Reciprocal signaling by Wnt and Notch specifies a muscle precursor in the C. elegans embryo

The mechanism(s) by which a cell becomes committed to a specific developmental fate is one of the central questions in biology, with direct significance for stem cell biology and regenerative medicine. One key insight comes from determining whether the fate of a cell is specified through a cell-autonomous or non-autonomous mechanism. This is often determined experimentally by isolating a cell and comparing the tissue type(s) generated from the isolated cell with those from an in vivo fate map (Sulston and White, 1980; Sulston et al., 1983). If a cell can produce in isolation all tissue types that it normally would in an intact embryo, its developmental potential is interpreted as being cell-autonomous. Such analyses in early studies suggested that many cell fates in C. elegans embryos are specified cell-autonomously (Sulston and White, 1980; Sulston et al., 1983). However, from follow-on analyses, it became clear that simple cell isolation experiments often overlooked the requirement of cell-cell interactions in cell fate specification, which are now known to be central during C. elegans embryogenesis (Priess et al., 1987; Priess and Thomson, 1987; Schierenberg, 1987; Wood, 1991; Bowerman et al., 1992b; Goldstein, 1992, 1993; Hutter and Schnabel, 1994, 1995; Schnabel, 1994, 1995; Priess, 2005; Sawa and Korswagen, 2013).

A good example of dramatic reversal of thinking concerns the ability of an early 8-cell stage blastomere, MS, to produce body wall muscles (BWMs). C. elegans embryos produce a total of 81 BWMs, of which 28 derive from MS (Sulston et al., 1983). Upon isolation, or following laser ablation of all other blastomeres, the MS blastomere still generates the wild-type number of 28 BWMs. This result was initially viewed as strong evidence that MS produces BWMs cell-autonomously (Priess and Thomson, 1987; Mello et al., 1992; Schnabel, 1994). However, cell-cell interactions were subsequently found to be required for MS-derived BWM. An inductive interaction was identified that functions by countering an earlier inhibitory interaction, and is dispensable if the inhibitory interaction is absent (Schnabel, 1994). Therefore, an isolated MS blastomere, which receives neither the inhibitory nor the activating interaction, produces the normal number of BWMs (Fig. 1B). The precise nature of the inhibitory or the inductive signal for MS-derived BWM cells has not been identified. However, blastomere ablation experiments have identified probable sources for these two interactions, as well as the developmental stages at which these interactions normally occur (Schnabel, 1994). The mother cell of MS is the EMS blastomere, which, in a 4-cell embryo, contacts the other three blastomeres – P2, ABp and ABa (Fig. 1A). Ablation experiments showed that ABp or its descendants are the sources of the inhibitory interaction, whereas two ABa-derived blastomeres, ABalp and ABara, provide the inductive interaction at the 12- to 14-cell stage (Schnabel, 1994). Interestingly, while MS produces additional tissue types, including pharyngeal muscle cells, somatic gonad and neurons, only BWMs derived from MS require the inductive interaction from the two ABa granddaughters.

Although neither the inhibitory nor the inductive signals had been identified, it was known that the inductive interaction required the expression of a Notch receptor, GLP-1 (Priess et al., 1987; Priess and Thomson, 1987; Yochem and Greenwald, 1989; Schnabel, 1994; Moskowitz and Rothman, 1996). GLP-1 functions in multiple cell-cell interactions in the early C. elegans embryo that specify cell fate, including specifying the fate of ABalp and ABara at the 12-cell stage (Priess et al., 1987; Priess and Thomson, 1987; Moskowitz et al., 1994; Priess, 2005). This result suggests that the generation of MS BWMs requires the expression of a protein in ABalp and ABara that is induced by Notch signal. Intriguingly, we now know that the specification of ABp also requires the Notch receptor GLP-1 (Mango et al., 1994; Mello et al., 1994; Moskowitz et al., 1994). This Notch signaling occurs at the 4-cell stage (Fig. 1A), induced by the ligand APX-1 which is expressed in the P2 blastomere (Mickey et al., 1996).

The two Notch signals, one at the 4-cell stage and the other at the 12-cell stage, appear to require different levels of Notch activity. The strong reduction-of-function mutation e2144 or RNAi of glp-1 prevents both signals from occurring (Kodoyianni et al., 1992; Mello et al., 1994), whereas the weak reduction-of-function mutation e2142 (Kodoyianni et al., 1992) prevents only the 12-cell stage signal (Mello et al., 1994). These two glp-1 mutations also differentially affect production of MS-derived BWMs. However, counterintuitively, it is only the ʻweak' glp-1 allele e2142, and not the ʻstrong' glp-1 allele e2144, that results in absence of MS-derived BWMs (Schnabel, 1994). The different phenotypes of the two glp-1 alleles with respect to MS BWM generation could result from Notch signaling regulating both the inhibitory and activating interactions for MS BWMs, either directly or indirectly, by differentially influencing the signaling ability of ABp and ABalp/ABara. In this scenario, the strong allele e2144 would reduce both the inhibitory and activating signals and therefore, analogous to the isolated MS blastomere, wild-type numbers of BWMs would be generated. The weak allele e2142 would reduce only the more sensitive activating signal and no BWMs would result. For simplicity, we will refer to the strong and weak glp-1 alleles as glp-1(S) for e2144 and glp-1(W) for e2142, respectively.

The C. elegans genome contains five Wnt genes, of which only one, mom-2, is known to function in the very early embryo (Sawa and Korswagen, 2013). Wnt signaling from P2 to EMS, mediated by translation of maternally supplied mom-2 mRNA, functions in the specification of the E blastomere as the sole intestinal precursor (Rocheleau et al., 1997; Thorpe et al., 1997; Sawa and Korswagen, 2013). EMS, the mother of MS and E, exhibits the potential to generate all cell types made by MS and E. However, this potential is differentially segregated to the two daughter cells. The P2-derived Wnt/MOM-2 signal is important for E to adopt a fate that is different from the MS blastomere. When either P2 or MOM-2 activity is absent, E does not produce intestinal cells and instead produces tissue types normally made by a wild-type MS blastomere.

Whereas maternally expressed MOM-2 is essential for embryonic viability, it is unclear whether MOM-2 is zygotically expressed and, if so, what function it might have. Genetic results support zygotic expression and that zygotic MOM-2 is dispensable in embryos that express maternally supplied MOM-2 (Rocheleau et al., 1997; Thorpe et al., 1997). However, despite apparent dispensability, zygotically expressed MOM-2 can partially rescue the morphogenesis defect, but not the E specification defect, exhibited by embryos that lack maternally supplied MOM-2. This suggests that maternal and zygotic MOM-2 share overlapping functions during morphogenesis (Thorpe et al., 1997).

We show here that zygotically expressed MOM-2 drives both the inhibitory as well as the inductive interactions regulating MS-derived BWMs. MOM-2 is zygotically expressed in the Notch signal-receiving cells as a result of both the 4-cell and 12-cell Notch-mediated interactions. In addition, we show that Notch/GLP-1 is required for zygotic mom-2 expression. Absence of zygotic mom-2 expression restores the generation of MS-derived BWMs in glp-1(W) embryos. Furthermore, we show that a defective 4-cell inhibitory Notch interaction leads to loss of lineage-restricted transcription in both the MS and E blastomeres, supporting a model in which the inhibitory interaction serves to repress ectopic (i.e. precocious) activation of both MS and E fates within the EMS blastomere.

Using a probe set specific to mom-2 mRNA in single-molecule fluorescence in situ hybridization (smFISH) experiments, we and others have shown that mom-2 mRNA is present at a high level in oocytes and early embryos (Harterink et al., 2011; Oldenbroek et al., 2013), consistent with maternal contribution. The level of mom-2 smFISH signal decreases in the embryo, beginning at the 4-cell stage and primarily in the somatic blastomeres, a pattern observed for many maternally supplied mRNAs in early embryos (Seydoux et al., 1996).

We observed, however, that the decrease in mom-2 smFISH signal was not uniform among AB-derived blastomeres. In the 8-cell embryo, when there are four AB-derived blastomeres (two ABa daughters and two ABp daughters), the smFISH signal was consistently 2- to 4-fold higher in the ABp-derived blastomeres (Fig. 2A-D, Table S1). In the 14- to 15-cell embryo, when there are eight AB-derived blastomeres, elevated smFISH signal was detected primarily in two of the four ABa granddaughters (ABalp and ABara), in addition to all ABp-derived blastomeres (Fig. 2E-H, Table S1). This uneven distribution could reflect either differential degradation of maternal mom-2 mRNA or zygotic transcription of mom-2 mRNA in selected somatic blastomeres. We believe the latter to be true for the following reasons. First, we analyzed two other maternal mRNAs, mex-3 and spn-4, both of which have been shown to be preferentially degraded in somatic blastomeres (Draper et al., 1996; Ogura et al., 2003), by smFISH, and we did not consistently detect any uneven distribution of smFISH signal among the AB-derived blastomeres with either probe (Fig. S1). Second, an extrachromosomal transgene in which gfp::histone h2b is driven by a 2.9 kb putative promoter region of mom-2 (Pmom-2::gfp::h2b) is expressed in the same subset of somatic blastomeres (ABp derivatives and ABara/ABalp) that exhibit the elevated mom-2 smFISH signal (Fig. 3A-H). As extrachromosomal transgenes generated by this method are silenced in the gonad (Mello and Fire, 1995; Kelly et al., 1997), transgene expression is likely to be the result of zygotic transcription. Third, and most conclusively, we observed the same blastomeres exhibiting elevated GFP fluorescence in the cross progeny when males carrying the Pmom-2::gfp::h2b reporter were crossed to hermaphrodites lacking the reporter. In this cross, reporter expression must result from zygotic transcription. We conclude that mom-2 mRNA, in addition to being provided maternally, is transcribed zygotically in a specific subset of blastomeres in early embryos.

The spatial and temporal pattern of Pmom-2::gfp::h2b expression resembles that of the six HES-like bHLH transcription factors, namely REF-1, HLH-25, -26, -27, -28 and -29, which we will refer to as the REF-1 family of genes or proteins (Neves and Priess, 2005). Members of the REF-1 gene family are known direct Notch targets (Neves and Priess, 2005). Expression of REF-1 family proteins in ABa- and ABp-derived blastomeres is dependent on Notch signaling and is abolished in embryos lacking glp-1 activity. Whereas REF-1 family expression in EMS descendants in later embryos is GLP-1 independent, it is dependent on SKN-1, an EMS-specific maternal transcription factor (Neves and Priess, 2005).

We find that expression of Pmom-2::gfp::h2b in ABa- and ABp-derived blastomeres is, like that of REF-1 family genes, also dependent on Notch signaling. Reporter expression is greatly reduced or eliminated in glp-1(S) embryos raised at a non-permissive temperature (25°C) (Fig. 3I-L). The reduction in reporter expression is less severe in glp-1(W) embryos, with GFP detectable at low levels in both ABp- and ABa-derived blastomeres. Like the REF-1 gene family, expression of Pmom-2::gfp::h2b in EMS descendants is GLP-1 independent and SKN-1 dependent (Fig. 3K,L,S,T). The ligand(s) for the 12-cell Notch interaction has not been identified, but is believed to require SKN-1 activity (Bowerman et al., 1992a; Shelton and Bowerman, 1996). We observed no Pmom-2::gfp::h2b expression in either of the ABa descendants or the EMS lineage in skn-1(zu67) embryos (Fig. 3S,T). The ligand for the 4-cell Notch signal is APX-1 (Mickey et al., 1996). In apx-1(zu183);skn-1(zu67) embryos, the GFP::H2B signal was completely abolished (Fig. 3O-R, Fig. S2). Together, these results demonstrate that two sequential Notch interactions in the early embryo induce zygotic expression of Pmom-2 gfp::h2b in ABp and a subset of ABa descendants.

As the P2 blastomere exhibits both Notch and Wnt signaling ability, we also analyzed whether maternally expressed Wnt/MOM-2 regulates the zygotic expression of Pmom-2::gfp::h2b in ABp. Although blastomere divisions are slightly altered in mom-2(or42) embryos (Rocheleau et al., 1997; Thorpe et al., 1997), we observed expression of GFP in all descendants of ABp and in the two ABa granddaughters that touch MS (Fig. 3M,N). We conclude, therefore, that maternal MOM-2 does not regulate zygotic expression of Pmom-2::gfp::h2b.

Interestingly, our analyses appear to have uncovered a previously unidentified SKN-1-dependent Notch signal from the EMS lineage to the two ABp granddaughter cells, ABprp and ABplp, that contact EMS descendants. In wild-type embryos, ABprp and ABplp exhibit a higher level of GFP::H2B from the mom-2 reporter than the other two ABp granddaughters (ABpra and ABpla) that do not contact EMS descendants (Fig. 3O,P Fig. S2). This difference in GFP levels was less pronounced in skn-1(zu67) embryos (Fig. 3O,P compared with 3S,T, Fig. S2). Furthermore, in apx-1(zu183) embryos, GFP::H2B is abolished in ABpra and ABpla but is only reduced in ABprp and ABplp (Fig. 3U,V, Fig. S2). All GFP signal is abolished in skn-1(zu67);apx-1(zu183) embryos (Fig. 3Q,R, Fig. S2). The function of this SKN-1-dependent Notch interaction is unclear. Wnt signaling has been proposed to have a direct role in gastrulation. Both ABprp and ABplp daughters surround the two E descendants at the time when gastrulation starts. This Notch interaction would place zygotic MOM-2-expressing cells at the right place and time to regulate this process. Further characterization of this novel SKN-1-dependent Notch interaction is currently under way.

The expression pattern of mom-2 suggests zygotic MOM-2 as a candidate ligand for the inhibitory interaction between ABp (or descendants) and EMS, as well as the later activating interaction between certain ABa descendants and MS. If zygotic MOM-2 were the ligand for both interactions, absence of zygotic mom-2 should remove the inhibitory interaction and thereby restore the formation of MS BWMs in glp-1(W) embryos, which are defective in the inducing interaction. However, absence of zygotic mom-2 would be predicted to have no effect on the ability of MS to generate wild-type numbers of BWMs in glp-1(S) embryos, which are already defective in both the inhibitory and the inducing interactions.

To test this possibility, we crossed glp-1(−) (III);mom-2(ok519) (V) males with glp-1(−) hermaphrodites homozygous for an insertion on LGV, teIs18(gfp), which express a GFP::H2B reporter in intestinal precursors (Shetty et al., 2005). F1 cross progeny would be homozygous for glp-1(−/−) and transheterozygous for mom-2(ok519) and teIs18 on LGV. All F2 embryos produced by these F1 hermaphrodites will be glp-1(−/−) and express maternal MOM-2 protein [maternal positive, or MOM-2(M+)]. However, ∼25% of F2 embryos would be homozygous for mom-2(ok519) and would not express zygotic MOM-2 protein [zygotic negative, or MOM-2(Z−)]. These MOM-2(M+ Z−) embryos would be identifiable by virtue of being negative for intestinal GFP expression, unless an extremely rare recombination event occurred between mom-2 and teIs18. We performed immunofluorescence on these F2 embryos using anti-myosin antibody (Priess and Thomson, 1987), which specifically stains BWMs.

3D image stacks were collected for embryos of various genotypes at 400-450 min post fertilization (Fig. 4A-J), and the numbers of myosin-positive BWMs were counted (Fig. 4K). Wild-type embryos generated on average 81 BWMs (n=10, s.d. 1.5) and glp-1(S) embryos generated on average 80 BWMs (n=7, s.d. 2.6). Control glp-1(W) embryos generated on average 58 BWMs (n=32, s.d. 4.6), which is less than the wild-type 81, as expected owing to the absence of MS-derived BWMs (Schnabel, 1994). Putative glp-1(W);mom-2(M+ Z−) GFP-negative embryos generated on average 80 BWMs (n=9, s.d. 1.9), very close to the number from wild-type embryos. Two additional GFP-negative embryos each generated 62 BWMs, indicating no rescue. A parallel control showed that putative mom-2(M+ Z−) embryos generated on average 80.5 BWMs (n=6, s.d. 1.0). That removal of zygotically expressed Wnt/MOM-2 suppressed the MS BWM defect in glp-1(W) embryos demonstrates that Wnt signaling mediates the inhibitory interaction for MS-derived BWMs.

It is curious that MS BWM fate must first be inhibited only to have the inhibition lifted by an activating interaction two cell cycles later. We investigated whether embryos lacking the inhibitory interaction suffer any developmental consequence in the EMS lineage. The MS and E developmental programs are driven by the lineage-restricted expression of two different sets of transcription factors (for a review, see Maduro, 2015; Maduro et al., 2015). The MS blastomere expresses the T-box transcription factor TBX-35, which, along with the homeodomain transcription factor CEH-51, drives both body wall and pharyngeal muscle programs (Broitman-Maduro et al., 2006, 2009). The E blastomere expresses two partially redundant GATA factors, END-1 and END-3, which are together necessary and individually sufficient for the intestinal program (Zhu et al., 1997; Maduro et al., 2005).

To test whether the APX-1-dependent inhibitory interaction between the ABp lineage and the EMS lineage serves to modulate transcription of MS and E lineage-specific genes, we analyzed the expression of a tbx-35 transgenic reporter or an end-1 transgenic reporter in apx-1(zu183) mutant embryos (Fig. 5). In wild-type embryos expressing the tbx-35 reporter, GFP is observed exclusively in the MS lineage, starting in the MS daughters (0% embryos with ectopic expression, n>250; Fig. 5A, Fig. S3). In apx-1(zu183) embryos, approximately one-quarter (21 of 89) exhibited ectopic GFP expression in the E lineage (Fig. 5B, Fig. S3). Similar ectopic lineage expression was also observed for the end-1 reporter (Fig. 5D,E, Fig. S3). The end-1 reporter is expressed almost exclusively in the E lineage in wild-type embryos, with rare ectopic expression observed (<1% of embryos, n>300). Analogous to the tbx-35 reporter, ectopic expression of the end-1 reporter in the MS lineage was observed in almost one-third (19 of 61) of apx-1(zu183) embryos. For both transgene reporters, the ectopic GFP levels detected were almost always weaker than those in the normal lineage of the same embryo. These results suggest that APX-1, and by extension the ABp-derived blastomeres, modulates lineage-restricted transcription in both the MS and E lineages.

What might be the nature of the ABp-derived lineage-specific repressive activity? Repression of the MS- and E-specific transcriptional programs must also occur in the EMS blastomere. Two types of DNA-binding proteins, both capable of mediating transcriptional repression, are known to be expressed in the EMS blastomere: the TCF protein POP-1 and the bHLH REF-1 family proteins (Lin et al., 1995; Neves and Priess, 2005; Coroian et al., 2006). We have shown previously that POP-1 is present at a higher nuclear level in MS, where it binds to the end-1 promoter and represses end-1 transcription (Shetty et al., 2005). Mutating the POP-1 binding site centered at −164 relative to the first base of the AUG translation start site on an end-1 transgene resulted in derepression of the end-1 reporter in the MS lineage (Shetty et al., 2005). Repression of end-1 transcription in MS by POP-1 requires the co-repressors UNC-37/Groucho and HDA-1/HDAC (Calvo et al., 2001). Depletion of pop-1, unc-37 or hda-1 resulted in a similar derepression of end-1 in MS (Calvo et al., 2001; Shetty et al., 2005). The tbx-35 promoter does not contain a readily identifiable POP-1 binding site. Therefore, MS-restricted transcription of tbx-35 presumably requires a different mode of suppression specifically in the E lineage. As there are a total of six REF-1 family genes, it is difficult to inactivate REF-1 activity completely (Neves and Priess, 2005). Therefore, to investigate whether the REF-1 family of DNA-binding proteins could contribute to lineage-restricted expression in MS, E, or both blastomeres, we performed mutational analyses on the promoters of the tbx-35 and end-1 reporters (Shetty et al., 2005).

The mammalian and fly REF homologs [HES, Hairy and E(spl)] have been shown to bind to DNA sequences CACGNG (Class C sites), CACNAG (N-box) and, to a lesser extent, CANGTG (Class B sites) (Ohsako et al., 1994; Van Doren et al., 1994; Fisher and Caudy, 1998; Jennings et al., 1999; Iso et al., 2003). C. elegans REFs have also been shown to bind TGCCACGTGTCCA in vitro (Neves and Priess, 2005). Here, we will refer to all three of these DNA sequences as candidate REF-1 sites. Interestingly, both the tbx-35 and end-1 promoters include a candidate REF-1 site between the most proximal of multiple MED-1/2 sites and the predicted AUG translation initiation codon (Fig. 5G,H). For tbx-35, this candidate REF-1 site (−152) overlaps with the MED-1/2 binding site at −155 (Broitman-Maduro et al., 2005). A second candidate REF-1 site (−162) is located between, and overlaps with, two MED-1/2 sites, one at −155 and another at −167. We considered the possibility that binding by REF-1 family proteins on the tbx-35 or end-1 promoter interferes with MED-1/2 binding or activity on the same promoter. We mutated the putative REF-1 site at −127 of the end-1 reporter (CACGAG to CTCGTA; mut-127), a mutation shown previously to abolish binding by REF-1 (Neves and Priess, 2005). This transgene was derepressed in the MS lineage with no defect in expression in the E lineage (Fig. 5F, Fig. S3). These mutagenesis results suggest that binding of REF protein at the −127 REF-1 site represses transcription of the end-1 transgene by interfering with MED-1/2 binding or activity at a nearby site.

Mutation of both the −152 (CTCGTG→CTCATA) and −162 (CAACTG→CTAAAG) candidate REF-1 sites on the tbx-35 transgene (mut-162, mut-152, Fig. 5G,H), without altering the consensus MED-1/2 binding sites, also abolished lineage-restricted expression, resulting in derepressed tbx-35::GFP expression in the E lineage (Fig. 5C, Fig. S3). Mutation of either candidate REF-1 site alone had only a minor effect (Figs S3 and S4). Together, these results suggest that REF-1 family proteins repress the transcription of end-1 in the MS lineage and tbx-35 in the E lineage and that this is likely to be by interfering with MED-1/2 binding to, or activity at, the same promoter.

How might zygotic MOM-2 induce MS-derived BWMs? BWMs are generated from two of the four MS granddaughter cells, MSap and MSpp, which are the posterior daughters of MSa and MSp, respectively (Sulston et al., 1983). The anterior daughters of MSa and MSp, however, produce only pharyngeal muscle. A crucial component of the anterior-posterior fate differences in most, if not all, cells in developing C. elegans embryos is the nuclear levels of the TCF protein POP-1 and its coactivator, the β-catenin SYS-1 (Lin et al., 1995, 1998; Kidd et al., 2005; Huang et al., 2007). Both proteins exhibit a reciprocal asymmetric distribution between the nuclei of anterior and posterior sisters, with the posterior nuclei exhibiting higher SYS-1 and lower POP-1 levels compared with the corresponding anterior nuclei (Lin et al., 1998; Huang et al., 2007; Phillips et al., 2007). We have shown previously that the ratio of nuclear SYS-1 to POP-1 is the key regulator of the differential anterior-posterior developmental fates in many, and possibly all, pairs of anterior-posterior sisters in the embryo (Huang et al., 2007). In addition, it was shown that an artificially induced high POP-1 level is incompatible with the production of BWM (Fukushige and Krause, 2005). Therefore, it is possible that specification of MS BWMs involves regulation of the levels, or activity, of SYS-1, POP-1, or both, in MSap and MSpp.

Wnt signaling plays an important, but not exclusive, role in regulating the levels of both POP-1 and SYS-1 in early blastomeres (Lin et al., 1998; Meneghini et al., 1999; Rocheleau et al., 1999; Park and Priess, 2003; Lo et al., 2004; Huang et al., 2007; Phillips et al., 2007). We investigated whether zygotically expressed MOM-2 modulates POP-1 or SYS-1 levels in the MSa and MSp daughters. We used a strain carrying an integrated transgene that expresses SYS-1::GFP in embryos, and performed immunofluorescence using anti-POP-1 and anti-GFP antibodies (Huang et al., 2007). In wild-type embryos we observed clear reciprocal asymmetric levels of POP-1 and SYS-1 between daughters of MSa and MSp (Huang et al., 2007). MSaa and MSpa have a higher level of nuclear POP-1 and lower level of SYS-1 than MSap and MSpp, respectively (Fig. 6A-C, Fig. S5). In embryos from hermaphrodites homozygous for mom-2(or42), which lack both maternal and zygotic MOM-2, we observed no asymmetry for SYS-1 and variable asymmetry for POP-1 between MSa and MSp daughters (Fig. 6D-I, Fig. S5). We asked whether providing zygotically expressed MOM-2 in these embryos could rescue the SYS-1 and/or POP-1 defects. We crossed males from a strain expressing teIs18 into mom-2(or42) homozygous hermaphrodites, and performed immunofluorescence on the embryos. Embryos that derived from the cross would express GFP::H2B from teIs18 and would have one wild-type copy of the mom-2 gene that can be expressed zygotically [MOM-2(M− Z+)]. Of five 28-cell embryos expressing GFP::H2B in the E lineage, all five exhibit the wild-type pattern of POP-1 asymmetry between daughters of MSa and MSp (Fig. 6J-L, Fig. S5). The effect on SYS-1::GFP asymmetry is difficult to evaluate in this experiment because the level of histone H2B overshadows that of SYS-1::GFP (Fig. 6K). This suggests that zygotic MOM-2 is sufficient to regulate POP-1 asymmetry in the daughters of MSa and MSp.

We also asked whether zygotic MOM-2 is necessary for SYS-1 or POP-1 asymmetry between daughters of MSa and MSp. We performed immunofluorescence of embryos derived from hermaphrodites transheterozygous for mom-2(or42) and teIs18 [MOM-2(M+ Z−)]. Of the two 28-cell embryos lacking GFP::H2B in the E lineage, which are likely to be homozygous for mom-2(or42), both have a wild-type pattern of POP-1 and SYS-1 asymmetry (Fig. 6M-O, Fig. S5). This shows that zygotic expression of MOM-2 is not required for POP-1 or SYS-1 asymmetry as long as maternal MOM-2 is present.

Together, our analyses suggest that the levels of POP-1 and SYS-1 in MSap and MSpp can be regulated by either maternally or zygotically expressed MOM-2.

We show here that reciprocal signaling by Wnt and Notch specifies the BWM fate of the MS blastomere in C. elegans embryos. Two distinct Notch-mediated interactions, separated both spatially and temporally, induce zygotic expression of the Wnt ligand MOM-2 in a subset of descendants from the AB blastomere. Zygotically expressed MOM-2 then mediates two distinct interactions, with the first inhibiting the ability of MS to generate BWMs and the second interaction, two cell cycles later, to lift the repression. Our data suggest a molecular model by which the repression and activation of MS-derived BWM fate can be achieved.

Several observations support our conclusion that zygotic mom-2 serves as a ligand for both the inhibitory interaction between ABp descendants and EMS, and the activating interaction between ABa descendants and MS that regulate MS-derived BWMs (Fig. 7A). First, the temporal and spatial expression patterns of zygotic mom-2 put it in the right place and at the right time to mediate these interactions. Second, the differences in mom-2 transgene expression in the glp-1(e2142) and glp-1(e2144) backgrounds can explain the differential MS-derived BWM production in these two mutants (Schnabel, 1994). This supports a model whereby e2144, the glp-1 ‘strong’ mutant, exhibits defects in both the inhibitory and activating interactions, whereas e2142, the glp-1 ‘weak’ mutant, exhibits a defect only in the activating interaction. Third, our strongest piece of data supporting this model is the suppression of the MS BWM defect in glp-1(e2142) embryos when zygotic expression of MOM-2 is prevented. This clearly demonstrates that the inhibitory interaction requires zygotic mom-2. We do not currently have the genetic tools necessary to test directly whether zygotic Wnt also mediates the activating interaction. This idea nevertheless receives strong indirect support from our result that zygotic Wnt can regulate the nuclear levels of both SYS-1 and POP-1 in MSap and MSpp.

At face value it appears to be an overly complex strategy for an embryo to specify BWM potential in MS through two sequential interactions, the first an inhibitory interaction that is relieved some minutes later by an activating interaction. Our results here suggest that the function of the initial inhibition is to prevent precocious activation of lineage-restricted genes in MS and E. During the first few hours of C. elegans embryogenesis, cell cycle times are short, at ∼10-15 min, and a number of cell lineages are specified early during embryogenesis (Sulston et al., 1983). Specification of these precursor blastomeres results from the coordinated action of both maternally supplied factors and the lineage-restricted zygotic expression of transcription factors (see reviews by Schnabel and Priess, 1997; Maduro, 2010). The short cell cycle times demand very precise timing of zygotic expression of lineage-restricted transcription factors in the correct blastomeres. Although MS and E produce mutually exclusive tissue types, they are sister blastomeres. Furthermore, transcriptional activation of both MS and E lineage-restricted genes requires zygotic expression of MED-1 and MED-2, both of which are transcribed in the mother cell, EMS (Maduro et al., 2001; Robertson et al., 2004; Broitman-Maduro et al., 2005, 2006, 2009). This clearly poses the risk of precocious transcriptional activation of MS- and E-restricted transcription factors, with the resultant loss of lineage restriction.

The data presented here suggest a possible mechanism by which such precocious transcriptional activation can be prevented in EMS (Fig. 7B). We propose that tbx-35 and end-1 are transcriptionally repressed in EMS and early MS and E cycles by REF-1 family proteins. This repression is a consequence of the arrangement of the respective binding sites on the end-1 and tbx-35 promoters, making it incompatible for MED-1/2 to bind or function when the REF-1–UNC-37 complex is bound. The observed ectopic expression of tbx-35 and end-1 in apx-1(zu183) embryos suggests that this REF-1-mediated repression is regulated by ABp or its descendants. We propose that a zygotic MOM-2-mediated interaction from ABp descendants to EMS regulates REF-1 repression of tbx-35 and end-1. How the ABp interaction regulates the REF-1-mediated repression in EMS remains to be determined.

This competitive binding model suggests that the repression by REF-1 would be relieved when MED-1 and MED-2 levels increase. However, as MED-1/2 proteins are present in both MS and E, why is tbx-35 not expressed in E, and end-1 not expressed in MS? For end-1, the repression in MS and expression in E has been explained by the asymmetric levels of POP-1 and SYS-1 between these two cells (Lo et al., 2004; Shetty et al., 2005; Huang et al., 2007). A high SYS-1 and low POP-1 level in E allows POP-1 to function as a transcriptional activator for end-1. In the MS blastomere, a high level of nuclear POP-1 binds to UNC-37 and HDC-1, together keeping the end-1 gene repressed (Calvo et al., 2001).

This same paradigm does not explain the MS-restricted expression of tbx-35. We propose that although the tbx-35 promoter does not contain an obvious POP-1 binding site, the differential nuclear levels of POP-1 still underlie lineage-restricted expression of tbx-35 between MS and E. Both POP-1 and REF-1 family proteins bind to UNC-37, which is required for transcriptional repression by each complex (Calvo et al., 2001; Neves and Priess, 2005). The levels of UNC-37 and HDC-1 are similar in MS and E nuclei (Calvo et al., 2001). If the level of UNC-37 is limiting compared with POP-1 and REF-1 family proteins, then the repressive activity of REF-1 in MS and E could differ as a consequence of different levels of nuclear POP-1 in these two cells. In the E blastomere where POP-1 nuclear level is low, REF-1 remains effective as a repressor for tbx-35 even as the level of MED-1/2 increases. In the MS blastomere, where POP-1 nuclear level is high and therefore the availability of unbound UNC-37 for REF-1 binding is low, an increased level of MED-1/2 would overcome the repression by REF-1 family proteins, allowing transcription of tbx-35.

Why is it that MS BWM cell fate, but not pharyngeal muscle cell fate, requires an activating interaction? It has been suggested previously that development of BWM fate in C. elegans embryos is sensitive to the level of POP-1 protein (Fukushige and Krause, 2005). In the MS lineage, BWMs derive from MSap and MSpp, two posterior daughters that in the wild type have low nuclear levels of POP-1 and high levels of SYS-1. We show here that, in MSap and MSpp, higher levels of SYS-1 and lower levels of nuclear POP-1 require either maternal or zygotic MOM-2. Our results supports a model whereby the activating interaction mediated by zygotic MOM-2 expressed in ABalp and ABara, or their daughters, regulates the levels of POP-1 and SYS-1 in MSap and MSpp, leading to the generation of BWMs (Fig. 7B). MS-derived BWMs line the head region of worms, helping to explain the previous observation that zygotic expression of MOM-2 can partially rescue the head morphogenesis defect of mom-2(or42) embryos (Thorpe et al., 1997). We show that embryos that express maternal MOM-2 but lack zygotic MOM-2 have normal POP-1 and SYS-1 levels in MSap and MSpp. This observation supports the notion that regulation of SYS-1 and POP-1 levels in MSap and MSpp is only needed when the inhibitory interaction mediated by zygotic MOM-2 occurs earlier in embryogenesis. Further investigation is needed to elucidate the molecular mechanism(s) by which the inhibitory interaction influences the levels of POP-1 and SYS-1.

This study highlights the extent of cell-cell interactions and the tremendous complexity of cell fate specification in an organism once thought to be totally hardwired. Like many other observations in biology, the phenomena often reflect the evolution of the pathways, rather than the simplest solution to a problem.

N2 was used as the wild-type strain. Genetic markers utilized in this study: LGIII: glp-1(e2142), glp-1(e2144), unc-119(ed3); LGIV: skn-1(zu67), DnT1(qIs51); LGV: mom-2(or42), mom-2(ok591), apx-1(zu183), dpy-11(e224), teIs18, DnT1(qIs51). Strains utilized in this study: TX895 [unc-119(ed3)III; him-3(e1147)IV; teIs84(Pend-3::gfp::H2B, pDPmm16)X] (Shetty et al., 2005), TX585 [unc-119(ed3)III; teIs18(Psdz-23::gfp::H2B, pDPmm016)V] (Robertson et al., 2004), TX1310 [unc-119(ed3)III; teEx268(Pend-1::gfp::H2B, pDPmm016)], TX964 [teIs98(Ppie-1::gfp::sys-1)], TX878 [unc-119(ed3)III; teEx368 (pRL1950 Pend-1^(mut-127)::gfp::H2B, pDPmm016)], TX929 [unc-119(ed3)III; teEx393 (pRL2075 Pend-1^{(mut-140,-127)}::gfp::H2B, pDPmm016)], TX989 [unc-119(ed3)III; teEx435 (pRL1582 Pmom-2::gfp::H2B, pDPmm016)], TX1038 [unc-119(ed3)III; teEx469 (pRL1377 Ptbx-35::gfp::H2B, pDPmm016)], TX1182 [unc-119(ed3)III; teEx576 (pPD93.92 Phlh-1::hlh-1::gfp, pDPmm016)], TX1971 [unc-119(ed3)III; teEx936 (pRL3467 Ptbx-35^{(mut-162,-152)}::gfp::H2B, pDPmm016)], TX1963 [unc-119(ed3)III; teEx928 (pRL3467 Ptbx-35^(mut-162)::gfp::H2B, pDPmm016)], TX1967 [unc-119(ed3)III; teEx932 (pRL3467 Ptbx-35^(mut-152)::gfp::H2B, pDPmm016)]. All other strains were obtained from the Caenorhabditis Genetics Center (CGC) and were grown at 20°C unless specified. All teEx strains were generated by injection and the transgenes are not integrated. For each construct, expression was analyzed and found to be consistent in at least two independent lines.

All images were acquired, processed and quantified as described previously (Huang et al., 2007). Expression of reporter GFP in wild-type and mutant embryos was analyzed as described previously (Shetty et al., 2005; Robertson et al., 2014). Embryos were prepared for immunofluorescence using a 1:50 dilution of the anti-myosin heavy chain A monoclonal antibody 5.6 as described (Miller et al., 1983; Priess and Thomson, 1987). smFISH of mom-2 mRNA was performed as described previously (Harterink et al., 2011; Oldenbroek et al., 2013).

All plasmids were constructed using Gateway technology as previously described (Guven-Ozkan et al., 2010). A 3.4 kb genomic sequence comprising −2876 to +555 of the mom-2 gene (promoter, first exon, first intron and part of the second exon) was cloned into pRL1075 upstream of, and in-frame with, gfp::h2b. A 54 bp region corresponding to amino acids 2-19 of MOM-2, the putative signal peptide, was then deleted to prevent interference with the nuclear localization of GFP::histone H2B. The two putative LAG1 binding sites (ATGGGAA at −1038 and −992) were mutated to AAGGCAA, with no detectable change in the GFP pattern. All mutations were generated with the QuikChange II Site-Directed Mutagenesis Kit (Stratagene/Agilent Technologies) and confirmed by sequencing. The end-1 and tbx-35 promoters used comprise 2.2 kb and 734 bp, respectively, from upstream of the ATG (Robertson et al., 2004). An end-1 promoter with three other candidate REF sites deleted (−367, −2064, −3100) did not result in derepression of the reporter or enhancement of the derepression when combined with mutation at −127.

We thank R.L. lab members for discussion; Heng Yu Fan, Gokalp Celik, Haochen Cui and Defen Lu for technical support; and the Caenorhabditis Genetics Center for strains.