Structure and Evolution of the C. elegans Embryonic Endomesoderm Network

Summary

The specification of the C. elegans endomesoderm has been the subject of study for more than 15 years. Specification of the 4-cell stage endomesoderm precursor, EMS, occurs as a result of the activation of a transcription factor cascade that starts with SKN-1, coupled with input from the Wnt/β-catenin asymmetry pathway through the nuclear effector POP-1. As development proceeds, transient cell fate factors are succeeded by stable, tissue/organ-specific regulators. The pathway is complex and uses motifs found in all transcriptional networks. Here, the regulators that function in the C. elegans endomesoderm network are described. An examination of the motifs in the network suggests how they may have evolved from simpler gene interactions. Flexibility in the network is evident from the multitude of parallel functions that can been identified, and from an apparent change in parts of the corresponding network in C. briggsae. Overall, the complexities of C. elegans endomesoderm specification build a picture of a network that is robust, complex, and still evolving.

1. Introduction

Triploblastic animals begin life as a single cell, which after many rounds of mitosis, will ultimately consist of a multitude of genetically equivalent cells. By adulthood, the majority of these will have selected a particular pathway of differentiation, each expressing a subset of the genes in the organism’s genetic complement that uniquely defines its type. At some point in embryogenesis, precursor cells become specified, and acquire transcriptional differences that set them apart from their neighbors. These differences will instruct their descendants as to their ultimate cell type, or at least restrict their choices until a later time.

In the nematode, C. elegans, cells acquire these differences very early, as seen in the stereotyped cleavage patterns that are the hallmark of its nearly-invariant cell lineage [1]. The point of sperm entry sets the posterior of the embryo, defining one of the three embryonic axes (reviewed in [2]). The first division produces a larger cell, AB, and a smaller posterior cell, P₁. Following division of AB and P₁, the embryo consists of the anterior and posterior daughters of AB (ABa and ABp, respectively), and the two daughters of P₁, called EMS and P₂ (Fig. 1). EMS, situated ventrally, is an endomesoderm precursor: It will divide to produce a posterior daughter, called E, and an anterior daughter, MS. The E cell will clonally generate the 20 larval cells of the midgut (endoderm), while MS generates many cells that are primarily mesodermal, which includes most cells in the posterior half of the pharynx, and many of the animal’s body muscles. The remaining portion of the pharynx is made by the anterior daughter of AB (ABa). Because many cells in the C. elegans lineage undergo anterior-posterior divisions to produce daughters that will acquire different fates, specification of MS and E is a platform for examining mechanisms that may operate throughout much of the animal’s development.

FIG. 1.

The C. elegans endomesoderm gene regulatory network. Ovals represent transcription factors, while rectangles indicate other types of proteins. Question marks denote hypothesized co-regulators or functions. Arrows denote, in most cases, direct regulatory interactions. Thinner arrows denote contributions that are hypothesized to be weaker based on genetic evidence. Overlapping transcription factor symbols denote common function and are not meant to imply physical interaction. Diagrams of the C. elegans embryo (4-cell and 8-cell stages), and anatomy of the digestive tract in a larva are shown, after ref. [3], with anterior to the left, and dorsal upwards. Sections of the digestive tract are labeled with the name of the blastomere whose descendants contribute to that region. The association of ‘anterior pharynx’ with ABa, and ‘posterior pharynx’ with MS is in reality not a precise distinction [9] and is shown here as such for simplicity. GLP-1/Notch-dependent cell-cell interactions are shown as ‘Notch’ with an arrow [9].

Work over the past 15+ years has identified multiple factors that specify the C. elegans endomesoderm. Essentially, there are two pathways that converge on EMS specification: The SKN-1/MED-1,2 pathway assigns an endomesodermal fate to EMS, while the Wnt/β-catenin asymmetry pathway makes E different from MS [3, 4]. Although the pathways that lead to MS and E specification look superficially like a simple cascade, the network contains much subtlety, crosstalk, redundancy, and flexibility. This review will examine the genes that specify MS and E, how deployment of their developmental programs is restricted to the appropriate lineages, and how the overall network may be evolving. A diagrammatic summary of the information flow in the network is presented in Fig. 1.

2. The endomesoderm network

The rapid development of C. elegans is considered derived within the phylum, and the rapid divisions in the early embryo are proposed to be correlated with the use of maternal factors to drive much of the early cell specification events [5–7]. Screens for maternal embryonic lethals, in which arrested embryos lack one or more major tissue types but still contain many differentiated nuclei, led to the identification of multiple factors, including the gene skn-1 [8].

2.1. Getting it all started: Maternal SKN-1 specifies EMS

Embryos from skn-1(−) mothers undergo a developmental arrest and lack pharynx 100% of the time, while endoderm is absent in approximately 70% of embryos [8]. Pharynx originates from descendants of both MS and ABa [1], and the absence of both AB- and MS-derived pharynx in skn-1 mutant embryos is attributed to the failure of a GLP-1/Notch-mediated cell induction that occurs between the MS cell and descendants of ABa [8, 9].

Antibody staining shows that SKN-1 protein is present in the EMS and P₂ nuclei at the 4-cell stage, placing it in the correct time and space to directly act in its specification [10]. As discussed below, SKN-1 does not act in P₂ due to the function of another maternal gene, pie-1 [11–13]. The skn-1 locus encodes a transcription factor that has domains similar to those found in bZIP and homeodomain proteins [14]. As its expression normally disappears during the MS and E cell cycles [10], SKN-1 is likely to be a factor that initiates a zygotic gene cascade that will specify MS and E. The observation that many skn-1(−) embryos still make endoderm points to the existence of parallel pathways that are capable of contributing to gut specification in the absence of SKN-1, and these will be discussed later.

2.2. Zygotic specification of endoderm by END-1 and END-3

Mutagenic screens for penetrant zygotic mutations that resulted in the absence of endoderm identified only a large genomic region on chromosome V, named the ‘Endoderm Determining Region’ or EDR [15]. EDR-deficient [EDR(Df)] embryos lack endoderm and show a transformation of E to a C-like cell [15]. Within a 30-kbp region located within the EDR, two GATA factor genes, end-1 and end-3, were identified that could individually restore endoderm development to EDR(Df) embryos, suggesting that both genes share overlapping function [15, 16]. Consistent with this, overexpression of either end-1 or end-3 can reprogram non-endodermal cells into gut precursors [16, 17]. Transgene fusion reporters for both genes are also expressed in the early E lineage, though expression goes away after several cell divisions [16, 18]. Hence, end-1 and end-3 are clearly paralogous, likely having arisen from an ancient gene duplication [16].

Two observations suggest that end-1 and end-3 have diverged somewhat, which might account for their maintenance [19]. First, while a null mutation of end-1 has no discernible phenotype, mutation of end-3 results in a weak endoderm specification defect, and in those embryos making endoderm, the number of gut cells frequently deviates from the 20 normally seen in wild-type [16, 20]. Second, both in situ hybridization studies and whole-embryo transcriptome experiments show that end-3 is activated slightly earlier than end-1 within the E cell cycle [20, 21]. There is further genetic evidence, discussed below, that END-3 also contributes to activation of end-1 [20].

Are there any other zygotic genes that specifically contribute to E specification? RNAi targeted to both end-1 and end-3 resulted in only approximately 43% of embryos that lacked gut, although this appears to be the result of a weak RNAi effect on end-1 [16]. To resolve the question of whether or not other genes can specify endoderm in addition to end-1 and end-3, we generated an end-1(−) end-3(−) strain using putative null mutants for both loci. All end-1,3(−) embryos lack differentiated endoderm (Fig. 2E), although the overall phenotype is surprisingly mild: Most embryos become enclosed in epidermis and undergo elongation, and some can hatch into arrested larvae as shown in Fig. 2B (M.M., unpublished results). This phenotype is very different from EDR(Df) embryos, which arrest well before elongation due to the simultaneous loss of many genes in addition to end-1 and end-3 [15, 22].

FIG. 2.

Appearance of wild-type and zygotic cell fate specification mutants and associated expression of a pha-4::GFP reporter (pseudocolored yellow; a gift from Jeb Gaudet, University of Calgary). (A) Wild-type L1, showing pharynx and rectum (red) and intestine (blue) and associated blastomeres as in Fig. 1. (B) Arrested end-1(ok558) end-3(ok1448) double mutant embryo, showing absence of gut and slightly reduced body length as compared with wild-type larvae. The vast majority of end-1,3(−) embryos elongate within the eggshell, but only some of these hatch. A more detailed characterization of these mutants will be presented elsewhere. (C) Arrested med-1(ok804); med-2(cx9744); end-1(ok558) end-3(ok1448) quadruple-null mutant. The most severely affected med-1,2(−) embryos arrest lacking gut and at a similar stage as the embryo shown here [25]. (D–F) Expression of pha-4::GFP in pharynx, rectum and intestine cells, corresponding to the DIC images in A–C. The hypodermis is indicated with a dotted blue line. Anterior is to the left in panels A, B, E and F, and up in panels C and F. All micrographs are shown at the same scale. In A–C, Adobe Photoshop was used to colorize regions. A C. elegans embryo is approximately 50μm along its long axis.

The behavior of the early E lineage cells explains the mild end-1,3(−) phenotype. One of the properties of the early E lineage is that the two E daughter cells (Ea and Ep) move into the interior of the embryo, which marks the onset of gastrulation [23]. Examination of the movements of the early E descendants in end-3(−), end-1,3(−) and EDR(Df) embryos, alone and in combination with Wnt pathway mutants, shows that Wnt components and the ENDs share overlapping function in gastrulation [15, 16, 23] [Jacob Sawyer and Bob Goldstein, personal communication]. Indeed, it has been proposed that in general, pathways that control morphogenesis will have greater redundancies than those that specify cell fate [24].

2.3. The MED-1 and MED-2 factors

The involvement of the GATA factors end-1 and end-3 in specification of E led to the hypothesis that a similar regulator might be responsible for specification of MS [25]. Searches of the partially-assembled C. elegans genome sequence led to the identification of med-1 and med-2, two unlinked but nearly-identical putative GATA factors, that seemed to confirm this [25]. Rather than function in MS specification analogous to the end genes, however, the med-1,2 genes appear to function ‘in between’ SKN-1 and end-1,3. RNA interference and double mutant studies with null alleles of both genes have shown that med-1,2(−) embryos lack MS-derived tissues all of the time, but also lack endoderm some of the time (15–50%) [3, 25–27]. Unlike loss of end-1,3, med mutant embryos do not complete elongation and arrest before hatching. Reporter gene, gel shift and in situ hybridization studies showed that the med genes are activated in the EMS cell, and that this expression is transient [20, 25]. Expression is dependent on interaction of SKN-1 with the med promoters, implying that SKN-1 directly activates med-1,2 in EMS [20, 25]. An additional med-1,2 expression component occurs in the maternal germ line that also is dependent on SKN-1 [20]. The germline expression appears to account for the observation that embryos only zygotically lacking in med-1,2 are more likely to specify endoderm, though these findings have been disputed by others based on indirect evidence [20, 26].

Sufficiency experiments show that overexpression of med-1 throughout the embryo results in widespread expression of end-1::GFP and end-3::GFP reporters, and embryonic lethality with excess pharynx muscle or endoderm [18, 25]. These results suggest that med-1 is sufficient to initiate a program of MS or E specification, perhaps in combination with other factors, and that with respect to endoderm, the MEDs function upstream of the end genes [25]. Terminal med-1,2; end-1,3 quadruple mutant embryos (Fig. 2C) lack gut and arrest with an appearance similar to the most profoundly affected med-1,2(−) embryos, consistent with placement of end-1,3 downstream of med-1,2 [25]. The weak endoderm phenotype of med-1,2(−) embryos is consistent with involvement of parallel inputs into end-1,3 activation, discussed later.

Additional evidence supports a role for MED-1,2 in activation of end-1,3. Both in vitro studies (DNaseI footprinting and gel shift assays) and in vivo experiments (visualization of subnuclear spots representing interaction of GFP-tagged MED-1 with target arrays) confirm a direct interaction of MED-1 with the promoters of end-1 and end-3 [28, 29]. Two sites in end-1, and four in end-3, were identified as regions of MED-1 interaction [28]. Unexpectedly, MED-1 appears to not recognize a canonical GATA binding site (HGATAR), but rather a related sequence (RAGTATAC) [28, 30].

There is some evidence that med-1 and med-2 are, like the end-1,3 pair, not completely redundant. First, germline med-2 mRNAs accumulate to apparently higher levels in a med-1(−) animal, than the reverse, although both show similar expression in EMS [20]. Second, of the four possible med-x; end-y double mutant combinations, only med-1; end-3 demonstrates a synergistic effect, resulting in a ~58% gutless phenotype, and a high proportion of embryos containing less than the wild-type number of 20 gut cells [20]. In contrast, a med-2; end-3 double mutant strain makes endoderm >95% of the time, similar to loss of end-3 alone. Therefore, even though the med genes are undoubtedly the result of a recent duplication, they have nonetheless diverged [20, 31].

2.4. MS specification and TBX-35

Under the hypothesis that MED-1,2 would likely bind putative MS regulators in the same manner as with the E targets end-1,3, the C. elegans genome sequence was searched for MED-1 binding site clusters, which identified (among other genes) tbx-35, a gene that contains seven putative MED-1 sites and which encodes a putative T-box transcription factor [28]. Consistent with med-1,2-dependent activation, tbx-35 transcripts are found in the MS cell, and a tbx-35::GFP reporter is expressed in the early MS lineage [32]. tbx-35 was also identified as an early MS-specific gene in a study that identified downstream embryonic targets of SKN-1 [33]. Recombinant MED-1 protein is able to shift fragments of the tbx-35 promoter, consistent with a direct interaction [32]. Mutant tbx-35(−) animals undergo developmental arrest as embryos or larvae, with the most extremely affected animals demonstrating a profound lack of MS-derived pharynx and body wall muscle [32]. The variable phenotype contrasts with the more stereotyped embryonic arrest of med-1,2(−) embryos, suggesting that many tbx-35(−) embryos can still make MS-derived tissues. Consistent with the existence of factors that work in parallel with TBX-35, we have found that the four embryonically-derived coelomocytes, which normally arise from MS descendants [1], are still present in tbx-35(−) embryos [Melissa Owraghi and M.M., unpublished observations]. Consistent with an ability to promote MS specification, overexpression of TBX-35 leads to the appearance of ectopic pharynx and body muscle [32] as well as coelomocytes [M.O. and M.M, unpublished]. Therefore, while TBX-35 is clearly an important regulator in MS specification, there must be additional factors that work in parallel.

3. Restriction of SKN-1 activity to EMS

Just as there are factors that assure timely activation of lineage-specific specification genes, other gene products act to prevent inappropriate activity of such factors in other cells. In the endomesoderm, these factors all act, collectively, in three functions: To restrict SKN-1 activity to the EMS nucleus, to prevent its translation in inappropriate parts of the embryo (or in the germline), and to promote its timely degradation. Hence, these components do not act directly in the network but are nonetheless vital for assuring its deployment in the EMS lineage alone.

3.1. PIE-1 and MEX-1

One of the factors that restricts SKN-1 function is the maternal factor PIE-1 [12]. In embryos lacking pie-1 function, P₂ develops like an EMS cell, generating SKN-1-dependent MS and E fates in its descendants [12]. Consistent with ectopic activation of the EMS specification pathway by the SKN-1 protein normally found in P₂, pie-1 mutant embryos display ectopic activation of med-1 and end-3 reporters in P₂ descendants [16, 25]. Whole-genome transcriptome analysis also detected elevated med-1,2 and end-1 transcripts in pie-1 mutant embryos, consistent with ectopic activation of the pathway downstream of SKN-1 [34]. PIE-1 is a CCCH zinc finger protein that is found in the P lineage, and functions by inhibiting transcription, explaining how the SKN-1 in P₂ normally does not activate EMS development [13, 35, 36].

At the other end of the embryo, the maternal factor MEX-1 restricts appearance of ectopic MS-like fates by preventing appearance of high levels of SKN-1 protein in the early AB lineage [10, 12, 37]. In embryos lacking mex-1 function, the AB granddaughters adopt MS-like fates, concomitant with ectopic expression of med-1,2 and tbx-35 [12, 25, 32, 38]. As with pie-1, function of SKN-1 is required for the appearance of ectopic MS-derived tissues in mex-1 mutant embryos, consistent with the ectopic activation of the normal MS specification pathway in the AB lineage [12]. MEX-1 is a CCCH-type zinc finger protein similar to PIE-1, and like PIE-1, is found in the P lineage, where it functions in PIE-1 localization [39]. Hence, the role of MEX-1 in preventing in AB-specific accumulation of SKN-1 is apparently indirect.

3.2. Regulation of OMA-1 degradation

A gain-of-function (gf) mutation in oma-1, a gene encoding another CCCH-type zinc finger protein, results in ectopic mis-specification of C, a somatic daughter of P₂ (Fig. 1), as an EMS-like cell [40]. In oma-1(gf) embryos, SKN-1 protein degradation is delayed compared with the wild type, and ectopic expression of a med-1::GFP reporter is observed in the early C lineage [40]. Other maternal proteins (e.g. PIE-1 and MEX-1) are found to perdure in oma-1(gf) embryos, consistent with a role for OMA-1 in negative regulation of timely degradation of cell fate specification factors in general [40]. A group of kinases (CDK-1, GSK-3, KIN-19 and MBK-2) has been identified that act upstream of OMA-1 [41, 42]. Loss of function of any of these components results in stabilization of OMA-1 and an ectopic SKN-1-dependent, ectopic endoderm phenotype similar to oma-1(gf) [41, 42]. Timely degradation of OMA-1 permits the proteolysis of cell fate determinants, dependent upon the Zinc Finger binding protein ZIF-1 [41]. These results provide an explanation for how loss of GSK-3 function was previously found to result in ectopic expression of a med-1 reporter in the C lineage, and specification of Cp as an E-like cell [25, 43]. As GSK-3 functions in post-embryonic regulation of SKN-1 nuclear localization, there may also be a more direct role for GSK-3 in regulation of SKN-1 activity in C, rather than only through regulation of OMA-1 stability [44].

3.3. Maintenance of germline totipotency

The germline must assure that cell fate pathways are not activated. Intestine-like cells have been observed within the germline in approximately 25% of animals lacking function of both mex-3 and gld-1, genes that encode RNA-binding proteins that function as translational repressors [45]. Although the appearance of intestine-like cells was not specifically associated with ectopic activity of SKN-1, transdifferentiation of other somatic cell types in this background was correlated with ectopic activation of other known tissue regulators (e.g., hlh-1 for body muscle, as discussed below), suggesting that this is likely to be the case [45].

4. Permissive functions for SKN-1 activity

Loss of function of either of two genes, pos-1 and spn-4, results in defects in EMS specification [46, 47]. In pos-1 mutant embryos, SKN-1 localization appears normal, but med-1::GFP is not expressed [25, 47]. spn-4 mutants also show normal SKN-1 localization, but apparently normal activation of a med-1 reporter, suggesting that spn-4 plays a different role in EMS development [46]. As pos-1 and spn-4 mutants have a number of other developmental defects outside of the EMS lineage, their roles would appear to be more permissive for EMS specification [46, 47]. Indeed, POS-1 and SPN-4 have been shown to interact, and have a role in regulating translation of maternal glp-1 mRNA [48].

5. Making MS and E different

The genes described above participate in specification of EMS as an endomesodermal precursor, but not in choosing between the alternate fates of mesoderm and endoderm. To make an MS or E cell from EMS, the SKN-1 pathway works with an evolutionarily conserved switching system that acts on sister cells to specify their fates as different, through a signaling cascade that is now called the Wnt/β-catenin asymmetry pathway [4, 49].

5.1. The Wnt/β-catenin asymmetry pathway and POP-1

When cultured in isolation, EMS divides asymmetrically to produce two MS-like cells, giving rise to excess pharynx muscle and no endoderm [50, 51]. When EMS is allowed to contact P₂ before it divides, it becomes polarized, such that the side of EMS that was in contact with P₂ will become the E cell, while its sister becomes MS [51]. Screens for mutations that could produce the same phenotype in intact embryos identified components of overlapping Wnt, MAPK and Src pathways [43, 52, 53]. The Wnt ligand MOM-2 and the receptor tyrosine kinase MES-1 function in parallel in the interaction between P₂ and EMS, though in slightly different ways: While MOM-2 is required only in P₂, MES-1 functions in both P₂ and EMS in an apparent dynamic interaction between the two cells [43, 53]. Through a network of downstream signal transduction, which also participates in the reorientation of the EMS spindle [54], the Wnt/MAPK and Src pathways ultimately converge on the differential localization and activity of the nuclear Wnt effector TCF/POP-1 [43, 55–57]. pop-1 was first identified by a maternal-effect mutation that results in the MS cell adopting the fate of E [58]. Antibody staining of POP-1 showed that it is widely expressed, and in sister cells that are born along an anterior-posterior cleavage axis, there is a higher level of POP-1 in the anterior daughter nucleus than the posterior one [59]. This ‘POP-1 asymmetry’ is the result of the nuclear export of POP-1 in Wnt-responsive cells [29, 60]. In addition to the MS/E decision, POP-1 has been found to be involved in a number of asymmetric cell divisions, including those of the postembryonic T hypodermal blast cell [61], and the Z1 and Z4 somatic gonad precursor cells [62]. Recently, it has been shown that the reduced levels of POP-1 in Wnt-signaled cells permits POP-1 to form a bipartite activator with limiting concentrations of the divergent β-catenin SYS-1, which shows a reciprocal posterior-anterior asymmetry compared with POP-1 [63–65]. POP-1 thus forms part of a binary switching system that can establish transcriptional differences of lineage-specific factors.

5.2. Multiple functions for POP-1

How does POP-1 participate in the MS/E decision? The phenotype of pop-1 loss is a transformation of MS to E, and the ectopically-specified MS blastomeres in mex-1 and pie-1 mutant embryos adopt an E-like fate when pop-1 function is simultaneously lost [58]. These results suggest that the main function of POP-1 is to repress endoderm specification in MS. Indeed, in situ and transgene reporter assays show that end-1 and end-3 are activated in the MS and E lineages in pop-1 mutant embryos [16, 20, 66]. A GFP-tagged form of POP-1 can interact in vivo with extrachromosomal arrays carrying the end-1 or end-3 promoters, suggesting that this repression is direct [29], and a repressor complex that includes the Groucho homolog UNC-37 mediates repression of end-1 in MS [67].

Multiple studies have shown that POP-1 also promotes endoderm fate in the E cell in parallel with SKN-1 and MED-1,2. Genetically, depletion of pop-1 greatly enhances the endoderm phenotype of skn-1 and med-1,2 mutants, suggesting that POP-1 functions in parallel with these factors to promote E fate [16, 65, 68]. Consistent with a role in activation of endoderm when these genes are not mutated, expression of multiple E-specific reporter transgenes, including end-1 and end-3, occurs in both the MS and E lineages at reduced levels in pop-1 mutants as compared with expression in wild types [66]. An end-1::GFP reporter requires optimal TCF/POP-1 binding sites to exhibit POP-1-dependent repression in MS and activation in E, confirming that POP-1 likely exerts its effects through direct interaction with end-1 [66]. As predicted by association of SYS-1 with POP-1 to form a bipartite activator in E, depletion of sys-1 significantly enhances the endoderm-defective phenotype of skn-1 mutant embryos [64, 65]. Accumulation of endogenous end-1 transcripts is slightly reduced in pop-1 mutant embryos, and almost eliminated in end-3; pop-1 mutants, confirming overlapping roles of both POP-1 and END-3 in end-1 activation in E [20]; as an aside, this synergistic effect on end-1 activation is consistent with greatly enhanced gutlessness in end-3; pop-1 mutants [16].

There is further evidence that POP-1 is capable of additional functions besides repression of endoderm in MS and contribution to activation of endoderm fate in E. Expression of a tbx-35::GFP reporter [33] is diminished in MS, and activated in E, in pop-1 mutant embryos, suggesting that some gene(s) might be able to respond to POP-1 in a reciprocal manner from end-1,3 [Premnath Shetty and Rueyling Lin, personal communication]. However, expression of endogenous tbx-35 transcripts, and a slightly different tbx-35::GFP reporter, are apparently unaffected by loss of pop-1, suggesting that there may be additional nuclear factors that respond to Wnt signaling [32].

To test for a possible requirement for pop-1 in MS specification beyond repression of end-1,3, we examined the phenotype of end-1,3(−) embryos in which pop-1 function was eliminated by RNAi. Preliminary results suggest that both MS and E adopt some properties of MS in pop-1(RNAi); end-1,3(−) embryos, suggesting that POP-1 is not strictly required to initiate a program of MS development [Melissa Owraghi and M.M., unpublished observations]. The apparent adoption of MS-like properties by the E cell in such embryos contrasts with the penetrant transformation of E to C in end-1,3(−) embryos [15, 16], and suggests a cryptic MS-repressive role for POP-1 in the E cell. Future work will no doubt shed light on these additional roles.

6. Moving from lineage- to tissue-based gene networks

As development proceeds, two types of specification mechanisms are thought to drive the process forward: Those that are lineage-based, in which mutation affects all descendants of a cell irrespective of the tissue types it produces, and those that are organ/tissue-based, in which mutation results in a defect in an entire organ/tissue irrespective of its lineal origin [69]. Global expression studies have identified several hundred genes activated in the pharynx [70], several thousand in the intestine [71, 72], and more than a thousand in muscle [73]. Activation of these tissue-specific networks results from the activation, in multiple lineages, of a small number of organ/tissue identity factors. In contrast with early cell fate specification genes, most of these factors remain activated through the lifespan of animal. These will be described only briefly below, as there are recent reviews that cover these networks [74–76].

6.1. Intestine fate: ELT-2 and gut identity

The GATA factor ELT-2 appears to be the intestine identity factor. elt-2 expression begins in the E daughter cells, downstream of end-1 and end-3, and continues through adulthood in all intestinal cells [16, 17, 77]. Consistent with the placement of ELT-2 at the very top of an intestine differentiation network, overexpression of elt-2 can drive specification of gut throughout the embryo, similar to ectopically-expressed end-1 or end-3 [16, 17, 77], and it maintains its own expression through positive autoregulation [78]. Two independent studies found that the vast majority of genes that are activated in the intestine have a recognizable GATA binding site, with a core sequence of TGATAA [71, 72]. Recombinant ELT-2 has been repeatedly shown to be able to interact with GATA sites required for ELT-2-dependent expression of numerous intestinal genes, for example ges-1, ftn-1, pho-1 and the intestine-specific, Notch-dependent component of ref-1 [79–82]. elt-2 has also been shown to be required for an innate immune response to microbial pathogens such as Pseudomonas [83, 84].

Within the intestine, ELT-2 works in combination with other genes to generate more restricted patterns of gene expression, either autonomously within the gut, or as a result of external inductions. For example, ELT-2 and POP-1 appear to collaborate to restrict expression of pho-1 to the posterior gut, or for anterior-specific expression of a deleted ges-1 promoter::reporter construct [81, 85]. An increase in the SYS-1::POP-1 ratio by an early E-lineage transgene can result in changes in anterior pho-1 expression consistent with a role of the Wnt/β-catenin asymmetry pathway in patterning expression within the gut [64]. Extrinsic cell signaling also interacts with gut-intrinsic factors, as Notch signaling and ELT-2 collaborate to activate expression of ref-1 in the left side of the primordial gut at the 4E and 8E stages, and the right side at the 16E stage [79].

There is evidence that, at least in the embryo, ELT-2 activates intestinal development with at least one other gene: Loss of elt-2 does not prevent gut formation in the embryo, though it does result in larval lethality from a failure to maintain gut integrity [77]. In earlier models of the endoderm specification pathway, it was hypothesized that another intestinal GATA factor, ELT-7, might work in parallel with ELT-2 in the early embryo [18]. This hypothesis made sense in light of the fact that elt-7 is apparently coexpressed with elt-2 through adulthood [18]. However, loss of either elt-7 or elt-4 (a partial duplication of elt-2) does not detectably affect intestine development, and only a slight enhancement of the elt-2 phenotype is evident in an elt-2; elt-7 double mutant, not a complete loss of gut [72]. Instead, perdurance of END-1 or END-3, or both, might be responsible for activating early expression of downstream intestine-specific factors [74]. This is plausible, as glo-1 and pgp-2, both of which function in gut granule biogenesis, commence transgene expression in the two E daughter cells prior to activation of an elt-2 reporter, suggesting that at least some gut-specific ‘differentiation’ factors are targets of END-1,3 [77, 86, 87]. As these regulators are all GATA factors, they would be expected to be able to bind to the same target sites. Hence, early intestine gene expression could be initiated by END-1/3 and ELT-2, and maintained by ELT-2 later, with some function contributed by ELT-7.

6.2. MS and pharynx, body muscle fates

Unlike E, the MS blastomere generates descendants of very different types, including pharynx cells, body muscle cells, the four embryonically-derived coelomocytes, the somatic gonad precursors Z1 and Z4, and even some neurons [1]. The gene networks that drive specification of these various cell types are undoubtedly complex, though there are two major differentiation pathways that can be considered: Of the 80 cells made by MS in the embryo, 28 are body muscle cells and 31 are pharynx cells, which together comprise the majority of the embryonic MS descendants [1].

The regulator FoxA/PHA-4 was identified by mutations that resulted in the absence of pharynx [88, 89]. Consistent with the placement of PHA-4 at the top of a network that specifies pharynx fate, ectopic PHA-4 is sufficient to induce formation of ectopic pharynx tissue [88]. pha-4 is also expressed throughout the intestine and rectum (Fig. 2), and is required for rectum development, though pha-4 loss does not profoundly affect gut development [89]. The pharynx itself contains numerous cell types, which includes muscles, neurons and epithelia [76], implying that there are additional factors that work with PHA-4 to generate the pharynx. Within pharynx muscle, the gene myo-2, which encodes a pharynx-specific myosin isoform, is regulated by distinct cis-regulatory modules that function within distinct pharynx muscle groups [90]. Both PHA-4 and the homeodomain transcription factor CEH-22 recognize cis-regulatory sites in myo-2, but whereas pha-4 is expressed throughout the pharynx, ceh-22 is activated only in pharynx muscle [89, 91]. Another factor, PEB-1, is found both within and outside the pharynx, but also contributes directly to myo-2 regulation [92]. The pharynx gland-specific gene hlh-6 contains three regulatory elements, one of which binds PHA-4, and all three cis-regulatory modules work in concert to produce cell type-specific activation [93]. Multiple other pharynx regulators have been identified, suggesting that pharynx organogenesis in general involves the activation of many complex sub-networks [70, 76, 94].

Development of body muscles has been shown to be the result of three-way redundancy among the factors HLH-1, HND-1 and UNC-120 [34, 95, 96]. Consistent with overlapping function, loss of any one of these factors produces only mild muscle phenotypes, but loss of all three together results in a profound failure of muscle specification [95]. Similarly, overexpression of hlh-1, hnd-1 or unc-120 can specify cells as muscle precursors [95, 97]. Within the C lineage, the factor Caudal/PAL-1 is proposed to activate zygotic pal-1, which activates a ‘muscle module’ consisting of hnd-1, unc-120 and hlh-1, specifying muscle progenitors [34, 75, 95]. HND-1 is proposed to act in early embryogenesis, participating in activation of hlh-1, while hlh-1 and unc-120 act later, mutually enforcing their expression [75]. Enrichment of embryonic muscle genes has identified more than 1300 muscle-enriched genes, suggesting that just as with pharynx and intestine, there are complex sub-networks that remain to be identified [98].

The basis for how the muscle and pharynx tissue networks are activated in appropriate MS-derived precursors, downstream of MED-1/2 and TBX-35, are not yet fully understood. It is likely that the Wnt/β-catenin asymmetry pathway plays a role based on multiple observations. First, POP-1 asymmetry is found in the MS daughters and grand-daughters [29, 59], and lower levels of POP-1 are permissive for specification of muscle fates by ectopic HLH-1 [97]. Second, transgenic SYS-1 in the early MS lineage can produce apparent MSa to MSp fate transformations that are manifested as changes in muscle fates within the pharynx [64]. Third, loss of the divergent β-catenin WRM-1, required for Wnt-dependent modification of POP-1, causes defects within the MS lineage [52]: Mutants in lit-1, a gene whose product works with WRM-1 to modify POP-1, show posterior-to-anterior transformations in multiple lineages, including MS [99]. These transformations are consistent with a fourfold, rather than twofold, increase in the number of ceh-22::GFP-expressing cells produced from E+MS in wrm-1(RNAi) embryos [32].

Analogous to the MS/E decision, there must be factors that work in combination with the Wnt/β-catenin asymmetry pathway to segregate fate potential within the early MS lineage. The C cell, cousin to MS and E, generates primarily muscle fates among its posterior granddaughters (Cxp), similar to the muscle fates made by the MS posterior granddaughters (MSxp) [1]. PAL-1, which normally promotes C fate, is found in the early MS and E lineages as well [68, 100]. An intriguing possibility, therefore, is that PAL-1 contributes to muscle specification in the MS lineage. Other candidates for such factors might also be found among the genes identified by their expression in the early EMS lineage [21, 33].

7. Conservation of endomesoderm genes among nematodes

In perhaps the most well-characterized gene regulatory network, that of the sea urchin endomesoderm [101], evolutionary changes are driven by mutations in cis-regulatory sites [102]. This phenomenon seems to be generalizable as in multiple taxa, there are many examples of evolution being driven by changes in cis-regulatory sites [103]. Based on the extant C. elegans endomesoderm network, is it possible to make any conclusions about how it may have evolved? In a recent review, homologs of genes in the network were sought using genome sequence information from related species [3]. In the nematodes Haemonchus contortus and Brugia malayi, maternal factors and tissue/organ identity factors were found to be the most conserved [3]. This might be expected, as SKN-1 and POP-1, have additional roles in the animal [49, 59, 104], and the organ identity factors (e.g. PHA-4, ELT-2) have so many targets that it would seem unlikely that their functions could be transferred to another regulator. There are substantial differences in the way nematodes apportion fates to early blastomeres, although later embryogenesis tends to be more similar [7]. For example, in Romanomermis, it is the AB blastomere that produces gut [105], while in Acrobeloides, endoderm fate can be reassigned to another blastomere if the AB cell is ablated [106]. These differences mean that gene networks upstream of tissue/organ identity factors would be expected to be the most divergent among nematode species.

Within the ‘Elegans group’ in the Caenorhabditis genus (elegans, briggsae, sp. 5, remanei, and brenneri), there are in most cases one-to-one homologs for all the known genes in the endomesoderm network, although duplications appear to be occurring more frequently among the early zygotic regulators [3, 107]. For example, elt-4 is a partial duplication of elt-2 that is found only within C. elegans, and C. briggsae carries a very recent, inverted duplication of end-3 [16, 108]. In the nematode Pristionchus pacificus there appears to be a pair of linked end-like genes whose transcripts accumulate in the early E lineage, suggesting that the end-1,3 pair is at least as old as the time to the common ancestor of Caenorhabditis and Pristionchus (George Hsu and M.M., unpublished).

The med genes show a unique pattern of evolution. First, outside of Caenorhabditis, there are no known med-like GATA factor genes at all, suggesting that they evolved very recently [31]. Second, all known med genes are intronless [31]. Third, in contrast to the other nematode GATA factors, the meds are undergoing rapid duplications: While C. elegans has only two med paralogs, the meds in C. remanei number at least seven genes, and in C. briggsae, at least three [31]. There is no reason to think that the frequency of duplications of med genes should be any different than other genes in the network, suggesting that some other selective pressure, as yet unknown, is maintaining the duplicates.

8. Evolution of the network

With the limited studies done in other nematodes, it is premature to make any conclusions about how connectivity, gene hierarchies and robustness arose in the C. elegans endomesoderm network. Here, we will speculate as to how parts of the network might have evolved by gene duplication and loss/gain of transcription factor binding sites, and mention evidence for flexibility in how the Wnt/β-catenin asymmetry pathway functions in the MS/E decision.

8.1. Intercalation of genes into a pathway

Studies of gene networks in other systems find that particular types of local connections, or motifs, occur at a high frequency [109]. Within the C. elegans endomesoderm network, many of these types of motifs, such as the regulatory chain motif, can be found [3]. Based on recent studies of motif evolution in the yeast transcriptional network [110], two speculative models are presented below to account for the formation of two motifs within the C. elegans endomesoderm network (Fig. 3). The evolutionary changes proposed are plausible based on findings in other systems [102, 103, 110].

FIG. 3.

Speculative models for stepwise evolution of two extant motifs in the C. elegans endomesoderm network, starting with single activator, single target interactions. (A) Stepwise formation of the SKN-1 → MED-1,2 → TBX-35 regulatory chain. (B) Formation of a feed-forward loop involving SKN-1 and two ENDs. Note that autoregulation of the first end gene is inferred but this is not known for end-1. Intercalation of the MEDs could occur at an early or later step by acquisition of MED sites in the end genes (not shown).

Consider first the regulatory chain of SKN-1 activating med-1,2, which then leads to activation of tbx-35 (Fig. 3A) [32]. Given that the meds are not found outside of Caenorhabditis, the ancestral pathway might very well have been direct activation of tbx-35 by SKN-1. If a med-like gene arose from duplication of a GATA factor, this gene could have acquired sites for SKN-1, forming a single-input motif. Next, tbx-35 could have acquired sites that allowed it to be recognized by the MED product, forming a feed-forward motif. If tbx-35 then lost the original sites for SKN-1, while the number of sites for MED-1 was increased, the result would be the regulatory chain SKN-1 → MED-1 → TBX-35. Duplication of med-1 would result in the extant pathway in C. elegans.

Next, consider the feed-forward loop whereby SKN-1 activity leads to activation of end-3, and SKN-1 and END-3 together activate end-1 (Fig. 3B) [20]. By analogy to the first example, the ancestral pathway could be taken to be SKN-1 activating a single ancestral end gene directly. A duplication of the ancestral end gene occurs. If one of the end genes then acquires additional sites for positive autoregulation, a feed-forward loop can be established. The intercalation of MED-1,2 could have occurred prior to end duplication (similar to the example above), or afterwards, to produce the relationships in the extant network.

8.2. Evidence for flexibility in how POP-1 influences the MS/E decision

In the above examples, the initial input (SKN-1) and final output (MS or E specification) remain the same after the pathways have undergone changes, and the evolutionary steps required to change motif types are simple enough that they might occur over short time periods. Post-embryonically, changes in relative importance of overlapping signal transduction pathways has been demonstrated in the vulval lineages, despite no apparent differences in developmental output, suggesting that such informational connectivity changes can and do occur [111].

We have begun to look for cryptic differences in the endomesoderm network in C. briggsae. C. elegans and the related species C. briggsae diverged approximately 100 million years ago, but their embryonic cell lineages have remained highly similar [112, 113]. Both genomes encode a single pop-1 orthologue. While loss of pop-1 in C. elegans leads to a mis-specification of MS as an E-like cell [58], RNA interference of C. briggsae pop-1 leads to a loss of endoderm, and an apparent transformation of E to an MS-like cell [Katy Lin and M.M., results to be reported elsewhere]. Surprisingly, the two end-3 paralogues in C. briggsae are not expressed ectopically in the MS lineage in Cb-pop-1(RNAi) embryos, and are in fact almost undetectable compared with wild-type embryos [Gina Broitman-Maduro and M.M., unpublished]. One explanation is that the positive input by POP-1 into endoderm specification, apparent in C. elegans only under certain conditions, is now a primary input in C. briggsae. Similarly, the role for POP-1 in repression of the end genes in MS might simply have been lost, if the Cb-end genes were no longer as responsive to the SKN-1 pathway due to changes in cis-regulatory sites. Regardless of the molecular details, it is clear that there is flexibility in how factors such as POP-1 participate in the same embryonic cell fate decision in related species. In other developmental events in nematodes, such differences thought to underlie the basis for robustness [114]. The prediction is that other such cryptic differences in phenotypes will be discovered as more comparative studies are performed.

9. Conclusion

When it comes to gene regulatory networks, the C. elegans embryonic endomesoderm network seems to have it all. Gene interactions have been characterized by forward genetics, reverse genetics, whole-genome transcriptional profiling, bioinformatics, transgene analysis and biochemistry. The network is replete with parallel pathways, redundant genes, combinatorial specification, feed-forward loops, regulatory chains, maternal genes, zygotic genes, repression, activation, cell-autonomous mechanisms and cell-cell interactions. Parallel pathways and redundancies almost certainly contribute to the robustness of the network, and they also enable alternate gene interactions to arise over time, even in the absence of any obvious outward phenotypic change. Other gene networks tend to use similar motifs and gene interactions, e.g. [109, 115, 116], suggesting that comparative studies with C. elegans and other related animals will yield new insights into their evolution. More specifically, it will be interesting to compare embryonic gene regulatory networks from related animals outside the nematode phylum, to test hypotheses of the mechanisms seem to be the main driving forces within the protostomes. To this end, studies in other ecdysozoa that partition fates early, such as the tardigrade Hypsibius dujardini, may prove fruitful [117].