Mutations in PAX2 Associate with Adult-Onset FSGS

Abstract

FSGS is characterized by the presence of partial sclerosis of some but not all glomeruli. Studies of familial FSGS have been instrumental in identifying podocytes as critical elements in maintaining glomerular function, but underlying mutations have not been identified for all forms of this genetically heterogeneous condition. Here, exome sequencing in members of an index family with dominant FSGS revealed a nonconservative, disease-segregating variant in the PAX2 transcription factor gene. Sequencing in probands of a familial FSGS cohort revealed seven rare and private heterozygous single nucleotide substitutions (4% of individuals). Further sequencing revealed seven private missense variants (8%) in a cohort of individuals with congenital abnormalities of the kidney and urinary tract. As predicted by in silico structural modeling analyses, in vitro functional studies documented that several of the FSGS-associated PAX2 mutations perturb protein function by affecting proper binding to DNA and transactivation activity or by altering the interaction of PAX2 with repressor proteins, resulting in enhanced repressor activity. Thus, mutations in PAX2 may contribute to adult-onset FSGS in the absence of overt extrarenal manifestations. These results expand the phenotypic spectrum associated with PAX2 mutations, which have been shown to lead to congenital abnormalities of the kidney and urinary tract as part of papillorenal syndrome. Moreover, these results indicate PAX2 mutations can cause disease through haploinsufficiency and dominant negative effects, which could have implications for tailoring individualized drug therapy in the future.

FSGS is a heterogeneous form of kidney injury defined by partial sclerosis of some but not all glomeruli.^1,2 FSGS can be idiopathic, a result of genetically determined changes in podocytes, or secondary to a variety of renal insults, including reduced nephron mass and vesicoureteral reflux. It is a condition marked by significant proteinuria with or without features of nephrotic syndrome. All forms of FSGS are challenging to treat and frequently lead to ESRD.

Only a minority of individuals with adult-onset FSGS have a family history of disease that suggests a monogenic origin. Nonetheless, the study of familial FSGS and the discovery of genes implicated in this disease, such as INF2, TRPC6, and ACTN4, have yielded important insight into our current understanding of the glomerular filter.^3–5 These studies have provided evidence that dysfunction in the podocyte is central to disease, serving a critical role in glomerular filtration. Monogenic adult-onset FSGS is genetically heterogeneous, with mutations in INF2, TRPC6, and ACTN4 accounting for 9%, 3%, and 2% of our own cohort of families, leaving a substantial number of unexplained pedigrees.⁶

Exome analysis is facilitating the discovery of disease-causing genetic alterations in small, previously uninformative families. To identify additional FSGS genes, we exploited this technology, coupled with high-throughput Sanger sequencing, in a cohort of FSGS families with unexplained genetic etiology. This sequencing effort identified a disease-segregating PAX2 missense mutation in a family designated FG-EQ (Figure 1). The product of the PAX2 gene, one of the nine members of the family of paired box (PAX) transcription factor genes, plays a critical role in kidney development.^7–9 Mutations in this gene have been associated with congenital abnormalities of the kidney and urinary tract (CAKUT) as part of a syndrome known as papillorenal syndrome (PRS) in which ocular manifestations also occur (Mendelian Inheritance in Man, 120330). We report that heterozygous PAX2 mutations account for 4% of adult FSGS and perturb PAX2 function by affecting proper binding to DNA or enhancing its interaction with repressor proteins.

Figure 1.

FG-EQ pedigree, sequencing and multisequence alignment demonstrating conservation of the affected residue, Gly189. (A) Pedigree for family FG-EQ. Affected individuals are indicated in gray. One indeterminate individual is indicated with a half-shaded icon. Individuals heterozygous for the PAX2 p.G189R mutation are denoted by a plus sign while individuals without the mutation are denoted by a minus sign. Individuals for whom no DNA was available have no notation. Exome sequencing was performed in individuals FG-EQ III(8) and FG-EQ IV(8). (B) Electropherograms obtained from Sanger sequencing of exon 5 of the PAX2 gene confirming the c.565G>A missense change (p.G189R). (C) Alignment of PAX2 orthologs across 7 species demonstrating conservation of the affected residue, Gly189 (asterisk).

Results

Genetics

Targeted enrichment was performed on genomic DNA obtained from two affected individuals from family FG-EQ, who were separated by three meioses (Figure 1A). Massively parallel sequencing resulted in 36,116,715 and 66,666,479 seventy-four–base pair single-end reads. Following alignment, target region coverage had an average sequencing depth of 30× and 52× for the two samples. Collectively, the total number of variants called was 86,328 (81,338 single-nucleotide polymorphisms [SNPs] and 4990 small indels). Among them, 75,753 SNPs and 4394 indels had been annotated in dbSNP137 (ftp://ftp.ncbi.nih.gov/snp). Variants annotated in the 1000 Genomes Project (ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/) and National Heart, Lung, and Blood Institute Exome Sequencing Project (http://evs.gs.washington.edu/EVS/) were removed, leading to a total of 137 SNPs and 15 indels that were previously unreported. Of these, 22 nonsynonymous SNPs and 3 indels shared by the affected individuals were retained in the analysis (Supplemental Table 1). An overall prioritization score was obtained for each variant using ranking parameters after functional annotation was performed, retrieving information from several data sources (Supplemental Tables 2 and 3). PAX2 was identified as the most promising candidate as a result of this analysis and given its known association with CAKUT and PRS.^10–12 The c.565G>A missense variant, predicting the p.Gly189Arg amino acid substitution, was validated by Sanger sequencing (Figure 1B). Genotyping of members of the family for whom DNA was available documented cosegregation of the variant with disease. Of note, Gly189 is a highly conserved residue within the octapeptide motif, a functionally important and conserved region (Figure 1C), and its substitution by arginine was predicted to be damaging with high confidence (PolyPhen-2 [http://genetics.bwh.harvard.edu/pph2/], score 0.99; SIFT [http://sift.jcvi.org/], score: 0).

To investigate the potential role of PAX2 variation in other FSGS families, Sanger sequencing of the entire coding region of the gene was performed using genomic DNA from 175 additional unrelated subjects with familial disease. We also performed PAX2 mutation screening in a second cohort of 85 individuals with CAKUT. This analysis identified seven heterozygous single nucleotide substitutions in each of the familial FSGS and CAKUT groups (4% and 8% of cases, respectively) (Table 1). None of the variants found were present in sequences of approximately 7592 nominally normal individuals available in public databases, including the Exome Sequencing Project, the 1000 Genomes Project, or dbSNP137. Two of the variants isolated in FSGS families (p.Arg104× and p.Thr164Asn) were previously identified in patients with CAKUT, with p.Thr164Asn considered benign.¹³ In the FSGS families, the mutation of interest was documented in all affected individuals where DNA was available. Incomplete penetrance was documented in one family (Supplemental Figure 1). Direct sequencing demonstrated the de novo origin of the PAX2 mutation in the three patients with CAKUT for which parental DNAs were available and short tandem repeat genotyping confirmed paternity (Table 1). Examination of the burden of coding nonsynonymous variants in the entire PAX2 gene in familial FSGS revealed significant enrichment for variants compared with 6503 controls in the Exome Sequencing Project (P<0.05 by two-tailed chi-squared test with Yates correction). A statistically significant difference was also found when we compared the burden of CAKUT variants to control variants (Supplemental Table 4). Specifically, the number of rare variants (defined as minor allele frequency ≤1%) divided by the total number of haplotypes was higher in the patient groups than in controls.

Table 1.

List of PAX2 variants in FSGS families and CAKUT cohorts

Family or Individual ID	DNA Change	Exon	Amino Acid Change	Domain	Polyphen2 Prediction
FSGS families
FG-BF	c.491C>A	4	p.Thr164Asn	In between paired and octapeptide	Probably damaging
FG-DG	c.239C>T	3	p.Pro80Leu	Paired: linker subdomain	Probably damaging
FG-EQ	c.565G>A	5	p.Gly189Arg	Octapeptide	Probably damaging
FG-GE	c.398C>T	3	p.Ser133Phe	Paired: C terminus subdomain	Probably damaging
FG-IX	c.167G>A	2	p.Arg56Gln	Paired: N terminus subdomain	Possibly damaging
FG-JO	c.448A>G	4	p.Thr150Ala	In between paired and octapeptide domain	Benign
FG-KV	c.310C>T	3	p.Arg104×	Paired: C terminus subdomain	NA
CAKUT individuals
CKT-11Ca	c.887T>C	8	p.Leu296Pro	Transactivation	Probably damaging
CKT-34C	c.415A>G	4	p.Ile139Val	Paired: C terminus subdomain and transactivation	Both predicted to be benign
	c.985A>G	8	p.Thr329Ala
CKT-35C	c.5A>G	1	p.Asp2Gly	Paired domain: before N terminus subdomain	Probably damaging
CKT-39Ca	c.892C>T	8	p.Pro298Ser	Transactivation	Benign
CKT-46C	c.887T>C	8	p.Leu296Pro	Transactivation	Probably damaging
CKT-54C	c.1240T>C	11	p.Tyr414His	Transactivation	Probably damaging
CKT-89Ca	c.884C>T	8	p.Ala295Val	Transactivation	Possibly damaging

Nucleotide and amino acid sequence changes are reported using the following National Center for Biotechnology Information RefSeq accession numbers (NM_003987 and NP_003978). NA, not available.

^aPatients with CAKUT for whom de novo mutations could be confirmed.

The PAX2 protein is a multidomain transcription factor characterized by an N-terminal DNA-binding paired domain and a transactivation domain at the C terminus (Figure 2). In the FSGS group, six of the seven heterozygous mutations were missense and altered amino acid residues located outside of the transactivation domain (Figure 2A). One nonsense mutation was identified that resulted in a premature stop codon within the C terminus of the paired domain. All of the mutations in the CAKUT cohort were missense, with the majority involving residues clustering within the transactivation domain (Figure 2A). We compared the location of these mutations to those listed in the PAX2 variant database (www.lovd.nl/PAX2), which catalogs published disease-causing mutations (Figure 2B, Supplemental Figure 2). Most PRS-causing mutations were found to be truncating (28 nonsense, frameshift, or splice site changes, representing 79% of total cases). Of note, all the PRS-associated missense changes and small in-frame indels (16 different changes, accounting for 21% of cases) affected the paired domain, in striking contrast with the distribution of CAKUT-associated mutations identified in the present study.

Figure 2.

PAX2 domain structure and localization of FSGS-associated and CAKUT- and PRS-causing mutations. (A) Location of mutations identified in the study are shown in the schematic PAX2 domain structure. PAX2 is characterized by an N-terminal paired domain consisting of the N terminus (residues 16–76, violet) and C terminus (88–142, blue) separated by a small linker region (cyan). The relative locations of the other domains, including the octapeptide motif (residues 185–192, green), homeodomain (yellow), and transactivation domain (red), are also indicated. Mutations identified in the familial FSGS cohort are shown above the cartoon, while those identified in the CAKUT cohort are reported below the cartoon. The asterisk indicates co-occurring mutations in the same individual. (B) Location of missense changes and small in-frame indels reported in the PAX2 variant database (www.lovd.nl/PAX2). Variants with uncertain clinical impact are not shown. Mutations identified in patients with PRS are indicated with red characters, while those occurring in patients with renal dysplasia/hypoplasia and isolated ocular involvement are in green and blue, respectively.

Clinical Characteristics

The clinical characteristics for the FSGS families and individuals with CAKUT are outlined in Tables 2 and 3. In the index family FE-EQ, ages of disease onset ranged from 17 years in FG-EQ IV(8) to 68 years in FG-EQ III(8). ESRD occurred in two of five affected individuals, at ages 40 and 58 years. Urography was performed in these two individuals, with evidence of bilateral renal pelvis dilatation in one. All five affected individuals had ultrasonographic examinations that identified no other structural abnormalities. No ocular or auditory abnormalities were documented. FG-EQ IV(7) had a slightly elevated 24-hour urine collection at age 39 years, but a repeat collection was normal. His status was defined as indeterminate. Paraffin-embedded kidney biopsy tissue from individual FG-EQ III(8) was examined. Periodic acid–Schiff staining revealed several segmentally sclerosed glomeruli; electron microscopy showed diffuse podocytopathy as evidenced by degenerative changes, microvillous transformation, vacuolization, and lysosome accumulation but with focal foot process effacement (Figure 3, A and B). These electron microscopy findings are similar to those observed in biopsy samples from individuals with other genetic causes, such as ACTN4-associated disease.¹⁴

Table 2.

Clinical features of FSGS families with PAX2 mutation–associated disease

Family ID	Self-Reported Ethnicity	Ages at Disease Onset (yr)	Persons Affected (n)	Patients with ESRD (n)	Ages at Development of ESRD (yr)	Ultrasonography Findings	Diagnosis	Patients with Biopsy (n)
FG-BF	White	8	2	1	Unknown	Increased echogenicity	FSGS	1
FG-DO	African American	7–11	2	Unknown	Unknown	Unknown	Proteinuria	Unknown
FG-EQ	European	17–68	6	2	40, 58	Dilated renal pelvis, small kidneys	FSGS	2
FG-GE	Unknown	Unknown	2	1	Unknown	Slightly small kidney, calyceal diverticulum	Proteinuria	1
FG-IX	Middle Eastern	36	4	Unknown	Unknown	Unknown	FSGS	Unknown
FG-JO	East Indian	31–32	5	4	30–36	Unknown	FSGS	3
FG-KV	European American	15–24	3	1	42	Unknown	FSGS, undiagnosed PRS	3

Table 3.

Clinical features of patients with CAKUT who have PAX2 mutation–associated disease

Individual ID	Age at Diagnosis (yr)	Clinical Features
CKT-11	8	Solitary kidney with moderate hydronephrosis and hydrocele
CKT-34	2	Unilateral UPJO. Associated nonrenal manifestations, including dysmorphic facial syndrome, ventriculomegaly, and seizures. Result of genetic testing for Opitz G syndrome was negative.
CKT-35	4	Unilateral UPJO and bilateral VUR
CKT-39	4	Solitary kidney and mild unilateral ventriculomegaly
CKT-46	3	Horseshoe, small ectopic left kidney. Multiple congenital abnormalities and dysmorphic features; hemifacial microsomia.
CKT-54	5	Unilateral UVJO/UPJO
CKT-89	3	Left UPJO, right VUR

UPJO, ureteropelvic junction obstruction; VUR, vesicoureteral reflux; UVJO, ureterovesical junction obstruction.

Figure 3.

Pathologic findings in a renal biopsy specimen from an affected individual in family FG-EQ demonstrated FSGS. (A) Representative image of a paraffin section from individual FG-EQ III(8) stained with periodic acid–Schiff. Glomeruli are enlarged, and several show segmental sclerosis (asterisk) and focal adhesion of the tuft to Bowman’s capsule (arrow). Periodic acid-Schiff–positive protein casts are present (arrowhead). Scale bar=20 µm. (B) Representative electron micrograph from the same kidney shown in part A, demonstrating damage to the glomerular filtration barrier and signs of diffuse podocyte injury characterized by extensive podocyte foot process effacement (arrowheads), microvillous transformation (arrow), and lysosome and vacuole formation within the cytoplasm (asterisk). Scale bar=5 µm (direct magnification: ×5000).

Clinical re-evaluation of the affected members of family FG-KV heterozygous for the nonsense mutation revealed a more severe phenotype, compatible with undiagnosed PRS. This same mutation was previously described in PRS.¹⁵

In Silico Structural Analysis

Detailed molecular modeling was performed to explore the effects of the identified PAX2 missense mutations. The domain structure of PAX2 can be inferred with high reliability because of its high sequence homology with other PAX family proteins for which crystallographic structures are available. The PAX2 protein comprises an N-terminal DNA-binding paired box domain consisting of two subdomains separated by a linker region, which is followed by a highly conserved octapeptide that is involved in functional modulation of the protein by specific interactions with Groucho/TLE/Grg proteins, converting PAX2 from a transcriptional activator to a potent repressor. The C-terminal half of the protein contains a partial so-called homeodomain that can function as a second DNA-binding motif, or as a protein–protein interaction motif, and a transactivation domain, which regulates gene transcription.^13,16–20

Three of the FSGS missense mutations, p.Arg56Gln, p.Pro80Leu, and p.Ser133Phe, affected residues located in the paired domain near the N terminus of the protein. Although no structural data for PAX2 are available, crystallographic structure of this highly conserved domain has been determined for PAX5 in complex with DNA and the transcription cofactor ETS-1 (PDB ID 1K78), allowing the use of this structure to predict the effect of PAX2 substitutions without the need to generate a homology model (Figure 4A).²¹ The sequence of this domain in PAX5 differs from that of PAX2 by just three residues (97, 122 and 123), which are far from those residues affected by the mutations identified here. All three FSGS mutations in this region were predicted to affect PAX2 binding to DNA. Specifically, both the p.Arg56Gln and p.Ser133Phe substitutions were expected to disrupt interactions at these sites that directly interact with DNA through an ion pair and H-bond, respectively. Furthermore, the p.Arg56Gln substitution was also predicted to interrupt the electrostatic interaction with ETS-1 cofactor at this site (Figure 4A). According to the structural model, the third mutation, p.Pro80Leu, located in the linker region between the N-terminal and C-terminal regions of the paired domain, does not interact directly with DNA; however, because of the peculiar conformational properties of proline, this substitution is expected to alter the flexibility and conformation of the linker (Figure 4A).²²

Figure 4.

Structural analysis and molecular modeling predict that FSGS mutations located in the paired domain of PAX2 affect DNA binding, while the lesion in the octapeptide region increases PAX2 affinity for TLE corepressors. (A) In the upper panel, the structure of a complex between the paired domain of PAX5 (green), DNA, and the ETS-1 cofactor (magenta) is shown. PAX5 in this domain differs from PAX2 by just three residues, which are located far from those affected by the FSGS paired domain mutations. The FSGS paired domain missense mutations are indicated in red. H-bonds or ion pairs formed by these residues are shown as thin cyan lines. (B) In the middle panel, location of FSGS, CAKUT, and PRS missense mutations in the paired domain are shown. Residues affected by FSGS and CAKUT mutations identified in this study are shown in red and blue, respectively. Residues affected by missense mutations previously reported in the PAX2 variants database are shown in cyan. (C) In the left lower panel, the homology model of the complex between the PAX2 p.Gly189Arg octapeptide and TLE1. TLE1 is shown in gray, with its hydrophobic residues interacting with the octapeptide represented as light blue spheres and side chains forming H-bonds with the octapeptide as light blue sticks. The backbone of the octapeptide is shown as a green ribbon, and its hydrophobic and hydrophilic side chains are reported as blue and yellow sticks, respectively. The mutated residue (p.Gly189Arg) is in magenta. (D) In the right lower panel, electrostatic interactions of p.Arg189 in the homology model of the complex between the octapeptide (green) and TLE1. The surface of TLE1 is colored according to the electrostatic potential (from red to blue for potential values increasing from −10 to 10 kT/e).

One of the identified PAX2 missense mutations in the FSGS group, p.Gly189Arg, was located in the octapeptide region. Experimental evidence supports the view that this motif plays a negative modulatory role on PAX2 transcriptional activity.²³ Such transcriptional repression has been shown to be mediated by its interaction with the Groucho/TLE/Grg family of corepressors.^19,20 On the basis of its location, we hypothesized that the p.Gly189Arg substitution might influence PAX2 interaction with TLE proteins, thus perturbing switching between its active and inhibited conformation. To explore this hypothesis, a homology model of the PAX2 octapeptide domain bound to TLE1 was constructed (Figure 4, B and C). Structural analysis of this model supports the idea that, by introducing a cationic side chain in correspondence of a region of negative electrostatic potential of the corepressor, the arginine residue improves the electrostatic interaction between PAX2 and TLE1. The p.Gly189Arg substitution is also predicted to affect the conformational freedom of the octapeptide when in solution. This region of PAX2 is probably unstructured when not associated with TLE proteins. Therefore, the p.Gly189Arg substitution would reduce the degree of conformational disorder in the free octapeptide and the entropic cost of peptide immobilization and helical structuring on the TLE1 surface. This would provide an additional contribution to an increased binding affinity for the mutant.

The remaining FSGS associated mutations, p.Thr150Ala and p.Thr164Asn, are located in the region between the paired domain and the octapeptide, where no structural data are available for modeling. For this reason, no predictions can be attempted on the structural effects of these lesions. However, it is interesting to note that phosphorylation prediction servers PhosphoMotif and KinasePhos 2.0 indicate both Thr150 and Thr164 as possible phosphorylation sites.^24,25

With respect to the missense variants identified in the CAKUT cohort, only one is located in the paired domain (p.Ile139Val) (Figure 4A). This variant is found in an individual harboring a mutation in the transactivation domain as well, and it is difficult to know the extent of each variants’ contribution to disease. In PAX2, Ile139 is part of the hydrophobic core of the C-terminal subdomain of the paired box domain. Valine is less bulky than isoleucine, and therefore the p.Ile139Val substitution could form a small cavity in this core but how this affects the biochemical and DNA binding properties of the PAX2 protein is difficult to predict.

No predictions are possible for the other CAKUT mutations identified in this study located in the transactivation domain because no structural data are available. Interestingly, similar to what was observed for Thr150 and Thr164, residue Thr329 (affected by the substitution p.Thr329Ala) was predicted as a possible phosphorylation site.

Functional Studies

To characterize the functional behavior of FSGS-associated PAX2 mutants, the Arg56Gln, Pro80Leu and Ser133Phe (DNA-binding domain), and Gly189Arg (octapeptide motif) amino acid substitutions were selected for further study. All mutants were efficiently expressed in HEK 293T cells, with levels that were similar to those of the wild-type protein, basally or in presence of TLE4, a well known repressor of PAX2 function (Supplemental Figure 3). The transactivation ability of each mutant, alone or in combination with TLE4, was measured by luciferase assays performed using transiently transfected HEK 293T cells (Figure 5A). As predicted by molecular modeling, the mutants exhibited different behavior in transactivation properties. Cells transiently expressing the PAX2^G189R mutant showed an efficient induction of luciferase expression basally, which was equivalent to that observed for the wild-type protein, indicating unaffected DNA-binding capability to the PRS4 consensus binding site. In contrast, cells expressing the PAX2^R56Q, PAX2^P80L, and PAX2^S133F mutants were characterized by a significantly weaker induction, which was consistent with the predicted perturbing effect of mutations on DNA binding. However, TLE4 was documented to dramatic reduce the transactivation activity of the PAX2^G189R mutant compared with wild-type (Figure 5A). This finding, consistent with our in silico structural analyses, supports the idea of a more stable interaction of this mutant with proteins of the TLE family. To confirm this hypothesis, coimmunoprecipitation assays were performed in HEK 293T cells coexpressing myc-tagged TLE4 together with hemagglutinin (HA)-tagged wild-type PAX2 or the PAX2^G189R mutant, which documented a dramatically enhanced interaction of the mutant with TLE4 (Figure 5B).

Figure 5.

Functional characterization of FSGS-associated PAX2 mutations revealed results consistent with structural analysis and molecular modeling predictions. (A) Transactivation assays. Induction of the luciferase reporter (PRS4-luc) in response to cotransfection with wild-type (WT) PAX2 or individual mutants, basally (left). The PAX2^G189R mutant exhibits a transactivation activity similar to that of the wild-type protein, while a significantly reduced increase in reporter expression is documented in cells expressing the other mutants. Luciferase induction observed in the presence of TLE4 is also shown, reported as the fold inhibition relative to wild-type protein (right). Coexpression of the repressor protein TLE4 results in an enhanced inhibition of luciferase induction in cells expressing the PAX2^G189R mutant, compared with wild-type PAX2 and the other FSGS-associated mutants. Values are expressed as the means±SEM of six independent experiments. *P≤0.05; **P≤0.01. All the experiments were normalized to a renilla internal control vector. (B) Coimmunoprecipitation assays. Coimmunoprecipitation of Myc-tagged TLE4 with HA-tagged PAX2 (wild-type or G189R mutant) using lysates from transiently transfected HEK 293T cells is shown. Anti-HA immunoprecipitates (top panels) and total cell lysates (bottom panels) were analyzed by Western blotting with the indicated antibodies. Note that the PAX2^G189R mutant forms a more stable complex with TLE4 compared with wild-type PAX2. GAPDH, glyceraldehyde 3-phosphate dehydrogenase. (C) Chromatin immunoprecipitation assays. National Institutes of Health 3T3 cells were transiently transfected with the PRS4-luciferase construct in the presence of HA-tagged PAX2 proteins, with or without Myc-tagged TLE4. Primer pairs against the PRS4 element were used for realtime quantitative PCR. Relative amounts of PCR products are expressed as fold of enrichment normalized versus input and versus background (mouse IgG antibodies). Wild-type PAX2 and the PAX2^G189R mutant more efficiently binds to PRS4 compared with the PAX2^R56Q, PAX2^P80L, and PAX2^S133F mutants. For all experiments, equivalence of PAX2 and TLE4 expression levels is shown in Supplemental Figure 3.

To demonstrate that the weak transactivation activity of the PAX2^R56Q, PAX2^P80L, and PAX2^S133F proteins was due to decreased DNA binding, chromatin from National Institutes of Health 3T3 cells expressing HA-tagged wild-type PAX2 or each mutants, in the presence or absence of TLE4, was prepared 48 hours after transfection for chromatin immunoprecipitation assays (Figure 5C). Consistent with the previous findings, binding to the PRS4 promoter sequence was documented for the wild-type protein and PAX2^G189R mutant, while a reduced binding to DNA was observed for the other mutants.

Overall, these data provide evidence for diverse perturbing effects of the FSGS-associated PAX2 mutations on protein function, affecting proper binding to DNA (p.Arg56Gln, p.Pro80Leu, and p.Ser133Phe) or enhanced interaction with repressors (p.Gly189Arg).

Discussion

Our genetic, in silico analysis and functional data suggest that PAX2 missense variants may lead to an expanded phenotypic spectrum that includes FSGS, through haploinsufficiency and/or dominant negative effects. Our data support the view that mutations in PAX2 might account for a significant proportion (approximately 4%) of families designated to have hereditary FSGS. Mutations in this key kidney development transcription factor have also been reported to cause congenital abnormalities of the kidney and urinary tract. Our screening demonstrates that PAX2 mutations accounts for disease in 8% of a CAKUT cohort.

We expand the phenotypic spectrum associated with PAX2 mutations to include not only CAKUT but also autosomal dominant adult-onset FSGS in the absence of other syndromic features. Given the known role of PAX2 in kidney development, it is not surprising that mutations in this gene can lead to congenital and structural disease. It is possible that some of the families in our cohort designated as having familial primary FSGS have this pathologic lesion due to subtle developmental abnormalities, such as reduced nephron mass, although we are unable to quantitate this. Furthermore, the presence of nephromegaly, which can be seen with reduced nephron mass, is also observed in primary FSGS. Nonetheless, this possibility demonstrates the limitations and heterogeneous nature of the pathologic diagnosis we call FSGS. It also highlights the potential benefits of clarifying diagnosis in such a heterogeneous disorder through genetic analysis to avoid unnecessary immunosuppressive agents. Of note, most PRS-causing PAX2 mutations result in truncated proteins, which is in contrast with the preponderance of missense mutations in our familial FSGS cohort. We speculate that these hypomorphic mutations may have a role in leading to less severe disease, as defined by phenotype and relatively late ages of clinical presentation. This may also help to explain the “incomplete penetrance” observed in one FSGS family, where carriers of the variant may have subtle undetectable clinical characteristics (Supplemental Figure 1). If true, this would be analogous to some individuals with PKD2 mutations who have a milder renal cystic phenotype that can often go undetected.²⁶

We also consider an alternate mechanism by which these variants may lead to segmental scarring. During kidney development, PAX2 is important in activating downstream targets that allow for communication between the ureteric bud and surrounding metanephric mesenchyme, the latter of which ultimately epithelializes to form podocytes.⁸ Evidence reported in the literature suggests that PAX2 strongly represses the expression of WT1, a nuclear protein expressed in podocytes, by interacting with Groucho/TLE/Grg proteins through its octapeptide motif, and activates it in their absence, while WT1 is a PAX2 repressor.^19,20 We hypothesize that dysregulation of PAX2 targets, such as WT1, may lead to FSGS by disrupting the development and/or function of the podocyte.^20,27–29

The functional data presented here support structural predictions that different missense mutations lead to FSGS through different mechanisms. Loss of function of a PAX2 allele has been previously described to be the inciting event in renal disease, in part due to the predominance of nonsense mutations reported in PAX2 related autosomal dominant conditions. Further support of PAX2 gene dosage being critical for normal development was obtained in mouse models where knockout of or nonsense mutations in the PAX2 gene led to a phenotype analogous to PRS.^30–32 We provide evidence that missense mutations can lead to disease through loss of function but by diverse mechanism(s) as well, which could have implications for individualizing drug therapy.²⁷

Concise Methods

Patients

Individuals belonging to 176 families with FSGS and 85 individuals with CAKUT were included in this study. Inherited cases were defined as families with two or more affected individuals. All families studied had an inheritance pattern consistent with autosomal dominance. Familial FSGS affected status was defined as having either a reported history of proteinuria with urine albumin-to-creatinine ratio >250 mg/g, nephrotic syndrome, or biopsy-proven FSGS in a family with at least one other case of documented FSGS or nephrotic syndrome. CAKUT-affected status was defined as having fetal or postnatal ultrasonographic evidence of the following: renal agenesis, renal dysplasia (undifferentiated renal tissue), renal hypoplasia, duplex kidney, horseshoe kidney, ureteropelvic junction obstruction with and without megaloureter (ureter is refluxing or obstructed), duplication of the ureter, ureteral agenesis, vesicoureteral reflux, and ectopic ureter (abnormally located terminal portion of the ureter, often ending in the urethra). Included CAKUT individuals did not have associated symptoms consistent with PRS. We obtained blood or saliva for DNA isolation as well as clinical information after receiving informed consent from participants in accordance with the Institutional Review Board at the Beth Israel Deaconess Medical Center and Children’s Hospital Boston. Clinical information was obtained from telephone interviews, questionnaires, and physician reports. Genomic DNA was extracted from blood or saliva samples using standard procedures.

Exome Sequencing and Sequence Data Analysis

Targeted enrichment and parallel sequencing was performed on genomic DNA belonging to two affected individuals from a family with FSGS. Exome capture was performed using NimbleGen SeqCap EZ Exome v2 (NimbleGen, Madison, WI), which is estimated to cover 98% of the human coding genome corresponding to the Consensus Conserved Domain Sequences database and 710 micro RNAs. Enriched libraries were sequenced by 74–base pair, single end read sequencing on an Illumina GAII machine (Illumina Inc., San Diego, CA). Next-generation sequencing reads were aligned to the most recent reference human genome (UCSC hg19), with the Burrows-Wheeler Aligner (v.0.5.9-rcl).³³ The Genome Analysis Toolkit was used to further process the aligned read data, and the same program was used to genotype the individuals from the processed read data.^34,35 Variants were filtered against dbSNP 137, the 1000 Genomes Project, and the Exome Sequencing Project. We further used the Integrative Genomics Viewer to manually identify some of the novel variants as artifacts.³⁶ Variants were filtered by comparison to the other related affected individual sent for exome sequencing—variants that appeared in both affected individuals remained in the analysis, even at low coverage, given the unlikely possibility of this happening by chance. If a variant was seen at low coverage in one sample but not in the other because of noncapture, it was kept in the analysis as well. Variants of interest were confirmed by Sanger sequencing. Functional annotation of candidate genes was performed by retrieving information from several data sources containing annotations on Gene Ontology terms (GO project), Online Mendelian Inheritance in Man reports, human phenotypes (Human Phenotype Ontology project), mouse phenotypes (Mouse Genome Database–MGD phenotypes), protein-protein interactions (STRING), pathways (KEGG PATHWAY), gene expression data (Gene Atlas), protein domains (Pfam, InterPro), and literature (National Center for Biotechnology Information’s PubMed). Candidate genes were then prioritized with GeneDistiller, using functional relationships to genes already known to be implicated in similar disease phenotypes (i.e., ACTN4, APOL1, CD2AP, INF2, MYO1E, NEDE, NPHS1, NPHS2, PLCE1, TRPC6, and WT1) as ranking parameters.³⁷

Sanger Sequencing

Sanger sequencing was performed on all FSGS and CAKUT samples using a Big Dye 3.1 terminator cycle sequencing kit (Life Technologies, Grand Island, NY) and analyzed with an ABI Prism 3730 XL DNA analyzer (Applied Biosystems, Foster City, CA). Primer sequences are available on request. Sequence chromatograms were analyzed using the Sequencher software (Gene Codes, Ann Arbor, MI). Specific variants identified in family probands were sequenced in all available affected family members to investigate whether the variant segregated with disease (Supplemental Figure 1). If an affected individual did not harbor the variant of interest, it was excluded as disease-causing.

Parental Status

DNA belonging to three individuals with CAKUT along with their parents was available for DNA profiling to establish biologic parental status. PowerPlex 16TM HS PCR Amplification Kit (Promega, Madison, WI) was used to amplify and genotype 16 loci (15 STR loci and amelogenin) in nine human genomic DNA samples according to protocol.

Molecular Modeling and Structural Analysis

The PAX2 octapeptide complex with the WD-repeat domain of TLE1 was modeled by homology to the structure of TLE1 bound to an eh1 motif peptide of human Goosecoid (PDB ID 2CE8).²¹ The sequence of this peptide (MFSIDNILA) is representative of the consensus octapeptide sequence (F/Y) XIXXILX (where X can be any amino acid), typical of Engrailed/Goosecoid/Nkx, and PAX proteins, with the addition of an N-terminal methionine. Most differences between this sequence and the octapeptide of PAX2 (YSINGILG) (D→N, N→G, A→G) involve residues facing the water phase, and not implicated in any specific interactions. The only exception is substitution F→Y, which is involved in hydrophobic interactions with protein residues. However, this substitution is rather conservative and is easily accommodated in the protein pore where this side chain inserts. The octapeptide sequence was mutated to that of wild-type PAX2 by using the program UCSF Chimera, and the system energy was minimized with the “repair” function of the program FoldX 3.0.^24,38 The same approach was also used to introduce the p.Gly189Arg substitution. The electrostatic potential generated by TLE1 on its surface was calculated with the program UCSF Chimera.³⁸

The position of mutated amino acid residues in the paired box domain, and their possible structural and functional effects, were analyzed using the crystallographic structure of PAX5 in complex with DNA and the transcription cofactor ETS-1 (PDB ID 1K78). The sequence of the paired domain of PAX5 differs from that of PAX2 by just three residues (97, 122, and 123), all relatively far from those affected by the mutations, so that the generation of a homology model was not necessary.³⁹ Phosphorylation sites were predicted with the servers PhosphoMotif and KinasePhos 2.0.^39,40

DNA Constructs

Full-length HA-tagged PAX2 and Myc-DDK tagged TLE4 cDNAs were provided by OriGene Technologies. The FSGS-associated PAX2 mutations resulting in the p.Arg56Gln (c.167G>A), p.Pro80Leu (c.239C>T), p.Ser133Phe (c.398C>T) and p.Gly189Arg (c.565G>A) amino acid substitutions (NP_003978.2, NM_003987) were introduced by site-directed mutagenesis (Agilent Technologies). The reporter plasmid, PRS4-Luc, containing five tandem repeats of the PAX2 DNA-binding site, cloned upstream of the herpes simplex virus thymidine kinase promoter, was provided by Dr. G.R. Dressler (University of Michigan Medical School, Ann Arbor, MI).

Transactivation Assays

HEK 293 cells (ATCC) were cultured in DMEM supplemented with 10% FBS, 100 U/ml penicillin, and 100 mg/ml streptomycin in 5% CO₂/95% air at 37°C. Cells were seeded in 12-well plates and transfected using 250 ng of reporter plasmid and 500 ng of wild-type PAX2 or each of the four mutant cDNAs, with or without 1 μg of TLE4 plasmid, using Fugene 6, according to the manufacturer’s protocol (Promega). Twenty-four hours after transfection, firefly luciferase and renilla luciferase activities were assayed using the Dual Luciferase Reporter Assay Kit (Promega) in a Lumat LB9501 luminometer (E&G Berthold).

Immunoprecipitation of PAX2/TLE4 Complex

HEK 293T cells were transfected with each of the PAX2 WT or p.Gly189Arg constructs, with or without TLE4 as indicated in the Figure 5 legend. Forty-eight hours after transfection, lysates were prepared in immunoprecipitation protocol (IP) buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 0.5% Triton X-100, and protease inhibitor mixture). Lysates were incubated with anti-HA monoclonal antibodies for 16 hours at 4°C. Antibodies were captured with protein A-Sepharose for 2 hours at 4°C and washed five times with IP-wash buffer (same as IP buffer except for containing 0.1% Triton X-100). Proteins were eluted from protein A-Sepharose by boiling in SDS-PAGE sample buffer, and Western blot was performed.

Chromatin Immunoprecipitation

National Institutes of Health 3T3 cells were transfected with each of the PAX2 constructs and PRS4-Luc, with or without TLE4. Forty-eight hours after transfection, cells were fixed with 1% formaldehyde in culture medium (10 minutes). Cross-linking was stopped by the addition of glycine to 0.125 M. Cell pellets were washed in PBS, suspended in cell lysis buffer (5 mM PIPES [pH 8.0], 85 mM KCl, 0.5% NP40, and protease inhibitors), incubated at 4°C for 5 minutes, and centrifuged at 6000 rpm for 5 minutes. The nuclei were resuspended in sonication buffer (50 mM Tris-HCl [pH 8.1], 10 mM EDTA, 1% SDS, and protease inhibitors), incubated at 4°C for 10 minutes, and then sonicated on ice with five 20-second pulses. Sonicated lysates were cleared by centrifugation at 4°C for 15 minutes. Chromatin was diluted in IP dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl [pH 8.1], and 167 mM NaCl) and precleared with 80 μl protein G-agarose. Each immunoprecipitation was performed using 5 μg anti-HA antibody (sc-805×; Santa Cruz Biotechnology). After overnight incubation at 4°C, 60 μl protein G-agarose were added. Following 2-hour incubation, the beads were sequentially washed two times in IP dilution buffer, two times in high-salt buffer (0.1% SDS, 1% Triton X-100, 2mM EDTA, 20 mM Tris-HCl [pH 8.1], 500 mM NaCl), two times in LiCl buffer (100 mM Tris-HCl [pH 8.1], 500 mM LiCl, 1% NP-40, 1% Na deoxycholate), and finally two times in Tris-EDTA. Bound complexes were then eluted by vortexing beads twice for 15 min at 25°C in 250 μl of elution buffer (50 mM Na bicarbonate and 1% SDS). NaCl, 5 M, was added to a final concentration of 0.2 M to the pooled eluates, and cross-links reversed by incubating samples at 65°C overnight. The samples were digested with proteinase K for 1 hour at 56°C, and DNA isolation was performed using column purification. Precipitated DNA was reconstituted in sterile water, and quantitative real-time PCR of precipitated genomic DNA relative to inputs was performed using the primers 5′-GCTACCGGACTCAGATCTCG-3′ (PRS4-forward) and 5′-TGCGAAGTGGACCTCGGACC-3′ (PRS4-reverse).

Histochemistry

Formalin-fixed human kidney tissue was paraffin-processed and sectioned at 4 μm. After processing for antigen retrieval (pressure cooker in citrate buffer [pH 6]), sections were stained with Periodic-acid Schiff. Images of representative glomeruli were taken with an Olympus BX53 microscope equipped with an Olympus DP72 camera.

Electron Microscopy

Ultrathin sections of resin-embedded kidney tissue were cut at 80 nm, mounted on 200 mesh copper grids, treated with uranyl acetate and lead citrate, and examined in a JEOL 1010 transmission electron microscope (Tokyo, Japan).

Disclosures

None.