Identification of TMEM206 proteins as pore of PAORAC/ASOR acid-sensitive chloride channels

Acid-sensing ion channels have important functions in physiology and pathology, but the molecular composition of acid-activated chloride channels had remained unclear. We now used a genome-wide siRNA screen to molecularly identify the widely expressed acid-sensitive outwardly-rectifying anion channel PAORAC/ASOR. ASOR is formed by TMEM206 proteins which display two transmembrane domains (TMs) and are expressed at the plasma membrane. Ion permeation-changing mutations along the length of TM2 and at the end of TM1 suggest that these segments line ASOR’s pore. While not belonging to a gene family, TMEM206 has orthologs in probably all vertebrates. Currents from evolutionarily distant orthologs share activation by protons, a feature essential for ASOR’s role in acid-induced cell death. TMEM206 defines a novel class of ion channels. Its identification will help to understand its physiological roles and the diverse ways by which anion-selective pores can be formed.


INTRODUCTION
Chloride is by far the most abundant anion in animals. Its concentration can vary substantially between the extracellular space, the cytoplasm and various intracellular organelles, resulting in concentration gradients across membranes separating these compartments. Negatively charged chloride can cross biological membranes only with the help of membrane-spanning proteins such as Clchannels, which allow passive diffusion of Clalong its electrochemical gradient, or transporter proteins that couple the movement of Clto that of other ions and can thereby establish electrochemical gradients.
Chloride channels fulfill a broad range of biological functions, including the homeostasis of cell volume, vesicular acidification, transepithelial transport and cellular signaling [1,2]. Elucidation of these roles has been greatly facilitated by the molecular identification of the underlying channel proteins, a discovery process that began in the late 1980's and is still ongoing. Chloride channels are molecularly and structurally very diverse and lack a defining 'signature' pattern. More than six unrelated Clchannel families are known, prominently including voltage-regulated CLC channels [2,3], bestrophin [4,5] and TMEM16 [6][7][8] Ca 2+ -activated Clchannels, as well as LRRC8/VRAC volumeregulated anion channels [9,10]. However, several Clcurrents biophysically characterized in mammalian cells still lack molecular correlates, severely hindering the elucidation of their cellular and organismal functions.
We set out to identify the protein(s) mediating a widely expressed, strongly outwardlyrectifying plasma membrane Clcurrent I Cl,H that is detectable only upon marked extracellular acidification and which may play a role in acid-induced cell death [11][12][13][14][15]. The underlying channel is most frequently called ASOR (for Acid-Sensitive Outwardly Rectifying anion channel [14]), although the earlier name PAORAC (Proton-Activated Outwardly Rectifying Anion Channel [12,16]) provides a better description. At room temperature, ASOR currents are only observable when external pH (pH o ) drops below 5.5, but the threshold of activation shifts to ~6.0 at 37 o C [15]. Analysis of the steep pHdependence suggests that 3-4 protons are required to open the channel [12,13]. ASOR currents are strongly outwardly rectifying with almost no currents being observable at negative-inside voltages. This rectification results both from voltage-dependent gating that operates on a 100 ms time scale and from an outwardly-rectifying single channel conductance [12] (although others described that single ASOR channels are not rectifying [14]). ASOR currents (I Cl,H ) display an SCN ->I ->NO 3 ->Br ->Clpermeability sequence. The channel can be blocked by various Cltransport inhibitors such as DIDS (4,4'-diisothiocyano-2,2'-stilbenedisulfonic acid) and niflumic acid [12,13] and by other compounds such as phloretin [14] and pregnenolone sulfate (PS) [17]. However, none of these inhibitors is specific for ASOR.
The observation of I Cl,H currents in every investigated mammalian cell type [11][12][13][14][15][16][18][19][20][21] suggests that ASOR may be expressed in all tissues. Such a wide expression pattern indicates that this channel has important physiological functions. However, only few mammalian cells are physiologically exposed to an extracellular pH that is acidic enough to open ASOR. It was therefore proposed [14] that the channel rather plays a role in pathologies such as cancer or ischemic stroke in which pH o can drop to pH 6.5 or below [22][23][24][25][26]. Indeed, ASOR inhibitors blunted cell swelling and cell death provoked by prolonged exposure to acidic pH o [14,27], suggesting that ASOR-mediated chloride influx may worsen the outcome of ischemic stroke. It seems, however, counterintuitive that a channel that enhances cell death under pathological conditions confers an evolutionary advantage. Alternatively, ASOR might not only be present at the plasma membrane, but also in intracellular compartments such as lysosomes where it would be exposed to an appropriate acidic pH [12].
The molecular identity of ASOR had remained unknown, with several candidates such as LRRC8/VRAC anion channels or ClC-3 Cl -/H + exchangers having been excluded previously [12,15,28]. Using a genome-wide siRNA screen we now identified TMEM206 as essential ASOR component. TMEM206, which has two transmembrane domains, was necessary and sufficient for the formation of functional ASOR channels.
All tested TMEM206 orthologs from other vertebrates form acid-activated channels which, however, differ moderately in their biophysical properties. Ion-selectivity changing mutations suggest that TMEM206's transmembrane domains, in particular TM2, form ASOR's pore. Finally, partial protection of TMEM206 −/− cells from acidinduced cell death indicates that ASOR/TMEM206 channels play a detrimental role in pathological conditions that are associated with tissue acidosis.

Identification of TMEM206 as crucial ASOR component
To identify the protein(s) constituting ASOR, we performed a genome-wide siRNA screen using an optical assay for acute ASOR-mediated iodide influx into HeLa cells (Fig. 1A). Intracellular iodide was detected by fluorescence quenching of inducibly expressed, halide-sensitive and relatively pH-insensitive E 2 GFP [29]. We activated ASOR by exposing cells to acidic extracellular pH (5.0) and simultaneously depolarized their plasma membrane to maximize ASOR currents. To this end, we used an approach which we had used previously [30] to activate ClC-5, a Cl -/H + -exchanger that displays a similarly strong outward rectification. In HeLa cells we stably co-expressed E 2 GFP with FaNaC [31], a ligand-gated Na + channel from the snail Cornu aspersum that is closed until it is activated by the cognate neurotransmitter peptide Phe-Met-Arg-Phe-NH 2 (FMRFamide). FMRFamide application leads to Na + influx that acutely depolarizes the membrane to inside-positive voltages. Concomitant application of 100 mM iodide, pH 5.0 and 20 µM FMRFamide to our engineered HeLa cells induced fast quenching of E 2 GFP fluorescence (Fig. 1B). Control experiments omitting either iodide, acidic pH o , or FMRFamide indicated that a major component of quenching is owed to iodide influx through an acidic pH-and depolarization-dependent process such as ASOR (Fig. 1B, Suppl. Fig. S1A-D). This notion was further supported by applying the inhibitors pregnenolone sulfate and DIDS [17] (Suppl. Fig. S1E,F).
Having established a sensitive assay for ASOR function, we performed a genome-wide siRNA screen using a commercial library containing pools of four siRNAs per gene (Fig.  1C). The screen was carried out in triplicate and results were ranked according to the maximal slope of quenching. The top hit was TMEM206, a so far uncharacterized membrane protein (Fig. 1D). The wide tissue expression pattern of TMEM206 indicated by public databases (Fig. 1E) agreed with the finding that I Cl,H has been found in every cell studied so far [11][12][13][14][15][16][18][19][20]. TMEM206 has orthologs in vertebrates and in the hemichordate Saccoglossus kowalevskii (acorn worm) (Fig. 1F), but there are no homologs in simpler animals or other kingdoms of life. In contrast to many other ion channels, TMEM206 does not form a gene family because it lacks paralogs within the same species (i.e. it is a 'single gene').
After transfection of GFP-tagged human TMEM206 (hTMEM206) into HeLa cells a significant fraction of the protein was found at the plasma membrane ( Fig. 2A). Additionally, TMEM206 appeared to be expressed in intracellular membranes, which, however, might result from heterologous overexpression. Hydropathy analysis suggested the presence of two transmembrane spans (Fig. 2B). To determine the transmembrane topology of TMEM206, we tested the accessibility of added epitopes by immunofluorescence. Detection of both N-or C-terminally added GFP required permeabilization of the plasma membrane (Fig. 2C,D), suggesting that both the aminoand the carboxy-terminus of TMEM206 face the cytosol.
A cytosolic localization of the N-terminus is further supported by public databases showing that the amino-terminus of human and rodent TMEM206 can be phosphorylated (Fig. 2G). An extracellular localization of the TM1-TM2 stretch was suggested by the immunofluorescent detection in non-permeabilized cells of an HAepitope inserted at position 271 (Fig. 2E,G). The observation that this stretch is glycosylated further supports its extracellular localization (Fig. 2F). Only this segment displays consensus sites (N148, N155, N162, N190) for N-linked glycosylation (Fig. 2G, Suppl. Fig. S4). Deglycosylation of WT and mutant TMEM206(Δglyc), in which all four consensus sites were eliminated, demonstrated that at least one of these sites are used (Fig. 2F) and hence face the lumen of the ER during biogenesis. We conclude that both N-and C-termini of TMEM206 reside in the cytosol, whereas the TM1-TM2 loop faces the extracellular space (Fig. 2G). GFP and FRMFamide-gated Na + channel FaNaC were acutely exposed to an acidic solution (pH 5) containing 20 μ M FMRFamide and 100 mM I -. Iodide influx through ASOR is stimulated both by acidic pH and the depolarization caused by FaNaC-mediated Na + influx and induces quenching of E 2 GFP fluorescence. Fluorescence quenching is reduced by ASOR knock-down. (B) Assay verification under the conditions used for screening. Addition of I and FRMFamide at pH 5 (arrow) induces rapid quenching of fluorescence. Less rapid quenching upon omission of either I -, FMRFamide or acidic pH suggests that it is caused by Iinflux through ASOR, as further supported by inhibition by PS (pregnenolone sulfate) and DIDS (4,4'-diisothiocyano-2,2'stilbenedisulfonic acid) (Suppl. Fig. 1). (C) Fluorescence curves from a 384-well plate treated with siRNA against 280 genes, including TMEM206. Fluorescence quenching is specifically slowed by siRNA against TMEM206. (D) Distribution of median Z-scores (mean of 3 replicates) from filtered hits (see Methods). TMEM206 was the top hit. (E) Tissue expression of TMEM206 extracted from the GTEx database (https://gtexportal.org/home/; TPM, transcripts per million). (F) Dendrogram depicting similarity between TMEM206 orthologs from human (Homo sapiens), African naked mole-rat (Heterocephalus glaber), chicken (Gallus gallus), green anole lizard (Anolis carolinensis), zebrafish (Danio rerio) and from the hemichordate acorn worm (Saccoglossus kowalewski). Amino-acid sequence identity to human TMEM206 given in brackets. Dendrogram based on a Clustal Omega protein alignment fed into the Simple Phylogeny tool at EMBL-EBI (https://www.ebi.ac.uk/services).

Human TMEM206 mediates typical ASOR currents
Overexpression of human TMEM206 in HEK cells increased acid-activated Clcurrents about 10-fold (Fig. 3A,B). Addition of GFP to either the amino-or carboxy-terminus of TMEM206 only moderately decreased current amplitudes compared to the untagged construct (Fig. 3B). Currents from overexpressed hTMEM206, irrespective of whether fused to GFP, resembled native I Cl,H . To confirm that TMEM206 is essential for the generation of I Cl,H , we disrupted the TMEM206 gene in both HEK and HeLa cells using CRISPR-Cas9 genomic editing. To exclude off-target effects, we used three different gRNAs to generate independent KO clones. None of the clonal HEK KO cell lines displayed measurable I Cl,H currents ( Fig. 3C-E). Hence, TMEM206 is an indispensable component of ASOR. showed the typical outward-rectification and activation kinetics of ASOR [12][13][14] after depolarizing voltage steps (Fig. 3C,F). The pH o -dependence of native I Cl,H was indistinguishable from that of TMEM206-transfected KO HEK cells (Fig. 3I,H). The extent of acidification needed for half-activation of I Cl,H (pH o ~5.3) agrees with previous studies [12][13][14][15]. Measurements of reversal potentials of hTMEM206-transfected cells (Fig. 4A,B) revealed an SCN ->I ->NO 3 ->Br ->Clpermeability sequence that is typical for ASOR [12,13]. Replacement of extracellular Clby either gluconate - (Fig. 4C) or SO 4 2- (Fig. 4D) not only abolished outward, but also inward currents which are carried by anion efflux.
Since we had replaced extracellular, but not intracellular Cl -, these observations suggest that gluconateand SO 4 2-(or HSO 4 -, a minor species with which SO 4 2is at equilibrium and whose abundance will be increased by the acidic pH) block the outward movement of Clthrough ASOR/TMEM206 channels. Alternatively, ASOR may require activation by extracellular chloride. Native I Cl,H is inhibited by several compounds [12,13,17], but none of these is specific for ASOR. Acid-activated currents from overexpressed TMEM206 were efficiently blocked by DIDS, niflumic acid, and pregnenolone sulfate (PS, Fig. 4E-H). As described for native I Cl,H [17], the block by PS was fast and reversible ( Fig. 4F) and affected both outward and inward currents (Fig. 4G). We conclude that overexpressed TMEM206 shares all essential properties with native ASOR currents (I Cl,H ).

TMEM206 proteins from other species mediate moderately different I Cl,H currents
We asked whether TMEM206 proteins from other clades also function as ASORs and transfected orthologs from the green anole lizard (Anolis carolinensis), chicken (Gallus gallus), zebrafish (Danio rerio) and from the hemichordate acorn worm (Saccoglossus kowalewski) into TMEM206 −/− HEK cells. We also tested TMEM206 from naked molerats (Heterocephalus glaber), mammals displaying various unusual properties including insensitivity to acidic pain [32][33][34] and decreased acid-induced neuronal cell death [35]. Except for the evolutionarily distant hemichordate protein, which was retained in the endoplasmic reticulum (ER), all orthologs were expressed at the plasma membrane (Suppl. Fig. S2). All plasma-membrane expressed orthologs mediated robust acidactivated outwardly rectifying I Cl,H currents ( Fig. 5A,C). These currents, however, differed in detail. Strikingly, TMEM206 from the reptile Anolis carolinensis mediated outwardly-rectifying currents already at pH o 7.4. These currents were further steeply activated when pH o dropped below 6 ( Fig. 5A-D). Whereas currents from all orthologs activated at a threshold of ~pH o 6, their currents decreased differently at pH o < 5 (Fig.  5D). The current decrease at more acidic pH questions the reliability of determining Hill coefficients for the activation of ASOR by H + ions [12,13,36]. Differences between orthologs were also observed in depolarization-activated gating and current rectification ( Reversal potentials E rev of indicated orthologs with external NaCl (E), Nal (F) and Na 2 SO 4 (G). Significant currents with Na 2 SO 4 could be measured only for green anole and zebrafish.
Δ E rev , difference to E rev for NaCl. All currents measured at pH o 5.

Cysteine-scanning mutagenesis reveals pore-lining residues of TMEM206
To identify pore-lining residues of ASOR we used a systematic substituted-cysteine accessibility approach [37]. After having ascertained that the cysteine reagent MTSES lacks significant effects on human WT channels (Fig. 6A,B), we generated point mutants in which all residues of both hTMEM206 transmembrane spans were singly replaced by cysteines (Fig. 6B). Except for two mutants in TM2 (L309C and K319C), all constructs elicited acid-activated I Cl,H currents when transfected into TMEM206 −/− HEK cells (Fig. 6B,C, Suppl. Fig. S3A,B). However, current amplitudes and properties were often changed (Suppl. Fig. S3A,B). Only one mutant in TM1 (L84C) showed a marked response to cysteine modification (Fig. 6B,C). Application of MTSES to this mutant, which changes a residue close to the predicted external end of TM1, increased currents selectively at negative voltages ( Fig. 6B,C). TM2, which is less hydrophobic than TM1 (Fig. 2B) and may form an amphipathic α -helix (Suppl. Fig. S4A), contained a larger number of responsive cysteine-substituted residues (Fig. 6B,C). These residues were not only found close to the external end of TM2 (W304), but also close to the center (G312) and more towards the cytoplasmic end of TM2 (L315, A316) (Fig. 6B,C). These residues failed to clearly cluster on one side of the postulated α -helix (Suppl. Fig. S4A). Depending on the mutant, MTSES exposure increased or decreased I Cl,H amplitudes ( Fig. 6A-C) or changed current rectification ( Fig. 6B-C). Consistent with a covalent modification of cysteine residues, the effects of MTSES appeared irreversible (Fig. 6A). Strikingly, several cysteine mutants showed robust outwardly-rectifying currents already at pH o 7.4 (Fig. 6D,E). These could be further stimulated by exposure to acidic pH o (Fig.  6F).
Of note, some cysteine mutants displayed changed ion selectivity (Fig. 7A,B,F-H). Whereas changes in Ipermeability were relatively small (most evident for L84C and R87C, Fig. 7B,F), the same two mutants in TM1 displayed a small sulfate conductance that was undetectable in WT hTMEM206, allowing us to determine reversal potentials (Fig. 7A,B,G-H). Residues giving an interesting effect upon cysteine modification were further mutated to other, mostly charged, residues. This yielded detectable I Cl,H in many cases (Suppl. Fig. S3C), with a number of mutants displaying slightly changed iodide permeability (Fig. 7C,D,F) or marked increases in SO 4 2-/HSO 4 conductance (Fig. 7C,G-H). Strikingly, replacement of L315 by negatively charged aspartate introduced a large ~22 mV shift of the reversal potential in NaCl medium (Fig. 7D,E), indicative of a loss of anion selectivity. Together with the strong MTSES effects on the L315C mutant ( Fig.  6B,C) these data suggest that this TM2 residue importantly determines pore properties of ASOR. Of note, with the exception of R87, all functionally important residues identified here in TM1 and TM2 are highly conserved (Suppl. Fig. S4B). In conclusion, residues along the length of TM2 and towards the external end of TM1 may be close to, or line, the pore of ASOR channels.    Reversal potentials E rev indicated by arrows. With WT and L315D TMEM206 currents in Na 2 SO 4 were too small for E rev determination. Currents were elicited by voltage ramps as in Fig. 6C. (E) Reversal potential E rev with extracellular NaCl (***, p < 0.001 vs. WT; one-way ANOVA, Bonferroni correction). (F, G) Shift of reversal potentials (ΔE rev ) with extracellular NaI (F) or Na 2 SO 4 (G) relative to that measured with NaCl (*, p < 0.033 and ***, p < 0.001 vs. WT; one-way ANOVA, Bonferroni correction; NA, not applicable (currents not significantly above background or E rev not stable over time (I307C)). (H) Current densities at +80 mV with extracellular Na 2 SO 4 at pH 5.25.

Role of TMEM206/ASOR channels in acidotoxicity
ASOR was suggested [14,27] to play a role in acidotoxicity because DIDS and phloretin, compounds inhibiting ASOR but also several other channels or transporters, reduced acid-induced necrotic cell death of HeLa cells [14] and cultured cortical neurons [27]. We exposed both WT HEK cells, and two separately derived TMEM206 −/− clones, for 2 hours to acidic medium (pH o 4.5) in the absence and presence of PS. Cell death was assessed by propidium iodide staining and normalized to the total number of cells as determined by Hoechst 33342 labeling. Indeed, disruption of TMEM206 partially protected against acid-induced cell death (Fig. 8A,B).  *** *** *** *** A roughly similar protection was seen with PS in WT but not in TMEM206 −/− cells (Fig.  8A,B), suggesting that the effect of PS on cell viability results from its ability to block ASOR [17]. In conclusion, ASOR/TMEM206 channels enhance acidotoxicity.
This form of cell death had been attributed to cell swelling caused by osmotic gradients generated by ASOR-mediated Clinflux [14,27]. When exposed to pH o 4.2, both WT and TMEM206 −/− HEK cells increased their volume within the first three minutes as indicated by increased calcein fluorescence (Fig. 8C). Agreeing with the hypothesis [14,27] that the parallel opening of ASOR and ASICs (acid-sensitive Na + -channels [38,39]) leads to an electrically coupled influx of osmotically active Na + and Cl -, WT cells swelled faster than TMEM206 −/− cells. Whereas WT cells subsequently decreased their volume, TMEM206 −/− cells failed to shrink (Fig. 8C). Likewise, the ASOR inhibitor pregnenolone sulfate (100 µM) [17] slowed the initial swelling and inhibited the subsequent shrinkage (Fig. 8D). Similar effects of ASOR disruption or block were observed with HeLa cells (Suppl. Fig. S5). Hence, contrasting with the conclusion of Wang et al. [14], our data suggest that the role of ASOR/TMEM206 in acidotoxicity cannot be explained by a sustained volume increase of acid-exposed cells.

DISCUSSION
We have identified TMEM206 as essential, pore-forming subunit of ASOR, a widely expressed acid-sensitive outwardly rectifying anion channel. TMEM206 lacks homologous proteins within any given species. Both transmembrane domains of TMEM206, most importantly TM2, likely participate in forming its anion-selective pore. TMEM206/ASOR channels are apparently present in all vertebrates and share activation by markedly acidic pH as prominent feature. TMEM206/ASOR channels play a role in pathologies associated with a marked decrease in extracellular pH. Although their wide expression pattern across tissues and vertebrates suggests important physiological functions for all cells, these roles remain to be determined. The identification of TMEM206 as ASOR channel-forming protein in this work and in a recently published independent study by Qiu and coworkers [36] is an essential step forward towards the elucidation of these roles.

TMEM206 fully constitutes ASOR
The dissimilar biophysical characteristics of currents elicited by TMEM206 orthologs and the changes observed with mutants demonstrate that TMEM206 proteins constitute the pore of ASOR channels. With only two exceptions, all cysteine-substituted channels gave currents. However, their amplitudes were reduced by several mutations which were distributed along the length of either TM1 or TM2 (Suppl. Fig. S3). Whereas this result only shows that TMEM206 is (somehow) important for ASOR currents, the observed functional effects of MTSES on various cysteine mutants additionally indicate that the respective cysteines are accessible from the aqueous phase. If such residues are located in a membrane-embedded region, they are probably close to, or line, the pore. More convincing evidence that TMEM206 proteins form the channel comes from mutations that change intrinsic channel properties such as rectification or ion selectivity (a property of the pore). Several mutations, both at the extracellular end of TM1 and at various positions along TM2, moderately changed ASOR's Ipermeability. Moreover, whereas WT hTMEM206 lacked measurable currents with extracellular Na 2 SO 4 , TMEM206 from anole and zebrafish, and a surprisingly large number of mutants, displayed robust currents with extracellular sulfate. These mutations may markedly increase the channel's SO 4 2-(or HSO 4 -) permeability, or relieve a channel block that was indicated by the absence of inward currents (efflux of Cl -) in the presence of extracellular sulfate. The most striking evidence for TMEM206 forming ASOR's pore comes from the loss of anion selectivity with the L315D mutant. The ~22 mV positive shift of the reversal potential indicates a marked increase in cation conductance (E rev for Na + or H + are >+120 mV and +113 mV, respectively) that might be due to electrostatic interactions of the newly introduced negatively charged aspartate with the permeant ion. Our data suggest that the whole length of TM2, with a participation of the outer end of TM1, lines the pore of ASOR. This is reminiscent of crystallographically resolved pore structures of ASIC channels with which ASOR shares the transmembrane topology, but no obvious sequence homology. Intriguingly, ASICs and other ENaC/Deg channels display a glycine residue (in the 'GxS' motif) in TM2 that plays an important role in determining pore properties and gating [40,41]. Whether G212, which is located at a similar distances from the ends of TM2, plays a comparable role must await determination of TMEM206/ASOR channel structures.
Heterologous expression of all tested vertebrate orthologs led to large acid-induced currents. As shown with human TMEM206, biophysical properties of these currents did not differ significantly from native I Cl,H . Hence completely functional ASOR channels may only require TMEM206. However, our results cannot strictly exclude that TMEM206 lines the pore together with other, unknown proteins. This scenario would require that such proteins are produced in excess endogenously since overexpression of hTMEM206 or its orthologs elicited currents >10-fold larger than native I Cl,H . It further requires that such a putative additional pore-forming protein from HEK cells is able to associate with all the vertebrate orthologs tested. If ASOR's pore is exclusively constituted by TMEM206 proteins that cross the membrane only twice, TMEM206 must form at least a dimer to build a protein-enclosed pore. ASICs and other members of the DEG/ENaC channel family, which display a transmembrane topology similar to TMEM206, form trimeric homo-or heteromeric channels [42]. Likewise ASOR channels might be formed by TMEM206 trimers.

Role of acid-activated chloride current
With the exception of the hemichordate TMEM206, which shows marked sequence divergence and was retained in the ER, all tested TMEM206 proteins gave strongly outwardly rectifying Clcurrents that were activated by extracellular acidification in excess of ~pH o 6. The strong rectification might be physiologically irrelevant since all orthologs gave currents also at negative voltages and because inside-positive voltages are rarely reached in vivo, in particular in non-excitable cells. By contrast, the conserved activation by acidic pH o appears crucial and poses the question where and when such pH values might be reached. Mammalian ASORs require slightly less acidic pH for activation at physiological temperatures (37 o C) [15,27], but this consideration seems irrelevant for the reptile and fish orthologs.
Only a few cell types in the mammalian organism are physiologically exposed to an extracellular pH that is more acidic than pH 6.0. These include cells in the stomach and duodenum, renal medullary collecting duct [43] and acid-secreting osteoclasts and macrophages [44], but it is unclear whether ASOR is expressed in the relevant membranes of these polarized cells. Moreover, it is unlikely that an important function in a small subset of cells has resulted in an almost ubiquitous expression of ASOR during evolution. On the other hand, intracellular compartments such as endosomes and lysosomes, which can be acidified down to pH 4.5, would provide an ideal environment for ASOR activity. Here ASOR might facilitate luminal acidification by shunting currents of the proton-ATPase as described for endosomal CLC Cltransporters [45][46][47].
Interestingly, vesicular CLCs also show strong outward-rectification, but unlike ASOR are 2Cl -/H + -exchangers that are inhibited by luminal (or extracellular) acidification [2,30,47]. Hence presence of both ASOR and CLCs on endolysosomes might allow for a differential regulation of vesicular membrane voltage and ion concentrations. However, ClC-7 is believed to be the main Cltransporter of lysosomes [48,49] and immunocytochemistry showed prominent plasma membrane expression of all examined vertebrate ASOR orthologs. Although the variable cytoplasmic labeling observed upon TMEM206 overexpression may have resulted from overexpression and showed no obvious co-localization with the lysosomal marker lamp-1 in preliminary experiments, we feel that a more thorough investigation of an additional localization of ASOR in intracellular compartments may be warranted.
When speculating about biological roles of ASOR it is instructive to compare it to acidsensitive ASIC channels [38,50]. Besides being cation channels, ASICs differ from ASOR by their larger molecular diversity (four different subunits can assemble to various homo-and hetero-trimers), rapid inactivation and the degree of pH-sensitivity. Depending on the isoform, their pH 50 for activation ranges between ~4.8 and ~6.6 [39,50], values that are up to more than 1 pH unit more alkaline than that of ASOR (pH 50 ~5.3). Hence several physiological functions of ASICs are unlikely to have equivalents with ASOR. For instance, postsynaptic ASICs may enhance excitatory synaptic transmission by opening in response to a drop in pH in the synaptic cleft that is caused by exocytosis of the acidic contents of synaptic vesicles [39,51,52]. Even if ASOR would be expressed postsynaptically, the estimated drop in pH o [39] seems to be too small for its activation. ASICs also have roles in peripheral pain sensation [38,50]. If expressed on appropriate sensory neurons, such a function is imaginable also for ASOR because the high intracellular Clconcentration of peripheral neurons would allow for a depolarizing Cl --efflux [53]. Strikingly, naked mole-rats are insensitive to acid-induced pain, but their ASIC channels show normal pH-sensitivity [54]. We likewise found here that TMEM206/ASOR channels of that species display normal pH-sensitivity. The pH-insensitivity of mole-rats may rather be due to a changed pH-sensitivity of the Na + -channel Na v 1.7 [54,55].
Of note, ASIC channels are involved in various neurological pathologies that are associated with acidic pH o [50,52], including stroke [24] and multiple sclerosis [56]. Likewise, based on protective effects of non-specific ASOR inhibitors, Okada and colleagues proposed that ASOR plays a role in acid-induced cell death of epithelial cells [14] and neurons [27]. Using TMEM206 −/− cells, we now ascertained that genetic ablation of TMEM206/ASOR channels partially protects cells from cell death provoked by strongly acidic pH o (4.5). Interestingly, when compared to mice, cortical neurons of naked mole rats are resistant to acid-induced cell death (at pH 5.0) [35]. Since neither the pH-response of ASICs [54] nor of ASOR (this work) is changed in that species, the explanation for this resistance may be due to channel expression levels or other factors.
It has been proposed [14] that ASOR-dependent, acid-induced cell death results from sustained cell swelling owed to the parallel influx of Na + and Clthrough acid-activated ASIC and ASOR channels, respectively. Whereas in our experiments WT cells initially swelled faster than TMEM206 −/− cells, they subsequently recovered their volume and even shrank. Initial swelling may indeed be owed to the opening of both ASIC and ASOR, with ASIC-mediated Na + -influx mediating the depolarization required for passive Clinflux. In contrast to ASOR, however, ASIC channels inactivate [39], leading to a reversal of the Clelectrochemical potential, passive efflux of Clthrough ASOR, and subsequent osmotic cell shrinkage. Although there is now little doubt that TMEM206/ASOR channels foster acid-induced cell death, the underlying signal transduction cascade likely is complex.
Recently, while the present work was being completed, Yang et al. reported on their independent identification and characterization of TMEM206 as integral component of the ASOR channel (which the authors renamed PAC for Proton-Activated Channel) [36]. They likewise reported protective effects of TMEM206 disruption on acid-induced cell death of HEK cells or cultured primary cortical neurons. Intriguingly, these authors showed that Tmem206 −/− mice showed smaller stroke areas in a middle cerebral artery occlusion model [36], although the previously reported decrease of pH o in the stroke area (down to ~6.4) [26] is slightly above the value needed for ASOR activation. In a cysteine modification scan restricted to the extracellular half of TM2, they also found that reaction of I307C with MTSES reduces ASOR currents, and showed that the I307A mutant displayed moderately decreased iodide permeability. These results agree with our conclusion that the entire TM2 and the extracellular end of TM1 line ASOR's pore in an oligomeric complex.

Conclusion
We have identified TMEM206 as essential subunit of the acid-sensitive outwardly rectifying anion channel ASOR. TMEM206 proteins constitute the pore of ASOR, probably in a homo-oligomeric complex. TMEM206, which lacks paralogs in any given species, defines a structurally novel class of ion channels. TMEM206/ASOR channels are involved in acid-induced cell death, but this role in pathology cannot explain its wide and possibly ubiquitous expression across tissues and vertebrate species. Its molecular identification in this work and in the parallel study of Yang et al. [36] is an important step to identify its physiological roles and to mechanistically understand the diverse ways by which anion-selective channels can be formed.

HeLa-E 2 GFP-2A-FaNaC Cell Line used in the siRNA Screen
The T-REx™ system (Life Technologies) was used to generate a stable HeLa cell line inducibly co-expressing the halide-sensitive E 2 GFP variant (GFP S65T/T203Y) [29] and the FMRFamide-gated cation channel FaNaC [31] (UniProtKB Q25011). A pcDNA5/FRT/TO-based plasmid containing the cDNAs encoding E 2 GFP and FaNaC, separated by a self-cleaving 2A peptide was generated for this purpose. Resulting clones were selected using 200 µg/ml hygromycin B (Santa Cruz) and 4 µg/ml blasticidin (InvivoGen). Monoclonal cell lines were tested for robust expression of E 2 GFP and FaNaC by fluorescence microscopy and patch clamp electrophysiology, respectively. Clone E2F-5 was chosen for the genome-wide screen. Cells were kept in DMEM supplemented with 10% tetracycline-free Hyclone™ FCS (Fisher Scientific) and the above-mentioned antibiotics.

Genome-wide siRNA Screen
The genome-wide siRNA screen was performed at the FMP Screening Unit using the Dharmacon ON-TARGETplus™ human siRNA library (Horizon Discovery) arrayed in 66 384-well plates, targeting 18090 genes by four pooled siRNAs each. The screen was performed three times. Analysis and processing of data obtained in the genome-wide screen were performed with the KNIME Analytics Platform (KNIME). The maximal slope of fluorescence change was determined by linear regression of 10 points in a sliding window between 20 s and 100 s. Slope was normalized to the averaged baseline and background fluorescence.
For plate-wise normalization a Z-score was calculated from the normalized slope value using robust statistics. The median of the three replicate Z-scores per target gene was used to rank the data. As a measure of cell viability, an additional Z-score was calculated from the baseline fluorescence. We filtered out all candidates in which the median Baseline START Z-score was lower than -0.5, indicating a significantly reduced cell number.

Electrophysiology
Cells were plated onto poly-L-lysine-coated coverslips and transfected using the Xfect (Clontech) transfection reagent. Whole-cell patch-clamp recordings were performed 1 day after transfection using an EPC-10 USB patch-clamp amplifier and PatchMaster software (HEKA Elektronik). All experiments were performed at 20-22°C. Patch pipettes had a resistance of 1-3 MΩ in NaCl/CsCl-based solutions. The holding potential was -30 mV. Cells with a series resistance above 8 MΩ were discarded, and series resistance was not compensated for. Currents were sampled at 1 kHz and low-pass filtered at 10 kHz. In some cases, a notch filter was applied to current traces postrecording to minimize line noise.
MTSES was purchased from Biotium and stored at -20°C as powder. MTSES stock solutions (250 mM in water) were freshly prepared on each day and kept on ice until final dilution in recording solutions right before experiments.
The standard protocol for measuring the time course of acid activated currents, applied every 4 s after membrane rupture, consisted of a 100-ms step to -80 mV followed by a 500-ms ramp from -80 to +80 mV. The read-out for current amplitudes was the current at -80 mV and +80 mV normalized to cell capacitance (current density, I/C). The voltage protocol, applied after complete activation of ASOR, consisted of 1-s steps starting from -80 mV to +80 mV in 20-mV intervals preceded and followed by a 0.5-s step to -80 mV every 4 s.
Liquid junction potentials were only considered in ion selectivity experiments, where they were measured for all solutions and corrected for after recording.

Molecular Biology
Human TMEM206 was cloned from cDNA obtained from a human brain cDNA library (Clontech) using the following primers: Forward: AGAAGCTTCCACCATGATCCGGCAGGAGCGCTCCAC Reverse: GAGGATCCTCAGCTTATGTGGCTCGTTGCCTG HindIII and BamHI restriction sites on the primers were used to clone the PCR product into pcDNA3.1(+) (Invitrogen). Sequencing of the entire ORF confirmed the sequence was identical to NCBI RefSeq NM_018252.2. Subsequently, the cDNA was subcloned into other vectors such as pEGFP-N1/C1 using PCR-based strategies. Point mutations were introduced using the QuikChange Kit (Agilent). All clones were verified by sequencing the entire ORF. The following TMEM206 ortholog cDNA clones (human codon-optimized) were synthesized by Life Technologies using: Danio rerio TMEM206

Generation of Monoclonal Knockout Cell Lines
Cell lines with disruptions in the TMEM206 gene were generated using CRISPR-Cas9 as described previously [10]. Briefly, target sgRNAs were cloned into the pSpCas9(BB)-2A-GFP (PX458, gift from Feng Zhang, MIT) vector and transfected into HEK293 and HeLa cells. 2-5 days post-transfection single GFP-positive cells were FACS-sorted into 96-well plates containing pre-conditioned DMEM medium. Monoclonal cell lines were raised and tested for sequence alterations using target-site-specific PCR on genomic DNA followed by Sanger-sequencing. sgRNA sequences, genetic modifications and resulting protein mutations are listed for each clone in Table 1.

Immunocytochemistry
To determine the transmembrane topology of TMEM206, TMEM206 −/− HeLa cells (clone 3F8) were transfected with plasmids encoding GFP-TMEM206, TMEM206-GFP or TMEM206 containing a HA tag (YPYDVPDYA) after residue 271 using Lipofectamine 2000. 24 h after transfection, cells were fixed with 2% PFA for 10 min and permeabilized with 0.1% Triton X-100 as indicated. Cells were incubated with chicken anti-GFP (1/1000, Aves Lab) or rabbit anti-HA (1/1000, Cell Signaling) for 2 h and subsequently with secondary antibodies coupled to AlexaFluor633 for 1 h. For the nonpermeabilized conditions, no Triton X-100 was added to the antibody solutions. Images were acquired with a confocal microscope with a 63x NA1.4 oil-immersion lens (LSM880, Zeiss).

Acid-Induced Cell Death Assay
Acid-induced cell death of WT and TMEM206 −/− HEK293 cells (clones 2E5 and 3D9) was assessed by double staining with Hoechst 33342 and propidium iodide (PI). WT and KO cells were incubated for 2 h at 37°C with solution at pH 7.4 (HEPES) or 4.5 (citrate, see Electrophysiology). Experimental solutions were subsequently replaced with pH 7.4 solution containing Hoechst 33342 (1 µg/ml, ImmunoChemistry Technologies) and PI (5 µg/ml, Thermo Fischer) and cells were incubated for 15 min at 37°C. After washing, cells were fixed and analyzed using a confocal microscope (LSM880, Zeiss). For quantification, several 10x microscopic fields were randomly chosen. Percentage of PI positive cells over the total number of cells, identified with Hoechst 33342, was determined using Fiji.

Cell Volume Measurements
Cell volume was measured semi-quantitatively using the calcein fluorescence method [57]. HEK293 cells were plated 2 days before measurements in a 96-well-plate at a density of 45,000 cells per well, HeLa cells were plated 1 day before measurements at a density of 35,000 cells per well. Cells were loaded with 10 μ M calcein-AM (eBioscience) in DMEM for 1 h at 37°C. Subsequently, cells were washed 3 times with 100 μ l of a pH 7.4 (HEPES, see Electrophysiology) bath solution. After washing, each well contained 50 μ l bath solution. Following a 5-min incubation period the plate was transferred into Safire² Multimode Microplate Reader (Tecan) and fluorescence measurements at λ = 495 nm were initiated. After baseline recording for 30 s, 125 μ l of a pH 4.0 (citrate, see Electrophysiology) solution, or pH 7.4 bath solution were added to the wells resulting in a final pH of 4.2 or 7.4 respectively. Calcein fluorescence was measured every 10 s for 10 min. Wells containing cells of the respective cell line not loaded with calcein-AM (but otherwise treated equally) were used for background subtraction, and fluorescence values were normalized to t = 0 s (after the pipetting procedure).

AKNOWLEDGEMENTS
We thank Sabrina Kleissle for excellent help with the genome-wide siRNA screen, Marc Wippich for help with data analysis, and Judith von Sivers for technical assistance. This work was supported by the European Research Council (ERC) Advanced Grant VOLSIGNAL (#740537) to TJJ.

AUTHOR CONTRIBUTIONS
FU developed the genomic siRNA screening procedure, optimized, performed and evaluated it together with KL, and performed all electrophysiology experiments. SB performed IHC, topology, and cell death experiments. TD performed cell volume measurements. The siRNA screen was supervised by JPvK. TJJ planned and evaluated experiments and wrote the paper. The manuscript was edited by all co-authors.

CONFLICT OF INTEREST
The authors declare no conflict of interest.