pH Dependency and desensitization kinetics of heterologously expressed combinations of acid-sensing ion channel subunits.

The exact subunit combinations of functional native acid-sensing ion channels (ASICs) have not been established yet, but both homomeric and heteromeric channels are likely to exist. To determine the ability of different subunits to assemble into heteromeric channels, a number of ASIC1a-, ASIC1b-, ASIC2a-, ASIC2b-, and ASIC3-containing homo- and heteromeric channels were studied by whole-cell patch clamp recordings with respect to pH sensitivity, desensitization kinetics, and level of sustained current normalized to peak current. Analyzing and comparing data for these three features demonstrated unique heteromeric channels in a number of co-expression experiments. Formation of heteromeric ASIC1a+2a and ASIC1b+2a channels was foremost supported by the desensitization characteristics that were independent of proton concentration, a feature none of the respective homomeric channels has. Several lines of evidence supported formation of ASIC1a+3, ASIC1b+3, and ASIC2a+3 heteromeric channels. The most compelling was the desensitization characteristics, which, besides being proton-independent, were faster than those of any of the respective homomeric channels. ASIC2b, which homomerically expressed is not activated by protons per se, did not appear to form unique heteromeric combinations with other subunits and in fact appeared to suppress the function of ASIC1b. Co-expression of three subunits such as ASIC1a+2a+3 and ASIC1b+2a+3 resulted in data that could best be explained by coexistence of multiple channel populations within the same cell. This observation seems to be in good agreement with the fact that ASIC-expressing sensory neurons display a variety of acid-evoked currents.

It is well established that tissue acidification, which may be present in inflammatory and ischemic conditions, causes pain (1)(2)(3). In line with this, peripheral sensory neurons exhibit sensitivity toward acid by activating several types of depolarizing currents (4 -6). Although the repertoire of ion channels responsible for these currents is not fully known, the family of acid-sensing ion channels (ASICs) 1 is believed to be an important constituent. The ASIC family is a member of the ENaC/ DEG superfamily, which also includes the amiloride-sensitive epithelial sodium channels (ENaCs), the mechanically gated degenerins of Caenorhabditis elegans (DEGs), and a neuropeptide-gated channel of Helix aspersa (FaNaC). The membrane topology of all channels within this superfamily comprises two transmembrane domains, intracellular N and C termini and a large extracellular loop with a number of conserved cysteine residues (7). Currently, four genes encoding six ASIC transcripts have been cloned and characterized from mammalian organisms. ASIC1a (BNaC2) and ASIC1b (ASIC1␤) are the products of alternatively spliced transcripts of the ASIC1 gene that differ in the N-terminal region, including the first transmembrane domain and the proximal part of the large extracellular domain (8 -10). ASIC2a (BNaC1, MDEG) and ASIC2b (MDEG2) are alternatively spliced forms of the ASIC2 gene product. These, too, have unique N termini and share the C-terminal amino acids (11). Splice variants have not yet been identified for ASIC3 (DRASIC) (12,13) and ASIC4 (SPASIC) (14,15). Except for ASIC2b and ASIC4, all subunits have the ability to form functional homomeric channels when expressed in Xenopus laevis oocytes or mammalian cells. The functional and pharmacological properties of homomeric ASIC1a match one type of acid-evoked current described in peripheral sensory neurons (16,17), and the properties of ASIC3-mediated currents mimic acid-evoked currents in cardiac sensory neurons, believed to mediate the pain of angina (18). Different ASIC subunits often co-localize within the same neurons (19), and heteromeric channels are likely to exist, as is the case for most ligand-gated ion channels. However, the exact composition of endogenous acid-sensitive channels is at present largely unknown.
To study heteromerization of ASICs, we examined co-expressions of different ASIC subunits in Chinese hamster ovary (CHO) cells. By analyzing currents obtained in whole-cell voltage clamp experiments using ultrafast solution exchange for providing acidification, data for all possible combinations of ASICs (except ASIC4) were obtained. These data reveal unique properties for several of the homomerically expressed ASIC subunits and clearly demonstrate the formation of heteromeric channels in a number of co-expression experiments. Interestingly, not all the tested co-expressions led to channels that could be identified as unique; in the case of experiments involving co-expression of three different subunits, the data suggested existence of multiple channel populations. To summarize, this is the first comprehensive study of ASICs co-expressed in a mammalian expression system.

EXPERIMENTAL PROCEDURES
Cloning of ASICs-Poly(A) ϩ mRNA was purified from fetal rat dorsal root ganglion (DRG) neurons using the Oligotex® direct mRNA mini kit (Qiagen) according to the manufacturer's protocol. ASIC1b, ASIC2b, and ASIC4 were cloned from fetal rat DRG poly(A) ϩ mRNA, whereas ASIC1a, ASIC2a, and ASIC3 were cloned from rat total brain poly(A) ϩ mRNA (Clontech) using reverse transcription PCR. First and second strand cDNA was obtained in a one-tube reverse transcriptase-PCR reaction using Avian myoblastosis virus reverse transcriptase, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The cDNAs were cloned into pSwas, which is a custom-designed vector derived from pZero TM (Invitrogen), and positive clones were sequenced bidirectionally. ASIC1a, ASIC2a, ASIC2b, and ASIC4 were subcloned directly into the pNS1z vector. The ASIC1b fragment, which encoded the N-terminal part of the channel, was subcloned into pNS1z-ASIC1a, substituting the corresponding part of ASIC1a. ASIC3 was cloned in two overlapping fragments and subcloned in pNS1z in a unique restriction site of ASIC3 (BsrGI) within the overlapping region. The pNS1z vector is a customized vector derived from pcDNA3 (Invitrogen) with expression of the insert under control of the cytomegalovirus promoter.
Electrophysiology-All experiments were performed under voltage clamp using conventional whole-cell patch clamp methods (20) at 20 -22°C. A Macintosh G4 computer was used to control an EPC-9 amplifier (HEKA-electronics) via an ITC-16 interface. The experimental conditions were defined by the Pulse software accompanying the amplifier, and data were sampled at 2 kHz and low pass-filtered at 667 Hz.
Pipettes were pulled from borosilicate glass (Modulohm) using a horizontal electrode puller (Zeitz-Instrumente). The pipette was filled with an intracellular solution containing 120 mM KCl, 31 mM KOH, 2 mM MgCl 2 , 10 mM EGTA, and 10 mM HEPES adjusted to pH 7.2. The pipette electrode was a chloridized silver wire, and the reference electrode was a silver chloride pellet electrode (In Vivo Metric) fixed to the experimental chamber. The electrodes were zeroed with the open pipette in the bath just prior to sealing, and the pipette resistances were 1.5-3.0 megohms. Series resistance compensation was set at 80%.
Coverslips with cells were transferred to the recording chamber (RC-25F, Warner Instruments) mounted on the stage of an inverted microscope (Olympus). Transfected cells were identified by the emission of green fluorescence when exposed to UV light. After giga-seal formation, the whole-cell configuration was attained by suction.
Cells were continuously superfused at a rate of 2.5 ml/min with an extracellular solution (Na-R) containing 140 mM NaCl, 4 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, and 5 mM MES adjusted to pH 7.4. Test solutions were made from standard Na-R, adjusting pH with HCl. Rapid pH changes were achieved by placing a piezo-driven doublebarreled application pipette (theta tube) in front of the cell. Using this system, complete solution exchange can be achieved in Ͻ1 ms, as measured by the liquid junction potential shift (data not shown). Currents were measured at the peak of the response and normalized to the current evoked by a pH 4 stimulus.
Data Analysis-Using the software GraphPad Prism® 3.0, pH doseresponse curves were fitted to the equation I/I max ϭ 1/(1ϩ(EC 50 /[H ϩ ]) n ), where EC 50 denotes the proton concentration [H ϩ ] yielding 50% activation and n is the Hill coefficient. In a few instances, the current was not fully saturated at pH 4 when fitting to this equation. In these cases, the current amplitudes were subsequently normalized to the value of I/I max at saturation as estimated from the fit to obtain a maximal I/I max of 1.0. Time constants of desensitization were calculated using IgorPro software (WaveMetrics) by fitting the falling phase of the evoked current to a single exponential function I(t) ϭ K 0 ϩ K 1 exp(ϪK 2 t), where the time constant of desensitization, desens , equals 1/K 2 , K 0 denotes the amplitude of the non-desensitizing current, and K 0 ϩ K 1 is the peak current; i.e. all ASIC subunit combinations yielded currents that desensitized incompletely at pH 4.0. Results are presented as mean Ϯ S.E., and comparisons were made using a two-tailed t test or one-way analysis of variance. A p value of less than 0.05 was considered to be significant.

RESULTS
ASIC subunits were transiently expressed, either singly or as combinations, in CHO cells. All functional channels yielded transient acid-evoked whole-cell currents that displayed varying levels of sustained current. These whole-cell currents were characterized with respect to three different properties: 1) the pH sensitivity (pH 50 and Hill slope); 2) the time constant of desensitization ( desens ), measured at various pH values; and 3) the amplitude of the residual non-desensitizing current at pH 4.0 (Table I). Some ASIC channels exhibited desensitization kinetics that were dependent on the applied proton concentration, whereas others had kinetics that were independent of proton concentration. When comparing channels with these different kinetic properties, values of desens determined at a proton concentration close to the pH 50 values or in some cases pH 100 values for activation were used.
Homomeric Expression of ASIC Subunits-The expression of ASIC1a, ASIC1b, ASIC2a, or ASIC3 in CHO cells resulted in functional acid-sensitive channels ( Fig. 1, A-E), but ASIC2b and ASIC4 could not be activated even at pH 4. Whereas ASIC1a, ASIC1b, and ASIC2a displayed transient activation characteristics, ASIC3 was found to have two different activation profiles. In some cells ASIC3 had a transient activation profile (Fig. 1D), but in other cells a slower activating current followed the transient current (Fig. 1E). The proton affinities for the various channels varied by two orders of magnitude ( Fig. 2A and Table I), with ASIC3 being the most sensitive subunit (pH 50 ϭ 6.4) and ASIC2a the least sensitive (pH 50 ϭ 4.5). Although pH 50 was similar for the two ASIC1 splice variants, activation of ASIC1a occurred across a larger pH range, whereas ASIC1b was activated and saturated within a narrow pH range ( Fig. 2A). This is clearly reflected in the Hill coefficients, with ASIC1a and ASIC1b having Hill slopes of 0.75 and 4.8, respectively. The Hill coefficients for ASIC2a and ASIC3 were intermediate to those of the two ASIC1 splice variants ( Table I).
The values of desens were strongly dependent upon the proton concentration used for activation of ASIC1a, ASIC1b, and ASIC2a, as the rate of desensitization increased with increasing proton concentrations (Fig. 2, B and C). This is a feature commonly observed for ligand-gated ion channels. In contrast, the desensitization rate of the transient current of ASIC3 was independent of the applied pH (pH 6.5-4.0; p ϭ 0.25) (Fig. 2B). Despite their different pH 50 and Hill slopes, no significant differences in the desensitization rates of ASIC1a and ASIC1b were observed at pH 50 (p Ͼ 0.14).
The sustained current remaining at pH 4.0 was in the range of 3-8%, relative to the peak current, for ASIC1a, -1b, and -2a (Table I). ASIC3 displayed a significantly larger sustained current of 26% (p Ͻ 0.001) ( Table I).
ASIC1a Co-expressed with ASIC1b, ASIC2a, ASIC2b, or ASIC3-Co-expression of ASIC1a and one other ASIC subunit gave rise to functional acid-sensitive channels in all cases (Fig.  1, F-I). Proton sensitivities of the different subunit combinations fell into two categories. The ASIC1aϩ1b, ASIC1aϩ2a, and ASIC1aϩ2b combinations all showed proton sensitivities similar to ASIC1a (Fig. 3, A and B, respectively), whereas the sensitivity of the ASIC1aϩ3 combination was indistinguishable from that of ASIC3 homomeric channels (Fig. 3C).
For three of four of these combinations desensitization kinetics were affected in one way or another relative to the respective homomeric receptors; only ASIC1aϩ2b appeared unchanged compared with the ASIC1a homomeric receptor. The desensitization rate of the ASIC1aϩ1b combination was dependent on the proton concentration as was the case for the respective homomeric channels (Fig. 4A). Contrary to this, the rates for the ASIC1aϩ2a and ASIC1aϩ3 combinations were independent of the proton concentration (p Ͼ 0.98 and p Ͼ 0.48, respectively) (Fig. 4, B and C). Comparing desensitization rates at pH 50 , ASIC1aϩ1b was significantly slower than ASIC1a and ASIC1b (p Ͻ 0.001 and p Ͻ 0.05, respectively), whereas ASIC1aϩ3 was significantly faster than ASIC1a and ASIC3 (p Ͻ 0.001 and p Ͻ 0.05, respectively). Due to the pH dependence of ASIC1a desensitization, a comparison of the rates of ASIC1aϩ2a and ASIC1a was complex with significant differences at some proton occupancy levels but not at others. However, measured at pH ϭ 4.0 (ϳpH 100 ), ASIC1aϩ2a desensitization was significantly slower than that of ASIC1a (p Ͻ 0.05) but faster than that of ASIC2a (p Ͻ 0.001).
The sustained current at pH 4.0 of all the combinations was in the range of 3-6% of peak currents, which is not significantly different from homomeric ASIC1a (Table I). Interestingly, when compared with homomeric ASIC3, the ASIC1aϩ3 combination displayed a major reduction of the level of sustained current (p Ͻ 0.001).
ASIC1b Co-expressed with ASIC2a, ASIC2b, or ASIC3-ASIC1b co-expressed with ASIC2a or ASIC3 in CHO cells yielded functional channels (Fig. 1, J-K); however, contrary to all other combinations of subunits, ASIC1bϩ2b did not exhibit proton-activated currents in 30 of 36 cells. In the transfection procedure for ASIC1bϩ2b, an increase of DNA levels to the highest level possible (before cellular death occurred) was tried, but still, in the few cells that did show a proton-evoked current, the amplitude was very low (Ͻ100 pA). For this reason it was concluded that ASIC1bϩ2b is a non-functional combination, and no data are presented. The proton sensitivity of the ASIC1bϩ2a combination (Fig. 3D) resembled that of ASIC2a, whereas the sensitivity of ASIC1bϩ3 (Fig. 3E) was almost identical to that of homomeric ASIC1b.
The values of desens of both ASIC1bϩ2a and ASIC1bϩ3 were independent of the proton concentration used for stimulation (p Ͼ 0.23 and p Ͼ 0.15, respectively) (Fig. 4, D and E). Again, comparing proton-independent with proton-dependent desensitization is problematic. In general, desensitization rates of ASIC1bϩ2a could be characterized as intermediate to the respective homomeric channels; ASIC1bϩ2a was significantly faster than ASIC2a at pH 50 (p Ͻ 0.05) but significantly slower than ASIC1b at pH 100 (p Ͻ 0.05). The general picture of ASIC1bϩ3 was one of a channel desensitizing faster than the respective homomeric channels (Fig. 4E). Indeed, it desensitized significantly faster than ASIC1b at pH 50 (p Ͻ 0.001); however, despite the pH independence of both ASIC3 and ASIC1bϩ3 significant differences were not observed for all proton occupancy levels. Still, at pH 100 (pH ϭ 4.0) ASIC1bϩ3 clearly desensitized significantly faster than ASIC3 (p ϭ 0.01).
For the ASIC1bϩ2a combination the level of sustained current at pH 4.0 relative to peak current was 14%, which represents a significant increase compared with either homomeric channel (p Ͻ 0.001). The level of sustained current for the ASIC1bϩ3 combination (15%) was intermediate to the levels determined for the respective homomeric channels (Table I).
Co-expression of ASIC2aϩ2b, ASIC2aϩ3, and ASIC2bϩ3-Acid-sensitive channels were formed from co-expressing dual combinations of ASIC2a, ASIC2b, and ASIC3 (Fig. 1, L-O). The activation characteristics of ASIC2aϩ3 resembled ASIC3 with two different profiles. In some cells ASIC2aϩ3 had an apparent simple transient activation profile (Fig. 1M), whereas in other cells a slower activating current followed the transient current (Fig. 1N). Despite functionality of all these subunit combinations, the acid sensitivity of ASIC2a and ASIC3 was largely unaffected by the presence of ASIC2b (Fig. 3, D and F, respectively), whereas the sensitivity of ASIC2aϩ3 was intermediate to either homomeric channel (Fig. 3F).
Overall, desensitization kinetics of ASIC2aϩ2b and ASIC2bϩ3 were identical to the kinetics of homomeric ASIC2a and ASIC3, respectively (p Ͼ 0.05) (Fig. 4, F and G). The desens values of ASIC2aϩ3 were independent on the proton concentration (p Ͼ 0.54) (Fig. 4G). At pH 50 ASIC2aϩ3 desensitized significantly faster than ASIC2a (p Ͻ 0.01); however, although desensitization of ASIC2aϩ3 always appeared faster than ASIC3 (Fig. 4G), this difference did not reach significance. Interestingly, the desensitization phase of ASIC2aϩ3 was masked at pH Յ 4.5 due to activation of a prominent sustained current (see below).
Whole-cell currents measured from cells co-expressing ASIC2aϩ2b maintained a level of sustained current of 17% at  4.0; this represents a significant increase relative to the 7.5% determined for homomeric ASIC2a (p Ͻ 0.0001) ( Table I). Interestingly the ASIC2aϩ3 combination displayed a pronounced mean sustained current of 105% at pH 4. The reason for a measured sustained current larger than the peak current is the second slower activating current observed in some cells, which frequently exceeds the amplitude of the transient peak. The sustained current of 105% was significantly greater than the values corresponding to both homomeric ASIC2a and ASIC3 (p Ͻ 0.05). The level of sustained current determined for ASIC2bϩ3 was of a similar magnitude to that of ASIC3 homomers (p Ͼ 0.16) (Table I).
Co-expression of ASIC1aϩ2aϩ3-Co-expressing these three subunits gave rise to channels with peculiar characteristics (Fig. 1P). The proton sensitivity of acid-evoked currents in these cells was highly unusual (Fig. 5A), with activation starting at pH 6.5 and not fully saturating at pH 4, which is also clearly reflected in a low Hill slope of 0.65 (Table I).
The desens values of ASIC1aϩ2aϩ3-mediated whole-cell currents displayed unusual pH dependence, in the range of pH 6 -4.5, in that desens actually increased with increasing proton concentration (p Ͻ 0.01) (Fig. 6A). The desensitization rate at pH 50 was similar to that of ASIC1aϩ2a (p Ͼ 0.05) but significantly slower when compared with the rates of ASIC1aϩ3 and ASIC2aϩ3 (p Ͻ 0.001).
The level of sustained current at pH 4.0 did not differ significantly from that of ASIC1aϩ2a and ASIC1aϩ3 (p Ͼ 0.05) but was significantly lower than that of ASIC2aϩ3 (p Ͻ 0.01) ( Table I).
Co-expression of ASIC1aϩ2bϩ3-Although this co-expression experiment yielded functional channels (Fig. 1Q) the channel characteristics were similar to those of ASIC1aϩ3. First, the proton sensitivity of acid-evoked currents was virtually identical to that of ASIC1aϩ3-mediated currents (Fig. 5B). Next, desensitization kinetics were not significantly different from those of ASIC1aϩ3 at any pH value (p Ͼ 0.05) (Fig. 6B). Finally, the level of sustained current normalized to peak current at pH 4.0 was not significantly different from that of ASIC1aϩ3 (p Ͼ 0.35) ( Table I).
Co-expression of ASIC1bϩ2aϩ3-The last triple co-expression tested also yielded functional acid-sensitive channels (Fig.  1R) but again with peculiar characteristics. With respect to the proton sensitivity, the ASIC1bϩ2aϩ3 combination was reminiscent of ASIC1aϩ2aϩ3 (Fig. 5A) and was thus less acidsensitive than any of the corresponding pairwise subunit combinations (Table I).
Time constants of desensitization for ASIC1bϩ2aϩ3-mediated currents were also unusual compared with other subunit combinations, as desens increased at increasing proton concentration (Fig. 6A). The desensitization rate at pH 50 was signifi-FIG. 2. Biophysical properties of homomeric ASICs. A, sensitivity of homomeric ASICs to lowering of extracellular pH from a holding of pH 7.4, showing that ASIC3 is most and ASIC2a least sensitive to low pH and that activation of ASIC1a and ASIC1b can be clearly separated. Currents are normalized to pH 4.0 currents (I max ), except ASIC2a currents, which are not fully saturated at pH 4 and therefore subsequently have been normalized to I/I max at saturation as estimated from the initial pH dose-response fit. B and C, desensitization kinetics of homomeric ASICs indicated by mean time constants of desensitization ( desens ), revealing that desensitization of ASIC1a, ASIC1b, and ASIC2a is increased when pH is lowered, whereas the rate of desensitization of the transient ASIC3 current is constant despite changes in pH. n ϭ at least four cells for each data point. Bars indicate mean Ϯ S.E. cantly faster than that of ASIC1bϩ2a (p Ͻ 0.01) but significantly slower than those of ASIC1bϩ3 (p Ͻ 0.05) and ASIC2aϩ3 (p Ͻ 0.05).
The mean level of sustained current at pH 4.0 was not significantly different from that of ASIC1bϩ2a (p Ͼ 0.05) and ASIC1bϩ3 (p Ͼ 0.05), whereas it was significantly reduced compared with that of ASIC2aϩ3 (p Ͻ 0.05) ( Table I). DISCUSSION Results from co-expressing subunits, such as the ASIC subunits, can potentially be quite complex. In the simplest system either homomeric or heteromeric channels are formed; however, besides the possibility of coexistence of homomeric and heteromeric channels, the heteromeric channels themselves could be a mixed population with different stoichiometries. Assuming a tetrameric assembly of the ASIC channel (21), co-expressing two subunits gives rise to the possibility of hav-ing two homomeric channels along with three different heteromeric channel assemblies. When co-expressing three subunits these numbers rise to a possibility of having three homomeric channels along with 12 heteromeric channels. The currents measured from co-expression experiments may therefore reflect the presence of homomeric and/or heteromeric channels, which may or may not have distinct characteristics. Thus, presence of heteromeric ASIC channels could only be identified to the extent that they displayed biophysical properties distinct from the homomeric channels. Generally, the features studied here (pH sensitivity, desensitization rate, and sustained current at pH 4) proved to be very informative in identifying the existence of heteromeric channels, and the present data demonstrate that a number of distinct heteromeric channels are formed.
Homomeric ASICs Show Variable Properties-Initially, the  4. Desensitization kinetics of double ASIC combinations. The rate of desensitization is indicated by mean time constants of desensitization ( desens ). A, ASIC1aϩ1b desensitizes more slowly than either homomeric channel, reflected in higher desens , and as observed for both homomers, desensitization is increased at decreasing pH. B, desensitization of ASIC1aϩ2a is independent of proton concentration and shows no obvious resemblance to either homomeric channel. There are no significant differences between desens of ASIC1a and ASIC1aϩ2b. C, ASIC1aϩ3 different homomeric channels were characterized to enable identification of channels formed from co-expression experiments. As described previously, ASIC1a, ASIC1b, ASIC2a, and ASIC3 all expressed well as homomeric channels (8, 10 -12), whereas ASIC2b and ASIC4 did not yield functional acid-sensitive channels (11).
Interestingly, the pH dose-response curves of ASIC1a and ASIC1b differed markedly despite similar pH 50 values; activation and saturation of ASIC1a occurred over a range of three pH units, whereas ASIC1b activated and saturated within a single pH unit. Waldmann et al. (8) reported a much steeper pH dose-response curve for ASIC1a expressed in Xenopus oocytes, and Chen et al. (9) showed more gradual activation of ASIC1b expressed in COS-7 cells. Hence, this feature may be critically dependent upon cell type and/or the solution application system used for whole-cell recording. In agreement with previous findings, ASIC2a homomeric channels were the least acidsensitive (i.e. displaying a low pH 50 ) (11), whereas ASIC3 homomers were the most acid-sensitive (i.e. high pH 50 ) (12).
The desensitization rate of ASIC1a, ASIC1b, and ASIC2a was clearly increased when the stimulation pH was lowered (i.e. proton concentration increased). Conversely, the time constant of desensitization for ASIC3 remained constant despite changes in pH.
Heteromeric ASIC Channels with Distinct Properties-A heteromeric ASIC1aϩ1b channel with functionally unique properties has not been reported previously; however, the present data support its existence. Currents from cells co-expressing ASIC1aϩ1b desensitized significantly slower at pH 50 than currents corresponding to either homomeric channel. Moreover, the proton sensitivity and Hill slope of ASIC1aϩ1b were indistinguishable from those of homomeric ASIC1a, which would appear inconsistent with a mixture of homomeric channels. If only homomers were present, the very steep proton dose-response curve of ASIC1b would be expected to affect the shallow dose-response curve of homomeric ASIC1a.
Co-expression of either ASIC1 splice variant with ASIC2a resulted in heteromeric channels clearly identifiable from the respective homomeric channels. Contrary to the homomeric channels, the ASIC1aϩ2a and ASIC1bϩ2a combinations both displayed pH-independent rates of desensitization. Moreover, proton sensitivities of activation for both combinations were desensitizes faster than both ASIC1a and ASIC3, reflected in lower desens , and the kinetics is independent of [H ϩ ]. D, ASIC1bϩ2a desensitizes slower than ASIC1b but faster than ASIC2a (not shown in this figure because of desens in the range of 1-5 s, Fig. 2C), and desensitization kinetics is constant across the applied pH range unlike ASIC1b and ASIC2a. E, ASIC1bϩ3 desensitizes faster than both ASIC1b and ASIC3, reflected in lower desens , and the kinetics is independent of [H ϩ ]. F, ASIC2b does not influence desens of ASIC2a significantly. G, desensitization of ASIC2aϩ3 is significantly faster than ASIC2a (compare with Fig. 2C) and ASIC3, whereas ASIC2b does not change desensitization kinetics of ASIC3 significantly. inconsistent with a predominance of homomeric channels. Considering the large separation of the pH dose-response curves and Hill slopes between the ASIC1 variants and ASIC2a, a mixture of independent homomeric channels would be anticipated to result in biphasic pH dose-response curves or curves with extremely low Hill slopes. However, this was not observed; ASIC1aϩ2a had a pH 50 value similar to homomeric ASIC1a but an increased Hill slope, whereas ASIC1bϩ2a had a pH 50 value similar to homomeric ASIC2a with an unaltered Hill slope.
As noted above, ASIC3 was characterized by displaying pHindependent desensitization kinetics. Interestingly, this phenotype was conferred to the channels resulting from the coexpressions of ASIC3 with ASIC1a, ASIC1b, or ASIC2a. Furthermore, these co-expressions all gave rise to channels with accelerated desensitization kinetics (ϳ0.2 s) relative to any of the respective homomeric channels. However, this observation might in each case argue in favor of assembly of heteromeric channels. In agreement with this, it was reported previously that desensitization of acid-evoked currents from DRG neurons of ASIC3-null mice was slowed compared with wild type animals, suggesting that presence of ASIC3 might shorten the response to acid in vivo (22,23). A striking observation in the co-expression experiments with ASIC3 was the large amplitude of sustained current evoked by stimulation of ASIC2aϩ3 at pH 4.0. The summed contributions of homomeric ASIC2a and ASIC3 channels cannot explain the magnitude of this current, strongly suggesting the assembly of functional ASIC2aϩ3 heteromeric channels. This was also noted by Babinski et al. (24), who co-expressed ASIC2a and ASIC3 in Xenopus oocytes.
Heteromerization of ASIC1aϩ3, ASIC1bϩ3, and ASIC2aϩ3 was supported by the fact that the DNA amounts used for these transient transfections were reduced considerably compared with expression of the same subunits singly. This could suggest very efficient assembly and/or transport of such heteromeric channel complexes to the cell membrane.
Interestingly, whenever co-expressing two subunit types, the pH dose-response curves usually ended up being very similar to one of the curves for the respective homomeric channels. Only in the case of ASIC2aϩ3 an easily identifiable "new" doseresponse curve appeared, which was located between the curves for the respective homomers.
Does ASIC2b Constitute a Functional Subunit?-The present data do not provide convincing evidence that ASIC2b participates in any functional channel complex, although it previously has been suggested to constitute a modulatory subunit (11). When ASIC2b was co-expressed with ASIC1a or ASIC3 or both in combination, minor changes to features such as pH doseresponse curves could be observed, but nothing significant. In accordance with this, Lingueglia et al. (11) reported that ASIC2b was devoid of effect on the identical set of properties of ASIC1a and ASIC3. However, these authors found that ASIC2b altered the permeation properties of ASIC3 from being Na ϩ -selective to being non-selective for cations, suggesting that heteromultimerization of these subunits had occurred. With regards to ASIC2aϩ2b co-expression, the relative level of sustained current at pH 4.0 appeared increased by the presence of ASIC2b, an observation also reported by Lingueglia et al. (11). However, there were no other clear indications suggesting that ASIC2b influenced the properties of ASIC2a. Strangely, the most obvious "effect" of ASIC2b appeared to be suppression of ASIC1b function. Compared with homomeric expression of ASIC1b, the fraction of cells displaying acidevoked currents was highly reduced by the presence of ASIC2b, and acid-evoked currents of the few responsive cells were of low amplitude. This might suggest a dominant negative effect of ASIC2b on ASIC1b.
Three Different Subunits Appear Not to Form a Distinct Channel-As described in the preceding sections, ASIC1a, ASIC2a, and ASIC3 are capable of assembling into heteromeric complexes when co-expressed pairwise. Therefore, we surmised that assembly of all three of these subunits into single ASIC channels could be a realistic possibility. However, experiments involving co-expression of the three subunits yielded no conclusive evidence in favor of such an interaction. Two lines of evidence rather point toward existence of multiple channel populations. The pH dose-response relationship was unusual with activation taking place over a broad pH range, thereby yielding a very low Hill coefficient; the desensitization rate for ASIC1aϩ2aϩ3 decreased with increasing proton concentration in a large part of the pH range examined, which is contrary to the properties of all other investigated combinations of subunits. These observations could be explained by assuming that multiple channel populations were present. Considering the very efficient assembly of in particular heteromeric ASIC1aϩ3 and ASIC2aϩ3 channels but also ASIC1aϩ2a, it might seem unlikely that homomeric channels contribute significantly. On the other hand, the most obvious way to create a shallow dose-response relation is by having a mixture of channels with dose-response relations widely separated over the pH range, which suggests a presence of ASIC2a homomeric channels. Thus, a mixture of ASIC1aϩ3 and ASIC2a channels could theoretically account for a large part of the unexpected observations. The difference in pH 50 of these two channel types would result in a low Hill coefficient of the activation profile. Furthermore, at high pH values (pH 5.5-6.5), the rapid desensitization of ASIC1aϩ3 would be observed, whereas the slower desensitization of ASIC2a would dominate at lower pH values (pH 4.0 -5.0). It is of course entirely possible that other combinations, such as ASIC1a homomers as well as ASIC1aϩ2a and ASIC2aϩ3 heteromers, also contribute to the data observed. It should be noted, however, that the relatively modest amplitude of sustained current at pH 4.0 determined for ASIC1aϩ2aϩ3 (14%; Table I) seems to argue against the presence of large amounts of ASIC2aϩ3 heteromers.
Co-expression of ASIC1b with ASIC2a and ASIC3 gave very similar results, suggesting that expression of either ASIC1a or ASIC1b with ASIC2a and ASIC3 resulted in channel populations that were alike. Again assuming a tetrameric assembly of the ASIC channel (21), the apparent inability of channel formation involving three different subunits combined with the highly efficient channel formation with two different subunits could perhaps suggest an assembly consisting of two dimers.
ASIC Channels in DRG Neurons-At least three different types of acid-evoked currents have been recorded from DRG neurons (4 -6). Benson et al. (23) showed that co-expression of ASIC1aϩ2aϩ3 could reproduce the biophysical properties of a rapidly desensitizing acid-evoked current of mouse DRG neurons. Indeed, the fast desensitization at pH 6 observed by these authors was reproduced in the present study; however, as noted above, the rate of desensitization was reduced at lower pH values. In this context, it should be noted that the pH value giving half-maximal activation of current in the present study (pH 50 ϭ 5.1) was substantially less than the value (pH 50 ϭ 6.4) reported by Benson et al. (23), a fact which may relate to the use of different expression systems and/or other differences in experimental conditions.
Understanding the molecular composition of native ASIC channels is an important step in understanding their physiological role in sensory transduction and also has clinical implications, as these channels provide potential targets for the pharmacological modulation of sensory stimuli, including pain. However, identification of native ASICs is complicated by the lack of pharmacological tools, as only a single specific ASIC1a antagonist has been reported (16). By using a combination of non-steroidal anti-inflammatory drugs, which show different patterns of antagonism toward ASICs, Voilley et al. (17) showed that acid-evoked currents of DRG neurons are likely to reflect the presence of at least four different populations of ASIC channels. It may therefore very well be that the ASIC currents recorded from native DRG neurons reflect the concerted activity of multiple types of ASIC channels with distinct molecular compositions. This hypothesis seems to be compatible with the present ambiguous data obtained for cells coexpressing ASIC1aϩ2aϩ3, which also indicate that multiple channels form when these subunits are allowed to interact.