daf-12 regulates developmental age and the dauer alternative in Caenorhabditis elegans

The life history of Caenorhabditis elegans has six recognized stages, embryo, first (L1) through fourth (L4) stage larvae, and adult, with options at specific points for alternative developmental programs or reversible arrest (diapause). From cell division to tissue morphogenesis, the sequence and timing of developmental events are nearly invariant (Sulston and Horvitz, 1977; Sulston et al., 1980, 1983; Kimble and Hirsh, 1979; Hedgecock and White, 1985). These descriptions of wild-type development in controlled conditions provide an integrated framework, spanning from the cellular to the organismal level, for studying the influence of environment and genotype on life history.

Physiological and environmental factors, including food, temperature and population density, can pause or divert postembryonic development at several points (Golden and Riddle, 1984; Johnson et al., 1984; Arasu et al., 1991). Most dramatically, larvae must choose between two mutually exclusive programs, continuous or dauer, for L3 stage development (Cassada and Russell, 1975). The dauer larva is specialized, both morphologically and behaviorally, for extended diapause and dispersal, allowing survival until conditions improve. Signals for this decision are integrated via the dauer-formation (daf) genes. Daf-defective mutants bypass dauer formation, while Daf-constitutive mutants select this option, regardless of environmental conditions. These genes identify multiple signaling pathways that converge on a nuclear hormone receptor gene, daf-12, possibly their ultimate target (Yeh, 1991; Riddle and Albert, 1997).

Heterochronic genes specify developmental age, i.e., select stage-appropriate developmental programs, in nematode tissues (Ambros, 1997). Mutations in these genes can advance or delay the expression of stage-specific developmental programs. The known heterochronic gene activities affect all extragonadal tissues, but have no known role in the somatic gonad and germline. Those genes that have been molecularly characterized encode presumptive transcriptional or translational regulators, generally distributed throughout the affected tissues (Ruvkun and Giusto, 1989; Lee et al., 1993; Rougvie and Ambros, 1995; Moss et al., 1997).

Here we describe novel heterochronic mutants, defining two loci, that delay gonadal development at L3 and later stages. Some mig-7 alleles, and the one existing mig-8 allele, delay gonadal maturation only, notably leader cell migrations. Other mig-7 alleles also delay extragonadal development; epistasis experiments suggest mig-7 mediates between lin-14 and lin-28 heterochronic activities in these tissues. Neither gene affects germline age. Unexpectedly, mig-7 mutants prove to be novel alleles of daf-12. The various alleles reveal two separable gene activities, one specifying L3 dauer programs in adverse conditions, and another specifying L3 continuous programs in favorable environments; in the absence of either activity, somatic cells repeat L2 programs as a default. We discuss how daf-12 and related genes could act as selectors for these programs. We propose the hierarchy of heterochronic genes form cellular memories regulating these selector functions. In particular, endocrine factors could trigger stage transitions and simultaneously advance these memories.

Except where noted otherwise, nematodes were cultured at 20°C on NGM media inoculated with Escherichia coli strain OP50 (Sulston and Hodgkin, 1988). him-5(e1490) strains were used to examine male phenotypes (Hodgkin et al., 1979).

daf-12 alleles rh61, rh62 and rh84, and mig-8 allele rh50, were isolated after ethylmethanesulfonate (EMS) mutagenesis in F₂ screens for mutants with distal-tip cell migration (Mig) phenotypes (Hedgecock et al., 1987). Allele rh193, which was found segregating in strain CB3400, presumably also resulted from EMS mutagenesis. Alleles rh62rh157 and rh61rh411 were isolated as spontaneous non-Mig revertants of rh62 and rh61, respectively. Seven additional, daf-12 alleles were obtained after EMS mutagenesis in non-complementation screens for Mig phenotypes. Briefly, crosses of EMS-treated N2 males with daf-12 unc-128(rh110) hermaphrodites were screened for rare Mig non-Unc hermaphrodite progeny. Alleles rh257 and rh258 were obtained from 6000 mutagenized chromosomes tested over daf-12(rh84). Alleles rh273 and rh274 were obtained from 20,000 chromosomes tested over daf-12(m20). Finally, alleles rh284, rh285 and rh286 were obtained from 25,000 chromosomes tested over daf-12(rh258).

Mapping data for daf-12 and mig-8, summarized in Fig. 1, were deposited with the Caenorhabditis Genetics Center.

During postembryonic development, each tissue undergoes a sequence of unique or repeated stage-specific events for growth and maturation (Sulston and Horvitz, 1977; Sulston et al., 1980). In the hypodermis, a new cuticle is synthesized at the beginning of each stage, followed by seam cell divisions and molt of the old cuticle (Singh and Sulston, 1978). This entire sequence of events, or molt cycle, typically requires just 9 hours at 20°C. However, the sequence is modified in the first molt cycle, where there is no old cuticle, and the fifth, where the seam cells fuse together, exiting permanently from the cell cycle (see Fig. 4). At characteristic stages, some seam cells undergo additional divisions to increase their numbers or generate peripheral neuroblasts. Finally, the composition and structure of the cuticle itself is specialized in the L1, dauer larva and adult stages. Growth and maturation of other somatic tissues are closely coordinated with the hypodermal molt cycle. Intestinal nuclei, for example, endoreplicate just before each molt (Hedgecock and White, 1985).

The somatic gonad and germline also undergo a sequence of unique stage-specific events for reproductive maturation. Mesoblasts Z1 and Z4 divide in L1 stage to generate the distal and proximal leader cells, and various gonadoblasts (Kimble and Hirsh, 1979). The gonadal leader cells, which are postmitotic, organize gonadogenesis, and polarize germline maturation, along their distoproximal axis. In the male, the proximal leader, or linker cell, undergoes a multistage migration along the larval body wall, which determines the shape and position of the mature gonad, while, in the hermaphrodite, the two distal leaders, or distal-tip cells, lead such migrations. The gonadoblasts divide in L3 and L4 stage to generate specific regions of the gonad. Beginning in L3, the proximalmost germline cells cease proliferation and commit to meiosis. In hermaphrodites, meiocytes commit to spermatogenesis or oogenesis at L4 and adult stages, respectively (Kimble and White, 1981; Austin and Kimble, 1987).

Development of gonadal and extragonadal tissues is closely coordinated over a range of environments. We examined the relative timing of gonadal development, notably leader cell migrations, under various culture conditions. Each leader cell describes a unique trajectory on the body wall which can be analyzed as a succession of unidirectional migrations on uniform substrata (Fig. 2). Under exhausted conditions, gonadoblast division, gonadal leader cell migration and germline proliferation cease during dauer diapause, indicating that germline, somatic gonad and extragonadal tissues all respond to a common signal for developmental arrest.

To describe the effects of genotype and environment on the relative timing of events, we must distinguish chronological and developmental stage. Following Ambros and Horvitz (1984), we reserve ‘L1’ through ‘L4’ for chronological stages as measured by elapsed molts, and use ‘S1’ through ‘S4’ for the corresponding developmental stages in wild type. In particular, ‘S3’ and ‘S3d’ refer to the characteristic developmental programs expressed in non-dauer or dauer larvae, respectively, during the third larval (L3) stage. Similarly, ‘A’ and ‘SA’ refer to chronological and developmental adult stages, respectively. Finally, we have adapted this notation to gonadal development, notably leader migrations (Fig. 2), with the proviso that some events occurring late within an intermolt are considered early events of the next developmental stage.

The original mig-7 and mig-8 alleles were isolated after ethylmethanesulfonate mutagenesis in screens for self-fertile hermaphrodites with abnormal distal-tip cell migrations (Hedgecock et al., 1987). Additional mig-7 alleles were obtained in non-complementation screens or as spontaneous intragenic revertants (see Methods). As described below, genetic mapping (Fig. 1) and complementation show mig-7 mutations represent new classes of daf-12 alleles that perturb developmental age as well as dauer-formation. Cellular phenotypes, described below by tissue, vary greatly by allele and cell type. The most severe alleles delay expression of S3 and later developmental programs throughout the soma. In many cases, cells express essentially normal S3 programs after variable delays; in other cases, cellular development arrests completely. Cells also make inappropriate stage commitments in these mutants. In some cases, cells repeat S2 programs one or more times before expression of later programs; in other cases, they express abnormal programs, interpreted as hybrids of S2 and S3 programs.

Hypodermal development appears normal in daf-12(rh61) through the L2 stage. At L3, however, seam cells frequently fail to express S3 programs, repeating S2 programs instead. In particular, Vn.pap and Vn.ppp undergo equational division followed by stem-cell division in early L3 (Figs 3A, 4). Occasional cells repeat S2 programs yet again at L4, but most express S3 or S4 programs (which are not easily distinguishable). Finally, a fraction of the seam cells delayed at L3 later fail to express SA programs at adult and repeat S4 programs instead. As a result, the alae of the adult cuticle have visible gaps (Figs. 3B, 4). In hermaphrodites whose V3 seam descendants are delayed, uterine, vulval and sex muscle cells fail to attach to the seam, allowing the uterus to evert at the onset of egg-laying (cf. Newman et al., 1996). Adults with many delayed seam cells may undergo an extra molt cycle in which delayed cells finally express SA programs, filling in any alae gaps. The penetrance of these seam phenotypes varies greatly by allele; the adult phenotype generally parallels the L3 phenotype, but is less penetrant (Table 1).

Vulval development is generally normal in daf-12(rh61) hermaphrodites. However, in another allele, rh193hs, the anchor-cell-dependent divisions of vulval precursors (cf. Kimble, 1981) are frequently delayed or absent. At 25°C, these divisions were delayed beyond the L3 molt in 6/26 larvae examined. Finally, in rh61 males, there are frequent defects in tail hypodermal development (cf. Sulston et al., 1980), including delayed division of ray precursor cells in the L3 stage, and incomplete tail-spike retraction and fan morphogenesis in the L4 stage. In particular, the anteriormost ray sensilla often fail to join the developing fan.

In daf-12(rh61) hermaphrodites, development of the somatic mesoblast M, including sex-myoblast migrations, are normal through the L2 stage (cf. Sulston and Horvitz, 1977). However, the subsequent divisions of the sex-myoblasts are often delayed (Fig. 3C). At the L3 molt, only 2/22 cells had completed all three division rounds on schedule, while 13/22 had completed two divisions and 7/22 just one. Finally, all three division rounds are completed by mid-L4. Despite these delayed divisions, differentiation of sex-myoblast descendants, the vulval and uterine muscles, are usually normal. Some adults, however, retain their eggs.

In wild type, intestinal cells int3 through int9 undergo nuclear division at L1 lethargus, becoming binucleate (Sulston and Horvitz, 1977; Hedgecock and White, 1985). In daf-12(rh61), intestinal development appears normal through the L2 stage. However, occasional cells become either tri- or tetranucleate as one or both nuclei undergo a second division at L2 lethargus (Fig. 3D). We interpret this phenotype as a repetition of S2 programs in the intestinal cells at the L3 stage (Table 1).

In daf-12(rh61) males, gonadal development appears normal through mid-L2 stage. In particular, the compaction of gonadoblasts behind the linker cell and its initial anterior migration are normal. Beginning in late L2, the linker cell migration is frequently abnormal. We have classified these migration errors according to larval stage, i.e., L3 or L4, and direction of movement, i.e., dorsoventral or anteroposterior (Fig. 5, Table 2). The mutant trajectories may reflect repetitions of S2 or S3 pathfinding programs in the linker cell at L3 and L4, respectively (see Discussion). In cases where the linker cell completes its S4 movements, it generally halts short of the cloaca and/or fails to be engulfed, suggesting its SA program may be delayed as well. Adult males are invariably infertile.

In daf-12(rh61) hermaphrodites, gonadal development appears normal through mid-L2 stage. However, the compaction of gonadoblasts at L2 lethargus, and their divisions in early L3, are variably delayed (Fig. 3E). We interpret these phenotypes as delay or failure to express S3 gonadoblast programs (Kimble and Hirsh, 1979). Later abnormalities, including irregular queuing of oocytes, production of undersized eggs and sterility, may be consequences of these L3 phenotypes, or reflect direct defects in the regulation of S4 or SA programs in these cells. Beginning in late L3, the distal-tip cell migrations are frequently abnormal (Figs. 3F, 5; Table 2).

The mutant trajectories may reflect repetition of S3 pathfinding programs in the distal-tip cells at L4 (see Discussion). Frequently, the basement membrane surrounding the distal ovary ruptures with escape of germline cells into the pseudocoelom. In most alleles, the distal-tip cells remain motile through mid-L4 as judged by their adherent morphology. In rh193 cultured at 25^oC, however, these cells cease active migration during L3.

In wild-type hermaphrodites, pachytene stage meiocytes are first evident in L3 lethargus (33 hours), while sperm and oocytes become evident at the onset of L4 lethargus (42 hours) and at ecdysis (45 hours), respectively (Kimble and White, 1981; Austin and Kimble, 1987). In daf-12(rh61) hermaphrodites, sperm and oocytes become evident about 6 and 2 hours, respectively, before L4 ecdysis. The brood size ranged from 0 to 41, with mean 10±12, for a sample of selfing hermaphrodites (n=22), compared with over 300 in wild type (Hodgkin et al., 1979).

The reference allele, daf-12(m20), and similarly selected alleles are Daf-defective and, by genetic epistasis experiments, act near the end of the dauer pathway (Riddle et al., 1981; Vowels and Thomas, 1992; Thomas et al., 1993). Many of the new classes of daf-12 mutations that we isolated affect the regulation of dauer formation; most alleles are Daf-defective, some are Daf-constitutive, while others have near normal dauer regulation. Epistasis experiments using these new allele classes also place daf-12 near the end of the dauer-signaling pathway, e.g. Daf-defective allele rh84 completely suppresses dauer formation in doubles with Daf-constitutive mutants daf-1(m40) and daf-7(e1372).

We compared dauer formation over a range of growth temperatures in both replete and exhausted cultures. Generally, no morphologically recognizable dauer larvae were ever observed in cultures of the various Daf-defective alleles. However, in exhausted cultures, allele rh257 forms occasional larvae with cuticular alae, and some of the radial shrinkage of body and pharynx, characteristic of dauer larvae. Under these conditions, wild-type dauer larvae remain arrested in diapause, and neither feed nor undergo cellular development. By contrast, these ‘partial-dauer’ larvae attempt to resume feeding and, consequently, are sensitive to 1% sodium dodecyl sulfate (SDS) (Cassada and Russell, 1975).

Daf-constitutive alleles rh62, rh273 and rh274 form partial-dauer larvae under replete conditions at all temperatures, whereas rh193 and rh285 can produce occasional partial-dauer larvae when cultured at 25°C or 15°C, respectively (Fig. 6; Table 1). Again, these larvae resume feeding and are generally SDS-sensitive. Somatic development, but not germline proliferation, are arrested in the partial-dauer larvae. Whereas dauer larvae in exhausted cultures generally arrest with fewer than 40 germ cells in wild type, partial-dauer larvae often have twice this number. Most partial-dauer larvae eventually progress to fertile adults, but up to one-third remain arrested indefinitely even at 15°C. Occasionally, these larvae have been observed to synthesize dauer-specific cuticle at L3, and then again, at L4 stage. Finally, all five of these alleles can form normal dauer larvae when cultured under exhausted conditions.

We have placed available daf-12 alleles into six, idealized classes based on their dauer-formation phenotypes, i.e., Daf-defective, -normal, or -constitutive, and secondarily, their heterochronic phenotypes in gonadal and extragonadal tissues (Table 1). Class 1, 2 and 3 alleles are Daf-defective. Class 1 alleles (rh61, rh84, rh257) have delayed gonadal and extragonadal development. Class 2 comprises a single allele rh62rh157 with nearly normal gonadal, but delayed extragonadal development; this mutation arose as a spontaneous non-Mig, intragenic revertant of rh62. Class 3 alleles (rh258, rh61rh411) have nearly normal gonadal and extragonadal development. Allele rh61rh411 arose as a spontaneous non-Mig, intragenic revertant of rh61. The original, Daf-defective alleles isolated by D. Riddle and others (e.g., m20, m25, m116, m583, sa156) all fall within this, largest class (Riddle et al., 1981; Thomas et al., 1993; Larsen et al., 1995). Class 4 and 5 have nearly normal dauer regulation. Class 4 alleles (rh193, rh285) have delayed gonadal and extragonadal development, while class 5 alleles (rh284, rh286) have delayed gonadal development only. Finally, class 6 alleles (rh62, rh273, rh274) are Daf-constitutive. These mutants have delayed gonadal, but nearly normal extragonadal development.

All daf-12 alleles examined are recessive to wild type for both dauer regulation and heterochronic phenotypes (Table 1). For dauer-formation, class 1, 2 and 3 alleles fail to complement, i.e., trans-heterozygotes are Daf-defective; moreover, these Daf-defective classes are recessive to class 6, i.e., trans-heterozygotes are Daf-constitutive. These mutations reveal a gene function required for S3d programs in adverse conditions designated daf-12b. We propose class 1, 2 and 3 alleles reduce or eliminate daf-12b, while class 6 alleles misregulate this activity (see Discussion).

For gonadal and extragonadal heterochrony, class 3 alleles partially complement class 1, i.e., the penetrance of these phenotypes follows the series rh61/rh61>rh61/m20> m20/m20>rh61/+ = m20/+ = +/+ (Table 1). Class 6 alleles complement extragonadal, but not gonadal heterochrony, of other classes; finally, class 3 alleles only poorly complement gonadal heterochrony of class 6. These mutations reveal a gene function required for S3 programs in replete conditions designated daf-12a. We propose class 3 alleles reduce or eliminate daf-12a function arising from the daf-12 locus, while class 1 alleles also antagonize daf-12a function arising from other sources (see Discussion).

Heterochronic and dauer-formation phenotypes vary significantly with growth temperature in many daf-12 alleles (Table 1). To learn when daf-12(+) activity is required during development, we performed temperature-shift experiments with two temperature-sensitive alleles. In rh285, seam cells express S3 programs at L3 stage when cultured at 15°C, but repeat S2 programs instead at 25°C (Table 1). Up until mid-L2, temperature upshifts and downshifts can reveal or rescue this phenotype, respectively (Fig. 7). In rh193, seam cells generally repeat S2 programs at L3 stage regardless of culture temperature; their descendants express SA programs at adult stage when cultured at 15°C, but repeat larval programs instead at 25°C (Table 1). Up until late L2, temperature upshifts can reveal the adult seam phenotype, while downshifts can rescue until about mid-L3 (Fig. 7). In rh193 hermaphrodites, the distal-tip cells express S4 pathfinding programs at L4 when cultured at 15°C, but repeat S3 programs instead at 25°C (Table 1). Up until late L2, temperature upshifts can reveal this phenotype, while downshifts can still rescue until mid-L3 (Fig. 7).

Using double mutants with class 1 allele daf-12(rh61), we examined possible functional interactions with the heterochronic genes lin-4, lin-14, lin-28 and lin-29 (Ambros, 1997). Hypodermal phenotypes of these double mutants are summarized below; gonadal phenotypes generally resemble daf-12 alone.

In lin-14(ma135) null mutants, hypodermal cells express S2-S3-S4-SA programs in succession, bypassing S1 entirely (Ambros and Horvitz, 1984). This phenotype reflects two separable activities, lin-14a, which prevents precocious expression of S2 programs in L1 stage, and lin-14b, which prevents precocious expression of S3 programs in L2 stage (Ambros and Horvitz, 1987). In lin-14 daf-12, most seam cells express S2-S2-S3-S4-SA programs in succession, indicating daf-12 acts downstream of lin-14; moreover, daf-12 prevents precocious vulval development (Table 3). We propose lin-14b inhibits daf-12a, possibly indirectly, to prevent precocious expression of S3 and later programs.

In lin-4(e912) null mutants, hypodermal cells fail to advance to S2 programs, repeating S1 programs indefinitely (Chalfie et al., 1981; Ambros, 1997). lin-4 lin-14 double mutants are indistinguishable from lin-14 alone, suggesting lin-4(+) inhibits lin-14(+) to advance S2 and later programs (Ambros, 1989). As expected, hypodermal development in lin-4 daf-12 double mutants resemble lin-4 alone (Table 3); also like lin-4 alone, lin-4 daf-12(rh62) double mutants are Daf-defective (data not shown). In the absence of lin-4(+), persistent lin-14b can evidently inhibit daf-12a and daf-12b indefinitely.

In lin-28(n719) null mutants, hypodermal cells express S1-S3-S4-SA programs in succession, bypassing S2 entirely (Ambros and Horvitz, 1984). In lin-28 daf-12, seam and vulval development are nearly indistinguishable from lin-28 alone, indicating lin-28 acts downstream of daf-12 (Table 3). We propose daf-12a inhibits lin-28(+), possibly indirectly, to advance S3 and later programs.

lin-4 enhances daf-12 hypodermal phenotypes in the lin-14(ma135) background, indicating lin-4(+) acts on a second target, downstream of lin-14b, to advance S3 programs. For example, seam cells often express SA programs at L4 stage in lin-14 daf-12 doubles, but never advance to these programs before adult stage in lin-14 daf-12 lin-4 triples (Table 3). Moss et al. (1997) have recently reported that lin-4(+) can inhibit lin-28(+) directly. Incorporating their observation, we propose lin-4(+) and daf-12a inhibit lin-28(+) through independent mechanisms to advance S3 and later programs.

mig-8 is represented by a single, recessive allele, rh50, with delayed expression of S4 gonadal programs but normal extragonadal development and dauer regulation (Tables 1, 2). In hermaphrodites, the compaction and divisions of gonadoblasts are normal through early L3, but thereafter, there are penetrant defects in distal-tip cell migrations and herniation of the distal ovary. The brood size ranged from 40 to 146, with mean 117±32, for a sample of selfing hermaphrodites (n=11). Similarly, in males, linker cell migrations are essentially normal through early L3, but then halt without undergoing their S4 movements.

mig-8 maps between mec-2 and mup-2 on chromosome X and fails to complement the deficiency stDf1 (Fig. 1). rh50/stDf1 hemizygotes are phenotypically indistinguishable from rh50/rh50 homozygotes, suggesting that rh50 could be null or near null. In particular, no enhancement of gonadal phenotypes, nor novel defects in extragonadal development or dauer regulation were observed. Using double mutants, we examined possible functional interactions with daf-12(+) (Table 1). Unexpectedly, daf-12(m20), which has no gonadal phenotype itself, can partially suppress the L4 distal-tip cell migration phenotype of mig-8. Conversely, mig-8, which itself has normal dauer regulation, can partially suppress the Daf-defective phenotype of daf-12(rh84). These results suggest that mig-8(+) somehow acts with daf-12(+) to regulate both life stage transitions and dauer formation.

In daf-12 individuals, each tissue is generally a mosaic of cells expressing normal and delayed developmental programs (Fig. 8). For example, individual seam cells may express S3 programs at L3 stage while adjacent cells within the same larva repeat S2 programs instead. Similarly, phenotypic mosaicism is observed for expression of seam SA, intestinal S3 and distal-tip S4 programs. In the partial-dauer larvae formed in Daf-constitutive alleles under replete culture, occasional seam cells express S3 rather than S3d programs. Whereas most seam cells produce dauer-specific alae and undergo mitotic arrest, these exceptional cells leave gaps in the alae and divide, bypassing S3d diapause. Interestingly, in both hypodermis and intestine, the frequency of delayed phenotypes is highest for cells in the mid-body and declines toward both the head and tail (Fig. 8).

In daf-12(rh84) hermaphrodites (n=100), 74% ± 4 of the left (posterior) and 39% ± 4 of the right (anterior) distal-tip cells undergo S4 dorsal movements (Fig. 8). The frequency of larvae in which both distal-tip cells migrate dorsally, 33% ± 5, is consistent with probabilistic independence of these events, 74%×39% = 29%. In mig-8(rh50) hermaphrodites (n=201), by comparison, 6.5±2% of the left and 7±2% of the right distaltip cells migrate correctly. In this case, the high frequency of larvae in which both distal-tip cells migrate dorsally, 4.5±1%, appears inconsistent with probabilistic independence, 6.5%×7% = 0.5%.

From a cellular perspective, the multicellular organism is a colony of cells whose coupled programs of gene expression collectively direct development. Few steps in these programs are strictly cell-autonomous; rather most are contingent upon intercellular signals or environmental factors encountered during development. Inductive signals, for example, coordinate cellular programs in adjacent tissues, while morphogens and hormones coordinate development over much greater distances. In special cases, hormones can advance cellular programs throughout the organism; such events are usually recognized as life stage transitions. Without such endocrine regulation, each individual would eventually become a mosaic of cells and tissues of incommensurate developmental ages, or incompatible life strategies. daf-12 regulates L3 stage development in C. elegans. We propose this gene coordinates cellular commitments throughout the soma, and subordinates them to organismal commitments, through an endocrine mechanism.

In daf-12 mutants, development is essentially normal until L3 commitment but, thereafter, somatic cells throughout the larva fail to express S3 and later programs on schedule (Tables 1, 2). In hypodermis and intestine, cells undergo extra proliferative divisions interpreted as repeated S2 programs. In the gonad, altered cell migrations likely reflect delays or repetitions in pathfinding programs of leader cells themselves (Antebi et al., 1997; M.-W. Su, M. Killeen, E. Hedgecock and J. Culotti, unpublished data). For example, the male linker cell often continues along the ventral body muscles without changing direction at L2 lethargus, interpreted as a repetition of its S2 pathfinding program (Fig. 5). In other cases, this cell executes dorsal movement but then fails to turn posteriorly, interpreted as hybrid S2/3 programs.

A few daf-12 phenotypes, e.g., the repeated execution of S2 programs in the intestine or linker cell, are already evident by L2 lethargus. Presumably, the cellular commitments for these programs occur even earlier. Indeed, in temperature-sensitive daf-12(rh285hs) larvae, the critical period for seam cell commitments to S3 programs occurs around the middle of the L2 intermolt (Fig. 7). Later stage phenotypes seen in some daf-12 mutants, e.g., failure to express S4 distal-tip cell programs at L4, or SA seam cell programs at adult stage, might be purely consequences of errors made at L3 commitment. In daf-12(rh193hs), however, the critical periods for these phenotypes occur within the L3 intermolt itself (Fig. 7), indicating either a direct requirement for daf-12(+) at L4 commitment, or inappropriately late gene activity in these mutants.

In wild type, germline and somatic development are closely coordinated. For example, both germline and somatic cell divisions are arrested during the dauer diapause. Like other known heterochronic genes, daf-12 probably has no direct role in germline development. The onsets of meiosis and gametogenesis, which roughly coincide with S4 and SA somatic programs in wild type, are not delayed in these mutants. Moreover, while somatic tissues are arrested in partial-dauer larvae arising in Daf-constitutive alleles, germline proliferation appears unaffected. Signals directly from the somatic gonad, which determine the spatial arrangement of germline maturation, might control germline cell age as well (Schedl, 1997).

Dauer larvae express an alternative program for L3 development, designated S3d, adapted for diapause and dispersal (Cassada and Russell, 1975). Several environmental factors, predictors of future starvation, regulate both the choice between S3 and S3d programs, and within the S3d program itself, the decision to exit from diapause (Golden and Riddle, 1984). Exudates from soil microbes, indicating potential food sources, bias L3 commitment toward S3 programs, while warm temperatures and pheromone, indicating potential overcrowding, bias this decision toward S3d programs. Committed dauer larvae cease feeding at L2 lethargus and undergo morphological adaptations for diapause, e.g., radial shrinkage of the body and maturation of a dauer-specific cuticle. On exit from diapause, these larvae resume feeding and reverse their radial shrinkage. At this time, cells throughout the animal resume programs of division, migration and differentiation that are indistinguishable from their S3 counterparts.

Environmental and physiological factors regulating dauer commitment and recovery are integrated via the dauer formation genes (Riddle and Albert, 1997). Of the known genes in this signaling pathway, daf-12 appears unique in three respects. First, this locus has both Daf-defective and Daf-constitutive alleles, and may therefore have an instructive role in the dauer decision. The Daf-constitutive alleles are dominant to Daf-defective alleles, and both classes are recessive to wild type, suggesting that daf-12(+) can promote or inhibit S3d programs under adverse and favorable growth conditions, respectively. Second, the other genes only affect the choice between S3 and S3d programs, not the advance from S2 to either S3 or S3d programs. Finally, epistasis experiments suggest that daf-12 is possibly the ultimate target of the dauer signaling pathway (Riddle et al., 1981; Thomas et al., 1993; Gottlieb and Ruvkun, 1994; Riddle and Albert, 1997). The terminal branches of this pathway are shown in Fig. 9B.

The ‘partial-dauer’ larvae formed in Daf-constitutive alleles of daf-12 under replete conditions never complete radial shrinkage, but instead resume feeding several hours after the onset of L2 lethargus. Paradoxically, developmental programs in the soma and gonad, excepting germline proliferation, remain arrested. These observations might be explained if commitment to S3d programs is incomplete in these individuals, affecting some but not all aspects of dauer formation. Instead, we suggest that these larvae initiate all S3d programs at L3 commitment, but sensing no threat of starvation, attempt to exit from the dauer-associated diapause at the earliest opportunity (cf. Vowels and Thomas, 1992; Gottlieb and Ruvkun, 1994). The continued developmental arrest of these larvae after resumption of feeding suggests daf-12 has a role in maintaining growth arrest during dauer diapause.

Environmental and physiological factors also regulate adult life history traits, e.g., size at maturity, reproductive schedule and lifespan, and behaviors, e.g., egg-laying and thermotaxis. In particular, the insulin-like receptor branch of the dauer signaling pathway (Fig. 9B) has been implicated in the regulation of adult lifespan (Kenyon et al., 1993; Dorman et al., 1995; Larsen et al., 1995; Morris et al., 1996; Kimura et al., 1997, Ogg et al., 1997, Lin et al., 1997). This pathway is proposed to integrate physiological signals regulating larval and adult life histories. We suggest these signals converge on an endocrine tissue, yet unidentified, making stage commitments for the entire organism, and then relayed hormonally throughout the soma.

The functional organization of the daf-12 locus is surprisingly complex (Table 1). In class 1 alleles, somatic cells cannot select S3 or S3d programs, but instead repeat S2 programs, in L3 stage. In class 4 alleles, these cells cannot select S3 programs, but express S2 or S3d programs in favorable and adverse environments, respectively. In class 3 alleles, they cannot select S3d programs, even in adverse environments. In class 6 alleles, they cannot select S3 programs, but express S3d programs even in favorable environments.

The various daf-12 mutants reveal two separable gene activities, i.e., daf-12a, which selects S3 programs in favorable environments, and daf-12b, which selects S3d programs in adverse environments (Fig. 9B). In the simplest model, these mutant alleles, which are all recessive to daf-12(+) for the phenotypes examined (Table 1), reflect a reduction of one or both gene activities. In particular, class 3 and 4 alleles selectively reduce daf-12b or daf-12a activity, respectively, while class 1 alleles reduce or eliminate both activities. Finally, in class 6 alleles, daf-12b is active even in favorable environments; whereas these Daf-constitutive alleles are dominant to the Daf-defective classes, they are fully recessive to daf-12(+).

Preliminary molecular studies suggest that class 3 alleles, rather than some other class, make no functional products (A. Antebi, M. Snow, W. Yeh, E. Hedgecock, D. Riddle, and P. Larsen, unpublished data). If so, daf-12b activity reflects a simple loss-of-gene-function in class 1 and 3 alleles, while daf-12a reflects a recessive gain-of-function in class 1 and 4 alleles. This could explain why class 3 alleles are relatively common among daf-12 mutants, and how class 1 alleles sometimes mutate spontaneously to class 3 (Table 1). Class 1 and 4 alleles are recessive to wild type, indicating daf-12(+) products can compete effectively against the mutant products to advance S3 programs; moreover, class 3 homozygotes have impenetrant heterochronic phenotypes, indicating daf-12(+) products must themselves provide some daf-12a activity. An obvious explanation is that daf-12a reflects partially redundant actions of daf-12(+) and another, unknown gene; only a genetic analysis of both genes can establish the relative contributions of each locus.

We have incorporated epistasis relationships (see Results) for both daf-12a and daf-12b activities into existing models of the heterochronic and dauer signaling pathways (cf. Moss et al., 1997; Riddle and Albert, 1997). In particular, daf-12a acts between lin-14 and lin-28 in extragonadal cells. In our model (see Fig. 9B), S1, S2 and S3/S3d programs form a default hierarchy favoring the highest available program. In extragonadal tissues, the regulatory circuit is initialized by embryonic expression of lin-14 and lin-28. S2 and S3 programs are repressed by lin-14a and lin-28, respectively, while lin-14b provides feed-forward inhibition of daf-12a and daf-12b. Consequently, at L1 commitment, cells must adopt S1 programs. By L2 commitment, inhibition of lin-14 by lin-4 has derepressed S2 programs. By L3 commitment, inhibition of lin-28 by lin-4 and daf-12a has derepressed S3 programs. As discussed above, daf-12 may also act at L4 commitments. Finally, lin-29 (not shown) is required to initiate SA programs at adult commitment (Ambros, 1989).

The molecular mechanisms for several heterochronic gene activities have been identified. lin-14 and lin-28 encode unstable nuclear and cytoplasmic proteins, respectively, expressed throughout the extragonadal tissues in L1 larvae (Ruvkun and Giusto, 1989; Moss et al., 1997). LIN-28, which has a cold-shock domain and CCHC zinc-finger motifs, is a presumptive RNA-binding protein, while LIN-14 appears entirely novel. After hatching into a replete environment, lin-4 expresses small RNAs that base pair with the 3′ UTR of lin-14 and lin-28 mRNAs, inhibiting their translation (Lee et al., 1993; Wightman et al., 1993; Moss et al., 1997). As a consequence, these protein levels decline significantly from mid-L1 through mid-L2.

Less is known about the molecular mechanisms for other heterochronic gene interactions. The critical period for lin-14b activity, which prevents precocious expression of S3 programs, occurs in mid-L1 (Ambros and Horvitz, 1987). We propose LIN-14 is required at this time to inhibit daf-12a and daf-12b (Fig. 9B); the mechanism of this inhibition is unknown. Similarly, the critical period for daf-12a activity, which prevents repeated expression of S2 programs, occurs in mid-L2. We propose daf-12a acts at this time to either lower LIN-28 levels or inhibit its actions; unlike the translational inhibition of lin-28 mRNA described above, this daf-12a action does not require lin-4(+). Indeed, in extragonadal tissues, overexpression of LIN-28 causes repeat of S2 programs at L3 stage, phenocopying daf-12 mutants (Moss et al., 1997). Finally, lin-29 encodes proteins with CCHH zinc-finger motifs that accumulate at later stages in various somatic nuclei, both gonadal and extragonadal (Rougvie and Ambros, 1995; Bettinger et al., 1996). In the hypodermis, LIN-29 accumulation is regulated by LIN-14 and LIN-28.

Under adverse conditions, daf-12b activity specifies the S3d alternative at L3 commitment. We favor a one-step model for commitment to dauer formation (see Fig. 9A). Under favorable conditions, the insulin-like receptor and TGF-β signaling pathways together select daf-12a, while under adverse conditions, when one or both pathways are inactive, daf-12b acts as a default (Fig. 9B). This model also explains why lin-14, but not lin-28 mutants, cause precocious expression of daf-12b activity and selection of S3d programs at L2 commitment (Liu and Ambros, 1989). Interestingly, the late stage phenotypes in daf-12, mig-8 and other heterochronic mutants are suppressed by dauer development (Table 1; Liu and Ambros, 1991; Euling and Ambros, 1996a). We suggest these phenotypes reflect misregulation of the daf-12a activity but are bypassed during dauer development by daf-12b.

A variety of evidence suggests that the hierarchy of heterochronic genes is initialized differently, and operates independently, in different tissues and, perhaps, individual cells. For example, lin-29 is expressed throughout the soma (Bettinger et al., 1996), whereas lin-14 and lin-28 appear inactive in the gonad (Ambros and Horvitz, 1984; Ruvkun and Giusto, 1989; Moss et al., 1997). Such differences could contribute to the tissue specificity of some daf-12 alleles, e.g., in class 5 and class 6 daf-12 alleles, and in mig-8(rh50), gonadal somatic cells, but not extragonadal cells, fail to express S3 or later programs on schedule (Table 1). Interestingly, mutations changing the relative timing of gonadal and extragonadal development are proposed to play a major role in the evolution of metazoan life histories (Gould, 1977).

Heterochronic genes select stage-appropriate developmental programs in nematode tissues (Ambros, 1997). In daf-12 and other mutants, individual cells within a tissue can vary in developmental age (cf. Ambros and Horvitz, 1984, 1987). Moreover, these differences, precocious or retarded, can propagate through lineal descendants over successive stages. By inference, each somatic cell maintains a partly autonomous memory of its developmental age. Heterochronic genes are likely components of these memory circuits as well as temporal selectors per se. We suggest these circuits, distributed throughout the soma, advance one step with each new life stage so that the heterochronic activities enumerate the chronological stages (Fig. 9B). At the same time, these activities are used to select stage-appropriate developmental programs. In this view, the heterochronic genes are analogous to cyclin-dependent protein kinases in cell-cycle regulation whose activities both specify a progression of stages and select stage-appropriate events.

The hypodermal molt cycle provides several convenient landmarks for staging nematode development (cf. Fig. 4). Ecdysis, the traditional boundary between molt cycles, occurs properly within a life stage, not between stages. A variety of evidence suggests that organismal and cellular commitments to new life stages precede each lethargus by about 5 hours (corrected to 20°C culture). First, new stage-specific programs, observed as changes in gene expression, cell-cycle commitment, or cell migration, commence several hours before lethargus (cf. Euling and Ambros, 1996b; Johnstone and Barry, 1996; Antebi et al., 1997). Cellular commitments are presumably completed before any execution of these programs. Second, critical periods of several heterochronic gene activities precede the lethargus by several hours. For example, lin-14b and daf-12a act several hours before L1 or L2 lethargus, respectively (Ambros and Horvitz, 1987); these critical periods likely reflect cellular commitment itself. Finally, a change of environmental conditions can reverse any bias toward S3 or S3d alternatives until up to 5 hours before L2 lethargus (Swanson and Riddle, 1981; Golden and Riddle, 1984). Phenotypic mosaicism has not been observed in wild type, suggesting that organismal commitment to a particular L3 alternative strictly precedes cellular commitments.

The nervous system, integrating several environmental factors, releases growth factors modulating the C. elegans life history (Riddle and Albert, 1997; Ren et al., 1996). Presumably, these graded neural signals are evaluated by an endocrine tissue, yet unidentified, that makes life history decisions for the entire organism. If heterchronic genes provide cellular memories of developmental age, a single hormone from this tissue could suffice to periodically coordinate cellular programs and advance their memories. In insects, for example, ecdysteroids periodically coordinate cellular programs (Thummel, 1995), but it is unknown if they also advance memories.

Switching circuits that step through a sequence of unique states in response to a repeating input are called counters. We suggest that nematode heterochronic genes form a life stage counter that advances at each new stage in response to a repeating endocrine input (Fig. 9B). Presumably, two distinct hormones are required at the L3 branchpoint to coordinate cellular programs as well as specify an S3/S3d alternative (Fig. 9A). A hormone released only in adverse environments, for example, might complement or replace the hormone released in favorable environments.

The discovery that daf-12 encodes a nuclear receptor subunit (Yeh, 1991; Riddle and Albert, 1997) suggests hormonal activation of nuclear receptor complexes could regulate cellular commitments for L3 and perhaps other stages. Selective activation of receptor complexes mediating daf-12a activity could select S3 programs in favorable environments, while activation of daf-12b could select S3d programs in adverse environments. Within the S3d program itself, endocrine signals for L4 commitment could also coordinate dauer recovery by activating daf-12a (cf. Fig. 9A). Without daf-12a or daf-12b, cells recommit to S2 programs as a default, suggesting that distinct nuclear receptor complexes could act as selectors for S2, S3 and S3d programs, respectively. It will be important to determine the molecular composition of daf-12 and related receptor complexes, and compare them to selector activities defined genetically.

We wish to thank V. Ambros, P. Larsen, W. H. Yeh and D. Riddle for sharing unpublished results and strains, D. Brown, V. Corces, A. Fire, H. Hutter, M. Montgomery and C. Norris for helpful comments on the manuscript, and S. Wimbuish for technical assistance. The Caenorhabditis Genetics Center, which is supported by National Institutes of Health National Center for Research Resources, provided strains for this work. A. A. received postdoctoral fellowships from the Damon Runyon-Walter Winchell Cancer Research Fund and the National Institutes of Health. This research was supported through a grant from the National Institutes of Health (NS26295) to E. H.