HOM-C/Hox genes and four interacting loci determine the morphogenetic properties of single cells in the nematode male tail

Two central processes in the development of metazoans are the production of multiple cell types and the morphogenesis of complex multicellular structures. Morphogenesis requires that reproducible homotypic and heterotypic cellular contacts and interactions be established following precise cell recognition events. In such morphogenetic interactions, cells of a single overt cell type may not all act the same. Pattern forming mech-anisms that generate non-equivalent cell properties thus play a crucial role in establishment of the form of the organism. This process is particularly acute in development of the nervous system, where neurons acquire distinct characteristics necessary to generate a complex network of cell contacts.

We have been studying development and morphogenesis of the copulatory structures of the C. elegans male tail (Emmons, 1992). Because its cellular structure is completely defined (Wood, 1988), C. elegans provides an opportunity for investi-gating, in detail, the relationship between pattern formation and morphogenesis. We have shown previously that the distinctive arrangement of sense organs in the male tail is dependent on a pattern formation mechanism in the epidermis (Baird et al., 1991). This pattern formation mechanism generates differ-ences between cells of the peripheral nervous system necessary for organelle assembly at reproducible positions. Here we show that two genes encoding Antennapedia-class home-odomain transcription factors, members of the C. elegans HOM-C/Hox gene cluster, as well as several genetic modifiers of HOM-C/Hox gene mutations, play a role in the pattern formation mechanism.

The C. elegans epidermis consists of a large syncytium covering most of the body, distinctly shaped cells and syncytia constructing the head and tail regions, and left and right lateral rows of seam cells (White, 1988). The male-specific sense organs that are the subject of this paper are generated by seam cells. At hatching, there are nine seam cells arrayed in single file on each side of the body. During postembryonic develop-ment, seam cells generally divide in a stem cell pattern to generate an anterior daughter that fuses with the large hypo-dermal syncytium, and a posterior daughter that remains as a seam cell (Sulston and Horvitz, 1977; Fig. 1A). In certain regions of the body, however, seam cells generate neuroblasts. Seam cell lineages can therefore serve as a model to study the mechanism of axial pattern formation in C. elegans. During the L1 larval stage in both sexes, the embryonic seam cell T generates the phasmid, a chemosensory sensillum. During the L2 larval stage in both sexes, a descendant of the embryonic seam cell V5 generates the postdeirid. The postdeirid is a neuronal structure of unknown function. During the L3 larval stage in males, descendants of V5, V6, and T generate ray neu-roblasts, termed ray precursor cells (Rn cells) (Fig. 1B). Each ray precursor cell expresses the ray sublineage, and gives rise to the cells of one ray (Fig. 1C). The rays are male-specific sensilla that extend outward from the tail in an acellular fan (Fig. 1D). They are necessary for copulation. As the rays differ from each other, there is further axial patterning within the tail region (Fig. 1E).

Each ray comprises the dendritic endings of two neurons and a support cell. The ends of the rays are located at reproducible positions in the cuticle, giving the male a characteristic, species-specific posterior morphology (Fig. 1D). In our previous studies we described the cellular mechanisms involved in the development of the adult ray pattern, and demonstrated a role of six genes essential to the process (Baird et al., 1991). From these studies we concluded that the rays were not morphogenetically equivalent. In spite of the fact that they arose from repetition of a stereotyped cell sublineage (Sulston and Horvitz, 1977) (Fig. 1C) and comprised cells of the same three cell types (Sulston et al., 1980), each ray behaved differently in its interactions with surrounding epidermal cells. Furthermore, cells of individual rays were able to recognize each other during organelle assembly, suggesting that they expressed distinct cell recognition functions. Both these properties were lost in mutants, resulting in misplace-ment of rays and fusion of one ray to another. The rays also differ in their individual morphology, their ultrastructure, and their neurotransmitter usage (Fig. 1E) (Sulston and Horvitz, 1977; Sulston et al., 1980). Thus in development of the rays we see an example of the specification of differences among cells of a single differentiated cell type, and of the dependence of morphology on these differences.

We report that mutations in two genes of the C. elegans HOM-C/Hox gene cluster, mab-5 and egl-5, affect the mor-phology of particular rays. In C. elegans, homeotic selector genes are organized into a gene cluster similar to the HOM-C/Hox clusters of fruitflies and vertebrates (Clark et al., 1993; Wang et al., 1993; McGinnis and Krumlauf, 1992). mab-5 is most similar to Drosophila HOM-C genes Antennapedia, Untrabithorax, and AbdominalA, while egl-5 is most similar to Drosophila HOM-C gene AbdominalB. In addition to mab-5 and egl-5, the C. elegans HOM-C gene cluster contains the gene lin-39, which shares similarities to Drosophila probosci-pedia, Sex combs reduced, and Deformed, and ceh-13, which shares similarities to Drosophila labial. Mutations in the C. elegans genes, with the exception of ceh-13, have been shown to affect the division patterns, fates, migration, fusion, and neu-rotransmitter expression of cells in particular body regions.

We describe a role of mab-5 and egl-5 in determining addi-tional morphogenetic properties of cells. In particular, we show that raising and lowering the gene dosage of mab-5 and egl-5 results in changes in ray properties that appear to reflect specific transformations of one ray morphogenetic type into another. For two of the six most anterior rays, ray morphology is influenced by the ratio of the number of copies of these two genes. Action of mab-5 and egl-5 in the different terminal branches of the ray lineages appears to be cell-autonomous, but the characteristic level of expression of these genes postulated to define each lineage branch may be influenced by interac-tions between seam cells. Mutations in several of the genes described previously are shown to be dominant modifiers of mutations in mab-5 and egl-5. This suggests that their action is related to implementing the pattern formation mechanism.

All nematodes used in this work were derivatives of strain Bristol, N2. They were reared following standard methods (Brenner, 1974; Sulston and Hodgkin, 1988). Unless otherwise noted, animals were grown at 20°C.

Most bx alleles of mab genes were isolated by EMS mutagenesis (Baird et al., 1991; this work), following the mutagenesis protocol of Brenner (1974). bx80 was isolated after mutagenesis with 10^—4 M 1,2,3,4-di-epoxy-butane (DEB) (Trent et al., 1991). The strains used generally carried the him-5(e1490) mutation, which elevates X chro-mosome non-disjunction (Hodgkin et al., 1979). The purpose was to increase the frequency of spontaneous males among the progeny of selfing hermaphrodites.

Strains bearing the following additional mutations were used: Linkage group I: mab-20(bx24), mab-20(bx61ts)

Linkage group II: tra-2(q279), mnC1[dpy-10(e128)unc-52(e444)] Linkage group III: egl-5(n486), egl-5(n945), mab-5(e1751), mab-5(bx54), mab-5(e1239), mab-21(bx53), mab-21(bx41), mab-21(sy155), qDp3, pal-1(e2091), unc-36(e251)

Linkage group IV: dpy-9(e12), lin-22(mu2), mab-26(bx80), unc-17(e245)

Linkage group X: mab-18(bx23)

Properties of these mutations have been or are described as follows: egl-5(n486) and egl-5(n945), Chisholm (1991); lin-22(mu2), Waring et al. (1992); mab-5(e1239), Kenyon (1986); mab-5(bx54), this work; mab-5(e1751), Hedgecock et al. (1987); alleles of mab-18 and mab-20, Baird et al. (1991); alleles of mab-21, Baird et al. (1991) and Chow and Emmons (in preparation); mab-26(bx80), this work; pal-1(e2091), Waring and Kenyon (1990). For dpy-9(e12), unc-17(e245), tra-2(q279), unc-36(e251), mnC1, and qDp3 see Hodgkin et al. (1988). Additional known properties of the mab-5 and egl-5 alleles relevant to the work described here are presented in Table 1.

The bx54 and bx80 mutations were isolated by screening for mor-phological defects among male F² progeny of him-5(e1490) her-maphrodites mutagenized with EMS (bx54) or DEB (bx80) (Brenner, 1974). The bx54 mutation was identified as a mutation causing loss of anterior rays. It was mapped to LG III and shown to be an allele of mab-5 by failure to complement mab-5(e1239). S. Salser and C. Kenyon have shown that bx54 is a E(135)→K missense mutation in the homeobox of mab-5 (personal communication).

The semi-dominant bx80 mutation was identified by its ray fusion phenotype: in homozygotes rays 1 through 6 are variably fused together. Males and hermaphrodites also have deformities of the body in the nose, posterior midbody, and tail regions. mab-26(bx80) was placed near to or to the left of dpy-9 on the left end of LGIV by a three factor cross with dpy-9 and unc-17 (data not shown).

Animals heterozygous for two unlinked genes were analyzed by mating him-5(e1490) males to homozygous double mutants and scoring the ray phenotype of cross progeny males. For the linked genes mab-21, mab-5, and egl-5, doubly heterozygous strains were first constructed by mating single mutants together. Lines segregating both homozygous and double heterozygous progeny were maintained and heterozygous males were scored for ray phenotype. For strains containing the duplication qDp3, either the chromosomal mutation unc-36(e251), which is covered by qDp3, was used to score for the presence of the duplication (Table 2, lines 1 and 5), or the presence of qDp3 was indicated by the presence of a full complement of rays and normal gross tail morphology (Table 2, lines 3, 11, and 12). To score the effect of heterozygosity of the X-linked gene mab-18, the transformer mutation tra-2(q279) II was used. XX animals, normally hermaphrodites, are transformed into phenotypic males by the recessive tra-2(q279) mutation. tra-2(q279) was maintained in het-erozygous strains balanced by the inversion chromosome mnC1; such strains segregated one quarter tra-2 homozygous males. tra-2 males were mated to tra-2/mnC1; mab-18 hermaphrodites, and cross-progeny males, identified as males with at least one wild-type ray 6, were scored for ray fusion phenotype.

Morphology of adult males and cell lineages were determined by Nomarski microscopy with a Zeiss Axioskop or Axioplan micro-scope. For scoring ray fusion, males were mounted by the Standard Mount procedure (Sulston and Hodgkin, 1988) without bacteria and observed at 400×. For cell lineage analysis, the coverslip was spread with bacteria, and animals were observed at 1000×. For laser ablation, the Anesthetic Mount was used (Sulston and Hodgkin, 1988). Cells were killed by multiple pulses with a VSL337 nitrogen laser equipped with a dye laser module containing 440 nm coumerin dye (Laser Science Inc., Cambridge, MA; see Avery and Horvitz, 1987). The laser beam was directed into the incident-light port of the Axioplan microscope with a VSL-LMA Laser Microscope Adapter (Laser Science Inc., Cambridge, MA).

The C. elegans HOM-C/Hox gene mab-5 is required for gen-eration of rays by descendant cells of seam cells V5 (ray 1) and V6 (rays 2–6), but not T (rays 7–9) (Kenyon, 1986) (Fig. 1). We found that the level of expression of mab-5 also affects the morphology of some of the rays descended from V5 and V6. (The rays generated by V5 and V6 will be referred to col-lectively as V-rays.) In males homozygous for strong mab-5 loss-of-function alleles, the ray sublineage is not expressed in the V5 and V6 lineages, and the L4 larval seam cells instead generate adult cuticular structures known as alae, as do lineages generated by more anterior seam cells (Kenyon, 1986). As no V-rays are generated, the role of mab-5 in spec-ifying ray morphology could not be assessed in this back-ground. In our screens for mutations affecting male tail mor-phology (Materials and Methods), we identified a weak allele of mab-5 that allowed us to overcome this difficulty.

The mab-5 mutation bx54 is a hypomorph by the criteria that it causes loss of only some V-rays, and affects ray morphology in a manner similar to a mab-5(0)/mab-5(+) heterozygote (described below). In mab-5(bx54) homozygotes, only anterior V-rays are lost, and it is possible to examine the role of mab-5 in specification of the morphologies of the remaining rays. By morphological criteria, mab-5(bx54) homozygous males lack rays 1, 2, and 4 (Fig. 2). However, by cell lineage analysis we determined that V5.ppppp, V6.papap, and V6.pappp, the normal precursors of rays 1, 2, and 3 (Fig. 1), failed to express the ray sublineage in mutant worms (6/6 sides examined). The most anterior ray was generated by V6.pppap, which normally gives rise to ray 4 (Fig. 1). Thus in the mab-5(bx54) mutant back-ground the ray generated by V6.pppap has moved to a position adjacent to the cloaca and extended to the margin of the fan; i.e., it has assumed the morphology of ray 3. This transformation was not the consequence of the loss of rays 1, 2, and 3, because these rays may be eliminated by cell ablation without a morphologi-cal transformation of ray 4 (Table 4, lines 21 and 22).

We also examined the effect of lowered mab-5 gene dosage in a mab-5(0)/mab-5(+) heterozygote. In nematodes carrying a single copy of the putative mab-5 null allele, e1239, the fourth ray formed more anteriorly, opened at the margin of the fan, and was fused to ray 3 in a small but significant percentage of sides (11.5%) (Fig. 2; Table 2, line 10). (As there was no evidence to the contrary, rays on the two sides of single animals are assumed to be independent; thus data throughout is reported as percentage of sides with a particular phenotype.) Fusion with ray 3 suggests that the fourth ray now not only extends to the fan margin at a more anterior position, but also expresses one or more cell recognition molecules in common with ray 3. Thus, when mab-5 function is decreased, several characteristics of the fourth ray undergo a posterior-to-anterior transformation. We refer to these several changes collectively as a change in the identity of the ray generated by V6.pppap. As the ray normally generated by this cell is referred to as ray 4, and since the new identity appears to be the same as that normally taken by ray 3, we refer to the phenotype in a mab-5 hypomorph or heterozy-gote as a transformation of the identity of ray 4 to that of ray 3. A similar shorthand is used throughout in referring to trans-formations affecting the other rays.

A gain-of-function allele of mab-5 resulted in an anterior-to-posterior transformation in ray identities opposite to that seen with mab-5 loss-of-function alleles. mab-5(e1751) is a tandem duplication covering the mab-5 gene that results in increased or ectopic expression of normal mab-5 gene product (Hedgecock et al., 1987; Salser and Kenyon, 1992; Salser et al., 1993). In males homozygous for mab-5(e1751), ray 1 was located more posteriorly, opened on the ventral surface of the fan or at the fan margin, and fused with ray 2 (74%, n = 200). Likewise, ray 3 was located more posteriorly, opened on the ventral surface of the fan (80%, n = 200), and sometimes (3%, n = 200) fused with ray 4 (Fig. 2). We interpret these mor-phological changes as anterior-to-posterior transformations in the identities of ray 1 to that of ray 2, and ray 3 to that of ray 4.

The results described above suggest that increased levels of mab-5 gene expression caused rays 1 and 3 to develop like their posterior neighbors, and decreased level of mab-5 gene expression caused ray 4 to develop like its anterior neighbor. This might be explained by the presence in the wild type of a gradient of mab-5 gene activity through the tail region, with increasing levels of activity determining more posterior ray development. In a simple model of this kind in which mab-5 alone plays a determining role, if mab-5 gene function were brought to its lowest level, or eliminated, all rays should take on a common anterior morphology.

It was possible to test this prediction in a background con-taining a mutation at the lin-22 locus. lin-22 mutations have a phenotype opposite to that of mab-5 mutations. That is, whereas in mab-5 mutants the ray sublineage is not expressed by the V5 and V6 lineages, in males homozygous for lin-22 loss-of-function mutations, the ray sublineage is expressed by anterior seam cells in addition to V5, V6, and T (Horvitz et al., 1983). In lin-22;mab-5 double mutants, some seam cells express the ray sublineage, but these are not confined to the tail region (Kenyon, 1986). This indicates that the combined function of mab-5 and lin-22 is to regulate the spatial expression of the ray sublineage. In mab-5;lin-22 double mutants, some V-rays as well as T-rays are present in the fan. We examined the morphology of such rays in a mab-5(e1239);lin-22(mu2) double mutant. We found V-rays of all types were generated at comparable, low frequencies (per-centage of sides with a given ray was as follows: ray 1, 6%; ray 2, 13%; ray 3, 17%; ray 4, 12%; ray 5, 15%; ray 6, 15%; n = 200). Thus, there is no simple relationship between mab-5 expression level and any one particular ray identity or mor-phology. Since posterior V-ray morphologies can develop even in the absence of mab-5 gene function, additional genetic functions must play a role in specification of ray morphology.

The AbdB-homolog egl-5 is required for expression of spe-cialized cell fates in the region of the cloaca and elsewhere (Chisholm, 1991). In males homozygous for egl-5(lf) mutations, the morphology of the tail is severely affected by transformations in the fates of several male-specific blast cells. Among these, V6.ppppa is transformed into a hypodermal cell and hence ray 6 is lost. Thus egl-5 plays a restricted role in specifying expression of the ray sublineage by a single branch of the V6 lineage (Chisholm, 1991).

In addition, in egl-5 mutants, rays 2 through 5 appear to loose their separate identities. During the late L4 larval stage, papillae of rays 2 through 5 cluster together in the lateral hypo-dermis, and after morphogenesis short rays clustered and fused together can be seen (Fig. 3). Fusion of rays 2 through 5 was more dramatically demonstrated by Chisholm (1991), who showed that in mosaic animals that were egl-5(—) in the V6 lineage but egl-5(+) elsewhere, all the rays were fully formed and rays 2 through 5 were fused. These observations suggest that in egl-5(—) rays 2 through 5 take on a single, common identity. For example, they might assume some ground state identity. Multiply fused rays usually extend to the fan margin, suggesting the minimal ground state might be that of ray 3. Ray 1 appeared to be unaffected (Fig. 3). We conclude that, in addition to being required for expression of the ray 6 sublin-eage, egl-5 is required for specification of at least three of the four identities taken by rays 2 through 5.

As was the case with mab-5, egl-5 is weakly haplo-insuffi-cient for expression of the identity of one of the rays. In an egl-5(0)/egl-5(+) heterozygote, in 7% of sides the morphology of ray 6 was transformed into that of ray 4 (Table 2, line 4). In these animals, ray 6 lost its distinctive, tapered morphology, moved anteriorly, and fused with ray 4 (Fig. 3). In fused rays, there were two ray openings, showing that the tip of ray 6, which is not open in wild type, was transformed to one typical of the other rays (Chow, Hall and Emmons, unpublished data). Specification of ray 6 identity therefore appears to be sensitive to the level of egl-5 gene activity.

In other systems HOM-C/Hox genes may act redundantly, combinatorially, or one may suppress the effects of another in the establishment of a positional code (McGinnis and Krumlauf, 1992; Morgan et al., 1992; Salser et al., 1993). To test whether mab-5 and egl-5 interact in specification of ray morphological identities, we analyzed strains carrying mutations in both of these genes. Since the phenotype of the mab-5 egl-5 double homozygote reflected the additive pheno-types of both mutations (absence of V-rays and deformed tail region; data not shown), it was necessary to examine the inter-action of these genes in double heterozygotes. We found that both genes have an effect on the specification of both ray 4 and ray 6 morphological identities, and that the ratio of the number of gene copies appears to be a critical factor.

The data for ray 6 are presented in Table 2, lines 1–4, 7, and 8. As already discussed, ray 6 morphology appeared to be sensitive to the level of egl-5 gene product; in an egl-5(0)/egl-5(+) heterozygote, ray 6 lost its distinctive morphology and fused with ray 4 at a frequency of 7% (Table 2, line 4). In such a heterozygote the ratio of mab-5(+) to egl-5(+) genes was 2:1. In order to examine whether mab-5 also played a role in spec-ification of ray 6, this ratio was increased by the introduction of a single copy of the dominant mab-5 tandem duplication allele e1751. In this background, the frequency of transforma-tion of ray 6 to ray 4 was increased to 13% (Table 2, line 2). The ratio of mab-5 to egl-5 was also increased by taking advantage of the free duplication qDp3, which covers the chromosomal region containing the entire C. elegans HOM-C/Hox gene cluster, including mab-5 and egl-5. A ratio of mab-5(+) to egl-5(+) of 3:1 was obtained by the introduction of qDp3 into a strain homozygous for a strong egl-5 loss-of-function allele. In this background the frequency of transfor-mation of ray 6 to ray 4 rose to 34% (Table 2, line 1). Evidence that this increase was due at least in part to the additional copy of mab-5(+) on qDp3 was obtained by analyzing a similar qDp3-containing, egl-5(—) strain in which one chromosomal wild-type copy of mab-5 was replaced by a mab-5 null allele. In this strain the ratio of mab-5(+) to egl-5(+) was 2:1 and the frequency of transformation of ray 6 to ray 4 was reduced to 16% (Table 2, line 3).

These results demonstrated that additional copies of mab-5(+) enhanced the effect on ray 6 of lowering the number of copies of egl-5(+). In a converse manner, lowering the number of copies of mab-5(+) suppressed the effect of lowering the number of copies of egl-5(+). In the double heterozygote (ratio of mab-5(+) to egl-5(+) of 1:1), there was no detectable trans-formation of ray 6 to ray 4 (Table 2, lines 7 and 8). Thus it is not the decreased number of egl-5 gene copies per se that causes the ray 6 phenotype, but rather it is the number of egl-5 gene copies relative to the number of mab-5 gene copies that is important.

A similar result was obtained showing an effect of egl-5 in specification of the morphology of ray 4 (Table 2, lines 7–12). As described above, reduction of the number of wild-type copies of mab-5 from two to one resulted in fusion of ray 4 with ray 3 at a frequency of 12% (Table 2, line 10). This effect of reducing mab-5 gene number was partially suppressed if the number of copies of egl-5 was also reduced (Table 2, line 8; 5.7% transformation of ray 4 to ray 3). Parallel results were obtained with the mab-5 hypomorphic allele bx54 (5.8% trans-formation in a heterozygote [Table 2, line 9], 1.3% transfor-mation in a double heterozygote with egl-5 [Table 2, line 7]). When qDp3 was added to a strain homozygous for a mab-5 null allele (ratio of mab-5(+) to egl-5(+) of 1:3) the frequency of transformation of ray 4 to ray 3 rose to 62% (Table 2, line 12). If one chromosomal egl-5(+) gene copy was replaced by egl-5(n945) in such a strain (ratio of mab-5(+)/egl-5(+) of 1:2), the frequency of transformation dropped to 30.5%, indi-cating that part of the effect of qDp3 was due to the increased dosage of egl-5(+) (Table 2, line 11).

Thus it appears that mab-5 and egl-5 both play roles in the specification of both ray 4 and ray 6 morphological identities. Their interaction is reciprocal: ray 6 is transformed with increasing frequency to ray 4 as the ratio of egl-5(+) to mab-5(+) falls below 1, and ray 4 is transformed with increasing frequency to ray 3 as the ratio of mab-5(+) to egl-5(+) falls below 1. Thus the ratio of mab-5 to egl-5 function may normally help to differentiate these structures. Other rays may not be affected in these experiments because the levels of mab-5 and egl-5 gene activity examined do not cross critical thresh-olds for these rays.

We described previously a set of six genes required for devel-opment of the wild-type ray pattern (Baird et al., 1991). Mutations in these genes caused transformations in ray mor-phology and fusion of rays that were interpreted to have resulted from failed specification or implementation of distinct ray morphological identities. For three of these genes, mab-18, mab-20, and mab-21, the mutant phenotypes were similar to those described here for mab-5 and egl-5. We have isolated an additional mutation defining a new gene, mab-26, also with a fused-ray phenotype. We asked whether these four genes acted in the same pathway as mab-5 and egl-5 in specification of ray morphology, by examining double mutants for evidence of genetic interactions. Enhancement or suppression of HOM-C/Hox gene effects would suggest action in the regulatory pathway specifying ray identity. Absence of a synergistic inter-action would suggest action independent of mab-5 and egl-5.

Two of the four genes have mutant phenotypes similar to that of egl-5. Mutations in mab-18 and mab-21 result in loss of the ray 6 morphology and fusion of this ray with ray 4. Mutations in both genes are recessive or show very weak dominant effects (Table 2, lines 14, 19-21) and are fully penetrant. mab-18 encodes a putative transcription factor con-taining a homeodomain most closely related to that of ver-tebrate Pax-6 (Y. Zhang and S.W. Emmons, unpublished data). Mutations in mab-21, in addition to affecting ray 6, also result in the transformation of one seam hypodermal cell into a ray precursor blast cell, which divides and gives rise to a tenth ray. This suggests that mab-21 is a regulatory gene required for specification of seam cell fates (K.L. Chow and S.W. Emmons, unpublished data). mab-21 encodes a novel protein which also has an essential embryonic function (K.L. Chow and S.W. Emmons, unpublished data).

Mutations in both mab-18 and mab-21 were dominant enhancers of egl-5. Reducing the number of copies of mab-18(+) or mab-21(+) from 2 to 1 in an egl-5(0)/egl-5(+) het-erozygote increased the frequency of transformation of ray 6 to ray 4 from 7% to around 30% (Table 2, compare lines 4, 13, and 18, also compare lines 2 and 17). Since the mab-21 and mab-18 alleles may not be null, these figures may underesti-mate the extent of the enhancement. Neither mutation inter-acted significantly with mab-5. There was little or no effect of making these genes heterozygous in a mab-5(0)/mab-5(+) het-erozygote (Table 2, compare lines 10, 16 and 25, compare lines 9 and 24). In the mab-18 mab-21 double heterozygote there may have been a weak expression of the ray 6 to ray 4 trans-formation phenotype (Table 2, compare lines 14, 15, and 19). We conclude that egl-5, mab-18, and mab-21 lie on a common genetic pathway leading to expression of the characteristi morphology of ray 6. The action of mab-18 and mab-21 appears to be either downstream of egl-5 or confined to cells leading to ray 6, since if mutations in these genes acted in cells leading to ray 4 either by lowering egl-5 activity or raising mab-5 activity, suppression of the heterozygous effect of mab-5 on ray 4 would have been expected.

The remaining two genes have mutant phenotypes affecting several rays. The gene mab-20 is defined by one strong (bx24) and one weak (bx61ts) allele (Baird et al., 1991). The strong allele causes variable fusions of many rays, whereas in homozygotes for the weak allele the most frequent fusion was of ray 4 to ray 3, similar to the transformation seen in a mab-5 heterozygote (Table 3, line 1). In the mab-20/+;mab-5/+ double heterozygote containing the strong mab-20 allele, there was no enhancement of the mab-5 effect on ray 4 (Table 2, compare lines 10 and 28), and a possible suppression of the effect of egl-5 on ray 6 (Table 2, compare lines 4 and 26). In a background homozygous for the weak mab-20 allele, intro-duction of a single mab-5 null mutation resulted in a signifi-cant increase in the frequency of fusions involving rays 1 through 4 (Table 3, compare lines 1 and 2). Thus mutations in mab-5 and mab-20 are mutually enhancing, and this suggests that mab-20 and mab-5 have related functions in specifying ray identities. One possibility is that mab-20 mutations result in lowered activity of the mab-5 gene in the ray identity pathway.

In a mab-26(bx80) homozygote, fusions of many rays occur at high frequency (Table 3, line 3). bx80 is semidominant, causing 25% transformation of ray 4 to ray 3 in heterozygotes (Table 3, line 4). In the absence of further information about the nature of the bx80 mutation and the mab-26 loss-of-function phenotype, it is not possible to infer a role of the wild type mab-26 gene in specification of ray morphology. Never-theless, interaction between the bx80 mutation and mutations in mab-5 and egl-5 indicates that this mutation affects a process associated with the action of these genes.

As for mab-20, the data for mab-26(bx80) suggest that this mutation might lower the activity of mab-5. In the mab-5;mab-26(bx80) double heterozygote, the frequency of transformation of ray 4 to ray 3 was increased to 74% (Table 2, line 31), whereas in the egl-5;mab-26(bx80) double heterozygote it was reduced to 5% (Table 2, line 29). (On the assumption of inde-pendent action, the frequency of transformation of ray 4 to ray 3 in the mab-5;mab-26 double heterozygote was expected to be 33%). In the egl-5;mab-26(bx80) double heterozygote the frequency of transformation of ray 6 to ray 4 was decreased from 7% to 0% (Table 2, compare lines 5 and 29).

In all of these phenotypic effects on ray fusion, mab-26(bx80) and mab-5 null mutations have similar effects in het-erozygotes, with mab-26(bx80) having the stronger effect. Hence, one possible explanation for the mab-26(bx80) phenotype was that this mutation decreased mab-5 gene expression. However, with respect to expression of the ray sub-lineage, mab-26(bx80) differs from mab-5 null mutations: in a mab-5(0) homozygote all the V-rays are lost, whereas there is little or no ray loss in a mab-26(bx80) homozygote. Thus if the effect of mab-26(bx80) is to decrease mab-5 gene expression, it does so in a way that only affects specification of ray mor-phology. Alternative possibilities are that mab-26(bx80) acts at the same step or downstream of mab-5 in a ray identity pathway, or acts in an independent, parallel pathway.

The transformations in ray morphology observed in the various mutant backgrounds described above could be the result of altered gene function in seam cells or in cells outside the seam. In the next section we present evidence against gene action outside the seam. Within the seam, these genes could act in one or more of the terminal cells generated by the ray sublin-eage, in the ray precursor cells (Rn cells), or in cells at earlier stages in the seam lineages. We conclude that the important sites of altered gene action in these mutants lie within cells that contribute to or generate only single rays. We reach this con-clusion in part by considering the nature of the ray transfor-mations seen in mutants. If the function of ray identity genes was to determine the fates or potentialities of cells ancestral to multiple rays, mutations would be expected to transform a subset of adjacent, lineally related rays into a subset of rays characteristic of a different lineage branch. However, the ray transformations observed were not of this sort. Instead, single rays of one branch were transformed into a ray descended from a different lineage branch, while other rays of the first branch remained unaffected. Thus, in mab-5 mutants, transformations affected ray 3, descended from V6.pap, and ray 4, descended from V6.ppp, while other rays generated by these precursor cells (rays 2, 5, and 6) were unchanged. Thus these transfor-mations cannot be the result of transformation of V6.pap into V6.ppp, or vice versa. In the egl-5(lf) background ray 6, descended from V6.pppp, was transformed to ray 4, normally a descendant of V6.pppa. However, ray 5, also descended from V6.pppp, was not affected. Thus this phenotype is not the result of transformation of V6.pppp to V6.pppa. These transforma-tions are most easily explained as transformations of terminal branches of the lineage, and hence suggest that mab-5, egl-5, and their genetic modifiers act within the terminal branches.

We investigated whether gene function critical for specifying ray morphology is cell-autonomous within the terminal branches leading to single rays. In some systems, pattern formation genes act cell-nonautonomously to generate signals from one cell that influence the fates of neighboring cells (Greenwald and Rubin, 1992). In order to determine whether such signals pass between seam cells, we analyzed the effects of laser ablations of ray precursor cells and their parents (Table 4, lines 1–12). We found it was possible to eliminate ray precursor cells or their parents in all combinations without altering the morphology of the remaining rays. The morphol-ogy of the ray generated by a given branch of the lineage was independent of the morphology of neighboring rays. Similar results have been reported previously (Sulston and Horvitz, 1977; Sulston and White, 1980). This indicates that for the terminal branches of the ray lineages, defined as branches giving rise to single rays, signals emanating from a ray of a given type necessary for the fate of a neighboring ray are not present. These results argue that the identity of each ray is specified cell-autonomously in the terminal cells of the ray lineages.

We considered two other modes of intercellular signaling that might help to determine the properties of the cells in the ray lineages. Cell signals might pass between seam cells early in the lineage to restrict the potential of lineal precursors to generate a subset of rays. Alternatively, positional signals extrinsic to the seam might influence the fates of seam cells later in the lineage.

We obtained evidence in favor of early seam cell signaling and against extrinsic positional signals by carrying out ablations of seam cells earlier in the lineage. Sulston and White (1980) showed that ablation of V6 in early L1 resulted in a transformation of the V5 lineage to one resembling that of V6. We confirmed that such a transformed lineage generates rays of the expected wild type morphologies (Fig. 4B; Table 4, line 13). Therefore regulation is complete, affecting both the pattern of cell divisions as well as the morphological pattern of the rays generated.

Because they generate a V6 lineage, such transformed V5 cells do not produce a postdeirid (Fig. 1A). We asked whether regulation of the V5 lineage could occur at a later stage, after V5.p was committed to produce a postdeirid. We found that after ablation of V6.p late in the L1 larval stage (Table 4, lines 14 and 15), or V6.pp plus V6.pa early in the L2 larval stage (Table 4, line 16), the fate of V5.ppp could be transformed to one resembling V6.pap. That is, in these ablations, after all seam cells of the V6 lineage branch were removed, V5.ppp generated a lineage identical to that normally generated by V6.pap, and gave rise to rays 2 and 3 (Figs 4C, 5A). Thus the fate of V5.ppp is initially not determined, but is restricted to generate ray 1 rather than rays 2 and 3 by the presence of more posterior seam cells. Ablation of V6 lineage cells after division of V6.pa and V6.pp resulted in no regulation of V5.ppp (Table 4, lines 19, 20, 24-26). Likewise, we observed no regulation within the V6 lineage after early ablations (Table 4, lines 17, 18, 23, 25, and 26; Figs 4D,E, 5B).

In order to determine whether the effect of V6-lineage cells on the fate of V5.ppp was to provide a signal acting on V5.ppp to influence its fate, or was instead a passive effect on the positions of V5.ppp or its descendants, thereby causing them to receive certain seam-extrinsic positional cues and not others, we followed the positions of cells in the V5.ppp lineage after ablation of V6 cells. We found a lack of correlation between the position of a cell and the type of ray it generated. After ablation of cells of the V6 lineage, descendants of V5 moved gradually to more posterior positions. In spite of the fact that V5.ppppp reached the normal position of R4 or R5, it generated ray 3 (Fig. 6). Likewise, V5.pppap generated ray 2 in spite of the fact that it occupied the normal position of R3.

No hybrid lineages were observed. V5.ppp either generated ray 1 or rays 2 and 3; in no case did it generate ray 1 and ray 2, or ray 3 and ray 4. These results are more consistent with a trans-formation in cell fate at the level of V5.ppp, affecting the type of lineage and rays it could generate, than they are with the presence of seam-extrinsic positional signals acting later on ray precursor cells.

We were similarly able to show that the rays generated by the normal precursors of rays 2 and 3 (R2 and R3) did not depend on these cells occupying their normal positions along the body. Although the lineage and rays generated by V6.pap were not affected by ablation of V6.ppp (Table 4, line 26; Figs 4D, 5B), the positions of V6.pap descendants along the body axis were affected. Thus, we observed V6.pappp (R3) in the normal position of R4 or R5, yet its fate was not affected, and it generated ray 3 as expected (Fig. 6).

In order to test further whether ray morphology was deter-mined by signals early or late in the lineage, we sought a genetic method of uncoupling lineage, cell position, and ray morphology. The genes pal-1 and lin-22 affect the response of V6 and other seam cells to intercellular signaling (Waring et al., 1992). In pal-1;lin-22 double mutants, V6 generates a novel lineage giving rise to two rays (Waring and Kenyon, 1990) (Fig. 5C). We found that these two rays were invariably rays 2 and 3 (Fig. 4F). This was in spite of the fact that the two ray precursor cells were posterior of the positions normally occupied by R2 and R3 (Fig. 6). This observation is consistent with the interpretation that in the pal-1;lin-22 background, V6 or V6.p is restricted to a fate that can generate rays 2 and 3 only. It is not consistent with the presence of seam-extrinsic positional cues acting later to determine the type of ray generated by each lineage branch, unless such positional cues were also affected by pal-1 and lin-22 mutations.

The sensory rays of the C. elegans male tail have distinct mor-phogenetic properties. As a result, in the adult male the rays are arrayed at reproducible positions within the acellular fan. We have shown that the pattern formation mechanism that guides the generation of ray differences depends in part on the action of the HOM-C/Hox genes mab-5 and egl-5. It appears that the level of activity of these two genes in the terminal branches of the ray lineages is one of the critical factors that determines what type of ray develops from each branch. These results present the opportunity to study in greater detail the steps in the morphogenetic pathway, including the mechanisms that set the level of mab-5 and egl-5 gene activity within a developmental field, the way in which mab-5 and egl-5 activity specifies a positional value or developmental fate within the field, the nature of additional upstream and downstream regu-latory functions essential to this process, and the identities of regulated downstream genes that execute the morphogenetic program.

We present several lines of evidence that the site of action of mab-5 and egl-5 is within the terminal branches of the seam lineages, and that ray morphological identities are autonomously determined within these branches. First, ray transformations in mab-5 and egl-5 mutants were most simply interpreted as transformations of one terminal branch into another, suggesting the level of gene activity was critical within each branch. Secondly, ablations of seam cells after the mid L2 larval stage resulted in no regulation. As any ray type could develop in the absence of any other ray type, this result indicated an absence of signals emanating from a ray of one type to determine the fate of another ray. Thus determination of multiple ray types differs from determination of multiple cell types in the Drosophila ommatidium, and the terminal branches of the ray lineages do not constitute an equivalence group (Greenwald and Rubin, 1992). Thirdly, there was no cor-relation between the position of a ray precursor cell and the morphology of the ray it generated, suggesting an absence of essential positional signals emanating from outside the seam. Sulston and White (1980) similarly could find no evidence for essential morphogenetic signals from outside the seam. They carried out an extensive series of ablations in search of an ‘organizer’ or ‘inducer’ in the male tail analogous to the anchor cell inducer of the vulva (Kimble, 1981). Ablations of the gonad, tail neurons, and male-specific blast cells in different animals resulted in no evident effects on the rays. Much of the above evidence is negative, and therefore must be treated with caution. However, taken altogether the results were most con-sistent with a late, cell autonomous action of mab-5 and egl-5 in defining the morphology of the V-rays. The levels of activity of these two genes within the terminal branches of the ray lineages appears to convey a particular positional or lineal value to each cell individually, and to control, in an appropri-ate manner, the expression of downstream target genes.

How is positional or lineal value specified by the levels of activity of mab-5 and egl-5? Analysis of the effects of raising and lowering the level of activity or the number of copies of these genes suggests that in wild type their activity may be graded through the region of the tail where the V5 and V6 rays form. A gradient of mab-5 activity in the body has previously been suggested to define the ray domain in the tail (Kenyon, 1986). A graded pattern of expression of MAB-5 protein has been observed in ventral cord neurons (Salser et al., 1993), and expression of MAB-5 protein in V6 has been reported (Austin and Kenyon, 1994). Alternatively, it might be the requirement for mab-5 or egl-5 activity that is graded. Thus, a different component of the pathway might be present in a gradient, and this results in a variable threshold for mab-5 or egl-5 activity.

The presence of two graded molecules might provide the basis of a code of positional or lineal values determining the pattern of the V-rays. Indeed, we provide evidence that the ratio of the number of mab-5 and egl-5 gene copies influences the specification of the ray 6 and ray 4 fates. Since, as we have argued above, mab-5 and egl-5 probably act cell autonomously within the terminal branches of the ray lineages, it is the ratio of function of these two genes within the terminal ray 4 and ray 6 lineages that determines cell fate. We examined males mutant in a third HOM-C/Hox gene, lin-39, and found they possessed normal rays (data not shown). A code resulting from overlapping expression of HOM-C/Hox genes has been proposed to determine the identities of digits in the vertebrate limb (Morgan et al., 1992). Increase or decrease of the level of activity of individual HOM-C/Hox genes resulted in anterior or posterior transformations of digit identities similar to the anterior and posterior transformations of ray identities demon-strated here. Future studies should reveal whether the mecha-nistic basis of a HOM-C/Hox code is similar in these two very diverse systems.

The opposite effects of raising and lowering the number of copies of mab-5 and egl-5 suggests the possibility of a cross-competitive interaction between their gene products. Cross-competition could be at the level of competition for binding to common sequences in critical target promoters, which would thereby become responsive to the ratio of nuclear concentra-tions of MAB-5 and EGL-5 proteins. Or cross-competition could be due to the formation of inactive heterodimers. Alter-natively, cross-competition could be at the level of expression of the HOM-C/Hox genes themselves. mab-5 and egl-5 might be mutually repressing, so that increasing the number of copies of one of these genes results in lowered expression of the other. In this model, promoters of target genes would not bind both mab-5 and egl-5 gene products. Repression of mab-5 gene expression by egl-5 in a ventral cord blast cell has been demon-strated (Salser et al., 1993). In other systems, there is consid-erable evidence for cross-regulatory interactions between HOM-C/Hox genes (see McGinnis and Krumlauf, 1992).

It is helpful in determining the roles of regulatory genes to examine the ground state of the system in the absence of gene function. It appears that in the absence of egl-5 gene product, two V-ray identities are specified, one taken by ray 1, and the other taken by rays 2 through 5. This may represent the ground state generated by the remaining functions, and indicates that in the presence of all other gene functions, egl-5 has an essential role in specifying the differences among rays 2 through 5.

As no V-rays are generated in a mab-5 null mutant, the ground state in this background cannot be assessed. However, we have found that rays of correct morphology can be generated in the presumed absence of mab-5 gene product when a mutation at the lin-22 locus is also present. This result appears paradoxical in view of the effects of altered mab-5 gene function discussed above. One possibility is that the mab-5 allele used (e1239), which contains an altered splice site (Table 1), still allows some gene function, and this low level is elevated significantly in the lin-22 background. This expla-nation seems unlikely, because the phenotype of the mab-5;lin-22 double mutant does not resemble a mab-5 hypomorph. Whereas in a mab-5 hypomorph there is a higher frequency of loss of anterior V-rays than posterior V-rays, in the lin-22;mab-5 double mutant all rays were lost with similar frequency. If e1239 is in fact null, then the mab-5 gene product is not essential for the specification of multiple ray morphologies in a lin-22 mutant background.

Although a central role of transcriptional regulation by HOM-C/Hox genes in the generation of biological pattern has been implicated for over a decade, few genes that are HOM-C/Hox transcriptional targets are known, apart from the HOM-C/Hox genes themselves (Andrew and Scott, 1992). It was recognized early that region-specific cell recognition and adhesion was one crucial aspect of cell differentiation regulated by HOM-C/Hox genes (Garcia-Bellido, 1968; Morata and Lawrence, 1975; Gauger et al., 1985). Yet the mechanisms by which this is brought about are largely unknown. Our results suggest that differentiation of the rays in the C. elegans male tail may provide a promising opportunity to study this problem. In the specification of ray identities by mab-5 and egl-5, there is no change in cell lineage or obvious change in cell type. Instead, the morphogenetic properties of the rays, and most cogently the recognition functions of their constituent cells, appear to be altered. This raises the possibility that in the ray lineages genes encoding cell recognition functions may be direct targets of mab-5 and egl-5 transcriptional regulation.

We sought to determine whether genes we identified previously, as well as one new gene, acted in the same pathway as mab-5 and egl-5 in specifying ray morphology, and thus might be candidates for mab-5 and egl-5 target genes. Our approach was to look for evidence of genetic interactions. Mutations in mab-18, mab-20, mab-21 and mab-26 cause loss of distinct ray identities, in some cases in patterns similar to mab-5 and egl-5 mutations. We have isolated multiple, recessive alleles of mab-18, mab-20 and mab-21 in screens of relatively small numbers of worms (Baird et al., 1991). This argues that these mutations are loss-of-function alleles, and on this basis we conclude that these genes play a role in specification of ray morphology in wild type. We are unable to draw such a con-clusion for mab-26, based on the single semi-dominant allele. As this mutation may cause ectopic or neomorphic gene activity, the normal function of mab-26 could be unrelated to morphogenesis of the rays. Resolution of this issue awaits the isolation and study of additional alleles at this locus. In the dis-cussion below, the general conclusions apply in the case of mab-26 to the mab-26(bx80) mutation only.

We found that mutations in mab-18, mab-20, mab-21, and mab-26 were enhancers or suppressors of the effects of mab-5 and egl-5 mutations on rays. This suggests that these genes do not act independently of mab-5 and egl-5 in specifying ray morphology. The cellular milieu in which they act is altered when mab-5 or egl-5 gene expression is lowered, indicating a proximity in their function in the ray morphogenetic pathway. These results are consistent with any of these genes being tran-scriptional targets of mab-5 or egl-5. However, other possibil-ities remain, including roles in regulation of HOM-C/Hox gene expression, or as HOM-C/Hox cofactors (McGinnis and Krumlauf, 1992). Some dosage-dependent modifiers of the Drosophila HOM-C/Hox gene Antennapedia are known to participate in regulation of HOM-C/Hox gene expression (Kennison and Tamkun, 1988). Alternatively, mab-18, mab-20, mab-21, and mab-26 might act in one or more parallel pathways partially redundant with the one in which mab-5 and egl-5 act.

Pattern formation in seam cell lineages involves spatially restricted generation of postdeirid, alae, rays, and types of rays. It is possible that a general patterning mechanism governs the differentiation of all these structures. Kenyon and coworkers have studied the formation of alae, postdeirid, and rays, and have identified several genes essential to specification of these structures (Kenyon, 1986; Waring and Kenyon, 1990, 1991; Waring et al., 1992; Austin and Kenyon, 1994). They have proposed that multiple mechanisms contribute to generation of the axial pattern of the seam cell lineages. First, a graded signal of unknown nature extending along the entire length of the body appears to influence the tendencies of seam cells to generate rays versus alae. Second, signals between seam cells refine the pattern by inducing or inhibiting certain develop-mental pathways. Third, a prepatterning of seam cells, brought about by an unknown mechanism, affects their individual responses to cell signals. These three mechanisms are postu-lated to influence the differentiation of seam cells in part by regulating the cell-autonomous expression level of mab-5, which is expressed in a gradient with the highest level in the tail.

Our results on generation of the ray pattern are consistent with these proposals. We suggest that gradients of mab-5 and egl-5 are present in the tail region, and that the levels of expression of these genes influences the pattern of ray types in a cell-autonomous manner. We show that a signal from cells of the V6 lineage determines the types of rays generated by V5.ppp. Our evidence is against a role of seam-extrinsic posi-tional signals in defining the final ray fates. Waring and Kenyon (1990) likewise favored signaling between seam cells over seam-extrinsic positional signals in the ray versus alae decisions of the V4 and V5 lineages.

While giving an indication of the nature of the ray pattern formation mechanism, our results leave unanswered a number of questions. For example, we cannot yet explain how the complete pattern of V-rays is generated, and likewise do not address the generation of multiple ray types by the T lineage. We have described one additional gene, mab-19, that could play a role in the T lineage similar to the roles of mab-5 and egl-5 in the V lineages (Sutherlin and Emmons, 1994). Further, we do not know whether the relative expression levels of mab-5 and egl-5 provide all the information required to discrimi-nate up to 6 V-ray types; nor do we understand how such an expression pattern could accomplish this. How HOM-C/Hox genes convey positional identities to cells in general is not known. One possibility is that multiple transcription factor concentrations dictate multiple cellular responses. This property has been demonstrated for the product of the bicoid gene in Drosophila, which is also a homeodomain-containing transcriptional regulator (Driever and Nüsslein-Volhard, 1989). Alternatively, or in addition, combinations of tran-scription factors may give a cellular outcome different from the sum of the single factors acting separately, thus generating a code. Such a mechanism has been demonstrated for cell fate determination in the C. elegans ventral cord (Clark et al., 1993; Wang et al., 1993).

In other well-studied examples of cellular pattern formation, multiple mechanisms are superimposed to generate a repro-ducible spatial pattern of cell fates (Greenwald and Rubin, 1992). For example, in generation of the C. elegans vulva, a graded diffusable signal and two different neighbor inhibitory signals participate in the specification of three cell types (Horvitz and Sternberg, 1991). Likewise, in generation of the various organs of the Drosophila peripheral nervous system, gross patterns generated by one pattern formation system are refined by nearest-neighbor signaling systems (Heitzler and Simpson, 1991). Thus it seems likely that several mechanisms will also be involved in specification of the ray pattern.

We have shown that the types of rays generated by V5.ppp are determined by a signal from its posterior neighbors. We did not observe regulation indicative of signaling between other branches of the ray lineages, or at later times. It is possible that for these later lineages segregation of lineage branches with distinct potentialities occurs by a cell-autonomous mechanism. However, a cell-autonomous mechanism begs the question as to how asymmetric cell divisions come to be correctly oriented within the body. It seems more likely to us that signals between seam cells occur at all levels. In this case, we have simply been unable to demonstrate their presence. Our results indicate that after committing to produce a postdeirid, the V5 lineage is capable of undergoing regulation with respect to ray type for only a brief interval (on the order of one or two hours) (Table 4; data not shown). If essential signals pass between neigh-boring cells during a shorter interval than this, possibly imme-diately after cells are born, it may be impossible to prevent their action by a cell ablation experiment.

We suggest that ubiquitous signaling between neighbors within the seam results in the generation of non-equivalent precursor cells at each point in the ray lineages. The levels of activity of the HOM-C/Hox genes mab-5 and egl-5 are likely to be at least one component of this non-equivalence. Partly as a result of differing mab-5 and egl-5 activity levels, each precursor cell is capable of generating a subset of rays, and this capability is continuously segregated into daughter cells until the unique single lineage branches are generated.

We are grateful to S. Baird, N. Baker, D. Fitch, D. Stein, and L. Stephens for their comments on the manuscript, and to members of the laboratory for helpful discussions. We thank L. Avery, A. Chisholm and S. Salser for sending nematode strains. Additional strains were received from the Caenorhabditis Genetic Center, which is funded by the NIH National Center for Research Resources (NCRR). K. L. C. was supported by a Martin Foundation Postdoc-toral Fellowship. This work was supported by NIH grant R01 GM39353.