Specification of muscle neurotransmitter sensitivity by a Paired-like homeodomain protein in Caenorhabditis elegans

Homeodomain transcription factors are involved at numerous levels of animal development: while HOX genes specify the relative position of entire domains of gene expression in the anteroposterior axis, at the other extreme there are types of homeobox genes that appear to determine the expression of only a few genes in a single cell type. For example, in Caenorhabditis elegans,Paired-like homeodomain proteins have been found to be involved in the fine-tuning of neuronal differentiation, specifying such features as cell-type-specific synaptic inputs (Miller et al., 1992; White et al.,1992; Winnier et al.,1999), neurotransmitter synthesis(Eastman et al., 1999; Jin et al., 1999) and axonal outgrowth (Brockie et al.,2001; Pujol et al.,2000).

By relating behavioral outputs to patterns of gene expression and the expression of unique differentiated cellular features, it is possible to study the molecular, genetic and cellular bases of behavior in C. elegans. For example, neural circuits regulating behaviors such as pharyngeal pumping,egg-laying and locomotion have all been studied in this manner (reviewed by Rankin, 2002; Whittaker and Sternberg,2004). The neuromuscular system that regulates the Defecation Motor Program (DMP) (Liu and Thomas,1994) is also beginning to be studied at this level. The DMP consists of three distinct, stereotyped steps(Thomas, 1990): the posterior body muscle contraction (pBoc), the anterior body muscle contraction (aBoc)and the expulsion (Exp), which is the most complex step as it consists of the coordinated contractions of two specialized enteric muscles, the intestinal and anal depressor muscles, and probably the relaxation of a third enteric muscle, the sphincter. The defecation cycle period, which is the interval between two series of these three contractions, is very regular over time in single animals, and from animal to animal, but is modulated by sensory inputs such as touch (Liu and Thomas,1994), food (Liu and Thomas,1994) and temperature (Branicky et al., 2001).

Two GABAergic neurons, AVL and DVB, send processes to the enteric muscles(the intestinal, sphincter and anal depressor muscles) and are required for the proper execution of the DMP. Laser ablation studies have revealed that these two neurons are partially redundant for enteric muscle contractions. AVL, but not DVB, also appears to be required for the aBoc step(McIntire et al., 1993b). The pBoc step may be regulated by a non-neural pathway, as neuronal ablations have failed to identify any neuron required for the pBoc (E. Jorgensen, personal communication). Interestingly, the neurotransmitter GABA, normally an inhibitory neurotransmitter, is an excitatory neurotransmitter for enteric muscle contraction. Indeed, loss of the GABA biosynthetic enzyme glutamic acid decarboxylase (GAD), encoded by the unc-25 gene, causes an expulsion defective (or Exp) phenotype, which can be rescued by exogenous GABA(Jin et al., 1999; McIntire et al., 1993a). In addition, mutations that disrupt GABA transport (unc-47, gat-1), and presynaptic GABA release (unc-2) also produce an Exp phenotype(Jiang et al., 2004; Mathews et al., 2003; McIntire et al., 1993a; McIntire et al., 1997). The excitatory effect of GABA on these muscles is mediated by exp-1, an unusual cation-selective GABA receptor, which is expressed in the intestinal and anal depressor muscles (Beg and Jorgensen, 2003). There is little known about the development and specification of these muscles, except that they are non-striated muscles that require Twist, encoded by the hlh-8 gene, for their formation. Loss of hlh-8 also produces an Exp phenotype, because these muscles are missing or are severely reduced in the mutant(Corsi et al., 2000).

In addition to mutants defective in the execution of the DMP steps, a number of mutants that affect the periodicity of the defecation cycle have been identified (e.g. Branicky et al.,2001; Iwasaki et al.,1995). In general, these mutants are distinct from those affecting the steps of the DMP, suggesting that the expression of the DMP is regulated by mechanisms that are distinct from those required for maintaining the periodicity of the DMP. The characterization of some of these genes has shown that calcium-dependent signals in the intestine play an important role in regulating the periodicity of the DMP. In particular, loss of itr-1(dec-4/lef-1), which encodes the worm inositol triphosphate receptor(IP₃ receptor) and regulates calcium release from the endoplasmic reticulum, slows down the defecation period(Dal Santo et al., 1999). Furthermore, the same authors showed that calcium levels peak in the intestine just before the first muscle contraction of the DMP, and that expression of the itr-1 in the intestine is sufficient for normal rhythm generation. Moreover, mutations in unc-43, the C. eleganscalcium-dependent calmodulin Kinase II (CaM Kinase II), which is widely expressed in neurons, muscles and the intestine, result in multiple behavioral defects, including defecation phenotypes(Reiner et al., 1999).

The characteristics of several other mutants also point to the intestine as the main modulator of the DMP periodicity. For example, flr-1 and flr-4 mutants, originally identified on the basis of their resistance to fluoride (Katsura et al.,1994), have very short defecation cycle lengths(Iwasaki et al., 1995). Intestinal expression of both flr-1, which encodes an ion channel of the degenerin/epithelial sodium channel superfamily, and flr-4, which encodes a serine/threonine protein kinase, is sufficient to rescue their respective defects (Take-Uchi et al.,1998; Take-Uchi et al.,2005). The elo-2 gene, which encodes a fatty acid elongation enzyme that is expressed in the intestine, also appears to be involved in regulating the defecation cycle length as RNAi-mediated suppression of elo-2 results in a shortened defecation cycle length(Kniazeva et al., 2003). Finally, loss of function of dsc-4, which encodes the worm microsomal triglyceride transfer protein (MTP), and is probably required in the intestine for lipid transport, also causes a shortened defecation cycle length(Branicky et al., 2001; Shibata et al., 2003).

We have studied the altered defecation cycle of the clk-1 mutants,in which a number of behavioral, developmental and physiological features are abnormal (Branicky et al.,2001; Wong et al.,1995). clk-1 encodes a highly conserved protein that is required for ubiquinone (UQ or CoQ) biosynthesis(Ewbank et al., 1997). We have analyzed the defecation behavior of clk-1 mutants in some detail and found that in the mutants the cycle length is both increased and insensitive to changes in temperature. To investigate these defects further, we have carried out a screen to identify suppressor mutations (which we call dsc, for defecation suppressor of clk-1) and identified two classes of suppressor mutants. The class I dsc mutants (which include dsc-4) suppress both the lengthened defecation cycle and the temperature insensitivity; the class II mutants (which include dsc-1)suppress only the lengthened defecation cycle, but does not restore normal reaction to changes in temperature(Branicky et al., 2001).

Here we report our characterization of dsc-1. We find that the dsc-1 mutants have a shortened defecation cycle length and are also expulsion defective. We cloned dsc-1 and found that it encodes a Paired-like homeobox transcription factor that is expressed in some sensory neurons and in the enteric muscles. dsc-1 is the first homeobox gene of its class that is found to be expressed in, and important for the differentiation of, non-neuronal cells in C. elegans. We show that expression of dsc-1 in the intestinal and anal depressor muscles alone is sufficient to rescue both the defecation cycle length and expulsion defects. However, a variety of experimental manipulations allow for the uncoupling of the effects of dsc-1 on these two phenotypes,suggesting a previously unanticipated role for the enteric muscles in regulating the defecation cycle length. We also show that dsc-1(+)activity is required for the expression of exp-1 in the intestinal and anal depressor muscles, and that the expulsion defect of the dsc-1 mutant is due to the loss of exp-1 expression.

Animals were cultured at 20°C as described(Brenner, 1974), and were fed the Escherichia coli strain OP50. The dsc-1(qm133) allele was isolated in an EMS mutagenesis screen(Branicky et al., 2001) and was outcrossed five times; the dsc-1(tm241) allele was obtained from the National Bioresource Project for the nematode, and was outcrossed three times.

The mutations used to refine the genetic position of dsc-1 include lin-15(n765), flr-4(ut7) and unc-7(e5). dsc-1 was also mapped relative to the mnDF20 deficiency using the SP278 strain (mnDp1 (X;V)/+V; mnDf20 X). Other mutations and strains used in this study include exp-1(sa6) I, EG1653 oxIs22 [Punc-49:UNC-49::GFP,lin-15(+)] II (Bamber et al.,1999), clk-1(qm30) III, unc-25(e156) III and hlh-8(nr2061) X (Liu et al.,1999).

Defecation was scored as described in Branicky et al.(Branicky et al., 2001). Each animal was scored for five consecutive cycles at 20°C. To quantify the Exp phenotype, animals were also scored for five consecutive defecation cycles,which includes six consecutive posterior body contractions (pBocs), and the number of pBocs followed by an expulsion was determined. Student's t-test was used to determine statistical significance and was carried out using Graphpad Prism 4.0 software.

dsc-1 had previously been mapped to the right arm of LG X, between unc-3 and lin-15(Branicky et al., 2001). Using 2-point, 3-point and deficiency mapping strategies, we further refined the genetic position of dsc-1 and determined that it is between the two cloned genes pag-3 and unc-7 (see www.wormbase.org). The six cosmids that correspond to this region were injected into the dsc-1 mutants as a single pool, in pairs, and singly, and it was determined that the dsc-1 mutants could be rescued by all injection mixes containing the C18B12 cosmid. The three predicted genes on the C18B12 cosmid were amplified from the cosmid by PCR and were tested for rescuing activity. A PCR product corresponding to the predicted gene C18B12.3 (from 30798 to 24611 on C18B12) was able to rescue the dsc-1 mutants. All cosmids and PCR products were injected at a concentration of 50 ng/μl with the co-injection marker ttx-3::gfp(Hobert et al., 1999) at a concentration of 150 ng/μl.

To determine the nature of the qm133 mutation, genomic DNA was extracted from dsc-1(qm133) mutants and the predicted coding region was sequenced from both strands. We identified a G-to-A transition at position 695 of the coding region relative to the ATG (position 26250 of the C18B12 cosmid) resulting in an R-to-H substitution at residue 232 of the protein.

First-strand cDNA libraries were generated by the reverse transcription of total RNA extracted from adult worms, using a polydT primer and reverse transcriptase. The dsc-1 transcript was amplified from the library with primers corresponding to the predicted transcript and primers complementary to the polyA tail. The dsc-1 transcript contains five exons and is 933 bp long. The 3′ UTR is 284 bp long, and probably uses one of two possible AAUAAA polyadenylation signals, located 14 or 18 bases upstream of the cleavage site. The dsc-1 transcript does not seem to be trans-spliced, as the transcript could not be amplified using primers corresponding to either the SL1 or the SL2 splice-leader sequences, although there is a consensus trans-splice site located 14 bases upstream of the start codon.

For transcriptional and translational dsc-1::gfp reporter fusions,portions of the dsc-1 promoter and genomic region were amplified from the C18B12 cosmid with primers containing synthetic PstI and NheI sites, and were cloned into the PstI and XbaI sites of the pPD95.77 vector, kindly provided by A. Fire. Unless otherwise indicated, constructs were injected into N2 and dsc-1 at a concentration of 50 ng/μl along with the transformation marker pRF4[rol-6(su1006)] or ttx-3::gfp at a concentration of 150 ng/μl.

The insert contains bases 30284-28330 of C18B12, which includes ∼2 kb upstream of the initiating ATG of dsc-1 and the first 18 bp.

The insert contains bases 30284-25275 of C18B12, which includes ∼2 kb upstream of the initiating ATG of dsc-1 and the entire genomic sequence, excluding the stop codon.

For this construct, gfp expression could be observed in 4/4 lines,all of which were rescued. However, rescued lines also produced ∼30%non-rescued transgenic worms.

The Nde box is a concatamerized element of the ceh-24 promoter that was previously found (Harfe and Fire,1998) and has been used to drive expression specifically in the enteric muscles (Bulow et al.,2002). It was amplified from the pBH10.21 clone and was cloned into the PstI and XbaI sites of the pPD95.77 vector. (The PstI site was introduced in the primer and the XbaI site was internal to the PCR product.) The dsc-1 cDNA (–19 to +930) was amplified from a cDNA library with primers containing synthetic KpnI sites, and was cloned into the KpnI site of pPD95.77. This construct was injected into dsc-1 at concentrations of 25 and 100 ng/μl along with the transformation marker ttx-3::gfp (at concentrations of 175 and 100 ng/μl). For this construct, gfp expression could only be observed in 3/5 transgenic lines; all three were rescued. However, rescued lines also produced ∼30% non-rescued transgenic worms.

This was kindly provided by E. Jorgensen(Beg and Jorgensen, 2003).

This fusion was created using a PCR approach essentially as described previously (Hobert, 2002); the construct is not a GFP fusion, so as not to interfere with the function of EXP-1. The Nde-box portion was amplified from the pBH10.21 clone and the exp-1 gene (from –60 to +4128, which includes the UTRs) from genomic DNA. The fusion PCR product was injected in exp-1 and dsc-1 mutants at a concentration of ∼1 ng/μl together with the transformation marker ttx-3::gfp at a concentration of 100 ng/μl. For both backgrounds 4/4 lines were rescued. Rescued lines produced ∼10%non-rescued transgenic worms.

A portion of the dsc-1 cDNA (+33 to + 933) was amplified from a first-strand cDNA library with primers containing synthetic NheI and PstI restriction sites and was cloned into the corresponding sites in the pPD129.36 vector, kindly provided by A. Fire. The cDNA flanked by the T7 promoter sites was then amplified from the clone and was used as the template for in vitro transcription (Promega Ribomax). Double-stranded RNA was injected into wild-type animals at a concentration of ∼1 μg/μl as described(Fire et al., 1998). The F1 progeny were scored for their defecation cycle length and number of pBocs associated with expulsions.

The dsc-1(qm133) mutant was identified in a screen for suppressors of the slow defecation phenotype of clk-1 mutants(Branicky et al., 2001). In addition to decreasing the defecation cycle length of clk-1 mutants, dsc-1 also significantly shortens the defecation cycle length in a wild-type background (Fig. 1A). In addition to affecting the periodicity of the cycle, dsc-1 is also required for the proper execution of the defecation motor program. dsc-1 mutants have normal posterior and anterior body muscle contractions, but lack the enteric muscle contractions(Fig. 1B). As this step is necessary for the expulsion of the gut contents, dsc-1 mutants also have a distended gut, as the gut lumen fills with bacterial debris(Fig. 2B). This phenotype,termed the `Constipated' or Con phenotype, is common to most if not all expulsion defective mutants (Fig. 2C,D) (Hobert et al.,1999; McIntire et al.,1993b; Reiner and Thomas,1995; Thomas,1990).

dsc-1 was previously mapped to LGX between unc-3 and lin-15 (Branicky et al.,2001). Using 2- and 3-point mapping strategies we determined that dsc-1 is tightly linked to, and to the left of both unc-7and flr-4. Using the chromosomal deficiency mnDf20, which deletes pag-3 but not dsc-1, we determined that dsc-1 is right of pag-3. The six cosmids that cover the region between pag-3 and flr-4 were injected into dsc-1 mutants as a single pool, then in pairs and then singly. All injection mixes containing the C18B12 cosmid were capable of rescuing all phenotypes of the dsc-1 mutants. Of the three predicted genes on the C18B12 cosmid, a PCR product corresponding to only one, C18B12.3, was capable of rescuing dsc-1 mutants (Fig. 1).

We sequenced the coding region of C18B12.3 and found a G-to-A point mutation that results in an R-to-H substitution at residue 232 of the protein(Fig. 3B). In addition, a strain harboring a deletion in the C18B12.3 gene, (C18B12.3(tm241)),phenocopied and failed to complement the dsc-1(qm133) mutant(Fig. 1 and data not shown). Finally, injection of double-stranded RNA corresponding to the C18B12.3 coding region into wild-type animals also phenocopied the dsc-1(qm133)mutation (Fig. 1). Together these observations indicate that dsc-1 corresponds to the predicted gene C18B12.3 and suggest that the dsc-1(qm133) mutation may be a genetic null.

dsc-1 encodes a Q50 Paired-like (or Prd-like) homeodomain protein. Paired-class proteins are defined by the presence of a homeodomain (HD)that resembles that encoded by the Drosophila prd gene and are characterized by six diagnostic residues in the HD. Three sub-classes can be defined according to the residue at position 50 of the HD. The Pax or Prd-type genes are characterized by a serine at position 50 and also contain a second DNA-binding domain, the so-called Paired domain (PD). The other two sub-classes lack the PD (and are therefore Prd-like) and have either a lysine(K₅₀) or a glutamine (Q₅₀) at position 50, the latter being considered to be the most ancestral of the three sub-families (reviewed by Galliot et al., 1999). Comparison between the homeodomain of DSC-1 with the consensus Q₅₀Prd-like homeodomain (Fig. 3C)reveals that the DSC-1 HD contains only four of the six diagnostic residues for the Prd class, indicating that it is somewhat diverged. In addition, dsc-1 does not appear to be orthologous to any of the more than 12 Q₅₀ Prd-like sub-families already described, suggesting that dsc-1 encodes a novel Q₅₀ Prd-like protein. In addition, dsc-1 does not appear to have orthologs in any species other than C. briggsae, suggesting that it might be a recently evolved,nematode-specific, or even Caenorhabditis-specific, gene.

The severity of phenotype of the qm133 mutant, which harbors an R-to-H substitution at position 53 of the homeodomain is the same as that of the tm241 deletion mutant, which lacks the HD entirely. This suggests that the qm133 mutation may entirely abolish the binding of the dsc-1 HD to its target DNA. Indeed, the Arg53 residue is one of the most highly conserved residues of the HD from a broad range of species and has been shown to contact the DNA phosphate backbone in the paired- and engrailed-DNA crystal structures(Kissinger et al., 1990; Wilson et al., 1995). In addition, it has been shown that mutations at position R53 of human Q₅₀ Prd-like genes CHX10 and HESX1, which result in microphthalmia and septo-optic dysplasia, respectively, entirely abolish binding to target DNA in vitro (Dattani et al.,1998; Ferda Percin et al.,2000).

In C. elegans, there are 13 genes encoding Prd-like homeobox proteins (six are K₅₀ Prd-like and seven are Q₅₀Prd-like genes), eight of which (unc-4, unc-30, unc-42, ceh-10, ceh-17,ceh-36, ceh-37 and ttx-1) have been studied genetically, and all have been shown to be required for particular features of a subset of neurons(Altun-Gultekin et al., 2001; Baran et al., 1999; Eastman et al., 1999; Jin et al., 1994; Lanjuin et al., 2003; Miller et al., 1992; Pujol et al., 2000; Satterlee et al., 2001; Svendsen and McGhee, 1995; White et al., 1992). Based on some of these analyses, it has been suggested that this class of homeoproteins may have essential roles in defining neuron subtype identity(Baran et al., 1999). In order to determine the focus of action of dsc-1, we examined the expression pattern of transcriptional and translational gfp reporter fusions.

We found that both the transcriptional and translational fusions are expressed in a subset of bilateral sensory neurons. Based on the positions of the cell bodies and processes we believe we have identified these neurons as AWA, AWB, AWC, ASE, FLP and PVD (Fig. 4A,B,D,E). GFP expression could also occasionally be observed in a sixth pair of unidentified neurons, located to the anterior of AWC (not shown). As expected, we observed expression of the transcriptional fusion in both the cell bodies and axons, and in the cytoplasm as well as in the nucleus, whereas expression of the translational fusion was mostly confined to the nucleus.

In addition to expression in neurons, we also observed expression in the enteric muscles. Expression of the transcriptional fusion could be observed only in the sphincter muscle (Fig. 4C), whereas expression of the translational fusion was observed in the nuclei of the two intestinal muscles (IM) and the anal depressor muscle(AD), but not in the sphincter. This suggests that sequences upstream and internal to the dsc-1 gene might participate in a combinatorial code of expression that identifies as a group all the muscles that are involved in defecation, and that there are sequences in the coding region that prevent expression of dsc-1 in the sphincter and promote expression of dsc-1 in the IM and AD (see Discussion).

In order to determine whether it was the expression of dsc-1 in the neurons, the enteric muscles or both that was necessary for the defecation phenotypes of dsc-1 mutants, we used a synthetic promoter element(the so-called NdE-box), derived from the concatamerization of an element in the ceh-24 promoter (Harfe and Fire, 1998) to specifically drive expression of the dsc-1cDNA in the intestinal and anal depressor muscles(Fig. 5A,B). Indeed, we found that expression of dsc-1(+) in only these two muscle types was sufficient to rescue both the defecation cycle length and expulsion defects of dsc-1 mutants. Moreover, we found that the degree of rescue obtained with this construct was the same as that obtained by a translational fusion that uses the dsc-1 promoter and the entire genomic region containing dsc-1 to drive dsc-1 expression in both the enteric muscles and neurons. This suggests that, for the defecation phenotypes that we have been studying in the dsc-1 mutants, dsc-1(+) is required only in the IM and AD. Consistent with this, we observed that dsc-1(RNAi) only knocks down the expression of Pdsc-1::dsc-1::gfp in these two enteric muscles and not in the dsc-1 expressing neurons. However, dsc-1(RNAi) can completely phenocopy the dsc-1(qm133, tm241) mutations(Fig. 1), indicating that loss of dsc-1 in only these two enteric muscles is responsible for the defecation defects in the mutants.

After having established that the expression of dsc-1(+) in the AD and IM is necessary for the regulation of the defecation cycle length and the expulsion step, we sought to determine the nature of the defect in the mutant muscles. Using a combination of polarized light microscopy,rhodamine-phalloidin staining and expression of a non-rescuing dsc-1(qm133)::gfp reporter, we could observe that the IM, AD and sphincter were all present in the mutants, and had an overall wild-type appearance (data not shown). Given the role of homeoproteins of the Prd-like class in defining neuron sub-type identity, we wondered whether dsc-1could have a similar role in these muscles. A number of genes affecting the expulsion step at the neuronal level, as well as some genes required for the development of the enteric muscles have been identified(Corsi et al., 2000; Hobert et al., 1999; Jin et al., 1999; Mathews et al., 2003). However, only one gene, exp-1, has been shown to be required in these muscles for their function in the execution of the expulsion step. exp-1 encodes an unusual cation-selective GABA receptor, which is expressed in the intestinal and anal depressor muscles as well as in a few other cells (Beg and Jorgensen,2003).

We found that expression of an exp-1::gfp transcriptional reporter could not be detected in the IM and AD in either dsc-1(qm133) or dsc-1(tm241) backgrounds, while gfp expression could be observed when the same transgene was transferred into a dsc-1(+)background (Fig. 6A,B and data not shown). exp-1::gfp expression could also be restored when a dsc-1-rescuing construct was introduced into the dsc-1mutant lines (data not shown). In addition, we observed that dsc-1(+)was not required for the expression of exp-1 in the PDA neuron, a neuron that is known to express exp-1 but is not known to have a role in defecation. Taken together, these results indicate that dsc-1(+)is required for the expression of exp-1 in an enteric-muscle-specific manner.

As a control, we also examined the expression of unc-49 in dsc-1 mutants (Fig. 6C,D). unc-49 encodes a typical anion-gated GABA channel(Bamber et al., 1999), which is expressed in the sphincter, and should promote relaxation of this muscle in response to GABA. However, the exact role of the sphincter in hermaphrodite defecation is less clear than for the other two muscles. Given that unc-49 mutants are not expulsion defective(McIntire et al., 1993a), the sphincter probably plays only a minor role in the expulsion step in hermaphrodites. However, in males there is a remodeling of the system in adulthood such that the expulsion step is achieved by relaxation of the sphincter, rather than by contraction of the IM and AD(Reiner and Thomas, 1995). The expression of unc-49 does not appear to be in any way altered in dsc-1 mutants, which is consistent with the expression pattern of dsc-1. Moreover, we found that dsc-1 mutant males do not have the Con phenotype that is associated with the Exp phenotype(Fig. 2F). This further supports a requirement for dsc-1 only in the IM and AD muscles, and only for the expression of genes such as exp-1, that act in the IM and AD. As has previously been reported(Reiner and Thomas, 1995), we found that exp-1 males are also not Con, whereas unc-25males are severely Con. This is consistent with a requirement for GABA for the expulsion in males, not by acting on EXP-1 to promote contraction of the IM and AD, but rather by acting on UNC-49 to promote relaxation of the sphincter.

In order to test whether the lack of exp-1 expression was responsible for the Exp defect of dsc-1, we used the Nde-box to drive exp-1 expression in the IM and AD muscles(Fig. 6E). We found that this construct was able to fully rescue the Exp defect of exp-1 mutants. We also found that this fusion was also able to rescue the Exp defect of dsc-1, which indicates that the expulsion defect of the dsc-1 mutants is indeed due to the lack of exp-1 expression. However, the rescue achieved with this construct was somewhat weaker in dsc-1 than in exp-1 and also weaker than the rescued achieved with dsc-1(+) expression(Fig. 5C). This suggests that dsc-1 may also be required for the expression of some positive regulators of exp-1 or some other genes that contribute, but are not essential, for enteric muscle function.

Consistent with the finding that the GABA-gated EXP-1 channel is responsible for the dsc-1 phenotype, neither exp-1 nor dsc-1 mutants can be rescued by exogenous GABA (data not shown).

We have described two defecation defects for the dsc-1 mutant: an expulsion defect and a cycle length defect, both of which result from a lack of dsc-1(+) expression in the IM and AD. One possibility is that the shortened defecation cycle length is in fact a secondary effect of the Exp defect. It has been shown for other Exp mutants, such as the unc-43(gf) mutant, that a lack of enteric muscle contractions leads to a progressive shortening of the cycle as the intestine fills with bacterial debris. This eventually leads to a forceful expulsion of the gut contents,which is generally followed by one or two longer cycles(Reiner et al., 1999). We also observed a similar oscillation in cycle length with dsc-1, albeit over a very small range (see legend of Fig. 7), suggesting that the Exp phenotype could contribute to the shortened cycle length of dsc-1, but only in a minor way. Consistent with this, we found that the Nde-box::exp-1 construct, which rescues the Exp defect of exp-1 and dsc-1, fully rescued the cycle length defect of exp-1, but only partially that of dsc-1(Fig. 6E).

Although the cycle length of dsc-1 mutants is affected by the absence of the expulsion step, several observations suggest that dsc-1 has a role in cycle length regulation that is independent of its effect on the expulsion step of the cycle. The first observation relates to the fact that dsc-1(qm133) was identified in a screen for suppressors of the slow defecation cycle phenotype of clk-1 mutants,in which we did not recover any other Exp mutant, although the screen may have begun to reach saturation (Branicky et al.,2001).

Secondly, we examined the defecation cycle lengths of other Exp mutants in both clk-1(+) and clk-1(qm30) backgrounds(Fig. 7). Our survey included unc-25, a mutant of GAD (Jin et al., 1999), hlh-8, a mutant that affects the development of the enteric muscles (Corsi et al.,2000), and exp-1, which, as described above, affects the enteric muscle contraction in response to GABA(Beg and Jorgensen, 2003). All mutants showed the oscillation in cycle length characteristic of Exp mutants(see legend of Fig. 7). We found that only exp-1, like dsc-1, significantly shortened the average cycle length in a wild-type background. The others(unc-25 and hlh-8) significantly increased the average cycle length (Fig. 7A), which suggests that an expulsion defect per se does not necessarily lead to an average shortening of the defecation cycle length. Moreover, we found that, by contrast to what is observed with dsc-1 mutants, none of the Exp mutants, including exp-1, could suppress the lengthened defecation cycle of clk-1 mutants (Fig. 7B).

Thirdly, we observed that overexpression of dsc-1 can significantly increase the defecation cycle length. The transgenic lines that express dsc-1, either under the control of the dsc-1promoter or the Nde-box, which presumably leads to overexpression, have significantly longer defecation cycle lengths than the wild type(Fig. 5C; Fig. 1B). As a loss of dsc-1 activity results in a shortened defecation cycle length, this suggests that the level of dsc-1(+) expression is somehow involved in setting the defecation rate. A similar observation has been made with the itr-1 gene, except that the effect is reversed (loss-of-function mutations slow down, whereas overexpression speeds up, the cycle)(Dal Santo et al., 1999).

Fourthly, we found that dsc-1 can also significantly shorten the lengthened defecation cycle of the itr-1(sa73) mutant (data not shown). This again supports a role for dsc-1 in cycle length regulation, and suggests that dsc-1 affects the cycle length by a mechanism that is independent of the regulation that depends on calcium signaling in the gut. However, given that itr-1(sa73) is not a null allele, potential interactions between dsc-1 and itr-1 need to be further investigated.

The Paired class of homeobox genes, which are related to the Drosophila gene paired (prd), can be divided into three sub-classes according to the residue at position 50 of the HD. The Pax or Prd-type genes are characterized by a serine at position 50 and also contain a second DNA-binding domain, the so-called Paired domain (PD). The other two Prd-like sub-classes lack the PD and have either a lysine(K₅₀) or a glutamine (Q₅₀) at position 50 (reviewed by Galliot et al., 1999). A number of orthologous gene families have been identified within each sub-class. For example, more than 15 orthologous families of K50 or Q50 Prd-like genes have been identified. In C. elegans, there are 13 Prd-like genes that fall into only eight of these families.

Caenorhabditis elegans Prd-like genes have mostly been studied for their roles in the nervous system and, with one exception, have all been shown to have roles in specifying neuron sub-type identity: that is, particular aspects of neuronal function such as choice of neurotransmitter synthesis,connectivity or sensory abilities. Briefly, unc-30, the worm Rx family ortholog, specifies the identities of 19 GABAergic neurons(Eastman et al., 1999; Jin et al., 1999). The worm Otx family orthologs ttx-1, ceh-36 and ceh-37 specify the identities of the ASH, the AWB and AWC, and the ASE sensory neurons,respectively (Lanjuin et al.,2003; Satterlee et al.,2001). ceh-10 specifies the identity of the AIY interneuron (Altun-Gultekin et al.,2001; Svendsen and McGhee,1995). unc-4 specifies the identities of the VA motoneurons (Miller et al.,1992; White et al.,1992). ceh-17, a worm Prx family member, also functions in a subset of neurons, but appears to be required for axonal outgrowth rather than subtype identity (Pujol et al.,2000).

Remarkably, the function of some of these genes is precisely conserved in vertebrates. For example, vertebrate orthologs of unc-30, ceh-36 and ceh-37, and ceh-17 (Pitx2, Otx1 and Phox2b, respectively)are all expressed in cell types similar to those in which they are expressed in the worm, and can complement the mutant worm genes(Lanjuin et al., 2003; Pujol et al., 2000; Westmoreland et al., 2001). However, this does not mean that they are not involved in other processes as well. For example, Pitx2 is also involved in the development of the pituitary gland, craniofacial region, eyes, heart, abdominal viscera and limbs(Gage et al., 1999; Kitamura et al., 1999; Lin et al., 1999; Liu et al., 2001).

dsc-1 encodes a Q50 Prd-like protein, which does not appear to have any clear ortholog in any species but C. briggsae. It is expressed in a few sensory neurons (Fig. 4), as well as in the non-striated enteric muscles. We have determined that dsc-1 is required in three of these muscles for the expression of exp-1, a GABA-gated cation channel. Our analysis suggests that, as in other systems, Prd-like genes in C. elegans may play key developmental roles in non-neuronal tissues. We speculate that other Prd-like genes in C. elegans might also be involved in specifying neurotransmitter sensitivity. Indeed, unc-42, another Prd-like gene in C. elegans that is without a clear vertebrate ortholog, appears to be required for regulating glutamate receptor expression in the nervous system(Baran et al., 1999).

Muscles in worms are of two broad types. The main class of muscles is the striated muscles, the so-called `body-wall muscles', which attach to the outer body wall and contract longitudinally for locomotion. The second class is the non-striated muscles, which are a diverse group of single-sarcomere muscles. These muscles include the pharyngeal muscles, which are used for feeding, the uterine and vulval muscles, which are used for egg-laying, the male-tail associated muscles, which are used for mating, and the enteric muscles we have discussed here (the AD, IM and sphincter), which are required for the expulsion part of the defecation behavior(www.wormatlas.org).

Only very few genes required for the development of the enteric muscles have been identified so far. The most extensively characterized is hlh-8, the C. elegans homolog of Twist, a basic helix-loop-helix (bHLH) transcription factor that has an evolutionarily conserved role in mesodermal patterning(Corsi et al., 2000; Harfe et al., 1998). hlh-8 mutants have defects in the patterning of the post-embryonic mesodermal lineages, and in the development of the IM and AD muscles, which develop embryonically. One downstream target of hlh-8 in the AD muscle is the NK-2 class homeobox gene ceh-24, whose expression is dependant on hlh-8(+) activity. However, ceh-24 mutants do not appear to have expulsion defects(Harfe and Fire, 1998),suggesting that the activity of ceh-24 is not necessary for the main aspects of muscle function. hlh-8 is not required for the development of the sphincter muscle (Corsi et al.,2000). In fact, although a number of genes have been reported to be expressed in the sphincter (WormBase Release WS147, www.wormbase.org),none has been shown to be required for its development.

Here we have shown that dsc-1 is required for an aspect of terminal differentiation of the AD and IM muscles; that is, the specification of their sensitivity to GABA. Like hlh-8, dsc-1 does not appear to have a role in the sphincter, although it does contain some elements in its promoter that can drive its expression there. Indeed, the transcriptional dsc-1::gfp reporter is expressed exclusively in the sphincter(Fig. 4C); conversely, the translational reporter is expressed only in the AD and the IM(Fig. 4F). This suggests that there are elements in the dsc-1 promoter that direct expression to the sphincter, while there are elements in the intronic and/or exonic regions that prevent expression in the sphincter but instead promote expression in the AD and IM. We speculate that dsc-1 participates in a combinatorial code of expression that identifies all muscles that are involved in defecation as a group, but also ensures that exp-1 is only expressed in the muscles that need to contract during defecation.

Two types of defecation mutants have been identified in C. elegans: those that affect the aBoc and/or Exp step, and those that affect periodicity. In general, these two classes of mutants are distinct,although a few mutants fall into both classes (e.g. the flr mutants and the unc-43(gf) mutant)(Reiner et al., 1999; Take-Uchi et al., 1998; Take-Uchi et al., 2005). Mutants that affect both the aBoc and Exp steps (defining the aexgenes) have general synaptic transmission defects(Doi and Iwasaki, 2002; Iwasaki et al., 1997). Similarly, there is a class of dominant Exp mutants, also egg-laying defective, that affect another general property, which is muscle excitability(Reiner et al., 1995). Most of the recessive Exp mutants reported are defective in either the function of the GABAergic DVB neuron (Hobert et al.,1999), the synthesis or transport of GABA(McIntire et al., 1993a; McIntire et al., 1993b), or the response of the enteric muscles to GABA(Bamber et al., 1999; Beg and Jorgensen, 2003). Consistently, we find that dsc-1 is required for an aspect of GABA neurotransmission; that is, the expression of the appropriate GABA receptor in the IM and AD.

As described in the Introduction, a number of genes affecting the periodicity of the defecation cycle have been identified: in particular, itr-1 (dec-4/lef-1), elo-2, dsc-4, flr-1 and flr-4. Although these genes appear to affect diverse processes, all affect defecation by acting in the gut, and presumably by altering some of its properties (Dal Santo et al.,1999; Kniazeva et al.,2003; Shibata et al.,2003; Take-Uchi et al.,1998; Take-Uchi et al.,2005). By contrast, we have found that dsc-1 modulates the defecation cycle length by acting in the enteric muscles. As the gut appears to control the rhythm of muscle contraction in this system, we speculate that dsc-1 is required for a property of these muscles that allows them to act in a feedback mechanism from the muscles to the gut. One type of feedback mechanism that is operating in this system works through gut distention. Indeed, a lack of enteric muscle contractions leads to a progressive shortening of the defecation cycle until the contents of the gut are forcefully expulsed by pressure alone, and is then followed by a few longer cycles. We speculate that dsc-1 is necessary for a feedback from the muscles that may involve a humoral signal that acts to coordinate the calcium signal generated in the gut with the timed contractions of the various muscles types. In this context, it is interesting to note that the flr-1 and flr-4 mutants, which have a shortened defecation cycle length, are also Exp. Furthermore, expression of both genes in the intestine alone is sufficient to rescue both the cycle length and the Exp defects. Possibly these genes could somehow be involved in transducing the putative humoral signal from the muscle to the gut.

Interestingly, we have found that dsc-1 suppresses the lengthened defecation cycle of clk-1(qm30) mutants. Although exp-1 is a target of dsc-1 in the IM and AD muscle for the expulsion step of the defecation behavior, dsc-1 probably acts through different targets for cycle length regulation. Indeed mutation of exp-1 cannot suppress the lengthened defecation cycle of clk-1. The identification of the downstream effectors of dsc-1 are likely to give new insights into the slow defecation phenotype of clk-1 mutants and, more generally,into mechanisms that regulate the defecation cycle length in C. elegans.

We thank David Shanks for some phenotypic analyses. We also thank A. Coulson, A. Fire, P. Loria, and E. Jorgensen for kindly providing clones and A. Corsi as well as the Caenorhabditis Genetics Center, which is funded by the National Institute of Health National Center for Research Resources (NIH NCRR), for providing strains. We gratefully acknowledge the NemaPharm Group of Axys Pharmaceuticals for providing the hlh-8(nr2061) deletion strain,the National Bioresource Project for the nematode and S. Mitani for providing the dsc-1(tm241) deletion strain. S.H. was a Canadian Institutes of Health Research (CIHR) Scientist and R.B. was supported by a Mary Louise Taylor McGill Major Fellowship.