Autonomy and nonautonomy of sex determination in triploid intersex mosaics of C. elegans

In the soma of adult C. elegans, more than 30% of the 959 cells in hermaphrodites and 40 % of the 1031 cells in males make up tissues that exhibit sexual dimorphism or sex-specific biochemical differences. These tissues include the hypodermis, musculature, intestine, somatic gonad, germ line and nervous system (Sulston and Horvitz, 1977; Sulston et al. 1980; Kimble and Sharrock, 1983). The hermaphrodite has a simple tail, a vulva arising from the midventral hypodermis, and a symmetrical gonad with an anterior and a posterior reflexed arm, in which the germ line produces sperm during the fourth larval (L4) stage and oocytes in the adult. In addition, the hermaphrodite intestine late in the L4 stage begins to produce vitellogenins, which are secreted into the pseudocoelom and taken up by developing oocytes during oogenesis (Sharrock, 1983; Kimble and Sharrock, 1983; Blumenthal et al. 1984). The male has a more complex tail and accompanying specialization of the nervous system for copulation, in which the male is the active partner. The male gonad consists of a single reflexed arm, and the germ fine produces sperm only. The male intestine does not synthesize vitellogenins.

The primary signal that determines sex in C. elegans is the ratio of the number of X chromosomes to the number of sets of autosomes (X/A ratio). Wild-type diploid hermaphrodites have two X chromosomes and two sets of autosomes (2X;2A), whereas diploid males have only one X chromosome (1X;2A). 2X;4A tetrapioids and 2X;3A triploids develop as fertile males. 3X;4A tetrapioids develop as fertile hermaphrodites, as do 4X;4A tetrapioids and 3X;3A triploids. Thus X/A ratios of ≧ 0.75 lead to hermaphrodite development, and ratios of g ≦ 0.67 lead to male development (Nigon, 1951; Madl and Herman, 1979).

The X/A ratio determines the activities of a set of interacting regulatory genes (Fig. 1), which ultimately dictate the activity of the tra-1 gene, the terminal regulator of sexual differentiation (Hodgkin, 1988; Villeneuve and Meyer, 1990). The activity of the tra-1 gene is required cell autonomously for hermaphrodite development (Hunter and Wood, 1990). However, it is not known whether the sex determination process that sets the state of tra-1 is executed autonomously by each cell in the embryo, or whether information regarding X/A or the states of the upstream sex determining genes is communicated between cells. We have approached this question by artificially setting the X/A ratio to approximately 0.7, between the values that unambiguously dictate male or hermaphrodite development, so that the animals develop as mosaics of male and hermaphrodite tissues (Mad1 and Herman, 1979). In these animals, we have examined the resulting patterns of sexual differentiation at the cellular level. If sexual identity decisions (tra-1 ON or OFF) are made autonomously then clones of lineally related cells will express the same sexual phenotype. Alternatively, if sexual identity decisions involve non-autonomous inter-actions then non-clonally related cells will consistently express identical sexual phenotypes. If decisions must be made early, then clones expressing the same sexual phenotype will be large; if the decision can be reversed or postponed until later in development, then clones may be small.

This approach was taken earlier with Drosophila (Bridges, 1925), in which sex is also determined by the X/A ratio, normally 2X:2A in females and 1X:2A in males. Triploid animals with two X chromosomes (2X:3A) develop as intersexuaf mosaics, with, large patches of male and female tissue in the same individual. This observation is consistent with an ambiguous X/A signal that is assessed autonomously by individual cells or groups of cells fairly early in development, at about the cellular blastoderm stage, such that sexually determined progenitors give rise to patches of contiguous, similarly sexed progeny cells (for review see Baker and Belote, 1983). More recent work indicates that assessment of the X/A ratio involves irreversibly setting the expression state of the cell-autonomous Sex lethal (Sx/) gene. In addition to regulating downstream sex determination genes, Sxl also sets the level of X chromosome expression, ensuring that similar levels of X-linked gene products are made in both XY and XX cells (Cline, 1984).

In C. elegans, Madl and Herman (1979) showed that if 2X;3A animals carry one copy of an X duplication representing about 20% of the chromosome, then they exhibit intersexual phenotypes. If such animals carry a smaller X duplication, then they develop as morpho-logically normal males, like 2X;3A animals without a duplication; if they carry an even larger duplication, they develop as complete hermaphrodites, like 3X;4A animals. These authors observed wide variability of sexual phenotypes in the intersexual animals, ranging from nearly fully hermaphrodite to nearly fully male, with examples of mosaic individuals. They interpreted this result to indicate that assessment of the X/A ratio could be made independently in different tissues during development. In earlier experiments we confirmed the production of several classes of mosaic intersexes in 2X;3A strains carrying duplications of about 20% of the X chromosome (Wood et al. 1985). We present here a more extensive analysis of such intersexual animals, whose patterns of mosaic development provide information regarding both the cell autonomy and timing of sex determination decisions.

The wild type of Caenorhabditis elegans var. Bristol (designated N2) used in these studies was originally obtained from the Cambridge England strain collection (Brenner 1974). Robert Herman (University of Minnesota) provided the tetrapioid strain SP344, of genotype dpy-11 (e224)\;unc-3(el51)X, and the two diploid strains SP75, of genotype mnDp25(X:T);unc-3(el51)X, and SP116, of genotype mnDp9(X:l);unc-3(el51)X. The duplications mnDp25 and mnDp9 each represent approximately 20% (genetically) of the X chromosome and are actually translocations stably attached to linkage group (LG) I (Herman et al. 1979).

Genetic nomenclature in this paper conforms to established conventions for C. elegans (Horvitz et al. 1979), with the following exception: because the convential representation of karyotypes (nA;mX) leads to confusion when discussing X/A ratios, we have reversed this representation to nX;mA, as used for Drosophila karyotypes; we urge that this latter convention be adopted for C. elegans as well.

General culture methods have been compiled by Sulston and Hodgkin (1988). To obtain synchronized animals, embryos isolated by hypochlorite treatment of gravid hermaphrodites were plated onto NGM plates seeded with OP50 and allowed to hatch at 20 °C. Every hour the plates were gently flooded with 5 ml of M9 salts to dislodge the newly hatched first-stage larvae, which were removed by aspiration and replated to provide synchronous populations.

Developmental age of worms was determined by microscopic (Nomarski) examination of gonads to determine stage of germ cell maturation, number of somatic gonad cells and number of vulval precursor cells (Hirsh et al. 1976; Wolf et al. 1978).

Triploid intersex animals were produced as follows, where Dp indicates either mnDp9 or mnDp25, both of which carry unc-3 (+)X (Madl and Herman, 1979). N2 males were mated to DpiDp,unc-3/unc-3 hermaphrodites, and the male progeny, of genotype Dp/ + ;unc-3/0, were mated to tetrapioid dpy-ll;unc-3 hermaphrodites. The latter mating yields the following classes of non-Dpy outcross progeny (in addition to Dpy Une tetrapioid hermaphrodite and occasional male self progeny):

1. Une triploid hermaphrodite (3X;3A, not carrying Dp), of genotype dpy-11 /dpy-11 /+;unc-3/unc-/unc-3,

2. Une triploid males (2X;3A, not carrying Dp), of genotype dpy-11 /dpy-11 /+;unc-3/unc-3/0,

3. non-Unc triploid hermaphrodites (3X;3A, carrying Dp) of genotype Dp[+/+;dpy-ll/dpy-11 /+;unc-3/unc-3/unc-3, and

4. non-Unc triploid males, intersexes, and possibly hermaphrodites (2X;3A, carrying Dp) of genotype Dp/+/ +,dpy-l 11 dpy-11/+;unc-3/unc-3 /0.

Non-Unc progeny identified in the dissecting microscope as males or intersexes (class 4) on the basis of morphology were picked for analysis by in situ hybridization and Nomarski microscopy. Note that non-Unc externally hermaphrodite animals of class 4, if present, could not have been distinguished from those of class 3 and therefore would not have been picked. As a result the population of triploid intersexes analyzed could have a bias against fully hermaphrodite phenotypes.

Triploid intersex animals were examined by Nomarski microscopy to determine the phenotypic sex of the somatic gonad, germ line, tail hypodermis and ventral hypodermis (presence or absence of vulva) based on gross morphology. Gonads with two reflexed arms were scored as hermaphrodite; those with a single arm were scored as male. Germ cells that were large, round, with prominent nuclei and usually arranged in single file in the proximal gonad were scored as oocytes. Small round germ cells with very compact nuclei were scored as sperm. A protrusion of the ventral hypodermis near the middle of the animal was scored as a vulva; no distinction was attempted between functional vulvae and nonfunctional protrusions of vulval cells. (Although the signal for vulva formation originates from the somatic gonad, the response of the ventral hypodermis appears to be sex-specific, resulting in a functional vulva if the hypodermis is female and a non-functional protrusion of hypodermal cells if it is male; Hodgkin, 1987.) Tails with a hermaphrodite-specific tail whip, either full length or stunted, were scored as hermaphrodite. Tails with any male-specific rays or fan structure were scored as male.

Intestines and gonads were dissected from adult animals according to the method of Kimble and Sharrock (1983), mounted on a subbed (0.1% gelatin, 0.01% potassium chromate) hybridization slide, lightly squashed with a coverslip, and frozen in liquid nitrogen as described by Edwards and Wood (1983). The coverslip was removed and the frozen tissues adhering to the slide were fixed in either ethanokacetic acid (3:1) at 4°C for 15 min or 4 % paraformaldehyde at 4°C for 15 min, and then dehydrated in ethanol.

Fixed tissues were hybridized to either a nick-translated plasmid DNA or a primer-extended purified fragment of either the vit-5 or vit-6 gene probes. The C. elegans vitellogenin clones used in this study, pACYC184 and pl3F9 – 2, were gifts from T. Blumenthal (Indiana University, Bloomington). pACYC184 contains a 1.3 kb EcoRI-Wzndlll fragment of the vit-5 5 ′ -upstream sequence and coding sequence cloned into PUC-8. pl3F9 – 2 is a 1.1 kb EcoRI-Hizidlll fragment of the vit-6 coding region cloned into PUC-8. The hybridization was carried out as described by Albertson (1984) with the following modifications. Probes were labelled with ³⁵S (10⁷ – 10⁸ctsmin⁻¹μ g⁻¹) rather than biotin and were resuspended in 70% formamide, 0.4 M NaCl, 1.6mM EDTA, 0.04M Na2HPO₄, 10mM DTT and 5 × Den-hardt’s solution at 100 – 200 μ l final volume. Probe solutions were heated at 65 °C for 5 min and quick-cooled on ice for 1 min. Approximately 10⁶ctsmin⁻¹ in 10 μ l of hybridization solution was applied to each sample, spread over the slide with an 18-mm round coverslip, and put in an air-tight box equilibrated with 70% formamide, at room temperature overnight. Slides were given 5 washes of 15 min each in 70 % formamide, I×SPE (0.165M NaCl, 20mM NaH₂PO₄, 1mM EDTA) one wash for 15min in I × SPE, and two more for 5min each in I×SPE, dried, and mounted for autoradiography.

Samples were mounted and autoradiographed according to Edwards and Wood (1983). Exposure time was 1 day to 2 weeks depending on specific activity of probe. To make cell nuclei visible by microscopy, developed slides were stained with 1 μ g ml⁻¹ aqueous solution of the DNA stain diamidino-phenylindole dihydrochloride (DAPI, obtained from Boehringer-Mannheim).

A Zeiss photomicroscope was used to photograph the slides. Autoradiographic grains were viewed either as bright spots on a black background using dark-field optics or as dark spots on a white background using bright-field optics. DAPI-stained nuclei were observed under 365 nm epi-illumination. Cell nuclei on DAPI-stained slides could be seen simultaneously with autoradiographic grains by viewing with both visible dark-field and 365 nm epi-illumination. In dissected, DAPI-stained intestines, cell pairs int2 through int9 were identified by counting nuclei posterior to inti. The inti quadruplet was identified either by its association with the pharynx or by the four characteristic inti nuclei. Only animals with 12 or more dissected intestinal nuclei were scored to obtain the data presented for in situ hydribidization experiments.

To analyze intersexual animals, we scored sex-specific differentiation in five tissues derived from four of the six embryonic founder cells: AB, MS, E, and P₄, which are generated during the first four rounds of cleavage division in embryogenesis (Fig. 2). The tail ectoderm, midventral hypodermis, somatic gonad and germ line we scored by Nomarski microscopy on the basis of morphological differences described in the Introduction and summarized in Table 1. In addition we could score the intestine for synthesis of vitellogenins, using cloned vit gene fragments (kindly furnished by T. Blumenthal, Indiana University) as probes in an in situ hybridization assay for presence of vitellogenin transcripts in individual cells. Of the six members in the vitellogenin gene family vit-1-vit-6, the vit-5 probe hybridizes to transcripts of the highly conserved vit-l-vit-5 sub-family, while the vit-6 probe is specific for the more divergent vit-6 transcripts (Blumenthal et al. 1984; Spieth and Blumenthal, 1985; Heine and Blumenthal, 1986).

To demonstrate the sex and tissue specificity of vitellogenin mRNAs, we hybridized the vit-5 probe to dissected intestines and gonads of both diploid and triploid male and hermaphrodite animals as described in Materials and methods. Among a total of 145 diploid (2X;2A) and 40 triploid (3X;3A) adult hermaphrodites, all but a single animal showed strong hybridization of the labeled probe to all cells in dissected portions of the intestine and no hybridization to the gonad (Fig. 3A and Table 2). In contrast, a total of 55 diploid (1X;2A) and 45 triploid (2X;3A) adult males showed no signal over either tissue (Fig. 3B and Table 2). These results confirm the sex specificity of vitellogenin gene expression observed by Blumenthal et al. (1984) using other methods. They also show that the tissue specificity of vitellogenin protein synthesis observed by Kimble and Sharrock (1983) reflects a corresponding specificity of vitellogenin gene transcription.

We obtained animals with an X/A ratio of about 0.7 as described in Materials and methods, by crossing diploid (1X;2A) males carrying an autosomally attached duplication of the right arm of X (mnDp9 or mnDp25), representing about 20% (genetically) of the chromosome, to marked autosomally tetrapioid hermaphrodites (4X;4A). We picked all unmarked progeny showing any discernible masculine characteristics as representing 2X;3A animals carrying the duplication (2X;3A+Dp). The markers used allowed us to distinguish these from all other progeny genotypes, except that externally fully hermaphrodite 2X;3A+Dp animals, if present, would have been indistinguishable from 3X;3A hermaphrodites and would not have been picked. Therefore, the populations of triploid intersexes analyzed may be biased against hermaphrodite phenotypes. We have pooled results obtained with mnDp9 and mnDp25 in the experiments described below, since both duplications resulted in similar degrees and variability of masculinization.

We scored tissue sexual phenotypes in a total of 386 animals identified according to the above criteria as triploid intersexes. Of these, 186 were, live animals, scored by Nomarski microscopy for hypodermal structures, gonad and germ line only; the remaining 200 were dissected and fixed for hybridization. In most of the fixed animals, we were able to score one or more of the above tissues by morphological criteria, in addition to scoring the intestine for vitellogenin transcripts.

The five tissues scored in these animals generally showed clear predominance of either male or hermaphrodite differentiation. Many animals were clearly mosaic, with both male and hermaphrodite tissues in the same individual. For example, Fig. 4A shows an animal whose intestine contains vitellogenin transcripts, yet whose single-armed gonad and germ line appear fully male. Fig. 4B shows the converse: an animal with a male intestine (no vitellogenin synthesis) and a female germ line. The germ line is scored as female because the germ cells are large, have large nuclei, and are ordered in a single row, all traits specific to oocytes. The oocytes are smaller than normal, possibly due to lack of vitellogenins.

The population of intersexes displayed a variety of sexual phenotypes, ranging from morphologically fully male to almost fully hermaphrodite. Between these extremes, we observed animals displaying all possible pairwise combinations of differently sexed tissues but one, as shown in Table 3. We conclude that in general the sex determination decision of one tissue is not dictated by the sex of another, that is, decisions appear to be tissue-autonomous. The ventral hypodermis was exceptional in that vulva formation correlated essentially completely with the sexual phenotype of the somatic gonad. In 208 of 209 animals, the presence of an hermaphrodite (two-armed) gonad was associated with the presence of a vulva, and the presence of a male gonad with no vulva, as expected from the known induction of vulval development by the anchor cell of the hermaphrodite gonad (Kimble, 1981; Sternberg and Horvitz, 1986).

Among the animals in the intersex population examined, different tissues showed different frequencies of male and hermaphrodite phenotypes (Table 4). For example, the percentages of animals showing the hermaphrodite phenotype for intestine, germ line, and somatic gonad were 62 %, 39 % and 29 %, respectively. Thus tissues appear to differ in their response to an intermediate primary sex-determining signal (see Discussion).

Most animals in the intersex population appeared to be predominantly male or predominantly hermaphrodite. To quantitate this apparent correlation between tissue sexual phenotypes, we compared the observed frequencies of various pairwise combinations with the frequencies expected assuming complete independence (Table 5), calculated from the data in Table 4. Correlations are apparent; for every pair of tissues, there is a clear bias toward the same-sex and against the oppositesex combination, showing that although tissue sexual identity decisions can be made independently (Table 3), these decisions are often correlated in individual animals. Differences in the degree of non-correlation are as expected from the apparently different responsiveness of the tissues described in the preceding section; for example, hermaphrodite intestine-male gonad is much more frequent than male intestinehermaphrodite gonad.

We examined 163 fixed intersex animals that had been dissected so that at least half of the intestinal cells were exposed, a sufficient number of cells to detect progeny from all terminal intestinal cell divisions (see Fig. 5). Among these animals, 26 exhibited mosaicism within the intestine; that is, some cells showed the high level of labeling characteristic of hermaphrodites and other cells showed no labeling (Table 2). Fig. 6 depicts such a mosaic intestine. The anterior end, up to and including the cell pair designated int4, shows hybridization, while the adjacent pair (int5) and cells posterior to it show none (see Fig. 5 for nomenclature of intestinal cells).

The patterns of intestinal mosaicism in these animals showed a consistent anterior-posterior polarity. In all 26 mosaics the intestinal cells, beginning at the anterior of the intestine, showed labeling to a certain point, beyond which no cells were labeled. In addition, labeled cells were always contiguous. The position of the boundary (/) between labeled and unlabeled cells varied among different individuals from int2/int3 to int6/int7. Cells within an intestinal pair also occasionally differed; Fig. 6C shows an intestine in which only one cell of the int4 pair contained vitellogenin transcripts. In some mosaic intestines, the level of labeling in individual hybridizing cells appeared graded, again always with the heavier labeling toward the anterior. This phenomenon was observed with both the generic vit-5 and the gene-specific vit-6 probes. It may represent the first case of truly intersexual cellular differentiation described in C. elegans.

In an attempt to control for an alternative explanation for the observed patterns, such as a general lack of metabolic activity in the posterior intestinal cells, we assayed intestines of triploid intersex and control animals for expression of two non-sex-limited markers, gut-specific esterase (Edgar and McGhee, 1986) and an intestinal antigen stained by the monoclonal antibody SP37 (S. Strome, personal communication). We found no mosaic intestines among 200 triploid intersexes scored for gut-specific esterase and 100 scored for expression of the intestinal antigen; both markers were expressed uniformly throughout the intestine in all animals.

To ask if the observed polarity of transcript presence in mosaic intestines might reflect a polarity in the initiation of vitellogenin gene expression, we examined the onset of vit transcript accumulation in normal diploid hermaphrodites. Vitellogenins are first produced in the hermaphrodite intestine in late L4 larvae, just before the onset of oogenesis (Sharrock, 1983; Kimble and Sharrock, 1983; Blumenthal, 1984). To determine whether vitellogenin transcripts are also first produced at this time, we dissected synchronized N2 larvae at various stages and assayed them by in situ hybridization with the vit-5 probe.

We could first detect vitellogenin transcripts in late L4 larvae, shortly after condensation of the germ-line nuclei undergoing spermatogenesis in the proximal arm of the gonad (Table 6). However, the transcripts appeared with a polarity opposite to that observed in adult mosaic intersexes. In animals assayed at the time of onset, only the posterior cells of the intestine were generally labeled; the most anterior cell pairs, inti and int2, were consistently unlabelled (Fig. 7). At slightly later stages, all cells were labeled, but a gradient of labeling intensity was still apparent until the L4/adult molt, when the labeling became uniform. Vitellogenins, assayed by gel electrophoresis of proteins in parallel experiments with the same synchronized populations, became detectable about two hours after the first vitellogenin transcripts (data not shown).

In the preceding analyses, we scored tissues other than the intestine only as hermaphrodite versus partially or completely male, with no attempt at detailed characterization. To detect possible internal mosaicism in these tissues, we examined over 300 additional intersex mosaics by Nomarski microscopy. We scored these animals for all the sex-specific structures deriving from the postembryonic blast cells, listed in Table 7 (Sulston and Horvitz, 1977; Sulston et al. 1980; 1983). In addition to the germ line, clonally derived from P₄, and the somatic gonad, derived from the MS lineage, these structures include the sex muscles (scored using polarizing optics), derived postembryonically from the M blast cell of the MS lineage, and several ectodermal structures derived postembryonically from blast cells of the AB lineage: the vulva in hermaphrodites and various components of the tail in males (Figs 3 and 7; see Hunter and Wood, 1990 for more detailed lineage diagrams).

In general the results corroborated those observed for the whole organ phenotypes described above; however, many of the lineages scored in these experiments diverge during late embryogenesis and larval development, increasing the resolution of the analysis. The structures deriving from a given blast cell were usually clearly hermaphrodite or male (Table 7). Although we again observed a strong preference for same-sex decisions among different blast-cell-derived structures in individual animals (Table 8), we also found instances of all pairwise combinations of sexual phenotypes, indicating that the postembryonic blast cells can make independent sexual identity decisions. For example, the lateral (V5, T) and ventral (P3-8, PIO) hypodermal cell lineages diverge at about the 300cell stage (Sulston et al. 1983). Since the structures derived from these cells can express different sexual phenotypes, these cells must be able to make sexual identity decisions as late as the end of the cell proliferation stage of embryogenesis.

Some of the phenotypes that we observed for the tissues and structures in Table 7 were intersexual, suggesting that sexual identity decisions can be made during postembryonic divisions in at least some blast cell lineages. In some cases the mosaic phenotypes suggested that, as in the intestine, anterior-posterior position can affect the sexual identity of cells that divide postembryonicaliy. For example, mosaicism within the germ-line was observed in five animals, all of which had the normal bilobed hermaphrodite somatic gonad morphology. In normal hermaphrodite development the germ-line precursor cells Z2 and Z3 undergo many rounds of proliferative division, beginning in the LI larval stage. At first the descendants of Z2 and Z3 intermix, so that each gonadal lobe contains germ cells descended from both blast cells In each lobe the first 150 germ cells to mature become sperm and those remaining mature as oocytes. The switch from spermatogenesis to oogenesis occurs just after the L4 to adult molt, slightly earlier in the anterior lobe than in the posterior lobe. In four of the mosaic animals the anterior lobe showed hermaphrodite germ-line differentiation into oocytes and sperm, while the posterior lobe showed male differentiation into sperm only. In the remaining mosaic, the anterior lobe contained only oocytes, while the posterior lobe contained oocytes and sperm. These animals were scored as mature adults, well after the switch to oogenesis. Thus the germ-line mosaics, like the intestinal mosaics, appear to show a consistent polarity with an anterior-hermaphrodite, posterior-male bias, as if the sex determination decision were positionally influenced. Moreover, germ-line sexual identity can be determined postembryonically, since Z2- and Z3-derived cells in the two gonadal lobes can make opposite decisions.

Intersexual or abnormal somatic gonads were observed in 17 animals. The predominant phenotype was a sexually indeterminate gonad consisting of a large sac, as if both distal tip cells or the linker cell failed to migrate (Fig. 8). Several intersexes showed morphologies that could be interpreted as gonad mosaicism, in which one lobe apparently initiated normal hermaphrodite development while the other failed to elongate, resulting in a single-lobed hermaphrodite gonad. Such a structure could result if one of the two somatic gonad precursor cells (Z1 and Z4) and its descendants adopted a hermaphrodite fate while the other adopted a male fate. In five of seven animals with mosaic gonads the anterior Z1 descendants apparently followed hermaphrodite fates and formed an anterior lobe while the posterior Z4 descendants apparently followed the male fate of no posterior migration, consistent with the gonad also being sensitive to positional influences.

Two exceptional animals scored as having a male gonad and a vulval protrusion in the mid-ventral hypodermis (one listed in Table 3 and another in Table 7) are likely to be gonad mosaics that developed a hermaphrodite anchor cell and male linker cell. The hermaphrodite anchor cell induces vulval formation (Kimble, 1981; Sternberg and Horvitz, 1986) while migration of the male linker cell leads the elongation of the developing gonad to generate the characteristic male morphology (Kimble and White, 1981).

The AB-derived tail structures scored arise from eight blast cells, which in the male produce the structures listed in Table 7 (Sulston and Horvitz, 1977; Sulston and White, 1980). In hermaphrodites these eight cells either do not divide or produce fewer progeny that generate less specialized structures. In the triploid intersexes, most of these blast cells exhibited either normal male or normal hermaphrodite fates. In individual animals all pairwise combinations of blast cell clones showed strong biases toward same-sex decisions (Table 8). However, exceptions were observed for every pair, arguing for the possibility of autonomous decisions and against obligate inductive effects between any of the sexually dimorphic structures scored.

The T and V6 blast cells produce tail sensory rays in males and posterior extensions of lateral alae in hermaphrodites. In our analysis we detected apparently intersexual fates for these blast cells (Table 7), seen generally as presence of some but not all the rays derived from a particular blast cell. These cases might simply represent instances of incomplete male differentiation. Alternatively, however, they could result from sexual mosaicism within these lineages. Since some of the progeny cells that apparently made opposite sexual identity decisions are generated during the L3 stage of postembryonic development, these decisions may be made near the point at which sex-specific cell differentiation begins.

Sex muscle development was often incomplete or sexually ambiguous (Table 7), and many intersexual animals appeared to lack sex muscles completely. The most common muscle phenotype observed among intersexes was both partial male and partial hermaphrodite sex muscle development. The hermaphrodite sex muscles scored form a crossed pattern around the vulva and function in egg-laying; the male sex muscles form a diagonal parallel array at the base of the tail that functions in mating, as well as a second group required for normal morphogenesis and movement of the spicules (Sulston and Horvitz, 1977; Sulston et al. 1980). The sex muscles arise postembryonically, beginning with division of the M blast cell during the LI stage to produce two sex myoblasts in hermaphrodites and six in males (Fig. 8). The hermaphrodite myoblasts migrate anteriorly during the L2 stage to a point near the developing gonad, where they divide during the L3 stage and later differentiate into vulval and uterine muscles. The male myoblasts begin to migrate posteriorly, divide during the L3 stage, and later differentiate to form the male diagonal sex muscles (Sulston and Horvitz, 1977). In one mosaic animal, we observed distinctly male diagonal sex muscles located midventrally at the hermaphrodite position. This phenotype suggests that sex myoblasts reversed their sexual identity during development, first migrating in the hermaphrodite mode and then differentiating in the male mode.

By constructing and analyzing triploids carrying a partial duplication of the X chromosome, we have extended the observations of Madl and Herman (1979) that intersexual phenotypes result from an X/A ratio of about 0.7, between the values known to signal normal male and hermaphrodite development, respectively. Our results can be summarized as follows. The animals are generally healthy, indicating that the X-chromosome dosage compensation mechanism can accommodate intermediate X/A ratios. The intersexual animals are mosaics of male and female tissues and cells within tissues. For any pairwise combination of tissues or cells there is a clear bias toward same-sex choices in individual animals. As expected, this bias is absolute for the pair of tissues somatic gonad-midventral hypodermis, since the hermaphrodite gonadal anchor cell is required to induce vulval development in the midventral hypodermis (Kimble, 1981; Sternberg and Horvitz, 1986). However, for all other combinations of tissues and cells, the bias is not absolute, and opposite-sex decisions were observed. The mosaic patterns indicate how late in development these decisions can be made for the various cells and tissues scored. In at least two tissues, intestine and germline, the decisions appear to show global positional influences, so that mosaicism within these tissues shows a consistent anterior-feminine, posterior-masculine polarity.

Our finding of all but one of the possible pairwise combinations of opposite-sex phenotypes among the tissues examined argues that in each of the sexually dimorphic embryonic lineages, with the expected exception noted above, the sexual identity decision can be made autonomously, after these lineages have diverged. However, several of our observations are not consistent with a simple model, such as that generally accepted for Drosophila, involving autonomous irreversible assessments of the X/A ratio in the early embryo that commit large clones of cells to one sexual identity or the other in triploid mosaic animals. Rather, our results suggest the possibility that sexual identity decisions in C. elegans can be influenced by cell interactions until quite late in development. We discuss this possibility and other interpretations below, in connection with our observations on the same-sex bias among tissues, the local same-sex biases among cells within, certain tissues, the apparent global positional influences on sexual identity, and the timing of sexual identity decisions.

One plausible explanation for the same-sex preference among tissues in an individual is that although the actual chromosomal X/A ratio is identical in each embryo, its initial assessment, giving rise to the primary signal, is imprecise in triploid intersexes, such that individual embryos are likely to have either a male or hermaphrodite bias. This would affect the sex determination decisions in all tissues; those likely to go against the bias would be the apparently most responsive to X/A increase in male-biased embryos and least responsive in. hermaphrodite-biased embryos (see below) as in fact we have observed. The initial biases could reflect variations in maternal contributions to the assessment mechanism, for which there is evidence although its nature is not understood (Villeneuve and Meyer, 1987; Plenefisch et al. 1989; Villeneuve and Meyer, 1990; R. Herman, personal communication). A similar same-sex bias of tissue phenotypes has been observed by Villeneuve and Meyer (1990) in individual intersexual animals resulting from mutations in the sdc-1 gene.

An alternative or additional effect contributing to same-sex biases could be some non-autonomy in the sex determination process, such that for example a hermaphrodite decision in one tissue could influence other tissues toward the same decision. The finding of samesex biases in the hermaphrodite direction despite a probable overall bias against hermaphrodite phenotypes in the population of animals analyzed would be consistent with both of these explanations.

Two other possible causes of same-sex bias seem less likely. One would be the influence, at threshold X/A levels, of ‘modifier genes’, which segregate in different proportions to different 2X;3A+Dp offspring in the crosses producing these animals. Although such effects are known to contribute to individual biases in Drosophila 2X:3A intersex mosaics (Baker and Belote, 1983), they should not be a factor for C. elegans, in which all genetic stocks derive from the ancestral N2 strain and should, therefore, be generally isogenic as well as homozygous at almost all loci. Another possibility would be that among the progeny of tetrapioids, which are karyotypically unstable, there may be differences in autosome composition between animals, resulting in actual X/A differences that could account for the observed biases. Since some degree of autosomal aneuploidy is known to be tolerated in C. elegans (Sigurdson et al. 1986; C. P. H. unpublished), we cannot rule out this possibility.

An additional puzzling observation at the tissue level in triploid intersex mosaics is the apparent difference in responsiveness of different tissues to feminization in response to the intermediate X/A ratio. These differences could be related to the nature of the various responses: feminization of the intestine as scored in our experiments could result from expression of a single vit gene, whereas feminization of the tail or somatic gonad would require alteration of a more complex morphogenetic process controlled by many genes. Alternately, responsiveness differences could be only apparent, resulting from biases in the scoring of different tissues as male or hermaphrodite. In the intestine, the tissue found most responsive, an intersexual phenotype, that is an intermediate level of vit transcription might well have been scored as hermaphrodite, whereas in tail and somatic gonad, the tissues found least responsive, an intersexual phenotype would probably be scored as male. However, it is also possible that intermediate X/A ratios cause intermediate levels of tra-1 activity (Fig. 1), and that different tissues have different thresholds for response to tra-1. If so, this response must in general be all or none; we did not observe clearly intersexual cellular phenotypes except in mosaic intestines, where cells at the border between vit gene expression and non-expression often showed intermediate hybridization levels.

At the local level, biases toward same-sex choices of cells within tissues is difficult to explain except by nonautonomy in the sex determination process. In the intestine, non-autonomy of the decision to express vit genes is indicated by our finding that all labelled cells were contiguous with each other, as were all unlabelled cells. This contiguity of labeled cells is unlikely to be an artifact of leakage or transfer of vit transcripts between cells, because individual unlabeled cells among labeled neighbors (and vice versa) can be clearly seen using similar procedures in the intestines of animals mosaic for tra-1 gene function (Hunter and Wood, 1990). The absence of individual unlabeled cells among labeled neighbors and vice versa rules out the possibility that clonally inherited commitments are made during generation of the intestine, because of a peculiarity of the intestinal lineage as determined by Sulston et al. (1983). The 20 cells of the intestine arise from the E founder cell in the embryo by the lineage diagrammed in Fig. 5. Because two cells on each side of the animal exchange places with their neighbors between the 16- and 20-E-cell stages of intestinal development, the descendants of certain precursor cells do not occupy adjacent positions in the adult tissue. For example, descendants of the Ea cell give rise to the intestinal units inti, int2, int3, and int5; the intervening unit int4 is derived from descendants of Ep. Therefore, if different heritable sexual commitments were made early in the lineage, for example in the E-cell daughters Ea and Ep, the result would be mosaic patterns with non-contiguous cells expressing vitellogenin genes (Fig. 5). Likewise, loss at one of the following left-right divisions would result in patterns with cells on one side of the intestine expressing and contiguous cells on the other not expressing vitellogenins. These predicted patterns are in fact observed in tra-1 mosaics (Hunter and Wood, 1990), but never in triploid intersexes.

Additional evidence for local non-autonomy comes from the same-sex preferences observed among blast cells that give rise to postembryonic lineages, for example in the tail (Tables?, 8). Among the animals exhibiting a male fate for any one of the blast cells V5L/R, V6L/R, and TL/R (ranging from 191 to 223 scored), 168 or about 80% showed male fates for all these cells (Table 7). If sexual fates were decided independently, the expected percentage of animals with all male fates for these cells, calculated from the probabilities of maleness for the individual cells (Table?), would be about 25%, far less than the observed value. The patterns’ observed in mosaic intestines as well as the biases observed in postembryonic blast cell lineages strongly suggest that cells in these tissues can influence the sexual identity decisions of their neighbors.

The observed polarity of the intestinal mosaics, with hermaphrodite fates always anterior to male fates, suggests a more global external influence on sex determination or sexual differentiation. Its significance is underscored by our observation, on smaller numbers of animals, of the same polarity in germ-fine mosaics and the majority of apparent somatic-gonad mosaics. The exclusive polarity observed in the intestine cannot simply be an artifact of the dissection and in situ hybridization procedures, since the same techniques showed labeling with the reverse polarity in intestines from diploid L4 hermaphrodites. The different responses of anterior and posterior cells could result from effects on either the sex determination process or on downstream genes that more directly control sexspecific differentiation. In the intestine, for example, the result could be explained by position-dependent differences in activity of the mab-3 gene, which negatively regulates vitellogenin synthesis in response to the state of the sex-determining tra-1 gene (Shen and Hodgkin, 1988).

We found no strict correlation among triploid intersexes in general between vitellogenin synthesis in the intestine and the sexual identity of any other single tissue. Therefore, the external influence responsible for the polarity of intestinal mosaics is not likely to be the signal controlling normal intestinal sexual differentiation, unless such a signal emanates from a tissue that was not scored or from several different tissues. Some possibilities, for example, a signal from the midventral hypodermis (scored only indirectly in most of our experiments; see Materials and methods) is ruled out by other experiments: animals in which the entire AB lineage and therefore the ventral hypodermis is male nevertheless express vitellogenin genes normally in the intestine (Hunter and Wood, 1990), showing that a female ventral hypodermis is not required to induce intestinal vitellogenin synthesis.

Thus we can conclude only that in certain tissues, when the primary sex-determining signal is poised at a threshold level, some difference in positional information along the anterior-posterior axis can tip the balance toward female differentiation in anterior cells and male differentiation in posterior cells. Precedent for such a phenomenon in C. elegans comes from analyses of the mab-5 gene, which functions cell-autonomously to control cell fates in the posterior of the animal as if responding to positional information (Kenyon, 1986; Waring and Kenyon, 1990).

There is no contradiction in our finding that the gradient pattern of intestinal vitellogenin gene expression during larval development of normal diploid hermaphrodites is opposite to the pattern in adult intersexes; the triploid intersex patterns show that intestinal cells can have position-dependent differences in adult levels of vitellogenin transcripts, while the pattern in L4 intestines reflects the kinetics of initiating vitellogenin transcription. These results suggest that vitellogenin transcription may be under the control of two temporally distinct signals in normal diploids. The first, establishing the sexual identity of intestinal cells, probably leads to expression in XX animals of the tra-1 gene (Hodgkin, 1987) by each intestinal cell (Hunter and Wood, 1990); this in turn may permit subsequent vit gene expression by preventing the action of the mab-3 gene product. The second, a temporal signal probably produced during the L4 stage and not necessarily hermaphrodite-specific, could then trigger the onset of vitellogenin gene transcription in hermaphrodite-committed intestinal cells, beginning at the posterior end. Two such signals are also implicated in Drosophila, where the hormones 20-hydroxyecdysone and juvenile hormone signal initiation of vitellogenin synthesis in the fat body of females, but have little effect on the fat body cells of males, even at abnormally high concentrations (Postlethwaite and Jowett, 1980).

If two cells in a mosaic animal adopt opposite sexual fates, then the choices must have been made after their lineages diverged. The mosaic patterns that we have observed indicate that choices of sexual identity can be made during late embryonic and also during postembryonic development. For example, different sexual identities of tail ectoderm hook and ray precursors in a mosaic animal must result from decisions made after their lineal divergence about midway through embryogenesis. Differences seen within mosaic germ lines, sex muscles and tail ectodermal lineages that generate multiple rays must result from decisions made after hatching, because these tissues are all generated postembryonically. The patterns observed indicate that some of these decisions can be made as late as the L3 stage of postembryonic development. In mosaic intestines, the finding of many examples in which the vit probe hybridized to only one of two sister cells in the adult (e.g. to int2V but not int5L; see Fig. 5) shows that the decision to express vitellogenin genes can be made after all 20 cells of the adult intestine are generated by the final E-cell divisions, which occur at the end of the cell proliferation phase about halfway through embryogenesis (Sulston et al. 1983). However, this decision could be made as late as the L4 stage when these genes are expressed. In other tissues, our ability to score cells reliably as either male, hermaphrodite, or intersexual may be the limiting factor in ascertaining how late sex determination decisions can be made.

These observations suggest the possibility that in animals with ambiguous X/A ratios, some cells may remain sexually uncommitted and able to vacillate between feminine and masculine states until expression of their sexually differentiated fates. This possibility is supported by our observation of an animal with male diagonal sex muscles at the mid-ventral hermaphrodite position, indicating that the sex myoblasts must have first migrated anteriorly in the hermaphrodite mode and subsequently undergone the male pattern of muscle differentiation. Because both migratory behavior and production of sexually dimorphic muscle patterns are probably cell-autonomous characteristics (Hunter and Wood, 1990), the sex myoblasts in this animal must have reversed their sexual identity between the L2 and L4 larval stages.

Temperature-shift experiments on animals carrying ts mutations in one of the sex-determining genes fem-2 (Kimble et al. 1984; Doniach and Hodgkin, 1984), tra-2 (Klass et al. 1979) and her-1 (P. S., P. Jonas, and W. B. W., in preparation) have also shown that sex determination decisions can be reversed during larval development and even in adults. However, in these experiments, done with normal diploid strains, X/A ratios were unambiguously assessed, and reversal of the normal sexual phenotype was caused by inactivation of a sex-determining gene.

X/A assessment in normal diploids must occur early in embryogenesis, so that dosage compensation can be set to either the male or the hermaphrodite mode (Villeneuve and Meyer, 1990). However, the above shift experiments and many other observations (Hodgkin, 1988) make it clear that X/A assessment does not represent an irreversible commitment to sexual identity, because it can be overridden by subsequent effects on the sex-determining genes that control tra-1 (Fig. 1). The commitment becomes irreversible only when tra-1 activity has been definitively set to ON or OFF, or if this activity is at an intermediate level, when the tra-1 signal has been perceived as present or absent by its target sexual differentiation genes. Thus our experiments measure not the timing of X/A assessment, but rather the timing of setting or perceiving the activity state of tra-1.

We have shown that tra-1 and its target genes act cell-autonomously to control sexual differentiation (Hunter and Wood, 1990). However, if one or more of the upstream genes that regulate tra-1 can act non-autonomously, then cells in triploid intersexes might change their sexual identity through cell interactions at any time in development, overriding any earlier X/A assessment. Our results would then indicate the timing of the last non-autonomous step in tra-1 regulation.

We conclude that some non-autonomy in the sex determination process could explain the same-sex biases of tissues and cells within tissues, the lateness of sexual commitments, and possibly though less straightforwardly the positional effects on sex determination observed in triploid intersex mosaics. If indeed present, such non-autonomy should be demonstrable by analysis of specific gene mosaics, which we are currently pursuing (Hunter and Wood, in preparation).

Triploid mosaic intersexes in Drosophila produce only large clones of oppositely sexed cells, indicating early (pre-gastrulation) sex determination decisions that appear irreversible (Baker and Belote, 1983) and probably involve setting the activity state of the Sx1 gene (Cline, 1984), which maintains this state by autoregulation. We have shown that such intersexes in C. elegans can produce small clones of oppositely sexed cells, indicating later, apparently unstable sex determination decisions and lack of an early commitment maintained by autoregulation. These contrasting results support the recent realization (Hodgkin, 1990) that although Drosophila and C. elegans share a common primary signal, the X/A ratio, these organisms differ fundamentally in their mechanisms of sex determination.

We are grateful to A. Carpenter and A. Chisolm for critical reading of the manuscript, to J. Taylor for help with its preparation, and in particular to the reviewers for insightful suggestions regarding revisions. This research was supported by the National Institutes of Health through grants HD-11762 and HD-14958 to W.B.W. and predoctoral traineeships to P.S. and C.P.H.