A regulatory network of T-box genes and the even-skippedhomologue vab-7 controls patterning and morphogenesis in C. elegans

T-box genes are a large family of transcriptional regulators unified by a conserved DNA binding, or T, domain that have important, yet diverse roles in patterning animal development (reviewed by Papaioannou, 2001). Mouse mutations in T-box genes such as Brachyury and Tbx6 are associated with notochord and mesoderm, or neural tube defects, respectively(Herrmann, 1995; Chapman and Papaioannou, 1998),and chick Tbx4 and Tbx5 have roles in limb specification(reviewed by Simon, 1999). In Xenopus, Brachyury (Xbra) is required for the development of mesoderm (Conlon et al.,1996), while VegT is required for induction of mesoderm and endoderm (Kofron et al.,1999; Zhang et al.,1998). The Drosophila Brachyury homologue brachyenteron (byn), or Trg, functions in the formation of hindgut and posterior mesoderm(Kispert et al., 1994; Singer et al., 1996; Kusch and Reuter, 1999), while optomotor blind (omb) is necessary for the development of anterior sensory structures such as the eye(Pflugfelder and Heisenberg,1995). In addition, several T-box genes are associated with human developmental abnormalities, including Holt-Oram syndrome(Basson et al., 1999),Ulnar-Mammary syndrome (Bamshad et al.,1999) and X-linked cleft palate(Braybrook et al., 2001),further highlighting the critical functions of members of this gene family.

All T-box proteins tested so far appear to bind to the same consensus DNA binding site (Kispert et al.,1995; Muller and Herrmann,1997). T-box genes can be divided into several subfamilies based on sequence comparisons, and some subfamilies are highly conserved across a wide range of phyla from Caenorhabditis elegans to humans(Papaioannou, 2001). The completed C. elegans genome sequence predicts 20 T-box genes in C. elegans, more than in any other species analysed so far, but functions for only two, mab-9 and mls-1, have been studied in detail(Woollard and Hodgkin, 2000; Kostas and Fire, 2002). Most of the C. elegans T-box genes appear to be highly diverged from those found in other species, with only four genes having obvious counterparts in other organisms (Papaioannou,2001). mab-9 is a member of the Tbx20 subfamily(Papaioannou, 2001) and is required for cell fate specification in the developing hindgut and for proper motor neuron function (Woollard and Hodgkin, 2000; Huang et al.,2002). mls-1, related to the Tbx1 subfamily, is necessary for correct muscle cell fate determination(Kostas and Fire, 2002). Two other T-box genes with orthologues in other organisms, F21H11.3(tbx-2), a member of the Tbx2 subfamily, and ZK328.6(tbx-7), related to the ascidian gene T2(Papaioannou, 2001), have not yet been characterised. Unlike other species tested so far, C. elegans does not have a recognisable Brachyury orthologue.

We have taken a reverse genetic approach to analysing T-box gene function in C. elegans. We used RNA interference (RNAi) to survey loss of function phenotypes for T-box genes and found that RNAi of tbx-9resulted in a phenotype similar to that of vab-7 mutants. vab-7 is a homologue of the Drosophila homeobox gene even-skipped (eve), required for posterior patterning of muscle and hypodermal cells in C. elegans(Ahringer, 1996); evehomologues also have roles in posterior development in several other species(Brown et al., 1997; Joly et al., 1993). We find that tbx-9, vab-7 and three other T-box genes act within a regulatory pathway important for embryonic development, with vab-7 functioning both upstream and downstream of T-box genes to regulate posterior patterning.

All C. elegans strains were derived from the wild-type Bristol strain N2. Routine maintenance of worms and genetic manipulations were performed as described (Sulston and Hodgkin, 1988).

PCR primers, including T7 or T3 RNA polymerase promoter sites, were designed to be specific to the gene to be silenced and to amplify typically 500-900 bp of mostly exonic sequence. dsRNA was synthesised directly from gel-purified PCR product as previously described(Fire et al., 1998) and injected into young N2 adult hermaphrodites at a concentration of ∼1 mg/ml. Injected worms were transferred to fresh plates 6 hours following injection and thereafter every 10-14 hours for 3 days.

One hundred mixed-stage embryos were isolated from the progeny of wild type, tbx-8(RNAi) and tbx-9(RNAi) hermaphrodites. Total RNA was extracted using the RNeasy® RNA extraction kit (Qiagen, Crawley, UK). RT-PCR amplification of tbx-8, tbx-9 and ama-1 (control)mRNA was carried out on these total RNA samples using Superscript™One-Step RT-PCR with Platinum® Taq (Invitrogen, Paisley, UK). Gene specific RT-PCR oligos (designed to anneal to either side of an intron to enable DNA and RNA amplification to be distinguished and to give RT-PCR products of around 200 bp) were as follows: tbx-8 (F cgtctt-gtcacttctgttcg, R ccctctcggaatccttggc), tbx-9 (F agtaacggcttaccagaacc, R tggggactgtgagttgctgc), ama-1 (F ttccaagcgccgctgcgcattgtctc, R cagaatttccagcactcgaggagcgga).

Plasmids were injected into the syncytial gonad of young adult hermaphrodite worms at concentrations of 1-50 ng/μl as described(Mello and Fire, 1995). Co-transformation markers were either rol-6 (plasmid pCes1943, gift of Diana Janke, University of British Columbia), in which case Rol progeny from N2 injected worms were picked and stable lines selected, or unc-119⁽⁺⁾ (plasmid pDP#MM016β), in which case non-Unc progeny from unc-119 (ed3) injected worms were selected. Where appropriate, integrated lines were generated by X-ray mutagenesis as described previously (Mello and Fire,1995). Integrated lines were outcrossed several times prior to analysis.

GFP fusions were made using pPD vectors kindly supplied by the Fire Lab(Carnegie Institute of Washington). All constructs were verified by sequencing. To make a tbx-8::GFP fusion, a PCR fragment (oligos F aaaactgcagaccggtttggcagctacac, R aaaactgcaggccattgatttc ctgctcaatatc)containing the entire coding and 5′ intergenic region minus the stop codon (7.7 kb) was first inserted into pBSKS⁺ (Stratagene) using PstI and then sub-cloned into pPD95.77 to make an in-frame GFP translational fusion (plasmid pAW232). Several transgenic lines were generated carrying this construct using 20-50 ng/μl DNA, giving very similar expression patterns. The strain described in this report is AW27(ouEx12[pAW232 + pCes1943]). To make a tbx-9::GFP fusion, a PCR fragment (oligos F aaaactgcaggattcaatcaaaacgggc, R aaaactgcaggcaccaacaatatcaatatcttc) was cloned directly into pPD95.75 using PstI to make an in-frame translational fusion (plasmid pJA57). Several transgenic lines were generated carrying this construct, using 1 ng/μl DNA (higher concentrations were found to be toxic). Transgenic lines gave similar expression patterns. The strain described in this report is AW22(unc-119 (ed3); ouEx8 [pJA57 + pDP#MM016β]). A vab-7::GFP reporter construct was made by cloning GFP in-frame into the SphI site of exon 1 of the vab-7 genomic rescuing construct pJA17 (14 kb), to generate the translational fusion construct pJA64,which was injected at a concentration of 20-50 ng/μl. This construct gave the same expression pattern in transgenic worms as the lacZ reporter and in-situ hybridisations previously described(Ahringer, 1996). The vab-7::GFP transgenic strain described in this report is AW23(unc-119 (ed3); ouEx9 [pJA64 + pDP#MM016β]). The integrated mab-9::GFP strain (eIs34) used in this study has been described previously (Woollard and Hodgkin, 2000). This was crossed into vab-7(e1562) to give the strain AW24. The integrated pal-1::GFP strain(ctIs33) used in this study was a gift from Lois Edgar (University of Colorado, Boulder). The elt-2::GFP reporter strain(wIs81[elt-2::GFP;pRF4(rol-6)]) was a gift from Joel Rothman(University of California, Santa Barbara).

Two hsp-16 driven tbx-9 constructs were made. pJA50 consists of the tbx-9 full-length cDNA (XbaI-KpnI insert from cDNA clone pYK337c3) cloned into pPD49.78 (hsp16-2driven), and pJA52 consists of the same cDNA insert cloned into pPD49.83(hsp16-41 driven). A transgenic line was derived containing both hsp16::tbx-9 constructs, weEx41 [pJA50 + pJA52 + pRF4(rol-6)]. An integrated strain was subsequently derived from this, weIs8 (strain JA1286). A similar approach was taken to construct a strain (JA1284, weIs6) containing an integrated copy of an extrachromosomal array containing hsp16-2 driven (pJA49) and hsp16-41 driven (pJA51) tbx-8 cDNAs, derived from the cDNA clone pYK325e8. Integrated heat-shock driven tbx-8 and tbx-9strains were then crossed into an unc-119 (ed3) background and injected with the vab-7::GFP construct described above (pJA64)together with unc-119⁽⁺⁾ to give the strains AW26(unc-119(ed3); weIs8 (hsp-16::tbx-9 + rol-6); ouEx11 [pJA64 +pDP#MM016β]) and AW41 (unc-119(ed3); weIs6 (hsp-16::tbx-8 + rol-6);ouEx14 [pJA64 + pDP#MM016β]). Non Unc, Rol progeny were selected in each case. Adult hermaphrodite worms were subsequently heat shocked at 33°C for 45 minutes and then incubated at 20°C for 2 hours and embryos dissected for examination.

A 4.3 kb PacI-PflMI subclone (plasmid pAW223) of the vab-7::GFP reporter pJA64 containing the putative T-box binding sites to be mutated was used for site-directed mutagenesis. Site B was mutated first(see legend to Fig. 8) to generate plasmid pAW224 using the Stratagene Quickchange kit and protocols. Mutagenesis PCR oligos were as follows: F gtacgcctcattgatgtaatgtaaaagagatgtg R catgcggagtaactacattacattttctctacac. Site A was subsequently mutated to generate pAW226, in which both putative T-box binding sites were mutated. PCR oligos were as follows: F cgagcggaaaagatgtatttgaaccttcc R gctcgccttttctaca taaacttggaagg. The mutated PacI-PflMI fragment was then re-inserted into the vab-7::GFP reporter construct pJA64 to generate the plasmid pAW231, a vab-7::GFP reporter in which both putative T-box binding sites in the vab-7 regulatory region were mutated into sites that would not be expected to permit T-box proteins to bind(Sinha et al., 2000). Transgenic lines were generated using this construct that gave similar expression patterns. The transgenic line described in this report is AW28(ouEx13 [pAW231 + pCes1943]).

To verify that the cells expressing tbx-8 and tbx-9::GFP fusions were the cell types they appeared to be by morphology and position,the following antibodies were used and co-localisation was assessed: a GFP antibody (chicken monoclonal from Chemicon, Hampshire, UK) for visualising TBX-8 and TBX-9::GFP fusions, LIN-26 antibody (gift from M. Labouesse,Strasbourg), which stains all hypodermal cells, and mouse mAb NE8/4C6.3, which stains the outlines of body wall muscle cells(Goh and Bogaert 1991). Secondary antibodies were FITC anti-chicken or Texas Red anti-mouse (Jackson Immunoresearch, Pennsylvania, USA). Immunostaining experiments were performed as follows: embryos were placed on poly-lysine coated slides, squashed under a coverslip and frozen on dry ice for 20 minutes. After freezing, the coverslip was flicked off and slides placed in methanol for 20 minutes at room temperature, washed with PBS for 5 minutes and PBS + Tween (0.2%) for 10 minutes. Primary antibodies were incubated overnight at 4°C in PBS+Tween. After washing, secondary antibodies were incubated for 1-2 hours at room temperature. Samples were mounted in mowiol.

A phylogenetic tree of C. elegans T-box genes together with representative members of each of the defined T-box subfamilies found in other organisms (Papaioannou, 2001)is shown in Fig. 1A. This illustrates the high degree of divergence of most of the C. elegansT-box genes, with the exception of the four genes discussed above. There are several putative paralagous pairs of T-box genes in C. elegans, which may be the result of recent gene duplications.

A survey of C. elegans T-box gene function by RNAi has revealed obvious phenotypes in only a few cases (Maxwell, Aslam and Woollard,unpublished observations; Ahringer, unpublished observations). The tbx-9(RNAi) phenotype consists of a characteristic swelling of the hermaphrodite tail hypodermis, giving a `bobbed tail' appearance highly reminiscent of the tail phenotype seen in vab-7(e1562) animals(Fig. 2). This phenotype is not very penetrant in N2 animals (∼10%). When RNAi experiments were conducted in the rrf-3 strain, which is more sensitive to RNAi(Simmer et al., 2002), the penetrance of tbx-9(RNAi) phenotypes was increased: 15-25% of the progeny of rrf-3 dsRNA injected hermaphrodites had a vab-7-like bobbed tail and, in addition, 10-15% of progeny had other morphological defects, including dorsal hypodermal bulges in the midbody and posterior region (data not shown).

The phylogenetic tree presented in Fig. 1 suggests that tbx-8 and tbx-9 may be paralogous. The two proteins are 59% identical in their T-box domains and are located only 3.75 kb apart on chromosome III (map position 2.41). They are divergently transcribed and thus may share common 5′ regulatory regions(Fig. 1B). RNAi of tbx-8 revealed no obvious abnormalities, either in N2 or rrf-3 animals (Fig. 2Band data not shown). To test the efficacy of RNAi for both tbx-8 and tbx-9, we assayed tbx-8 and tbx-9 mRNA levels in embryos by RT-PCR and found that both mRNAs were undetectable following injection of the corresponding dsRNA (Fig. 2G). We also found that tbx-8 and tbx-9 RNAi abolished the GFP signals of tbx-8::GFP and tbx-9::GFPtransgenic worms, respectively (data not shown).

To test whether tbx-8 and tbx-9 have overlapping functions, double RNAi experiments were performed. Silencing of both genes simultaneously in N2 animals gave rise to 100% embryonic or early larval lethality (50-60% embryonic arrest, 40-50% L1 arrest, n=130), with gross morphological defects in the midbody region and posterior(Fig. 2E, Fig. 4). There was a failure of body elongation, and hatched animals had large dorsal bulges in the hypodermis(Fig. 2E). We compared phenotypes observed in double RNAi experiments with those of a tbx-8knockout allele (ok656, a presumed null allele available from the C. elegans Knockout Consortium, Oklahoma, USA), which had been subjected to tbx-9 RNAi and found that they were identical: tbx-8(ok656); tbx-9(RNAi) animals also arrest as embryos or early larvae in similar proportions (53% embryonic arrest, 46% L1 arrest, n=220), with identical morphological defects (data not shown). This demonstrates the efficacy of silencing tbx-8 and tbx-9simultaneously by double RNAi. tbx-8(ok656) worms, like tbx-8(RNAi) animals, have no obvious defects on their own (data not shown).

The dorsal bulges displayed by tbx-8/tbx-9(RNAi) animals suggested that hypodermal morphogenesis might have been affected. In wild-type embryos,the two rows of dorsal hypodermal cells, born around 240 minutes after first cleavage, intercalate into a single dorsal row around 290-340 minutes(Fig. 3A). During dorsal intercalation, dorsal hypodermal cells become wedge-shaped, with their pointed ends oriented towards the dorsal midline. The pointed tips of these cells interdigitate contralaterally and elongate so that the advancing edges eventually contact the opposing lateral hypodermal (seam) cells, which flank the dorsal hypodermis (Fig. 3A). Migration of dorsal nuclei follows behind the extending cell tips, so that the nuclei end up close to the lateral seam cells. The hypodermis then extends ventrally, spreading to enclose the embryo, with ventral hypodermal cells eventually sealing up the ventral pocket by a`purse-string' mechanism at 310-360 minutes. Dorsal hypodermal cells subsequently undergo fusion to form a syncytium, and circumferential contraction within the hypodermis causes a fourfold elongation of the embryo(360-600 minutes). The rearrangement and movements of groups of hypodermal cells during embryonic development is instrumental in setting the shape of the worm (reviewed by Chin-Sang and Chisholm,2000; Simske and Hardin,2001).

To analyse hypodermal cell positions and migrations, we used the ajm-1::GFP (pka jam-1::GFP) reporter (strain SU93), which marks adherens junctions, allowing visualisation of the outlines of hypodermal cells (Mohler et al., 1998). In tbx-8/tbx-9(RNAi) embryos, dorsal intercalation is defective(Fig. 3). In these embryos,dorsal intercalation is relatively normal at the anterior end of the embryo but fails to complete towards the posterior(Fig. 3B-D). Some of the cells appear square, or even rounded, rather than wedge-shaped, and intercalation appears to arrest at this point (Fig. 3B-D). Observation of the same embryos 90 minutes later showed that dorsal intercalation does not proceed beyond this stage (data not shown). No obvious defects in ventral enclosure were evident (data not shown);however, defects in the arrangement of lateral hypodermal (seam) cells could be seen. In wild-type embryos, lateral hypodermal cells form two linear rows with 10 cells on each side of the embryo(Fig. 3E). In tbx-8/tbx-9(RNAi) embryos, lateral rows are interrupted,with one to four cells pinched out of line(Fig. 3F). However, seam cell number is not affected (data not shown). This misalignment of seam cells can also be seen in those RNAi worms that survive to hatching(Fig. 3G). tbx-8/tbx-9(RNAi) embryos fail to elongate properly and die either as late-stage embryos (50-60%) or misshapen, short L1 larvae (40-50%),with large bulges typically, though not exclusively, on the dorsal side(Fig. 2E).

Muscle cells are also affected in tbx-8/tbx-9(RNAi)animals. In wild-type animals, body wall muscles are present as two regular rows on the dorsal and ventral sides of the animal and can be visualised using hlh-1::GFP or myo-3::GFP transgenic markers. In tbx-8/tbx-9(RNAi) embryos and hatched larvae, these rows are not regular. Muscles are often seen out of line and rows are sometimes broken down completely, with muscle cells being bunched together laterally(Fig. 4B,D,F). The overall number of muscle cells, as assessed by counting nuclei expressing the muscle marker hlh-1::GFP at the 2-fold stage, is unchanged (wild type, 58(n=22), tbx-8/tbx-9(RNAi), 58 (n=17)),suggesting that this is a positioning, rather than a cell fate commitment,defect. The earliest observable muscle-positioning defect was seen around the 400 cell stage, as dorsal hypodermal intercalation was proceeding(Fig. 4B). We cannot exclude the possibility, therefore, that muscle-positioning defects are caused by abberant dorsal hypodermal morphogenesis, rather than being a cell autonomous defect. Likewise, it is possible that some of the hypodermal defects observed were a secondary consequence of muscle-positioning defects. However, it is known from ablation experiments that dorsal hypodermal intercalation proceeds normally in the absence of the underlying muscle tissue(Heid et al., 2001),consistent with our view that hypodermal defects in tbx-8/tbx-9(RNAi)embryos are unlikely to be a secondary consequence of muscle-positioning defects. Furthermore, bulges on the surface of tbx-8/tbx-9(RNAi)animals that survived to L1 were always correlated with gross disorganisation of hypodermis, while some bulges did not contain muscle cells (data not shown), suggesting that hypodermal defects were the primary cause.

tbx-8/tbx-9(RNAi) animals that survived to hatching had abnormal gut morphology (Fig. 4G,H,J). Using an elt-2::GFP intestinal cell marker, we examined the number and positions of intestinal nuclei in wild-type and tbx-8/tbx-9(RNAi)worms. We found that the number of intestinal cells was unchanged at the 1.5-fold stage (wild type, 19 (n=14), tbx-8/tbx-9(RNAi), 19(n=22)). These cells were correctly positioned at this stage (data not shown) but in hatched tbx-8/tbx-9(RNAi) animals intestinal cells were mispositioned (Fig. 4Jcompared with Fig. 4I). This suggests a potential role for tbx-8 and tbx-9 in intestinal morphogenesis, although it may be that tbx-8 and tbx-9 do not have a direct role in intestinal cell positioning; positioning defects during late embryogenesis could be a consequence of aberrant morphogenesis. Likewise, it is possible that intestinal defects in tbx-8/tbx-9(RNAi)worms are a secondary consequence of defective embryonic elongation.

The expression patterns of tbx-8 and tbx-9 GFP translational fusions are shown in Fig. 5. tbx-8 and tbx-9 are both expressed in three different cell types in the embryo, gut, muscle and hypodermis, in a largely overlapping pattern. In all cases expression was found to be nuclear, as would be expected for putative transcription factors. The earliest embryonic expression of both tbx-8 and tbx-9 was detected in the E lineage gut cell precursors Ea and Ep at the beginning of gastrulation(∼100 minutes after first cleavage)(Fig. 5A,B). Expression was seen in the direct descendents of Ea and Ep but was not obvious later in E lineage nuclei (data not shown). Although the expression of tbx-8 and tbx-9 in intestinal cells is consistent with a possible role for tbx-8 and tbx-9 in gut development, it is not clear how the early gut expression of tbx-8 and tbx-9 in gut cell precursors relates to the intestinal defects seen in tbx-8/tbx-9(RNAi) worms, which only become apparent later in embryogenesis, as discussed above.

tbx-8 and tbx-9 were subsequently expressed in muscle and hypodermal cells (Fig. 5C-I). The expression of both tbx-8 and tbx-9 was stronger in dorsal hypodermal nuclei, which were just starting to intercalate, compared with those at the anterior, which were intercalated, or those at the posterior, which had yet to intercalate(Fig. 5E-G). This is particularly evident in embryos that have been co-stained with LIN-26 and GFP antibodies (Fig. 5I) and is consistent with a role for tbx-8 and tbx-9 in the process of dorsal intercalation, as suggested by the phenotypic analysis presented above. Expression of tbx-8 and tbx-9 could also be seen in dorsal hypodermal nuclei undergoing contralateral migration following cellular intercalation (data not shown), and in lateral hypodermal cells during dorsal intercalation (Fig. 5E-G),although expression in lateral hypodermal cells at this stage was weaker. Expression in lateral hypodermal (seam) cells could also be seen later in embryogenesis until the 1.5-fold stage(Fig. 5C). No expression of either tbx-8 or tbx-9 was observed in ventral hypodermal cells. The earliest detectable muscle cell expression of tbx-8 and tbx-9 was seen around the 400-cell stage, when muscle cells were present as two rows on the left and right sides of the embryo(Fig. 6A-B). Expression of tbx-8 and tbx-9 in hypodermal and muscle cells around the time when defects in these two cell types are first observed supports the view that hypodermal and muscle defects in tbx-8/tbx-9(RNAi)animals are likely to be largely cell autonomous, rather than one being a consequence of the other.

vab-7 is required for the correct patterning of posterior body muscles and hypodermis and is expressed in posterior body muscle precursors and in hypodermal cells (Ahringer,1996). The similarity in phenotype between vab-7 mutants and tbx-9(RNAi) animals, and the overlapping expression domains of vab-7, tbx-8 and tbx-9(Fig. 6A-C)(Ahringer, 1996), suggest a possible regulatory interaction. To investigate this, we asked whether RNAi silencing of tbx-8 and tbx-9 affects the expression pattern of a vab-7::GFP reporter. Muscle expression of vab-7::GFPwas abolished in these embryos, indicating that tbx-8 and tbx-9 function to promote muscle expression of vab-7, either directly or indirectly (Fig. 6E, compared with Fig. 6D). At the comma stage, vab-7::GFP is normally expressed in four to five posterior hypodermal nuclei (4.7±0.1, n=18)(Fig. 6D)(Esmaeili et al., 2002), but in tbx-8/9(RNAi) embryos expression was seen in only one or two posterior hypodermal nuclei (1.7±0.1, n=13, Fig. 6E). Silencing either tbx-8 or tbx-9 alone did not significantly reduce vab-7 expression (data not shown), indicating that these two genes have redundant roles in vab-7 activation. It is possible, however,that the low penetrance vab-7-like bobbed-tail phenotype in tbx-9(RNAi) animals is the result of a slight reduction of vab-7 expression in posterior hypodermis not detectable with the reporter.

To see if tbx-8 and tbx-9 are sufficient for vab-7 expression, we generated transgenic worms in which either tbx-8 or tbx-9 expression was driven by two strong inducible heat-shock promoters hsp16-2 and hsp-16-41, in addition to carrying a vab-7::GFP reporter construct (see Materials and methods). Following heat treatment, ectopic vab-7::GFP expression driven by either TBX-8 or TBX-9 could be observed up to the 400-cell stage of embryogenesis (Fig. 7A-D). No ectopic vab-7 expression was seen as a result of heat treatment in the absence of the tbx-8 or tbx-9 heat-shock constructs(Fig. 7E-F). Therefore, tbx-8 and tbx-9 are both necessary and sufficient for vab-7 expression in embryos.

An obvious question is whether or not the activation of vab-7expression by tbx-8/tbx-9 is direct. All T-box proteins analysed so far interact with a characteristic DNA binding site in target genes with the core consensus sequence GGTGTGAA (reviewed by Papaioannou, 2001; Smith, 1999). The vab-7 5′ regulatory region contains two close-match T-box binding sites situated close together, ∼2 kb from the start of transcription (Fig. 8A). As these sites are good candidates for TBX8/9 binding, we engineered a vab-7::GFP reporter construct in which both sites were mutated. We found, however, that the posterior muscle and hypodermal expression pattern of vab-7 was unchanged in transgenic embryos carrying the mutated vab-7::GFP reporter (Fig. 8B,D,F). This suggests either that TBX-8 and TBX-9 do not regulate vab-7 directly, or that TBX-8/9 do activate vab-7 directly,but bind to alternative or additional binding sites, perhaps with a more degenerate sequence. T-box proteins in other organisms have been found to exhibit complex binding interactions with in-vivo target sites, often with a high degree of redundancy (Kusch et al.,2002; reviewed by Papaioannou,2001).

Although posterior expression of vab-7 was unchanged by mutating the two putative T-box binding sites, we did detect a marked difference in vab-7 expression at the anterior of embryos carrying the mutated reporter. Normally, no vab-7 expression is observable in the anterior of embryos at any stage of development(Ahringer, 1996) (this report). In transgenic lines carrying the mutated vab-7 reporter, however,significant anterior vab-7 expression is reproducibly seen (77% of embryos, n=48), at several different embryonic stages(Fig. 8B,D,F). This raises the possibility that there is a T-box factor in C. elegans that represses vab-7 expression in the anterior of embryos and that this factor normally binds directly to the T-box sites we have mutated. In order to investigate this, we silenced each of the remaining 18 T-box genes singly by RNAi and monitored the effect on wild-type vab-7 expression. We found that silencing tbx-30 (Y59E9AR.3) resulted in comparable ectopic anterior expression of vab-7::GFP(Fig. 8C,E,G), thereby phenocopying the effect of mutating the T-box binding sites in the vab-7 reporter. There were no obvious morphological phenotypes associated with silencing tbx-30. Silencing the other 17 C. elegans T-box genes had no effect on vab-7::GFP expression (data not shown). Thus, tbx-30 encodes a novel, anterior repressor of vab-7. Examination of the amino acid sequences of TBX-8, TBX-9 and TBX-30 revealed that there are highly acidic domains at the C-termini of TBX-8 and TBX-9 (data not shown), consistent with a role for these two genes in transcriptional activation. No such activation domain is present in TBX-30,however, as we would expect for a transcriptional repressor.

In a separate screen for genes that regulate the expression of the T-box gene mab-9 (Pocock and Woollard, unpublished observations), we found that mab-9 expression was altered in a vab-7 mutant background. mab-9 is normally first expressed in embryos at the 1.5-fold stage in the nuclei of three cells around the presumptive rectum, B,F and one hyp 7 nucleus (Fig. 9B) (Woollard and Hodgkin,2000). In a vab-7(e1562) mutant background, however, the domain of mab-9 expression was more extensive, with three to four extra mab-9 expressing nuclei appearing more posterior to the usual mab-9 expressing nuclei (Fig. 9D). It is difficult to unambiguously assign these nuclei in a vab-7 background, because of the disruption in normal cell positioning in this mutant, but they appear to be muscle nuclei, and would normally be expected to express vab-7. We examined mab-9expression in tbx-8/tbx-9(RNAi) embryos and found similar ectopic posterior expression (data not shown). This is consistent with the notion that TBX-8 and TBX-9 are required for the expression of vab-7 in posterior muscle cells, and that VAB-7 is required, in turn, to repress mab-9expression in these cells.

Our data suggest that tbx-8 and tbx-9 have crucial,overlapping roles in controlling several aspects of embryonic morphogenesis and patterning. These two genes are required for the correct arrangement of dorsal and lateral hypodermal cells during intercalation and elongation and also form part of a regulatory network that includes two other T-box genes, tbx-30 and mab-9, and the homeobox gene vab-7. This network of T-box genes and vab-7 is important for the correct patterning of posterior cells.

Embryos lacking both tbx-8 and tbx-9 fail to complete dorsal hypodermal intercalation, arresting as late embryos or short L1 larvae with gross morphological defects. Consistent with this phenotype, we found that tbx-8 and tbx-9 were both expressed at high levels in dorsal hypodermal cells undergoing intercalation, suggesting that tbx-8/tbx-9 are likely to be acting cell autonomously within the dorsal hypodermis to control these rearrangements. Although no obvious defect was seen on the ventral side of embryos during enclosure, the lateral hypodermal (seam) cells were often mispositioned, or pinched out of line. This may be a consequence of the dorsal intercalation defect, with the seam cells not being held in line properly by the dorsal syncytium.

There are some known mutants in C. elegans with similar dorsal hypodermal phenotypes, which might point to biochemical pathways in which tbx-8 and tbx-9 are acting. For instance, die-1(dorsal intercalation and elongation defective)mutants also initiate, but fail to complete, dorsal intercalation(Heid et al., 2001). tbx-8, tbx-9 and die-1 are all expressed within dorsal hypodermal cells undergoing intercalation(Heid et al., 2001 and this report). die-1 encodes a zinc finger protein thought to act as a transcriptional regulator, although it is not known at present what the transcriptional targets of die-1 might be, or how this gene might be regulated. It is also noteworthy that tbx-8/tbx-9(RNAi) animals display other phenotypes in common with die-1 mutants. These include lateral hypodermal cells being pinched out of line, body wall muscle cell positioning defects and abnormalities in gut morphogenesis(Heid et al., 2001) (this report). This would support the idea that tbx-8/tbx-9 and die-1 may act in the same pathway to regulate several aspects of embryonic morphogenesis.

Dorsal hypodermal intercalation in C. elegans has been likened to the process of convergent extension (directed intercalation of cells towards an axis of extension) in vertebrates(Chin-Sang and Chisholm, 2000). Intriguingly, it has been reported that the T-box gene spadetail is required in zebrafish embryos for lateral mesoderm cells to undergo the convergent extension movements of gastrulation(Griffin et al., 1998; Ho and Kane, 1990). Furthermore, Brachyury in Xenopus has also been shown to be required for convergent extension movements(Conlon et al., 1996; Conlon and Smith, 1999). Therefore, it seems that regulation of cell intercalation movements during embryonic morphogenesis is a conserved T-box gene function. Since spadetail and Brachyury have no obvious counterparts in C. elegans, and tbx-8 and tbx-9 have no obvious counterparts in vertebrates, we suggest that this may be an ancient role, with its origin in a common ancestral gene.

We have shown that: (1) tbx-8, tbx-9 and vab-7 have overlapping expression domains in embryos in posterior muscle and hypodermal cells; (2) vab-7 expression in embryos requires tbx-8/tbx-9activity; and (3) overexpression of tbx-8 or tbx-9 is sufficient to drive ectopic vab-7 expression throughout the embryo. This suggests that tbx-8, tbx-9 and vab-7 function in a common genetic pathway to control embryonic patterning events, with tbx-8 and tbx-9 acting as upstream activators of vab-7.

Posterior muscles (those derived from the C blastomere) in vab-7mutants are disorganised. Muscle cells are often bunched together laterally or even form rings with neighbouring rows(Ahringer, 1996). We found similar defects in tbx-8/tbx-9(RNAi) animals, with muscle cells being present in the correct number, but severely misplaced, consistent with the notion that tbx-8 and tbx-9 are required for vab-7 expression and subsequent patterning activity in C blastomere-derived muscle cells. The muscle defects seen in tbx-8/tbx-9(RNAi) animals were not confined to the posterior, however, so it is likely that there are other muscle-specific targets of tbx-8/tbx-9 besides vab-7. It is also possible that some of the muscle defects seen in tbx-8/tbx-9(RNAi) animals are a secondary consequence of the other morphogenetic abnormalities. The bobbed tail appearance of tbx-9(RNAi) and vab-7(e1562) animals is likely to be the outcome of a hypodermal defect that is common to both situations. However, tbx-8/tbx-9(RNAi) worms also have more severe hypodermal defects during dorsal intercalation that are not seen in vab-7 mutants,suggesting that TBX-8 and TBX-9 are likely to influence the expression of multiple target genes involved in hypodermal cell rearrangements as embryogenesis proceeds.

Overexpression of tbx-8 or tbx-9 from the strong,inducible heat-shock promoter is sufficient to drive ectopic vab-7expression throughout the embryo. The caudal orthologue pal-1 has also been shown to be capable of driving ectopic vab-7 expression, and, like tbx-8/tbx-9, is required for vab-7 expression(Ahringer, 1997). It is therefore of interest to consider whether tbx-8/tbx-9 act in the same pathway as pal-1 to regulate vab-7 expression. We have performed experiments, however, that show that tbx-8 and tbx-9 were still expressed in pal-1(RNAi) embryos (data not shown), albeit in a disorganised pattern, as would be expected if pal-1 expression was silenced(Hunter and Kenyon, 1996). Likewise, pal-1::GFP is still expressed in tbx-8/tbx-9(RNAi) embryos (data not shown); therefore we conclude that tbx-8/tbx-9 and pal-1 probably act in separate, complementary pathways to regulate vab-7 expression.

The activation of evenskipped in posterior cells by T-box genes has been reported in other systems, suggesting that this may be a conserved mechanism for controlling posterior pattern formation. For example, brachyenteron is required for eve expression in the hindgut and anal pad primordia of Drosophila(Kusch and Reuter, 1999; Singer et al., 1996), and in mouse, evx-1 expression has been shown to be dependent on Brachyury in the posterior tail bud(Rashbass et al., 1994). Likewise, no tail (ntl-zebrafish Brachyury) is required for the maintenance of eve1 expression during tail extension in zebrafish (Joly et al.,1993). Whether these examples of T-box gene mediated regulation of eve genes involve direct or indirect effects remains to be seen. It is also noteworthy that ectopic Xbra expression in Xenopuscauses marked induction of the eve homologue Xhox3(Cunliffe and Smith, 1992),similar to the induction of vab-7 expression caused by tbx-8or tbx-9 overexpression reported here. Perhaps the regulation of eve has been taken over by tbx-8/tbx-9 in C. elegans in the absence of a bona fide Brachyury homologue.

If tbx-8 and tbx-9 directly activate vab-7, then this is not through the two T-box binding sites we have described, as mutating these sites did not eliminate vab-7 expression. By contrast, we found that mutating these T-box binding sites caused ectopic anterior expression of vab-7, suggesting that vab-7 may be subject to direct inhibitory regulation by another T-box gene. T-box genes have been shown to act as either activators or inhibitors of target gene expression (reviewed by Papaioannou, 2001). By silencing each of the C. elegans T-box genes in turn, we found that tbx-30 normally repressed vab-7 expression in anterior cells during embryogenesis. tbx-30 probably directly represses vab-7 at the anterior, through the T-box binding sites we have defined, as mutating these binding sites had exactly the same effect on vab-7 expression as silencing tbx-30.

eve orthologues are known to function as transcriptional repressors in a wide variety of situations and are thought to act by interacting with the basal transcriptional machinery at specific promoters(Han and Manley, 1993; Li and Manley, 1998). The VAB-7 protein contains two putative repressor domains that are also found in other members of the eve family. There is a polyalanine stretch near the N-terminus of VAB-7 and a region rich in proline and serine residues at the C-terminus (Ahringer,1996). Alanine-rich and proline-rich regions have both been shown to function in transcriptional repression(Um et al., 1995). Our data suggest that VAB-7 acts to repress mab-9 expression in cells at the very posterior of embryos at the 1.5-fold stage. Ectopic expression of mab-9 is generally deleterious during development, resulting in embryonic arrest or animals with posterior deformations(Woollard and Hodgkin, 2000). It is possible, therefore, that inappropriate, ectopic expression of mab-9 in posterior cells contributes to the abnormal posterior phenotype of vab-7 mutant animals.

Overall, we have described a situation in which the activities of at least four different T-box genes in C. elegans (tbx-8, tbx-9,tbx-30 and mab-9) are linked by vab-7. A model for these regulatory interactions is depicted in Fig. 10. Although positive regulation of eve homeobox genes by T-box proteins has been reported in other systems, this is the first demonstration that an evehomeobox gene is also sensitive to T-box mediated repression, and that T-box genes themselves can be subject to repression by eve. The potential evolutionary association between the role of tbx-8 and tbx-9in dorsal hypodermal intercalation in C. elegans and the role of T-box genes in convergent extension movements of cells in other organisms is a second intriguing connection. Perhaps the involvement of T-box genes in these kinds of cell rearrangements, which are an essential prelude to metazoan morphogenesis, will turn out to be widespread.

We would like to thank Jonathan Hodgkin and Sarah Newbury for helpful comments on the manuscript. R.P. and A.W. were funded by the UK Medical Research Council, the Biotechnology and Biological Sciences Research Council,and the Royal Society. M.M. and J.A. were funded by Wellcome Trust Fellowships(Nos 054523 and 045515). We would also like to thank Behrooz Esmaeili(University of Oxford) for making the vab-7::GFP reporter construct used in this study.