Positive and Negative Regulation of Muscle Cell Identity by Members of the hedgehog and TGF-β Gene Families

Embryos were staged by hours (h) after fertilization at 28.5°C (Kimmel et al., 1995; available at World Wide Web address http://zfish.uoregon.edu). Chorions were removed with watchmaker's forceps, and embryos were maintained in Ringer's solution (Westerfield, 1995). Older embryos were anesthetized in a 0.6 mM solution of tricaine (Sigma Chemical Co., St. Louis, MO) to inhibit movement during observation.

To link the tiggy-winkle hedgehog (twhh)¹ promoter to the gene encoding nuclear β-galactosidase (nβ-gal), plasmid pGEM-7Z-twhh5.5 containing a 5.5-kb genomic fragment of the twhh gene (Ekker et al., 1995b) was digested with SacI and BamHI and then deleted at the 3′ end of the promoter to position −49 with respect to the ATG translation start codon by EXO-III nuclease using an Erase-a-Base kit (Promega Corp., Madison, WI). The resulting plasmid, pGEM-7Z- twhh5.2, containing the 5.2-kb twhh promoter, was partially digested with BstXI and EcoRI to release the DNA insert. The EcoRI site at the 5′ end of the insert was then blunted by Klenow DNA polymerase and subcloned into the pBluescript-SK BstXI site that had been partially blunted by T4 DNA polymerase. The resulting plasmid, pBluescript-SK-twhh5.2, was digested with SacI and then blunted by T4 DNA polymerase. This linearized plasmid, pBluescript-SK-twhh5.2, was subsequently digested with SalI. The DNA insert containing the 5.2-kb twhh promoter sequence was purified and cloned into plasmid pCS-nβ-gal (a gift from D. Turner, R. Rupp, J. Lee, and H. Weintraub, Fred Hutchinson Cancer Research Center, Seattle, WA) at the SalI and HindIII sites, the latter site having been blunted by Klenow DNA polymerase. The resulting plasmid, pCS-twhh-β-gal, contains the 5.2-kb twhh promoter and the nuclear β-gal reporter gene.

To make the twhh promoter into a convenient expression vector for expressing heterologous cDNAs, the BamHI site upstream of the twhh promoter in the plasmid pCS-twhh-β-gal was deleted. The resulting plasmid, which retains the BamHI site between the promoter and the β-gal, was isolated and designated pCS-twhh-β-gal-vec. Genes of interest can be cloned into the BamHI and XhoI sites of this vector by replacing the β-gal sequence.

The reporter gene, “bright” green fluorescent protein (bGFP) with a serine 65 to threonine mutation (Heim et al., 1995), was cloned into vector pCS-twhh-β-gal-vec BamHI/XhoI sites by blunt end ligation. The resulting plasmid was named pCS-twhh-bGFP.

dorsalin-1 cDNA was amplified from 9-d chick embryos by reverse transcriptase PCR using primers based on the published chick dsl-1 DNA sequence (Basler et al., 1993). The sequences for the 5′ and 3′ PCR primers were 5′ CTCTGTCTGTAAAGATTCAAC 3′ and 5′ GTACAGTTTCACAGACAGCAG 3′, respectively. The PCR product was subcloned into the pCR-II vector (Invitrogen, San Diego, CA). The c-myc–tagged derivative (dsl-1^myc) was constructed as previously described (Basler et al., 1993), with all subcloning steps carried out in the pCR-II vector. To place dsl-1^myc after the twhh promoter, the DNA insert of dsl-1^myc was first subcloned into the EcoRI site of expression vector pCS²⁺ (a gift from D. Turner, R. Rupp, J. Lee, and H. Weintraub), which has a cytomegalovirus promoter and a polyadenylation site. The resulting plasmids were named pCS²⁺-dsl-1^myc. To link dsl-1^myc to the twhh promoter, the dsl-1^myc insert was released from pCS²⁺-dsl-1^myc by BamHI and XhoI digestion and then subcloned into pCS-twhh-β-gal-vec BamHI and XhoI sites by replacing the β-gal sequence. The final construct, pCS-twhh-dsl-1^myc, contains the 5.2-kb twhh promoter and dsl-1^myc.

Plasmid pT7TSshh and pT7TStwhh contain the zebrafish shh and twhh cDNAs, respectively (Ekker et al., 1995b); plasmid pT7TS-X-shhfs contains the Xenopus shh with a single base pair insertion, resulting in a frame shift in amino acid position No. 39 (Ekker et al., 1995a). Plasmid pCS2⁺dnPKA-GFP contains the dominant negative form of PKA regulatory subunit (Ungar and Moon, 1996). Plasmid pSP64T-PKA* contains the constitutively active PKA catalytic subunit (Hammerschmidt et al., 1996a).

shh and twhh RNAs were transcribed from DNA plasmid T7TSshh or T7TStwhh as described (Ekker et al., 1995b). Capped mRNAs were transcribed from linearized DNA template with a T7 RNA polymerase in vitro transcription kit (mMESSAGE mMACHINE T7, Ambion, Inc., Austin, TX) according to the manufacturer's instructions.

β-Gal labeling was carried out with minor modification of published procedures (Westerfield et al., 1992). Embryos were fixed for 30 min at room temperature with rotation, with or without removing the chorion, in 4% paraformaldehyde, 0.2% glutaraldehyde, 4% sucrose, 0.15 mM CaCl₂, and 1× PBS. To visualize β-gal activity, embryos were rinsed twice for 15 min with PBS containing 0.1% Triton X-100 and then incubated in reaction solution containing 0.04% x-gal (bromo-4-chloro-indoxyl-β-d-galactoside), 1 mM MgCl₂, 3.3 mM K₄[Fe₃(CN)₆], and 3.3 mM K₃[Fe₂(CN)₆] at room temperature for 1–2 h. The reaction was stopped by replacing the substrate solution with PBS.

For DNA microinjection, linearized DNA was dissolved in distilled H₂O to a final concentration of 50 μg/ml. For mRNA injection, mRNA was dissolved in distilled H₂O to a final concentration of 100 μg/ml. A final concentration of 0.1% phenol red was added to the DNA or RNA solution to facilitate visualization during microinjection. Approximately 2 nl of DNA or RNA solution was microinjected into the cytoplasm of zebrafish embryos at the one- or two-cell stage.

For antibody labeling, embryos were fixed in 4% paraformaldehyde in 1× PBS for 2 h at room temperature. Embryos were washed twice for 5 min in 1× PBS and once for 5 min in water. Embryos were then soaked in cold acetone for 10 min at −20°C. Embryos were washed once with water for 5 min, and twice with 1× PBS for 5 min each, and once with BDP (0.1% bovine serum albumin, 1% dimethylsulfoxide, 1× PBS) for 5 min. For labeling using monoclonal antibodies, the embryos were incubated with avidin-blocking reagent (Vector Laboratories, Burlingame, CA) and 10% goat serum in BDP for 30 min at room temperature. The embryos were then rinsed twice for 5 min in BDP and subsequently incubated with 1:5 diluted antiengrailed monoclonal antibody 4D9 (Patel et al., 1989) and/or 1:500 diluted anti–c-myc monoclonal antibody (Oncogene Science, Cambridge, MA) together with biotin-blocking reagent (Vector Laboratories) overnight at 4°C with shaking. The embryos were then washed three times for 30 min with BDP, followed by incubation with 1:500 diluted biotin-labeled secondary antibody (Vector Laboratories) in BDP for 1 h at 37°C. Embryos were then washed three times for 30 min with BDP and incubated with 1:1 diluted ABC solution (avidin biotin complex) for 30 min at room temperature. Embryos were washed three times for 30 min in BDP and then presoaked in DAB solution (0.05% diaminobezidine, 1% DMSO, 1× PBS) for 10 min at room temperature. The DAB soaking solution was then replaced by DAB staining solution containing 0.003% of H₂O₂ in DAB soaking solution. The staining was monitored and stopped by washing twice for 10 min with BDP. Embryos were photographed in PBS.

Immunofluorescent labeling of sections with the F59 and 4D9 monoclonal antibodies was done as previously described (Crow and Stockdale, 1986; Devoto et al., 1996b).

Three distinct types of embryonic muscle fibers can be identified in zebrafish based on position, gene expression, and pattern of immunoreactivity with several monoclonal antibodies. Their development is summarized in Fig. 1. Slow muscle precursors, known as adaxial cells, develop adjacent to the notochord and then migrate radially through the somite to become a monolayer of muscle cells on the surface of the myotome (Devoto et al., 1996b). A subset of the slow muscle precursors, located at the future horizontal myoseptum, remain in contact with the notochord and become flattened cells that extend from the notochord to the lateral surface of the myotome (Waterman, 1969; van Raamsdonk et al., 1974; Devoto et al., 1996b). These cells, called muscle pioneers (Felsenfeld et al., 1991), are the only slow muscle precursors to express the engrailed1 and engrailed2 genes at high levels (Hatta et al., 1991; Ekker et al., 1992). Fast muscle precursors, in contrast, develop from lateral presomitic cells and remain deep within the myotome.

Previous studies have shown that ectopic expression of Hedgehog induces MyoD and supernumerary muscle pioneer cells, suggesting that it may play an important role in muscle development in zebrafish (Currie and Ingham, 1996; Hammerschmidt et al., 1996a; Weinberg et al., 1996). To examine directly whether hedgehog genes influence the development of muscle fiber type identity, we expressed zebrafish sonic hedgehog or tiggy-winkle hedgehog ectopically by injection of RNA into cleavage stage embryos. We then examined the developing embryos for induction of slow muscle cells using several monoclonal antibodies that recognize the entire population of slow muscle cells, including the muscle pioneers. F59 recognizes myosin heavy chain in fish (Miller et al., 1989); in zebrafish it specifically labels slow muscle fibers during the first day of development, and then later it also weakly labels fast muscle fibers (Devoto et al., 1996b). We also used zn5 (Trevarrow et al., 1990) and S58 (Crow and Stockdale, 1986) antibodies that also label slow but never label fast muscle fibers in zebrafish (Devoto et al., 1996b). We found that both Sonic hedgehog and Tiggy-winkle hedgehog induced the development of many extra slow muscle cells. Specifically, as in uninjected embryos, only one layer of slow muscle cells was present in the superficial layer of the somite in control embryos injected with frame-shifted sonic hedgehog RNA (Fig. 2,A), whereas in embryos injected with sonic hedgehog (Fig. 2,B) or tiggy-winkle hedgehog (Fig. 2,C) RNA, almost all cells in the somite differentiated into slow muscle. These ectopic slow muscle cells were also labeled by the S58 and zn5 antibodies, indicating that these cells had fully differentiated as slow muscle fibers (data not shown). Presumably, these extra slow muscle cells are formed at the expense of fast muscle because they occupy the locations where fast muscle cells normally form, and because nearly all the muscle cells in the somite exhibited these slow muscle properties. Both Sonic hedgehog (Fig. 2, E and H) and Tiggy-winkle hedgehog (Fig. 2, F and I) also induced extra muscle pioneer cells, as determined by labeling with the anti–engrailed monoclonal antibody, 4D9. In control embryos injected with frame-shifted sonic hedgehog, two to six muscle pioneer cells were normally present in each somite (Fig. 2, D and G) as in uninjected embryos, whereas Sonic hedgehog induced an average of 20 muscle pioneer cells per somite (88%, n = 87; Fig. 2, E and H), and Tiggy-winkle hedgehog induced an average of 10 muscle pioneer cells per somite (75%, n = 105; Fig. 2, F and I).

Protein kinase A (PKA) is an integral part of the Hedgehog signaling pathway (for review see Perrimon, 1995). PKA constitutively represses Hedgehog target genes, and Hedgehog acts to relieve this repression. Thus, expression of a dominant negative isoform of PKA mimics Hedgehog signaling in both Drosophila (Jiang and Struhl, 1995; Li et al., 1995; Pan and Rubin, 1995) and in vertebrates (Fan et al., 1995; Hammerschmidt et al., 1996a; Ungar and Moon, 1996). Our results (Fig. 2) suggested that Hedgehog is sufficient to trigger slow muscle development. To test whether Hedgehog signaling is required for slow muscle development, we ectopically expressed the constitutively active PKA isoform (Orellana and McKnight, 1992). Compared with control embryos (Fig. 3,A), slow muscle cells labeled with F59 antibody appeared to be absent in embryos injected with RNA encoding the constitutively active isoform of PKA (Fig. 3,B). Frequently, injected RNAs are localized to one region of the embryo (Hammerschmidt et al., 1996a). Consistent with this, transverse sections through control (Fig. 3,C) and active PKA-injected embryos demonstrated a local loss of slow muscle cells in the active PKA injected embryos (Fig. 3,D). Together with the Hedgehog ectopic expression data (Fig. 2), this result suggests that Hedgehog signaling is required for the development of all slow muscle cells, including muscle pioneer cells (Hammerschmidt et al., 1996a, and data not shown).

Interestingly, we observed that the ectopic muscle pioneer cells induced by Hedgehogs appeared only in the region of the somite nearest the notochord; ectopic muscle pioneers were absent in the dorsal or ventral thirds of the somite. Because Hedgehogs were active in the induction of non– muscle pioneer slow muscle cells in these regions, this suggested that inhibitory signals from tissues near these regions may antagonize the activity of Hedgehogs.

Competition between BMP4 expressed in the dorsal neural tube and Sonic hedgehog expressed in the ventral neural tube has been shown to play an important role in dorsoventral patterning of the spinal cord (Basler et al., 1993; Liem et al., 1995). Somite patterning may also be regulated by competing positive and negative signals, including BMP4 (Fan and Tessier-Lavigne, 1994; Pourquié et al., 1996). To learn whether a BMP4-like protein can affect the development of muscle cell identity in zebrafish, we tested whether ectopic expression of Dorsalin-1, a BMP4-like factor, would inhibit the formation of muscle pioneer cells in surrounding somites. We used the chick Dorsalin-1 in this study for several reasons. First, the dorsal neural tube (Basler et al., 1993), a tissue known to play a role in somite patterning (Lassar and Munsterberg, 1996; Pourquié, et al., 1996), expresses Dorsalin-1. Second, Dorsalin-1 can antagonize Hedgehog signaling in the dorsoventral patterning of the neural tube (Basler et al., 1993; Liem et al., 1995). Third, at the time we initiated this study, no gene encoding BMP or a BMP-like protein expressed in the neural tube had been isolated from zebrafish. More recently, a BMP-like gene named radar was reported in zebrafish; however, this clone contains only the partial coding region (Rissi et al., 1995).

In initial experiments, we found that injection of Dorsalin-1 mRNA had a severe ventralizing effect during gastrulation, similar to that caused by injection of BMP4 mRNA (Hammerschmidt, et al., 1996b). Thus, to assess potential later effects on somite patterning, we expressed Dorsalin-1 in the notochord after gastrulation. Additionally, expression of Dorsalin-1 in the notochord localized the protein to the region of the somites, where we anticipated the lowest activity of the putative inhibitor of Hedgehog signaling. To express a potential inhibitor specifically in this region of the somites, we put dorsalin-1 under the control of a promoter from the tiggy-winkle hedgehog gene. The floor plate normally expresses Tiggy-winkle hedgehog. Paradoxically, we found that 5.2 kb of the 5′-flanking sequence from the tiggy-winkle hedgehog gene leads to expression of heterologous proteins, including β-galactosidase, specifically in the notochord (Fig. 4 A; a further characterization of this promoter is in progress). Thus, we used this promoter fragment to express Dorsalin-1 in the notochord.

Embryos injected with the dorsalin-1 DNA construct (twhh-dsl-1^myc) developed with apparently normal anteroposterior and dorsoventral axes. As expected, the tiggy-winkle hedgehog promoter drove expression of Dorsalin-1 specifically in notochord cells (Fig. 4,B, arrowheads), consistent with the expression pattern of the β-gal reporter gene under control of the same promoter (Fig. 4,A). To analyze whether Dorsalin-1 has an inhibitory effect on the development of muscle pioneer cells, we examined embryos injected with the twhh-dsl-1^myc for differentiation of muscle pioneer cells labeled with the 4D9 antibody. As shown by the bracket in Fig. 4,D, muscle pioneer cells (arrows) were absent in the somites adjacent to notochord cells expressing the twhh-dsl-1^myc construct. In contrast, muscle pioneer cells developed normally in embryos injected with the control construct, twhh-bGFP (Fig. 4 C, arrows). A single Dorsalin-1–expressing cell in the notochord was able to inhibit the formation of muscle pioneer cells in the flanking two to four somites. Usually there were more somites affected rostral than caudal to the Dorsalin-1–expressing notochord cell (data not shown). This is probably because notochord cells shift caudally relative to the somites, from about 12 h to at least 48 h (Devoto, S.H., and M. Westerfield, in preparation). This correlation between Dorsalin-1 expression in the notochord and the absence of muscle pioneer cells in adjacent somites indicates that the differentiation of muscle pioneer cells can be blocked by a BMP-like signal, establishing a BMP-like molecule as a viable candidate for an inhibitory signal that prevents muscle pioneer differentiation in the dorsal and ventral regions of the somite.

Muscle pioneers are derived from a subset of slow muscle precursor cells, whereas most of the precursor cells develop into non–muscle pioneer slow muscle cells (Devoto et al., 1996b). To learn whether ectopic expression of Dorsalin-1 in the notochord inhibits the development of muscle pioneer cells specifically or whether non–muscle pioneer slow muscle cells are also affected, we injected embryos with twhh-dsl-1^myc DNA and labeled with the F59 antibody, which recognizes all of the slow muscle cells (Fig. 1; Devoto et al., 1996b). As shown by the bracket in Fig. 5,A, there was a gap in F59 labeling in the middle of some of the somites in embryos injected with twhh-dsl-1^myc. Transverse sections through unaffected regions (Fig. 5,B) and affected regions (Fig. 5,C) demonstrated that this gap in labeling is a result of the absence of the muscle pioneer population of slow muscle cells, which are normally located adjacent to the notochord (Fig. 5,B, arrows). In contrast, the dorsal and ventral populations of slow muscle cells (the non–muscle pioneer slow muscle cells; Fig. 5, B and C, arrowheads) are apparently unaffected by Dorsalin-1. These data demonstrate that notochord expression of Dorsalin-1 specifically interferes with the development of muscle pioneer cell identity and does not affect the development of the non–muscle pioneer slow muscle cells from adaxial cells.

These results demonstrate that Hedgehogs can induce slow muscle cells, including both muscle pioneers and non–muscle pioneer slow muscle cells, and that Dorsalin-1 can specifically inhibit the development of muscle pioneer cells. Dorsalin-1 could act by inhibiting the expression of hedgehog genes in the notochord, or by antagonizing Hedgehog protein activity. If Dorsalin-1 represses expression of hedgehog genes, then overexpression of Hedgehog should overcome the inhibitory effect of Dorsalin-1 on muscle pioneer formation. We tested this prediction by coinjecting Hedgehog RNAs with twhh-dsl-1^myc DNA and analyzing the injected embryos by double labeling with anti-myc antibody (labeling myc-tagged Dorsalin-1) and with the 4D9 antibody (labeling muscle pioneer cells). Compared to embryos injected with control RNA (Fig. 6,A), the expression of Dorsalin-1 in the notochord (Fig. 6,C, arrowhead) inhibited development of muscle pioneers in adjacent somites (Fig. 6,C, bracket), regardless of whether embryos were coinjected with RNA encoding Tiggy-winkle hedgehog (100%, n = 65; not shown) or Sonic hedgehog (100%, n = 43). In contrast, induction of muscle pioneers by Hedgehogs was unaffected in embryos coinjected with Sonic hedgehog RNA and the control DNA construct twhh-bGFP (Fig. 6 B). These data suggest that Dorsalin-1 blocks the differentiation of slow muscle precursor cells into muscle pioneer cells by antagonizing the activity of Hedgehogs rather than by simply inhibiting their expression.

As discussed above, Hedgehog signaling is mediated by inhibition of PKA activity (for review see Perrimon, 1995). PKA constitutively represses Hedgehog target genes, and inhibition of PKA by Hedgehog alleviates this repression. To test whether blocking PKA activity induces slow muscle cells, we injected RNA encoding a dominant negative mutant form of the PKA regulatory subunit (dnPKA; Ungar and Moon, 1996) or frame-shifted Sonic hedgehog as an injection control and then examined the somites by antibody labeling for muscle pioneer cells and for other slow muscle cells. dnPKA induced the development of extra slow muscle cells (94%, n = 102), including both the muscle pioneer (Fig. 7, E and H) and non–muscle pioneer slow muscle cells (Fig. 7,B). In contrast, injection of frame-shifted sonic hedgehog as a control had no effect on the development of slow muscle cells (Fig. 7, A, D, and G).

To learn whether Dorsalin-1 can antagonize the ability of dnPKA to induce both muscle pioneer cells and non– muscle pioneer slow muscle cells, we coinjected dnPKA RNA and twhh-dsl-1^myc DNA into zebrafish embryos and labeled serial sections or whole-mount embryos with the 4D9 antibody (to detect muscle pioneer cells) or with the F59 antibody (to detect all slow muscle cells). We found that Dorsalin-1 inhibited muscle pioneer induction by dnPKA (100%, n = 84; Fig. 7,F compared to E, and Fig. 7,I compared to H; arrowhead in I indicates Dorsalin-1–expressing cell). This result suggests that the inhibitory effect of Dorsalin-1 on muscle pioneers is downstream of PKA activity. In contrast, induction of non–muscle pioneer slow muscle cells by dnPKA was apparently unaffected by Dorsalin-1 (Fig. 7,C compared to B). In many dnPKA/twhh-dsl-1^myc coinjected embryos, the entire somite was transformed into slow muscle cells, even in regions where there were no muscle pioneers because of the action of Dorsalin-1. This is consistent with the results obtained in embryos injected with the twhh-dsl-1^myc DNA construct alone (Fig. 5). Dorsalin-1 apparently affected only the muscle pioneer population of slow muscle cells. These data confirm that the inhibitory effect of Dorsalin-1 on formation of slow muscle cells is specific for muscle pioneer cells. Together, these data suggest that Hedgehog signaling induces slow muscle cells and that BMP-like signaling is involved in the specification of distinct slow muscle cell identities.

We have investigated the mechanisms regulating the induction and differentiation of slow muscle fibers in zebrafish. Our results suggest that Hedgehog signals are involved in the initial induction of slow muscle precursor cells, whereas the subsequent differentiation of these precursors into distinct types of embryonic slow muscle cells may involve an inhibitory TGF-β signal. This proposed inhibitory signal antagonizes the Hedgehog activity in dorsal and ventral regions of the somite. Our data suggest that opposing actions of hedgehog and TGF-β gene family members may regulate the differentiation of specific slow muscle fiber cell types in the zebrafish somite.

We have shown that ectopic expression of members of the hedgehog gene family during early zebrafish development induces extra slow muscle cells, suggesting that Hedgehog signaling participates in the establishment of slow muscle cell identity. This is further supported by our observation that inhibition of PKA, likely to occur during Hedgehog signaling, mimics the activity of Hedgehog in slow muscle induction and that constitutive activation of PKA blocks the development of slow muscle cells. Several observations support the hypothesis that one or more Hedgehogs are the endogenous factors that induce the formation of slow muscle precursors during normal development. First, slow muscle precursors develop adjacent to the notochord, becoming apparent after notochord precursor cells begin to express hedgehog genes (Krauss et al., 1993; Roelink et al., 1994; Currie and Ingham, 1996; Devoto et al., 1996b). Second, all slow muscle precursors strongly express the patched gene, which is induced by Hedgehog signaling (Concordet et al., 1996), suggesting that they receive and respond to Hedgehog. Third, there is a loss of slow muscle cells in mutants in which Hedgehog signaling is reduced (Talbot et al., 1995; Devoto et al., 1996a; Weinberg et al., 1996). Together with the results reported here, these observations provide compelling evidence that Hedgehogs induce slow muscle cells.

We found that ectopic expression of either Sonic hedgehog or Tiggy-winkle hedgehog induced ectopic muscle pioneer cells. Our data differ from a previous report that ectopic expression of Sonic hedgehog was unable to induce muscle pioneers, unless another member of the Hedgehog family, Echidna hedgehog, was coexpressed (Currie and Ingham, 1996). The reason for this discrepancy is unclear, although the two studies used different plasmids that generate RNAs with different untranslated regions. It is possible that these differences affected the stability or translation of the RNA. Our results are consistent with those reported by Hammerschmidt et al. (1996a), who found that ectopic expression of either mouse Sonic hedgehog or Indian hedgehog induced extra muscle pioneer cells in zebrafish embryos.

We propose that early signaling by Hedgehogs is sufficient to trigger the development of slow muscle identity, but that muscle pioneer development requires additional later exposure to Hedgehogs (Fig. 8 A). This hypothesis is supported by the following observations. First, slow muscle precursors are distinct from the other presomitic cells before muscle pioneers become distinct from the other slow muscle precursors. Second, injection of Hedgehog RNA was consistently more effective at inducing non– muscle pioneer slow muscle cells than muscle pioneer cells. Third, hedgehog RNA injection induces muscle pioneers more effectively in anterior somites than in posterior somites of the embryo (data not shown). If the injected hedgehog RNA is degraded over time, then there would consistently be more ectopic hedgehog early in development (in anterior somites) than there would be later in development (in posterior somites). Finally, in several mutants (no tail, floating head), Hedgehog expression becomes progressively reduced, relative to wild-type embryos, as development proceeds. In these mutants, muscle pioneers are missing, whereas other slow muscle cells develop relatively normally, especially in the earlier developing anterior somites (Devoto et al., 1996a).

During normal zebrafish embryogenesis, slow muscle precursor cells that move dorsally and ventrally in the somite develop into non–muscle pioneer slow muscle cells. Although ectopic expression of Hedgehogs and dnPKA induces ectopic non–muscle pioneer slow muscle cells throughout the somite, neither induces ectopic muscle pioneers in the dorsal or ventral thirds of the somite. We showed that expression in the notochord of dorsalin-1, a BMP4-like gene normally expressed in the neural tube (Basler et al., 1993), can inhibit the development of muscle pioneer cells. This suggests that an inhibitory signal such as Dorsalin-1 might normally prevent the development of muscle pioneers in the dorsal and ventral portions of the somite (Fig. 8 B). In addition to inhibitory BMP-like signals originating from the neural tube, other BMP-like factors expressed ventral to the notochord (Rissi et al., 1995) and elsewhere (Liem et al., 1995; Pourquié et al., 1996) are candidates for antagonizing Hedgehog signaling.

We have shown that when it is expressed in the notochord, Dorsalin-1 has a specific effect on muscle pioneer identity and does not affect the differentiation of non– muscle pioneer slow muscle fibers. Several studies have suggested that BMPs have a limited range of diffusion, raising the question of whether non–muscle pioneer cells are also exposed to Dorsalin-1 protein when it is expressed in the notochord. We think that the specificity of Dorsalin-1 action on muscle pioneer cells but not on the non–muscle pioneer cells due to a difference in the exposure to Dorsalin-1 is unlikely for several reasons. First, Dorsalin-1 was expressed in the notochord just before the migration of adaxial cells away from the notochord, and thus all slow muscle precursors would be exposed to Dorsalin-1 (data not shown). Second, in the case of dominant negative PKA and Dorsalin-1 coinjection, non–muscle pioneer slow muscle cells were induced in the region adjacent to the notochord cells expressing Dorsalin-1, whereas muscle pioneer cells were inhibited in this region (compare Fig. 7, C with F), suggesting that the lack of effect on non–muscle pioneer cells is unlikely the result of a limited range of Dorsalin-1 action.

It is likely that slow muscle identity is established earlier than muscle pioneer cell identity. Slow muscle precursors are morphologically and molecularly distinct at the end of gastrulation (Devoto, et al., 1996b; Weinberg, et al., 1996), whereas muscle pioneers are not identifiably separate from the other slow muscle precursors until the time of somite formation, when they express engrailed genes and develop their distinctive morphology (Felsenfeld et al., 1991; Hatta et al., 1991; Ekker et al., 1992). Expression of dorsalin-1 from the twhh promoter begins before the appearance of muscle pioneer identity and before the migration of the slow muscle precursors away from the notochord (data not shown). Thus, all of the slow muscle cells are likely to be exposed to Dorsalin-1; this suggests that the development of non–muscle pioneer slow muscle precursors is unaffected by exposure to Dorsalin-1 at this time, perhaps because they are already committed to a slow muscle fate. We have not tested whether BMP-like signals acting earlier, during gastrulation, influence the development of non–muscle pioneer slow muscle cells.

In this study, we showed that the 5.2-kb twhh promoter could drive expression of heterologous cDNAs specifically in the notochord. This notochord specificity was unexpected considering that the endogenous twhh gene is exclusively expressed in the floor plate. It is possible that the 5.2-kb twhh promoter we isolated and used may lack a repressor sequence present in the twhh gene that inhibits the notochord expression of the twhh gene. Our results highlight the power of using tissue-specific promoters to direct expression of proteins to specific tissue types, at specific times, in the zebrafish embryo. This will be generally useful for analyzing later functions of genes that also have functions during gastrulation (see also Kroll and Amaya, 1996).

Based on our results and studies by other laboratories, we propose that the differentiation of slow muscle cells in zebrafish is regulated by at least two signals, Hedgehogs and BMP-like proteins (Fig. 8). During early stages of development, Hedgehogs secreted from midline cells induce paraxial mesodermal cells to become adaxial cells, the precursors of slow muscle. We propose that this early exposure to Hedgehogs is sufficient to signal the development of non–muscle pioneer slow muscle cells; however, it is insufficient for the development of muscle pioneer cells. Differentiation of muscle pioneers requires prolonged exposure to the inductive Hedgehog signals, and minimal exposure to an inhibitory BMP-like signal. In the dorsal and ventral regions of the somite, an inhibitory BMP4-like signal blocks the response to Hedgehog, whereas in the middle region of the somite, this inhibitory BMP-like activity is absent or very low and consequently restricts the development of muscle pioneer cells to the middle region of the somite. The mechanism for setting up this low BMP-like activity in the middle region of the somite is unknown. One possibility is an uneven distribution of the BMPs and BMP-like protein within the somite. The dorsal and ventral regions are exposed to a high concentration of the BMP-like protein, while the middle region is exposed to a low concentration of BMP-like protein. Evidence supporting this hypothesis comes from studies of mediolateral patterning of the chick somite. In chick embryos, the lateral part of the somite is exposed to a high concentration of BMP4 expressed in the lateral plate mesoderm (Pourquié et al., 1996). BMP4 from this source acts as a diffusible lateralizing signal to specify the hypaxial muscle lineage (Pourquié et al., 1996). In zebrafish, several BMPs and BMP-like proteins have been shown to be expressed in tissues near the somite during segmentation stages. For example, a BMP-like gene, radar, is specifically expressed in the dorsal neural tube and hypochord cells (Rissi et al., 1995), and BMP2 and BMP4 are expressed primarily in the mesenchyme of dorsal and ventral fins (Nikaido et al., 1997). Therefore, it is likely that there is a gradient in the distribution of BMP and BMP-like proteins within the somite, with higher concentrations in the dorsal and ventral regions of the somite and lower concentrations in the middle region of the somite. Alternatively, the BMP inhibitory activity might be reduced in the middle region of the somite (around the notochord) by an opposing signal from the notochord that blocks the BMP-like activity in this region. Chordin (Piccolo et al., 1996), Noggin (Zimmerman et al., 1996), and Follistatin (Hemmati-Brivaniou et al., 1994) can each bind to and inactivate BMPs and other TGF-β family members, and in Xenopus these genes are also expressed in the notochord (Smith and Harland, 1992; Hemmati-Brivanlou et al., 1994; Sasai et al., 1994). Thus, Chordin, Noggin, or Follistatin could repress the BMP-like activity in the notochord region, allowing the development of muscle pioneer cells. Further experiments are required to learn which mechanism is correct, and possibly both mechanisms are used to establish an uneven distribution of BMP-like inhibition of muscle pioneer development. This BMP4-like signal may also stimulate migration of adaxial cells toward the surface of the somite, where they are consequently exposed to a lower concentration of Hedgehogs. Regardless of the mechanism, this model predicts that these opposing signals determine slow muscle cell identities; adaxial cells that remain near the notochord express engrailed and develop into muscle pioneers, whereas the adaxial cells in the dorsal and ventral regions of the somite do not develop into muscle pioneers.

Our model is similar to that proposed for the dorsoventral patterning of the spinal cord (Liem et al., 1995; Ericson et al., 1996). Sonic hedgehog expressed by the notochord induces ventral cell types, such as floor plate and motorneurons, whereas BMP4 expressed in the dorsal neural tube induces dorsal cell types, such as neural crest, roof plate cells, and dorsal commissural neurons (Liem et al., 1995). The activity of Hedgehog and BMP4 are mutually antagonistic; Hedgehog inhibits the responses to BMP4, and BMP4 in turn inhibits the responses to Hedgehog (Liem et al., 1995). It has been suggested that patterning of the chick somite also involves the opposing actions of signals from surrounding tissues, including the neural tube and notochord (Fan and Tessier-Lavigne, 1994; Pourquié et al., 1996). These signals include Sonic hedgehog and BMP4 (Munsterberg et al., 1995; Pourquié et al., 1996). Our results suggest that in addition to regulating the broad subdivisions of the somite into sclerotome and dermamyotome, the opposing actions of hedgehog and TGF-β gene family members also regulate the development of embryonic muscle fiber-type identity.

We are indebted to C.B. Kimmel (University of Oregon, Eugene, OR), D. Raible, H. Roelink, and A. Ungar for their comments on the manuscript. We thank C.S. Goodman (University of California, Berkeley) for the monoclonal antibody, 4D9, F. Stockdale (Stanford University, Stanford, CA) for the F59 monoclonal antibody, and D. Turner and R. Rupp for the CS²⁺ vector and the cytomegalovirus-nβ-gal DNA construct, and A.P. McMahon (Harvard University, Cambridge, MA) for the pSP64T-PKA* construct. We thank Michele McDowell and Ruth BreMiller for help with histology, and the members of our laboratories for their helpful discussions.

This work was supported by Public Health Services Grants HD29360, HD22486, and NS21132. R.T. Moon is an investigator of the Howard Hughes Medical Institute.

β-gal

β-galactosidase

bGFP

bright green fluorescence protein

PKA

protein kinase A

twhh

tiggy-winkle hedgehog.