Alternative Splicing and Interaction with Di- and Polyvalent Cations Control the Dynamic Range of Acid-sensing Ion Channel 1 (ASIC1)*

Homomeric acid-sensing ion channel 1 (ASIC1) can be activated by extracellular H+ in the physiological pH range and may, therefore, contribute to neurotransmission and peripheral pain perception. ASIC1a and ASIC1b are alternative splice products of the ASIC1 gene. Here we show that both splice variants show steady-state inactivation when exposed to slightly decreased pH, limiting their operational range. Compared with ASIC1a, steady-state inactivation and pH activation of ASIC1b are shifted to more acidic values by 0.25 and 0.7 pH units, respectively, extending the dynamic range of ASIC1. Shifts of inactivation and activation are intimately linked; only two amino acids in the ectodomain, which are exchanged by alternative splicing, control both properties. Moreover, we show that extracellular, divalent cations like Ca2+ and Mg2+ as well as the polyvalent cation spermine shift the steady-state inactivation of ASIC1a and ASIC1b to more acidic values. This leads to a potentiation of the channel response and is due to a stabilization of the resting state. Our results indicate that ASIC1b is an effective sensor of transient H+ signals during slight acidosis and that, in addition to alternative splicing, interaction with di- and polyvalent cations extends the dynamic range of ASIC H+ sensors.

H ϩ -gated Na ϩ channels (acid-sensing ion channels; ASICs) 1 belong to the Deg/ENaC super family of ion channels (1). Members of this super gene family form Na ϩ -selective ion channels (P Na /P K , 8 -100) that can be blocked by amiloride (IC 50 , 0.2-20 M). All family members show some common hallmarks including two hydrophobic domains, short intracellular N and C termini, and a large extracellular loop containing conserved cysteines. Channels of this gene family probably form tetramers (2,3).
To date, six different members of the ASIC subfamily have been cloned (ASIC1a, ASIC1b, ASIC2a, ASIC2b, ASIC3, and ASIC4), which are encoded by four genes. ASIC1a and ASIC1b are alternative splice products of the ASIC1 gene (4,5). Splicing exchanges approximately the first third of the protein, including the first transmembrane domain and the proximal part of the large ectodomain. In contrast, the C-terminal twothirds are identical. We will use the term ASIC1 when we do not refer to a specific splice variant. All ASICs with the exception of ASIC4 (6) are expressed in sensory neurons of the dorsal root ganglion. Proposed functions in sensory neurons include peripheral pain perception (1,7,8) and mechanotransduction (9,10). Although some evidence suggests that some of the native H ϩ -gated currents in sensory neurons are mediated by heteromeric ASICs (11,12), part of these currents are probably mediated by homomeric ASIC1 and ASIC3 (8,11).
ASIC1a is expressed in sensory neurons and throughout the brain (13), whereas ASIC1b is specifically expressed in sensory neurons (5). Both subunits are Na ϩ -selective (P Na /P K Ϸ 10 -15), and only ASIC1a has a low Ca 2ϩ permeability (P Na /P Ca Ϸ 15) (4). ASIC1a and ASIC1b form rapidly activating and completely desensitizing ion channels ( act , ϳ10 msec; desens , ϳ1 s) (4). The expression of ASIC1 in small diameter, capsaicinsensitive sensory neurons (5,11,14,15) has led to the proposal that ASIC1a mediates excitation during tissue acidosis, which accompanies inflammation and ischemia. However, complete desensitization of ASIC1 makes it difficult to imagine how this channel can sense H ϩ signals during inflammation and ischemia when the pH persistently falls.
Here we show that both ASIC1a and ASIC1b undergo steady-state inactivation. The steady-state inactivation curve for ASIC1b is shifted by 0.25 pH units to more acidic values as compared with ASIC1a, showing that ASIC1b can operate at a more acidic resting pH. pH activation is shifted by 0.7 pH units. Differences in the sensitivity of activation and inactivation by protons are intimately linked, as both are controlled by only two amino acids in the ectodomain. These two amino acids are exchanged by alternative splicing. Moreover, we show that Ca 2ϩ , Mg 2ϩ , and spermine, when applied during the steadystate, shift the steady-state inactivation curves of both ASIC1a and ASIC1b. This leads to a potentiation of the current. Modulation by di-and polyvalent cations may be a means to adapt ASIC1 activity to changes in the extracellular concentration of these ions. Our results show that ASIC1b can replace ASIC1a as a sensor of transient proton signals during slight acidosis and that di-and polyvalent cations are modulators of ASIC activity.

EXPERIMENTAL PROCEDURES
cDNAs and Site-directed Mutagenesis-cDNAs for ASIC1a and ASIC1b have been described previously (4). Chimeric molecules between ASIC1a and ASIC1b and point mutations were constructed by recombinant PCR using Pwo DNA polymerase (Roche Molecular Biochemicals). N-terminal sequences of ASIC1b were from ASIC1b-M3 (4). All PCR-derived fragments were entirely sequenced.
The bath solution for the two-electrode voltage clamp contained 140 mM NaCl and 10 mM HEPES; concentration of divalent cations was 1.8 mM CaCl 2 and 1.0 mM MgCl 2 if not otherwise indicated. Where appropriate, HEPES was replaced by MES buffer. Holding potential was Ϫ60 or Ϫ70 mV. Solutions with low concentrations of divalent cations contained 0.1 mM flufenamic acid or niflumic acid to block the large conductance induced in oocytes by divalent-free extracellular solutions.
For steady-state inactivation curves and Ca 2ϩ dose-response curves, solution exchange was gravity driven. Whole cell currents were recorded with a TurboTec 01C amplifier (NPI Electronic Instruments, Tamm, Germany), and data were stored on hard disk. For H ϩ doseresponse curves and recovery measurements, we used an automated, pump-driven solution exchange system together with the oocyte testing carousel controlled by the interface OTC-20 (NPI Electronic Instruments). Complete solution exchange at whole oocytes with this system can be obtained within less than 1 s (16). Currents were recorded and filtered at 20 Hz with a TurboTec 03X amplifier (NPI Electronic Instruments) digitized at 100 Hz using the AD/DA interface PCI 1200 (National Instruments, Austin, Texas) and stored on hard disk. Data acquisition and solution exchange were managed using the CellWorks 5.1.1 software (NPI Electronic Instruments). Amiloride, collagenase, flufenamic acid, and niflumic acid were purchased from Sigma.
Data Analysis-Data was analyzed using IgorPro software (Wave-Metrics, Lake Oswego, OR). Steady-state inactivation, dose-response curves for potentiation by divalent cations, and activation by H ϩ were fitted using the following equation, where I max is the maximal current, a is the residual current, [B] is the concentration of H ϩ or of the divalent cation, C 50 is the concentration at which half-maximal response occurs, and n is the Hill coefficient. Time constants characterizing recovery from desensitization and steadystate inactivation were determined using a monoexponential fit. Statistical analysis was done with Student's t test.
All values reported represent the mean Ϯ S.D. from n individual measurements. Currents from individual measurements were either normalized to the I max obtained from a fit to the logistic function (1) or to the maximal value obtained (pH activation curves, recovery from desensitization). The mean of these normalized data from different measurements was then used to do the fit reported in the figures. Error bars on graphs indicate the standard error (except Fig. 2).

Splice Variants ASIC1a and ASIC1b
Have Different Dynamic Ranges-Using Xenopus oocytes, we first investigated the activation of ASIC1a and ASIC1b by protons. As shown in Fig. 1A, application of an acidic solution to ASIC1 expressing oocytes elicited transient, fast activating, and completely desensitizing inward currents. Current amplitude of ASIC1a and ASIC1b increased over a pH range of 1 unit, corresponding to a 10-fold increase in agonist concentration (Fig. 1). Thus, the sensitivity of ASIC1 to its agonist is comparable with that of other ligand-gated ion channels. ASIC1b has a lower apparent affinity for H ϩ (EC 50 , pH 5.84 Ϯ 0.09, n ϭ 8) than ASIC1a (EC 50 , pH 6.56 Ϯ 0.06, n ϭ 11). Both values for pH of halfmaximal activation are in good agreement with values obtained from measurements with smaller cells like COS-7 (5,8) showing that activation curves can be confidently determined with whole oocytes. The Hill coefficient was slightly higher for ASIC1a compared with ASIC1b (3.24 Ϯ 0.44 and 2.52 Ϯ 0.47, respectively). A Hill coefficient bigger than 1 suggests cooperative binding of H ϩ to more than one subunit. As can be seen in Fig. 1B, the shift in the activation curve for ASIC1b leads to an efficient activation approximately at the pH at which the activation of ASIC1a starts to saturate.
We then analyzed the steady-state inactivation of ASIC1a and ASIC1b. In this study, the term inactivation rather than desensitization will be used whenever there is no macroscopic activation observable. Oocytes were superfused for 2 min in an extracellular solution with a pH ranging from 7.8 to 6.7, and channels were then activated by a constant test pH of 6.0. The steady-state inactivation curves are plotted in Fig. 1B together with the curves for activation. The steady-state inactivation curves of both splice variants were very steep, leading to complete inactivation in a range of 0.4 pH units corresponding to a 2.5-fold increase in agonist concentration. Correspondingly, Hill coefficients were large (10.90 Ϯ 6.15 and 8.63 Ϯ 4.31 for ASIC1a and ASIC1b, respectively) suggesting high cooperativity for inactivation. The inactivation curves were shifted to more acidic values for ASIC1b (IC 50 , pH 7.00 Ϯ 0.04, n ϭ 10) compared with ASIC1a (IC 50 , pH 7.24 Ϯ 0.05, n ϭ 12). Such a shift would lead to the start of inactivation of ASIC1b at a pH at which ASIC1a is already almost completely inactivated ( Fig.  1), suggesting that ASIC1b replaces ASIC1a when the dynamic range of ASIC1a is saturated.
A Region in the Proximal Part of the Ectodomain Controls Apparent H ϩ Affinity of Both Activation and Steady-state Inactivation-To determine the structural elements in the primary structure of the protein controlling apparent H ϩ affinity of ASIC1, we constructed a series of chimeric molecules that exchanged parts of the N terminus of ASIC1a by the corresponding part of ASIC1b (Fig. 2). These chimeras exchange the cytoplasmic N terminus, the first transmembrane domain, and various parts of the extracellular loop. All of the chimeras were functional and showed current amplitudes comparable with wild type ASIC1a. First, we determined activation by H ϩ for the chimeras. The activation curves of all chimeras had the same overall shape as those of the wild type with similar Hill FIG. 1. Activation and steady-state inactivation of ASIC1. A, top, representative current traces obtained by repetitive short (3.5 s) activation of ASIC1 with pH as indicated. pH between activation was 7.4. Bottom, representative traces of ASIC1 currents elicited by application of pH 6.0 and conditioning pH varying between 7.8 and 6.7. B, Dose-response relationship for pH activation (n ϭ 7-11; ASIC1a, filled squares and ASIC1b, filled diamonds) and steady-state inactivation (n ϭ 10 -12; ASIC1a, filled circles and ASIC1b, filled triangles).
coefficients. Although the EC 50 of most chimeras did not correspond exactly to the value of either splice variant, the chimeras can be clearly divided into two groups. As can be seen from Fig. 2, currents through the chimeric channels in which up to 98 amino acids were replaced showed an EC 50 comparable with ASIC1a (values for chimeras C62-C98 ranged from 6.32 to 6.85, n ϭ 6 -10). A chimera in which the first 106 amino acids were replaced, however, showed a strong shift in the EC 50 to a value even lower than for ASIC1b (5.42 Ϯ 0.10, n ϭ 9). The EC 50 for all chimeras that contained larger stretches of ASIC1b resembled the EC 50 of ASIC1b (values for chimeras C112-C148 ranged from 5.52 to 6.00, n ϭ 6 -9). This suggests that amino acids from 98 to 106 are critical for determining the apparent H ϩ affinity of ASIC1 activation.
Next, we analyzed steady-state inactivation for the chimeras. Again, half-maximal, steady-state inactivation (IC 50 ) of most chimeras did not correspond exactly to the value of either splice variant, but they can be divided into the same two groups (Fig. 2). Chimeras C62, C76, and C92 showed half-maximal, steady-state inactivation at a pH similar to that of the ASIC1a wild type (values for chimeras C62-C92 ranged from 7.17 to 7.35, n ϭ 8 -9), whereas chimera C98 showed a significant shift of half-maximal steady-state inactivation (7.02 Ϯ 0.02, n ϭ 8 -9). However, the other chimeras showed a much stronger shift in steady-state inactivation to pH values similar to or more acidic than ASIC1b (values for chimeras C106 -C148 ranged from 6.74 -6.93, n ϭ 3-6). This is in agreement with the conclusion that the region comprising amino acids 98 to 106 of ASIC1 contains a critical element controlling the apparent affinity of the channel to H ϩ for both activation and steadystate inactivation.
In this region only four amino acids are different between ASIC1a and ASIC1b (Fig. 2). We substituted these four amino acids in ASIC1a individually by those found in ASIC1b and determined the pH of half-maximal activation as well as steady state inactivation of the substituted variants. Variants ASIC1a F100L and V103L had values similar to ASIC1a wild type (pH 6.60 Ϯ 0.13 and 6.64 Ϯ 0.08 for activation and pH 7.19 Ϯ 0.07 and 7.23 Ϯ 0.06 for steady state inactivation, respectively, n ϭ 8; Fig. 2). The other two variants (K105Y and N106P), however, showed a significant shift of both pH activation and steady state inactivation to values similar to ASIC1b wild type (pH 6.27 Ϯ 0.15 and 5.93 Ϯ 0.13 for activation and pH 7.05 Ϯ 0.02 and 6.95 Ϯ 0.04 for steady state inactivation, respectively, n ϭ 8; Fig. 2). These results indicate that these two residues are critical for setting the apparent H ϩ affinity of ASIC1.
Ca 2ϩ Potentiates Currents through ASIC1a at Physiologic Concentrations-Next, we investigated the interaction of ASIC1 with divalent cations. To have constant open channel properties, we used a protocol in which we varied the Ca 2ϩ concentration only during the conditioning period at pH 7.4 and held it constant at 1.8 mM during low pH (6.0) activation. When we increased the Ca 2ϩ concentration during the conditioning period at pH 7.4, we observed a significant increase in ASIC1a current amplitude when pH 6.0 solution was applied (Fig. 3A). Because we used different Ca 2ϩ concentrations during the conditioning period and low pH activation, the potentiation by Ca 2ϩ cannot be precisely quantified. With this reservation, we observed half-maximal potentiation around the physiological concentration of Ca 2ϩ (EC 50 , 2.10 Ϯ 0.29 mM, n ϭ 8; figure 3), suggesting that this interaction with Ca 2ϩ is of physiological relevance. We observed a similar potentiation of currents through ASIC1a by both Ba 2ϩ and Mg 2ϩ (EC 50 , 2.90 Ϯ 0.52 and 1.99 Ϯ 0.69 mM, respectively, n ϭ 7-8; figure 3). The Hill-coefficient for the potentiation by divalent cations was ϳ2 (2.45, 2.50, and 2.01 for Ca 2ϩ , Ba 2ϩ , and Mg 2ϩ , respectively), suggesting that more than one ion is necessary to potentiate the current. Potentiation was voltage independent (data not shown) suggesting a binding site outside the membrane electric N-terminal sequences from ASIC1a are drawn in light gray, those from ASIC1b in darker gray. The first transmembrane domain TM1 is indicated; the common C terminus is drawn in black, only its first part is shown. Names of the chimeras denote the number of amino acids, which have been replaced in ASIC1a by those from ASIC1b. N-terminal sequences of ASIC1b were from ASIC1b-M3 (4). Amino acid sequence of the critical region from 92 to 106 is shown below the chimeras. field. Under identical conditions, ASIC1b currents were only slightly potentiated by Ca 2ϩ (Fig. 3B).
Extracellular Divalent Cations Shift the Steady-state Inactivation Curve of ASIC1a and ASIC1b-We then addressed whether extracellular Ca 2ϩ changes the affinity of the channel to H ϩ . Low extracellular concentrations of divalent cations (0.1 mM Ca 2ϩ , 0 mM Mg 2ϩ ) during the conditioning period did not significantly shift the pH dose-response curve for ASIC1a and ASIC1b (EC 50 ϭ 6.53 Ϯ 0.11, n ϭ 11, and 5.86 Ϯ 0.08, n ϭ 7, respectively; Fig. 4A). For ASIC1a, at any given pH there was an increase in current amplitude with high concentrations of divalent cations compared with low concentrations (Fig. 4A,  inset). Thus, Ca 2ϩ potentiation is not due to a shift in the pH activation curve. We then decreased Ca 2ϩ concentration during low pH activation (0.1 mM Ca 2ϩ ) and held it constant (1.8 mM Ca 2ϩ , 1.0 mM Mg 2ϩ ) in the conditioning period at pH 7.4. Under these conditions, we observed a shift in the pH activation curve to a lower agonist concentration for ASIC1a (EC 50 ϭ 6.83 Ϯ 0.02, n ϭ 8, p Ͻ Ͻ 0.01; Fig. 4B) and to a lesser extent for ASIC1b (EC 50 ϭ 5.95 Ϯ 0.12, n ϭ 8, p ϭ 0.05; Fig. 4B). The shift of the activation curve during activation but not steady-state suggests fast on and off rates of Ca 2ϩ at this modulating site. The observed shift of the activation curve would lead to an inhibition by Ca 2ϩ during low pH activation, which has already been reported for ASIC3 (17). This inhibition by Ca 2ϩ reveals a complex interaction of ASICs with divalent cations.
Next, steady-state inactivation curves were determined with low concentrations of divalent cations (0.1 mM Ca 2ϩ ) between activation steps. For both splice variants, there was a significant shift of the steady-state inactivation curves by about 0.2 pH units to higher pH values (IC 50 , 7.48 Ϯ 0.07 for ASIC1a, n ϭ 12, and 7.22 Ϯ 0.04 for ASIC1b, n ϭ 10, p Ͻ Ͻ 0.01; Fig. 4C) compared with normal concentrations of divalent cations. This shift of the steady-state inactivation curves will lead to a sig-nificantly lower number of available ASIC1a channels at pH 7.4 and low concentrations of extracellular divalent cations. This accounts for the current potentiation by Ca 2ϩ at this pH.
Because the steady-state inactivation curve for ASIC1b is shifted to more acidic values compared with ASIC1a, this suggests that ASIC1b may be potentiated by Ca 2ϩ at a more acidic conditioning pH. In fact, at a conditioning pH of 7.1 we observed a strong potentiation for ASIC1b also (Fig. 4D). At this pH, current amplitude rose with increasing Ca 2ϩ concentration. Half-maximal potentiation was reached at a Ca 2ϩ concentration similar to ASIC1a (2.00 Ϯ 0.68 mM, n ϭ 9), suggesting: (i) a conserved binding site for Ca 2ϩ ; and (ii) a physiologic relevance for Ca 2ϩ potentiation of ASIC1b. Because Ca 2ϩ potentiation of ASIC1b can be seen only at a slightly acidic resting pH, ASIC1b is a more effective sensor of transient H ϩ signals during slight acidosis.
Shift in Steady-state Inactivation Is Due to a Stabilization of the Resting State-The shift in the steady-state inactivation curves by divalent cations can be explained either by a difference in the rate of recovery from desensitization and/or by a difference in the rate of steady-state inactivation. First, we determined time constants of recovery. We used a conditioning pH of 7.4. At this pH value we did not observe significant steady-state inactivation for ASIC1b with both low and normal concentrations of divalent cations (see Fig. 3); for ASIC1a we observed substantial inactivation at pH 7.4, but recovery at pH 7.7 was too fast (complete in less than 3 s) to be reliably determined. Time constants of recovery were similar for ASIC1a and ASIC1b under physiological concentrations of divalent cations (5.23 Ϯ 2.37 s, n ϭ 27, and 7.66 Ϯ 1.87 s, n ϭ 24, respectively; Fig. 5A). With low concentrations of divalent cations (0.3 mM Ca 2ϩ ), we observed a significantly slower recovery from desensitization for both ASIC1a and ASIC1b (15.39 Ϯ 14.14 s, n ϭ 23, and 17.03 Ϯ 3.96 s, n ϭ 19, respectively, p Ͻ 0.01).
The time constant of the onset of steady-state inactivation was determined by a varying period at pH 7.05 (ASIC1a) or 6.5 (ASIC1b) and normal concentrations of divalent cations or a varying period at pH 7.15 or 6.6 and low concentrations of divalent cations. These pH values were chosen to have maximal inactivation without significant activation (see Fig. 1B). The number of available channels was then assessed by a short (3.5 s) test pulse of pH 6.0, after which channels were allowed to recover from inactivation for 100 s at pH 7.4 (Fig. 5B). With low concentrations of divalent cations (0.3 mM Ca 2ϩ ), we observed a significantly faster steady-state inactivation for ASIC1a and a slightly faster inactivation for ASIC1b (ASIC1a, 14.08 Ϯ 3.99 s, n ϭ 13, compared with 78.92 Ϯ 54.48 s, n ϭ 16, respectively, p Ͻ Ͻ 0.01; ASIC1b, 38.95 Ϯ 8.63 s, n ϭ 23, compared with 47.48 Ϯ 18.88 s, n ϭ 18, respectively, p ϭ 0.06). Thus, Ca 2ϩ seems to favor the resting state of the channel, thereby increasing the rate of recovery as well as slowing the rate of inactivation during steady-state. Both mechanisms work in concert to increase the number of available channels during steady-state, which explains the shift in the steadystate inactivation curves.
The Polyvalent Cation Spermine Also Potentiates ASIC1 Currents-The extracellular concentration of the polyvalent cation spermine can significantly vary, and it has been shown that spermine modulates different ion channels. Potentiation of ASIC1 currents by divalent cations led us to ask whether spermine also interacts with ASIC1. We determined steadystate inactivation in the presence of 0.25 mM spermine and 0.1 mM Ca 2ϩ during the conditioning period. As can be seen in Fig.  6, 0.25 mM spermine led to a significant shift of the steady-state inactivation curves of both ASIC1a and ASIC1b. This shift was in the order of 0.2 pH units (half-maximal inactivation at pH 7.25 Ϯ 0.03 for ASIC1a and 7.04 Ϯ 0.03 for ASIC1b, n ϭ 9), showing that currents through both splice variants can be strongly potentiated by spermine and that interaction with spermine significantly extends the dynamic range of ASIC1. DISCUSSION Our study has two key findings. 1) ASIC1b is an effective sensor of transient H ϩ signals under slight acidosis, extending the dynamic range of ASIC1. 2) We show that extracellular diand polyvalent cations interact with ASIC1, leading to a potentiation of the current.
Physiological Relevance of Splice Products ASIC1a and ASIC1b-Our results indicate that agonist sensitivity is the major difference between ASIC1a and 1b. We show that ASIC1b can still function at a resting pH down to 6.9, which is of special importance because ASIC1b is specifically expressed in sensory neurons (5). This property may, thus, enable ASIC1b to sense acidic transients even under conditions of slight acidosis, for example during inflammation. The finding that Ca 2ϩ potentiation of ASIC1b is efficient only at a resting pH below 7.15 additionally shows that ASIC1b has its full operational capacity only at slightly acidic resting pH. ASIC1a is expressed in the dorsal root ganglion mainly in small diameter neurons (5,13,14), whereas ASIC1b is expressed in both small and large diameter neurons (5). Whether both splice variants are co-expressed in some small-diameter neurons is at present unknown. Irrespective of differential expression, our results suggest that alternative splicing at the ASIC1 gene increases the operational range of H ϩ receptors in sensory neurons.
Alvarez de la Rosa et al. (15) recently reported an EC 50 value for activation by H ϩ of ASIC1a expressed in Xenopus oocytes that is substantially higher (pH 5.3) than the value we obtained. ASIC1 activates within few msec (4) and, therefore, fast solution exchange is essential for concerted activation of ASICs. This is especially relevant for large cells like Xenopus oocytes. One possible explanation for lower agonist affinity reported in their study is, thus, slow solution exchange in their experimental set-up.
Physiological Relevance of Ca 2ϩ and Spermine Modulation-Our results point to an important role for extracellular Ca 2ϩ in modulation of neuronal receptor channels. The Ca 2ϩ -sensing receptor (CaR), a G-protein-coupled cell-surface receptor for divalent cations, is activated by Ca 2ϩ over a concentration range of 0.5 mM to 10 mM (18,19). ASIC1 has a Ca 2ϩ sensitivity comparable with the CaR, which is specialized in the sensing of extracellular Ca 2ϩ . The CaR has a role in body homeostasis of Ca 2ϩ but is also significantly expressed in different brain regions, suggesting that fluctuations of extracellular Ca 2ϩ concentrations do occur in the brain. In a recent elegant study, Hofer et al. have shown that mobilization of Ca 2ϩ in one cell will lead to extrusion of Ca 2ϩ and concurrent activation of neighboring cells expressing the CaR (20). This suggests that there is a form of intercellular communication using extracellular Ca 2ϩ . In addition to these local changes in Ca 2ϩ concentration, experiments with Ca 2ϩ selective microelectrodes have revealed more global changes in Ca 2ϩ concentration in the brain during neuronal activity (21)(22)(23). The extracellular Ca 2ϩ concentration can decrease from a resting value around 1.2 mM to values as low as 0.08 mM in extreme conditions (22). Although Ca 2ϩ concentrations usually decreased by only about one fourth of the resting concentration, local changes may well be higher.
In the skin it has been shown that there is a standing gradient of extracellular Ca 2ϩ (24) and it has been proposed that tissue injury may cause a spread of high Ca 2ϩ from the outer epidermis to the inner epidermis (25), which would then increase the number of available ASIC1. Chelation of divalent cations by metabolites like lactate may also contribute to vari-ations in the extracellular concentration of divalent cations (17).
The Ca 2ϩ and Mg 2ϩ concentration at which half-maximal potentiation of ASIC1 occurred (EC 50 , about 2 mM) matches well with the resting concentration of these ions. Therefore, small variations in the physiologic extracellular Ca 2ϩ and/or Mg 2ϩ concentration will have significant effects on channel activity. The inhibition of ASIC1 by Ca 2ϩ during H ϩ -activation and the potentiation in the steady-state endows this channel with the property to translate the time when Ca 2ϩ concentration changes into two opposing effects. Changes in Ca 2ϩ concentration both in the conditioning period and during low-pH activation have only a small net effect on channel activity (results not shown), adding to the possible responses of ASICs to changes in extracellular Ca 2ϩ concentration.
Natural polyamines, spermine, spermidine and putrescine, are synthesized and released upon brain trauma (26,27). The concentration of spermine in secretory granules and nerve terminals from ox neurohypophyses has been calculated to be 0.26 and 0.52 mM, respectively (28). Our results show that concentrations of spermine similar to these physiological values shift the steady-state inactivation curve of ASIC1, which would increase the number of available channels. Increasing the number of available channels would be especially relevant during tissue acidosis and may contribute to increased excitotoxicity during seizures and ischemia when the extracellular pH falls. A more detailed analysis will show whether spermine acts through the same or a different mechanism than divalent cations and what the apparent affinity is of ASIC1 for polyvalent cations.
Ca 2ϩ Modulation of Other Ion Channels-ASICs share the property of Ca 2ϩ potentiation with P2X receptors (25), nicotinic acetylcholine receptors (29), and metabotropic glutamate receptors (30). They also share the property of Mg 2ϩ potentiation with the NMDA receptor (31). This underlines the relevance of potentiation. Potentiation of P2X 3 receptors by Ca 2ϩ is especially interesting because P2X 3 is an ion channel that is specifically expressed in small-diameter sensory neurons. P2X 3 currents are potentiated by speeding recovery from desensitization, and potentiation has the special property of endowing P2X 3 , which shows a very slow (Ͼ 5 min) recovery, with a memory for Ca 2ϩ transients (25). Because recovery of ASIC1 is much faster and, additionally, the memory will vanish because of steady-state inactivation, the memory for Ca 2ϩ transients during steady-state will be short (several seconds) for ASIC1.
Mechanism of Ca 2ϩ Potentiation-The most likely explana- FIG. 5. Effect of Ca 2ϩ on recovery from desensitization and onset of steady-state inactivation. A, Recovery from desensitization. Left, to achieve complete desensitization of ASIC1, oocytes were exposed to pH 6.0 for 30 s (horizontal bars). These periods of activation were interrupted by increasing periods (3,5,15,25,50, and 100 s) with pH 7.4, allowing ASIC1 channels to recover from desensitization. This protocol was performed with low and normal concentrations of divalent cations during the conditioning period at pH 7.4. We observed in some measurements significant channel rundown. Right, current peak amplitudes were normalized to the largest current amplitude during a recovery protocol and are plotted against the corresponding time of recovery. B, Onset of steady-state inactivation. Left, ASIC1 was activated shortly (pH 6.0 for 3.5 s, black bars) and was then allowed to recover completely from desensitization (pH 7.4 for 100 s). Next, oocytes were exposed to a conditioning pH for varying periods of time (hatched bars). Right, current peak amplitudes were normalized to the largest current amplitude at the end of a steady-state inactivation protocol (time in conditioning pH ϭ 0 s) and are plotted against the corresponding time in conditioning pH. tion of the potentiation by Ca 2ϩ and other di-and polyvalent cations is a direct competitive binding with H ϩ to a common binding site. Alternatively, there may be a distinct binding site for di-and polyvalent cations, which would change the affinity at the H ϩ binding site.
The H ϩ sensor or ligand binding domain of ASICs is unknown. It has been suggested that His-72 may be involved in activation of ASIC2a (32). However, this residue is not the ligand binding site of ASIC1 or ASIC3 (32). Our results suggest that residues 105 and 106 of ASIC1a are critical in modulating the pK a of the H ϩ binding site. We propose that they are contained within a pocket creating an altered electrostatic environment thereby changing pK a of the H ϩ binding site. This does not necessarily imply that they are close to the H ϩ binding site in the primary structure of the channel. The similar EC 50 of Ca 2ϩ potentiation of ASIC1a and ASIC1b suggests a conserved binding site for Ca 2ϩ . Negatively charged amino acids are the best candidates to contribute to this Ca 2ϩ binding site. Site-directed mutagenesis will show whether any of the numerous aspartates/glutamates in the extracellular loop of ASICs are involved in Ca 2ϩ binding and if these residues constitute also the H ϩ sensor.
The fact that steady-state inactivation occurs at a pH at which there is no macroscopic activation has been taken as evidence that the inactivated state can be reached without prior opening of the channel (7,15). However, models where the channel has to open before inactivation occurs cannot be ruled out. Our results (not shown) indicate that time constant of recovery from steady-state inactivation is not significantly different from time constant of recovery from desensitization suggesting that there is only one type of inactivated state and we favor a simple model where closed, open and inactivated states are linearly connected.
In summary, our results show that ASIC1b is still active at more acidic resting pH than ASIC1a, suggesting that it is a sensor of transient H ϩ signals during slight acidosis. Moreover, we show that di-and polyvalent cations modulate gating of ASIC1, which may link ASIC activity to the extracellular concentrations of these ions.