Abstract
We describe a genetic model for neurodegeneration in the nematodeCaenorhabditis elegans. Constitutive activation of the GTP-binding protein Gαs induces neurodegeneration. Neuron loss occurs in two phases whereby affected cells undergo a swelling response in young larvae and subsequently die sometime during larval development. Different neural cell types vary greatly in their susceptibility to Gαs-induced cytotoxicity, ranging from 0 to 88% of cells affected. Mutations that prevent programmed cell death do not prevent Gαs-induced killing, suggesting that these deaths do not occur by apoptosis. Mutations in three genes protect against Gαs-induced cell deaths. Theacy-1 gene is absolutely required for neurodegeneration, and the predicted ACY-1 protein is highly similar (40% identical) to mammalian adenylyl cyclases. Thus, Gs-induced neurodegeneration is mediated by the second messenger cAMP. Mutations in the unc-36 and eat-4 genes are partially neuroprotective, which indicates that endogenous signaling modulates the severity of the neurotoxic effects of Gαs. These experiments define an intracellular signaling cascade that triggers a necrotic form of neurodegeneration.
Neuronal cell death is a prominent feature both of normal brain development and of particular pathological states. Neuron cell deaths can be placed into two general categories (apoptotic and necrotic) on the basis of a variety of criteria. Although a great deal is known about the molecular pathways leading to apoptotic deaths, the pathways leading to necrotic cell deaths are less well understood. Most examples of necrotic neurodegenerative deaths occur in pathological states, e.g., stroke (or other cerebrovascular injury) or neurological disorders. One hallmark of these neurodegenerative disorders is that in each case specific classes of neurons are targeted for degeneration. For example, dopaminergic neurons of the substantia nigra are lost in Parkinson’s disease, whereas in epilepsy and focal ischemia CA1 and CA3 neurons of the hippocampus are killed selectively (Pulsinelli et al., 1982; Ben-Ari, 1985).
Genes involved in inherited neurodegenerative disorders have been characterized in humans as well as in a several model organisms. Nine genes involved in inherited human neurodegenerative disorders have been cloned, including huntingtin, ataxin, and SOD. Genetic models for neurodegeneration also have been described in mice (Mullen et al., 1976; Herrup and Wilczynski, 1982; Norman et al., 1995), flies (Grether et al., 1995; Hay et al., 1995; Rangnathanan et al., 1995; Chen et al., 1996; White et al., 1996), and Caenorhabditis elegans(Driscoll and Kaplan, 1997; Hengartner, 1997). From these model systems several genes have been identified that apparently control neurodegenerative cell deaths. For example, four genes leading to necrotic forms of neuron death in worms have been described, each encoding proteins that are similar to mammalian ion channel subunits—the epithelial sodium channel homologs (MEC-4, MEC-10, and DEG-1) and a neuronal acetylcholine receptor homolog (DEG-3) (Chalfie and Wolinsky, 1990; Driscoll and Chalfie, 1991; Huang and Chalfie, 1994; Treinin and Chalfie, 1995). In these cases neuron death is thought to occur by exaggerated or toxic influx of ions, perhaps akin to excitotoxicity in mammals.
Although these genetic studies successfully have identified the many genes involved in neurodegenerative deaths, in most cases the identity of these genes has not implicated directly a defined signal transduction pathway in neurodegeneration. Alterations in G-protein-coupled phospholipase C signaling lead to retinal degeneration in Drosophila (for review, see Rangnathanan et al., 1995). A second example of G-protein-induced neurodegeneration has been reported recently. Constitutive activation of the heterotrimeric G-protein Gs causes neuronal degeneration in C. elegans (Korswagen et al., 1997). Here we show that cells differ greatly in their susceptibility to Gαs-induced killing, that the neurotoxic effects of Gαs are mediated byacy-1 (which encodes a protein that is 40% identical to mammalian adenylyl cyclases), and that endogenous neural signaling modulates the severity of Gαs-induced killing. These results define an intracellular signaling pathway by which Gs triggers a necrotic form of neuron death.
MATERIALS AND METHODS
Plasmid construction and transgene expression. The Gαs expression vector (KP#20) was constructed by ligating a 1.5 kb NcoI–XhoI fragment encoding a GTPase defective (Q227L) mutant rat Gαs cDNA containing the hemagglutinin (HA) epitope into the glr-1 expression vector CX#1 (Hart et al., 1995; Maricq et al., 1995). Amec-7:: αs(gf) expression plasmid (KP#7) was constructed by ligating the 1.5 kbNcoI–XhoI Gαs(Q227L) into themec-7 expression vector pPD52.102. Transgenic animals were prepared by microinjecting various expression constructs together with a glr-1:: gfp plasmid (KP#6), usinglin-15 as a transformation marker (Huang et al., 1994). Two stable lines carrying glr-1 expression constructs for both green fluorescent protein (GFP) and the GTPase defective Gαs (nuIs3 and nuIs5) were isolated after γ-irradiation. Unless otherwise noted, data reported in the text refer to nuIs5, which is referred to asαs(gf). Integrated lines typically greatly overexpress the transgenes. Thus, it is likely that thenuIs5-encoded Gαs is more abundant than the endogenous C. elegans Gαs.
Characterization of the neurodegeneration phenotype. Swollen or missing cells were identified by examining the morphology of GFP-expressing cells. The glr-1-expressing cells were identified previously (Hart et al., 1995; Maricq et al., 1995). Inαs(gf) animals ∼90% of the PVC neurons degenerate. Other cells degenerate at lower frequencies, including AVA, AVD, AVE, AVG, PVQ, RIG, and SMD. Cell deaths with similar morphology and similar cell type specificity were observed in both nuIs3 and nuIs5 animals. Gαs-induced neurodegeneration in various genetic backgrounds was quantitated as the number of swollen PVC neurons in L1 larvae and as the percentage of PVC neurons that are missing or are swollen in adult hermaphrodites. Statistical differences between genotypes was determined by the method of attributable risk (Devore, 1987). We compensated for multiple comparisons by settingp < 0.005 as the threshold for significance. Confidence intervals of 95% were calculated as 1.96* (SEM).
Antibody staining. Gαs expression in transgenic animals was monitored by staining fixed animals with anti-HA antibodies. Fixation and antibody staining were done according to a protocol devised by M. Nonet. Briefly, worms were fixed and their cuticles reduced in Bouin’s solution with β-mercaptoethanol (BME). Worms were washed sequentially in BTB [1× borate buffer, pH 9.2 (20 mm H3BO3 and 10 mmNaOH), 0.5% Triton X-100, and 2% BME], BT (1× borate buffer and 0.5% Triton X-100), and finally in AbA solution (1× PBS, 1% BSA, 0.5% Triton X-100, 0.05% sodium azide, and 1 mm EDTA). The anti-HA monoclonal antibody 12CA5 (Boehringer Mannheim, Indianapolis, IN) and goat anti-mouse rhodamine-conjugated antibody (Cappel, Cochranville, PA) were used at a 1:100 dilution, and incubations were done in AbA solution overnight at room temperature.
Isolation of acy-1 mutations. Mutations that restore normal locomotion rates toαs(gf) homozygotes were isolated from the F2 self-progeny of EMS mutagenized hermaphrodites. Candidate suppressor mutants subsequently were screened for the reduction of Gαs-induced swelling in L1 larvae. In a screen of 7500 haploid genomes, the nu327, nu329, andnu343 alleles were isolated. Other suppressor mutations isolated in this screen will be described elsewhere.
Positional cloning of acy-1. The acy-1 mutationsnu327, nu329, and nu343 are all linked to dpy-17 in two-factor mapping experiments. Three-factor mapping placed acy-1 between emb-5 anddpy-17: (nu327 dpy-17) 37/37unc-32; (nu329 dpy-17) 16/16unc-32; (nu343 dpy-17) 4/4unc-32; unc-79 (6/14) MJ#NEC2 (5/14)nu329 (3/14) dpy-17; emb-5 (1/16)nu327 (15/16) dpy-17. The cosmid F17C8 (which carries ACY-1) was microinjected into acy-1(nu327); nuIs5animals, and transgenic lines were isolated by usinggoa-1:: gfp (KP#13) as a transformation marker (Ségalat et al., 1995). Four independent lines were obtained, two of which were rescued for the acy-1 phenotype, i.e., they had increased degeneration of the PVC neurons. Sequences spanning the GENEFINDER-predicted exons of ACY-1 (accession number Z35719) were amplified from the mutants nu327, nu329, andnu343, and the resulting fragments were sequenced directly by cycle sequencing. The GENEFINDER prediction for the first exon was confirmed by isolating partial cDNA clones from the Barstead RB1 cDNA library by PCR amplification.
Analysis of acy-1 expression. A deleted derivative (KP#106) of the cosmid F17C8 was isolated by digestion withAflII and religation. KP#106 contains the entire 8.35 kb of the acy-1 genomic region, together with 5.2 kb 5′ and 4.9 kb 3′ flanking sequences. An acy-1:: gfp expression vector (KP#107) was constructed by PCR-amplifying a 1.7 kb fragment containing the GFP coding region and the unc-54transcription terminator from pPD95.75 and ligating this fragment into the unique Asp718 site in KP#106, creating a fusion protein containing the first six exons of acy-1 fused to GFP. The ACY-1:: GFP fusion protein contains six predicted transmembrane domains of ACY-1 and hence is membrane-localized. Transgenic animals carrying KP#107 were isolated by microinjection, using lin-15 as a transformation marker (Huang et al., 1994). Expressing cells were identified on the basis of their morphology and nuclear positions.
Isolation and responsiveness of eat-4 mutations.We screened 11,000 mutagenized haploid genomes for animals that failed to respond to nose touch. Mutants isolated were subjected to a series of secondary screens, including dye filling of the amphid sensory neurons and responsiveness to osmotic shock and volatile repellents. Six alleles of eat-4 (n2458,n2474, nu2, nu142, nu143, and nu146) were isolated in this screen. A seventh allele, eat-4(ky5), was isolated by C. Bargmann (University of California at San Francisco, San Francisco, CA) as a chemotaxis defective mutant. All seven eat-4 alleles are normal for dye filling but are defective for all three ASH sensory behaviors.
Behavioral assays. ASH sensory responses were assayed as previously described (Kaplan and Horvitz, 1993; Hart et al., 1995;Maricq et al., 1995; Troemel et al., 1995). For nose touch, animals were tested 10 times each, with a positive response being scored when animals either halted forward movement or initiated backward movement after the stimulus. Osmotic avoidance assays were performed with either of two protocols. Assays in Table 3 were done as described previously (Hart et al., 1995). Assays in Table 2 were done with a modified protocol described by C. de Vries and R. Plasterk (personal communication). Briefly, worms were washed twice with S-basal and once with water and placed inside a semicircle of 60% glycerol with bromphenol blue. A microliter of diacetyl at a 1:100 dilution was placed outside the semicircle at the opposite edge of the plate. The percentage of adult worms that crossed the osmotic barrier after 10 min was calculated. For ASH-mediated volatile avoidance, an eyelash was dipped in 1-octanol and held near an animal’s nose; responses were quantitated by recording the length of time that elapsed before an animal reversed locomotion.
Chemotaxis assays were performed as previously described (Bargmann et al., 1990). Assay plates contained 2% Difco-agar, 5 mmpotassium phosphate, pH 6.0, 1 mm calcium chloride, and 1 mm magnesium sulfate. Animals were placed at the center of the plate, with 1 ml of diluted attractant and 1 ml of 1 msodium azide at one edge of the plate; 1 ml of ethanol and 1 ml of sodium azide were placed at the opposite edge. Dilutions in ethanol were as follows: 1:1000 for diacetyl, 1:100 for isoamyl alcohol, and 1:200 for benzaldehyde. The chemotaxis index was calculated after 1 hr as the (number of worms at attractant − number of worms at control solvent)/total number of animals.
RESULTS
Gαs-induced neurodegeneration
While studying signaling by the G-protein Gs, we made the observation that expression of a constitutively active rat Gαs cDNA caused neurodegeneration in C. elegans. Mutations that diminish the GTPase activity of Gαs have been shown to cause constitutive, agonist-independent signaling (Landis et al., 1989; Lyons et al., 1990;Wong et al., 1991). We expressed a rat cDNA encoding a GTPase-defective (Q227L) Gαs subunit, hereafter referred to asαs(gf), in C. elegansneurons by using the glr-1 glutamate receptor (GluR) promoter. We chose the glr-1 promoter because glutamate-responsive cells might be prone to neurodegeneration, because it is highly expressed, and because glr-1-expressing cells control locomotion, an easily assayed behavior. The glr-1promoter is expressed in 17 classes of neurons, including interneurons required for locomotion (Hart et al., 1995; Maricq et al., 1995). Gαs was coexpressed with the GFP protein ofAequorea victoria (Chalfie et al., 1994), which allowed us to examine the morphology of Gαs-expressing cells. Transgenic glr-1:: αs(gf)animals were paralyzed and a subset of the expressing neurons swelled to several times their normal diameter and eventually disappeared, presumably because they died (Fig. 1, Table 1). These results suggest that exaggerated Gαs signaling kills neurons.
Gαs-expressing neurons differed greatly in their susceptibility to Gαs-induced toxicity. In first-stage (L1) glr-1:: αs(gf)larvae, the swelling of different cell types occurred at very different frequencies, and these differences were seen in two independentαs(gf) transgenes: innuIs5, PVC 88%, AVD 34%, and RIG 7%; in nuIs3, PVC 77%, AVD 44%, and RIG 6%. One potential mechanism for this apparent cell type specificity would be that cells differ substantially in their levels of Gαs expression. Because the rat Gαs cDNA contains the HA epitope, we were able to test this possibility by staining αs(gf)animals with anti-HA antibodies (Fig. 1E). Differences in Gαs expression correlated well with differences in toxicity for some cells, but not for others. For example, RIG neurons expressed much less Gαs and swelled much less frequently than PVC neurons, whereas AVD and PVC neurons expressed equivalent amounts of Gαs but swelled at significantly different frequencies. In general, 10 neurons in the lateral ganglion expressed levels of Gαs similar to those seen in the PVC, whereas mostαs(gf) animals have only one or two dying cells in the lateral ganglion (Fig. 1A). Thus, many more cells express Gαs than are found to die, and differences in Gαs-induced toxicity do not always correlate with differences in Gαs expression. To demonstrate further the specificity of Gαs-induced neurodegeneration, we expressedαs(gf) by using themec-7 promoter. MEC-7 tubulin is expressed abundantly in five neurons that sense light touch to the worm’s body, which are called touch cells (Savage et al., 1989; Hamelin et al., 1992; Mitani et al., 1993). Animals expressing themec-7:: αs(gf) transgene are indistinguishable from wild-type animals, having no obvious defect in touch sensitivity nor in the morphology of the touch cells (data not shown). Therefore, both the glr-1 and the mec-7expression constructs support the notion that the effects of Gαs on neural activity and on neurodegeneration are cell type-specific.
The pattern of cell deaths that was observed does not explain the severity of the locomotion defect inαs(gf) animals. The GLR-1-expressing interneurons AVA, AVB, AVD, and PVC play an important role in locomotion; hence these cell deaths could, in principle, explain the locomotion defects (Chalfie et al., 1985). However, the sluggish locomotion defect is apparent in 100% of theαs(gf) animals, although cell deaths are found in only a subset of animals (see above). For example, a significant fraction of uncoordinated animals can be found in which only the PVC neurons have died (A. Berger and J. Kaplan, unpublished observations). Because killing the PVC neurons with a laser microbeam is not sufficient to cause uncoordinated locomotion (Chalfie et al., 1985), these results suggest that theαs(gf) transgene inhibits the function of these interneurons in addition to causing a subset of these cells to die.
ACY-1 mediates Gαs-induced neurodegeneration
To identify the targets of Gαs, we isolated mutations that block Gαs-induced paralysis and neurodegeneration. In a screen of 7500 haploid genomes, we isolated three semidominant mutations that map to the cluster of chromosome III (Fig. 2, Table 1). Via a series of experiments we showed that these mutations occur in an adenylyl cyclase gene, which we have named acy-1. First, two of these mutations were mapped to a 1.5 cm genetic interval between MJ#NEC2 anddpy-17 (Fig. 2A). Second, a transgene carrying a cosmid from this interval (F17C8) corrected the mutant phenotype of acy-1(nu327) animals (Fig.2B). Third, all three alleles corresponded to mutations in the predicted exons of the gene F17C8.1 (Fig.2C,D), one of two predicted adenylyl cyclase genes in theC. elegans genome database. It is unclear why these mutations are partially dominant. The molecular nature of the mutations, together with the fact that the mutant phenotype is rescued by a wild-type copy of the F17C8 cosmid, suggests that these mutations reduce ACY-1 activity. For example, nu329 andnu343 are predicted to disrupt pre-mRNA splicing. Thus, it is possible that αs(gf)animals are highly sensitive to changes in cAMP levels; however, because none of the genetic deficiencies in this region uncovers theacy-1 gene, we cannot test directly whether acy-1is haplo-insufficient. These results suggest that Gαsneurodegeneration is mediated by changes in intracellular cAMP.
The expression pattern of acy-1 was determined by analyzing a GFP reporter construct. The acy-1:: gfp fusion protein is expressed in virtually all neurons and body muscles; however, it does not appear to be expressed in other tissues (Fig.3). Therefore, the cell type specificity of Gαs-induced neurodegeneration cannot be explained by the expression pattern of the acy-1:: gfp reporter. One caveat to this conclusion is that the expression of reporter genes can differ from that of the endogenous genes if, for example, some critical regulatory elements are missing in the reporter constructs.
Role of ACY-1 in sensory behaviors
cAMP has been implicated in a wide variety of signaling pathways, and ACY-1 is expressed in most if not all neurons and muscles in the worm. Therefore, we would expect that acy-1 mutants would have defects in behavior or development. In particular, the C. elegans Gαs subunit was shown previously to be essential for viability as well as for regulating locomotion and egg laying (Korswagen et al., 1997); hence we would expect to find similar defects in acy-1 mutants. Surprisingly, acy-1mutants are overtly normal, with no obvious defects in development, fertility, egg laying, or male mating behaviors (data not shown). Because the deaths of GLR-1-expressing cells are prevented byacy-1 mutations, we tested whether other behaviors mediated by GLR-1-expressing cells are impaired by acy-1 mutations. GLR-1-expressing cells are required for response to body touch and for locomotion (Chalfie et al., 1985); however, these behaviors are not affected in acy-1 mutants (see Table 4) (our unpublished observations). Three sensory behaviors mediated by the ASH neurons (nose touch, osmotic shock, and volatile repellent responses) are also likely to be mediated by the GLR-1-expressing cells (Bargmann et al., 1990; Hart et al., 1995; Maricq et al., 1995; Troemel et al., 1995), yet none of these behaviors is impaired in acy-1 mutants (Table 2).
Recently, two putative cyclic nucleotide-gated ion channels (TAX-2 and TAX-4) were shown to be involved in axon morphogenesis, thermotaxis, and olfaction (Coburn and Bargmann, 1996; Komatsu et al., 1996). Because these channels in principle could be activated by the cAMP produced by ACY-1, we examined acy-1 mutants for defects in these processes (Table 2). Mutations in tax-2 andtax-4 cause certain chemosensory neurons to grow out supernumerary axonal processes, which can be visualized by staining the animals with the fluorescent dye DiI (Coburn and Bargmann, 1996;Komatsu et al., 1996). We found no abnormal sensory axon morphologies in DiI-stained acy-1 mutants (data not shown). Chemotaxis toward isoamyl alcohol and benzaldehyde requires TAX-2 and TAX-4 channels (Coburn and Bargmann, 1996; Komatsu et al., 1996). We found that the response of acy-1 mutants to benzaldehyde was normal, whereas their response to isoamyl alcohol was impaired slightly. Finally, the thermotactic behavior of acy-1mutants was also normal (A. Berger and O. Hobert, unpublished observations). Because acy-1 mutants lack the axon and olfactory defects seen in tax-2 and tax-4mutants, it is unlikely that ACY-1 is the sole source of cyclic nucleotides required for the activation of TAX-2 and TAX-4. These results are consistent with the fact that TAX-4 channels are activated selectively by cGMP (Komatsu et al., 1996). Furthermore, these results do not exclude the possibility that cAMP is required for these behaviors, because there is at least one other adenylyl cyclase in the worm genome.
Cytotoxic targets of cAMP
Several previously identified genes were considered good candidates for mediating the toxic effects of Gαs (see Table 1). The putative cyclic nucleotide-gated ion channels TAX-2 and TAX-4 are not expressed in glr-1-expressing cells (Coburn and Bargmann, 1996; Komatsu et al., 1996) and hence are unlikely targets in this case. The mec-6, unc-8, anddeg-1 genes were implicated previously in neurodegeneration (Chalfie and Wolinsky, 1990; Driscoll and Chalfie, 1991; Shreffler et al., 1995; Tavernarakis et al., 1997), and the DEG-1 and UNC-8 proteins are similar to mammalian epithelial sodium channel subunits (ENaC), which are activated potently by cAMP-dependent protein kinase (PKA) (Sariban-Sohraby et al., 1988; Oh et al., 1993; Bubien et al., 1994). The unc-2, unc-36, and egl-19 genes encode subunits of voltage-dependent Ca2+ channels (Schafer and Kenyon, 1995; Lee et al., 1997) that are likely to be regulated by PKA (Curtis and Catterall, 1985) and also have been implicated in neurodegeneration. The glr-1 gene encodes an ionotropic GluR (Hart et al., 1995; Maricq et al., 1995), GluRs have been implicated in neurodegeneration in mammals (Olney, 1986; Choi, 1988), and PKA augments the response of mammalian neurons to glutamatergic agonists (Greengard et al., 1991; Wang et al., 1991;Colwell and Levine, 1995).
Of these candidate genes the unc-36 mutations (e251 and e837) conferred a slight but significant reduction in Gαs-induced cytotoxicity (Table1). Interestingly, the unc-36 mutations had no effect on cell swelling. These results suggest that either Ca2+ influx or depolarization of the affected cells modulates susceptibility to Gαs cytotoxicity. All other candidate genes had no effect on either neuron swelling or deaths inglr-1:: αs(gf) animals (Table 1). Our results do not exclude the possibility that these other candidate PKA targets play a role in Gαs-induced toxicity. For example, more than one type of channel may be capable of mediating the toxic effects of Gαs, in which case neurodegeneration would be prevented only in multiply mutant animals.
Role of apoptosis and necrosis in Gαs neurodegeneration
Whether neurodegeneration occurs by apoptosis or by necrosis has remained controversial (Choi, 1996). The Gαs-induced deaths appear to be necrotic (i.e., undergoing cell swelling and delayed cytotoxicity) rather than apoptotic. We directly tested the role of apoptosis in these deaths by testing the effects of mutations in cell death genes (Hengartner, 1997). We found that mutations in the cell death genes ced-3 and ced-4, both of which are required for apoptosis (Ellis and Horvitz, 1986), had no effect on Gαs-induced swelling or killing (Table 1). Thus, apoptosis is not required for Gαs-induced killing.
Role of endogenous neural signaling in Gαs-induced neurodegeneration
Because Gαs often couples to neurotransmitter receptors thereby producing or altering synaptic transmission, we wondered whether endogenous neural activity would regulate Gαs-induced neurodegeneration. The weakly neuroprotective effect of unc-36 mutations is consistent with this hypothesis. Decreased calcium influx or decreased cell excitability inunc-36 mutants could explain this neuroprotective effect. In addition, unc-36 mutations have been shown to decrease endogenous synaptic transmission, albeit by an uncharacterized mechanism (Nguyen et al., 1995). Because UNC-36 channels play a role in many different aspects of neural activity, we reasoned that mutations that more specifically perturb the cytotoxic mechanism might produce a more profound neuroprotective effect.
To test directly the role of synaptic activity on these cell deaths, we tested the effect of mutations in the unc-104 andsnt-1 genes, which encode phylogenetically conserved proteins (kinesin heavy chain and synaptotagmin) that are required for synaptic vesicle transport and exocytosis (Otsuka et al., 1991; Nonet et al., 1993). We found that neither the unc-104 norsnt-1 mutations reduced Gαs-induced cell killing. This result suggests that the overall levels of synaptic input are not required for killing, per se.
Given its role in excitotoxicity in mammals, we wondered whether endogenous glutamate signaling is required for Gαsneurodegeneration. We found that a loss-of-function glr-1mutation did not reduce Gαs-induced cell swelling or cytotoxicity (Table 1). This result does not exclude the possibility that glutamate neurotransmission mediates cAMP-induced cytotoxicity. The C. elegans genome sequence (currently ∼85% complete) predicts eight additional ionotropic GluR subunits; therefore, theglr-1 mutation is unlikely to eliminate glutamate signalingin vivo.
eat-4 mutations are neuroprotective
Because GLR-1 receptors are required for several mechanosensory behaviors (Hart et al., 1995; Maricq et al., 1995), we reasoned that other mutations affecting these behaviors might be neuroprotective. Previous work has shown that ASH sensory neurons mediate an aversive response to three distinct stimuli (nose touch, osmotic shock, and volatile repellents) and that the ASH-mediated touch response requires functional GLR-1 glutamate receptors in synaptic targets of ASH (Hart et al., 1995; Maricq et al., 1995; Troemel et al., 1995). We isolated seven alleles of the eat-4 gene, originally identified because of its function in pharyngeal pumping (Avery, 1993), in a screen for mutations that eliminate ASH-mediated touch sensitivity. All seven eat-4 strains have similar behavioral defects. In particular, they have severe defects in the ASH-mediated touch, osmosensory, and volatile repellent responses (Table3). We found that eat-4mutations significantly reduced Gαs-induced cytotoxicity but had no apparent effect on cell swelling (see Table 1). PVC cytotoxicity in eat-4 unc-36; αs(gf) triple mutants (55 ± 5%) was not significantly different from that seen in unc-36; αs(gf) double mutants, suggesting thateat-4 and unc-36 act in a single pathway regulating Gαs-induced killing. In addition,eat-4 mutations dramatically improved the locomotion rate ofαs(gf) animals (Table4). Thus, the eat-4 gene plays a significant role in Gαs-induced effects on neurodegeneration (reducing cytotoxicity) and on neural activity (reducing paralysis).
DISCUSSION
We and others (Korswagen et al., 1997) have shown that the expression of a constitutively active form of Gαs induces a form of neurodegeneration in the nematode C. elegans. The Gαs-induced deaths appear to be necrotic, because the affected cells swell and subsequently lyse and because these deaths are not prevented by mutations in the cell death genes ced-3 andced-4. We provide here a detailed analysis of the Gαs killing pathway. Neurons differ greatly in their susceptibility to Gαs-induced neurodegeneration, ranging from 0 to 88% killed. Three genes (acy-1, eat-4, and unc-36) that contribute to Gαs-induced neurodegeneration are identified. Gαs killing appears to be mediated by the second messenger cAMP, because acy-1 mutations block killing. Our data also suggest that the two phases of Gαsneurodegeneration can be distinguished genetically, becauseacy-1 mutations block both swelling and cytotoxicity, whereas other mutations (i.e., unc-36 andeat-4) reduce cytotoxicity but have no effect on swelling. However, because we have not quantitated the extent or duration of cell swelling, it remains possible that there is a more subtle correlation between the extent of swelling and cytotoxicity. Finally, killing is not dependent on synaptic transmission, but it is modulated by endogenous neural signaling.
Function of ACY-1
We have isolated mutants that lack ACY-1, an adenylyl cyclase. ACY-1 appears to be expressed in nearly all neurons and muscles, but not in other tissues. These results suggest that ACY-1 adenylyl cyclase is likely to participate in many neural signaling pathways. Therefore, we would expect that acy-1 mutants would have defects in behavior or development. Consistent with this notion is that mutations that inactivate the C. elegans Gαs subunit (GSA-1) are homozygous lethal (Korswagen et al., 1997). Surprisingly, the behavior and morphology of acy-1 homozygotes are nearly indistinguishable from wild-type animals. Several behaviors were examined in greater detail, including behaviors mediated by GLR-1-expressing cells, behaviors regulated by GSA-1, and those that require the cyclic nucleotide-gated channels TAX-2 and TAX-4. We found that acy-1 mutants were proficient in all of these behaviors. These results suggest that acy-1 alleles do not eliminate ACY-1 activity, that the function of GSA-1 required for viability and for regulating behaviors is mediated by another adenylyl cyclase, or that ACY-1 acts redundantly with other forms of adenylyl cyclase. The genome sequence predicts at least one other adenylyl cyclase gene, which could account for the discrepancy between thegsa-1 and acy-1 mutant phenotypes. Our results also suggest that ACY-1 is not the sole source of cyclic nucleotides required for the activation of TAX-2 and TAX-4, which is consistent with previous studies suggesting that TAX-4 channels are activated selectively by cGMP.
Role of cAMP in neurodegeneration
Several possible mechanisms could explain the toxic effects of activated Gαs. We believe that the most likely explanation is that cAMP regulation of ion channels or ion transporters grossly alters membrane permeability, leading to cell swelling and death. In other systems many ion channels have been shown to be potently regulated by cAMP. We tested the C. eleganshomologs of several of these potential cAMP targets, finding that mutations reducing the activity of UNC-36 calcium channels had a modest but significant neuroprotective effect. Because voltage-dependent calcium channels correspond to heteromultimers of several types of subunits, it is difficult to predict the extent to which alteration of the UNC-36 α2 subunit reduces the overall calcium permeability of cells in vivo. Although other candidate mutations were not neuroprotective, these results do not exclude a role for these genes, e.g., if several genes play redundant roles in inducing neurodegeneration.
An alternative explanation for Gαs-induced neurodegeneration is that cells expressing the Gαstransgene have taken on the fates of cells that undergo developmentally programmed cell deaths. Several facts argue against this model. First, all of the cell deaths that occur during normal C. elegansdevelopment are apoptotic, by both morphological and genetic criteria (Hengartner, 1997). Therefore, the necrotic cell deaths produced by Gαs are not seen in normal development. Second, theglr-1 promoter used to express Gαs is expressed starting in the threefold embryo (Hart et al., 1995; Maricq et al., 1995) after most cell fate choices (and in fact most cell deaths) already have occurred (Sulston et al., 1983). Third, Gαs-expressing cells continue to express theglr-1 promoter, indicating that at least some aspects of cell fate have not been altered.
It is also possible that overproduction of cAMP diminishes ATP levels, leading to a metabolic crisis and cell death. We think that this model is unlikely for three reasons. First, Gαs-induced cell deaths are cell type-specific, which would not be predicted by this model. Second, suppression by acy-1 mutations is semidominant, which implies that cells are sensitive to subtle changes in cAMP levels. Third, mutations that alter endogenous neural signaling (i.e., unc-36 and eat-4) modulate Gαs toxicity, implying that the toxic signal is mediated by normal signaling pathways. These results suggest that, rather than creating a catastrophic metabolic event, Gαs kills via endogenous signaling pathways. On the other hand, the terminal phases of cytotoxicity must include the gross alteration of cellular metabolism and cellular integrity. Thus, in addition to inducing a cytotoxic signal, overproduction of cAMP also might hasten death by acting as a metabolic sink.
cAMP has been implicated in growth control and cell death in other systems. For example, elevated cAMP levels are cytotoxic in S49 lymphoma cells (Coffino et al., 1975), and these deaths subsequently have been shown to occur by apoptosis (Lanotte et al., 1991; Duprez et al., 1993). The molecular target of cAMP in the induction of lymphoid apoptosis has not been determined. In other cell culture systems cAMP induces growth arrest in G1 of the cell cycle (Khan et al., 1996). To our knowledge this is the first report of cAMP-induced neurodegeneration. In fact, in several other models of neurodegeneration cAMP has been shown to be neuroprotective (D’Mello et al., 1993; Dockwerth and Johnson, 1993; Kawakami et al., 1996;Michel and Agid, 1996).
Role of eat-4 and unc-36
We found that eat-4 and unc-36 mutations are partially protective against the cytotoxic and paralytic effects of the Gαs transgene. Although these protective effects are admittedly subtle, we believe that they are real for several reasons. First, all of our data were compared by statistical methods, using a relatively strict threshold for significance (p< 0.005). Second, we have examined many other genetic backgrounds and found no similar protective effects, which indicates that effects of this magnitude are uncommon. Third, for both eat-4 andunc-36 we found the neuroprotective effect in two unrelated strains carrying different alleles of these genes. Therefore, it is highly likely that these protective effects are caused by the mutations in unc-36 and eat-4 rather than by some uncharacterized mutation in the genetic background. By all of these measures the effects of the eat-4 and unc-36mutations are real; therefore, we conclude that these genes play some role in determining the severity of Gαs-induced killing.
The eat-4 gene was identified initially in screens for mutations that disrupt eating behavior (Avery, 1993). Theeat-4 eating defect is caused by the elimination of a glutamate-induced inhibitory synaptic signal (mediated by the M3 motor neuron), which can be observed in extracellular recordings of pharyngeal muscle activity (Raizen and Avery, 1994). The defect in pharyngeal neurotransmission is likely to be caused by a presynaptic defect in M3, because pharyngeal muscles isolated from eat-4mutants are responsive to glutamate iontophoresis (Dent et al., 1997). We have shown that eat-4 mutants are defective for three ASH-mediated sensory behaviors, one of which (nose touch) is mediated by GLR-1 GluRs. Thus, both the eating defects and the sensory defects observed in eat-4 mutants could be explained by an underlying defect in glutamate neurotransmission. The eat-4gene has been cloned (R. Lee, E. Sawin, M. Chalfie, H. R. Horvitz, and L. Avery, personal communication); however, the molecular identity of EAT-4 does not reveal what role it plays in neuronal signaling.
Two sorts of models could explain the neuroprotective effects ofeat-4 and unc-36 mutations. First, these mutations could be neuroprotective, because Gαs-induced deaths are mediated in part by endogenous glutamate neurotransmission. Alternatively, EAT-4 and UNC-36 could act in the dying cells, directly or indirectly mediating the cytotoxic effects of cAMP. Gαs-induced killing is not diminished by mutations that impair synaptic transmission, which would favor the model that EAT-4 and UNC-36 act in the dying cells. However, EAT-4 is not expressed in PVC neurons (R. Lee, E. Sawin, M. Chalfie, H. R. Horvitz, and L. Avery, personal communication), which favors the idea that EAT-4 acts in the presynaptic partner. Neither of these results conclusively tests these models. The reported EAT-4 expression pattern may be incomplete. Moreover, several results indicate that residual synaptic transmission occurs in unc-104 and snt-1 mutants. Theunc-104 allele used in this study is a partial loss of function, because null alleles are homozygous lethal (Hall and Hedgecock, 1991), and residual synaptic transmission has been documented in synaptotagmin null mutants in worms, flies, and mice (DiAntonio et al., 1993; Littleton et al., 1993; Nonet et al., 1993;Geppert et al., 1994). Moreover, some glutamate may be released by a nonvesicular mechanism (Attwell et al., 1993). In fact, cell swelling might stimulate glutamate efflux through volume-sensitive osmolyte channels (Jackson and Strange, 1993). Finally, it is also possible that defects in exocytosis make cells more susceptible to necrosis, for example by preventing the addition of new membranes to swelling cells. Further experiments will be required to distinguish among these models.
Similarities to other forms of neurodegeneration
Several other C. elegans mutations have been described that cause a necrotic form of neurodegeneration (Chalfie and Wolinsky, 1990; Driscoll and Chalfie, 1991; Treinin and Chalfie, 1995). In particular, mutations in the deg-1 gene also cause a specific subset of neurons to undergo necrosis, including the PVC neuron (Chalfie and Wolinsky, 1990; Garcia-Anoveros et al., 1995). Our results suggest that the deg-1-induced and Gαs-induced cell deaths occur by distinct mechanisms because mec-6 mutations block deg-1-induced deaths, but not Gαs-induced cell deaths (Chalfie and Wolinsky, 1990); however, it remains possible that ACY-1 is required for deg-1-induced deaths.
Gαs-induced cell deaths also share some properties with glutamate-induced excitotoxicity in mammals. Neurodegeneration occurs in two phases (swelling and killing), and killing is apparently cell type-specific. Reducing the activity of voltage-dependent calcium channels is neuroprotective in both cases; however, the protective effect of unc-36 mutations is modest (albeit statistically significant). Finally, the neuroprotective effects of eat-4mutations imply that glutamate may play a role in Gαs-induced neurodegeneration. Further evidence will be required to determine whether Gαs-induced cell deaths and excitotoxicity are related mechanistically.
Footnotes
This work was supported by Grants from National Institutes of Health (NS32196) and the Human Frontiers Science Program to J.K. A.B. was a Howard Hughes Medical Institute predoctoral fellow. A.H. was supported by a fellowship from the Medical Foundation. J.K. is a Pew Scholar in the Biomedical Sciences. We thank L. Chen for excellent technical assistance; A. Fire, F. Kolakowski, and H. Bourne for plasmids; A. Coulson for cosmids; R. Barstead for cDNA libraries; R. Horvitz, M. Chalfie, J. Miwa, E. Jorgensen, and the C. elegans Genetics Stock Center for strains; M. Nonet for providing the antibody staining protocol; and C. Bargmann for theeat-4(ky5) allele.
Correspondence should be addressed to Dr. Kaplan at his present address: Department of Molecular and Cell Biology, 361 Life Sciences Addition, University of California, Berkeley, CA 94720-3200.
Dr. Hart’s present address: Department of Pathology, Harvard Medical School and Massachusetts General Hospital Cancer Center, 149-7202 13th Street, Charlestown, MA 02129.