Chloride and the endosomal–lysosomal pathway: emerging roles of CLC chloride transporters

Abstract

Several members of the CLC family of Cl⁻ channels and transporters are expressed in vesicles of the endocytotic–lysosomal pathway, all of which are acidified by V-type proton pumps. These CLC proteins are thought to facilitate vesicular acidification by neutralizing the electric current of the H⁺-ATPase. Indeed, the disruption of ClC-5 impaired the acidification of endosomes, and the knock-out (KO) of ClC-3 that of endosomes and synaptic vesicles. KO mice are available for all vesicular CLCs (ClC-3 to ClC-7), and ClC-5 and ClC-7, as well as its β-subunit Ostm1, are mutated in human disease. The associated mouse and human pathologies, ranging from impaired endocytosis and nephrolithiasis (ClC-5) to neurodegeneration (ClC-3), lysosomal storage disease (ClC-6, ClC-7/Ostm1) and osteopetrosis (ClC-7/Ostm1), were crucial in identifying the physiological roles of vesicular CLCs. Whereas the intracellular localization of ClC-6 and ClC-7/Ostm1 precluded biophysical studies, the partial expression of ClC-4 and -5 at the cell surface allowed the detection of strongly outwardly rectifying currents that depended on anions and pH. Surprisingly, ClC-4 and ClC-5 (and probably ClC-3) do not function as Cl⁻ channels, but rather as electrogenic Cl⁻–H⁺ exchangers. This hints at an important role for luminal chloride in the endosomal–lysosomal system.

The lumen of the vesicles of the endosomal pathway is increasingly acidified along their way from the plasma membrane to lysosomes. Likewise, synaptic vesicles and vesicles in the exocytotic pathway display an acidic interior. Their low luminal pH is important for several processes like receptor–ligand interactions, trafficking along the endosomal pathway, and luminal enzymatic activities. Although other transport processes like Na⁺–H⁺ exchangers may play a role in acidifying some vesicle populations, the active accumulation of protons in these compartments is mainly accomplished by V-type H⁺-ATPases, which function as ‘proton pumps’. As these pumps are electrogenic, i.e. transport net positive charge, they would create a lumen-positive voltage in the absence of a neutralizing current. This voltage would soon inhibit further acidification on energetic grounds. It has been known for decades that chloride needs to be present in the medium to achieve an efficient luminal acidification of diverse vesicle preparations. This suggested that the neutralizing current may be mediated by vesicular chloride channels. The molecular identity of these channels, however, remained obscure for a long time.

The molecular identification in my laboratory of ClC-0, a voltage-gated chloride channel from the electric organ of the marine ray Torpedo marmorata (Jentsch et al. 1990), led to the identification of all nine members of the mammalian CLC gene family (Jentsch et al. 2005). By their degree of homology, they can be grouped into three branches. Similar to ClC-0, members of the first branch (ClC-1, -2, -Ka, -Kb) are plasma membrane Cl⁻ channels, whereas members of the two other branches (ClC-3, -4 and -5, and ClC-6 and -7, respectively) reside predominantly on intracellular vesicles (Fig. 1). ClC-4 and ClC-5 are now known to operate as voltage-dependent Cl⁻–H⁺ exchangers (Picollo & Pusch, 2005; Scheel et al. 2005).

Figure 1.

Subcellular localization of vesicular CLC proteins A, proposed localization of different CLC isoforms along the endosomal–lysosomal pathway. Vesicles are progressively acidified from the neutral extracellular pH (∼7.4) to the acidic pH (∼4.5) of lysosomes. Acidification is performed by an ATP-driven proton pump that needs a net influx of negative charge for electroneutrality. The neutralizing current is thought to be mediated by CLC isoforms. ClC-4 and -5 mediate nCl⁻–H⁺ exchange (with an imprecisely known stoichiometry n, which might be n = 2 as in ClC-e1) and this very likely applies for ClC-3 as well. ClC-6 and ClC-7/Ostm1 are also shown as antiporters, although this remains to be demonstrated. The localization of ClC-4 is rather uncertain, but may also be endosomal (Suzuki et al. 2006). B, in osteoclasts attached to bone, ClC-7/Ostm1 is co-inserted with the proton pump into the highly infolded ‘ruffled border’ that acidifies the underlying resorption lacuna.

CLC proteins, whether ion channels or exchangers, function as (homo)dimers (Ludewig et al. 1996; Middleton et al. 1996), with each of the two ion translocation pathways entirely contained within a single subunit (Weinreich & Jentsch, 2001). This was beautifully confirmed by the crystal structure of bacterial CLC proteins (Dutzler et al. 2002). To some degree, different CLC isoforms may assemble to heteromers that are functional in heterologous expression systems (Lorenz et al. 1996; Weinreich & Jentsch, 2001). So far, only two accessory and structurally unrelated β-subunits of CLC proteins are known: Barttin for both ClC-K isoforms (Estévez et al. 2001) and Ostm1 for ClC-7 (Lange et al. 2006).

Loss of ClC-5: impaired renal endocytosis leads to kidney stones

Human mutations in ClCN-5 lead to low molecular weight proteinuria, urinary loss of phosphate and calcium, and frequently to kidney stones in a syndrome called Dent's disease (Lloyd et al. 1996). ClC-5 is expressed in renal proximal tubule (PT) cells. In these cells, ClC-5 co-localizes with proton pumps in apical endosomes, suggesting that the lack of ClC-5 might impair endosomal acidification, thereby compromising the reabsorption of filtered proteins and causing proteinuria (Günther et al. 1998). Indeed, ClC-5 KO mice from two independent strains (Piwon et al. 2000; Wang et al. 2000) lose low molecular weight proteins into the urine. In vivo endocytosis experiments revealed that the broad defect in endocytosis is cell-autonomous. It affects receptor-mediated endocytosis, fluid-phase endocytosis and the retrieval of plasma membrane proteins such as the sodium–phosphate cotransporter NaPi-2a and the Na⁺–H⁺ exchanger NHE3 (Piwon et al. 2000). The abundance of the endocytotic receptor megalin was decreased in KO PT cells, possibly pointing to a defect in recycling it back to the surface (Piwon et al. 2000). As a consequence, receptor-mediated endocytosis is reduced more severely than fluid-phase endocytosis. The role of ClC-5 in endosomal acidification was tested with suspensions of renal cortical endosomes (Piwon et al. 2000; Günther et al. 2003) and in cell culture (Hara-Chikuma et al. 2005a). In both protocols, the disruption of ClC-5 reduced endosomal acidification. The concomitant increase in luminal Cl⁻ concentration was blunted as well (Hara-Chikuma et al. 2005a).

The impairment of proximal tubular endocytosis might be explained by a reduced electrical shunt for endosomal proton pumps, but how can one explain the kidney stones observed in many, but not all patients with Dent's disease? A unifying, experimentally supported hypothesis (Piwon et al. 2000) links these symptoms to the primary defect in endocytosis (Fig. 2). The small peptide hormone parathyroid hormone (PTH) passes the glomerular filter into the primary urine and is normally endocytosed by PT cells (Fig. 2A). The lack of ClC-5 severely impairs PTH endocytosis, leading to higher than normal PTH levels in the lumen of the nephron even when the blood concentration of PT is normal. This leads to an excessive stimulation of apical PTH receptors in the late PT, triggering the endocytic removal of the phosphate transporter NaPi-2a from the plasma membrane and thereby causing hosphaturia (Piwon et al. 2000) (Fig. 2B). The increased stimulation of apical PTH receptors in the PT also augments the transcription of 1α-25(OH)-VitD₃-hydroxylase (Günther et al. 2003), a mitochondrial enzyme that converts the inactive precursor 25(OH)-VitD₃ into the active form 1,25-(OH)₂-VitD₃. This active hormone, which is slightly increased in the serum of patients with Dent's disease (Scheinman, 1998), is expected to stimulate the intestinal resorption of Ca²⁺, which then must be eliminated by the kidney. On the other hand, the main supply of 25(OH)-VitD₃ to the activating hydroxylase occurs through apical, megalin-dependent endocytosis, which is severely impaired in the absence of ClC-5. The outcome of these opposing mechanisms (increase in hydroxylase activity versus decreased precursor availability; Fig. 2B) determines whether there is an increase in serum 1,25-(OH)₂-VitD₃ and hence hypercalciuria and kidney stones. This model thereby can account for the clinical variability of Dent's disease.

Figure 2.

Model for renal pathology in Dent's disease (due to a loss of ClC-5) The small peptide PTH (parathyroid hormone) and various forms of vitamin D (VitD; bound to their binding protein) pass the glomerular filter into the early proximal tubule (A). PTH and VitD-binding protein complexes are normally endocytosed by proximal tubular cells after binding to megalin. PTH is degraded in lysosomes, whereas 1,25-VitD reaches VitD receptors that regulate the transcription of nuclear target genes (not shown) and 25-VitD is metabolized by mitochondrial enzymes: 1α-hydroxylase activates the precursor to the active hormone 1,25(OH)₂-vitaminD₃ (1,25-VitD), whereas the 24-hydroxylase inactivates VitD. The endocytotic uptake of PTH and VitD is severely impaired in the absence of ClC-5 (A and B; indicated by red minus symbols). As a consequence, their concentration increases in the lumen of later nephron segments compared to wild-type (A–C). This leads to an enhanced stimulation of apical PTH receptors (B; green plus symbols), causing the endocytosis and degradation of the sodium–phosphate cotransporter NaPi-2a (leading to phosphaturia) and increasing α-hydroxylase while decreasing 24-hydroxylase levels. Similar changes in hydroxylase levels also result from the impaired apical endocytosis of VitD, which acts on the transcription of the respective genes (not shown). The altered hydroxylase activities tend to produce more active hormone (1,25-VitD). Such an increase, however, is counteracted by a decrease in precursor availability due to its defective endocytotic uptake. Thus, depending on other factors, the amount of 1,25-VitD released into the blood may be decreased or increased (B). C, in contrast to proximal tubules, in which VitD is taken up primarily by apical endocytosis of its binding protein and is hence decreased in ClC-5 KO cells, it enters distal tubular cells mainly by diffusion of the free hormone. The increased luminal concentration of 1,25-VitD in distal segments of Clcn5⁻ nephrons enhances the transcription of distal VitD target genes like that of the ion channel TRPV5 (Maritzen et al. 2006).

This hypothesis has recently been put to more tests (Maritzen et al. 2006). Expression profiling and quantitative PCR indicated that VitD target genes are down-regulated in proximal tubules, but up-regulated in more distal nephron segments of Clcn5^−/− mice. This finding is consistent with the decreased endocytotic access of VitD₃ to PT cells and its increased concentration in the lumen of later segments (Fig. 2C). KO mice also showed increased transcription of distal tubular target genes of retinoic acid, pointing to a more generalized importance of changes in hormone concentration in the nephron lumen in pathological states (Maritzen et al. 2006).

Lack of ClC-3, ClC-6 or ClC-7 causes neurodegeneration

Surprisingly, the disruption in mice of ClC-3, -6, or -7 led to a neurodegeneration in the central nervous system (Stobrawa et al. 2001; Kasper et al. 2005; Poët et al. 2006). The severity and the morphological and biochemical characteristics of the degeneration differed significantly between the genotypes. In mice lacking ClC-6 or ClC-7, neurons displayed intracellular, electron-dense deposits that stained for lysosomal marker proteins and the subunit c of ATP-synthase, a protein typically accumulated in a subset of human lysosomal storage disease called neuronal ceroid lipofuscinosis (NCL). Storage occurred throughout neuronal cell bodies in ClC-7 KO mice (Kasper et al. 2005), whereas it accumulated specifically in initial axon segments of mice lacking ClC-6 (Poët et al. 2006) (Fig. 3). In comparison, no severe intraneuronal storage was observed in ClC-3 KO mice (Stobrawa et al. 2001; Kasper et al. 2005), although Clcn3^−/− mice were reported to display NCL-like features (Yoshikawa et al. 2002). The neuronal cell loss was severe in mice lacking ClC-3 or ClC-7, leading to a complete loss of the hippocampus in adult Clcn3^−/− mice (Stobrawa et al. 2001). These latter mice, however, survived quite happily for more than a year, whereas Clcn7^−/− mice died after about 2 months even if their osteopetrosis (see below) was cured (Kasper et al. 2005). Neuronal cell loss was nearly absent in ClC-6 KO mice, which showed nearly normal life span and only mild neurological abnormalities that included a reduced sensitivity to pain (Poët et al. 2006). Whereas the combination of severe osteopetrosis with lysosomal storage disease may also occur in human patients homozygous for severe CLCN7 mutations, it is currently unclear whether human mutations in CLCN3 or CLCN6 may cause phenotypes similar to ClC-3 and ClC-6 KO mice.

Figure 3.

Lysosomal storage in neurons of mice lacking ClC-7 or ClC-6 Electron-dense lysosomal storage material (indicated by arrows) is present throughout the cytoplasm of Clcn7^−/− neurons (A), whereas it accumulates exclusively in initial axon segments of Clcn6^−/− neurons (B). Scale bars, 1.5 μm in A and 0.4 μm in B. The plasma membrane of the neuron in B is highlighted by a dashed line. nuc, nucleus. Modified images from Kasper et al. (2005) and Poët et al. (2006).

The lysosomal pathologies observed with a disruption of either ClC-6 or ClC-7 raise the question of the subcellular localization of these CLC proteins (Fig. 1A). Immunocytochemistry and subcellular fractionation demonstrated the presence of ClC-7 in lysosomes and late endosomes (Kornak et al. 2001; Kasper et al. 2005). Staining for ClC-7 and the late endosomal–lysosomal protein lamp-1 overlapped almost completely, whereas there was only a partial overlap with ClC-6 (Poët et al. 2006). ClC-6 and ClC-3 may both be expressed predominantly on late endosomes (Stobrawa et al. 2001; Poët et al. 2006), a conclusion also supported by localizing epitope-tagged CLCs in transfected cells (Suzuki et al. 2006). A pre-lysosomal localization of ClC-3 and -6 is further suggested by the partial shift of both proteins, but not of ClC-4, into lysosomal fractions in Clcn7^−/− mice (Poët et al. 2006). In neuronal cells, ClC-3 is additionally expressed on synaptic vesicles (Stobrawa et al. 2001). The subcellular localization of ClC-6 and -7 suggests that the lysosomal storage in the respective KO mice is a consequence of impaired late endosomal or lysosomal function. As the neurodegeneration of Clcn3^−/− mice does not display the typical features of a lysosomal storage disease (Stobrawa et al. 2001; Kasper et al. 2005), it may be caused by a different mechanism. For instance, abnormal intracellular trafficking as in ClC-5 KO mice might change the subcellular localization of various proteins in a manner that is detrimental for the cell.

Thick bones, grey hair and lysosomal storage: identification of a new ClC-7 β-subunit

The immediately obvious phenotype of Clcn7^−/− mice, however, is not lysosomal storage, but rather a severe osteopetrosis that leads to skeletal deformities, lack of tooth eruption, and impaired growth (Kornak et al. 2001). The bone marrow cavity is replaced by calcified bone. ClC-7 is highly expressed in osteoclasts, where it can be inserted into the ‘ruffled border’ together with the vacuolar proton pump (Fig. 1B). This specialized membrane secretes acid into the resorption lacuna that directly faces the calcified bone and that may be regarded as an ‘extracellular lysosome’. The acidic pH in this extracellular compartment is necessary both for the chemical dissolution of inorganic bone and for the activity of secreted lysosomal enzymes that degrade organic bone material. Clcn7^−/− osteoclasts still attached to ivory (a bone surrogate) in vitro, but did neither acidify a resorption lacuna nor degrade bone like wild-type (Kornak et al. 2001). As their ‘ruffled border’ was underdeveloped, ClC-7 might contribute to the exocytotic build-up of this membrane. Our mouse model suggested that ClC-7 might also play a role in human osteopetrosis. Indeed, CLCN7 was mutated on both alleles in patients with malignant infantile osteopetrosis (Kornak et al. 2001), and heterozygous missense mutations of CLCN7 underlie the less severe osteopetrosis of the dominant Albers-Schönberg disease (Cleiren et al. 2001). Mutations in the a3 subunit of the H⁺-ATPase, a subunit highly expressed in osteoclasts, may also underlie severe infantile osteopetrosis (Frattini et al. 2000; Kornak et al. 2000), again highlighting the importance of having an acidic resorption lacuna.

Grey lethal mice, a spontaneous, severely osteopetrotic mouse mutant, resemble Clcn7^−/− mice not only in their severe osteopetrosis, but also in having grey fur in an agouti background. We therefore searched for a possible link between ClC-7 and Ostm1, the protein encoded by the grey lethal gene (Chalhoub et al. 2003). Ostm1 is a type I membrane protein with a heavily glycosylated amino-terminal portion and a short cytoplasmic tail (Lange et al. 2006). It co-localizes with ClC-7 in lysosomes and in the ruffled border of osteoclasts. Ostm1 needs ClC-7 to travel to lysosomes, whereas ClC-7 reaches lysosomes also without Ostm1. ClC-7 could be co-immunoprecipitated with Ostm1 and vice versa, identifying Ostm1 as a novel β-subunit of ClC-7 (Lange et al. 2006). Importantly, the stability of either protein depends on the coexpression with its partner. The pathology observed upon a loss of Ostm1 may be entirely due to the ensuing instability of the ClC-7 chloride transporter. Indeed, grey lethal mice do not only resemble Clcn7^−/− mice in their osteopetrosis, but also display similar lysosomal storage disease (Lange et al. 2006). The grey hair of either mouse model is not yet fully understood, but may be related to a dysfunction of melanosomes, lysosome-related organelles.

Lysosomal storage, but normal lysosomal pH: a clue for Cl⁻–H⁺ exchange?

The subcellular localization of ClC-6 and ClC-7 indicated that the lysosomal storage disease observed upon their loss might be a consequence of impaired late endosomal–lysosomal function. Because ClC-5 facilitates the acidification of renal endosomes (Günther et al. 2003), and as ClC-3 plays a similar role in endosomes (Hara-Chikuma et al. 2005b) and synaptic vesicles (Stobrawa et al. 2001), we expected Clcn6^−/− and Clcn7^−/− lysosomes to be more alkaline. However, steady-state lysosomal pH was unchanged in either mouse model (Kasper et al. 2005; Poët et al. 2006). Several hypotheses may reconcile these findings with the assumed role of vesicular CLCs: even if ClC-6 and -7 would account for the bulk of lysosomal membrane conductance, other smaller conductances remaining after their elimination may suffice to neutralize proton pump currents over the long run, leading to identical lysosomal pH under our experimental conditions where we had chased a pH-sensitive dye into lysosomes overnight (Kasper et al. 2005; Poët et al. 2006). The observed pathology might be due to a slower rate of acidification on the way to lysosomes (fitting to the likely late endosomal localization of ClC-6), or might hint at an important role of lysosomal chloride.

A very relevant recent discovery is that ClC-4 and ClC-5, just like the E. coli protein ClC-e1 (Accardi & Miller, 2004), function as electrogenic Cl⁻–H⁺ exchangers rather than being Cl⁻ channels (Picollo & Pusch, 2005; Scheel et al. 2005). While residing mainly in intracellular vesicles, a minor portion of ClC-4 and ClC-5 reaches the plasma membrane, at least upon heterologous expression. This localization allowed biophysical studies that showed that either protein mediates anion currents which decrease with extracellular (and, by extension, luminal) acidification (Friedrich et al. 1999). These currents displayed a Cl⁻ > I⁻ conductance sequence, as found in other CLC proteins. These currents, as it has emerged now, reflect an electrogenic exchange of Cl⁻ for H⁺ (Fig. 4). Whereas the stoichiometry of the ion exchange could not be determined precisely for ClC-4 and -5 (Picollo & Pusch, 2005; Scheel et al. 2005), it is 2Cl⁻ for 1H⁺ in the bacterial ClC-e1 (Accardi & Miller, 2004). Mutating a key ‘gating’ glutamate in ClC-e1, ClC-4 or ClC-5 uncouples the Cl⁻ conductance from H⁺ countertransport and abolishes the rectification of ClC-4 and -5 currents (Friedrich et al. 1999; Accardi & Miller, 2004; Picollo & Pusch, 2005; Scheel et al. 2005). It is unknown whether ClC-6 and ClC-7 also function as Cl⁻–H⁺ exchangers, but the presence of another glutamate typically found in CLC exchangers, but not channels (Accardi et al. 2005), suggests that they do. Several mutually incompatible channel functions were assigned to ClC-3 (e.g. Duan et al. 1997; Wang et al. 2006), but it is very likely that ClC-3 rather functions as a Cl⁻–H⁺ antiport as well. ClC-3, -4, and -5 share the same high degree of sequence homology (including the two critical glutamate residues), and Weinman's group reported ClC-3 currents that were nearly identical to those of ClC-4 and -5 (Li et al. 2000). Importantly, these currents were affected similarly by mutating the key ‘gating’ glutamate (Friedrich et al. 1999; Li et al. 2002). Unfortunately, the low currents obtained with ClC-3 did not allow direct tests for Cl⁻–H⁺ exchange activity (Picollo & Pusch, 2005).

Figure 4.

ClC-5 is a Cl⁻–H⁺ exchanger, whereas ClC-0 is a Cl⁻ channel HEK cells transfected with ClC-5 (A) or ClC-0 (B) were clamped to different voltages (lower traces) using the gramicidin-perforated patch clamp technique to minimize the equilibration of their internal pH (pH_i) with the patch pipette. The middle traces shows clamp currents, while the upper panels show pH_i (measured using a ratiometric pH-sensitive dye). When ClC-5-transfected cells were clamped to voltages more positive than +30 mV, pH_i increased, indicating an exit of H⁺ in exchange for Cl⁻ entry. Consistent with the steep outward rectification of ClC-5 currents (Friedrich et al. 1999), the rate of intracellular alkalinization increased steeply with inside-positive voltage. By contrast, a similar voltage-clamp protocol did not change the pH_i of cells expressing the Torpedo channel ClC-0 (B). Panels taken from Scheel et al. (2005).

At first glance, it seems counterproductive to use Cl⁻–H⁺ exchangers rather than Cl⁻ channels to compensate proton pump currents, since a portion of the pumped protons will leave the vesicles via the transporter. Nonetheless, electrogenic Cl⁻–H⁺ exchangers will facilitate acidification, albeit at the expense of consuming more metabolic energy. The most important difference may be the direct coupling of Cl⁻ gradients to H⁺ gradients. H⁺-ATPases and Cl⁻ channels operating in parallel will accumulate Cl⁻ during active acidification, but the antiport will keep vesicular Cl⁻ high under steady-state conditions as long as the inside-acidic pH gradient is maintained. In indirect support for this notion, atClCa, an anion–H⁺ exchanger from Arabidopsis thaliana, has recently been shown to use the pH gradient of the plant vacuole to accumulate nitrate in that organelle (De Angeli et al. 2006). Hence, although the pH of Clcn6^−/− and Clcn7^−/− lysosomes was normal (Kasper et al. 2005; Poët et al. 2006), their luminal Cl⁻ concentration might be changed. This altered Cl⁻ concentration might contribute to the lysosomal storage disease of Clcn6^−/− and Clcn7^−/− mice. Our work may foreshadow an important role of vesicular chloride that has not been recognized previously.