The Caenorhabditis elegans lim-6 LIM homeobox gene regulates neurite outgrowth and function of particular GABAergic neurons

Dedicated neural circuits which mediate specific behaviors have been defined in Caenorhabditis elegans. The description of the complete neural connectivity of C. elegans as well as the behavioral defects induced by laser or genetic ablation of specific neurons has led to the definition of neural circuits that mediate, for example, mechanosensory, locomotory, thermosensory and rhythmic defecation behaviors (Avery and Thomas, 1997; Chalfie et al., 1985; Hart et al., 1995; McIntire et al., 1993a,b; Miller et al., 1992; Mori and Ohshima, 1995; Thomas, 1990; White et al., 1992). Not surprisingly, specific neurotransmitter synthesis, synaptic signaling, and receptor signaling pathways have been implicated in the function of most of the pathways (Bargmann and Kaplan, 1998). But genetic analysis has also revealed that transcriptional control mechanisms regulate both early stages of neural cell fate determination and events late in neural differentiation and in the mature nervous system. For example, the unc-4 homeobox gene determines the pattern of interneuron connectivity to motor neurons from its site of expression within motorneurons (Miller et al., 1992; White et al., 1992), and the unc-55 nuclear hormone receptor is required for the correct wiring of VD type motorneurons (Zhou and Walthall, 1998). The analysis of several LIM homeobox genes also revealed that they regulate terminal steps in the differentiation of specific components of the mechanosensory (Way and Chalfie, 1988, 1989) or thermosensory circuits (Hobert et al., 1998, 1997).

LIM homeobox genes represent a large subfamily of homeobox genes; although they act in a variety of different developmental contexts, a common theme is their regulatory role in neural development (Dawid et al., 1998). In Drosophila, a requirement for LIM homeobox genes in axonal fasciculation, pathfinding and neurotransmitter choice was demonstrated for the apterous, lim3 and islet genes, respectively (Benveniste et al., 1998; Lundgren et al., 1995; Thor and Thomas, 1997; Thor et al., 1999). In vertebrates, LIM homeobox genes appear to play a similar role in terminal steps of neural differentiation, although it is clear that they have also been recruited to several additional functions outside the nervous system, such as early embryonic inductions or limb patterning (Dawid et al., 1998). In the vertebrate spinal cord, the temporal and spatial patterns of LIM homeobox gene expression implicates them in control of motorneuron differentiation (Tsuchida et al., 1994). Loss of function of the mouse Isl-1 gene leads to a failure of motorneuron differentiation (Pfaff et al., 1996), Lhx-3 and Lhx-4 determine motorneuron subtype identities (Sharma et al., 1998) whereas loss of function of the Lhx-2 gene causes a variety of brain developmental defects (Porter et al., 1997).

Here we analyse the function of the C. elegans LIM homeobox gene lim-6 and compare it with its mammalian orthologs Lmx-1a and Lmx-1b. The Lmx-1a LIM homeodomain protein was originally identified as a regulator of insulin gene expression (German et al., 1992). The closely related Lmx-1b gene (100% identity in homeodomain) is expressed in the central nervous system, including regions of the hindbrain and in particular domains in the spinal cord (Chen et al., 1998; Matise and Joyner, 1997; Riddle et al., 1995). Lmx-1b mutant mice display CNS patterning defects (R. Johnson, personal communication). However, the analysis of Lmx-1b has so far been focused on its function outside the nervous system, mainly in limb and kidney development, where it is required for specific patterning and differentiation events (Chen et al., 1998; Dreyer et al., 1998; Riddle et al., 1995; Vogel et al., 1995). Moreover, Nail Patella Syndrome, an autosomal dominant disorder characterized by renal defects and various forms of skin and limb hypoplasias is caused by a heterozygous null mutation in human Lmx-1b (Dreyer et al., 1998). Thus the normal two copy gene dosage of Lmx-1b is essential for these patterning events. We show that the single C. elegans Lmx-1a/b ortholog lim-6 is expressed in particular neurons, many of which are GABAergic, in the excretory system and in epithelial cells of the uterus. Using a lim-6 deletion mutant we show that lim-6 is required for the terminal differentiation of sensory- and motorneurons, and for morphological aspects of uterine development. The behavioral and neuroendocrine defects of the lim-6 mutant animals are mostly due to defects in maturation of GABAergic neurons that depend on lim-6 function. The neuronal defects of lim-6 mutant animals resemble those of other LIM homeobox genes in C. elegans, thus pointing to a common theme in the function of this gene family in C. elegans. Moreover, the sites of Lmx-1 and lim-6 expression also suggest that specific functions of these genes may be conserved during evolution.

Wild type were N2 Bristol, lim-6(nr2073), unc-25(e156), unc-30(e191), daf-7(e1372), unc-1(e719), unc-33(e204), unc-36(e251), unc-73(e936), unc-76(e911), egl-38(n578), lin- 11(n389), lin-17(n671), che-3(e1379), pha-1(e2123ts), aex-2(sa3), sem-4(n2654).

Transgenic lines were either created in a wild-type background using pRF4 (rol-6(su1000)) at 100 ng/μl or mec-7::GFP at 50 ng/μl as an injection marker or in a pha-1(e2123ts) mutant background using pBX (pha-1 wild-type expression construct) at 100 ng/μl as the rescuing construct for the pha-1 lethality (Granato et al., 1994). The lines are:

pha-1(e2123ts); mgEx[lim-6r::GFP; pBX] (4 independent lines)

pha-1(e2123ts); mgEx[lim-6prom::GFP; pBX] (3 independent lines)

pha-1(e2123ts); mgEx[lim-6int3::GFP; pBX] (8 independent lines)

pha-1(e2123ts); mgEx[lim-6up::GFP; pBX] (1 line)

N2; mgEx[GADcompl::GFP; pRF4] (8 independent lines)

mgEx446, mgEx447 = N2; mgEx[fl-lim-6-2; mec-7::GFP] (2 independent lines)

mgEx408, mgEx409 = N2; mgEx[fl-lim-6VP16-2; mec-7::GFP] (2 independent lines)

juIs8: integrated pSC381 (Jin et al., 1999)

oxIs12: McIntire et al. (1997) 

nuIs1: Hart et al. (1995) 

nuIs26: S. Nurrish and J. Kaplan (unpublished)

lin-15(n765) X; adEx1262 [lin-15(+) gcy-5::GFP] (Yu et al., 1997)

lin-15(n765) X; adEx1297 [lin-15(+) gcy-6::GFP] (Yu et al., 1997)

lin-15(n765) X; adEx1288 [lin-15(+) gcy-7::GFP] (Yu et al., 1997)

Unless noted otherwise below, the expression plasmids were constructed by amplifying the respective sequences either from wild-type genomic DNA or the K03E6 cosmid and subcloning of the PCR products into the pPD95.75 expression vector. The respective constructs contain the following sequences (the numbering originates from an arbitrary numbering of the K03E6 cosmid; the ATG startcodon is at position 14254, the stop codon ends at position 10541).

lim-6r::GFP: bp 18290-10544

lim-6prom::GFP: bp 18290-12846

lim-6int3::GFP: bp 12785-11584

lim-6up::GFP: bp 20219-16193

fl-lim-6-2: bp 18290-10544 (+ unc-54 3′UTR, without GFP)

lim-6VP16-2: As fl-lim-6-2, except that the acidic activation domain of VP16 (aa 411-aa 490) was cloned into a BamHI site that had been engineered at the C terminus of lim-6.

GADcompl::GFP: The GAD/unc-25 gene is from YAC y37d8contig.03366. Its structure was predicted by Genefinder. The GFP fusion was generated by PCR amplification of the complete gene including 1846 bp of upstream sequence (up to the preceding predicted gene) and the full coding sequence. The amplification product was fused to GFP using a PCR fusion approach with overlapping PCR primers (O. H. and G. R., unpublished). The fusion amplification product was directly injected into adult animals. The sites of expression of this GADcompl::GFP construct was similar to the expression construct pSC381/juIs8 described by Jin et al. (1999).

lim-6(nr2073) was kindly provided by NemaPharm, Inc. It was isolated from an EMS-induced C. elegans deletion library which included approximately 400,000 mutagenized chromosomes using a PCR based sib-selection procedure (Jansen et al., 1997). The mutant strain was backcrossed five times. Its genotype was confirmed using a triplex PCR with 3 primers, two of which flank the deletion (yielding a PCR product of 0.9 kb on the deleted chromosome), the third is located within the deletion (yielding a PCR product of 1.1 kb on the wild-type chromosome). The exact deletion point was determined by DNA sequencing. All the phenotypes reported here were linked with the lim-6 deletion nr2073 through the five backcrosses. Moreover, the nr2073 mutant phenotypes were complemented by a wild type lim- 6(+) transgene and phenocopied by expression of a dominant negative lim-6 gene product (described below). For the rescuing approach, we introduced two independent extrachromosomal arrays, mgEx446 and mgEx447, which contain a wild-type copy of lim-6, termed fl-lim-6-2 into wild-type animals. fl-lim-6-2 contains the same genomic region as the GFP construct lim-6r::GFP shown in Table 1. To cross these arrays into lim-6(nr2073) the mgEx446 and mgEx447 transgenes were mated into unc-1(e719). Then these animals were crossed with the lim-6(nr2073) strain and the closely linked unc-1 chromosome was segregated away to yield a homozygous lim-6 mutant carrying the transgene. The genotype was verified by PCR. The resultant strain lim-6(nr2073) mgEx446/447 at least partially rescues the uterine defects and defecation defects. We noted that the same mgEx446/447 arrays cause moderate defecation defects in some wild-type animals, suggesting that either overexpression of lim-6 or promoter titration effects cause similar defects to its loss of function. This observation is a potential explanation for why not all lim-6(nr2073) mgEx446 animals are completely rescued. We also constructed a lim-6::VP16 expression construct (lim-6VP16-2), which by increasing transactivation function of LIM-6 was expected to enhance rescue (as exemplified by ttx-3::VP16, O. H., I. Mori and G. R., unpublished data). Extrachromosomal arrays expressing the VP16 fusion constructs (mgEx408 and mgEx409) indeed completely rescued the lim-6 mutant phenotype; 20/20 animals showed intact uterine lumens; 19/20 animals restored gcy-5 expression in ASER; 3/3 showed intact EMC cycles.

The second approach to show linkage of the defects with lim-6 consisted of expressing a dominant negative derivative of lim-6 in which we deleted from the fl-lim-6-2 expression construct the C- terminal 72 amino acids (including parts of the homeodomain) and replaced it by GFP. Expression was similar to that seen for lim-6r::GFP (Table 1). We observed strong and highly penetrate constipation defects and brood size reductions, which thus phenocopies the lim-6(nr2073) defects.

FMRFamide antibody staining was performed as previously described by Schinkmann and Li (1992).

The defecation assay was described by J. Thomas (1990). Well fed animals were scored as young adults. Muscle contractions of individual animals were scored using a stereomicroscope at 50x magnification. Each animal was observed for 5 to 15 defecation cycles. The time elapsing between the pBoc and the EMC and the intercycle time was recorded. An EMC was only scored as complete if the release of gut contents was observed.

Egg laying behavior was scored by picking single L4 staged animals to a fresh plate and transferring them every day to a fresh plate until no further progeny was produced. The total amount of progeny was counted.

Dauer arrest was scored with non-starved, non- crowded animals. Adult animals of the respective genotype were allowed to lay eggs for 4-12 hours; those eggs were kept at the respective temperature and the larvae hatching from these eggs were scored for dauer characteristics 3-7 days after the egg lay.

To reveal the expression pattern of lim-6, we fused lim-6 genomic regions to green fluorescent protein (GFP) and generated transgenic lines bearing these fusion genes (Table 1). One of these reporter genes, lim-6r::GFP, contains precisely the same lim-6 non-coding and coding regions as a genomic region that complements the null phenotype of lim-6 (fl-lim-6-2; Table 1); moreover, most cells expressing lim-6 reporter genes show defects in lim-6 null mutant animals (described below). Thus, with the usual caveats of reporter gene expression studies, these observations indicate that the lim-6r::GFP constructs reveals authentic sites of endogenous lim-6 expression.

The lim-6r::GFP fusion gene reveals expression in a restricted set of neurons, epithelial cells of the uterus and the excretory system (Table 1; Fig. 2). Reporter gene expression in the nervous system begins late in embryogenesis at about 300 minutes of development, which is after these neurons have been generated and while they initiate neurite outgrowth (Sulston, 1983). After hatching, lim-6r::GFP is expressed in one chemosensory neuron, ASEL, and in eight inter- and motorneurons (Fig. 2A-E). Most of these neurons are GABAergic, namely RMEL/R, AVL, RIS and DVB (McIntire et al., 1993b). RMEL/R, AVL and DVB are motorneurons which innervate specific sets of head and enteric muscles, respectively, whereas RIS is an interneuron (White et al., 1986). The other three neurons, PVT and RIGL/R, express the neuropeptide FMRFamide (Schinkmann and Li, 1992). The expression of lim-6 in all neurons continues throughout adulthood, indicating that lim-6 may continue to function in the mature nervous system.

The expression of lim-6 in the uterus is dynamic. lim-6 reporter gene constructs (lim-6prom::GFP, lim-6up::GFP, lim-6r::GFP, Table 1) start to be expressed in two uterine cells during the late L3/early L4 stage and the expression widens during the L4 stage to include the uv2 and uv3 cells, several uterine toroid (ut) cells, which form the lumen of the uterus and at least one cell type (sujn) of the spermatheca-uterine junction (Fig. 2F,G). Occasionally, weaker and less consistent expression can be observed in some cells of the distal side of the spermatheca, which connect the spermatheca to the rest of the somatic gonad. Expression of lim-6 in the uterus is absent in adults suggesting that lim-6 function is specific for the stages of uterine development. Consistent with these uterine cell identities, we could not observe any major changes of lim-6 expression in animals that are mutant for genes that affect other cells of the developing uterus, such as egl-38, which affects the fate of the uv1 cells (Chamberlin et al., 1997) or lin-11, which affects the utse cells (Newman et al., 1996) (data not shown).

Considering the expression and function of vertebrate Lmx- 1b in the kidney, it is intriguing that lim-6 is expressed in the excretory system of C. elegans. The C. elegans excretory system is composed of four cell types (Nelson et al., 1983), one of which, the A-shaped excretory gland cell, expresses lim-6 from late embryogenesis throughout adulthood (Fig. 2B).

Analysis of lim-6 promoter deletion derivatives reveals that the lim-6 promoter is composed of separable regulatory elements that are specific for individual cell types as shown in Table 1. The regulatory element for uterine and sensory neuron expression is localized upstream of the transcriptional start site, while the regulatory element for lim-6 expression in several GABAergic neurons is located in the third intron.

To determine the function of lim-6, a lim-6 deletion mutant was isolated and phenotypically characterized. This deletion allele, nr2073, was isolated by NemaPharm, Inc. (Cambridge, MA) using a PCR screen of a mutagen induced deletion library. The nr2073 allele harbors a 1.7 kb deletion in the lim-6 gene which deletes three quarters of the lim-6 coding region, including the second LIM domain and the homeodomain (Fig. 1A). Since the DNA binding homeodomain represents the crucial functional feature of LIM homeodomain proteins, it is very likely that nr2073 represents a strong loss-of-function allele, presumably a null allele. After extensive backcrossing of the mutant, we undertook a phenotypic analysis of lim-6(nr2073). lim-6(nr2073) animals are viable, move normally and show no gross morphological abnormalities. However, the animals do not expel the gut contents normally and have reproductive defects. In addition, upon a detailed characterization of particular neural markers, lim-6 mutant animals show defects in the pattern of gene expression and function of GABAergic neurons (see below). The defects of the mutant animals are summarized in Table 2. All of the cellular defects that we observe correspond to the defects expected based on the sites of lim-6 expression described above. Both the neural and non- neural defects of lim-6(nr2073) could be rescued with a genomic lim-6 transgene that contains all the regulatory elements described above in the lim-6r::GFP reporter gene construct (see Material and Methods). Moreover, most of the defects can be phenocopied by expressing, in wild-type animals, a dominant-negative version of the LIM-6 protein under its own promoter (data not shown; see Materials and Methods), further supporting that the nr2073 deletion allele causes the defects that we observe.

lim-6 mutant animals display defective defecation behavior, as revealed by the bloated appearance of their gut (Fig. 3A). Periodic contractions of the enteric muscle are required to release the gut content (Avery and Thomas, 1997). The enteric muscle contractions (EMC) are regulated by the motorneurons AVL and DVB which innervate the enteric muscles (Liu and Thomas, 1994; McIntire et al., 1993b). The expression of lim-6 in AVL and DVB and the bloated visible phenotype of the lim- 6 mutant animals suggested that lim-6(nr2073) mutants may have defects in the rhythmically executed EMCs triggered by AVL and DVB. In wild-type animals, the EMCs are intricately linked to two other muscle contraction steps, posterior body wall contractions (pBoc) and anterior body wall contractions (aBoc). The activation of these muscle contractions is precisely timed and represents an ultradian rhythm that is temperature compensated and can be entrained by external stimuli (Avery and Thomas, 1997; Liu and Thomas, 1994; Thomas, 1990). The EMC step of this motor program is activated 3-5 seconds after the pBoc step of the motor program; after a 40-55 second intercycle time, the pBoc (and its closely linked aBoc step) is activated again and is again followed by the EMC and so on. While in wild-type animals the EMC is activated in 88% (n=82) of all cycles with a correct timing of 3-5 sec after the pBoc, in lim-6 mutant animals the EMC is activated only in 25% (n=205) of the cycles; in 75% of the cycles, the EMC is either entirely absent (59%) or significantly delayed relative to the stereotyped pBoc step (16 %) (Fig. 3). The overall cycle length, i.e. the time between the intact pBoc steps, is not affected.

The GABAergic AVL and DVB neurons are essential for the muscle contraction step of the defecation cycle and proteins involved in GABAergic signaling, such as the GABA- synthesizing enzyme UNC-25 or the GABA transporter UNC- 47 are expressed in these neurons and required for the execution of the cycle (Jin et al., 1999; Liu and Thomas, 1994; McIntire et al., 1993b, 1997; Thomas, 1990). The expression of lim-6 in these neurons suggests that the focus of lim-6 action in the defecation cycle is these neurons. The impact of lim-6 loss of function on the neuroanatomy of AVL and DVB is consistent with this model (see below).

The expression of lim-6 in RMEL/R, two out of a set of four GABAergic head motorneurons (RMEL, RMER, RMED, RMEV) required for head foraging (Hart et al., 1995) indicates a possible lim-6 function in these neurons. Although we have not tested head foraging behavior in detail, a preliminary analysis suggests that lim-6 mutant animals do indeed display abnormal nose movements (S. Nurrish, personal communication).

The lim-6 mutant defects in GABAergic mediated behaviors suggests that the LIM-6 transcription factor mediates a developmental step in the differentiation of these neurons. We attempted to define the cellular basis of these defects by examining the neural fate and the neuroanatomy of lim-6- expressing neurons in the lim-6 mutant animals. As a first step to monitor the cell fate of the lim-6-expressing neurons, we crossed the lim-6int3::GFP and lim-6prom::GFP reporter gene constructs (Table 1) into lim-6 mutant animals. The expression of these reporter genes as well as other reporter genes described below is largely unaffected in the lim-6 mutant, thus revealing that lim-6-expressing neurons are generated and continue to survive in the mutant (data not shown).

The cell fate of lim-6-expressing neurons was further analyzed using three differentiation markers of the neural types that express lim-6. First, we examined the expression of the GABA vesicular transporter unc-47 (McIntire et al., 1997) in lim-6(nr2073) and found it to be correctly expressed in AVL and DVB, as well as the other GABAergic neurons that express lim-6, RMEL/R and RIS (Figs 4, 5). Thus, expression of the GABA transporter unc-47 is not dependent on lim-6 gene activity. Second, we examined the expression of the glutamate receptor glr-1, which is a cell fate marker for the DVB and RME motorneurons (Hart et al., 1995); like lim-6, glr-1 is only expressed in the right and left types of the otherwise fourfold symmetric RME motorneurons. lim-6 mutant animal containing a chromosomally integrated glr-1::GFP reporter gene, nuIs1, show unaltered expression of this neurotransmitter receptor in RMEL/R and DVB compared to wild-type animals (data not shown). Third, we observed FMRFamide expression in lim-6(nr2073) animals using anti- FMRFamide antibodies. FMRFamides are neuromodulatory peptides expressed in a subset of neurons in C. elegans, including the DVB, PVT and RIGL/R neurons (Schinkmann and Li, 1992); FMRFamide expression is correctly specified in DVB, PVT and RIGL/R in lim-6(nr2073) mutant animals (Fig. 4). In summary, this analysis demonstrates that lim-6 is not required for the lim-6- expressing neurons to be generated nor to assume certain aspects of their respective identities.

Close examination of the neuroanatomy of the DVB motorneuron, using the unc-47::GFP fusion gene reveals severe neuroanatomical defects in the lim-6 mutant animals (Table 3). 73% of the animals display additional small axons of varying lengths that either emanate directly from the DVB cell body or from the main axonal process in close vicinity of the cell body (Fig. 4). These extra axons often have additional branches. This axonal morphology is very rarely observed in wild-type animals (Table 3). Due to the proximity of AVL to other neurons in the head ganglia, we were unable to observe whether AVL displays similar sprouting defects.

The main axonal trajectories of the DVB and AVL motorneurons are also affected by the lim-6 null mutant. Both of these neurons are monopolar and extend their single axonal process along the ventral nerve cord (VNC) (White et al., 1986). The visualization of these axons by unc-47::GFP is obscured by the processes of the GABAergic D-type motorneurons in the VNC. However, the expression of unc- 47::GFP in the D-type neurons can be disrupted without affecting expression in the other GABAergic neurons by a mutation in the unc-30 homeobox gene. In unc-30(e191) animals, the only VNC axons that contain unc-47::GFP are those of the RMEV, AVL and DVB motorneurons (Basson and Horvitz, 1996; Jin et al., 1994), and the axon projections of these neurons are normal (Fig. 4). We observed, however, that in more than half of the unc-30(e191); lim-6(nr2073) mutant animals, the main axonal process of the DVB motorneuron displays abnormalities along the VNC (Fig. 5; Table 3). Either the axon does not terminate at its normal termination position at the vulva, or it takes an abnormal path outside the VNC, or it turns at the vulva and projects aberrantly (Fig. 5). The main axon of the AVL motorneuron, which runs along the entire length of the VNC was more difficult to examine since it is obscured in the anterior quarter of the VNC by the RMEV motorneuron, which also expresses unc-47::GFP (but not lim-6). Nevertheless, in the majority of lim-6 mutant animals, the AVL motorneuron does not extend past the RMEV motorneuron (Table 3; Fig. 5). In summary, both the DVB and the AVL motorneurons require lim-6 either to acquire or maintain intact axon morphology.

The PVT neuron is a presumptive guidepost of the VNC and is possibly involved in directing axons of the ventral cord (Wadsworth et al., 1996). Since PVT expresses lim-6, we examined whether a presumptive PVT function in VNC organization requires lim-6. VNC organization was monitored using a cat-1::GFP reporter gene construct (S. Nurrish and J. Kaplan, personal communication) and a lin-11::GFP reporter gene construct (Hobert et al., 1998), each of which is expressed in axons of the right and left VNC. Transgenic lim-6 mutant animals expressing these reporter genes revealed no obvious defect in VNC organization.

We conclude that lim-6 is not necessary for neurons to be generated and to execute certain aspects of their differentiation program. However, their patterns of neurite outgrowth depend on lim-6 gene activity.

Glutamic acid decarboxylase (GAD) is the rate limiting enzyme for synthesis of the neurotransmitter γ-aminobutyric acid (GABA) and is encoded by the C. elegans unc-25 gene (Jin et al., 1999). A translational fusion of the promoter and the first 13 codons of unc-25 to GFP is expressed in all GABAergic neurons (Jin et al., 1999). The expression of this fusion gene in lim-6(nr2073) mutant animals is significantly decreased in the GABAergic neurons DVB and RIS and to a lesser extent in AVL (Fig. 6; Table 4). unc-25::GFP expression is normal in the RMEL/R motorneurons, which normally express lim-6, and in the RMED/V motorneuron and the D- type motorneurons of the VNC, neither of which express lim-6. A translational GFP fusion to unc-25, that encompasses the full coding sequence of unc-25 (see Material and Methods) showed a similar pattern of expression in GABAergic neurons and was also affected by the absence of lim-6 gene activity; while 96% (n=23) of wild-type animals show strong expression in DVB, only 28% (n=25) of lim-6(nr2073) mutant animals show equally strong expression, 44% show weak expression and 28% show no expression (data not shown). We conclude that lim-6 contributes to the cell-type-specific regulation of unc-25 expression in the DVB, RIS and AVL neurons but because the lim-6 effects are quantitative, other factors also contribute to unc-25 expression.

We tested whether the neuroanatomical defects of the DVB motorneuron in the lim-6 mutant might be related to defects in GABAergic signaling, perhaps due to LIM-6 regulation of unc-25 expression. For example, active DVB synaptic signaling may be required for normal axon morphology in a manner similar to other activity-dependent structural changes in the nervous system (Dahm and Landmesser, 1988, 1991). To observe DVB morphology in animals lacking GABA signaling, we observed DVB axon morphology using the unc-47::GFP reporter gene in GABA-deficient unc-25 mutant. We found that even in the complete absence of GABA, the DVB motorneuron maintained its correct axon morphology (Table 3). An electron microscopical reconstruction of the synaptic connectivity of unc-25 mutant animals has led to a similar conclusion (Jin et al., 1999).

It is conceivable that lim-6 indirectly affects unc-25 expression through the effects of lim-6 on axonal pathfinding. However, this possibility is unlikely since several genes known to affect axonal pathfinding of GABAergic neurons, such as unc-34, unc-71 or unc-76, do not affect GABA synthesis and thus unc-25 expression (McIntire et al., 1992). Thus, it is most likely that the effect of lim-6 on DVB axonal morphology represents a pathway that is parallel to its regulation of unc-25 expression.

Sensory inputs to C. elegans not only elicit motor outputs. As in other animals, there are endocrine outputs of the nervous system. For example, sensory detection of a pheromone controls neuroendocrine outputs mediated by insulin-like and TGFβ-like signaling molecules that in turn regulate the metabolism of the whole animal and regulate whether development is arrested at the dauer diapause stage or continues to the reproductive adult stages (Riddle and Albert, 1997). A variety of sensory inputs are integrated by distinct neural circuits, for example food signals, crowding and temperature, to affect the neuroendocrine control of dauer stage entry (Riddle and Albert, 1997). We noted that unlike wild-type animals lim-6(nr2073) animals occasionally form dauers under conditions in which the animals are fairly crowded yet not completely starved; their pheromone responsiveness was unaffected, however (data not shown). As these dauer inducing conditions were difficult to reproduce, we investigated lim-6 regulation of dauer arrest by sensitizing the genetic background to make the animals more prone to enter the dauer stage. We used a strain mutant for the daf-7/TGFβ gene, a neuroendocrine signaling molecule that synergizes with an insulin-like signaling cascade to repress entry into the dauer stage (Kimura et al., 1997; Ogg et al., 1997; Ren et al., 1996). In the absence of daf-7/TGFβ signaling, the animals are sensitized to enter the dauer stage; e.g. they now become hypersensitive to specific environmental stimuli such as changes in their ambient temperature (Riddle and Albert, 1997). Consistent with the enhanced dauer arrest under weakly dauer inducing conditions of the lim-6 mutant alone, lim-6(nr2073); daf-7(e1372) double mutant animals are significantly enhanced for dauer arrest compared to daf- 7(e1372) at 15°C (Table 5).

In vertebrates, the neurotransmitter GABA is expressed in several neuroendocrine tissues and affects the release of insulin (Gu et al., 1993; Sorenson et al., 1991). Considering the expression of lim-6 in GABAergic neurons we examined whether defects in GABAergic signaling could phenocopy the effects of lim-6(nr2073) on dauer arrest. unc-25(e156) mutant animals, which are GABA deficient (McIntire et al., 1993a), are neither defective in dauer arrest, nor do they enter the dauer stage constitutively (Table 5). However, like lim-6(nr2073), the unc-25(e156) mutation strongly enhances daf-7(e1372) at 15°C (Table 5). This effect of unc-25 is not due to a requirement for unc-25 in the D-type ventral cord motorneurons, since unc-30, a gene that abolishes GABAergic cell fate in the D-type motorneurons, but not in the other GABAergic neurons (Jin et al., 1994), does not enhance daf-7(e1372) (Table 5). The enhancement of daf-7(e1372) by unc-25(e156) is unlikely to be a non-specific secondary consequence of constipation, since neither aex-2(sa3) mutant animals, which are 100% EMC defective (Liu and Thomas, 1994) nor sem-4(n2654) mutant animals, which are 84% EMC defective (Basson and Horvitz, 1996) enhance daf-7 induced dauer formation to a similar degree as unc-25 (Table 5). We conclude that GABA signaling normally activates reproductive development or represses dauer arrest. Since the absence of functional D-type motorneurons in unc-30 does not effect dauer formation, GABA regulation of dauer arrest is most likely mediated through either the RME, AVL, RIS or DVB neurons. As each of these neurons expresses lim-6 and as the absence of lim-6 gene activity phenocopies the absence of GABAergic function we consider it most likely that lim-6 activity in these neurons is required for their functional specification to in turn affect GABA control of dauer arrest. For example, neurons postsynaptic to these GABA neurons may secrete DAF-7 or one of the dozens of insulin-like genes that are likely to regulate dauer arrest (S. Pierce, L. Liu and G. R., unpublished).

Many neurons in the nervous system of C. elegans come in bilaterally symmetric pairs (White et al., 1986), yet lim- 6::GFP reporter gene constructs are only expressed in the left neuron (ASEL) of the ASER and ASEL chemosensory pair (Fig. 2, Table 1; Bargmann and Horvitz, 1991). Using the ASEL-expressed lim-6prom::GFP reporter gene construct, we found that ASEL is generated and pathfinds normally in lim-6 mutant animals (data not shown). We examined the expression of the two ASE differentiation markers gcy-6 and gcy-7, two putative guanylyl cyclase-type sensory receptors that like lim-6 are exclusively expressed in ASEL, but not in ASER (Yu et al., 1997). We found that gcy-6 and gcy-7 GFP reporter gene constructs are correctly expressed in lim-6 mutant animals (Fig. 7), suggesting that lim-6 gene activity is not necessary for these aspects of ASEL neural maturation. However, lim-6 activity is required for the correct expression of another putative guanylyl cyclase-type sensory receptor, gcy-5. A gcy- 5 reporter gene construct is normally expressed in the right neuron, ASER, but not in ASEL (Yu et al., 1997). In lim-6(nr2073) mutant animals, the gcy-5 reporter gene is ectopically expressed in ASEL (Fig. 7). Thus, lim-6 is normally required for repression of gcy-5 expression in ASEL. The LIM- 6 transcription factor could either activate a repressor that directly acts on the gcy-5 promoter or LIM-6 could directly bind and repress the gcy-5 promoter. It is not known to what sensory inputs the gcy-5 receptor responds. Consequently, the physiological significance of disregulated gcy-5 receptor expression in lim-6 mutant animals is hard to assess at the moment. We did, however, examine lim-6(nr2073) animals for their responsiveness to various water-soluble chemicals that are sensed by ASEL and ASER, but could not detect any defects (data not shown).

The asymmetric expression of lim-6 in ASEL does not appear to be neural activity dependent. The abrogation of sensory inputs or the disturbance of correct connectivity have no influence on the asymmetric expression, as revealed by crossing lim-6prom::GFP in various che and unc mutants (che- 3, daf-19, unc-33, unc-36, unc-73; data not shown). It may be relevant that besides ASEL most of the neurons that express lim-6 (RIS, AVL, DVB, PVT) are asymmetrically generated in the sense that they do not come as bilaterally symmetric pairs.

The expression of the vertebrate lim-6 homolog Lmx-1 is regulated by a vertebrate wingless homolog (Riddle et al., 1995; Vogel et al., 1995); wingless signaling in C. elegans is involved in determining asymmetric features of cell division (Sawa et al., 1996). We tested whether lin-17/wingless mutants affect the asymmetry of lim-6 expression, but found no effects (data not shown). Other wingless like genes that are present within the C. elegans genome (Ruvkun and Hobert, 1998) might be involved in regulating asymmetric features of lim-6 expression.

The excretory gland cell displays a neuron-like appearance; it contains specific types of secretory vesicles of unknown function and extends thin processes towards the nerve ring, where it receives synaptic input (Nelson et al., 1983). To examine the effects of the lim-6(nr2073) mutation on excretory gland cell morphology and function, we attempted to observe the cell with the lim-6prom::GFP reporter gene, which labels the entire gland cell in wild-type animals. Interestingly, the lim-6 reporter gene construct is not expressed in the excretory gland cell of lim-6 mutant animals, whereas its expression in the other neurons and epithelial cells is unaffected (data not shown). This suggests that LIM-6 may autoregulate its own expression in the excretory gland cell or that there are severe differentiation defects of the excretory gland cell. Because laser ablation of the excretory gland cell has no significant impact on growth, moulting, osmoregulation, fertility, longevity, and dauer larva formation (Nelson and Riddle, 1984), we did not assess the functional consequences of the absence of lim-6 activity in the excretory gland cell. Vertebrate Lmx-1 is required for the specification of basement membrane structures in the kidney. An ultrastructural analysis of lim-6 mutant animals will be required to examine whether lim-6 might have a similar function in C. elegans.

We found that the extension of excretory gland cell processes uses guidance mechanisms similar to those used by neurons, since several neural pathfinding genes, such as unc-33, unc-73 and unc-76 (Hedgecock et al., 1987), affect excretory gland cell morphology as monitored by crossing lim- 6prom::GFP into these mutants (data not shown). Since lim-6 is required to specify the correct neuroanatomy of several neurons, as described above, it is conceivable that by regulating a similar set of target genes lim-6 also affects the process morphology of the excretory gland cell.

lim-6 mutant animals have a significantly reduced brood size (Table 6). Moreover, most of the progeny of lim-6 mutant animals hatch within the parent animal. The uterine expression of lim-6 prompted us to examine the morphology of the uterus in the lim-6(nr2073) mutant. The main tissue types of the uterus are the epithelial uterine toroid (ut) cells, which line the uterine lumen and which express lim-6 (see above), and specialized utse and uv cells, which constitute the connection between the uterus and the vulva (Newman et al., 1996). Using DIC microscopy we found that the uterine lumen is not formed correctly in 78% of lim-6 mutant animals and appears to be clogged (Fig. 8; Table 6). The extent of clogging varies from animal to animal and can range from a simple ‘bridge’ as shown in Fig. 8 to a complete closure of the uterine lumen. The tissue that clogs the uterine lumen expresses lim-6 and could represent ut cells that have not separated correctly (Fig. 8). The lim-6(nr2073) egg laying defects could be explained by the animal’s inability to accommodate eggs in the uterus. A hindered passage through the uterus would congest the gonad and inhibit further fertilizations, thus explaining the low brood size of the animals (which is even smaller than those of vulvaless animals, such as lin-11(n389); Freyd, 1991). However, the uterine closure defect is not 100% penetrant, while the low brood size clearly is. This argues for additional defects in the reproductive system. The expression of lim-6 in the junction cells of the spermatheca described above could indicate that eggs in lim-6(nr2073) mutants might have problems in passing through the spermatheca thus reducing fertility. In summary, we have shown that lim-6 is required for uterine morphogenesis.

lim-6 is one of seven C. elegans LIM homeobox genes revealed by the genome sequence. Most of these LIM homeobox genes detect mammalian orthologs (Fig. 1). The determination of the function of these C. elegans LIM homeobox genes with precise cellular identifications and correlations with neuron type and function should broadly indicate the function of these genes across phylogeny. lim-6 is an interesting case because while the function of the mammalian ortholog Lmx-1b was much better understood than lim-6 before this study, our genetic and cellular analysis suggests new functions to search for in mammals.

Our results show that lim-6 functions mainly in GABAergic neurons that regulate enteric muscles and endocrine outputs, and uterine development. It is not yet clear which of the C. elegans lim-6 functions are generalizable to its mammalian orthologs Lmx-1a and Lmx-1b. It is possible that some of the lim-6 functions are served by one ortholog and others by the other. Consistent with such a model, the regulation of neuroendocrine signals by both C. elegans lim-6 (dauer regulation) and mammalian Lmx-1a (insulin regulation) suggests common, ancestrally related functions. Similarly, the requirement for Lmx-1b in mammalian kidney function and the lim-6 expression in the C. elegans excretory gland cell suggests a common function in excretory system differentiation for this LIM homeobox gene. Our data also suggests that expression of mammalian Lmx-1a and Lmx-1b in the central and peripheral nervous system should be correlated with GABAergic neural markers. Lmx-1b knock-out mice exist (Chen et al., 1998) but have not been analyzed for of GABAergic neuroendocrine or enteric muscle control defects. The C. elegans function of this class of LIM homeobox genes suggests that GABAergic function (for example sensitivity to GABA-related drugs such as valium) in the Lmx-1a or Lmx-1b knockout mice should be assayed. Our data also suggests that uterine development should be analyzed. The expression of vertebrate Lmx-1 in the uterus has not been reported. However, Wnt-7a, which induces Lmx-1b expression in the developing limb (Riddle et al., 1995; Vogel et al., 1995), has recently been shown to act in uterine development (Miller and Sassoon, 1998). As in limb development the expression of Wnt-7a in the uterus might similarly be coupled to the induction of Lmx-1b expression in the uterus. However, it is also possible that some of the lim-6 functions in C. elegans may not be primitive and universal.

The most prominent site of expression of lim-6 in the nervous system are GABAergic neurons; five of the nine lim-6- expressing neurons are GABAergic. The complete GABAergic nervous system of C. elegans consists of 26 neurons (McIntire et al., 19993b). They can be functionally subdivided into the D-motorneurons of the ventral nerve cord, which are required for locomotion, the AVL/DVB motorneurons required for enteric muscle contraction, the RME motorneurons required for foraging and the RIS interneuron with an unknown function. The unc-30 homeobox gene is exclusively expressed in and required by one specific subset of the GABAergic neurons, the D-type motorneurons, for the acquisition of the GABAergic neural cell fate (Jin et al., 1994). Interestingly, the expression pattern and function of lim-6 is almost entirely complementary to the the expression pattern and function of unc-30 in the GABAergic nervous system; lim-6 is not expressed in the D-type motorneurons, but is expressed in all other GABAergic neurons (with the exception of two dorsal and ventral types of the fourfold symmetric RME motorneurons). However, lim-6 does not appear to be the functional counterpart of unc-30 in these neurons, since unlike unc-30 the lim-6 gene is not as tightly required for the expression of all GABAergic-specific cell specializations. For example, the expression of the unc-47 GABA transporter in the non-D-type GABAergic neurons is largely unaffected and the expression of the unc-25 /GAD gene is only partially affected by the lim-6 null mutation. Presumably other transcription factors can also regulate unc-25 and unc-47 expression in the non D-type motorneurons, suggesting that the genes that define GABAergic neural identity are regulated by different regulatory mechanisms in different subtypes of GABAergic cells.

The functional and neuroanatomical defects of the AVL and DVB neurons in the lim-6 null mutant strongly suggest that the function of these neurons is defective. The neuroanatomical defects which manifest as sprouting defects or the inability of the main axonal projection to follow correct paths could either represent an inability of the neuron to correctly engage its axonal outgrowth machinery, or alternatively, could represent a secondary consequence of neural activity defects. Neural activity defects have been shown before to cause axonal misrouting in both C. elegans and vertebrates (Coburn and Bargmann, 1996; Coburn et al., 1998; Dahm and Landmesser, 1988, 1991; Shatz and Stryker, 1988). lim-6 could be required for neural activity of GABAergic neurons by regulating the expression of genes directly involved in generating electric activity or regulating genes required to generate the synaptic connectivity required to receive and transmit electrical signals. The defects in non-D-type GABAergic function of the lim- 6 mutant can account for the defects in defecation cycle that the animals show. Defecation behavior requires the activation of the defecation motor program. This program represents a rhythmic behavior composed of three stereotyped muscle contractions; one of these rhythmic contractions, the enteric muscle contraction is controlled by two GABAergic motorneurons, AVL and DVB (Liu and Thomas, 1994; McIntire et al., 1993b). Mutations in GABAergic signaling, for example in the unc-47 GABA transporter or the unc-25 GABA synthesizing enzyme, cause very similar enteric muscle contraction defects (McIntire et al., 1993a). Similarly, GABAergic neurons are also part of the neural circuit that controls gut function in vertebrates (Jessen et al., 1986) and enteric GABA has been shown to regulate the peristaltic effects in the vertebrate enteric system (Grider and Makhlouf, 1992). Moreover, muscle contractions of the enteric system are also rhythmically active in vertebrates. This rhythmicity is regulated by the electric pacemaker activity of the interstitial cells of Cajal, which function as an intermediary between enteric nerves and smooth muscle cells (Sanders, 1996). It will be interesting to see if human patients with Nail Patella syndrome, a haploinsufficency of Lmx-1b, or mouse carrying Lmx-1a or Lmx-1b knockout mutations, also show defects in enteric motor function, perhaps due to a common GABAergic input to enteric muscles.

Our data favors an output from the GABAergic neurons to the neuroendocrine regulation of dauer formation which depends on lim-6 activity. We find that lim-6 loss of function strongly enhances the neuroendocrine signaling defects caused by a daf-7/TGFβ mutation. This neuroendocrine function of lim-6 is likely to be due to its function in GABAergic neurons, since unc-25 mutant animals with defects in GABA synthesis show the same phenotype. Because animals with defective D- type motor neurons do not have these neuroendocrine defects, these data argue that it is the lim-6-regulated GABAergic neurons that regulate C. elegans endocrine signaling. Although we can not exclude that lim-6 affects dauer arrest through other, as yet unidentified sites of expression, the simplest model is that lim-6 specifies functional aspects of the non-D-type GABAergic neurons that in turn regulate dauer arrest via GABAergic signaling.

The expression of lim-6 in the GABAergic neuroendocrine system also reveals an interesting similarity to the expression of its vertebrate homlog Lmx-1a. Vertebrate Lmx-1a participates in the control of insulin gene expression in the β cells of the pancreas (German et al., 1992) which also contain high levels of the neurotransmitter GABA. GABA regulates insulin secretion from the β cells via an autocrine loop (Gu et al., 1993; Sorenson et al., 1991). We find that both lim-6 and non D-type GABAergic neurons in C. elegans regulate dauer arrest, which is regulated by an insulin-like neuroendocrine pathway. It is possible that lim-6 regulates the development of the particular GABAergic neurons that in turn regulate insulin expression and secretion in C. elegans. For example, these GABAergic neurons may couple temperature or food or other sensory inputs to the endocrine output of dauer regulatory proteins such as the many insulin-like proteins that may converge on the single insulin-like receptor that clearly regulates dauer arrest. The synergizing effects of mutations in lim-6 and GABA signaling on the daf-7/TGFβ control of dauer formation indeed mirror the synergism between insulin signaling and daf-7/TGFβ signaling (Ogg et al., 1997). However, the complexity of the insulin family in C. elegans (Duret et al., 1998) complicates the matter significantly.

The major head ganglia of C. elegans display an obvious overall bilateral symmetry. The majority of neural cell types have both a left and right representative that share a comparable lineage history and morphology and make similar synaptic connections (Sulston, 1983; White et al., 1986). In those cases tested, both left and right neurons are required for a full functional response of, for example, a given pair a chemosensory neurons (Bargmann and Horvitz, 1991). The asymmetric expression of lim-6 in ASEL, but not in ASER thus represents a surprising observation. It may suggest a potential diversification of apparently symmetric sensory neurons possibly leading to a diversification in sensory function. This notion is supported by the asymmetric expression of several receptor-type guanylyl cyclases in the ASER and ASEL neurons (Yu et al., 1997), one of which we have shown to be under control of lim-6 activity. As the function of these receptor-type guanylyl cyclases are as yet unknown, we could not assess the physiological consequences of ectopic activation of one of these receptors. Gross defects in ASE sensory functions were not obvious in lim-6 mutant animals (data not shown).

It will be interesting to determine how the asymmetry of lim-6 expression is established. In vertebrates, Lmx-1b expression is under control of wingless signaling (Riddle et al., 1995; Vogel et al., 1995); wingless signaling in C. elegans is involved in determining the polarity of certain cell divisions (Sawa et al., 1996). Although we have not found an impact of the wingless-like lin-17 gene on lim-6 expression (data not shown), it is possible that any other of the five wingless-like genes present in the C. elegans genome (Ruvkun and Hobert, 1998) are involved in regulating lim-6 expression and possibly its asymmetry. The asymmetrically expressed lim-6 reporter gene construct is a tool for the genetic identification of mutants involved in establishing asymmetry in the nervous system.

Homeobox genes have been shown to act at a variety of different stages of neurogenesis, including such early steps as the determination of neural identity and later steps such as neural differentiation (Manak and Scott, 1994). The analysis of 4 of the 7 LIM homeobox genes in C. elegans demonstrates that they are required for the terminal differentiation of the neurons that express them (Hobert et al., 1998, 1997; Way and Chalfie, 1988, 1989; this study). In many respects the neural defects caused by mutations in these genes are similar. Mutations in lim-6, like mutations in the ttx-3 and lin-11 LIM homeobox genes do not affect the generation of the neurons that express them (Hobert et al., 1998, 1997) nor do they affect several aspects of neurotransmitter choice (this study and our unpublished data). However, the lim-6-, ttx-3- and lin-11- expressing neurons are functionally as well as structurally defective (Hobert et al., 1998, 1997; this report). Similarly, mec-3 is required for the structural integrity of mechanosensory neurons (Way and Chalfie, 1998). These similarities in LIM homeobox gene function suggest a common theme in the action of this class of transcriptional regulators. We speculate that each gene is required in its respective neuron to make a specific target choice and that in the absence of intact signaling partners, retrograde signaling events induce the neuron to find a signaling partner by sprouting additional processes or by inducing abnormal turns of the main axonal process. Alternatively, each gene might be required within its given neuron to directly regulate its axon outgrowth machinery. Our finding that defects in the GABA neurotransmitter synthesis do not cause axonal sprouting and the report of Jin et al. that GABAergic neurons and their postsynaptic partners show no defects in synaptic connectivity in unc-25 mutants (Jin et al., 1999) indicate that defects in synaptic outputs are unlikely to feedback on neurite outgrowth functions. However, defects in other synaptic outputs, such as signaling by neuropeptides might be feeding back into axonal sprouting or, alternatively, defects in synaptic inputs to the GABAergic neurons, caused for example by connectivity defects, could induce axonal sprouting.

Another common feature of all C. elegans LIM homeobox genes studied to date is their onset of expression in postmitotic neurons and the maintenance of their expression throughout adulthood (Hobert et al., 1998, 1997; Way and Chalfie, 1989). These observations are unlikely to be a reporter gene artefact, since the expression in other tissues, such as lim-6 in the uterus or lin-11 in the vulva is dynamic and transient. Maintained expression of a regulatory gene throughout the life of a given cell suggests, but does not prove an involvement of the particular gene in maintenance of the differentiated features of the given cell. In the case of LIM homeobox genes it is possible that they are required for the maintenance of such neural features as synaptic connectivity.

Lastly, all of the C. elegans LIM homeobox genes described so far are expressed in a largely non-overlapping subsets of neurons (Hobert et al., 1998, 1997; Way and Chalfie, 1989; this study; D. H. and G. R., unpublished). The most prominent non- neural tissue that expresses C. elegans LIM homeobox genes is the epithelium of the somatic gonad and the vulva. Intriguingly, the LIM homeobox genes are also expressed in a complementary, non-overlapping pattern in this epithelium. lin-11 is expressed in the vulva and cells that connect the uterus to the vulva (Freyd, 1991; Hobert et al., 1998), lim-6 is expressed in uterine toroid cells and in spermathecal junction cells (this study), ceh-14 is expressed in the epithelial cells of the spermatheca (T. Bürglin, personal communication) and we found another LIM homeobox gene, lim-7, to be expressed in the gonadal sheath cells (O. H. and G. R., unpublished). The unifying feature of all these cells is their highly polarized morphology and their engagement in complex morphogenetic events for which specific cell-cell contacts are required (Newman et al., 1996). These morphological features are shared by neurons as well. Moreover, epithelial cells and neurons utilize similar sorting and targeting mechanisms for specific cell surface molecules (Bredt, 1998; Rongo et al., 1998). It is conceivable that by regulating the expression of specific cell surface proteins, LIM homeodomain proteins in C. elegans are generally involved in determining the specificity of cell-cell recognition and attachment to neighboring cells both in the nervous system and in epithelial cells. The structural defects of uterine cells in lim-6 mutants indeed point to a role at least of lim-6 in cell recognition and adhesion.

We thank NemaPharm, Inc., a subsidiary of Axys Pharmaceuticals (South San Francisco, CA) for isolating the lim-6(nr2073) allele, the Caenorhabditis Genetics Center (funded by the NIH Center for Research Resources) for providing strains, Y. Liu for expert technical assistance, S. Nurrish and J. Kaplan for nuIs26, S. Nurrish for examining lim-6(nr2073) for foraging defects, C. Li for anti- FMRFamide antibodies, E. Jorgensen for oxIs12 and for comments, A. Hart and J. Kaplan for providing nuIs1, Y. Jin for providing juEx61 and juIs8, R. Johnson and T. Bürglin for communicating unpublished results and members of the Ruvkun lab for comments on the manuscript. O. H. was supported by a postdoctoral fellowship from the Human Frontiers Science Program and from the MGH Fund for Medical Discovery. K. T. was supported by the Studienstiftung des Deutschen Volkes.