EGFR signal transduction is downregulated in C. elegans vulval precursor cells during dauer diapause

Caenorhabditis elegans larvae display remarkable developmental plasticity in response to environmental conditions, and integrate information about population density, temperature, and food availability to make decisions about developmental progression (reviewed by Baugh and Hu, 2020; Fielenbach and Antebi, 2008; Hu, 2007) (Fig. 1A). In favorable conditions, larvae develop continuously through four larval stages (L1-L4), separated by molts, to become reproductive adults. Adverse conditions in the L1 stage lead to a prolonged second larval stage (L2d), and if they persist, L2d larvae molt into dauer larvae. Although dauer has been called an ‘alternative L3 stage’, it is a distinct larval state, with morphological and behavioral specializations that facilitate survival and dispersal (Golden and Riddle, 1984). Dauer larvae can survive for several months – many times the lifespan of a worm that develops continuously (Klass and Hirsh, 1976). Favorable conditions induce recovery and developmental progression to reproductive adulthood. Post-dauer worms appear overtly similar to continuously developing worms (Cassada and Russell, 1975), implying the existence of robust protective mechanisms during developmental arrest.

Protective mechanisms include the maintenance of multipotency of precursor cells that give rise to the vulva (Fig. 1B,C). Six vulval precursor cells (VPCs), named P3.p-P8.p. have the potential to generate vulval cells. The specification of the VPCs by cell–cell interactions has provided many important insights into epidermal growth factor (EGF) receptor (EGFR) signaling and its regulation (reviewed by Gauthier and Rocheleau, 2017; Shin and Reiner, 2018; Sternberg, 2005; Sundaram, 2006, 2013). During continuous development, beginning in L2, an EGF signal from the anchor cell (AC) of the gonad activates EGFR and the canonical signal transduction cascade (Ras-Raf-MEK-ERK) in the nearest VPC, P6.p. EGF signaling plays different roles in L2 versus L3. During the L2 stage, EGF and Wnt signaling prevent fusion of P6.p with the hypodermal syncytium (Eisenmann et al., 1998; Myers and Greenwald, 2007). During the L3 stage, EGF signaling induces P6.p to adopt the primary (1°) vulval fate, characterized by expression of EGFR pathway target genes, a distinctive cell division pattern in the L3 stage, and different terminal cell types in the L4 stage.

In dauer life history, the VPCs undergo a reprogramming-like event that preserves their multipotency (Fig. 1D). As first described in mutants that exhibit premature induction of P6.p in the L2 or L2d stages, such that P6.p produces daughters or granddaughters before dauer entry, the lineage does not simply resume after recovery; instead, one of the descendants is induced to adopt the 1° fate by the EGF signal from the AC, suggesting that committed descendants of P6.p are reprogrammed back to a multipotent VPC-like state (Euling and Ambros, 1996). Similarly, in wild type, transcriptional targets of EGFR signaling are expressed in the L2d-dauer molt and then silenced upon dauer entry, suggesting that restoration to multipotency is a normal feature of dauer life history (Karp and Greenwald, 2013).

Dauer development induces changes to gene expression mediated by both global and gene-specific regulation. Global mechanisms are suggested by the observations that dauers have decreased transcription and mRNA abundance overall (Dalley and Golomb, 1992; Snutch and Baillie, 1983), and by changes to the histone landscape implying dauer-specific chromatin remodeling (Hall et al., 2010; Jones et al., 2001). Gene-specific changes are also seen, suggesting that different gene regulatory circuitry is imposed during dauer. For example, dauer alters the complement of microRNAs that regulate the transcription factor HBL-1 (Karp and Ambros, 2012). The reprogramming of VPCs in dauer larvae involves losing expression of transcriptional targets of the EGFR-Ras-ERK pathway (Karp and Greenwald, 2013), which could reflect global or context-specific mechanisms that operate at the level of target gene expression. However, it is unknown whether EGFR-Ras-ERK signal transduction decreases in dauer VPCs to cause silencing exclusively or in parallel to mechanisms that operate at the level of gene expression.

Here, we used a genetically encoded biosensor for ERK activity and found that that MPK-1/ERK activation is downregulated in P6.p beginning in the L2d-dauer molt and that activity remains low in dauer larvae. We interrogated the nature of ERK activity downregulation and found that dauer VPCs are profoundly desensitized to EGF despite the presence and correct localization of EGFR in VPCs. In dauer, ERK activity can be increased by an allele that encodes constitutively active Ras, but not by an allele that encodes constitutively active EGFR, suggesting there is a mechanism that prevents EGFR signal transduction at a step upstream of Ras activation. Increasing ERK activity by constitutively activating Ras or Raf does not release VPCs from quiescence and cause them to divide, indicating that desensitization of VPCs to EGF is distinct from, and not the result of, a single global mechanism to prevent VPCs from developmental progression during dauer diapause. Finally, during our analysis, we also found that the gene regulatory circuitry used to maintain VPC competence in dauer differs from that of continuous development. In summary, our analysis indicates that dauer VPCs have some L2-like features rather than L3-like features of continuous development as well as unique features specific to the dauer stage.

The six multipotential VPCs are spatially patterned by differential activation of EGFR and LIN-12/Notch. During wild-type development, P6.p is induced to have a high level of EGFR-Ras-ERK signaling and produces ligands that activate LIN-12/Notch in the neighboring VPCs, P5.p and P7.p. The “outer” VPCs, P3.p, P4.p, and P8.p, do not experience significant activation of either EGFR-Ras-ERK or LIN-12/Notch. In this study, we were primarily concerned with the regulation of EGFR-Ras-ERK activity in P6.p during dauer life history, and we used P4.p as a neutral VPC for normalization or comparison as indicated below.

In this study, we used a temperature-sensitive mutant that compromises TGFβ signaling [daf-7(e1372ts), referred to hereafter as daf-7(ts)], to drive constitutive dauer formation at 25°C, and corroborated key findings in dauer larvae induced by adverse conditions in a daf-7(+) background (‘natural dauers’). The use of daf-7(ts) greatly facilitates the study of pre-dauer larvae and the examination of animals that have been in dauer for a defined amount of time (Karp, 2018; Larsen et al., 1995; Vowels and Thomas, 1992); daf-7(ts) was also used to drive dauer formation in prior studies of VPC quiescence that provided the impetus for this work (Karp and Greenwald, 2013).

To examine EGFR-Ras-ERK signal transduction directly, we used ERK-nKTR, a fluorescent biosensor for quantifying MPK-1/ERK activity in vivo (de la Cova et al., 2017), validated in VPCs and other tissues in C. elegans (de la Cova et al., 2017, 2020; Maxeiner et al., 2019; Mereu et al., 2020; Robinson-Thiewes et al., 2021). The ERK-nKTR contains a kinase translocation reporter (KTR) containing ERK-dependent phosphorylation sites that promote nuclear export, resulting in cytoplasmic localization of a green fluorescent protein when MPK-1 is active (Fig. 2A), produced from a bicistronic transcript that also encodes a red fluorescent histone protein that marks the nucleus. The red/green ratio of the static red marker to the shuttling green reporter in nuclei is a quantitative measurement of net MPK-1/ERK activity. In continuously developing L2 and L3 larvae, the red/green value is high in P6.p, the VPC closest to the LIN-3/EGF-producing AC, and low in the other VPCs (de la Cova et al., 2017).

We imaged the ERK-nKTR in daf-7(ts) dauers at four different time points: L2d, the L2d-dauer molt, ∼2 days into dauer (young dauers) and ∼12 days into dauer (older dauers) (Fig. 2B,C). To compare ERK activity between stages, we normalized ERK activity measured in P6.p to activity measured in P4.p, a cell in which ERK is not activated by the inductive signal (de la Cova et al., 2017). We found that normalized ERK activity in P6.p decreases in wild-type animals during the L2d-dauer molt and remains low in dauer, a phenotype in stark contrast to the persistent, high level of ERK activity specifically in P6.p during continuous development L2 and L3 (de la Cova et al., 2017).

Similarly, we observed strongly downregulated ERK activity in P6.p in dauer larvae that were induced by other treatments. We drove dauer formation using the dauer constitutive temperature-sensitive mutant daf-2(e1370ts) [hereafter daf-2(ts)], which reduces insulin signaling, and found that ERK activity in P6.p is significantly lower in dauer than in L2d (Fig. 2D), indicating that downregulation of ERK activity is not specific to driving dauer formation by reducing daf-7. In addition, ERK activity in P6.p is low in natural dauers compared with continuous development L2 (Fig. 2E), verifying that decreased ERK activity is a property of the dauer state and not a secondary consequence of daf-7(ts) or daf-2(ts) constitutive dauer formation.

Although ERK activity in P6.p was decreased in dauer, the average ERK activity remained higher in P6.p than in P4.p in all dauer-promoting conditions. This suggested that ERK activity in P6.p is lower during dauer compared with L2d but not completely abrogated. If so, we would expect ERK activation to be further decreased in P6.p, but not in P4.p, in the absence of the EGF-like inductive signal. We removed the inductive signal using the mutant hlh-2(ar614), which carries a deletion of hlh-2 regulatory sequences and fails to specify an AC (Attner et al., 2019) (Fig. S1) (see Materials and Methods). Indeed, ERK activity in P6.p was significantly lower in hlh-2(ar614) AC-less dauers than in wild-type dauers, whereas ERK activity in P4.p was not altered by loss of the AC (Fig. 2F).

Because ERK activity is significantly lower in P6.p in dauer larvae compared with continuous development but is not completely absent, we infer that there is a mechanism that attenuates but does not eliminate either the production of the inductive signal or the response to it. In a previous study, loss of daf-16/FOXO in VPCs was found to release P6.p from cell cycle arrest and allow the EGFR-Ras-ERK target gene lag-2 to be expressed in about half of dauer larvae examined (Karp and Greenwald, 2013). As expected, the inferred mechanism appears to be regulated by daf-16, as ERK activity is elevated in daf-16(0) dauers (Fig. S2). We therefore proceeded to investigate the level at which such attenuation occurs.

A simple hypothesis to explain low ERK activity in dauer VPCs is that expression of the EGF-like ligand LIN-3 in the AC is downregulated. A transcriptional reporter transgene that has been used historically to assay lin-3 transcription (Abdus-Saboor et al., 2011; Hwang and Sternberg, 2004) showed bright expression in the AC in natural dauers (Fig. S3A). However, as lin-3 encodes several predicted isoforms with different regulatory regions, we also wanted to examine endogenous transcription. To do so, we created a synthetic operon by inserting sl2::nls::TdTomato::nls immediately after the lin-3 termination codon, as the last exon is common to all isoforms. This endogenous transcriptional reporter, lin-3(ar656[lin-3::sl2::nls::TdTomato::nls]), is expressed in the four proximal gonadal cells that are initially competent to become the AC and remains bright in the AC in daf-7(ts) constitutive dauers and natural dauers (Fig. 3A, Fig. S3B,D), indicating that transcriptional regulation of lin-3 is not the proximate cause for ERK activity downregulation in dauer larvae.

To test potential post-transcriptional levels of regulation, we used heterologous sequences to drive high-level expression of two different forms of LIN-3 in all somatic gonad cells (Fig. 3B). One is the natural transmembrane LIN-3 isoform S [‘LIN-3(S)’], which efficiently rescues vulval formation in a lin-3(0) mutant (Pu et al., 2017); the other is an engineered secreted EGF moiety [‘LIN-3(EGF)’] that is sufficient to activate EGFR signaling in VPCs (Katz et al., 1995), and should bypass any potential requirement for proteolytic processing of a transmembrane EGF precursor in the AC. For these experiments, we used tissue-specific expression of Cre recombinase (ckb-3p::Cre) to excise a flexon stop cassette (Shaffer and Greenwald, 2022), thereby enabling expression of lin-3 under the strong ubiquitous rps-27 (ribosomal protein) promoter specifically in all 12 cells of the somatic gonad primordium in dauer larvae (Fig. 3C, Fig. S3E). Strong expression of LIN-3(S) or LIN-3(EGF) from the somatic gonad in continuous development resulted in a highly penetrant multivulva (Muv) phenotype due to P4.p and other ‘outer’ VPCs ectopically adopting vulval fates in response to excess ligand, confirming that these ligands effectively activate EGFR-Ras-ERK signaling (Fig. 3D). In contrast, in dauer larvae, these forms of LIN-3 did not result in ectopic activation of ERK in P4.p or increased ERK activity in P6.p compared with ‘wild-type’ control dauers (Fig. 3E). These observations suggest that dauer VPCs are not readily responsive to ligand produced by the gonad even when the expression level is increased above the endogenous level.

We also considered whether changes to the basement membrane of the somatic gonad or the pseudocoelomic basal lamina might prevent the EGF signal from reaching EGFR on the VPCs. To test this possibility, we examined lin-15(n309), a mutant in which LIN-3 is expressed ectopically in the major hypodermal syncytium, hyp7 (Cui et al., 2006), on the same side of the pseudocoelomic cavity as the VPCs. Using the endogenous transcriptional reporter lin-3(ar656), we confirmed that lin-15(n309) leads to ectopic lin-3 transcription in hyp7 in dauer (Fig. S3F). We then found that lin-15(n309) did not result in ectopic activation of ERK in P4.p in dauer or significantly increase activity in P6.p (Fig. 3F), suggesting that a basement membrane barrier is not the likely reason that VPCs fail to respond to EGF produced from the AC in dauer.

VPCs are polarized epithelial cells. LET-23/EGFR trafficking and localization are regulated during continuous development, and altering these processes affects signaling strength. Importantly, the default localization of the receptor is the apical membrane, and its trafficking to the basolateral membrane, facing the EGF-producing AC, is crucial for signaling (Kaech et al., 1998). Therefore, we examined the expression and localization of LET-23 in P6.p in dauer larvae to determine whether a change in the overt level or subcellular localization of the receptor may underlie the downregulation of ERK activity. We note that expression of DLG-1::RFP, which localizes to the C. elegans apical junction (CeAJ), appears similar in both L2d and dauer (Fig. S6) (Diogon et al., 2007), suggesting that the cells maintain their intrinsic apical-basolateral polarity.

Endogenously tagged LET-23 is difficult to visualize in VPCs in live worms (Gauthier and Rocheleau, 2021a). We therefore used zhIs035, which overexpresses LET-23::GFP and has been used to examine trafficking of LET-23 in several studies and recapitulates the pattern seen using antibody staining of fixed worms; thus, the degree of LET-23::GFP overexpression does not appear to interfere with normal trafficking while facilitating its detection in live worms (Haag et al., 2014; Skorobogata et al., 2014). In continuous development, basolateral versus apical trafficking of LET-23::GFP in the L2 and L3 stages has been quantified as a ratio of basal/apical peak fluorescence intensity (B/A) (Gauthier and Rocheleau, 2021a,b).

Adapting this method, we examined LET-23::GFP distribution in P6.p during continuous (L2, the L2-L3 molt, and L3) and dauer (L2d, the L2d-dauer molt, and dauer) development. In continuous development, our average B/A values were consistent with the measurements of Gauthier and Rocheleau (2021a) for the L2 and L3 stages. We found a decrease in average B/A LET-23::GFP from L2 to L3 (Fig. 4A-D). In dauer development, the B/A ratio in L2d was very similar to the B/A ratio of continuous L2, and there was a decrease in the B/A ratio during the intervening L2d-dauer molt. However, in dauer larvae, the average B/A LET-23::GFP ratio increased to a value comparable to L2d. We interpret these results as suggesting that a persistent, lower B/A ratio is associated with VPC developmental progression. Furthermore, we speculate that the return to a higher B/A ratio in dauer reflects a return to an ‘L2-like’ multipotent state, consistent with a reprogramming event. These results indicate that simple loss of LET-23 at the basolateral membrane is not responsible for decreased sensitivity to EGF ligand and downregulation of ERK activity.

To gain insight into the mechanism of ERK activity downregulation in dauer, we tested whether mutations that result in strong constitutive activity of EGFR, Ras or Raf in continuous development can increase ERK activity in dauer VPCs. We reasoned that constitutively activating a component at or downstream of the step at which ERK activity is downregulated should increase ERK activity in dauer.

In continuous development, let-23(sa62) results in ligand-independent, constitutive activity of LET-23/EGFR, termed ‘EGFR(Act)’ here. This mutation results in strong constitutive activity in continuous development, even in the absence of the gonad (Katz et al., 1996). Although the biochemical mechanism by which this mutation causes constitutive activity is not clear, the critical observation for this study is that, in contrast to continuous development, we did not observe increased ERK activity in P6.p in EGFR(Act) dauers compared with ‘wild-type’ control dauers, implying that EGFR(Act) constitutive activity is opposed in dauer (Fig. 5A). In contrast, a mutation that constitutively activates Ras, let-60(n1046) [Ras(Act)], and a transgene that expresses constitutively active Raf in the VPCs [Raf(Act)] caused increased ERK activity in P6.p (Fig. 5A,B). Furthermore, Ras(Act) and Raf(Act), but not EGFR(Act), also caused ectopic ERK activation in P4.p (Fig. 5A,B). In summary, this analysis indicates that signal transduction is opposed (1) downstream of the step in LET-23 activation that is mimicked by the sa62 mutation, and (2) at or upstream of Ras activation (Fig. 5C).

Importantly, all genetic manipulations to increase signaling tested here cause highly penetrant Muv phenotypes in continuous development, suggesting that they have comparable effects on pathway activity, and that the different response of VPCs in dauer to activated forms used here is not simply attributable to the intrinsic strength of the alleles or transgenes used. We examined three different conditions that increase EGF signaling: (1) the LIN-3(S) and LIN-3(EGF) flexon transgenes and lin-15(n309); (2) an allele that constitutively activates LET-23; and (3) two different constitutively activated signal transduction cascade components, the Ras(Act) mutation and the Raf(Act) transgene. We found that there is a clean divide: elevating ligand or receptor activity does not activate ERK, whereas elevating Ras or Raf activity does. Together, our analysis implies that dauer-specific downregulation of signaling operates at or upstream of Ras activation and downstream of the step of LET-23 activation mechanism that is elevated by the let-23(sa62) mutant. Unfortunately, there are no strongly constitutively active mutants for two conserved pathway components known to link LET-23/EGFR to LET-60/Ras during signal transduction, SEM-5/Grb2 and SOS-1/Sos, so we could not resolve the step at which attenuated signaling occurs further (see also Fig. S5A,B).

We also directly compared ERK activity of lin-15(n309), let-60(n1046) and a ‘wild-type’ control strain in L2d, a stage that should reveal the intrinsic signaling activity of the mutations in animals that are still developmentally progressing but are destined to be dauer larvae. All three strains had indistinguishable levels of ERK activity in P6.p during L2d, suggesting that the maximum activation of the pathway in P6.p had been reached (Fig. 5D). Furthermore, P4.p, the cell that does not normally receive EGF signal, in L2d, had higher ERK activity in lin-15(n309) than in let-60(n1046) mutants (Fig. 5D), suggesting that lin-15(n309) is better able to activate ERK activity in a neutral VPC. However, in dauer, ERK activity in both P4.p and P6.p returned to control levels in lin-15(n309) but ERK activity remained increased in let-60(n1046) (Figs 3F, 5A, Fig. S4). Therefore, the activated position within the signal transduction cascade appears to be more relevant than the ‘strength’ of the allele.

We also note that we infer that no core component is completely missing, because low ERK activity persists in P6.p (Fig. 2F). However, we were unable to increase ERK activity in P6.p or cause ectopic activation in P4.p (Fig. S5C) by removing negative regulators of EGFR-Ras-ERK signaling that can suppress hypomorphic alleles of pathway components in continuous development, suggesting that the residual signaling in dauer may be lower than in such hypomorphs or that paralogous regulators or other genetic circuits operate in dauer.

Use of the ERK-nKTR to assess ERK activity during recovery from the dauer state has afforded insights into the natural reprogramming-like mechanism that restores multipotency and maintains competence in the dauer state. We described above how ERK activity in L2d parallels the L2, with high ERK activity in P6.p. Differences emerge in the L2d-dauer molt and in dauer itself, when ERK activity decreases and remains low in P6.p, compared with L3, when activity continues to be high. When we examined normalized ERK activity in natural dauers during recovery from the dauer state, we observed that normalized ERK activity was significantly increased in P6.p by ∼6 h into dauer recovery at 25°C (Fig. 6A). In addition, ERK activation occurred before the VPCs divided, mimicking the sequence of events seen in continuous development and consistent with the reinduction of 1° fate after a natural reprogramming-like event restores VPCs to their initial multipotent state (Euling and Ambros, 1996; Karp and Greenwald, 2013).

VPC competence refers to the mechanism that prevents VPCs from fusing to the hyp7 syncytium in the L2 stage, unlike their anterior and posterior Pn.p lineal homologs (reviewed by Sternberg, 2005). In wild type, there is natural variability in the competence of P3.p during continuous development: P3.p fuses to the hypodermis ∼50% of the time in the L2 stage (Delattre and Félix, 2001; Eisenmann et al., 1998; Myers and Greenwald, 2007). Previous work has shown that the division frequency of P3.p is sensitive to environmental conditions (Braendle and Félix, 2008). Inherent properties of ERK-nKTR make it a useful fusion marker: in competent, unfused VPCs, mClover is visible in the cytoplasm of the cell and mCherry-H2B is expressed in the nucleus. When a VPC loses competence and fuses to hyp7, mClover in the cytoplasm disappears as it gets diluted in the syncytium, whereas mCherry-H2B persists in the nucleus (Fig. 6B). Over the course of our analysis using ERK-nKTR, we incidentally observed that P3.p was generally maintained unfused throughout dauer diapause in every genotype scored (Fig. S7). We first ascertained that P3.p still fused without dividing in 50% of worms in continuous development (n=5/10). Therefore, the suppression of P3.p fusion we observed in dauer life history cannot be explained by a difference in genetic background.

During L2, lin-39 expression in the VPCs, promoted by both Wnt and EGFR signaling, maintains VPC competence (Eisenmann et al., 1998; Myers and Greenwald, 2007). Thus, the observation that P3.p always remains competent in dauer life history raises the question of whether dauer-destined larvae have increased Wnt signaling to help maintain VPC competence, as we already know that EGFR signaling is decreased. To test this possibility, we examined P3.p fusion in a mutant lacking cwn-1, which encodes a ligand of the Wnt family that is essential to maintain P3.p competence in continuous development (Gleason et al., 2006; Myers and Greenwald, 2007). Surprisingly, we found that cwn-1 is not required to maintain VPC competence in dauer life history: in 48/49 cwn-1(0) daf-7(ts) 1- and 2-day dauers, P3.p remained unfused. This observation suggests that there may be another Wnt that maintains competence in dauer larvae or an alternative mechanism to maintain competence.

When natural dauers made by starvation and crowding were induced to recover, we found that P3.p fused directly with hyp7 in about half of the post-dauer larvae (n=10/24). Therefore, P3.p, which remains competent throughout all of L2d, the L2d-dauer molt, and dauer, behaves as it does in the L2 stage of continuous development, suggesting that reprogramming to multipotency does not cause a permanent change in P3.p competence. Surprisingly, though, we found that cwn-1 is not required to prevent fusion during recovery from dauer (Fig. 6C), indicating that passage through dauer may produce a lasting genetic circuitry change.

Using a genetically encoded biosensor for ERK activity, ERK-nKTR, we found that the EGFR-Ras-ERK pathway activity is regulated in VPCs during dauer life history. In L2d, signaling is similar to that in continuous L2, but signaling is downregulated in P6.p during dauer entry, and remains downregulated for the duration of the dauer state. We also found that during dauer the VPCs become desensitized to EGF: excess LIN-3/EGF production from the gonad or ectopic cellular sources is unable to activate signal transduction in VPCs. Similarly, a missense mutation in LET-23/EGFR that causes strong constitutive signaling in continuous development cannot activate signaling in dauer VPCs. In contrast, mutations that cause strong constitutive activity of LET-60/Ras or LIN-45/Raf in continuous development are able to activate signal transduction in dauer. Together, our results indicate that the response to EGF is attenuated at or upstream of Ras activation, and is likely to involve modulation of receptor activity.

Studies of mammalian EGFR suggest mechanisms that operate at the level of the receptor or cell membrane that could be engaged in dauer VPCs. Two major mechanisms for regulating mammalian orthologs of EGFR are differential stabilization of receptor dimerization (Freed et al., 2017) and increased endocytic trafficking of the receptor to degradative compartments (Sigismund et al., 2008). However, these mechanisms seem unlikely to be relevant to the downregulation of LET-23 signaling we have described here because, unlike its mammalian orthologs, LET-23 is thought to be a constitutive dimer that responds to a single ligand (Freed et al., 2015), and LET-23::GFP is present at the basolateral membrane in dauer larvae. However, three other regulatory mechanisms described in mammalian cells may apply, and potentially could operate alone or redundantly in dauer larvae.

First, in some contexts, receptor tyrosine kinases, including EGFR, have been shown to continue signaling after internalization and downstream core signaling components colocalize in endosomes (Miaczynska and Bar-Sagi, 2010; Platta and Stenmark, 2011). These ‘signaling endosomes’ may amplify signals or act as RTK signaling platforms (Miaczynska, 2013; von Zastrow and Sorkin, 2007). A recent C. elegans study suggested that suppression of certain constitutive endocytosis pathways is a general feature of dauer larvae (Richardson et al., 2019), making it plausible that decreased endocytosis may reduce the ability of a cell to sustain high ERK activity.

Second, parallels between dauer VPCs and aged senescent human diploid fibroblasts (HDFs) suggest that, in dauer, EGFR on the basolateral membrane could be in a microenvironment that is not permissive for signaling. Senescent HDFs are insensitive to EGF (Park et al., 2000), even though the levels of both EGFR and ERK are similar to the levels in young, EGF-responsive cells. Caveolae may play a role in reduced signal transduction, as the level of caveolin is elevated in aged cells, and ectopic expression of caveolin 1 in young HDFs is sufficient to attenuate EGFR signaling (Park et al., 2000). It is not yet clear whether C. elegans has caveolae, although C. elegans has two caveolin genes, and their products are localized in membrane-associated puncta (Parker et al., 2009; Sato et al., 2006; Sloan and Bembenek, 2020). However, null alleles of caveolin genes do not cause overt vulval defects, so at this time it is an open question as to whether there may be modulatory effect on signal reception, and it is further possible that compensatory mechanisms are induced in null mutant backgrounds.

Finally, the extensive metabolic changes that occur in dauer larvae (Baugh and Hu, 2020) may result in membrane lipid composition unfavorable for signaling, as lipid composition and crowding of the membrane are thought to contribute to Ras nanocluster formation, which concentrates Ras and its downstream effectors to promote efficient signaling (Nussinov et al., 2021; Zhou et al., 2018).

When dauer larvae recover, they enter post-dauer L3 (PD L3) without an intervening molt; they subsequently molt into post-dauer L4. The lack of a dauer-PD L3 molt has contributed to the view that dauer is an ‘alternative L3 stage’, but there are many specializations of dauer anatomy and physiology that make dauer distinct, and the reprogramming-like event was the first indication that there are special mechanisms for preserving quiescence and multipotency of VPCs in dauer (Euling and Ambros, 1996; Karp and Greenwald, 2013). We found that dauer VPCs share more similarities with L2 VPCs than with L3 VPCs, but also have unique characteristics consistent with a unique dauer-stage identity.

We have identified two L2-like features here. First, the distribution of EGFR (LET-23::GFP) in dauer is more L2-like than L3-like. Second, increasing ERK activity by bypassing opposition to signal transduction does not cause cell cycle progression, which parallels the ability of VPCs to sustain high ERK activity in L2 without dividing (de la Cova et al., 2017). In addition, a previous observation about the transcription factor LIN-1/ELK1 may also be viewed as an L2-like feature. LIN-1 is phosphorylated by MPK-1 in response to the inductive signal (Jacobs et al., 1998, 1999; Miley et al., 2004); during continuous development, it is required during the L3 stage, but not the L2, stage, for repression of lag-2 in VPCs other than P6.p (Underwood et al., 2017; Zhang and Greenwald, 2011). However, in dauer, LIN-1 is no longer required for repression of lag-2 in VPCs (Karp and Greenwald, 2013), so the lack of a requirement for LIN-1 in dauer may also reflect VPC reprogramming to an ‘L2-like’ state.

The preservation or restoration of L2-like features after the L2-dauer molt is consistent with a reprogramming event that restores multipotency, a phenomenon that is unique to dauer larvae and is itself evidence for a unique state different from L2 or L3. However, we have identified additional features that suggest that VPCs have a uniquely dauer-stage identity that do not have L2 or L3 counterparts. Our major focus was on regulation of EGFR-Ras-ERK signaling, and we found that, unlike L2, L2d or L3, ERK activity in dauer remains low in dauer even though the EGF signal is present. We also found that, despite the presence of EGFR on the basolateral membrane, the VPCs are desensitized to EGF ligands and to mutational activation of LET-23 that causes a constitutive response in continuous development. Furthermore, we found that there is a difference in the competence of VPCs, as P3.p never fuses with hyp7 in dauer as it would it in about 50% of continuous L2 hermaphrodites, and that CWN-1, a Wnt ligand that is absolutely required for P3.p (and P4.p) to remain competent in L2, is not required to maintain competence in dauer. A better understanding of the dauer state may reveal how quiescence and multipotency are maintained over long periods of time in other animals, and may provide insights into basic features of stem cell biology.

A full list of the C. elegans strains used in this work can be found in Table S1. Transgenes used can be found in Table S2. Plasmids can be found in Table S3. Statistical tests and P-values can be found in Table S4. Some strains and transgenes were provided by the Caenorhabditis Genetics Center (CGC) at the University of Minnesota. The N2 Bristol Caenorhabditis elegans strain was used as the starting strain for all crosses and injections. Strains were maintained on Escherichia coli OP50; all strains were maintained at 15°C (due to temperature-sensitive dauer-constitutive mutations) and experiments were performed at 25°C as described below.

The following previously published alleles and single-copy insertion transgenes were used in this study: daf-7(e1372ts) and daf-2(e1370ts) cause constitutive dauer formation at 25°C (Karp, 2018; Larsen et al., 1995; Vowels and Thomas, 1992); hlh-2(ar614) has a small deletion ∼5 kb upstream of the coding region that removes hlh-2 expression specifically from the proximal gonad (Attner et al., 2019) such that an AC is not specified [additionally, hlh-2 is required for expression of lin-3 in the AC (Hwang and Sternberg, 2004); therefore, this allele effectively acts as a null allele of lin-3/EGF for vulval induction (Fig. S3C)]; hlh-2(ar623) was generated by CRISPR/Cas9-mediated genome engineering and encodes GFP::HLH-2 (Attner et al., 2019); let-23(sa62) [EGFR(Act)] encodes a constitutively activated receptor as a result of a point mutation in the extracellular cysteine-rich domain, near the major ligand-binding domain (Katz et al., 1996); let-60(n1046) [Ras(Act)] encodes a constitutively activated Ras protein caused by point mutation that may favor maintenance of the Ras-GTP active state (Beitel et al., 1990); lin-15(n309) results in ectopic expression of lin-3 from the major hypodermal syncytium, hyp7, and other cells (Cui et al., 2006; Fig. S3F); the negative regulators of EGFR-Ras-ERK signaling tested were gap-1(ga133) (Hajnal et al., 1997), dep-1(zh34) (Berset et al., 2005), lip-1(ok154) (Das et al., 2022) and sli-1(sy143) (Yoon et al., 1995), which are all predicted to be null or strong loss of function alleles; sos-1(pd10) is a missense mutation isolated as a suppressor of sem-5(n1619) lethality, but has no phenotype on its own (described by Pu et al., 2017); daf-16(mgDf50) is a large deletion and is predicted to be a molecular null allele (Ogg et al., 1997); cwn-1(ok546) (C. elegans Deletion Mutant Consortium, 2012) contains a large deletion and is predicted to be a molecular null allele (Zinovyeva and Forrester, 2005); arTi31 [Raf(Act)] is a single-copy insertion transgene that expresses an activated and stabilized form of LIN-45 in the VPCs [lin-31p::lin-45(T432A, S436A, V627E)::unc-54 3′UTR] (de la Cova et al., 2017); we used two ERK-nKTR single-copy insertion transgenes, arTi87 and arTi85, that express the ERK-nKTR biosensor in the VPCs (de la Cova et al., 2017, 2020) and are identical except for their insertion site (wild-type controls for a given experiment always carried the same ERK-nKTR transgene as mutants).

In this work, we looked at hermaphrodite dauers formed using mutations that result in constitutive dauer formation or at dauers isolated from starved and crowded plates (‘natural dauers’). Unless otherwise noted, all dauers were isolated by 1% SDS selection before scoring, the most stringent test of dauer formation, and assessed for dauer characteristics, such as radial constriction and alae formation.

Strains were maintained at 15°C, the permissive temperature for dauer-constitutive mutations. To score dauer larvae in daf-7(e1372ts) strains, worms and eggs were washed off continuously growing plates and treated with sodium hydroxide and bleach to isolate eggs. Eggs were plated on an OP50-seeded plate and shifted to 25°C to induce dauer formation. Unless otherwise noted, dauers were scored ∼96 h after egg preparation, when the animals had been in dauer for approximately 2 days (‘2-day dauers’). For scoring daf-2(e1370ts), which develop more slowly than daf-7(e1372ts), we scored ∼120 h after egg preparation to ensure worms had been in dauer for at least 2 days. In order to obtain natural dauers, ten gravid hermaphrodites were picked onto an OP50-seeded plate and the plate was incubated at 25°C. By ∼10 days, the food had been consumed and selection with 1% SDS yielded many dauers.

A modified protocol was necessary to isolate daf-7(e1372ts); let-23(sa62) and lin-3(ar656); lin-15(n309) dauer larvae because the strains were sick and dauers could not be obtained using the above protocol. The strains were maintained at 15°C until crowded and then the crowded plate was shifted to 25°C. Three days later, we performed SDS selection and scored the resulting dauers. We verified that this modified protocol did not have unexpected effects on ERK activity: we assessed daf-7(e1372ts); arTi87(ERK-nKTR) and daf-7(e1372ts); let-60(n1046); arTi87(ERK-nKTR) dauers using this protocol and obtained the same results as in our usual dauer isolation protocol (Fig. 5A, Fig. S3).

When scoring ERK-nKTR in dauer stages, all mutant or natural dauer larvae were scored within 2 h of 1% SDS selection. Because dauer larvae are very sensitive to the environment, for all experiments in which ERK activity in P6.p was to be compared directly, the control strain was set up and stored in parallel with the experimental strain and scoring was performed back-to-back.

In order to score ERK-nKTR in L2, L2d and the L2d-Dauer molt, tighter synchronization was used. Isolation of eggs was followed by picking worms that hatched within 2-4 h increments. L2 larvae were scored ∼15 h post hatching. L2d larvae were scored ∼20 h post hatching and L2d-Dauer molt larvae were scored ∼30 h post hatching (Karp, 2018). Morphological characteristics were also used to confirm stages such as gonad size, the presence of a buccal plug in the L2d-dauer molt, and a lack of radial constriction. Animals carrying the daf-7(ts) mutation still enter a prolonged L2/L2d stage and can sometimes enter dauer at the lower permissive temperature (Swanson and Riddle, 1981; reviewed by Karp, 2018). Therefore, only dauer life history is studied in worms containing the daf-7(ts) mutation.

To score recovering dauers, we performed SDS selection to isolate natural dauers, plated them onto fresh OP50-seeded plates, and allowed recovery to occur at 25°C. These animals appeared to recover synchronously and were then scored at designated time points. Post-dauer (PD) L4 worms were scored to determine whether P3.p had fused without dividing. To obtain cwn-1(0); daf-7(ts) PD L4 hermaphrodites, we performed SDS selection and plated dauers onto fresh OP50-seeded plates and allowed recovery to occur at 25°C. The SDS wash and fresh plating was sufficient to induce recovery, even at 25°C, although recovery was somewhat asynchronous. We picked PD L4 worms based on morphology and scored whether P3.p had fused without dividing.

To score LET-23::GFP in continuous and dauer life histories, we isolated eggs as described above and scored the population of larvae over time. In continuous development, L2-L3 molt worms were identified by picking non-pumping worms in lethargus. Constitutive 2-day dauers in this assay were not SDS-selected.

Worms were immobilized in 10 mM levamisole on 3-4% agarose pads, unless otherwise noted. All fluorescence imaging was carried out using a Zeiss AxioObserver Z1 inverted spinning disk confocal microscope (Carl Zeiss) with a 63× oil objective.

ERK-nKTR was imaged using a dual camera setup such that red and green fluorescence were imaged simultaneously at the same exposure though a z-stack of 26 slices using 488 nm and 561 nm lasers. For comparison between stages, imaging parameters were modified between stages owing to differences in transgene brightness over time and data imaged at different parameters were normalized (see description of quantification below). For comparison at the same stage, imaging parameters were the same between control and experimental animals being imaged on the same day. For every ERK-nKTR experiment, we also captured ‘blank’ images for flat fielding using the same parameters. In general, dauer larvae were on their backs and ventral views were imaged.

Strains that did not contain ERK-nKTR were imaged using a single camera set up using either 488 nm or 561 nm lasers or both sequentially.

To image the Muv phenotype using differential interference contrast microscopy (Fig. 3D), we used a Zeiss Axio Imager Z1 microscope with a Hamamatsu Orca-ER camera and a 40× oil objective.

Quantification of ERK-nKTR was carried out as described by de la Cova et al. (2017). However, we found that the baseline of the red/green ratio increases over time in dauer life history, and, as a consequence, the pixel by pixel red/green ratio calculated and scaled as described by de la Cova et al. (2017, 2020) did not give sufficient dynamic range. Therefore, we have instead reported upper quartile intensity (UQI) Nuclear Green/ UQI Nuclear Red, unscaled, which gives us greater dynamic range. Here, this is simply called ‘red/green’.

To compare between developmental stages, the red/green value in P6.p was normalized to the red/green value for P4.p, a cell that does not normally have mpk-1 activity, within each animal to look for differences in the pattern of ERK activity (de la Cova et al., 2017). Within the same stage, red/green values for a given VPC (P4.p or P6.p) were directly compared between mutants and controls. We excluded animals for which we were unable to use our pipeline to segment both P4.p or P6.p, for example because of worm movement.

To quantify LET-23::GFP protein localization, we adapted the method of Gauthier and Rocheleau (2021b). In order to compare basolateral and apical membranes, we only scored animals that were laying on their side (lateral view). To increase the number of dauers on their side, we decreased the concentration of agarose in the pad to 2% for that stage. We took 20 z-stack confocal images of P6.p in strains containing zhIs035. Using ImageJ, we created maximum intensity projections for each image, drew a thick line of a consistent width – about the size of the nucleus – through the center of the cell and collected the average GFP values across that line. This resulted in two peak values per cell, one at the apical and one at the basal membrane. These two peak fluorescence values were then used to determine the peak basal/apical fluorescence, giving one data point per animal. There were two settings used for experiments, one for imaging L2, L2/L3 molt and L3 and another for imaging L2d, the L2d-dauer molt and dauer, because we found that the apical expression in L3 gets so bright that we had to decrease laser power for continuous life history experiments.

The endogenous LIN-3 transcriptional reporter was made using a self-excising drug selection cassette according to the protocol published by Dickinson et al. (2015) and using their published reagents. The ‘sl2::nls::TdTomato::nls’ insert was modeled after lag-1(ar618) (Luo et al., 2020). We used two previously unpublished plasmids, pKL142 and pKL143, made by Katherine Luo in our lab, that expressed the guide RNA and Cas9. The guide sequences were: GATCCCCGTCTCGAAGTTC and TCATTGGTTCACATGTTGC.

In order to overexpress different forms of LIN-3 from the somatic gonad, we created miniMos single-copy insertion transgenes using the protocol described by Frøkjær-Jensen et al. (2014). Two plasmids were used to generate insertions. The first plasmid, pCO66, is designed to encode a synthetic signal sequence and the EGF domain as described by Katz et al. (1995). The second plasmid, pCO67, contains cDNA encoding the lin-3 isoform ‘S’ as defined by Pu et al. (2017), and is based on the cDNA sequences from Van Buskirk and Sternberg (2007). Both plasmids contain the rps-27 promoter, a form of LIN-3, and the unc-54 3′UTR cloned into the miniMos vector pCFJ1662, which also contains a hygromycin-resistance gene. Both pCO66 and pCO67 have the LIN-3 coding sequence interrupted by a flexon stop cassette as described by Shaffer and Greenwald (2022).

Two single-copy miniMos transgenes, arTi424 (from pCO66) and arTi425 (from pCO67), were generated. The flexon prevents functional protein production in the absence of Cre recombinase. Each transgene was combined with arTi237 (somatic gonad Cre), and, after excision of the artificial exon, strong expression of functional EGF or LIN-3(S) (via the rps-27 promoter) results specifically in the somatic gonad and causes a Muv phenotype. We note that arTi424 produces an incompletely penetrant phenotype in the absence of the Cre recombinase driver, and the phenotype is enhanced by the presence of Cre recombinase.

A summary of statistical tests and P-values can be found in Table S4. All plots and statistical analysis were performed in GraphPad Prism. Some of our data was not normally distributed so we performed non-parametric tests on all data. To compare normalized ERK activity across more than two stages or red/green values between more than two strains, we used a Kruskal–Wallis test with Dunn's multiple comparisons test. We also used this test to compare LET-23::GFP A/B ratio across more than two stages. To compare ERK activity (red/green) in P6.p (or P4.p) between two genotypes we used a Kolmogrov–Smirnov test. Data displayed on graph includes bars to show the median and interquartile range. Each point represents one animal. We used a Fisher's exact test to compare the P3.p fates in Fig. 6C. For experiments in which we did not normalize P6.p to P4.p, we only considered P-values <0.01 to be significant, although we noted in Table S4 all P-values <0.05 that were obtained. We chose this more stringent cutoff because there was some variability in the baseline values for each experiment when normalization was not able to be performed because the experimental conditions affecting intrinsic signaling would also affect P4.p.

We are indebted to Claire de la Cova for valuable advice on quantification of the ERK-nKTR and comments on this manuscript and to Xantha Karp for insightful discussions about dauer larvae. We also thank Barth Grant for helpful discussions about EGFR trafficking, Katherine Luo and Justin Shaffer for sharing strains and plasmids, and Alex Hajnal, Guangshuo Ou and Zhiwen Zhu for strains. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440).