Introduction

Bicarbonate (HCO₃^-), the conjugate base of carbonic acid, is a waste product of the citric acid cycle and an important buffer in the intra- and extracellular environment. Its transport across cellular membranes thus constitutes an important mechanism for the control of intracellular pH and the increased buffer capacity of bodily fluids and secretions (Alka & Casey, 2014). This process is mediated by distinct channels and secondary transporters, the latter comprising paralogs of the SLC4 and SLC26 families. The SLC26 transporters constitute an abundant class of anion transport proteins that are expressed in all kingdoms of life with a considerable degree of sequence conservation (Alper & Sharma, 2013; Dorwart, Shcheynikov, Yang, & Muallem, 2008). Whereas homologs in prokaryotes and plants were classified as symporters, where the uptake of mono- and divalent anions is coupled to the import of either sodium ions or protons (Geertsma et al., 2015; C. Wang et al., 2019; L. Wang, Chen, & Zhou, 2021), the transporters in animals show mechanistic differences. Among the ten functional SLC26 paralogs in humans, we find an astonishing mechanistic breadth. Family members function as either coupled anion exchangers, anion transporters with channel-like properties and in the case of SLC26A5 (Prestin), as a motor protein in cochlear outer hair cells (Alper & Sharma, 2013). Among the paralogs which have retained their transport function, the distinction between coupled and uncoupled transporters is most pronounced. Whereas most family members operate as exchangers of different anions including the monovalent ions chloride (Cl^-), bicarbonate (HCO₃^-), iodide and formate and the divalent ions sulfate and oxalate (Alper & Sharma, 2013), the protein SLC26A9 functions as fast passive transporter of Cl^- (Dorwart, Shcheynikov, Wang, Stippec, & Muallem, 2007; Walter et al., 2019). Our previous study has characterized the structural and functional properties of the murine paralog of SLC26A9 (Walter et al., 2019). The protein adopts a homodimeric architecture that is representative for the entire SLC26 family. Each subunit consists of a membrane inserted transport domain (TM) followed by a cytoplasmic STAS domain, which together with the extended N-terminus mediates the bulk of the subunit interactions with few contacts in the membrane-inserted region (Walter et al., 2019). The TMs share an architecture that was initially identified in the prokaryotic transporter UraA (Lu et al., 2011), and that was later defined in the prokaryotic homologue SLC26Dg as common framework for the entire family (Geertsma et al., 2015). In this architecture, the TM segregates into two separate motifs that adopt distinct roles during transport, respectively termed ‘core’ and ‘gate’ domains. The gate domain forms a contiguous rigid scaffold with the cytosolic STAS domain that remains static during transport. Conversely the mobile core domain harbors a selective anion binding site and acts as mobile unit during the transport cycle. The core domain alternately exposes the substrate site to different sides of the membrane as expected for a protein working by an elevator-type transport mechanism (Drew & Boudker, 2016; Walter et al., 2019). Despite its classification as membrane transporter, SLC26A9 mediates large channel-like chloride currents that saturate at high mM concentration and it efficiently discriminates against HCO₃^- (Dorwart et al., 2007; Walter et al., 2019). Its function as a fast electrogenic chloride transporter and selection against HCO₃^- is astonishing since several other family members have been classified as Cl^-/ HCO₃^- exchangers (Alper & Sharma, 2013). One of the family members sharing this property is the protein SLC26A6 (Knauf et al., 2001; Lohi et al., 2000). SLC26A6 is widely expressed in different tissues including the heart, kidney, pancreas and intestine (J. Wang, Wang, Wang, & Tuo, 2020). Besides its function as HCO₃^- transporter, it also plays an important role in the extrusion of oxalate to prevent the formation of kidney stones (Clark et al., 2008; Jiang et al., 2006; Ohana, Shcheynikov, Moe, & Muallem, 2013). Similar to SLC26A9, the protein acts in synergy with the chloride channel CFTR and was proposed to engage in direct interaction and mutual regulation (Bertrand et al., 2017; Dorwart et al., 2008; Shcheynikov, Ko, et al., 2006; Y. Wang et al., 2006). The stoichiometry of HCO₃^- and oxalate transport by SLC26 is still controversial, as transport of both ions has been described in different studies as either electrogenic or electroneutral (Chernova et al., 2005; Clark et al., 2008; Kim, Myers, Sihn, Rafizadeh, & Zhang, 2013; Ohana, Yang, Shcheynikov, & Muallem, 2009; Shcheynikov, Wang, et al., 2006; Xie, Welch, Mercado, Romero, & Mount, 2002). This question is particularly pertinent in light of the divergent net charge of both substrates, with HCO₃^- being monovalent and oxalate divalent under physiological conditions. Consequently, the structural basis for the mechanistic distinction between the uncoupled SLC26A9 and the coupled SLC26A6 and the basis for their pronounced selectivity, with Cl^- acting as common substrate of both proteins, has remained elusive.

To address these open questions, we here present a thorough structural and functional characterization of the human transporter SLC26A6 combining cryo-electron microscopy with electrophysiology and liposomal transport assays. With respect to its overall structure, SLC26A6 closely resembles its paralog SLC26A9 although it shows a distinct transport mechanism. The protein mediates electroneutral Cl^-/ HCO₃^- and electrogenic Cl^-/oxalate exchange, which is in stark contrast to the uncoupled Cl^- transport of SLC26A9 and which is accomplished by an altered ion binding site located in the mobile core domain of the transmembrane transport unit.

Results

Functional characterization of SLC26A6

We have previously characterized the transport properties of murine SLC26A9 by patch clamp electrophysiology and found comparably large and selective Cl^- currents. SLC26A9 currents were shown to reverse at the Nernst potential of the conducting anion, which distinguishes the protein as an uncoupled chloride transporter (Walter et al., 2019) (Figure 1A, Figure 1-Figure Supplement 1A). The high magnitude of these ohmic currents, which lack any time-dependence in response to voltage changes, originate from an unusually active protein that transports anions with channel-like properties (Figure 1-Figure Supplement 1A). SLC26A9 shows a lyotropic permeability sequence that facilitates the transport of Cl^-, isothiocyanate, I^- and nitrate but not sulfate or HCO₃^- (Walter et al., 2019) (Figure 1A, Figure 1-Figure Supplement 1A). The inability to conduct HCO₃^- is an exception in a family where most members work as transporters of this abundant anion (Alper & Sharma, 2013; Ohana et al., 2009). To investigate mechanistic differences between this passive Cl^- transporter and its paralogs functioning as coupled antiporters, we have turned our attention towards the protein SLC26A6, which was proposed to mediate electrogenic Cl^-/HCO₃^- exchange (Ohana et al., 2009; Shcheynikov, Wang, et al., 2006). Hence, we initially attempted to use patch clamp electrophysiology in the whole-cell configuration to characterize its transport properties. However, unlike for SLC26A9, we did not detect specific currents that could be attributed to SLC26A6 in a wide voltage range. This was irrespective of whether we used symmetric chloride solutions or asymmetric conditions, where the major anion in the extracellular solution was changed to bicarbonate (Figure 1B, Figure 1-Figure Supplement 1B). To confirm the expression of SLC26A6 at the plasma membrane, we employed surface biotinylation and found a similar level as observed for its electrogenic paralog SLC26A9 (Figure 1C). We thus concluded that the transporter would either be inactive, exceedingly slow or that transport would be electroneutral as a consequence of the strict 1:1 exchange of two monovalent anions.

Figure 1.

Consequently, we turned to in vitro assays where we have reconstituted SLC26A6 into proteoliposomes and monitored its activity with ion selective fluorophores. Initially we investigated whether we would be able to detect any sign of uncoupled chloride flow by using a fluorometric assay based on the fluorophore ACMA. This assay monitors pH changes following the ionophore-mediated H⁺ influx to compensate for the buildup of a membrane potential during Cl^- uptake, which would enable us to observe transport even if the slow kinetics prevents a characterization by electrophysiology. In this way, we have previously monitored SLC26A9-mediated anion transport into proteoliposomes (Walter et al., 2019). However, in contrast to SLC26A9, no acidification was observed for SLC26A6, emphasizing that the transporter is either inactive or, alternatively, transport would be strictly coupled and electroneutral (Figure 1D).

Next, we were interested in whether Cl^- would be exchanged by HCO₃^- by a mechanism where the charges of transported anions are balanced. To this end, we directly monitored Cl^- influx into HCO₃^--loaded proteoliposomes using the fluorophore lucigenin, whose fluorescence is selectively quenched by the former anion. In this assay, we found a robust time-dependent fluorescence decrease at high Cl^- concentration, demonstrating its transport by SLC26A6 (Figure 1E, Figure 1-figure Supplement 1C, D). The process did not require the addition of valinomycin to dissipate the membrane potential, thus underlining that transport is electroneutral. In the investigation of the Cl^- concentration-dependence of transport, we found a saturation with an apparent K_M of about 16-37 mM (Figure 1F, Figure 1-figure Supplement 1E). This indicated low affinity for the transported anion is in the same range as observed for Cl^- transport in SLC26A9 (Walter et al., 2019).

Finally, we probed whether HCO₃^- itself would be a transported substrate. To this end, we employed a novel HCO₃^--selective europium probe and incorporated this probe within proteoliposomes to assay HCO₃^- influx (Figure 1G) (Martínez-Crespo et al., 2021). The addition of 10 mM HCO₃^- to the outside in presence of a symmetric 200 mM concentration of Cl^- on both sides of the membrane caused a pronounced increase in luminescence that is well above the small non-specific signal observed in liposomes not containing any protein (Figure 1G). This increase is substantially enhanced when using a reduced outside Cl^- concentration of 50 mM to generate an outwardly directed driving force (Figure 1G). Together these experiments demonstrate that HCO₃^- is a transported anion and that its import into proteoliposomes is coupled to the counterflow of Cl^-. Taken together, our experiments underline the function of SLC26A6 as electroneutral coupled Cl^-/HCO₃^- exchanger.

Besides HCO₃^-, SLC26A6 was also proposed to be involved in the transport of oxalate in exchange for Cl^- and thus to participate in the excretion of this small divalent anion in both the intestine and kidneys (Clark et al., 2008; Ohana et al., 2013). However, if assuming a coupled 1:1 stoichiometry as observed in case of Cl^-/HCO₃^- exchange, the transported charge would in this case no longer be balanced, thus converting an electroneutral into an electrogenic transport process. As there are no suitable fluorophores for the selective detection of oxalate and since the small dicarboxylic acid also interferes with the detection of Cl^- by lucigenin, we decided to investigate whether a potentially electrogenic Cl^-/oxalate exchange could be assayed with the fluorophore ACMA (Figure 1H, Figure 1-Figure Supplement 1F). In these experiments, we found a robust decrease of the ACMA fluorescence upon the addition of extracellular oxalate to Cl^--loaded proteoliposomes containing SLC26A6, with the concentration dependence and saturation of the fluorescence decay further emphasizing a specific transport process (Figure 1H). We also attempted to follow electrogenic Cl^-/oxalate exchange by patch clamp electrophysiology in the whole cell configuration but did not measure pronounced currents that would be above background, likely due to the slow kinetics of the process (Figure 1-Figure Supplement 1G, H). Together our functional experiments characterize SLC26A6 as a strictly coupled transporter that exchanges Cl^- with HCO₃^- and presumably also oxalate at a 1:1 stoichiometric ratio.

SLC26A6 structure

To reveal the molecular basis of the described transport properties of SLC26A6, we have determined its structure by cryo-electron microscopy (cryo-EM). A dataset of the protein purified in the detergent glycol-diosgenin (GDN) at 3.3 Å was of high quality and allowed the unambiguous interpretation of the cryo-EM density by an atomic model (Figure 2A, Figure 2-figure supplements 1 and 2). In its organization, the SLC26A6 structure shows familiar features that have previously been observed in various eukaryotic family members of known structure including human and murine SLC26A9 (Chi et al., 2020; Walter et al., 2019), SLC26A5 (Bavi et al., 2021; Butan et al., 2022; Futamata et al., 2022; Ge et al., 2021), SLC26A4 (Liu et al., 2022) and the plant transporter AtSULTR4 (L. Wang et al., 2021) (Figure 2B, Figure 2-Figure Supplement 3). The SLC26A6 subunit consists of a membrane-inserted unit (TM) followed by a cytoplasmic STAS domain. In the homodimeric protein, subunit interactions are predominantly formed by the domain-swapped cytoplasmic parts including an extended N-terminal region and the C-terminal STAS domain, whereas the few contacts between the transmembrane transport units are confined to the end of the terminal membrane-spanning helix α14 (Figure 2B). A region connecting secondary structure elements of the STAS domain (located between Cα1 and β4) termed the intervening sequence (IVS) is present in the characterized protein. Previously, removal of this region increased both the stability and surface expression of a construct of the murine SLC26A9 while retaining its transport properties (Walter et al., 2019). In the density, the N-terminal part of the IVS folds into an α-helix encompassing amino acids 569-593 (Cα^IVS) whereas the following 71 residues are unstructured and thus not resolved in the density (Figure 2A). On the C-terminus of the IVS, the density re-appears at Ser 655 at the region preceding β4, which interacts with Cα^IVS (Figure 2B, C). A generally similar arrangement was also observed in a full-length construct of human SLC26A9 (Chi et al., 2020) (Figure 2D). In both structures, Cα^IVS runs about parallel to the membrane plane and contacts the loop connecting α8 and α9 of the mobile core domain. This presumably influences its conformational preferences thereby potentially impacting transport (Figure 2B-D).

Figure 2.

The transmembrane domain

In the dimeric transporter, the two TMs share minimal contacts and thus supposedly function as independent units (Figure 2B). Consequently, we expect structural differences underlying the distinct transport properties of SLC26 paralogs to be manifested in this part of the protein. Like in other pro- and eukaryotic family members of known structure, the TMs of SLC26A6 consist of two topologically related repeats of seven transmembrane segments, which span the membrane with opposite orientation (Figure 2-Figure Supplement 3B). Elements of both repeats are intertwined to form two segregated modules, a gate domain and a core domain (Figure 3A). The equivalence between the SLC26A9 and SLC26A6 conformations is illustrated in the overlay of the respective TMs whose Cα positions superimpose with an RMSD of 1.3 Å (Figure 3B). Both conformations represent inward-facing states of transporters presumably functioning by an alternate access mechanism. The substrate binding site located in the center of the core domain is accessible from the cytoplasm via a spacious aqueous cavity, whereas the extracellular exit is sealed by extended contacts between core and gate domains (Figure 3C, D, Figure 3-figure supplement 1A, B). The resemblance between individual sub-domains even exceeds the overall similarity of the TM (with core- and gate domains superimposing with RMSDs of 1.13 Å and 1.2 Å, respectively) distinguishing them as largely independent units of a modular protein (Figure 3B). In the comparison of both paralogs, the gate domains do not show pronounced differences (Figure 3B). Together with the cytoplasmic STAS domains, which constitute the bulk of the dimer interface, these sub-domains form a contiguous rigid scaffold of the dimeric protein that presumably remains static during transport (Figure 2B, Figure 3-figure supplement 1C). In contrast, both core domains represent the mobile elements of the protein (Figure 3-figure supplement 1D). Each core domain carries a substrate binding site located in a pocket at the center of the unit facing the gate domain (Figures 3C, D and 4A). The presumable location of a bound chloride ion is resolved in the structure of SLC26A6 at a low contour of the map (Figure 4B, Figure 2-figure supplement 2B). This position concurs with positions inferred from known structures of paralogs, which show binding of Cl^- and HCO₃ to an equivalent location (Chi et al., 2020; Futamata et al., 2022; Ge et al., 2021; Liu et al., 2022) (Figure 4B, Figure 4-figure supplement 1). The anion binding site is placed between the opposed short helices α3 and α10, which partly span the membrane and whose N-termini are aligned to face each other in the center of the core domain (Figure 4A). Together with the sidechains of surrounding residues, both α-helix termini form a selective anion binding site (Figure 4B-G, Figure 4-figure supplement 1). In complex with its bound cargo, the core domain is believed to shuttle as rigid unit between two extreme conformations where the binding sites are either exposed to the cytoplasm or the extracellular environment (Figure 3C, D, Figure 3-figure supplement 1D).

Figure 3.

Figure 4.

As carrier of the substrate binding site, the core domains presumably display the characteristic features accounting for the distinct functional properties of SLC26A6 and SLC26A9. Their conservation within the SLC26 family is illustrated in the comparison of the respective domains of SLC26A6 and SLC26A5 (Prestin) (Figure 4C, Figure 4-figure supplement 1A, C, D, E, G). Latter belongs to a protein that, instead of transporting ions, confers electromotility and which in the best defined structure (PDBID 7LGU) resides in a distinct conformation where the substrate binding site has become buried between core and gate domain (leading to an RMSD of 2.9 Å for the entire TMD) (Ge et al., 2021). Despite the functional and conformational differences, the core domains of both proteins superimpose with an RMSD of only 1.13 Å, illustrating their close structural resemblance (Figure 4-figure supplement 1A, C). Although not working as transporter, both proteins share similar ion binding properties that extends to Cl^- and HCO₃^-, latter of which is not a substrate of SLC26A9 (Walter et al., 2019). In contrast to the general similarity with SLC26A5, the comparison of the same units of SLC26A6 and SLC26A9 shows distinct features that likely underlie the differences in substrate selectivity and transport mechanism (Figure 4-figure supplement 1B, C). These are confined to the conformation of α10, which at its N-terminus forms direct interactions with transported ions, whereas the remainder of the domain is very similar (Figure 4-figure supplement 1B, C). The difference in the conformations of this short membrane-inserted helix can be approximated by a 12-14° rotation of α10 of SLC26A9 around an axis located at its C-terminal end towards α-helix 1 (Figure 4F, Figure 4-figure supplement 1B, C). The replacement of residues within the contact region of SLC26A9 in comparison to other family members acts to lower their side chain volume at the contact region between both helices (Figure 4C-F). These include Ser 96 and Val 100 on SLC26A6, which are both replaced by glycines at equivalent positions in SLC26A9 (i.e., Gly 77 and Gly 81) and a substitution of Val 407, Gln 408 and Thr 411 by Ala 396, Val 397 and Ala 400 (Figure 4C-F). In addition to the described substitutions, there are also differences in the side-chains surrounding the ion binding site, which together determine the size and polarity of the substrate binding pocket (Figure 4C, G-I). These differences extend to Ser 401 of SLC26A6, which is replaced by an alanine in SLC26A9 (Ala 390), and most prominent, the large Arg 404 of SLC26A6, which is replaced by Val 393 in SLC26A9 (Figure 4C, D-I, Figure 4-figure supplement 1D-G). As a result, the pocket is considerably deeper in SLC26A9 compared SLC26A6, where it is delimited by the side chain of Arg 404 that interacts with the residues preceding α helix 3 to form the back side of the shallow binding site (Figure 4D-I, Figure 4-figure supplement 1D-H). In addition to its structural role, Arg 404 likely also contributes to the stabilization of bound anions by coulombic interactions, with its guanidium group potentially engaging in direct interactions with bicarbonate or oxalate. The altered geometry of the site leads to the binding of Cl^- deeper in the pocket in case of SLC26A9 (Figure 4G-I, Figure 4-figure supplement 1H) and presumably explains why bicarbonate interacts with SLC26A6 but not SLC26A9.

Structure-function relationships

Whereas our functional data has defined SLC26A6 as a coupled antiporter that exchanges Cl^- with either HCO₃^- or oxalate at equimolar stoichiometry, its structure has revealed the architecture of a transport protein in an inward-facing conformation where the transport properties are presumably determined by the detailed organization of the mobile core domain. This protein module shows pronounced structural differences compared to the equivalent unit of the uncoupled transporter SLC26A9, resulting in a remodeled anion binding site, which in SLC26A6 is lined by a conserved basic residue (i.e., Arg 404) that is ubiquitous in other mammalian paralogs (Figure 4C and 5A). We thus decided to mutate this position to the amino acid found in SLC26A9 and to characterize the transport properties of the SLC26A6 mutant R404V. This mutation is well-tolerated and does not interfere with protein integrity as judged by its expression level and biochemical properties. We initially investigated whether SLC26A6 R404V would conduct ions as a consequence of compromised coupling properties and thus studied the protein by patch-clamp electrophysiology. In whole-cell patch clamp experiments, we detected small currents in symmetric Cl^- solutions, which are low compared to SLC26A9, despite the comparable expression of the mutant at the cell-surface (Figure 5B-D). Irrespective of their magnitude, these currents are indicative for the capability of the protein to mediate electrogenic Cl^- transport. We thus turned to our proteoliposome-based ACMA assay to confirm that sidechain replacement has indeed conferred the ability to SLC26A6 to pass downhill Cl^- transport (Figure 5E). After demonstrating uncoupled chloride transport, we tested whether the mutant has retained its ability of mediating coupled Cl^-/HCO₃^- exchange using the fluorophore lucigenin but did not find pronounced Cl^- flux in this case (Figure 5F, G). Together our data suggest that the mutation of a conserved basic position in the ion binding site has compromised the ability of SLC26A6 to mediate coupled Cl^-/HCO₃^- exchange and instead facilitated uncoupled Cl^- transport with slow kinetics. While these findings underline the importance of Arg 404 for anion interactions and coupling, they also emphasize that the mutation of a single residue is insufficient to confer the complete functional phenotype of SLC26A9 as a fast uncoupled Cl^- conductor. Nevertheless, our results further underline the importance of the anion binding site in SLC26 transporters for the transport mechanism of its members.

Figure 5.

Discussion

By combining structural data obtained by cryo-EM with electrophysiology and transport assays, our study has elucidated the previously unknown architecture of SLC26A6, defined its transport properties and revealed the features that underlie the mechanistic distinction to the paralog SLC26A9. By using a novel HCO₃^- selective europium probe (Martínez-Crespo et al., 2021), we have directly demonstrated the transport of this important physiological anion (Figure 1G). We have further clarified the controversy concerning the transport mechanism of SLC26A6 (Ohana et al., 2009) by detecting a strict equimolar exchange of anions binding to a conserved site in the mobile core domain of the transmembrane transport unit (Figure 4B, H). Consequently, this exchange is electroneutral in case of the monovalent HCO₃^- and electrogenic in case of the divalent oxalate (Figure 1E-H).

Regarding their function, SLC26A6 and SLC26A9 show distinct features with respect to substrate selectivity and the general transport mechanism (Ohana et al., 2009). Remarkably, both proteins are capable of interacting with Cl^- with similar mM affinity (Figure 1F) (Walter et al., 2019). However, they do this via distinct interactions in equivalent binding regions. In SLC26A9, Cl^- binds in a deep pocket with low field strength, whereas the same pocket is shallower in SLC26A6, where a conserved arginine (Arg 404) increases the positive electrostatic potential and thus likely stabilizes the bound anion (Figures 4H, I and 6). Together with the presence of additional residues at the binding site (i.e., Ser 401 at the N-terminal end of α11), the captured Cl^- in SLC26A6 thus can rely on a larger density of polar interactions (Figure 4-extended data fig. 1E, F). The same site likely also provides a suitable environment for HCO₃^- and oxalate, neither of which are substrates of SLC26A9 (Dorwart et al., 2007; Walter et al., 2019). A similar anion interaction observed in SLC26A6 is also found in the paralogs SLC26A4 (Pendrin) (Liu et al., 2022) and SLC26A5 (Prestin) (Futamata et al., 2022; Ge et al., 2021), which both share a comparable substrate preference. The similarity between the core domains of SLC26A6 and Prestin is remarkable in light of the altered function of the latter, which is no longer capable of transport and instead operates as motor protein in cochlear outer hair cells of mammals (Zheng et al., 2000). While to some degree Prestin undergoes comparable conformational changes as other family members, as illustrated in the similarity of the structures of intermediate conformations of SLC26A9 and SLC26A5, the full transition into an outward-facing conformation to release the bound anion to the extracellular environment appears in this case to be prohibited (Bavi et al., 2021; Ge et al., 2021).

Besides the absence of a residue carrying a positive charge, the altered binding site in SLC26A9 is also a consequence of concerted replacements of interacting amino acids on α1 and α10, which change the immediate environment of the bound ion and allow for a reorientation of the partly inserted helix α10 whose N-terminus contributes part of the interactions (Figure 4C, D). Collectively, these changes presumably lead to the altered ion selectivity. The importance of the described anion binding site for the transport mechanism is illustrated in the point mutation R404V in SLC26A6. The replacement of this basic residue, which was considered a hallmark in the distinction between SLC26A9 and other family members (Walter et al., 2019) has compromised coupled Cl^-/ HCO₃^- exchange and instead mediates uncoupled Cl^- transport, although with slower kinetics than SLC26A9 (Figures 5B-G, 6).

From a mechanistic viewpoint SLC26A9 and SLC26A6 are distinguished by the sequence of conformations that together define the transport cycle (Figure 6). In the uncoupled SLC26A9, the transporter is able to transition between inward- and outward facing states at a high rate, which underlies the pronounced currents observed in patch clamp experiments (Figure 6B, Figure 1-figure supplement 1A). The same change of the access in an unloaded conformation is kinetically unfavorable in the coupled exchanger SLC26A6, as illustrated in the lack of current in electrophysiological recordings and the absence of a Cl^- leak in a vesicle-based assay (Figures 1B-D and 6A, Figure 1-figure supplement 1B). The basis for the distinctive kinetic features of both paralogs is currently still unclear, and we do at this stage not exclude the existence of a leaky channel-like conformation in SLC26A9. Such a conformation could potentially promote Cl^- flow across a continuous aqueous pore without obligatory protein movement, although there is currently no structural evidence for its existence. The nature of the energy barrier, which prevents the shuttling of the unloaded state of SLC26A6 and other paralogs with similar functional phenotype (i.e., the transporters SLC26A1-4) has thus far remained elusive. It is likely that conformational changes are dictated by interactions between the mobile core domain with the immobile gate domain, which are both believed to operate as semi-independent entities (Geertsma et al., 2015; Walter et al., 2019). Remarkably, both units bury a similar contact area in the inward-facing conformations of SLC26A6 and SLC26A9 and do not exhibit pronounced differences in their interactions (Figure 3-figure supplement 1A, B). Additional contacts in SLC26A6 are established with the structured N-terminal part of the IVS region of the STAS domain of the interacting subunit, which appear to further stabilize the observed inward-facing conformation (Figure 2C). Similar interactions have also been found in the full-length transporter SLC26A9 (Chi et al., 2020), which, next to the increased surface expression, could account for the strongly increased activity observed in the construct SLC26A9^T where the IVS sequence of the protein was removed (Walter et al., 2019). Besides the effect of the IVS, also the truncation of the C-terminus of SLC26A9 might have contributed to this high activity. These unstructured C-termini are weakly conserved and contain a PDZ binding motif at its end, which is believed to tether either transporter to potential partners such as the anion channel CFTR or the Na⁺ dependent citrate carrier SLC13A2 via interacting scaffolding proteins (Ohana et al., 2013; Shcheynikov, Ko, et al., 2006). In the isoform b of SLC26A9, a C-terminal extension was located in the aqueous cavity leading to the substrate binding site in the inward-facing conformation of SLC26A9, thereby locking the transporter in the observed conformation (Chi et al., 2020) (Figure 2-Figure Supplement 3A). There is no similar extension found in any SLC26A6 isoforms, and the role of this interaction in a cellular environment is currently unclear, although it hints at a possible regulation in SLC26A9. Collectively our study has provided key insight into the structural basis for substrate selectivity and the mechanistic distinction between uncoupled and coupled transporters of the SLC26 family and thus provides a foundation for future investigations.

Figure 6.

Material and Methods

Key resources table

Construct Generation

The open reading frame of human SLC26A6 (isoform 3) (GenBank accession number: BC017697) and mouse SLC26A9 (GenBank accession number: BC160193) was amplified from a cDNA clone (Source BioScience) and cloned into into a pcDNA 3.1 vector (Invitrogen) modified to be compatible with the FX-cloning (Geertsma & Dutzler, 2011). The truncated murine SLC26A9 construct (SLC26A9T) was generated as previously detailed (Walter et al., 2019). All FX-modified expression constructs contained a C-terminal C3 protease cleavage site, venus YFP, myc-tag and streptavidin-binding protein (SBP). To introduce the R404V mutation to hSLC26A6, the QuickChange (Aligent) mutagenesis method was used with the following primers: Forward: 5’-GCT CTA TGT CTG TGA GCC TGG TAC-3’ and Reverse: 5’-GTA CCA GGC TCA CAG ACA TAG AGC-3’.

Protein Expression

Suspension HEK HEK293S GnTI^- cells (ATCC CRL-3022) were grown and maintained in HyClone TransFx-H (GE Healthcare) media supplemented with 1% FBS, 1% Kolliphor P188, 100 U ml^-1 penicillin/streptomycin & 2mM Glutamine. Cells were incubated at 37 °C & 5% CO₂ in TubeSpin Bioreactor 600 vessels (TPP) and shaken at 185 rpm.

For transient transfections, cells were split for transfection 24 h prior from a cell density of 1-1.5 10⁶ ml^-1 to 0.4-0.6 10⁶ ml^-1. A ratio of 1.2µg Plasmid DNA per 10⁶ cells and 40kDa linear PEI MAX (Polysciences) transfection reagent were utilised for protein expression. On the day of transfection, a 10µg.ml^-1 DNA dilution was incubated in non-supplemented DMEM media (Sigma). Linear PEI MAX was incubated with DNA to a ratio of 1:3 (w/w) from a 1 mg ml^-1 stock. After a 15-minute incubation, the DNA-PEI mix was added to suspension cells and co-incubated 3mM valproic acid (VPA). Cells were harvested after 48 h expression and flash-frozen in liquid N₂ prior to storage at -80 °C.

Protein Purification

For protein purification, all protocols were performed at 4°C. Cell pellets were thawed and mixed with extraction buffer (25 mM HEPES, pH 7.4, 200 mM NaCl, 5% glycerol, 2 mM CaCl₂, 2 mM MgCl₂, 2% GDN, benzamidine 1 mM, leupeptin 10 µM, pepstatin A 1 µM, PMSF 100 µM). Cell lysis was incubated for 2 h at 4 °C prior to ultracentrifugation (85,0000 g, 30 min, 4 °C) to remove insoluble material. Lysate was filtered (5µm filter, Sartorius) prior to binding to Streptactin Superflow high-capacity resin (IBA Lifesciences) equilibrated in wash buffer (25 mM HEPES, pH 7.4, 200 mM NaCl, 5% glycerol, 0.02% GDN) for 2 h.

After discarding the flow-through of clarified lysate through gravity flow, Streptactin resin was washed with 40 column volumes (CV) wash buffer and SBP-tagged protein was eluted using wash buffer supplemented with 10 mM desthiobiotin (Sigma). Eluate was concentrated to 500 µl using a 100 KDa MWCO concentrator (amicon), pre-equilibrated in wash buffer. C-terminal purification and detection tags were cleaved by 3C protease at a 2:1 protease:protein mass ratio. Prior to size-exclusion chromatography, samples were passed through a 0.22 µm filter (Millipore). Samples were injected on the Äkta prime plus chromatography system (Cytivia) and separated on a Superose 6 10/300 column (GE Healthcare) in SEC buffer (10 mM HEPES, pH 7.4, 200 mM NaCl, 0.02% GDN). Protein containing peak samples were collected and used for cryo-EM sample preparation at a 2 mg ml^-1 concentration or used immediately for liposome reconstitution.

Surface biotinylation

Adherent HEK293T cells were transfected at 80% confluency were transfected with 10 µg of plasmid encoding SLC26A6 WT and R404V constructs fused with a c-terminal SBP-Myc-Venus tag per 10 cm culture dish. Transfection was performed using PEI MAX at a DNA:PEI ratio of 1:4. The Pierce™ Cell Surface Biotinylation and Isolation Kit (Themo Fisher) was used according to manufacturer’s guidelines. Cells were washed with PBS from 10 ml culture per construct and biotinylated following 24 h of expression using EZ-Link Sulfo-NHS-SS-Biotin. Quenching was then performed using 750 µl quenching buffer and cells were harvested. Following a PBS wash-step and centrifugation at 1000 g, cells were re-suspended and lysed in 200 µl extraction buffer (25 mM HEPES, pH 7.4, 200 mM NaCl, 5% glycerol, 2 mM CaCl₂, 2 mM MgCl₂, 2% GDN, benzamidine 1 mM, leupeptin 10 µM, pepstatin A 1 µM, PMSF 100 µM) by gentle agitation for 1 h at 4 °C. Insoluble fractions were removed by centrifugation at 50,000 g for 30 min and supernatant was incubated with 200 µl Neutravidin agarose slurry pre-equilibrated in wash buffer (25 mM HEPES, pH 7.4, 200 mM NaCl, 5% glycerol, 0.02% GDN) for 1 h at 4°C. The resin was washed 3 times with 200 µl of wash buffer and surface-biotinylated proteins were eluted by incubation with 200 µl wash buffer containing 50 mM fresh DTT for 1 h at 4 °C with gentle mixing of the sample. 20 µl of input, flow-through and surface-eluted samples were separated by SDS-PAGE. Samples were then transferred to a PVDF membrane and analyzed by Western blot using a mouse-anti-myc primary antibody (Sigma) and HRP-coupled goat-anti-mouse secondary antibody (Jackson ImmunoResearch). Antibodies were diluted 1:10,000 in TBS-T with 5% skimmed milk. Chemoluminescence signal was developed using ECL prime reagent (Cytiva) and imaged using a Fusion FX7 imaging system (Vilber).

Liposome Reconstitution

For liposome reconstitutions using the [Eu.L1⁺] bicarbonate-selective probe, lipid preparation was performed on the same day as protein purification. POPE:Cholesterol lipids (7:3 w/w ratio, Avanti) were pooled in a round-bottom flask and initially dried under N₂ flow to form a lipid film. Lipids were further dried under vacuum for 1 h prior to re-hydration with [Eu.L1⁺] buffer (50 µM [Eu.L1⁺], 200 mM NaCl, 5 mM HEPES, pH 7.4). The resulting suspension was gently sonicated and stirred for 2 h to generate a homogeneous mixture. Ten freeze-thaw cycles were performed to generate multilamellar liposomes before extrusion 29 times through x2 polycarbonate 400 nm filters to give unilamellar liposomes.

For protein incorporation, incremental 10 µl additions of 10% Triton X-100 (w/v) were performed and monitoring of absorbance at 540 nm was used to detect liposome destabilization. After the formation of stabilized liposomes, purified SLC26A6 was added at an lipid-to-protein ratio (LPR) of 50:1. To counteract [Eu.L1⁺] probe leakage, an additional equimolar 50 µM [Eu.L1⁺] was added. The subsequent mixture was gently rotated at RT for 20 min prior to an addition of 250 mg SM-2 Bio-Beads (Bio-Rad) per 5 ml of sample. The sample was then incubated for 30 min at RT prior to incubation at 4 °C. Bio-Bead additions were performed over a three-day period every 24 h. Biobeads were removed by gravity filtration and the sample was used immediately for assay measurements.

For liposome reconstitutions for the Lucigenin and ACMA assays, a similar procedure was used with the following adjustments. POPE:POPG lipids (3:1 w/w) were first pooled in a round-bottom flask and washed with diethyl ether prior to drying under N₂ for 1.5 h. Lipids were then dissolved to 20 mg/ml in 10 mM HEPES, 150 mM KCl prior to sonication to form a homogeneous mixture. After three freeze-thaw cycles the sample was frozen in liquid N₂ and stored at -80 °C until further use. Purified SLC6A9T was incorporated utilizing the triton-destabilisation method previously detailed at a LPR of 80:1 (w/w). Both SLC26A6 wildtype and the R404V mutant were incorporated at a 50:1 (w/w) ratio. Subsequently, successful incorporation was confirmed by SDS-PAGE gel quantification. For all reconstitution procedures, mock empty liposomes lacking any purified protein were generated using the same protocols.

[Eu.L1⁺] Bicarbonate Transport Assay

For all measurements of HCO₃^- transport, buffers used were degassed to remove atmospheric CO₂. After Bio-Bead removal via gravity filtration, proteoliposomes were pelleted and the excess [Eu.L1⁺] probe (Martínez-Crespo et al., 2021) washed by sequential ultracentrifugation and resuspension steps in outside buffer (200 mM NaCl, 5 mM HEPES, pH 7.4). Where 50 mM internal NaCl was required, liposomes were pelleted by ultracentrifugation and resuspended in asymmetric internal buffer (50 µM [Eu.L1⁺], 50 mM NaCl, 5 mM HEPES, pH7.4). Liposomes were further subjected to five freeze-thaw cycles and extruded 29 times using 400 nm polycarbonate filters. Sequential ultracentrifugation wash steps were performed before resuspension in outside buffer.

For measurements of electroneutral Cl^-/ HCO₃^- exchange, 100 µl of SLC26A6 proteoliposomes were added to a 96-well flat bottom black plate (Grenier). Prior to measurement, liposome samples were gently bubbled under N₂ for 2 min to remove atmospheric CO₂ and residual HCO₃^- from samples. Following the recording of a baseline emission for 20 cycles (lasting 0.5 s each), 10 mM HCO₃^- was added and normalized [Eu.L1+] emission (excitation = 332, emission= 617) was determined following lysis of liposomes by 0.4 % Triton after cycle 300. Time gating of the recorded emission was performed using a 150µs lag-time between excitation and signal integration with an overall 850µs integration time. HCO₃^- transport by was measured in the presence of both four-fold NaCl gradient (200 mM inside/50 mM outside) and symmetrical conditions (200 mM inside/200 mM outside).

ACMA Assay

To compare uncoupled chloride properties of SLC26A6 to SLC26A9, proteoliposomes were prepared to obtain an internal buffer concentration of 50 mM Na₂SO₄, 10 mM HEPES, pH 7. Proteoiposomes were sonicated in 20 µl aliquots to form unilamellar vesicles prior to dilution into 9-amino-6-chloro-2-methoxyacridine (ACMA) assay buffer (2 µM ACMA, 75 mM NaCl, 10 mM HEPES, pH 7.4). 100 µl aliquots were added to a 96-well plate and after a baseline measurement period for 13 cycles (0.5 s per cycle), the protonophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP) was added to dissipate the membrane potential and permit transporter-mediated anion-movement. ACMA fluorescence (excitation = 412 nm and emission = 482 nm), as a determinant of Cl^- transport, was measured using the Tecan Infinite M1000 Pro Plate Reader over 495 seconds or using the Tecan Spark Plate reader over 1,800 s. Data are normalized to the point of CCCP addition.

To measure electrogenic oxalate transport, the same protocol was applied with the following exception: SLC26A6 liposomes consisted of an internal concentration of 150 mM KCl, 10 mM HEPES, pH 7.4. 40 µl aliquots were added to a to a black 384-well plate (Grenier) and ACMA fluorescence was measured on a TECAN Spark Plate Reader over 1,200 s.

Lucigenin Assay

For measurement of anion transport mediated by Cl^-/ HCO₃^- exchange, the halide-sensitive lucigenin dye (400 µM) was incorporated into proteoliposomes containing either SLC26A6 or SLC26A6 R404V mutant and empty mock liposomes (all with a POPE:POPG composition of 3:1 w/w) in 200 mM HCO₃^-, 10 mM HEPES, pH 7.4. To permit fluorophore incorporation, liposomes were diluted in the above buffer and three freeze-thaw cycles were performed prior to extrusion (17 times) using 50 nm polycarbonate filters. Sequential ultracentrifugation and resuspension in assay buffer containing no lucigenin dye were performed to remove excess exterior lucigenin. 40 µl aliquots were transferred to a black 384-well plate and after 20 cycles (0.2 second per cycle), 0 to 240 mM NaCl diluted in assay buffer was added. Quenching of lucigenin fluorescence (excitation = 430 nm, emission = 505 nm) was measured as a determinant of chloride influx for 300 cycles prior to lysis by 0.4% Triton. Data are normalized to the point of NaCl addition and corrected for signal observed under mock liposome conditions.

Electrophysiology

For electrophysiology recordings, adherent HEK293T cells (ATCC) were maintained in DMEM (Gibco) media and incubated at 37 °C and 5% CO₂. Media was supplemented with 10% FBS, 1 mM L-glutamine, 4.5 g l^-1 Glucose, 1 mM sodium pyruvate and 100 U ml^-1 penicillin/streptomycin (Sigma). Cells were transfected at between 40-60% confluency and split into 60 x 15 mm dishes (Corning) the day prior to transfection. Plasmid DNA was transfected at a concentration of 5 µg DNA together with linear PEI (25 kDa) at a ratio of 1:2.5 (w/w) DNA to PEI. DNA and PEI were mixed in non-supplemented media, incubated at RT for 10 min and added dropwise to cells with 3 mM VPA. Venus-fused constructs of human SLC26A6 and mouse SLC26A9 were used for recordings to allow for detection of expression through fluorescence microscopy. Cells were recorded from between 16-30 h post-transfection.

Patch pipettes were formed from pulled and polished borosilicate glass capillaries (OD = 1.5 mm, ID = 0.86 mm, Sutter). When backfilled with 150 mM CsCl, patch pipettes yielded a resistance of between 2-4 MΩ. For voltage-clamp experiments, currents were recorded with an Axopatch 200B amplifier and digitized using a Digidata 1440A A/D converter. Signals were filtered at 5 kHz and sampled at 20 kHz prior to acquisition with Clampex 10.6 (Molecular Devices). Recorded cell capacitance upon break-in varied between 10-30 pF with series resistance <10 MΩ and calculated liquid junction potentials (JPCalcW, Molecular Devices) never surpassed 5 mV.

For whole-cell patch-clamp experiments, the intracellular pipette solution consisted of 146 mM CsCl, 2 mM MgCl2, 5 mM EGTA, 10 mM HEPES and pH 7.4. Extracellular bath solutions consisted of either 146 mM CsCl, 2 mM MgCl2, 5 mM EGTA, 10 mM HEPES or 150 mM CsHCO₃/150 mM Cs₂Oxalate, 2mM Mg(OAc)₂, 5 mM EGTA, 10 mM HEPES with pH adjusted using either CsOH or methanesulfonate. The pH of HCO₃^- was monitored to ensure the production of CO₂ and decline of HCO₃^- never exceeded 20%. All acquired electrophysiology data were analyzed using Clampfit 11.0.3 (Molecular Devices), Excel (Microsoft), and GraphPad Prism 9 (GraphPad).

Cryo-EM Data Acquisition

To proceed with cryo-EM structural determination, 2.5 µl samples of GDN-purified SLC26A6 was applied to glow-discharged holey carbon grids (Quantifoil R1.2/1.3 Au 200 mesh) at a concentration of 2 mg ml^-1. Blotting of the sample (3-6 s) was performed at 4 °C at a relative humidity of 75% and grids were flash-frozen in liquid propane-ethane mix with a Vitrobot Mark IV (Thermo Fisher Scientific). A total of 11,962 images from two combined datasets were collected on a 300 kV Titan Krios G3i using a 100 μm objective aperture. Data were collected in super-resolution mode using a post-column quantum energy filter (Gatan) with a 20 eV slit and K3 Gatan direct detector. Data were acquired using EPU2.7 with aberration-free image shift (AFIS) and a defocus range of -0.8 to -2.4. The dataset was recorded using a pixel size of 0.651 Å/pixel (0.3225 Å/pixel in super-resolution mode) with a total exposure time of 1s (38 individual frames) and a dose of 1.696 e^-/Ų/frame. The total electron dose for the specimen was 61 e^-/Ų for dataset 1 and 67 e^-/Ų for dataset 2.

Cryo-EM data processing

Cryo-EM data were processed using CryoSPARC v3.2.0-4.0 (Punjani, Rubinstein, Fleet, & Brubaker, 2017) (Figure 2—figure supplement 1). All movies were subjected to motion correction using patch motion correction. Following patch CTF estimation, high quality micrographs were identified based on relative ice thickness, CTF resolution estimation and total full frame motion. 1,749,907 particles were picked following the generation of 2D templates for automated template picking. Subsequently, picked particles were binned x2.4 using a box-size of 360 pixels for 2D classification (pixel size 1.56 Å/pixel). Following three rounds of 2D classifications, particles were selected based upon visual inspection. Classes were selected that displayed general structural features characteristic for the SLC26 family in various orientations. From this, 256,581 particles were used to generate a ‘good’ ab initio model and a portion of rejected particles (16,030 particles) were used to generate a ‘junk’ class. 967,713 selected particles from 2D classification were subjected to heterogenous refinement using the selected good models as a ‘template’ and the junk model as a ‘decoy’ model. After several rounds of heterogenous refinement, the selected 507,174 particles and associated heterogeneously refined map were subjected to multiple rounds of non-uniform refinement with imposed C2 symmetry. Following iterative rounds of non-uniform and local CTF refinement a reconstruction of a nominal resolution at 3.55 Å was generated. These particles were then subjected to a 4-class ab initio for further sorting resulting in one good map comprised of 173,834 particles. Particles were re-extracted with a bin-1 pixel size of 0.651 Å/pixel using 432 pixel-sized box and subjected to non-uniform refinement prior to a final 3-class ab initio resulting in a single good map comprised of 93,169 particles. Multiple rounds of non-uniform refinement and local CTF refinement were performed. Finally, the resultant reconstruction maps were sharpened with the DeepEMhancer tool using the HighRes deep-learning model (Sanchez-Garcia et al., 2021). The quality of the map was evaluated and evaluated using 3DFSC (Tan et al., 2017) for FSC validation and local resolution estimation.

Cryo-EM model building and refinement

Map interpretation was performed in Coot (Emsley & Cowtan, 2004) using a model of SLC26A6 obtained from the AlphaFold Protein Structure Database (Jumper et al., 2021; Varadi et al., 2022) as template. The quality of the map allowed for the unambiguous assignment of residues 28-49, 62-594 and 655-747. The model was iteratively improved by real space refinement in PHENIX (Afonine et al., 2018) maintaining NCS and secondary structure constrains throughout. Figures were generated using ChimeraX (Pettersen et al., 2004) and Dino (http://www.dino3d.org). Surfaces were generated with MSMS (Sanner, Olson, & Spehner, 1996).

Data availability

The cryo-EM density map of hSLC26A6 has been deposited in the Electron Microscopy Data Bank under ID code XXX. The coordinates for the atomic model of hSLC26A6 has been deposited in the Protein Data Bank under ID code YYY. Transport experiments are provided as source data.

Author Contributions

D.N.T has cloned, expressed and purified proteins, carried out electrophysiology and transport experiments, prepared samples for cryo-EM, processed cryo-EM data and built models. M.S. assisted in cryo-EM data collection and processing. C.B. and S.J.B. synthesized probe used for HCO₃^- transport experiments. D.N.T and R.D. jointly planned experiments and wrote the manuscript with input from all authors.

Author Information

The authors declare no competing financial interests. Correspondence and requests for materials should be addressed to R.D. (dutzler@bioc.uzh.ch).

Acknowledgements

We thank the Center for Microscopy and Image Analysis (ZMB) of the University of Zurich for the support and access to the electron microscopes. All members of the Dutzler lab are acknowledged for their help at various stages of the project. We thank Anastasiia Sukalskia, Elena Lehmann, Katarzyna Drozdzyk and Melanie Arndt for their help in cryo-EM data acquisition. Justin Walter is acknowledged for the cloning of initial constructs and advice in the SLC26 project. This research was supported by a grant from the Swiss National Science Foundation (No. 31003A_163421) to R.D. SJB acknowledges the Engineering and Physical Sciences Research Council (EPSRC) for funding [EP/S032339/1].

The following figure supplements are available for figure 1:

Figure supplement 1. Transport data.

Figure 1—figure supplement 1.

The following figure supplements are available for figure 2:

Figure supplement 1. Cryo-EM reconstruction of SLC26A6.

Figure supplement 2. Section of the cryo-EM density of SLC26A6.

Figure supplement 3. Sequence and Topology.

Figure 2—figure supplement 1.

Figure 2—figure supplement 2.

Figure 2—figure supplement 3.

The following figure supplements are available for figure 3:

Figure supplement 1. TM domain features.

Figure 3—figure supplement 1.

The following figure supplements are available for figure 4:

Figure supplement 1. Binding site comparisons.

Figure 4—figure supplement 1.

Table 1.