The K⁺ channel K_IR2.1 functions in tandem with proton influx to mediate sour taste transduction

To determine whether sour taste cells respond electrically to intracellular acidification, we measured action potentials in cell-attached recording from freshly isolated taste receptor cells from mouse circumvallate papillae (CV). Sour taste cells were identified by expression of YFP from the Pkd2l1 promoter (PKD2L1 cells), and responses were compared with those obtained from nonsour taste cells, identified by GFP expression from the Trpm5 (transient receptor potential M5) promoter in a double-transgenic mouse (24, 25). Healthy, electrically excitable cells were identified using 2 mM Ba²⁺, which blocks resting K⁺ channels and elicits action potentials in both cell types (Fig. 1 A and B). To acidify the cell cytosol, we used acetic acid (AA) and propionic acid (PA), weak acids that partition in the lipid bilayer in their protonated form and are able to shuttle protons across the cell membrane (26–28). Each acid was used at a concentration of 100 mM and pH of 7.4, which generates ∼0.3 mM of the membrane permeable form. As can be seen in Fig. 1 A and B, the two weak acids evoked robust action potential firing in PKD2L1 cells, whereas no action potentials were elicited in response to the two weak acids in TRPM5 cells. As a control for removal of Cl⁻, we used methane sulfonic acid (MA), which is a strong acid and cannot shuttle protons across the cell membrane. MA was without effect on PKD2L1 cells (Fig. 1 A and B). We also measured responses from PKD2L1 cells isolated from foliate (FL) papillae on the sides of the tongue, and compared the results with those obtained using cells isolated from the CV (Fig. S1); no difference in responses were detected; thus, all additional experiments were done using cells from the CV. We conclude that intracellular acidification is sufficient to activate taste cells that express PKD2L1 and mediate sour taste, but not taste cells that express TRPM5 and mediate bitter, sweet, and umami taste.

Intracellular acidification could generate membrane depolarization either by activating excitatory, Na⁺- or Ca²⁺-permeable, channels, or by inhibiting K⁺ channels (3, 28). We previously tested whether weak acids could activate an inward Na⁺- or Ca²⁺-permeable current in PKD2L1 cells and failed to find any difference in the magnitude or reversal potential of the inward current evoked in response to pH 5 with or without acetic acid (16). We also tested whether the channel complex formed from PKD2L1/PKD1L3 contributes to the response to weak acids (29). Cells isolated from Pkd1l3^−/− animals responded vigorously to 100 mM AA and 100 mM PA (pH 7.4), and no difference in response intensity could be detected between knockout animals and wild-type (WT) littermates (two-way ANOVA, P = 0.37; Fig. 1 C and D). Because PKD1L3 is required for membrane trafficking of the PKD2L1 (9), we conclude that PKD2L1/PKD1L3 does not contribute to the depolarization in response to intracellular acidification.

We next tested whether, instead, cytosolic acidification had an effect on the resting K⁺ currents in PKD2L1 cells, measured by elevating extracellular K⁺ to 50–140 mM. The inward K⁺ current (at −80 mV) was unaffected by lowering the pH of the extracellular solution (to pH 6.0), but was strongly inhibited by 20 mM acetic acid, pH 6.0, a condition that acidifies the cell cytosol (27) (Fig. 2 A and B). The block by acetic acid developed over tens of seconds and recovered over a similar time course (Fig. 2A), as expected if the protonated form of the acid must enter the cell to block the channel.

To identify candidates to mediate the resting K⁺ current in PKD2L1 cells, we analyzed the transcriptome of lingual epithelium containing circumvallate papillae and compared it with the transcriptome of nontaste epithelium (NT) (Fig. 2C). We focused on inward rectifier channels (K_IR family) and two-pore domain K⁺ channels (K_2P), based on our preliminary results, and known functions of these channels in mediating the resting potential in other cell types. Altogether, we detected moderate-to-high expression of eight K_IR channels and five K_2P channels in taste tissue (Fig. 2C); four additional K_2P channels were detected at low levels and were not pursued further. The eight K_IR channels (K_IR2.1, 2.2, 3.1, 3.4, 1.1, 4.2, 6.1, and 6.2) included two that were previously identified in taste cells: K_IR1.1 (ROMK1), which is expressed by type I taste cells (30), and K_IR6.1, a component of a K_ATP channel, expressed by type II taste cells (31). Among the five K_2P channels (Task2, Trek1, Twik1, Twik2, and Twik3), we observed particularly high expression of Twik1 (KCNK1) and lower expression of Task-2, which were previously reported to be expressed in taste tissue (22, 23). As a control, we examined expression of Pkd2l1 and Trpm5; both were highly and specifically expressed in taste tissue, as expected (Fig. 2C).

We next used pharmacological profiling to determine which among the 13 K_IR and K_2P channels mediates the resting K⁺ current in PKD2L1 cells. Of the K⁺ channel blockers tested (32, 33), the resting K⁺ current in PKD2L1 cells was sensitive only to 1 mM Ba²⁺ and 1 mM Cs⁺ (Fig. 2 D and E; one-way ANOVA, P < 0.0001). Notably, the current was insensitive to quinine (Fig. 2E), excluding contributions from Twik1-3 (sensitivity to Cs⁺ excludes the remaining K_2P channels). The current was also insensitive to tertiapin-LQ (targets K_IR1.1 and K_IR3.1/3.4), glibenclamide (targets K_IR6.1 and 6.2), tetraethylammonium (TEA), 4-aminopyridine (4-AP), and linopiridine (target Kv and KCNQ channels) (Fig. 2E). Of the channels expressed in taste tissue, only the inward rectifiers K_IR2.1, K_IR2.2, and K_IR4.2 have this pharmacological profile.

To narrow down the candidates further, we assessed the sensitivity of the resting K⁺ current in PKD2L1 cells to block by Ba²⁺ and Cs⁺ (Fig. 3 A–D). Block of the current by Cs⁺ was strongly voltage dependent (Figs. 2D and 3 A and B) with an IC₅₀ of 211 ± 35 μM measured at −80 mV. A similar sensitivity to Cs⁺ was measured for heterologously expressed K_IR2.1 and 4.2 (IC₅₀ values of 169 ± 8 and 194 ± 19 μM, respectively), whereas heterologously expressed K_IR2.2 showed a significantly higher sensitivity to Cs⁺ (118 ± 9; P < 0.05 by one-way ANOVA followed by Tukey’s multiple-comparison test). Sensitivity to Ba²⁺ was even more informative. Ba²⁺ blocked the K⁺ current in PKD2L1 cells with an IC₅₀ of 2.1 ± 0.4 μM (measured at −80 mV), which was not significantly different from the IC₅₀ for inhibition of K_IR2.1 (1.4 ± 0.2 μM; Fig. 3 C and D), but considerably different from the IC₅₀ of either K_IR2.2 or K_IR4.2 (0.23 ± 0.06 and 5.8 ± 0.8 μM, respectively; P < 0.0001 and P < 0.01 by one-way ANOVA followed by Tukey’s multiple-comparison test). Finally, we tested the K_IR2-specific blocker ML133, which has a reported IC₅₀ of 1.9 μM for K_IR2.1 (34). ML133 (50 μM) blocked the resting K⁺ current in PKD2L1 cells by ∼90%, similar to its effect on K_IR2.1 and K_IR2.2, whereas K_IR4.2 was virtually insensitive to ML133 (Fig. 3 E and F). Moreover, we observed an increase in the rate of inhibition by ML133 when the pH of the extracellular solution was raised to 8.5, as previously described for K_IR2.1 (Fig. S2) (34). Thus, based on sensitivity to Ba²⁺, Cs⁺, and ML133, we conclude that the resting K⁺ current in PKD2L1 cells is mediated by K_IR2.1.

Sensitivity to intracellular acidification is not a well-documented feature of K_IR2.1 as it is for other K_IR and K_2P channels (35, 36). We therefore tested whether heterologously expressed K_IR2.1 channels could be blocked by intracellular acidification. Indeed, K_IR2.1 currents were potently inhibited by 20 mM acetic acid, pH 6, and to a similar degree as currents in PKD2L1 cells (Fig. 4A; compare with Fig. 2A), but not by extracellular acid alone (Fig. 4 A–C). Moreover, in excised patches of cells heterologously expressing K_IR2.1 channels, we found that exposure of the cytoplasmic surface to acidic pH (pH <6) produced a rapid and partially reversible block of channel activity (Fig. 4D). Thus, among K⁺ channels expressed in taste epithelium, K_IR2.1, with a pharmacological profile nearly identical to that of the resting K⁺ current in PKD2L1 cells, including a similar sensitivity to intracellular acidification, is a strong candidate for mediating the resting K⁺ current of PKD2L1 cells.

To further determine whether K_IR2.1 accounts for the resting K⁺ current in PKD2L1 cells, we measured currents in cells isolated from mice carrying a targeted deletion of the gene. Because knockout of K_IR2.1 is neonatal lethal (37) and it is difficult to identify sour taste cells by expression of YFP at birth, we chose to use a tissue-specific gene deletion strategy. To this end, we used a mouse strain carrying a floxed allele of Kcnj2 (Methods) and crossed it to a mouse strain in which the Pkd2l1 promoter drives expression of Cre recombinase. Based on use of a floxed Tdt reporter, Cre is expected to be active in ∼79% of the Pkd2l1-expressing cells, and only in Pkd2l1-expressing cells in the circumvallate papillae (Fig. 5A and Fig. S3). We confirmed that Kcnj2 was inactivated in taste tissue using a PCR strategy designed to detect the deletion event (Fig. S4).

The resting K⁺ current in animals carrying two floxed alleles of Kcnj2 and one allele of Pkd2l1-Cre (cKO) was compared with that in littermates that carried either the floxed allele or the Pkd2l1-Cre driver alone. As shown in Fig. 5 B–D, the average magnitude of the resting K⁺ current in the cKO animals was significantly reduced compared with either control group (one-way ANOVA followed by Tukey’s post hoc test, P < 0.001 compared with Cre⁺ and P < 0.01 compared with Kcnj2^fl/fl). We also measured sensitivity of the current to 1 μM Ba²⁺, a concentration that is expected to produce ∼50% block of K_IR2.1. Ba²⁺ produced a substantial (>15%) block of the inward K⁺ current in all cells isolated from WT mice (18 of 18), whereas in a significant proportion cells from the cKO, the inward current was Ba²⁺-insensitive (4 of 18; P < 0.05 by one-tailed χ² test; Fig. 5E). These cells tended to be the ones with the smallest inward current, and thus likely represent cases where the Kcnj2 gene was completely excised and the K_IR2.1 protein eliminated. In the remaining cells, the observation that the residual current retained sensitivity to Ba²⁺ indicates that it is not a product of a compensatory increase in the expression of a different subtype of ion channel, but instead likely represents incomplete elimination of the Kcnj2 gene product. Thus, tissue-specific knockout of Kcnj2 significantly reduces the magnitude of the resting K⁺ current in PKD2L1 cells and eliminates the current in a subset of cells, lending support to the conclusion that K_IR2.1 mediates this current.

These data argue that, in response to weak acids, inhibition of K_IR2.1 by intracellular acidification produces membrane depolarization that drives action potentials in PKD2L1 cells. To directly test this hypothesis would require replacing K_IR2.1 with an acid-insensitive variant, which is not currently possible. Instead, we reasoned that there should be a correlation between sensitivity to weak acids and expression of K_IR2.1, such that a cell that is insensitive to weak acids should not express K_IR2.1 or a similar acid-sensitive K⁺ channel. TRPM5 taste cells are insensitive to weak acids (Fig. 1), and we therefore measured the resting K⁺ current in these cells. To our surprise, we found that the resting K⁺ current in TRPM5 cells was inhibited by exposure to weak acids (Fig. 6A), to a similar extent to that of the current in PKD2L1 cells. Moreover, among known blockers of K⁺ channels tested, the resting K⁺ current in TRPM5 cells was sensitive to only Cs⁺ and Ba²⁺, like the resting K⁺ current PKD2L1 cells (Fig. 6B; compare with Fig. 2A), indicating that it is likely mediated by K_IR2.1. To directly test this hypothesis, we generated a cell type-specific knockout of Kcnj2, this time using the Advillin promoter (38), which drove expression of Cre in ∼71% of TRPM5 cells (Fig. 7A and Fig. S5). We then measured the resting K⁺ current in cKO and control cells (Fig. 7 B and C). Although there was no change in the average magnitude of the current (Fig. 7D), the Ba²⁺-sensitive resting K⁺ current was eliminated in a subpopulation of cKO cells (5 of 24) but not in any control cells (0 of 13; P < 0.05; one-tailed χ² test) (Fig. 7E). This supports the interpretation that K_IR2.1 contributes to the resting K⁺ current in TRPM5 cells.

We next determined whether TRPM5 cells and PKD2L1 cells acidified to the same extent in response to the weak acid stimuli. Freshly isolated cells were loaded with the pH-sensitive dye phrodo red, and fluorescent intensity was monitored under the same conditions used to measure action potentials in cell-attached recording. As we previously reported, PKD2L1 cells had an initial starting pH that was acidified compared with that of TRPM5, a phenomenon that can be attributed to proton leak through ungated proton channels in these cells (21). There was, however, no difference in the final pH reached in response to the weak acid solutions between cell types (pH 6.47 ± 0.08 in TRPM5 cells versus 6.34 ± 0.05 in PKD2L1 cells, P = 0.43 by two-way ANOVA followed by Sidak’s multiple-comparison test; Fig. S6).

Thus, TRPM5 cells express a resting K⁺ channel sensitive to intracellular pH and show a drop in intracellular pH in response to weak acids, similar to that in PKD2L1 cells. Why then do they not fire action potentials to weak acids? In contemplating this question, we noticed that there was one important difference between the resting K⁺ currents in PKD2L1 and TRPM5 cells—the K⁺ currents were severalfold larger in TRPM5 cells compared with PKD2L1 cells (Fig. 8 A and B). The larger K⁺ current is expected to more effectively clamp the resting potential of TRPM5 cells, reducing sensitivity to K⁺ channel blockers such as intracellular acidification. To test this hypothesis, we used a low concentration of Ba²⁺ (10 μM) to reduce the magnitude of the resting K⁺ current in the TRPM5 cells to the level observed in PKD2L1 cells (Fig. 8 A and B). We then asked whether under these conditions TRPM5 cells would fire action potentials to weak acids. Indeed, when primed with 10 μM Ba²⁺, which alone did not stimulate action potentials, the TRPM5 cells fired action potentials to weak acid stimulation (Fig. 8 C and D). These data are consistent with the hypothesis that the acid-sensitive K_IR2.1 channels mediate responses to weak acids in PKD2L1 cells; the small magnitude of this current, relative to other conductances, and not its identity, restricts acid sensitivity to these cells.

The pH sensitivity of K_IR2.1 makes it well suited to be part of the sour transduction machinery, acting to transduce cellular acidification into a change in membrane potential. Intracellular acidification can be elicited in PKD2L1 cells when protons enter through an apical conductance. In the previous experiments, we took safeguards to prevent proton entry through the proton conductance, which would have generated an increase in the inward current and confounded interpretation of the results. Thus, whenever the extracellular solution was acidified, we added Zn²⁺ (2 mM) to block the proton current (16, 21). To determine whether entry of protons through the proton conductance could block the resting K⁺ current in PKD2L1 cells, we simply measured the response to extracellular acidification in the absence of Zn²⁺. As seen in Fig. 9 A and B, the resting K⁺ current was inhibited when the pH of the extracellular solution was lowered to 6.0 in the absence of Zn²⁺ (48 ± 7% at 120 s), but not in its presence (13 ± 5%). A similar block of the K⁺ current in TRPM5 cells was not observed (Fig. 9A), consistent with the observation that these cells do not have a proton current (16, 21).

These data suggest that there are two components to the sensory response to sour stimuli: H⁺ entry through a proton channel and intracellular acidification, which blocks resting K⁺ channels. To determine whether the two components act synergistically, we tested the response to two subthreshold stimuli: mild extracellular acidification (pH 7.0) and a low concentration of acetic acid (0.1 mM of protonated acid; Methods) either alone or in combination. Neither alone was sufficient to stimulate action potentials, but together they strongly activated PKD2L1 cells (Fig. 9 C and D). At more acidic pH (6.6), action potentials were evoked even in the absence of the weak acid as we previously reported (21), and addition of the weak acid did not produce a significant change in firing rate.

Together, these results suggest that protons serve not only as charge carriers to generate membrane depolarization but also as second messengers that amplify the primary sensory response (Fig. 10).

Block of K⁺ channels that are open at the resting membrane potential (resting K⁺ channels) has previously been suggested as a potential mechanism for sour transduction, although the specific elements of such a pathway have been uncertain. Using the mudpuppy, it was found that extracellular protons block apically located K⁺ channels (41, 42), but this mechanism does not appear to apply to other more conventional model systems. More recently, K⁺ (K_2P) channels, which are sensitive to either intracellular or extracellular protons and thus are well suited to play a role in sour transduction were identified in rodent taste cells (22, 23). However, although electrophysiological properties of each type of taste cell have been described (43, 44), before our study, very little was known about the molecular nature of the K⁺ channels that set the resting potential in taste cells and their sensitivity to intracellular or extracellular acidification.

Our data show that there is an acid-sensitive resting K⁺ current in sour cells, but that this current is not mediated by the K_2P channel Twik1/KCNK1 (22, 23). Although we found that KCNK1 was expressed at significantly higher levels than other K_2P or K_IR channels, we were unable to detect a current with the pharmacological profile of KCNK1. We calculate that KCNK1 contributes no more than 1% to the resting K⁺ current (Fig. S7).

Instead, pharmacological and molecular profiling indicate that K_IR2.1 channels mediate the resting, acid-sensitive K⁺ current in sour taste cells. This is supported by the observation that the resting K⁺ currents in PKD2L1 cells from a tissue-specific K_IR 2.1 KO mouse are significantly reduced in magnitude, and in a subset of PKD2L1 cells, the K⁺ current is completely eliminated. We did not expect to observe a complete elimination of the current in all cells because a reporter for the driver showed activity in only 70% of PKD2L1 cells. The same argument applies to TRPM5 cells, where the promoter that we used to drive expression of Cre, is effective at most in ∼70% of cells and fewer if one considers only cells that show strong expression of the reporter (∼50%). Moreover because the Pkd2l1 Cre driver is expressed by mature neurons, it is only expected to excise the floxed allele at a time when there may already be significant expression of ion channels; under these circumstances, elimination of the current would only be observed if the existing protein has turned over and both alleles of Kcnj2 were inactivated. Given the short lifetime of taste cells (∼22 d) (45), the likelihood that all these events occur before a cell dies is low. Nonetheless, the significant reduction in the resting K⁺ current in the tissue-specific KO mice, together with our pharmacological and expression profiling data, provide strong evidence that K_IR2.1 is required for the resting K⁺ current in sour taste cells.

K_IR2.1 is an unexpected player in sour taste transduction, as the channel was not previously known to be expressed in taste cells or to be associated with sensitivity to intracellular acidification. Notably, K_IR2.1 is much less sensitive to intracellular acidification than other inward rectifiers, such as K_IR2.3 and K_IR1.1 (35). Nonetheless, K_IR2.1 can be inhibited at a sufficiently low intracellular pH (≤5.5; our data and ref. 35). Moreover, we find that K_IR2.1 can be blocked in whole-cell recording by weak acids that activate sour taste cells. Under these conditions, the block is slow because the concentration of the protonated form of the weak acid is low, and protons can diffuse out of the cell and into the pipette. We speculate that the reduced sensitivity of K_IR2.1 to intracellular pH may be advantageous in the context of the sour taste cell, which is acidified at rest (21). Indeed, we do not know the extent of intracellular acidification that occurs in intact taste cells during sensory stimulation, but assuming that proton entry occurs through channels located on sensory microvilli, it is likely that local concentrations of protons could be very high (46). Under these conditions, a less sensitive channel, like K_IR2.1, would be better able to sense dynamic changes in intracellular pH.

Our experiments show that, among taste cells, only those that mediate sour taste respond to intracellular acidification with action potentials. However, quite unexpectedly this response cannot be attributed to the expression of a highly specialized ion channel whose distribution is restricted to sour taste cells. Instead, sensitivity to intracellular acidification is conferred by a relatively ubiquitous ion channel, K_IR2.1. K_IR2.1 (IRK1/KCNJ2) was the first member of the family of inward rectifier ion channels to be identified (47), and it is expressed broadly in the brain, heart, vascular smooth muscle cells, skeletal muscle, and throughout the body (47). Our data also indicate that K_IR2.1 mediates the resting K⁺ current in TRPM5 cells, which are unresponsive to weak acids that stimulate action potentials in PKD2L1 cells. In an attempt to reconcile these findings, we found one way in which the resting K⁺ currents in TRPM5 and PKD2L1 cells differed substantially—in their magnitude. As previously reported (44), the resting K⁺ currents in TRPM5 cells are severalfold larger than those of PKD2L1 cells. This difference in the magnitude of the K⁺ current can completely account for the difference in sensitivity to intracellular acidification between the two cells types as inhibition of the K⁺ currents in TRPM5 cells to the level observed in PKD2L1 cells rendered the TRPM5 cells responsive to intracellular acidification. Thus, a smaller K⁺ current, as found in PKD2L1 cells, makes the cells more responsive to intracellular acidification.

To gain a quantitative appreciation of this phenomenon, we consider two hypothetical cells that vary only in the magnitude of their resting K⁺ current. The first cell (representing the PKD2L1 cell) has a mixed cation conductance of 0.25 nS and a resting K⁺ conductance of 1 nS. A resting potential of −67 mV is calculated from the equation: V_m = [(G_K*E_K) + (G_C*E_C)]/(G_K + G_C), where G_K and E_K are the K⁺ conductance and equilibrium potential, G_C and E_C are the cation conductance and cation equilibrium potential, and the K⁺ concentration is assumed to be 5 mM outside the cell and 140 mM inside it (48). The same equation can be used to show that it would be necessary to block 60% of the K⁺ current to reach threshold for firing action potentials (assumed to be −50 mV). Now consider another cell (a TRPM5 cell), with a similar resting cation conductance, but a threefold larger K⁺ conductance (3 nS). The resting potential of this cell is calculated to be −77 mV. Block of 50% of the current will depolarize the cell to −72 mV. In this cell, it is necessary to block 87% of the current to depolarize the cell to a threshold of −50 mV. Thus, one can see that, in moving from a cell with a small resting K⁺ current to one with a larger current, the fractional block of the current required to reach threshold has increased substantially, from 60% to 87%. If one further assumes a single-site binding site model for block of the channel, this translates into a need for a ∼4.5-fold higher concentration of intracellular protons to bring the TRPM5 cell to threshold compared with the PKD2L1 cell (Methods). Thus, without considering other conductances, the smaller K⁺ current in PKD2L1 cells is expected to make the cells much more sensitive to intracellular acidification compared with TRPM5 cells.

It is also worth noting that, among a variety of cells tested, only PKD2L1 cells have an inward proton conductance, which can act upstream of the acid-sensitive K⁺ channel (21). Thus, in other cells in the taste bud, the only way that the cytosol can be acidified is by penetration of weak acids. Although these factors make it unlikely that in vivo TRPM5 taste cells will fire action potentials to acids, we cannot rule out the possibility that intracellular acidification might modulate the response of TRPM5 cells to bitter, sweet, and umami stimuli. Moreover, because we looked only at the aggregate behavior of TRPM5 cells, it is possible that a small subset of TRPM5 cells responsive to one taste quality could show heightened sensitivity to intracellular acidification, and possibly even fire action potentials under some conditions.

It was noted nearly a century ago that at the same pH, organic weak acids evoke a more intense sour sensation than strong mineral acids (18). This has been attributed to the ability of weak acids, in the protonated form, to penetrate the cell membrane, and upon entering the cell, release a proton (28). Cytosolic acidification could then affect membrane excitability and/or transmitter release. Consistent with this hypothesis, a previous study found that the gustatory nerve response was better correlated with the ability of sour stimuli to acidify the cytosol, rather than the absolute concentration of free protons they contained (5). Similar results were found in recordings from a slice preparation of lingual epithelium, where focal application of sour stimuli elicited calcium responses in taste cells that correlated with the degree of intracellular acidification (20) To our knowledge, our experiments are the first to directly show that intracellular acidification alone, in the absence of extracellular acidification, increases cellular excitability in identified sour taste cells. For these experiments, we used two organic weak acids, acetic acid and propionic acid, which due to their small size and relatively high pK_a values (4.76 and 4.88, respectively) can acidify the cell cytosol when used at neutral pH (27). Application of these acids at neutral pH caused a similar degree of intracellular acidification in both TRPM5 and PKD2L1 cells, but only PKD2L1 cells fired action potentials. However, it is worth noting that, in the intact epithelium where taste buds are surrounded by a relatively impenetrable barrier, weak acids may not enter taste cells unimpeded (49). Alternatively, weak acids may be effective sour stimuli because they buffer pH, protecting protons from absorption by salivary proteins. The extent to which penetration of weak acids into taste cells and the buffering capacity of weak acids contributes to the intensity of sour taste perception remains to be determined.

The ability of K_IR2.1 to respond to an elevation of protons makes it well suited to participate in sour transduction, downstream of the previously characterized Zn²⁺-sensitive proton conductance (Fig. 10). Indeed, we find that the K_IR2.1 current in sour taste cells can be blocked by extracellular acidification, when protons are allowed to enter the cell through the Zn²⁺- sensitive conductance. Block of resting K⁺ channels would reduce the extent of proton entry needed to depolarize the sour taste cell to threshold, and thereby enhance the detection of small signals. This may explain how taste cells can respond to bath acidification as mild as pH 6.7, which produces a proton current of only a few picoamperes (21). Moreover, it is possible that K_IR2.1 channels are localized in close proximity to proton channels, facilitating interaction between the two molecules.

Amplification is a common theme in sensory biology and is usually associated with signaling by G-protein–coupled receptors that activate a cascade of intracellular events (50, 51). Our proposed mechanism, on the other hand, is most similar to the mechanism of amplification observed in the olfactory system where calcium entering through cyclic-nucleotide gated ion channels acts on chloride channels, which further depolarize the cell membrane (52, 53). This mechanism allows the cell to respond to odorants under circumstances where extracellular ions are limited. Similarly, the use of an intracellular amplification step for sour taste transduction is expected to enhance signaling while limiting the need for excessive proton entry that might be cytotoxic.

Taken together, our experiments provide evidence for the participation of the inward rectifier K⁺ channel, K_IR2.1, as a key element in sour taste transduction. Notably, we find that K_IR2.1 is blocked by intracellular acidification and acts downstream of proton channels to enhance electrical excitability. Working in tandem, the proton conductance and K_IR2.1 channel constitute an elegant mechanism for amplification of the sensory response to sour stimuli.

All methods of mouse handling were approved by the University of Southern California Institutional Animal Care and Use Committee. Mouse strains Pkd2l1-YFP, Pkd1l3^−/− were previously described (14, 16). Mouse strains Kcnj2^fl/fl, Pkd2l1-Cre are described in SI Methods. Avil-Cre mice were a kind gift from the Peter Heppenstall Laboratory, European Molecular Biology Laboratory, Monterotondo, Rome (38).

Taste cells were isolated from adult mice (6–10 wk old) as previously described (16, 21). In brief, Tyrode’s solution containing 1 mg/mL elastase (Worthington), 2.5 mg/mL dispase II (Roche), and 1 mg/mL trypsin inhibitor (Sigma-Aldrich) was injected between the epithelium and the muscle of the isolated tongue, which was incubated for 20 min in Tyrode's solution bubbled with 95% O₂/5% CO₂. The epithelium was then peeled off from the tongue. The small piece of epithelium containing the papillae was further trimmed and incubated in the enzyme mixture for 20 min at room temperature. The reaction was stopped by washing the tissue in Ca²⁺-free saline. Single cells were isolated by trituration in Tyrode’s solution with a fire-polished Pasteur pipette and were used within 6 h.

Taste cells were prepared for patch-clamp recording as previously described (16, 21). Data were collected with an Axopatch 200B amplifier using pClamp software suite (Axon Instruments), and time course measurements and I–V curve were graphed with Origin 6 (OriginLab). For cell-attached recording, we selected for cells that were healthy and able to fire action potentials, using 2 mM Ba²⁺, which elicited action potentials more reliably in both PKD2L1 and TRPM5 cells compared with high K⁺ (16). In addition, we selected for cells with intact apical processes, based on the flask-like shape of the cell, and presence of apically located YFP. Firing rates were measured within the interval of 0.4 to 4 s following application of stimuli unless otherwise stated. The evoked frequency was calculated after subtraction of the background frequency, measured during the 10 s before application of the stimulus. Note that the amplitude of the spikes measured in a single cell varied during recordings; this can be attributed to changes in the availability of Na⁺ channels, which may not fully recover from inactivation when resting K⁺ channels are blocked to changes in the seal resistance, which can affect spike amplitude. For whole-cell recording, the membrane potential was held at −80 mV, the voltage was ramped from −80 to +80 mV (1 V/s), and the current was measured at −80 mV. For testing the dose dependence of inhibition by Ba²⁺ and Cs⁺ of the resting K⁺ current (Fig. 3 A–D), series resistance compensation was performed.

Solutions for electrophysiology were exchanged using a piezo-driven stepper motor system (SF-77B; Warner Instruments). The pipette solution in all experiments except those shown in Figs. 2A, 4A, 6A, and 9A contained the following (in mM): 120 KAsp, 7 KCl, 8 NaCl, 10 Hepes, 2 MgATP, 0.3 GTP-Tris, 5 EGTA, 2.4 CaCl₂ (∼100 nM free Ca²⁺), pH 7.4 with KOH. For the experiments shown in Figs. 2A, 4A, and 6A, to minimize contributions due to changes in intracellular Ca²⁺ as a result of competition between H⁺ and Ca²⁺ for binding to EGTA, a Ca²⁺-free pipette solution was used containing the following (in mM): 120 KAsp, 20 KCl, 10 Hepes, 2 MgATP, 0.3 GTP-Tris, 0.5 EGTA, pH 7.3; in Fig. 9A, Hepes concentration was 10 μM. Tyrode’s solution contained the following (in mM): 140 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, 10 Hepes, and 10 d-glucose, pH 7.4. Sodium solution was as follows (in mM): 150 NaCl, 10 Hepes, and 2 CaCl₂, pH 7.4. For recording action potentials, the tested solutions contained the indicated concentration of acid (in mM), which replaced equimolar Cl⁻, and the pH was adjusted with NaOH. In Fig. 9 C and D, the pH 7.4, pH 7.0, and pH 6.6 solutions have 43.7, 17.5, and 7 mM overall acetic acid, respectively, to generate a final concentration of 0.1 mM protonated form. For whole-cell recordings, 50 mM KCl (Figs. 2E, 3 E and F, 5, 6B, 7, 9 A and B), 100 mM KCl (Fig. 3 A–D and 8 A and B), or 140 mM KCl (Figs. 2 A and D, and 6A) and labeled concentrations of TEA replaced equimolar NaCl in the extracellular solution. For solutions of pH ≤6.6, Mes replaced Hepes. In excised patch, the pipette solution was as follows (in mM): 150 KCl, 2 CaCl₂, 10 Hepes, pH 7.4. The extracellular solution was 150 KCl, 10 Hepes or 10 Mes, 1 EDTA, 1 EGTA, pH 7.4 with KOH or HCl.

Glibenclamide and Tertiapin-LQ, bupivacaine, and linopirdine were purchased from Tocris; ML133 was a gift of Craig Lindsley, Vanderbilt University, Nashville, TN. All other chemicals used for pharmacological profiling were from Sigma-Aldrich.

Bar and summary scatter plots were prepared in Prism (GraphPad Software). Significance was determined with Student’s t test or ANOVA as indicated. In all cases, data represent the mean ± SEM, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Heterologous expression, mouse genetics, immunocytochemistry, transcriptome analysis, intracellular pH imaging, and other experimental details are described in SI Methods.

Mouse handling, genotyping, and methods of killing were approved by the University of Southern California Institutional Animal Care and Use Committee. Mouse strains Pkd2l1-YFP, Pkd1l3^−/− and Trpm5-GFP were previously described (14, 16, 25). Mouse strains Kcnj2^fl/fl, Pkd2l1-Cre are described below. Avil-Cre mice were a kind gift from the Peter Heppenstall Laboratory, European Molecular Biology Laboratory, Monterotondo, Rome (38). For the validation of the cell-specific expression of Cre recombinase, the two strains were each bred into a Rosa26^Tdt strain (The Jackson Laboratory; stock 007908) (54), and immunocytochemistry (see below) was used to identify the Td-Tomato–expressing cell population. To monitor expression of Cre from the Pkd2l1 promoter in nonsour taste cells, the Pkd21l-Cre::Rosa26^Tdt strain was bred into a Trpm5-GFP strain (25). To monitor expression of Cre from the Avil promoter in sour taste cells, the Avil-Cre::Rosa26^Tdt was bred into a pkd211-YFP strain (16).

Gene targeting was used to insert an IRES-Cre transgene immediately after the stop codon of Pkd2l1 gene in the mouse genome. The targeting vector was constructed by cloning a 5.2-kb genomic fragment encompassing the last exon of Pkd2l1 gene into pBluescript and inserting IRES-Cre immediately after the stop codon. A floxed-neomycin resistance cassette was cloned at a site immediately downstream of the IRES-Cre. The targeting vector was linearized by XmaI digestion and electroporated into mouse ES cells (EC7.1 line). Recombinant ES cells were selected by G418 resistance in culture, and 282 clones were isolated for PCR screening to identify correct targeting events. Two correctly targeted clones were isolated and were used for microinjection to generate germ-line chimeras, and through standard breeding, the F₁ generation of mice. To monitor expression in nonsour taste cells, the Pkd21l-Cre::Rosa26^Tdt strain was bred into a Trpm5-GFP strain (25), and confocal images were taken, as shown in Fig. S3C.

Floxed Kcnj2 mice were generated by the M.T.N. Laboratory (University of Vermont, Burlington, VT) and will be described elsewhere in more detail. In brief, the second exon of Kcnj2 was flanked by LoxP sites and a downstream Neo cassette was flanked by FRT sites. The neo cassette was removed by breeding to a FLP-deleter mouse (The Jackson Laboratory). To generate a Kcnj2 sour taste cell-specific knockout in a background of Pkd2l1-YFP, the Pkd2l1-Cre and Pkd2l1-YFP strains were each bred with the Kcnj2^fl strain and the offspring were further bred to generate Pkd2l1-YFP, Pkd2l1-Cre, Kcnj2^fl/fl progeny. To generate a knockout of Kcnj2 in TRPM5 cells in a background of Trpm5-GFP, the Avil-Cre strain and Trpm5-GFP strain were each bred with the Kcnj2^fl strain, and the offspring were further bred to generate Trpm5-GFP, Avil-Cre, Kcnj2^fl/fl progeny. Genotyping methods for Pkd2l1-YFP strain were previously described (16). The primers for detecting Cre, Tdt, and Kcnj2^fl are as follows: Cre (GCACTGATTTCGACCAGGTT, GAGTCATCCTTAGCGCCGTA; 420-bp product), Rosa26^Tdt (CTGTTCCTGTACGGCATGG, GGCATTAAAGCAGCGTATCC; 196-bp product), Kcnj2 distal loxP (TCCGATGACACTGAGAACCTGGA, TGGATGCTTCCGAGAACCTTGG; WT allele: 520 bp; floxed allele: 600 bp), Kcnj2 neo deletion (CTGACTGAACACACAGGTCCAGGG, GGGACCATCAAGCCCTGGTAATGG; floxed allele: 800 bp; WT allele: 600 bp).

K_IR2.1 was kindly provided by Zhe Lu, University of Pennsylvania, Philadelphia; K_IR2.2, K_IR4.2 were amplified from mouse skeleton muscle and kidney cDNA. The amplified fragments were inserted into pcDNA3 vector with the In-Fusion method (Clontech). KCNK1 was mutated to increase membrane trafficking (KCNK1-I293A,I294A; KCNK1-AA) (55) and was fused to GFP by removing the stop codon and inserting the gene upstream in frame with the GFP coding sequence. Clones used for recording were fully sequenced (Genewiz).

HEK 293 cells were cultured with DMEM containing 10% (vol/vol) FBS and 0.05% gentamicin. K_IR2.1, K_IR2.2, and K_IR4.2 were cotransfected with GFP (ratio 20:1), whereas KCNK1-GFP and KCNK1-AA-GFP were transfected alone, using TransIT-293 Transfection Reagent (Mirus Bio). Cells were lifted with trypsin-EDTA 24 h after transfection and used for recording within 6 h. All cell culture reagents were from Life Technologies.

We assumed a single-site binding model for block of the K⁺ channel by intracellular protons:Solving this equation for the case of the PKD2L1 cell where I_blocked/I_initial = 0.60 yields a value of [H⁺]_i = 1.5*IC₅₀ and solving for the case of the TRPM5 cell where I_blocked/I_initial = 0.87 yields a value of [H⁺]_i = 6.7*IC₅₀. The ratio of the two is 4.5.

We rearranged a set of linear functions assuming that the Rb⁺ and K⁺ inward currents are generated solely from the contributions of KCNK1 and K_IR2.1, and that the ratio of Rb⁺ to K⁺ permeability was preserved. Rearranging to solve for IK_KCNK1, the potassium current contributed by KCNK1, results in the following equation:where IK_taste is the resting K⁺ current amplitude in PKD2L1 taste cells, and r is the ratio of Rb⁺ to K⁺ current for KCNK1, K_IR2.1, or PKD2L1 taste cell as indicated in subscript.

A small region of the tongue containing circumvallate papillae and control regions devoid of taste buds was microdissected, and RNA was extracted with TRIzol (Life Technologies). Library construction and sequencing were carried out at the Norris Cancer Center Next Gen sequencing core. RNA was visualized on a Bio-Rad Experion to ensure RNA Quality Index values were greater than 8. For library construction, the Illumina TruSeq, version 2, mRNA kit was used. The protocol was followed according to manufacturer’s instructions except the final number of PCR cycles was 12 and not 15. Following construction, libraries were visualized by Bioanalyzer (Agilent) using the High Sensitivity Chip and quantified for pooling and sequencing using Kapa Biosystems qPCR quantitation kit according to manufacturer’s instructions. For sequencing, libraries were diluted to 16 pM, and then applied to a V3 flowcell using the Illumina cBot according to manufacturer’s instructions. Sequencing was carried out on the HiSeq 2000 using HSCS, version 1.5.15.1. Image analysis and base calling were carried out using RTA 1.13.48.0, and deconvolution and fastq file production were performed with CASAVA 1.8.2. Reads were mapped to the mouse genome (mm9) summed and converted to reads per kilobase per million reads (RPKM) using the equation: RPKM = [# of mapped reads]/([length of transcript]/1,000)/([total reads]/10^6).

For measuring the expression of the PKD2L1-Cre reporter, circumvallate taste tissue was incubated in 4% paraformaldehyde for 1.5 h before being cryoprotected in 20% sucrose solution at 4 °C overnight. Tissue was embedded in OCT at −60 °C and sectioned at 12 µm. Slides were blocked with 2% normal donkey serum in buffer containing 0.3% Triton for 1 h before being incubated in primary antibody overnight at 4 °C and secondary antibody for 3 h at room temperature (RT). Slides were mounted with Fluoromount-G (SouthernBiotech) and imaged on a Leica confocal microscope. Hiroaki Matsunami, Duke University, Durham, NC generously provided the primary anti-PKD2L1 1:500, which was previously described (9). All z-stack images were acquired on a Leica confocal. To quantify immunohistochemical images, we defined a cell as a distinct shape featuring: (i) a clear nuclear region and (ii) an elongation along the pore-to-base axis of the taste bud. We thresholded images using a modified Otsu method (56) and counted positive cells by eye, tracing each through the image stack via optical dissection.

For measuring expression of the Avil-Cre reporter, circumvallate papillae from triple transgenic Avil-Cre::Rosa26^Tdt::Pkd2l1-YFP mice were dissected and incubated in 4% paraformaldehyde for 30 min at RT. The tissue was cryoprotected in 30% sucrose at 4 °C overnight, embedded in OCT at −80 °C and sectioned at 14 μm. Immunohistochemistry was performed by permeabilizing the tissue with 0.1% Triton X-100 for 15 min, blocking with BlockAid (Life Technologies) plus 0.1% Triton X-100 for 1 h and incubating in primary antibody overnight at 4 °C or RT for 1 h and secondary antibody at RT for 1 h. Slides were mounted with Fluoroshield (Sigma-Aldrich). The primary antibodies were chicken anti-GFP (1:1,000) (Aves Labs), guinea pig anti-TRPM5 (1:500) (24); the secondary antibodies were goat anti-chicken Alexa Fluor 488 and goat anti-guinea pig Alexa Fluor 546 (Life Technologies). All z-stack images were acquired on an Olympus confocal. To quantify immunohistochemical images, two observers independently scored cells as Tdt+, TRPM5+, or PKD2L1-GFP+.

The pipette solution contained the following (in mM): 120 KAsp, 7 KCl, 8 NaCl, 10 Hepes, 2 MgATP, 0.3 GTP-Tris, 5 EGTA, 2.4 CaCl₂ (∼100 nM free Ca²⁺), pH 7.4 with KOH. The standard bath solution contained the following (in mM):130 NaCl, 20 TEA-Cl, 10 Hepes, 2 CaCl₂, 1 MgCl₂, pH 7.4. Rb solution contained the following (in mM): 130 RbCl, 20 TEA-Cl, 10 Hepes, 2 CaCl₂, 1 MgCl₂, pH 7.4. K solution contained the following (in mM): 130 KCl, 20 TEA-Cl, 10 Hepes, 2 CaCl₂, 1 MgCl₂, pH 7.4. CsCl was added at the concentrations indicated.

Dissociated taste cells were loaded with the intracellular pH indicator pHrodo Red AM using PowerLoad concentrate accordingly to the manufacturer’s instructions (Molecular Probes). TRPM5-GFP and PKD2L1-YFP cells were distinguished as previously described (21). During pH imaging, cells were excited every 5 s with light of 560 nm, and emission at 630 nm was detected by a Hamamatsu digital CCD camera attached to an Olympus IX71 microscope using a U-N31004 Texas Red/Cy3.5 filter cube (Chroma Technologies). After exposure to MA and PA (both pH balanced to 7.4 using NaOH), intracellular pH was calibrated with high K⁺ solution containing the following (in mM): 140 KCl, 4.5 NaCl, 2 CaCl₂, 1 MgCl₂, 10 glucose, 10 Mes or 10 Hepes, and 0.02 nigericin (Molecular Probes), pH adjusted with NaOH to pH 6.0, 6.7, or 7.4. PHrodo Red fluorescence intensity at baseline, when exposed to MA, and when exposed to PA was converted to pH_i using linear calibration curves constructed for each cell using fluorescence intensity data in the presence of calibration solutions under the assumption that pH_i = pH_o.