Three hundred and seventy-one unrelated individuals diagnosed with FSGS were studied. Of these, 122 (33%) had at least one affected relative, while the remaining 249 (67%) were sporadic cases without a family history. Most of these samples were ascertained after referral by a nephrologist caring for the subject and one or more family members.

Two hundred (54%) of the subjects were Caucasian. The remaining 171 were either African American (n = 68, 18.2%), Hispanic (n = 26, 7%), Asian (n = 6, 1.6%), American Indian (n = 1, 0.2%) or of unreported ethnicity (n = 70, 19%). Families and sporadic cases with radiologic, clinical or histopathologic findings consistent with secondary forms of FSGS were excluded from the analysis. The median age of onset of disease in the subjects studied was 25 years (range 0.25-to-65 years of age). The median age of onset of FSGS in the familial cases was 16 years (range 1.20-to 50 years of age), while that of sporadic patients was 26 years (range 1.25-to-69 years).

We genotyped samples from women enrolled in the Nurses' Health Study I or II 11. The Nurses' Health Study (NHS) samples are large prospective cohorts of female registered nurses. We also genotyped 1988 samples from three distinct study groups, the University of Utah diabetes study, the Genetics of Kidney Disease in Diabetes study (GoKinD), and Australian endocrine diabetics. University of Utah diabetes study: n = 280; 41 type 1 diabetic individuals and 239 with type 2 diabetes. Twenty patients (49%) with type 1 diabetes and 89 patients (37%) with type 2 diabetes had end-stage renal disease. Diabetic patients without end-stage renal disease were used as controls (21/41, 51% of type 1 diabetics; 150/239, 63% of type 2 diabetics). Median age of type 2 diabetic cases was 22 years, and controls with type 2 diabetes was 19 years. The median age of the type 1 diabetic cases and controls was not available). Genetics of Kidney Disease in Diabetes (GoKinD) study group: n = 1,279; 455 "case" samples have end-stage renal disease and 824 who do not have end-stage renal disease were used as "control" subjects. All GoKinD subjects have long-standing (10+ years) of type 1 diabetes. Median age of the cases was 44, and median age of controls was 40 years. Australian endocrine diabetes group: n = 429; 67 type 1 diabetic subjects and 362 with type 2 disease. Fifty-one of the type 1 diabetics had an urinary albumin excretion rate (AER) <20 μg/min, 10 had an AER between 20 μg/min and 200 μg/min, and the remaining 6 had an AER >20 μg/min). Two hundred and twenty-eight of the type 2 diabetics had an AER<20 μg/min, 83 had an AER between 20 μg/min and 200 μg/min, and the remaining 46 had an AER >20 μg/min. The median age of the type 1 microalbuminurics was 55 years, that of macroalbuminurics was also 55 years and normoalbuminurics was 53 years. The median age of the type 2 diabetics with microalbuminuria was 67 years, that of macroalbuminurics was 68 years and normoalbuminurics was 70 years.

Studies were performed in accordance with human subject protocols approved by the human research committees at each of the institutions.

The frequency distributions of alleles was assessed using StatView for Windows, version 5.0 and SAS version 9.0.

We directly sequenced the entire coding region of the NPHS2 gene in PCR amplified DNA from 371 individuals, 122 of whom also had at least 1 relative with FSGS. We identified fifteen alleles that changed the predicted amino acid sequence in 63 patients (or 17% of the samples screened). These were either missense (p.R10T (c.29G>C), p.V127W (c.379G>T), p.R138Q (c.413G>A), p.V180M (c.538G>A), p.R229Q (c.686G>A), p.T232I (c.694C>T), p.E237Q (c.709G>C), p.A242V (c.725 C>T), p.L270F (c.810G>T), p.A284V (c.851C>T), p.L312V (c.934C>G) and p.L327F (c.1048C>T), truncation (p.Q215X (c.643C>T) or frameshift variants (397delA and 855-856delAA) (Table 1). Twelve of these 63 patients had non-synonymous variants in two alleles; 3 were homozygous for p.V127W, p.R138Q or p.V180M, and the remaining 9 were compound heterozygous for p.R138Q and/or p.R229Q and a variety of other alleles (Table 2).

In 3 of the 122 families studied (2.5%), the p.R138Q, p.V127W and p.V180M alleles were demonstrated in a homozygous state (Table 2). In families with multiple affected individuals available for genetic analysis, these alleles segregated in a pattern consistent with autosomal recessive transmission. We identified 9 compound heterozygous events. A majority (7/9 or 78%) of these consisted of p.R229Q, inherited together either with a rarer allele (n = 4), p.R138Q (n = 2) or p.A284V (n = 3). These mutations were identified in 2 index cases with a family history of disease and 7 sporadic cases (Table 2). In the proband's family, the alleles segregated with disease in an autosomal recessive manner. In addition, we identified p.R229Q and p.A242V in homozygous and compound heterozygous states. The p.R229Q and p.A242V homozygous events were identified in two unrelated sporadic patients (CPMC-96 and FG-GC-1112 respectively), and the p.R229Q/p.A242V compound heterozygous state was found in another unrelated sporadic patient (CH-1).

The p.R10T, p.V127W, p.Q215X and p.L270F alleles (present in a compound heterozygous state with either p.R138Q or p.R229Q (Table 2)) were not observed in 362 non-diseased control alleles. Only 1 proband with 2 deleterious NPHS2 alleles other than p.R229Q developed disease in adulthood (Table 2). The 4 other probands with adult onset disease and significant NPHS2 variants in both alleles had one p.R229Q variant. This supports the notion that in the absence of a p.R229Q NPHS2 variant, NPHS2 defects are unlikely to be the cause of disease in an adult with FSGS undergoing genetic analysis.

Fifty-one patients had a single heterozygous NPHS2 allele predicted to alter the encoded protein. These variants (and the respective number of patients they were demonstrated in) were: p.R138Q (n = 3), p.R229Q (n = 32), p. T232I (n = 1), p.E237Q (n = 1), p.A242V (n = 13) and p.L312V (n = 1). Both the p.T232I (c.694C>T) and p.L312V (c.934C>G) alleles were absent in 362 non-diseased control alleles sequenced, while all the others have been demonstrated by previous studies 67812131415161718192021222324. The p.T232I variant was identified in an 8 year-old African-American boy (UAB-023) with no family history who developed FSGS at the age of 2.5 years. The p.L312V mutation was identified in 2 of 4 siblings in a heterozygous state, both of whom had biopsy-confirmed FSGS (Family FG-DK). No second mutant allele was found. Both parents were deceased, the father from kidney failure secondary to prostrate cancer and the mother for unknown reasons.

The p.R229Q and p.A242V variants were the most common non-synonymous variants demonstrated in our FSGS patient cohort (with allele frequencies of 0.02 and 0.06 respectively). The frequency of p.R229Q did not differ between the FSGS races studied, whereas the p.A242V variant was present in a higher proportion of African Americans with FSGS compared to other races (chi-square with Yates correction 9.79, p = 0.00057). We were interested in whether both the p.R229Q and p.A242V variants might contribute to the risk of renal impairment in the general population by investigating whether these variants associate with increased urinary albumin/creatinine ratio, an early marker of renal disease.

We genotyped the p.R229Q variant in a cohort of 2,596 women aged over 50 years who were enrolled in 1976 to participate in the Nurses' Health Study I (of which 97% were of Western european descent). We genotyped the p.A242V variant in 1559 participants in the Nurses' Health Study II (age 44 or greater; 94% of Western european descent). The allele frequencies of the p.R229Q and p.A242V variants in these cohorts were 0.0352 and 0.034, respectively. The genotype frequencies of p.R229Q and p.A242V heterozygotes were 6.5% and 7.0% respectively. Homozygous events were rare for both variants, with genotype frequencies of 0.1% and 0.26% (for p.R229Q and p.A242V respectively). We observed no association between the presence of p.R229Q and p.A242V in either of the homozygous or heterozygous states with increased urinary albumin/creatinine ratio.

We were also interested in investigating whether p.R229Q might represent a modifying allele for renal disease in a more common cause of renal impairment, diabetic nephropathy. The p.R229Q variant has been associated with microalbuminuric events (designated by semi quantitative protocol) in an urban population isolated from Brazil 25 and has shown to cause FSGS in the compound heterozygous state along with rarer non-synonymous variants 6. We genotyped p.R229Q in 1988 diabetic patients from 3 different cohorts and investigated for associations between p.R229Q with end-stage renal disease and proteinuria (Table 3).

We identified 51 patients (51/371, 14%) that had a heterozygous allele that altered the amino acid sequence without any other identified NPHS2 allele. The majority of these were the p.R229Q and p.A242V alleles (Table 1). Two of these alleles (p.T232I (c.694C>T) and p.L312V (c.934C>G)) are private non-conservative alleles that have not been identified in previous NPHS2 resequencing studies. The contribution of these rare alleles to disease is unclear. Neither was observed in any of the control alleles genotyped. The p.T232I allele was demonstrated in a subject with sporadic disease. The p.L312V allele was present in all affected siblings in a family with biopsy-confirmed FSGS. We cannot out rule the possibility of autosomal dominant disease in this family, as the father reportedly died from kidney impairment secondary to prostate cancer and the mother died of unknown causes. Interestingly, the p.T232I variant is predicted to be not tolerated by SIFT and possibly damaging by PolyPhen, and p.L327V is tolerated by SIFT and benign using PolyPhen. Rare heterozygous alleles in NPHS2 may possibly affect protein function, but these will need to be studied in cell culture to verify this.

Rare NPHS2 alleles have also been demonstrated in subjects with thin basement membrane nephropathy (TBMN). Tonna et al. 2003 identified a rare heterozygous variant (p.R224H (c.672G>A)) in only one patient with both TBMN and proteinuria, and like p.T232I the allele was not demonstrated in any control sample 28. Without knowing the frequency with which non-proteinuric control groups carry single, rare, non-synonymous variants, it is difficult to know the clinical significance of such variants.

A small number of studies have attempted to show that heterozygous mutations in both the NPHS2 and NPHS1 genes in a single patient can cause FSGS 7816. It is unclear, however, what the frequency of such digenic events is in non-proteinuric (control) individuals. It is a straightforward hypothesis that digenic or perhaps even trigenic combinations of non-synonymous variants may occur in either of the NPHS2, ACTN4, TRPC6, CD2AP, or even PLCE1 genes in FSGS patients (especially those with sporadic disease). This may be in fact be the case in some of these patients with a single non-synonymous NPHS2 allele (especially if the allele is rare or is known to cause disease in the homozygous or compound heterozygous state), but confirming such a hypothesis will require extensive resequencing of many genes in FSGS cases as well as controls. It is however clear that 2.8% of patients with late onset, non-familial, FSGS have disease attributable to two mutant NPHS2 alleles.

The most common non-synonymous changes we identified were the p.R229Q and p.A242V variants with population frequencies of 0.02 and 0.06, respectively, in the FSGS probands. We found one FSGS patient homozygous for p.R229Q, one homozygous for p.A242V, and one with compound heterozygosity for each of these alleles. Weber et al identified 3 families (2 of which are consanguineous) and 2 sporadic cases of FSGS with homozygosity for p.R229Q, 17. This substitution is one of the most commonly reported NPHS2 alleles 681415202125 with a greater frequency in Europeans (0.036) compared to the frequency in African Americans and Brazilians 61725. In another study, p.R229Q heterozygotes were found to have a 2.77 fold increased risk of developing microalbuminuria compared with controls 25, but it has been unclear whether this allele represents a genetic modifier for renal impairment in other diseases such as diabetic nephropathy.

To clarify whether these common non-synonymous variants cause or contribute to the development of albuminuria, we genotyped both of these alleles in cohorts of women from the Nurses' Health Study I and II. We observed that neither the p.R229Q nor the p.A242V allele were associated with increases in urine albumin/creatinine ratio in either the homozygous or heterozygous states. We found that p.A242V was present in an allele frequency of 0.034 in the Nurses' Health Study II, but in only 0.02 in the FSGS cohort. We note that this frequency in the control group (NHSII) is higher than that previously reported in control groups of similar ethnicity 29. In the FSGS sample set, this variant was present in a higher frequency in persons of African descent compared to other ethnicities (chi-square with Yates correction 9.79, p = 0.00057), consistent with other studies 17). It is not clear how the p.R229Q allele causes FSGS in the compound heterozygote state when inherited together with a second mutant allele 68141516171922, but not in the homozygous state, but this study confirms earlier suspicion that p.R229Q causes disease only in conjunction with a second more detrimental allele 6. An earlier study reported that p.R229Q associates with microalbuminuric events in the population 25. However, in the present study, we saw no association of the p.R229Q allele with albuminuria in either the Nurses' Health Study I nor in the diabetic populations analyzed. Further, the p.R229Q and p.A242V alleles in the homozygous state do not cause FSGS. The results of the genotyping of p.R229Q and p.A242V in the Nurses' Health Study reinforces the importance of evaluating potentially pathogenic variants in large populations.

The biological effect of twelve mutant NPHS2 alleles (p.P118L, p.R138Q, p.R138X, p.D160G, p.R168H, p.R168C, p.R168S, p.V180M, p.R238S, DelLER (aa 237–239), p.V260E, and p.R291W) and two relatively common variants (p.P20L and p.G92C) have been studied in cell culture 303132. All but two of the mutants (p.V180M and p.R238S) fail to reach the plasma membrane in contrast to wild type podocin, whereas both the p.P20L and p.G92C variants do reach the plasma membrane. The lack of proper targeting of mutant NPHS2 to the plasma membrane has been shown to affect nephrin trafficking 303132. Other variants may have less avid binding to nephrin, as has been demonstrated in the case of p.R229Q 6. However, lacking a systematic study of the frequency and biological effects of non-synonymous variants, it remains unclear which altered functions are meaningful markers of clinical pathogenicity.