A Family of K (cid:1) Channel Ancillary Subunits Regulate Taste Sensitivity in Caenorhabditis elegans *

We have identified a family of ancillary subunits of K (cid:1) channels in Caenorhabditis elegans . MPS-1 and its related members MPS-2, MPS-3, and MPS-4 are detected in the nervous system of the nematode. Electrophysiologi-cal analysis in ASE neurons and mammalian cells and epigenetic inactivation by double-stranded RNA inter-ference (RNAi) in vivo show that each MPS can associate with and functionally endow the voltage-gated K (cid:1) channel KVS-1. In the chemosensory neuron ADF, three different MPS subunits combine with KVS-1 to form both binary (MPS-1 (cid:1) KVS-1) and ternary (MPS-2 (cid:1) MPS-3 (cid:1) KVS-1) complexes. RNAi of mps-2 , mps-3 , or both, enhance the taste of the animal for sodium without altering the susceptibility to other attractants. When sodium is introduced in the test plate as background or as antagonist attractant, the nematode loses the ability to recognize a second attractant. Thus, it appears that the chemosensory apparatus of C. elegans uses sensory thresholds and that a voltage-gated K (cid:1) channel is specifically re-quired for this mechanism.

erate functional diversity, and we hypothesized that, by analogy with other species (6,7), MPS-1-related proteins might exist and operate in C. elegans. This effort led us to discover three novel mps members that establish, together with MPS-1, the C. elegans family of KCNE ancillary subunits of K ϩ channels. MPS-2, -3, and -4 are shown to associate with and to modulate the A-type K ϩ current mediated by KVS-1 in vitro and in vivo. Moreover we show that one of these complexes, resulting from the assembly of MPS-2 and MPS-3 with KVS-1, is specifically involved in tuning the responsiveness of the animal for sodium.

Cloning of MPS-2.a, MPS-2.b, MPS-3, and MPS-4 -Cloning was
performed with a Smart Race kit (Clontech) using poly(A) ϩ mRNA extracted from total C. elegans RNA with a Oligotex kit (Quiagen). cDNA was amplified by PCR and inserted in pCI-neo vector (Promega) for functional expression in Chinese hamster ovary (CHO) 1 cells. All sequences were confirmed by automated DNA sequencing. Transcripts were quantified with spectroscopy and compared with control samples that were separated by agarose gel electrophoresis and stained with ethidium bromide. The novel genes have been assigned the following accession numbers by the Genome Database Nomenclature Committee: AY255667 (mps-2.a), AY255666 (mps-2.b), AY255668 (mps-3), and AY255665 (mps-4).
Construction of Tagged Reporter Fusions to GFP-To obtain transgenic nematodes expressing GFP-tagged MPS-2, MPS-3, and MPS-4, the last ϳ1 kb of these genes was amplified by PCR from genomic DNA and joined in-frame to the GFP reporter gene in the pPD 95.75 vector (Fire Vector Kit). The reporter constructs and the cosmids were linearized, annealed at 65°C for 5 min, and co-injected with the transformation marker lin- 15(ϩ) into the syncitial gonad of adult hermaphrodite lin-15(Ϫ) nematodes at a concentration of 25, 100, and 50 ng/l, respectively. Five transgenic lines carrying extrachromosomal arrays were identified for each construct. Because the constructs intentionally lack the promoter and the initial methionine, they are not translated without recombination with the cosmid (which contains the entire gene and its promoter). Worms were analyzed and photographed with an Olympus BX61 microscope equipped with a digital camera.
Dye-filling Experiments-To identify amphid neurons, transgenic worms were picked to a plate containing 1,1Ј-dioctodecyl-3,3,3Ј,3Ј-tetramethylindodicarbocyanine, 4-chlorobenzinesulfonate (Molecular Probes) diluted in M9 buffer (0.01 mg/ml), and allowed to stain for 2-3 h at room temperature. Worms were then transferred to an agar plate and allowed to crawl on the bacterial lawn for about 15 min to destain.
RNA Interference-For double-stranded RNA (dsRNA) production, ϳ0.5-kb regions of mps-2, mps-3, and mps-4 genomic DNA were amplified by PCR with oligonucleotides that added 5Ј T7 promoter sequence. dsRNA in vitro synthesis was with MEGAscript kit (Ambion) using the PCR products as templates. The reactions were annealed at 37°C for 30 min after denaturation (68°C for 10 min). RNA was analyzed by agarose gel electrophoresis to verify that it was double stranded. Approximately 100 pl of dsRNA (1.5 g/l in H 2 O) was injected into both gonads of young adults. Worms were allowed to lay the eggs contained in the uterus for 2-3 h and then transferred separately onto fresh plates. F1 progeny of injected worms were analyzed.
Embryonic Cultured Cells-Cultured ASE right (ASER) neurons were prepared from gyc-5 strains using a method developed by Christensen et al. (8). Gravid adult worms were lysed using 0.5 M NaOH and 1% NaOCl . Released eggs were washed three times with sterile egg  buffer containing 118 mM NaCl, 48 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 ,  and 25 mM Hepes (pH 7.3, 340 mosM), and adult carcasses were separated from washed eggs by centrifugation in sterile 30% sucrose. Eggshells were removed by resuspending pelleted eggs in a sterile egg buffer containing 1 unit/ml chitinase at room temperature for 20 min. Embryos were resuspended in L-15 cell culture medium containing 10% fetal bovine serum, 50 units/ml penicillin, and 50 g/ml streptomycin (Sigma) and dissociated by gentle pipetting. The osmolarity of the culture medium was adjusted to 340 mosM with sucrose and filtersterilized. Intact embryos, clumps of cells, and larvae were removed from the cell suspension by filtration. Dissociated cells were plated on glass coverslips previously coated with peanut lectin (0.1 mg/ml) dissolved in water.
Electrophysiology-CHO cells were transiently transfected with cDNA ligated into pCI-neo using Superfect kit (Qiagen) and studied after 24 -36 h. Data were recorded with an Axopatch 200B (Axon), a personal computer (Dell), and Clampex software (Axon); filtered at 1 kHz; and sampled at 2.5 kHz. Bath solution was (in mM) 4 KCl, 100 NaCl, 10 Hepes (pH 7.5 with NaOH), 1. , 2% agar) were devoid of Na ϩ . Chunks were put back in the plate overnight to allow equilibration and formation of a gradient. Twenty worms were placed between this spot and a control spot on the opposite side of the plate. After 1 h of chemotaxis, animals on the two different sides of the plate (the one with the attractant and the one with the control) were counted, and a chemotaxis index (CI) was calculated as where N test , N cont. , and N tot indicated the number of animals at the test spot and in the control spot and the total number of animals, respectively. A positive CI indicated an attraction to the attractant.

RESULTS
Cloning the Family of MPS Proteins-We blasted MPS-1 sequence against the C. elegans genome, and we identified four new MPS-1-related proteins, MPS-2.a and MPS-2.b (K01A2.8.a,.b), MPS-3 (T06A4.2), and MPS-4 (F09G8.9), and confirmed their expression and primary sequences by analysis of reverse transcription-PCR products (Fig. 1A). All sequences, with the exception of K01A2.8b, turned out to be as predicted, although a third putative K01A2.8 isoform, K01A2,8.c, was not present in our libraries. Like other KCNE proteins, C. elegans members exhibit the strongest conservation in the transmembrane spanning domain. Overall, however, the C. elegans genes appeared to be more related to each other than to genes of other species.
MPS Proteins Are Expressed in the Nervous System of the Nematode-To determine the cellular expression patterns of the novel genes we constructed transgenic nematodes harboring translational GFP reporter fusions (9). We detected strong MPS-2 signals in the ADF amphid sensory neurons (Fig. 1B). MPS-2 was expressed in the enteric muscle (Fig. 1E), albeit signals were fainter compared with those of ADF cells. MPS-3 was also intensely expressed in ADF neurons (Fig. 1C) and, in addition, in two neurons of the pharynx anterior bulb (Fig. 1F), in PVC and PVN neurons (not shown). MPS-4 was present in the neurons that we tentatively identified as AIA (Fig. 1D) and AUA. The latter have been shown to regulate social feeding behavior (10). Less intense but perceptible MPS-4 signals were also detected in enteric muscle (Fig. 1G) and PVC neurons (not shown).
Each MPS Can Interact with K ϩ Channel KVS-1 in Mammalian Cells-Some cells, including the ADF neurons and the enteric muscle, express the voltage-gated K ϩ channel KVS-1, a previously identified partner for MPS-1 (4). This suggests that in addition to MPS-1 the other members of the family might also interact with this channel. To test whether the novel MPS proteins could work as ion channel subunits we co-expressed them with KVS-1 in Chinese hamster ovary cells and used the whole-cell configuration of the patch clamp to characterize currents. Each MPS altered the characteristics of the current, suggesting interaction ( Fig. 2A). Fig. 2B shows the peak current density at ϩ120 mV of KVS-1 alone and with each of the four MPS subunits (MPS-2.b gave results similar to the "a" isoforms, not shown). MPS-1 and MPS-4 inhibited the current roughly 3-5-fold, and MPS-3 increased this quantity significantly (3-fold). Inactivation, a physiologically relevant characteristic of A-type channels, was also affected by the MPS proteins (Fig. 2C). The time course of inactivation at ϩ120 mV, fitted to a single exponential, varied from ϭ 50 ms for channels formed with MPS-3 to ϭ 7 ms with MPS-4, a 7-fold change. Recovery from inactivation was quantified by using a 1.2-s and a 0.2-s depolarizing pulse at ϩ60 mV spaced out by progressively longer periods at Ϫ80 mV. Recovery was slowed down by MPS-3 and speeded up by the other members (Fig.  2D). Moreover, MPS-1 and MPS-3 co-immunoprecipitated with KVS-1 (not-shown), suggesting the formation of stable complexes between the pore-forming subunit and the MPS accessory proteins.
KVS-1 Forms Functional Complexes with the MPS Proteins in Native Cells-To corroborate the notion that KVS-1 and the MPS proteins form functional complexes in native cells we recorded currents from cultured C. elegans ASE neurons (8) that express KVS-1 and MPS-1 (4). Cells were obtained from P gyc-5 ::gfp nematodes that express GFP in the ASE right neuron under the guanylyl cyclase gene gcy-5 promoter (11). To our knowledge, specific markers for other cells expressing KVS-1 and the MPS proteins such as ADF neurons are not currently available, and further, the GFP fluorescence of cells cultured from mps-2::gfp or mps-3::gfp transformants was not sufficiently intense to allow reliable identification using our patch clamp microscope. Voltage pulses from Ϫ80 mV to ϩ80 mV evoked robust outward currents (Fig. 3A) that reversed at Ϫ44.3 Ϯ 9.1 mV. Unlike untreated neurons, cells created by kvs-1 RNAi (Fig. 3B) or mps-1 RNAi (Fig. 3C) were characterized by slow activation and lack of inactivation (Fig. 3D). Treatment by RNAi (Fig. 3E) or application of 10 mM 4-aminopyridine (not shown) led to an ϳ35% decrease of the peak current suggesting, first, that a MPS-1/KVS-1 channel complex is functional in ASER neurons, and second, that knockdown of each subunit is sufficient to destabilize the entire channel. We showed previously (4), that mps-1 RNAi in kvs-1::gfp animals (that express full KVS-1 proteins fused to GFP) suppressed KVS-1 signals only in the neurons in which the two subunits co-localized, and similar effects were observed with kvs-1 RNAi in mps-1::gfp strains (see also Fig. 4, B and C). Moreover RNAi had no effect on transgenic strains expressing transcriptional reporters (P kvs-1 ::gfp, and P mps-1 ::gfp) that express GFP protein driven by kvs-1 or mps-1 promoter sequences (not shown), indicating that RNAi does not interfere with gene transcrip-tion. Probably, if two subunits are assembled in the rough endoplasmic reticulum, disruption of one subunit is sufficient to interrupt the biosynthesis of the complex. Because it is reasonable to assume that the assembly of KVS-1 with the other MPS subunits follows a route similar to that of MPS-1 we used the RNAi method to detect subunit-subunit interactions in native neurons. Thus, kvs-1 RNAi suppressed mps-2::gfp and mps-3::gfp fluorescence in ADF neurons (Fig. 4, A, D, and E), and mps-2 and mps-3 RNAi also affected kvs-1::gfp signals in the same neurons albeit to a lesser extent (Fig. 4B). Moreover, kvs-1 RNAi was very effective against mps-4::gfp signals in AUA and vice versa (not shown). Thus, each MPS appears to be

FIG. 3. A MPS-1⅐KVS-1 complex is functional in ASER neurons.
A, wholecell currents recorded from cultured ASER neurons. Currents were elicited by voltage jumps from Ϫ80 mV to ϩ80 mV in 20-mV increments. B-C, same as in A from cells incubated with kvs-1 or mps-1 dsRNA (15 g/ml). D, macroscopic currents could be fitted to a single exponential function, I 0 ϩ I 1 exp(Ϫt/) (inset). Time constants for control (E, 9 cells) or for cells incubated with kvs-1 RNAi (•, 8 cells) or mps-1 RNAi (f, 4 cells). E, I-V relationships for peak currents measured 30 ms after voltage stimulation from control (9 cells) or from cells incubated with kvs-1 RNAi (8 cells) or mps-1 RNAi (4 cells). Peak currents were normalized to the steady-state value calculated at the end of the voltage jump. capable of forming a complex with KVS-1 in native neurons, consistent with the previous finding ( Fig. 2) that in CHO cells each MPS modulates KVS-1.

MPS-2 and MPS-3 Form a Ternary Channel Complex with KVS-1 in ADF Neurons-Some cells co-express multiple MPS
subunits, suggesting the possibility that they might form mixed complexes with KVS-1, that is, complexes containing more than one MPS member. In particular, in ADF neurons we detected KVS-1 and three MPS proteins (MPS-1, MPS-2, and MPS-3). Therefore, these sensory cells represented an optimal system to investigate the interactions between the MPS and KVS-1. Thus, mps-2 RNAi suppressed mps-3::gfp fluorescence only in ADF neurons, and a similar effect was observed with mps-3 RNAi in mps-2::gfp animals (Fig. 4, D and E). In contrast, mps-1 RNAi failed to affect either MPS-2 or MPS-3 fluorescence (Fig. 4, D and E). These data lead us to suggest that KVS-1 can form binary and ternary complexes with the MPS subunits in ADF neurons.
Functional Properties of a Ternary Complex in CHO Cells-To test whether MPS-2, MPS-3, and KVS-1 could form a functional ternary complex we transfected the cDNA encoding these subunits in CHO cells and used the whole-cell configuration of the patch clamp to study the currents as done before. The magnitude of the current density of MPS-2/MPS-3/KVS-1 channels was similar to that of binary complexes formed by MPS-3 and KVS-1 (Fig. 5A), whereas inactivation kinetics were like those of complexes formed with MPS-2 (Fig. 5, B and C). Despite the fact that these currents probably have contributions from populations of binary complexes, their mixed characteristics suggest the formation of functional MPS-2⅐MPS-3⅐KVS-1 ternary complexes.
A MPS-2⅐MPS-3⅐KVS-1 K ϩ Channel Complex Tunes the Responsiveness to Sensory Cues-The current model for chemotaxis to water-soluble attractants predicts that this sensory function is controlled by a network of five amphid neurons organized hierarchically (1,12). Thus a single pair of neurons, the ASEs, determine the primary response, and the group composed by the remaining four pairs, ADF ASG, ASI, and ASK, mediate a residual response (3). To elucidate the physiological role of K ϩ channels formed by the assembly of MPS-1, MPS-2, MPS-3, and KVS-1 in ADF neurons we evaluated the ability of animals treated by RNAi to recognize classic attractants such as Na ϩ , Cl Ϫ , cAMP, biotin, and lysine. Inactivation of kvs-1 or mps-1, which is expressed in several neurons of the network including the ASE pair, leads to defective chemotaxis to these cues (4). In contrast, mps-2, mps-3, or mps-2ϩmps-3 RNAi did not alter susceptibility to Cl Ϫ , cAMP biotin, and lysine (Fig. 6A). These nematodes exhibited an unexpected behavior, however, in that their taste for Na ϩ was enhanced (Fig. 6B). Co-injection of dsRNAs encoding mps-2 and mps-3 did not further augment Na ϩ sensibility, suggesting that the two subunits contribute similarly to this behavior probably as FIG. 4. KVS-1, MPS-1, MPS-2, and MPS-3 form ternary complexes in ADF amphid neurons. A, representative pictures of mps-3::gfp transgenic nematodes. These transformants display GFP fluorescence in ADF neurons (arrow). When the same animals are given injections of kvs-1 or mps-2 dsRNA, fluorescence becomes very faint and cannot be appreciated by eye (a fluorescence ratio (FR) of ϳ0.3 is background), suggesting that all three subunits contribute to form a unique complex. B-E, normalized fluorescence ratios in ADF neurons of kvs-1::gfp (B), mps-1::gfp (C), mps-2::gfp (D), and mps-3::gfp (E) nematodes treated with RNAi against the indicated genes. Fluorescence was quantified by ImageJ software (available at rsb.info.nih.gov/ nih-image/). Fluorescence ratio was normalized to control nematodes. Data are from groups of 20 or more worms. Statistically significant differences from control animals are indicated with **, p Յ 0.01 (unpaired t test). parts of the same channel complex. Moreover, when mps-2 was co-injected with mps-1 the nematodes retained the phenotype of mps-1 animals. We explain the dominant effect of mps-1 by the fact that this gene is expressed in several neurons of the network including the ASE pair (4). The broad MPS-1 expression pattern also limited our ability to further dissect the specific role of this complex in ADF neurons. Fig. 6C shows that mps-2 nematodes were more responsive to lower Na ϩ gradients than were wild-type animals. When the concentration was increased, however, both groups exhibited the same phenotype. Thus, the chemosensory apparatus of C. elegans is such that it can detect few sodium ions (2) (ϳ700,000 ions/mm 2 in a 1.0 M sodium acetate agar medium), but a behavioral response is triggered only above a certain threshold. We next sought to evaluate the impact of lowering the sodium threshold on the other sensory activities. In a first group of experiments we FIG . 5. MPS-2, MPS-3, and KVS-1 form a functional complex in CHO cells. A, peak current density for ternary MPS-2⅐MPS-3⅐KVS-1 (n ϭ 11) complexes and for comparison, for MPS-2⅐KVS-1 and MPS-3⅐KVS-1 binary complexes. CHO cells were transfected with an equal amount of MPS-2 and MPS-3 cDNA and studied 24 h after transfection. Peak current densities were obtained by normalizing peak currents at ϩ120 mV to the cell capacitance. Statistically significant differences are indicated with *, p Յ 0. 01 (unpaired t test). B, inactivation rates for MPS-2⅐MPS-3⅐KVS-1 (diamonds, n ϭ 11) and for comparison for MPS-2⅐KVS-1 (solid line), and MPS-3⅐KVS-1 (dotted line). C, dependence of the normalized peak current on the length of the recovery phase for MPS-2⅐MPS-3⅐KVS-1 (diamonds, n ϭ 11) and for comparison for MPS-2⅐KVS-1 (fit, solid line) and MPS-3⅐KVS-1 (fit, dotted line).
FIG. 6. MPS-2⅐MPS-3⅐KVS-1 complexes determine the threshold of sensitivity for Na ؉ ions. A, chemotaxis to classic attractants of animals that have received injections of mps-2, mps-3, or mps-2ϩmps-3 RNAi (three assays each). Attraction to water-soluble attractants is mainly mediated by ASE and to a lesser extent by ASI, ADF, ASG, and ASK neurons. B, chemotaxis to Na ϩ . In these assays a 0.4 M gradient sodium acetate was used. The last two columns indicate coinjection of mps-2 and mps-3 and mps-1 and mps-2 dsRNAs (five to nine assays each). C, chemotaxis to Na ϩ for nematodes injected with mps-2 dsRNA in the presence of the indicated sodium gradients (0.2 and 0.4 M, four to seven assays; 1.0 and 1.5 M, three to five assays). Red and gray, N2 animals and animals treated by mps-2 RNAi, respectively. D, chemotaxis to cAMP in the presence of the indicated sodium phosphate in agar for N2 and mps-2 RNAi animals (four assays). E, chemotaxis in the presence of two attractants for N2 and mps-2 animals (three assays). In these assays the control spot was replaced with cAMP (0.2 M). The CI was calculated as the difference between the number of animals in the Na ϩ spot minus the animals in the cAMP spot. Thus, a negative CI indicates attraction for cAMP and a positive CI attraction for Na ϩ . Animals were tested for chemotaxis to a point source of each odorant. A CI of 1.0 indicates complete attraction; a CI of 0 indicates a random distribution of worms on the assay plate. If not otherwise stated test plates were devoid of Na ϩ . Each column represents the average of three to eight independent assays using a minimum of 100 animals/assay distributed in four to five test plates. Unpaired t tests were calculated for each individual assay. *, p Յ 0.05; **, p Յ 0.01. introduced a Na ϩ background in the test plate. As expected, chemotaxis to cAMP of wild-type animals was not significantly hindered ( Fig. 6D) (2). Conversely, nematodes treated with mps-2 RNAi exhibited a striking loss of attraction for cAMP at Na ϩ backgrounds as low as 0.1 mM (Fig. 6D). Untreated animals require sodium concentrations as high as 100 mM to show detectable loss of attraction (12), which represents a loss of sensitivity of 4 orders of magnitude. Furthermore, attraction to biotin was also hindered under the same conditions (not shown). In a second set of experiments, the nematodes were exposed to a fixed cAMP gradient and to sodium gradients of various intensities (Fig. 6E). Wild-type animals migrated more favorably toward the spot containing cAMP when the sodium gradient was low (this effect corresponds to a negative CI in Fig. 6E). In contrast, mps-2 animals displayed a significant preference for Na ϩ . Taking into account that knockdown of mps-2 or mps-3 had no apparent effect on chemotaxis to other attractants in the absence of Na ϩ (Fig. 6A), these results suggest that when the threshold for Na ϩ is lowered the signal becomes intense and overwhelms the other signals. DISCUSSION KCNE proteins are an emerging family of accessory subunits of K ϩ channels. The ubiquity of their expression together with the ability to assemble with multiple K ϩ channels (13) underscores a potentially very important role of these proteins in human physiology. In fact, mutations in KCNE1, KCNE2, KCNE3, and KCNE5 genes have been linked to congenital and acquired disease (6, 14 -17). Several analogies appear to exist between the C. elegans and the mammalian subfamilies. Like human KCNE proteins that individually assemble with the voltage-gated K ϩ channel KCNQ1 (18 -23), each MPS endows KVS-1. We suggest that this class of ancillary subunits is an integral part of K ϩ channels and that these subunits probably operate through conserved mechanisms.
C. elegans MPS proteins can form ternary complexes with KVS-1, that is, complexes containing more than one MPS member. It is interesting to note that the subunit composition of ternary complexes appears to be tightly regulated because combinations such as MPS-1/MPS-2 and MPS-1/MPS-3 are apparently forbidden. The existence of these complexes is an observation relevant to mammalian physiology because it raises the possibility that the KCNE proteins that have been shown to associate individually with KCNQ1 might also form ternary complexes in human tissues (18 -23). For instance Lundquist et al. (24) recently speculated that in human heart, a balance of KCNE accessory subunits may be important for cardiac K(V) channel function.
Taken together these data provide insights into the mechanisms determining chemotaxis in C. elegans. Our findings imply that the net of interactions between the five neurons of the sensory network is very complex and that neurons thought to be marginal instead have an important role. The use of sensory thresholds controlled by secondary neurons might be an efficient strategy to filter and to balance inputs coming from the diverse sources present in the natural environment of the animal. This mechanism would increase fidelity, and it may be seen as a means to lower the background noise because it is likely that traces of sodium and other attractants are always present in the environment of the animal. Interestingly, a molecular component of the ADF neuronal K ϩ current is linked to a particular sensory function and apparently is not related to other activities. Because K ϩ currents dampen cell excitability, it is reasonable to assume that destabilization of the MPS-2⅐MPS-3⅐KVS-1 complex would make ADF cells more excitable with the result of lowering the threshold for sodium. In fact, we observe this phenotype only at low Na ϩ concentrations. We conclude that sensory pathways require specific genes not only at the level of the receptor molecule but also at the end of the sensory cascade.
The remarkable variety of K ϩ channel genes in mammalian neurons has led to the working hypothesis that the electrical properties of those cells are dynamic and constantly subjected to fine-tuning. In C. elegans a voltage-gated pore-forming subunit is functionally differentiated by assembly with small accessory subunits. These findings disclose an unexpected complexity in C. elegans neurons and underscore a similarity to mammals.