Neurotoxic unc-8 mutants encode constitutively active DEG/ENaC channels that are blocked by divalent cations

DEG/ENaC channel subunits (named after the degenerin MEC-4 and the mammalian epithelial Na⁺ channel; Driscoll and Chalfie, 1991; Canessa et al., 1993) contain two transmembrane domains, a large extracellular loop and short intracellular N and C termini that form trimeric voltage-independent Na⁺ or/and Ca²⁺ channels. Crystal structure analysis of chicken ASIC1a revealed that three DEG/ENaC subunits come together to form a channel, and that the large extracellular domain of each subunit forms a “clenched-hand” structure that protrudes from the plasma membrane (Jasti et al., 2007). Structural and functional analysis has shown that this extracellular domain specifies channel regulation by protons and other ions, and that it undergoes substantial movement during channel gating (Jasti et al., 2007; Zhang et al., 2008; Li et al., 2009; Passero et al., 2009; Sherwood et al., 2009; Wang and Bianchi, 2009; Shi et al., 2011). DEG/ENaC channels are comprised of either homomeric or heteromeric trimers, and at least some DEG/ENaC channels are associated with auxiliary subunits (Lingueglia et al., 1997; Bianchi and Driscoll, 2002; Chelur et al., 2002; Goodman et al., 2002; Askwith et al., 2004).

DEG/ENaC channels are expressed in a wide variety of cell types including epithelial cells (Canessa et al., 1993), neurons (Bianchi and Driscoll, 2002), and glia (Wang et al., 2008; Hitomi et al., 2009; Calavia et al., 2010; Han et al., 2013), where they function to mediate physiological processes as diverse as Na⁺ reabsorption (Hummler et al., 1996), memory formation, and sensory perception (Tavernarakis et al., 1997; Price et al., 2000, 2001; Sluka et al., 2003; Wemmie et al., 2003; Wang et al., 2008; Chandrashekar et al., 2010). Some DEG/ENaC channels, including Caenorhabditis elegans MEC-4 and mammalian ASIC1a, have been shown to undergo “hyperactivation” as a result of genetic mutations (Driscoll and Chalfie, 1991; García-Añoveros et al., 1995) and chronic exposure to protons (Xiong et al., 2004), respectively. In these conditions, DEG/ENaC channels switch to a high open probability mode that results in large whole cell currents (Goodman et al., 2002; Brown et al., 2007). The resultant influx of cations induces cell swelling and death through a mechanism that involves release of calcium from the ER and activation of calcium-sensitive proteases such as calpains and cathepsins (Xu et al., 2001; Syntichaki et al., 2002). Interestingly, both C. elegans MEC-4 and mouse ASIC1a show some calcium permeability, suggesting that, at least in some cases, calcium influx through hyperactivated DEG/ENaCs triggers the cell death program (Zhang and Canessa, 2002; Bianchi et al., 2004; Xiong et al., 2004).

Mutations in the C. elegans DEG/ENaC subunit UNC-8 have been isolated in genetic screens as uncoordinated animals (Brenner, 1974; Park and Horvitz, 1986; Shreffler et al., 1995). The dominant mutants, unc-8(n491) and unc-8(e15), of which both encode the single mutation UNC-8(G387E), display a strong backward movement defect. Microscopic analysis of unc-8(n491) and unc-8(e15) animals detected neuronal swelling during early larval development (Shreffler et al., 1995). Similar cell swelling resulting from hyperactivated DEG/ENaC channels such as MEC-4(d) (Driscoll and Chalfie, 1991) suggests that unc-8(n491) and unc-8(e15) may encode constitutively active UNC-8 channels. Cell swelling in these unc-8(d) mutants is less severe, however, than that observed for mec-4(d) animals. In addition, a majority of the affected motor neurons in unc-8(d) mutants appear to recover by the adult stage, whereas neurons that express MEC-4(d) channels typically undergo necrotic cell death (Driscoll and Chalfie, 1991; Shreffler et al., 1995). A milder neurotoxic phenotype is also caused by expression of MEC-10(d) in touch neurons (Huang and Chalfie, 1994). MEC-10 cannot form channels on its own but associates with MEC-4 to modulate its properties (Goodman et al., 2002). Thus, the milder neurotoxic phenotype of MEC-10(d) was interpreted to be the result of the MEC-10 accessory role in the channel complex.

We undertook this study to better define the neuronal toxicity of unc-8(n491), unc-8(e15), and unc-8(n49) (encoding A586T mutation) mutants and to establish the physiological properties of UNC-8(d) channels that they encode. We show that DA and DB motor neurons swell in unc-8(n491) L1 larval stage and that ∼35% of them degenerate by adulthood. We also show by expression in Xenopus laevis oocytes that UNC-8 forms homomeric channels and that neurotoxic UNC-8 mutations induce channel hyperactivation. However, we found that the overall current amplitude is small because of the high affinity blockage of UNC-8 channels by extracellular divalent cations. Removal of divalent cations results in strong currents and induces Xenopus oocyte cell death. Expression of an UNC-8 intragenic mutation that suppresses neuronal swelling reduces current amplitude. UNC-8(A586T), which corresponds to the relatively weak unc-8(n49) allele, is inhibited by extracellular calcium and acidic pH with higher affinity than the more toxic UNC-8(G387E). Collectively, our data are consistent with the proposal that hyperactive cation transport accounts for the neurotoxicity of UNC-8 gain-of-function mutants, and that this activity is likely regulated in vivo by extracellular divalent cations and pH.

Nematode strains were maintained at 20°C on standard nematode growth medium seeded with Escherichia coli strain OP50 (Brenner, 1974). Wild-type animals were N2 Bristol. Strains used in this work include: CZ1200 (juIs76 [unc-25p::GFP + lin-15(+)] II;lin-15B(n765) X) to label GABA motor neurons with GFP and RP1(trIs10 [myo-3p::MB::YFP + myo-2p::YFP + ceh-23::HcRed + unc-25::dsRed + unc-129nsp::CFP] I) to label GABA motor neurons with dsRed and cholinergic DA/DB neurons with CFP. unc-8(n491) IV was crossed into the marker strains to generate NC2782 [punc-25::GFP (II); unc-8(n491) IV] and NC2799 [unc-8(n491) IV; trIs10 [myo-3p::MB::YFP + myo-2p::YFP + ceh-23::HcRed + unc-25::DsRed + unc-129nsp::CFP] I].

Animals were immobilized with 15 mM levamisole on 2% agarose pads with M9 buffer as described previously (Smith et al., 2012). L1 animals were imaged with a microscope (Axioplan; Carl Zeiss) equipped with a camera (ORCA; Hammamatsu) and ImageJ Micro-Manager software. DIC images of ventral cord motor neurons were collected with a 100× oil objective; young adult animals were evaluated with a 63× oil objective. Images in Fig. 1 (C and D) were acquired with a confocal microscope (TCS SP5; Leica). Z stacks were collected with a 40× oil objective (1 µm/step). Single-plane projections were generated with Application Suite Advanced Fluorescence software (Leica). Insets of the CFP-labeled DA/DB neurons shown in Fig. 1 (C and D) were generated with the ImageJ plug-in “straighten curved objects” (Kocsis et al., 1991). Image overlays were performed with Adobe Photoshop.

To evaluate ventral cord motor neurons in the L1 stage for potential swelling, punc-25::GFP was used to mark GABA neurons in wild-type and unc-8(n491) animals. To generate synchronized populations of L1 larvae, 100 wild-type (CZ1200) and NC2782 [punc-25::GFP;unc-8(n491)] adults were picked to single 60-mm plates and allowed to lay eggs for 1 h. The midpoint of the hour was considered to be t = 0. Plates were maintained at 20°C. At 18 h after lay, early L1 animals were picked from plates and imaged (as above) on a microscope (Axioplan; Carl Zeiss).

RP1 (trIs10) was used to visualize embryonic motor neurons (DA, DB, DD) in the ventral nerve cord in wild-type and unc-8(n491) adults. CFP-labeled DA/DB neurons and dsRed-labeled DD neurons were scored in the ventral cord regions between DD1 through DD6. Results were pooled from three separate experiments.

UNC-8 cDNA was cloned by RT-PCR using the one-step RT-PCR kit Titanium (Takara Bio Inc.) in two parts. We cloned the last ∼1,500 bp (from the HincII site to the stop codon) using UNC-8–specific primers designed according to the UNC-8–predicted cDNA sequence available on WormBase, and the 5′ end (up to the HincII site) using a reverse-specific primer and the splice leader primer SL1 (Graham et al., 1988) (Fig. S2). Total RNA from a C. elegans mixed-age population was used as template. PCR products were initially cloned into TOPO PCR2.1 vector for sequence verification and amplification. Sequence-verified cDNA fragments were then ligated together into PGEM vector (Goodman et al., 2002) to reconstitute the full-length UNC-8 cDNA. Mutations were introduced in UNC-8 cDNA sequence using the Quickchange mutagenesis kit (Agilent Technologies) according to the manufacturer’s procedures. Mutant DNA were then sequenced and used for in vitro cRNA synthesis.

cRNAs were synthesized using T7 mMESSAGE mMACHINE kit (Invitrogen). cRNAs were then purified and run on denaturating agarose gels for size and integrity verification. cRNA quantification was performed spectroscopically. Stage V–VI oocytes were selected among multi-staged oocytes dissected by a 2-h collagenase treatment (2 mg/ml in Ca²⁺-free OR2 solution) from Xenopus ovaries (provided by G. Dahl or C.W. Luetje, University of Miami, Coral Gables, FL). Oocytes were injected with 10 ng/oocyte of each cRNA and incubated in OR2 medium plus 300 µM amiloride. OR2 solution consists of 82.5 mM NaCl, 2.5 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, 1 mM Na₂HPO₄, 0.5 g/liter polyvinyl pyrolidone, and 5 mM HEPES, pH 7.2, supplemented with penicillin and streptomycin (0.1 mg/ml) and 2 mM Na-pyruvate at 20°C for 2–3 d before recordings. Currents were measured using a two-electrode voltage-clamp amplifier (GeneClamp 500B; Axon Instruments) at room temperature. Electrodes (0.2–0.5 MΩ) were filled with 3 M KCl, and oocytes were perfused with a physiological NaCl solution containing (mM) 100 NaCl, 2 KCl, 1 CaCl₂, 2 MgCl₂, and 10 HEPES, pH 7.2. The divalent cation–free plus EGTA NaCl solution contained (mM): 110 NaCl, 2 KCl, 1 EGTA, and 10 HEPES, pH 7.2. CaCl₂ was added to this solution to reach the desired concentrations (Schoenmakers et al., 1992). pH was adjusted at the indicated values using NMDG-Cl or HCl. When we tested solutions at pH lower than 6, we used 10 mM Mes instead of 10 mM HEPES to buffer the solutions. The divalent cation–free NaCl solution contained (mM): 110 NaCl, 2 KCl, and 10 HEPES, pH 7.2. MgCl₂ was added to this solution to reach the desired concentrations. The solution to test calcium permeability contained (mM): 73 CaCl₂, 2 KCl, and 10 HEPES, pH 7.2. The solutions to test UNC-8(G387E) selectivity contained (mM): 112 NaCl (LiCl, KCl, or NMDG-Cl), 1 EGTA, and 10 HEPES, pH 7.2. The oocyte membrane was clamped at −30 mV and stepped from −160 to +100 mV. Amiloride was added to the solutions from a stock of 10 mM, except for when 1 and 5 mM amiloride were used. In this case, amiloride was dissolved directly into the perfusing solution. Saturating concentration of amiloride (1 and 5 mM for experiments in physiological and divalent-free solutions, respectively) was added at the end of each experiment to make sure that endogenous/leak currents were similar in amplitude to noninjected oocytes within each oocyte batch. Oocytes that had larger endogenous/leak currents were not further analyzed. We used the pCLAMP suite of programs (Axon Instruments) for data acquisition and analysis. Currents were filtered at 200 Hz and sampled at 1 kHz.

Occytes were incubated in regular OR2 or OR2 containing 5 µM Ca²⁺ plus or minus 500 µM amiloride for 24 h in individual wells (96-well plate), after injection with UNC-8 wild-type and UNC-8(G387E) cRNAs and overnight incubation in regular OR2. After a 24-h incubation, oocytes were examined under the dissection microscope to establish if they were still intact. Oocytes were scored as “intact” if no cytoplasmic material was visible outside the oocyte.

Fig. S1 shows the fraction of unc-8(n491) animals that lack a given DA, DB, or DD motor neuron versus the location of each motor neuron along the ventral nerve cord from anterior (DD1) to posterior (DD6). Fig. S2 shows UNC-8 protein sequence and alignment with other DEG/ENaC channels, unc-8 gene structure, sequence of the first two exons, and corresponding translation and location of the mutations studied.

We observed significant cell swelling in the ventral nerve cords of unc-8(n491) L1 stage larvae (Fig. 1). Swollen cells were visible in most unc-8(n491) animals at this stage but were rarely observed later in larval development. Swollen cells were also detected in other body regions (e.g., PVC neuron). Our results agree with a previous report that described transient swelling in the ventral nerve cord of L1 stage unc-8(n491) and unc-8(e15) animals (Shreffler et al., 1995). The ventral nerve cords of newly hatched L1 larvae contain three types of motor neurons: the cholinergic DAs (nine neurons), DBs (seven neurons), and GABAergic DDs (six neurons). We relied on the invariant locations of each member of these motor neuron classes (White et al., 1986) and the GABA-specific marker, unc-25::GFP, to determine that swelling is limited to DA and DB motor neurons and was never observed for the DD class in unc-8(n491) L1 animals (Fig. 1 B). We also noted occasional swelling of P cell precursors in the ventral nerve cord (Fig. 1 B). The motor neuron–specific effects are correlated with our finding that a fraction of DA and DB motor neurons (∼36%) is missing from the adult ventral nerve cord, but no absences were observed for DD motor neurons (Fig. 1, C–E, and Fig. S1). The death of DA and DB motor neurons was not previously detected from counts of ventral cord cell nuclei stained with the DNA dye DAPI perhaps because this effect is limited to <10% of all ventral cord motor neurons in the adult and therefore difficult to detect with this approach (Shreffler et al., 1995). Our results suggest that constitutive UNC-8 activity can result in cell death and is thus consistent with similar observations for dominant mutants of the DEG/ENaC channels DEG-1 and MEC-4 (Chalfie and Wolinsky, 1990; Driscoll and Chalfie, 1991).

Tavernarakis et al. (1997) showed that unc-8 encodes a DEG/ENaC subunit and that both n491 and e15 alleles contain the G387E mutation. G387 is located in the UNC-8 extracellular domain in a poorly conserved loop of the “finger” domain (Gessmann et al., 2010) (Fig. S2, A and C). The neuronal swelling phenotype typical of neurons expressing hyperactive DEG/ENaC channels such as MEC-4(d) (Driscoll and Chalfie, 1991) suggested that the G387E mutation may result in a hyperactive UNC-8 channel. To test this hypothesis, we cloned UNC-8 cDNA for heterologous expression in Xenopus oocytes. Using the splice leader sequence primer SL1 (Graham et al., 1988), we identified an UNC-8 transcript that differed from the predicted gene structure (WormBase) in its first two exons (Fig. S2 B) as the most abundant UNC-8 mRNA. Using this UNC-8 cDNA as a template, we generated the UNC-8(G387E) mutant and expressed wild-type and mutant UNC-8 in Xenopus oocytes for electrophysiological analysis. We found that oocytes injected with wild-type UNC-8 did not produce currents above the background level. In contrast, UNC-8(G387E)–injected oocytes expressed an amiloride-sensitive current that reversed at ∼0 mV (Fig. 2, A–D). These results are consistent with the idea that the G387E mutation enhances UNC-8 channel cation transport above the wild-type level and thus corresponds to a “hyperactive” mutation. From a dose–response curve, we calculated an amiloride Ki of 7.8 µM at −100 mV, as expected for a current that depends on DEG/ENaC channel activity (Fig. 2 E). Interestingly, UNC-8(G387E) currents displayed a strong outward rectification with larger values for positive potentials than for voltage clamps below 0 mV.

The outward rectification of the amiloride-sensitive current in UNC-8(G387E)–expressing oocytes (Fig. 2 D) resembles that of other hyperactive DEG/ENaC channels including C. elegans UNC-105(d) and MEC-4(d) (García-Añoveros et al., 1998; Bianchi et al., 2004; Brown et al., 2008) that are negatively regulated by extracellular calcium. To test UNC-8(G387E) for similar regulation, we perfused oocytes with a solution prepared without divalent cations, and with a solution without divalent cations to which 1 mM EGTA was added to chelate residual calcium. In these divalent cation–free solutions, the UNC-8(G387E) current did not rectify and showed a linear response to voltage from −160 to +100 mV (Fig. 3, A–D). Strikingly, currents were also ∼20 times larger both when the membrane was hyperpolarized and depolarized, suggesting that both inward and outward currents are blocked by divalent cations (Fig. 3 D). This behavior differs significantly from that of UNC-105(d) and MEC-4(d) channels for which inhibition by extracellular calcium is limited to currents at negative membrane potentials. We conducted similar experiments in divalent cation–free solution for oocytes injected with wild-type UNC-8 but did not detect currents above the oocyte background level, even when twice as much cRNA was injected (Fig. 3 D and not depicted). These results suggest that wild-type UNC-8 requires a specific stimulus to be activated. Several members of the DEG/ENaC family are gated by extracellular protons (Bianchi and Driscoll, 2002). Thus, we tested whether wild-type UNC-8 was activated by perfusion with a solution at pH 5. However, we found no increase of currents above the background level, suggesting that UNC-8 is gated by stimuli other than protons (not depicted). One possibility is that UNC-8 is gated by mechanical forces, as suggested by Tavernarakis et al. (1997).

Because removal of extracellular divalent cations results in a large increase in UNC-8(G387E) current, we predicted that the resting potential of these oocytes would become more depolarized, consistent with Na⁺ becoming the predominant cation transported into the oocyte, upon perfusion with the divalent cation–free plus EGTA solution. Indeed, we found that oocytes injected with UNC-8(G387E) underwent a significant depolarization with the switch from physiological to divalent cation–free plus EGTA NaCl solution. Conversely, noninjected oocytes depolarized only a few millivolts under the same experimental conditions (Fig. 3 E). Amiloride binding was substantially weakened in the divalent cation–free plus EGTA solution with a Ki of 106 µM (compare Fig. 3 F with Fig. 2 E). Interestingly, although amiloride binding appears to display negative cooperativity in physiological solution (fit with Hill’s equation, n = 0.5; Fig. 2 E), it shows no cooperativity in the divalent cation–free plus EGTA solution (n = 1). Collectively, these results suggest that extracellular divalent cations influence amiloride binding and/or blockage.

Although UNC-8(G387E) generates strongly outwardly rectifying currents, MEC-4(d) currents are not outwardly rectifying (Goodman et al., 2002). Moreover, although removal of divalent cations from the extracellular solution has little effect on MEC-4(d) currents, UNC-8(G387E) currents are increased >20-fold by this treatment (Fig. 3, A–D) (Bianchi et al., 2004; Brown et al., 2008). Interestingly, although MEC-4(d) causes death of all six touch neurons in which it is expressed, UNC-8(G387E) appears to kill only a fraction of DA/DB motor neurons (Fig. 1 E and Fig. S1). Thus, we hypothesized that the amplitude of the current may reflect the level of toxicity of these two mutant channels. We also predicted that relief or reduction of the divalent block of UNC-8(G387E) would exacerbate toxicity. To test this idea, we incubated Xenopus oocytes expressing UNC-8(G387E) in OR2 solution containing only 5 µM Ca²⁺ and no Mg²⁺; 68% of these oocytes appeared swollen and burst after 24 h of incubation. In contrast, the same treatment did not cause bursting of noninjected oocytes or oocytes injected with wild-type UNC-8 (Fig. 4, A and B). The addition of 500 µM amiloride was protective and rescued oocytes from bursting (Fig. 4 B). These results suggest that the block exerted by divalent cations limits the toxicity of the UNC-8(G387E) mutant channel.

Neurotoxic DEG/ENaC channels MEC-4(d) and ASIC1a show some permeability to Ca²⁺ ions (Bianchi et al., 2004; Xiong et al., 2004). However, it is not clear whether Ca²⁺ permeability is necessary for DEG/ENaC-induced neurotoxicity. We tested whether neurotoxic UNC-8(G387E) is Ca²⁺ permeable by perfusing oocytes with a CaCl₂ solution and by monitoring the activation of the endogenous Ca²⁺-activated Cl⁻ current (Bianchi et al., 2004). However, we did not detect currents above the background level (Fig. 4, C and D), indicating that UNC-8(G387E) is not permeable to Ca²⁺ ions. We further analyzed UNC-8(G387E) selectivity properties by measuring current amplitude in oocytes perfused with extracellular solutions in which NaCl was replaced with LiCl, KCl, and NMDG-Cl. We found that the selectivity series for UNC-8(G387E) for monovalent cations was Na⁺ > Li⁺ > K⁺ >> NMDG⁺ (Fig. 4 E). The relatively high K⁺ permeability may explain why the reversal potential of the UNC-8(G387E) was closer to 0 mV than expected for a selective Na⁺ channel. To conclude, UNC-8(G387E) is permeable to both Na⁺ and K⁺, but it is not permeable to the larger monovalent cation NMDG⁺ or to the divalent cation Ca²⁺.

For quantitative measurements of calcium and magnesium affinity, we measured UNC-8(G387E) currents while perfusing the oocytes with increasing concentrations of each divalent cation. These experiments showed that UNC-8(G387E) is blocked by extracellular calcium with a Ki of 6 µM at −100 mV (Fig. 5, A–C). To establish whether calcium binds the channel within the membrane electric field, we determined the voltage dependence of Ki. By fitting data points with the Woodhull model (Woodhull, 1973), we established that the calcium-binding site in UNC-8(G387E) is 6% within the membrane electric field (Fig. 5 D). Note that Ca²⁺ blocks the outward current as well (Fig. 5 B). This result suggests either that Ca²⁺ binds within a crevice of the channel from where it cannot escape during membrane depolarization, or that the channel changes conformation once Ca²⁺ is bound, thereby trapping the ion. On the basis of these results, we conclude that Ca²⁺ block of the outward current likely accounts for the inhibition of outward current in physiological solution.

UNC-8(G387E) was blocked by magnesium with a Ki of 598 µM at −100 mV (Fig. 5, E–G). Interestingly, the voltage dependence of magnesium blockage is suggestive of a binding site deeper within the membrane electric field as compared with calcium. Indeed, by fitting data points with the Woodhull model, we obtained a δ of 0.13 (δ was 0.06 for calcium) (Fig. 5 H). Interestingly, Mg²⁺ does not block the outward current (Fig. 5 F), suggesting that Mg²⁺ is dislocated from its binding site upon depolarization. Collectively, our data show that UNC-8(G387E) is blocked by extracellular calcium and magnesium. However, the channel is more sensitive to calcium, and the two ions appear to bind to different sites within the channel: calcium binds a more external site, whereas magnesium binds to a location deeper within the pore, consistent with the smaller size of the magnesium ion.

We next wondered whether UNC-8(G387E) high sensitivity to extracellular divalent cations is a general feature of UNC-8 or if it is related to this specific mutation. To discriminate between these two possibilities, we expressed a second UNC-8 mutant protein, UNC-8(A586T), in Xenopus oocytes (Tavernarakis et al., 1997). The A586T missense mutation is located in the extracellular region (Fig. S2). A586T is encoded by the e49 allele, which also results in uncoordinated movement (Shreffler et al., 1995). Our measurements showed that UNC-8(A586T) produces outwardly rectifying currents that are 20 times larger in the absence of divalent cations than in physiological NaCl solution (Fig. 6, A–D). These results support the idea that the blockage by extracellular divalent cations is likely an intrinsic feature of UNC-8 channels.

An intragenic suppressor mutation was isolated in a genetic screen for mutations that influence unc-8(d)–induced neuronal swelling (Shreffler et al., 1995). H114Y (e15lb129 allele) suppresses the Unc-8(d) neuronal swelling and uncoordinated phenotypes in either the cis or trans configuration to the G387E mutation. This result is consistent with a model in which the H114Y mutation disrupts either assembly or function of UNC-8 in a multimeric channel. H114 corresponds to a highly conserved residue located in the intracellular N-terminal region adjacent to the first transmembrane domain (Fig. S2, A and C). Based on our results, we reasoned that H114Y may reduce neuronal damage by suppressing UNC-8(G387E) currents. To test this idea, we coexpressed UNC-8(H114Y) with UNC-8(G387E) in Xenopus oocytes and analyzed current amplitudes. We found that UNC-8(G387E) plus UNC-8(H114Y) currents were smaller than that of UNC-8(G387E) alone (Fig. 7, A–C). In fact, UNC-8(G387E) currents were suppressed ∼85% by coexpression with UNC-8(H114Y). Consistent with a smaller current amplitude, the membrane potential of these oocytes on average depolarized to a lower potential upon switching to a divalent cation–free plus EGTA solution as compared with oocytes injected with UNC-8(G387E) alone (compare Fig. 7 D with Fig. 3 E). To test the effect of H114Y mutation within the same UNC-8 subunit (in cis), we built the UNC-8(G387E/H114Y) double mutant. We found that these channels were not functional in oocytes (Fig. 7 C). Notably, UNC-8(H114Y) did not produce currents when expressed alone, suggesting that this mutant subunit is not hyperactivated (Fig. 7 C). Interestingly, wild-type UNC-8 also suppressed UNC-8(G387E) currents but by a lower degree, which is consistent with the lower neurotoxicity observed in unc-8(n491)/+ heterozygotes (Shreffler and Wolinsky, 1997). Current suppression by UNC-8(H114Y) could be the result of effects on channel function or trafficking to the cell surface. In any case, our data are consistent with the idea that the level of neurotoxicity of hyperactivated UNC-8 is defined by the amplitude of UNC-8 mutant ionic current.

Allele n49, which encodes the A586T mutation, causes a less severe uncoordinated phenotype than the n491 and e15 alleles that encode G387E mutation (Shreffler et al., 1995). This difference is caused by a lower number of swollen neurons in n49 versus n491 and e15 mutants (Shreffler et al., 1995). Based on our results, we hypothesized that in vivo cationic currents through UNC-8(A586T) channels should be smaller than currents through UNC-8(G387E). We thus compared current amplitudes in oocytes expressing UNC-8(A586T) and UNC-8(G387E), perfused with a divalent cation–free plus EGTA solution. However, we found no statistical difference (10.1 ± 1.8 and 13.3 ± 3.1 µA for A586T and G387E, respectively, at −100 mV, expressed as mean ± SD; n = 9 and 15). We next tested whether calcium sensitivity was different in the two mutants. Indeed, because calcium inhibits UNC-8(d) currents, a different sensitivity to this divalent cation could manifest in different current amplitude in vivo. We found that the Ki for calcium was 1.66 µM at −100 mV, which is lower than the Ki we determined for G387E mutant (6 µM at −100 mV).

We next wondered whether UNC-8(G387E) and UNC-8(A586T) currents were sensitive to pH and if a difference in pH sensitivity could contribute to the difference in the in vivo phenotypes. We analyzed UNC-8(G387E) and UNC-8(A586T) pH sensitivity in Xenopus oocytes perfused with solutions with a pH ranging from 5 to 8. We found that both currents were inhibited by acidic pH. However, although UNC-8(G387E) pH_0.5 was 5.6, UNC-8(A586T) pH_0.5 was 6.6 (Fig. 8). These data suggest that a smaller fraction of UNC-8(G387E) channels than UNC-8(A586T) channels is inhibited at physiological pH. This effect could account for the observation that unc-8(n491) and unc-8(e15) corresponding to the G387E mutation result in more neuronal swelling than the unc-8(n49) allele that encodes UNC-8(A586T). To conclude, our results show that UNC-8(d) is inhibited by acidic pH and suggest that differences both in calcium and pH sensitivity between the G387E and A586T mutant proteins could contribute to their different toxicities in vivo.

We undertook this study to better characterize neurotoxic mutants of the C. elegans DEG/ENaC channel UNC-8 and to establish whether a correlation exists between their level of toxicity and channel properties. By expression in Xenopus oocytes, we show that neurotoxic UNC-8(d) mutations induce channel hyperactivation. The strong block exerted by extracellular divalent cations on UNC-8(d) channels diminishes current amplitude and could therefore account for limiting the cellular toxicity of UNC-8(d) channels. Indeed, reduction of extracellular divalent cations concentration resulted in swelling and death of Xenopus oocytes expressing UNC-8(d) channels. Further evidence that the current amplitude through the mutant UNC-8 channels defines their level of toxicity is provided by the our findings that: (a) coexpression with an UNC-8 intragenic mutation that suppresses neuronal swelling reduces current amplitude, and (b) the less toxic mutation, UNC-8(A586T), is associated with higher sensitivity to calcium and acidic pH and is therefore expected to produce smaller currents in vivo. Collectively, our data are consistent with the idea that the neurotoxicity of UNC-8 gain-of-function mutants is caused by hyperactive cation transport and that this effect may be quelled in vivo by extracellular divalent cations and pH.

Hyperactivated DEG/ENaC channels can induce swelling and death of the cells in which they are expressed (Driscoll and Chalfie, 1991; Huang and Chalfie, 1994; Xiong et al., 2004). For example, hyperactivation of mouse ASIC1a by prolonged stroke-induced tissue acidosis contributes to neuronal demise (Xiong et al., 2004). Similarly, a mutation in the pore domain hyperactivates the C. elegans MEC-4 channel and results in the swelling and death of the touch neurons in which MEC-4 is expressed (Driscoll and Chalfie, 1991; Goodman et al., 2002). Excess ion influx caused by hyperactivation of DEG/ENaCs is thought to result in an ionic imbalance that is ultimately toxic to the cell. MEC-4(d) and ASIC1a have been shown to be calcium permeable, suggesting that at least some neurotoxic DEG/ENaC channels contribute directly to elevation of intracellular calcium and/or trigger calcium release from the ER (Bianchi et al., 2004; Xiong et al., 2004). Blockage of hyperactivated DEG/ENaC channels by the application of amiloride or PcTx1 toxin, an ASIC1-specific inhibitor, significantly improves neuronal survival, supporting the idea that ion influx through these channels is the trigger for neuronal death (Bianchi et al., 2004; Xiong et al., 2004). However, to what extent the amount of Na⁺ and/or Ca²⁺ influx is toxic and what intrinsic channel properties define its level of toxicity remained unclear.

We report here the functional characterization of neurotoxic mutations in C. elegans DEG/ENaC channel UNC-8. Neurotoxic mutations in UNC-8 are unique in that they cause the death of only a fraction of the neurons in which UNC-8 is expressed (Shreffler et al., 1995) (Fig. 1 E). Indeed, DA/DB motor neurons swell during the L1 larval stage, but most (∼65%) recover and survive into adulthood. This result suggests that the neurotoxicity of the UNC-8 mutants is lower than that of other DEG/ENaC channels including MEC-4(d), DEG-1(d), and ASIC1a (Driscoll and Chalfie, 1991; Huang and Chalfie, 1994; Xiong et al., 2004), and that it may be regulated in a cell-specific manner. We found that UNC-8 subunits form homomeric channels in Xenopus oocytes and that neurotoxic mutations G387E and A586T cause channel hyperactivation. However, ionic currents are small because of the block exerted by extracellular divalent cations calcium and magnesium. Indeed, with a Ki for calcium of <10 µM, the UNC-8(d) channels G387E and A586T are blocked at physiological calcium concentrations, at least at negative potentials. Perhaps this effect contributes to the relatively low toxicity of UNC-8(d) channels in vivo. On the other hand, the neuronal swelling of UNC-8(d) mutants also suggests that endogenous factors and/or other subunits may at least partially relieve the calcium block of UNC-8(d) in vivo. When calcium and magnesium are removed, UNC-8(d)–dependent currents in Xenopus oocytes are significantly enhanced, and this effect is correlated with oocyte toxicity. Thus, our data support the idea that the current amplitude through neurotoxic DEG/ENaC channels specifies the level of toxicity.

Interestingly, UNC-8 gain-of-function mutants are calcium impermeable, suggesting that excess Na⁺ influx is sufficient to induce the toxic insult. Elevated Na⁺ influx could induce reversal of the Na⁺/Ca²⁺ exchanger and/or membrane depolarization, resulting in activation of voltage-gated calcium channels. Both of these events are expected to elevate intracellular Ca²⁺. Importantly, block by extracellular divalent cations was observed in both UNC-8(G387E) and UNC-8(A586T) mutants, suggesting that the high sensitivity to extracellular divalent cations is a property of UNC-8 channels and that it is not introduced by either mutation per se.

Our data show that calcium acts as an open channel blocker but does not permeate through UNC-8 gain-of-function mutant channels. In contrast, calcium acts as a permeant blocker on MEC-4(d) (Bianchi et al., 2004; Brown et al., 2008). Interestingly, although the Ki for calcium block in UNC-8(G387E) is in the micromolar range (6 µM at −100 mV), it is in the millimolar range for the MEC-4(d)–MEC-10(d) channel complex (Brown et al., 2008). Moreover, although calcium binds very close to the external mouth of the pore in UNC-8(G387E), it binds to a site deeper in the pore in the MEC-4(d)–MEC-10(d) channel complex (Brown et al., 2008). Other members of the DEG/ENaC family were shown either to be blocked by calcium or to be calcium permeable. For example, calcium-impermeable ASIC3 is blocked by calcium in the nanomolar range, and low pH releases the calcium block (Immke and McCleskey, 2003). Conversely, calcium-permeable ASIC1a is blocked by extracellular calcium with a Ki of ∼5 mM (Paukert et al., 2004). Based on these observations, we speculate that DEG/ENaC channels that have a strong affinity for calcium may not be calcium permeable and vice versa. This prediction is consistent with the fact that tightly bound ions are unlikely to leave a high affinity site to permeate through the channel pore.

Importantly, our data show that despite a strong affinity for calcium and magnesium, caused by the voltage dependence of the block, currents through gain-of-function UNC-8 mutants are still substantial at depolarized potentials. Based on the fact that two different UNC-8 gain-of-function mutations show strong affinity for extracellular divalent cations, we suggest that divalent cation sensitivity is likely a property intrinsic of the UNC-8 channel. Thus, we speculate that wild-type UNC-8 is similarly blocked by extracellular divalent cations. If this is the case, currents through UNC-8 may be regulated by voltage in an indirect way. For example, membrane depolarization associated with neuronal and/or synaptic activity would release calcium and magnesium block, allowing flow of ions through UNC-8. This mechanism could contribute to the recently discovered role for UNC-8 in an activity-dependent mechanism of synaptic remodeling (Petersen et al., 2011; Thompson-Peer et al., 2012; unpublished data).

The two gain-of function UNC-8 mutations that hyperactivate channel function and induce neuronal swelling reside in the UNC-8 extracellular domain. G387E is located in a poorly conserved loop in the “finger” domain, and A586T is positioned in a loop between α 4 and α 5 of the “thumb” domain (Jasti et al., 2007; Gessmann et al., 2010; Eastwood and Goodman, 2012) (Fig. S2 C). Previous studies have shown that the thumb domain participates significantly in proton binding (Immke and McCleskey, 2003; Paukert et al., 2004; Jasti et al., 2007; Wang and Bianchi, 2009; Yang et al., 2009), and that it undergoes substantial movement during gating (Passero et al., 2009). In ASICs, the movement of the thumb domain is translated into channel gating by the π–π interaction between a conserved extracellular tyrosine adjacent to TM1 (Y72 in ASIC1a) and a tryptophan located in a loop between β sheets 9 and 10 (W288 in ASIC1a) (Li et al., 2009). These gating residues are conserved in UNC-8 (Fig. S2 A). Our data are consistent with the idea that in UNC-8 channels, mutations that may perturb the movement of the thumb domain shift the channel into a hyperactivated mode, thereby stabilizing the channel open state, as suggested by previous studies (Hong et al., 2000; Yang et al., 2009). Thus, our data and previous published work support the idea that the thumb domain plays an important role in DEG/ENaC channel gating by diverse stimuli including mechanical forces. Intriguingly, several mec-4 mutations that result in touch insensitivity cluster around this extracellular region (Tavernarakis et al., 1997; García-Añoveros et al., 1998; Hong et al., 2000).

To conclude, we have characterized the functional properties of the neurotoxic UNC-8 DEG/ENaC protein. Our findings are consistent with the proposal that strong blockage exerted by extracellular divalent cations inhibits UNC-8 channel activity in vivo and that this effect limits the toxicity of UNC-8(d) channels in vivo. We suggest that the extent of divalent cation blockage, in addition to calcium permeability, is a DEG/ENaC channel property that sets the level of toxicity in vivo.

We thank Gerhard Dahl and Charles W. Luetje for providing Xenopus oocytes and Jeff Grant for critical reading of the manuscript. Some of the strains used in this work were provided by the C. elegans Genetics Center, which is supported by the US National Institutes of Health (NIH) National Center for Research Resources.

This work was supported by American Cancer Society RGS-09-043-01-DDC and NIH RO1NS070969 grants (to L. Bianchi), NIH RO1NS26115 and R21MH077302 grants (to D.M. Miller), and by Vanderbilt Silvio O. Conte Center grant MH078028 (to T. Miller).

Lawrence G. Palmer served as editor.