I. INTRODUCTION

Top Previous Next References

Anion channels are proteinaceous pores in biological membranes that allow the passive diffusion of negatively charged ions along their electrochemical gradient. Although these channels may conduct other anions (e.g., I or NO) better than Cl, they are often called Cl channels because Cl is the most abundant anion in organisms and hence is the predominant permeating species under most circumstances. Cl channel gating may depend on the transmembrane voltage (in voltage-gated channels), on cell swelling, on the binding of signaling molecules (as in ligand-gated anion channels of postsynaptic membranes), on various ions [e.g., anions, H⁺ (pH), or Ca²⁺], on the phosphorylation of intracellular residues by various protein kinases, or on the binding or hydrolysis of ATP.

Like other ion channels, Cl channels may perform their functions in the plasma membrane or in membranes of intracellular organelles. On the one hand, these functions are related to the transport of charge, i.e., to the electric current flowing through the channel, and on the other hand to the transport of matter. For instance, plasma membrane Cl currents are important for the regulation of excitability in nerve and muscle. Currents flowing through intracellular Cl channels are thought to ensure the overall electroneutral transport of the electrogenic H⁺-ATPase that acidifies several intracellular compartments. On the other hand, bulk flow of chloride is important for cell volume regulation, as well as for transepithelial transport. Unlike Ca²⁺, Cl does not seem to play a role as intracellular messenger. However, the regulation of Cl channel activity by anions (90, 495, 538) also implies that changes in intracellular Cl concentration ([Cl]_i) may have a regulatory role. A recent report (114) additionally suggested that [Cl] may serve as an allosteric effector in post-Golgi compartments.

Patch-clamp studies have revealed a bewildering variety of anion channels that differ in their single-channel conductance, anion selectivity, and mechanism of regulation. Although differences in experimental conditions make comparisons often difficult, this suggests a large molecular diversity of Cl channels. Cl channels may be classified as to their localization (plasma membrane vs. vesicular), single-channel conductance, or mechanism of regulation. However, such classification schemes are ambiguous. For instance, the same channel may reside in the plasma membrane and in intracellular organelles, or the mechanisms of activation may overlap. Furthermore, with the exception of GABA and glycine receptors, such a classification is unlikely to correlate with the underlying gene families.

The most logical classification of Cl channels will be based on their molecular structures. However, the large variety of biophysically identified Cl channels is not yet matched by a similar number of known Cl channel genes, suggesting that entire gene families of anion channels remain to be discovered. For instance, we probably do not yet know the gene encoding the channel mediating the swelling-activated Cl current (I_Cl,swell) (volume-sensitive organic anion channel, volume-regulated anion channel), and many investigators would agree that the genes encoding the archetypal Ca²⁺-activated Cl channels have not yet been identified.

The correlation of a cloned gene with an ion channel function is often problematic due to the presence of endogenous channels in the expression system. For instance, it now appears that neither mdr (652) nor pI_Cln (469) represents the swelling-activated Cl channel (460, 490). Furthermore, several reports on currents elicited by CLC proteins (which form a well-established Cl channel family) have probably described currents that are endogenous to the expression system (75, 127, 359, 366).

So far, we know three well-established gene families of Cl channels. In mammals, the CLC gene family of chloride channels has nine members that may function in the plasma membrane or in intracellular compartments. CLC proteins were thought to have probably 10 or 12 transmembrane domains (Fig. 1A, top). This model has now to be revised because Dutzler et al. (131a) recently reported the three-dimensional crystal structure of bacterial CLC proteins (Fig. 1A, bottom). As already indicated by a combined approach of mutagenesis and biophysical analysis, CLC channels are dimers in which each monomer has one pore (double-barreled channels). This has been fully confirmed by the crystal structure of bacterial CLCs. Because the crystal structure (131a) was published after the review was accepted, we still refer to the old nomenclature of protein regions throughout this review. Some CLC proteins associate with crucial -subunits, as recently shown (147) for ClC-K channels that need barttin (47) for functional expression. The cystic fibrosis transmembrane conductance regulator (CFTR) has 12 transmembrane domains, two nucleotide binding folds (NBFs), and a regulatory R domain (Fig. 1B). The opening of this channel is controlled by intracellular ATP and through phosphorylation by cAMP- or cGMP-dependent kinases. Quite surprisingly, it is the only member of the large gene family of ABC transporters that is known to function as an ion channel. Finally, the largest known family of Cl channels is formed by the ligand-gated GABA- and glycine-receptor Cl channels. These subunits have four transmembrane domains (Fig. 1C) and combine to form pentameric channels.

View larger version (27K): [in this window] [in a new window]   Fig. 1. Topology models for the established Cl channel families. A, top: CLC Cl channel model based on biochemical analysis (552). Conflicting results exist in the D4/D5 region (156). The broad hydrophobic region between D9 and D12 was difficult to investigate experimentally, but it was clear that it has an odd number of membrane crossings. The carboxy terminus of all eukaryotic CLC proteins has two CBS domains (30, 484) that have a so far unspecified role in protein-protein interaction. ClC-K proteins associate with the -subunit barttin, which spans the membrane twice (147) (shown at right). A, bottom: model of CLC Cl channel derived from three-dimensional crystal structure of a bacterial CLC protein shows that the membrane-associated part of the protein is composed of 17 -helices (helix A is not inserted into the membrane). Inspection of the crystal (131a) reveals that most of these helices are not perpendicular to the membrane, but severely tilted. Many of these helices do not span the width of the bilayer. This even serves an important function, as Cl is coordinated in the pore by helices extending from either side of the membrane into the center plane. For comparison and reference, the previous nomenclature of CLC domains (D1-D12) is indicated by shaded areas and dashed lines. B: topology model of cystic fibrosis transmembrane conductance regulator (CFTR), a member of the ABC transporter superfamily. It has two blocks of six putative transmembrane spanning domains each, which are separated by a cytoplasmic region that contains the first nucleotide binding fold (NBF1) and the regulatory R domain. A second NBF is present in the carboxy terminus. It is not yet firmly established whether CFTR functions as a monomer or as a dimer. C: topology model of ligand-gated anion channels. These proteins have four transmembrane domains and assemble to homo- and heteromeric pentameric channels.

In addition, a family of putative intracellular Cl channels that have a single putative transmembrane domain has been identified (the CLIC family) (43, 165, 345, 346, 503, 637). Another gene family that encodes proteins with four or five putative transmembrane domains (the CLCA or CaCC family) was suggested to encode Ca²⁺-activated Cl channels (468). However, in both cases, the evidence that these proteins form channels is not as watertight as with the gene families mentioned above. For instance, no mutants with changed permeation properties have been reported. Furthermore, the presence of only one transmembrane domain in CLIC proteins is highly unusual for ion channels.

After a short overview of the cellular functions of Cl channels, we focus on molecularly identified Cl channels and their physiological roles. The cloning of the genes encoding these channels has enabled detailed studies concerning their structure and function. It has also provided insights into their physiological functions by the subsequent generation of knock-out mouse models and the discovery of novel ion channel diseases ("channelopathies"). Because recent, excellent, and exhaustive reviews on CFTR (66, 115, 183, 319, 480, 560, 565, 575) and ligand-gated Cl channels (45, 86, 143, 162, 238, 411, 714) are available, these channels are discussed concisely and emphasis is put on CLC channels. We also provide short reviews of swelling-activated and Ca²⁺-activated Cl channels, a family of putative intracellular Cl channels (the CLIC family), and finally give a short overview of the pharmacology of Cl channels.

	II. CELLULAR FUNCTIONS OF CHLORIDE CHANNELS

Top Previous Next References

Cl channels are needed for the transport of salt and fluid across many epithelia. The polarized expression of Cl channels and secondary active Cl uptake mechanisms ensures the directionality of transport. For example, airway epithelia, acinar cells of many glands, and the intestine can actively secrete Cl across their apical membrane. Because Cl channels only permit passive transport by diffusion, the intracellular Cl concentration is raised above equilibrium by Na⁺-K⁺-2Cl cotransporters that often need K⁺ channels for recycling potassium (Fig. 2, B and C). In the chloride reabsorptive thick ascending loop of Henle, an apical cotransporter raises [Cl]_i, which then leaves the cell via basolateral Cl channels that are probably identical to ClC-Kb/barttin (Fig. 2B). This is discussed in detail in section IIIE. In contrast, intestinal crypt cells secrete Cl (Fig. 2C). In these cells, the Na⁺-K⁺-2Cl cotransporter, together with the K⁺ channel needed for recycling, is located basolaterally, and Cl leaves the cell apically via CFTR Cl channels (discussed in section IV). Both CFTR and the basolateral KCNQ1/KCNE3 K⁺ channel are stimulated by cAMP, resulting in an efficient regulation of transepithelial transport. In acinar cells, regulation of Cl secretion depends on intracellular Ca²⁺. Accordingly, the apical Cl channel is activated by Ca²⁺ (472). While these chloride secretory and reabsorptive epithelia must recycle K⁺ transported by the cotransporter, the K⁺-secretory stria vascularis of the inner ear uses Cl channels to recycle the chloride ions that are not transported across the epithelium (Fig. 2A).

View larger version (10K): [in this window] [in a new window]   Fig. 2. Transepithelial transport models. A: potassium secretion in the stria vascularis of the cochlea needs basolateral Cl channels for recycling Cl that is transported into the cell by a Na+-K+-2Cl cotransporter (NKCC1). K+ is secreted apically via KCNQ1/KCNE1 potassium channels. The basolateral membrane most likely contains parallel heteromeric ClC-Ka/barttin and ClC-Kb/barttin Cl channels (147). Mutations in KCNQ1, KCNE1, NKCC1, and BSND (encoding barttin) cause deafness, but mutations in either ClC-Ka or ClC-Kb alone do not. B: chloride reabsorption in the thick ascending limb of Henle's loop involves an apical Na+-K+-2Cl cotransporter (NKCC2) that needs a parallel K+ channel (ROMK1, Kir1.1) for recycling potassium. Cl leaves the cell passively across the basolateral membrane through the ClC-Kb/barttin Cl channel (147). Mutations in NKCC2, ROMK, or ClC-Kb cause variants of the same disorder, Bartter's syndrome. Mutations in the -subunit barttin (BSND) cause Bartter syndrome with deafness, as its loss of function affects both ClC-Ka and ClC-Kb. C: chloride secretion in intestinal crypt cells. Intracellular Cl concentration ([Cl]i) is raised above equilibrium by a Na+-K+-2Cl cotransporter that needs a parallel K+ channel (KCNQ1/KCNE3) for recycling and passively leaves the cell via the apical cAMP-stimulated Cl channel CFTR.

Another important function of chloride channels is the regulation of membrane electrical excitability. For voltage-gated Cl channels, this is most obvious for the skeletal muscle Cl channel ClC-1. As discussed in detail in section IIIC, ClC-1 stabilizes the resting potential of skeletal muscle. Accordingly, the loss of ClC-1 function leads to myotonia, an intrinsic muscle hyperexcitability. Also the electrical activity of other cells may be modulated by Cl channels. For instance, Ca²⁺-activated Cl channels (described in sect. VI) may be important for amplifying the sensory response of olfactory cells (379). The voltage-gated Cl channel ClC-2 (see sect. IIID) was hypothesized to codetermine neuronal [Cl]_i (596).

In contrast to skeletal muscle, in smooth muscle the electrochemical potential for chloride (E_Cl) is significantly higher than the resting potential (96). Thus an opening of Cl channels (e.g., of Ca²⁺-activated or of swelling activated channels) will lead to a depolarization that may be strong enough to cause influx of Ca²⁺ through voltage-activated Ca²⁺ channels. This may be important for the response of vascular resistance to mechanical stress (306) or to regulators of vasoconstriction such as norepinephrine (11, 76, 464).

The intracellular Cl concentration of neurons determines the response to the neurotransmitters glycine and GABA. Because glycine, GABA_A, and GABA_C receptors (discussed in sect. VIII) are ligand-gated Cl channels, their activation can lead to a passive influx or efflux of chloride, depending on the electrochemical potential for Cl. Their activation can therefore lead to an excitatory, or to the more commonly observed inhibitory, response.

Several roles of ion channels, e.g., action potential generation, volume regulation, and transepithelial transport, are specific for the plasma membrane. This does not mean, however, that there is no need for anion fluxes (and anion channels) in internal membranes. First of all, anion channels (or transporters) are needed for the passage of anionic substrates like phosphate and sulfate out of degradative as well as biosynthetic compartments, e.g., lysosomes and the Golgi apparatus. A large-conductance anion channel of cardiac sarcoplasmatic reticulum was shown to conduct adenine nucleotides, but the physiological role of this conductance remains elusive (292).

Second, anion channels are implicated in organellar volume regulation. Mitochondria are subject to volume changes, depending on the metabolic state of the cell. This is probably mediated by the flux of K⁺ and Cl across the inner mitochondrial membrane. A vesicular volume increase was reported to accompany the exocytosis of secretory granules in mast cells (110) and in pancreatic acinar cells (269), which was also mediated by the uptake of potassium chloride.

Apart from organellar volume regulation, Cl channels play an important role in maintaining electroneutrality. Electrogenic uptake of protons or calcium ions into intracellular compartments will very soon create a charge imbalance hampering further uptake. This is true for the Ca²⁺-ATPase of endoplasmic and sarcoplasmic reticulum as well as for the V-type H⁺-ATPase of the Golgi lamellae as well as endosomal and synaptic vesicles. To build up the necessary calcium or proton gradients, the excess positive charge in these organelles has to be neutralized. In principle, this may be achieved either by import of chloride (via anion channels) or by export of potassium (via cation channels). From studies on isolated endosomes it is known that acidification is more efficient in the presence of extravesicular chloride (176, 657). In situ studies with secretory and recycling endosomes of the trans-Golgi network indicated a dependence of the acidification rate on both potassium and chloride in the cytosol (118). This demonstrates the requirement for a chloride conductance in the acidification of these intracellular organelles.

The characterization of anion channels in intracellular membranes usually requires the isolation of the membrane under study, often in the form of small vesicles. These may then be studied in tracer-flux assays, fused to a lipid bilayer for electrophysiological investigation or fused to other vesicles and subsequently studied by patch-clamp techniques. With these methods, contamination with other membrane fractions is often a problem that cannot be solved satisfactorily. The purification of the channel protein and subsequent reconstitution is an alternative, but this entails the loss of the native environment and possibly conformational changes of the protein. In a few cases, the direct observation of intracellular ion channels in intact membranes has been reported (280, 616, 650), but this is technically very demanding. A much simpler way to study intracellular channels would be to redirect them to the plasma membrane. By overexpressing them, some of the intracellular CLC channels (ClC-3, -4, -5) are incorporated into the plasma membrane (171, 359, 600), but this does not work for all intracellular channels.

In synaptic vesicles from rat brain (543) and from Torpedo electric organ (295), voltage-dependent anion channels of intermediate conductance (10-100 pS) were found. These channels were present in every synaptic vesicle (295). Reconstitution of endoplasmic reticulum membranes from rat hepatocytes (138) yielded a large-conductance (150-200 pS) anion channel, which was also voltage dependent. A different type of anion channel has been found in sheep brain endoplasmic reticulum membranes, where it is colocalized with calcium release channels (586). Recently, a Cl channel in the Golgi complex was characterized, which was present even in the absence of protein translation, indicating that these channels are not en route to the plasma membrane, but endogenous to this compartment (454).

The outer membrane of mitochondria contains a Cl-selective porin, the so-called voltage-dependent anion channel (VDAC) (546). This ~0.6-nS outer membrane channel may be transformed to a ~2-nS unselective pore after association with proapoptotic proteins of the BCl-2 family (581, 582). The 2-nS pore was shown to be permeable to cytochrome c, which triggers apoptosis when released from the intermembrane space into the cytosol. Surprisingly, VDAC has also been found in the plasma membrane of several cell types (reviewed in Ref. 626). Plasma membrane porins apparently are confined to specialized domains such as the postsynaptic density (428) or caveolae (32). In accordance with a possible role of caveolae in transcytosis, it was speculated that plasma membrane porin may become active only after vesicle formation, being largely closed while located in the plasma membrane. Several different types of Cl channels are present in the inner mitochondrial membrane (25, 719), but their physiological function is unclear.

With the exception of the VDAC porin, all of these channels are known only on a functional basis, i.e., their molecular identity remains unknown. Recently, it became clear that many CLC Cl channels reside primarily in intracellular compartments. The generation and analysis of corresponding knock-out mice has shed considerable light on their role in endocytosis and acidification (320, 481, 601). This is discussed indetail in section III, F, H, and J.

	III. THE CLC CHLORIDE CHANNEL FAMILY

Top Previous Next References

The CLC chloride channel family was initially identified by the expression cloning of the voltage-gated Cl channel ClC-0 from the electric organ of the marine electric ray Torpedo marmorata (273). It is now clear that CLC genes are present both in prokaryotes and eukaryotes.

There are nine different CLC genes in mammals. Based on sequence homology, they can be grouped into three branches (Fig. 3). The first branch comprises plasma membrane channels, whereas the proteins encoded by the two other branches are thought to reside predominantly in intracellular membranes. Some of these vesicular channels, however, may be trafficked to the plasma membrane under special circumstances. For instance, the late-endosomal/lysosomal ClC-7 is inserted into the ruffled border of osteoclasts that are attached to bone (320). ClC-4 and ClC-5 also reach the plasma membrane upon heterologous expression (171), but it is not yet clear whether this also occurs in native cells.

View larger version (19K): [in this window] [in a new window]   Fig. 3. The CLC family of Cl channels in mammals. Based on homology, the nine mammalian CLC proteins can be grouped into three branches, as shown by the dendrogram (left). Channels of the first branch predominantly reside in the plasma membrane, whereas channels from the two other branches are thought to be predominantly intracellular. The localization on human chromosomes is indicated below the channel names. The next columns indicate the most important features of their tissue distribution, their presumed functions, the phenotype of the corresponding knock-out (KO) mouse model, and the name of the human disease associated with the channel, respectively. The asterisk indicates that mutations in barttin, a -subunit for ClC-Ka and ClC-Kb (147), cause Bartter syndrome with sensorineural deafness and kidney failure (47) because it compromises the function of both ClC-Ka and ClC-Kb in the kidney and the inner ear (147).

Many CLC channels (e.g., ClC-0, -1, and -2) yield sizeable currents when expressed alone, but ClC-K channels need the -subunit barttin (147). It is currently unclear whether other CLC proteins need -subunits.

Many, but possibly not all, CLC chloride channels are gated in a voltage-dependent manner. Although not universally accepted (cf. sect. IIF on ClC-3), all CLC channels that have been studied display a Cl > I conductance sequence. This may even be true for a bacterial CLC (396). Currents of many CLC channels are additionally modulated by anions and pH, but there are only a few reports describing a regulation by intracellular messengers or protein kinases.

The physiological and pathophysiological roles of several CLC channels are impressively illustrated by four human inherited diseases caused by mutations in their genes (313, 320, 373, 587). Additionally, human disease also results from mutations in barttin (47), a -subunit of ClC-K channels (147). Recently, five different CLC genes were disrupted in mice (63, 320, 406, 481, 601, 676), leading to important and often unexpected insights into their physiological functions.

The ClC-1 Cl channel provides the bulk of resting conductance of the plasma membrane of skeletal muscle. As a consequence, its mutational inactivation leads to myotonia in humans and mice (313, 597). The role of ClC-2 is less clear. The testicular and retinal degeneration resulting from its disruption in mice may suggest a role in transepithelial transport (63). The two renal ClC-K channels function (in a heteromeric complex with barttin, Ref. 147) in transepithelial transport across different nephron segments, as demonstrated by Bartter's syndrome in humans (587) and renal diabetes insipidus in mice (406). In addition, both ClC-Ka/barttin and ClC-Kb/barttin are important for inner ear K⁺ secretion (147). Accordingly, human mutations in barttin lead to Bartter syndrome associated with deafness (47).

The knock-out of ClC-3 in mice led to a severe degeneration of the hippocampus and the retina (601). ClC-3 is present in endosomes and synaptic vesicles, but whether the degeneration is due to the observed impairment of synaptic vesicle acidification is currently unclear (601). Mutations in ClC-5 underlie Dent's disease (373), an inherited disorder characterized by kidney stones and proteinuria. Both symptoms are a consequence of a reduced proximal tubular endocytosis, as revealed by a recent ClC-5 knock-out (KO) mouse model (481). Probably similar to ClC-3, ClC-5 provides a shunt for the H⁺-ATPase that is necessary for the efficient acidification of endosomes. Finally, mutations in ClC-7 lead to osteopetrosis, as first recognized in a mouse model and then confirmed for humans (320).

In the following sections, we first describe general or typical properties of CLC chloride channels. These properties were mostly gleaned from experiments with ClC-0 and ClC-1. These points will not be repeated in the following sections that discuss the individual mammalian channels in some detail. Particular emphasis is laid on their physiological function and the recently described mouse models. A final section deals shortly with CLC channels in model organisms like yeast and the nematode Caenorhabditis elegans.

A.  General Features of CLC Channels

1.  Topology of CLC channels

The recently identified crystal structure of CLC channels now gives a definitive picture of the topology (131a). The crystal reveals that the bacterial CLC protein is composed of 18 helices, most of which do not cross the membrane entirely. None of the helices is perpendicular to the membrane plane, but severely tilted. Not recognized previously, each subunit has an internal repeat pattern, with amino- and carboxy-terminal halves having opposite orientations in the membrane. Previous analysis of CLC topology by various biochemical methods yielded a confusing picture. In the following, we first describe the topology derived from site-directed mutagenesis, glycosylation scanning, protease protection assays, and cysteine modification experiments and then compare these results with the crystal structure. This comparison illustrates the methodological difficulties of biochemical topology analysis, which failed in several regions of CLC proteins where helices only partially span the membrane and are inserted obliquely.

Hydropathy analysis of ClC-0 initially suggested 13 hydrophobic stretches that might be able to cross the membrane and that were called D1 through D13 (273). D13 is now known to be part of the second of two CBS domains that are present in the cytosolic carboxy termini of every known eukaryotic CLC protein. CBS domains (named after cystathionine--synthase, one of the proteins in which these domains occur) are structural domains of unknown function that are conserved in a wide range of proteins (30, 484).

Site-directed mutagenesis of ClC-2 indicated that D13 (CBS2) does not cross the membrane and that both amino and carboxy termini reside in the cytosol (211). Furthermore, the loop between D8 and D9 turned out to be glycosylated, placing it firmly on the extracellular side (297, 417). Comparative analysis of newly identified CLC proteins indicated that D4 is poorly conserved and lacks significant hydrophobicity in ClC-3 to ClC-7 (67, 272). Thus a topology model was proposed in which D4 is extracellular and in which D9 to D12 cross the membrane either three or five times (272).

Schmidt-Rose and Jentsch (552) used glycosylation scanning and protease protection assays to assess the transmembrane topology of ClC-1. This confirmed the notion (272) that the loops between D1 and D2, between D6 and D7, and between D8 and D9 are extracellular, whereas D2/D3, D5/D6, D7/D8, and D10/D11, as well as the carboxy terminus after D12 are intracellular (552).

Conflicting evidence was obtained for the region between D3 and D5. Although an epitope inserted (in a truncated construct) after D3 could be partially protected against cytosolic proteases (suggesting it is extracellular), this region could not be glycosylated in a full-length construct. Glycosylation was observed after D4, consistent with it being extracellular (552). This was supported by the reaction of extracellular Zn²⁺ with cysteines located at both ends of D4 (340). However, an epitope inserted after D4 (in a truncated construct) was not protected against proteolysis, suggesting that it is cytosolic (552). This discrepancy may be due to a concerted membrane insertion of D3-D5 (552). On the other hand, cysteine modification experiments by Fahlke et al. (156) showed that residues at the end of D4 and the beginning of D5 are accessible to internal, but not to external, membrane-impermeable reagents. This indicated an intracellular location. While agreeing with protease protection, it contradicts the glycosylation experiment (552). In the light of these experiments, it was unclear which of these conflicting predictions of the D4-D5 region is correct. None of the methods is without problems. For instance, truncated proteins may not insert correctly into the membrane (381), and cysteine-modification experiments have sometimes led to incorrect predictions of channel pores (338, 619).

In addition to the unclear topology in the D3-D5 region, the broad hydrophobic region between D9 and D12 poses daunting problems. D9 enters the membrane from the exterior (297, 417, 552), and the end of D12 is intracellular (418), as is probably the D10-D11 linker (552). Hence D9-D10 may span the membrane just once.

The recently derived crystal structure now gives a high-resolution picture of the molecular structure of bacterial CLC proteins (131a). It reveals the presence of 18 -helices that exhibit a complex topology (Fig. 1A, bottom). The unambiguous predictions of previous biochemical topology analysis turned out to be correct. Given the intermingling of tilted protein helices, many of which only partially cross the lipid bilayer, it is not surprising that biochemical analysis had severe problems in some areas. The crystal shows that D3 and D4 partially span the membrane. D5 is split into two -helices that enter and leave the membrane at the intracellular side of the membrane. The broad hydrophobic region at the carboxy terminus (D9-D12) is composed of six -helices that cross the membrane several times.

All CLC channels that were examined are dimers. This conclusion was based on the coexpression of mutant and wild-type (WT) subunits of ClC-1 (152, 598) and on sedimentation studies of ClC-0 (417) and ClC-1 (152). Single-channel analysis of mutant/WT ClC-0 heteromers (387, 418), as well as of ClC-0/ClC-1 and ClC-0/ClC-2 concatemers (679), provided compelling evidence for a dimeric structure of CLC channels. Even EcClCa, a bacterial CLC protein from Escherichia coli which is also called YadQ or EriC, is a dimer as shown by chemical cross-linking, gel filtration, and velocity sedimentation (396). Importantly, the projection structure of two-dimensional EriC crystals by Mindell et al. (424) also suggested dimers. The three-dimensional crystal structure now unambiguously shows the dimeric double-barreled structure of CLC channels (131a). Both subunits are in contact at a broad interface that is formed by four helices each.

Those CLC channels that were studied at a single-channel level (ClC-0, ClC-1, and ClC-2) display two equally spaced conductance levels that are almost certainly due to the presence of two physically distinct, identical pores in the dimer (387, 418, 424, 545, 679). Each of these pores appears to be formed within a single CLC protein, and not at the interface between the two constituent subunits (387, 679).

When Miller and colleagues (224, 421, 422) analyzed single-channel currents through chloride channels directly reconstituted from Torpedo electric organ, they observed long periods of zero current that were interrupted by bursts of channel activity (Fig. 4A). During these bursts, two equally spaced conductance levels of ~10 and ~20 pS were found in addition to the zero-current state. The probability to find nonzero conductances within the burst increased with depolarization. At sufficiently positive voltages, the channel resided mostly in the ~20-pS state, with only a few short transitions to the ~10-pS state. In contrast, the probability to observe "bursts" of channel activity increased with hyperpolarization. These results could be reproduced by expressing the cloned ClC-0 channel (33).

View larger version (48K): [in this window] [in a new window]   Fig. 4. The double-barreled structure of CLC channels. A: a simple model of a CLC channel. As best exemplified by the Torpedo channel ClC-0, CLC channels are believed to be dimers that have two largely independent pores. These pores can be gated individually or can be closed together by a common gate. In ClC-0, both pores have identical properties, and their individual gates are independent. B: single-channel recordings supporting the double barrel model. Top: a recording from a native ClC-0 channel incorporated into a lipid bilayer. Note that there are long periods with zero current flow, attributed to a closed slow gate that closes both pores. An opening of this gate leads to "bursting" activity in which the equally spaced conductance levels of the individual pores become apparent. (From Miller C and Edwards EA. Chloride Channels and Carriers in Nerve, Muscle, and Glial Cells, edited by Alvárez-Leefmans FJ and Russel JM. New York: Plenum, 1990, p. 383-420.) Middle: excised patch containing a concatemer of a wild-type (WT) and a mutant (S123T) ClC-0 protein. Note that the recording can be explained by a large pore with WT conductance and a small mutant pore. In the recording to the right, bromide was substituted for chloride. As known for homomeric WT and mutant channels, WT ClC-0 conducts Cl better than Br, but this selectivity is lost in the mutant. This is faithfully reflected in the concatemer, showing that the permeation properties of both pores are independent. [From Ludewig et al. (387).] Bottom: registration of a ClC-0/ClC-2 concatemer. The recording can be explained by a ~8.5 pS ClC-0 pore attached to a ~2.5 pS ClC-2 pore. These values correspond to those of the corresponding homodimers, arguing even more strongly that pores are formed within the individual subunits. [From Weinreich and Jentsch (679).] C: the projection structure of two-dimensional crystals of the E. coli channel EcClCa (EriC, YadQ) reveals a symmetric structure with off-axis regions of reduced electron densities that might represent the two individual pores of the dimeric channel. [From Mindell et al. (424).]

A detailed biophysical analysis led to the "double-barrel" model (421), which states that ClC-0 has two identical pores. Thus the ~10-pS and ~20-pS conductance levels reflect the opening of one and two pores, respectively. Each of these pores can be gated independently by a process that is fast (with time constants in the 10-ms range) and opens the channel upon depolarization. In addition, there is a common "gate" that closes both pores at the same time. This gate is very slow (in the 10 s to minute range) and is opened by hyperpolarization. It leads to long closed periods that separate the bursts of channel opening.

This channel model is highly unusual. It requires solid evidence to convincingly distinguish it from a single pore that has two subconductance states. Although many of the arguments for a double-barrel structure do not constitute decisive proof, the sum of the experimental evidence overwhelmingly argues for a double-pore architecture. 1) The ratio of the ClC-0 conductance levels equals 2 to high precision. This does not depend on ionic conditions and is valid over a large voltage range. 2) The substates show binomial distribution, exactly as expected from two pores that are gated independently (33, 90, 224, 365, 386, 417, 421). Finally, 3) DIDS inhibited ClC-0 in a two-hit process, leading at first to the disappearance of the 20-pS state, and then followed by a total inhibition (422). This suggested that one molecule inhibited one pore at a time.

In more recent studies, mutagenesis was used to change the properties of only one pore in WT/mutant ClC-0 heteromers (387, 418). Several point mutations resulted in homomeric channels that had reduced single-channel conductance, an altered ion selectivity, and changed gating time constants (387, 418). Homomeric mutant channels retained two equally spaced non-zero conductance states, compatible with the presence of two identical, altered pores. The central question now asked was the following: Will the coexpression of WT and mutant channel cDNAs result in a channel with a large (WT) and a small (mutant) pore, as predicted by the double-barrel model? This was indeed observed (Fig. 4B, middle trace). The conductance levels corresponded to those observed in the respective homomers. Not only that, they retained their respective WT or mutant halide selectivity (387) and their time constants of fast gating (386), independent of their association with a WT or mutant subunit. Furthermore, cysteine modification of a single mutated residue in a WT/mutant heteromers changed the conductance of just one pore (364, 418).

The two pores might be formed either at the interface between the two proteins of the dimer (i.e., the first half of subunit 1 and the second of subunit 2 form one pore, or vice versa) or may be contained within a single protein. Single-channel experiments of mutant/WT and mutant/mutant heteromers (carrying mutation in different parts of the protein) supported the latter model (387). In a more radical approach, ClC-0 was linked covalently to either ClC-1 or ClC-2 in concatemeric constructs (679). Single-channel analysis revealed the presence of a ~8.5-pS ClC-0 pore alongside either a ~1.5-pS ClC-1 pore or a ~ 2.5-pS ClC-2 pore (Fig. 4B, bottom trace). It seems impossible to explain these results by a single pore with two subconductance states. Moreover, as ClC-0, -1, and -2 are only ~60% identical, it is highly unlikely that pores are located at the interface between both constituent subunits; rather, a pore is formed within a single subunit (679).

While these experiments are next to proof, some investigators may only be convinced by a crystal structure. The projection structure of two-dimensional crystals of the EriC (EcClCa) protein was recently resolved at 6.5-Å resolution by Mindell et al. (424). Although it was not yet possible to identify the pore(s) or transmembrane domains, the pictures revealed a twofold symmetry and off-axis areas of reduced electron density (Fig. 4C). This was compatible with a dimeric structure and suggested the presence of two off-axis pores (424). Recently, the higher resolution X-ray structure of Dutzler et al. (131a) confirmed these predictions. The CLC channel is a homodimer. Each subunit within the dimer forms its own ion conduction pore, and both subunits are interacting at a broad interface (131a).

A fundamentally different view of CLC pores was held by Fahlke et al. (153). While agreeing that ClC-1 is a dimer (152, 153), they suggested that ClC-1 has a single pore formed by both subunits. By extension, this should also apply to ClC-0 and other CLC channels. Thus the equally spaced conductance levels seen in ClC-0, ClC-1, and ClC-2 would represent subconductance states of a single pore. This clashed with the overwhelming evidence for a double-barreled structure of ClC-0 (33, 90, 224, 364, 365, 386, 387, 417, 418, 421, 422), the single-channel analysis of ClC-1 (545) and ClC-2 (679) and of concatemeric channels combining ClC-0 with ClC-1 or ClC-2 (679), and the crystal structure of a CLC protein (131a, 424). The arguments of Fahlke et al. (153) are based on the inhibition of macroscopic ClC-1 currents by the modification of cysteines introduced into the highly conserved region between D3 and D4. Using concatemers, they presented evidence that the side chains of such residues in the first subunit were close to the side chain of the equivalent residue of the second subunit (153). This suggested that these are located at an axis of symmetry between both subunits. As the authors assumed that these residues directly projected into the pore (153, 156), they concluded that there is a single pore formed by both subunits (156). However, the evidence that this segment directly lines the inner pore is weak (395), and the mutations used (156) had drastic effects on gating as well. A complete inhibition of macroscopic currents by modifying a cysteine on just one subunit does not prove that it blocks the pore but may result from an effect on gating. Indeed, Lin and Chen (364) mutated a ClC-0 residue (K165) to cysteine that is equivalent to a ClC-1 residue (K231) mutated by Fahlke and co-workers (153, 156). Lin and Chen (364) agreed that this mutation influenced "pore properties," but their single-channel analysis showed that modification of this cysteine in a WT/K165C heteromer affected just one conductance level (i.e., a single pore). Importantly, they observed effects on the "fast" gating of individual pores, as well as effects on the slow, common gate (364). Thus the nearly complete inhibition of macroscopic ClC-1 currents upon cysteine modification of only one subunit (153) might be explained by an effect on the common gate.

Both ClC-0 (397) and ClC-1 (551) were expressed as "split channel," where cDNAs encoding complementary fragments were expressed singly or in combination in Xenopus oocytes. This revealed that 1) several parts of the protein can fold independently and assemble to functional channels without a covalent link and 2) there is an important role of the cytoplasmic carboxy terminus, in particular, the CBS domains (named after cystathionine--synthase, one of the proteins in which these domains occur) (30, 484).

ClC-1 could be reconstituted from fragments that resulted from splits between transmembrane domains D7 and D8, D8 and D9, but not between D10 and D11 (551). None of the channel fragments gave rise to channel activity by themselves. Truncating the channel in the cytoplasmic carboxy terminus between CBS1 and CBS2 resulted in nonfunctional channels, which could be rescued functionally by expressing the lacking part containing CBS2. Likewise, ClC-0 was nonfunctional when truncated at several positions between CBS1 and CBS2 but could be rescued by coexpressing the lacking carboxy-terminal fragment (397). When cut in CBS1, however, coexpression of both parts did not yield currents. In an important experiment (397), oocytes were injected with a bacterial fusion protein representing the CBS2-containing carboxy terminus of ClC-0 2 days after they had been injected with cRNA encoding ClC-0 truncated after CBS1. This restored currents even when translation was inhibited by cycloheximide before injecting the fusion protein. This strongly suggested that the carboxy terminus, probably CBS2, interacted with other parts (possibly CBS1) of the truncated channel. It is currently obscure whether CBS2 is necessary for cellular trafficking or for channel function proper. The first possibility is suggested by mutagenesis of the CBS domains of the yeast scClC (Gef1p) (563). This entailed a mislocalization of the protein and a failure to complement the gef1 phenotype. The second possibility is supported by the observation that chimeras (169) or point mutations (36, 397) in the carboxy termini of ClC-0 and ClC-1 can affect gating. However, CBS2 is not absolutely required. Even though a large part of ClC-1 CBS2 was deleted, typical ClC-1 currents were recorded in Sf9 insect cells (248). Interestingly, CBS2 does not need to be close to CBS1 in the primary sequence. The function of ScClC was restored when CBS2, which was deleted from the carboxy terminus, was added back to the amino terminus (563).

These experiments point to an important, but largely unknown role of CBS domains in eukaryotic CLC channels. It is currently unclear whether these CBS domains bind to each other and/or to associated proteins. Because a bacterial CLC lacking CBS domains is a dimer (396, 424), CBS domains are not essential for dimerization. Interactions of CLC carboxy termini with other proteins, however, are not restricted to CBS domains. For instance, a proline-rich stretch located between the two CBS domains of ClC-5 probably interacts with ubiquitin protein ligases (562).

Some CLC proteins can combine to form heteromeric channels in vitro, but it is unclear whether this occurs in vivo. When ClC-1 and ClC-2 were coexpressed in Xenopus oocytes, the resulting macroscopic currents were incompatible with a linear superposition of currents from the respective homomeric channels (377). Instead, they could be explained within a double-barrel model in which an "open" ClC-2 pore operates in parallel to a smaller ClC-1 pore. The ClC-2 pore was suggested to have lost its voltage-dependent gating (possibly the "common" gate) by associating with ClC-1. It was noted that also ClC-0 and ClC-1 can form heteromers with novel properties (377).

These nonphysiological ClC-0/ClC-1 heteromers, as well as ClC-0/ClC-2 heteromers, were studied in detail in concatemeric constructs (679). The functional interaction between the ClC-0 and the ClC-1 or ClC-2 pore, respectively, seemed to be restricted to gating. This probably reflects the common gate that depends on both subunits (364, 387). Thus the double-barrel architecture of CLC channels allows for much less functional diversity compared with tetrameric K⁺ channels where pore properties depend on all four subunits. It also severely limits dominant-negative approaches to knock down CLC channel function.

Whereas there is strong evidence that several (possibly all) CLC channels are dimers with two pores, up to now it was difficult to identify the protein segments that line the pores. This is largely a consequence of the fact that a CLC pore is probably formed by a single subunit (679). In contrast, e.g., to tetrameric K⁺ channels, where four identical or homologous "P loops" contribute to the permeation pathway, the pore of CLC channels must be lined by different, nonhomologous parts of a single protein. Accordingly, mutations in various regions of the protein changed pore properties. However, this does not prove that the mutated residues directly line the pore. As a further complication, permeation and gating are tightly coupled in CLC channels (498). These factors combined make the identification of the pore by mutational analysis exceedingly difficult.

The crystal structure shows that various regions of the protein come together to form the pore. Four antiparallel helices extend from the inside and the outside into the center plane of the membrane. The Cl is coordinated by residues at the ends of these helices, which contain highly conserved residues. This includes the sequences GSGIP (end of D2), GK/REGP (between D3 and D4), GXFXP (between D9 and D10), and in addition a Y (end of D12). Interestingly, these regions are always oriented with their amino terminus pointed toward the binding site. Due to the helix dipole, or the amino-terminal positive end charge, this arrangement of helices might create an electrostatically favorable environment for anion binding (131a). A similar principle was also used in the K⁺ channel selectivity filter, but with reversed polarity (123a). In another contrast to cation channels, there is no water-filled cavity at one side of the pore, but the permeation pathway has the shape of an oblique hourglass.

Jentsch and co-workers discovered that mutations in the conserved D2-D3 linker (387), in the conserved region between D3 and D4 (598), and in a region after D12 (498) changed the ion selectivity and/or single-channel conductance of ClC-0 or ClC-1. These parameters are considered as pore properties, but indirect effects of the mutated residues could not be excluded. As mentioned above, the role of these residues in pore formation was confirmed by the crystal structure. Fahlke and co-workers (149, 156) later focused on the D3-D4 region of ClC-1 and proposed that it forms the narrowest part of the pore (156). Several point mutations in the D3/D4 region and in D5 drastically changed gating, often inverting the direction of voltage dependence. The anion selectivity of several mutants was changed, sometimes leading to a reversal of the Cl > I sequence of WT ClC-1 (156). Cysteine accessibility studies suggested that the D3/D4 region, as well as the carboxy-terminal part of D5, forms a diffusional barrier for the access of reagents from the either side of the membrane.

An important argument for the hypothesis that the D3-D5 region directly lines the pore was transplantation experiments (156). Fahlke et al. (156) substituted the ClC-1 D3-D5 region by that of ClC-3 and observed a reversal of the Cl > I selectivity of ClC-1. Because ClC-3 was believed to have an I > Cl selectivity (156), it was concluded that this segment transferred isoform-specific pore properties from ClC-3 to ClC-1. However, it now seems that ClC-3 has a Cl > I selectivity like other CLC channels (359) and that previously measured currents (127, 156, 293, 294) are endogenous to the expressing cells (171, 359, 601, 681). Thus the effect of the transplantation (156) may rather be due to indirect, possibly long-range effects. This was also suggested by a recent study of chimeric ClC-K channels (669).

Mutagenesis of the D2/D3 linker (150, 387) and in the region at the end of D12 (384, 387, 418, 498) revealed residues whose mutations can result in altered single-channel conductance, ion selectivity, and gating. Furthermore, a missense mutation in D10 of ClC-1 reduced single-channel conductance (692). Thus it seems fair to say that the D3-D5 region probably plays an important but poorly understood role in permeation and gating and that several other regions of the protein may contribute to the formation of the pore. This problem is unlikely to be solved by site-directed mutagenesis alone.

Most CLC protein that could be expressed functionally showed voltage-dependent gating. Compared with S4-type cation channels, the voltage dependence is ~5- to 10-fold weaker. At least in ClC-0 and ClC-1, there are two different gating processes, one of which acts on each individual pore (also called "fast gating" or "activation gating" for ClC-0), and one of which acts on both pores as a common gate (also called "slow gate" or "inactivation gate" for ClC-0). Two different gating processes were also found in the worm channel CeClC-3 (559), but it is not yet known whether they correspond to "common" and "individual" gates.

The primary sequence of CLC channels does not reveal any conspicuous charged transmembrane domain like the S4 segment that acts as a voltage sensor in a superfamily of cation channels (606). However, this does not rule out that charged amino acids in CLC transmembrane domains may act as voltage sensors. Indeed, it was proposed that an aspartic acid at the extracellular end of D1 acts as a voltage sensor in ClC-1 (155). When mutated to glycine, as in a patient with recessive myotonia (234), ClC-1 shows an inverted voltage dependence (155). However, several point mutations in various regions of either ClC-0 (385, 397) or ClC-1 (156, 692, 715) have similar effects. One such mutation even affects a residue close to the end of the long cytoplasmic carboxy-terminal tail and which is therefore unable to sense transmembrane voltage (397). Thus it is very unlikely that all these residues represent "voltage sensors." Mutations at these positions may rather reveal an intrinsic ability of CLC channels for inverted voltage-dependent gating. The structural basis for this effect is completely unknown.

The voltage-dependent gating of many CLCs is strongly modulated by extracellular anions and pH (90, 224, 498, 524, 536, 538, 559). Gating was most thoroughly studied in ClC-0 because of its relatively high single-channel conductance (~10 pS) and because its gating is relatively simple. The vastly different kinetics of the common, slow gate, and the fast gates that operate on individual pores allow an easy separation of these gates both in single-channel studies and in macroscopic current measurements. Furthermore, the fast gating is apparently a two-state process with monoexponential kinetics. Fast gating of ClC-0 is strongly dependent on extracellular chloride, with a shift in the open probability (p_open) curve toward more positive voltages by ~50 mV per 10-fold reduction in extracellular Cl concentration ([Cl]_o) (498). Thus ClC-0 opening is promoted by its substrate, chloride. Pusch et al. (498) proposed an unusual gating model in which the binding of chloride to a site deep within the pore promotes the (voltage-independent) opening of the channel. This results in voltage-dependent gating as chloride has to travel along the electric field to reach this site. Hence, both depolarization and an elevation of [Cl]_o will increase the local concentration of chloride at the binding site and promote channel opening. Cl was thus proposed to be the gating charge, with the steepness of the voltage dependence depending on the (electrical) distance of the putative binding site from the outside (498). The nominal gating charge derived from the macroscopic voltage dependence is close to 1, which could reflect a single Cl moving through the entire voltage drop. This very simple model could well describe p_open as a function of voltage and [Cl]_o. With the use of different anions and a mutant with altered anion selectivity, it was argued that only permeant anions promote the opening of the channel (498). This notion was further supported in experiments exploiting the anomalous mole fraction behavior of ClC-0 (498).

Chen and Miller (90) extended and modified this model. They reconstituted ClC-0 into lipid bilayers and measured gating at the single-channel level. They confirmed that external Cl acts as gating charge and showed that it increases the rate of channel opening (90). The closing rate was much more sensitive to intracellular than to extracellular Cl. Measurements over a Cl concentration range that was larger than in the previous study (498) suggested a saturation of Cl binding. It was concluded that Cl binds in a voltage-independent manner to a site in the vicinity of the outer opening of the closed pore and that a subsequent conformational change, which involves Cl as a gating charge, leads to another closed state which then opens very quickly (90). Alternatively, these data might be explained by a model with two Cl binding sites in the pore [which is supported by the anomalous mole fraction behavior (498)] and where, as originally proposed, Cl must reach the binding site by moving in the electric field (493).

A direct consequence of this activation of channel opening by Cl is that ClC-0 gating is not at thermodynamic equilibrium. This was indeed shown in an analysis of single channels from the reconstituted Torpedo protein (524) and is discussed in detail in a recent review (395). Exploiting the presence of the two gates, Richard and Miller (524) demonstrated a violation of microscopic reversibility of gating transitions. This resulted in a predominant cycling in one direction between observable states. The ratio of clockwise to counterclockwise transition rates varied with the magnitude of the Cl gradient. Although this observation agrees well with the notion that permeating anions are involved in gating, we are far from having a detailed understanding of this process.

The common (or inactivation) gate of ClC-0 is still less understood. It is exceedingly slow (in the 10 to 100 s time scale) and very sensitive to temperature (a Q₁₀ of ~40) (89, 169, 421, 497). It does not lead to a complete channel closure at positive voltages. Slow gating can be described by a Markovian process with at least two open and two closed states (497). Like the fast gate, also the slow gate is influenced by cytoplasmic pH and extracellular anions (90, 495). The mechanism of slow gating and its relation to fast gating is currently unclear. Several mutations in the transmembrane block and the carboxy terminus change or abolish slow gating (89, 169, 365). As expected for a gate acting on both pores of the channel, it can be influenced by mutations in only one of the subunits of the dimer (387). Interestingly, single-channel recordings of ClC-0 show very rare events in which only one of the two pores is closed over a long period (in the range of seconds) (386). This was most often observed at very negative voltages and may represent a third gating process.

An interesting analogy to ClC-0 is provided by the C. elegans channel CeClC-3. It has two easily distinguishable gating processes, at least one of which depends on extracellular chloride (559). A slow, anion-dependent process activates CeClC-3 by depolarization. A faster inactivation gate, however, closes the channel quickly at positive voltages such that practically no outward currents can be measured. When stepping back to negative voltages, the channel recovers from inactivation within ~10 ms. Unlike the activation process at depolarizing voltages, the inward peak current did not depend on anions. At the negative voltages, the channel closed again slowly in an anion-dependent manner. Thus this suggests a depolarization activated gate that is slow and anion dependent, and a much faster gate of opposite voltage dependence that is largely independent of anions (559). This provides a delightful contrast to ClC-0. In the absence of single-channel recordings, it is unclear whether one of these gates acts on two pores.

Many of the properties of ClC channels have been difficult to discover in the absence of high-resolution structural data. This gap has been closed by the elegant work of Dutzler, MacKinnon, and co-workers (131a). The structure of two bacterial ClC channels from S. typhimurium (StClC) and E. coli (EcClC) were solved with a resolution of 3.0 Å. The channel is formed by two identical subunits. The entire channel with two subunits is shaped like a rhombus with diameters of 100 and 55 Å and a thickness of ~65 Å as the helical extension protrude into the aqueous solution on both sides of the membrane plane. As predicted by the analysis of concatemeric channels (387, 679), the pore is not formed at the interface between subunits, but each subunit forms its own pore and selectivity filter. The core structure of a CLC channel subunit contains 18 -helices, nearly all of which are not perpendicular to the membrane, but severely tilted. Many of the helices do not cross the membrane and therefore do not qualify as classical "transmembrane helices." Interestingly, the three-dimensional structure reveals an internal repeated pattern as the amino-terminal half is structurally related to the carboxy-terminal half. These two halves wrap around each other. As mentioned above, amino acids conserved in all CLC channels form an ion-binding site near the membrane center by bringing together the ends of four -helices. The favorable electrostatic environment for Cl arises from partial positive charges.

Starting in the late 1970s, Miller and colleagues (421, 422, 524, 683, 684), by reconstituting electric organ membranes into lipid bilayers, discovered and characterized the activity of this Cl channel. After attempts to identify the channel protein by inhibitor binding had failed (271), Jentsch and colleagues isolated its cDNA by expression cloning in Xenopus oocytes (273) and later called it ClC-0 (Cl Channel 0) (599). It is now known to belong to a large gene family of CLC channels with nine