The Caenorhabditis elegans NK-2 class homeoprotein CEH-22 is involved in combinatorial activation of gene expression in pharyngeal muscle

Vertebrates contain three muscle types: skeletal muscle, cardiac muscle and smooth muscle. Although these muscle types express many of the same genes, they are morphologically distinct and arise from separate regions of the developing embryo. A group of helix-loop-helix transcription factors, collectively referred to as the MyoD family, has been implicated in skeletal muscle differentiation (for recent review see Emerson, 1993). None of the identified MyoD family members are detected in cardiac and smooth muscle. Therefore differentiation of these muscle types must involve either divergent basic-helix-loop-helix factors or members of other transcription factor families. Several candidates for cardiac muscle differentiation factors have been proposed, including the MEF-2 family of MADS box transcription factors, the homeodomain proteins MHox and Nkx-2.5/Csx and the zinc-finger protein HF-1b (Yu et al., 1992; Cserjesi et al., 1992; Lints et al., 1993; Komuro and Izumo, 1993; Zhu et al., 1993).

Like vertebrates, the muscles in the nematode Caenorhabditis elegans can be divided into distinct classes: body wall muscle, pharyngeal muscle and several groups of minor muscles (for review see White, 1988). In structure and function, body wall muscle may be most analogous to vertebrate skeletal muscle. Screens for MyoD homologs in C.elegans have identified a single family member, designated hlh-1, expressed specifically in body wall muscles and their clonal precursors (Krause et al., 1990). No hlh-1 expression is seen in pharyngeal muscle or the minor muscles, suggesting that differentiation of these muscle types may be analogous to that of vertebrate cardiac or smooth muscle.

We have examined the regulation of myosin heavy chain gene expression as an initial step to analyze myogenesis in C. elegans. Two myosin heavy chain genes, myo-1 and myo-2, are specifically expressed in pharyngeal muscle (Miller et al., 1986; Ardizzi and Epstein, 1987). Regulatory regions controlling the expression of myo-2 have been characterized in most detail (Okkema et al., 1993). The myo-2 gene contains at least two independent tissue-specific regulatory elements: a promoter sufficient for low level pharyngeal muscle-specific expression is located near the transcriptional start site, and a separable pharyngeal muscle-specific enhancer is located approximately 300 bp upstream of the start site. Neither the promoter nor the enhancer contains consensus binding sites associated with vertebrate muscle-specific genes.

In this paper, we describe the modular structure of the myo-2 enhancer. At least three subelements contribute to enhancer activity. Two of these display distinct cell type specificities: one is active in a subset of pharyngeal muscles; the second is more generally active in all cell types in the pharynx. In a screen for factors regulating the myo-2 enhancer, we identified an NK-2 class homeodomain factor that appears to play a key role in pharyngeal muscle-specificity of the myo-2 enhancer.

Standard methods were used to manipulate plasmids DNAs, oligonucleotides and RNA (Ausubel et al., 1990). The parental promoter::lacZ fusions used to assay enhancer activity are described in Okkema et al. (1993): pPD26.02 (myo-3::lacZ); pPD26.50 (glp-1::lacZ); pOK1134 (Δmyo-2::lacZ; this construct is similar to pOK5.56 with a StyI restriction site inserted upstream of the myo-2 sequences to facilitate cloning of oligonucleotides). Oligonucleotides were designed with 5′, non-palindromic overhangs to allow ligation as head-to-tail concatenates and ligation into the StyI site of pOK5.56. Inserts of plasmids containing oligonucleotides from the B and C fragments were sequenced to determine concatamer number and confirm the integrity of the cloned oligonucleotides. Sequences of plasmid constructs are available from the authors.

When hybridized to an ordered array of YACs spanning the C. elegans genome (Coulson et al., 1991), the ceh-22 cDNA identified four overlapping YACs (Y24E3, Y59H10, Y61D5, Y33F10) located between her-1 and act-1,2,3 on chromosome V. ceh-22 genomic DNA was subcloned from cosmid WB2 (kindly provided by T. Bürglin). To construct the ceh-22::lacZ fusion pOK29.02, a 4 kb fragment con-taining ceh-22 5′-flanking sequences and part of the 5′-UTR was subcloned into the lacZ expression vector pPD22.04 (Fire et al., 1990).

C. elegans strain Bristol (N2) was grown under standard conditions (Sulston and Hodgkin, 1988). F₁ expression assays were done as previously described (Okkema et al., 1993). Plasmid DNAs were injected at 100 μg/ml into the germ line of adult hermaphrodites (Mello et al., 1991) and F₁ progeny stained for β-gal activity as larvae and adults (Fire, 1993). We have also tested activity of the B+B and C+C enhancers in transgenic lines (Mello et al., 1991). As has been observed with other enhancer assay constructs, the B+B and C+C enhancer constructs function poorly in high copy extrachromosomal arrays in transformed lines. We do not know whether the lack of observed enhancer function in these lines is a result of copy number, the co-selected marker gene, or the structure of the transforming DNA. Stable lines expressing ceh-22::lacZ were obtained by coinjecting pOK29.02 with pRF4 (Mello et al., 1991); the resulting extrachromosomal array was integrated into a chromosome by X-ray irradiation (Krause et al., 1990).

cDNA libraries in λgt11 were constructed with oligo(dT)-primed cDNA synthesized from poly(A)+ RNA isolated from either embryos or mixed stage animals. cDNAs encoding candidate enhancer binding factors were isolated from the embryo cDNA library by screening with concatenated, double-stranded oligonucleotides (Vinson et al., 1988, Singh et al., 1988) using conditions optimized by V. Jantsch-Plunger (Jantsch-Plunger, 1993). Phage were grown 3-4 hours on E. coli Y1090 at 42°C. The plate was overlayed with a nitrocellulose filter soaked in 100 mM IPTG and transfered to 37°C for 6 hours. The filter was removed and a second IPTG soaked filter placed on the plate for 10-12 hours at 37°C. Filter lifts were processed by modification of the procedure described by Vinson et al. (1988). Filters were sequentially agitated in prebinding buffer [10 mM Hepes (pH7.9), 10 mM MgCl₂, 50 mM KCl, 1 mM DTT] containing 6 M, 4 M, 2.7 M, B1.3 M, 0.67 M, 0.33 M and 0.17 M guanidine hydrochloride, followed by two washes in prebinding buffer (10 minutes each, 4°C). Filters were blocked in prebinding buffer containing 5% non-fat dry milk (Carnation) (45 minutes, 4°C) and probed in binding buffer [10 mM Hepes (pH7.9), 10 mM MgCl₂, 100 mM KCl, 1mM DTT] containing 2% non-fat dry milk, 1.5 μg/ml double-stranded salmon sperm DNA, 3 μg/ml denatured salmon sperm DNA and 10⁵ cts/minute/ml nicktranslated concatenates of B207 and C183 (overnight, 4°C). Filters were washed twice in binding buffer containing 2% non-fat dry milk and twice in TBS (10 minutes, 4°C).

Two different recombinant CEH-22 fusion proteins were expressed in E. coli BL21. A glutathione-S-transferase::CEH-22 fusion protein (GST::CEH-22) encoding CEH-22 amino acids 1-346 was purified according to Smith and Johnson (1988). A phage T7 gene 10::CEH-22 fusion protein containing a polyhistidine tract (poly-his::CEH-22) encoding CEH-22 amino acids 79-346 and a derivative of polyhis::CEH-22, deleted for CEH-22 amino acids 217-264, were purified by affinity chromatography to immobilized Ni²⁺ under denaturing conditions (Hochuli et al., 1990).

DNAse protections assays were performed by modification of a protocol described by Ausubel et al. (1990). Poly-his::CEH-22 was bound to an end-labeled B fragment probe in 10 mM Hepes (pH 7.9), 10 mM MgCl₂, 100 mM KCl, 1 mM DTT, 1.5 μg/ml double-stranded salmon sperm DNA, 3 μg/ml denatured salmon sperm DNA and 100 μg/ml BSA.

Rabbit polyclonal antibodies were raised separately against GST::CEH-22 (antiserum c184) and poly-his::CEH-22 (antiserum c187). Antibodies that specifically recognize CEH-22 were affinity purified (Harlow and Lane, 1988; N. Patel, personal communication) from c184 by binding to immobilized poly-his::CEH-22 protein and from c187 by binding immobilized GST::CEH-22. Affinity-purified anti-CEH-22 antibodies were used undiluted.

Embryos [isolated by hypochlorite digestion (Sulston and Hodgkin, 1988)] were fixed 30 minutes (in PBS containing 55 mM Pipes (pH 6.95), 1.1 mM EGTA, 0.5 mM MgSO₄, 2.3% formaldehyde) under a coverglass on a microscope slide coated with 3% polylysine. The slide was then frozen on an aluminum block cooled in dry ice, separated from the coverglass, dipped for 4 seconds in −20°C methanol and rinsed 3× 5 minutes in TTBS (TBS, 0.1% Tween 20). The embryos were incubated overnight at 4°C under a drop of primary antibody, washed 4× 20 minutes in TTBS, incubated 4 hours at 25°C with a secondary antibody, washed 4× 20 minutes in TTBS and mounted in 70% glycerol, 1 mg/ml phenylenediamine, 0.02% NaN₃ for microscopy.

In an initial analysis of cis-acting sequences regulating myosin gene expression, we identified a 395 bp fragment from myo-2 that functions as a pharyngeal muscle-specific enhancer (Okkema et al., 1993). This enhancer was defined by its ability to induce pharyngeal muscle expression from a myo-3::lacZ fusion which is normally expressed only in body wall muscle (Fig. 1A). For these experiments, we used an F₁ expression assay to characterize enhancer function (see Okkema et al., 1993 for discussion). Using this assay to delimit the enhancer further, we identified two overlapping fragments that also function as strong pharyngeal enhancers (Fig. 1B, pOK4.50 and pOK3.16). When assayed alone, the region of overlap functions only very weakly (pOK6.22). This analysis suggests the myo-2 enhancer contains several functional components that cooperate to activate transcription.

Based on the deletion analysis, we provisionally divided the myo-2 enhancer into three overlapping fragments, designated A, B and C (Fig. 1B). These fragments are individually inactive, but segments spanning A+B or B+C function as strong pharyngeal enhancers. A plausible working model is that the A, B and C fragments each contain a discrete subelement, with two or more subelements necessary for activity. Consistent with this model, we found that duplications of A, B or C function as pharyngeal enhancers (Fig. 1C). Thus, each of these fragments contains sufficient information to indepen-dently activate transcription when duplicated. A combination of the C and A fragments also functions as a pharyngeal enhancer (Fig. 1C), indicating that the three fragments are mutually synergistic.

To test promoter requirements for the individual subelements, we assayed these duplications and combinations of fragments upstream of two additional promoters (data not shown). The glp-1 promoter fragment used shows a background of rare sporadic staining with no bias towards muscle, while a deleted myo-2 promoter (Δmyo-2) shows essentially background activity (Okkema et al., 1993). Both of these promoter segments are sensitive to transcriptional enhancement in a variety of tissues (Okkema et al., 1993; P. O., V. Jantsch-Plunger and A. F., unpublished data). Both the B and C fragments function identically with all promoters tested: a single copy of either displays little or no enhancer activity; while two copies induce abundant pharyngeal expression. The B+C enhancer likewise activates the glp-1 and Δmyo-2 promoters. In contrast, A+A, C+A and A+B are unable to enhance expression from either the glp-1 or Δmyo-2 promoters, although each is able to activate myo-3. Therefore, function of the A fragment with the myo-3 promoter appears to require specific enhancer-promoter interactions.

The pharyngeal muscles of C. elegans can be grouped into 8 types (m1-m8) based on cell morphology and position (described in detail in Albertson and Thomson, 1976). The muscles are arranged in layers along the anterior-posterior axis that are three-fold rotationally symmetric, each containing 1-3 cells of a single type (Fig. 2A). The m3-m7 muscles are large and define the overall contour of the pharynx. Smaller m1, m2 and m8 muscles are located at the anterior or posterior ends of the pharynx. The pharynx also contains epithelial cells, neurons, specialized marginal cells and secretory gland cells.

The cell type specificity of various enhancer constructs was analyzed by identification of individual cells expressing enhancer driven lacZ fusions. The F₁ expression assay used for this analysis generates animals that exhibit mosaic expression of the transforming DNA. We have used the frequency of staining as a measure of enhancer strength in each cell type (e.g., Weintraub, 1988). A myo-2::lacZ fusion containing the entire enhancer and promoter region is expressed in all pharyngeal muscle cell types (pPD20.97, Figs 2B, 3A; Okkema et al., 1993). Expression of this fusion is most frequent in muscle cells m3-m7, with less frequent expression in m1, m2 and m8. The A+B+C enhancer gives a similar expression pattern when assayed with Δmyo-2 or glp-1 promoters, although the relative frequency of expression is somewhat reduced in m1, m2, m6 and m8 (Figs 2B, 3B).

The B and C fragments activate transcription in distinct sets of cells in the pharynx. The B+B enhancer (pOK17.13) activates frequent expression only in pharyngeal muscles m3, m4, m5 and m7, with occasional expression in m1 (Figs 2B,3C). No expression has been observed in m6 or m2. In contrast, the C+C enhancer activates frequent expression in all pharyngeal muscles (Figs 2B, 3D). This distinction between the C and B fragments is particularly apparent in m1, m2 and m6, in which C+C is very active, while B+B is almost completely inactive. Unlike the other enhancers assayed, the C+C enhancer also activates expression in non-muscle cells in the pharynx. Fig. 3D shows an animal expressing β-gal in the e1 and e2 epithelial cells, as well as muscles m1, m2 and m4. We have observed expression induced by the C+C enhancer in all pharyngeal cell types including gland cells, neurons and marginal cells as well as muscles. These results define C as an organ-specific subelement.

Given the differences between the expression patterns induced by the B+B and C+C enhancers, it was of interest to determine the expression pattern of the B+C enhancer. This pattern is indicative of a restrictive rather than an additive interaction. The B+C (pOK18.51) enhancer is active only in cells where both B and C can be active (m3, m4, m5 and m7; Figs 2B, 3E).

To map precisely the subelements within the B and C fragments, we synthesized a set of overlapping doublestranded oligonucleotides spanning B and C (Fig. 4). The oligonucleotides were individually ligated to form head-to-tail multimers that were assayed for enhancer function upstream of the Δmyo-2::lacZ fusion (Table 1).

We might expect to find both general and cell-type-specific elements using this assay. Indeed, three of the oligonucleotides appear to contain general transcriptional activator elements. B201, B203 and B205 activate expression at low frequency in a variety of tissues and cell types (Table 1). Given our goal of understanding cell-type-specific regulation of myo-2, we have not investigated these general activities further.

Multimers of a single oligonucleotide from the B fragment (B207) activate pharyngeal expression in a pattern very similar to that observed with the duplicated B fragment, with expression predominately in pharyngeal muscles m3, m4, m5 and m7 (Table 1, Figs 2B, 3F). Unlike the larger B fragment, the B207 oligonucleotide occasionally activates additional expression in cells other than pharyngeal muscle (Table 1).

Multimers of either of two overlapping oligonucleotides from the C fragment exhibit enhancer activity identical to that of the duplicated C fragment. C181 and C183 induce frequent expression in both muscle and non-muscle cells in the pharynx (Table 1; Figs 2B, 3G). Like the intact C fragment, these oligonucleotides activate expression only in the pharynx.

To identify regions in the B and C oligonucleotides necessary for transcriptional activation, we assayed a set of mutated oligonucleotides for enhancer activity (Fig. 5; Table 1). Mutations near each end of the B oligonucleotide B207 (Bmut1, Bmut2 and Bmut4) and each end of the C oligonucleotide C183 (Cmut1, Cmut2 and Cmut4) eliminate activity in the pharynx. In contrast, internal mutations in B207 and C183 (Bmut3 and Cmut3, respectively) have little effect, suggesting that the activities of both the B and C subelements might require multiple binding sites. These analyses also suggest that the left end of B plays a critical role in specificity, since a mutation in this region (Bmut1) eliminates pharyngeal muscle expression without affecting the occasional non-pharyngeal expression (Table 1). Mutations to the right (Bmut2 and Bmut4) drastically reduce both pharyngeal and non-pharyngeal activity.

We used the B207 and C183 oligonucleotides to identify candidate genes that regulate myo-2 enhancer activity. Phage plaques from a λgt11 cDNA library expressing C. elegans proteins were immobilized on nitrocellulose and probed with multimers of B207 and C183 (Singh et al., 1988; Vinson et al., 1988). Screening approximately 4×10⁵ recombinant clones, we isolated three related cDNAs whose products specifically bind B207. Complete sequencing of two of these clones and restriction analysis of the third indicates that they are co-linear cDNAs encoded by a single gene that we have named ceh-22 (ceh =C. eleganshomeobox; see below). In addition to the three ceh-22 clones, two unrelated clones with binding specificities distinct from that of ceh-22 were isolated. Analysis of the latter cDNAs is in progress and will be described elsewhere.

The predicted CEH-22 protein contains a homeodomain DNA-binding motif belonging to the phylogenetically conserved NK-2 family (Fig. 6A). The CEH-22 homeodomain contains 8 of 9 conserved residues that define the NK-2 family and is most similar (87% identical) to the homeodomain of the Drosophila NK-2 gene (Kim and Nirenberg, 1990; Nardelli-Haefliger and Shankland, 1993). The CEH-22 homeodomain is 60% identical to those of tin/NK-4 and bag/NK-3 (Kim and Nirenberg, 1990; Bodmer et al., 1990; Azpiazu and Frasch, 1993). By contrast, CEH-22 shares only 27-50% identity with published C. elegans homeodomain sequences (data not shown). CEH-22 differs from other NK-2 class proteins in containing a serine rather than a conserved glutamine at homeodomain position 22 and an alanine rather than a conserved histidine at position 33. These amino acids are located in helices 1 and 2 of the homeodomain, respectively, and are not predicted to contact DNA (Kissinger et al., 1990).

Outside the homeodomain, CEH-22 shares no significant identity with any of the NK-2 family members. In particular, CEH-22 lacks both a conserved 17 amino acid peptide found downstream of the homeodomain and a decapeptide found upstream of the homeodomain in several family members (Price et al., 1992; Azpiazu and Frasch, 1993; Lints et al., 1993; Saha et al., 1993). An acidic region is located just upstream of the CEH-22 homeodomain (Fig. 6B). Highly acidic regions are also found upstream of the NK-1, NK-2 and Dth-1 homeodomains (Kim and Nirenberg, 1990; GarciaFernàndez et al., 1991).

Using reagents provided by the C. elegans genome project (Coulson et al., 1991), we mapped ceh-22 to chromosome V, between her-1 and the actin gene cluster. In a hybridization screen for homeoboxes, Bürglin and co-workers identified a hybridization signal in this region (locus 29; Bürglin et al., 1989). Further analysis of a cosmid from the region indicates the ceh-22 homeobox indeed corresponds to locus 29 (data not shown; T. Bürglin, personal communication). A genomic fragment from this cosmid was sequenced, revealing that the ceh-22 cDNA is derived from 7 exons spanning 3.2 kb (Fig. 6B). The homeobox is split by an intron within codon 53. Although this intron is absent in the NK-2 class homeoboxes for which genomic DNA sequence is available [NK-1,-2, bag/NK-3, tin/NK-4, Dth-1 and Dth-2 (Kim and Nirenberg, 1989; Garcia-Fernàndez et al., 1993)], an intron in this position is present in at least one non-NK-2 class homeobox in C. elegans (ceh-19; Bürglin, 1993).

ceh-22 encodes two RNAs that are most abundant in embryos and present in decreasing amounts throughout development (Fig. 7). The relative abundance of the two ceh-22 RNAs is modulated during development: the 1.5 kb RNA is present at higher levels than the 1.45 kb RNA in embryos; while the two are present at roughly equal levels in late larvae. Structural or functional differences between the two RNAs have not yet been determined. A low level of ceh-22 RNA is detected in adults. At least a fraction of this adult expression is in the soma (data not shown), since the RNA is detected in mutant glp-4(bn2) animals that contain very few germ cells (Beanan and Strome, 1992).

To define the CEH-22-binding site, we used recombinant CEH-22 protein purified from E. coli to generate a DNAseI footprint on the B fragment of the myo-2 enhancer (Fig. 8). At high concentrations, recombinant CEH-22 protects the sequence TAAAGTGGTTGTGTG, which overlaps the 5′ end of B207 by 13 bp. This sequence contains a TNNAGTG which is present in consensus binding sites for the NK-2 class homeoproteins TTF-1 and NK-2 (Guazzi et al., 1990; M. Nirenberg, personal communication). The mutation Bmut1, which affects this consensus sequence, eliminates B subelement activity in vivo (Fig. 5). An identically prepared protein containing a deletion removing helix 3 of the homeodomain fails to footprint the B fragment (data not shown).

Immunostaining was used to determine the temporal and spatial expression pattern of CEH-22. Antibodies were raised separately against two CEH-22 fusion proteins purified from E. coli by different protocols and affinity purified (see Methods). Identical staining patterns were observed with both antisera. Staining is limited to nuclei within the pharynx and is detected from the beginning of morphogenesis onwards (Fig. 9A-C). When CEH-22 is first detected (approximately 330 minutes after fertilization; the lima bean stage), all 37 pharyngeal muscle nuclei are present (Sulston et al., 1983). At this stage, CEH-22 is detected in 11-14 pharyngeal nuclei (Fig. 9A). Positive identification of the CEH-22 containing nuclei is difficult; however, their positions suggest they are pharyngeal muscles. As the embryo elongates to the 1G-fold stage, the number of CEH-22-positive nuclei in the pharynx increases to 14-23 (Fig. 9B). The wide range of staining nuclei suggests that, within a relatively short time period, a number of nuclei begin to accumulate CEH-22. CEH-22-positive cells were identified in these animals as muscles m3, m4, m5 and m7 by double staining with the monoclonal antibody 3NB12 (data not shown); 3NB12 had previously been shown to recognize a surface antigen in this set of pharyngeal muscles (Priess and Thompson, 1987). In embryos that have completed elongation (the pretzel stage), CEH-22-positive nuclei can be recognized by their characteristic positions as m3, m4, m5 and m7 (Fig. 9C). CEH-22 is also detected in 6 additional pharyngeal nuclei, which we believe are the m1 muscles. After hatching, CEH-22 remains detectable in m1, m3, m4, m5 and m7, but remains absent in m6 and m2 (data not shown).

We have also examined expression of a ceh-22::lacZ fusion in transgenic C. elegans. This construct contains approximately 4 kb of ceh-22 5′-flanking DNA fused to lacZ within the ceh-22 5′-UTR (see Fig. 6B). The timing and distribution of β-galactosidase expression in transgenic animals containing this fusion is identical to the staining pattern observed using anti-CEH-22 antibodies (Fig. 9D-F).

We have identified three regions of the myo-2 enhancer, which we call A, B and C, that function together to activate transcription. Within the B and C regions, we have identified short oligonucleotides with activities virtually identical to the intact fragments. These results support the model that discrete subelements are combined to form the myo-2 enhancer.

The B and C subelements appear to contain multiple sites necessary for activity. Mutations at each end of these subele-ments eliminate activity, whereas mutations nearer the center have no effect. Thus two levels of organization apparently exist in the myo-2 enhancer: subelements separated by approximately 100-200 bp are each composed of a cluster of binding sites for interacting transcription factors.

Similar modular organization is a common feature of transcriptional enhancers. The structure of the SV40 enhancer has been analyzed in detail (reviewed in Wildeman, 1988). It is composed of several elements of differing specificity that cooperate to activate transcription (Herr and Clarke, 1986; Ondek et al., 1987). The elements each in turn contain multiple sites necessary for activity (Ondek et al., 1988). This type of organization has also been demonstrated for a vertebrate muscle-specific enhancer from the creatine kinase gene (Cserjesi et al., 1992).

The B and C subelements of the myo-2 enhancer exhibit distinct cell type specificities. The B subelement is primarily active in the pharyngeal muscles m3, m4, m5 and m7. The C subelement is active in all pharyngeal cells. Thus distinct muscle-type-specific and organ-specific pathways converge to activate the intact enhancer.

What benefits might be realized by constructing the myo-2 enhancer from muscle-specific and organ-specific subelements? The B+C combination is active in the same cells as the duplicated B subelement. Thus, cooperation with C does not refine cell type specificity of B. Perhaps the combination of the B and C subelements coordinates timing of pharyngeal muscle differentiation with that of other cell types in the pharynx. Proper formation of the pharynx might require differentiation of many cell types to be synchronized. The C subelement could be a target of a factor that synchronizes differentiation of the entire organ.

Independent evidence exists for a pathway specifying organ identity in the pharynx: the genes pha-1 (Schnabel and Schnabel, 1990) and pha-4 (S. Mango, E. Lambie and J. Kimble, personal communication) are specifically required for differentiation of many cell types in the pharynx. These genes might participate in pharyngeal differentiation by directly or indirectly activating regulatory sequences similar to the myo2 C subelement.

We have identified a new homeobox-containing gene, ceh-22, which appears to be a key factor in activating myo-2 expression through the B subelement. This conclusion is based upon both in vitro DNA-binding specificity and the in vivo expression pattern of CEH-22.

The ceh-22 gene was initially identified by screening a mixed tissue cDNA expression library for clones encoding proteins that specifically bind the B subelement. No attempt was made to enrich the library for pharyngeal muscle cDNAs. After screening approximately 4×10⁵ recombinant phage, we identified three clones encoding CEH-22 (of which at least two are independent). Recombinant CEH-22 protein footprints sequences necessary for in vivo activity of the B subelement. Moreover, the mutation Bmut1 in this region, which eliminates in vivo activity of B, also decreases CEH-22 binding (unpublished observations). Thus a functional CEH-22-binding site appears necessary for transcriptional activity of the B subelement.

We have examined the ceh-22 expression pattern using antibodies against CEH-22 protein and examining expression of a ceh-22::lacZ fusion. Both analyses indicate ceh-22 is expressed in pharyngeal muscle prior to the onset of myo-2 expression. To date, ceh-22 expression is the earliest known marker of pharyngeal muscle differentiation. Strikingly, ceh-22 expression is limited to the same subset of pharyngeal muscle cells in which the B subelement is active. This corre-lation is most apparent in the major muscles m3-m7: both ceh-22 expression and activity of the B subelement are limited to m3, m4, m5 and m7; neither ceh-22 expression nor frequent B subelement activity has been observed in m6. We expect that expression of other genes might be specifically activated by CEH-22 in m3, m4, m5 and m7; one candidate for this regulation is the gene encoding the antigen recognized by the monoclonal antibody 3NB12, which is expressed in this subset of pharyngeal muscles (Priess and Thompson, 1987).

Potential CEH-22-binding sites are present at additional locations in myo-2 and in the other pharyngeal myosin gene, myo-1. In myo-2, one site (located at position 1504) is within a 33 bp segment necessary for activity of the pharyngeal muscle-specific promoter (Okkema et al., 1993). In myo-1, a site (located at position 3742) is within a 138 bp segment necessary for activity of the pharyngeal muscle-specific enhancer (Okkema et al., 1993). We have not yet tested whether these sites can bind CEH-22.

The fact that B subelement activity correlates very well with ceh-22 expression suggests that cell type specificity of this subelement is largely determined by the expression pattern of ceh-22. While CEH-22 binding appears necessary for B subelement transcriptional activity, it is not sufficient. Three oligonucleotides (B205, Bmut2 and Bmut4), which contain the CEH-22-binding site and bind CEH-22 protein in vitro, fail to activate transcription in vivo (unpublished observations; Fig. 5). These observations suggest a second factor binding to the B subelement is essential for activity. We have not yet identified that factor but we are currently screening for proteins that bind this second critical site.

Previous workers have reported morphological and immunological differences between pharyngeal muscle m6 and other pharyngeal muscles. Albertson and Thompson (1976) observed vesicles in m6 that were absent in other pharyngeal muscle types. They suggested m6 may be specialized to secrete the pharyngeal grinder, a cuticular structure associated with this cell type. Priess and Thompson (1987) noted the monoclonal antibody 3NB12 recognizes a surface antigen in all the large pharyngeal muscle cells except m6.

Our results indicate the m6 cells also activate myo-2 expression differently than other pharyngeal muscles, via a CEH-22-independent mechanism. How then is myo-2 expression induced in m6? The complete myo-2 enhancer can activate expression in all large pharyngeal muscle cells including m6, whereas an enhancer composed of the B and C subelements activates transcription only in m3, m4, m5 and m7. Thus sequences in the A subelement may contribute to expression in m6, perhaps in combination with the C subelement.

Comparison with known homeodomain sequences identifies CEH-22 as a member of NK-2 family of homeodomain factors (Kim and Nirenberg, 1990; Nardelli-Haefliger and Shankland, 1993). Members of this family are expressed in a variety of tissues, suggesting they play diverse roles in development. The NK-2 class homeodomains most similar to that of CEH-22 (77-87% identical) are preferentially expressed in the central nervous system (CNS). The Drosophila NK-2, mouse Nkx-2.2 and TTF-1, Xenopus XlNK-2 and leech Lox10 genes are all expressed in the developing CNS (M. Nirenberg, personal communication; Price et al., 1992; Saha et al., 1993; Nardelli-Haefliger and Shankland, 1993). In addition to the nervous system expression, Lox10 and NK-2 are expressed in midgut (Nardelli-Haefliger and Shankland, 1993; M. Nirenberg, personal communication), while TTF-1 is also expressed in thyroid and lung (Guazzi et al., 1990). The planarian Dth-1 and Dth-2 genes, whose homeodomains are approximately 75% identical to that of CEH-22, are expressed in intestine and unidentified peripheral parenchymal cells, respectively (Garcia-Fernàndez et al., 1993).

In contrast, three homeodomains that are somewhat less similar to that of CEH-22 (60-67% identical) are expressed in muscle tissues similarly to CEH-22. The mouse Nkx-2.5/Csx gene is expressed in cardiac muscle progenitors prior to myogenic differentiation and continues throughout development (Lints et al., 1993; Komuro and Izumo, 1993). Nkx-2.5/Csx expression is also detected in a subset of pharyngeal endoderm adjacent to the cardiac mesoderm, tongue muscle, visceral muscle in the stomach and spleen. The Drosophila gene tinman (tin) is initially expressed throughout the presumptive mesoderm and becomes restricted to cardiac and visceral muscle (Bodmer et al., 1990; Azpiazu and Frasch, 1993). Loss of tin activity results in absence of cardiac and midgut visceral muscle, and defects in a subset of dorsal body wall muscles (Azpiazu and Frasch, 1993; Bodmer, 1993). Likewise, the Drosophila gene bagpipe (bag) is expressed in a segmented pattern in visceral muscle and in a subset of cardiac muscles (Azpiazu and Frasch, 1993). Loss of bag activity results in segmental gaps in midgut visceral muscle (Azpiazu and Frasch, 1993).

The similarity in expression patterns between ceh-22, Nkx-2.5/Csx, tin and bag suggests that the function of these genes may be conserved. Interestingly, the developmental programs in vertebrate cardiac muscle, Drosophila cardiac and visceral muscle, and C. elegans pharyngeal muscle all occur without myogenic factors related to MyoD (Emerson, 1993; Michelson et al., 1990; Krause et al., 1990). An attractive hypothesis from the analysis of ceh-22 and homologs in other species is that these homeodomain factors function in a phylogenetically conserved myogenic pathway occurring in muscle types that do not utilize the MyoD family.

We are indebted to V. Jantsch-Plunger for advice throughout this work, and to J. Ahnn, L. Avery, T. Bürglin, L. Chen, S. Dymecki, R. Jehan, W. Kelly, N. Patel, A. Pinder, G. Seydoux and C. Thompson for their help and suggestions. Sequence comparisons were done at the NCBI using the BLAST network service. This work was supported by the NIH (HD07532-03 to P. G. O.; R01-GM37706 to A. F.), the Carnegie Institution of Washington. A. F. is a Rita Allen Foundation scholar.

The ceh-22 cDNA and genomic DNA sequences have been submitted to GenBank (accession numbers, U10080 and U10081, respectively).