Structure and gating of CLC channels and exchangers

Structure of the transmembrane region of the CLC exchangers

Over the last 13 years the high-resolution crystal structures of four CLC exchangers have revealed the structural organization of the CLCs and elucidated the molecular basis of anion binding and transport. Three of these four homologues are from prokaryotic organisms, CLC-ec1, CLC-st1 (Dutzler et al. 2002, 2003) and CLC-cy1 (Jayaram et al. 2011), while the fourth, cmCLC (Feng et al. 2010, 2012), is from the thermophilic alga Cyanidioschyzon merolae. The structures of these four family members are nearly identical (Fig. 1A–C), apart from the position of a critical glutamate residue found within the ion permeation pathway (Fig. 1D–E). Remarkably, the structures of these CLC exchangers shed profound insights into the transport and gating of the CLC channels, indicating that the two subtypes share a common structural fold despite functioning according to opposite thermodynamic paradigms.

Figure 1

Structure of the CLCs

A, structure of the CLC-ec1 dimer viewed from the plane of the membrane. The internal symmetry repeats of the left monomer are shown in blue and cyan. Cl ^{ions are shown as magenta spheres. Neighbouring helices A and R from opposing subunits are labelled.} B, surface representation of a CLC-ec1 monomer viewed from the intracellular side to highlight the contribution of the internal symmetry repeats to the formation of the ion transport pathway. The repeats and Cl ^{ions are coloured as in} A. C, structure of the cmCLC dimer viewed from the plane of the membrane. D–F, close up views of the three conformations of the ion binding region of WT CLC-ec1 (D), E148Q CLC-ec1 (E) and WT cmCLC (F). Glu_ex is shown in magenta, Tyr_cen and Ser_cen are shown in cyan and the Cl ^{ions are shown as green spheres. PDB ID: WT CLC-ec1 (1OTS), E148Q CLC-ec1 (1OTU) and WT cmCLC (3ORG).}

Consistent with previous biochemical and electrophysiological data (Ludewig et al. 1996; Middleton et al. 1996; Maduke et al. 1999) all CLCs assemble, function and crystallize as homodimers (Fig. 1A and C). Unexpectedly, recent work has shown that CLC-ec1 can also fold, exist in a membrane and function as a monomer (Robertson et al. 2010). Whether all – or any other – CLC homologues can be split into their monomeric constituents while preserving function remains to be seen. Each subunit comprises 18 α-helices, helices A–R, which form an internal antiparallel repeat (Fig. 1A and B), reminiscent of those found in many other transport proteins (Fu et al. 2000; Yernool et al. 2003; Yamashita et al. 2005; Forrest & Rudnick, 2009; Forrest et al. 2011). The first and last helices, A and R, are amphipathic and helix A from one subunit contacts helix R of the other in a small domain swap (Fig. 1A). The functional implications of this interaction are unclear as in the crystal structure of the CLC-ec1 monomer, helix A faces the inter-subunit interface in a drastically different conformation (Robertson et al. 2010), with little impact on function. Furthermore, in cmCLC helix A is not resolved and therefore its position within the context of a full-length CLC homologue is unclear. The remaining 16 transmembrane helices form the Cl ^{transport pathway at the interface between the two internal repeats (}Fig. 1B).

The anion transport pathway

Each CLC monomer forms three anionic binding sites (Fig. 1D–F): ions bound to the internal and external sites, S_in and S_ex, are directly exposed to the intra- and extracellular solutions, respectively. The central site, S_cen, is located near the centre of the bilayer and is shielded from extracellular water molecules by a conserved glutamate, Glu_ex, and is cut-off from the intracellular solution by a constriction formed by the side chains of a conserved serine, Ser_cen, and tyrosine, Tyr_cen (Fig. 1D). S_ex and S_cen can be alternatively occupied by a dehydrated Cl ^{ion or by the, presumably deprotonated, side chain Glu}_ex. The ion binding region can exist in three different conformations (Fig. 1D–F), which differ almost exclusively for the position of Glu_ex. In WT CLC-ec1 (Fig. 1D) a Cl ^{ion occupies S}_cen and Glu_ex occupies S_ex and occludes the conduction pathway towards the outside. The structure of the E148Q mutant of CLC-ec1 (Fig. 1E), which presumably reflects the state assumed by the pathway when E148 becomes protonated, shows a conformation in which the Q148 side chain moves out of the pathway, opening it towards the extracellular solution, and all three binding sites are occupied by Cl ^{ions. This state is also thought to reflect the open conformation of the CLC channels, consistent with the finding that the charge-neutralized mutants of Glu}_ex in CLC-0, -1 and -2 are constitutively open channels (Dutzler et al. 2003; Traverso et al. 2003; Zúñiga et al. 2004). The third state is revealed by the structure of cmCLC (Fig. 1F): in this configuration a Cl ^{occupies S}_ex while Glu_ex slides into S_cen, where it is presumably stabilized by a hydrogen bond with Tyr_cen (Feng et al. 2010), and occludes the pathway towards the intracellular side.

The three binding sites display distinct ion coordination: anions bound to S_cen are coordinated by the hydroxyl oxygen atoms of Ser_cen and Tyr_cen and by the main-chain amide nitrogen atoms of G149, I356 and F357 (CLC-ec1 numbering) (Dutzler et al. 2002). In contrast anions in S_ex and S_in are mostly coordinated by backbone atoms, with the exception of a hydrogen bond between Ser_cen and a Cl ^{in S}_in, and remain partly hydrated (Dutzler et al. 2003; Lobet & Dutzler, 2006). Two of the three sites, S_cen and S_ex, have similar Cl ^{binding affinities,} K_d ∼1 mm, while S_in binds ions very weakly, K_d > 30 mm (Lobet & Dutzler, 2006; Picollo et al. 2009). Ion binding to S_cen is disrupted when Tyr_cen is replaced by residues with smaller side chains, such as leucine or alanine, with the former increasing the K_d to ∼4 mm and the latter abolishing binding almost completely (Picollo et al. 2009). While mutations that selectively affect ion binding to S_in have not yet been identified, the S107G/Y445A/E148A triple mutant of CLC-ec1 simultaneously disrupts S_cen and S_in (Lobet & Dutzler, 2006), suggesting that Ser_cen might be important for binding to S_in. It is not known whether the S107G mutation alone is sufficient to disrupt S_in, but the S107P mutant nearly eliminates Cl ^{binding in CLC-ec1 suggesting that this residue might play a role in regulating ion binding to both S}_in and S_cen (Picollo et al. 2009).

Tunable substrate specificity of the ion pathway

Despite poor sequence conservation among CLC family members, the residues that directly line the ion binding/transport pathway are relatively well conserved, for example the amino acids that form S_cen are among the most highly conserved within the family: Ser_cen is found in ∼42% of the deposited CLC sequences, Glu_ex in ∼88%, F357 in ∼57%, while at position 356 I, L or V are found in 73% of the sequences. Consistent with these observations, most CLCs share a signature selectivity sequence of SCN ^{> Cl > Br > NO}₃ ^{> F > I (Jentsch} et al. 1995; Rychkov et al. 1998; Accardi et al. 2004; Nguitragool & Miller, 2006), regardless of whether they are channels or transporters. The two notable exceptions to this pattern yielded surprising insights into the mechanisms of substrate selectivity in the CLC family: atCLC-a, a homologue from the plant Arabidopsis thaliana, is a 2:1 NO₃^{:H exchanger (De Angeli} et al. 2006) and members of a newly identified clade of prokaryotic CLCs, the CLC-F, are fluoride selective and operate as 1:1 F^{:H exchangers (Stockbridge} et al. 2012).

While the CLCs have three anionic binding sites, it appears that S_cen is primarily responsible for their inter-anionic selectivity as neither S_ex nor S_in are very selective (Picollo et al. 2009), with the latter displaying a slight preference for SCN ^{over Cl (Nguitragool & Miller,} 2006; Picollo et al. 2009). Ions bound to S_cen are coordinated by a combination of backbone and side chain interactions (Fig. 1D). Consistent with early electrophysiological work, Ser_cen is the most important determinant of substrate specificity in both the CLC channels and transporters: its replacement with a proline in the CLC-5 and CLC-ec1 transporters and in the CLC-0 channel switches selectivity from Cl ^{to NO}₃ ^(Bergsdorf et al. 2009; Picollo et al. 2009; Zifarelli & Pusch, 2009a) and conversely, the P160S mutant of atCLC-a is Cl ^{selective rather than NO}₃ ^{selective like its WT counterpart and impairs the ability of plants expressing this mutant to accumulate NO}₃ ^{in vacuoles (Bergsdorf} et al. 2009; Wege et al. 2010).

Interestingly, the newly identified CLC-F clade of transporters diverge drastically from their Cl ^{or NO}₃ ^{selective counterparts in the C–D loop where Ser}_cen is located. In these transporters, the canonical G(S/P)GIP sequence is replaced by X-G(N/M)X and helix C is shortened by three amino acids (Stockbridge et al. 2012; Brammer et al. 2014), possibly enlarging the cavity around S_cen. Thus, more drastic structural rearrangements might be needed to accommodate F ^{ions rather than Cl or NO}₃^{, likely to be because of the higher energy costs associated to F dehydration, ΔH ∼−125 kcal mol, compared to Cl, ∼90 kcal mol. How the CLC-Fs succeed in selectively dehydrating, binding and transporting F but not Cl remains unclear as does whether permeating F ions need to shed their water molecules.}

Taken together these results indicate that the substrate specificity of the CLCs is remarkably tunable and that, unlike what is seen in K ^{and Na channels (Doyle} et al. 1998; Zhou et al. 2001; Payandeh et al. 2011), specific interactions between side chains and the ions themselves play a key role in determining the selectivity of these transport proteins. The varied and tunable substrate specificity of the CLC family allows them to adapt and fulfil diverse physiological roles, such as controlling Cl ^{homeostasis in cellular compartments, nutrient accumulation and extrusion of the toxic F ion from the cytosol of bacteria.}

Gating mechanisms

Two mechanisms regulate gating of the CLC channels and transporters: a common gate, which simultaneously regulates the activity of both monomers, and a single pore gate, which regulates the activity of the individual subunits (Miller, 1982; Jentsch et al. 1990). In most CLCs, both processes are voltage dependent, although even the sign of this dependence varies between different homologues (Accardi & Pusch, 2000; Niemeyer et al. 2003, 2009; Zúñiga et al. 2004; de Santiago et al. 2005; Ludwig et al. 2013). Opening of the common gate requires a concerted conformational change of both subunits and involves both the transmembrane and C-termini (Bykova et al. 2006; Garcia-Olivares et al. 2008). In contrast, gating of the individual monomers entails rearrangements of a smaller scale (Dutzler et al. 2003). In addition to voltage, the CLCs are also modulated by changes in the intra- and extracellular concentrations of H ^{and Cl (Hanke & Miller,} 1983; Pusch et al. 1995; Chen & Miller, 1996; Rychkov et al. 1998; Chen & Chen, 2001, 2003; Zdebik et al. 2008; Zifarelli et al. 2008; Picollo et al. 2010; Leisle et al. 2011). While each homologue has distinct properties, gating of the CLC channels shares many characteristics with the transport cycle of the transporters, suggesting a strong evolutionary link between channel opening and the transport cycle (Miller, 2006).

The available crystal structures have provided insights into the ion conduction and selectivity mechanisms and, together with extensive functional work, have highlighted the central role played by Glu_ex in CLC channel gating and transport: in order for the channels to gate or for coupled transport to occur this position must be occupied by a protonatable residue (Dutzler et al. 2003; Traverso et al. 2003, 2006; Accardi & Miller, 2004; Feng et al. 2010). However, since the transmembrane regions of the various homologues and mutants adopt identical conformations, the question of what, if any, conformational changes – in addition to the above described movement of Glu_ex – are required for transport remains open. The lack of crystallographically resolved structural rearrangements led to the hypothesis that Glu_ex is the sole gate regulating ion flux through both CLC channels and transporters and that its movements in and out of the pathway (Fig. 1D–F) are the only relevant conformational changes taking place during gating (Dutzler et al. 2003; Feng et al. 2010, 2012). This minimal gating mechanism is consistent with functional investigations showing that each subunit functions independently (Robertson et al. 2010) and that transport does not require major conformational rearrangements between them (Nguitragool & Miller, 2007; Zdebik et al. 2008). A single gate hypothesis is also appealing as it readily rationalizes how one structural scaffold can support both secondary active transport and passive diffusion. However, such a mechanism breaks one of the basic rules of active transport: substrates should never be in simultaneous contact with both sides of the membrane (Jardetzky, 1966). Violation of this rule would result in uncontrolled substrate slippage, causing the dissipation of the gradients that the transporters create. Indeed, in most transporters extensive conformational rearrangements ensure the alternate exposure of substrates (Morth et al. 2007; Olesen et al. 2007; Reyes et al. 2009; Forrest et al. 2011; Kaback et al. 2011; Krishnamurthy & Gouaux, 2012; Perez et al. 2012; Shi, 2013) and disruption of these coordinated movements adversely impacts function. Thus, a mechanism to prevent free diffusion and slippage is necessary.

Recently, two models have been proposed to solve this conundrum. In the first model (Fig. 2A) ion movement is regulated at one end of the pathway by Glu_ex and at the other by a static kinetic barrier (Feng et al. 2010, 2012), possibly formed by the steric constriction in the Cl ^{transport pathway created by Ser}_cen and Tyr_cen. This barrier slows free diffusion of Cl ^{enough that during a protonation–deprotonation event of Glu}_ex, only two Cl ^{ions, on average, will go through the ‘open’ CLC transporter. The second, more conventional, model (}Fig. 2B) proposes that in addition to the Glu_ex gate the CLCs also have an intracellular gate (i.e. a barrier whose height depends on the state of the transporter) which is formed by Tyr_cen. I will describe the two models and discuss their relative strengths and weaknesses.

Figure 2

Models for the transport cycle of the 2 Cl^{: 1 H CLC exchangers}

A, Model I is drawn according to the proposal of Feng et al. (2010). B, Model II is drawn according to the proposal of Basilio et al. (2014).

The six states comprising Model I (Fig. 2A) are based on the crystal structures of cmCLC and CLC-ec1: states I and II reflect the unprotonated and protonated forms of the conformation observed in CmCLC (Fig. 1F), state III is a transient intermediate towards states IV and V, which respectively correspond to the protonated and unprotonated conformation of the E148Q mutant of CLC-ec1 (Fig. 1E), and state VI represents the structure of WT CLC-ec1 (Fig. 1D). The cycle starts when an intracellular proton reaches Glu_ex, which occupies S_cen (I–II), leading to its displacement (II–III) allowing two Cl ^{ions to enter the ion pathway from the extracellular side (III–IV). Deprotonation of Glu}_ex (IV–V) favours its transition back into the permeation pathway displacing two Cl ^{ions towards the intracellular side (V–VI–I). In this scheme the key variable is the balance between the height of the kinetic barrier and the deprotonation rate of Glu}_ex in state IV. The former has to be sufficiently high to prevent significant diffusion of Cl ^{while Glu}_ex is protonated and at the same time is has to be sufficiently low to allow for the rapid turnover seen in some CLCs. Simulations show that this model accounts for the fundamental properties of the CLC exchangers: their 2 Cl^{:1 H stoichiometry and the ability of either ion to drive uphill transport of the other (Feng} et al. 2010, 2012). Importantly, model I is also appealing for its unifying view of the CLC channels and transporters, as the former are nothing more than CLCs in which the intracellular kinetic barrier is so low that millions of ions can cross it during a protonation event of Glu_ex. However, Model I does not capture well other key properties of the CLCs. For example, it predicts that the transporters should become progressively uncoupled in acidic environments, while CLC-ec1 remains coupled at pH values as low as 3 (Accardi et al. 2004; Accardi & Miller, 2004) and most CLC transporters carry out their physiological functions in acidic environments. Similarly, it is not clear how in Model I the stoichiometry can remain constant when the turnover rate of the transporters varies by over three orders of magnitude, from ∼10 s ^{to ∼10 s, between different CLC homologues (Accardi & Miller,} 2004; Zdebik et al. 2008; Jayaram et al. 2011; Phillips et al. 2012). Finally, the extremely local conformational changes assumed in Model I (just the movement of Glu_ex) are difficult to reconcile with experiments showing that conformational changes in regions distal to the ion pathway occur during transport (Accardi & Pusch, 2003; Bell et al. 2006; Osteen & Mindell, 2008; Elvington et al. 2009). Despite these shortcomings, Model I is appealing for its elegant simplicity and because it quantitatively reproduces some of the key features of the CLC transporters.

More recently, a second model has been proposed to describe transport by the CLCs (Fig. 2B) (Basilio et al. 2014). This model accounts for the finding that Tyr_cen (and possibly other residues nearby) act as the intracellular gate of the CLC exchangers and that a conformational change of helix O is required for transport, likely as a step in the allosteric pathway coupling the inner and outer gates. These new findings led to the formulation of an alternative model. In Model II, state I represents the transporter in its apo and occluded state. Opening of the inner gate (I–II) allows two Cl ^{ions to bind from the intracellular side displacing Glu}_ex (II–III). This favours its protonation from the outside and triggers the closure of the inner gate, thus yielding a fully loaded state of the transporter in its outward-facing conformation (III–IV). Then Glu_ex re-enters the ion transport pathway, while still protonated (IV–V), and the proton is transferred from Glu_ex to the intracellular solution via the internal glutamate (E203 in CLC-ec1) (V–VI) to be released towards the intracellular solution and returning the transporter to its apo configuration (VI–I). In addition to describing the basic features of CLC transport, such as the stoichiometry and the tight coupling, Model II also captures some of the more unusual mechanistic features of the CLCs. For example, the synergism between Cl ^{and H binding gives rise to a fully loaded state (IV) in which 2 Cl and 1 H are simultaneously bound to the transporter, a state that is explicitly forbidden for classical ‘ping-pong’ alternating access exchange mechanisms. Within the framework of Model II, the CLC channels would originate from transporters in which the coupling between the two gates is lost. Consistent with this hypothesis, recent work has shown that in the muscle homologue CLC-1, Tyr}_cen plays a key role in the common pore gate (Bennetts & Parker, 2013). One of the key features of Model II is that Tyr_cen serves as the pH-independent inner gate, a structural element that would allow the CLCs to maintain a constant exchange stoichiometry regardless of pH and of the turnover rate. In order to function, however, Model II makes several assumptions that require further testing: first, what triggers opening of the inner gate in the apo state (I–II)? Second, how are the two gates coupled? Besides the observation that helix O plays a role in this process (Basilio et al. 2014), there is no clear evidence in the literature suggesting how this could occur or if the gates are coupled solely via the ions in the permeation pathway.

Despite their significant differences, both models share numerous similarities and capture the essential features of the CLC transporters (exchange stoichiometry and tight coupling) as well as offering reasonable frameworks for the evolutionary relationship between the channels and exchangers found within the CLC family. Both models also make some assumptions that remain untested and additional work is required to further refine our understanding of how these proteins gate.

Structure of the cytoplasmic domains

In addition to the membrane-embedded transport domain, all eukaryotic CLCs and nearly half of the prokaryotic ones have large cytoplasmic domains which contain two cystathionine β-synthetase (CBS) domains. These domains regulate transport of both CLC channels and exchangers and numerous disease-causing mutations cluster to the CBS regions. Indeed, nucleotide binding to these domains regulates the activity of CLC-1 (Bennetts et al. 2005, 2007, 2012; Tseng et al. 2007; Zhang et al. 2008), CLC-5 (Wellhauser et al. 2006; Meyer et al. 2007; Zifarelli & Pusch, a2009; Wellhauser et al. 2011) and atCLC-a (De Angeli et al. 2009) and conformational changes in these regions have been linked to common gating in CLC-0 and CLC-1 (Bykova et al. 2006). Furthermore, mutations in the CBS domains of the channels CLC-1 and CLC-Ka/-Kb respectively cause dominantly inherited myotonia (Thomsen’s disease) (Steinmeyer et al. 1991; Pusch, 2002; Estévez et al. 2004) and Bartter’s syndrome (Estévez et al. 2001; Krämer et al. 2008). Similarly, mutations in the cytosolic regions of the CLC-5 and CLC-7 exchangers result in Dent’s disease (Lloyd et al. 1996, 1997a,b1997) and infantile malignant osteopetrosis (Kornak et al. 2001, 2006).

The structures of the isolated CBS domains from CLC-0, CLC-Ka and CLC-5 (Meyer & Dutzler, 2006; Markovic & Dutzler, 2007; Meyer et al. 2007) together with the recent structure of the full length eukaryotic transporter cmCLC (Feng et al. 2010) revealed the molecular organization of these cytosolic regions and their interactions with the transmembrane domain (Fig. 1C). The CBS are small domains, ∼50 amino acids long, with low sequence conservation and characterized by a fold composed of three β strands and two α helices (Estévez et al. 2004). The cytoplasmic domains of CLC-5 (Fig. 3A) and cmCLC (Fig. 3B) show that the C-termini assemble to form a dimer with an extensive interaction surface, mostly formed by the β2 and β3 sheets of the CBS2 domains, and characterized by a 2-fold axis of non-crystallographic symmetry. Mutagenesis, crosslinking and analytical ultracentrifugation experiments demonstrated that this surface is conserved also in CLC-0, even though its cytoplasmic domain crystallized in monomeric form (Markovic & Dutzler, 2007).

Figure 3

Structures of the CBS domains with and without ATP

A, structure of the isolated CBS domains from CLC-5. Bound ATP molecules are shown as CPK coloured, spacefilled models. B, structure of the CBS domains of cmCLC. C, close up view of the interactions between the C-terminal and TM domains of cmCLC. The important helices are labelled. D, detailed view of the interactions between ATP and the CBS domains of CLC-5: the α-helices and β-sheets from each CBS domain are coloured (the α-helices from CBS1 and 2 are respectively shown in yellow and orange and the β-sheets are shown in green and cyan). For clarity only the secondary structural elements from CBS1 are labelled. The ATP molecule is shown as a CPK coloured, spacefilled model. PDB ID: WT cmCLC (3ORG) and the isolated C-terminus of CLC-5 in complex with ATP (2J91).

In cmCLC the extensive interactions between the transmembrane domain and the cytosolic region are mostly mediated by CBS2, while CBS1 is located far from the membrane and the linker connecting it to helix R winds around CBS2 (Fig. 3C). The extensive interface between the TM and cytosolic regions of cmCLC has a high complementarity index indicating a tight interaction between the two regions (Feng et al. 2010). The CBS domains interact with the cmCLC TM region at three locations (Fig. 3C): helix R is directly connected to CBS1, CBS2 contacts helix D and the loop connecting helices H and I at the dimer interface. Since helices D and R directly line the ion pathway, respectively contributing Ser_cen and Tyr_cen (Fig. 1D), it is plausible to hypothesize these contacts are responsible for transducing rearrangements of the CBS domains directly to the ion pathway. Similarly, the contact between CBS2 and the H–I loop can rationalize how the cytosolic domains influence processes involving both subunits, such as the common gating process displayed by CLC channels. Indeed, several mutations in human CLCs that are thought to act through the common gate localize to the inter-subunit interface between the TM and CBS domains (Estévez & Jentsch, 2002; Feng et al. 2010).

Structural basis of nucleotide gating of the CLCs

Although the structure of the full length cmCLC and those of the isolated domains revealed the molecular organization of these domains, the mechanisms by which the CBS domains modulate transport remain to be elucidated. Recent findings showing that nucleotides regulate CLC gating by binding to the cytosolic CBS domains (Bennetts et al. 2005, 2007, 2012; Wellhauser et al. 2006; Meyer et al. 2007; Tseng et al. 2007; Zhang et al. 2008; De Angeli et al. 2009; Zifarelli & Pusch, 2009a). Interestingly, ATP binding to different CLC homologues elicits divergent responses: it inhibits the CLC-1 channel and the atCLC-a exchanger while it potentiates the CLC-5 transporter. Despite these differences, the interactions of the nucleotides with the CBS domains of the CLCs share the same basic characteristics: binding occurs with low affinity (100–1000 μm), nucleotides with the same base bind with similar affinities (irrespective of the number of phosphates), the binding site is non-catalytic, and the nucleotides are allosteric modulators of the common gate (Bennetts et al. 2007; De Angeli et al. 2009; Zifarelli & Pusch, 2009b). Not all CLCs bind and are modulated by nucleotides, for example CLC-0, CLC-Ka, CLC-7 and cmCLC are unaffected by ATP or its analogues (Meyer & Dutzler, 2006; Markovic & Dutzler, 2007; Feng et al. 2010; Leisle et al. 2011).

The structures of the C-terminal domain of CLC-5 in complex with ATP (Fig. 3D) or with ADP revealed the principles of nucleotide binding to the CLC CBS domains. The nucleotide binding site is located at the interface between CBS1 and -2 in a head-in configuration (Fig. 3D) with the adenine base inserted into a deep crevice formed by hydrophobic residues from both CBS domains. The phosphate tail is partly water exposed and makes only weak contacts with the protein, rationalizing the observation that ATP, ADP and AMP have similar affinities for CLC-5 and CLC-1 (Bennetts et al. 2005; Meyer et al. 2007) while nucleotides with different bases (cAMP, IMP and GTP) bind weakly or are inert (Bennetts et al. 2005).

Functionally, nucleotides have very diverse effects on different CLC homologues. While application of millimolar concentrations of intracellular ATP potentiates CLC-5 currents by increasing the probability of the transporter to be in an active conformation (Zifarelli & Pusch, b2009) the same manoeuvre inhibits the skeletal muscle channel CLC-1 by reducing the open probability of the common pore gate (Bennetts et al. 2005). This inhibitory action is exerted through two distinct mechanisms: ATP shifts activations towards positive voltages and lowers the minimal open probability of the common gate. The inhibitory potency of ATP strongly depends on pH and redox potential (Bennetts et al. 2007; Tseng et al. 2007; Zhang et al. 2008). Despite extensive mutagenesis experiments, the amino acids responsible for these modulatory effects have not yet been identified suggesting that multiple residues might be involved in these processes (Zhang et al. 2008). While the precise physiological implications of the nucleotide modulation of CLC-1 remain to be elucidated, it has been proposed that they would allow the channel to be a finely tuned sensor of the metabolic state of skeletal muscle cells. Finally, the plant CLC homologue atCLC-a is inhibited by saturating intracellular ATP, for a ∼2-fold reduction in current (De Angeli et al. 2009). In contrast to CLC-1 and CLC-5, in which only the nucleotide base matters, for atCLC-a the number of phosphates in the nucleotides also play a key role: ATP inhibits transport, ADP is inert and AMP, while incapable of eliciting a response on its own, competes with ATP and diminishes its inhibitory effect (De Angeli et al. 2009). The structural bases for the differential effects of nucleotides on atCLC-a remain unclear since all key residues for nucleotide binding are conserved and homology modelling suggested that nucleotides bind to the two domains in similar configurations.

Outlook

Ten years after the molecular identification of CLC-0, the first high-resolution structure of a CLC transport protein completely revolutionized our understanding of CLC function. In the nearly 15 years since, numerous structures elucidated the complex architecture of the CLCs, allowed the identification of key regions regulating transport, provided glimpses into the selectivity mechanism and revealed the basis for nucleotide binding and modulation. The unexpected discovery that the CLC family is split between channels and exchangers provided insights into the mechanistic boundary separating these two types of transport proteins and upended our understanding of the physiological roles of intracellular CLCs. Yet, despite these major strides, the original and fundamental question of what does a CLC channel look like remains unanswered. Which structural features distinguish the channels from the exchangers? What rearrangements occur during transport and gating? Answering these questions is likely to lead to more unexpected discoveries and point us in unforeseeable directions.

Competing interests

There are no competing financial interests.