Negative feedback by conserved kinases patterns the degradation of Caenorhabditis elegans Raf in vulval fate patterning

Genetic analysis of Caenorhabditis elegans vulval cell fate patterning has provided many insights into EGF receptor (EGFR) signal transduction and its regulation (Gauthier and Rocheleau, 2017; Shin and Reiner, 2018; Sternberg, 2005; Sundaram, 2013). Six multipotential vulval precursor cells (VPCs), designated P3.p to P8.p, have the potential to adopt one of three fates, termed primary, secondary or tertiary (Fig. 1A). Normally, an EGF-like inductive signal from the anchor cell (AC) of the gonad initiates pattern formation by activating LET-23/EGFR and a canonical Ras-Raf-ERK cascade in P6.p, the VPC nearest the AC. Activated ERK in P6.p promotes the primary fate and the production of Delta/Serrate/LAG-2 (DSL) ligands that activate LIN-12/Notch in the flanking VPCs, P5.p and P7.p, specifying them to adopt the secondary fate. The outer VPCs, in which neither EGFR nor Notch is active, adopt the tertiary fate and do not contribute to the vulva. This apparently simple fate pattern is invariant, precise and robust, because many different regulatory mechanisms ensure appropriate initiation, intensity, duration and integration of these LET-23/EGFR- and LIN-12/Notch-mediated signaling events.

Signal transduction by EGFR occurs via LET-60/Ras and a conserved kinase cascade composed of LIN-45/Raf, MEK-2/MEK and MPK-1/ERK (Gauthier and Rocheleau, 2017; Shin and Reiner, 2018; Sternberg, 2005; Sundaram, 2013). In this study, we investigate a negative-feedback mechanism in which high MPK-1 activity in P6.p leads to degradation of LIN-45 (de la Cova and Greenwald, 2012; see also Fig. 1B). We originally identified this negative regulatory mechanism via a computational search for potential substrates of SEL-10/FBXW7, a highly conserved substrate-recognition subunit of a multiprotein Skp-Cullin-F-box (SCF) E3 ubiquitin ligase and a commonly mutated tumor suppressor gene (Yumimoto and Nakayama, 2020). FBXW7 substrates contain a sequence called a ‘CPD’ (Cdc4 phosphodegron) that, when phosphorylated at two key residues, acts as a high-affinity binding site for FBXW7, thereby targeting that substrate for ubiquitin-mediated degradation (Nash et al., 2001; Orlicky et al., 2003). We found that LIN-45 has a CPD consensus sequence that is conserved in mammalian BRAF (de la Cova and Greenwald, 2012; see also Fig. 1C). By examining the consequences for stability and activity of LIN-45 when the CPD is mutated, and in mutants lacking sel-10, we validated LIN-45 as a SEL-10 substrate. In VPCs, LIN-45 protein degradation requires MPK-1 activity and occurs specifically in P6.p, where MPK-1 activation is highest (Fig. 1A). This negative-feedback mechanism also appears to operate in mammalian cells (see Discussion).

There is also a temporal dimension to cell fate specification of the multipotent VPCs. They are born in the first larval (L1) stage, and remain in a G0/G1-like cell cycle phase throughout the second larval (L2) stage (Ambros, 1999; Euling and Ambros, 1996; Hong et al., 1998). EGFR signal transduction in P6.p begins soon after entry into the L2 stage (de la Cova et al., 2017), but VPC fates are not fixed until later, in the third larval (L3) stage (Ferguson et al., 1987; Greenwald et al., 1983; Sternberg and Horvitz, 1986). Maintenance of cell cycle quiescence during L2 is controlled by ‘heterochronic’ genes (Euling and Ambros, 1996; Hong et al., 1998) that allow expression of genes that inhibit cell cycle cyclin-dependent kinases (CDKs) (Buck et al., 2009; Saito et al., 2004). In the early L3 stage, VPCs enter S phase and divide in the mid-L3 stage (Ambros, 1999; Sulston and Horvitz, 1977). Hydroxyurea (HU) blockade of S phase reveals that P6.p is committed to the primary fate prior to DNA replication (Ambros, 1999; Euling and Ambros, 1996; Hong et al., 1998), implying coordination between the cell cycle and VPC fate specification.

SEL-10-mediated LIN-45 protein degradation in P6.p also occurs in the L3 stage, coincident with the time of commitment to the primary fate, rather than earlier in response to MPK-1 activation during the L2 stage. This timing suggests that LIN-45 degradation is temporally gated and requires more than just MPK-1 activation, and also raises the question of how the onset of negative feedback is coordinated with developmental timing. To understand this at a mechanistic level, we turned to analysis of the LIN-45 CPD. For all known SEL-10/FBXW7 substrates, high-affinity binding requires phosphorylation of two specific residues, a ‘central threonine’ (referred to throughout as +0 Thr) and a serine at position +4 relative to the +0 Thr (+4 Ser) of the substrate CPD. The analysis presented here implicates two potential proline-directed phosphorylation sites near the CPD that are required for degradation, suggesting that phosphorylation at sites other than the +0 Thr and +4 Ser residues impinges on SEL-10-dependent degradation. We also performed a targeted RNAi screen of the 248 C. elegans protein kinase-encoding genes that have human orthologs (Deng et al., 2019; Kim et al., 2018) and identified two kinases – GSK-3 and CDK-2 – that, in addition to MPK-1, are required for LIN-45 degradation in P6.p. Using different forms of LIN-45 and an in vivo biosensor to quantify MPK-1 activity, we analyzed the roles of GSK-3 and CDK-2 in restraining Raf-MEK-ERK signalling and identified a temporal block that prevents LIN-45 degradation in the L2 stage due to cell cycle quiescence, which is relieved during the L3 stage by activation of CDKs. We propose a model wherein negative feedback acting on Raf-MEK-ERK signal transduction in VPCs is imposed by kinase-regulated protein degradation of LIN-45, with spatial and temporal patterning provided by MPK-1 and CDK-2.

To define sequences that regulate LIN-45 degradation in cis, we created single-copy transgenes that express wild-type, truncated or mutant forms of yellow fluorescent protein (YFP)-tagged LIN-45 under the control of transcriptional regulatory sequences from lin-31, referred to here as lin-31p (Tan et al., 1998; see Materials and Methods). This method enables approximately uniform expression in all six VPCs from the early L2 stage and after VPCs divide in the L3 stage (de la Cova and Greenwald, 2012; see Fig. 1A), allowing us to focus on protein stability. When YFP-tagged wild-type LIN-45 (YFP-LIN-45) is expressed using lin-31p, it is stable in all VPCs at the L2 stage and degraded in P6.p prior to cell division (Fig. 1B). Degradation in P6.p depends on sel-10 and mpk-1 and the presence of the +0 Thr and +4 Ser phospho-acceptor sites of the CPD (Fig. 1C; de la Cova and Greenwald, 2012), consistent with the sequence requirements of high-affinity CPDs (Nash et al., 2001).

We previously showed that a 262-amino-acid region of LIN-45 lacking domains for Ras binding, kinase activity and dimerization, termed LIN-45(218-480), is sufficient to promote degradation in P6.p (de la Cova and Greenwald, 2012). To further define sequence requirements for degradation, we first analyzed YFP-LIN-45(417-480), which encompasses the CPD. This fragment was efficiently targeted for degradation (Fig. 1C,D, see also Fig. 2B), and the CPD +0 Thr and +4 Ser were required (Fig. S1), indicating that it contains the LIN-45 degron required for SEL-10-mediated negative feedback. A shorter derivative lacking sequences N-terminal to the CPD, YFP-LIN-45(430-480), was also robustly degraded in P6.p (Fig. 1D), so we focused on mutating residues C-terminal to the CPD to identify other potentially important sites for LIN-45 degron function (Fig. 1D; Fig. S1). Deletion of a C-terminal region with two potential Ser-Pro (S-P) sites at S443 and S453 resulted in stabilization of the YFP reporter in P6.p (Fig. 1D).

To test whether S443 and S453 act to promote LIN-45 degradation, we mutated them individually or together in the full-length YFP-LIN-45 context (Fig. 1E) and found that they act redundantly, because only the double mutant S443A+S453A was strongly stabilized in P6.p. Both are S-P sites, suggesting that regulation of LIN-45 stability may occur via a proline-directed kinase. Mutation of F446 and F447 to alanine also resulted in stabilization in P6.p (Fig. 1E), suggesting that they are required for an additional protein-protein interaction or conformation required for LIN-45 degradation; we did not study this site further, but it is tempting to speculate that these residues are part of a binding site for one or both of the kinases we identified below.

We note that we previously proposed that amino acids H426 to L431 of LIN-45 function as an MPK-1 docking site, because this region resembles a ‘D domain’-type ERK docking site consensus (Biondi and Nebreda, 2003); we also provided support for this proposal by mutating L429 and L431, and found that alanine substitutions at these sites stabilize full-length LIN-45 (de la Cova and Greenwald, 2012). The observation here that YFP-LIN-45(430-480) retains degron function suggests that this sequence does not function as an MPK-1 docking site. Subsequent to our previous observations, it has been reported that FBXW7 substrate binding requires the ‘−1 leucine’ residue, which corresponds to L429 in LIN-45 (Welcker et al., 2013). Thus, we now believe that the L429A+L431A mutations stabilized LIN-45 because CPD function was compromised, rather than because MPK-1 docking site function was eliminated.

The presence of a phospho-regulated CPD and requirement for additional S-P sites in the LIN-45 degron suggested that protein kinases regulate LIN-45 protein stability. We performed an RNAi screen of the 248 kinases conserved between C. elegans and humans (Deng et al., 2019; Materials and Methods) for their ability to stabilize the truncated reporter YFP-LIN-45(417-480), a form that was brightly and reliably expressed in VPCs, and robustly degraded in P6.p (Fig. 1D). This screen identified two proline-directed kinases, GSK-3 and CDK-2, as required for YFP-LIN-45 degradation in P6.p in L3 stage larvae. To confirm the RNAi results, and for further analysis, we obtained gsk-3(tm2223) and cdk-2(tm2244) mutants [referred to hereafter as gsk-3(0) and cdk-2(0), respectively], both of which have large deletions and are predicted null alleles. Here and below, we included the reporter arIs222[lag-2p::2×NLS-tagRFP] (Sallee and Greenwald, 2015); because lag-2 expression in P6.p results from MPK-1-mediated phosphorylation of LIN-1/Elk1 (Jacobs et al., 1998; Underwood et al., 2017; Zhang and Greenwald, 2011). Expression of this RFP marker indicated that P6.p was appropriately induced by the inductive signal in the individuals we scored.

GSK-3, the ortholog of human GSK3B (glycogen synthase kinase 3 beta), acts in many cell processes and signal transduction pathways (Cohen and Frame, 2001). In contrast to wild-type L3 stage larvae (Fig. 2A,B), in gsk-3(0) mutants, YFP-LIN-45 protein was not degraded and was still present in P6.p (Fig. 2A,C). A transgene in which lin-31p drives expression of wild-type gsk-3(+) specifically in VPCs [lin-31p::gsk-3(+)] significantly restored YFP-LIN-45 degradation in gsk-3(0) mutants (Fig. 2C), consistent with a role for GSK-3 in the VPCs.

CDK-2 is a cyclin-dependent kinase related to the human cell cycle kinases CDK2, CDK1 and CDK3 (Kim et al., 2018), with functional similarities to the G1-to-S-promoting kinase CDK2 (Fox et al., 2011; Korzelius et al., 2011). In cdk-2(0) mutant L3 stage larvae, YFP-LIN-45 protein was not degraded in P6.p (Fig. 2A,D). Similarly, cye-1(ar95), a null allele of cyclin E (Fay and Han, 2000) referred to here as cye-1(0), also abrogated YFP-LIN-45 degradation (Fig. 2E). Expression of wild-type cye-1(+) in VPCs [lin-31p::cye-1(+)] significantly rescued YFP-LIN-45 degradation in the cye-1(0) background (Fig. 2E), supporting a role for CDK2 and cyclin E in VPCs.

CDK-2 and CYE-1 promote progression through the S phase of the cell cycle (Kipreos and van den Heuvel, 2019), raising the question of whether failure to degrade LIN-45 in the mutants resulted from failure of cell cycle progression. However, VPCs of cdk-2(0) and cye-1(0) mutants were capable of cell divisions (Fig. 2A) (Fay and Han, 2000), suggesting that CDK2-mediated LIN-45 degradation is not inextricably coupled to cell division. This conclusion was further supported by using HU to block S-phase progression and observing that YFP-LIN-45(417-480) was not stabilized (Fig. S2), indicating that CDK2 promotes degradation of LIN-45 independently of its role in promoting cell cycle progression.

We next assessed the effect of gsk-3(0) and cdk-2(0) on LIN-45 biological activity, using cell fate as a readout. In wild-type larvae, VPCs that do not receive EGFR- or Notch-mediated inputs (P3.p, P4.p and P8.p) adopt the tertiary (non-vulval) fate (Fig. 1A). Mutations that increase Ras-Raf-ERK pathway activity cause these cells to behave as if they have received EGF and to adopt the primary fate. In adults, this ectopic induction manifests as a ‘Multivulva’ (Muv) phenotype, which is characterized by large protrusions formed of ectopic vulval tissue flanking the normal vulva.

The fates of individual VPCs can be assessed during larval development using the primary fate marker arIs222[lag-2p::2×NLS-tagRFP], which is normally expressed only in P6.p and its descendants upon activation of ERK by the EGF-like inductive signal (Fig. 3A). Expression of this marker was unaffected in gsk-3(0) mutants (Fig. 3B). In cdk-2(0) mutants, a small but significant percentage of larvae displayed ectopic primary fate marker expression, although most VPCs did not (Fig. 3C). The lack of an overt effect on vulval cell fates by single mutants has been observed for many other negative regulators of EGFR-Ras-ERK activity, including sel-10, because multiple mechanisms operate redundantly (de la Cova and Greenwald, 2012; Hajnal et al., 1997; Hopper et al., 2000; Jongeward et al., 1995; Yoo et al., 2004).

LIN-45(V627E) is equivalent to human BRAF(V600E), which has constitutive kinase activity and is the product of the most common mutation of human BRAF in melanomas (Lavoie and Therrien, 2015). Using a single-copy insertion transgene (see Materials and Methods), we assessed the effect of LIN-45(V627E) expression on the fates of individual VPCs, based on the presence and pattern of ectopic primary fate marker expression. If expression is spaced in alternating VPCs, then the level of ERK activity is not high enough to overcome the opposing effects of LIN-12/Notch (Berset et al., 2001; Underwood et al., 2017; Yoo et al., 2004). When P5.p and/or P7.p adopt the primary fate despite being adjacent to P6.p, ERK signaling is high enough to prevail. Larvae expressing LIN-45(V627E) displayed ectopic primary fate in P3.p, P4.p or P8.p in 31% of animals (Fig. 3D). Notably, ectopic primary fate was never seen in P5.p or P7.p, nor in two adjacent VPCs, as expected for a moderate level of ERK activation. Larvae expressing LIN-45(V627E) in gsk-3(0) or cdk-2(0) mutants not only displayed significantly higher frequencies of ectopic primary fate marker expression (Fig. 3E,F), but also expressed this marker in adjacent VPCs (Fig. 3E,F) and occasionally even in P5.p or P7.p, consistent with high levels of Raf-MEK-ERK signaling. Thus, both gsk-3 and cdk-2 appear to act as negative regulators of Raf-MEK-ERK signaling and LIN-45 stability, consistent with a role in promoting LIN-45 degradation.

In adults, the penetrance of the Muv phenotype caused by expressing LIN-45(V627E) in VPCs was low (Fig. 3G,H). Loss of sel-10 increases the penetrance of LIN-45(V627E), causing a highly penetrant Muv phenotype (de la Cova and Greenwald, 2012). Similarly, loss of gsk-3, cdk-2 or cye-1 increased the penetrance and expressivity of the Muv phenotype of LIN-45(V627E)-expressing adults (Fig. 3G,H), consistent with a role in a negative feedback mechanism.

Because the ectopic primary fate and Muv phenotypes arise from ERK activation in VPCs other than P6.p, these results indicate that GSK-3- and CDK-2-mediated negative feedback can operate in other VPCs. As we have suggested previously for sel-10 (de la Cova and Greenwald, 2012), opposition to Ras pathway activation may normally help set a high threshold to ensure that only P6.p adopts the primary fate.

We assessed the effect of gsk-3(0) and cdk-2(0) on the level of MPK-1/ERK activity using ERK-nKTR-mClover (ERK-nKTR), a genetically encoded in vivo biosensor, as a readout. ERK-nKTR allows for quantitation of the degree of MPK-1 activation based on the nucleocytoplasmic localization of a fluorescent protein, mClover, fused to a module that is phosphorylated by MPK-1. When the module is unphosphorylated, mClover is enriched in the nucleus, whereas when the module is phosphorylated, mClover is exported and excluded from the nucleus (de la Cova et al., 2017; Regot et al., 2014). The ERK-nKTR sensor is expressed as a bicistronic transcript with mCherry-histone H2B from a single-copy insertion transgene, ensuring equivalent expression of mClover and mCherry (see Materials and Methods). The degree of MPK-1 activity may be quantified by determining the mCherry:mClover ratio of fluorescence intensity per nucleus (see Materials and Methods; de la Cova et al., 2017).

Because LIN-45 degradation occurs in early L3 (de la Cova and Greenwald, 2012), we examined ERK-nKTR at that time, before VPC division (see Materials and Methods). In wild-type L3 VPCs, ERK-nKTR was excluded from nuclei of P6.p and enriched in the nuclei of P4.p, P5.p, P7.p and P8.p, and quantification revealed that MPK-1 activity was significantly higher in P6.p than in other VPCs (Fig. 4A). Compared with wild-type larvae, LIN-45(V627E) larvae displayed significantly reduced nuclear ERK-nKTR in P4.p, P5.p, P7.p and P8.p, indicating that MPK-1 activity was increased in these cells (Fig. 4B).

We next analyzed ERK-nKTR in gsk-3(0) mutants. In a lin-45(+) background, gsk-3(0) had no obvious effect on ERK-nKTR localization, and MPK-1 activity remained restricted to P6.p (Fig. 4C). When LIN-45(V627E) was expressed in VPCs, quantification revealed that gsk-3(0); lin-45(V627E) had increased MPK-1 activity in anterior VPCs (P4.p and P5.p) but not posterior VPCs (P7.p and P8.p) compared with the activity in gsk-3(+); lin-45(V627E) (Fig. 4D). This anterior-posterior difference may be related to the proximity of posterior VPCs to a strong posterior Wnt signaling source (Sawa and Korswagen, 2013).

By contrast, cdk-2(0) affected MPK-1 activity even in a lin-45(+) background: in cdk-2(0) mutants, ERK-nKTR was less nuclear enriched, and quantification indicated that MPK-1 activity in P5.p and P7.p was significantly increased compared with that in the cdk-2(+) background (Fig. 4E), suggesting a role for CDK-2 in fine-tuning the pattern of MPK-1 activation by restricting it to P6.p. However, in cdk-2(0); lin-45(V627E) mutants, the level of MPK-1 activity was not increased compared with that of cdk-2(+); lin-45(V627E) larvae, despite the high penetrance of ectopic primary fate observed in this genotype (Fig. 4F). This result suggests that the Muv and ectopic primary fate phenotypes observed may be caused by an alternative mechanism and not further elevated MPK-1 activity. We consider the role of cdk-2 further below.

MPK-1 is necessary for degradation of LIN-45 in P6.p (de la Cova and Greenwald, 2012). To test whether activating MPK-1 is sufficient to promote degradation, we used the transgene gaIs37, referred to hereafter as ERK(act), which constitutively activates MPK-1 in all VPCs and causes a highly penetrant Muv phenotype (Eisenmann et al., 1998; Lackner and Kim, 1998). ERK(act) produced ectopic expression of the primary fate marker lag-2p::2×NLS-tagRFP at high frequency (Fig. 5A; Fig. S3) and caused YFP-LIN-45 protein levels to be very low, or not visible, in VPCs (Fig. 5A,B). Thus, ERK(act) is sufficient to induce LIN-45 degradation in all VPCs. We found that in L3 stage gsk-3(0) mutants carrying YFP-LIN-45 and ERK(act), YFP-LIN-45 was frequently stable in P5.p, P6.p and P7.p (Fig. 5B), indicating that strong activation of MPK-1 does not bypass the need for gsk-3 activity in LIN-45 degradation.

Because animals containing ERK(act) and cdk-2(0) grew poorly due to apparent pleiotropic effects, we reduced cdk-2 activity by an alternative approach that does not compromise viability or fertility. Expression of the CDK-interacting protein/kinase inhibitory protein (Cip/Kip) CKI-1 using the egl-17 promoter (egl-17p) causes CDK inactivation and results in incompletely penetrant cell cycle arrest (Hong et al., 1998). The egl-17 promoter is activated in P6.p in response to MPK-1 activation (Burdine et al., 1998). Using a egl-17p::cki-1 transgene (see Materials and Methods) combined with YFP-LIN-45, we identified individual larvae in which P6.p was arrested while P5.p and P7.p had divided (Fig. 5C), and other individuals in which P6.p was not arrested. We found that LIN-45 degradation was abrogated when P6.p was arrested (Fig. 5C,D) but usually did occur when P6.p was not arrested. Furthermore, when we combined the egl-17p::cki-1 and ERK(act) transgenes, VPCs other than P6.p also displayed incompletely penetrant cell cycle arrest (Fig. 5E), because ERK(act) drives egl-17p::cki-1 expression in all VPCs. In VPCs that were arrested, YFP-LIN-45 degradation was prevented (Fig. 5F). Thus, as for gsk-3, high MPK-1 activity does not bypass the requirement for cdk-2 in LIN-45 degradation.

In the L2 stage, VPCs remain in a G0/G1-like phase as a result of CDK inhibition by CKI-1; in the L3 stage, diminished expression of CKI-1 allows CDK activation and progression through the cell cycle to mitosis (Fig. 6A) (Kipreos and van den Heuvel, 2019). We previously showed that MPK-1 is activated in P6.p early in the L2 stage (de la Cova et al., 2017), and that LIN-45-YFP is degraded in the L3 stage (de la Cova and Greenwald, 2012). We therefore hypothesized that the timing of LIN-45 degradation might be controlled by the onset of CDK-2 activity.

We first examined YFP-LIN-45 protein expression in larvae staged by milestones of somatic gonad development (see Materials and Methods), and found that all VPCs in wild-type mid-L2 stage larvae express YFP-LIN-45 (Fig. 6B), a time point when MPK-1 is active in P6.p (de la Cova et al., 2017). This result suggests that LIN-45 protein is protected from degradation in the L2 stage, consistent with the hypothesis that CDK-2 activity provides a temporal gate for LIN-45 degradation. We then tested an alternative hypothesis, that LIN-45 is stable in the L2 stage due to insufficient MPK-1 activation, by examining YFP-LIN-45 in L2 stage larvae carrying ERK(act). YFP-LIN-45 was present in all VPCs (Fig. 6F,G), indicating that strong MPK-1 activation is not sufficient to overcome the L2 block to LIN-45 degradation.

If CDK-2 provides a temporal gate, we would expect that precocious CDK activation would cause precocious LIN-45 degradation. In C. elegans, the phosphatase CDC-14 is required to maintain cell cycle quiescence of blast cells. In cdc-14 null mutants, referred to here as cdc-14(0), VPCs fail to express CKI-1 and display incompletely penetrant premature cell division during the L2 stage (Saito et al., 2004). Precocious cell cycle progression of VPCs in the L2 stage does not cause precocious expression of fate markers or overt abnormalities of adult vulval morphology, suggesting that the only consequence of removing cdc-14 activity is CDK activation (Karp and Greenwald, 2013; Saito et al., 2004). Indeed, when cdc-14(0) VPCs divide precociously in the L2 stage, their daughter cells also behave like uncommitted VPCs (Clayton et al., 2008; Saito et al., 2004).

We took advantage of the incomplete penetrance of the cdc-14(0) mutant phenotype. When P6.p did not divide, we inferred that CDK activity had remained low; when P6.p did divide, we inferred that CDK activity had been elevated. In cdc-14(0) mutants, when P6.p had not divided prematurely, it displayed YFP-LIN-45, as in wild type (Fig. 6B,D). In contrast, when P6.p had divided prematurely, YFP-LIN-45 was degraded in one or both P6.p VPC-like daughters at a significant frequency (Fig. 6C,E). These results suggest that, normally, the onset of CDK-2 activation in the L3 stage is the temporal signal that triggers MPK-1-dependent LIN-45 degradation. In cases where P6.p divided but only one of its daughters displayed YFP-LIN-45 degradation, it may be that the different outcomes in each VPC-like daughter reflect the incomplete penetrance of the cdc-14(0) mutant phenotype, or that there is an additional regulatory mechanism in the L2 stage that also opposes degradation.

We previously showed that MPK-1 activity is necessary for LIN-45 degradation in P6.p as part of a negative-feedback loop in response to the EGF-like inductive signal (de la Cova and Greenwald, 2012). This negative feedback requires SEL-10, the substrate recognition component of the multiprotein SCF E3 ubiquitin ligase, and a sequence in LIN-45 called the CPD, which would be expected to act as a high-affinity binding site for SEL-10. Here, we identified sequence requirements other than the CPD for LIN-45 degradation in P6.p, including potential phosphorylation sites, showed that ectopic activation of MPK-1 is capable of promoting LIN-45 degradation in all VPCs and identified two additional kinases, GSK-3 and CDK-2, as essential for the negative-feedback mechanism in P6.p and threshold-setting in other VPCs. In this Discussion, we (1) propose a model for how the combined activities of MPK-1, GSK-3 and CDK-2 are required for LIN-45 degradation, and how MPK-1 and CDK-2 provide spatial and temporal information; (2) consider potential molecular mechanisms by which these kinases operate based on studies of other SEL-10/FBXW7 substrates; and (3) consider to what extent the SEL-10/FBXW7- and MPK-1/ERK-mediated regulation of LIN-45 is conserved in the case of human BRAF regulation.

During vulval induction, negative feedback acting on LIN-45 in VPCs is spatially restricted to P6.p and temporally restricted to the L3 stage. We propose a model for how the activities of three kinases, MPK-1, GSK-3 and CDK-2, combine to provide this information (Fig. 7A,B).

MPK-1 was an obvious candidate for providing spatial information because it is more highly activated in P6.p compared with other VPCs due to proximity to the AC. Indeed, engineering high MPK-1 activity in all VPCs is sufficient to cause LIN-45 protein degradation in all VPCs in the L3 stage, supporting this role in defining the spatial pattern in VPCs. However, activated MPK-1 does not bypass the requirement for GSK-3 and CDK-2, suggesting that it functions upstream of or in parallel to them, and does not cause premature degradation in the L2 stage, indicating that there is a temporal gate to LIN-45 degradation.

Our analysis indicates that CDK-2, with its associated cyclin, cyclin E, provides a temporal cue that restricts LIN-45 degradation to the L3 stage, even though MPK-1 is active in P6.p in the L2 stage (de la Cova et al., 2017). In the cdc-14 mutant, when CDKs are inappropriately activated during the L2 stage (Saito et al., 2004), we observed precocious LIN-45 degradation in P6.p. Thus, in wild-type VPCs, LIN-45 protein is degraded where MPK-1 and CDK-2 kinase activities coincide: in P6.p during the L3 stage.

Although GSK-3 does not appear to provide spatial or temporal information per se, its potential role is suggested by mammalian GSK3B. As described more fully in the next section, GSK3B phosphorylates the CPDs of many different FBXW7 substrates on the +0 Thr CPD residue of a previously phosphorylated ‘primed’ degron, resulting in a fully phosphorylated CPD that allows for recognition by FBXW7. Thus, in our model, we propose a specific mechanistic role for GSK-3 as a +0 Thr kinase, with MPK-1 and CDK-2 as potential priming kinases for initiation of LIN-45 degradation (Fig. 7C).

The three kinases we identified as regulating the spatial and temporal stability of LIN-45 have also been implicated in the regulation of the stability of other FBXW7 substrates. In humans, the best-studied substrates of FBXW7 are oncoproteins that regulate the cell cycle and growth of mammalian cells, including MYC, JUN, cyclin E and NOTCH1 proteins (reviewed by Yumimoto and Nakayama, 2020). Some of these substrates are also known to be modulated by GSK3B, CDK2 or ERK. Each of these kinases has substrate docking or phospho-acceptor site preferences that we review below; however, the degeneracy found in actual substrates makes it difficult to exclude direct phosphorylation using prediction alone. Nevertheless, the roles of GSK3B, CDK2 and ERK in regulating other FBXW7 substrates provide lessons on degradation of LIN-45.

GSK3B regulates FBXW7 binding and degradation of human MYC, JUN, cyclin E and NOTCH1, in each case by phosphorylating the +0 Thr of the substrate CPD (Espinosa et al., 2003; Wei et al., 2005; Welcker et al., 2004, 2003; Yada et al., 2004). As is true for a large number of GSK3B substrates, these sites conform to the consensus [S/T]-x-x-x-p[S/T], where a prior ‘priming’ phosphorylation present at the +4 position is recognized by a phosphate-binding pocket of GSK3B and greatly increases affinity (Frame et al., 2001). We propose that GSK-3 promotes LIN-45 degradation through phosphorylation at the +0 Thr residue, providing potential roles for MPK-1 and/or CDK-2 in priming (Fig. 7C). Intriguingly, in human cells, ERK promotes priming phosphorylation at the +4 Ser of the MYC CPD, but it is unclear whether this phosphorylation is direct (Sears et al., 2000). Indeed, other kinases, including CDK2, phosphorylate the MYC CPD +4 Ser in vitro, and contribute to MYC phosphorylation and degradation in vivo (Hydbring et al., 2010). Some FBXW7 substrates, such as SREBP, contain additional priming sites C-terminal to the +4 Ser (Bengoechea-Alonso and Ericsson, 2009). The two proline-directed sites we identified downstream of the LIN-45 CPD may therefore serve a priming function for one or both phosphorylations of the CPD.

As a proline-directed kinase, CDK2 strongly favors [S/T]-P phospho-acceptor sites (Chi et al., 2008), often containing a basic residue at positions C-terminal to the phospho-acceptor (Brown et al., 1999; Hodeify et al., 2011; Lees et al., 1992). However, there are many exceptions: for example, in human cells, CDK2 phosphorylates the +4 Ser residue of a high-affinity CPD of cyclin E that lacks a proline (Welcker et al., 2003). Thus, CDK-2 may be capable of phosphorylating several sites of the C. elegans LIN-45, including the +4 Ser, which does not have an adjacent proline, and/or the two S-P sites we identified in this work.

Many well-characterized MPK-1/ERK substrates in mammals and C. elegans contain docking sites with a ‘D domain’ motif (also known as a KIM or DEJL sequence) characterized by the consensus sequence [K/R]-x(0-5)-L-x-L, and/or a ‘DEF’ motif with the consensus F-x-F-P (Arur et al., 2009; Fantz et al., 2001; Jacobs et al., 1999). Although the LIN-45(430-480) region that is sufficient for P6.p-specific degradation does not contain an obvious docking site, it is formally possible that MPK-1 acts as a priming kinase through a non-canonical site. However, insights into the mechanism underlying the spatial specificity provided by MPK-1 may be gleaned from mammalian studies, where ERK impacts activities of FBXW7, CDK2 and GSK3B by a number of mechanisms. For example, in some human cell contexts, ERK signaling stimulates CDK2 activity, likely by an indirect mechanism mediated by the upstream CDK-activating kinase CDK7 (Chiariello et al., 2000; Lents, 2002). In C. elegans, Arur et al. (2009) found that both GSK-3 and CDK-7 proteins are direct substrates of MPK-1, and knockdown of either kinase produces germline phenotypes consistent with a role in negative feedback on MPK-1 signaling. Because activated MPK-1 does not bypass the need for gsk-3 or cdk-2, we propose that GSK-3 and CDK-2 do not act upstream of MPK-1.

In mammalian cells, Raf-MEK-ERK cascade activation results in negative feedback via phosphorylation of BRAF, inactivation and degradation (Grbovic et al., 2006; Hernandez et al., 2016; Ritt et al., 2010; Saei et al., 2018), as we have observed for LIN-45/Raf. Indeed, Hernandez et al. (2016) provide strong support for conservation of the MPK-1- and CPD-dependent degradation that we described for C. elegans LIN-45, showing that the level and half-life of BRAF depends on ERK activity and that mutation of the +0 Thr or +4 Ser residues of the conserved CPD increases BRAF stability in a mouse embryonic fibroblast cell line.

At least in some cell contexts, this negative feedback involves FBXW7. Saei et al. (2018), using human HEK293 cells, demonstrated that FBXW7 physically interacts with BRAF, and observed a loss of polyubiquitylation of BRAF protein after siRNA-mediated knockdown of FBXW7. Although Hernandez et al. (2016) concluded that ERK- and CPD-dependent negative-feedback regulation of BRAF is not mediated by FBXW7, their conclusion was based on the finding that BRAF is still degraded in the presence of a missense mutation in the substrate-binding domain, FBXW7 (R482Q). Because FBXW7 (R482Q) can still bind to BRAF and retains the F-box domain that mediates SCF complex interaction, the depletion experiments of Saei et al. (2018) appear more compelling, and suggest that the negative-feedback loop we have identified in P6.p may operate via SEL-10/FBXW7 in some cell contexts. We therefore propose that our findings of additional kinases that mediate the degradation of LIN-45 in C. elegans may also be relevant to such mammalian contexts.

All C. elegans strain names and full genotypes are listed in Table S1. The following mutant alleles were obtained from the National BioResource Project C. elegans (NBRP C. elegans, https://shigen.nig.ac.jp/c.elegans/). LGI: cdk-2(tm2244); gsk-3(tm2223). The following alleles were obtained from the Caenorhabditis Genetics Center (CGC; https://cgc.umn.edu/). LGII: cdc-14(he141) (Saito, et al., 2004). LGIII: the transgene gaIs37 [Ef1a::dSor, hs::mpk-1] (Eisenmann, et al., 1998). The allele cye-1(ar95) was previously described (Fay and Han, 2000). The RNAi-sensitized mutant nre-1(hd20) lin-15B(hd126) displays enhanced RNAi effectiveness in VPCs (Deng et al., 2019; Schmitz et al., 2007). The transgene arIs222 [lag-2p::2×NLS-tagRFP::unc-54 3′UTR] was previously described (Sallee and Greenwald, 2015; Underwood et al., 2017). Because homozygous individuals were sterile, animals carrying gsk-3(tm2223), cdk-2(tm2244) or cye-1(ar95) alleles were maintained as balanced heterozygotes. Homozygous individuals were scored based on lack of a fluorescent GFP-tagged balancer chromosome.

Plasmids for lin-45 reporter transgenes were generated following a method described in de la Cova and Greenwald (2012). Specifically, the YFP-coding sequence was fused in frame with lin-45 cDNA (Wormbase sequence Y73B6A.5a) to produce a sequence encoding N-terminally tagged YFP-LIN-45 protein. For expression of YFP-LIN-45 in VPCs, cDNAs were cloned into pCC249, an expression vector in a miniMos backbone (Frøkjær-Jensen et al., 2014) that contains a hybrid promoter derived from 5′ and intronic elements of the lin-31 gene, designated lin-31p (Tan et al., 1998), and a 3′ UTR derived from the unc-54 gene. Other than the specified mutations generated in YFP-LIN-45, all regulatory and surrounding sequences are identical in the plasmids reported. For rescue of gsk-3 and cye-1 mutants, gsk-3 and cye-1 cDNAs were amplified from the ProQuest C. elegans cDNA library (Invitrogen) and cloned into the pCC249 expression vector.

New transgenes generated in this work are described in Table S2. Single-copy miniMos insertion transgenes were generated by germline injection of N2 strain hermaphrodites with 10 ng/μl reporter plasmid and a co-injection mixture of pCFJ601, pCFJ90, pGH8 and pMA122 Frøkjær-Jensen et al. (2014); strains carrying integrated reporters were isolated as described by Frøkjær-Jensen et al. (2014). Extrachromosomal transgenes were generated by germline injection of pha-1(e2123) mutant hermaphrodites with a mixture of ScaI-digested reporter plasmid (1 ng/μl), pCW2.1 [ceh-22p::GFP] (1 ng/μl), pBX [pha-1(+)] (1 ng/μl) and digested N2 genomic DNA (50 ng/μl), as described previously (de la Cova and Greenwald, 2012).

The single-copy integrated transgene arTi88 (this work) contains the lin-31p elements described above, and drives expression of 2×NLS-YFP. Expression from lin-31p is approximately uniform in L3 stage VPCs (de la Cova and Greenwald, 2012). We scored arTi88 expression in wild type larvae and in gsk-3, cdk-2 and cye-1 mutants at the L3 stage and found that lin-31p-driven transcription was active in all VPCs of these genotypes (Fig. S4).

The single-copy integrated transgene arTi94 (this work) uses lin-31p to drive expression of a full-length YFP-tagged LIN-45. Protein accumulation of YFP-LIN-45 in L2 and L3 stage VPCs is described in this work and by de la Cova and Greenwald (2012).

The integrated transgene arIs222 (Sallee and Greenwald, 2015) uses 5′ sequences of the lag-2 gene to drive expression of 2×NLS-tagRFP. This reporter is expressed exclusively in P6.p and daughters at the L3 stage. For data reported on YFP-LIN-45 expression, we examined arIs222 in the same larvae to ensure that lag-2 was expressed in P6.p, an indicator that Raf-MEK-ERK signal transduction and subsequent primary cell fate induction had occurred.

The single-copy integrated transgene arTi11 (this work) uses the lin-31p elements to drive expression of a mutant cDNA, lin-45(V627E), followed by a 3′ UTR derived from the unc-54 gene.

The single-copy integrated transgene arTi129 (this work) uses the lin-31p elements to drive expression of wild-type gsk-3 cDNA (Wormbase sequence Y18D10A.5), followed by a 3′ UTR derived from the unc-54 gene.

The single-copy integrated transgene arTi7 (this work) uses the lin-31p elements to drive expression of wild-type cye-1 cDNA (Wormbase sequence C37A2.4a), followed by a 3′ UTR derived from the unc-54 gene.

The single-copy integrated transgene arTi87 (this work) uses the lin-31p elements to drive expression of a bi-cistronic ERK-KTR-mClover-T2A-mCherry-his-11 cDNA, followed by a 3′ UTR derived from the unc-54 gene. This transcript produces two separate reporter proteins, ERK-KTR-mClover and mCherry-HIS-11, as described previously (de la Cova, et al., 2017).

The integrated transgene gaIs37 was described previously (Eisenmann et al., 1998). The Multivulva phenotype and adult lethality produced by gaIs37 is most severe at higher temperatures. To avoid lethality, strains carrying gaIs37 were maintained at 15°C; larvae scored for all experiments were collected and grown at 20°C.

Elements derived from sequence 5′ of the egl-17 gene are capable of driving expression in P6.p during the L3 stage (Burdine et al., 1998). The extrachromosomal transgene arEx2192 (this work) uses these elements from egl-17 to drive expression of wild-type cki-1 cDNA (Wormbase sequence T05A6.1), followed by a 3′ UTR derived from the unc-54 gene. Like other reported egl-17p::cki-1 transgenes (Hong et al., 1998; Nusser-Stein et al., 2012), arEx2192 produces cell cycle arrest at incomplete penetrance. P6.p was scored as arrested if P6.p was undivided while P5.p and P7.p had divided to produce at least two daughters.

RNAi-feeding bacteria strains that target 248 different C. elegans protein kinases that have known or predicted orthology with human kinases (Deng et al., 2019; Kim et al., 2018) were used for our RNAi-feeding screen. Each bacteria strain was grown in liquid culture and spread on NGM plates containing IPTG inducer, as previously described (Kamath et al., 2003). The extrachromosomal array arEx1820 contains lin-31p-driven yfp-lin-45(417-480). This array produces very bright expression of the truncated YFP-LIN-45(417-480) protein, making it useful for our visual screen. The C. elegans strain GS7213, genotype pha-1(e2123); nre-1(hd20) lin-15B(hd126); arEx1820[yfp-lin-45(417-480)], was treated with a standard bleach/sodium hydroxide protocol to prepare eggs (Stiernagle, 2006). Approximately 200 eggs were placed on each RNAi-feeding plate and grown at 22°C. L3 stage larvae were examined after ∼40 h, a time point when larvae of early, mid and late L3 stage were easily found. Larvae were anesthetized in M9 buffer containing 10 mM levamisole, and the presence of YFP fluorescence in P6.p, daughter cells (P6.px) or granddaughter cells (P6.pxx) was scored qualitatively using 40× magnification and a Zeiss Axio Imager microscope. At least 30 L3 stage larvae were scored for each RNAi treatment.

To assess the Multivulva phenotype caused by arTi11[lin-31p::lin-45(V627E)], L4 hermaphrodites from uncrowded cultures were picked to fresh plates; adults were examined ∼24 h later using a dissecting microscope and scored for the presence of a normal vulva and the number of pseudovulvae. We note that cdk-2(0) mutants display the abnormal vulval eversion (Evl) phenotype previously reported for cye-1 mutants (Fay and Han, 2000; Seydoux et al., 1993), characterized by a single protrusion at the normal vulval position. To score ectopic vulval induction for cdk-2(0) and cye-1(0) mutants, the single, central protrusion was treated as the vulva, and supernumerary vulva-like ventral protrusions were scored as pseudovulvae. All lin-45(+) and lin-45(V627E) animals scored for the Multivulva phenotype were grown at 20°C.

To obtain L2 and L3 stage larvae, egg collections were performed by allowing adult hermaphrodites to lay eggs for 8-16 h. For all experiments with YFP-LIN-45 transgenes and arIs222[lag-2p::tagRFP], larvae were grown at 20°C. For scoring and imaging, larvae were anesthetized in M9 buffer containing 10 mM levamisole.

L2 stage larvae were examined ∼24 h after egg collection and more precisely staged based on the location and appearance of presumptive AC/VU cells of the somatic gonad. Specifically, during the early L2, expression of the arIs222[lag-2p::tagRFP] transgene is visible in four gonad cells arranged in an anterior-posterior line in the ventral gonad. By mid L2, two of these cells migrate left and right, while the presumptive AC/VU cells remain together at the midline. For data on YFP-LIN-45 protein in the L2 stage, we report larvae scored at mid and late L2.

L3 stage larvae were examined ∼36 h after egg collection. To ensure that the inductive signal was intact in animals of the genotypes scored, we examined larvae for expression of the arIs222 lag-2p::tagRFP reporter in P6.p. In addition, larvae were staged based on whether VPCs had divided, a milestone that occurs at the time of cell fate commitment. Unless otherwise specified, for data on YFP-LIN-45 protein in the L3 stage, we report larvae in which P6.p was positive for lag-2p::tagRFP and VPCs had divided to produce at least two daughters.

YFP and RFP fluorescence was scored at 40× magnification using a Zeiss Axio Imager microscope, equipped with either a Hamamatsu ORCA-ER or Zeiss Axiocam. For all YFP-LIN-45 reporters, a 500 ms exposure was used to capture the YFP image and qualitatively score VPCs as positive or negative.

To obtain L3 stage larvae for analysis of the ERK-nKTR biosensor, egg collections were performed by allowing adult hermaphrodites to lay eggs for 2 h. Larvae were grown at 25°C to enhance expression and brightness of the fluorescent ERK-nKTR biosensor, and examined 28-30 h after egg collection. For imaging, larvae were anesthetized in M9 buffer containing 0.5 mM levamisole. Z-stacks of mClover and mCherry fluorescence in VPCs were captured simultaneously at the same exposure time and laser power (500 ms, 50% laser), using 63× magnification and a Zeiss spinning-disk dual-camera system. A workflow of ERK-nKTR image quantification has been described previously (de la Cova et al., 2017). Briefly, a ratiometric image called ‘Red/Green’ was produced by dividing mCherry intensity values by mClover intensity values for each pixel. Image segmentation was performed using CellProfiler (Kamentsky et al., 2011). Nuclei were segmented based on the nuclear histone mCherry-H2B signal. For each nucleus, the upper quartile intensity of the ratiometric Red/Green image was obtained, and the mean of the five most equatorial z sections is reported as the measurement ‘Red/Green’.

When comparing two genotypes for differences in phenotype penetrance, such as YFP-LIN-45 stability, expression of the lag-2 arIs222 reporter and the Multivulva phenotype, we used the Fisher's exact test to calculate a two-tailed P-value. When comparing multiple genotypes for differences in mean Red/Green ratio, we performed a one-way ANOVA, followed by pairwise comparisons and P-value correction using the Bonferroni method.

We thank Katherine Luo, Kennedy Stacy and Elizabeth Johnson for assistance in generating reporter plasmids and transgenic strains, and Daniel Shaye and Catherine O'Keeffe for their valuable comments on this manuscript. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440), and by the NBRP C. elegans of the MEXT, Japan.