Positive autoregulation of lag-1 in response to LIN-12 activation in cell fate decisions during C. elegans reproductive system development

The intracellular domain of LIN-12/Notch is essentially a membrane-tethered transcriptional activator released by proteolytic cleavage after ligand binding, and is conserved in all animals (Greenwald and Kovall, 2013). LIN-12/Notch activation is initiated when a ligand of the Delta/Serrate/LAG-2 (DSL) family binds to the extracellular domain of LIN-12/Notch and triggers two cleavage events that result in release of the LIN-12/Notch intracellular domain. This domain becomes translocated to the nucleus, where it forms a core nuclear complex with a CSL (Cbf1/Su(H)/LAG-1) family DNA binding protein and a mastermind family protein (Jeffries et al., 2002; Petcherski and Kimble, 2000; Wilson and Kovall, 2006). This core nuclear complex binds to the DNA through CSL, which recognizes a TGGGAA or YRTGRGAA motif (Bailey and Posakony, 1995; Chen et al., 2020; Christensen et al., 1996; Lai et al., 2000a,b; Lecourtois and Schweisguth, 1995; Nellesen et al., 1999). Genome-wide studies suggest that chromatin state and other transcription factors influence which genes containing consensus CSL binding sites are bona fide Notch targets in a particular tissue (Castel et al., 2013; Chan et al., 2017; Pillidge and Bray, 2019; Skalska et al., 2015). Target genes may also be sensitive to the level of activated Notch, which has been found to affect the dynamics of CSL association with binding sites (Castel et al., 2013; Krejci and Bray, 2007).

Genes that are master regulators of cell specification are often positively or negatively autoregulated, a mechanism that improves stability and control of gene circuits (Becskei and Serrano, 2000). Such positive autoregulation has been described in some Notch-mediated decisions. A positive autoregulatory loop of CSL is required for normal mechanosensation in Drosophila melanogaster through a conserved element called ASE; the initiation but not the maintenance of this autoregulation is Notch-mediated (Barolo et al., 2000). Further analysis of ASE revealed a combinatorial effect of various conserved regions to produce the final spatiotemporal pattern of the CSL protein Su(H) (Liu and Posakony, 2014). In Caenorhabditis elegans, a study of LIN-12/Notch transgenes suggested that lin-12 is positively autoregulated through a conserved binding site for LAG-1 and autoregulation is important for a cell fate decision during early gonadogenesis (Wilkinson et al., 1994). A recent report identified multiple LAG-1 ChIP-seq peaks in the lag-1 gene using whole-worm ChIP-seq of L4 larvae; however, a lag-1-positive transcriptional feedback loop was shown not to be active in the germline (Chen et al., 2020). This observation suggests that if there is meaningful binding of LAG-1 to sites in the lag-1 gene, it may be for autoregulation during somatic cell fate decisions.

Here, we show that in C. elegans, LAG-1 positively autoregulates in response to LIN-12/Notch activation in three different somatic cell fate decisions important for reproductive system development. We further identify an enhancer region containing multiple binding sites for LAG-1 that is necessary and sufficient to mediate autoregulation in these contexts. The binding sites that mediate autoregulation are embedded in a conserved high occupancy target (HOT) region, a stretch of open chromatin that is promiscuously pulled down in ChIP-seq experiments, including those performed with LAG-1 (Chen et al., 2020; Wreczycka et al., 2019). Deletion of the HOT region from endogenous lag-1 using CRISPR/Cas9 reduced the overall level of lag-1 transcription and abrogated positive autoregulation, but did not compromise viability or produce overt cell fate transformations characteristic of loss of LIN-12/Notch or LAG-1 activity. However, we found that adult hermaphrodites have temperature-sensitive defects in egg-laying and vulval eversion. We suggest that the HOT region and positive transcriptional autoregulation of lag-1 contribute to the robustness of the reproductive system to environmental perturbations.

Six vulval precursor cells (VPCs) initially have the potential to adopt one of three fates. The anchor cell (AC) of the gonad produces the LIN-3/EGF ligand, which induces vulval development. LIN-3/EGF activates a canonical EGFR-Ras-ERK cascade in P6.p, thereby specifying the primary fate and promoting the expression of ligands that activate LIN-12/Notch in the neighboring VPCs, P5.p and P7.p, specifying them to adopt the secondary fate. The remaining VPCs – P3.p, P4.p, and P8.p – adopt the tertiary fate, dividing to produce daughters that fuse to the major hypodermal syncytium. EGF signaling begins during the L2 stage, but the fates of the VPCs are not fixed until the L3 stage (Fig. 1A) (de la Cova et al., 2017; Sternberg, 2005).

This sequential signaling process also engages feedback mechanisms to ensure precise and robust spatial patterning. In primary VPCs, there are mechanisms to reinforce EGFR-Ras-ERK activity and to counter potential LIN-12 activation (Berset et al., 2005; Shaye and Greenwald, 2002; Stetak et al., 2006; Underwood et al., 2017). In secondary VPCs, activation of LIN-12/Notch leads to expression of direct transcriptional target genes that encode negative regulators of EGFR-Ras-ERK signaling (Berset et al., 2001; Yoo et al., 2004) and a microRNA that targets vav-1/Vav, a negative regulator of LIN-12/Notch signaling, so as to positively stimulate lin-12 activity (Yoo and Greenwald, 2005). Because such LIN-12 target genes are preferentially expressed or upregulated in secondary VPCs compared with other VPCs, we refer to this characteristic herein as a ‘secondary fate pattern’.

We generated two alleles of lag-1 that were endogenously tagged with fluorescent proteins using CRISPR/Cas9 (see Materials and Methods): lag-1(ar611[lag-1::gfp]) and lag-1(ar613[lag-1::mKate2]) (Fig. 1B and Fig. S1A). The position of the tag at the carboxy-terminus should capture all known isoforms. Both alleles appear to be phenotypically wild type and there is no apparent difference in expression between them. Initially, we performed a qualitative analysis of LAG-1::mKate2, which revealed a dynamic expression pattern in the VPCs during vulval induction (Fig. S1A). In order to achieve a more quantitative analysis, and potentially reveal more subtle dynamics, we quantitated LAG-1::GFP fluorescence, as GFP appeared brighter overall than mKate2. We observed that both LAG-1::GFP (Fig. 1C) and LAG-1::mKate2 (Fig. S1A) accumulated to a higher level in P5.p and P7.p compared with other VPCs (secondary fate pattern). This accumulation pattern is consistent with LIN-12 activation in the VPCs and suggests a positive feedback mechanism.

To test the possibility of positive feedback, we manipulated LIN-12 activity in the VPCs and observed the effect on LAG-1::GFP accumulation. Our results support the interpretation that LIN-12 activity increases the expression and/or stability of LAG-1::GFP in VPCs. First, we used the transgene arTi102[lin-31p::lin-12(intraΔP)] to ectopically express a constitutively active and stable nuclear form of LIN-12, LIN-12(intraΔP), which is known to activate lin-12 target gene reporters in all VPCs (Deng and Greenwald, 2016; Underwood et al., 2017). The presence of constitutively active LIN-12 resulted in the loss of the secondary fate pattern. The level of LAG-1::GFP in non-secondary VPCs (P4.p, P6.p and P8.p) was increased to be comparable with the level achieved normally in the secondary VPCs (P5.p and P7.p) (Fig. 1B,D). We note that animals carrying this transgene produce a functional vulva, indicating that P6.p adopts its normal primary fate, consistent with mechanisms that resist LIN-12/Notch when EGFR is active (Underwood et al., 2017) and suggesting that the increase in LAG-1::GFP is a consequence of LIN-12 activation as opposed to an overt cell fate transformation.

We then reduced lin-12 activity using two different approaches: the null allele lin-12(n941) to eliminate lin-12 activity, and genetic ablation of the AC using the proximal gonad-specific null allele hlh-2(ar614) (Attner et al., 2019) to prevent the production of LIN-12/Notch ligands by P6.p to activate LIN-12(+) in P5.p and P7.p. Both treatments resulted in loss of secondary fate pattern, and we observed uniformly low LAG-1::GFP accumulation in all VPCs (Figs. 1E-H).

We made two additional observations during this analysis. One observation pertains to the level of lag-1 expression in VPCs in which LIN-12 is not active: in a lin-12(+) background, the level of LAG-1::GFP in P6.p, where EGFR-Ras-ERK signaling is active, is similar to the level in P4.p and P8.p, where EGFR-Ras-ERK signaling is not active (de la Cova et al., 2017). This result suggests that mechanisms that counter potential LIN-12 activation in the presumptive primary VPC do not do so by negatively regulating LAG-1 accumulation. The other observation is that both methods used to eliminate lin-12 activity caused all VPCs to have a lower level of LAG-1::GFP accumulation than tertiary VPCs in a lin-12(+) background. This observation suggests that a basal level of lin-12 activity, inferred from genetic data (Greenwald and Seydoux, 1990; Greenwald et al., 1983), helps set the initial level of lag-1 expression.

To determine whether the increase of LAG-1 levels in VPCs where LIN-12 is active is a special property of VPCs, we examined two other well-studied paradigms in which LIN-12 activity specifies cell fate during reproductive system development: the developing gonad and the developing sex musculature. In both cases, we observed a higher level of LAG-1::GFP in cells in which LIN-12 is active, suggesting that the inferred positive feedback mechanism may be general in somatic reproductive system development.

In the central region of the developing somatic gonad, four cells in the L2 stage initially have the potential to be the AC or a ventral uterine (VU) precursor cell (Fig. 2A). When the somatic primordium forms, the two outer cells, called β cells, are rapidly specified as VUs and continue to express LIN-12; the two inner cells, called α cells, interact with each other via LIN-12/Notch to resolve which will be the AC and which will be another VU. The LIN-12/Notch-mediated interaction between the α cells is called the AC/VU decision. During the AC/VU decision, lin-12 and the gene encoding its ligand, lag-2, are initially expressed in all four cells; lin-12 transcription is maintained in presumptive VUs via autoregulation, and lag-2 transcription is lost in the presumptive VUs through degradation of its transcriptional activator, HLH-2 (Karp and Greenwald, 2003; Sallee and Greenwald, 2015; Wilkinson et al., 1994). Ultimately, all three specified VUs exhibit high LIN-12 activity, whereas the specified AC does not (Fig. 2A).

We marked the α and β cells of the developing somatic gonad primordium with GFP using arTi22[hlh-2(prox)p::GFP::H2B]; this marker is initially expressed in all four cells with AC potential, and then becomes progressively restricted from presumptive VUs until it is expressed only in the presumptive AC (Sallee and Greenwald, 2015). GFP expression therefore allowed us to examine the expression of LAG-1::mKate2 protein as a function of the AC/VU decision. Initially, LAG-1::mKate2 is undetectable in the α and β cells (Fig. 2B). Later in development, but still before resolution of the AC/VU decision, LAG-1::mKate2 is present at similar levels in the α and β cells (Fig. 2C). After resolution of the AC/VU decision, the level of LAG-1::mKate2 was higher in the three VUs than in the AC, suggesting that LAG-1 is preferentially expressed in cells with high LIN-12 activity (Fig. 2D). A similar expression profile was observed for the LAG-1::GFP allele (Fig. S2).

The postembryonic blast cell M is present at hatching, and undergoes a dorsal-ventral division in the L1 stage; the dorsal lineage is specified through TGF-β signaling, whereas the ventral lineage is specified by LIN-12/Notch, via a ligand made by adjacent ventral hypodermal cells (Foehr and Liu, 2008). The stages of the lineage are designated based on the number of descendants. LIN-12 protein is present both dorsal and ventral lineages, starting at the ‘4-M’ stage and continuing through the ‘18-M’ stage, but lin-12 activity is restricted to the ventral lineage because signal transduction is activated by LIN-12/Notch ligands present ventrally but not dorsally (Foehr and Liu, 2008) (Fig. 2E). lin-12 activity is required to specify the fates of the sex myoblasts (SMs), which arise in the ventral lineage at the 18-M stage, and migrate to the midbody, flanking the gonad, where later they generate egg-laying vulval muscles. Their lineal cognates in the dorsal lineage are coelomocytes (CC). We observed that LAG-1::GFP is preferentially expressed in the ventral descendants of the M-cell from 4-M to 18-M compared with the dorsal homologs (Fig. 2F).

Together with the observations in VPCs, the analysis of LAG-1::GFP expression in the developing gonad and sex muscle paradigms suggest that preferential LAG-1::GFP expression in cells with LIN-12 signaling may be a general feature of lag-1 regulation in the somatic reproductive system.

Approximately 70% of all C. elegans mRNAs are trans-spliced by the addition of a 22nt sequence, either ‘spliced leader 1’ (SL1) or ‘spliced leader 2’ (SL2). SL2 is trans-spliced to downstream genes located in naturally occurring operons in the worm genome, with the SL2 acceptor sequence located in the intergenic region between the upstream and downstream genes (Blumenthal, 2012). Synthetic operons can be generated by placing an intergenic region taken from the CEOPX036 operon between two separate coding sequences. This strategy has been used in fosmid-based contexts to make transcriptional reporters by fusing the SL2 acceptor sequence and the coding region for a fluorescent protein 3′ to the gene of interest (Tursun et al., 2009). We adapted that approach to test whether the increase in LAG-1::GFP levels observed in cells where LIN-12 is active reflected increased lag-1 transcription.

We used CRISPR/Cas9 genome editing to construct lag-1(ar618), a bicistronic lag-1::gfp::sl2acc::nls::tdTomato::nls allele, where ‘sl2acc’ represents the aforementioned intergenic trans-splicing acceptor region from CEOPX036 (see Materials and Methods for details). Two separate mRNAs are created via SL2-mediated trans-splicing: one encoding LAG-1::GFP with a heterologous 3′UTR and the other NLS::tdTomato::NLS with the endogenous lag-1 3′UTR (Fig. 3A). If the secondary pattern of LAG-1::GFP we observed in lag-1(ar611[lag-1::gfp]) reflects regulation at the transcriptional level, then we would expect to observe that both LAG-1::GFP and NLS::tdTomato::NLS exhibit the same secondary pattern in L3 VPCs. However, if regulation occurs post-transcriptionally, we would expect to see differences between the pattern of LAG-1::GFP and NLS::tdTomato::NLS accumulation, as they share transcriptional regulation but differ in their 3′UTR and protein sequences.

We found that in VPCs during the L3 stage, both LAG-1::GFP and NLS::tdTomato::NLS exhibited a secondary pattern (Fig. 3B,C), suggesting that the secondary pattern of LAG-1::GFP observed in lag-1(ar611) reflects regulation at the level of transcription.

A simple hypothesis to account for increased transcription of lag-1 in cells in which LIN-12 is active is that there is positive autoregulation of lag-1 transcription. To test this possibility, we generated a series of transcriptional reporters. The resulting transgenes were generated using CRISPR/Cas9 single-copy insertion techniques at a well-described site on LG I (see Materials and Methods) to control for positional and copy-number effects between different transgenes. The gcy-5 minimal promoter [gcy-5p(min)] provides a heterologous transcriptional start site, and was chosen because it alone does not drive expression in VPCs, but allows expression in VPCs when combined with other VPC gene regulatory elements (Zhang and Greenwald, 2011) (Fig. S4A).

The lag-1 gene has four predicted isoforms, with the a/b isoforms starting ∼9k upstream of isoform d (Fig. 4A). A transcriptional reporter corresponding to the 5′ intergenic region of the a/b isoform did not drive expression in the VPCs (Fig. S4B). We therefore analyzed lag-1 for conserved potential LAG-1 binding sites (LBSs) that would be located in the potential 5′ regulatory regions for the c or d isoforms, which also constitute introns for the a/b isoforms. We performed the search using the consensus sequences TGGGAA or YRTGRGAA, which were derived from the original electrophoretic mobility shift assay (EMSA) studies and also match the motif from a recently published LAG-1 ChIP-seq (Bailey and Posakony, 1995; Chen et al., 2020; Lai et al., 2000a,b; Lecourtois and Schweisguth, 1995; Nellesen et al., 1999). In addition, these motifs have been shown to predict functional LBSs in C. elegans VPCs (Yoo et al., 2004; Yoo and Greenwald, 2005) (see also Materials and Methods). We identified a cluster of nine LBSs within a sequence of 700-800 bp that is conserved among multiple Caenorhabditis species (Fig. S3 and Fig. 4A). The high degree of conservation suggested functional importance, and the large number of conserved LBSs made this region especially attractive to analyze as potentially mediating autoregulation.

This LBS-rich conserved region was additionally attractive based on ChIP-seq studies. First, it is embedded within a 2 kb region that was significantly (P<0.01) enriched in all available ChIP-seq tracks in C. elegans at the time we began this study, and therefore may be considered a ‘HOT region’ (Wreczycka et al., 2019). Second, Chen et al. showed that LAG-1 also binds to this HOT region, strengthening our inference based on predicted conserved LBSs that it is a potential site of autoregulation (Chen et al., 2020).

We initially analyzed a 2.7 kb region encompassing this conserved region to study the basis for the secondary pattern of lag-1 transcription (Fig. 4A). This enhancer region contains 18 LBSs and extends through to a Hox site (Roiz et al., 2016) immediately 3′ to the HOT region. A transcriptional reporter containing the intact 2.7 kb enhancer combined with gcy-5(min) produced a secondary fate pattern (Fig. 4B), faithfully reporting the expression pattern of endogenously-tagged LAG-1 and the ‘knock-in’ artificial operon.

We then performed additional analysis to identify a minimal region that is sufficient to allow for basal expression in all VPCs and upregulation in secondary VPCs. This analysis established the ‘1.6 kb enhancer’ region, which begins at the start of the conserved region and extends to the 3′ end of the 2.7 kb region (Fig. 4C). Smaller regions were not sufficient or drove weaker expression (Fig. 4G,H). We note that non-conserved LBSs, and sequences other than LBSs both within and outside of the conserved region, may influence lag-1 expression, but the evidence indicates that the conserved LBSs in the 1.6 kb enhancer region are the main elements that confer the secondary pattern.

The 1.6 kb enhancer exhibited preferential expression in VUs and in the ventral lineage of the M-cell from 4-M to early 18-M similar to LAG-1::GFP (Fig. 5A,C), suggesting it is a candidate for mediating autoregulation in response to LIN-12 activation in these cell contexts as well.

Upregulation of lag-1 transcription in cells where LIN-12 is active could reflect direct positive autoregulation via LBSs in lag-1 or be an indirect effect mediated by other transcription factors. To test the requirement for LBSs, we mutated the LBSs from TGGGAA to AGGGAA and YRTGRGAA to YRAGRGAA, mutations shown to disrupt LAG-1/CSL binding in EMSA studies (Christensen et al., 1996) and to disrupt lin-12 target gene reporter expression in vivo in C. elegans (Yoo et al., 2004; Yoo and Greenwald, 2005). Importantly, mutating the LBSs in cis in these reporters does not disrupt cell fate specification, avoiding the potentially confounding effect of cell fate transformation that would occur if LIN-12-mediated cell fate specification is disrupted by depleting lag-1 in trans.

Our mutational analysis suggests that, in this reporter context, the LBSs in the conserved region of the 2.7 kb enhancer appear to be essential for both basal expression and the secondary VPC pattern. The 2.7 kb enhancer contains 18 LBSs, nine of which reside in the conserved region (Fig. 4B). When we mutated the nine LBSs in the conserved region and left the nine LBSs outside of the conserved region intact, in addition to no longer observing a secondary pattern, we were unable to detect any expression from the 2.7 kb enhancer in the VPCs during the L3 stage (Fig. 4F). Conversely, when we left the nine LBSs in the conserved region intact and mutated the nine LBSs outside the conserved region, a stereotypical secondary pattern was detected in the L3 stage (Fig. 4E), including basal reporter expression in P4.p, P6.p, and P8.p.

The abrogation of the basal level of expression of lag-1 in VPCs when the nine LBSs in the conserved region of the 2.7 kb enhancer are mutated was also observed when we prevented LIN-12 signal transduction in VPCs in a lin-12(0) mutant or in the absence of an AC. Together, these results suggest that basal level or maintenance of lag-1 expression is achieved via an autoregulatory mechanism. The source of ligand may be nearby neurons that express lag-2 (Li and Greenwald, 2010; Takacs-Vellai et al., 2007) or a low level of LIN-3/EGF from the AC that is below the threshold for primary fate and associated mechanisms that oppose lin-12 activity (Barkoulas et al., 2013; Berset et al., 2001; Underwood et al., 2017; Yoo et al., 2004).

We next performed mutational analysis using the 1.6 kb enhancer region. When all 13 LBSs were mutated in the context of the 1.6 kb enhancer, the secondary fate pattern was compromised, whereas basal VPC expression was maintained. The level of expression in both P5.p and P7.p was reduced relative to the wild-type enhancer, with the level in P7.p reduced to the level observed in other VPCs (Fig. 4D); the residual elevation in P5.p may reflect other inputs into this enhancer that are normally masked by upregulation via LBSs. Furthermore, this mutant enhancer did not display the characteristic preferential expression in cells where LIN-12 is active in the somatic gonad and M lineages (Fig. 5B,D,E). Together, this analysis indicates that upregulation of lag-1 transcription in cells in which LIN-12 is active reflects direct positive autoregulation.

In Drosophila and in mammals, CSL proteins can act as default repressors of Notch target genes when Notch is inactive (reviewed by Lai, 2002). Such default repression was not observed in the C. elegans germline, based on the absence of overlap between genes displaying elevated expression in the absence of lag-1 and germline LAG-1 ChIP-seq peaks (Chen et al., 2020), or in VPCs, in which loss of lin-12 activity does not increase LAG-1 expression in VPCs (Fig. 1) and expression from the intact and LBSmut versions of the 1.6 kb enhancer has a similar baseline level in non-secondary VPCs (Fig. 4).

However, our analysis suggests default repression in the AC/VU context. Expression of GFP is qualitatively (Fig. 5A,B) and quantitatively (Fig. 5E) greater in the AC when driven by the LBSmut version of the 1.6 kb enhancer than by the intact 1.6 kb enhancer (Fig. 5E). In addition, the level of expression in the AC and VU appears similar when driven by the LBS mutant enhancer and there is more variability (Fig. 5E), consistent with ‘noisier’ expression and potential dysregulation resulting from loss of LAG-1-modulated transcription. Qualitatively, it appears that default repression may also be observed in the SM mother and CC lineal homologs, but we were unable to quantify expression in these cells owing to technical limitations during image processing. Overall, these results are consistent with other findings that CSL is crucial for expression of Notch-activated enhancers in certain cells, but removal of CSL also led to a broadening of expression of such enhancers into other cell types (Morel and Schweisguth, 2000).

Our reporter analysis suggested that the LBSs within the conserved HOT region were likely to be necessary for expression and positive autoregulation of lag-1 in the somatic reproductive system. We did not succeed in attempts to use CRISPR/Cas9 to make multiple specific point mutations in this region, so we instead deleted the 2 kb HOT region that encompasses the conserved region of the allele lag-1(ar611[lag-1::gfp]) (Fig. 6A). Surprisingly, the resulting mutant, lag-1(ar611ar647[lag-1(ΔHOT)::gfp]), was homozygous viable. Nevertheless, there were overt phenotypic consequences consistent with defects in development of the reproductive system: lag-1(ΔHOT)::gfp) hermaphrodites were egg-laying defective at all temperatures (n=63/66 at 25°C) and displayed a temperature-dependent abnormal vulval eversion (Evl) defect (n=121/122 at 15°C, n=36/52 at 25°C) (Fig. 6C, Fig. S5A-C). Such defects are also seen in lin-12 hypomorphs, and may have several different underlying causes, including incompletely penetrant cell fate transformations (Sundaram and Greenwald, 1993). The defects we see may therefore reflect less effective LIN-12 activity due to reduction of lag-1 expression, but it remains possible that they reflect other roles of lag-1 in later reproductive system development.

We looked for evidence of defects in the AC/VU decision and VPC secondary fate specification in lag-1(ΔHOT)::gfp hermaphrodites. We used arIs222[lag-2p::NLS::tagRFP] expression to mark the AC, and observed a single AC in the L3 stage of hermaphrodites grown at 25°C, indicating that the AC/VU decision had been made normally.

To assess secondary VPC fate specification, we first quantitated the fluorescence of LAG-1::GFP in the VPCs of lag-1(ΔHOT)::gfp hermaphrodites at 25°C, as in our analysis above. We found that LAG-1::GFP levels were significantly decreased in all VPCs, below basal levels seen for the intact locus, consistent with observations described above suggesting the basal expression in VPCs is mediated by LIN-12/Notch activation of LBSs located in the HOT region, and that there was no secondary pattern, consistent with loss of positive autoregulation (Fig. 6D,E).

The reduction in lag-1::gfp expression in lag-1(ΔHOT)::gfp hermaphrodites does not compromise the overt specification of the secondary fate. Loss or strong reduction of lin-12 activity results in a failure of lateral signaling, such that P5.p and/or P7.p adopt the primary or tertiary vulval fate instead of the secondary vulval fate (Greenwald et al., 1983; Sundaram and Greenwald, 1993). Partial failure of lateral signaling can be assessed by the ectopic expression of primary fate markers in prospective secondary VPCs even if they otherwise have secondary character (Berset et al., 2001; Yoo et al., 2004). The arIs222[lag-2p::NLS::tagRFP] marker of AC fate is also a marker of primary VPC fate, and we did not observe ectopic expression of tagRFP in P5.p and P7.p, suggesting that secondary fate specification occurred normally. A difference in expression between lag-1(ΔHOT)::gfp and lag-1::gfp was noted later: as the lineage progressed from mid to late L3, ectopic expression of arIs222[lag-2p::NLS::tagRFP] became increasingly evident: 18% of animals at the ‘Pn.px’ stage, and 45% at the ‘Pn.pxx’ stage (Fig. 6F,G). This observation suggests that execution of the secondary fate is compromised.

Abnormalities in secondary fate execution suggested by ectopic lag-2 reporter expression in VPC descendants may contribute to abnormal vulval eversion, but the penetrance of the egg-laying defect in animals grown at 25°C is higher than any vulval abnormalities observed. Defects in later diversification of the egg-laying musculature (Hale et al., 2014) as well as defects in terminal features of gonadal and vulval cell types from early or persistent deficit in lag-1 transcription, may underlie the egg-laying defect as well as contributing to abnormal vulval morphology.

This study began with the observation that the accumulation of endogenous LAG-1::GFP becomes higher in cells in which LIN-12/Notch is active than in cognate cells with low LIN-12/Notch activity in three important cell fate decisions in reproductive system development. We constructed a synthetic operon, in which the spliced leader sequence SL2 was used to drive tdTomato expression in tandem with endogenous lag-1 transcription, and demonstrated that this LAG-1::GFP upregulation was at the level of transcription. We then identified an enhancer rich in LBS sequences and, using transgenes, showed that these LBS sequences are required for upregulation in response to LIN-12/Notch activation. We further observed that this enhancer region is embedded in a HOT region identified in ChIP-seq experiments for many different transcription factors, including LAG-1 (Chen et al., 2020; Wreczycka et al., 2019).

When we deleted the entire HOT region from the endogenous lag-1 gene, we saw greatly reduced expression from the locus, as well as failure to upregulate activity in cells in which LIN-12 is active. Despite these effects on gene expression, we did not observe hallmark phenotypes associated with loss of Notch activity in C. elegans: animals were viable, indicating that embryonic cell fate decisions mediated redundantly by lin-12/Notch and glp-1/Notch were normal; animals were fertile, indicating that glp-1/Notch activity in the germline was sufficient to prevent premature entry into meiosis; cell fate marker expression indicated that the AC/VU decision and secondary VPC fate specification mediated by lin-12/Notch during reproductive system development were also normal.

However, we observed that somatic reproductive system development was affected, and that hermaphrodites had an egg-laying defect and abnormal vulval eversion. These defects are consistent with defects in later development of the reproductive system, which involves coordination and conjunction of gonad, vulva and sex muscles, each of which may be individually affected by reduction in lag-1 activity, such that the overall adult phenotype may reflect additive or cumulative effects.

The temperature-sensitivity of these defects suggests that transcriptional regulation of lag-1 may contribute to the robustness of reproductive system development. Robustness is defined as the ability of an organism to resist stochastic, environmental and genetic perturbations to its development (Felix and Wagner, 2008). This evolutionary mechanism is thought to be vital for maintaining developmental integrity while allowing for cryptic variation to give rise to increasingly fit organisms (Kienle and Sommer, 2013). VPC specification in C. elegans has been used as a model to study robustness given that divergence from a normal specification pattern is rare not only in response to environmental stimuli such as temperature, starvation and differing food sources, but also tolerance for genetic variation (Barkoulas et al., 2013; Braendle and Felix, 2008; Grimbert and Braendle, 2014; Zauner and Sommer, 2007). Indeed, the robustness of VPC patterning can withstand significant perturbation of the activity of LIN-3/EGF and LIN-12/Notch (Barkoulas et al., 2013). Notably, even when a marked effect on gene expression was observed, the range of phenotypic variation was still buffered (Braendle and Felix, 2008; (Barkoulas et al., 2013). Thus, it is plausible that the regulated level of lag-1 activity contributes to robustness to environmental effects for different lin-12/Notch-mediated events in somatic reproductive system development.

A full list of strains and genotypes used in this study can be found in Table S1. lag-1(ar611), lag-1(ar613) and lag-1(ar618) are endogenously-tagged lag-1 alleles generated by CRISPR/Cas9 for this study (see below). Strains carrying these alleles were viable and fertile, appear overtly wild type, and had no detectable cell fate transformations involving the AC/VU decision, VPC fate patterning or SM specification.

lin-12(n941) is a null allele (Greenwald et al., 1983); because homozygotes are sterile, strains containing lin-12(n941) were maintained using the rescuing extrachromosomal array arEx1442, which expresses lin-12(+) and is marked myo-3p::mCherry (Sallee et al., 2015), allowing the identification of homozygous segregants that lack the array based on lack of mCherry expression in body wall muscles.

Transgenes generated during the course of this work are described in the section ‘lag-1 enhancer analysis’ below. The following transgenes were used to mark cells of interest or to ascertain cell fate. VPCs: jccTi1[lin-31p::mCherry::H2B] (de la Cova et al., 2017) and arTi253[lin-31p::mTurquoise::H2B] [generated by miniMOS (Frokjaer-Jensen et al., 2014) for single-copy genomic insertion of pKL44] are expressed in VPCs during the L2 and L3 stages. The transgene arIs222[lag-2p::NLS::tagRFP] is expressed in primary VPCs in response to the inductive signal and therefore marks primary fate. Proximal somatic gonad: arTi112[ckb-3p:: mCherry::H2B] labels all somatic gonad cells in the L2 and early L3 stages, until the somatic gonad blast cells divide and dilute the fluorescent histone (Attner et al., 2019; Tenen and Greenwald, 2019). arIs222[lag-2p::NLS::tagRFP] is expressed in all four cells with AC potential and resolves to the AC by the mid-L3 stage (Sallee et al., 2015). arTi22[hlh-2(prox)p::GFP::H2B] was used to label the somatic gonad during L2; expression is initially equivalent in α and β cells, and gradually becomes restricted to the α cells only during the AC/VU decision and after β specification to VUs, and finally to the AC only following resolution of the AC/VU decision (Sallee et al., 2017). M-lineage: jjIs3900[hlh-8p::NLS::mCherry] marks all M-lineage descendants in the L1 and L2 stages (Shen et al., 2017).

lag-1(ar611[lag-1::gfp]) was generated through an unc-22 co-CRISPR strategy, using sgRNAs targeting both the lag-1 locus and unc-22 (Kim et al., 2014). The allele was modeled off of a lag-1 fosmid, the tag of which contains a 2×TY1-GFP-frt-2×FLAG (Sarov et al., 2012). A silent mutation was incorporated into the homology repair template to mutate the protospacer adjacent motif (PAM) site. A hygromycin resistance cassette, flanked with loxP sites, was inserted 3′ to the lag-1 3′UTR.

lag-1(ar613[lag-1::mKate2]) was generated using the self-excising cassette strategy (Dickinson et al., 2015), using a sgRNA that targeted the junction between the last exon of lag-1 and the lag-1 3′UTR, alleviating the need to mutate the PAM site.

lag-1(ar618[lag-1::gfp::sl2acc::nls::tdTomato::nls]) was made using a similar strategy to lag-1(ar613). The NLS::tdTomato::NLS sequence contains an N-terminal SV40 NLS and a C-terminal egl-13 NLS, followed by TEV-3×FLAG.

lag-1(ar611ar647[lag-1(ΔHOT)::gfp]) was achieved by injecting preformed Cas9-tracrRNA-crRNA snRNP complexes with the homology repair along pRF4 [rol-6(su1006)] (Dokshin et al., 2018) into lag-1(ar611); jccTi1[lin-31p::mCherry::H2B]. We used PCR-based screening to identify the appropriate mutant, followed by sequence confirmation of the deletion. A list of sequences targeted for editing by sgRNA or crRNA is provided in Table S2, and a list of primers used to generate the homology arms is listed in Table S3. For ttTi4348 CRISPR/Cas9-medited transgenic insertions, please see Pani and Goldstein (2018).

All larvae used for image quantification were grown at 25°C. L3 larvae were taken at 30 h after a 2 h egg-lay and mounted in 12.5 µM levamisole (Sigma-Aldrich) on agarose pads. Image stacks were collected from a Zeiss Cell Observer spinning disk confocal microscope at 0.25 µm intervals. All photomicrographs used for quantification were taken using a VPC marker, jccTi1[lin-31p::mCherry::H2B::unc-54 3′UTR] or arTi253[lin-31p::mTurquoise::H2B::unc-54 3′UTR]; the latter in the case of quantification of lag-1(ar618[lag-1::gfp::sl2acc::nls::tdTomato::nls]. It was noted that there no bleed-through of mTurquoise into the GFP channel, but there was bleed-through of GFP into the CFP channel. An image ‘blank’ was taken at each experiment for flatfielding purposes. A dual camera system (Carl Zeiss) was used to acquire GFP/YFP and mCherry simultaneously. For all images, exposure times used for GFP was 500 ms, mCherry was 150 ms, YFP was 500 ms, RFP from the NLS::tdTomato::NLS was 500 ms and mTurquoise was 1000 ms.

Images were processed by a pipeline as described in de la Cova et al. (2017) using ImageAnalysis software from the Covert Lab to identify VPCs P4.p through P8.p (de la Cova et al., 2017; Regot et al., 2014). A custom Matlab script (https://github.com/kluo91/Image-Quantification/) was used to sum the maximum five integrated intensity values of each VPC within an animal to obtain the total fluorescence intensity value. To compare the degree of patterning relative to P6.p, the summed value of each VPC was normalized to the summed value of P6.p within the same animal.

Because our sample data did not pass the Shapiro-Wilk test of normality, the non-parametric Wilcoxon Signed Rank was used to compare P5.p and P7.p with P6.p, and Mann–Whitney U-tests were used to compare VPCs from different sets of data [e.g. P4.p expression from the 1.6 kb enhancer with P4.p expression from the 1.6 kb enhancer (LBSmut)].

We analyzed LAG-1::GFP and wild-type or LBSmut transgenes formed from 1.6 kb enhancer reporters in the proximal somatic gonad and in the M lineage. To score expression in the somatic gonad, we marked gonadal cells with arTi112[ckb-3p::mCherry::H2B] and examined hermaphrodites in the late-L2 stage, after the somatic gonad primordium had formed (Kimble and Hirsh, 1979). Characteristic morphology and position (5R and 5L) in the somatic primordium was used to identify the AC or VUs, and then relative intensities was assessed qualitatively.

To score expression in cells of the M lineage, we marked cells with jjIs3900[hlh-8p::NLS::mCherry], which was kindly provided by Jun (Kelly) Liu of Cornell University, NY, USA (Shen et al., 2017). Morphology was first determined by counting the number of cells expressing NLS::mCherry to determine the stage of the M-lineage descendants. Only 4-M through 18-M (Foehr and Liu, 2008) (before anterior migration of the SM) stages were used for scoring. Preferential expression was then qualitatively assessed through the appropriate GFP or YFP channel (depending on whether the LAG-1::GFP or the enhancers were being assessed) on a spinning disk confocal microscope.

The enhancer analysis was performed by inserting enhancer transgenes into the defined site on linkage group I, ttTi4348, using reagents provided by Ariel Pani and Bob Goldstein (Pani and Goldstein, 2018). This method not only allowed for single-copy analysis but also a controlled genomic environment for quantitation.

The germline injection mixture is as described previously (de la Cova et al., 2017). The homology repair templates were generated using pWZ111, which contained the homology arms specific to ttTi4348 along with a self-excising cassette that was used to identify successful insertions. The homology repair was injected at 10 ng/µl along with the sgRNA plasmid pAP082 (65 ng/µl), and fluorescent coinjection plasmids pCFJ90[myo-2p::mCherry] (2.5 ng/µl) and pGH8[rab-3p::mCherry] (10 ng/µl). Injected hermaphrodites were incubated for 2 days at 25°C before a hygromycin solution was added to a final concentration of 250 μg/ml. Surviving F1 and F2s were then assayed for insertion of the transgene.

The 2.7 kb enhancer region was selected to encompass the greatest density of LBSs (defined as TGGGAA or YRTGRGAA, see Results for more information) in the lag-1 genomic locus. This region is located in the intron between the first and second exons of isoforms a/b, and in the 5′ region upstream of isoforms c and d. This region differs from that of a previous preliminary enhancer analysis performed using extrachromosomal arrays and a Δpes-10 minimal promoter (Choi et al., 2013) in that it contains an additional 800-900 bp to the 3′ end, which we found increases the strength of expression and was crucial for the detailed analysis described herein.

The 2.7 kb enhancer contains 18 LBSs. We performed a BLAT alignment (Kent, 2002) against eight other Caenorhabditis species: C. sinica, C. tropicalis, C. nigoni, C. remanei, C. brenneri, C. briggsae, C. angaria and C. japonica. Of these species, only C. japonica and C. angaria failed to align with the enhancer region. The regions that had aligned to the 2.7 kb enhancer were also all found to be in the 5′ intergenic region of the lag-1 homolog of the corresponding Caenorhabditis species. None of these regions was located in the intron of any lag-1 isoform, as with C. elegans, but were primarily located in promoter or intergenic regions, with the exception of C. tropicalis, the conserved region of which was located in the intron of a gene that was adjacent to its lag-1 homolog. This gene was found to have no C. elegans homologs, nor did it align (using BLASTx) to the annotated coding genes of other Caenorhabditis species. We defined the conserved region as the overlapping sequence that was common to all six Caenorhabditis species that had possessed an alignment using Clustal Omega.

We identified the HOT region by downloading the HOT region track from Wreckzycka et al. (2019), which had defined a HOT region as the 99th percentile in binding (equivalent to P<0.01). We found the track overlapped with the conserved region and was located within the larger 2.7 kb enhancer we had determined to contain the desired cluster of eighteen LBSs that were sufficient to reproduce the LIN-12-dependent expression patterns that were analyzed.

The following transgenes were created from the 2.7 kb enhancer and fused to the 200 bp gcy-5 minimal promoter to drive expression of 2×NLS::YFP using Gibson assembly cloning (Gibson, 2011): arSi31 – this transgene contained the minimal 200 bp gcy-5 promoter only; arSi35 – the full 2.7 kb enhancer was used in this transgene; arSi59 – the region 5′ to the conserved region was removed from the 2.7 kb enhancer to create the 1.6 kb enhancer; arSi60 – this region is the same as arSi59 except with all 13 LBSs mutated (TGGGAA→AGGGAA and YRTGRGAA→YRAGRGAA), thus generating the 1.6 kb enhancer (LBSmut); arSi70 – this transgene is the same as arSi59 except all the LBSs not located in the HOT region were mutated (TGGGAA→AGGGAA and YRTGRGAA→YRAGRGAA); arSi53 – the conserved region only was used in this transgene; arSi55 – the conserved region with an additional ∼200 bp on each 5′ and 3′ end was used to make this transgene; arSi74 – the full 2.7 kb enhancer was used in this transgene with only LBSs located in the non-conserved region mutated; arSi81 – the full 2.7 kb enhancer was used in this transgene with only LBSs located in the conserved region mutated.

In addition, arSi86 is a transgene comprised of a 1.4 kb gene-to-gene intergenic region from lag-1 5′ upstream neighbor cpi-1 down to the ATG of the lag-1 first exon. This transgene is followed by a 2×NLS::YFP, but does not contain the 200 bp gcy-5 minimal promoter, because the lag-1 promoter would provide its own transcriptional start site. All transgenes generated by insertion into ttTi4348 are listed in Table S4. For a list of primers used to make these transgenes, along with those of the lag-1 CRISPR alleles, please see Table S5.

We thank Jessica Chan for assistance in generating the GFP-tagged lag-1(ar611) allele. We also thank Julie Canman and Jun (Kelly) Liu for markers that facilitated this analysis, Tim Schedl and Jian Chen for communicating the results of their LAG-1 ChIP-seq study before publication, Nuria Flames and Miren Maicas Iragarai for stimulating discussions, Claire de la Cova for providing initial guidance and instruction on image quantification, Peter Sims and Julie Canman for additional guidance on image analysis, Michelle Attner for helpful discussion and comments on the manuscript, and Claudia Tenen and Claire de la Cova for additional thoughtful discussions. Some strains were provided by the CGC, which is funded by National Institutes of Health Office of Research Infrastructure Programs (P40 OD010440).