Familial idiopathic nephrotic syndromes represent a heterogeneous group of kidney disorders, and include autosomal recessive steroid-resistant nephrotic syndrome, which is characterized by early childhood onset of proteinuria, rapid progression to end-stage renal disease and focal segmental glomerulosclerosis. A causative gene for this disease, NPHS2, was mapped to 1q25–31 and we report here its identification by positional cloning. NPHS2 is almost exclusively expressed in the podocytes of fetal and mature kidney glomeruli, and encodes a new integral membrane protein, podocin, belonging to the stomatin protein family. We found ten different NPHS2 mutations, comprising nonsense, frameshift and missense mutations, to segregate with the disease, demonstrating a crucial role for podocin in the function of the glomerular filtration barrier.
Plasma ultrafiltration during primary urine formation in the glomerulus is a central function of the kidney. The structurally complex glomerular capillary wall is responsible for this function and is composed of a basement membrane covered by fenestrated endothelium on the inner surface and highly specialized epithelial cells called podocytes (due to their characteristic interdigitated foot processes, on the outer surface1). Dysfunction of the glomerular filter, resulting in extensive leakage of plasma proteins and diffuse effacement of podocyte foot processes detected by electron microscopy, is observed in many acquired and inherited nephropathies. The clinical hallmark is nephrotic syndrome characterized by heavy proteinuria, edemas, hypoalbuminaemia and hyperlipidaemia. Progression to end-stage renal disease (ESRD) is associated with the development of focal segmental glomerulosclerosis2 (FSGS).
The study of genetic diseases with filtration barrier defects provides useful models for deciphering the physiopathology of the glomerular filtration process. The most severe such disorder is congenital nephrotic syndrome of the Finnish type (CNF), caused by mutations in the recently identified gene NPHS1 (ref. 3). NPHS1 encodes a transmembrane protein of the Ig superfamily called nephrin3, which is exclusively localized at the slit diaphragm joining the interdigitated podocyte foot processes4,5,6. Less severe cases of familial proteinuria or nephrotic syndrome with FSGS have also been described. These forms have an adult onset and a slow progression towards renal failure. Two loci for autosomal dominant forms of these ‘familial’ FSGS have been mapped to 19q13 (ref. 7) and 11q21–q22 (ref. 8), respectively, but the underlying genes have not yet been identified.
Previously, we delineated a new type of familial steroid-resistant idiopathic nephrotic syndrome (SRN; MIM 600995) characterized by an autosomal recessive mode of inheritance, onset between three months and five years, resistance to steroid therapy, rapid progression to ESRD, absence of recurrence after renal transplantation and absence of extra-renal disorders. Histologically, minimal glomerular changes are observed on early biopsy samples and FSGS is present at later stages. Although we demonstrated genetic heterogeneity in this subset of patients, we mapped a gene for SRN, NPHS2(formerly SRN1), to 1q25–q31 between the markers D1S452 and D1S466 , a region spanning approximately 12 cM (ref. 9).
Here we use positional cloning to identify the causative gene. We found that NPHS2 is exclusively expressed in the podocytes and encodes an integral membrane protein, which we named podocin.
Physical mapping of the candidate region and refinement of the NPHS2 interval
By linkage analysis of our previously reported families using new microsatellite markers10 as well as analysis of novel families, we refined the NPHS2 locus between the markers D1S480 and D1S2883, a region of 2.7 cM. A YAC contig (20 clones) spanning this region was retrieved from the Whitehead Institute server. Using the sequence-tagged site (STS) content provided by various databases and YAC-end sequences as new STSs, we constructed a PAC contig spanning the NPHS2 interval, with five gaps partially bridged by cosmid clones. A renewed linkage analysis of two consanguineous families (8 and 10) using new microsatellite markers (D1S1640, D1S3758, D1S3760, D1S215 and D1S3759) assigned to the contig (Fig. 1) localized the SRN locus between the markers D1S1640 and D1S3759, an interval estimated at 2.5 Mb.
We assigned ESTs belonging to 14 UniGene clusters, 5 independent ESTs, 4 known genes (NGAP, SOAT1, ABL2 and KIAA0475) and 1 pseudogene (RPS14P1) to the contig (Fig. 1). The four genes encode a Ras GTPase activating protein; an acyl-CoA:cholesterol acyltransferase; a tyrosine-protein kinase; and an unknown protein, KIAA0475, respectively. As we found NGAP and KIAA0475 to be expressed in kidney as well as other tissues by northern-blot analysis (data not shown), we undertook mutation screening in nine affected individuals. We detected no pathogenic mutations in the regions tested. SOAT1 and ABL2, as well as three ESTs and four UniGene clusters that were not expressed in kidney, were not further analysed. The remaining UniGene clusters and EST were considered for further analysis.
Identification of NPHS2
We localized the EST AA398634, belonging to one of these UniGene clusters (Hs. 97398) and mapping to the BAC clone 545A16 (The Sanger Centre, United Kingdom), to the cosmid 28e17 and PAC 302d13 clones of the contig (Fig. 1). Using RT–PCR we found that this EST, originating from an adult testis library, was also expressed in kidney. By 5′-and 3′-RACE–PCR of a human fetal kidney adaptator-ligated cDNA library, a transcript with a putative ORF and homology with six known ESTs (UniGene cluster Hs. 192657), all from a kidney cDNA library, was amplified. Subsequent screening of a human fetal kidney cDNA library allowed the reconstitution of a 1,853-bp transcript. The five isolated cDNA clones start between positions 1 and 5 of the reconstituted cDNA sequence and contain a methionine codon located 70 nt downstream from the first nucleotide. This ATG codon lies within a Kozak translation initiation consensus sequence11. Sequence analysis of 710 bp of genomic sequence upstream of the ATG codon, using the programs TSSG and TSSW (ref. 12), predicted the presence of a transcription start site located 83 bp upstream of the ATG codon (TSSG and TSSW scores of 21.97 and 13.08, respectively). These features suggested that this ATG is the translation initiation codon. The complete NPHS2 ORF is 1,149 bp and is followed by a 635-bp 3′ UTR containing an atypical polyadenylation signal (AATTAAA) situated 13 nt upstream of the poly(A) tail. The sequence of the EST AI672038 (belonging to the UniGene cluster Hs.192657), however, shows that one of two overlapping consensus polyadenylation signals (AATAAA) located 44 bp downstream of the aforementioned signal can also be used.
The length of the isolated cDNA sequence is in agreement with the size of the transcript (∼2 kb) detected by northern-blot analysis. The transcript was strongly expressed in human fetal and adult kidney with no signal being observed in other tissues (Fig. 2a,b). This was further confirmed by hybridization of a normalized human poly(A) + RNA dot blot containing RNA from 50 different human tissues. We observed a strong hybridization signal in mature and fetal kidney, whereas a weak signal in the adult testis, fetal heart and fetal liver was seen only after a 24-hour exposure (Fig. 2c). The expression pattern suggested that this gene was likely to be NPHS2.
Screening for mutations
We next amplified total RNA from an end-stage kidney of an affected member of family 8 using RT–PCR with cDNA-specific primers. We observed a barely detectable amplification product (Fig. 3a), suggesting that the transcript of interest was expressed at a very low level due to a mutation in the corresponding gene. We determined the gene structure by sequencing DNA from the cosmid 28e17 and PAC 302d13 clones using cDNA-specific primers (see Table 1, http://genetics.nature.com/supplementary_info/ ), screened the 8 coding exons and intron-exon junctions of 16 unrelated affected individuals for mutations by SSCP, and detected 10 different mutations in 14 families (Table 2). All mutations co-segregated with the disease, and none were identified in 80 control chromosomes. Three mutations resulted in a frameshift or a premature stop codon ( Fig. 3b,c), potentially null mutations. Six were missense mutations. Of these, two (R138Q and V180M), which occur in a CpG pair, were detected in un-related families. We found R138Q in the heterozygous state in two families (4 and 12) and in the homozygous state in four families with no known consanguinity (families 6, 7, 11 and 13). Five of these patients originated from Germany or The Netherlands. The mutation was not found in 28 additional control DNA samples from unrelated individuals from these countries. These data indicated that this gene was indeed NPHS2.
Characterization of the protein product of NPHS2
The full-length NPHS2 cDNA encodes a putative 383–amino-acid protein of approximately 42 kD, which we named podocin. Database comparisons of this sequence detected a region of extensive similarity between the central region of podocin and proteins of the band-7-stomatin family. The band-7-stomatin protein family signature was found between amino acids 238 and 266 of podocin by comparison with the Prosite database13. The strongest homologies were found with human stomatin (47% identity and 67% similarity over 253 aa) and Caenorhabditis elegans MEC-2 (44% identity and 65% similarity over 275 aa; Fig. 4). Analysis of the podocin amino-acid sequence using the PSORTII program14 suggested that NPHS2 encodes an integral membrane protein with one transmembrane domain (aa 105–121) and a 262–amino-acid carboxy-terminal cytoplasmic tail. The transmembrane domain and most of the cytoplasmic tail are homologous to the corresponding regions of the stomatin family proteins. Furthermore, the cysteine residue located four amino acids upstream of the transmembrane domain, which is conserved among stomatins and MEC-2 and has been shown to be the major palmitoylation site in stomatin15, was also conserved in podocin. Podocin, however, extends beyond the stomatin-like sequences at both the N and C termini (97 aa and 36 aa, respectively), which show no homology with any known protein sequence.
In situ hybridization
We performed in situ hybridization on normal fetal and post-natal kidney samples using digoxigenin- or [35S]-labelled riboprobes (Fig. 5). We observed strong signals exclusively in glomeruli. In the fetuses, no signal was detected in the earlier stages of nephron development, whereas intense signals were seen in the future podocytes of the inferior segment of the S-body, concomitant with vascularization of the inferior cleft (Fig. 5a). This expression was maintained at the same level in immature glomeruli and persisted in mature glomeruli of the deep cortex ( Fig. 5c). All glomeruli were still strongly labelled in post-natal kidneys. Under high magnification, transcript expression was restricted to within the podocytes, at the periphery of the glomerular tuft ( Fig. 5b,d). No specific signals were obtained with the sense probes (Fig. 5e).
Here we identified by positional cloning a new gene, NPHS2, which is mutated in a subset of childhood steroid-resistant nephrotic syndrome with FSGS. This gene was identified by characterization of ESTs in the critical interval at 1q25–q31 and considered to be a candidate gene due to its almost exclusive expression in fetal and mature kidney. The detection of mutations leading to premature termination of the protein in affected individuals confirmed it as the causative gene.
NPHS2 encodes a new member of the stomatin protein family, which we named podocin due to the specific expression of NPHS2 in podocytes. Stomatin, also known as band-7 protein, is an integral membrane phosphoprotein initially isolated from erythrocyte plasma membranes and subsequently found to have a more widespread expression16. Its function has not yet been determined. Like MEC-2, another member of this family which is required for the function of touch-receptor neurons in C. elegans17, it has been suggested to have a role in primary sensory neurons by linking ion channels to the cytoskeleton17,18.
Podocin has a structure similar to that of all stomatin-like proteins, which consists of an N-terminal domain, a short transmembrane domain and a cytosolic C-terminal domain. Both the N- and C-terminal domains of stomatin and MEC-2 are cytosolic17,19, suggesting that their transmembrane domains form a hairpin-like structure. Stomatin forms homo-oligomeric complexes via its C terminus20. This oligomerization, coupled with the presence of palmitoylated cysteine residues, is thought to greatly increase the affinity between stomatin and the plasma membrane15. Given the sequence homology between stomatin and podocin, a similar monotopic conformation might be predicted for podocin. Other integral membrane proteins, such as caveolins, possess a hairpin-like membrane domain and form homo-oligomers21. Due to the side-by-side interactions of these oligomers, resulting in the formation of complex structures, caveolin is thought to act as scaffolding in membrane of caveolae21. These elements suggest that podocin may form homo-oligomers resulting in a widespread structure which, in turn, may interact with other podocyte proteins and function as a linker between the plasma membrane and the cytoskeleton.
Although the subcellular localization of podocin is not yet known, it is tempting to speculate that it might interact directly or indirectly with nephrin, which is involved in CNF (ref. 3). It has recently been shown that nephrin can associate with CMS (known as Cd2ap in mice) at the slit diaphragm of podocyte foot processes22,23. CMS is an adapter molecule involved in cytoskeletal rearrangement and, when inactivated, results in a congenital nephrotic syndrome in the mouse22. In contrast, no direct interactions were found between nephrin and ZO-1 despite the fact that they are both localized at the slit diaphragm5. Thus, podocin may have a role in stabilizing interactions between these proteins and anchoring them to the cytoskeleton. Along this line, the podocin C terminus contains six proline residues that may be SH3-binding sites and may therefore interact with one SH3 domain of CMS. Another possibility is that podocin interacts with the focal adhesion proteins involved in cell-matrix interactions at the basolateral membrane of the podocyte. Of note, mice lacking laminin β2 display a nephrotic syndrome (ref. 24).
Regardless of how podocin functions, it has a role in regulating glomerular permeability, as determined by patient phenotype. We have shown, however, that NPHS2 is expressed early during kidney development in the metanephros, and it remains to be determined why inactivation of NPHS2 does not lead to a congenital nephrotic syndrome as is the case with NPHS1.
We detected NPHS2 mutations in 14 of 16 families with SRN for which haplotype analysis was compatible with linkage to the NPHS2 locus. Mutations include premature protein termination as well as missense mutations. The latter are predicted to cause either non-conservative substitutions (P20L, G92C, R291W) or replacement of residues that are highly conserved among the stomatin-like protein family members (R138Q and D160G) and are probably crucial for podocin function. One mutation, G92C, also involves the last nucleotide of exon 1 and probably affects the splicing process. These mutations are therefore expected to be pathogenic. The V180M variant, detected in three affected individuals of different geographic origins involves two non-polar amino acids in a non-conserved amino acid. Its absence from 40 control individual suggests that it is pathogenic.
The R138Q mutation was found in approximately one-third of the patients and, if this prevalence is confirmed using a larger affected population, may be useful in mutation screening. Additional cases need to be tested to clarify whether this mutation is due to a founder effect or recurrent mutation; its occurrence in six families from the same part of Europe favours the former scenario.
Although other loci have been mapped for autosomal dominant FSGS (refs 7,8), NPHS2 is the first gene involved in familial FSGS to be identified. It is likely that NPHS2 mutations will be found in sporadic cases of steroid-resistant idiopathic nephrotic syndrome, which represent an important cause of childhood ESRD. The small number of NPHS2 exons and the detection of a frequent mutation will allow rapid screening of these individuals for mutation of NPHS2. The detection of NPHS2 mutations is of clinical utility; it would prescribe against unnecessary immunosuppressive therapy and permit the prediction of an absence of disease recurrence after kidney transplantation. Furthermore, it is also possible that NPHS2 might have a role in the development of secondary FSGS observed in diseases such as diabetic nephropathy25, HIV nephropathy26 or morbid obesity27.
PAC and cosmid contig construction and isolation of polymorphic markers.
A human genomic PAC library (RPCI 1; ref. 28) and a chromosome 1-specific cosmid library (LL01NC01; ref. 29) were provided by the UK HGMP Resource centre. We isolated PAC clones by PCR screening using microsatellite markers, STS and ESTs spanning the chromosomal region between D1S1640 and D1S3759. Cosmid clones were isolated by hybridization of cosmid library filters with probes corresponding to PAC ends. Cosmid and PAC clones were directly sequenced as described30. ESTs reported in GeneMap'96 and '98 and those listed in the Sanger Centre database as mapping to the gene interval were assigned to the contig by PCR amplification. D1S215 and D1S1640 were retrieved from the Genethon genetic map and the Genome Database, respectively. The D1S3760 marker is localized within intron 12 of SOAT1 as described31. We identified D1S3758 and D1S3759 while sequencing the NGAP intron-exon boundaries and HindIII-SacII subclones of PAC 183f10. PCR amplification of these two markers was carried out using the primers 5′–AACATAAACTCATCCCACC–3′ and 5′–AATAGTAGCTAACTGCCACC–3′ for D1S3758 and 5′–CTTGTAAAGGCTTAGGAATG–3′ and 5′–AGGCAGTCACAGTAGAGGT–3′ for D1S3759at an annealing temperature of 55 °C.
We extracted total RNA from human fetal kidney using the RNAeasy Maxi kit (Qiagen) and purified poly(A)+ RNA with the mRNA purification kit (Pharmacia Biotech). We amplified cDNA from human fetal kidney poly(A)+ RNA by several rounds of 5′ and 3′ RACE–PCR using the Marathon cDNA Amplification kit (Clontech) according to the manufacturer's instructions. Primary and nested PCRs were carried out using the adaptor primers (AP1 and AP2) coupled with primers specific to known cDNA sequences. Subsequent rounds of amplification were performed with primers specific to the newly isolated RACE–PCR clones. RACE–PCR amplification products were subcloned using the pGEM-Teasy vector system II (Promega) and the corresponding clones sequenced. To obtain the 5′ region of the NPHS2 cDNA and to confirm the reconstructed cDNA sequence, we screened a human fetal kidney cDNA library (Clontech) with a 1,065-bp cDNA probe containing the last four exons of NPHS2(positions 728 to 1,792).
We isolated total RNA from human fetal and adult kidney, adult liver and end-stage renal kidneys as described above. After a 30-min incubation with RQ1 DNAse (Boehringer), cDNA synthesis was performed using RNA (5 μg), random primers (600 ng; Promega) and MMLV reverse transcriptase (200 U; Gibco Life Technologies) according to standard procedures. We carried out PCR amplification using NPHS2 cDNA primers, 5′–AGGTGGTGGCGCTGTTGGAG–3′ and 5′–GAAGCAGATGTCCCAGTCGGAATAT–3′, and GAPD-specific primers32 at an annealing temperature of 58 °C.
Nucleotide comparison of the NPHS2 cDNA sequence and those of known genes and ESTs listed in a non-redundant compilation of the EMBL and GenBank databases were performed using the BLAST program33. We carried out amino-acid comparisons with the non-redundant Swiss-Prot database using the BLASTP program. Multiple alignments were performed using CLUSTAL W (ref. 34), protein-domain homologies and motifs predicted using the Prosite database13, membrane topology using the PSORT program14 and the PolII promoter sequences using the TSSG and TSSW programs12.
Genotyping and mutation screening.
We obtained blood samples after informed consent from 16 families originating from Europe, North Africa, Egypt, Turkey and Saudi Arabia. Of these families, nine have already been reported, families 2-9 (ref. 9) and family 10 (ref. 35). All patients presented with SRN according to defined criteria9. Haplotype analysis was compatible with linkage to the NPHS2 locus in all families: five were consanguineous and in one additional family (family 13), affected individuals were homozygous for five consecutive markers spanning the NPHS2 locus. Genotyping of DNA with microsatellite markers was carried out as described9. For the SSCP analysis, exons were amplified by PCR using flanking intronic primers. We selected primers and PCR conditions using the Oligo 5.0 program (NBI): exon 1, 5′–GCAGCGACTCCACAGGGACT–3′ and 5′–TCAGTGGGTCTCGTGGGGAT–3′; exon 2, 5′–AGGCAGTGAATACAGTGAAG–3′ and 5′–GGCCTCAGGAAATTACCTA–3′; exon 3, 5′–TTCTGGGAGTGATTTGAAAG–3′ and 5′–TGAAGAAATTGGCAAGTCAG–3′; exon 4, 5′–AAGGTGAAACCCAAACAGC–3′ and 5′–CGGTAGGTAGACCATGGAAA–3′; exon 5, 5′–CATAGGAAAGGAGCCCAAGA–3′ and 5′–TTTCAGCATATTGGCCATTA–3′; exon 6, 5′–CTCCCACTGACATCTGA–3′ and 5′–AATTTAAAATGAAACCAGAA–3′; exon 7, 5′–CTAAATCATGGCTGCACACC–3′ and 5′–CTTCCTAAAGGGCAGTCTGG–3′; exon 8, 5′–GGTGAAGCCTTCAGGGAATG–3′ and 5′–TTCTATGGCAGGCCCCTTTA–3′; at annealing temperatures of 50 °C (exon 6), 55 °C (exons 2, 3, 4 and 5) and 60 °C (exons 1, 7 and 8). Due to the high GC content of exon 1, we carried out PCR using the Qiagen Taq polymerase and Q-Solution according to the manufacturer's instructions. Moreover, due to its size, the PCR product of exon 1 was digested by SmaI into two fragments before gel electrophoresis. Electrophoresis was carried out for 2 h at 600 V, 25 mA and 15 W with the Genephor Electrophoresis Unit using the GeneGel Excel 12.5/24 kit (Pharmacia). Staining was performed in a GeneStain Automated Gel Stainer using PlusOne Silver Staining kit (Pharmacia).
In situ hybridization.
We paraffin-embedded and sectioned (6 μm) kidneys from three fetuses (12–30 weeks of gestation) and six children or adults (ranging from 1 month to 51 y). After deparaffinization and rehydration, we treated sections by microwave heating in sodium citrate buffer (0.01 M, pH 6) to enhance the hybridization signal. The NPHS2 riboprobes were synthesized from a 1,065-bp PCR product (spanning bases 728 to 1,792 in the NPHS2 cDNA) subcloned into the vector PGEM-Teasy (Promega). The antisense probe was synthesized after digestion with SalI using the T7 polymerase and the sense probe after digestion with SacII using the Sp6 polymerase. The riboprobes were labelled with either digoxigenin-11-UTP (Roche) according to the manufacturer's instructions or [35S]UTP as described36. In situhybridization was carried out as described for the digoxigenin-11-UTP (ref. 37) and the [35S]UTP-labelled probes (ref. 38).
GenBank accession numbers.
Note: supplementary information is available on the Nature Genetics web site http://genetics.nature.com/supplementary_info/).
We thank the patients and their families for participation; E. Al-Sabban, L. Alsford, J.L. André, F. Bouissou, S. Caliskan, E. Kuwertz-Bröcking, B. Lange, J. Nauta and W. Proesmans for referring patients; V. Kalatzis and L. Heidet for critical reading of the manuscript; V. Chauvet for help with in situ hybridization; and Y. Deris for assistance with figure preparation. This study was supported in part by the Association pour l'Utilisation du Rein Artificiel and the Fondation pour la Recherche Médicale. N.B. was supported by grants from the Association Française contre les Myopathies and, subsequently, the Programme Hospitalier de Recherche Clinique.
About this article
Note added in proof:
During the publishing process, J. Kaplan et al. have shown that mutations in ACTN4, mapped to 19q13 and encoding a-actinin-4, an actin-filament cross linking protein, cause autosomal FSGS (ref. 40).