PAC, an evolutionarily conserved membrane protein, is a proton-activated chloride channel

See allHide authors and affiliations

Science  26 Apr 2019:
Vol. 364, Issue 6438, pp. 395-399
DOI: 10.1126/science.aav9739

A new protein channel family

Exposure of cells to acidic conditions outside the cell activates chloride-conducting channels that influence physiological and pathological processes. Yang et al. used an unbiased RNA interference screen to identify the channel protein that allows such proton-activated ion conductance. The protein, called PAC for proton-activated channel, has two predicted transmembrane domains and is unlike any previously identified channel. Identification of the channel should advance studies of its physiological roles, which range from tissue injury after stroke to adaptation of Tibetan natives to a high-altitude environment.

Science, this issue p. 395


Severe local acidosis causes tissue damage and pain, and is one of the hallmarks of many diseases including ischemia, cancer, and inflammation. However, the molecular mechanisms of the cellular response to acid are not fully understood. We performed an unbiased RNA interference screen and identified PAC (TMEM206) as being essential for the widely observed proton-activated Cl (PAC) currents (ICl,H). Overexpression of human PAC in PAC knockout cells generated ICl,H with the same characteristics as the endogenous ones. Zebrafish PAC encodes a PAC channel with distinct properties. Knockout of mouse Pac abolished ICl,H in neurons and attenuated brain damage after ischemic stroke. The wide expression of PAC suggests a broad role for this conserved Cl channel family in physiological and pathological processes associated with acidic pH.

Chloride is the most abundant free anion in animal cells and Cl channels have fundamental roles in physiology and disease (1). One type of Cl channel is activated by extracellular acidic pH. The acid- or proton-activated Cl (PAC) currents are present in a wide range of mammalian cells (2). Although the threshold to elicit ICl,H is relatively low, ~pH 5.5 at room temperature (RT), the channels become activated under less acidic conditions of ~pH 6.0 at 37°C (3, 4). Tissue pH can fall below 6.0 under many pathological conditions (57) such as ischemic stroke, in which acidosis is thought to be an important mechanism promoting neuronal death and brain injury (8, 9). By mediating Cl influx and subsequent cell swelling, the noninactivating ICl,H currents have been implicated in acidosis-induced cell death (4, 10). The PAC channel displays characteristic biophysical properties (2), including a strong outwardly rectifying current-voltage (IV) relationship (for this reason, the channel is also referred to as ASOR for “acid-sensitive outwardly rectifying anion channel”) (10), time-dependent activation at positive membrane potentials, and an anion permeability sequence of I > Br > Cl. Previous candidate gene approaches focusing on known Cl channel families have failed to identify the protein underlying ICl,H (2, 3, 11), so its molecular identity has remained elusive.

To search for genes encoding the PAC channel, we established a high-throughput cell-based fluorescence assay for ICl,H activity (12). Human embryonic kidney 293 (HEK293) cells stably expressing the iodide-sensitive yellow fluorescent protein (YFP) were stimulated with acidic pH solutions to elicit the endogenous ICl,H (13). I influx through the PAC channel triggered a rapid fluorescence decrease (Fig. 1A). The degree of fluorescence quenching correlated with the strength of the acidic stimuli (Fig. 1B). We selected pH 5.0 for the acid-induced YFP quenching assay and performed an unbiased RNA interference screen with an arrayed small interfering RNA (siRNA) library (four siRNAs per gene; one siRNA per well) targeting 2725 human proteins predicted to have at least two transmembrane (TM) domains (a characteristic shared by all known ion channels). Although most candidate genes had only one effective siRNA, a single previously uncharacterized gene, TMEM206 (transmembrane protein 206 or C1orf75), which we now call PAC (proton-activated Cl channel), stood out, with three of four siRNAs causing a decreased quenching response (>4 standard deviations above the mean) (Fig. 1C). We confirmed the attenuation of the quenching response and the reduction of the mRNA level by PAC siRNA transfection (Fig. 1, D to F).

Fig. 1 RNA interference (RNAi) screen identifies PAC as being essential for the PAC currents.

(A) Acid-induced and I-dependent fluorescence change (normalized to the baseline) in HEK293-YFP cells. (B) Acid-induced quenching response-to-pH relationship (n = 12 wells). Unapparent error bars are smaller than symbols. (C) Z-score (the number of standard deviations from the mean) of individual siRNAs (PAC siRNAs are in red) plotted according to the rank order. See supplementary materials and methods for details. (D) Acid-induced fluorescence change, (E) quenching response (n = 6 wells), and (F) PAC mRNA knockdown (n = 3 wells) in HEK293-YFP cells transfected with control or PAC siRNA. (G) Whole-cell currents induced by extracellular pH 4.6 in WT HEK293 cells. The transient inward current at −100 mV is blocked by 100 μM amiloride, an ASIC channel blocker. (H) ICl,H monitored by voltage-step (left) and voltage-ramp (right) protocols in WT HEK293 cells. (I) Whole-cell currents induced by extracellular pH 4.6 in PAC KO HEK293 cells. The normal amiloride-sensitive ASIC current is apparent at −100 mV. (J) ICl,H monitored by voltage-step (left) and voltage-ramp (right) protocols in PAC KO HEK293 cells. (K) Current densities at +100 mV induced by extracellular pH 4.6 in WT and two PAC KO HEK293 cells. KO1 was used throughout the study unless stated otherwise. Data points or bars represent mean ± SEM. **P < 0.01, ***P < 0.001 [two-tailed Student’s t test in (E) and (F); one-way analysis of variance (ANOVA) with Bonferroni post hoc test in (K)].

We used CRISPR-Cas9 technology to inactivate the PAC gene in HEK293 cells (fig. S1) and performed whole-cell patch-clamp recording to assay ICl,H. Perfusion of acidic solutions elicited large, stable conductances in wild-type (WT) HEK293 cells (Fig. 1G). The rapid activation and inactivation upon perfusion of different pH solutions suggested direct proton binding as the likely channel-gating mechanism (Fig. 1G). The currents exhibited time-dependent facilitation at positive membrane potentials and their IV relationships showed a steep outward rectification (Fig. 1H). The normalized current-to-pH relationship was fitted with a Hill equation, yielding a pH50 of 5.0 and a Hill coefficient of 4 (fig. S2A), indicating current activation by cooperative binding of protons. Furthermore, the currents were inhibited by 4,4ʹdiisothiocyano-2,2ʹ-stilbenedisulfonic acid (DIDS), a known blocker of ICl,H (fig. S2B). These features are characteristic of the widely described PAC channel. ICl,H currents were abolished in two independent PAC knockout (KO) HEK293 cell lines (Fig. 1, I to K). This effect appeared to be specific to the PAC channel because the transient inward currents at −100 mV mediated by the amiloride-sensitive acid-sensing ion channel (ASIC) were unaffected (Fig. 1I). HeLa cells also expressed a typical PAC channel, and deletion of PAC eliminated their ICl,H (figs. S1 and S3). We conclude that PAC is an indispensable component of the PAC channel or is necessary for its activation in human cells.

On the basis of hydrophobicity analysis, the human PAC (hPAC) protein is predicted to have two TM domains (Fig. 2A). Its plasma membrane localization was detected by immunostaining in nonpermeabilized cells expressing hPAC tagged with FLAG in the putative TM1–TM2 loop (Fig. 2B). By contrast, C-terminal FLAG was detected only upon cell permeabilization, consistent with a topology of N and C termini of PAC both facing the cytosolic side (Fig. 2, A and B). To test whether hPAC overexpression would rescue and enhance ICl,H, we transiently transfected hPAC cDNA in PAC KO HEK293 cells. Perfusion of acidic bath solutions evoked robust whole-cell currents several times larger than the endogenous ICl,H with the same properties: time-dependent facilitation and strong outward rectification (Fig. 2, C and D). The normalized current-to-pH relationship yielded a pH50 of 5.0 and a Hill coefficient of 4, which were identical to those of the native currents recorded at RT (Fig. 2E and fig. S2A). Like the endogenous ICl,H (3, 4), the pH sensitivity of hPAC was also temperature sensitive with a shift of pH50 (5.5) and activation threshold (~pH 6.0) to less acidic pH at 37°C (Fig. 2, E and F). Furthermore, hPAC was blocked by several ICl,H inhibitors, including DIDS, niflumic acid (NFA), and 5-nitro-2-(3-phenylpropylamino)benzoic acid (NPPB) (fig. S4). Substitution of extracellular Na+ with impermeable N-methyl-D-glucamine(NMDG+) did not alter ICl,H, but replacing extracellular Cl with the impermeable anion gluconate abolished the currents (fig. S5), indicating that hPAC confers an anion channel conductance. We estimated the permeability ratios from the shift in reversal potentials observed when equimolar amounts of I, Br, or Cl were the only permeable extracellular anion (Fig. 2G). The relative permeability (I > Br > Cl) of hPAC was essentially the same as that of native ICl,H (Fig. 2H). Therefore, the biophysical properties and pharmacological profile of hPAC fully match with those of the endogenous PAC channel in human cells.

Fig. 2 Human PAC recapitulates the properties of the endogenous PAC channel.

(A) Hydrophobicity plot (top) and predicted 2 TM topology (bottom) of hPAC. Red circles indicate FLAG insertion sites. (B) Nonpermeabilized and permeabilized HEK293 cells expressing FLAG-tagged hPAC were immunostained with antibody against FLAG. Scale bars: 10 μm. (C) Time courses (left) and current densities (right) of extracellular pH 4.6–induced whole-cell currents at +100 mV in vector- or hPAC-transfected PAC KO HEK293 cells. (D) hPAC-mediated currents at different pH values monitored by voltage-step (left) and voltage-ramp (right) protocols. (E) Normalized current-to-pH relationship of hPAC recorded at RT (n = 10 to 15 cells) or 37°C (n = 11 to 12 cells). Current at pH 4.6 and +100 mV is set to 1.0. (F) pH 6.0-induced hPAC current recorded at 37°C. (G) Representative IV relationship recorded in extracellular I, Br, or Cl pH 4.6 solution and (H) anion selectivity in WT (n = 10 cells) and PAC KO HEK293 cells expressing hPAC (n = 14 cells). Arrows in (G) indicate the reversal potentials. Data points or bars represent mean ± SEM. n.s., not significant [two-tailed Student’s t test in (H)].

PAC might assemble directly into a channel; however, it is also possible that PAC might regulate ICl,H activity indirectly, for example, by regulating channel activation or expression. To differentiate these possibilities, we used the substituted-cysteine accessibility method to probe channel-pore positions (14). Because it has a relatively low hydrophobicity (Fig. 2A), we tested whether the highly conserved TM2 of PAC might be part of the pore-lining domain. We generated cysteine substitution mutants of eight residues (amino acids 302 to 309) in the extracellular side of TM2 and found that four of them had normal current amplitudes (fig. S6, A and B). We tested the accessibility of these functional mutants by the thiol-reactive reagent 2-sulfonatoethyl methanethiosulfonate (MTSES). Currents mediated by one of the mutants (I307C, substitution of Ile307 by Cys) were rapidly blocked by the bulky MTSES, which were significantly reversed only upon application of the reducing agent dithiothreitol (DTT) (fig. S6, C to E), indicating a covalent modification of I307C by MTSES. To further analyze whether I307 is located near the putative pore, we measured the I to Cl permeability ratio of I307C and its alanine substitution mutant (I307A). Although I307C was normal, I307A showed a reduction of I permeability compared with WT hPAC (fig. S6, F and G). Therefore, PAC is an integral component of the PAC channel and I307 appears to be close to the ion-conducting pathway.

Highly similar homologs of PAC (TMEM206) are widespread among vertebrates (Fig. 3A), with more divergent homologs identified in the polychaete worm Capitella teleta and the sponge Amphimedon queenslandica. Zebrafish PAC (fPAC) shares 54% amino acid identity with hPAC and its overexpression in PAC KO HEK293 cells generated large ICl,H (Fig. 3B). fPAC had a pH50 (4.8) very similar to that of its human counterpart (5.0), indicating that pH sensitivity of PAC family proteins is conserved in vertebrates (Fig. 3C). However, the Hill coefficient of fPAC was close to 1, suggesting independent proton binding to the channel subunits instead of the highly positive cooperativity of hPAC (Fig. 3C). fPAC also exhibited altered kinetics: The time-dependent facilitation of fPAC was slower than that of hPAC (Fig. 3, D and E); its current decay after returning to the holding potential was also slower, resulting in prominent tail currents (Fig. 3B). Additionally, the preference of fPAC for I over Cl was decreased (Fig. 3, F and G). Therefore, fPAC generates ICl,H in human cells with channel gating and conducting properties distinct from those of hPAC. Together with the mutagenesis analysis (fig. S6), these results provide evidence that PAC forms the PAC channel pore directly.

Fig. 3 Zebrafish PAC encodes a PAC channel with distinct properties.

(A) Phylogenetic tree of PAC from different vertebrate animals (sequence similarity to human) generated with the ClustalW program. (B) fPAC-mediated currents at different pH values monitored by voltage-step (left) and voltage-ramp (right) protocols. Arrows point to the tail currents compared with those of hPAC (Fig. 2D). (C) Normalized current-to-pH relationship of hPAC (current at pH 4.6 is set to 1.0, n = 10 to 15 cells) and fPAC (current at pH 4.0 is set to 1.0, n = 5 to 12 cells) recorded at RT. (D) Examples and (E) time constant values (τ) of time-dependent facilitation at +20- to +100-mV voltage steps for hPAC (n = 7 cells) and fPAC (n = 11 cells). τ values were determined by fitting the currents with a single exponential function. (F) Representative IV relationship recorded in extracellular pH 4.6 containing I or Cl solution and (G) I versus Cl permeability for hPAC and fPAC. Arrows in (F) indicate the reversal potentials. Data points or bars represent mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 [two-way ANOVA with Bonferroni post hoc test in (E); two-tailed Student’s t-test in (G)].

To explore the role of PAC in acid-induced cell death, we treated HEK293 cells with acidic solutions. Treatment at pH 5.0, but not at pH 7.3, for 2 hours caused necrotic cell death in WT cells. By contrast, PAC KO cells were partially protected (fig. S7A). Overexpression of hPAC in WT cells increased their susceptibility to cell death induced by acid treatment (fig. S7B). These data suggest that PAC contributes to acid-induced cell death in HEK293 cells. Consistent with the wide presence of ICl,H, human PAC mRNA is expressed broadly in various tissues (15), with the highest amounts detected in the brain (Fig. 4A). Indeed, acid perfusion in the presence of amiloride induced the outwardly rectifying ICl,H in primary mouse cortical neurons (Fig. 4B). To determine whether Pac is essential for neuronal ICl,H, we generated Pac KO mice (fig. S8). The homozygous mutant mice were viable and appeared mostly normal, indicating that Pac is not required for animal viability. Cortical neurons isolated from Pac KO mice completely lacked ICl,H (Fig. 4, B and C), suggesting that Pac is essential for neuronal PAC channel activity. Pac KO neurons were partially but significantly protected from delayed cell death induced by 1-hour acid treatment (pH 5.6 and pH 6.0) (Fig. 4, D and E). To test whether ICl,H functions in acidosis-mediated neuronal toxicity in vivo, we subjected Pac KO mice to permanent middle cerebral artery occlusion (pMCAO), which induced severe ischemic brain injury. Compared with their WT littermates, Pac KO mice exhibited significantly smaller total brain infarct volume 1 day after pMCAO (Fig. 4, F and G). Their neurological scores were also consistently improved (Fig. 4H). Therefore, inhibition of the PAC channel appears to limit the pathogenesis of ischemic brain injury.

Fig. 4 PAC contributes to acid-induced cell death and ischemic brain injury.

(A) Transcripts per million (TPM) of PAC gene in selective human tissues from RNA-sequencing data (n = 2 to 7 biological replicates) (15). (B) ICl,H and (C) baseline-subtracted pH 4.6–induced currents at +100 mV in WT or Pac KO primary mouse neurons. Recordings were performed in the presence of 100 μM amiloride to block ASICs. (D) Images of neurons stained with Hoechst 33342 (blue) for nuclei of all neurons and with propidium iodide (red) for nuclei of dead neurons. Scale bar: 50 μm. (E) Percentage of acid-induced neuronal cell death (n = 8 wells, 1-hour acid treatment and 24-hour recovery in culture medium) (9). (F) Triphenyltetrazolium chloride staining (necrotic tissue in white), (G) total infarct volume (normalized to total volume of the ipsilateral hemisphere as 100%), and (H) neurological score 1 day after pMCAO (n = 12 and 13 for WT and Pac KO mice, respectively). Bars represent mean ± SEM. *P < 0.05, ***P < 0.001 [two-tailed Student’s t test in (E) and (G); Mann–Whitney U test in (H)].

We identified PAC, a previously uncharacterized membrane protein family with no sequence similarity to other ion channels, as the protein-activated Cl channel. Our findings provide the basis for elucidating the PAC-gating mechanisms by the simplest ligand, the proton, and for identifying channel modulators. Like ASIC channels (16), PAC may represent a potential drug target for stroke and other acidosis-associated diseases. The threshold for PAC activation is too acidic to be found commonly under physiological conditions except for the endocytic pathways and secretory vesicles, which have a luminal pH of 4.5 to 6.5. PAC may traffic from the plasma membrane to intracellular organelles (17) and regulate their function by modulating the luminal Cl concentration (18, 19). Exome sequencing of Tibetan highlanders identified PAC (TMEM206) as one of the genes with highest frequency change compared with those of ethnic Han Chinese individuals (20). The PAC genomic region is also under increased natural selection in pigs adapted to high altitude (21). Therefore, PAC could have a conserved role in the adaptation to hypoxia.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (2225)

References and Notes

Acknowledgments: We thank A. Patapoutian, W. B. Guggino, and K. W. Yau for helpful discussion; A. Orth for the siRNA library; Y. Zhou for assistance with RNAi data analysis; A. Dubin and S. Murthy for advice on electrophysiology; and M. Tian for help with primary neuron culture. This paper is dedicated to the memory of our dear friend and wonderful colleague, Dr. Maria del Carmen Vitery, who recently passed away. Funding: This work was supported by NIH grants R35 GM124824 (Z.Q.) and R01 NS107347 (S.S.); AHA fellowships 19POST34410020 (J. Chen) and 18PRE34060025 (J.O.-O.); and NIH fellowships F31 NS108658 (M.C.V.) and F32 AG059358 (H.Y.). Author contributions: Z.Q. conceived the study, performed the RNAi screen, generated PAC KO cells, and wrote the paper with J.Y. J.Y. performed electrophysiological recordings. J. Chen implemented the stroke model. M.C.V. generated Pac KO mice and conducted the neuronal cell death assay. J.O.-O. and S.S. carried out molecular cloning. J.O.-O. performed immunostaining and the HEK cell death assay. J. Chu performed neuronal recording. H.Y. performed ES cell electroporation. Competing interests: None declared. Data and materials availability: All data are available in the main text or the supplementary materials.

Correction (26 April 2019): Figure 4D has been corrected to improve the color contrast.

View Abstract

Stay Connected to Science


Navigate This Article