Common Gating of Both CLC Transporter Subunits Underlies Voltage-dependent Activation of the 2Cl−/1H+ Exchanger ClC-7/Ostm1*

Background: ClC-7 is a homodimeric lysosomal chloride transporter important for lysosomal function and bone degradation. Results: Altered gating kinetics of one subunit affect the kinetics of the other subunit. Conclusion: Gating of ClC-7 involves both CLC subunits and requires noncovalent binding of cytoplasmic domains. Significance: Osteopetrosis and lysosomal storage disease are associated with accelerating mutations in the ClC-7 C terminus and the contacting intramembrane part. CLC anion transporters form dimers that function either as Cl− channels or as electrogenic Cl−/H+ exchangers. CLC channels display two different types of “gates,” “protopore” gates that open and close the two pores of a CLC dimer independently of each other and common gates that act on both pores simultaneously. ClC-7/Ostm1 is a lysosomal 2Cl−/1H+ exchanger that is slowly activated by depolarization. This gating process is drastically accelerated by many CLCN7 mutations underlying human osteopetrosis. Making use of some of these mutants, we now investigate whether slow voltage activation of plasma membrane-targeted ClC-7/Ostm1 involves protopore or common gates. Voltage activation of wild-type ClC-7 subunits was accelerated by co-expressing an excess of ClC-7 subunits carrying an accelerating mutation together with a point mutation rendering these subunits transport-deficient. Conversely, voltage activation of a fast ClC-7 mutant could be slowed by co-expressing an excess of a transport-deficient mutant. These effects did not depend on whether the accelerating mutation localized to the transmembrane part or to cytoplasmic cystathionine-β-synthase (CBS) domains of ClC-7. Combining accelerating mutations in the same subunit did not speed up gating further. No currents were observed when ClC-7 was truncated after the last intramembrane helix. Currents and slow gating were restored when the C terminus was co-expressed by itself or fused to the C terminus of the β-subunit Ostm1. We conclude that common gating underlies the slow voltage activation of ClC-7. It depends on the CBS domain-containing C terminus that does not require covalent binding to the membrane domain of ClC-7.

The CLC gene family, first identified by the cloning of the voltage-dependent Cl Ϫ channel ClC-0 from Torpedo electric organ (1), not only encodes Cl Ϫ channels, but also encodes anion/proton exchangers (2)(3)(4)(5). Both CLC Cl Ϫ channels and Cl Ϫ /H ϩ exchangers function as homodimers with one ion translocation pathway per pore, as shown by mutational structure/function analysis (6 -8) and x-ray crystallography of prokaryotic and algal CLC proteins (9,10). Four of the nine mammalian CLC proteins associate with distinct small accessory subunits that display one or two transmembrane domains (11)(12)(13). For instance, the lysosomal 2Cl Ϫ /1H ϩ exchanger ClC-7 requires binding to the type-I transmembrane protein Ostm1 for protein stability and ion transport activity (12,14).
The ion transport activity of both CLC Cl Ϫ channels and Cl Ϫ /H ϩ exchangers can strongly depend on the transmembrane voltage. Voltage-dependent gating has been most thoroughly studied with the fish ClC-0 Cl Ϫ channel. Macroscopic currents and single-channel analysis provided clear evidence for two distinct gating processes in ClC-0: a fast, depolarization-activated protopore gate acting independently on each of the two pores of the dimer and a slow, hyperpolarization-activated common gate closing both pores simultaneously (15)(16)(17). Likewise, mammalian ClC-1 and ClC-2 display protopore and common gating, but their time constants and voltage dependences, as well as the low single-channel conductances of these channels, render the identification and separation of these gating modes more difficult (18 -22). The crystal structure of bacterial EcClC-1, together with mutagenesis, showed that a certain "gating" glutamate, whose side chain protrudes into the ion translocation pathway, plays a pivotal role in protopore gating of CLC channels (23) and in Cl Ϫ /H ϩ coupling of CLC exchangers (5), findings that were amply confirmed in other studies with prokaryotic and eukaryotic CLCs. The structural basis of common gating is much less understood, but the gating glutamate may also play a role in closing the permeation pathway after a common rearrangement of both subunits (24,25). Early mutagenesis studies have implicated the cystathionine-␤-synthase (CBS) 3 domain-containing C terminus of ClC-0 in its * This work was supported, in part, by grants from the Deutsche Forschungs-slow common gating (26), and more recent studies revealed C-terminal movements during common gating of ClC-0 and ClC-1 (27,28). However, mutations in other regions also affect common gating. For instance, the C212S mutation that totally abolishes slow gating of ClC-0 (29) involves a residue close to the extracellular face of the transmembrane segment.
The mammalian CLC Cl Ϫ /H ϩ exchangers (ClC-3 through ClC-7), which are predominantly expressed in the endosomallysosomal system (30), are strongly voltage-dependent and mediate measurable ion transport only at positive cytoplasmic potentials (14,(31)(32)(33)(34). As currents mediated by ClC-3 through ClC-6 respond very quickly to changes in membrane potential, it is difficult to attribute their voltage dependence either to a gating process that turns the ion exchange on and off or to an intrinsic voltage dependence of the exchange process. Using a plasma membrane-targeted mutant of the lysosomal ClC-7 anion/proton exchanger (35), we recently found that ClC-7/ Ostm1 currents respond very slowly to voltage changes (14). These slow current relaxations enabled us to show that the strong outward rectification of ClC-7/Ostm1 is owed to a voltage-dependent gating of the 2Cl Ϫ /1H ϩ exchange process that intrinsically shows an almost linear voltage dependence (14). Mutations either in ClC-7 or in its ␤-subunit Ostm1 lead to osteopetrosis, lysosomal storage disease, and neurodegeneration in mice and humans (12, 36 -39). Interestingly, several human CLCN7 mutations identified in osteopetrosis drastically accelerate the gating of ClC-7/Ostm1, suggesting that slow gating may be needed for its physiological function (14). Several of these mutations affect residues in the cytoplasmic CBS domains or in nearby loops of the membrane part of ClC-7. In the present work, we use some of these accelerating mutations, in combination with other mutants and biophysical analysis, to show that the gating of ClC-7/Ostm1 involves the common gate. By extension, these findings probably also apply to the other mammalian 2Cl Ϫ /1H ϩ exchangers in which gating cannot be studied easily.

EXPERIMENTAL PROCEDURES
Expression Constructs-All constructs for heterologous expression in Xenopus laevis oocytes were cloned into the pTLN expression vector (40) as described (14). Point mutations, insertions, and deletions were introduced into human ClC-7 by recombinant PCR. All constructs were confirmed by sequencing the complete open reading frame. To delete the ClC-7 C terminus, a stop codon was introduced at position 624 (ClC-7 C624X , or ClC-7 ⌬CT for short). C-terminal residues 617-805 were either expressed as a bona fide soluble protein (ClC-7 CT ) or fused to the last amino acid of Ostm1 (Ostm1-ClC-7 CT ).

RESULTS
As common gating is associated with a coordinate conformational change of both subunits of functional CLC dimers, mutations altering common gating may affect the gating of a wildtype (WT) subunit when present in just one subunit of a WT/mutant heteromer. Indeed, CLCN1 mutations identified in patients with dominant myotonia that shift the voltage of half-maximal activation (V1 ⁄ 2 ) of homodimeric ClC-1 also shifted the voltage dependence of WT subunits in WT/mutant heteromers (41) by changing common gating (18). In a more extreme case, the slow activation by hyperpolarization of ClC-2 was abolished in ClC-1/ClC-2 heterodimeric channels (40). We therefore argued that human CLCN7 mutations that lead to faster gating in ClC-7 homodimers (14) should accelerate gating of WT subunits in WT/mutant heterodimers if the activation process involves a common gate. However, changed gating kinetics of different subunits are difficult to disentangle upon 1:1 co-expression of WT and fast mutant subunits as we have to deal with a 1:2:1 mixture of WT/WT, WT/mutant, and mutant/ mutant subunits. One may try to solve this problem by forcing the formation of heterodimers by constructing concatemers. Although this approach was successfully applied previously in mixing CLC subunits of different unitary conductances (6,8,42), gating kinetics may easily be changed by linking the C terminus of one subunit to the amino terminus of the second one. Indeed, a human ClC-7 frameshift mutant (G796fs) (43) that replaced the last 10 C-terminal residues by 129 foreign ones accelerated ClC-7 gating (14).
We therefore opted to combine in a single subunit a fast mutation with another mutation that abolishes ion transport through the same subunit (see scheme in Fig. 1A). Co-expression of this virtually fast, but transport-deficient ("td") mutant at molar excess with WT subunits should result in currents that are dominated by WT subunits present in ClC-7/ClC-7 fast,td heteromers. We used the E314A mutation, which will be further referred to as td, to abolish ClC-7/Ostm1 transport activity (14). This mutation neutralizes a "proton glutamate" at the cytoplasmic face of the transmembrane part of ClC-7 that is thought to be required for the transport of cytoplasmic protons to the Cl Ϫ /H ϩ exchange site at the gating glutamate (44). Mammalian Cl Ϫ /H ϩ exchangers carrying such proton glutamate mutations do not measurably transport Cl Ϫ either (14,34,42), probably because Cl Ϫ lacks its exchange partner. This mechanism suggests that the block of ion transport is confined to subunits carrying the td mutation. Although not indicated in the following text, all ClC-7 subunits additionally carried the L23A,L24A and L68A,L69A mutations in N-terminal AP-and GGA-binding sites, respectively, which lead to partial plasma membrane localization of ClC-7 (14,35). Furthermore, all constructs were co-expressed with the essential ␤-subunit Ostm1 (12,14).
We first performed several control measurements to validate our experimental strategy. To confirm that currents of ClC-7/ ClC-7 td heteromers were only carried by subunits not containing the td mutation, we exploited the effect of the pore mutation S204P that selectively increases nitrate conductance (14,45,46). Coexpression of ClC-7 S204P with a molar excess of ClC-7 td gave rise to currents whose nitrate selectivity was indistinguishable from that of ClC-7 S204P homomers (Fig. 1B). Hence the td mutation abolishes currents only in the subunit in which it has been inserted, and this effect does not require that the adjacent subunit of the dimer also carries this mutation. In other control experiments, we ascertained that the td mutation abolished ClC-7 transport activity also when present together with the accelerating mutations R762Q or R286Q in the same subunit (Fig. 1C).
For our analysis, we chose three previously characterized (14) fast mutants of ClC-7: R762Q (36) and R767Q (47), mutations in the second CBS domain found in patients with recessive infantile osteopetrosis, and R286Q, which changes a residue in the intracellular loop between intramembrane helices F and G and that has been found in dominant osteopetrosis (47) (for localization of these mutations, see Fig. 1, D and E). We co-injected oocytes with ClC-7 and ClC-7 fast,td cRNAs at a 1:4 ratio (Fig. 2). ClC-7 currents were elicited by 2-s voltage steps to 80 mV. Activation and deactivation kinetics were determined by fitting single exponential equations to currents at the beginning and at the end of the depolarizing voltage step, respectively. Although present in a transport-deficient ClC-7 td subunit, all three fast mutations (R286Q, R762Q, and R767Q) markedly accelerated the activation kinetics of ClC-7 currents in trans through the attached WT subunit in ClC-7/ClC-7 td,fast heteromers. This acceleration did not reach the same levels as observed with homomeric ClC-7 fast /ClC-7 fast transporters (Fig.  2, A-D). However, this apparent difference in kinetics is at least partially owed to currents from WT ClC-7/ClC-7 homodimers, which are estimated to contribute about 20% with our 1:4 ClC-7:ClC-7 td,fast co-injection scheme. In the reverse experiment, we asked whether ClC-7 fast currents could be slowed down in trans by 1:4 co-expression with ClC-7 td mutants. This was indeed the case, and gating kinetics were similar between ClC-7/ClC-7 fast,td and ClC-7 fast /ClC-7 td expression schemes for all three accelerating mutants (Fig. 2, A-D). Likewise, deactivation kinetics of ClC-7 were accelerated in trans by ClC-7 td,fast mutants (Fig. 2, E-G). In control experiments, ClC-7 td did not affect gating kinetics of co-injected ClC-7, and ClC-7 fast /ClC-7 td,fast displayed gating kinetics similar to ClC-7 fast (Fig. 2).
We next asked whether we could further accelerate ClC-7 gating by inserting two fast mutations into the same subunit. We generated the ClC-7 R286Q,R762Q, R286Q,R767Q, and R762Q,R767Q double mutants, out of which only ClC-7 R286Q,R762Q gave rise to robust currents (Fig. 3A). Time constants of activation of ClC-7 R286Q,R762Q and ClC-7 R762Q did not differ from each other, showing that the effects of these mutations were not additive (Fig. 3B).
Most mutations accelerating the gating of ClC-7 localize to the interface between the cytosolic C terminus and transmembrane domains, where they are thought to disturb gating-relevant interactions (14). On the other hand, the CBS domaincontaining C terminus of ClC-7 is covalently attached to intramembrane helix R, which directly coordinates a Cl Ϫ ion in the CLC pore through the side chain of a tyrosine close to its N terminus (9). It is tempting to speculate that the ClC-7 C terminus influences gating by changing the position of the R-helix. We therefore asked whether the length of the R-CBS1-linker might influence ClC-7 gating. Both deletion and insertion of several amino acids after threonine 621 (Fig. 4A) indeed accelerated gating, whereas deleting 6 amino acids abolished ClC-7/ Ostm1 transport function (Fig. 4, B and C).
These results seemed to support the hypothesis that CBS domains modulate gating by "pulling" on the R-helix. We were curious to know what would happen if we entirely eliminated this possibility by disrupting the peptide bond between helix R and CBS1, using a "split transporter" approach akin to the "split channel" approach as used first with the ClC-1 Cl Ϫ channel (48). Expression of a ClC-7 mutant truncated shortly after the R-helix and thus lacking both CBS domains (ClC-7 ⌬CT ) did not yield currents (Fig. 5, A and B).
Surprisingly, currents of ClC-7 ⌬CT could be rescued by coexpressing the cytoplasmic C terminus (ClC-7 CT ) from a separate cRNA (Fig. 5, A and B). Resulting currents reached amplitudes similar to WT ClC-7, but their activation kinetics were slightly faster (Fig. 5, B and C). Instead of expressing the C terminus only as a soluble protein, we also fused it to the obligate ClC-7 ␤-subunit Ostm1 directly after its last residue to anchor it to the plasma membrane in close vicinity to the ClC-7 backbone. Co-expressing this fusion protein (Ostm1-ClC-7 CT ) with ClC-7 ⌬CT (without additional WT Ostm1) yielded currents that were indistinguishable from WT ClC-7/Ostm1 in their kinetics (Fig. 5C) and often reached similar amplitudes. We also inserted the fast R762L mutation into CBS2 of the isolated C terminus ClC-7 CT and into the Ostm1-ClC-7 CT fusion protein. Although the combination of ClC-7 ⌬CT ϩ ClC-7 CT(R762L) failed to yield currents (data not shown), split channels in which the mutant C terminus was attached to Ostm1 (ClC-7 ⌬CT ϩ Ostm1-ClC-7 CT(R762L) ) produced robust currents that displayed activation time constants indistinguishable from the fast mutant ClC-7 R762L ϩ Ostm1 (Fig. 5C). We conclude that the role of the ClC-7 C terminus in ClC-7 gating does not depend on a covalent linkage to the last intramembrane helix R.
In view of these results, we re-examined the effects of linker deletions (Fig. 4) and expressed split transporters in which the polypeptide chain following helix R was truncated at different positions (Fig. 5D). These experiments revealed that gating kinetics became significantly faster when ClC-7 was truncated at position 621 or earlier. Hence the results of our linker shortening might be owed to specific effects of deleting particular   Fig. 1. B, individual and averaged activation time constants ( act ) obtained from those experiments. Thick lines indicate arithmetic mean, and thin lines indicate Ϯ S.E. The ␤-subunit Ostm1 was always co-expressed. The voltage clamp protocol is as in Fig. 1. Thick lines indicate arithmetic mean, and thin lines indicate Ϯ S.E. FIGURE 5. The ClC-7 C terminus is crucial for ClC-7 function and can be supplied as a separate polypeptide chain. A, typical recordings of oocytes expressing ClC-7 ⌬CT in which ClC-7 is truncated shortly after the last membrane helix and of oocytes co-expressing this mutant with either the C terminus (ClC-7 CT ) or a chimeric protein (Ostm1-ClC-7 CT ) in which the C terminus is fused to the ␤-subunit Ostm1. WT Ostm1 was co-expressed in the first two cases. The voltage clamp protocol is as in Fig. 1. B, mean near steady-state current-voltage relationships (evaluated at 2 s after voltage steps) obtained from truncated constructs compared with controls. C, activation time constants ( act ) of currents from experiments as above, and from ClC-7 ⌬CT ϩ Ostm1-ClC-7 CT-R762L in which a ClC-7 C terminus carrying the fast mutation R762L in CBS2 is fused to the Ostm1 ␤-subunit. D, activation time constants of WT ClC-7 compared with split transporters in which the truncation point after helix R is varied. ClC-7 1-623 corresponds to ClC-7 ⌬CT . Constructs were always co-expressed with ClC-7 CT and with Ostm1. Thick lines indicate arithmetic mean, and thin lines indicate Ϯ S.E. Significance levels (one-way analysis of variance): ***, p Ͻ 0.001. amino acids, although the mechanism by which this changes gating remains unclear.

DISCUSSION
A fascinating aspect of the CLC family is that its members transport ions in modes traditionally thought to be radically different; they function either as ion channels or as strictly coupled ion exchangers. However, remnants of an ion exchange activity can be found in CLC Cl Ϫ channels because the gating of ClC-0 may be associated with H ϩ transport (49). Further, the strong voltage dependence of Cl Ϫ /H ϩ exchange activity of vesicular CLCs resembles the gating of ion channels, as recently demonstrated for plasma membrane-expressed ClC-7 (14). The intrinsically almost linear anion/proton exchange activity of this transporter is very slowly turned on by depolarization in a process that strongly resembles gating of ion channels.
We now asked whether the gating of ClC-7/Ostm1 resembles the ClC-0 protopore gate, which affects one pore at a time, or its common gate, which closes both pores of the doublebarreled channel simultaneously. We argued that mutations accelerating ClC-7 gating might impose their faster gating kinetics on associated WT subunits in heterodimers if the gating process involves both subunits. We devised experiments in which one of the subunits of the dimer was rendered transportdeficient by another mutation to better isolate the effect on the currents of the other subunit. Such transport-deficient mutants indeed partially imposed their gating kinetics on the other, transporting subunit, rendering their activation and deactivation faster or slower, depending on the particular combination. These results resemble previous work in which dominant ClC-1 mutants partially imposed their shifted voltage dependence (as observed in homomeric channels) on the gating of heteromers through an effect on common gating (18,41). We conclude that voltage-dependent activation of ClC-7/Ostm1 prominently involves common gating of both subunits. A minor contribution of protopore gating, however, cannot be excluded.
Many accelerating mutations cluster in cytoplasmic CBS domains and the cytoplasmic surface of the ClC-7 transmembrane segment (14), suggesting that interactions between these parts of the protein influence gating. We have also explored another possible mechanism by which mutations in the C terminus could affect gating, i.e. through a direct "pull" of the C terminus on the last intramembrane helix R. This speculation is attractive because the N-terminal part of helix R participates in coordinating a Cl Ϫ ion in a binding site in the permeation pathway as revealed by crystal structures of prokaryotic CLCs (9,23). Consistent with this idea, Förster resonance energy transfer (FRET) experiments using ClC-0 and ClC-1 to whose C termini GFP variants had been fused indicated large movements of those domains during gating (27,28). Although initial experiments in which we varied the length of the R helix-CBS1 linker seemed to support this hypothesis, our subsequent split transporter experiments clearly indicated that this is not the case.
These split transporter experiments indicated that the ClC-7 C terminus binds efficiently to the ClC-7 backbone even without being anchored nearby at the plasma membrane (as in the Ostm1-ClC-7 CT fusion protein). The crystal structure of the CBS domain-containing algal CLC (10) suggests that this binding prominently involves CBS2, a domain in which several osteopetrosis-causing mutations were found (14,36,43,47). The similar gating kinetics of currents obtained from WT and ClC-7 ⌬CT co-expressed with either Ostm1-ClC-7 CT or the detached C terminus ClC-7 CT suggest that the C terminus does not fully dissociate from the ClC-7 backbone during gating because its dilution in the cytoplasm is expected to severely slow down its reassociation. This should then be accompanied by a severe deceleration of either activation or deactivation, depending on whether the binding of the C terminus "closes" or "opens" the transporter, respectively. Interestingly, when we inserted fast mutations into CBS2, we observed functional split transporters only when the mutated C terminus was attached to Ostm1 (Fig. 5C), but not when it was supplied as a soluble protein. Hence these mutations may destabilize the binding of the C terminus to the transporter backbone with reasonably efficient binding occurring only when the local concentration of the C terminus is drastically increased by tethering it to Ostm1.
Split channel (respectively, transporter) approaches have been used previously with ClC-0 (50), ClC-1 (48,(51)(52)(53)(54), and ClC-5 (55). These studies agreed in that the function of CLC proteins that were truncated after CBS1 could be rescued by co-expressing the missing fragment containing CBS2. This complementation can be easily explained by the tight CBS1-CBS2 binding that was revealed by crystal structures of CLC CBS domains (10,56). Indeed, CBS domains also form internal dimers in other CBS domain-containing proteins (57). A complementation similar to the one described here, i.e. the functional rescue of a CLC truncated directly after the R helix by the complete soluble C terminus, had been tried previously with the skeletal muscle Cl Ϫ channel ClC-1 (48,54). However, both studies agreed that the coexpression of both fragments failed to yield functional channels. Hence the binding affinity of the backbone to the C-terminal tail seems to be stronger in the case of ClC-7.
Common gating requires a coordinate conformational change of both subunits of the dimer that ultimately leads to an occlusion of the permeation pathway. This necessitates both changes at the interface between both subunits and of residues in the pore region. CLC monomers contact each other not only at their broad membrane-embedded interface (9,10), but also between the CBS domains of either subunit (10,58). Consistent with these considerations, FRET detected large movements of CBS domains during common gating of ClC-0 (27), 19 F NMR indicated substrate-driven changes in the position of a tyrosine at the membrane-embedded dimer interface of EcClC-1 (59), and mutations in the gating glutamate in the pore abolish both common and protopore gating of e.g. ClC-2 (21). However, our picture of common gating remains appallingly incomplete. In particular, the way in which mutations in CBS domains affect this gating process remains enigmatic.
Many mutations, not just those found in patients with osteopetrosis, accelerate ClC-7 gating, whereas we have not yet obtained ClC-7 mutants that gate more slowly than WT. 4 Evolution may have maximized the time constant of ClC-7 gating that is several orders of magnitude slower than with ClC-3 to ClC-6. Together with the fact that many disease-causing mutations accelerate ClC-7 gating (14), these observations suggest that slow gating of ClC-7 may be important for its biological function. If accelerated gating by itself were pathogenic, the acceleration of common gating in WT/fast heteromers might contribute to the dominant inheritance pattern of some osteopetrosis-causing CLCN7 mutations. However, some accelerating mutations may also decrease protein stability, as shown previously (36) for the R762Q mutant we have studied here. Indeed, this particular accelerating mutation was found in a patient with recessive osteopetrosis (36).
In summary, the gating of ClC-7/Ostm1 bears parallels to the common, slow gate of the Torpedo Cl Ϫ channel ClC-0 that affects both pores of the dimeric channel simultaneously. It depends on the presence of the CBS domain-containing C terminus that interacts with the transmembrane backbone and modulates gating kinetics even in the absence of a covalent link. Our work extends the role of common gating from CLC Cl Ϫ channels to Cl Ϫ /H ϩ exchangers.