Characterization of a dominant negative C. elegans Twist mutant protein with implications for human Saethre-Chotzen syndrome

Understanding how basic helix-loop-helix (bHLH) transcription factors cause changes in gene expression is important for revealing how transcriptional cascades function in development. Several bHLH transcription factors such as MyoD and NeuroD have been shown to play important roles in promoting myogenesis and neurogenesis, respectively (for a review, see Massari and Murre, 2000). These transcription factors work as homo- or heterodimers through contacts made in the helix-loop-helix (HLH) domain. The dimers then bind promoter DNA at canonical sequences called E boxes by virtue of the now close proximity of their individual basic DNA-binding domains (Murre et al., 1989). In general, bHLH heterodimers are composed of one member with tissue-specific expression and another member with more broad expression throughout the organism (Lassar et al., 1991). One tissue-specific bHLH family member, Twist, plays a role in the development of the mesoderm in all organisms examined including humans, mice, C. elegans and Drosophila melanogaster (where the gene was first discovered) (Simpson, 1983; Nüsslein-Volhard, 1984; Chen and Behringer, 1995; Wang et al., 1997; Harfe et al., 1998b; Corsi et al., 2000). In several organisms, the broadly expressed bHLH family members, E (in vertebrates) and Daughterless (in invertebrates), are thought to heterodimerize with Twist (Spicer et al., 1996; El Ghouzzi et al., 2000; Castanon et al., 2001). In C. elegans, the E/Daughterless (DA) homolog is called CeE/DA and, although embryonic expression is ubiquitous, the protein has a restricted expression pattern postembryonically that includes mesodermal tissue expressing CeTwist (Krause et al., 1997).

We are interested in elucidating aspects of mesodermal development and have been studying the function of Twist in the model organism C. elegans. Of the 35 conventional bHLHs identified in the C. elegans genome, a single Twist homolog exists (Ruvkun and Hobert, 1998; Ledent and Vervoort, 2001). Initially, we characterized a putative null allele of the CeTwist gene, hlh-8 and found several defects in the mesoderm of the mutant animals (Corsi et al., 2000).

The C. elegans mesoderm is predominantly composed of body wall muscles used for movement, macrophage-like coelomocytes, vulval and uterine muscles used for egg laying, enteric muscles used for defecation, pharyngeal muscles used for eating, and somatic gonad used for reproduction (Waterston, 1988; Moerman and Fire, 1997). CeTwist is required for the formation of three out of 126 mesodermal cells that differentiate in the embryo (notably, striated body wall muscles were formed in the absence of CeTwist) (Corsi et al., 2000). The three cells that require CeTwist for proper differentiation in the embryo are the non-striated enteric muscles: two intestinal muscles and the anal depressor (Corsi et al., 2000). In postembryonic development, CeTwist is important for patterning of the mesoderm lineage and is necessary for the proper formation of the vulval and uterine muscles (Corsi et al., 2000).

In humans, mutations in the TWIST gene are associated with an autosomal dominant developmental syndrome called Saethre-Chotzen syndrome (for a review, see Wilkie, 1997). The incidence of this disease is 1/25,000-1/50,000 live births (Paznekas et al., 1998). Individuals with this syndrome have craniofacial and digit defects, including drooping eyelids, short fingers, the presence of soft webbing between the fingers and premature fusion of the cranial sutures (causing an abnormally shaped head) (Reardon and Winter, 1994; Cohen, 2000). Greater than 40 unique mutations in the human TWIST gene have been identified, with the majority being stop codons or mis-sense mutations in the highly-conserved bHLH domains (reviewed by Gripp et al., 2000). Most of these mutations are considered to be haploinsufficient. The best evidence for haploinsufficiency is that humans with a heterozygous deletion of the TWIST locus exhibit Saethre-Chotzen phenotypes (Johnson et al., 1998).

We have been studying an interesting allele in C. elegans hlh-8 called n2170 that may have parallels to the human Saethre-Chotzen disease. This allele behaves as a semidominant mutation. Surprisingly, although we found that the mutation was an amino acid substitution in the basic DNA-binding domain, the mutant CeTwist protein still bound DNA in vitro but could not activate transcription in vivo. Overexpression of the mutant CeTwist protein caused a phenotype similar to the n2170 allele, suggesting the mutant could act as a dominant-negative protein. Cellular characterization of the mesodermal defects in mutant hlh-8 (n2170) animals provided an assay to test for dominant-negative function of several basic domain mutations in the human TWIST gene of some individuals with Saethre-Chotzen syndrome.

We determined the location of the point mutation in the hlh-8 locus by amplifying the 2 kb genomic DNA of the hlh-8 locus in two fragments using single animal PCR (Williams, 1995) and sequencing the five exons and splice junctions (Corsi et al., 2000). The G to A transition at base pair 85 was found in four independent animals and the remainder of the hlh-8 locus was wild-type in the n2170sd animals.

We used standard genetic methods to introduce a set of characterized, integrated gfp fusion transgenes into n2170sd and nr2061 (–) genetic backgrounds (Corsi et al., 2000). Analysis was carried out with animals that were homozygous for both the gfp transgene and hlh-8 mutations. gfp reporter strains used were the following: hlh-8::gfp ayIs7(IV) (Harfe et al., 1998b), egl-15::gfp ayIs2(IV) (Harfe et al., 1998a), from the ceh-24 promoter NdEbox::gfp ccIs4656(IV) (Harfe and Fire, 1998), arg-1::gfp ccIs4443 (A. Melendez, I. Greenwald, S. Kostas, J. Liu and A. Fire, personal communication), rgs-2::gfp LX354 (Dong et al., 2000), intrinsic cc:: gfp (J. Fares and S. Kostas, personal communication) and myo-3::gfp ccIs4251(I) (Fire et al., 1998). We confirmed that the egl-15::gfp and NdEbox::gfp reporters were present and homozygous, but not expressed, in n2170sd strains by outcrossing the strains to wild-type (N2) males and observing that all of the cross progeny expressed gfp. For egl-15::gfp, we cloned the heterozygous animals from the N2 cross and observed whether each plate had wild-type animals expressing the gfp to determine whether the parent was homozygous for egl-15::gfp. Rescue experiments of n2170sd animals were performed as described elsewhere (Corsi et al., 2000).

An expression plasmid (pRSETA with CeTwist cDNA) (Harfe et al., 1998b) was used to construct new plasmids containing the n2170sd point mutation or missing the basic domain of CeTwist using the QuikChange Site-Directed Mutagenesis Kit from Stratagene (La Jolla, CA) and appropriate primers (sequences available upon request). cDNA inserts from plasmids containing mutations identified in individuals with Saethre-Chotzen syndrome (described in the next section) were PCR-amplified to create BamHI/EcoRI fragments. The PCR fragments were then cloned into pRSETA and the inserts were sequenced. The mutant proteins along with wild-type CeTwist and CeE/DA were expressed in E. coli and purified under denaturing conditions using Ni²⁺ affinity column chromatography according to Harfe et al. (Harfe et al., 1998b). Gel shift assays using a CAGGTG E box-based oligo were performed as previously described (Krause et al., 1997; Harfe et al., 1998b).

Plasmids were constructed that would express wild-type and CeTwist mutant proteins in the M lineage in order to evaluate their ability to cause a phenotype in the presence of wild-type chromosomal hlh-8. Wild-type and n2170sd cDNA from the pRSETA vectors described above were PCR-amplified to create XbaI/EcoRI fragments. After restriction digest, these fragments were moved into the plasmid pBH56.55 (gift from B. Harfe and A. Fire, Carnegie Institution of Washington). The plasmid pBH56.55 was made by adding ∼500 bp upstream of the hlh-8 coding region into the gfp expression vector pPD95.69 (A. Fire, G. Seydoux, J. Ahnn and S. Q. Xu, personal communication). The cDNA fragments replaced the gfp in the vector. The final wild-type plasmid was expressed in the M lineage based on the criteria that the plasmid rescued the egg-laying defective phenotype of nr2061 (–) animals. The amino acid residues analogous to human TWIST mutations shown in Fig. 1 were also created in this same plasmid backbone by using the QuikChange Site-Directed Mutagenesis Kit from Stratagene. All cDNA inserts that were subjected to PCR were sequenced to detect any PCR-induced errors. All of these constructs will potentially express full-length CeTwist with two additional amino acids (a Ser and Arg directly after the initiator methionine). The mutant and wild-type plasmids were injected using standard techniques (Mello and Fire, 1995) at 50 μg/ml into animals that already had integrated hlh-8::gfp (Harfe et al., 1998b). The hlh-8::gfp contains only hlh-8 promoter sequences and is not a rescuing construct. The dominant marker rol-6 (pRF4) (Mello et al., 1991) was used to follow the presence of injected DNA.

In order to observe the ability of the mutant CeTwist E29K protein to activate various promoters, we constructed strains containing the n2170sd mutation and either egl-15::gfp (Harfe et al., 1998a), rgs-2::gfp (Dong et al., 2000), NdEbox::gfp (Harfe et al., 1998b) or arg-1::gfp (Mello et al., 1994) (A. Melendez, I. Greenwald, S. Kostas, J. Liu and A. Fire, personal communication). These reporter lines were observed for gfp expression in the mutant animals. In the case of strains containing egl-15::gfp and NdEbox::gfp, no expression was seen in n2170sd animals. Expression was restored in transgenic lines (two lines for each reporter) with extrachromosomal arrays containing genomic hlh-8 DNA.

The ability of the mutant CeTwist E29K protein to activate egl-15::gfp in conjunction with CeE/DA was assayed by expressing the proteins singly or in combination from cDNAs using a heat shock promoter from the vector pPD49.83 (Mello and Fire, 1995). Embryos and larvae from stable lines were subjected to a 33°C heat shock for 4 hours followed by a 2-4 hour room temperature recovery period. Ability to activate the reporter was assayed by observing the level of gfp expressed using a GFP dissecting scope to scan a population of animals. The embryos were then processed for antibody staining with polyclonal anti-CeTwist and anti-CeE/DA as previously described (Krause et al., 1990; Krause et al., 1997). The same methods were used to stain n2170sd animals with anti-CeTwist in Fig. 3. The antiserum was raised against a peptide of the last 14 amino acids of CeTwist and was affinity purified.

We have been taking a genetic approach to understand CeTwist function in C. elegans development. Previously, we showed that animals with a deletion mutation in hlh-8, called nr2061, that removes the HLH domain of CeTwist are viable and fertile (Corsi et al., 2000). The nr2061 mutant animals lack nonstriated muscle function required for egg-laying and for defecation and are, therefore, egg-laying defective and constipated (Corsi et al., 2000). Because the nr2061 animals are most probably null for hlh-8 function, we were interested in finding non-null alleles that might provide further insight into CeTwist function. A potential allele, n2170, has been isolated by G. Garriga and H. R. Horvitz (personal communication) in a screen for egg-laying defective animals. The n2170 animals are also constipated. We have characterized this allele in detail.

n2170 and nr2061 homozygous animals appeared identical under the dissecting microscope: they were visibly constipated with an expanded intestine and failed to lay eggs. Both strains were fertile with embryos developing inside the hermaphrodite uterus causing a ‘bag-of-worms’ phenotype. The two alleles failed to complement. Trans-heterozygous animals (n2170/nr2061) were constipated and egg-laying defective (data not shown). Heterozygous nr2061/+ animals were wild type, indicating that nr2061 is recessive (Corsi et al., 2000). By contrast, n2170/+ heterozygous animals were egg-laying defective; they could lay their eggs but not at a wild-type rate and hermaphrodites eventually bloated with unlaid embryos. However, n2170/+ heterozygotes were not visibly constipated, thus we considered the allele semidominant rather than dominant. For the remainder of this paper, hlh-8 (n2170) will be referred to as n2170sd and hlh-8 (nr2061) as nr2061(–).

We mapped the n2170sd mutation to the X chromosome and found that the allele is covered by the small duplication, stDp2, in the region of the hlh-8 locus. The hlh-8-coding sequence in n2170sd mutant animals contained a single base pair change resulting in a glutamic acid to lysine (E29K) alteration in the basic domain (Fig. 1). The affected amino acid is conserved among Twist family members and less-related bHLH family members such as C. elegans MyoD and E/Daughterless homologs (Fig. 1). An independently identified allele, n1897 (M. Stern and H. R. Horvitz, personal communication) had the identical point mutation in the hlh-8 locus, and n1897 was not characterized further. The egg-laying and constipation phenotypes of n2170sd animals were fully rescued by an extrachromosomal array containing wild-type CeTwist expressed from a 15 kb genomic DNA fragment (with approximately 9 kb of 5′-UTR and 3 kb of 3′-UTR from the hlh-8 locus) (Corsi et al., 2000). It has previously been shown that the expression of wild-type CeTwist is necessary for the activation of two target gene reporters, an FGF receptor homolog, egl-15, and an NK homeodomain homolog, ceh-24 (Harfe et al., 1998b; Corsi et al., 2000). Expression of egl-15 and ceh-24 reporters was not detected in n2170sd animals, but was restored to wild-type pattern in n2170sd animals that carried the extrachromosomal array containing wild-type genomic CeTwist DNA (data not shown). Altogether, these data indicate the phenotype of the n2170sd animals was due to the point mutation we found in hlh-8.

We wanted to understand the cellular cause of the defecation and egg-laying defects observed in homozygous n2170sd animals. We previously performed an extensive characterization of mesodermal specification and patterning in nr2061 (–) putative null animals (Corsi et al., 2000). We observed similar cellular defects in mesodermal patterning in n2170sd and nr2061 (–) animals, with one notable exception (see next section). n2170sd animals, like nr2061 (–) animals, lack three out of four non-striated enteric muscles. The loss of enteric muscles in the n2170sd animals accounted for their constipation phenotype, as is true for the nr2061 (–) animals (data not shown) (Corsi et al., 2000).

The major nongonadal components of the wild-type C. elegans postembryonic mesoderm develop from a single blast cell, M (Sulston and Horvitz, 1977) (Fig. 2A). The M mesoblast, located in the posterior of the newly hatched animal, consistently goes through 4 sets of cell divisions in a stereotypical spatial and temporal manner to form 18 descendants that will become two coelomocytes (cc), 14 body wall muscles (bwm) and two sex myoblasts (SMs). The two SMs migrate to the center of the animal, where the vulva is forming, and divide three times each to form 16 descendants that will become eight vulval and eight uterine muscles (collectively called sex muscles). All of these events can be followed in live animals using the hlh-8::gfp reporter that contains only promoter sequence and is not a rescuing construct (Harfe et al., 1998b). As observed in nr2061 (–) animals, three major defects were found in the M lineage in n2170sd animals: the initial cell divisions of M did not occur in the proper division planes, an incorrect number of M descendants were formed (e.g. extra SMs in some animals possibly due to bwm to SM fate transformations), and non-striated sex muscles were not formed properly (Corsi et al., 2000) (Tables 1, 2). The lack of vulval muscles explained why the n2170sd animals were unable to lay any eggs.

Although many of the phenotypes observed in nr2061 (–) animals were also apparent in n2170sd animals, a striking difference existed in the postembryonic development of the two strains. The SMs in n2170sd animals had a different phenotype from nr2061 (–) or wild-type animals. In wild-type animals, the SMs were born in the posterior of the animal and migrated to the center of hermaphrodites where they proliferated and differentiated into vulval and uterine muscles by the end of the fourth larval stage (Fig. 2A,B) (Sulston and Horvitz, 1977). In both n2170sd and nr2061 (–) animals, the SM number may differ from 2 (Table 1). In n2170sd animals, the SMs often failed to divide, whereas the SMs in nr2061 (–) animals divided but sex muscles did not differentiate properly (Table 1; Fig. 2C,D). Some of the SMs in n2170sd animals migrated to, or past, the region of the developing vulva where they could receive nonautonomous signals required for division, yet they still did not divide (Fig. 2D,E). These undivided SMs sometimes persisted in n2170sd young adults (Fig. 2E). Initially, wild-type and hlh-8 mutant animals were observed at the L4 stage, where, in wild-type animals, 16 SM descendants were apparent by hlh-8::gfp. Small GFP-expressing cells characteristic of SM descendants were found in nr2061 (–) animals and large SM-blast like cells were found in n2170sd animals at this stage and later (Table 1; Fig. 2C,D). The cells with SM morphology in n2170sd animals could result from underdivided or undivided SM blast cells. These possibilities were tested by observing single animals at stages younger than L4. We observed hlh-8::gfp in single wild type (n=6), nr2061(–) (n=7) and n2170sd (n=25) animals at 15-30 minute intervals over a 3.5-4 hour time period. During this time period, wild-type and nr2061(–) SMs divided at least one time in every animal (data not shown). By contrast, the SMs divided at least once in only 36% of n2170sd animals (data not shown). Only those animals showing evidence of vulval development by the end of the experiment were reported to ensure that the time interval included the period when the SMs could possibly be dividing. Although the inability to lay eggs because of the loss of properly formed sex muscles was the same in both hlh-8 mutant strains, the underlying cellular defects were distinct in the two strains.

To examine whether the defects in n2170sd animals resulted from E29K CeTwist protein instability, improper localization or a loss of hlh-8 expression, we performed immunofluorescence studies on the mutant animals. The mutant E29K protein was observed in nuclei of large cells in the center of late fourth larval stage hermaphrodites (Fig. 3). The size and position of these cells suggested they were the SM-like cells found in n2170sd animals at this stage. The mutant protein was localized to nuclei by the coincident staining with the DNA dye, DAPI (Fig. 3D). The subcellular localization of the E29K CeTwist was indistinguishable from that of the wild-type protein (data not shown). Therefore, the mutant protein was expressed in the appropriate cells and located in the nucleus.

To test whether the mutant CeTwist E29K acted as a dominant-negative protein, we expressed the mutant protein from a plasmid in animals that were wild type for hlh-8 at the chromosomal locus. We looked for defects in the postembryonic mesoderm that could be observed with hlh-8::gfp in the transgenic animals during the third (L3) and fourth (L4) stages of larval development. Animals that were expressing extra CeTwist from a plasmid containing the wild-type CeTwist cDNA had some defects (4% of animals at L3 stage and 12% at L4 stage) compared with animals not expressing extra CeTwist (0% at the L3 stage and 5% at the L4 stage) (Fig. 4A). By contrast, at the L3 stage, almost 70% of the animals expressing the mutant CeTwist E29K protein had supernumerary sex myoblasts. This phenotype was similar to nontransgenic animals from nr2061 (–) or n2170sd strains (Table 1). At the L4 stage, 16% of animals expressing CeTwist E29K had extra SM descendants (Fig. 4A). This phenotype is observed in animals with a loss of function of CeTwist (Corsi et al., 2000). Similar to the n2170sd homozygotes, 41% of the transgenic animals expressing the E29K mutant CeTwist at the L4 stage had SMs that did not divide. Therefore, expressing mutant CeTwist E29K in the presence of wild-type chromosomal hlh-8 could mimic the semidominant phenotype in the n2170sd animals.

As the n2170sd mis-sense mutation in the DNA-binding domain was able to dominantly cause a phenotype when expressed from a transgene, we reasoned that other mis-sense mutations in the same domain might also cause similar phenotypes. We performed a similar experiment using several selected alleles that had been identified in individuals diagnosed with Saethre-Chotzen syndrome [hR116W (Paznekas et al., 1998); hR118H (Rose et al., 1997); hQ119P (Howard et al., 1997)]. Although the DNA-binding domain in human Twist is not 100% identical to the same domain in CeTwist, the amino acids that have been altered in individuals with Saethre-Chotzen syndrome are conserved between humans and C. elegans so we made the mis-sense mutations directly in the C. elegans hlh-8 cDNA (Fig. 1). We expressed transgenes with the engineered mutant version of the C. elegans hlh-8 cDNA in wild-type animals and examined the post-embryonic mesodermal development. We found that expressing CeTwist with the hQ119P mutation caused phenotypes similar to CeTwist E29K (Fig. 4A,F-I). These animals also had a high number of underdivided SMs (Fig. 4A). The hR116W mutation had a slight mutant phenotype at the L3 stage (twice as many animals had extra SMs than those expressing wild-type CeTwist), and these transgenic animals had undivided SMs and extra SM descendants at the L4 stage (Fig. 4A). Expression of the third mutation, hR118H had no phenotype at the L3 stage but did have a similar phenotype to hR116W at the L4 stage (Fig. 4A). Therefore, the effect of expressing CeTwist with a mutant basic domain can be observed at the cellular level (Fig. 4B-I), providing a powerful assay for dominant negative activity of TWIST point mutations.

In order to explore the biochemical nature of the E to K mutation in n2170sd animals, we investigated the behavior of the mutant protein in an in vitro DNA-binding assay. Because the n2170sd point mutation creates a charge change in the basic domain, we predicted that DNA binding by the mutant protein might be disrupted. To test whether the protein would bind DNA, we purified a His-tagged mutant protein in parallel with wild-type His-tagged CeTwist in an E. coli expression system. As previously shown by Harfe et al. (Harfe et al., 1998b), CeTwist can bind DNA containing an E box sequence as a homodimer or cooperate with CeE/DA to bind as a heterodimer (Fig. 5). The mutant CeTwist E29K protein was also able to homodimerize and heterodimerize with CeE/DA to bind the same sequence in our gel shift conditions (Fig. 5). To be certain that the binding activity we observed with CeTwist E29K was not due to impurities in the protein purification procedure, we also purified His-tagged CeTwist that was missing the entire basic domain (ΔB). The ΔB mutant protein did not shift the same oligo as a homodimer (Fig. 5). Therefore, in a qualitative in vitro assay, the mutant CeTwist E29K protein behaved similar to wild-type CeTwist and could bind DNA on its own as well as with CeE/DA.

The human Twist DNA-binding domain mutations were also purified from E. coli as His-tagged CeTwist proteins. None of these mutants was able to appreciably bind to the E box DNA as a homodimer (Fig. 5). Like ΔB, the human mutations were able to bind DNA as a heterodimer with CeE/DA in our in vitro assay. The ability of all the mutant CeTwist proteins to form heterodimers may account for their dominant negative activity in vivo.

As the mutant CeTwist E29K protein could heterodimerize with CeE/DA in vitro to bind DNA, we tested whether the mutant protein could also function in an in vivo transcription assay. CeTwist and CeE/DA that are ectopically expressed from heat shock promoters can promote robust expression of an egl-15::gfp reporter (Harfe et al., 1998a; Harfe et al., 1998b) (Fig. 6A,C). In order to test whether the mutant CeTwist E29K protein was capable of the same coactivation, we performed an equivalent experiment expressing the mutant protein from the heat shock promoter. The CeTwist E29K protein was unable to activate the egl-15::gfp either on its own or with CeE/DA (data not shown) (Fig. 6E,G). We detected expression of the mutant CeTwist E29K protein from the heat shock promoter by immunostaining with polyclonal CeTwist antibodies (Fig. 6F). This observation ruled out the possibilities that the lack of expression of CeTwist E29K or perhaps the increased degradation of the mutant protein was responsible for the inability to coactivate the egl-15::gfp reporter with CeE/DA. Therefore, the mutant protein is unable to cooperate in this in vivo assay with CeE/DA to promote egl-15::gfp transcription.

We have shown that the hlh-8 (n2170) strain is a semi-dominant allele. Because n2170sd/+ heterozygotes are egg-laying defective, we investigated whether a defect in vulval or uterine muscle development was causing this phenotype in heterozygous animals. We observed the uterine muscles with rgs-2::gfp (regulator of G protein signaling) (Dong et al., 2000), and we found wild-type, nr2061 (–)/+ and n2170sd/+ heterozygotes to be indistinguishable (Fig. 7D,H,L). In wild-type animals, the four vm1 type vulval muscles express egl-15::gfp (Harfe et al., 1998a). Using the egl-15::gfp reporter, we observed that n2170sd/+ heterozygotes never expressed GFP from the egl-15 promoter in the vm1 vulval muscles (n>50), whereas wild-type and nr2061 (–)/+ heterozygotes always expressed the reporter in the vm1 cells (n>50) (Fig. 7B,F,J). In order to distinguish whether the vulval muscles were not formed or whether the muscles were made but the reporter gene was not expressed in that tissue, we used a third gfp reporter arg-1::gfp (apx-1 related gene), which also is expressed in the vm1 vulval muscles (apx-1 and arg-1 are both Notch pathway ligands) (Mello et al., 1994) (A. Melendez, I. Greenwald, S. Kostas, J. Liu and A. Fire, personal communication). We observed wild-type expression of arg-1::gfp in the vm1 muscles in n2170sd/+ heterozygotes (data not shown). This observation indicated that the vulval muscles were made in the n2170sd/+ heterozygotes. Therefore, there was differential gene expression from the two gfp reporters that are expressed in the vm1 vulval muscles when both wild-type and mutant CeTwist E29K proteins are present in the same animal. The egg-laying defect in the n2170sd/+ heterozygotes does not result from a lack of vm1 muscles but was more likely to be due to a problem in their proper functioning caused by differential regulation of CeTwist target genes.

CeTwist plays a role in patterning and formation of a subset of C. elegans mesodermal tissues. By studying the n2170sd allele, more can be learned about how CeTwist functions. In n2170sd animals, the mutant CeTwist protein has a single glutamic acid to lysine change at a conserved position in the CeTwist basic DNA-binding domain. This charge change makes the DNA-binding region even more basic and still allows the mutant CeTwist E29K protein to bind DNA. The mutant CeTwist E29K protein binds to a canonical bHLH E box binding site in vitro as a homodimer and as a heterodimer with the Daughterless homolog CeE/DA protein. However, the cooperation with CeE/DA does not extend to an in vivo assay because mutant CeTwist E29K cannot coactivate an in vivo reporter gene with CeE/DA. Therefore, we propose that the mutant CeTwist E29K can bind to promoters alone as a homodimer or as a heterodimer with CeE/DA but cannot activate transcription, perhaps because of the inability to interact with the transcriptional machinery or with other transcription factors important at CeTwist target promoters (see model below).

How might the point mutation in the basic DNA-binding domain of E29K CeTwist affect the function of the mutant protein? Studies of basic domain mutations in other bHLH proteins suggests several possible ways the E29K mutation might interfere with function. For example, E29K could affect dissociation of the mutant bHLH from target promoters, could cause more stable DNA binding or could cause an altered protein conformation, resulting in defective transcriptional activation. The crystal structure of the bHLH domains of mouse MyoD and E47 has revealed that the amino acid corresponding to E29 in CeTwist plays an especially important role in making crucial contacts with DNA. E47 and E12 are the vertebrate E partners and are splice variants from the same locus. The glutamic acid (E118 in MyoD; E345 in E47) at the position that corresponds to E29 in CeTwist contacts DNA at the cytosine and adenine bases in the E boxes CAGCTG for MyoD or CACCTG for E47 (Ma et al., 1994; Ellenberger et al., 1994). In vitro DNA-binding studies have been performed with several E12 mutants. Sieber and Alleman (Sieber and Alleman, 1998) found that an E345Q E12 mutant protein dissociated less well from both specific and nonspecific E12 binding sites. Vitola and colleagues have found that mutations in the DNA-binding domain of E12 that cause the overall charge change of the domain to be more positive result in mutant proteins that have increased and more stable DNA-binding activity (Vitola et al., 1996). Basic domain mutations in MyoD reveal the importance of the domain for DNA binding as well as for transcriptional activation (Bengal et al., 1994). Some mutations in the DNA-binding domain of MyoD can still bind DNA as homodimers and as heterodimers with E47, but these mutants are not able to activate transcription in vitro (Bengal et al., 1994). For example, an A114N mutation in the DNA-binding domain of MyoD that still binds DNA has been proposed to alter the conformation of the activation domain to block transcription (Huang et al., 1998). Similarly, the E29K mutant CeTwist might have an altered conformation that is blocking the ability of the mutant protein to participate in transcriptional activation. Further biochemical characterization of the CeTwist E29K protein would be required to provide evidence for this possibility.

We propose that the n2170sd mutation created a dominant-negative CeTwist mutant protein. Because the null mutation hlh-8 (nr2061) is fully recessive, it is unlikely that haploinsufficiency at the hlh-8 locus could cause a semidominant mutant. Additionally, in homozygous hlh-8 (n2170sd) animals, a protein that is recognized by a CeTwist antibody is present in a wild-type pattern of expression so these animals are not suffering from a lack of CeTwist caused by nonexpression of the hlh-8 locus or protein instability caused by the point mutation. We have observed that a plasmid containing genomic CeTwist DNA can fully rescue the egg-laying and constipation phenotypes of n2170sd animals. As extra CeTwist can rescue, it is unlikely that the n2170sd allele results in too much hlh-8 activity that would be characteristic of a strong gain-of-function mutation. The mutation is classified genetically as semidominant because we observed some but not all of the same defects in both homozygotes and heterozygotes. For example, both homozygotes and heterozygotes have egg-laying defects but the homozygotes cannot lay eggs at all, whereas the heterozygotes lay eggs at a slower than wild-type rate. The phenotypes we observed in the mesoderm of n2170sd animals were all in cells where wild-type CeTwist is normally found, and the phenotypes were rescued by overexpressing wild-type CeTwist, suggesting that n2170sd is not a neomorphic mutation. Naturally occurring dominant negative bHLH proteins exist. The vertebrate protein Id and the Drosophila protein Emc are HLH proteins that lack a basic domain (reviewed by Massari and Murre, 2000). These proteins can still heterodimerize with other bHLH proteins but the resulting heterodimer cannot bind DNA. In effect, these proteins sequester their partner bHLH from functioning. The mutant CeTwist E29K protein does not appear to act by the same mechanism as Id and Emc, as the protein can heterodimerize with CeE/DA and bind to DNA in vitro. Rather, the protein may be sitting unproductively at promoter sequences with CeE/DA and/or other proteins.

While observing n2170sd animals, we discovered an earlier cellular defect not seen in the nr2061 (–) animals. In wild-type animals, the SMs that are born in the postembryonic mesodermal lineage divide to become 16 sex muscles required for egg laying (Fig. 2A). These SM cells either did not divide at all or did not divide at the wild-type rate in the n2170sd animals, whereas the cells divided but did not form proper sex muscles in nr2061 (–) animals. In both mutant strains, the animals could not lay any eggs because the muscles that open the vulva were not present and functioning. However, the cellular defects were distinct, and we would argue are more severe in the n2170sd animals, suggesting an interference with, or titration of, a factor or pathway required for SMs to divide properly. In heterozygous n2170sd/+ animals, the SMs can divide and differentiate as we observed both vulval and uterine muscles that are the products of the SMs in adult animals (Fig. 7). This observation demonstrates that a single copy of the n2170sd allele is insufficient to block all SM divisions.

The decreased amount of divisions of the SMs in homozygous n2170sd animals cannot be explained by the lack of migration of the SMs. Although some of the undivided SMs did not migrate and remained in the posterior of n2170sd animals, some of the SMs migrated to the appropriate region of the developing vulva and still did not divide properly (Fig. 2D,E). If spatial signals were necessary to initiate the cell divisions, then at least those SMs near the vulva should have divided in the same way as nr2061 (–) SM descendants in n2170sd animals. In wild-type animals at the young adult stage, the SM descendants no longer express hlh-8::gfp because the cells have differentiated into sex muscles and have turned off the reporter (Harfe et al., 1998b). In n2170sd young adult animals, the reporter expression often persisted in the undivided SMs (Fig. 2E). This aberrant perdurance could be due to the lack of divisions of the SMs that were now uncoupled with the age of the n2170sd animal or could be due to the inability of the CeTwist E29K protein to regulate hlh-8 expression negatively.

A working model is presented to explain the phenotypes and predict how CeTwist is influencing gene expression based on our observations with hlh-8 mutant animals (Fig. 8). In wild-type animals, CeTwist (along with a partner protein), which in many cases will be CeE/DA, heterodimerizes to promote gene expression from promoters such as egl-15 as indicated in Fig. 8A. A factor that physically interacts with CeTwist, i.e. the heterodimeric partner or a different protein, is responsible for allowing or promoting the SMs to divide in wild-type animals. This factor may act alone or in combination with other factors to turn on cell division targets. The factor(s) is freely available to promote or allow SM divisions when CeTwist is absent in nr2061 (–) homozygotes (Fig. 8B). As there is no CeTwist made in nr2061 (–) homozygotes, no target genes can be activated (i.e. no egl-15::gfp). In nr2061 (–)/+ heterozygotes, half of the wild-type amount of CeTwist is sufficient to turn on downstream targets resulting in animals that are indistinguishable from wild-type animals (Fig. 8C). In n2170sd homozygotes that are producing a dominant negative version of CeTwist, the mutant protein may bind to DNA in vivo but is unable to turn on downstream target genes (Fig. 8D). DNA binding may not be important for dominant negative activity, however, because DNA-binding domain mutations that cause phenotypes (Fig. 4) do not bind DNA efficiently in vitro (Fig. 5). The mutant CeTwist E29K protein is likely to titrate away the factor(s) that promote SM divisions, as we frequently saw SMs that did not divide or were underdivided in n2170sd homozygotes when compared with wild-type animals. n2170sd/+ heterozygotes had a subset of the defects seen in n2170sd homozygotes. Half the amount of defective CeTwist E29K cannot completely sequester the factor(s) required for SM divisions as SM descendants were observed in n2170sd/+ heterozygotes (Fig. 8E). However, some targets, such as the potential target arg-1::gfp, could be activated by the wild-type CeTwist present in these animals. Other promoters such as egl-15::gfp were not turned on, suggesting an important difference among CeTwist promoters that may be revealed in future experiments. The egg-laying defects that were observed in n2170sd/+ heterozygotes may be due to this lack of functional gene products from the egl-15 and other target promoters.

In an alternative model not shown, our results could be explained by invoking CeTwist homodimers rather than heterodimers as the transcriptional regulator. In this case, the default state would be SMs dividing because of other factors, such as cell cycle proteins that have no physical or direct interaction with CeTwist. If CeTwist is not available [e.g. in nr2061 (–) animals] the SMs still divide with this model. The mutant CeTwist E29K protein would have to inhibit the SM divisions actively, perhaps through some neomorphic function or misexpression of the mutant protein. This model does not require a partner or other factor that would interact directly with CeTwist. Because we have been able to create multiple semidominant alleles, we do not favor this neomorphic function model.

It has recently been shown in Drosophila that Twist activity is influenced by its dimerization partner at different stages of development. Castanon and colleagues (Castanon et al., 2001) have demonstrated, using covalently linked dimers, that Twist homodimers promote specification of the mesoderm and allocation of precursors to somatic muscle fate early in development. Additionally, they have shown that linked Twist-Daughterless heterodimers are involved in repressing somatic myogenesis later in development. In C. elegans, overexpression of CeTwist by itself causes a less robust activation of target genes than overexpressing both CeTwist and CeE/DA (Harfe et al., 1998b), suggesting that CeTwist homodimers may not be the primary mode for transcriptional activation at least in this developmental context.

We have seen a range of phenotypes in the three human TWIST mutations that were examined in our C. elegans cellular assay. Although no genotype to phenotype correlation has been noted for any human TWIST mutation in the literature, our results suggest that there could be differences between individuals with different basic domain mutations or mutations elsewhere in the protein. The finding that basic domain mutations analogous to mutations found in human Saethre-Chotzen syndrome caused a phenotype in C. elegans is exciting for two reasons. First, although some human TWIST mutations are surely haploinsufficient because of the early aberrant stop codons, at least one of the human basic domain mutations is acting as a strong dominant negative in our C. elegans assay and thus might act similarly in humans. Second, the genetically manipulable organism C. elegans can provide a useful assay for evaluating proteins that might interact or be downstream from human Twist. The Saethre-Chotzen disorder belongs to a class of disorders causing craniosynostosis (premature fusion of cranial sutures). These disorders occur once in every 2500 births (Wilkie, 1997). Out of about 100 related forms of craniosynostosis about 30 are thought to arise from single gene disorders and the specific gene has been identified in less than 10 disorders (Reardon and Winter, 1995; Cohen, 2000). The two known CeTwist targets are an NK class homeodomain protein and an FGF receptor homolog (Harfe et al., 1998b; Corsi et al., 2000). In humans, other proteins known to be mutant in craniosynostotic disorders are Msx2 (an NK class homeodomain protein) and FGFR1, FGFR2 and FGFR3 (for reviews, see Wilkie, 1997; Hehr and Muenke, 1999; Lajeunie et al., 1999). These strikingly similar pathways suggest that any other downstream targets or partners of Twist that are identified in C. elegans will be good candidates for craniosynostotic disorders whose underlying cause has yet to be molecularly identified in humans.

We have characterized a semidominant mutation in the CeTwist gene called hlh-8 (n2170sd). We found that the n2170sd mutation caused a single amino acid substitution (E29K) in the basic DNA-binding domain of CeTwist. Animals with this mutation resembled a null hlh-8 mutation, except for the improper division of blast cells that will become sex muscles in wild-type animals. This blast cell phenotype can be mimicked in otherwise wild-type animals expressing hlh-8 (n2170sd) or hlh-8 with mutations analogous to those found in the DNA-binding domain of human Twist in individuals with Saethre-Chotzen syndrome who have digit and cranium developmental defects. The mutant CeTwist E29K protein could still bind DNA but could not coactivate transcription of an in vivo reporter with its potential partner CeE/DA (the Daughterless homolog). These results suggest that the defects in the hlh-8 (n2170sd) animals were due to either an inability to interact properly with CeE/DA or an inability to cooperate with other unknown partner proteins required for CeTwist to activate transcription. These mutant animals will be useful in future genetic screens that should identify partners and downstream target genes of CeTwist and will provide candidate genes for other human developmental syndromes.

We are grateful to G. Garriga, M. Stern and H. R. Horvitz for isolating the n2170sd and n1897 alleles. We also thank S. Kostas and A. Fire for providing the body wall muscle counts in Table 2; M. Dong and M. Koelle for the rgs-2::gfp strain; A. Melendez, I. Greenwald, S. Kostas, J. Liu and A. Fire for the arg-1::gfp strain; J. Fares and S. Kostas for the intrinsic cc::gfp strain; R. Littlejohn for technical help; and S. Kostas, J. Liu, A. Fire, A. Golden and J. Campbell for many useful discussions and for manuscript improvements.