Primary sex determination in the nematode C. elegans

The nematode Caenorhabditis elegans has a sexual system that at first sight appears to be profoundly different from the sexual systems found in higher vertebrates. First, the two sexes normally encountered are hermaphrodite and male (Fig. 1), rather than the more usual female and male, and the animal can reproduce both by self-fertilization and by cross fertilization (Fig. 2). Second, the primary sex-determining signal is not a dominant sex chromosome like the Y chromosome in mammals, or the W chromosome in birds. Instead, sex is determined by X-chromosome dosage, so that diploid animals with two X chromosomes (XX) are hermaphrodite and those with one X chromosome (XO) are males. Males can arise spontaneously by X-chromosome loss in a self fertilizing hermaphrodite population, clearly demonstrating that there is no Y chromosome in this system. Sex determination in C. elegans has much more in common with sex determination in Drosophila than it has with mammalian mechanisms; nevertheless, there are some underlying similarities between all three systems, and it has been possible to modify the C. elegans system in a variety of ways so as to make it resemble the Drosophila system or the vertebrate system more closely.

An X-chromosome dosage mechanism could rely either on a chromosome-counting mechanism, in which the absolute number of chromosomes is important, or on a ratio mechanism, in which the relative number of X chromosomes to autosomes is the determinative variable. The first kind of system is found in species with haploid male/diploid female sex determination, such as Hymenoptera; the second kind is the system found in Drosophila. Classical work by Nigon (1951) showed that tetrapioid C. elegans carrying three or four X chromosomes (4A;3X or 4A;4X) were hermaphrodite, but those with two X chromosomes (4A;2X) were male. This shows that it is the X/A ratio that determines sex, not the absolute number of X chromosomes. Madl & Herman (1979) confirmed Nigon’s findings and demonstrated that 3A;2X triploids were male, in contrast to the 4A;3X tetrapioid hermaphrodites. The ratio-measuring mechanism is therefore capable of distinguishing between X/A ratios as similar as 0·67 and 0·75. They also showed that there must be at least three different sites on the X chromosome that were involved in computing the ratio, by observing the effect of small X-chromosome duplications on the sex of 3A;2X animals. This work is further discussed below.

Increasing the X/A ratio, as in a 2A;3X animal (ratio 1·5) does not cause feminization: such animals are viable self-fertile hermaphrodites, like 2A;2X animals (Hodgkin. Horvitz & Brenner, 1979). However, both evolutionary considerations and genetic analysis indicate that the C. elegans sexual system is a modified male/female system, and that the hermaphrodite is in essence a female that has the ability to make both sperm and eggs. The number of species in the phylum Nematoda is unknown, but probably well over Iff; most of these have an XX female/XO male system (White. 1973). Several species within the genus Cae-norhabditis itself have XX females and XO males (e.g. C. remanei:Sudhaus. 1974). so it is reasonable to suspect that hermaphroditism is a secondary specialization. This would be of obvious value to a fastgrowing ubiquitous soil organism like C. elegans. permitting more efficient exploitation of transient habitats.

Genetic analysis has supported the view that the hermaphrodite is a modified female and provided a preliminary explanation of how spermatogenesis can be allowed to proceed in a female body. The starting point of this analysis has been the isolation and analysis of (by now) hundreds of mutations that alter the normal process of sex determination, causing XX animals to develop into males or females (rather than the normal hermaphrodite), or causing XO animals to develop into hermaphrodites or females (rather than the normal males). Most of these mutations fall into seven autosomal complementation groups, each of which we have studied in some detail. Table 1 summarizes the phenotypes of some of these mutants and the locations of the genes are shown in Fig. 3. The epistatic interactions between these genes have been extensively investigated and have been interpreted in terms of the basic model for sex determination shown in Fig. 4. The autosomal genes appear to be organized in a cascade of at least four regulatory steps: the X/A ratio controls activity of her-1, her-1 activity controls tra-2. and so on. As a result, the regulatory genes can take on one of two activity states, depending on the X/A ratio. For most of the genes, putative null mutations are available: these have little or no effect on sexual phenotype in one chromosomal sex, but cause a transformation of the other sex. Therefore, it is possible that this gene activity is either low or absent during normal development of the unaffected sex. For four of the genes, both loss-of-function (putative null) mutations and gain-of-function mutations have been isolated. The loss-of-function mutations cause transformation in one direction and the gain-of-function mutations cause transformation in the other. An example is shown in Fig. 5: a her-l(O) (null mutation, loss-of-function) mutation has no effect on XX animals (normal hermaphrodites), but transforms XO animals into hermaphrodites (Hodgkin, 1980). Conversely, a rare her-l (gf) (gain-of-function) mutation has no effect on XO animals (normal males), but partly masculinizes XX animals (Trent, Tsung & Horvitz, 1983). This mutation is also weakly dominant, as expected for a gain-of-function allele, whereas the her-l(O) mutation is recessive. The existence of two classes of mutation causing opposite transformations provides strong evidence for a model in which these genes play key roles in normal sex determination: the loss-of-function mutations show that her-l activity is necessary for XO male sex determination, and the her-l(gf) mutation shows that her-l activity is sufficient for at least part of male sex determination. However, the effect of both kinds of mutation can be completely suppressed by appropriate mutations of tra-2: her-l (0); tra-2 (0) animals are masculinized, and her-l(gf):tra-2(gf) animals are feminized (Hodgkin, 1980; Doniach, 1986). Therefore her-l must act by regulating, either directly or indirectly, the tra-2 gene or gene product. Similar arguments have led to the proposed cascade in its present form.

At step one, the her-l gene appears to control a choice between male and hermaphrodite development, but at step three, the fem genes control a choice between male and female development. Both XX and XO fem mutants develop into fertile females, in which the early gametes normally destined for spermatogenesis develop into oocytes instead. This suggested that spermatogenesis in the hermaphrodite was achieved by a germ-line-specific modification of the normal regulatory cascade, acting at a step prior to the fem gene control, most probably tra-2. Support for this idea comes from the properties of certain dominant tra-2 mutations (tra-2(gf)) (Doniach, 1986; Hodgkin, 1986), which eliminate spermatogenesis from XX animals but not from XO animals. Homozygous tra-2(gf) stocks therefore grow as XX female/ XO male cross-fertilizing populations. The interpretation is that these tra-2(gf) mutations render tra-2 insensitive to the germ-line-specific modulation that normally occurs in the XX animal, but does not affect the control exerted by her-l action in the XO animal; consequently tra-2(gf) XO animals are fertile males. If this interpretation is correct, one might expect to find a gene activity that would be responsible for germ-line modulation, and a plausible candidate has been identified in the gene fog-2 V (fog standing for ‘feminization of germ line’: Schedl & Kimble, 1987). Recessive (probably loss-of-function) mutations of fog-2 have the same effect as tra-2(gf) mutations, causing feminization of XX but not of XO. It appears that fog-2 is active in both XX and XO animals (though irrelevant in the latter), but that it acts only during early germ-line development, resulting in the initial phase of hermaphrodite spermatogenesis. During later development, fog-2 does not or cannot act, so the hermaphrodite germ line switches to sustained oogenesis.

This model explains how hermaphrodite spermatogenesis is achieved, and also provides a possible justification for the rather elaborate cascade of gene interactions. That is, the cascade may be necessary in order to permit germline modulation, and that a simpler system would permit development of only males and females. Molecular probes for several of the genes in the cascade are becoming available, so it may be possible to determine how much of this regulatory cascade is operative in related nematode species with male/female systems, the presumed ancestral arrangement.

The fog-2 and tra-2(gf) mutations demonstrate how the sexual system can be modified in one step from a male/hermaphrodite system to a male/female system. Such an artificial strain has XX female/XO male sex determination, like most nematodes or many other animals – for example, the Drosophila species Drosophila annulimana.

It is also possible to provide C. elegans with a Y chromosome, as shown by Sigurdson et al. (1986). The X-autosome translocation mnT12 attaches the whole of LGIV to the X chromosome. This translocation is homozygous viable, so it is possible to construct a strain with only four pairs of autosomes, plus a neo-X (mnT12) and a neo-Y (the original LGIV). In this strain, hermaphrodites are XX and males are XY. If such a strain were made homozygous for a fog-2 mutation, the result would be an XX female/XY male strain, essentially like Drosophila melanogaster (in which a Y chromosome is present, but does not determine sex).

By using different mutations, it is possible to achieve a more extreme alteration. As indicated in Table 1, complete sexual transformation in either direction can be achieved by appropriate mutations of the last gene in the cascade, tra-l. For tra-l(O) alleles, both XX and XO develop into males, and for tra-l(gf). both XX and XO develop into females. Because tra-l(gf) is dominant to tra-l(O), heterozygous tra-l(gf)/tra-1(0) animals are also female. They can be cross-fertilized by tra-l (0) males to yield equal numbers of tra-l(gf)/tra-l (0) females, and tra-l(0) males. This establishes a stable male/female population which can be grown indefinitely, and in which sex is determined by a dominant W system (Hodgkin. 19836). The autosome bearing tra-I(gf) is in essence a W chromosome, and the autosome bearing tra-l(O) is a Z chromosome, so we have a WZ/ZZ system, as in birds and many reptiles. In these strains, the X chromosome dosage does not affect sex. Instead, primary sex determination is achieved by the presence or absence of a single active gene, which appears to be generally the case in mammals. Thus, sexual systems that appear to be extremely different on the surface may have significant underlying similarities.

The detailed interactions underlying this cascade and the subsequent events required for correct sexual development of many different tissues, are being actively investigated at both genetic and molecular levels in several different laboratories. It is reasonable to hope that a molecular definition of some of the postulated regulatory interactions will soon be possible. Also, molecular analysis should reveal whether there are significant homologies between sex determination in nematodes and sex determination in other organisms.

Sex determination in C. elegans depends on a chromosomal dosage mechanism, so it has been important to establish whether there is dosage compensation for the different numbers of X chromosomes in the two sexes and, if so, how dosage compensation and sex determination are integrated. In mammals, dosage compensation occurs by X-chromosome inactivation, a process wholly independent from sex determination. For example, the single X is never inactivated in XO female mammals, whereas only one X is active in XXY male mammals. In Drosophila, dosage compensation is achieved by increasing transcription from the single X chromosome of XY flies, to match the total output of the two X chromosomes of XX flies. Failure to compensate correctly is lethal. Both sex and compensation are controlled by a single ‘master control gene’, Sxl (Sex-lethal), which has been the subject of a long and fascinating series of studies by Cline (1983. 1984, 1986). His experiments demonstrate that in XX flies, Sxl adopts a state which dictates both female development and low X chromosome transcription; in XY flies, Sxl adopts a state dictating male development and high X transcription. As in C. elegans, the primary signal is the X/A ratio, which in some manner not yet understood (see Nothiger, this volume) sets the state of Sxl; this state is thereafter propagated by an autoregulatory mechanism, independent of X/A ratio (Cline. 1984). The arrangement in Drosophila therefore neatly solves a problem that would otherwise arise: compensation is necessary for viability, but perfect compensation removes the difference in X-chromosome dosage that is required for sex determination. There are, of course, other possible solutions to this problem: for example, the genes involved in determining X/A ratio might be specifically uncompensated, so that the perceived X/A ratio would not be affected by compensation.

Both genetic and molecular investigations have shown that C. elegans does indeed compensate for the difference in X-chromosome dosage between XX and XO animals (Wood. Meneely, Schedin & Donahue, 1985; Meyer & Casson. 1986). Most (but not all) sex-linked genes are expressed at similar levels in both XX and XO animals, and both X chromosomes in an XX hermaphrodite are active, so a mechanism of the Drosophila type is likely. Compensation appears to occur at the mRNA level: Meyer and Casson (1986) found that transcript levels for three different sex-linked genes were at equal levels in XX and XO animals (comparing either wild-type hermaphrodites and males, or tra-1 XX males and wild-type XO males). A fourth sex-linked gene examined in this study was found to be uncompensated; also, three sex-linked amber suppressors appear to be uncompensated. because suppression is higher in homozygous XX animals than in hemizygous XO animals (Hodgkin. 1985 and unpublished data).

Several autosomal genes have been identified that are involved in the process of dosage compensation (Table 2). Mutations of these genes, for example dpy-21 and dpy-26. have little or no effect on XO animals, but cause either dumpiness or lethality to XX animals (Hodgkin, 1983c). It was proposed that these effects arise from increased levels of X-chromosome expression, and this has been shown to be the case for dpy-21 (Meyer & Casson, 1986; Meneely & Wood, 1987). Two other mutations with effects similar to dpy-26 have been identified subsequently: rhlS (isolated by E. Hedgecock) and yl (isolated by J. Plenefisch & B. J. Meyer). The rh!8 mutation was found to complement dpy-26, and maps between the LGlll markers unc-93 and dpy-I7 (J. H.. unpublished data); it is therefore assigned to a new gene, dpy-27. They/ mutation also maps to LGlll. but to a different location, and yl complements rh!8. so yl defines a fourth gene in this class, dpy-28 (L. DeLong, J. Plenefisch & B. J. Meyer, personal communication). Both dpy-27 and dpy-28 mutations cause elevated levels of transcripts from sex-linked genes, like dpy-21 (Meyer & Casson, 1986).

These “chauvinist dumpy’ genes are significantly different from dosage compensation genes identified in Drosophila (reviewed by Baker & Belote. 1983). The fly genes mle, msl-1, insl-2 and msl-3 are required for increasing X-chromosome transcription in flies with only one X chromosome, so mutations of these loci do not affect female flies (XX) but are lethal to male flies (XY); hence the gene names maleless (mle) and male-specific lethal (msl). In contrast, the nematode genes dpy-21, dpy-26 etc. appear to be required for decreasing X-chromosome transcription, since mutations are deleterious to XX rather than XO. Also, three of the genes exhibit a strong maternal rescue effect: dpy-26 XX progeny derived from dpy-26/ + mothers show little abnormality, but those derived from dpy-26 homozygous mothers usually die as embryos or young larvae. The dpy-27 and dpy-28 mutations have a similar, but somewhat weaker phenotype (J. H., unpublished data; L. DeLong & B. J. Meyer, unpublished data). Mutations of dpy-21. in contrast, do not normally show any maternal effect.

Some possible candidates for genes with the opposite functional role have been identified, but have been harder to analyse than the XX-deleterious class. Mutations of two sex-linked genes, dpy-22 and dpy-23 cause abnormal phenotypes in both XX and XO, but the XO phenotype is much worse (Hodgkin & Brenner, 1977). Some evidence for decreased X-chromosome expression in dpy-22 and dpy-23 mutants has been obtained (Meneely & Wood, 1987). Another sex-linked mutation, y9, has a more clean-cut effect, leading to lethality in XO hemizygotes but not in XX homozygotes; it was isolated as a suppressor of XX lethality in dpy-28 mutants, so there is good reason to believe that it is involved in dosage compensation (L. Miller & B. J. Meyer, personal communication).

The mutations described in the previous section do not have any conspicuous effect on sex determination, although they do affect dosage compensation. A sex-linked locus with significant effects on both processes has recently been characterized, sdc-J (formerly egl-J6 (Trent et al. 1983); sdc-l stands for ‘sex and dosage compensation’: Villeneuve & Meyer, 1987). Two mutations of sdc-l have been studied, which result in partial masculinization and also in elevated X-chromosome expression and transcript levels. The masculinization effects, but not the altered dosage compensation, can be suppressed by her-l(O) mutations, indicating that sdc-l acts at an early step in the sex-determination pathway, and also acts on dosage compensation. The phenotypes are stronger in sdc-l/mnDfl XX heterozygotes, which indicates that these mutations do not completely eliminate sdc-l function. Also, bothxdc-7 alleles show a maternal effect: little masculinization is seen in sdc-l daughters of sdc-l/+ mothers.

There is a formal similarity between sdc-J and the Drosophila gene Sxl: both are sex-linked loci which appear to be involved at an early step in sex determination and dosage compensation. For both genes, loss-of-function mutations lead to masculinization and elevated X-chromosome transcription in XX animals, but have no obvious effect on XO (or XY) animals. However, as discussed by Villeneuve & Meyer (1987) there are other possible roles for this gene: it could be one of the elements measured in the assessment of the X/A ratio. It cannot be the only such element, because a deficiency for the sdc-l region has no dominant effect on sexual phenotype in XX animals (normal hermaphrodites), and a duplication for sdc-l has also no effect on XO animals (normal males). Nevertheless, it could be one of a number of such elements. More detailed analysis of this locus will be necessary to define its role. A different locus on the X chromosome, with properties related to those of sdc-l, has been identified by Wood et al. (1987).

The remainder of this paper will be devoted to analysis of the natural primary sex-determining signal, the X/A ratio. This is a problem which may have wider biological significance than is at first apparent, because it involves the translation of a quantitative signal into a qualitative choice. Most biological switch mechanisms that we understand at present operate on qualitative distinctions, as in the presence or absence of the mammalian Y determinant. However, there are also cases where small quantitative differences appear to trigger very different consequences. Nervous systems obviously have the computational ability to make such distinctions, but quantitative discrimination is a much harder task at the level of a single cell or part of a cell. It is widely believed that some kind of morphogen gradient is responsible for many events in pattern formation; the correct interpretation and response to such a gradient is likely to require accurate measurements by the cells in the morphogenetic field. No convincing example of a natural morphogen gradient in a multicellular organism has yet been put forward, so this is a highly speculative area of developmental biology. In the case of X/A ratio assessment in Drosophila and Caenorhabditis, we have examples of small differences in ratio leading to entirely different consequences. In these systems, it may be possible to define the signals precisely and determine how they are measured. It is also at this point that the sexdetermination systems of C. elegans and Drosophila seem most similar, and one can hope that insights gained from these two organisms will be mutually illuminating, even if they shed little light on mammalian sex determination.

The original analysis of X/A ratio by Madl & Herman (1979) and subsequent unpublished work by R. K. Herman (personal communication), suggested that there were at least four sites on the right half of the X chromosome that contributed to the ‘numerator’ of the X/A ratio. They examined the effect of partial X duplications on the sexual phenotype of triploid 3A;2X animals, which are normally male. A small duplication (mnDp8: see Fig. 6) had no obvious effect on 3A;2X animals, but larger duplications did have an effect: 3A.;2X;mnDp9 animals were usually intersexual and 3N’,2X;mnDplO animals were hermaphrodite. This showed that there must be at least two sites, one included in mnDplO but not mnDp9, and another within mnDp9 but not mnDp8. Furthermore. since diploid 2A;X0;mnDplO animals are male, there must be at least one more site not included in mnDplO. In these experiments, 3A;2X animals were generated by crossing tetrapioid hermaphrodites (4A;4X) with diploid males carrying the duplication; in the reverse cross, mating diploid 2A.;2X;mnDp8 hermaphrodites with 4A;2X tetra-ploid males, several of the progeny were hermaphrodite or intersex. This suggests that there is a fourth site within mnDp8 and also that there can be a maternal effect on the measurement of X/A ratio. Note that the gene sdc-J lies in this region, so it would be of interest to repeat these experiments using sdc-1 ;mnDp mothers.

Subsequent to this work a number of additional X chromosome duplications have become available, with extents approximately as shown in Fig. 6. These have not yet been investigated for their effect on 3A;2X animals. None of them appears to cause any hermaphroditization in 2A;XO;Dp males.

So far. it has been possible to infer rather little about the nature of the autosomal part of the X/A ratio. If there were a single site responsible for this ‘denominator signal’, it might be expected to show dosage effects in aneuploid animals: a deficiency for such a site might cause a shift towards hermaphrodite development in D// + . XO animals, and a duplication might cause masculinization of +/+;Dp;XX animals. A large number of autosomal duplications and deficiencies are now available in C. elegans. with approximate extents shown in Fig. 3. About 80% of the map is covered by duplications and about 50% by deficiencies. None of these have been shown to affect sexual phenotype in diploid animals. Possibly, they might have effects in animals with intermediate X/A ratios, such as 3A;2X males or 4A;3X hermaphrodites, but such an analysis would be hard to carry out, and has not yet been attempted. It is difficult to make a large quantitative shift in the X/A ratio by means of autosomal alterations.

The possibility that the first gene in the sex determination pathway, her-1. is involved directly in measuring X/A ratio can be considered, but is unlikely for several reasons. A simple model might be that the her-1 locus produces a fixed amount of some product which is titrated by X-chromosome sites or products. At high X dosage, the her-l product is completely titrated out and therefore inactive; at low X dosage only partial titration is achieved, so some active her-1 product remains and is able to put the pathway into the male state. However, with such a model one would expect to see dosage effects for her-1, and these are not usually observed. Both her-1 /+ and mDfl/ + XO animals are normal fertile males (Hodgkin. 1980 and unpublished data; mDfl is a deficiency that fails to complement her-1 and flanking markers). Doniach (1986) did observe an effect on sexual phenotype in tra-2(gf)/ + ;her-l/ + XO animals, which are more feminized than the corresponding lra-2(gf)/ + ;+/ + XO animals, but this is the only dominant effect so far observed.

One can sustain the her-1 titration model by proposing that the her-1 gene is able to autocompensate in some way. so that equal amounts of product are made both by her-1 /+ and by wild-type animals. Alternatively, it could be that none of the her-1 loss-of-function alleles is completely null for her-1 product. There is some heterogeneity amongst these alleles (Hodgkin, 1980, 1983c), suggesting that at least some are not null: for example, her-l (e1520) XO animals are less fertile than her-l(e1518) XO animals. A deficiency for the region. mDfl. has become available (D. L. Riddle, unpublished data) and has been used in complementation tests with these two alleles. Both e1518/ mDfl XO and e1520/mDfl XO animals are hermaphrodites, with similar very low fertility. This test therefore does not show which of the two alleles is closer to a null, because it is possible that dominant effects due to mDfl are causing a general reduction in fertility. C. Trent & W. B. Wood (personal communication) have carried out tests with weak alleles of her-l which suggest that e1520 does have residual her-l activity. Nevertheless, it is still possible that some of the other recessive her-l alleles are null alleles.

Expression of the gain-of-function her-l mutation, n695 (Trent et al. 1983) can be affected by X-chromosome dose or activity, as expected if her-l were a denominator element. For example, the phenotypes of 2A;2X and 2A;3X animals were compared, by examining the progeny of a 2A;3X hermaphrodite of genotype him-5(el467) her-l (n695). 16/28 XX animals showed some masculinization (as in Fig. 5E), but only 1/13 XXX animals. Also, a strain of genotype her-l(n695) dpy-21(e428) XX was constructed and found to show almost no masculinization (2/55 animals examined had a slightly masculinized tail). However, it may be that these effects are indirect or insignificant, given that n695 is an incompletely penetrant mutation; its expression may therefore be very sensitive to genetic background.

The fact that sdc-1 appears to act upstream of her-l and to control both sex and dosage compensation might be taken as evidence against a role for her-l as denominator, but the argument is not compelling. First, the function of sdc-1 is not certain at this point, being based on only two alleles. Second, there is no a priori reason why sex and dosage compensation should be regulated together; it could be that sdc-1 is a numerator element for both processes, and her-l is the denominator for sex alone, there being some other denominator for dosage compensation. Another (less likely) possibility is that her-l does have a function in dosage compensation control, but that none of the her-l mutations so far obtained affects this function.

The her-l gene cannot be the only gene responding to X/A ratio, because there is a ‘minor pathway’ (indicated by the dotted line in Fig. 4), whereby X/A ratio is able to influence sexual phenotype independent of her-l and tra-2 (see Hodgkin, 1980). It is still not clear what gene activities are involved in this minor pathway or whether it plays any role during the normal process of sex determination.

Wood et al. (1987) have proposed that the autosomal ‘chauvinist dumpy’ genes constitute the denominator elements. This proposal is based in part on an analogy with the Drosophila gene da (daughterless), which is believed to act in setting the state of Sxl, and therefore to have some kind of denominator function. Like da, the C. elegans genes dpy-26, dpy-27 and dpy-28 have an extreme maternal effect, being lethal or almost lethal to XX progeny but not to XO progeny. However, unlike da, these genes do not show a strict maternal effect, because mating dpy-26 mothers with wild-type males yields perfectly normal viable dpy-26/ + XX progeny. No dosage effects have been demonstrated for the three loci dpy-26, dpy-27, and dpy-28, for which both duplications and deficiencies are available. Moreover, in diploid animals the mutations of these genes do not cause any noticeable feminization of XO animals: a double mutant dpy-26(nl99); dpy-21 (e428) XO is still a fertile male (Hodgkin, 1983b).

Meneely & Wood (1984) reported that one allele of dpy-21, ct16, did cause feminization of diploid XO animals, leading to an abnormal male phenotype. This effect, however, is due to a second mutation. A strain BW139, genotype supposedly him-5(e1467) dpy-21 (ct16) V was obtained from P. Meneely and found to have a predominantly non-dumpy hermaphrodite phenotype, but to segregate males with a Mab (male abnormal) phenotype. A few dumpy hermaphrodites were also segregated; these appeared to be viable 2A;3X animals, as expected for a strain carrying the meiotic non-disjunction mutation him-5(e1467) (Hodgkin et al. 1979). When BW139 hermaphrodites were mated with dpy-21(e428) XO males, all cross progeny were completely wild type. The strain was also mated with wild-type males and the segregation of the mab mutation followed: it was unlinked to him-5 and was eventually mapped to LG1I by M. Shen (personal communication). He has found that it fails to complement the known mab mutation mab-3(el240) II, and therefore is assigned to this gene, being given a new allele designation e2093 (to avoid confusion). The ‘ctl6’ Mab phenotype illustrated by Meneely & Wood (1984) is strikingly similar to that of mab-3(e1240) (illustrated by Hodgkin, 1983A; Hodgkin, Doniach & Shen, 1985), so there is no doubt that the abnormal male phenotype described by these workers is due to mab-3, not dpy-21.

Meneely & Wood (1984) also described a different feminizing effect of dpy-21: in diploid XO animals with large X duplications, (such as mnDplO’), dpy-21 mutations caused a shift in phenotype from male to intersex. Related experiments are reported here which demonstrate a more extreme effect with both dpy-21 and dpy-27 (which has stronger effects than dpy-21).

In these experiments (Table 3), 3A;2X progeny are generated by mating 2A;2X hermaphrodites with 4A;2X males. Normally (cross 6), these animals are all male; the occasional hermaphrodites are probably 3A;3X animals, resulting from imperfect disjunction in the tetrapioid male parent (i.e. a few sperm are probably 2A;2X rather than 2A:1X). However, if the diploid mother carries either dpy-21 or dpy-27, a considerable shift is seen towards either an intersexual or a hermaphrodite phenotype (crosses 9-14). If both mother and father carry dpy-27, most of the progeny are hermaphrodite (cross 14).

Given the information that these dpy-21 and dpy-27 mutations cause elevated levels of most X-chromosome transcripts (Meyer & Casson, 1986), the obvious interpretation is that the ‘perceived’ X/A ratio has been shifted, leading to the change in sexual phenotype. In a diploid XO animal, the shift is insufficient to reach the critical level, but in a 3A;2X animal, a much smaller shift is required, so sexual phenotype is switched. If this interpretation is correct. then it follows that X-chromosome transcripts (rather than sites) are being measured in the computation of X/A ratio and that there can be a significant maternal contribution to this computation (because there is asymmetry between crosses 7 and 9, and 8 and 12).

Various other conclusions can be inferred from the data in Table 3. For example, it appears that X-chromosome disjunction is slightly impaired in tetrapioid dpy-21 and dpy-27 males. A meiotic effect has already been seen for dpy-26 (Hodgkin. 1983c) but not hitherto for dpy-21 or dpy-27. As a result of this, some 3A;1X male animals are generated in crosses 7, 8 etc.; they are recognizable because they express the matroclinous unc-7 marker. Such animals had previously been thought to be inviable. (Madl & Herman, 1979), but at least some appear to survive, although they are thin, small and pale. A cross to confirm this was carried out. which generated non-Unc 3A;1X males (Table 4). Fewer than the expected number were found, and none was able to mate successfully (unlike 3A;2X males), confirming this conclusion. The presence of a dpy-27 or a dpy-21 mutation did not seem to improve the phenotype (Table 3, cross 10 and cross 14), suggesting that the compensation machinery is only able to work within certain limits (2A;3X animals are also abnormal and 2A;4X animals die as embryos: Hodgkin et al. 1979).

It is possible that the autosomes also fail to disjoin perfectly in tetrapioid males, so that some of the resulting 3A;2X animals have either two or four copies of a particular autosome, rather than three. This could account for some of the variability in these crosses – for example, the occasional hermaphrodites and intersexes generated in the control cross (Table 3. cross 6). Conceivably one could identify which autosomes are important in computing X/A ratio, by repeating these crosses with animals carrying autosomal markers.

Many, and perhaps all, of the ‘intersexes’ generated in these crosses are ‘mosaic’ intersexes, containing a variety of mixtures of male and female parts. They have not been examined in detail, but it seems likely that this could result from independent assessment of X/A ratio in different tissues. When the signal is ambiguous (around 0·7), different tissues might make different decisions, resulting in a mosaic intersex, as in triploid Drosophila (Cline, 1983).

The main conclusion from these experiments is that the dosage compensation genes can affect assessment of the X/A ratio, so the X-chromosome part of the ratio is not uncompensated. This returns us to the problem mentioned above: compensation will tend to abolish the primary sex determining signal, so the system should not work, unless some other process is involved. At present, a likely possibility is that C. elegans uses the same solution as Drosophila. If the sdc gene(s) are the nematode analogues of Sxl, then one would predict that they, too, should exhibit autoregulation, and become insensitive to X/A ratio once their initial state has been set. Given the present tools available for molecular and genetic analysis of C. elegans, experiments to test these ideas should soon be possible.

I am grateful to Ed Hedgecock, Bob Herman, Judith Kimble. Phil Meneely. Barbara Meyer, Michael Shen and Bill Wood for discussion, for communicating unpublished results, and for providing strains.