Acid-sensing ion channels ASIC1a and ASIC1b are ligand-gated ion channels that are activated by H+ in the physiological range of pH. The apparent affinity for H+ of ASIC1a and 1b is modulated by extracellular Ca2+ through a competition between Ca2+ and H+. Here we show that, in addition to modulating the apparent H+ affinity, Ca2+ blocks ASIC1a in the open state (IC50 ∼ 3.9 mM at pH 5.5), whereas ASIC1b is blocked with reduced affinity (IC50 > 10 mM at pH 4.7). Moreover, we report the identification of the site that mediates this open channel block by Ca2+. ASICs have two transmembrane domains. The second transmembrane domain M2 has been shown to form the ion pore of the related epithelial Na+ channel. Conserved topology and high homology in M2 suggests that M2 forms the ion pore also of ASICs. Combined substitution of an aspartate and a glutamate residue at the beginning of M2 completely abolished block by Ca2+ of ASIC1a, showing that these two amino acids (E425 and D432) are crucial for Ca2+ block. It has previously been suggested that relief of Ca2+ block opens ASIC3 channels. However, substitutions of E425 or D432 individually or in combination did not open channels constitutively and did not abolish gating by H+ and modulation of H+ affinity by Ca2+. These results show that channel block by Ca2+ and H+ gating are not intrinsically linked.
Introduction
Acid-sensing ion channels (ASICs) belong to the DEG/ENaC superfamily of ion channels (Waldmann and Lazdunski, 1998). Members of this supergene family form Na+-selective ion channels (PNa/PK, 8–100), which can be blocked by amiloride (IC50, 0.2–20 μM) (Alvarez de la Rosa et al., 2000; Kellenberger and Schild, 2002).
Members of the DEG/ENaC family likely have a similar overall structure. Amino acids of members of the ASIC subfamily are >45% identical to each other and ∼20–25% identical to the subunits of the epithelial sodium channel (ENaC). The primary structure of all family members shows some common hallmarks, including two hydrophobic domains, short NH2- and COOH termini, and a large loop containing conserved cysteines between the hydrophobic domains. A topology with two transmembrane-spanning domains and intracellular termini has been experimentally verified for ENaC subunits (Canessa et al., 1994; Renard et al., 1994; Snyder et al., 1994). According to the current model of the pore structure, amino acids preceding the second transmembrane (M2) domain form the binding site for amiloride (Schild et al., 1997; Sheng et al., 2000), and amino acids within the M2 domain form the selectivity filter of ENaC (Kellenberger et al., 1999a,b; Snyder et al., 1999; Sheng et al., 2000, 2001). The preM2 and the M2 domain of ASICs are almost completely conserved between different ASIC subunits and show high homology to ENaC subunits, suggesting that they contribute to the amiloride binding site and the selectivity filter of ASICs. There are, though, some differences in pore properties of ENaC and ASICs. ENaC is virtually impermeable to K+ (PNa/PK ∼ 100) and Ca2+, and is blocked only by unphysiologically high concentrations of divalent cations (IC50 > 10 mM) (Schild et al., 1997). In contrast, ASICs discriminate less between Na+ and K+ (PNa/PK ≤ 15), and ASIC1a shows a low but significant Ca2+ permeability (PNa/PCa ∼ 15) (Bässler et al., 2001).
ASICs are activated by a rise in the extracellular concentration of H+. Proposed functions of ASICs include modulation of neuronal activity by extracellular pH (Waldmann and Lazdunski, 1998), peripheral perception of pain (Sutherland et al., 2001; Chen et al., 2002), and perception of taste (Ugawa et al., 2003). So far, six different members of this subfamily have been cloned (ASIC1a, ASIC1b, ASIC2a, ASIC2b, ASIC3, and ASIC4), which are encoded by four genes (Price et al., 1996; Waldmann et al., 1996, 1997a,b; Garcia-Anoveros et al., 1997; Akopian et al., 2000; Gründer et al., 2000).
ASIC1b is a splice variant of ASIC1a (Chen et al., 1998; Bässler et al., 2001), in which approximately the first third of the protein is exchanged, whereas the COOH-terminal two thirds are identical. ASIC1a is expressed throughout the brain and in sensory neurons of the dorsal root ganglion (Waldmann et al., 1997a), whereas ASIC1b is specifically expressed in sensory neurons (Chen et al., 1998). Both subunits form rapidly activating and completely desensitizing ion channels (τact ∼ 10 ms, τinact ∼ 1 s) (Bässler et al., 2001) that are activated by H+ in the physiological pH range (pH0.5, 5.9–6.4) (Babini et al., 2002). The molecular identity of the H+ sensor of ASICs is unknown.
We have previously shown that Ca2+ stabilizes the closed state of ASIC1 (Babini et al., 2002) and have proposed that Ca2+ binds to an extracellular site modulating the apparent affinity to H+. Recently, Immke and McCleskey (2003) extended our findings for ASIC3 and proposed an elegant model, in which Ca2+ blocks the open pore. Relief of the Ca2+ block by the competitive binding of H+ would open the channel, thereby accounting for H+ gating. Thus, this model predicts that disruption of the blocking site should constitutively open the channels. The aim of this study was to further address the relation between Ca2+ block and H+ gating of ASICs. We report here that substitution of two negatively charged amino acids at the beginning of the second transmembrane domain of ASIC1a reliefs Ca2+ block. Our results show that the relief of Ca2+ block is not sufficient to open the pore of ASIC1a.
Materials And Methods
cDNAs and Site-directed Mutagenesis
cDNAs for rat ASIC1a and ASIC1b have been described previously (Bässler et al., 2001). Point mutations were introduced by recombinant PCR using Pwo DNA polymerase (Roche). In brief, two fragments were amplified with primers containing the desired mutations in an overlapping region, joined by recombinant PCR, digested by appropriate restriction endonucleases (ASIC1a: MscI/XhoI; ASIC1b: SexAI/XhoI), and ligated into the corresponding cDNA in the vector pRSSP. All PCR-derived fragments were entirely sequenced.
Whole-oocyte Electrophysiology
Using the mMessage mMachine kit (Ambion), capped cRNA was synthesized by SP6 RNA polymerase from ASIC1a, ASIC1b, and mutant cDNA, which had been linearized by NaeI. cRNA was injected into stage V to VI oocytes of Xenopus laevis, and oocytes were kept in OR-2 medium (concentrations in mM: 82.5 NaCl, 2.5 KCl, 1.0 Na2HPO4, 5.0 HEPES, 1.0 MgCl2, 1.0 CaCl2, and 0.5 g/l PVP) for 1–5 d. We injected 0.01–0.1 ng of ASIC1a wild-type cRNA and 1–10 ng of the other cRNAs; currents ranged from 0.5 to 20 μA.
The bath solution for two-electrode voltage clamp contained (in mM): 140 NaCl, 10 HEPES; concentrations of divalent cations (BaCl2, MgCl2, or CaCl2) were as indicated in the figure legends. Since H+ affinity of ASIC1 is modulated by extracellular Ca2+ (Babini et al., 2002), we kept the Ca2+ concentration always constant (1.8 mM) between low-pH activation and changed it only during low pH activation. Therefore, Ca2+ may not completely reach a steady-state equilibrium at its binding site during low pH activation. This may slightly affect the shape of the blocking curve and the IC50. For measurements determining permeability to divalent cations, bath solution contained (in mM): 50 BaCl2, 10 HEPES. pH was adjusted using NaOH or Ba(OH)2; HEPES was replaced by MES buffer where appropriate. Holding potential was −60 mV or −70 mV.
Ca2+ blocking curves and H+ dose–response curves were registered using an automated, pump-driven solution exchange system together with the oocyte testing carousel controlled by the interface OTC-20 (npi electronic GmbH). With this system, 90% of the solution surrounding an oocyte can be exchanged within ∼100 ms. Currents were recorded with a TurboTec 03X amplifier (npi electronic GmbH) with the filter set to 20 Hz, digitized at 1 kHz using the AD/DA interface PCI 1200 (National Instruments), and stored on hard disk. Data acquisition and solution exchange were managed using the software CellWorks 5.1.1 (npi electronic GmbH). Steady-state inactivation curves were registered using gravity-driven solution exchange. Currents were recorded with a TurboTec 01C amplifier (npi electronic GmbH), data stored on hard disk, and analyzed using IgorPro software (WaveMetrics). Amiloride, EDTA, niflumic acid, and flufenamic acid were from Sigma-Aldrich.
Outside-out Patch-clamp Measurements
Oocytes injected with 1–10 ng of ASIC1a wild type or ASIC1aE425GD432C cRNA were used. Following shrinkage of oocytes in a hypertonic solution (300 mM K aspartate), the vitelline membrane was manually removed using forceps. The oocytes were allowed to recover from shrinkage for several minutes in the recording chamber containing the following solution (in mM): 100 KCl, 2 MgCl2, 10 EGTA, 10 HEPES; pH was adjusted to 7.2 with KOH. This solution was also used for seal and outside-out patch formation. Thick-walled borosilicate glass capillaries (Science Products GmbH) were used to pull pipettes with a resistance of 6–10 MΩ. Patch pipettes were backfilled with a solution containing (in mM): 140 KCl, 2 MgCl2, 5 EGTA, 10 HEPES; pH was adjusted to 7.4 with KOH. Following outside-out patch formation, the patch pipette was placed in front of a piezo-driven double-barreled application pipette enabling fast solution exchange (Bässler et al., 2001). Gravity-driven test and control solution flowing out of the application pipette contained (in mM): 140 NaCl, 10 HEPES; concentration of divalent cations and pH (adjusted with NaOH) were as described in results. Patches were clamped to −100 mV. Data were acquired using an Axopatch 200A amplifier (Axon Instruments), filtered with the built-in Bessel filter at 1 kHz, digitized at 10 kHz, and stored on hard disk. All experiments were conducted at room temperature.
Data Analysis
ASIC1a currents were characterized by a rundown in whole oocyte experiments (current amplitude of the eighth low-pH application ∼60% of the first application); this current rundown can be appreciated in Fig. 6 B, right panel. To compensate this rundown, we always analyzed an equal number of experiments in which we measured the current amplitude with increasing and with decreasing Ca2+ concentrations. In addition, current values were normalized to currents that were obtained under identical experimental conditions but with identical Ca2+ concentration (1.3 mM) during channel activation. Current values were then normalized to the current measured with 0.1 mM Ca2+ and data were fitted for each experiment using the following equation:
where Imax is the maximal current, a is the residual current, which is insensitive to Ca2+ block, [B] is the concentration of Ca2+, IC50 is the concentration at which half-maximal block occurs, and n is the Hill coefficient. This type of analysis allowed a good qualitative comparison of wild-type and mutant channels. For their quantitative comparison, we focused our analysis on the residual current with 10 mM Ca2+, expressed as a fraction of the current measured with 0.1 mM Ca2+. Our determination of the apparent binding constant (IC50) of Ca2+ neglects possible contributions of surface charge and competition with Na+ and H+ ions to the interaction of Ca2+ ions with the channel protein. While such interactions likely affect the precise value of the apparent binding constant, they do not alter the fundamental conclusions of this paper that are based on the large difference of the IC50 of wild type and mutants.
For determination of competition between Ca2+ and amiloride, channels were activated every 30 s with a solution of pH 5.5 that contained increasing concentrations of amiloride. Experiments were done with the acidic solutions containing either 0.1 or 10 mM Ca2+ as the only divalent cation. The current amplitudes with amiloride were normalized to and then subtracted from the amplitude without amiloride to obtain the fraction of the blocked current. These data were not compensated for rundown and were fit to a Hill equation like Eq. 1.
Voltage dependence of the block was analyzed using a Boltzmann equation (Woodhull, 1973):
where Ki(V) is the voltage-dependent inhibitory constant, Ki(0) is the inhibitory constant at 0 mV, and z′ is the product of the valence z of the blocking ion and the fraction of the membrane potential δ acting on the ion. F, R, and T have their usual meanings.
Xenopus oocytes possess an endogenous Ca2+-activated Cl− channel in their membrane. To exclude a contamination of the measured current by a Cl− current, which could have been activated by Ca2+ flowing through ASIC1a, we compared the amplitude of H+-activated currents in the presence of either Mg2+ or Ca2+ in the extracellular solution. We used concentrations of Mg2+ and Ca2+ that produced comparable block of ASIC1a (5 mM and 1.8 mM, respectively). These currents were of comparable amplitude (not depicted), demonstrating that the Ca2+ permeability of ASIC1a is too low to efficiently activate the Ca2+-activated Cl− channel of Xenopus oocytes. However, we did not systematically investigate activation of the Cl− channel by ASIC1a. Therefore, we cannot exclude a small contribution of a Cl− current to the total current at high extracellular Ca2+ concentrations or at potentials more negative than −70 mV.
Reversal potentials with extracellular Na+ were determined by running a fast (200 ms) voltage ramp from −70 to +40 mV. Currents were leak subtracted by running the ramp before and during low pH (5.0) activation. Only oocytes with current amplitudes <2 μA were analyzed to avoid clamp artifacts. In whole cell measurements with BaCl2 in the bath, we determined the current amplitude in 5-mV steps between −55 mV and −75 mV. Reversal potential was then calculated using a linear fit between the current values between which current reversed its sign.
Single-channel recordings were analyzed with the software Ana (available at http://www.ge.cnr.it/ICB/conti_moran_pusch/programs-pusch/software-mik.htm). Segments of channel openings from individual recordings under the same condition were pooled and contributed to the respective amplitude histogram. The amplitude distribution was fitted to a sum of Gaussian functions.
Results are reported as means ± SEM or, for the amplitude histograms, as means ± σ. They represent the mean of n individual measurements on different oocytes or different patches. For each condition, oocytes from at least two different frogs and patches from at least two different oocytes were analyzed. Statistical analysis was done with the unpaired t test.
Results
Block of ASIC1 by Divalent Cations
We investigated the block by Ca2+ of ASIC1a and 1b expressed in Xenopus oocytes. Since Ca2+ shifts the steady-state inactivation as well as the pH activation curve (Babini et al., 2002), we held the concentration of Ca2+ constant (1.8 mM) between activation steps and used a saturating pH for activation that leads to full activation at all Ca2+ concentrations used (ASIC1a: pH 5.5; ASIC1b: pH 4.7). This acidic application solution contained varying Ca2+ concentrations. As is shown in Fig. 1 (A and C), increasing Ca2+ concentrations during channel activation reduced the current amplitude of ASIC1a progressively. 10 mM Ca2+ blocked 60 ± 3% of the ASIC1a current (n = 16; Fig. 1, A and C). This reduction of the current amplitude cannot be explained by incomplete recovery. With low concentrations of divalent cations (0.3 mM), recovery of ASIC1a is complete in ∼1 min (Babini et al., 2002). Recovery with 1.8 mM Ca2+ as used in our experiments will be significantly faster (Babini et al., 2002). Since we used 1 min between activation steps, all channels should have recovered from inactivation during this time. In addition, as mentioned above, at pH 5.5, the reduction by Ca2+ of the current amplitude cannot be due to a shift in the dose–effect curve. Hence, most likely, Ca2+ blocked ASIC1a in the open conformation and is, therefore, a permeant blocker of ASIC1a. Fit to a single-site Hill equation (Eq. 1) of the Ca2+ dose–response curve of ASIC1a revealed a half-maximal block (IC50) at 3.9 ± 1.0 mM with a Hill coefficient of 1.3 (n = 16; Fig. 1, A and C), consistent with the idea that one Ca2+ ion blocks ASIC1a channels. ASIC1a currents were also blocked by Ba2+ and Mg2+ (affinity for block by Ba2+ was similar to block by Ca2+, and the affinity for block by Mg2+ was approximately threefold lower; not depicted) but only slightly by the divalent cations Zn2+ and Cd2+ (40 ± 2% of the current blocked by 10 mM Zn2+, n = 3, and 14 ± 2% of the current blocked by 3 mM Cd2+, n = 4). In contrast to ASIC1a, current amplitude of ASIC1b was only slightly reduced by Ca2+ (37 ± 5% of the current blocked by 10 mM Ca2+, n = 8; Fig. 1, B and C).
In addition to the reduction of the current amplitude, Ca2+ also influenced the desensitization of ASIC1a. The time constant τ of desensitization significantly decreased from τ = 2.2 ± 0.4 s with 0.1 mM Ca2+ to τ = 1.3 ± 0.1 s with 10 mM Ca2+ (n = 9, P < 0.05). Desensitization of ASIC1b was faster than desensitization of ASIC1a with 0.1 mM Ca2+ (τ = 0.7 ± 0.1 s), but not significantly accelerated by 10 mM Ca2+ (τ = 0.9 ± 0.2 s; n = 8, P = 0.1).
Ca2+ Reduces the Apparent Single Channel Amplitude of ASIC1a
We further confirmed that the observed inhibitory effect of Ca2+ on ASIC1a is due to an open channel block using single channel analysis. A low-affinity block of an ion channel pore is often associated with a flickering between the blocked and unblocked state that is faster than can be resolved by the patch-clamp technique. Therefore such a flickering block becomes apparent as a reduced single channel amplitude. A flickering block by Ca2+ has already been shown for ASIC3 (Immke and McCleskey, 2003). To open single channels in an outside-out patch from Xenopus oocytes, we switched from a control solution of pH 7.4 containing 1.8 mM Ca2+ and 1.0 mM Mg2+ to a test solution of pH 7.05 containing 1.8 mM Ca2+ or to a solution of pH 7.15 containing 0.1 mM Ca2+. These test pH values are in the range where the steady-state inactivation curve and the pH activation curve of ASIC1a overlap (Babini et al., 2002; Fig. 6). Hence a small steady-state current can be expected at these pH values. Indeed, under these conditions, single channel openings could often be detected also several seconds after the initial channel desensitization. Fig. 2 A shows, on the left, an example trace and, on the right, the amplitude histogram that was generated from several recordings obtained from outside-out patches of ASIC1a-expressing oocytes in the presence of 1.8 mM extracellular Ca2+. The histogram revealed a single channel amplitude of −1.3 ± 0.4 pA for ASIC1a wild type (n = 4 individual patches). Occasionally, very short current events (shorter than 2 ms) with an amplitude around 4 pA were observed that, due to their short duration, did not appear in the histogram. Frequency and duration of channel openings were variable between recordings from individual patches, not allowing the analysis of the mean open time. In contrast to the measurements with 1.8 mM Ca2+, in the presence of 0.1 mM extracellular Ca2+, single ASIC1a channels had an amplitude of −4.9 ± 0.9 pA (Fig. 2 B; n = 3 individual patches). The reduced amplitude of single ASIC1a channels with 1.8 mM Ca2+ compared with 0.1 mM Ca2+ confirms that Ca2+ is an open channel blocker of ASIC1a.
A reduction of the single channel amplitude from −4.9 pA with 0.1 mM Ca2+ to −1.3 pA with 1.8 mM Ca2+ predicts an apparent Ca2+ affinity of the blocking site that is considerably higher than 3.9 mM, the value obtained by the voltage clamp analysis on whole oocytes. We explain this apparent discrepancy by a pH-dependence of the Ca2+ affinity of the blocking site. As shown below, we identified aspartate and glutamate residues as crucial components of the Ca2+ blocking site. We propose that neutralization of the side chains of these amino acids by the low pH used for activation of ASIC1a will reduce the apparent affinity to Ca2+. Since we used lower pH for the whole cell experiments than for the patch clamp experiments, the affinity to Ca2+ will appear lower in the whole cell experiments than in the patch clamp experiments (see discussion).
Voltage Dependence of Ca2+ Block
Next, we determined the voltage dependence of Ca2+ block. Block of ASIC1a was decreased at hyperpolarized potentials with Ca2+ concentrations ranging from 1 to 10 mM (Fig. 3 A). Since ASIC1a is not only blocked by Ca2+ but also permeable for Ca2+ (Bässler et al., 2001), this most likely reflects a voltage dependence of the Ca2+ permeation. For the Ca2+-impermeable ASIC1b, the voltage dependence of Ca2+ block was investigated only for a Ca2+ concentration of 10 mM, which significantly blocked ASIC1b (Fig. 1, B and C). In contrast to ASIC1a, block of ASIC1b was slightly increased at hyperpolarized potentials (Fig. 3 B). Only at potentials more negative than −100 mV was a decreased block observed, which is most likely due to some unspecific Ca2+ influx at this strong hyperpolarization with the high Ca2+ concentration. By fitting the voltage dependence between 0 mV and −100 mV to a Boltzmann function (Eq. 2), we estimated the fraction of the transmembrane electric field sensed by the blocking Ca2+ ion to be ∼5% (δ = 0.054 ± 0.007, n = 9). Together, our results show that Ca2+ block of ASIC1 does not strongly depend on voltage, suggesting a binding site outside the deep parts of the membrane-spanning regions of the channel and therefore most likely outside the narrow part of the ion pore.
Identification of the Ca2+ Binding Site
Since the second transmembrane domain M2 is supposed to line the ion pore, we considered amino acids preceding M2 or at the beginning of M2 as candidate blocking sites. M2 of ASIC1a was predicted to begin at isoleucine 433 (I433) (TMpred at http://www.ch.embnet.org/). This prediction is in good agreement with experimental data for other ion channels of the DEG/ENaC gene family (see discussion). We therefore considered glutamate 425 (E425), aspartate 432 (D432), and glutamine 436 (Q436) as possible binding sites for Ca2+ (Fig. 4 A). We replaced E425 by a glycine (E425G), an amino acid that is found in several other members of the supergene family, D432 by a cysteine (D432C), and Q436 by an asparagine (Q436N), a conservative substitution. For mutant Q436N, 59 ± 3% of the current was blocked by 10 mM Ca2+, similar to ASIC1a wild type (IC50 = 2.4 ± 1.0 mM, n = 12; Fig. 4 B). Mutant channels containing either substitution E425G or D432C, however, showed a significant reduction of Ca2+ block (P < 0.005 for 10 mM Ca2+). Only ∼41 ± 6% of the currents through channels containing the E425G substitution were blocked by 10 mM Ca2+, and only ∼11 ± 2% of the currents through channels containing the substitution D432C (Fig. 4 B). The apparent affinity of Ca2+ block was slightly lower for mutant E425G than for the wild type (IC50 = 5.1 ± 2.4 mM, n = 12). For mutant D432C, the apparent Ca2+ affinity could not be determined with fidelity since the block by Ca2+ of this mutant was too weak. Since no individual mutation completely relieved Ca2+ block of ASIC1a, we also engineered a channel that contained both substitutions together. Channels containing the double substitution of E425 and D432 (ASIC1aE425GD432C) were not at all blocked by Ca2+ up to 10 mM (n = 14–16, Fig. 4 B), suggesting that both amino acids contribute to the blocking site for Ca2+.
Since substitution of D432 had the strongest effect on Ca2+ block, we additionally replaced D432 by an alanine (mutant D432A), an asparagine (D432N), a glutamate (D432E), a valine (D432V), a leucine (D432L), and a lysine (D432K). Mutant D432A was only weakly blocked by Ca2+ (7 ± 3% of the current blocked by 10 mM Ca2+, n = 14), similar to D432C. Mutants D432N and D432E showed rather low current amplitude and were, therefore, not analyzed in detail. Substitutions D432V, D432L, and D432K could not be activated by low pH. They also did not increase the background current, suggesting that these channels were not constitutively active (see below). The reason for the nonfunctionality of these mutant channels is not known. Also, since the functional mutants showed significantly reduced current amplitude compared with the wild type (∼100-fold higher cRNA concentrations induced a similar current amplitude), it may be due to impaired surface expression. Alternatively, nonfunctionality of these mutants may be due to a strong constriction of the ion permeation pathway or to some effect on channel gating. Finally, substitution of the corresponding aspartate in ASIC1b (D465N) also further reduced Ca2+ block in ASIC1b (unpublished data). Due to the low expression of channels mutated at position 432, we were unable to determine the Ca2+ dependence of the apparent single channel amplitude for the double mutant ASIC1a E425GD432C.
Competition Between Ca2+ and Amiloride in ASIC1a Wild Type but Not In ASIC1a E425GD432C
We next assessed if there is a competition between the open channel blocker amiloride and Ca2+ and if this competition is influenced by the E425GD432C mutation. In ASIC1a wild-type channels, Ca2+ reduced the apparent affinity for amiloride. In the presence of 0.1 mM Ca2+, amiloride blocked ASIC1a with an IC50 of 11.6 ± 2.4 μM (n = 15), while in the presence of 10 mM Ca2+, the IC50 was slightly but significantly increased to 33.3 ± 10.0 μM (n = 15, P < 0.05; Fig. 5 A). This result is consistent with the idea of a competition between Ca2+ and amiloride at the entrance to the ion pore. In contrast, in the E425GD432C substitution, amiloride block was hardly affected by Ca2+ (IC50 = 17.2 ± 3.7 μM with 0.1 mM Ca2+ and IC50 = 18.4 ± 3.9 μM with 10 mM Ca2+, n = 14, P = 0.99). Thus, in the ASIC1a E425GD432C mutant channel, a competition between Ca2+ and amiloride was no longer observed, lending additional support to the interpretation that E425 and D432 contribute to a binding site for Ca2+ at the outer entrance to the ion pore.
Substitution of D432 and E425 Does Not Prevent Modulation of H+ Affinity by Ca2+
The basic gating characteristics of ASICs were preserved in the ASIC1a substitutions E425G, D432C, and the double-substitution E425GD432C. First, channels containing these substitutions were not constitutively active (see also below) and were activated by extracellular acidification. As can be seen from Figs. 4 and 6, a rise in the extracellular H+ concentration induced a transient inward current, similar to ASIC1a wild type. Second, steady-state inactivation and pH activation curves were still modulated by Ca2+. As can be seen from Table I and Fig. 6, for all three mutants, the apparent H+ affinity was decreased by increasing Ca2+ concentrations. This decrease was in all cases significant (P < 0.05), except for the activation curves of D432C (P = 0.09). However, also for this substitution, the shift by Ca2+ of the H+ sensitivity of steady-state inactivation was highly significant (P < 0.001).
However, there were also notable changes in the gating properties of these substitutions. First, the apparent H+ affinity of the substituted variants was more acidic than of the wild-type channel (Table I, P < 0.05 for pH activation). Second, the Hill coefficient of the activation curves was smaller (1.9 ± 0.1 for E425GD432C compared with 2.9 ± 0.3 for the wild type with physiologic concentrations of divalent cations, P < 0.005), possibly indicating the loss of one or more H+ binding sites. And finally, the Ca2+-induced shift of the curves was smaller than for the wild-type channel (Table I), possibly indicating the loss of a Ca2+ binding site.
Other Pore Properties of Substitutions at D432 or E425
We also addressed the question whether substitutions at D432 or E425 influenced other pore properties. First, we investigated ion selectivity by running a voltage ramp from −70 mV to +40 mV. The extracellular solution contained 40 mM NaCl and 100 mM NMDGCl in order to shift the Na+ equilibrium potential to less positive values and to avoid activation of the large depolarization-induced Na+ conductance that is endogenous to oocytes (Baud et al., 1982). Reversal potentials were significantly less positive for both D432C and E425G than for the wild type (Erev, 22.1 ± 4.5 mV for ASIC1a wild type, 7.0 ± 6.7 for D432C, and 15.1 ± 7.6 for E425G, n = 10–17; P < 0.05), indicating a 1.5–2-fold reduction of the relative permeability PNa/PK. Next, we addressed if the apparent loss of the Ca2+ block by substitutions of D432 and E425 could be due to an increased permeability to Ca2+. Since any strong increase in the Ca2+ permeability should be accompanied by an increase in the Ba2+ permeability, we assessed the Ca2+ permeability of the mutants indirectly using 50 mM Ba2+ as the only cation in the bath solution. Under these conditions, ASIC1a currents reversed at a potential of −66.3 mV ± 4.6 mV (n = 6), as has previously been reported (Bässler et al., 2001). Since mutants ASIC1aD432C and ASIC1aE425G were characterized by reduced current amplitude, only a few measurements could be analyzed using Ba2+ as the main charge carrier. In these measurements, the reversal potential was slightly more negative than for the wild type (−70.9 mV ± 0.3 mV, n = 2, and −78.8 mV ± 4.3 mV, n = 4, respectively), indicating that the permeability to divalent cations was not increased by these substitutions.
Channels Containing the E425 and the D432 Substitutions Can Be Opened by Removal of Extracellular Ca2+
Recently, it has been shown that complete removal of extracellular Ca2+ opens ASIC3 channels (Immke and McCleskey, 2003). To address whether complete removal of extracellular Ca2+ also opens ASIC1a channels, we superfused oocytes with a solution that was nominally free of Ca2+ and in which any residual Ca2+ was chelated by the addition of EDTA. Superfusion of the oocytes with this solution indeed opened ASIC1a channels as can be seen in Fig. 7 from the appearance of a small current that was sensitive to block by amiloride (1 mM). The small current amplitude after removal of Ca2+ compared with the amplitude after activation with H+ has also been observed for ASIC3 (Immke and McCleskey, 2003) and has been attributed to the existence of two open states, a protonated one that makes large current and a deprotonated one that makes small current (Immke and McCleskey, 2003). The current induced by removal of Ca2+ did not desensitize at pH 8.0 but at pH 7.4 (unpublished data). This behavior is similar to the behavior described for ASIC3 after removal of Ca2+. In noninjected control oocytes, removal of Ca2+ did not induce a current (unpublished data).
ASIC1a substitutions E425G, D432C, and the double substitution E425GD432C were not constitutively active, since they did not induce any amiloride-sensitive current at neutral pH in a Ca2+-containing solution (unpublished data). Since the Ca2+ blocking site had been disrupted by the double substitution, this finding indicates that removal of the Ca2+ ion that blocks the pore is not sufficient to open the channel. This reasoning predicts that another Ca2+ ion is bound to the channel and inhibits its activity. We could confirm this prediction by superfusion of oocytes expressing the double substitution E425GD432C with a Ca2+-free solution. This Ca2+-free solution also opened these mutant channels (Fig. 7). The current induced by removal of Ca2+ was amiloride sensitive (Fig. 7) and did not desensitize at pH 8.0 but at pH 7.4 (not depicted). Thus, we conclude that the mutant channels still had a Ca2+ ion bound and that binding of this Ca2+ ion did inhibit channel activity. Removal of this Ca2+ ion, which did not block the open pore, was sufficient to open ASIC1a.
Discussion
Identification of the Ca2+ Blocking Site
In this study we identify two amino acids that are crucial for the block of ASIC1a by Ca2+: E425 and D432. The evidence for this conclusion is twofold. First, combined substitution of E425 and D432 completely abolished Ca2+ block in whole cell experiments. Second, combined substitution of E425 and D432 abolished the reduction by Ca2+ of the apparent affinity for amiloride. As discussed in the next paragraph, D432 is most likely located at the entrance to the ion pore. Since this is an ideal location for a blocking site, we propose that D432 and E425 are part of a binding site for Ca2+.
Several observations support the assumption that D432 is located at the entrance to the ion pore. First, amiloride binds within the pore of ENaC close to a residue that corresponds to G438 of ASIC1a (Schild et al., 1997). Since amiloride senses between 15 and 30% of the transmembrane electric field (Palmer, 1985; McNicholas and Canessa, 1997; Fyfe and Canessa, 1998), this residue should be located within the electric field. Second, a Ca2+ binding site that had been engineered in ENaC at a residue corresponding to G435 of ASIC1a leads to a Ca2+ block in which Ca2+ senses 15% of the transmembrane electric field (Schild et al., 1997). Therefore, this residue should also be located within the electric field. Assuming an overall conserved topology of the ion pore between ASICs and ENaC and considering the slight voltage dependence that we measured for the block of ASIC1b, the mentioned results suggest that D432 is located just at the level of the beginning of the transmembrane electric field. We propose that D432 constitutes the beginning of M2. Thus, D432 would form a ring of negative charges around the entrance to the ion pore.
E425 would form a similar ring of negative charges just slightly more toward the extracellular milieu. Although we observed the strongest reduction of Ca2+ block with substitutions of D432, block was completely abolished only by the combined substitution of both D432 and E425. We propose that Ca2+ binding to either site blocks the channel. Binding to D432 more strongly blocks the channel perhaps because the pore diameter is smaller at this location and Ca2+ therefore constricts the ion pore more strongly. But the apparent difference with respect to Ca2+ block of substitutions at E425 or D432 may also be due to a different pKa of the side chain of these amino acids (see below).
Several other studies have previously reported a block of ASIC1 by Ca2+ (Waldmann et al., 1997a; de Weille and Bassilana, 2001; Zhang and Canessa, 2002). In particular, a decrease of the single channel conductance by Ca2+ was reported (de Weille and Bassilana, 2001; Zhang and Canessa, 2002).
Relation of Ca2+ Block and Channel Gating by H+
One inherent technical problem when studying Ca2+ block with a channel that is activated by H+ is that the blocking site itself is modified by H+. The pKa of the side chain of free glutamate is ∼4.3 and that of free aspartate is ∼3.9. Thus, the pH that we used to activate ASICs in the whole cell experiments (pH 5.5) is close to the pKa of these amino acids. Considering that the pKa of these side chains may well be larger in a protein where several negative charges are in close proximity, it is likely that part of the negative charges of D432 and E425 had been neutralized by the low pH used to activate ASIC1a. Thus, activation of ASICs by H+ will interfere with Ca2+ block. There are several observations that strongly support this notion. First, in our single channel experiments, unitary current amplitude was more strongly reduced by 1.8 mM Ca2+ compared with 0.1 mM Ca2+ than we expected from the whole cell experiments (by ∼75%; expected was ∼25%). There are several possible explanations for this apparently increased Ca2+ affinity in the single channel experiments compared with the whole cell experiments. However, we believe that it is most likely due to the higher pH used to activate channels in the single channel experiments compared with the pH used to activate channels in the whole cell experiments (pH 7.05 compared with pH 5.5). This also suggests that at pH 7.4, physiologic concentrations of Ca2+ may completely block ASICs (see below). Second, in the whole cell experiments, even 10 mM Ca2+ did not completely block the channel. The incomplete block also with a saturating Ca2+ concentration was not unexpected for a Ca2+-permeable channel like ASIC1a. However, since Ca2+ permeability of ASIC1a is low (Bässler et al., 2001), we expected a stronger block. And finally, the block of ASIC1b was less complete than block of ASIC1a, although the putative pore-forming region M2 is identical between both channels. Since ASIC1b is less sensitive to H+, we had to use a more acidic pH for its activation than for ASIC1a (pH 4.7 rather than pH 5.5). This will probably have neutralized more of the negative charges of the blocking sites. On the same line, in preliminary experiments (unpublished data), we did not observe a strong Ca2+ block of ASIC2a. However, the amino acids that mediate Ca2+ block in ASIC1a are conserved in ASIC2a. This apparent discrepancy can again be explained by the fact that we used pH 4.0 to activate ASIC2a. All these observations are consistent with the idea that increasing H+ concentrations decrease the apparent Ca2+ affinity of the blocking site.
Recently, Immke and McCleskey (2003) proposed a detailed model to explain the opening of the ion pore of ASIC3. Their model predicts a complete block of ASIC3 by Ca2+ at neutral pH. The block would be achieved by binding of Ca2+ to a ring of negative charges at the outer mouth of the ion pore. Titration of the negative charges by H+ would relieve the Ca2+ block, thereby opening the ASIC3 ion pore without accompanying conformational changes. This single Ca2+ binding site could explain both block of the channel as well as modulation of its apparent H+ sensitivity by Ca2+. Its simplicity makes this model very attractive. Several basic findings of Immke and McCleskey could be confirmed in the present study for ASIC1a. However, our results clearly show that a single Ca2+ binding site is not sufficient to explain the activation of ASIC1a by H+. We present strong evidence that E425 and D432 are crucial parts of the Ca2+ blocking site. Yet, disruption of this site did not constitutively open ASIC1 channels as would be expected from the Immke-McCleskey model. Moreover, ASIC1 channels without a Ca2+ blocking site retained the principal gating characteristics of ASIC1 wild-type channels: activation by H+ and modulation of the apparent H+ affinity by Ca2+. However, there is some evidence indicating that the blocking site does indeed contribute to the opening of ASIC1 wild-type channels by H+. First, the shift of the apparent H+ affinity to more acidic values of channels without the Ca2+ blocking site as well as the decreased Hill coefficient of the activation curve suggest the loss of a H+ binding site. And second, the reduced shift by Ca2+ of the H+ activation curve of these channels suggests the loss of a Ca2+ binding site. In other words, the loss of a H+ and a Ca2+ binding site at the blocking site has direct impact on the activation curve. Thus, the blocking site likely contributes to the activation of ASICs by H+ but it cannot fully account for it. As outlined below, we explain this with the presence of a second site that binds H+ and Ca2+ in a competitive manner.
We show that channels without a Ca2+ blocking site can be slightly opened by complete removal of extracellular Ca2+. This finding as well as the remaining Ca2+ modulation of the apparent H+ affinity predicts another binding site for Ca2+. We would like to call this other site the modulating site to distinguish it from the blocking site that we identified in this study. Binding of Ca2+ to the modulating site competes with the binding of H+ to the channel, just as binding of Ca2+ to the blocking site does. Thus, the apparent affinity to Ca2+ of the two sites depends on the pH. We previously estimated the apparent Ca2+ affinity of the modulating site at pH 7.4 to be ∼2 mM (Babini et al., 2002). Thus, the modulating site would be a low-affinity site. We determined in the present study the apparent Ca2+ affinity of the blocking site to also be in the low millimolar range at pH 5.5. But at pH 7.05/7.15, as used for the single channel experiments, it was already considerably higher. Immke and McCleskey estimated the apparent affinity to Ca2+ of the blocking site of ASIC3 to be in the low micromolar range at pH 7.4 (Immke and McCleskey, 2003). Thus, it seems that the blocking site is a high-affinity site.
In addition to the existence of a second Ca2+ binding site, the modulating site, our study also suggests the existence of a second H+ binding site. This site would account for the H+ gating of channels without a blocking site and is probably identical to the modulating site. We believe that binding of H+ to the modulating site will not unblock the channel but rather that it releases Ca2+ from this site, which may induce a conformational change that opens the channel. The molecular identification of the modulating site will therefore further increase our understanding of the molecular mechanism of H+ gating of ASICs and of the role of Ca2+ for gating.
Acknowledgments
We thank H.-S. Geisler and M. Siba for expert technical assistance.
This work was supported by grants of the Attempto research group program of the Universitätsklinikum Tübingen (FG 1-0-0) and the DFG (GR1771) to S. Gründer, and the Italian “Ministero dell'Istruzione, dell'Università e della Ricerca” (FIRB RBAU01PJMS) to M. Pusch.
Olaf S. Andersen served as editor.
E. Babini's present address is Istituto di Biofisica, CNR, Via de Marini 6, I-16149 Genova, Italy.
Abbreviations used in this paper: ASIC, acid-sensing ion channel; ENaC, epithelial sodium channel.