Genetic studies of sup-9, unc-93, and sup-10 strongly suggest that these genes encode components of a multi-subunit protein complex that coordinates muscle contraction in Caenorhabditis elegans. We cloned sup-9 and sup-10 and found that they encode a two-pore K+ channel and a novel transmembrane protein, respectively. We also found that UNC-93 and SUP-10 colocalize with SUP-9 within muscle cells, and that UNC-93 is a member of a novel multigene family that is conserved among C. elegans, Drosophila, and humans. Our results indicate that SUP-9 and perhaps other two-pore K+ channels function as multiprotein complexes, and that UNC-93 and SUP-10 likely define new classes of ion channel regulatory proteins.
The sup-9, sup-10, and unc-93 genes of Caenorhabditis elegans can affect the regulation of muscle contraction. Mutants carrying semidominant gain-of-function (gf) mutations in sup-9, unc-93, or sup-10 move sluggishly, are defective in egg-laying and defecation, and display a rubberband uncoordinated (unc) response: when prodded on the head, a mutant worm contracts and then relaxes along its entire body without moving backwards, whereas a wild-type worm contracts its anterior end and backs away (Greenwald and Horvitz, 1980, 1986; Levin and Horvitz, 1993). Although capable of muscle contraction, these mutants appear to be unable to mount a coordinated muscle response. In contrast, mutants defective in muscle structural genes are defective in muscle contraction (Waterston et al., 1980). Worms with loss-of-function (lf) mutations in sup-9, sup-10, or unc-93 display no gross phenotypic abnormalities, suggesting that these genes have nonessential functions and may act in parallel to another gene or set of genes (Greenwald and Horvitz, 1980). Mosaic analysis of sup-10 indicates that it functions in muscle (Herman, 1984).
Three lines of genetic evidence strongly argue that the SUP-9, SUP-10, and UNC-93 proteins physically interact as members of a protein complex. First, lf mutations in sup-9, sup-10, or unc-93 suppress the rubberband phenotype caused by gf mutations in any of these three genes (Greenwald and Horvitz, 1980, 1986; Levin and Horvitz, 1993; De Stasio et al., 1997). Because the effects of each of these three genes are dependent on the functions of the other two genes, the gene products of all three genes must act at the same step. Second, this mutual suppression can be observed weakly in lf hetezozygotes, suggesting that a stochiometric relationship among SUP-9, SUP-10, and UNC-93 is important for their function (Levin and Horvitz, 1993). Third, certain sup-9 alleles display gene- and allele-specific interactions with sup-10(gf) and unc-93(gf) alleles (Levin and Horvitz, 1993); such genetic interactions are a hallmark of genes that encode proteins that physically interact (Hartman and Roth, 1973).
A fourth gene, sup-18, was identified as an lf suppressor of the sup-10(gf) rubberband Unc phenotype and shown to be a partial suppressor of sup-9(gf) and unc-93(gf) rubberband Unc mutants (Greenwald and Horvitz, 1986). sup-18 may encode a component, regulator, or effector of a SUP-9-SUP-10-UNC-93 complex. unc-93 encodes a novel, putative, multipass transmembrane protein of unknown biochemical function (Levin and Horvitz, 1992, 1993).
Here, we report that sup-9 encodes a two-pore K+ channel with similarity to hTASK-1 and hTASK-3 (TASK, TWIK-related acid-sensitive K+ channel) and that sup-10 encodes a novel single-pass transmembrane protein. Mammalian TASK-1 and TASK-3 form a subfamily of two-pore K+ channels that are activated by high pH, volatile anesthetics, and neurotransmitters (Duprat et al., 1997; Leonoudakis et al., 1998; Patel et al., 1999; Kim et al., 2000; Millar et al., 2000; Rajan et al., 2000; Talley et al., 2000). Our findings indicate that UNC-93 and SUP-10 associate with a SUP-9 two-pore K+ channel and suggest that UNC-93 and SUP-10 may be regulatory subunits of this channel.
Materials and Methods
Strains and genetics. C. elegans strains were cultured as described previously (Brenner, 1974), except that the Escherichia coli strain HB101 was used, instead of OP50, as a food source. N2 was the wild-type strain (Brenner, 1974). Strains were grown at 20°C unless otherwise noted. The following mutations were used in this study: linkage group (LG) II: lin-42(n1089), sup-9(e2655gf, e2661gf, lr1, lr100, lr11, lr129, lr142, lr30, lr35, lr38, lr45, lr57, lr73, n180, n186, n188, n189, n190, n191, n213, n219, n222, n223, n229, n233, n241, n264, n266, n271, n292, n345, n350, n508, n659, n688, n1009, n1012, n1016, n1020, n1023, n1025, n1026, n1028, n1037, n1428, n1435, n1469, n1472, n1549, n1550gf, n1553, n1557, n1913, n1914, n2174, n2175, n2176, n2276, n2278, n2279, n2281, n2282, n2283, n2284, n2285, n2286, n2287, n2288, n2291, n2292, n2294, n2296, n2343, n2344, n2345, n2346, n2347, n2348, n2349, n2350, n2351, n2352, n2353, n2354, n2355, n2356, n2357, n2358, n2359, n2360, n2361, n3310gf, nP136, lin-31(n301); LGIII: unc-93(e1500gf, lr12, n234), nIs124, sup-18(n1030); LGX: dpy-6(e14), sup-10(e2127, n983gf, n183, n240, n247, n250, n251, n342, n619, n1007, n1008, n1017, n1468, n1626, n1906, n2297, n3558, n3564), lin-15(n765ts). The sources of the sup-9 and sup-10 alleles are indicated Tables 1, 2, 3, 4.
Cloning of sup-9. We found a 5.5 kb Tc1-containing EcoRI polymorphism, nP136, that cosegregated with the sup-9 suppressor phenotype after nine backcrosses in sup-9(n1428);unc-93(e1500) mutants. We cloned 943 bp of DNA flanking the nP136 polymorphism using a vectorette-cloning approach (Korswagen et al., 1996) with the following modifications. We used an agarose gel to purify EcoRI-digested genomic DNA from the sup-9(n1428) nP136;unc-93(e1500) strain and ligated 40 ng to 10 pmol EcoRI vectorette in a 100 μl volume; 3 μl was used as a template for a 100 μl PCR reaction of 30 cycles using Tc1-specific primers L2 or R2 (Zwaal et al., 1993) and the universal vectorette primer 224 (Riley et al., 1990). Nested PCRs were performed with primer 224 and the Tc1-inverted-repeat primer N412 (Korswagen et al., 1996). The PCR product generated from the L2-nested reaction was cloned, its DNA sequence was determined using an automated ABI373A DNA sequencer (Applied Biosystems, Foster City, CA), and was mapped to the overlapping cosmids, F34D6 and F19E8, using database searches of genomic sequence (C. elegans Sequencing Consortium, 1998). The following primers were used in PCR reactions to amplify the sup-9 cDNA from mixed-stage cDNA: 5′-GTGTGAGCTCAGCAGCTTCT-3′ and 5′-TACTTCAAGAGATTGCAGC-3′. Rapid amplification of cDNA ends (RACE) reactions were performed using 5′ and 3′ RACE kits (Invitrogen, San Diego, CA).
Expression constructs. The sup-9::gfp fusion was obtained by subcloning the HpaI-rescuing sup-9 genomic fragment, which includes 2.9 kb of sup-9 promoter sequence, into the EcoRV site of Bluescript (Stratagene, La Jolla, CA). A BamHI site was introduced by PCR amplification immediately before the stop codon, and the genomic fragment subcloned into the SalI-BamHI sites of the green fluorescent protein (GFP) expression vector pPD95.77 (provided by Dr. A. Fire, Carnegie Institute of Washington, Baltimore, MD), thereby fusing gfp at the C-terminal end of sup-9 to create pIP201. The unc-54 3′ untranslated region (UTR) from pPD95.77 was replaced with an 808 bp region of genomic sequence immediately following the stop codon of sup-9 by standard PCR and cloning techniques. Animals carrying the sup-9::gfp transgenic array were treated with γ radiation to isolate a stable integrant (Fire, 1986).
The unc-93::gfp fusion was created by cloning a 12.9 kb SphI-KpnI genomic fragment from cosmid C46F11 containing 5.3 kb of unc-93 upstream sequence, the entire coding sequence of unc-93 and 3.2 kb of unc-93 downstream region into the SphI-KpnI sites of pPD95.79 (provided by Dr. A. Fire) creating pIP314. A KpnI site was introduced immediately preceding the unc-93 stop codon using PCR, and the KpnI fragment containing the unc-93 3′ region was excised, creating plasmid pIP321 with a GFP fusion at the C terminus of UNC-93.
The sup-10::gfp fusion used in colocalization studies was constructed by subcloning a 7.3 kb MfeI genomic fragment from cosmid C27G6 containing sup-10 into the EcoRI site of pBSKII. A 6.4 kb PstI fragment was subcloned from this vector into p95.77, which contained 3.5 kb of promoter sequence and the sup-10 coding region. Through PCR techniques, we introduced a SalI site immediately preceding the stop codon of sup-10 to create an in-frame fusion with the gfp coding sequence.
To make the ectopic expression constructs, the sup-9 cDNA was amplified by PCR from the start to the stop codon, with the introduction of NheI-SacI sites at the 5′ and 3′ ends, respectively, and subcloned into pPD95.86 (provided by Dr. A. Fire) or an unc-76 promoter vector (Bloom and Horvitz, 1997) to generate myo-3::sup-9 and unc-76::sup-9, respectively. The n1550 mutation was introduced into myo-3::sup-9 using mutagenic PCR primers.
Transgenic animals. Germline transformation experiments were performed using standard methods (Mello et al., 1991). Transformations were done using strains carrying lin-15(n765ts) with the coinjection marker pL15EK(lin-15) (Clark et al., 1994) at 100 ng/μl or with the muscle gfp marker pPD93.97 (provided by Dr. A. Fire) and the experimental DNA at 30-50 ng/μl. Transformants were recognized by their non-Muv phenotypes at 22.5°C or by their GFP fluorescence.
Antibodies and immunostaining. Codons 1-118 of sup-9 were fused to the glutathione S-transferase (GST) gene in the vector pGEX-2T (Amersham Biosciences, Piscataway, NJ), and codons 1-110 were fused to the maltose-binding protein (MBP) gene in the vector pMal-c2 (NEB). The GST fusion protein was expressed in E. coli, and the insoluble protein was purified by SDS-PAGE and used to immunize rabbits. Antisera were purified against the MBP fusion protein immobilized on nitrocellulose strips and eluted with 100 mm glycine-HCl, pH 2.5.
For immunofluorescence experiments, mixed-stage worms were fixed in 1% paraformaldehyde for 2 hr at 4°C and permeabilized as described previously (Finney and Ruvkun, 1990). For colocalization studies, transgenic lines were stained with anti-SUP-9 serum at 1:150 dilution and mouse anti-GFP serum at 1:100 dilution (Quantum Biotechnologies, Montreal, Canada). A secondary goat-anti-rabbit antibody conjugated to Texas Red (Jackson ImmunoResearch, West Grove, PA) and a secondary goat-anti-rabbit antibody conjugated to FITC (Jackson ImmunoResearch) were used at a 1:150 dilution. Worms were viewed using confocal microscopy (Zeiss LSM510; Zeiss, Thornwood, NY) or epifluorescence (Zeiss Axioskop2).
Neuronal cell identifications. To aid in the identification of neurons expressing SUP-9::GFP, we constructed a sup-9::gfp transcriptional fusion containing the same promoter and 3′UTR region as the sup-9::gfp translational reporter described above but lacking any sup-9 coding region. This reporter was brightly expressed in the cell bodies and processes of 4 of the ∼15 neurons that expressed the SUP-9 translational GFP fusion. Each neuron extended a single axonal process into the nerve ring and turned posteriorly, two along the dorsal and two along the ventral cords, before ending in the posterior of the animal between the vulva and the anus. This axonal morphology was consistent with that of the four SIA interneurons (White et al., 1986). We confirmed this identification by immunostaining SUP-9::GFP-expressing animals and UNC-93::GFP-expressing animals with antisera raised against the homeodomain transcription factor CEH-17, which is expressed in the four SIA neurons and the ALA neuron (Pujol et al., 2000).
Muscimol and rubberband Unc assays. Muscimol assays were done using 24-well plates (Costar, Cambridge, MA). A 200 mm stock of muscimol (Sigma, St. Louis, MO) was added to 1 ml of melted NGM agar (Brenner, 1974). After 4-6 hr, HB101 bacteria were streaked in each well. Two hours later, 10-12 worms were placed into each well and allowed to equilibrate for 1 hr. The rubberband Unc response was scored by touching each worm with an eyelash across its body, just posterior to the pharynx. Each worm was scored five or six times. The responses to touch were scored according to the presence of a contraction-relaxation cycle and backward movement in the following manner: 0, worms did not contract and relax but moved away from the touch; 1, worms quickly contracted and relaxed and moved away from the touch; 2, worms contracted and relaxed while concurrently generating a small backward displacement (less than one-half of body length); 3, worms contracted and relaxed but failed to move backwards; 4, worms incompletely contracted and relaxed and produced no displacement.
sup-9, sup-10, and unc-93 likely encode components of a protein complex
To facilitate our analyses of the interactions among sup-9, sup-10, and unc-93, we quantified the locomotory behavior of strains carrying mutations in one or more of these three genes (Table 1). Our data confirmed previous conclusions reached through qualitative examinations of mutant strains, including the following: (1) mutations that cause an abnormal (“altered”) function of sup-9, sup-10, or unc-93 result in a rubberband Unc paralysis and (2) the paralysis caused by an altered function mutation in any of these three genes can be suppressed by either an lf mutation in that gene or by an lf mutation in either of the other two genes. Such reciprocal suppression indicates that the activity of each of these three genes requires the functions of the other two genes and establishes that the products of the three genes act together at a single step, most likely as components of a protein complex.
sup-9 encodes a two-pore K+ channel
We cloned sup-9 using transposon tagging. sup-9(n1428) was previously isolated as a suppressor of the dominant rubberband Unc phenotype caused by the unc-93(e1500) gf allele in the mutagenic mut-2 background, in which the spontaneous excision and transposition of the mariner-type transposable element Tc1 is greatly increased (Levin and Horvitz, 1992). We isolated a Tc1 insertion within the first exon of predicted gene F34D6.3 associated with the sup-9(n1428) mutation (see Materials and Methods). To determine whether F34D6.3 corresponded to sup-9, we performed transformation rescue experiments. Because sup-9(lf) mutants have a wild-type phenotype, they cannot be used to assay sup-9 wild-type activity. Instead, we used sup-9(lf);sup-10(gf) mutants, in which the sup-10(gf) Unc phenotype is suppressed by the sup-9(lf) mutation; this suppression can be eliminated by theaddition of sup-9 wild-type activity. Cosmid F19E8, as well as a 6.8 kb subclone of F19E8 containing the predicted gene F34D6.3, restored sup-9 activity in transgenic sup-9(n180);sup-10(n983) animals (Fig. 1A).
The sup-9 minimal rescuing fragment contains a single complete predicted gene. Using reverse transcriptase (RT)-PCR and 5′ and 3′ RACE, we cloned a full-length sup-9 cDNA (Fig. 1B). sup-9 encodes a predicted protein of 329 amino acids with sequence similarity to the TASK subfamily of two-pore K+ channels (Fig. 1C). SUP-9 is 49-51% identical in amino acid sequence to human TASK-1(KCNK3), TASK-3(KCNK9), and TASK-5(KCNK15) channels as well as to two predicted Drosophila proteins. Mammalian TASK-1 and TASK-3 behave as pH-sensitive background K+ channels when expressed heterologously in mammalian cell lines (Duprat et al., 1997; Kim et al., 1998; Leonoudakis et al., 1998; Rajan et al., 2000), whereas the properties of TASK-5 channels remain unknown. All other identified human two-pore K+ channels, such as TREK-1(KCNK2), share <30% amino acid identity with SUP-9. Most of the amino acid identities among SUP-9 and the human TASK channels are restricted to the P domains, highly conserved loops that determine the ion selectivity of K+ channels (Heginbotham et al., 1994), and to the first, second, and fourth transmembrane domains.
Characterization of sup-9 mutant alleles
Suppressors of the rubberband Unc phenotypes of sup-9, unc-93, and sup-10 gf mutants have defined many sup-9 lf alleles. We determined the sequences of the open-reading frame and intron-exon boundaries of sup-9 from 81 strains carrying lf mutations (Table 2). These mutations include 47 missense, 15 nonsense, and 11 splice-site mutations as well as 3 deletions, 1 three-base-pair insertion, and 1 Tc1 insertion. We did not find molecular lesions in three sup-9 mutants; these alleles may contain mutations in the promoter region or in other regulatory elements not contained within the sup-9 exons. The identified sup-9 missense mutations include lesions in all four putative transmembrane domains, the M1-P1 linker, which joins the first transmembrane domain and the first P domain, and both putative P domains. sup-9(n2282) encodes an isoleucine instead of the initiator methionine and thus is likely to be a null allele. The mutations V41A, V44E, D58V, I61S, and A74V cause changes in the M1-P1 linker, which has been shown to act as a dimerization domain in vitro in the human TWIK-1 two-pore channel (Lesage et al., 1996).
The sup-9(n1550) gf allele that leads to the rubberband Unc phenotype causes an A236T substitution in the fourth transmembrane domain. We determined the sequences of three additional independently isolated sup-9 gf alleles. All three, e2655, e2661, and n3310, contain the same nucleotide change and therefore cause the same A236T substitution (Table 3).
The partial lf alleles n2359, n2360, n2361, and n2288, which were isolated as suppressors of sup-9(n1550) defects (Levin and Horvitz, 1993), contain second-site mutations in addition to the A236T sup-9 (n1550) substitution (Table 3). sup-9(n1550 n2360) mutants do not display the rubberband Unc phenotype of sup-9(n1550) animals but rather appear wild-type. However, unlike both sup-9(+) and sup-9(0), sup-9(n1550 n2360) partially suppresses the sup-10(gf) mutant phenotype, indicating that this allele is neither completely wild-type nor null. The n2360 and n2361 alleles contain a point mutation in the same codon mutated in n1550 animals, leading to a substitution of methionine in place of the wild-type alanine and the n1550 threonine. Two other partial lf alleles, sup-9(n1550 n2288) and sup-9(n1550 n2359), contain missense mutations affecting the third transmembrane domain, G173E and A174T, respectively. These amino acid changes may suppress the sup-9(n1550) phenotype by altering an interaction between the third transmembrane domain and the mutant threonine in the fourth transmembrane domain.
sup-10 encodes a novel putative transmembrane protein
sup-10 was previously mapped to the right end of LGX near mec-4 (Greenwald and Horvitz, 1980; A. Villenevue, personal communication to all). We analyzed restriction fragment-length polymorphisms in sup-10 lf mutants generated by γ-ray and transposon mutagenesis using cosmid ZK54, which we found in unrelated studies to contain the mlc-1 gene (mlc-1 encodes a myosin light chain, and we regarded this gene as a candidate for corresponding to sup-10) (Fig. 2A). We found that sup-10(n247), sup-10(n251), and sup-10(n1906) carry allele-specific polymorphisms to the right of mlc-1. In transformation rescue experiments, both cosmid C27G6, which is located 23 kb to the right of ZK54, and a 7.3 kb subclone of C27G6 containing a single predicted gene suppressed the semidominant rubberband Unc paralysis of sup-10(n983) animals as transgenes (Fig. 2A). We isolated a full-length cDNA (GenBank accession number U43891) corresponding to this predicted gene by screening a mixed-stage cDNA library (Fig. 2B). sup-10 encodes a predicted protein of 332 amino acids containing a putative signal sequence at its N terminus and a hydrophobic region near its C terminus, suggesting it is a type-1 transmembrane protein (Fig. 2C). We searched the GenBank protein database with the SUP-10 sequence but did not find any proteins sharing significant amino acid identity with SUP-10.
We identified the molecular lesions of 14 sup-10(lf) alleles (Table 3). sup-10(n1008), sup-10(n1017), sup-10(n2297), and sup-19(n1626) contain nonsense mutations and likely represent null alleles. sup-10(n619) encodes a protein with a glutamate-to-lysine substitution in the predicted extracellular domain of SUP-10, whereas a second missense mutation sup-10(n240) causes a glycine-to-arginine substitution at residue 323 at the beginning of the predicted intracellular domain (Table 4). The gf sup-10(n983) allele contains an amber stop mutation at position 322, resulting in a predicted truncation of the extreme C-terminal 11 amino acids. Because these 11 amino acids span the region following the putative transmembrane domain, they may form an intracellular domain of SUP-10. Consistent with this identification, we found that the locomotory defect of sup-10(n983) gf mutants was partially suppressed by the amber suppressor mutation sup-5(e1464sd) (Table 1). We suggest that the absence of the final 11 C-terminal amino acids of SUP-10 results in altered SUP-10 function.
sup-9::gfp, sup-10::gfp, and unc-93::gfp are expressed predominantly in muscle
To study the expression of sup-9, we constructed a translational fusion of SUP-9 to the GFP. This fusion included 2.9 kb of sup-9 upstream promoter sequence and 0.8 kb of sup-9 downstream sequence (see Materials and Methods). When expressed in a sup-9(lf);unc-93(gf) mutant, the sup-9::gfp reporter could restore the rubberband Unc phenotype caused by an unc-93(gf) allele, indicating that this reporter fusion protein is functional.
We observed SUP-9::GFP expression along the surface of body-wall muscle cells, with a punctate stripe pattern characteristic of dense bodies (Fig. 3A), structures functionally analogous to vertebrate Z-lines, which connect the myofibril lattice to the cell membrane (Waterston et al., 1980). The body-wall muscle staining became apparent at the 3.5-fold stage of embryogenesis, was most apparent in late embryos and L1 stage larvae, and persisted to adulthood. The vulval muscles, predominantly the four Vm1 cells, and the intestinal muscles also displayed GFP fluorescence (Fig. 3B,C), consistent with the egg-laying and defecation defects of sup-9(gf) mutants. We observed weaker fluorescence in the anal depressor and anal sphincter muscles (data not shown). We also observed GFP expression in 12-15 head neurons in each animal (Fig. 3D) (data not shown), including in the SIADL, SIADR, SIAVL, and SIAVR neurons (see Materials and Methods). Additional staining was observed in muscle arms, which project from the muscle body toward neurons to form synapses (Fig. 3E).
To determine whether the tissue and subcellular localization of SUP-9 is dependent on UNC-93, SUP-10, or SUP-18, we examined the expression of the sup-9::gfp reporter in unc-93(lr12), sup-10(n1008), and sup-18(n1030) lf mutants, all likely null alleles (De Stasio et al., 1997; I. Perez de la Cruz and H. R. Horvitz, unpublished results). We found that SUP-9::GFP was expressed and localized as in the wild type in all three mutants (data not shown), indicating that the SUP-9 K+ channel can localize properly in the absence of UNC-93, SUP-10, or SUP-18.
To study the expression of unc-93, we created transgenic lines carrying an unc-93::gfp reporter containing 5.3 kb of unc-93 upstream sequence, the entire unc-93 coding region, and the gfp coding region fused just before the stop codon of unc-93. This construct restored the rubberband Unc phenotype caused by a sup-10(gf) mutation in an unc-93(n234);sup-10(n983) mutant, indicating that the fusion protein was functional. We found GFP expression in the body-wall muscle membranes and dense bodies (Fig. 3F), as well as in all eight vulval and intestinal muscles (Fig. 3G,H). As with the sup-9::gfp fusion, the unc-93::gfp fusion resulted in neuronal GFP expression (Fig. 3I) (data not shown), including in the SIADL, SIADR, SIAVL, and SIAVR neurons. We tested the expression of this reporter in sup-9(n1913), sup-10(n1008), and sup-18(n1030) null backgrounds and found that the localization of UNC-93::GFP remained unchanged (data not shown).
To study the expression of sup-10, we created transgenic lines carrying a sup-10::gfp reporter containing 3.5 kb of sup-10 upstream sequence, the entire sup-10 coding region, and the gfp coding region fused just before the stop codon of sup-10. This construct restored the rubberband Unc phenotype caused by a sup-10 mutation in an unc-93(e1500);sup-10(n183) mutant, indicating that the fusion protein was functional. We found GFP expression in the body-wall muscle membranes and dense bodies (Fig. 3J), eight vulval muscles, intestinal muscles, and anal depressor muscle (Fig. 3K,L,M). As with sup-9::gfp, we observed GFP expression in muscle arms (Fig. 3N).
sup-9, sup-10, and unc-93 function in muscles
To confirm their structure and muscle expression, we overexpressed the sup-9, sup-10, and unc-93 cDNAs under the myo-3 muscle-specific promoter, which expresses in all nonpharyngeal muscle groups (Okkema et al., 1993). We tested whether the myo-3::sup-9 transgene was capable of restoring the sluggish movement and rubberband Unc phenotype of an unc-93(gf) allele in a sup-9(lf);unc-93(gf) mutant. We found that sup-9(n1913);unc-93(e1500) transgenic lines expressing SUP-9 in muscle exhibited a reduced locomotory rate compared with control transgenic lines expressing only GFP (Fig. 4A). The rubberband Unc response was also restored to these lines but not to the control lines expressing only GFP (data not shown). The reduction of locomotory rate by sup-9 overexpression in muscle was not caused by a nonspecific effect of transgene overexpression but rather to an interaction between sup-9 and unc-93(gf), because the myo-3::sup-9 transgene in an unc-93(+) background did not cause a reduced locomotory rate (Fig. 4B). Likewise, we tested whether unc-93 and sup-10 expression in muscle would restore the sup-9(gf) phenotype to a sup-9(gf);unc-93(lf) or a sup-9(gf);sup-10(lf) mutant, respectively. Lines transgenic for myo-3::unc-93 or myo-3::sup-10 showed a severe locomotory defect compared with control myo-3::gfp animals (Fig. 4C,D). We conclude that sup-9, sup-10, and unc-93 function in muscle cells.
SUP-9, SUP-10, and UNC-93 colocalize intracellularly
If sup-9, sup-10, and unc-93 encode components of a multi-subunit K+ channel complex, we expect them to colocalize within muscle membranes in vivo. To test for colocalization, we generated rabbit antibodies against SUP-9. Although these antibodies did not allow us to detect endogenous SUP-9 in whole-mount stainings of wild-type worms (Perez de la Cruz and Horvitz, unpublished observations), likely because of low expression, the antibodies did yield specific staining in transgenic worms overexpressing SUP-9 under the myo-3 promoter (Fig. 5A). Double-label whole-mount stainings of fixed worms with anti-SUP-9 and anti-GFP sera revealed identical muscle membrane distributions of SUP-9 and UNC-93::GFP (Fig. 5A). These proteins colocalized both to the dense bodies and more diffusely throughout the cell surface. Some muscle cells contained brightly staining clusters of SUP-9 protein, which may represent mislocalized protein resulting from overexpression. UNC-93::GFP was colocalized with SUP-9 in these clusters (Fig. 5B). Similarly, SUP-9 and SUP-10::GFP showed identical localization patterns in double-label experiments using animals carrying SUP-9 and SUP-10::GFP transgenes (Fig. 5C).
We compared the localization of SUP-9 with that of PAT-3, an integrin β subunit that is a structural component of dense bodies (Gettner et al., 1995), to test whether overexpression of another muscle membrane protein also resulted in its colocalization with SUP-9. We found that although both SUP-9 and PAT-3::GFP were localized to dense bodies, their distribution patterns were only partially superimposable. Unlike PAT-3::GFP, SUP-9 was present in the spaces between adjacent dense bodies, weakly diffused throughout the muscle membrane, and primarily excluded from the M-lines (Fig. 5D). This result is consistent with the hypothesis that UNC-93 and SUP-10, but not PAT-3, interact with SUP-9 in vivo.
Muscimol can phenocopy the rubberband Unc phenotype
We hypothesized that the rubberband Unc phenotype of sup-9(gf), sup-10(gf), and unc-93(gf) animals may be caused by an inappropriately open K+ channel. Increased K+ efflux from muscle would lead to an accumulation of negative charge inside the cell and could result in membrane hyperpolarization and reduced muscle contraction. Electrophysiological patch-clamp recordings of C. elegans muscles reveal that treatment of body-wall muscle cells with muscimol, a GABA Cl- channel agonist, leads to hyperpolarization by an inward Cl- flux that is dependent on the GABA receptor channel UNC-49 (Richmond and Jorgensen, 1999). We tested whether pharmacologically hyperpolarizing the muscles of wild-type worms would result in a rubberband Unc phenotype similar to that of sup-9(gf), sup-10(gf), or unc-93(gf) mutants.
Worms treated with muscimol displayed a rubberband Unc response in a concentration-dependent manner (Fig. 6A). Saturation of the rubberband response was reached at a concentration of 2.5 mm muscimol (data not shown). The responses of wild-type animals at high drug concentrations resembled those of the strong sup-9(n1550) and unc-93(e1500) mutants, whereas lower doses resembled those of the weaker sup-10(n983) and unc-93(n200) mutants (Fig. 6B). Additionally, worms treated with muscimol displayed uncoordinated movement and flaccid, extended body postures similar to those of the sup-9(gf) and unc-93(gf) mutants (data not shown).
We tested whether sup-9, unc-93, sup-10, or sup-18 were required for the rubberband Unc response caused by muscimol. Worms carrying lf mutations in these genes all displayed the rubberband Unc response (Fig. 6C), whereas, in the absence of muscimol, these lf mutants behaved identically to wild-type animals in showing no rubberband Unc responses (data not shown). Finally, we tested whether the rubberband Unc response caused by muscimol was additive with that caused by the sup-9(gf) mutations. We found that treatment with 1 mm muscimol enhanced the rubberband Unc phenotype of sup-9(n1550)/+ mutants (Fig. 6 D). Collectively, these results establish that muscimol induces the rubberband response by a mechanism that is independent of sup-9, unc-93, sup-10, and sup-18.
Overexpression of SUP-9(n1550gf) does not bypass the requirement for UNC-93
Why does the gf SUP-9(n1550) channel require UNC-93 to cause muscle paralysis? One possibility is that UNC-93 acts as a chaperone to increase levels of the mutant SUP-9 channel at the cell surface. Alternatively, UNC-93 may have little effect on the number of membrane SUP-9 channels but rather increase their K+-transporting activity by stabilizing their open state. To distinguish between these models, we overexpressed the sup-9(n1550) gf cDNA under the control of the myo-3 muscle-specific promoter in transgenic arrays in an attempt to bypass the requirement for unc-93. Three independent lines overexpressing sup-9(gf) in an unc-93(lr12) null background displayed identical locomotion rates to the wild-type strain (26.6, 26.8, and 26.0 vs 26.3 bends per minute, respectively) (Fig. 7A). Thus, in the absence of UNC-93, excess amounts of SUP-9(gf) channel did not hyperpolarize muscle cells. As expected, introduction of a single wild-type copy of unc-93 into each transgenic strain resulted in a severe paralysis (0.5-1.5 bends per minute) (Fig. 7A). To confirm that the SUP-9(gf) channel was being overexpressed, we immunostained these transgenic lines with anti-SUP-9 antiserum (Fig. 7B). As expected, sup-9(n1550)-transgenic animals, but not their nontransgenic siblings, displayed a robust signal on their muscle surface, confirming the overexpression of SUP-9(n1550) channels in these animals. We conclude that UNC-93 likely does not function as a chaperone for SUP-9.
unc-93 belongs to a large C. elegans gene family and is conserved in flies and mammals
We searched the C. elegans genome for predicted genes with sequence similarities to UNC-93. We found that UNC-93 defines a family of 17 worm genes, of which UNC-93 is a relatively divergent member (Fig. 8). UNC-93 contains a highly charged 245 amino acid N-terminal domain followed by 5-10 putative transmembrane domains (Levin and Horvitz, 1992). This N-terminal domain is unique to UNC-93 and appears not to be conserved in the other C. elegans members of this family.
In addition, by searching nucleotide and protein databases, we identified four Drosophila melanogaster and three mouse and three human genes with predicted products with sequence similarity to UNC-93 (Fig. 8). Three pairs of mouse and human genes are likely orthologs, because they are more closely related to each other than to other UNC-93-like genes in their respective species. Of the UNC-93-like genes, the human predicted gene 366N23.1/.2 was the most similar to UNC-93, sharing 30% amino acid identity. Mouse ET8 and human ET22 are more similar to the C. elegans predicted gene C27C12.4 than to UNC-93. The existence of UNC-93-like genes in other organisms suggests that regulation of two-pore K+ channels by regulatory subunits may be a common mechanism.
The C. elegans genome project has identified more than 80 K+ channels, ∼50 of which belong to the two-pore structural class of background or “leak” channels (Bargmann, 1998). Mutants defective in several C. elegans voltage-gated K+ channels have been characterized, including the Kv channels exp-2 (Davis et al., 1999) and egl-36 (Elkes et al., 1997; Johnstone et al., 1997), the HERG channel egl-2 (Weinshenker et al., 1999) and the large-conductance Ca2+-activated channel slo-1 (Wang et al., 2001). While two-pore channels define the largest family of K+ channels, mutations affecting only one gene in this family, twk-18, have been reported (Kunkel et al., 2000).
We cloned the muscle regulatory gene sup-9 and found that it encodes a two-pore K+ channel similar to the mammalian TASK-1 and TASK-3 channels. Genetic evidence strongly indicates that the SUP-9 protein forms a protein complex with at least two other proteins, SUP-10 and UNC-93 (Greenwald and Horvitz, 1980, 1982; Levin and Horvitz, 1993; De Stasio et al., 1997). SUP-10 and UNC-93 are novel transmembrane proteins. SUP-9, SUP-10, and UNC-93 function in muscle and colocalize in muscle membranes. Given these findings, we propose that SUP-9, SUP-10, and UNC-93 form a multi-subunit K+ channel complex that regulates C. elegans muscle contraction. We identified a large family of unc-93-like genes in C. elegans as well as related genes in Drosophila and mammals. We suggest that the regulation of two-pore channels by auxiliary subunits is evolutionarily conserved.
The SUP-9(gf) K+ channel may be stabilized in an open conformation
The molecular identity of sup-9 as a K+ channel, together with the rubberband Unc phenotype of wild-type animals treated with muscimol, suggests that the rubberband Unc phenotype of sup-9(gf) mutants is the result of an increased K+ efflux and a hyperpolarization of muscle cells. Consistent with this model, gf mutations in the muscle two-pore K+ channel twk-18 result in a rubberband Unc paralysis that is indistinguishable from that of sup-9(gf) mutants (Kunkel et al., 2000; Perez de la Cruz and Horvitz, unpublished observations). Heterologous expression of wild-type and mutant forms of TWK-18 channels in Xenopus oocytes reveals 30-fold greater K+ currents of gf channels than wild-type channels (Kunkel et al., 2000). We propose that gf mutations in sup-9 result in greater K+ efflux and hyperpolarization of muscle cells.
What is the mechanism by which an alanine-to-threonine substitution at amino acid 236 in SUP-9 might increase K+ channel activity? We considered three distinct mechanisms: by an increase in the number of SUP-9 channels at the cell surface, by an increase in the unitary conductance of each channel, or by an increase in the open probability of SUP-9. Because overexpression of wild-type SUP-9 under the myo-3 promoter did not result in an Unc phenotype in transgenic animals (Fig. 4), it is unlikely that the gf mutation in sup-9 causes an Unc phenotype by increasing the number of channels at the cell surface. We postulate that the gf mutation in sup-9 results in either a higher unitary conductance or a higher open probability.
All four gf mutations in sup-9 cause the same A236T amino acid substitution within the C-terminal half of the fourth transmembrane domain of SUP-9. A comparison of the crystal structures of two bacterial Kir-like channels, the closed KcsA channel (Doyle et al., 1998) and the open MthK channel (Jiang et al., 2002a), suggests that during channel opening, there is a rotation of the transmembrane domains that follow the P-domain (Jiang et al., 2002b). Because the gf substitution occurs within this gating domain, we postulate that the mutant threonine may stabilize the fourth transmembrane domain in its rotated conformation, resulting in a SUP-9 channel that is constitutively open.
We determined that the second-site compensatory mutations n2288 and n2359, which counteract the effects the gf A236T substitution (Levin and Horvitz, 1993), affect adjacent amino acids, G173E, and A174T, respectively, in the third transmembrane domain of SUP-9 (Table 3). In the two K+ channels for which a crystal structure has been solved, KcsA and MthK, the transmembrane helices flanking the P-domain physically interact (Doyle et al., 1998; Jiang et al., 2002a). Thus, these second-site compensatory mutations may counteract the effects of the gf threonine substitution by directly interacting with its side chain and neutralizing its stabilization of the open state. Alternatively, these second-site mutations may interact with other residues in the fourth transmembrane domain or induce a conformational change in the third transmembrane domain that stabilizes the closed conformation of the channel. We postulate that interactions between the third and fourth transmembrane domains of SUP-9 are important in channel gating.
SUP-10 and UNC-93 may regulate SUP-9 K+ channel gating
Although 14 mammalian two-pore channels have been cloned and characterized thus far (Goldstein et al., 2001), no two-pore channel regulatory subunits have been reported. Three cloned mammalian two-pore channels, KCNK7, KCNK13, and KCNK15, have not produced K+ currents when expressed in a heterologous system, suggesting that association with other proteins may be necessary for their functioning (Rajan et al., 2001). Similarly, we have been unable to obtain K+ currents from SUP-9 or SUP-9(n1550), expressed alone or together with SUP-10 and UNC-93, in Xenopus oocytes or human embryonic kidney (HEK) cells using whole-cell voltage-clamp configuration, although our control experiments demonstrated a robust K+ activity from the mammalian rat TASK-1 channel expressed in both cell types (Perez de la Cruz and Horvitz, unpublished observations). Our experiments were limited to detecting pH-sensitive currents, a hallmark of TASK-1 and TASK-3 channels; we did not vary other factors such as temperature, neurotransmitters, or anesthetics. Our immunohistochemical stainings of HEK cells transfected with SUP-9, SUP-10, and UNC-93 using anti-SUP-9 antisera indicate that much of SUP-9 was not targeted to the cell membrane but instead appeared to be trapped in the endoplasmic reticulum-Golgi complex (Perez de la Cruz and Horvitz, unpublished observations). Mutations in two additional genes, sup-18 and sup-11, suppress the rubberband-Unc phenotype of sup-9(gf), sup-10(gf), and unc-93(gf) mutants and thus may encode additional regulators of the proposed channel complex. It is possible that the SUP-18 and SUP-11 proteins are required for channel activity in heterologous expression systems (Greenwald and Horvitz, 1982, 1986; Levin and Horvitz, 1993).
K+ channel regulatory subunits play diverse roles during their association with the pore-forming α subunits of channels from other families. The cytoplasmic regulatory subunit Kvβ2 associates with Shaker-like Kv channels and increases their surface expression in heterologous systems (Shi et al., 1996) while exerting small effects on channel activity (Rettig et al., 1994; Heinemann et al., 1996). In addition to regulating K+ channels through chaperone activity or altering gating kinetics, some channel subunits, such as the SUR1 subunit of Kir6.2 channels or the β1 subunits of Slo channels, alter the sensitivities of their associated channels to activating signals, such as voltage, ATP, or Ca2+ (McManus et al., 1995; Tucker et al., 1997).
Three lines of evidence suggest that SUP-10 and UNC-93 do not behave simply as chaperones to increase SUP-9 surface expression. First, we found that in both sup-10(lf) and unc-93(lf) mutants, SUP-9::GFP is expressed on the cell surface of muscle cells, suggesting that association of SUP-9 with SUP-10 or UNC-93 is not required for SUP-9 cell-surface expression. Second, overexpression of SUP-9(gf) in transgenic unc-93 null mutants did not result in a rubberband Unc paralysis, indicating that UNC-93 is required for SUP-9 channel activity. Third, genetic analysis of gf mutations in sup-9, sup-10, and unc-93 strongly indicates that these three genes encode subunits of a protein complex (Greenwald and Horvitz, 1980, 1986; Levin and Horvitz, 1993). Because gf mutations in the proposed channel subunits sup-10 and unc-93 result in a muscle paralysis similar to that found in sup-9(gf) channel mutants, SUP-10 and UNC-93 likely function in the regulation of SUP-9 channel activity.
Implications for K+ channel biology
UNC-93 and SUP-10 may represent new classes of K+ channel regulatory subunits. Sixteen other C. elegans genes encode proteins with sequence similarity to UNC-93. Members of this family could associate with the more than 50 C. elegans two-pore K+ channels (Bargmann, 1998) to create a striking level of functional diversity. No interacting genes have been identified through genetic screens for the C. elegans two-pore K+ channel encoded by twk-18 (Kunkel et al., 2000), suggesting that unlike sup-9, some two-pore K+ channels in C. elegans may function without regulatory subunits. Alternatively, such subunits might be either functionally redundant or essential, and for this reason have not been identified. The presence of UNC-93-like genes in the genomes of flies and humans suggests that UNC-93-like regulatory subunits represent a conserved mechanism for the regulation of two-pore K+ channels. In contrast, we found no genes with similarity to sup-10 in the C. elegans genomic sequence or in mammalian expressed sequence tag or genomic databases, suggesting that SUP-10 may be a specialized regulatory subunit specific to the SUP-9-UNC-93 channel complex. Additional studies of the interactions between two-pore K+ channels and their presumptive regulatory subunits should enhance our understanding of this major family of K+ channels.
This work was supported by National Institutes of Health (NIH) Grant GM24663. I.P. was supported by an NIH predoctoral training grant. H.R.H. is an investigator for the Howard Hughes Medical Institute. We thank J. Hodgkin for the sup-9(e2655) and sup-9(e2661) mutants, C. Ceol for the sup-9(n3310) mutant, B. DeStasio for the sup-9(lr) lf mutants, B. Williams for supplying the pat-3::gfp plasmid, and A. Fire for worm expression plasmids. We thank E. Jorgensen for making the initial observation that muscimol induces the rubberband Unc response, B. Castor for help with determining the sequence of sup-9 alleles, R. Ranganathan, E. Speliotes, and B. DeStasio for critically reading this manuscript, and members of the Horvitz laboratory for suggestions during the course of this work.
Correspondence should be addressed to H. Robert Horvitz, Howard Hughes Medical Institute, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139. E-mail:.
Copyright © 2003 Society for Neuroscience 0270-6474/03/239133-13$15.00/0