596 - Allelic heterogeneity in a single cohort of 208 individuals with NPHS2-related steroid-resistant nephrotic syndrome and a novel podocin structure model

Publication Number: 596.3963

Nils D. Mertens, Boston Children's Hospital, Cambridge, MA, United States; Camille Nicolas Frank, Boston Children's Hospital, Boston, MA, United States; Leah Bolsius, Boston Children's Hospital, Cambridge, MA, United States; Shirlee Shril, Boston Children's Hospital, Boston, MA, United States; Florian Buerger, Boston Children's Hospital, Boston, MA, United States; Friedhelm Hildebrandt, Boston Children's Hospital, Boston, MA, United States

Background: Steroid-resistant nephrotic syndrome (SRNS) is the second leading cause of pediatric end-stage kidney disease. The most common molecular cause of SRNS is biallelic variants in the NPHS2 gene, encoding for podocin, a critical component of the glomerular slit diaphragm. The onset of SRNS related to NPHS2 is variable and ranges from infancy to adulthood. Functional studies modeling biallelic NPHS2 variants in cell lines, kidney organoids, and mouse models suggest that allele-specific disease mechanisms lead to podocytopathy.

Objective: 1) Assess the spectrum of disease-causing alleles and allele configurations in 208 individuals with SRNS due to NPHS2 variants
2) Predict and interrogate a novel podocin homo-oligomer structure model
2) Test the hypothesis that genotype-phenotype correlations exist between SRNS onset and a) allele configurations, b) position of affected residues in a novel protein structure model, and c) current in silico pathogenicity prediction scores.

Design/Methods: We re-evaluated genetic sequencing data of our single cohort of 208 individuals with SRNS and biallelic variants in NPHS2 and classified variants according to current ACMG sequence interpretation guidelines. AI-based prediction of podocin homo-oligomer models was performed, and structures were evaluated using PyMOL and pre-defined evidence criteria. Genotype-phenotype correlations were made via basic statics including linear regression models and Kaplan-Meier statistics.

Results: We found a wide spectrum of causative NPHS2 alleles and allele configurations in our single cohort of 208 individuals with SRNS. We detected 75 different pathogenic alleles. Specific genotypes convey a later SRNS onset, including homozygous p.V180M, p.R138Q trans-configured to p.V180M or p.V290M, and pathogenic alleles trans-configured to p.R229Q. We provide robust evidence for podocin to form a circular chalice-like podocin homo-oligomer structure with a hydrophobic pore. Within our novel model, we define new homo-oligomer interaction sides that challenge current models that are used for clinical variant classification. We describe evidence for a structural allelism between the structural location of residues in our model affected by homozygous missense variants and SRNS onset. REVEL and EVE scores for homozygous NPHS2 missense variants could not predict SRNS onset to a clinically relevant extent.

Conclusion(s): Allelic genotype-phenotype correlations for NPHS2-related podocytopathy exist. Novel AI tools may aid in research, as well as the interpretation of rare alleles, and have the potential to detect inter-allelic differences in pathogenicity.

Figure 1. SRNS onset in 208 individuals with biallelic NPHS2 variants ranged from infancy to adulthood.
PAS_figure1.pdfCohort (a) Bar graph showing the range of SRNS onset in 208 individuals with biallelic NPHS2 variants. (b) Pie chart visualizing the frequency distribution of major groups of pathogenic NPHS2 alleles and their respective allele configurations. Homozygous variants (134/208, 65%) are highlighted in shades of blue, and compound heterozygous variants (74/208, 35%) in shades of orange. Predicted loss-of-function (pLoF) variants include frameshift, stop-loss, stop-gain, and splice variants. (c) Kaplan-Meier analysis comparing the SRNS-free survival proportions of all homozygous NPHS2 missense variants detected in ≥4 individuals in our SRNS cohort. (d) Kaplan-Meier analysis comparing the SRNS-free survival proportions of homozygous pLoF variants with homozygous p.R138Q variants. (e) Kaplan-Meier analysis comparing the SRNS-free survival proportions of males and females with homozygous p.R138Q variants. (f) Lower panel showing the secondary protein domain structure of human podocin and upper panel the AlphaFold3-predicted podocin tertiary structure with color-coding corresponding to the domains. Residues R138 and R229 are depicted as spheres and highlighted with black arrows.

Figure 2. SRNS onset in 90 individuals with homozygous NPHS2 missense variants correlates with the position of respective affected amino acid residues
PAS_Figure2.pdf(a and b) Simple linear regression model correlating REVEL and EVE score for homozygous missense variants with SRNS onset, respectively. The REVEL (rare exome variant ensemble learner) score is an ensemble method, using 13 individual tools to predict deleteriousness of all missense variants. The EVE (evolutionary model of variant effect) score is a prediction model for clinical significance of human variants based on fully unsupervised deep learning trained on sequences of 140k species (evemodel.org). (c) Predicted podocin protein structure from Fig 1f. The 20 highlighted amino acid residues represent those affected by 90 homozygous missense variants. Coloring of spheres follows 3-bin grouping of SRNS onset. For variants affecting the same amino acid, the median age of SRNS onset was used. Color coding as follows: Infantile-onset (≤1 year, red), early-onset (1-6 years, blue), and late-onset (>6 years, green). The PHB domain is highlighted with a black box. (d) Predicted podocin structure is color-coded according to mean EVE scores per amino acid residue. Residues affected by homozygous missense variants in individuals with SRNS are highlighted as spheres similar to (c). (e) Scatter plot shows all 5,434 different podocin amino acid substitutions for which EVE scores are available. Each substitution is plotted as a grey asterisk. 36 likely pathogenic or pathogenic amino acid substitutions identified in our cohort are highlighted as red spheres. Homozygous pathogenic substitutions are highlighted as bigger yellow spheres. All alleles identified in individuals within the population control database gnomAD are highlighted as blue spheres, and homozygous gnomAD variants in cyan.

Figure 3. SRNS onset in 61 individuals with compound heterozygous NPHS2 variants trans-configured to p.R229Q or p.R138Q.
PAS_Figure3.pdf(a) Scatter plot correlating SRNS onset with the position of affected amino acid residues of 12 different pathogenic alleles trans-configured to p.R229Q (n=31). (b) Kaplan-Meier analysis comparing the SRNS-free survival proportions of individuals with pathogenic alleles trans-configured to p.R229Q, individuals with homozygous p.R138Q variants, individuals with compound heterozygous missense, and pLoF variants trans-configured to p.R138Q. (c) Stacked frequency distribution of individuals with infantile, early, or late SRNS in individuals with compound heterozygous variants trans-configured to p.R138Q. (d) Scatter plot correlating SRNS onset with the position of affected amino acid residues of 5 different pathogenic alleles trans-configured to p.R138Q (n=12).