Combinatorial control of touch receptor neuron expression in Caenorhabditis elegans

Combinatorial specification has often been proposed as a means of generating the large variety of cell types seen in animals and plants, but the combinatorial factors specifying cell fate are largely unknown (e.g., Gierer, 1974; Yamamoto, 1985; Johnson and McKnight, 1989; Simmons et al., 1990; Benfey et al., 1990). Genetic experiments on the specification of nerve cell fate in the nematode Caenorhabditis elegans also support the view of development as a combinatorial process. In particular, studies of two cell types, a set of six sensory receptors for gentle touch (Chalfie and Au, 1989; Way and Chalfie, 1989) and a pair of motor neurons needed for egg-laying (Desai et al., 1988), suggest that genes regulating cell fate are not expressed in a cell-specific fashion. The question that we address here is how cell fate, in this case the ability to become a C. elegans touch receptor neuron, is restricted to specific cells, i.e. what combinatorial factors result in cell-specific differentiation and how is the final number of these cells determined?

The six touch receptors (ALML, ALMR, PLML, PLMR, AVM and PVM; Fig. 1) can be distinguished from other C. elegans cells by several identifying features, including their position, structure and possession of large-diameter (15-protofilament) microtubules and an associated extracellular material called the mantle (Chalfie and Sulston, 1981; Chalfie and Thomson, 1982). Screens for touch-insensitive mutants (Chalfie and Sulston, 1981; Chalfie and Au, 1989) identified several genes that are needed for touch cell function. These include mec-7, a gene encoding a β-tubulin that is required for the production of the 15-protofilament microtubules (Savage et al., 1989), and mec-4, a gene encoding a putative membrane protein that can be mutated to cause the specific death of the touch cells (Driscoll and Chalfie, 1991). As we show below, these genes are expressed predominantly in the six touch receptor neurons and, thus, are excellent markers for touch cell differentiation.

Previous genetic studies also identified three genes that were needed for the production of the touch receptors. Mutations in two genes, lin-32 and unc-86, result in abnormal precursor cells such that touch receptors are not generated (Chalfie et al., 1981; Sulston et al., 1983; Chalfie and Au, 1989; E. Hedgecock and C. Kenyon, personal communication). The third gene, mec-3, is not needed for the production of precursors but for the differentiation of the cells as touch receptors; the cells develop as neurons, but these cells do not have touch receptor features (Chalfie and Sulston, 1981; Way and Chalfie, 1988; Chalfie and Au, 1989). All three genes are needed for the development of additional cells (Chalfie et al., 1981; Chalfie and Au, 1989; Way and Chalfie, 1989; E. Hedgecock and C. Kenyon, pers. comm.). In particular, mec-3, the most specific of the genes, is expressed and required in two other pairs of neurons (the FLP and PVD cells; Way and Chalfie, 1989; J. Kaplan and H.R. Horvitz, personal communication; see below). unc-86 is also expressed in these cells (and others) in wild-type animals (Finney and Ruvkun, 1990) and is needed for their development (Hamelin et al., 1992; Xue et al., 1992).

Multiple alleles have been identified for almost all of the genes that mutate to a touch-insensitive phenotype (Chalfie and Au, 1989). Thus, it is unlikely that any nonredundant genes specifically controlling touch cell development will be found. Genes having more general patterns of expression, however, could help specify touch cell fate. Previously, we had speculated that lin-14, based on its mutant phenotype, might be one of these genes (Way and Chalfie, 1989). In this paper we show that at least five genes (lin-4, lin-14, egl-44, egl-46 and sem-4) are needed in addition to lin-32, unc-86 and mec-3 to direct the correct cellular expression of touch cell characteristics. Together these genes provide a combinatorial basis for the specification of this single cell type. The number of cells expressing the touch receptor fate is further restricted by programmed cell death.

Wild-type C. elegans (var. Bristol, N2) and mutant stocks were grown at 25°C as before (Brenner, 1974; Way and Chalfie, 1988). Animals with temperature-sensitive phenotypes were grown at both 25°C and 15°C for at least two generations before testing.

Strains with the following mutations were used:

L.G. I: lin-17(n671), lin-44(n1792), lin-35(n745), sem-4(n1378, n1971, n2087), lin-10(e1439), lin-28(n719), lin-11(n389), unc-59(e261)

L.G. II: lin-8(n111), lin-31(n301), egl-44(n998, n1080, n1087), lin-23(e1883), lin-4(e912), lin-26(n156), lin-5(e1348), unc-4(e120), lin-29(n333, n836, n1440), lin-7(e1413)

L.G. III: ced-4(n1162), lin-16(e1743), lin-37(n758), lin-13(n387), mab-5(e1239), egl-5(u202, n945), lin-36(n766), unc-86(e1416), lin-9(n112), lin-12(n941, n302), lin-19(e1756), lin-30(e1908), lin-39(n1792)

L.G. IV: lin-1(e1275), lin-3(e1417), lin-22(n372), lin-33(n1043), mec-17(u265), mec-3(e1338, u6), lin-34(n1041), lin-24(n432), ced-3(n717)

L.G. V: dpy-11(e224), egl-46(n1075, n1076, n1127), lin-25(e1446), him-5(e1490), unc-61(e228)

L.G. X: lin-32(u282), lin-18(e620), mec-7(u443, u448), vab-3(e1796) (a.k.a lin-20), lin-14(n179, n536, n355, n355n531, n536n540), lin-2(e1309), lin-15(n309,n767), mec-4(e1611) References for the mutations are: ced mutations: Ellis et al. (1986); egl and sem-4 mutations: Trent et al. (1983) and Desai et al. (1988); him-5 mutation: Hodgkin et al. (1979); mab-5: Hodgkin (1983); mec, lin-32 and unc-86 mutations: Chalfie and Sulston (1981) and Chalfie and Au (1989); other lin mutations: Horvitz and Sulston (1980), Ambros and Horvitz (1984), Ferguson and Horvitz (1985), and M. Chalfie (unpublished data) and E. Hedgecock (personal communication); and other unc and the dpy-11 mutations: Brenner (1974). Strains containing these mutation were either in our collection or generously provided by the Caenorhab-ditis Genetics Center, Victor Ambros, Michael Basson, Gian Garriga, Ed Hedgecock and Bob Horvitz. Some of the double mutant strains were provided by these same individuals: lin-8(n111); lin-9(n112), lin-36(n766); lin-15(767), lin-37(n758); lin-15(n767) and egl-44(n1080); unc-4(e120) egl-46(n1075). The remaining multiple mutant strains were constructed by standard procedures (Brenner, 1974).

Digoxigenin in situ hybridization using an antisense oligonucleotide to the 3′ end of the mec-7 sequence (5′ GAACGCTTCG-GCGGCATCTT 3′) followed the procedure in the Boehringer Mannheim Genius Kit and of Tautz and Pfeifle (1989) except that after a short fixation animals were treated with β-mercaptoethanol (Cox et al., 1981) and proteinase K to permeabilize the cuticle. To stain nuclei, 0.5 μg/ml diamidinophenylindole (DAPI) was included in the final wash.

The hybridization is mec-7 dependent; it is absent in animals with mutations that delete the gene (u443 and u448; Savage et al., 1989). mec-7 mRNA is first detected in L2 larvae. The signal increases until the early L4 stage, when it begins to decline. Only the PLM cells stain in egg-laying adults. Usually L2-L4 larvae grown at 25°C (Way and Chalfie, 1989) were examined.

Animals were prepared for immunofluorescence microscopy according to the method of Finney and Ruvkun (1990) as modified by these authors (personal communication). Basically animals are fixed in 2% formaldehyde on ice for 30 minutes in the presence of EGTA, spermidine, 25% methanol. Fixation is followed by reduction first with β-mercaptoethanol and then with dithiothreitol. The animals are subsequently oxidized with H₂O₂.

Samples were incubated with shaking for 24 hours at room temperature with a 1:450 dilution of a rabbit anti-mec-7 antibody (C. Savage and M. Chalfie, unpublished data), washed 7 times with buffer AbA (Finney and Ruvkun, 1990), incubated with rhodamine isothiocyanate-conjugated goat anti-rabbit IgG antibody (Cooper Biochemical) and washed an additional 8 times with buffer AbB (Finney and Ruvkun, 1990). The primary antibody was preab-sorbed with acetone powders prepared by the method of Johnson (1989) from whole mec-7(u443) worms (the u443 mutation deletes the mec-7 gene; Savage et al., 1989). The secondary antibody was preabsorbed with acetone powders prepared from wild-type animals. Animals were observed with a Bio-Rad MRC-600 Confocal imaging system.

Only the touch receptor cells (ALML, ALMR, AVM, PVM, PLML and PLMR) stain intensely in wild-type animals; other cells stain less strongly (no cells stain in u443 mutants). To quantitate the differences in intensity, we measured fluorescence output of individual cell bodies using the histogram software provided with the confocal microscope (the mean of three readings taken from cells in composite images of optically sectioned animals was used). The relative intensity of staining of the AVM cells compared to the staining of the ALM cells is 0.87±0.16 (7) [mean±s.e.m. (number of animals)]. Non-touch receptor cells have 5-9% of the staining of the touch cells in wild-type animals [FLP: 0.07±0.01 (4); PVD: 0.09±0.01 (4); BDU: 0.05±0.01 (4)]. This low-level staining is referred to as ‘weak.’ Other weakly staining cells include one to four cells in the tail and two cells in the ventral ganglion that are also seen in mec-3 and unc-86 mutants. Given their position, number and bipolar shape, these latter neurons are likely to be the AVF cells (White et al., 1986). In contrast, the transformed FLP cells in egl-46(n1076) mutants (see RESULTS) display fluores-cence comparable to that of the touch receptor neurons [0.71±0.17 (5); AVM cells in these mutants have a relative intensity of 0.86±0.09 (4)].

The mec-4lacZ fusion vector (TU#44) was constructed using the mec-4 genomic DNA sequence of TU#12 (Driscoll and Chalfie, 1991) and the lacZ-containing plasmid pPD22.04 (Fire et al., 1990). We introduced an SphI site by site-directed mutagenesis (Kunkel, 1985) after position 4639 in TU#12 using the oligonu-cleotide 5′ AAGGCATGCAAAAAT 3′. Introduction of this site results a 4.6 kb HindIII-SphI restriction fragment that contains the mec-4 5′ regulatory sequences and genomic DNA encoding all but the last seven amino acids of the protein [as well as a substitution of Cys for Lys at position 490 (Driscoll and Chalfie, 1991]. This fragment was ligated to pPD22.84 that had also been digested with HindIII and SphI.

TU# 44 and pRF4, a plasmid that includes the dominant rol-6 allele su1006 [the rol-6 mutations causes animals to roll and serves as a marker for transformation (Kramer et al., 1990; Mello et al., 1991)], were coinjected at 50 μg/ml into wild-type animals. A strain that segregated approximately 80% rollers was obtained from the progeny of the injected animals. The concatemerized array of rol-6(su1006) and mec-4lacZ DNA in this strain (TU1422) was designated as uEx86.

To generate a strain in which rol-6(su1006) and mec-4lacZ were stably integrated into one of the C. elegans chromosomes, 40 rollers of strain TU1422 were irradiated at 330 rads/minute from a ¹³⁷Cs source for 12 minutes. Irradiated parents were allowed to lay eggs and 125 F₁ progeny were picked onto individual plates. Six F₂ rollers from each plate were put onto individual plates and three animals (from different F₁ parents) that produced only roller progeny were identified. BZ2 is one of the strains that contains the integrated DNA bzIs1.

Standard genetic procedures (Brenner, 1974) were used to place either uEx86 or bzIs1 into various genetic backgrounds. Animals containing either fusion were stained for β-galactosidase activity by the procedure of A. Fire (personal communication, see Xue et al., 1992).

Animals were fixed in buffered 2.5% glutaraldehyde, stained in 1% osmium tetroxide, positioned within a small agarose block, dehydrated and embedded in Medcast resin (cf. Sulston et al., 1983). Intermittent transverse thin sections and thick sections were cut through either the head or the tail, so that ganglia and nerve cords could be examined at 5-10 μm intervals, using a Philips CM10 electron microscope. Thin sections were poststained with uranyl acetate and lead citrate before microscopy.

The sampling procedure was generally thorough enough to confirm that extra processes were localized longitudinally in the same region as the expected touch neurons, as judged from the mec-7 antibody staining experiments. However, no attempt was made to preselect animals with extra touch receptors, nor to trace processes in serial sections back to their origin. These extra touch processes did not extend very far into the midbody region and often had fewer microtubules and more patchy mantle than their wild-type counterparts.

Both ALM cells and the AVM precursor, QR, were killed in one hour old larvae with a Laserscience laser (Chalfie and Sulston, 1981; Seydoux and Greenwald, 1989). Touch sensitivity (Chalfie and Sulston, 1981) of the resulting animals was tested in double blind tests over the next few days.

In order to follow cell fate rather than the expression of a particular gene, we have employed several methods to characterize touch cell differentiation. Initially we used in situ hybridization to mec-7 mRNA (Table 1). This mRNA is detected only in the six touch receptor neurons in wild-type animals. To examine cell morphology, we have used a serum antibody specific to the mec-7 β-tubulin (Table 2; Fig. 1). This is a more sensitive method: intense staining is seen in wild-type animals only in the six touch receptor neurons, while weak staining is seen in the PVD cells, the FLP cells, the BDU cells (the sister cells to the ALM touch cells) and a few other cells. We have also examined mec-4 expression using a mec-4lacZ fusion (Table 3; Fig. 2). This fusion is expressed in wild-type animals primarily in the six touch receptor neurons, although the FLP, PVD and BDU cells stain infrequently. In some cases, we also determined whether the degeneration-causing mutation mec-4(e1611), which causes the death of the touch receptor neurons (Chalfie and Sulston, 1981; Driscoll and Chalfie, 1991), caused the deaths of ectopic touch receptor-like cells. Finally, we examined cells by electron microscopy (Fig. 3). This allows us to determine whether the large diameter microtubules and mantle are present and how the processes are positioned relative to other neuronal processes. Examination of mutants known to be disrupted in touch cell development with these methods gave results that were consistent with (or enlarged upon) the previously noted phenotypes of the mutants (Tables 1-3; Figs 1, 2). For the mutants described in the following sections, all of these methods gave similar results, suggesting that the ectopic touch receptor-like cells that we find are cells whose fates have been changed.

To identify genes that regulate touch cell fate, we used these methods to examine mutants defective in 48 genes that had previously been shown to be needed for cell development (Tables 1-4; Figs 1-3). Because several genes are discussed in the following sections, we will invert the normal order of presenting information by providing a model that describes our results (Fig. 4) and then presenting the data on which the model is based.

In this model, the unc-86, mec-3 and lin-14 genes are needed to activate the expression of touch cell characteristics. The number of touch-receptor neurons is restricted to six in wild-type animals because of negative regulation of these positively acting genes or their effects. Specifically, the other mec-3-and unc-86-expressing cells do not develop as touch cells because either lin-4 represses lin-14 activity (PVD neurons) or the effects of these three genes (not their expression) are repressed by the egl-44 and egl-46 genes (FLP neurons). Another pair of neurons, tentatively identified as the PHC cells, also fails to develop as touch receptors (and do not express mec-3) because sem-4 represses mec-3 activity. Four additional cells that could express touch cell features die by programmed cell death, a process involving the genes ced-3 and ced-4.

Gain-of-function mutations in the heterochronic gene lin-14 cause the repetition of several first larval stage (L1) lineages, while loss-of-function mutations result in the premature appearance of later stage lineages (Ambros and Horvitz, 1984). The lin-14 gene appears to be negatively regulated by the lin-4 gene (Ambros, 1989; Arusa et al., 1991); a lin-4 mutation causes similar lineage defects to those produce by gain-of-function lin-14 mutations (Chalfie et al., 1981; Ambros and Horvitz, 1984). lin-14 protein is normally found in cells of the embryo and L1 larva, but not the L2 larva (Ruvkun and Giusto, 1989); its production is presumably turned off by lin-4 at this latter time.

We find that lin-14 activity is needed for the choice of whether the two postembryonic touch cells (AVM and PVM) or the two PVD cells are produced. These two pairs of cells are generated from different precursors during the L1 and L2 stages, respectively (Sulston and Horvitz, 1977). lin-14 loss-of-function mutants lack the AVM and PVM cells (Tables 1, 2); the cells that should give rise to the AVM and PVM cells produce a PVD-type cell lineage instead (Ambros and Horvitz, 1984). In contrast, both a gain-of-function lin-14 mutation, n355, which results in the continued production of lin-14 protein (Ruvkun and Giusto, 1989), and a lin-4 mutation produce an effect opposite to that of the lin-14 loss-of-function mutations: two (or more) extra cells express mec-7 and mec-4 in the PVD region in 4-40% of the animals (Tables 1-3). It seems likely that the PVD lineage has been altered so as to produce AVM-/PVM-like cells. Precursors giving rise to migrating cells (normally found in the L1 AVM/PVM lineages) are sometimes found in second stage larvae with lin-14 gain-of-function mutations (V. Ambros, pers. comm.). The PVD lineage is also transformed in lin-4 animals (Chalfie et al., 1981). The infrequent appearance of the extra touch-like cells may be due to the absence of necessary precursors because of earlier lineage defects in the mutants. Consistent with the model that lin-4 negatively controls lin-14 expression, double mutants containing the lin-4 mutation and lin-14 loss-of-function mutations detectably expresses mec-7 only in the ALM and PLM cells (Tables 1, 2).

These experiments suggest that lin-14, controlled by lin-4, acts as a genetic switch controlling the choice between the AVM/PVM and PVD lineages [two other heterochronic genes, lin-28 and lin-29 (Ambros and Horvitz, 1984) do not appear to be required; Table 1). We do not know whether lin-14 affects only the precursors of the AVM, PVM and PVD cells or also acts on the cells directly, but the results with the egl-44 and egl-46 mutants suggest that this latter role is possible (see below).

One paradox concerning the role of lin-14 as a coactivator in touch cell development is that lin-14 mutations do not apparently affect the function or development of the ALM and PLM cells (lin-14 mutants are touch sensitive; Ruvkun and Giusto, 1989; M. Chalfie, unpublished data). Since lin-14 is needed for the development of ectopic touch receptor-like cells (see below), its activity may be redundant in the ALM and PLM cells.

The egl-44 and egl-46 genes, which are needed for the differentiation of the HSN cells [a pair of neurons needed for egg-laying (Desai et al., 1988; Desai and Horvitz, 1989)], appear to encode negatively acting regulators that prevent the expression of touch cell features. Mutations in either gene result in the appearance of a pair of touch receptor-like cells in the region of the second pharyngeal bulb (Tables 1-3; Figs 1-3). In addition, mutants containing the dominant, degeneration-causing mutation mec-4(e1611) and a mutation in either egl-44 and egl-46 have one or two additional dying cells in the position of the FLP cells in newly hatched larvae (7/13 egl-44 animals and 11/17 egl-46 animals). It is likely that the ectopic cells are transformed FLP neurons, since they are in the normal FLP position and are the only cells in this position in egl-46 mutants to express a mec-3-lacZ fusion (D. Xue and M. Chalfie, unpublished data). The egl-44(n998); egl-46(n1076) double mutant did not display a greater transformation.

This transformation requires the lin-32, unc-86, mec-3 and lin-14 genes (Table 1). The requirement for lin-14 is somewhat surprising, even though lin-14 protein is found in wild-type FLP cells (Ruvkun and Giusto, 1989; G. Ruvkun, personal communication), since the FLP cells are non-dividing, embryonically derived cells (Sulston et al., 1983) and lin-14 defects had previously been described only for cells that divide in larvae (Ambros and Horvitz, 1984). These results suggest that the involvement of lin-14 in touch cell differentiation may be more direct than simply acting to determine the fate of precursor cells. In this way, lin-14 may act similarly to unc-86, which is needed at several stages in touch cell development.

We tested the extent of the FLP transformation by ablating the ALM and AVM cells (to abolish anterior touch sensitivity; Chalfie and Sulston, 1981) in egl-44(n998) (n=12) and egl-46(n1127) (n=5) mutants. The resulting animals were not detectably touch sensitive, perhaps because the ectopic touch receptor-like cells are unable to make appropriate synapses (these mutations also cause errors in neuronal outgrowth; Desai et al., 1988; Fig. 2).

The sem-4 gene, which is needed for the development of the HSN egg-laying neurons (Desai et al., 1988), also acts as a negative regulator in touch cell development. sem-4 mutants have two additional cells in the tail with touch cell-like features, including the large-diameter micro-tubules and mantle (Tables 1-3; Figs 1-3). This ectopic touch cell expression requires lin-14, lin-32, unc-86 and mec-3 (Table 1). Similar electron microscopy observations to ours have been made by M. Basson and H. R. Horvitz (personal communication), who have suggested that these extra cells arise from altered lineages that normally generate the PHC neurons. Since unc-86 (Finney and Ruvkun, 1990), but not mec-3 (Way and Chalfie, 1989), is expressed in the PHC cells in wild-type animals, we suggest that the wild-type product of the sem-4 gene may negatively regulate mec-3 expression in these cells.

Prevention of programmed cell death by ced-3 and ced-4 mutations leads to the appearance of additional differentiated cells (Ellis and Horvitz, 1986; Avery and Horvitz, 1987; Chalfie and Wolinsky, 1990; White et al., 1991). We find that these mutations result in the appearance of four additional touch receptor-like cells (Tables 1-3; Figs 1-3) in regions where several cells normally die (Sulston and Horvitz, 1977; Sulston et al., 1983). Two cells are found in the tail near the PLM cells. These cells also degenerate in ced-4(n1162); mec-4(e1611) double mutants. The ectopic mec-7 mRNA expression in these cells requires lin-32, unc-86 and mec-3 (Table 1; the effect of lin-14 mutations has not been tested). The two other cells, which look like the AVM and PVM cells and are situated near them, arise less frequently than the cells in the tail (Tables 1-3; Fig. 1). Their relative rarity may indicate that the transformation is less complete in these cells (incomplete function for cells prevented from dying has been seen previously by Avery and Horvitz, 1987).

Because the four extra touch receptor-like cells in these mutants are located near the PLMR, PLML, AVM and PVM cells, it seems likely that the new cells arise from the same lineages as the wild-type touch receptors. Two cells die in each of these four lineages (Sulston and Horvitz, 1977; Sulston et al., 1983), but only one (Q(R/L).pp and AB.p(r/l)apapppppp) expresses detectable unc-86 protein (Finney and Ruvkun, 1990) and is, thus, likely to be the extra touch receptor-like cell.

Our results suggest that C. elegans has solved the problem of making four different types of neurons by using combinations of positively and negatively acting factors (Fig. 4). A further restriction in cell number is caused by programmed cell death. None of the genes that specify or restrict touch cell expression acts in a cell-specific fashion (Desai et al., 1988; Chalfie and Au, 1989; Desai and Horvitz, 1989; Way and Chalfie, 1989; Finney and Ruvkun, 1990).

There are at least two stages in the specification of the touch cells: the expression of the regulatory gene mec-3 (controlled by unc-86 and sem-4) and the expression of touch cell features [controlled by mec-3, lin-14, egl-44, egl-46 and unc-86 (Hamelin et al., 1992; present results ; A. Duggan and M. Chalfie, unpublished data)]. The specification of cell fate, however, does not appear to be a strictly linear process; the cumulative effect of successively acting regulatory factors is also important. The low level expression of mec-7 protein in unc-86 and mec-3 mutants (RESULTS and Hamelin et al., 1992) and the requirement for lin-14 in the touch cell-like differentiation of FLP cells in egl-44 and egl-46 mutants suggest that unc-86, lin-14 and, perhaps, other genes act at more than one regulatory stage.

We do not know whether the genetic interactions diagrammed in Fig. 4 are direct or indirect. unc-86 and mec-3 encode DNA-binding proteins (Xue et al., 1992) found in the touch cells (Way and Chalfie, 1989; Finney and Ruvkun, 1990) and other genes may regulate the expression of genes within these cells. As argued above, the lin-14 gene, which encodes a nuclearly localized protein of unknown molecular function (Ruvkun and Giusto, 1989), may be required in the FLP cells for them to acquire touch receptor characteristics in egl-44 and egl-46 mutants. Alternatively, some genes, e.g. lin-14, may (or may also) regulate the production of touch cell features indirectly by affecting the fate of the immediate (parental and grandparental) precursor cells. The roles of genes such as egl-44, egl-46 and sem-4 should become better understood when these gens are cloned and their expression patterns examined.

Because mutations in lin-14, egl-44, egl-46 and sem-4 affect expression of mec-4 and mec-7 and the ultrastructure of the affected cells, these genes specify cell fate. However, the changes in the cell fate produced by these mutations do not appear to be complete: many of the extra cells in lin-4 or lin-14 (gain-of-function) mutants do not migrate to the AVM position, the extra cells in the egl-44 and egl-46 mutants cannot mediate a touch response, and the extra cells in several mutants sometimes have fewer microtubules and less mantle than wild-type cells. We do not know whether these defects result from incomplete transformation because necessary factors are lacking, inappropriate transformation because other factors are present (discussed in Dickinson, 1988), pleiotropic effects of the mutations, or inappropriate cell interactions or timing (the PVD cells arise later and the FLP cells are positioned more anterior than any of the touch receptor neurons). Pleiotropy may explain why egl-44 and egl-46 are needed both to repress touch receptor features in the FLP cells and to allow proper development of the AVM touch cells: the effects on nerve outgrowth, for AVM and perhaps the transformed FLP cells and other cells (Desai et al., 1988), may be indirect. In addition, cell interactions are necessary for the proper outgrowth of the AVM touch cell (Chalfie et al., 1983; Walthall and Chalfie, 1988).

We have identified several of the components needed for the combinatorial specification of touch cell differentiation. However, since we only examined the effects of known cell differentiation mutations, we would be surprised if we had identified all the genes needed for this process. At least one other factor may make lin-14 redundant in the ALM and PLM touch cells (see above), and other genes may be needed to produce the different synapses made by the touch receptor neurons (Chalfie et al., 1985). Other candidate genes include lin-32, which appears from the existing three alleles to be minimally needed for the development of the ALM touch cells (Chalfie and Au, 1989), and ceh-18, a recently identified POU-type homeobox gene of unknown function that is expressed in the ALM and AVM cells but not in other neurons (D. Greenstein, S. Hird and G. Ruvkun, personal communication).

Although future experiments will undoubtedly refine our picture of touch cell differentiation (e.g., screens for mutations that alter the pattern of mec-7 and mec-4 expression may reveal other genes in this developmental pathway), it is clear that combinatorial control is needed to direct cell differentiation. The genetic interactions that we have found in touch receptor differentiation are similar to the genetic circuits deduced for embryonic development in Drosophila (reviewed e.g., in Akam, 1987; Ingram, 1988). Our results show that similar developmental pathways also direct the terminal specification of cell fate.

We thank Geraldine Seydoux and Iva Greenwald for assistance with laser ablations. S. Mitani is grateful to Professors K. Takahashi and S. Sassa for encouragement. This work was supported by NIH Grant GM31997 and a McKnight Development Award to M. C. and a fellowship for research abroad from the Japan Society for the Promotion of Science to S. M.