The GATA factor ELT-3 specifies endoderm in Caenorhabditis angaria in an ancestral gene network

Gene regulatory networks drive development in metazoan systems and are subject to large- and small-scale changes over evolutionary time (True and Haag, 2001; Davidson and Levine, 2008). Two important mechanisms driving changes in developmental gene networks are changes in cis-regulation (rewiring) and gene duplication followed by subfunctionalization. One example of cis-regulatory changes occurs in endomesoderm specification in echinoderms through differences in responsiveness to the T-box factor Tbrain (Hinman et al., 2007). In another example, the MADF-BESS family is specifically amplified to 16 genes in Drosophila where the derived paralogues play overlapping roles in wing hinge development (Shukla et al., 2014). In an example of both duplication and rewiring, the Drosophila anterior embryo specification factor bicoid, found only in Cyclorrhaphan flies, arose through duplication of an ancient Hox gene followed by changes in expression and loss of the role of the ancestral factor (Stauber et al., 1999, 2002). Such cases in which the outward phenotype is maintained constitute examples of what is known as developmental system drift (True and Haag, 2001).

One of the most-studied gene networks in animals is that which specifies the Caenorhabditis elegans gut cell progenitor E and promotes intestine differentiation in its descendants (Fig. 1A,B). The zygotic portion of this network consists of a cascade of structurally similar transcription factors that are found only among close relatives of C. elegans, suggesting that they may be the result of duplication and subfunctionalization (Eurmsirilerd and Maduro, 2020; Maduro, 2020). At the top of the network, the maternal SKN-1/Nrf factor, acting partially through its zygotic effectors MED-1,2, along with the maternal Wnt/β-catenin asymmetry pathway through its effector POP-1/TCF, activate early E lineage expression of the end-3 and end-1 genes (Bowerman et al., 1992; Lin et al., 1995; Rocheleau et al., 1997; Thorpe et al., 1997; Maduro et al., 2001, 2002; Shetty et al., 2005; Bhambhani et al., 2014). Downstream of these transiently expressed factors, elt-2 and its paralogue elt-7 drive gut development and differentiation, and their expression is maintained through adulthood (Fukushige et al., 1998; Sommermann et al., 2010; Dineen et al., 2018).

As might be expected for a network with structurally similar genes, the factors between SKN-1 and elt-2 demonstrate complex patterns of partial or complete redundancy. Some, like end-1 and elt-7, can be individually deleted with no apparent phenotype, whereas other single- and double-mutant combinations result in stochastic expression and many embryos lacking gut (Maduro et al., 2005a; Raj et al., 2010; Sommermann et al., 2010; Ewe et al., 2022). The most-penetrant zygotic defect results from mutants lacking both end-1 and end-3 together, which fail to specify gut 100% of the time (Zhu et al., 1997; Maduro et al., 2005a; Owraghi et al., 2010). Generally, any genotype leading to partially compromised specification leads to a loss of robustness of elt-2 activation and a failure to develop a completely normal intestine in terms of gut differentiation, gut cell number, and metabolic function, showing that the network of factors evolved to make development robust (Maduro et al., 2007, 2015; Raj et al., 2010; Ewe et al., 2022).

The zygotic endoderm genes all encode GATA factors, a family of transcription factors that bind the canonical sequence HGATAR (Lowry and Atchley, 2000; Wiesenfahrt et al., 2015; Du et al., 2016). The MED factors are a divergent subfamily, binding to a related RAGTATAC core sequence (Broitman-Maduro et al., 2005; Lowry et al., 2009). The canonical embryonic C. elegans GATA factors, including the endodermal END-1,3 and ELT-2,7 factors, have highly similar DNA-binding domains (DBDs) (Fig. 1C,D) and recognize nearly identical sequences (Weirauch et al., 2014; Wiesenfahrt et al., 2015; Du et al., 2016). In two interesting demonstrations of functional overlap, forced early endoderm expression of ELT-2, using the end-1 promoter (end-1p::ELT-2), can functionally replace the upstream function of end-3, end-1 and elt-7; furthermore, the function of all of end-1,3 and elt-2,7 can be replaced by a double-transgenic combination of end-1p::ELT-7 and elt-2p::ELT-7 (Wiesenfahrt et al., 2015; Dineen et al., 2018). The functional overlap does have limits, however, as high copy numbers of these transgenes are required for their function, suggesting there are factor-specific activities that depend on regions upstream of the DBDs (Wiesenfahrt et al., 2015; Dineen et al., 2018).

The recent availability of high-quality genome sequences for dozens of Caenorhabditis species has enabled genome-level analysis of evolution of gene families (Félix et al., 2014; Slos et al., 2017; Stevens et al., 2019). In prior work using these sequences, we found no apparent orthologues of the med, end and elt-7 genes outside of the Elegans supergroup of species, suggesting these genes evolved over a short time period at its base (Eurmsirilerd and Maduro, 2020; Maduro, 2020) (Fig. 1E). Most nematodes in the broad clade of Rhabditids that includes Caenorhabditis have only four ‘core’ embryonic GATA factors that are orthologous to factors found in C. elegans (Eurmsirilerd and Maduro, 2020). Aside from ELT-2, there are the ELT-1 and ELT-3 factors, which both function in hypodermal specification, and ELT-5 (EGL-18), which specifies hypodermal cells in the lateral seam (Page et al., 1997; Gilleard and McGhee, 2001; Koh and Rothman, 2001; Koh et al., 2002).

In this work, we examine gut specification outside of the Elegans supergroup using C. angaria (Kiontke et al., 2011; Sudhaus et al., 2011). This species has several advantages for study, including its robust growth under laboratory conditions similar to those used for C. elegans, and the fact that it has been used in comparative studies by multiple laboratories, with some examples here (Jud et al., 2007; Kuntz et al., 2008; Brauchle et al., 2009; Nuez and Félix, 2012; Barkoulas et al., 2016; Macchietto et al., 2017). RNA interference (RNAi) has shown some success in C. angaria (Nuez and Félix, 2012). Embryos of C. angaria resemble those of C. elegans and undergo a similar development in just over 11 h at 24°C with minor variations in the times at which particular milestones are reached (Macchietto et al., 2017). Using a combination of C. elegans transgenics, single-molecule fluorescence in situ hybridization (FISH) detection, genetics, and RNAi in C. angaria, we present multiple lines of evidence that gut specification in C. angaria occurs via Can-POP-1-dependent activation of Can-elt-3, and that, in turn, Can-ELT-3 activates Can-elt-2 to drive gut differentiation. The results suggest an evolutionary origin of the endoderm gene regulatory network in the Elegans supergroup from a simpler GATA factor cascade, representing an example of developmental system drift by both gene duplication and rewiring.

To facilitate identification of orthologous genes in C. angaria, we sequenced and assembled the genome of PS1010 by a combination of Nanopore long reads, Illumina short reads, and Hi-C technology to produce a six-piece chromosome-level assembly (see Materials and Methods). A Hi-C image is shown in Fig. S1. The new sequence represents an improved assembly compared with a previously published draft sequence (Table S1) (Mortazavi et al., 2010).

To elucidate a pathway for gut specification outside the Elegans supergroup, we began by testing the possibility that the C. angaria orthologues of SKN-1 and POP-1 might play a role in gut specification. The Can-skn-1 and Can-pop-1 orthologues appear to be maternally expressed, as single-embryo RNA-sequencing (RNA-seq) experiments in very early embryos recovered transcripts for these (Macchietto et al., 2017). To deplete function of these genes individually, we used RNAi. Progeny of animals fed control dsRNA displayed normal development and gut granules (Fig. 2A,B; n=123). In contrast, we observed a penetrant embryonic lethality with Can-pop-1(RNAi). After >24 h of growth of L4/adult animals on Can-pop-1 dsRNA-expressing bacteria, 90% of progeny (n=252) showed a uniform embryonic arrest at one-fold elongation with several hundred nuclei but no morphogenesis, and an absence of gut granules (Fig. 2C,D). RNAi by injection resulted in the same phenotype, although only 149/234 (64%) of progeny embryos were affected, likely because we injected only a single gonad arm per female to favour survival. The lack of gut in Can-pop-1(RNAi) was immediately striking to us, as RNAi of pop-1 in C. briggsae resulted in a similar one-fold gutless phenotype (Lin et al., 2009; Zhao et al., 2010). In those experiments, E adopted the fate of MS, which produces extra pharynx and muscle. We tested for such a transformation by looking for ectopic pharyngeal tissue using single-molecule inexpensive FISH (smiFISH) to detect expression of the C. angaria orthologue of the pharyngeal myosin gene myo-2 (Okkema and Fire, 1994; Tsanov et al., 2016; Parker et al., 2021). However, we did not find evidence of extra Can-myo-2 expression (0/20 embryos; Fig. S2).

We next attempted RNAi of Can-skn-1. Although Can-pop-1(RNAi) resulted in a highly penetrant embryonic arrest, Can-skn-1(RNAi) resulted in no apparent phenotype (n=120 progeny), using both dsRNA injection and RNAi by feeding and with two different targeting sequences. Occasionally, unusual embryos or larvae were observed in less than 5% of progeny that had various morphological or elongation defects; however, these were also observed at a similar frequency following control dsRNA injection, control RNAi by feeding, or no treatment. Because C. angaria is a male-female species, these rare embryos likely result from a reduction in developmental robustness due to inbreeding depression (Nuez and Félix, 2012). Regardless, even these rare animals contained differentiated intestine as visualized by gut granules. To control for effectiveness of Can-skn-1(RNAi), we used smiFISH. Expression of the skn-1 mRNA in C. elegans and C. angaria was detected throughout four-cell-stage embryos (Fig. S3A,B). We detected Can-skn-1 mRNA by smiFISH in 97% (n=34) of untreated embryos, but in only 6% (n=32) of RNAi-treated embryos (Fig. S3C). Hence, Can-skn-1(RNAi) treatment was effective at knocking down Can-skn-1 transcripts. We interpret the lack of Can-skn-1(RNAi) phenotype to mean that, unlike in C. elegans, the skn-1 orthologue is dispensable in C. angaria.

Orthologues of the intermediate endodermal GATA factors from the Elegans supergroup are absent in C. angaria. Hence, we next examined Can-elt-2. ELT-2 is widely conserved among nematodes (Eurmsirilerd and Maduro, 2020). Haemonchus contortus, a parasitic nematode within the Rhabditida order, encodes an apparent elt-2 orthologue that can promote gut fate when overexpressed in C. elegans (Couthier et al., 2004). We therefore predicted that Can-elt-2 drives intestinal differentiation.

We examined expression of Can-elt-2 and Cel-elt-2 using smiFISH (Fig. 3). Consistent with a similar role in intestinal differentiation downstream of specification, we detected Can-elt-2 transcripts starting at the 2E stage, after the two E daughters had moved into the interior of the embryo, and continuing in the E lineage and intestine at later stages (Fig. 3A-D). This expression is similar to that of Cel-elt-2 in C. elegans, except that Cel-elt-2 appeared to be activated slightly later, at the 4E stage (Fig. 3E-H). To confirm intestinal elt-2 expression in other species outside of the Elegans supergroup, we examined C. portoensis, a distant relative of C. angaria, and C. monodelphis, an even more distant species that is considered basal for the genus (Félix et al., 2014; Slos et al., 2017; Stevens et al., 2019). As shown in Fig. 3I-L, the elt-2 orthologues were expressed in the early E lineage and later gut in both species.

We next tested whether the entire Can-elt-2 gene, when introduced into C. elegans, is capable of intestinal expression. We expected that activation of Can-elt-2 in C. elegans would occur through endogenous ELT-2 and ELT-7 acting through autoregulatory GATA sites in the Can-elt-2 promoter (Fukushige et al., 1999; Wiesenfahrt et al., 2015). We amplified the Can-elt-2 gene with 5.0 kbp of its upstream flanking DNA, the entire coding region including introns, and 231 bp downstream of the stop codon. We inserted the coding region for GFP just before the stop codon. In a wild-type background, the Can-ELT-2::GFP transgene was indeed expressed only in intestinal nuclei in C. elegans, starting in the early embryo and continuing through adulthood, similar to expression of a Cel-elt-2 reporter (Fig. 4A-D). We regularly observed a small subnuclear spot of Can-ELT-2::GFP, which was particularly prominent in the gut of young adults (Fig. 4D; 36% of 547 gut nuclei examined in 20 worms). These were reminiscent of spots observed from autoregulatory interaction of C. elegans ELT-2::GFP protein with the elt-2 promoter DNA on a multicopy transgene array (Fukushige et al., 1999). The nuclear spots suggest, therefore, that the C. angaria elt-2 gene is capable of positive autoregulation. A smaller construct with 3.0 kbp of upstream promoter showed identical intestinal expression, and, for the assays described below, we used either transgene interchangeably.

We used smiFISH to determine more precisely when Can-ELT-2::GFP was being activated in C. elegans. As shown in Fig. 4E,F, the earliest transcripts of Can-ELT-2::GFP were detected in the nuclei of Ea and Ep after the E daughters had ingressed into the embryo, slightly earlier than when Cel-elt-2 transcripts become detectable, but later than when Cel-end-3 transcripts first appear (Raj et al., 2010; Nair et al., 2013). This timing suggested that Can-ELT-2::GFP was being activated by END-1,3. To test this, we crossed the Can-ELT-2::GFP transgene into a double-mutant end-1(ok558) end-3(ok1448) strain that is maintained by an end-3(+) array marked with unc-119::mCherry (Owraghi et al., 2010). We examined embryos in which mCherry was absent, hence are double mutant end-1 end-3, but which express unc-119::CFP, confirming the presence of the Can-ELT-2::GFP array. Of 112 embryos lacking unc-119::mCherry, all (100%) lacked Can-ELT-2::GFP expression and visible evidence of gut differentiation (Fig. 4G, Table 1). To test whether END-1 by itself was sufficient to activate Can-ELT-2::GFP, we introduced the transgene into an end-3(ok1448) single mutant. In this background, Can-ELT-2::GFP was still expressed in 93% of transgenic animals with gut (n=73). We conclude that expression of Can-ELT-2::GFP in C. elegans requires prior specification of gut by end-1,3, an unexpected result because C. angaria lacks orthologues of these genes.

The early activation of Can-ELT-2::GFP, and its possible autoregulation, suggested that Can-ELT-2::GFP could substitute for endogenous Cel-elt-2. We introduced the Can-ELT-2::GFP transgene into a C. elegans elt-2(ca15); elt-7(tm840) double-null mutant background, in which animals arrest as first-stage larvae with incompletely developed intestines (Sommermann et al., 2010). As anticipated, Can-ELT-2::GFP rescued the larval lethality of the strain to complete viability in 89% (n=123) of transgenic animals (Table 1). The ability of Can-ELT-2::GFP to rescue a Cel-elt-2; elt-7 double mutant confirms that Can-ELT-2::GFP can drive gut development in C. elegans downstream of Cel-end-1,3.

To explain the activation of Can-ELT-2::GFP in C. elegans by END-1,3, we speculated that within C. angaria endogenous Can-elt-2 is activated by another GATA factor. We examined Can-elt-1 and Can-elt-5 by smiFISH and found that these showed expression similar to their C. elegans orthologues with no early E lineage-specific signal (Fig. S4). We next considered Can-elt-3, which at first seemed an unlikely candidate. From previous work in C. elegans, elt-3 is expressed only in hypodermal cells beginning in mid-embryogenesis (Gilleard et al., 1999). ELT-3 has since been shown to be part of a gene network that drives epidermal specification (Gilleard and McGhee, 2001; Shao et al., 2013). Subsequent studies have found roles for Cel-elt-3 in oxidative stress responses and regulation of cuticle collagen genes (Budovskaya et al., 2008; Hu et al., 2017; Mesbahi et al., 2020). As shown in Fig. 5, two major isoforms are known for Cel-elt-3: a shorter ‘a’ isoform of 226 amino acids (ELT-3A) and a longer ‘b’ isoform of 317 amino acids (ELT-3B) (Li et al., 2020). We predicted a Can-elt-3 gene model that includes both orthologues, and designed probe sets for smiFISH that would allow detection of both isoforms (probe set 1) or only the longer one (probe set 2).

Analysis by smiFISH showed that Can-elt-3 exhibits both endodermal and hypodermal expression (Fig. 6A-H). Using probe set 1, faint maternal transcripts for Can-elt-3 were detected in very early embryos (Fig. 6A), appearing much weaker than maternal Can-skn-1 transcripts (compare with Fig. S3B). Much stronger signal was detectable in the E cell, just after its birth, and the early E descendants up to the 4E stage (Fig. 6B-D), and later in the embryonic hypodermis (Fig. 6E). All of these expression components were previously detected at similar stages by single-embryo RNA-seq (Macchietto et al., 2017). Transcripts were primarily cytoplasmic in most cells; however, we regularly saw one or two bright foci of nuclear staining in E, as well as Ea and Ep, likely representing nascent bursts of transcription of the Can-elt-3 gene itself (Seydoux and Fire, 1994). Can-elt-3, like Cel-elt-3, is X-linked; hence, the foci are consistent with nascent transcripts on two X chromosomes in females and one X chromosome in males. When we repeated the staining using probe set 2, we observed only early E lineage expression (Fig. 6F-H), suggesting that the longer Can-elt-3B isoform is endoderm specific.

Our earlier observation that Can-pop-1(RNAi) results in the loss of gut prompted us to determine whether Can-pop-1 acts upstream or downstream of Can-elt-3. We used smiFISH to detect Can-elt-3 transcripts in control and Can-pop-1(RNAi) embryos. We observed Can-elt-3 expression in the early E lineage in control embryos from the eight-cell to the ∼50-cell stage (100%, n=20), but expression was eliminated in 85% (n=20) of similarly staged embryos in Can-pop-1(RNAi) (Fig. 6I). The small fraction that did show staining is consistent with our prior measurement of ∼10% of embryos that were unaffected by Can-pop-1(RNAi). We also confirmed that knockdown of Can-skn-1, which did not exhibit a phenotype, also did not affect Can-elt-3 expression (14/14 embryos; Fig. 6J). We conclude that Can-pop-1 is required for Can-elt-3 expression and therefore acts upstream of gut specification, similar to the pop-1 orthologues in C. elegans and C. briggsae (Shetty et al., 2005; Lin et al., 2009; Zhao et al., 2010).

To determine whether early E lineage expression of elt-3 is likely to be broadly conserved outside the Elegans supergroup, we examined expression in C. portoensis and C. monodelphis. The former encodes a single elt-3 orthologue (see Fig. 1E), whereas C. monodelphis encodes two (Eurmsirilerd and Maduro, 2020). Cmo-elt-3.1 showed no embryonic expression (Fig. S5A,B), but we did observe early E lineage expression for Cpo-elt-3 and for Cmo-elt-3.2 (Fig. 6K-N). Both also showed later hypodermal expression (Fig. S5C,D). These results are consistent with a widespread role of elt-3 in gut specification outside of the Elegans supergroup, especially considering the basal placement of C. monodelphis in the phylogeny (Slos et al., 2017).

Finally, we examined expression of the end genes and the orthologous elt-3 gene in C. elegans. We first examined expression of Cel-end-3 and Cel-end-1 to confirm that their overlapping expression patterns (E to 2E, and 2E to 4E, respectively) resemble the expression of Can-elt-3 in C. angaria by smiFISH (E to 4E; Fig. 6O-R). We then examined Cel-elt-3 to confirm the absence of expression in the early E lineage. We did not detect signal in early embryos (Fig. 6S,T); however, we observed later expression in hypodermal lineages (Fig. 6U), consistent with prior work (Gilleard et al., 1999). Taken together, the data suggest that the endodermal expression of elt-3 was lost at the base of the Elegans supergroup, but the hypodermal expression has been retained.

Because Can-pop-1(RNAi) results in a penetrant loss of Can-elt-3 expression and gut, we hypothesized that Can-elt-3 specifies gut in C. angaria. To test this directly, we performed RNAi by gonadal injection of Can-elt-3 dsRNA. Whereas control animals always developed intestine (Fig. 7A-D; n=102), Can-elt-3(RNAi) resulted in arrested embryos and larvae in 76/122 (62%) of progeny in a time window 24-48 h after injection (Fig. 7E-H). We examined these for the presence of birefringent gut granules, ‘fried-egg’ nuclei typical of gut cells, an intestinal lumen, and basement membrane surrounding the intestine. In almost all cases, these features of differentiated gut were completely absent (Fig. 7E-H,M,N). In a small number of embryos, we observed rare gut-like nuclei; however, we could not see a polarized epithelium and gut granule birefringence, and no lumen was visible. Except for these few cases, arrested embryos and larvae were strongly reminiscent of C. elegans end-1(ok558) end-3(ok1448) double-null mutants (Fig. 7I-L) (Owraghi et al., 2010). Unlike Cel-end-1,3(-) embryos, however, which show variable elongation of two to three times the length, arrested Can-elt-3(RNAi) embryos tended to be fully elongated. As well, Cel-end-1,3(-) embryos often contain internal hypodermis-lined cavities that result from the transformation of E to a C-like cell when Cel-end-1 and Cel-end-3 are absent (Sulston et al., 1983; Zhu et al., 1997; Maduro et al., 2005a). Such cavities were not obvious in Can-elt-3(RNAi), although we did see hypodermal defects, visible as a deformation of part of the cuticle, in 34% (n=29) of embryos (Fig. 7E, arrows). These could be the result of loss of Can-elt-3 in the hypodermis, or from defects in morphogenesis associated with loss of E specification.

We used CRISPR/Cas9 mutagenesis to generate a deletion of Can-elt-3 in strain PS1010 using a protocol optimized for C. elegans (Ghanta and Mello, 2020). We obtained a mutant, ir79, that deletes 2916 bp of Can-elt-3 and is a putative null (shown on the gene model in Fig. 5A). Because Can-elt-3 is X-linked, the mutant is maintained through heterozygous females; mating with males will produce one out of four hemizygous ir79 progeny. Of 140 progeny of Can-elt-3(ir79)/+ females crossed to wild-type males, 32 embryos arrested without gut (23%; P=0.6 with expected 25%) and resembled Can-elt-3(RNAi) embryos (Fig. 7O,P). These results are consistent with a fully penetrant embryonic lethality of Can-elt-3(ir79), confirming that Can-elt-3 is zygotically required for gut specification. In addition to the endoderm defect, 72% of gutless animals (n=25) showed a hypodermal defect, suggesting that ir79 mutants have a stronger phenotype than Can-elt-3(RNAi). These results also show that Can-elt-3(RNAi) phenotypes are not the result of depletion of maternal Can-elt-3 mRNA.

The essential role of C. angaria elt-3 contrasts with the absence of developmental phenotype seen in a C. elegans elt-3 null mutant (Gilleard and McGhee, 2001). To confirm that Cel-elt-3 plays no minor role in gut specification, we combined the elt-3(gk121) null mutant with null mutants in each of end-1, end-3 and elt-7 to look for possible synergistic effects (Table S2). As expected, we found no evidence of synergy.

We next confirmed that Can-elt-2 functions similarly to Cel-elt-2 by examining Can-elt-2(RNAi) using gonadal dsRNA injection. We observed a penetrant larval lethality in 39/89 (44%) of progeny embryos examined 24-72 h after injection. In these arrested larvae, although intestine was present, we observed a variety of differentiation defects, including a partial intestinal lumen and patches of intestine lacking gut granules (Fig. 7S,T). The phenotype was highly reminiscent of the C. elegans Cel-elt-2(ca15); elt-7(tm840) double mutant (Fig. 7U,V; compare with Fig. 7Q,R control). We conclude that Can-elt-2 is required for gut differentiation in C. angaria, as expected.

Prior studies in C. elegans showed that endodermal GATA factors are individually able to promote widespread gut specification when overexpressed throughout early embryos (Fukushige et al., 1998; Zhu et al., 1998; Maduro et al., 2001, 2005a; Sommermann et al., 2010). We wished to test whether widespread expression of Can-ELT-3 within C. elegans is sufficient to do so. We constructed heat shock (hs) hs-Can-ELT-3B::CFP and hs-Can-ELT-3A::CFP transgenes to express each isoform conditionally throughout embryos. The transgenes were individually introduced into a C. elegans elt-2(ca15); elt-7(tm840); Ex[Can-ELT-2::GFP] strain. We heat-shocked mixed-stage early embryos (<100 cells) for 20 min at 34°C. In both cases, within 75-90 min, widespread nuclear CFP was observed, indicating expression of the transgene [49% (n=35) of hs-Can-ELT-3A::CFP and 61% (n=18) of hs-Can-ELT-3B::CFP]. The CFP disappeared by 3 h after heat shock. In the case of hs-Can-ELT-3A, most embryos arrested with either no gut or a small patch of gut (n=85%, n=39), with a small number showing some gut granules and Can-ELT-2::GFP-expressing nuclei that were consistent with dispersal of a normal number of gut cells (15%, n=39; Fig. 8A-C). In contrast, with hs-Can-ELT-3B we observed 37% (n=43) of embryos that exhibited one-fold arrest with widespread Can-ELT-2::GFP with >50 nuclei (Fig. 8D-F). Parallel treatment of the rescued elt-2,7 strain carrying Can-ELT-2::GFP, without a heat-shock transgene, showed 12% (n=50) embryonic arrest but no ectopic gut. These results show that overexpressed Can-ELT-3B, but not Can-ELT-3A, is sufficient to promote gut specification outside of its normal context in C. elegans.

We next tested whether expression of Can-ELT-3B in the early E lineage could specify gut in C. elegans. We constructed a Cel-end-3promoter::Can-ELT-3B::CFP fusion transgene and introduced it, along with an unc-119::mCherry marker, into a triple mutant elt-7(tm840) end-1(ok558) end-3(ok1448) strain rescued by an unc-119::YFP-marked array. We obtained several viable transmitting lines in which the original unc-119::YFP array had been replaced by the end-3::Can-ELT-3B::CFP array, confirming rescue of specification. The best line rescued 68% (n=85) of transgenic animals to complete viability and fertility (Table 1). We next tested whether the combination of end-3::Can-ELT-3B::CFP and Can-ELT-2::GFP could rescue a strain in which all of elt-7, end-1, end-3 and elt-2 had been mutated. We were able to construct such strains, either using separate arrays containing end-3::Can-ELT-3B::CFP and Can-ELT-2::GFP, or with a single array containing both transgenes (Fig. 8G,H, Table 1). With two separate arrays, 15% (n=84) of double-transgenic embryos were rescued. In the single-array strain, rescue was strongest at 25°C with 50% (n=181) of transgenic embryos rescued to full viability, whereas at 20°C rescue dropped to 28% (n=257). These striking results demonstrate the ability of the simpler C. angaria gut network to replace the core gut specification and differentiation pathway of C. elegans.

In this work, we have elucidated a core pathway for gut specification and differentiation in a species outside of the Elegans supergroup. This solves a long-standing question about gut specification in Caenorhabditis, and adds a new example of a pathway that exhibits developmental system drift (True and Haag, 2001). The simpler pathway consists of a single zygotic specification factor, ELT-3, that serves the function of the three GATA factors END-1, END-3 and ELT-7 that drive endoderm development in C. elegans (Fig. 9). Both retain the terminal regulator, ELT-2, which is functionally interchangeable across the evolutionary distance between C. angaria and C. elegans. Consistent with the essentiality of the network components, loss of Can-elt-3 by mutation resembles loss of Cel-end-1,3, and loss of Can-elt-2 resembles loss of Cel-elt-2,7. When forcibly expressed in C. elegans, Can-ELT-3B can activate either endogenous Cel-elt-2 or transgenic Can-elt-2 and drive gut development. The simpler network of C. angaria is reminiscent of gut development in Drosophila, in which two GATA factors act in a similar cascade: serpent (srp) specifies gut fate upstream of GATAe, which executes and maintains this fate (Reuter, 1994; Okumura et al., 2005).

Both pathways in C. elegans and C. angaria share at least one maternal activator, POP-1/TCF. In C. elegans, the positive contribution of POP-1 is secondary to a stronger input by SKN-1, whereas in the close relative C. briggsae maternal input from POP-1 and SKN-1 are individually essential (Maduro et al., 2005b; Shetty et al., 2005; Lin et al., 2009; Zhao et al., 2010; Bhambhani et al., 2014). As a result, the phenotype of pop-1(RNAi) was similar between C. angaria and C. briggsae, namely a failure to activate the early E lineage specification factors, Can-elt-3 in the former, and Cel-end-1,3 in the latter.

We did not observe a phenotype for Can-skn-1(RNAi), perhaps because it is redundant with another factor, or because Can-SKN-1 does not have a role in endomesoderm specification in C. angaria as it does in C. elegans. The ancestral role of SKN-1 may not be in embryonic cell specification: SKN-1 is known to be a major effector of postembryonic responses to physiological stress, and, in an unexpected convergence of function, Cel-elt-3 interacts genetically with Cel-skn-1 in regulation of genes in the oxidative stress response (Blackwell et al., 2015; Hu et al., 2017). Therefore, it may be that Can-skn-1 was recruited into endoderm specification at the base of the Elegans supergroup, perhaps concomitantly with the emergence of the MED GATA factors. Further experiments to elucidate contributions of other maternal regulators of cell fate in C. angaria, and a more detailed understanding of the fate of the E cell in both Can-pop-1(RNAi) and Can-elt-3(ir79), may shed light on how combinatorial mechanisms of cell specification work in C. angaria that could explain the lack of a Can-skn-1(RNAi) phenotype.

ELT-3 in C. elegans has been associated with hypodermal expression and function, although a null mutation has no developmental phenotype (Gilleard et al., 1999; Gilleard and McGhee, 2001; Shao et al., 2013). In the early E lineage, though not in E itself, a low level of Cel-elt-3 transcripts has been observed by single-cell transcriptomics (Hashimshony et al., 2012; Tintori et al., 2016). In this study, we failed to observe such expression in intact embryos, and, moreover, found no evidence for even a cryptic role of Cel-elt-3 in gut specification. Later expression of Cel-elt-3 in the intestine has been reported, though this has been controversial (Budovskaya et al., 2008; Tonsaker et al., 2012). Potentially, such expression could be conditional, arising in response to oxidative stress as it appears to function in the hypodermis (Budovskaya et al., 2008; Shao et al., 2013; Hu et al., 2017). Our observation of hypodermal expression of Can-elt-3 in embryos suggests that hypodermal ELT-3 function is conserved in the genus.

Overexpression of ELT-3 throughout C. elegans embryos, or in the early E lineage, was previously found to promote widespread hypodermal fates, and not endodermal fates (Fukushige et al., 1998; Gilleard and McGhee, 2001; Wiesenfahrt et al., 2015). This contrasts with our results showing that overexpression of Can-ELT-3 throughout early C. elegans embryos, or in the early E lineage, is sufficient to drive gut development. The paradox is resolved by our evidence that a longer isoform, Can-ELT-3B, is endoderm specific, both in its expression in C. angaria, and in the ability of this isoform, and not Can-ELT-3A, to activate gut expression when expressed in C. elegans. The prior studies in C. elegans used only the shorter Cel-ELT-3A isoform (Fukushige et al., 1998; Gilleard and McGhee, 2001). It will be of interest to determine in future studies whether the longer isoform of the C. elegans ELT-3 harbours a cryptic ability to activate Cel-elt-2, and how the amino-terminal regions found only in the long isoforms, which are conserved between Can-elt-3 and Cel-elt-3, might be important for this activity (Fig. 5B). In human and mouse, protein-protein interactions outside of the DBDs, and combinatorial interactions at promoters, explain differential activities of otherwise similar GATA factors (Romano and Miccio, 2020). Also, a role for vertebrate GATA3 as a pioneer factor was recently described (Tanaka et al., 2020). Hence, it is plausible that the amino-terminal portion of Can-ELT-3B is important for interaction with co-factors, or for a possible role as a pioneer factor in establishing an active transcription state of Can-elt-2 in the early embryo.

The role of ELT-3 in gut specification is likely to be ancestral. The absence of med, end and elt-7 orthologues outside of the Elegans supergroup shows that the C. elegans gut network must be derived (Eurmsirilerd and Maduro, 2020; Maduro, 2020). We found that expression of elt-3 in the early E lineage also occurs in C. portoensis and in the basal species C. monodelphis, which is further consistent with the ancestral nature of the simpler pathway. The alternative, secondary simplification of the C. elegans-type pathway by loss and consolidation of factors, seems far less likely: although some individual genes are dispensable, loss of pairs of regulators in C. elegans is inviable or nearly so, and even minor disruption to timely activation of elt-2 results in abnormal gut development and metabolism (Maduro et al., 2007; Owraghi et al., 2010; Raj et al., 2010; Choi et al., 2017; Ewe et al., 2022). Loss of the upstream MED factors that directly activate end-1,3 would also result in a failure to specify MS (Maduro et al., 2001). Hence, the C. elegans-type expanded network may be evolutionarily fixed.

There are several differences between the C. angaria-type network and the derived C. elegans one that must have occurred over a very short time span. What originated as a simpler network involving only one upstream GATA factor, ELT-3, must have been rapidly replaced by several regulators, END-1, END-3 and ELT-7. We previously suggested that the end and elt-7 genes might have originated from a duplication of elt-2, through a successive cascade of upstream duplication and intercalation into the network through temporal refinement of expression and changes in cis-regulation (Maduro, 2020). Our results here suggest that the end genes and elt-7 originated as duplications of elt-3. Of the C. elegans GATA factors, ELT-3 has features that make it more ‘endodermal’ than hypodermal. For one, it has an intron in the same position as the endodermal GATA factors, between the first two cysteines in the zinc finger (Maduro, 2020). The elt-1 and elt-5 genes lack this intron and instead have one farther downstream in the basic domain. Furthermore, the short carboxyl end of ELT-3, which terminates abruptly after the basic domain, is a feature found only among the MED and END factors and ELT-7, as ELT-1, ELT-2 and ELT-5 contain extended regions after the basic domain (Maduro, 2020). As the presumptive paralogs of ELT-3 evolved, the original ELT-3 also had to lose its endoderm specification role, while the upstream MED factors evolved as activators of the ENDs and specifiers of MS fate. The similarity of the MEDs to END-3 suggests that these arose late and were derived from duplication and divergence of an END-3-like factor (Maduro, 2020).

The expansion of an ancestral elt-3 therefore likely occurred by duplication followed by divergence/specialization. Rapid expansion of gene families is known to have occurred in C. elegans (Lipinski et al., 2011; Konrad et al., 2018). The expansion of multiple GLP-1/Notch factor paralogues at the base of the Elegans supergroup is a similar example of expansion of an ancestral factor by gene duplication within the genus (Stevens et al., 2019). In addition, amplification of F-box genes has been observed among four Elegans supergroup species, in which tandem duplication was found to be an important mechanism (Wang et al., 2021). Indeed, among the Elegans supergroup endodermal GATAs, tandem duplication has also been observed for many med genes, and there is also the likely ancestral tandem duplication that generated end-1 and end-3, which are found within ∼50 kbp of each other in many species (Maduro, 2020).

Why would a simpler network undergo expansion? Perhaps the expanded network in the Elegans supergroup resulted from evolutionary pressure to accelerate development (Maduro, 2020). Early developmental timing events are slightly accelerated in C. elegans relative to C. angaria, although later developmental milestones are similarly timed (Macchietto et al., 2017). An increased number of regulators could amplify early specification and assure rapid, robust activation of elt-2, permitting development to speed up without sacrificing robustness. The expansion of genes in the Elegans supergroup endoderm network may thus resemble the emergence of bicoid in Drosophila, in which expansion of a gene network enabled a more rapid embryonic development to occur while maintaining robustness (McGregor, 2005). One intriguing observation from this study may offer an alternative explanation. When we introduced end-3promoter::Can-ELT-3::CFP and Can-ELT-2::GFP transgenes to rescue gut development in a quadruple elt-7 end-1 end-3; elt-2 background, rescue was cold sensitive, occurring at lower efficiency at 20°C than at 25°C (Table 1). C. angaria is phoretically associated with the weevil Metamasius hemipterus, a pest of sugar cane in South Florida and the Caribbean, places with tropical climates (Sudhaus et al., 2011). An intriguing idea is that a more complex network for gut specification enabled Caenorhabditis species to maintain robust development as they spread to different locales. With a core gut network now known for the genus, comparative studies within and outside of Caenorhabditis can now begin to explore this new system for studying developmental system drift.

Animals were grown on nematode growth medium (NGM) agar seeded with Escherichia coli OP50 for 5 days. Mixed-stage worms were collected from the culture and washed three times with M9 buffer complemented with Anti/Anti (Gibco). The worms were transferred to a worm lysis solution [QIAGEN buffer G2 with 400 µg/ml proteinase K, 50 mM dithiothreitol (QIAGEN) and 0.5 mg/ml RNase A (Invitrogen)] and incubated at 55°C for 4 h. High-molecular-weight genomic DNA was spooled from ethanol precipitation following phenol-chloroform extraction and dissolved in 10 mM Tris (pH 8.0). A Nanopore library was prepared using 1 µg genomic DNA using a ligation sequencing kit (SQK-LSK109, Oxford Nanopore Technologies) according to the manufacturer's protocol. A single 48-h sequencing run was performed with MinION R9.4.1 flow cell to obtain 5.0 Gb of sequence data (444,000 reads; N50, 23 kb). The Nanopore reads were base-called to generate FASTQ files using the Guppy v4.0.15 basecaller (Oxford Nanopore Technologies) with the supplied dna_r9.4.1_450 bps_hac configuration and were quality checked using NanoPlot v1.31.0 (De Coster et al., 2018). An Illumina paired-end sequencing library was prepared from 100 ng of DNA using the Nextera DNA library prep kit according to the manufacturer's instructions. A total of 1.5 Gb of paired-end reads (95 bp×2) were generated by library sequencing on an Illumina MiSeq instrument with the MiSeq reagent kit v3 according to the manufacturer's protocol. The Hi-C library was prepared from ∼2000 fresh worms using an Arima-HiC kit (Arima Genomics) and a Collibri ES DNA library prep kit (Thermo Fisher Scientific) according to the manufacturers' protocols and was sequenced using a MiSeq instrument with the MiSeq reagent kit v3 (101 bp×2), and the 4.9 million short reads were quality checked using the Hi-C quality control pipeline (https://phasegenomics.github.io/2019/09/19/hic-alignment-and-qc.html). The Nanopore long reads were assembled using Nextdenovo v2.4.0 (https://github.com/Nextomics/NextDenovo) with the parameters genome-size=70 M and read_cutoff=5 k. After base correction by three rounds of Pilon v1.23 (Walker et al., 2014) with the Illumina paired-end reads, the assembly was further scaffolded using the 3D-DNA pipeline v180114 (Dudchenko et al., 2017) without a misjoin correction process, and the chromosome-length scaffolds were extracted via manual curation using Juicebox v1.11.08 (Durand et al., 2016). A Hi-C plot is shown in Fig. S1.

Orthologous GATA factors and other orthologous genes were identified by BLAST searches as in prior work (Lin et al., 2009; Maduro, 2020). The identity of individual GATA factors was confirmed using defining features, including location of introns in the coding region, signature amino acids within the DBDs, and reciprocal search back to the C. elegans genome (Eurmsirilerd and Maduro, 2020). For some genes, we made use of gene predictions from a previously published sequence of C. angaria (Mortazavi et al., 2010; Macchietto et al., 2017) and from its close relative C. castelli (downloaded from The Caenorhabditis Genomes Project in November, 2020) (Félix et al., 2014). Genome sequences and annotation files for other species were downloaded from The Caenorhabditis Genomes Project and WormBase ParaSite. See Supplementary Materials and Methods for further details.

Strains used were: C. angaria, PS1010 and RGD1; C. portoensis, EG4788; C. monodelphis, JU1667. C. elegans strains were constructed by standard crosses and microinjections to generate transgene arrays (Brenner, 1974; Mello et al., 1991). Mutations were: LG III: unc-119(ed4); LG IV: him-8(e1489); LG V: dpy-11(e224), unc-76(e911), elt-7(tm840), end-1(ok558), end-3(ok1448); LG X: elt-2(ca15), elt-3(gk121). Genotypes were confirmed using a combination of progeny testing and PCR with allele-specific primers. Transgene arrays were: irEx498 [end-3(+) (pMM768), unc-119::mCherry (pMM824)], irEx798 [Can-elt-2_5kbp_promoter::ELT-2genomic::GFP::elt-2_3′UTR (pGB598), unc-119::CFP (pMM809), unc-119(+) (pMM016B)], irEx804 [Cel-end-3promoter::END-3genomic::Can-ELT-3_DNA-binding domain::CFP::Cel-end-3_3′UTR (pGB612), unc-119::YFP (pMM531), unc-119(+) (pMM016B)], irEx808 [hsp16-41::Can-ELT-3(isoform_b)::CFP (pGB619), rol-6D (pRF4)], irEx809 [hsp16-41::Can-ELT-3(isoform_a)::CFP (pGB620), rol-6D (pRF4)], irEx813 [pGB608(Can-elt-2_3.5kbp_promoter::Can-ELT-2::GFP)+pGB618(Cel-end-3::Can-ELT-3B::CFP)+pMM824(unc-119::mCherry)], irEx814 [pGB618(Cel-end-3::Can-ELT-3B::CFP)+pMM824(unc-119::mCherry)]. Strains used in this work are listed in Table S3.

We constructed transgenes using Gibson assembly (Gibson et al., 2009). The coding region for GFP was amplified from pPD95.67 (a gift from Andrew Fire, Stanford University, CA, USA). Plasmids containing coding regions for the fast-folding fluorescent proteins sCFP3A and Venus/YFP (Balleza et al., 2018) were obtained from Addgene (plasmids #103970 and #103986, respectively). Partial or complete coding regions for Can-elt-3, Can-skn-1 and Can-pop-1 were synthesized by IDT. See Supplementary Materials and Methods for further details.

We used Gibson assembly to construct intronless heat-shock Can-ELT-3A::CFP and heat-shock Can-ELT-3B::CFP transgenes using the heat-shock promoter obtained from vector pPD49.83. See Supplementary Materials and Methods for further details. We injected each transgene along with the rol-6^D marker (plasmid pRF4) into the elt-2(ca15); elt-7(tm840) genetic background rescued with the Can-ELT-2::GFP transgene.

Triple mutant elt-7(tm840) end-1(ok558) end-3(ok1448) hermaphrodites rescued by irEx813 or irEx814 were crossed to males from a him-8(e1489); elt-7(tm840) end-1(ok558) end-3(ok1448) strain rescued by irEx804. Progeny males carrying irEx813 or irEx814 and lacking irEx804, recognized by expression of unc-119::mCherry and absence of unc-119::YFP, were crossed to dpy-11(e224) unc-76(e911); elt-2(ca15); irEx798 hermaphrodites. F₁ males carrying irEx813, or irEx814 and irEx798 (the latter recognizable by unc-119::CFP), were backcrossed to dpy-11 unc-76; elt-2 hermaphrodites. Non-Dpy, non-Unc progeny were allowed to self-fertilize and non-Dpy, non-Unc progeny carrying irEx813 alone, or irEx814 together with irEx798, were singled to identify animals that never segregated Dpy Unc. These were confirmed by PCR and progeny testing to be quadruple elt-7(tm840) end-1(ok558) end-3(ok1448); elt-2(ca15) and rescued by irEx813 or [irEx814+irEx798].

For RNAi experiments, genomic DNA fragments or synthesized cDNA sequences were cloned into the feeding-based RNAi vector pPD129.36. For feeding-based RNAi, we used standard protocols (Timmons and Fire, 1998). To synthesize dsRNA for injection, we used primers L4440A (gagcgcagcgagtcagtgagcg) and L4440B (cccagtcacgacgttgtaaaacg) to PCR-amplify a template for synthesis of RNA using the T7 MEGAscript kit (Thermo Fisher Scientific). dsRNA at a concentration of ∼2 μg/μl was injected into one gonad arm per female. To prevent possible cross-interference with mRNA of the other GATA factors, we targeted sequences upstream of the coding regions for the DBDs. See Supplementary Materials and Methods for further details.

Images were obtained using either a Canon EOS 77D or Canon EOS RP camera with an LMscope adapter (Micro Tech Labs) on either of two Olympus BX-51 fluorescence microscopes equipped with DIC optics. Images were processed for contrast and colour uniformly across images using Adobe Photoshop.

We used a C. elegans protocol (Ghanta and Mello, 2020) with crRNAs to target genomic sequences 5′-gtgcttgaatgcggtgagtttgg-3′ and 5′-gaatttctccaccaactacatgg-3′. All CRISPR reagents were ordered from IDT. We injected 20 females and mated them individually with five males each. One plate had arrested embryos and PCR identified a putative deletion in Can-elt-3. We singled 30 mated females and obtained three plates with one out of four dead eggs lacking endoderm. See Supplementary Materials and Methods for further details.

We used the smiFISH protocol (Tsanov et al., 2016; Calvo et al., 2021) adapted for use in Caenorhabditis by Parker et al. (2021) using our previously described fixation protocol (Broitman-Maduro and Maduro, 2011). Probes consist of complementary FLAP-X sequence (5′-CCTCCTAAGTTTCGAGCTGGACTCAGTG-3′) followed by complementary gene-specific antisense sequence of 16-24 additional bases (Tsanov et al., 2016). These were generated using the Stellaris Probe designer (Biosearch Technologies) and are listed in Supplementary Materials and Methods. Conjugated FLAP-X oligos (CACTGAGTCCAGCTCGAAACTTAGGAGG) that were 5′ and 3′ end-labelled with Quasar 570 or Cal Fluor 610 were synthesized by Biosearch Technologies. FLAP-X oligos 5′ and 3′ end-labelled with Cy5 or Cy3 were synthesized by IDT. To detect fluorescent smiFISH signals, we used filter sets obtained from Chroma: for Quasar 570, we used the Gold FISH 49304 ET set; for Cy5, the Narrow-Excitation Cy5 49009 ET set; for Cal Fluor 610, set 31002 or Red#2 FISH set 49310 ET. We imaged co-stained embryos that had both Quasar 570 and Cy5 probes in order of increasing wavelength, i.e. DAPI→Gold FISH→Cy5, to prevent imaging of photoconverted Cy5 (Cho et al., 2021). In our hands, staining was highly consistent among a set of fixed embryos, such that when signal was detected in embryos of a particular stage, signal was seen in most other embryos of that stage. Rare embryos (<5%) that did not show staining were usually visibly damaged and were more likely to be younger than the four-cell stage. One exception was detection of Can-elt-2 transcripts in the Can-ELT-2::GFP strain, in which ∼60% of embryos showed staining, consistent with the transmission frequency of the extrachromosomal array. We performed smiFISH following RNAi by feeding in C. angaria and included controls for permeabilization and staining in each case. For Can-pop-1(RNAi), we simultaneously stained for Can-elt-3 using Quasar 570, and for Can-eef1A.1, the orthologue of Cel-eef1A.1 (also known as Cel-eft-3), using Cy5. From single-embryo RNA-seq data, Can-eef1A.1 is expressed at all embryonic stages from zygote through hatching (Macchietto et al., 2017). For Can-skn-1(RNAi), we stained for Can-skn-1 using Quasar 570 and Can-elt-3 using Cy5. Because Can-skn-1(RNAi) did not result in a loss of gut specification, we reasoned that Can-elt-3 expression would be unaffected and hence this served both as confirmation of this hypothesis as well as a control for staining of Can-skn-1 transcripts following Can-skn-1(RNAi). For each probe set, we examined 30-100 embryos. See Supplementary Materials and Methods for further details.

We are indebted to Mark Blaxter, Lewis Stevens and colleagues at The Caenorhabditis Genomes Project for prepublication access to genome sequences. We thank Jordan Ward, UC Santa Cruz, for helpful suggestions; and Esmeralda Rivera and Joshua Miguel Carreon, undergraduates at UC Riverside, for performing some of the RNAi controls. Some strains were provided by the Caenorhabditis Genetics Center, which is funded by the NIH Office of Research Infrastructure Programs (P40 OD010440).