middle of a larval stage coincides with peak accumulation of LIN-42 (Fig. 2C). Moreover, using endogenously tagged GFP::LIN-42 and wrmScarlet::KIN-20, we found that the two proteins co-localized in the nuclei of epidermal seam and hyp7 cells (Fig. 6B). Hence, we wondered if LIN-42 contributed to the change in KIN-20 accumulation. To test this notion, we examined how the lack of LIN-42-binding affected KIN-20 accumulation dynamics. Specifically, we visualized GFP::KIN-20 in lin-42∆CK1BD mutant animals during mid-L4 stage. We observed reduced nuclear GFP signal relative to wild-type animals (Fig. 6C,E) in both hypodermal and seam cell nuclei and little change in cytoplasmic levels. We also visualized GFP::KIN-20 in lin-42∆Tail mutant animals. Surprisingly, this deletion also reduced KIN-20 nuclear accumulation similar to the lin-42∆CK1BD mutants (Fig. 6D,E). These data indicate that LIN-42 promotes nuclear accumulation of KIN-20, but that additional mechanisms may contribute to the extensive arrhythmia of lin-42∆CK1BD molting.

The PAS domains are the most notable feature of LIN-42 sequence conservation and were thought to be an interaction platform for a developmental stage-specific binding partner (Jeon et al, 1999). Previous work characterizing the lin-42(ok2385) and n1089 alleles suggested that the PAS domains were necessary for the heterochronic functions of LIN-42 (Monsalve et al, 2011; Tennessen et al, 2006). Yet, our precise deletion of the LIN-42 PAS domains caused only a mild precocious alae phenotype with significantly less penetrance than in lin-42(n1089) mutants, suggesting that this model is incorrect. While lin-42(n1089) removes the sequence of the PAS domains, it also results in a truncation of lin-42b and removal of part of the lin-42a promoter sequence (Monsalve et al, 2011; Tennessen et al, 2006). The n1089 phenotype could reflect loss of the PAS domains combined with potential reduction in expression of other elements such as the C-terminal tail or an NHR-85 interaction site (Kinney et al, 2023).

The weak phenotypes seen with our precise deletion are particularly striking when considering that the PAS domains, and specifically the highly conserved PAS-B domain, can mediate LIN-42 dimerization (Lamberti et al, 2024)—a function that is important for PER activity in circadian rhythms (Zheng et al, 1999). Although it remains to be determined whether the PAS domains are at all necessary for LIN-42 dimerization in vivo, it seems possible that LIN-42's developmental timing function does not require dimerization.

Contrasting with the PAS domain, we found the LIN-42 SYQ/LT regions to be required for rhythmic molting, and we could demonstrate in vitro that they constitute a functional CK1BD that anchors CK1, promotes LIN-42 phosphorylation, and mediates kinase inhibition. Moreover, the patterns of phosphorylation that CK1 generates on LIN-42 in vitro overlap with those that we observed on LIN-42 in vivo. We note that it remains to be demonstrated to which extent the in vivo phosphorylation events depend on KIN-20. Our attempts to address this question by examining LIN-42 phosphorylation in strains lacking active KIN-20 were thwarted by the slow growth, sickness, and arrhythmic molting observed in these animals.

The two CK1BD subdomains have separable functions in vitro. CK1BD-B plays a more prominent role in kinase binding, mirroring the anchoring function of its mammalian and Drosophila equivalents (An et al, 2022; Marzoll et al, 2022; Philpott et al, 2023). Anchoring allows CK1 to target lower affinity, non-consensus motifs that are key regulatory sites in fly/mammalian circadian clocks (An et al, 2022; Kim et al, 2007; Marzoll et al, 2022; Nawathean et al, 2007; Philpott et al, 2023). However, without the PER CK1BD-B, not all capacity for phosphorylation is lost, suggesting that high-affinity binding is not essential for phosphorylation but promotes it (Marzoll et al, 2022). For LIN-42 CK1BD-∆B, we also see low levels of phosphorylation by CK1 in vitro (Fig. 4), suggesting that it may play a similar role in controlling KIN-20 activity.

The LIN-42 CK1BD-A plays a less prominent role in binding but rather is involved in CK1 inhibition. CK1δ-dependent phosphorylation of both mammalian and Drosophila PER at sites near the CK1BD leads to feedback inhibition of CK1 through conserved anion binding sites (Philpott et al, 2023). Our in vitro LIN-42 phosphorylation data reveal a decrease in enzyme catalytic activity as substrate concentration increases, dependent on the CK1BD-A, consistent with feedback product inhibition of CK1 (Fig. 4). Deletion of the LIN-42 CK1BD-A motif relieved the inhibition, presumably by supporting release of phosphorylated product and thereby preventing trapping of the enzyme in an inactive complex. Further enzymatic and structural studies will be required to distinguish between these modes of regulation. Although our experiments used human CK1 for technical reasons, given the high degree of conservation between CK1 and KIN-20 including complete conservation of the active site (Fig. EV2), we predict that these results will be transferable to C. elegans KIN-20.

Many models of circadian clock function highlight phosphorylation of PER by CK1 as the central consequence of their interaction, controlling PER stability and thereby setting the period length. Intriguingly, although CK1 heavily phosphorylates the LIN-42 C-terminal tail, including on serine and threonine residues that are conserved among Caenorhabditis nematodes, but not on more distantly related nematodes (Fig. EV5), deletion of this sequence impacts molting rhythmicity only modestly, affecting mostly the fourth molt. These data suggest either phosphosite redundancy, or that regulated phosphorylation of LIN-42 is not the key function of this complex for controlling molting timing. Consistent with the latter notion, loss of KIN-20 or its catalytic activity impair rhythmic molting much more dramatically than loss of binding to LIN-42 through the CK1BD.

For PER–CK1, several lines of evidence show that reciprocal to CK1-mediated regulation of PER, PER also regulates CK1 activity, both on itself and potentially other targets such as CLOCK. Indeed, PER and CRY function as a bridging complex to provide CK1 access to CLOCK (Aryal et al, 2017; Cao et al, 2021), and mutation of the CK1BD causes defects in PER and CK1 nuclear accumulation as well as CK1-mediated phosphorylation of CLOCK (Cao et al, 2021; Chiou et al, 2016). Product inhibition through phosphorylated PER may thus limit phosphorylation not only of PER itself, but potentially also of other substrates, including CLOCK. Strikingly, KIN-20 also exhibits dynamic subcellular accumulation, with nuclear accumulation depending on LIN-42 binding. Whether this nuclear accumulation reflects co-transport, nuclear retention, or stabilization of the nuclear KIN-20 pool is not known. We note that reducing nuclear KIN-20 levels appears insufficient to explain the arrhythmic molting phenotype of lin-42∆CK1BD animals since such a reduction also occurs in lin-42∆Tail animals that exhibit rhythmic molting. Hence, we speculate that it is the combination of promoting nuclear KIN-20 accumulation and regulating its activity that explains the importance of the CK1BD for rhythmic molting. Such a scenario may also explain the previous observation that forced expression of the short LIN-42a isoform, largely comprising the CK1BD + Tail, caused extended molt durations (Monsalve et al, 2011), as this might sequester KIN-20 and/or inhibit its kinase activity. However, it remains possible that dynamic and regulated nuclear accumulation of KIN-20 is not at the heart of the LIN-42–KIN-20 complex function.

Our data not only support the existence of LIN-42 and KIN-20 in a stable complex that is important for molting timing but also suggest that LIN-42 and KIN-20 can function independently of one another in some contexts. Neither kin-20(lf) nor lin-42∆CK1BD mutations cause precocious alae, a phenotype observed with several other lin-42 mutant alleles, including n1089 and deletion of the LIN-42 C-terminal tail.

These findings agree with previous genetic data that suggested independent functions of lin-42 and kin-20 in regulating the expression of the heterochronic let-7 microRNA (Rhodehouse et al, 2018). A recent study on LIN-42's function in regulating another heterochronic miRNA, lin-4, indicated that physical interaction with the NHR-85 transcription factor may support LIN-42's activity in this process, and perhaps the heterochronic pathway more generally (Kinney et al, 2023). Relatively mild effects upon nhr-85 deletion, and LIN-42's ability to interact with additional transcription factors in a yeast two-hybrid assay suggest that additional functionally relevant interactions partners remain to be identified. Exploring the relative contributions of the structured N-terminus and C-terminal tail to the heterochronic functions of LIN-42 are a particularly interesting future direction.

The apparent differences in wiring among components of the heterochronic pathway and the molting timer may reflect their different modes of timekeeping where the heterochronic pathway is chiefly concerned with the order of events and the molting timer—similar to the circadian clock—with the tempo and/or robustness of their execution. Indeed, although LIN-42/PER have previously been suggested to provide an evolutionary link between circadian (PER) and heterochronic (LIN-42) timing systems (Jeon et al, 1999), our data suggest that the link primarily derives from functional conservation between circadian and molting timing. Hence, we may speculate that the heterochronic function of LIN-42, which appears to involve neither rhythmic activity nor conserved interactions with orthologues of other circadian clock proteins (Banerjee et al, 2005; Gissendanner et al, 2004; Hasegawa et al, 2005; Jeon et al, 1999; Kostrouchova et al, 1998; Migliori et al, 2023) arose secondarily to, or independent from, a rhythmic timing function. Finally, we note with interest that the phosphorylation of PER by CK1 is an important component of temperature compensation in the circadian clock (Narasimamurthy and Virshup, 2021) that allows this clock to maintain its period despite variation in temperature. By contrast, the period of C. elegans molting and oscillatory gene expression changes with temperature. Our data that the LIN-42 tail, where the bulk of CK1 mediated phosphorylation occurs, is dispensable for rhythmic molting may thus reflect the lack of robust temperature compensation in the C. elegans developmental clock.