A Zn-finger/FH2-domain containing protein, FOZI-1, acts redundantly with CeMyoD to specify striated body wall muscle fates in the Caenorhabditis elegans postembryonic mesoderm

Mesodermal development is a complex multistep process in which the pluripotent precursor cells become increasingly specialized to form specific cell types. This process requires a coordination of multiple transcriptional pathways and cellular signaling mechanisms. The Caenorhabditis elegans postembryonic mesodermal lineage, the M lineage, allows us to study these mechanisms at single cell resolution. The M lineage arises from a single blast cell, the M mesoblast, which is born during embryogenesis and remains dormant until the larva hatches(Sulston and Horvitz, 1977; Sulston et al., 1983). During larval development this pluripotent blast cell undergoes a series of reproducible divisions to give rise to all of the postembryonically derived,non-gonadal mesodermal cells in the animal(Sulston and Horvitz, 1977). In hermaphrodites, the M mesoblast undergoes a series of divisions to give rise to a total of 18 cells. Fourteen of these cells become striated body wall muscles (BWMs) and two become non-muscle coelomocytes (CCs)(Sulston and Horvitz, 1977)(Fig. 1). The remaining two cells become sex myoblasts (SMs), which migrate to the presumptive vulval region and further proliferate to give rise to eight vulval muscles and eight uterine muscles (Sulston and Horvitz,1977) (Fig. 1).

Previous studies have identified a number of transcriptional regulators that are required for proper CC and BWM fate specification in the M lineage. These factors include the Hox factor MAB-5 and its co-factor CEH-20(Liu and Fire, 2000), the HMX homeodomain protein MLS-2 (Jiang et al.,2005), the MyoD homolog HLH-1(Harfe et al., 1998a) and the Twist ortholog HLH-8 (Harfe et al.,1998b). MAB-5 and CEH-20 directly activate expression of HLH-8(Liu and Fire, 2000), while MLS-2 regulates the expression of CeMyoD (HLH-1) in the M lineage(Jiang et al., 2005). Among these transcriptional regulators, HLH-1 is required specifically for BWM and CC fate specification in the M lineage(Harfe et al., 1998a). Lack of HLH-1 in the M lineage does not cause any early defects within the lineage. Instead, it causes stochastic cell fate transformations of several BWMs and the two CCs to SMs (Harfe et al.,1998a). HLH-1 is the only member of the myogenic basic helix-loop-helix (bHLH) family in C. elegans(Krause et al., 1990). Its requirement for BWM fate specification in the M lineage suggests a functional similarity between HLH-1 and the vertebrate myogenic bHLH proteins(Pownall et al., 2002). However, as only a subset of the BWM cells are transformed in mutant animals completely lacking HLH-1 in the M lineage, additional factors must exist for the proper specification of BWM fates.

To identify these additional factors, we screened for mutants with defects in the M lineage (mesodermal lineage specification, or mls,mutants). In this paper, we describe one gene, fozi-1 (formin zinc finger protein-1), that was identified through these screens. fozi-1 encodes a novel nuclear protein with motifs characteristic of transcription factors and actin-binding proteins. Like hlh-1, fozi-1is expressed in the M lineage and functions within the M lineage for proper CC and BWM cell fate specification. We show that FOZI-1 and the Hox factor MAB-5 function redundantly with HLH-1 to specify BWM fate in the M lineage.

Strains were maintained and manipulated using standard conditions(Brenner, 1974). Analyses were performed at 20°C, unless otherwise noted. The following strain was generated to visualize the M lineage in fozi-1(cc609) mutant animals throughout postembryonic development: LW0679 [fozi-1(cc609) ccIs4438(intrinsic CC::gfp) III; ayIs2(egl-15::gfp) IV; ayIs6(hlh-8::gfp)X]. The strain LW0081 [ccIs4438(intrinsic CC::gfp) III; ayIs2(egl-15::gfp) IV; ayIs6(hlh-8::gfp)X] was used in parallel as a wild-type control. The strain LW0064[cup-5(ar465) fozi-1(cc609) III; jjIs64(arg-1::gfp) V; arIs39(secreted CC::gfp) X] was used to generate transgenic lines for rescue. Intrinsic CC::gfp is a twist-derived coelomocyte marker(Harfe et al., 1998b). Secreted CC::gfp is another coelomocyte marker using a myo-3::secreted GFP construct(Harfe et al., 1998a). Additional M lineage-specific reporters were as described in Kostas and Fire(Kostas and Fire, 2002). The M lineage was followed in live animals under a fluorescence stereomicroscope and confirmed using a compound microscope. Other strains used in this work are:

LG II: hlh-1(cc561ts), hlh-1(cc450)/mIn1[dpy-10(e128) mIs14] II(Harfe et al., 1998a)

LG III: mab-5(e1239) (Kenyon,1986), fozi-1(tm0563) (gift from Shohei Mitani, Tokyo Women's Medical University School of Medicine, Japan), ceh-20(n2513)(Liu and Fire, 2000), WM31: dpy-17(e164) vab-7(e1562)

LG V: him-5(e1467) (Hodgkin et al., 1979)

LG X: mls-2(cc615) (Jiang et al., 2005), hlh-8(nr2061)(Corsi et al., 2000) The CB4856 strain (Hodgkin and Doniach,1997) was used for snip-SNP mapping of fozi-1.

In screens for mutants with M lineage defects(Jiang et al., 2005; Foehr et al., 2006), four recessive mutations (cc607, cc608, cc609 and cc610) were isolated that failed to complement one another. These alleles define the fozi-1 locus. Mapping and complementation analysis of fozi-1was carried out using standard methods by monitoring the number of CCs using the CC::gfp markers described above. Three-factor mapping using dpy-17 and vab-7 and snip-SNP mapping(Wicks et al., 2001) placed fozi-1 between cosmids K04H4 and F54G8 on chromosome 3. The molecular lesions of fozi-1 mutant alleles were identified through sequencing PCR products of genomic fragments spanning the entire coding region of K01B6.1.

A full-length fozi-1 cDNA was generated by piecing together two cDNA clones, yk288g3 and yk779b02 (gifts from Yuji Kohara,National Institute of Genetics, Japan). A 5 kb fragment spanning the entire coding region and 3′ UTR of fozi-1 was PCR amplified from N2 genomic DNA using the iProof™ High-Fidelity DNA Polymerase (Bio-Rad). The cDNA and PCR product were used to generate the following constructs:

Forced expression constructs:

pNMA37, hlh-8p::fozi-1::fozi-1 3′UTR

pNMA36, hsp-16p::fozi-1::fozi-1 3′UTR

Recombinant fusion constructs for protein expression in bacteria:

pNMA04, GST-fozi-1 cDNA (amino acids 1-171)

pNMA27, GST-fozi-1 cDNA (full length)

pNMA28, GST-fozi-1 cDNA (amino acids 366-732)

pGEX-6P-3-cyk-1 cDNA (amino acids 681-1435 of cyk-1)

All constructs were confirmed by sequencing. Details on all constructs are available upon request. Transgenic lines were generated using the plasmid pRF4(Mello et al., 1991) as a marker.

Plasmids yk288g3 and yk779b02 were used as templates for synthesizing dsRNA against fozi-1 as described by Fire and colleagues(Fire et al., 1998). Plasmid yk116b7 was used for synthesizing dsRNA against M01A8.2. All yk clones are from Yuji Kohara (National Institute of Genetics, Japan). dsRNAs for hlh-1, mab-5 and ceh-20 were synthesized using plasmids pVZ1200 (gift from Mike Krause, NIDDK, NIH, USA), pJKL718.2 and pJKL422.1(Liu and Fire, 2000),respectively. fozi-1, M01A8.2 or ceh-20 dsRNA was injected into gravid adults of our wild-type reference strain LW0081. Progeny from the injected animals were scored for M lineage phenotypes and, in the case of ceh-20, for FOZI-1 expression via immunostaining.

For hlh-1(RNAi) and mab-5(RNAi), synchronized L1 animals expressing various M lineage GFP markers were soaked in hlh-1 dsRNA, mab-5 dsRNA, or both, for 24-48 hours following the protocol of Maeda and colleagues (Maeda et al.,2001). Animals were allowed to recover at 20°C and scored for M lineage phenotypes. Water was used as a soaking control.

The following strains were generated and used in the heat-shock experiments:

ccIs4251(myo-3::gfp); jjEx[hsp::hlh-1(pPD50.63) + rol-6(d)]

ccIs4251(myo-3::gfp); jjEx[hsp::fozi-1(pNMA36) + rol-6(d)]

ccIs4438(intrinsic CC::gfp); ayIs2(egl-15::gfp); ayIs6(hlh-8::gfp);jjEx[hsp::hlh-1 + rol-6(d)]

ccIs4438(intrinsic CC::gfp); ayIs2(egl-15::gfp); ayIs6(hlh-8::gfp);jjEx[hsp::fozi-1 + rol-6(d)]

ccIs4438(intrinsic CC::gfp); ayIs2(egl-15::gfp); ayIs6(hlh-8::gfp);jjEx[hsp::hlh-1 + hsp::fozi-1 + rol-6(d)]

Two different heat-shock protocols were used to test the myogenic potential of FOZI-1. In each case, global ectopic expression of FOZI-1 was confirmed by anti-FOZI-1 antibody staining 4 hours after heat shock. The first protocol was the same as described in Fukushige and Krause(Fukushige and Krause, 2005). One- to two-cell embryos were harvested and incubated at 20°C for 30 minutes. Embryos were then heat shocked for 30 minutes at 37°C and allowed to recover at 20°C. Embryos were observed periodically over a 20 hour window for ectopic myo-3::gfp expression and an arrested development phenotype. For the second protocol, transgenic animals were given multiple heat-shock pulses (five to seven pulses) starting from mid-embryo to early L1 stage through to adulthood. Heat-shock pulses were performed for 30 minutes at 37°C followed by 3 to 4 hours at 20°C before subsequent heat-shock treatments. Heat-shocked animals were scored for their M lineage phenotypes. In both experiments, heat-shocked non-transgenic animals were used as negative controls.

Plasmids pNMA27, pNMA28 and pGEX-6P-3-cyk-1^FH1FH2COOHwere used to generate GST-fusion proteins in BL21(DE3) cells for FOZI-1,FOZI-1 FH2 domain and CYK-1 FH1FH2 domains plus COOH-terminal residues,respectively. All GST-fusion proteins were bound to glutathione sepharose 4B beads (Amersham Biosciences) and cleaved from GST by GST-3CPro precision protease (Amersham Biosciences) and soluble FOZI-1, FOZI-1^FH2, or CYK-1^FH1FH2COOH were extracted. GST-Bni1p^FH1FH2COOH was purified as previously described (Pruyne et al., 2002). Samples were analyzed by SDS-PAGE.

The effects of FOZI-1 on rates of actin polymerization and barbed end capping of actin filaments were examined using pyrene-actin assays(Pollard, 1983) with the above-purified recombinant proteins. Recombinant FOZI-1 proteins were incubated with F-actin in co-sedimentation assays to test whether FOZI-1 binds filamentous actin (Bretscher,1981). Details on all actin-based assays are available upon request.

The N-terminus of FOZI-1 (amino acids 1-171) was cloned into pGEX-4T-1(Smith and Johnson, 1988). The resulting plasmid pNMA04 was transformed into BL21(DE3)pLysS cells. Fusion proteins were purified from glutathione sepharose 4B beads (Amersham Biosciences) and further purified by SDS-PAGE. Gel slices containing purified FOZI-1 protein were used to immunize guinea pigs (Cocalico Biologicals, PA). Resulting antiserum was tested by western blot analysis using bacterially expressed GST-FOZI-1 fusion proteins. Antibodies were affinity purified against GST-FOZI-1 bound to a nitrocellulose membrane(Olmsted, 1981; Smith and Fisher, 1984).

Animal fixation and immunostaining were performed following the protocol of Hurd and Kemphues (Hurd and Kemphues,2003). For immunostaining with HLH-1 antibodies, animals were fixed and stained following the protocol of Harfe and colleagues(Harfe et al., 1998a). The following antibodies were used: affinity purified anti-FOZI-1 (CUMC-GP19;1:50), goat anti-GFP (Rockland Immunochemicals; 1:5000), rat anti-MLS-2(Jiang et al., 2005) (1:600)and rabbit anti-HLH-1 (Harfe et al.,1998a) (1:400). All secondary antibodies were from Jackson ImmunoResearch Laboratories and used in a dilution of 1:100 to 1:200. Differential interference contrast and epifluorescence microscopy were performed using a Leica DMRA2 compound microscope. Images were captured with a Hamamatsu Orca-ER Camera using the Openlab software (Improvision). Subsequent image analysis was performed using Adobe Photoshop 7.0 and Adobe Illustrator CS.

In screens to identify mutants affecting M lineage development (see Materials and methods), we isolated four recessive mutations that failed to complement one another: cc607, cc608, cc609 and cc610. These mutations defined the fozi-1 locus (see below). Animals homozygous for these mutations lacked M-derived CCs, with 100% penetrance in cc607,cc608 and cc609 and 81% penetrance in cc610(Table 1, Fig. 2).

To determine the basis for the missing CCs in these mutants, we followed the M lineage in cc609 mutant animals using a combination of cell-type-specific markers. Using the hlh-8::gfp reporter as a marker for undifferentiated cells of the M lineage(Harfe et al., 1998b), we found that the M mesoblast in cc609 mutant animals gave rise to a total of 18 descendants (Fig. 2J-L, data not shown), like wild type(Fig. 2I). However, the fates of these 18 cells in cc609 mutants were not specified correctly. Of the 18 cells that arose from M in wild-type animals, 14 became BWMs, two became CCs and two became SMs (Fig. 1). In cc609 mutants, however, both M-derived CCs(visualized using the intrinsic cc::gfp marker)(Harfe et al., 1998b) were missing, fewer than 14 M-derived BWMs (visualized using myo-3::gfp)(Fire et al., 1998) were present (Table 1, Fig. 2J-L), and more than two of the 18 M-derived cells displayed SM-like characteristics(hlh-8::gfp expression, SM shape, enlarged size and potential to migrate to the presumptive vulval region)(Table 1, Fig. 2A-D). Two of the SM-like cells in cc609 mutants always migrated to the ventral side near the presumptive vulval region and generated the correct number of sex muscles that attached properly, as visualized by egl-15::gfp, arg-1::gfp,myo-3::gfp and polarized light microscopy. The remaining SM-like cells divided one to three times and gave rise to extra sex muscles either around the vulva or in the posterior region of the animal(Table 1, Fig. 2G-H, data not shown). Thus cc609 mutants exhibit cell fate transformations from M-derived CCs and BWMs to SMs (Fig. 2J-L). Similar M lineage defects were also observed in cc607,cc608 and cc610 mutant animals(Table 1).

We mapped cc609 to an interval between cosmids K04H4 and F54G8 in the middle of chromosome 3. Two overlapping cosmids in this region, M01A8 and K01B6, rescued the M lineage defects of cc609 mutants. We performed RNAi for each of the two genes located at the junction between M01A8 and K01B6, M01A8.2 and K01B6.1 (see Materials and methods). RNAi of M01A8.2 gave no M lineage defects. However, RNAi of K01B6.1 in wild-type animals resulted in similar M lineage defects to those seen in cc609 mutant animals. Sequencing the coding regions of K01B6.1 in cc607, cc608, cc609 and cc610 mutant animals showed that they all contained molecular lesions in the K01B6.1 coding region (Fig. 3A). cc607 and cc608 contained the same nonsense mutation at residue 112 (CAG to TAG, Gln to amber stop). The cc609mutant contained a deletion that spanned residues 160 to 260 of K01B6.1 and caused a frameshift in the remaining coding sequence(Fig. 3A). In cc610mutants, there was a missense mutation at residue 207 (CGA to TGA, Pro to Leu)(Fig. 3A). During the course of our studies, the National Bioresource Project for the Experimental Animal Nematode C. elegans generated another deletion allele of K01B6.1, tm0563, which deletes a larger region than cc609 and also results in a frameshift (Fig. 3A). tm0563 mutant animals displayed similar M lineage phenotypes to cc609 animals (Table 1). As cc607, cc608, cc609 and tm0563 all exhibited 100% penetrance in their M lineage defects(Table 1), and they contained either early termination codons or deletions in K01B6.1(Fig. 3A), all four alleles are probably null alleles. Additional evidence supporting this is provided by the immunostaining results discussed below.

The predicted protein encoded by K01B6.1 is FOZI-1, a 732-amino acid protein with three distinct motifs (Fig. 3B): a glutamine-rich region (residues 88-133), two C₂H₂ zinc finger motifs (residues 177-201) and an FH2 formin homology domain (residues 369-732). The glutamine-rich region and the zinc finger motifs are characteristics of transcription factors(Matsuzaki et al., 1995; Iuchi, 2001; Stepchenko and Nirenberg,2004). The FH2 domain is conserved among a large family of proteins called formins, which are present in a range of organisms, from yeast to humans (Higgs and Peterson,2005). Formins have been shown to use their FH2 domains to form dimers that bind actin and play crucial roles in multiple processes of actin dynamics, such as nucleation, capping, severing and bundling of actin filaments (Pruyne et al.,2002; Sagot et al.,2002; Kovar et al.,2003; Zigmond et al.,2003; Harris et al.,2004; Xu et al.,2004; Michelot et al.,2005; Moseley and Goode,2005). The FH2 domain of FOZI-1 is unique in that it contains many of the conserved residues involved in dimerization, but lacks the key residues crucial for actin binding (see Fig. S1A in the supplementary material). We tested whether the FH2 domain of FOZI-1 or the full-length FOZI-1 protein could bind filamentous actin, induce actin polymerization or cap the barbed-ends of actin filaments in vitro. We found that FOZI-1 was unable to polymerize actin filaments, while the C. elegans CYK-1 FH1FH2 protein domains were fully active in this process (see Fig. S1B in the supplementary material). Furthermore, FOZI-1 was unable to bind actin or show capping activity in vitro (data not shown). As residues crucial for dimerization in the FH2 domain of FOZI-1 are conserved, the FOZI-1 FH2 domain probably functions as a dimerization domain, consistent with in vitro data obtained by Johnston and colleagues (Johnston et al.,2006).

As described above, fozi-1 mutant animals exhibited cell fate transformations from M-derived CCs and BWMs to SMs. To begin to understand how FOZI-1 functions in regulating fate specification in the M lineage, we examined the expression pattern of FOZI-1 during development. We generated antibodies to FOZI-1 and performed immunostaining to detect the endogenous FOZI-1 protein (see Materials and methods). The antibodies specifically recognized a protein of the predicted size for FOZI-1 (∼80 kD) on western blots using worm extracts (data not shown), and failed to detect any signal in cc609 mutant embryos and larvae(Fig. 4C,D,T).

Using these antibodies, we found that FOZI-1 protein was localized in nuclei of a distinct set of cells in wild-type animals. Expression of FOZI-1 was first detectable in a small subset of nuclei (probably neuroblasts) in embryos at the 2-fold stage (Fig. 4A-B). During larval development, FOZI-1 expression was restricted to a subset of unidentified neurons in the head (seven to 12 cells) and along the ventral nerve cord (five to seven cells)(Fig. 4E-F). FOZI-1 was also present transiently during the L1 larval stage in cells derived from the M lineage (see below).

To examine the M lineage expression pattern of FOZI-1, we performed double-labeling experiments using anti-FOZI-1 antibodies and the M lineage-specific hlh-8::gfp marker(Harfe et al., 1998b). FOZI-1 expression in the M lineage was first detectable in the nuclei of M.d and M.v at the 2-M stage (in 43% of the animals stained, n=21, Fig. 4G-I). This expression persisted through the next three rounds of cell divisions, such that all animals (n>50) at the 4-M through to the 16-M stage showed FOZI-1 expression in the M lineage (Fig. 4J-L). At the 16-M stage, two cells on the ventral side (M.vlpa and M.vrpa) divide one more time to produce two SMs and two BWM precursors and reach the 18-M stage (Fig. 1)(Sulston and Horvitz, 1977). At the 18-M stage, FOZI-1 was still detectable in the undifferentiated BWMs and CCs, but not in the SMs (n=22; Fig. 4M-O). This expression of FOZI-1 at the 18-M stage appeared transiently and quickly became undetectable in the differentiated BWMs and CCs. FOZI-1 expression was not detected in the SM lineage (n>50; Fig. 4P-R, data not shown). The expression pattern of FOZI-1 in the M lineage is summarized in Fig. 4S. Thus, expression of FOZI-1 is tightly regulated within the M lineage at the time when BWM and CC cell fate specification occurs.

To determine whether FOZI-1 expression and subcellular localization were lost in fozi-1 mutants, we performed FOZI-1 staining in all of the fozi-1 mutants at all developmental stages. As discussed above, cc609 animals had no detectable levels of FOZI-1 at any stage during development (Fig. 4C,D,T). cc607 and tm0563 also showed no FOZI-1 immunostaining at any stage (data not shown; we did not examine cc608 animals, because cc607 and cc608 contain the same molecular lesion). These results are consistent with our hypothesis that cc607, cc609 and tm0563 are null alleles. Animals mutant for cc610, however,showed nuclear anti-FOZI-1 staining in a pattern identical to wild-type animals in the head neurons, ventral nerve cord and the M lineage(Fig. 4U; n>50). As cc610 mutants exhibited M lineage defects(Table 1) and the missense mutation in cc610 is located six residues downstream of the second histidine residue of the second zinc finger(Fig. 3A), we conclude that the residues surrounding the C₂H₂ zinc fingers are required for proper FOZI-1 function in the M lineage.

The expression pattern of fozi-1 and the M lineage defects of fozi-1 mutants suggest that FOZI-1 functions within the M lineage to specify the fates of the CCs and BWMs. To directly test this hypothesis, we expressed fozi-1 specifically in the M lineage in cc609mutants using the hlh-8 promoter(Harfe et al., 1998b). Transgenic animals carrying an hlh-8p::fozi-1 transgene(Fig. 3C) in the cc609mutant background showed M lineage-specific expression of fozi-1(data not shown). This M lineage-specific expression was sufficient to restore M-derived CCs in cc609 mutant animals (23.2%, n=56). The low level of rescue may be attributed to the difference between endogenous hlh-8 and fozi-1 expression patterns: fozi-1persists slightly longer than hlh-8 in the BWM and CC precursors(Harfe et al., 1998b)(Fig. 4M-O). Thus, FOZI-1 appears to function cell autonomously within the M lineage for proper mesodermal cell fate specification.

Previous studies (Harfe et al.,1998a) have shown that the C. elegans MyoD ortholog,Ce-MyoD (HLH-1), is expressed in the M lineage in a pattern like that of FOZI-1 (except that HLH-1 persists in all differentiated M-derived BWMs,whereas FOZI-1 does not) and that hlh-1 mutants showed similar CC and BWM to SM fate transformation as fozi-1 mutants. To examine the functional relationship between FOZI-1 and HLH-1, we examined the expression and nuclear localization of HLH-1 in fozi-1(cc609) mutants and of FOZI-1 in the temperature-sensitive hlh-1(cc561ts) mutants at the non-permissive temperature (25°C). As shown in Fig. 5A-D, expression and nuclear localization of FOZI-1 in hlh-1(cc561ts) and HLH-1 in fozi-1(cc609) were normal. Thus, HLH-1 and FOZI-1 do not regulate the expression or subcellular distribution of each other.

As both hlh-1 and fozi-1 mutant animals exhibit loss of some, but not all, BWMs derived from the M lineage(Harfe et al., 1998a) (this work), we investigated the possibility that hlh-1 and fozi-1act redundantly. Because hlh-1 null mutants are embryonic/L1 lethal due to the essential functions of HLH-1 in proper differentiation of the embryonically derived BWMs (Chen et al.,1992; Chen et al.,1994), we used two approaches to assess the possible redundancy of HLH-1 and FOZI-1 in the M lineage. We first soaked fozi-1(cc609) L1 larvae with hlh-1 dsRNA to deplete levels of HLH-1 in fozi-1(cc609) worms [referred to as hlh-1(RNAi);fozi-1(cc609); see Materials and methods]. We also generated hlh-1(cc561ts); fozi-1(cc609) double mutants and examined them at the restrictive temperature (25°C) for cc561ts. Both approaches gave almost identical results. The hlh-1(cc561ts); fozi-1(cc609) or hlh-1(RNAi); fozi-1(cc609) animals had a significant increase in the number of SM-like cells (Fig. 6A) and a concomitant decrease in the number of M-derived BWMs(Fig. 6B) when compared with fozi-1(cc609), hlh-1(cc561ts) or hlh-1(RNAi) single mutants. The increased number of SM-like cells in hlh-1(cc609ts);fozi-1(cc609) or hlh-1(RNAi); fozi-1(cc609) animals was correlated with the significantly higher number of type I vulval-muscle-like cells (vm1-like cells) that express egl-15::gfp (data not shown). To distinguish whether these changes in the numbers of SMs and BWMs arose from cell fate transformations or cell proliferation defects, we followed the M lineage closely in three hlh-1(RNAi); fozi-1(cc609) mutant animals using a combination of hlh-8::gfp and DIC optics. We did not observe any proliferation defects in the first four rounds of cell divisions of the M mesoblast in these animals. Instead, all three animals exhibited a fate transformation from M-derived BWMs and CCs to SMs. There were eight SM-like cells and one M-derived BWM on the right side of animal #1, and nine SM-like cells and 0 M-derived BWMs on the right side of animal #2. Animal #3 was monitored on both the right and the left sides. There were a total of 12 SM-like cells and four M-derived BWMs in this animal. All three animals had the correct number of embryonically derived BWMs (data not shown). These observations indicate that FOZI-1 and HLH-1 function redundantly to specify striated BWM fate in the postembryonic mesoderm.

In addition to HLH-1 and FOZI-1, mutations in the Hox gene mab-5also result in the fate transformation of both M-derived CCs and some, but not all, M-derived BWMs to SMs (Harfe et al.,1998b) (Fig. 6A,B). We therefore examined the relationship among MAB-5, FOZI-1 and HLH-1 in M-derived BWM fate specification. Because mab-5 and fozi-1map very close to each other on chromosome 3, we used RNAi to knock down the expression of mab-5 in hlh-1(cc561ts), fozi-1(cc609) or hlh-1(cc561ts); fozi-1(cc609) mutants (see Materials and methods). We first established that mab-5(RNAi) caused similar M lineage phenotypes to the null mab-5(e1239) mutant(Fig. 6, data not shown). We then followed the M lineage and counted the number of M-derived BWMs and SMs in various mutants soaked with mab-5 dsRNA. All mutant animals exhibited normal proliferation in the early M lineage (data not shown). However, their terminal M lineage phenotypes differed from each other. As shown in Fig. 6A,B, fozi-1(cc609) mab-5(RNAi) animals behaved just like fozi-1(cc609) or mab-5(RNAi) single mutant animals with only a few M-derived BWMs transformed to SMs, whereas hlh-1(cc561ts);mab-5(RNAi) animals behaved just like hlh-1(cc561ts);fozi-1(cc609) animals in that most, if not all, M-derived BWMs were transformed to SMs (Fig. 6A,B). hlh-1(cc561ts); fozi-1(cc609) mab-5(RNAi) animals were indistinguishable from hlh-1(cc561ts); fozi-1(cc609) animals or hlh-1(cc561ts); mab-5(RNAi) animals in their BWM or SM phenotype(Fig. 6A,B). Collectively,these results suggest that both MAB-5 and FOZI-1 function redundantly with HLH-1 to properly specify striated BWM fates in the M lineage and that MAB-5 and FOZI-1 act in the same process.

mab-5 is expressed in the M mesoblast, and its expression persists throughout the M lineage (S. J. Salser, PhD thesis, University of California,1995). As mab-5 expression precedes fozi-1 expression in the M lineage, we asked if mab-5 is required for fozi-1expression. We stained mab-5(e1239) null mutant animals with anti-FOZI-1 antibodies and found a wild-type FOZI-1 expression and subcellular localization pattern (Fig. 5E),indicating that mab-5 is not required for the proper expression and localization of FOZI-1. It is unlikely that fozi-1 is required for mab-5 expression, as fozi-1(cc609) mutants do not display cleavage orientation defects or loss of hlh-8 expression in the M lineage (Fig. 2, data not shown), phenotypes exhibited by mab-5(e1239) single mutants(Harfe et al., 1998b). Thus,FOZI-1 and MAB-5 do not regulate the expression of each other in the M lineage.

In addition to MAB-5, three other transcription factors crucial for proper M lineage development, CEH-20, MLS-2 and HLH-8, are expressed in the M lineage beginning at the 1-M (or the M mesoblast) stage(Harfe et al., 1998b; Yang et al., 2005; Jiang et al., 2005). We asked whether these factors are required for fozi-1 expression in the M lineage. We stained animals mutant for the presumed null alleles mls-2(cc615) and hlh-8(nr2061) using anti-FOZI-1 antibodies and found that fozi-1 expression level and pattern were not altered(Fig. 5F, data not shown). Similarly, expression of mls-2 and hlh-8 in the M lineage does not depend on FOZI-1 (data not shown).

By contrast, when we stained animals from the strong loss-of-function allele ceh-20(n2513), we found that close to 90% of the animals(n=205) lacked fozi-1 expression in the M lineage(Fig. 5G). M lineage expression of fozi-1 was also lost in progeny of animals injected with ceh-20 dsRNA (data not shown). In both ceh-20(n2513) and ceh-20(RNAi) animals, fozi-1 expression was still detectable in the head neurons and the ventral nerve cord(Fig. 5G, data not shown),indicating that CEH-20 is specifically required for M lineage expression of fozi-1.

We have identified a gene, fozi-1, encoding a unique zinc finger-FH2 domain-containing protein that is required for proper fate specification in the C. elegans postembryonic mesoderm. We have shown that the FH2 domain of the FOZI-1 protein does not bind actin (see Fig. S1 in the supplementary material) and that FOZI-1 is localized to the nucleus(Fig. 4). As FOZI-1 contains a Q-rich region and two C₂H₂ zinc finger motifs that are characteristic of transcription factors(Matsuzaki et al., 1995; Iuchi, 2001; Stepchenko and Nirenberg,2004), we propose that FOZI-1 acts as a transcription factor required for proper fate specification in the M lineage.

In animals lacking FOZI-1, some M lineage-derived BWMs and non-muscle CCs are transformed to SM-like cells (Fig. 2). As M lineage-specific expression of fozi-1 rescued these M lineage defects of fozi-1(cc609) mutants, we conclude that FOZI-1 functions cell autonomously within the M lineage to specify both BWM and CC fates. FOZI-1 is present in the early M lineage as well as in all M-derived BWM and CC precursors, but not SMs(Fig. 5). It is not clear at present whether FOZI-1 functions early in the M lineage in the multipotent precursors or in the early cell cycle of the terminal BWM, CC or SM cells.

The stochastic loss of a fraction of M-derived BWMs in fozi-1(cc609) mutants is very similar to phenotypes exhibited by animals lacking HLH-1 in the M lineage(Harfe et al., 1998a). We have shown that hlh-1(cc561ts); fozi-1(cc609) and hlh-1(RNAi);fozi-1(cc609) animals exhibit normal proliferation in the early M lineage, yet most, if not all, of their M-derived BWMs are transformed to SM-like cells. These results demonstrate that FOZI-1 and HLH-1 function redundantly to specify M-derived BWM fates. Although the presence of a few M-derived BWMs in some hlh-1(cc561ts); fozi-1(cc609) or hlh-1(RNAi); fozi-1(cc609) animals could be due to the partial loss-of-function nature of the hlh-1(cc561ts) mutation or the inefficiency of hlh-1(RNAi), we cannot rule out the possibility that other factor(s) may also contribute to the specification of BWM fate in the M lineage.

The synergistic loss of M-derived BWMs in hlh-1(RNAi);fozi-1(cc609) animals suggests functional overlap between FOZI-1 and HLH-1 in specifying BWM fate. However, each factor alone is clearly required for proper BWM fate specification, as fozi-1 or hlh-1 single mutants lack a small fraction of M-derived BWMs(Harfe et al., 1998a) (this work). Both fozi-1 and hlh-1 are expressed in the early M lineage as well as in the BWM and CC precursors at the time when these cell fates are specified (Harfe et al.,1998a) (this work). Although FOZI-1 is no longer detectable after these cells differentiate, HLH-1 continues to be expressed in differentiated BWM cells (Harfe et al.,1998a) (this work). We have shown that fozi-1 and hlh-1 do not regulate each other's expression and subcellular localization in the M lineage (Fig. 5). It is possible that in the M lineage, FOZI-1 and HLH-1 together contribute to process(es) required to specify BWM and CC fates that compete with an opposing process leading to SM fate specification. A slight shift in the balance between these two processes, as in fozi-1 or hlh-1 single mutants, can lead to transformation of the CC fate and loss of a fraction of M-derived BWMs. Loss of both FOZI-1 and HLH-1, however,shifts the balance strongly in favor of SMs. This interpretation requires that the M-derived CC fate is more sensitive than the BWM fate to the imbalance of these two opposing processes.

HLH-1 is a myogenic factor and has been shown to be capable of inducing widespread myogenesis when ectopically expressed during early embryogenesis(Fukushige and Krause, 2005). FOZI-1, however, does not appear to be a myogenic factor. Forced global expression of FOZI-1 did not lead to precocious myogenesis during embryogenesis, nor did it cause an SM to BWM fate transformation within the M lineage (data not shown, see Materials and methods for heat-shock conditions). The failure of FOZI-1 to induce myogenesis is not unexpected, as FOZI-1 is also expressed in a subset of head neurons and along the ventral nerve cord(Fig. 4E,F), and FOZI-1 has been shown to function in the specification of left-right asymmetric fates of the ASE neurons (Johnston et al.,2006). Thus FOZI-1 is required not only for the specification of striated BWM fates, but also for the specification of non-muscle cells such as CCs and ASE neurons. It is possible that different co-factors are required for FOZI-1 to exert its different functions in different cell types.

We believe that the Hox factor MAB-5 is one of the factors that function together with FOZI-1 in specifying the striated BWM fate in the M lineage. MAB-5 is present throughout the M lineage. mab-5(0) mutation results in multiple defects in the M lineage, including abnormal cleavage orientations, loss of expression of the CeTwist homolog HLH-8 in the early M lineage, and BWM and CC to SM fate transformations (S. J. Salser, PhD thesis,University of California, 1995) (Harfe et al., 1998b). We have shown that fozi-1(cc609)mab-5(RNAi) animals behaved similarly to fozi-1(cc609) or mab-5(0) animals regarding BWM fate specification, whereas hlh-1(cc561ts); mab-5(RNAi), like hlh-1(cc561ts);fozi-1(cc609) animals, exhibited synergistic loss of M-derived BWMs(Fig. 6A,B). As MAB-5 and FOZI-1 do not appear to regulate the expression of each other(Fig. 5E), our results suggest that MAB-5 and FOZI-1 function either together or in parallel in specifying M-derived BWM fate and that they do this by affecting a common downstream target that functions redundantly with HLH-1.

Previous studies have shown that the homeodomain protein MLS-2 is required to activate the expression of HLH-1 in the M lineage and that mls-2(0) mutants have defects in BWM and CC fate specification(Jiang et al., 2005). We found that MLS-2 is not required for M lineage expression of FOZI-1, indicating that HLH-1 and FOZI-1 are regulated by different sets of transcription factors in the M lineage. We found that the Pbx/Exd homolog CEH-20 is required for the M lineage expression of FOZI-1 and that this regulation is independent of the Hox factor MAB-5 (Fig. 5). Interestingly, CEH-20 is also required for the expression of MLS-2 in the M lineage (Y. Jiang and J.L., unpublished). These observations place CEH-20 upstream of both HLH-1 and FOZI-1. As ceh-20 mutant animals lack all M-derived BWMs (Liu and Fire,2000), it is likely that any additional factors required for specifying M-derived BWM fate would also be downstream targets of CEH-20.

Based on all our results, we propose a model for how M-derived BWMs are specified, as depicted in Fig. 7. In this model, at least two redundant mechanisms are involved in specifying myogenic fates in the M lineage. First, MLS-2 regulates the expression of HLH-1, a myogenic factor. Second, FOZI-1 and MAB-5 regulate an unknown myogenic factor X that functions redundantly with HLH-1. At present,it is not clear whether MAB-5 functions together or in parallel with FOZI-1 in regulating factor X. Our data do not exclude the possibility of a third factor(Y) in addition to HLH-1 and factor X to promote myogenic fate specification. The Pbx/Exd homolog CEH-20 appears to act upstream of all of these myogenic pathways in the M lineage.

The C. elegans striated musculature consists of 81 embryonically derived and 14 M lineage-derived BWMs(Sulston and Horvitz, 1977; Sulston et al., 1983). Despite their different lineage history, all 95 BWMs appear morphologically and functionally equivalent. It has been previously shown that although HLH-1 is sufficient to induce striated BWM fate(Fukushige and Krause, 2005), hlh-1(0) mutants still make all 81 embryonically derived BWMs(Chen et al., 1992; Chen et al., 1994). These studies suggest that other factors share redundant functions with HLH-1 in specifying these 81 BWMs. Neither FOZI-1 nor MAB-5 appears to be one of these factors, as FOZI-1 and MAB-5 are not expressed in the 81 embryonically derived BWMs or their precursors, and fozi-1(cc609), mab-5(0), hlh-1(cc561ts);fozi-1(cc609), hlh-1(cc561ts); mab-5(RNAi) and hlh-1(cc561ts);fozi-1(cc609) mab-5(RNAi) double and triple mutant animals do not have any defect in the specification of these 81 BWMs. Thus, although HLH-1 functions in the development of all BWMs, embryonic and postembryonic BWMs require distinct sets of transcription factors that function redundantly with HLH-1. Therefore there is a remarkable level of complexity for the production of a simple striated musculature in C. elegans.

Functional redundancy in specification of striated muscle fates has been observed in both vertebrates and invertebrates. In vertebrates, the redundancy is limited to proteins in the MyoD family. There are multiple members of the MyoD family of myogenic regulatory factors (MRFs) in vertebrates, including MyoD, Myf-5, myogenin and MRF4 (reviewed by Buckingham, 2001; Pownall et al., 2002; Buckingham et al., 2003). Each of these factors is capable of inducing the transcription of muscle-specific genes when overexpressed, but they also share functional redundancy in striated muscle development (reviewed by Pownall et al., 2002). For example, mice lacking either Myf-5 or MyoD are viable and fertile(Braun et al., 1992; Rudnicki et al., 1992; Kaul et al., 2000). However,double mutants lacking both Myf-5 and MyoD do not form any skeletal muscles(Rudnicki et al., 1993).

Different from vertebrates, many invertebrate organisms have only one MyoD homolog, including ascidians (AMD-1)(Araki et al., 1994), sea urchins (sum-1) (Venuti et al.,1991) and Drosophila (nautilus)(Michelson et al., 1990). In Drosophila, nautilus is present in a subset of muscle precursors and differentiated muscle fibers (Michelson et al., 1990; Paterson et al.,1991). nautilus loss-of-function mutants have defects in only a distinct subset of cells normally expressing nautilus, and the defects appear to be a combination of abnormal differentiation of some muscle fibers and inappropriate fate specification(Keller et al., 1998; Balagopalan et al., 2001). These observations indicate that there are factor(s) functioning redundantly with nautilus in muscle fate specification and differentiation. The identity of these factors is currently unknown. It is possible that some of the muscle identity genes such as slouch, apterous, muscle segment homeobox, ladybird and Krüppel(Baylies et al., 1998; Frasch, 1999) could share redundant functions with nautilus in specifying the fate of certain muscle fibers.

Our work shows that the combinatorial functions of HLH-1 with FOZI-1 and MAB-5 are required to specify the M-derived BWM fate. This is similar to the situation in Drosophila, in which unique combinations of different muscle identity genes specify distinct muscle fiber fates, and loss-of-function mutations in these genes cause fate transformation from one type of muscle fiber to another (Bourgouin et al., 1992; Ruiz-Gomez et al., 1997; Jagla et al.,1998; Nose et al.,1998; Knirr et al.,1999). Notably, both MAB-5 and FOZI-1 are expressed in other cell types outside of the M lineage in C. elegans (S. J. Salser, PhD thesis, University of California, 1995; this work). The expression pattern of muscle identity genes in Drosophila is also not restricted only to muscle cells (Bourgouin et al.,1992; Jagla et al.,1998; Nose et al.,1998; Knirr et al.,1999). What is unique in our study is that hlh-1(cc561ts);mab-5(RNAi) and hlh-1(cc561ts); fozi-1(cc609) animals show clear fate transformations from M-derived BWMs to non-muscle SM-like cells, which not only display the morphology of SMs, but also go on to proliferate and produce differentiated non-striated sex muscles. This muscle to non-muscle fate transformation may be due to these cells adopting a developmental ground state represented by the SMs when myogenic factors are missing from the M lineage.

We thank Marisa Foehr for repeating the RNAi experiments to confirm the identity of fozi-1; Mike Krause, Yuji Kohara, Shohei Mitani and the Sanger Center for C. elegans strains, cosmids, cDNA clones and antibodies; Marisa Foehr, Yuan Jiang, Ken Kemphues, Siu Sylvia Lee and Diane Morton for helpful discussions, suggestions and valuable comments on the manuscript. J.L. appreciates the advice and support of Andy Fire, in whose lab the fozi-1 mutants were isolated. Some strains used in this study were obtained from the C. elegans Genetics Center (CGC),which is supported by a grant from the NIH National Center for Research Resources. This work was supported by NIH R01 GM066953 (to J.L.) and GM39066(to D.P. and A.B.). During part of this work, N.M.A. was supported by NIH T32 GM07617 and a GAANN fellowship sponsored by the U.S. Department of Education. D.T. was supported by the Summer Research Fellowship from Weill Cornell Medical College in Qatar.