pha-4 is Ce-fkh-1, a fork head/HNF-3α,β,γ homolog that functions in organogenesis of the C. elegans pharynx

Organogenesis is a complex and critical feature of animal development, involving cell fate specifications, cell interactions, tissue assembly and intricate morphogenesis. During the past several years, key regulatory molecules have been identified in the formation of organs as diverse as insect eyes (Quiring et al., 1994) and mammalian livers, hearts and kidneys (Zaret, 1996; Fishman and Chien, 1997; Vainio and Muller, 1997). Despite this wealth of new information, it is far from clear whether different organs in different animals obey common developmental rules.

The Caenorhabditis elegans pharynx provides a relatively simple experimental system to study the molecular basis of organogenesis. The pharynx is a neuromuscular feeding apparatus, containing 80 nuclei within five distinct cell types: muscles, neurons, gland cells, epithelial cells and structural cells called marginal cells (Albertson and Thomson, 1976). Two distinct developmental pathways involving maternal-effect genes and precise cell-cell interactions act in the earliest stages of pharynx formation: an ‘inductive’ pathway, by which cells of the ABa lineage are induced to produce the anterior pharynx, and an ‘autonomous’ pathway, by which cells of the MS lineage produce the posterior pharynx (reviewed in Schnabel and Priess, 1997). What happens next is not clear: how do the apparently undifferentiated cells of the pharynx primordium adopt the five distinct cellular phenotypes and assemble into the mature functioning organ?

One could imagine two extreme models to explain pharyngeal organogenesis. In one model, the five distinct cell types differentiate and assemble into the organ without expressing a common organ-specific regulator. In the opposite extreme, all cells that form the organ could express an ‘organ identity factor’ that coordinates development of the different cell types and directs their assembly into the final organ. The properties of such an organ identity factor should include the following: (1) in the absence of the factor, organ formation should be blocked at the earliest stages; (2) the factor should be present in all cells of the organ, beginning at the earliest stages of organ formation; (3) the factor should lie downstream of (and possibly integrate) pathways that are known to produce the organ primordium (in the case of the pharynx, the ‘inductive’ ABa and ‘autonomous’ MS pathways); and (4) ectopically expressed factor might induce ectopic organ formation, or at least ectopic expression of organ-specific genes.

Two candidates have already been proposed for zygotically produced pharynx-specific organ identity factors: PHA-1 and PHA-4. The important common feature of these two genes is that they influence formation of the pharynx as an organ, not as individual cell types, particular cell lineages or embryonic domains. In embryos mutant for the pha-1 gene, the pharynx primordium forms but does not elongate correctly; some early pharynx differentiation can be detected but later differentiation is abolished (Schnabel and Schnabel, 1990; Granato et al., 1994). pha-1 encodes a protein with similarity to bZIP transcription factors, suggesting it can act as a transcriptional regulator (Granato et al., 1994). Embryos mutant for the zygotically active pha-4 gene arrest pharynx formation at an even earlier stage than do pha-1 embryos; essentially no markers characteristic of the pharynx are expressed (Mango et al., 1994). Mango et al. (1994) showed that pha-4 acts genetically upstream of pha-1 but downstream of the maternal-effect genes that specify early blastomere fate.

In the present paper, we identify the pha-4 gene as Ce-fkh-1, a gene that we have previously cloned (Azzaria et al., 1996) based on its high degree of similarity (approx. 75% amino acid identity) to the winged helix DNA-binding domains of the Drosophila fork head gene and the mammalian genes HNF-3α,β and γ. Mutants in Drosophila fork head have severe defects in digestive tract development, in particular the foregut and hindgut (Weigel et al., 1989a,b). HNF-3α,β and γ genes were originally identified as liver-specific transcription factors in adult rodents (Lai et al., 1990) and have subsequently been shown to be expressed in endoderm and axial structures of the early mouse embryo, as well as in the organizing centers necessary for axis formation (Ang et al., 1993; Monaghan et al., 1993; Ang and Rossant, 1994; Weinstein et al., 1994). The present study describes common features of expression patterns and mutant phenotypes for this set of fork head/HNF-3α,β,γ/pha-4 genes and provides evidence for a highly conserved pathway of digestive tract formation in all (triploblastic) metazoa. However, our most important conclusion is that, during formation of the C. elegans pharynx, pha-4 conforms to the criteria set out above for an ‘organ identity factor.’

The following C. elegans strains were used: wild type, Bristol N2; JK892, unc-32(e189) glp-1(q231ts) III; JJ1057, pop-1(zu189) dpy-5(e61)/hT1 I; him-5(e1490)/hT1 V; EU1, skn-1(zu67) IV/nT1[unc-?(n754) let-?] (IV;V); JJ529, rol-1(e91) mex-1(zu121)/mnC1 [dpy-10(e128) unc-52(e444)] II; JJ532, pie-1(zu154) unc-25(e156)/qC1 [dpy-19(e1259) glp-1(q339)] III; JK1521, fog-2(q71) pha-4(q490)/stu-3(q265) rol-9(sc148) V; GE42, vab-7(e1562) pha-1(e2123ts) III.

Embryos produced by skn-1, mex-1, pie-1 and pop-1 homozygous hermaphrodites were collected at room temperature. Homozygous glp-1(q231ts) and pha-1(e2123ts) worms were shifted from 16°C to 25°C as young adults and embryos collected at 25°C.

DNA constructs were transformed into N2 worms as described by Mello et al. (1991), using either unc-22 antisense or rol-6 as transformation markers (Fire et al., 1991; Mello et al., 1991). At least three transformed lines, usually more, were examined for each construct. When appropriate, transformed constructs were integrated into the genome by γ-irradiation to make 100% heritable lines (Egan et al., 1995).

To determine if pha-4 and fkh-1 are the same gene, single embryo PCR was performed on homozygous pha-4(q490) embryos, identified among the progeny of JK1521 by the Pha-4 phenotype; primers were designed to amplify portions of the known Ce-fkh-1 genomic sequence (Azzaria et al., 1996). Both strands of the entire coding region were sequenced; only the single mutation described in the text was found.

A 175-base-pair BamHI-HincII fragment from pha-4 cDNA (Ce-fkh-1 coordinates 1173-1347 in Azzaria et al., 1996), encoding 58 amino acids located towards the C terminus end, outside of the DNA binding domain, was cloned in-frame into the BamHI-SmaI sites of pGEX-1 (Pharmacia Biotechnology) as well as into the SmaI site of pTRXfus (Invitrogen). A GST/PHA-4 fusion protein was induced and purified following manufacturer’s instructions and used to immunize rabbits following standard injection protocols (Harlow and Lane, 1988). Thioredoxin/PHA-4 fusion protein was induced and purified by osmotic shock, following manufacturer’s instructions, and used to affinity-purify PHA-4 antibodies from immunoblots (Harlow and Lane, 1988). Antiserum specificity is demonstrated by lack of staining in pha-4 mutants and by extensive ectopic staining when a transgenic pha-4 cDNA construct is activated by a heat-shock promoter (data not shown). Immunofluorescence analysis on embryos, larvae and adults was performed essentially as described in Miller and Shakes (1996). Primary antibodies were added for an overnight incubation either at 4°C or room temperature; slides were incubated for 2 hours at room temperature with a secondary antibody, either Cy3-labelled goat anti-rabbit IgG or FITC-labelled donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories). Pharyngeal muscle cells were identified by double staining myo-2::lacZ (cc228) larvae and adults (Okkema and Fire, 1994) with antibodies against PHA-4 and β-galactosidase (Promega). Rectal epithelial cells were identified by double-staining worms transformed with the enhancer trap construct pUL#38E12 (Hope, 1991), which expresses β-galactosidase in rectal epithelial cells (Mango et al., 1994). Other cell types were identified by position. Animals were stained for β-galactosidase activity as described in Fire (1992). PHA-4-staining cells were counted using deconvolution confocal microscopy.

Full-length pha-4 cDNA (corresponding to polypeptide I of Azzaria et al., 1996), was placed under the control of the hsp16-2 promoter (Stringham et al., 1992) in the vector pPD49.78 (Mello and Fire, 1996). The construct was transformed into worms and integrated into the genome by γ-irradiation to produce the strain JM70. Males from this strain were crossed to: (1) OK0039 (cuIs2), a transformed line containing integrated arrays of GFP under the control of eight copies of the C183 oligonucleotide (Okkema and Fire, 1994), and (2) PD8260 (ccIs8260), a line with integrated arrays of a ceh-22::lacZ fusion (Okkema and Fire, 1994). Since all transgenic arrays have rol-6 as a marker, the presence of transgenes was determined either by single worm PCR using transgene specific primers, or by directly examining embryos for reporter gene expression and for lethality upon heat shock. Embryos from hermaphrodites transgenic for hsp16-2::pha-4 and either reporter transgene were aged 30 to 60 minutes after being laid, incubated at 33°C for 1 hour and then allowed to recover at 16°C for approx. 7 hours before observing GFP expression or staining for β-galactosidase activity. This procedure results in developmentally arrested embryos with PHA-4 protein located in the majority of nuclei (data not shown).

As a control, the above crosses were repeated using strain JM53 (Fukushige et al., 1998), which contains an integrated transgenic array of a hsp16-2::elt-1 cDNA fusion instead of hsp16-2::pha-4. ELT-1 is a GATA transcription factor (Spieth et al., 1991) involved in hypodermal development but not pharyngeal development (Page et al., 1997).

For experiments in which the monoclonal antibody 9.2.1 was used to detect ectopic pharyngeal myosin (Miller et al., 1986; Ardizzi and Epstein, 1987), 2-to 4-cell stage hsp16-2::pha-4 embryos (JM70) were collected, aged 6 hours at 20°C, heat-shocked at 33°C for 2 hours, and allowed to recover at 20°C for 10 hours before staining. Control N2 embryos and embryos containing the hsp16-2::elt-1 cDNA construct were treated identically.

pha-4 reporter gene expression driven by ectopic ELT-2 was examined as described in Fukushige et al. (1998) using embryos from worms carrying two independently integrated arrays: hsp16-2::elt-2 (from JM57) and pha-4::lacZ from strain JM50 (containing pJM50; see below).

PHA-4 protein was produced by a coupled transcription-translation system (Promega T_NT) using as template pha-4 cDNA (encoding polypeptide I) cloned into pT7TAG (Pollock and Treisman, 1990). The electrophoretic mobility shift assay was performed essentially as described in Stroeher et al. (1994) using ³²P-end-labelled (double-stranded) C183 (Okkema and Fire, 1994), 100 ng poly(dIdC) non-specific competitor and a 40-fold excess of either specific (C183, double-stranded) or mutated (Cmut2, double-stranded) competitor.

Full-length ELT-2 (Hawkins and McGhee, 1995) was also produced by in vitro transcription/translation and the electrophoretic mobility shift assay conducted as above, with the following modifications. The wild-type probe sequence was 5′-GTGTCAAACTCTCTTATCAA-TCAGTGCTCTTTTTCTCT; the mutated probe sequence was 5′-GT-GTCAAACTCTCGGCGACATCAGTGCTCTTTTTCTCT (where the GATA and mutated sites are depicted in bold); both wild type and mutant competitors (double-stranded) were present in an 80-fold excess over the probe (double-stranded).

Uni-directional deletions from the 5′ end of the 7 kb pha-4 promoter were made as in Aamodt et al. (1991) beginning with plasmid pJM50 (a 9.0 kb BamHI-EcoR1 pha-4 fragment fused to lacZ at the second pha-4 exon inserted into the vector pPD22.11 (Fire et al., 1990); pJM50 was referred to as Y4 in Azzaria et al. (1996)). Internal deletions were made by PCR overlap extension (Horton et al., 1993).

Ce-fkh-1 maps physically (Azzaria et al., 1996) and pha-4 maps genetically (Mango et al., 1994) to the right side of C. elegans chromosome V. Because of the pharyngeal and rectal defects found in pha-4 mutants (Mango et al., 1994) and because Ce-fkh-1 is expressed throughout the embryonic digestive tract (Azzaria et al., 1996), it seemed possible that the two genes were one and the same. We sequenced the strong pha-4 allele q490 provided by S. Mango (University of Utah, USA); a single mutation (CAG→TAG; Q→Stop) was detected at amino acid 99 of the longest coding sequence (Azzaria et al., 1996). This mutation leads to truncation well upstream of the DNA binding domain of all three polypeptides predicted for this gene. Horner et al. have shown that other pha-4 alleles also cause changes in the Ce-fkh-1 coding sequence and that Ce-fkh-1 RNA interference injections produce a pha-4 phenocopy (M. A. Horner, S. Quintin, M. E. Domeier, J. Kimble, M. Labouesse and S. Mango, unpublished). We conclude that pha-4 and Ce-fkh-1 are the same gene, hereafter referred to as pha-4.

Fig. 1 summarizes the cell lineage of the C. elegans pharynx (Sulston et al., 1983). By the comma stage (approx. 560 total cells and 430 minutes post-fertilization at 20°C), the final cell divisions that produce the 80 nuclei of the pharynx have taken place and the pharyngeal precursor cells cluster together to form the undifferentiated pharynx primordium (Sulston et al., 1983; Schnabel and Schnabel, 1990; Mango et al., 1994; Schnabel et al., 1997). Fig. 2A shows a typical comma-stage wild-type embryo (430 minutes) stained with an affinity-purified rabbit polyclonal PHA-4 antibody; the DAPI image of the same embryo is shown in Fig. 2B. In six comma-stage embryos, we counted an average of 80.5 (± 2.5 s.d.) PHA-4-staining pharyngeal nuclei by confocal microscopy (maximum count = 85). Analysis of later stage animals (see below) shows clearly that all six cells of the pharyngeal-intestinal valve (Albertson and Thomson, 1976) contain PHA-4 protein; Mango et al. (1994) have previously noted valve defects in pha-4 embryos. Thus the total number of potentially staining nuclei (pharynx + valve) should be equal to 86. We conclude that, within the error associated with counting closely apposed nuclei, all cells of the pharyngeal primordium (+ valve) contain the nuclear factor PHA-4. The same embryos also show nuclear PHA-4 in 6-8 rectal cells, including the two rectal valve cells and the three rectal epithelial cells. At the lima bean and comma stages of embryogenesis, a low but significant level of PHA-4 can be detected in the gut (Fig. 2C,D).

Do pharyngeal precursor cells express pha-4 prior to the formation of the pharynx primordium? Fig. 2E,F shows an embryo that contains 177 DAPI-staining nuclei and hence corresponds to approx. 200 minutes of development (aligned as embryo 2E on the cell lineage in Fig. 1). At this stage, there are 28 cells whose lineages will ultimately give rise to pharyngeal cells (including the pharyngeal-intestinal valve cells) and 24 of these lineages are clonal, i.e. will give rise only to pharyngeal cells (Sulston et al., 1983). The embryo shown in Fig. 2E contains 21 nuclei with detectable levels of PHA-4 protein, as counted by confocal microscopy. In embryos one cell division earlier (approx. 100 total cells; 16 pharyngeal precursor cells, of which 11 give rise only to pharyngeal cells), we can detect PHA-4 in 3-10 nuclei but at very low levels (data not shown). Fig. 2G,H shows an embryo at approx. 250 minutes of development, when there are 52 pharyngeal precursor cells of which 49 give rise to only pharyngeal cells; 40-50 strongly PHA-4-staining nuclei were counted in embryos at this stage. From these and similar observations, we conclude that the pha-4 gene is expressed in pharyngeal precursor cells well before the formation of the pharynx primordium. Our data are consistent with a model in which PHA-4 first appears at the point when pharyngeal lineages become clonal and produce only pharyngeal precursor cells, independently of the particular pharyngeal cell type. However, considering the natural variability of normal embryogenesis (Schnabel et al., 1997), this assertion must be verified by direct lineage analysis of individual embryos.

PHA-4 can be detected immunologically in most but not all nuclei of the larval and adult pharynx. PHA-4 is detected in nuclei of all epithelial cells, muscle cells, marginal cells, gland cells and pharyngeal intestinal valve cells, but is detected only in about 8 of 19 neuronal nuclei (Fig. 2I,L). Based on reduced counts of total staining nuclei, we suggest that these neuronal cells had ceased pha-4 expression at mid to late embryogenesis.

PHA-4 is also detected in one (comma stage) to several (pretzel stage and later) head cells outside of the pharynx (data not shown). As expected from previous transgenic analysis (Azzaria et al., 1996), PHA-4 can also be detected in nuclei of the developing somatic gonad, including the distal tip cell and ventral uterine cells (data not shown).

PHA-4 distribution was examined in embryos mutant for maternal-effect genes that establish the two pathways of pharynx production (Schnabel and Priess, 1997). In glp-1 embryos, the induction of anterior pharynx is abolished (Austin and Kimble, 1987; Priess et al., 1987); as shown in Fig. 3A,B, PHA-4 is detected only in the posterior pharynges of glp-1(q231ts) embryos. The autonomously produced posterior pharynx is abolished in pop-1 embryos (Lin et al., 1995); Fig. 3C,D show that PHA-4 is detected only in the anterior pharynges of pop-1(zu189) embryos, plus rectal cells. In skn-1 embryos, no pharynx forms (Bowerman et al., 1992); as shown in Fig. 3E,F, no PHA-4 is detected in the pharynx of skn-1(zu67) embryos, but rectal staining persists as expected. mex-1 or pie-1 mutations cause the production of ectopic pharyngeal cells (Mello et al., 1992); excess PHA-4 staining is detected in mex-1(zu121) (Fig. 3G,H) and pie-1(zu154) embryos (data not shown). In summary, the pha-4 expression patterns detected in all the above mutants suggest that pha-4 must lie downstream of the convergence of the two distinct inductive and autonomous regulatory pathways determining the pharynx primordium. The distribution of PHA-4 molecules is thus completely consistent with genetic epistasis tests (Mango et al., 1994). Neither mutant nor wild-type embryos show obvious differences in the PHA-4 level or appearance times in the ABa-derived or MS-derived cells of the pharynx.

pha-4 expression was examined in embryos mutant for the zygotic gene pha-1, which is required for pharyngeal morphogenesis and expression of late pharyngeal markers (Schnabel and Schnabel, 1990; Granato et al., 1994). PHA-4 distribution in pha-1(e2123ts) embryos is essentially wild type if assayed 5-9 hours after embryos are laid (Fig. 3I,J). However, PHA-4 is not detected in pha-1(e2123ts) embryos when assayed later (data not shown), suggesting that pha-1 is required for the maintenance of PHA-4 levels. pha-1 is also required for the maintenance of ceh-22 expression (Okkema et al., 1997). Thus the role of pha-1 might be to increase the long-term stability of various pharyngeal transcription factors. It is interesting that the temperature-sensitive period and transgenic expression studies suggest that pha-1 does not seem to be required later than the 1.5-fold stage (Schnabel and Schnabel, 1990; Granato et al., 1994).

Of the 86 nuclei of the mature pharynx (including the six pharyngeal-intestinal valve cells), 37 are muscles and express the pharyngeal myosin myo-2 gene (Dibb et al., 1989). Okkema and Fire (1994) have analyzed the myo-2 promoter in transgenic nematodes and have identified two enhancer modules, denoted the B and C subelements. Dimers of the B subelement direct reporter gene expression in a subset of pharyngeal-muscle cells but not in non-muscle cells; sequences in the B subelement have been shown to bind the NK-2 class homeodomain protein CEH-22, a pharyngeal-muscle transcription factor (Okkema and Fire, 1994; Okkema et al., 1997). In contrast, dimers of the C subelement direct reporter gene expression into all cell types of the pharynx. Okkema and Fire (1994) suggested that the ‘pan-pharyngeal’ enhancer in the C subelement could be the direct target of a gene such as pha-4, because mutations in pha-4 caused such widespread pharyngeal defects. We now show that this suggestion appears to be correct.

Fig. 4A shows the sequence of a short region of the myo-2 C subelement, oligonucleotide C183 in Okkema and Fire (1994). Multiples of C183 can direct pharynx-wide reporter gene expression in wild-type embryos; a mutation (Cmut2) in this sequence abolishes this activity (Okkema and Fire, 1994). Multiples of C183 show no activity in pha-4 embryos (data not shown). As shown in Fig. 4A, the Cmut2 mutation lies within a sequence that agrees well with consensus binding sites determined for HNF-3α,β,γ factors (Costa, 1994) and for winged-helix proteins in general (Kaufmann et al., 1995). This suggests a model in which PHA-4 binds directly to the C183 sequence. As a test of this model, PHA-4 protein was produced in vitro and shown to form a complex with (double-stranded) C183, as detected using an electrophoretic mobility shift assay (Fig. 4B). This binding is effectively self-competed but is poorly competed by a double-stranded oligonucleotide containing the Cmut2 mutation.

Given that PHA-4 binds to C183 in vitro, we wished to demonstrate pha-4-dependent enhancer function in vivo. Worms were produced containing two independent transgenic arrays: the first array contained eight tandem copies of the C183 oligonucleotide fused to a GFP reporter; the second array contained the pha-4 cDNA under the control of the C. elegans hsp-16-2 promoter (Materials and methods). As shown in Fig. 5A, doubly transgenic embryos arrest after heat shock and express the reporter construct in most cells of the embryo. In heat-shocked control embryos that contain only the reporter construct array (i.e. do not contain hsp16-2::pha-4), reporter expression remains restricted to the presumptive pharynx and rectum (Fig. 5B). As an additional control, we examined embryos in which ectopic ELT-1, a GATA transcription factor essential for hypodermis development but uninvolved in pharyngeal development (Page et al., 1997), is produced throughout the embryo by heat shock. In heat-shocked hsp-16-2::elt-1 embryos, C183 reporter gene expression remains restricted to the pharynx and rectum (Fig. 5C).

We next wanted to determine if ectopic pha-4 could activate expression of the endogenous myo-2 gene. Heat shock of embryos transformed with hsp16-2::pha-4 results in the expression of pharyngeal myosin (detected by the monoclonal antibody 9.2.1) not only in the pharynx as expected, but also in dorsal and ventral regions of the embryo (probably body wall muscles) in approximately 20% of the embryos (Fig. 5D). When wild-type embryos or hsp16-2::elt-1 transformed embryos are heat shocked under the same conditions and stained with 9.2.1, pharyngeal myosin is detected only in the pharynx (Fig. 5E and F, respectively).

In summary, we suggest that PHA-4 binds to the C subelement of the myo-2 promoter and thereby participates directly in the activation of myo-2 expression in vivo.

ceh-22 encodes a NK-2 homeodomain protein that is found in a subset of pharyngeal muscle cells. CEH-22 binds to the B subelement of the myo-2 promoter and can drive expression of myo-2 when produced ectopically (Okkema and Fire, 1994; Okkema et al., 1997). Okkema and Fire (1994) have suggested that during normal pharyngeal development, it is the combination of pha-4 and ceh-22 that restricts myo-2 expression to muscle cells. CEH-22 is first detected in a subset of pharyngeal-muscle cells at the lima bean stage of embryogenesis, 330 minutes after fertilization. As demonstrated above, PHA-4 is already present in pharyngeal precursor cells at the time that CEH-22 first appears in pharyngeal muscle (Okkema and Fire, 1994); furthermore, CEH-22 is not detected in pha-4 mutants (Mango et al., 1994). This raises the possibility that pha-4 might also activate ceh-22.

Fig. 6A shows that ectopically produced PHA-4 can indeed induce significant levels of ectopic expression of a ceh-22::lacZ reporter gene outside of the presumptive pharynx. In heat-shocked embryos transformed either with the ceh-22::lacZ construct alone, or doubly transformed with ceh-22::lacZ and hsp16-2::elt-1, reporter expression remains confined to the pharyngeal muscle cells (Fig. 6B,C). Cell counts confirm the increase in ceh-22::lacZ-expressing cells in embryos with ectopic PHA-4. Heat-shocked embryos containing only the reporter gene show an average of 17.1 (± 1.5 s.d.; n = 25) total staining nuclei. Heat-shocked embryos containing the reporter gene and the hsp16-2::elt-1 control construct show an average of 13.3 (± 2.7 s.d.; n = 26) total staining nuclei. In contrast, heat-shocked embryos containing both the reporter gene construct and hsp16-2::pha-4 construct show an average of 41.6 (± 8.2 s.d.; n=15) total nuclei staining. In these latter embryos, a maximum of 54 staining nuclei was observed: a number greater than 21, the total number of nuclei in which CEH-22 is found in the mature pharynx (Okkema and Fire, 1994), and 37, the total number of all nuclei in pharyngeal muscles.

PHA-4 protein can be detected at low, but significant, levels in the gut of lima bean- and comma-stage embryos (Fig. 2C,D), and pha-4 mutant embryos show a mild gut phenotype. Staining with the monoclonal antibody MH-33 (Francis and Waterston, 1985) reveals the presumptive gut lumen to appear wider and more diffuse in pha-4 embryos than in wild-type controls (Fig. 7A,B respectively). This is a subtle defect compared to defects found in the pharynx and rectum but is nonetheless reproducible. It is not clear whether the phenotype is a direct result of the absence of PHA-4 in the gut or an indirect effect caused by the failure of pha-4 mutants to elongate.

Although endogenous PHA-4 protein (above) and pha-4 transcripts (Azzaria et al., 1996) are present at low levels in the gut, we previously showed that 7 kb of pha-4 promoter drives reporter gene expression strongly in gut cells (Azzaria et al., 1996). This discrepancy is likely to be caused by defects in transcriptional control and not by stable reporter protein, because in situ hybridization with transgenic embryos shows that reporter gene transcripts are present at high levels in the gut throughout embryogenesis (data not shown). We interpret the gut expression of the pha-4 transgene to reflect missing transcriptional gut repressor elements located outside of the 7 kb pha-4 promoter region used for the reporter constructs.

An intriguing parallel can be drawn between the expression of the pha-4 reporter gene and the expression of the gut esterase gene ges-1. ges-1 is normally expressed exclusively in the gut (Edgar and McGhee, 1986). However, deletion of a 36-base-pair region containing a tandem pair of GATA elements from the ges-1 promoter abolishes ges-1 expression in the gut and activates expression in cells of the pharynx and rectum (Egan et al., 1995; Fukushige et al., 1996). The pha-4 reporter constructs seem to show an inverse behavior: a dramatic increase in gut expression of the normally pharynx/rectum-specific gene. Overall, such observations suggest that each digestive tract module may contain factors to repress genes expressed in other modules of the digestive tract.

To investigate what controls pha-4 expression in the gut, a series of uni-directional 5′-deletions of the pha-4 promoter fused to a lacZ reporter gene was introduced into worms to produce stably transformed lines, whose embryos were then examined for β-Galactosidase activity. As shown in Fig. 7C, deletion of the pha-4 promoter to only 306 base pairs before the first PHA-4 ATG still allows reporter expression in the gut; however, deletion to 209 base pairs before the first ATG abolishes gut expression. Within this 97-base-pair gut-activating region, there is a sequence with significant similarity to the promoter region that controls the gut to pharynx/rectum switch in ges-1 (Fig. 7D) (Egan et al., 1995). The sequence in the pha-4 gut-activating region has a single GATA element along with an additional 9/9 match to a sequence adjacent to the GATA elements, both in reverse orientation. The gut-activating ability of this 97-base-pair pha-4 promoter region is redundant: when this precise region is deleted in the context of the entire 7 kb pha-4 promoter, the reporter gene is still expressed in the gut (data not shown).

The sequence matches between the pha-4 and ges-1 gut-activating regions suggest that the same factor might be involved in their control. The ELT-2 GATA factor is an excellent candidate to control ges-1 gut expression by binding to the GATA elements found in the ges-1 promoter (Hawkins and McGhee, 1995; Fukushige et al., 1998). As shown in Fig. 7E, in vitro-produced ELT-2 does indeed bind to the gut-activating region in the pha-4 promoter, as assayed by electrophoretic mobility shift. Furthermore, embryos that express ectopic ELT-2 under heat-shock control also express a pha-4::lacZ reporter construct throughout the embryos (Fig. 7F), compared to control embryos (Fig. 7G). Both of these observations suggest that elt-2 might activate gut expression of pha-4 in vivo.

The pha-4 gene (which we had previously cloned as Ce-fkh-1) meets most of the criteria proposed above for an ‘organ identity factor’: (1) in the absence of pha-4, the pharynx does not form (Mango et al., 1994); (2) PHA-4 is present in essentially all cells of the pharynx primordium and appears early in the pharyngeal precursor cell lineages; (3) pha-4 lies genetically (Mango et al., 1994) and molecularly downstream of the point at which the ABa-derived and MS-derived pathways of pharynx formation converge; and (4) although we were not able to show that ectopic PHA-4 causes formation of entire ectopic pharynges, ectopic expression of PHA-4 does lead to ectopic expression of at least two pharyngeal muscle-specific markers.

Both the persistence of PHA-4 protein into adults and the identification of ceh-22 and myo-2 as direct targets of pha-4 regulation suggest that PHA-4 participates at multiple steps of organ formation, from the earliest activation of subsidiary regulatory factors to the final activation of structural genes. Fig. 8 contains a simple schematic model illustrating how PHA-4 might function during pharynx organogenesis, more specifically in muscle lineages. We suggest that integration of the (induced) ABa-derived and (autonomous) MS-derived pharyngeal pathways occurs at the level of the pha-4 promoter. We further propose that PHA-4 is involved in succeeding waves of gene transcription throughout pharynx formation and that regulatory factors activated by PHA-4 at one stage of development (e.g. ceh-22) participate with PHA-4 in activating transcription of later genes (e.g. myo-2). Such progressive accumulation of transcription factors, which then cooperate in subsequent gene activation, might be a common feature in organogenesis.

It is clear that the regulatory scheme shown in Fig. 8 can reflect only a small fraction of the complexity inherent in building a pharynx. Our analysis of genes controlled by pha-4 applies only to (a subset of) pharyngeal muscles, but we suggest that similar schemes involving pha-4 may apply to other pharyngeal cell types (as well as to cells of the rectum). A major question is how the five different cell types are established within the pha-4-expressing cells of the pharyngeal primordium. For example, since ceh-22 expression is restricted to a subset of muscle cells, there must be at least one gene (depicted as X on Fig. 8) lying between the initial activation of pha-4 and activation of ceh-22. An additional complication is whether the three distinct PHA-4 polypeptides produced by alternative splicing (Azzaria et al., 1996) have different expression patterns or different functions.

Cloned C. elegans genes encoding proteins with the fork head/HNF-3α,β,γ DNA binding domain include lin-31, pes-1 and daf-16 (Miller et al., 1993; Hope, 1994; Ogg et al., 1997; Lin et al., 1997). Searches of the C. elegans genomic sequence reveal at least a dozen predicted genes encoding proteins with winged-helix DNA binding domains. However, PHA-4 remains by far the most similar protein to Drosophila fork head (Azzaria et al., 1996).

C. elegans pha-4 and Drosophila fork head share obvious common features of their mutant phenotypes, in particular severe derangement of the pharynx/foregut and rectum/hindgut (Weigel et al., 1989a,b; Mango et al., 1994). Furthermore, as illustrated in Fig. 9, the expression pattern of pha-4 in the C. elegans embryo bears a remarkable resemblance to the expression pattern of fork head in the digestive tract of the early Drosophila embryo (Weigel et al., 1989a,b). There are high levels of expression in domains at the digestive tract termini (pharynx/foregut and rectum/hindgut) and reduced expression in the gut proper.

Although comparison of expression patterns between pha-4 and the mammalian genes HNF-3α,β,γ is necessarily more subjective, it might be pertinent that two of the early points of HNF-3α,β,γ expression in the definitive endoderm of the mouse embryo are in the invaginating foregut and hindgut (Ang et al., 1993; Monaghan et al., 1993); in addition, foregut morphogenesis is severely affected in HNF-3β (−/−) mouse embryos (Ang and Rossant, 1994; Weinstein et al., 1994).

There may be additional commonalities between digestive tract development in C. elegans and in Drosophila. In C. elegans, we have shown that elt-2 may be involved in the gut expression of pha-4. ELT-2 is a gut-specific GATA factor (Hawkins and McGhee, 1995; Fukushige et al., 1998), highly similar to the Drosophila GATA factor serpent (Rehorn et al., 1996). serpent is required for Drosophila midgut development and acts genetically upstream of fork head (Reuter, 1994). This raises the possibility that HNF-3α,β,γ in vertebrate digestive tracts may be regulated by gut-associated GATA factors, such as GATA-4,5,6 (Laverriere et al., 1994).

In conclusion, we suggest that the pha-4-centered model proposed here for C. elegans pharynx formation may also apply to organogenesis regulated by fork head/HNF-3α,β,γ genes in the digestive tract of all (triploblastic) metazoa.

We thank S. Mango (University of Utah, USA) for providing pha-4(q490), and for helpful discussions and communicating results prior to publication. We also thank S. Marshall (University of Calgary, Canada) for providing pT7TAG-pha-4 plasmid, M. Chung (University of Calgary, Canada) for injecting promoter deletion constructs, I. Hope (Leeds) for providing the plasmid pUL#38E12, I. Ortiz and H. Epstein (Baylor), D. Miller (Vanderbilt) and R. Francis (Washington University) for providing antibodies, H. Wilkinson (Merck) for help in identifying somatic gonad cells and P. Mains for comments on the manuscript. Some nematode strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources (NCRR). This work was supported by the Medical Research Council of Canada, by the Howard Hughes Medical Institute and by the Alberta Heritage Foundation for Medical Research.