Kidney International, Vol. 63 (2003), pp. 1426–1432

Modeling study of human renal (hCLC-5) mutations suggests a structural-functional relationship

FIONA WU,1 PHILIPPE ROCHE,1 PAUL T. CHRISTIE,NELLIE Y. LOH,ANITA A.C. REED, ROBERT M. ESNOUF, and RAJESH V. THAKKER

Molecular Endocrinology Group, Nuffield Department of Clinical Medicine, University of Oxford, Botnar Research Centre, Nuffield Orthopaedic Centre, Headington, Oxford, United Kingdom; and Division of Structural Biology, University of Oxford, Henry Wellcome Building for Genomic Medicine, Roosevelt Drive, Oxford, United Kingdom

Modeling study of human renal chloride channel (hCLC-5) characterised by low-molecular-weight (Ͻ40 kD) pro- mutations suggests a structural-functional relationship. teinuria (e.g., ␣ and ␤ microglobulinuria), hypercalci- Background. Dent’s disease, a renal tubular disorder char- 1 2 acterized by low-molecular-weight proteinuria, hypercalciuria, uria, nephrocalcinosis, nephrolithiasis, and progressive and nephrolithiasis, is due to inactivating mutations in the renal failure. In addition, other renal proximal tubular X-linked renal-specific chloride channel, hCLC-5. The x-ray defects, which include aminoaciduria, phosphaturia, gly- crystal structures of two bacterial chloride channels (CLCs) cosuria, kaliuresis, uricosuria, and an acquired impair- have recently been established, thereby allowing us to construct a model for hCLC-5 and further examine the role of its muta- ment of urinary acidification may also occur [3, 4]. CLCN5, tions. which in humans is expressed predominantly in the kid- Methods. The data regarding 49 hCLC-5 mutations were ney, belongs to the family of mammalian voltage-gated reviewed. Thirty-four mutations that predicted absent or trun- chloride channel (CLCN1 to CLCN7, CLCKa and cated channels were excluded. The remaining 15 mutations (one in-frame insertion and 14 missense mutations), 12 of which CLCKb) that encode (CLC-1 to CLC-7, CLC- have been studied electrophysiologically, were assessed. The Ka and CLC-Kb) with transmembrane domains. Heter- hCLC-5 sequence was aligned with the Salmonella typhimurium ologous expression of six of the mammalian chloride and Escherichia coli sequences and used to map the hCLC-5 channels (CLC-1 to CLC-5, and CLC-Ka) and that of mutations onto a three-dimensional model. the fish Torpedo marmorata (CLC-0) has demonstrated Results. hCLC-5 is a homodimeric , with each subunit Ϫ Ϫ consisting of 18 helices. None of the missense mutations in- conductance of chloride (Cl ) and other anions (e.g., I , Ϫ Ϫ Ϫ volved the chloride (Cl ) selectivity filter, but 12 of the 15 Br ,NO3 , and glutamate [5–7]. These chloride channels mutations were found to be clustered at the interface of the are important in the control of membrane excitability, two subunits. Six of these mutations occurred in two of the helices that either form part of the interface or lie in close the regulation of cell volume, and transepithelial trans- proximity to the interface, and three other mutations that did port. For example, human CLC-5 (hCLC-5), which is Ϫ not lead to complete loss of Cl conductance were at the edge expressed at multiple sites in the human nephron, includ- of the interface. ing the proximal tubules (PT), the thick ascending limb Conclusion. These results demonstrate a crucial role for the ␣ interaction between the two subunits at the interface of the (TAL) of the loop of Henle, and the -intercalated cells homodimeric hCLC-5. of the collecting ducts (CD) [8] is involved in transepithe- lial solute transport (e.g., the receptor-mediated endo- cytic pathway of the PT cells). Indeed, the renal tubular Mutations in the renal-specific chloride channel abnormalities observed in patients with Dent’s disease (CLCN5) located on Xp11.22 result in are consistent with a functional loss of CLC-5 in such a Dent’s disease [1, 2], which is a renal tubular disorder reabsorptive pathway, and heterologous expression in Xenopus oocytes of the mutant hCLC-5s associated with 1 These authors contributed equally to this work Dent’s disease has been shown to abolish or markedly Ϫ Key words: X-linked renal disease, nephrolithiasis, proteinuria, renal reduce Cl conductance [9–13]. However, it has not been failure, tubular dysfunction, Dent’s disease. possible to establish correlations between the type of hCLC-5 mutation (e.g., premature truncation or mis- Received for publication July 19, 2002 and in revised form September 12, 2002 sense), the altered amino acid and its position in the Accepted for publication November 13, 2002 sequence, and the reduction in ClϪ conductance and the  2003 by the International Society of Nephrology clinical features. We postulated that the three-dimen-

1426 Wu and Roche et al: Structural analysis of hCLC-5 mutations 1427

Fig. 1. Sequence alignment of CLC channels to illustrate locations of hCLC-5 missense mutations (triangles) associated with Dent’s disease. The numbering corresponds to Salmonella typhimurium sequence (StCLC, GenBank number AE008704), and the regions of 100% identity or high homology between this, the Homo sapiens (hCLC-5, NM_000084) and Escherichia coli (EcCLC, NC_000913) CLC channels, are highlighted in black and gray boxes, respectively. The secondary structure (helices A to R) is shown above the sequences, and the residues forming the chloride ion selectivity filter (see Fig. 3) are indicated by stars (residues 106 to 110, 146 to 150, 356 to 360, and 445). StCLC and EcCLC each consist of 473 residues, while hCLC-5 consists of 746 residues. Insertions that are larger than ten residues were removed from the alignment (shown by XXX). This alignment has minor differences from that previously reported [14]. The figure was generated with ESPript [32]. 1428 Wu and Roche et al: Structural analysis of hCLC-5 mutations

Fig. 2. Schematic representation of 47 of the 49 hCLC-5 mutations [9 to 13, 16 to 24]. hCLC-5 consists of 18 ␣ helices, A to R, which are indicated by the boxed areas, with the cytoplasmic region below and the extracytoplasmic region above. hCLC-5 was previously considered to have 13 helical transmembrane domains [5, 9], and the residues forming these are represented in gray. It is important to note that the for helix A, which in the bacterial CLC is not inserted into the membrane [14, 33] is low and, thus, helix A of hCLC-5 is represented as an approximation (dotted box). Every 50th amino acid of the 746 amino acid hCLC-5 and the consensus phosphorylation (arrow heads) and glycosylation (branch) sites at codons 349, 350, and 408, respectively, are indicated. The residues involved in forming the chloride ion selectivity filter (pore) [14] are indicated by asterisks, and the hCLC-5 mutations reported in patients with Dent’s disease are shown in black. The two remaining mutations, which are complete deletions of CLCN5, are not shown [9, 16]. sional structure of hCLC-5 may provide additional in- METHODS sights about the roles of these mutations and their associ- hCLC-5 mutations ated phenotypes. The recent characterization of the The published data [9–13, 16–24] regarding the 49 crystal structures of homologous CLCs from Salmonella enterica serovar typhimurium (StCLC) and Escherichia hCLC-5 mutations associated with Dent’s disease were coli (EcCLC) [14] has enabled us to explore this further. reviewed with respect to type and electrophysiologic Analysis of the three-dimensional structures of these two functional effects following heterologous expression in bacterial CLCs revealed that they function as homodi- Xenopus oocytes. We specifically selected the electro- mers. Each subunit is roughly an equilateral triangle with physiologic data from the Xenopus oocyte experiments, a pore at its center and the diamond-shaped homodimer as these are the most comprehensive and, indeed, only is formed by an edge-to-edge interaction between the data available for mutant hCLC-5 expression. However, triangles. Four helices from each subunit make the main it is important to note that extrapolation of these findings contribution to the extensive symmetric interface region, from Xenopus oocytes to the more complex situation yielding a “double-barrelled” structure [14, 15]. In order that exists in polarized PT cells, where the mutant CLC-5 to investigate possible structural-functional relation- may be misfolded and not reach its target on the endoso- ships, we performed studies to define the likely positions mal membrane [8], requires cautious interpretation. within the three-dimensional structure of the hCLC-5 Thus, the 49 selected mutations consisted of 10 nonsense, mutations associated with Dent’s disease. 6 splice-site, 2 frameshift insertions, 11 frameshift dele- Wu and Roche et al: Structural analysis of hCLC-5 mutations 1429

Table 1. Human hCLC-5 missense mutations associated with Dent’s hCLC-1 mutations were used to facilitate a comparison disease and their corresponding residues in StCLC that were used between hCLC-1 and hCLC-5 (details available on re- to determine locations in the secondary structure quest). Numbera hCLC-5b StCLC residuec Helical locationd Reference 1 G57Ve I35 B [11] hCLC-5 structural analysis 2 L200R G137 E [9] The amino acid sequences of hCLC-5 (GenBank 3 S244L A179 G [9] 4 S270R R205 H to I loop [23] number NM_000084), StCLC (AE008704), and EcCLC 5 L278Fe I215 I [12] (NC_000913) were aligned using CLUSTAL W [27] with 6 R280Pe A217 I [10] subsequent manual adjustments. At the N-terminus, where 7 G506E G393 O [9] 8 G512R A399 O [11] the sequence identity is low, a tentative alignment was 9 G513Ef A400 O [16] based on the secondary structure prediction of a helical 10 R516Wf R403 O to P loop [16] region using PSIPRED [28]. For the whole alignment, 11 S520P T407 P [9] 12 I524Kf E414 P [24] the sequence identity between hCLC-5 and StCLC was 13 E527D L411 P [11] 22% overall, and 26% for residues 94 to 410 (numbering 14 W22G —g Ag [22] refers to StCLC). The equivalent identities between 15 30insHe —g Ag [11] hCLC-5 and EcCLC were 24% and 27%, respectively. a Mutation number as indicated in Figure 3 b Amino acid shown by one letter code: A, Ala; D, Asp; E, Glu; F, Phe; G, The level of conservation was quite evenly distributed Gly; H, His; I, Ile; K, Lys; L, Leu; P, Pro; R, Arg; S, Ser; T, Thr; V, Val; W, along the whole sequence, and at this level of identity, Trp; ins, insertion c Corresponding residue in StCLC the major secondary structural features would be ex- d Helices as shown in Figures 1 and 2 pected to be conserved, thereby making it possible to e hCLC-5 showing 50% to 70% reduction in channel activity as compared with wild-type when heterologously expressed in Xenopus oocytes use the StCLC structure as a model for hCLC-5. Using f Functional studies not performed these alignments, the 13 hCLC-5 missense mutations g Located in the amino terminus portion of helix A whose position has not been well characterized (Table 1) associated with Dent’s disease were mapped onto the StCLC sequence (Fig. 1) and analyzed with respect to its three-dimensional structure. tions, 3 intragenic deletions (resulting in loss of exons 4, 5, 6, 7, and 8), 2 complete deletions of the gene, 1 in- RESULTS frame insertion, and 14 missense mutations (Figs. 1 and An examination of the sequence alignments of 2). Thirty-four of these 49 hCLC-5 mutations are either hCLC-5, StCLC, and EcCLC (Fig. 1) together with the deletions, nonsense, splice-site or frameshift mutations secondary structures for the latter two channels (Fig. 2), that, if translated, would predict absent or truncated suggested that the topology of hCLC-5, which has been channels and which have been shown to abolish ClϪ previously thought to have 12 to 13 helices forming trans- conductance when expressed in Xenopus oocytes [9–13]. membrane domains [5, 9], needed to be revised to one Such truncated and inactivated channels were not ana- containing 18 helices, labeled A to R in Figures 1 and lyzed further as they were considered unlikely to yield 2. Interestingly, 13 of the 15 (i.e., 87%) hCLC-5 missense useful information in our analysis of the relationships mutations are located within the 18 ␣ helices. This is in between structure and function. The study focused on marked contrast (␹2 ϭ 5.8, P Ͻ 0.02) to the situation for the remaining 15 hCLC-5 mutations (14 missense and truncating hCLC-5 mutations, in which only 16 of the 1 in-frame insertion, Table 1) that were predicted, if 32 (i.e., 50%) are located in the helices. Indeed, the translated, to yield full-length hCLC-5 proteins that ei- truncating hCLC-5 mutations are scattered throughout ther reduced or abolished ClϪ currents [9–12, 16, 22–24]. the channel with no apparent hot spots [9–13, 16–24]. Two of these mutations (W22G and 30insH) were further An inspection of the hCLC-5 tertiary structure (Fig. 3) excluded from analysis, as they are located in the N-ter- and the locations of these missense mutations (Table 1) minal region [11, 22], which is poorly conserved between reveal four additional findings of interest. First, all except hCLC-5, StCLC, and EcCLC, thereby rendering the se- one (L200R) of the hCLC-5 missense mutations are at quence alignment unreliable. Thus, our analysis of the or near the dimer interface. This clustering of mutations structure and function relationships was based on the near the interface was unexpected and could not have remaining 13 missense mutations (Table 1) and this in- been predicted from the secondary structure of the chan- cluded three mutations (G513E, R516W, and I524K), nel alone. The only missense hCLC-5 mutation not oc- which are known to cause Dent’s disease, but which curring at the interface is L200R, and this is predicted have not been functionally assessed for ClϪ conductance to disrupt helix E (Fig. 3). The role of helix E in hCLC-5 [16, 24]. A similar analysis of the 66 published hCLC-1 function remains to be elucidated, but it is of note that mutations [25], associated with Thomsen and Becker the L200R mutation results in a loss of ClϪ conductance myotonias [26] was also undertaken and the 41 missense and a disease phenotype that is indistinguishable from 1430 Wu and Roche et al: Structural analysis of hCLC-5 mutations

Fig. 3. Three-dimensional model of the hCLC-5 dimer based upon the structure of StCLC dimer [14] showing the locations of missense mutations associated with Dent’s disease. Views of the hCLC-5 dimer [stick (A) and ribbon (B) representations] from the extracytoplasmic side. The two subunits are shown in blue and red with a ClϪ ion (large green sphere) located in the pore of each subunit. The locations of the mutated residues are shown by the smaller yellow spheres (A) and the number beside each refers to the missense mutation shown in Table 1. Helices O and P, which harbor Ͼ40% of the missense mutations (Fig. 2) are shown in green (B), and helices B and I, which harbor mutations that do not abolish channel function but reduce it by 70% are shown as yellow ribbons (B). View of hCLC-5 dimer (C) (stick model) from within the membrane with the extracytoplasmic surface above. The channel is rotated by 90Њ about the x- and y-axes relative to (A). The black line (35 A˚ ) indicates the approximate thickness of the membrane. N and C indicate the amino and carboxy termini of each subunit, respectively. Note that 12 of the 13 missense mutations cluster at the interface of the two subunits. This figure was prepared with Bobscript [29] and Raster 3D [30].

that of the other mutations [9]. Second, none of the helix P, whose sequence is highly conserved across the missense mutations alters the amino acids from the four members of the CLC family (20% identity, 73% conser- helices (D, F, N, and R) (Fig. 2) that are brought together vative changes between hCLC-5, StCLC, and EcCLC), near the channel center to form the ClϪ selectivity filter is one of the four major helices (H, I, P, and Q) that are (Figs. 2 and 3). Third, more than 40% of the missense involved in the formation of the interface (Fig. 3). Fourth, mutations are found in helices O and P, or the interven- the three missense mutations, G57V, L278F, and R280P ing loop (Table 1), thereby emphasizing the importance (Table 1), which, unlike the other 12 mutations, are not of this region as a hot spot for such mutations. Indeed, associated with a complete abolition of ClϪ currents, but Wu and Roche et al: Structural analysis of hCLC-5 mutations 1431 instead lead to an approximate 70% reduction in channel Reprint requests to Professor R.V. Thakker, Molecular Endocrinol- ogy Group, Nuffield Department of Clinical Medicine, University of activity [9, 12], are located as a subgroup at the edge of Oxford, Botnar Research Centre, Nuffield Orthopaedic Centre, Head- the interface. These three mutations are situated in heli- ington, Oxford, UK, OX3 7LD. ces B and I, and their positions along one of the sides E-mail: [email protected] (yellow ribbons, Fig. 3B) of the triangular subunit con- trasts with those of the other mutations in helices O and REFERENCES P (green ribbons, Fig. 3B), which are located more cen- 1. Scheinman SJ: X-linked hypercalciuric nephrolithiasis: Clinical trally at the interface of the homodimer. syndromes and chloride channel mutations. Kidney Int 53:3–17, 1998 A similar analysis of the skeletal chloride channel, 2. Thakker RV: Pathogenesis of disease and related syndromes of hCLC-1, mutations associated with the disor- X-linked nephrolithiasis. Kidney Int 57:787–793, 2000 3. Pook MA, Wrong O, Wooding C, et al: Dent’s disease, a renal ders of Thomsen and Becker [26], revealed that only Falconi Syndrome with nephrocalcinosis and kidney stones, is asso- one half of the 41 missense mutations are near the inter- ciated with a microdeletion involving DXS255 and maps to face, with the remainder being scattered throughout the Xp11.22. Hum Molec Genet 2:2129–2134, 1993 4. Wrong OM, Norden AG, Feest TG: Dent’s disease: A familial three-dimensional structure of the protein. The signifi- proximal renal tubular syndrome with low molecular weight pro- cance of these differences between hCLC-1 and hCLC-5 teinuria, nephrocalcinosis, metabolic bone disease, progressive re- remains to be elucidated, but it emphasizes that our nal failure and marked male predominance. Q J Med 87:473–493, 1994 observations regarding the hCLC-5 mutations associated 5. Jentsch TJ, Gunther W, Pusch M, Schwappach B: Properties of with Dent’s disease cannot be readily extrapolated to voltage-gated chloride channels of the CLC gene family. J Physiol the other CLC channels and their disease-associated mu- 482:19S–25S, 1995 6. Fahlke C: Ion permeation and selectivity in CLC-type chloride tations. channels. Am J Physiol 280:F748–F757, 2001 7. Jentsch TJ, Stein V, Weinreich F, Zdebik A: Molecular structure and physiological function of the chloride channels. Physiol Rev DISCUSSION 82:503–568, 2002 8. Devuyst O, Christie PT, Courtoy PJ, et al: Intra-renal and subcel- Our analysis of hCLC-5 mutations causing Dent’s dis- lular distribution of the human chloride channel, CLC5, reveals a ease in relation to a three-dimensional model for hCLC-5 pathological basis for Dent’s disease. Hum Mol Genet 8:247–257, 1999 provides support for the proposal that interaction be- 9. Lloyd SE, Pearce SHS, Fisher SE, et al: A common molecular tween the two subunits is crucial for the functioning basis for three inherited kidney stone diseases. Nature 379:445–449, of the hCLC-5 channel as a homodimer. The majority 1996 Ͼ 10. Lloyd SE, Pearce SHS, Kawaguchi H, et al: Idiopathic low molec- ( 90%) of the missense mutations in hCLC-5 involve ular weight proteinuria associated with hypercalciuric nephrocal- residues at or near the interface of the homodimer and cinosis in Japanese children is due to mutations of the renal chloride cause a total loss of, or great reduction in, channel activ- channel (CLCN5). J Clin Invest 99:967–974, 1997 11. Lloyd SE, Gunther W, Pearce SHS, et al: Characterisation of ity. Indeed, more than 35% of these mutated residues renal chloride channel, CLCN5, mutations in hypercalciuric neph- are directly involved in the formation of the interface, rolithiasis (kidney stones) disorders. Hum Mol Genet 6:1233–1239, while the three mutations associated with residual (i.e., 1997 12. Igarashi T, Gunther W, Sekine T, et al: Functional characteriza- 30% of wild-type) channel function are located toward tion of renal chloride channel, CLCN5, mutations associated with the periphery of the dimer interface and, hence, may Dent’s Japan disease. Kidney Int 54:1850–1856, 1998 13. Yamamoto K, Cox JPDT, Friedrich T, et al: Characterisation of cause a lesser disruption to structure and function. Muta- renal chloride channel (CLCN5) mutations in Dent’s disease. J tions of the residues at or near the interface may interfere Am Soc Nephrol 11:1460–1468, 2000 with the physical contact between the subunits, which 14. Dutzler R, Campbell EB, Cadene M, et al: X-ray structure of a CLC chloride channel at 3.0A reveals the molecular basis of anion itself may be of importance, either in holding each sub- selectivity. Nature 415:287–294, 2002 unit together with its pore in a correct configuration, or 15. Jentsch TJ: Chloride channels are different. Nature 415:276–277, in some other regulatory process such as a common gate 2002 16. Akuta N, Lloyd SE, Igarashi T, et al: Mutations of CLCN5 in [15, 31]. Thus, the elucidation of the three-dimensional Japanese children with idiopathic low molecular weight protein- structures of the StCLC and EcCLC channels has helped uria, hypercalciuria and nephrocalcinosis. Kidney Int 52:911–916, to provide insights about hCLC-5 function and the man- 1997 17. Nakazato H, Hattori S, Furuse A, et al: Mutations in the CLCN5 ner in which some hCLC-5 missense mutations may dis- gene in Japanese patients with familial idiopathic low-molecular rupt the interface between the two subunits and lead to weight proteinuria. Kidney Int 52:895–900, 1997 Dent’s disease. 18. Oudet C, Martin-Coignard DM, Pannetier S, et al: A second family of XLRH displays the mutation S244L in the CLCN5 gene. Hum Genet 99:781–784, 1997 ACKNOWLEDGMENTS 19. Hoopes RR, Jr, Hueber PA, Reid RJ, et al: CLCN5 chloride channel mutations in six new North American families with We are grateful to the Wellcome Trust (F.W., N.Y.L., A.A.C.R., X-linked nephrolithiasis. Kidney Int 54:698–705, 1998 R.V.T.), and Medical Research Council (P.T.C., R.M.E., R.V.T) for 20. Kelleher CL, Buckalew VM, Frederickson ED, et al: CLCN5 their support. P.R. is supported by the Centre National de la Recherche mutation Ser244Leu is associated with X-linked renal failure with- Scientifique (CNRS, France) and F.W. is a Nuffield Medical Trust out X-linked recessive hypophosphatemic rickets. Kidney Int 53: Fellow. 31–37, 1998 1432 Wu and Roche et al: Structural analysis of hCLC-5 mutations

21. Cox JPD, Yamamoto K, Christie PT, et al: Renal chloride channel, 27. Thompson JD, Higgins DG, Gibson TJ: Clustal W: Improving CLCN5, mutations in Dent’s disease. J Bone Miner Res 14:1536– the sensitivity of progressive multiple sequence alignment through 1542, 1999 weighting, positions-specific gap penalties and weight matrix 22. Morimoto T, Uchida S, Sakamoto H, et al: Mutations in CLCN5 choice. Nucleic Acids Res 22:4673–4680, 1994 chloride channel in Japanese patients with low molecular weight 28. McGuffin LJ, Bryson K, Jones DT: The PSIPRED protein struc- proteinuria. J Am Soc Nephrol 9:811–818, 1998 ture prediction server. Bioinformatics 16:404–405, 2000 23. Igarashi T, Inatomi J, Ohara T, et al: Clinical and genetic studies 29. Esnouf RM: Further additions to Molscript version 1.4 including of CLCN5 mutations in Japanese families with Dent’s disease. reading and contouring of electron density maps. Acta Crystallogr Kidney Int 58:520–527, 2000 D Biol Crystallogr 55:938–940, 1999 24. Takemura T, Hino S, Ikeda M, et al: Identification of two novel 30. Merrit EA, Bacon DJ: Raster 3D: Photorealistic molecular graph- ics. Methods Enzymol 277:505–524, 1997 mutations in the CLCN5 gene in Japanese patients with familial 31. Pusch M, Ludewig U, Jentsch TJ: Temperature dependence of idiopathic low molecular weight proteinuria (Japanese Dent’s dis- fast and slow gating relaxations of CLC-0 chloride channels. J Gen ease). Am J Kidney Dis 37:138–143, 2001 Physiol 109:105–116, 1997 25. Pusch M: Myotonia caused by mutations in the muscle chloride 32. Gouet P, Courcelle E, Stuart DI, Metoz F: ESPript: Analysis channel gene CLCN1. Hum Mutat 19:423–434, 2002 of multiple sequence alignments in PostScript. Bioinformatics 26. Koch MC, Steinmyer K, Lorenz C, et al: The skeletal muscle 15:305–308, 1999 chloride channel in dominant and recessive myotonia. Science 33. Schmidt-Rose T, Jentsch TJ: Transmembrane topology of a CLC 257:797–800, 1992 chloride channel. Proc Natl Acad Sci USA 94:7633–7638, 1997