An antagonistic role for the C. elegans Schnurri homolog SMA-9 in modulating TGFβ signaling during mesodermal patterning

During the development of multicellular organisms, a small number of signaling pathways (TGFβ, Notch, Wnt, Hedgehog, Jak/STAT, receptor tyrosine kinases and nuclear hormone receptors) are used repeatedly in many different developmental contexts to regulate distinct target genes(Barolo and Posakony, 2002). The TGFβ signaling pathway regulates many cellular and developmental processes (Feng and Derynck,2005; Massague et al.,2000; Shi and Massague,2003). The core components of the TGFβ signaling pathway have been well studied; these include type I and type II transmembrane receptors and intracellular Smad proteins that transduce the TGFβ signal from the cell membrane to the nucleus (Feng and Derynck, 2005; Shi and Massague, 2003). Despite a general understanding of the underlying signal transduction pathways, our understanding of the mechanisms involved in regulating the specific signaling output in different cell types is far from complete.

There are two canonical TGFβ-related signaling pathways in C. elegans: the Sma/Mab pathway that regulates body size and male tail patterning; and the dauer pathway that controls formation of dauer larvae(Patterson and Padgett, 2000; Savage-Dunn, 2001). In addition, a TGFβ-related molecule, UNC-129, is required to guide pioneer motor axons along the dorsoventral axis(Colavita et al., 1998). Studies in both Drosophila and vertebrates have shown that the TGFβ signaling pathway plays crucial roles in regulating dorsoventral patterning (De Robertis and Kuroda,2004). Despite the role of UNC-129 in guiding the dorsal migration of motor axons, a role of the canonical TGFβ signaling pathways in dorsoventral patterning is unclear in C. elegans. DBL-1, the ligand of the Sma/Mab pathway, has been proposed to specify dorsal cell fates in male sensory ray patterning (Suzuki et al.,1999). However, the fate transformations observed in dbl-1 mutants (rays 5, 7 and 9 to rays 4, 6 and 8, respectively) can equally be interpreted as a posterior to anterior transformation.

The C. elegans post-embryonic mesodermal lineage, the M lineage,provides an excellent system with which to study dorsoventral patterning and cell fate specification at single cell resolution. The M lineage arises from a single precursor cell, the M mesoblast, which is born during embryogenesis(Sulston and Horvitz, 1977). During larval development, the M mesoblast first divides along the dorsoventral axis to generate two daughter cells: M.d and M.v. These two cells subsequently give rise to distinct dorsal and ventral cell types. The dorsal cell, M.d, gives rise to six body wall muscles (BWMs) and two non-muscle coelomocytes (CCs), whereas the ventral cell, M.v, gives rise to eight BWMs and two sex myoblasts (SMs). The SMs subsequently migrate from the ventral posterior to the presumptive vulval region where they each undergo three rounds of cell division and differentiate into vulval and uterine muscles(Fig. 1E).

In a screen for mutants with M lineage patterning defects, we isolated two loss-of-function mutants of the sma-9 locus. These mutants exhibit a loss of dorsally M-derived CCs and a duplication of ventrally M-derived SMs(Fig. 1B). We show that SMA-9 functions within the M lineage to regulate this asymmetry, yet SMA-9 itself is not asymmetrically localized along the dorsoventral axis within the M lineage. sma-9 encodes a homolog of the Drosophila protein Schnurri(Shn), and has been previously shown to function as a downstream component of the Sma/Mab TGFβ signaling pathway in regulating body size and male tail patterning (Liang et al.,2003). We have found that none of the TGFβ pathway mutants exhibit any M lineage defects; however, mutations in any of the core components of the Sma/Mab TGFβ signaling pathway suppressed the M lineage defects of sma-9 mutants. These findings uncovered a role of the Sma/Mab TGFβ signaling pathway in patterning the M lineage along the dorsoventral axis. Studies in Drosophila have shown that Shn is a transcriptional repressor that forms a complex with the Smad proteins Mad and Medea to repress transcription of genes such as brinker(Marty et al., 2000; Pyrowolakis et al., 2004). Our data suggest that SMA-9 may act independently of complex formation with the Smad proteins to antagonize their function, adding another potential mode of action for the Shn family of proteins.

Strains were maintained and manipulated under standard conditions as described by Brenner (Brenner,1974). Analyses were performed at 20°C, unless otherwise noted. The following strains were used in this work.

LG I: rnt-1(e1241) (Link et al., 1988), rnt-1(ok351)(Lee et al., 2004).

LG II: sma-6(e1482) (Brenner,1974), sma-6(wk7)(Krishna et al., 1999).

LG III: lin-6(e1466) dpy-5(e61)/hT2[qIs48](Wang and Kimble, 2001), cup-5(ar465) (Fares and Greenwald, 2001), sma-2(e502) unc-32(e189)(Manser and Wood, 1990), sma-3(e491) unc-32(e189), sma-4(e729) and lon-1(e185)(Brenner, 1974), sma-3(wk28) (Savage-Dunn et al.,2000), daf-4(e1364) unc-32(e189) and daf-4(m63)(Riddle, 1977).

LG IV: daf-1(m213) and daf-1(m40)(Riddle, 1977), unc-129(ev554), unc-129(ev557) and unc-129(ev566)(Colavita et al., 1998), daf-14(m77) (Riddle et al.,1981).

LG V: him-5(e1467) (Hodgkin et al., 1979), dbl-1(wk70)(Suzuki et al., 1999), dbl-1(nk3) (Morita et al.,1999).

LG X: dpy-6(e14) unc-9(e101), lon-2(e678)(Brenner, 1974), daf-3(e1376) and daf-3(mgDf90)(Patterson et al., 1997), sma-9 (wk55), sma-9(wk62), sma-9(wk71), sma-9(wk82), sma-9(qc1),sma-9(qc3), sma-9(qc5), sma-9(qc6), sma-9(qc7), sma-9(qc8), sma-9(qc9),sma-9(qc10) and sma-9(qc11)(Liang et al., 2003), sma-9(tm572) (gift from Shohei Mitani, Tokyo Women's Medical University School of Medicine, Japan).

LW0081: ccIs4438(intrinsic CC::gfp) III;ayIs2(egl-15::gfp) IV; ayIs6(hlh-8p::gfp) X(Jiang et al., 2005). Intrinsic CC::gfp is a twist-derived coelomocyte marker(Harfe et al., 1998b).

BW1935: unc-119(ed3) III; ctIs43[pDP#MM016B, pNY+nls, pMY-nls(dbl-1::GFP)] him-5(e1490) V. (Suzuki et al., 1999).

Additional cell-type-specific reporters for the M lineage were as described in Kostas and Fire (Kostas and Fire,2002).

The lon-2(e678) sma-9(cc604) unc-9(e101) triple mutant was generated from a lon-2(e678) egl-15(n484)/sma-9(cc604)unc-9(e101) heterozygote by identifying Lon-non-Egl recombinants. Most of the Lon-non-Egl recombinants segregated lon-2(e678) sma-9(cc604)unc-9(e101) triple mutant animals. We then confirmed the presence of both the lon-2(e678) mutation (deletion) and the sma-9(cc604)mutation in the strain by PCR and sequencing. All other strains generated in this work (see Table 4) were generated through standard genetic crosses using appropriate balancers to follow individual chromosomes.

sma-9 alleles were isolated in an EMS screen for mutants with altered numbers of coelomocytes, as described by Yanowitz and Fire(Yanowitz and Fire, 2005). Two recessive mutations, cc606 and cc604, lacked M-derived coelomocytes. Both mapped to the X chromosome and failed to complement each other. cc604 was further mapped between dpy-6 and unc-9, and failed to complement two sma-9 alleles, sma-9(wk55) and sma-9(wk62)(Liang et al., 2003). The molecular lesions of cc606 and cc604 were identified through direct sequencing of PCR fragments encompassing the entire sma-9genomic region. Detailed information on the primers used is available upon request.

arIs37 [secreted CC::gfp (gfp with a signal peptide driven by the myo-3 promoter taken up by the coelomocytes)] I; cup-5(ar465) III; sma-9(cc604) X animals lacking M-derived coelomocytes were mutagenized by EMS. Individual F1 animals and their clonal F2 progeny were screened for the restoration of M lineage-derived coelomocytes by direct visual examination using a fluorescence stereomicroscope. Approximately 2800 haploid genomes were examined. Seven recessive suppressor mutations(jj1-jj7) were obtained. jj1 and jj3 were mapped via snip-SNP mapping (Wicks et al.,2001). The molecular lesions of jj1 and jj3 were identified via direct sequencing of PCR fragments encompassing the entire sma-6 and sma-3 genomic regions respectively. Detailed information on the primers used is available upon request.

We generated three genomic-cDNA hybrid constructs, one for each C-terminal isoform of sma-9 as described in Liang et al.(Liang et al., 2003). These constructs were referred to as C1 (Class I), C2 (Class II) and C3 (Class III). C1 was generated by ligating the cDNA fragment from yk128a08 (Class I) to a genomic fragment containing sequences from the predicted ATG to exon 20. C2 and C3 were generated by ligating a genomic fragment containing sequences from the predicted ATG to exon 10 to cDNA fragments from yk1103h10 (Class II) and yk328c9 (Class III), respectively. A 3763 bp fragment immediately upstream of the predicted ATG was used as the sma-9 promoter. To generate the GFP tagged constructs, GFP was added to the C-terminal end of the sma-9-coding region. All genomic fragments were generated through long-range PCR (Expand Long Template PCR system, Roche) using cosmid T01A5 as template. The following constructs were made.

pMX15: 3.7kb sma-9p::sma-9C1::unc-54 3′UTR

pMX16: 3.7kb sma-9p::sma-9C2::unc-54 3′UTR

pMX9: 3.7kb sma-9p::sma-9C1::gfp::unc-54 3′UTR

pMX10: 3.7kb sma-9p::sma-9C2::gfp::unc-54 3′UTR

pMX11: 3.7kb sma-9p::sma-9C3::gfp::unc-54 3′UTR

Forced expression constructs

pMLF28: hlh-8p::sma-9C1::unc-54 3′UTR

pMLF30: hlh-8p::sma-9C2::unc-54 3′UTR

pMLF27: hsp16p::sma-9C1::unc-54 3′UTR

pMLF29: hsp16p::sma-9C2::unc-54 3′UTR

pMLF31: hlh-8p::sma-3::unc-54 3′UTR

pMLF33: hlh-8p::sma9C1::gfp::unc-54 3′UTR

Detailed information on all these constructs is available upon request.

Transgenic lines were generated as described by Mello and Fire(Mello and Fire, 1995).

Heat shock was applied to transgenic lines that carried the hsp-16p::sma-9C1::unc-54 3′UTR construct (pMLF27) or the hsp-16p::sma-9C2::unc-54 3′UTR construct (pMLF29)in the sma-9(cc604) background. Stage-synchronized animals were collected at each of the following stages based on hlh-8::gfpexpression: embryos, larvae at the 1-M, 2-M, 4-M, 6-M, 8-M, 10- to 16-M and 2-SM stages. L1 larvae usually stay at the 1-M stage for about 8 hours, with subsequent cell divisions each taking about 2 hours leading to the 16-M stage(Sulston and Horvitz, 1977). We divided animals at the 1-M stage into different groups: 1-2 hours, 2-4 hours and 5-7 hours post-hatching. Animals of the same stage were transferred to the same plate for heat-shock treatment (37°C for 1 hour). Animals were examined post heat-shock to ensure that they were still at the indicated stage. They were allowed to recover at 20°C. Their terminal M lineage phenotypes were examined at the young adult stage. Non-heat-shocked and non-transgenic but heat-shocked animals were used as controls.

The following plasmids were used to generate dsRNAs used in the RNAi experiments. All yk clones are gifts from Yuji Kohara.

sma-9: yk127d10 and pJKL619, which has the N terminus of sma-9 from yk1185a11 cloned into pBS II SK+.

sma-2: pJKL716, which contains exons 6-7 of sma-2 cloned into pBS II SK+.

sma-4: pMLF35, which contains exons 7-10 of sma-4 cloned into pBS II SK+.

dsRNA was injected into LW0081 or sma-9 mutant animals following the protocol of Fire and colleagues (Fire et al., 1998). Progeny of injected animals were scored for defects in the M lineage.

Animals were fixed following the protocol in Hurd and Kemphues(Hurd and Kemphues, 2003). Goat anti-GFP antibodies (Rockland Immunochemicals) and rat anti-MLS-2 antibodies (Jiang et al.,2005) were both used in 1:500 dilution. Anti-SMA-9 antibodies(CUMC-R2 TB2, recognizing the C2 isoform)(Liang et al., 2003) were used in 1:2000 dilution. All secondary antibodies were from Jackson Immunoresearch Laboratories and used in 1:50 to 1:200 dilution. Differential interference contrast and epifluorescence microscopy were performed using a Leica DMRA2 compound microscope. Images were captured by a Hamamatsu Orca-ER camera using the Openlab software (version 4.0.1, Improvision). Subsequent image analysis was performed using Adobe Photoshop 7.0.

To identify genes involved in patterning of the M lineage, we screened for mutant animals with altered numbers of coelomocytes (CC) (see Materials and methods). Two recessive mutations, cc606 and cc604, were isolated that failed to produce M lineage-derived CCs(Fig. 1, Table 1). Both mutations mapped to the X chromosome, failed to complement each other, and exhibited smaller body sizes, resembling the small body phenotype of another mutation in the region: sma-9(wk55) (Liang et al., 2003). Six lines of evidence suggest that these alleles are loss or reduction of function mutations in the sma-9 locus. First,all 13 previously described sma-9 alleles lacked M lineage-derived CCs (see Table 1 for wk55, data not shown for qc1, qc3, qc5, qc6, qc7, qc8, qc9, qc10,qc11, wk62, wk71 and wk82). Two additional sma-9alleles isolated from our ongoing screen for mutants affecting the M lineage, jj25 and jj29, also lacked M lineage-derived CCs (data not shown). Second, both cc606 and cc604 failed to complement the M lineage and body size defects of two sma-9 alleles, wk55 and wk62. Third, both cc606 and cc604contained molecular lesions in the sma-9-coding region(Fig. 2). cc606 and cc604 each result in a change of a Gln codon to stop, at positions 78 and 1676, respectively [numbers based on the system of Liang et al.(Liang et al., 2003)]. Fourth,a sma-9-containing cosmid, T05A10, rescued the M lineage defects of both cc606 and cc604 mutants(Fig. 2B; data not shown). Fifth, sma-9(RNAi) resulted in the same M lineage defects as those of the sma-9 mutant alleles (Table 1). Sixth, affinity-purified anti-SMA-9 antibodies (CUMC-R2 TB2)(Liang et al., 2003) failed to detect any signal in sma-9(cc604) mutant animals (data not shown). Together, these results provide strong evidence that cc606 and cc604 are mutant alleles of sma-9 and that cc604behaves as a strong reduction or loss-of-function allele. We therefore chose cc604 for all of our analysis described below.

To determine the fate of the missing CCs in sma-9 mutants, we examined the M lineage in cc604 mutants using several markers that specifically label various M lineage-derived cells (see Materials and methods). Using an integrated hlh-8::gfp that labels all undifferentiated cells in the M lineage(Harfe et al., 1998b), we observed no defects in the division patterns up to the 16-M stage (data not shown). At the 16-M stage, however, the two CC precursors on the dorsal side of the animal (M.dlpa and M.drpa) behaved like their ventral counterparts(M.vlpa and M.vrpa). Both cells (M.dlpa and M.drpa) underwent another round of division to produce two body wall muscles (BWMs) and two sex myoblasts (SMs). This resulted in the loss of two M-derived CCs and the production of two extra SMs and two extra BWMs on the dorsal side of cc604 mutants(Fig. 1B,F). The two SMs on the dorsal side behaved as bona fide SMs. They migrated as normal SMs to the future vulval region (albeit on the dorsal side most of the time), divided three times to produce 16 sex muscle precursors, which then differentiated into sixteen uterine and vulval muscles (um and vm)(Fig. 1D,F; data not shown). The differentiated cell types generated from the M lineage in cc604mutants were confirmed using DIC optics in combination with cell type-specific markers, including myo-3::gfp (labeling BWMs)(Fire et al., 1998) (data not shown), a twist-derived CC::gfp (labeling coelomocytes)(Harfe et al., 1998b)(Fig. 1A-D), egl-15::gfp (labeling type I vulval muscles)(Harfe et al., 1998a)(Fig. 1C,D), arg-1::gfp (labeling both type I and type II vulval muscles)(Kostas and Fire, 2002) (data not shown) and rgs-2::gfp (preferentially labeling uterine muscles)(Dong et al., 2000) (data not shown). Thus, sma-9 mutants exhibit a loss of dorsal, and a duplication of ventral, M-derived cell fates.

Outside of the M lineage, sma-9(cc604) mutants did not show defects in other mesodermally derived cells, including the number and pattern of embryonically derived BWMs, the enteric muscles and the head mesodermal cell, although a low penetrance (<10%) of distal tip cell migration defects(extra turns) was observed (data not shown).

sma-9 encodes a predicted large C₂H₂zinc-finger protein with complex splicing isoforms that differ at both the N and C termini (Liang et al.,2003). As SMA-9 functions in the M lineage and in regulating body size and male tail patterning, we asked whether some SMA-9 isoforms are specifically required for proper M lineage development.

sma-9 is a complex locus that generates a variety of different cDNA species, with alternative splicing contributing to potential heterogeneity at both the N- and C-terminal regions of the protein(Liang et al., 2003). No full-length cDNA clones that encompass the entire predicted sma-9locus have been isolated. Instead, there are cDNA clones that represent two major N-terminal and three major C-terminal isoforms(Liang et al., 2003). When we generated artificial full-length cDNA constructs by piecing together different combinations of the N- and C-terminal cDNA clones, none of these constructs could rescue either the M lineage or body size defects of sma-9mutants. A set of constructs was then made that would maintain what could be an important physiological diversity at the N terminus, while constraining the C-terminal structure (where the Zn fingers reside) to one of several isoforms. These genomic-cDNA hybrid constructs contained genomic sequences from the N terminus and cDNA sequences for each of the three different sma-9C-terminal isoforms, which we termed C1, C2 and C3(Fig. 2A, Materials and methods). C1 is predicted to contain all seven zinc-finger (ZF1-7) domains. C2 contains a unique 70 amino acid domain instead of ZF6 and ZF7 at its C-terminal end. C3 is the shortest isoform and lacks ZF3-7. For each of the three constructs (C1, C2 and C3), we placed a GFP tag at the C terminus and used the predicted sma-9 promoter to drive its expression(Fig. 2A, Materials and methods).

In the wild-type background, all three constructs showed similar GFP expression pattern and subcellular localization: GFP was localized in nuclei of a wide variety of cell types (Fig. 3A), similar to what we have previously reported using C2 isoform-specific anti-SMA-9 antibodies(Liang et al., 2003). Both the Cl and C2 constructs, but not the C3 construct, rescued the M lineage and body size defects of sma-9(cc604) mutants(Fig. 2B). The rescue efficiency for both C1 and C2 was even better than that for the sma-9-containing cosmid T01A5(Fig. 2B). These results suggest that the structures unique to the C1 and C2 isoforms, including ZF6-7(for C1) and the 70 amino acid domain (for C2), are not crucial for SMA-9 function (Fig. 2B). Instead,the structures missing in the C3 isoform but present in both C1 and C2,including the ZF3-5 zinc-finger cluster, are apparently required for proper SMA-9 function in regulating both body size and M lineage patterning.

In addition to the functional plasticity at the C terminus observed above,some plasticity in the N-terminal region is also tolerated for sma-9. In particular, we observed wild-type M lineage and body size in a sma-9 deletion allele, tm572, from the Japanese knockout consortium(http://www.shigen.nig.ac.jp/c.elegans/). tm572 contains a deletion of the predicted exons 6, 7 and part of 8 in sma-9 (Fig. 2A),and it appears to make a functional SMA-9 protein as tm572/cc604heterozygous animals had a normal M lineage and wild-type-like body size(Table 1). Furthermore, when we injected dsRNA against predicted exons 1-4 (RNAi-N) and exons 11-20(RNAi-C) into tm572 animals, we observed M lineage defects and a small body size, similar to phenotypes observed in all known sma-9 mutant alleles (data not shown). These results suggest that the sequences deleted in tm572 mutants, including exons 6-8, are not required for sma-9 function in regulating M lineage patterning and body size.

We used the rescuing constructs described above to examine functional requirements for SMA-9 in the M lineage. To first obtain a detailed expression pattern, we double stained transgenic animals carrying the C1::GFP construct with antibodies against GFP and antibodies against an M-lineage specific marker MLS-2 (Jiang et al.,2005). We detected GFP signals in the M lineage at both the 1-M and 2-M stages (Fig. 3B-D) and up to the 8-M stage (data not shown). Notably, both within and outside of the M lineage, the dorsal and ventral cells showed similar distribution and intensity of the SMA-9::GFP signals.

To test if SMA-9 functions within the M lineage to regulate its proper patterning, we replaced the sma-9 promoter with the M lineage-specific hlh-8 promoter in the C1 and C2 rescuing constructs. We confirmed that these constructs are expressed only in the M lineage by examining the expression pattern of GFP in transgenic animals carrying the hlh-8p::C1::GFP construct (data not shown). As shown in Fig. 2B, both constructs specifically rescued the M lineage defects without rescuing the body size defects of sma-9(cc604) mutants. Although these M lineage-specific rescue results cannot exclude the possibility that SMA-9 activity outside of the M lineage could contribute to patterning of the M lineage, they certainly illustrate the ability of SMA-9 produced within the M lineage to establish proper patterning.

To determine the crucial time period when SMA-9 function is required for the proper development of the M lineage, we generated transgenic lines of sma-9(cc604) animals carrying either C1 or C2 SMA-9 constructs driven by a heat-shock inducible promoter. These animals were heat-shocked at distinct stages during larval development (1-M, 2-M etc. Fig. 1E) and assayed for rescue of the M lineage defects of cc604 mutants (see Materials and methods for heat-shock conditions).

As shown in Table 2,heat-shocking between the 1-M and 6-M stage resulted in rescue of the M lineage defects of cc604 mutants. The most efficient rescue was observed when animals were heat-shocked between the 2-M (C1: 93%, n=27; C2: 80%, n=5) and 4-M (C1: 100%, n=6; C2:88%, n=8) stages. No rescue was observed when animals were heat-shocked at or after the 8-M stage(Table 2). Both the C1 and C2 SMA-9 isoforms behaved similarly, although the rescuing efficiency for C1 appeared higher than that of C2 (Table 2). As rescue was observed even when SMA-9 expression was induced after the 4-M stage (at the 6-M stage, C1: 43%, n=7; C2: 17%, n=6), we concluded that SMA-9 is sufficient to act after the first cell division of the M lineage for its proper cell-fate specification. The lack of rescue in animals heat-shocked at or after the 8-M stage suggests that SMA-9 function is most probably required prior to the 8-M stage for proper M lineage development.

sma-9 encodes a homolog of the Drosophila protein Schnurri (Shn), and has been previously shown to function as a downstream component of the Sma/Mab TGFβ signaling pathway in regulating body size and male tail patterning (Liang et al.,2003). To test whether SMA-9 functions in a similar manner in the M lineage, we examined the M lineage phenotypes in mutants of known components of the TGFβ signaling pathway in C. elegans. As shown in Table 3, none of the mutants examined showed any M lineage defects, suggesting that SMA-9 might function through different mechanisms in the M lineage than its function in regulating body size and male tail patterning.

To further understand how SMA-9 functions in the M lineage, we carried out a suppressor screen for mutations that can restore the two M-derived CCs in sma-9(cc604) mutants (see Materials and methods). Seven recessive suppressor mutations (jj1-jj7) were isolated. All seven specifically suppressed the M lineage defects without suppressing the body size defects of cc604 mutants (Fig. 4A-A′,B-B′; Table 4, data not shown). We mapped jj1 and jj3 via snip-SNP mapping (Wicks et al.,2001). jj1 maps to chromosome II between -6.25 mu and+1.85 mu; jj3 maps to chromosome III between -1.45 mu and +5.38 mu. Subsequent out-crossing of jj1 and jj3 from the sma-9(cc604) mutation revealed that jj1 and jj3homozygous animals had a wild-type M lineage but a small body size. Complementation assays between jj1 and a sma mutation on chromosome II [sma-6(e1482)], and between jj3 and three sma mutations on chromosome III [sma-2(e502), sma-3(e491)and sma-4(e729)] showed that jj1 failed to complement sma-6(e1482) and jj3 failed to complement sma-3(e491). We further identified the molecular lesions in jj1 and jj3. jj1 contains a G to A change in sma-6,resulting in a Trp (TGG) to stop (TAG) mutation at amino acid 328. This mutation is predicted to result in a truncated SMA-6 protein lacking the intracellular kinase domain and is probably a null allele of sma-6. jj3 contains a T to A substitution at the beginning of intron 9 in sma-3, which is likely to cause aberrant splicing, and results in a truncation of the highly conserved MH2 domain. Thus, jj3 represents a likely null allele of sma-3.

To further test whether the interactions between sma-3(jj3) and sma-9(cc604), and between sma-6(jj1) and sma-9(cc604) were allele specific, we generated double mutants between sma-9(cc604) and other existing alleles of sma-3 and sma-6: sma-3(e491), sma-3(wk28) (a putative null allele), sma-6(e1482) and sma-6(wk7) (a null allele)(Brenner, 1974; Krishna et al., 1999; Savage-Dunn et al., 2000). We also generated double mutants between sma-3(jj3) and two other alleles of sma-9: sma-9(wk55) and sma-9(cc606). As shown in Table 4, all these double mutants exhibited a wild-type M lineage while retaining a small body size,suggesting that the genetic interactions between sma-3 and sma-9 and between sma-6 and sma-9 are not allele specific.

sma-3 and sma-6 both encode components of the Sma/Mab TGFβ signaling pathway. SMA-3 is one of the regulatory Smad proteins and SMA-6 is the type I receptor (Savage-Dunn,2001). The genetic interaction between sma-9 and these two genes suggested that SMA-9 may participate in regulating the output of the Sma/Mab TGFβ signaling pathway in the M lineage. To test this hypothesis,we generated double mutants between sma-9(cc604) and available mutations in other core members of the Sma/Mab TGFβ signaling pathway. These include the Sma/Mab ligand DBL-1, the type II receptor DAF-4, the other regulatory Smad (SMA-2) and the co-Smad, SMA-4(Estevez et al., 1993; Morita et al., 1999; Savage et al., 1996; Savage-Dunn, 2001; Suzuki et al., 1999). None of the genes encoding these components when mutated display any M lineage defects on their own (Table 3). However, mutations in any of them completely or nearly completely suppress the M lineage defect of sma-9(cc604) mutants. All double mutants with sma-9(cc604) were small, indicating that the small body size defect of sma-9(cc604) mutants was not suppressed(Table 4). In the case of dbl-1, we observed a low level of suppression (14%) of sma-9(cc604) M lineage defects in the dbl-1(wk70)/+background (Table 4). This suggests that simply lowering the level of DBL-1 is sufficient to rescue the M lineage phenotypes of sma-9 mutants. Because mutations in the core components of the Sma/Mab TGFβ signaling pathway all suppressed the M lineage defects of sma-9 mutants, we conclude that SMA-9 functions to antagonize the core Sma/Mab TGFβ signaling pathway for the correct patterning of the M lineage.

The inability of the TGFβ mutants to suppress the body size defects of sma-9 mutants suggests that the mechanism of SMA-9 function in the M lineage might be different from its function in regulating body size. We further examined this by generating double mutants between sma-9(cc604) and mutations in additional genes that appear to function in regulating body size, including lon-1(e185), two alleles of rnt-1,e1241 and ok351, and lon-2(e678)(Ji et al., 2004; Maduzia et al., 2002; Morita et al., 2002; Brenner, 1974). lon-1and rnt-1 function downstream (Ji et al., 2004; Maduzia et al.,2002), while lon-2 is proposed to function upstream(Brenner, 1974), of the core Sma/Mab TGFβ pathway. We found that neither of the two rnt-1alleles suppresses the M lineage defects of sma-9, and that lon-1(e185) suppresses the M lineage defects of sma-9 at a very low penetrance (7%) (Table 4). However, lon-2(e678), a predicted null allele of lon-2 (R. Padgett, personal communication), suppresses the M lineage defects of sma-9 at a relatively high penetrance (88%)(Table 4). We also noticed that the lon-2(e678) sma-9(cc604) double mutant animals appeared wild type with respect to body size (Table 4). Although we do not understand why the lon-2 mutation would suppress the M lineage defects of sma-9 mutants, the M lineage and the body size phenotypes of lon-2(e678) sma-9(cc604) mutants that we observed suggest that SMA-9 and LON-2 are likely to function in parallel to each other. Collectively, our data also indicate that not all genes involved in body size regulation are involved in M lineage patterning.

To further determine if the interaction between sma-9 and the Sma/Mab TGFβ pathway is specific, we generated double mutants between sma-9(cc604) and daf-1(m40). daf-1 encodes the type I receptor unique to the TGFβ pathway in controlling dauer development(Georgi et al., 1990). We found that daf-1(m40) does not suppress the M lineage defects of sma-9(cc604) (Table 4). Similarly, egl-20(n585) and lin-17(n671),mutations in the Wnt signaling pathway(Maloof et al., 1999; Sawa et al., 1996), do not suppress the M lineage defects of sma-9(cc604) (data not shown). These results indicate a specific role for SMA-9 in modulating the core Sma/Mab TGFβ pathway in M lineage patterning.

To test whether the antagonism between SMA-9 and the TGFβ pathway is occurring within M lineage cells, we used the M lineage-specific hlh-8 promoter to express sma-3 exclusively in the M lineage of sma-3(jj3); sma-9(cc604) double mutants. If SMA-3 functions cell autonomously in the M lineage, we would expect that introducing sma-3specifically in the M lineage would eliminate the suppression phenotype and restore the sma-9 mutant phenotype. We found that in four independent lines, greater than 92% (n=279) of transgenic animals were missing both M-derived CCs, indicating that expressing sma-3 exclusively in the M lineage is sufficient to eliminate suppression(Fig. 4C, Table 4). These data suggest that the interaction between the Sma/Mab TGFβ pathway and SMA-9 occurs cell-autonomously within the M lineage.

We have shown that SMA-9 is required for the correct dorsoventral patterning of the C. elegans post-embryonic mesoderm. In sma-9 mutants, the dorsal M lineage-derived cells are transformed into ventral M lineage-derived cells. This ventralization appears to be specific to the M lineage, as we did not observe any other defects in overall dorsoventral patterning in sma-9 mutants (data not shown).

SMA-9 has previously been shown to function in regulating body size and male tail patterning (Liang et al.,2003). Our finding adds another example to the multiple functions of SMA-9 in C. elegans. SMA-9 is a homolog of the Drosophilaprotein Schnurri (Shn) (Liang et al.,2003). Like SMA-9, Shn has multiple functions during Drosophila development, such as dorsoventral patterning of the early embryo, proliferation and differentiation in the wing, and maintenance of germline stem cells (Arora et al.,1995; Grieder et al.,1995; Shivdasani and Ingham,2003; Staehling-Hampton et al., 1995; Torres-Vazquez et al., 2000; Xie and Spradling,1998). Our data provides evidence for a conserved function of SMA-9 and Shn in dorsoventral patterning. In both animals, SMA-9 and Shn are required to specify dorsal cell fates. Vertebrates have multiple Shn/SMA-9 homologs, including Shn1, Shn2 and Shn3(Fan and Maniatis, 1990; Gascoigne, 2001; Jin et al., 2006; Lallemand et al., 2002; Nakamura et al., 1990; Oukka et al., 2002; Seeler et al., 1994; Takagi et al., 2001; van 't Veer et al., 1992);however, no reports have shown their functions in dorsoventral patterning. Our results raise the possibility that one or more members of this protein family also function in dorsoventral patterning in vertebrates.

sma-9 has multiple splicing isoforms(Liang et al., 2003). We have found that SMA-9 with either of two different C-terminal ends (C1 or C2) can rescue the M lineage and body size defects of sma-9 mutants(Fig. 2). These results suggest that the structures unique to either C1 or C2 are not crucial, although each alone could be sufficient, for SMA-9 function in regulating the M lineage and body size. The lack of rescue by the C3 isoform suggests that the common motifs shared by C1 and C2, including the three zinc fingers (ZF3-5), are crucial for SMA-9 function. Understanding how these zinc fingers contribute to SMA-9 function will help us better understand how SMA-9 regulates M lineage patterning.

Previous studies have shown that sma-9 mutants exhibit a small body size and defects in male tail patterning and that these defects are similar, but not identical, to those exhibited by mutations in the core Sma/Mab TGFβ signaling pathway (Liang et al., 2003). Furthermore, sma-9 acts downstream of the DBL-1 ligand in regulating both body size and male tail patterning(Liang et al., 2003). These studies clearly showed a contributory role for SMA-9 in the TGFβ response pathway. In this work, we uncovered a novel and unexpected role for the Sma/Mab TGFβ signaling pathway that appears to be antagonized by SMA-9. We have shown that none of the core Sma/Mab TGFβ signaling pathway members have any M-lineage defects when mutated(Table 3). However, mutations in any of them suppress the M lineage defects of sma-9 without suppressing its body size defect (Table 4). Furthermore, this suppression is specific to the Sma/Mab TGFβ signaling pathway and it is occurring specifically in the M lineage(Table 4).

There are two possible scenarios for how SMA-9 might antagonize the TGFβ signaling pathway in M lineage development. One possibility is that SMA-9 functions upstream of the ligand DBL-1 to reduce its expression or activity. However, we have shown that both SMA-9 and SMA-3 function within the M lineage for its correct patterning (Figs 3, 4). Thus, it is unlikely that SMA-9 functions upstream of DBL-1 in regulating M lineage patterning, unless DBL-1 functions in an autocrine manner. Supporting this notion, we have found that DBL-1 expression is not altered in sma-9(cc604) mutants (data not shown).

A second possibility is that SMA-9 acts to repress TGFβ target gene expression. Previous studies in Drosophila have shown that Shn mediates BMP signaling by forming a repressor complex with the Smad proteins Mad and Medea to repress gene expression(Marty et al., 2000; Muller et al., 2003; Pyrowolakis et al., 2004). Recently, it has been shown that mouse Shn-2 can bind to Smad1/4 to activate PPAR γ2 gene expression during adipocyte differentiation (Jin et al.,2006). We have found that the suppression of the sma-9 M lineage phenotype by sma-3 is not allele specific(Table 4). Additionally,mutations in sma-2 and sma-4 as well as RNAi knockdown of these two genes all suppressed the M lineage defects of sma-9 mutants(Table 4). Although these observations cannot completely rule out the possibility that SMA-9 functions by forming a complex with SMA-2, SMA-3 or SMA-4 in the M lineage, we favor the model that SMA-9 functions in parallel to the TGFβ pathway and directly binds to the regulatory regions of TGFβ targets in the M lineage to block the actions of the Smad proteins. Consistent with this, studies by Wang and colleagues (Wang et al., 2005)failed to detect a direct interaction between SMA-9 and SMA-2, SMA-3 or SMA-4 using the yeast two-hybrid system.

The antagonism between SMA-9 and the core Sma/Mab TGFβ signaling pathway may not be restricted to the M lineage. Liang and colleagues(Liang et al., 2003) have previously reported that sma-9; Sma/Mab pathway double mutants exhibit phenotypes similar to those of Sma/Mab pathway single mutant in terms of male ray patterning (Liang et al.,2003). In particular, the moderate reduction in the frequency of serotonergic fate expression in R9B and weak ectopic expression of this fate in R5B and R7B phenotypes of sma-9 mutants were completely suppressed by Sma/Mab pathway mutants. This was interpreted to suggest that SMA-9 acts downstream of the Sma/Mab pathway (Liang et al., 2003). In light of our current findings, these results may actually suggest that SMA-9 antagonizes the function of the Sma/Mab pathway in male ray patterning. Further molecular analysis of SMA-9 and TGFβ pathway targets will shed light on how SMA-9 antagonizes the function of the TGFβsignaling pathway in various cell types.

One surprising finding from our studies is that the lon-2 mutation also suppresses the M lineage defects of sma-9 mutants(Table 4). LON-2 has been previously proposed to function upstream, and possibly as a negative regulator, of the Smad proteins in regulating body size as sma-2;lon-2 double mutant animals are small(Brenner, 1974). We have found that a null allele of lon-2, e678 (R. Padgett, personal communication), partially suppresses the M lineage defects of sma-9(cc604) mutants (Table 4). Furthermore, lon-2(e678) sma-9(cc604) double mutant animals appear to have a wild-type body size(Table 4). These observations indicate that SMA-9 and LON-2 could function in parallel pathways and that SMA-9 functions differently from the Smad proteins in regulating body size and M lineage patterning.

SMs and CCs are descendants of two M daughter cells, M.v and M.d, after the first cell division in the M lineage (Fig. 1). Because forced expression of sma-9 between the 2-M and 6-M stages using a heat-shock inducible promoter can rescue the M lineage defects of sma-9 mutants (Table 2), we believe that dorsoventral asymmetry of the M lineage is not established during the first division of the M mesoblast. Instead, M.v and M.d are probably born with the same potential to generate either the ventral SM fate or the dorsal CC fate. The different developmental fate of these two cells is probably due to influences of the different environments that these two cells reside in. We believe that the decision of cells becoming either the CC or SM precursors is most likely made by the 8-M stage because forced expression of SMA-9 after the 8-M stage fails to rescue the M lineage defects of sma-9 mutants (Table 2). As SMA-9 is not asymmetrically expressed in the M lineage, we favor the model that the Sma/Mab TGFβ signaling pathway acts to promote SM cell fate dorsally in the M lineage, and that this action is antagonized by SMA-9, resulting in CCs being formed. We are currently investigating when and where the Sma/Mab TGFβ signaling pathway is activated with respect to the M lineage.

SMA-9 and the Sma/Mab TGFβ signaling pathway alone cannot be exclusively responsible for specifying all the fates of the M lineage,especially in the ventral M lineage. Previous studies (e.g. Greenwald et al., 1983) have shown that LIN-12/Notch is required for the proper formation of the ventral CC fates. It will be interesting to determine how SMA-9, the Sma/Mab TGFβsignaling pathway and the LIN-12/Notch signaling pathway are integrated to properly specify the correct dorsal versus ventral cell fates in the M lineage.

We thank Jun Liang and Cathy Savage-Dunn for identifying the molecular lesion in cc604; Sung-Eun Lim for complementation tests between jj25, jj29 and cc604; Yuji Kohara, Shohei Mitani,Cathy Savage-Dunn and Bill Wood for C. elegans strains and cDNA clones; and Melissa Beers, Iva Greenwald, Yuan Jiang, Ken Kemphues, Diane Morton, Cathy Savage-Dunn and Adam Zahand for helpful discussions and valuable comments on the manuscript. We also thank two anonymous reviewers for valuable suggestions on the manuscript. Some strains used in this study were obtained from the C. elegans Genetics Center (CGC), which is supported by NIH. This work was supported by NIH R01 GM066953 (to J.L.). During part of this work, M.L.F. and N.M.A. were supported by NIH T32 GM07617, and A.S.L. was a Howard Hughes Undergraduate Research Scholar at Cornell University.