The solute carrier SLC9C1 is a Na+/H+-exchanger gated by an S4-type voltage-sensor and cyclic-nucleotide binding

Voltage-sensing (VSD) and cyclic nucleotide-binding domains (CNBD) gate ion channels for rapid electrical signaling. By contrast, solute carriers (SLCs) that passively redistribute substrates are gated by their substrates themselves. Here, we study the orphan sperm-specific solute carriers SLC9C1 that feature a unique tripartite structure: an exchanger domain, a VSD, and a CNBD. Voltage-clamp fluorimetry shows that SLC9C1 is a genuine Na+/H+ exchanger gated by voltage. The cellular messenger cAMP shifts the voltage range of activation. Mutations in the transport domain, the VSD, or the CNBD strongly affect Na+/H+ exchange, voltage gating, or cAMP sensitivity, respectively. Our results establish SLC9C1 as a phylogenetic chimaera that combines the ion-exchange mechanism of solute carriers with the gating mechanism of ion channels. Classic SLCs slowly readjust changes in the intra- and extracellular milieu, whereas voltage gating endows the Na+/H+ exchanger with the ability to produce a rapid pH response that enables downstream signaling events.

S olute carriers (SLC), only second to GPCRs, form one of the largest gene families in vertebrates, comprising about 450 members in the human genome. Yet, compared to other gene families, SLCs are understudied and many isoforms represent orphan proteins, highlighting our ignorance 1 . A case in point is the subfamily SLC9C1, also referred to as sNHE. SLC9C1 has been suspected to serve as Na + /H + exchanger that controls intracellular pH (pH i ) in mammalian sperm 2 .
Changes in pH i are key to sperm signaling [3][4][5][6][7][8][9][10][11] , but it is not known if SLC9C1 indeed promotes Na + /H + exchange, how its activity is controlled, and whether it contributes to pH i regulation. Disrupting the mouse slc9c1 gene renders sperm immotile and male mice infertile 2,12 , demonstrating that SLC9C1 is required for sperm function and fertilization in mammals. However, the pH i of SLC9C1 −/− sperm is not altered 2 , and a clear-cut conclusion is compounded by the unexpected observation that cAMP synthesis is impaired in SLC9C1 −/− sperm and that the motility defect can be rescued by cAMP 12,13 . These observations suggest that the prime defect of SLC9C1 −/− sperm might be in cAMP-rather than pH i signaling. Finally, mouse SLC9C1 is non-functional in heterologous systems and attempts to study SLC9C1-mediated Na + /H + exchange by pH i fluorimetry in mouse sperm were unsuccessful 2 . Thus, the function of SLC9C1 as a Na + /H + exchanger and its role in pH i regulation of mammalian sperm is left in limbo.
On a different yet related note, in sea urchin sperm, chemoattractants stimulate a rapid rise of pH i 9,14-20 , which serves as a switch to activate the pH-sensitive CatSper Ca 2+ channel that controls chemotaxis 9 . The molecule underlying this alkalinization is not known. However, a Na + /H + exchange mechanism was described that is activated by hyperpolarization rather than by changes in the extracellular pH or Na + concentrations 14,15,17 . Ever since its discovery, the activation of Na + /H + exchange by voltage and the underlying mechanism and molecules have remained unexplained. We report here final success in solving these fundamental and long-standing questions. We demonstrate that the Strongylocentrotus purpuratus homolog (SpSLC9C1) exists in sperm and represents a genuine Na + /H + antiporter. Unlike solute carriers, Na + /H + exchange by SpSLC9C1 is gated by voltage via a voltagesensing domain (VSD) and directly modulated by cAMP via a cyclic nucleotide-binding domain (CNBD). Thus, we deorphanize the SLC9C1 family and identify Na + /H + exchange as a target for cAMP signaling and a mechanism of adaptive interaction between pH i and cAMP. On a broader perspective, our results now enable future studies of the commonalities and differences of voltage sensing and cAMP modulation between ion channels and a solute carrier and, thereby, gain insight into the evolution of gating mechanisms.

Results
Overall protein topology of SLC9C1. Sperm-specific Na + /H + exchangers share with SLC9 family members the exchanger domain that carries substrates across membranes. In addition, SLC9C1 holds a putative voltage-sensing domain (VSD) and a putative cyclic nucleotide-binding domain (CNBD) that are absent in other SLC9 members (Fig. 1a). The exchange domain of SpSLC9C1 is predicted to encompass 14 transmembrane segments (TM), whereas bacterial and archaeal Na + /H + exchangers feature 12 and 13 TMs, respectively [21][22][23] . A sequence alignment illustrates that SpSLC9C1 carries an additional TM at the Nterminal end ( Fig. 1a and Supplementary Figure 1). The Na + -binding site in archaeal SLC9 forms a trigonal bipyramid 22,23 . In Methanocaldococcus jannaschii NhaP1, two carboxyl groups (D132 and D161) and a hydroxyl group (S157) form a triangle around Na + ; the main-chain carbonyl of T131 is positioned at one bipyramid tip and T76/E154 at the other tip (Fig. 1b). All but one of the important Na + -coordinating residues are conserved in SpSLC9C1, including an Asn/Asp motif (Asn237/Asp238) that is diagnostic for electroneutral exchangers 24 (Fig. 1a, b). Residues occupying the (T76/E154) pyramid tip are less conserved, even among archaeal SLCs. Furthermore, two Arg residues in TMs 12 and 13 that are functionally important in other Na + /H + exchangers are conserved in SpSLC9C1 ( Fig. 1a and Supplementary Figure 1).
The full-fledged VSD (four transmembrane domains S1-S4) of SpSLC9C1 carries seven conserved Arg or Lys residues in the S4 motif and four conserved Glu or Asp residues in S1-S3 (Fig. 1c). Finally, the CNBD features the hallmarks of a cyclic nucleotidebinding fold: three α-helices (αA, αB, and αC), eight β-strands (β1-β8), and a phosphate-binding cassette (PBC) (Fig. 1d). Key residues that interact with cyclic nucleotides are conserved, including the purine-binding residues Val and Leu (β4 and β5), the ribofuranose-binding residues Gly/Glu (β6), the phosphatebinding Arg between β6 and β7, and the purine-binding residues Arg/Lys in αC 25,26 . The presence of exchanger, VSD, and CNBD domains suggests that SLC9C1 promotes Na + /H + exchange controlled by voltage and cyclic nucleotides. We studied by heterologous expression of SpSLC9C1 its voltage sensitivity, Na + /H + exchange activity, and regulation by cyclic nucleotides.
The voltage-sensing domain produces gating currents. For electrophysiological experiments, we used CHO cells stably expressing HA-tagged SpSLC9C1. An anti-HA antibody stained sheets of plasma membrane (Supplementary Figure 2), showing that SpSLC9C1 reaches the cell membrane. We tested by wholecell patch-clamping whether the VSD is functional and displays charge movements during voltage steps. In fact, several different ion channels, e.g., CNG channels, carry a VSD, yet are not gated by voltage and are not voltage-dependent. In voltage-activated ion channels, the movement of charged amino acids in S4 during activation produces so-called gating currents 27 . Brief voltage pulses (−15 to −155 mV) evoked transient negative and positive gating currents at the onset and termination of the voltage pulse, respectively (Fig. 2a). In control cells, voltage steps did not evoke gating currents (Fig. 2a). A fit of the Boltzmann function to the integrated off-gating currents yielded a voltage of half-maximal activation (V ½ ) and slope factor (s) of −94.7 ± 2.9 and 8.5 ± 0.8 mV, respectively, corresponding to a gating charge q g of 3.1 e o (Fig. 2a, b and Table 1). In conclusion, the VSD in SpSLC9C1 is functional.
In voltage-gated K + channels, substituting Arg residues in the S4 motif for neutral residues shifts V ½ to more negative values [28][29][30][31][32] . Replacing the third Arg residue in the S4 segment of SpSLC9C1 by Gln (R803Q) shifted the V ½ of gating-current activation by −24 to −117.9 ± 7.1 mV (Fig. 2c and Supplementary Figure 4a); concomitantly, the number of gating charges was lowered to 2.0 e o (Fig. 2c and Table 1). Arg803 apparently contributes one equivalent gating charge, indicating that it may cross the entire transmembrane electric field, similar to the homologous Arg368 in Shaker K + channels 32 .
Voltage dependence of exchange activity. We determined the voltage dependence of SpSLC9C1 by recording Na + /H + exchange activity at different voltages from one and the same cell. However, after stepping the voltage back to V hold , the pH i returned only slowly to resting levels ( Fig. 3a, b). Moreover, pH i signals started to run down during repetitive SpSLC9C1 activation, which prevented recording a complete voltage-response relation under stable conditions (Fig. 3e). To overcome these limitations, we coexpressed SpSLC9C1 with the proton channel Hv1 and recorded SpSLC9C1 activity in the reverse mode. The rationale underlying this strategy was: in the reverse mode, SpSLC9C1 activity acidifies the cell (Fig. 3b); subsequently, opening of Hv1 channels causes H + efflux and restores the original pH i . This prediction is borne out by experiments; upon hyperpolarization, SpSLC9C1 activity acidifies the cell, and activation of H + outward currents via Hv1 by a subsequent depolarization 37 hastens the recovery from acidification (Fig. 3f). Because the operative voltage regimes of Hv1 and SpSLC9C1 do not overlap, H + efflux via Hv1 does not interfere with H + import via SpSLC9C1. SpSLC9C1 is activated at V m < −50 mV, whereas for the pH gradients used here, Hv1 opens at V m ≥ 0 mV 38 . Furthermore, Hv1 only supports H + outward currents 38 . In CHO cells expressing only Hv1, stepping V m from values between −23 and −113 mV to +47 mV produced a rapid Hv1-mediated increase of pH i , whereas stepping voltage from +47 mV to negative values did not change pH i (Supplementary Figure 3c).
When SpSLC9C1 and Hv1 were co-expressed, the voltage protocol started with Hv1 activation to produce an initial alkalinization. This protocol provides a larger dynamic range of the BCECF dye, and the initial pH i value from which the SpSLC9C1-mediated acidification started was always identical. In addition, this activation protocol ensures that the voltage dependence of Hv1 does not overlap with that of the SpSLC9C1 protein, because alkalinisation shifts the voltage dependence of Hv1 activation to more positive voltages. Repetitive Hv1 activation (for 10 s at about +50 mV) followed by SpSLC9C1 activation (for 10 s at about −100 mV) produced sawtooth-like cycles of alkalinization and acidification that were highly reproducible (Fig. 3f). Therefore, SpSLC9C1 activity at different voltages can be compared quantitatively. The voltage dependence of pH i responses was determined by stepping V m from a holding potential of +47 mV to values from −23 to −103 mV (Fig. 3g). A Boltzmann function fitted to the initial slope ΔR s −1 vs. V m yielded a mean V ½ of −70.9 ± 2.5 mV and slope factor s = 3.3 ± 0.9 mV ( Fig. 3h and Table 1). The pH i responses were similar when the V m protocol was reversed (Supplementary Figure 3d).
We tested several generic inhibitors of Na + /H + exchangers present in somatic cells 39 . None of these drugs affected the voltage-gated Na + /H + exchange activity or voltage dependence of SpSLC9C1 activation (Supplementary Figure 5a, b, c, d), most likely because the sequence motifs and regions to which these NHE inhibitors bind 40 are either lacking or different in SpSLC9C1. The insensitivity of SpSLC9C1 to amiloride provided an opportunity to confirm that endogenous Na + /H + exchangers in CHO cells are silent under our measuring conditions. An acidload experiment using control CHO cells 41 42 . We tried to estimate the resting activity of SpSLC9C1 at depolarized voltages (−30 mV) in the presence of amiloride (500 µM) by exchanging solutions from symmetrical with respect to Na + and H + to asymmetrical. Under these conditions, we observed no significant change in pH i (Supplementary Figure 7a, n = 5 experiments). A subsequent voltage step to −100 mV again elicited SpSLC9C1 activity. The normalized and background corrected resting activity under these conditions was 4.3 × 10 −3 ± 2.0 × 10 −2 (range −2.5 × 10 −2 to 3.1 × 10 −2 ). Therefore, we believe that it is safe to estimate that the resting activity is below 3% of its maximum value (Supplementary Figure 7b).  V 1/2 refers to the potential where ΔQ(V) = Q max /2 or ΔR (V) = ΔR max /2 and s refers to the slope of the Boltzmann fit; n is the number of experiments and N q the number of charges involved in the gating process (the calculation is described in the Methods section). All values are given as mean ± SD Cyclic AMP modulates gating currents and exchange activity.
Although the CNBD domain of SpSLC9C1 is suggestive, several ion channels and protein kinase A (PKA) orthologues that carry a CNBD domain are, in fact, not gated or activated by cyclic nucleotides 5,43,44 . Therefore, we examined the action of cyclic nucleotides on gating currents and Na + /H + exchange (Figs. 2 and 4). Cyclic AMP (1 mM) in the pipette shifted the V ½ of gating currents by 20 mV to −74.4 ± 6.4 mV, whereas cGMP (1 mM) was much less effective (Fig. 2b). The number of gating charges was similar in the absence and presence of cAMP or cGMP. Similarly, cAMP shifted the V ½ of gating-current activation of the VSD mutant R803Q by 21 mV to −96.8 ± 6.6 mV ( Fig. 2c; Table 1). When cNMP binding was disabled by replacing a key Arg with a neutral residue in the phosphate-binding cassette of the CNBD (R1053Q) [45][46][47] , charge movement in the absence of cAMP was not affected, but the cAMP-induced V ½ shift was abolished (Fig. 2d). Finally, this V 1/2 shift was not affected by the R399A mutation in the exchanger domain (Fig. 2e). This result demonstrates that binding of cAMP to the CNBD changes the VSD equilibrium.
Next, we examined whether cyclic nucleotides control exchange activity (Fig. 4). With cAMP (1 mM) in the pipette, the V ½ of exchange activity was shifted by 15 mV to −56.8 ± 2.7 mV (Fig. 4a, b; Table 1). Again, cGMP was much less effective (Fig. 4b). To compare Na + /H + exchange without and with cAMP in the same cell, we rapidly photo-released cAMP from caged derivatives of cAMP (BCMCM-cAMP and BECMCM-cAMP) 48 . At −63 mV, reverse-mode activity produced only a small acidification. Photo-release of cAMP instantaneously stimulated exchange activity (Fig. 4c). The action of cAMP saturated after a few flashes (Fig. 4d). At higher light intensity, the number of flashes required to saturate the response was lower (Fig. 4d). The V ½ and s of activation after photolysis (−55.3 ± 5.1 mV and 5.0 ± 0.6 mV (n = 3), respectively) was similar to that recorded in the presence of cAMP in the pipette (Fig. 4e). Finally, in the R1053Q mutant with disabled CNBD, the activation of Na + /H + exchange by voltage was not altered, but the V ½ shift by cAMP was abolished (Fig. 4f). These results demonstrate that binding of cAMP to the CNBD modulates Na + /H + exchange by shifting the voltage dependence of activation.  Table 1 SpSLC9C1 mediates the chemoattractant-induced alkalinization. Stimulation of sperm from S. purpuratus and Arbacia punctulata with chemoattractant peptides evokes a transient hyperpolarization and a rapid alkalinization 9,14-18,49-51 . When the hyperpolarization was abolished, the alkalinization was abolished as well 17 . We studied whether this alkalinization is caused by Na + /H + exchange via SpSLC9C1. Using a rapidmixing technique, we followed kinetically H + efflux and Na + influx in S. purpuratus sperm. Changes in pH i and intracellular Na + concentration ([Na + ] i ) were measured using BCECF and Asante Natrium Green-2, respectively. Stimulation with the chemoattractant speract elevated pH i and [Na + ] i with a similar dose dependence, time course, and latency (Fig. 5a, b). The relation between the latencies of Na + and pH i responses was linear (slope of 1) over two orders of speract concentrations (Fig. 5b inset), indicating that H + efflux and Na + influx are mechanistically coupled. To investigate whether SpSLC9C1 mediates Na + /H + exchange in sea urchin sperm, we studied the cAMP dependence of ion exchange by loading sperm with DEACM-caged cAMP 52 . We studied the action of cAMP photorelease on the resact-evoked pH i responses (Fig. 5c). For all speract concentrations, the pH i response was faster and larger when cAMP was released (Fig. 5c). The exchange activity was analyzed by plotting the maximal slope of the pH i change in the absence (black) or presence (red) of cAMP uncaging (Fig. 5d). When cAMP was released, exchange activity was enhanced at all speract concentrations (Fig. 5d). As control, we studied speractevoked voltage signals with and without uncaging cAMP. Voltage responses were not affected by uncaging cAMP (Fig. 5e), demonstrating that the enhanced exchanger activity is not due to a larger hyperpolarization.
Blockers of SLC9A exchangers, including amiloride, failed to inhibit SpSLC9C1 in heterologous systems. Similarly, in sea urchin sperm, the speract-induced pH i signals were largely unaffected by amiloride (500 µM, Supplementary Figure 5e, f).
In western blots of flagellar membranes, a polyclonal antibody raised against SpSLC9C1 recognized three proteins with apparent M w of 132, 137, and 146 kDa similar to the M w of 146.6 kDa predicted for SpSLC91 and to the apparent M w of HA-tagged SpSLC9C1 in CHO cells (Fig. 6a and Supplementary Figure 8). The two prominent protein bands at higher M w might represent dimers and tetramers. SpSLC9C1 was less abundant in head Finally, in immunocytochemistry, the antibody stained the flagellum and also partially the head (Fig. 6b). These results confirm the presence of SpSLC9C1 in Strongylocentrotus purpuratus sperm 53 . Altogether these experiments suggest that SpSLC9C1 mediates the chemoattractantinduced alkalinization.

Discussion
Slc9 genes encode a large family of cation/proton-coupled antiporters, which fall into three subgroups: classical Na + /H + exchangers of the SLC9A (also referred to as NHE) subtype, the SLC9B (NHA) subtype, and SLC9C (sNHE) subtype. The SLC9A subfamily has been studied extensively, whereas, apart from geneinactivation studies, little is known about SLC9B and SLC9C (SLC9B1 54 ; SLC9C1 2 ). Here, 15 years after the discovery of the SLC9C1 gene, we demonstrate that SpSLC9C1 represents an electroneutral Na + /H + exchanger with a truly remarkable activation mechanism.
Like voltage-activated ion channels, SpSLC9C1 is gated by movement of a voltage sensor; this movement is controlled by cAMP binding to a C-terminal cyclic nucleotide-binding domain. The VSD of SpSLC9C1 harbors seven positively charged amino acids in the S4 motif. Assuming a simple two-state Boltzmann mechanism of VSD movement, we estimate for SpSLC9C1 activation an effective valence or gating charge g q of 3.1 e 0, which is similar to that of Shaker K + channels (g q values of 2.5 e 0 per subunit 55 or 12-13 e 0 per tetrameric channel 56 ). Thus, the number of charges transported across the membrane during gating is similar in Shaker channels and SpSLC9C1. For the only other non-channel VSD in a lipid phosphatase of Ciona intestinalis sperm, g q values are 1-1.6 e 0 57,58 .
How is SpSLC9C1 activity controlled by the VSD? In general, solute carriers undergo cycles of conformational changes for upload and release of substrates on opposite sides of the membrane-a mechanism known as alternating access model 59 or rocking mechanism. A defining feature of this model is that ions or substrates themselves gate the conformational change and that solutes passively redistribute in response to extra-or intracellular changes in substrate concentrations. By contrast, in SpSLC9C1, movement of a channel-like VSD couples membrane voltage to ion exchange. We envisage two different gating mechanisms. In one model, at resting voltage, either ions cannot reach the binding site, because access is obstructed by a physical gate, or ions cannot be uploaded, because the binding affinity is low (non-accessible, Fig. 7a). Hyperpolarization opens the gate or enhances the binding affinity. Alternatively, at rest, ions can access and be uploaded to the binding site, but the rocking motion is blocked (non-rocking, Fig. 7a) and hyperpolarization enables this exchange motion. These mechanisms are summarized by an extended kinetic scheme of exchange activity 60 (Supplementary Figure 7c). A corollary of this scheme is that the exchanger-like ion channels-may have a small basal activity. During their lifetime, sperm are exposed to vastly different aqueous conditions and compartments, involving large changes in external pH o and [Na + ] o 61 . Gating of Na + /H + exchange by voltage implies that extracellular changes are not relayed to the cytosol, unless triggered by a change in voltage that initiates rapid substrate redistribution. This mechanism might safeguard sperm from their environment and is functionally important for rapid periodic stimulation, while navigating in a chemical gradient. In conclusion, SLC9C1 represents a rapidly responding signaling molecule rather than a transporter for housekeeping pH i homeostasis.
This study also provides insight into the puzzling function of cAMP in sperm from marine invertebrates. Whereas the role of cGMP in chemotactic signaling has been established 62 , the physiological function of cAMP remains elusive (Fig. 7b). With the regulation of SpSLC9C1 activity by cAMP, previously unexplained observations fall into place. For example, the hyperpolarization-evoked alkalinization persists for several tens of seconds 9 , whereas the hyperpolarization succumbs within a second 50 . The cAMP-induced V ½ shift of SpSLC9C1 activation from −72 to −55 mV implies that Na + /H + exchange now also operates and maintains alkaline pH i near resting V m (about −50 mV). The alkalinization primes CatSper channels to open and promotes Ca 2+ entry 9 . Thus, cAMP, by maintaining alkaline pH i , keeps CatSper channels primed. This mechanism enables CatSper to translate rapid periodic V m changes into periodic Ca 2+ signals while sperm swim on a periodic path in a chemical gradient 9 .
How is cAMP synthesis controlled? A soluble adenylate cyclase (SACY) is the predominant source for cAMP in sperm 63,64 . There is evidence for a direct physical interaction between SLC9C1 and sAC 12,64 . What is the functional role of this interaction? Whereas in many species, activity of SACY is controlled by bicarbonate and Ca 2+65-67 , circumstantial evidence indicates that cAMP synthesis in sea urchin sperm might be directly or indirectly regulated by membrane voltage 49 , or by pH i 68 , or some other mechanism. We speculate that SpSLC9C1 confers voltage sensitivity to SACY either directly via its VSD or indirectly via alkalinization (Fig. 7b). In turn, cAMP regulates Na + /H + exchange (Fig. 7b). Future studies need to address this potential reciprocal control of SLC9C1 and SACY in sperm of sea urchin, vertebrates, and mammals.
Functionally important amino-acid residues in the exchange domain are conserved between archaeal Na + /H + exchangers, mammalian SLC9A members, and SpSLC9C1 (Supplementary Figure 1). Moreover, the VSD and CNBD domains of SpSLC9C1 are functional and feature all structural hallmarks of the respective domains in voltage-gated ion channels and cyclic nucleotidesensitive proteins ( Fig. 1 and Supplementary Figure 1). In comparison, mammalian SLC9C1 members display several variations of functionally important residues. For example, the ND motif, which is part of the cation-binding site and which is characteristic of 1:1 Na + /H + exchangers, is replaced by a TS motif in mammalian SLC9C1 members. This might be a clue that mammalian SLC9C1 proteins acquired different functions. Finally, three of the four N-terminally located Arg or Lys residues in the S4 motif of SpSLC9C1 are missing in mouse and human orthologues (Fig. 1c), suggesting that voltage activation of mammalian SLC9C1 proteins may be different. Thus, our work provides a technical and conceptual blueprint for future studies of mammalian SLC9C1 orthologues.

Methods
Strongylocentrotus purpuratus sperm samples. Collection of dry sperm was described previously 62 . In brief, 0.2-0.5 ml of 0.5 M KCl was injected into the sea urchin cavity to induce spawning. Spawned sperm (dry sperm) were collected using a Pasteur pipette and stored on ice.  (Fig. 5a) or 4 ml s −1 (Fig. 5c) Generation of a polyclonal anti-SpSLC9C1 antibody. Peptides comprising amino acids (aa) 574-591 (EFADMMEEARLRMLKAEK), aa 857-871 (MVDNKKIL-RELKHIS), aa 937-957 (KMKRLMNAPSSIPPPPPENLL), and aa 1111-1126 (GWTQEKVKLHLERGYL) were synthesized and coupled to BSA via a cysteine that was introduced at the N terminus of peptides. Rabbit antibodies directed against a mixture of these peptides, resulted in two polyclonal antibodies SU1 and SU2. The SpSLC9C1 antibody was purified from antisera by affinity purification using the four peptides. All steps for antibody production were performed by Davids Biotechnology, Regensburg, Germany.

Measurement of [Na
Preparation of heads and flagella from S. purpuratus. Sperm flagella and heads were separated as described 9 with one modification: instead of shearing with a 24-G needle, the sperm suspension was sheared 20 times by centrifugation for 30 s at 75 × g and 4°C through the net of a 40-µm cell strainer (BD Biosciences). All three resulting PCR fragments were cloned together into vector pcDNA3.1 (+) (Invitrogen, Carlsbad, USA) to obtain the full-length clone.
Generation of stable CHO cell lines of SpSLC9C1 and hHv1. CHO K1 cells were electroporated with pc3 sNHE-HA or with pc3 hHv1 using the Neon 100 Kit (Invitrogen, Carlsbad, USA) and a MicroPorator (Digital Bio) according to the manufacturer's protocol (3 × 1650 mV pulses with a 10-ms pulse width). Cells were transferred into complete medium composed of F12 plus GlutaMAX (Invitrogen) and 10% fetal bovine serum (Biochrom, Berlin, Germany). To select monoclonal cells stably expressing SpSLC9C1 or hHv1, the antibiotic G418 (1200 mg ml −1 ; Invitrogen) was added to the cell culture medium 24 h after the electroporation. Monoclonal cell lines were identified by immunocytochemistry using a rat-anti-HA antibody (Roche Applied Science) or by electrophysiological recordings.
Single-cell fluorimetry. We recorded in the whole-cell configuration changes in pH i from CHO cells expressing SpSLC91 (wt), either stably expressed or using transient expression, or the respective SpSLC91 mutants, using transient expression.
For the determination of the voltage dependence of transport activity, we coexpressed hHv1 with SpSLC91 (wt), or the mutants SpSLC9C1-R1053Q or SpSLC9C1-R803Q. Cells were loaded with BCECF (10 µM) via the pipette and were excited with a Photon Technology International DeltaRam X TM monochromator (PTI, New Jersey, USA). BCECF fluorescence was recorded in dual-excitation mode at 440 ± 6.25 nm and at 480 ± 6.25 nm with 5-Hz frequency (100 ms nm −1 ). Emitted light passed a dichroic mirror (500 nm LP) and a 525/15 nm filter (Semrock) and was detected by a photomultiplier system (Model 814, PTI). The pH i signals represent the ratio of F 480 /F 440 . We used a gravity-driven perfusion system. Temperature of ES and ES-NMDG solutions was set to 28°C by a HPT-2 Heated Perfusion Tube (ALA Scientific Instruments Inc. St. Farmingdale, USA). For analysis of V ½ and slope (s) of SpSLC9C1 activity, the initial slope of the ΔR signal was fitted by linear regression. The resulting slopes were plotted against voltage and were fitted to a Boltzmann equation ΔR V ð Þ ¼ ΔR max = 1 þ expðV À V 1=2 Þ=s with s = (kT)/(q g × q e ). k = 1.38 × 10 −23 J K −1 , T = 301.15 K (28°C), and q e = 1.6 × 10 −19 As. V ½ is the potential where ΔR (V) = ΔR max /2. To determine mean ΔR values across data sets, we normalized each dataset to the corresponding parameters of the Boltzmann fit. For experiments with caged compounds, we loaded cells with membrane-permeant caged cyclic BECMCM-caged cAMP (10 µM) for 30 min prior to measurement and included also 100 µM BCMCM-caged cAMP in the pipette solution. Flash photolysis was achieved with short UV pulses (~1 ms) via a Xenon flash lamp System (JML-C2, Rapp OptoElectronic, Hamburg, Germany); light was passed through a UV filter (UV-2 250-375, Rapp OptoElectronic). Light energy was adjusted through the loading voltage of the lamp's capacitor and by neutral density gray filters (300 V, OD1 in Supplementary Figure 3a; 200 V, OD0.4 in Supplementary Figure 3b).
The calibration procedure for BCECF fluorescence to yield pH i by the pseudonull-point method 33 was described previously 9 . Briefly, wild-type or SpSLC9C1 expressing CHO cells seated on glass coverslips were loaded with 10 µM BCECF-AM for 10 min. Coverslips were placed into a home-built perfusion chamber on an Olympus cell R single-cell imaging system. Fluorescence at 540 nm was recorded ratiometrically by alternating excitation using 430/20 nm and 470-490 nm filters, yielding R = F480/F430. The pH i pseudo-null-point solutions contained defined concentrations of weak acid (propionic acid) and weak base (ammonium chloride) in ES solution. In the neutral form, weak acids and bases permeate the plasma membrane resulting in a transient change in pH i . The extent and direction of the change in pH i can be predicted by pH null ¼ pH o À 0:5 log½A=½B, wherein [A] refers to an acid and [B] to a base. The concentrations of propionic acid (5 mM) and ammonium chloride (0.05, 0.32, 1.99, 12.56, 79.25 mM) yielded pH null solutions of 6.4, 6.8, 7.2, 7.6, and 8.0, respectively. To prevent side effects due to osmolarity issues, the NaCl concentration of the ES solution was respectively adjusted (134.95, 134.68, 133.01, 123.44, 55.75 mM for pH null of 6.4, 6.8, 7.2, 7.6, and 8.0).
Gating currents. We recorded gating currents in CHO cells expressing SpSLC9C1 (wt), SpSLC9C1-R1053Q, SpSLC9C1-R399A, or SpSLC9C1-R803Q in ES at 28°C. Online P/N leak subtraction was performed with four pre-pulses (P/4) opposite to pulse polarity to subtract linear currents due to leakage or capacitive artifacts (Clampex V1.10.2.0.12, MDS Analytical Technologies). Voltage steps ranging from +15 to −155 mV in steps of 10 mV were applied. Off-gating currents were integrated over time to yield net charges (Q). To quantify the voltage dependence of charge movement, we fitted Q/V curves to a Boltzmann equation, defined as Q V ð Þ ¼ 1= 1 þ exp V À V 1=2 =s with s = (kT)/(q g × q e ). Mean gating charges (Q/Q max ) were determined by normalizing each data set to the corresponding parameters of the Boltzmann fit. For the SpSLC9C1-R399A mutant, not all gating -current recordings were performed with the same voltage protocol. The number of charges (q g ) involved in the gating process was determined from q g = (kT)/(s × q e ).
Data availability. The data that support the findings of this study are available from the corresponding authors upon request.