ARTICLE https://doi.org/10.1038/s41467-020-15383-w OPEN The genetic architecture of membranous nephropathy and its potential to improve non-invasive diagnosis Jingyuan Xie et al.# 1234567890():,; Membranous Nephropathy (MN) is a rare autoimmune cause of kidney failure. Here we report a genome-wide association study (GWAS) for primary MN in 3,782 cases and 9,038 controls of East Asian and European ancestries. We discover two previously unreported loci, NFKB1 (rs230540, OR = 1.25, P = 3.4 × 10−12) and IRF4 (rs9405192, OR = 1.29, P = 1.4 × 10−14), fine- map the PLA2R1 locus (rs17831251, OR = 2.25, P = 4.7 × 10−103) and report ancestry-specific effects of three classical HLA alleles: DRB1*1501 in East Asians (OR = 3.81, P = 2.0 × 10−49), DQA1*0501 in Europeans (OR = 2.88, P = 5.7 × 10−93), and DRB1*0301 in both ethnicities (OR = 3.50, P = 9.2 × 10−23 and OR = 3.39, P = 5.2 × 10−82, respectively). GWAS loci explain 32% of disease risk in East Asians and 25% in Europeans, and correctly re-classify 20–37% of the cases in validation cohorts that are antibody-negative by the serum anti-PLA2R ELISA diagnostic test. Our findings highlight an unusual genetic architecture of MN, with four loci and their interactions accounting for nearly one-third of the disease risk. # A full list of authors and their affiliations appears at the end of the paper. NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w M embranous Nephropathy (MN) is a rare cause of kidney in both ethnicities. We describe a risk haplotype at the PLA2R1 failure, manifesting as nephrotic syndrome with a peak locus that has a regulatory function and exhibits strong genetic incidence between 30 and 50 years of age1. The land- interactions with the HLA-DRB1 risk alleles. Lastly, we calculate a mark discoveries of pathogenic antibodies against neutral endo- genetic risk score (GRS) based on these findings which, when peptidase in antenatal MN2, and anti-phospholipase A2 receptor used in combination with a serum anti-PLA2R ELISA (a serologic (PLA2R) antibodies in adult MN3 have established MN as the test for MN currently in clinical use), shows superior perfor- disease of autoantibodies directed against podocyte antigens. mance in discriminating cases and controls than the ELISA or Several studies have confirmed the presence of autoantibodies GRS alone. We validate the performance of this combined risk against PLA2R in ~60–70% of cases of primary MN4, with score (CRS) in external validation cohorts. Our results demon- another 3–5% potentially explained by antibodies against strate that a combined serum-genetic test can potentially be used thrombospondin type 1 domain-containing 7A5. to establish a new diagnosis of primary MN, obviating the need Previous genome-wide association study (GWAS) for MN for a high risk kidney biopsy procedure in the majority of cases. conducted in 75 French, 146 Dutch and 335 British cases geno- typed with low resolution arrays identified impressively strong associations of the HLA region and the PLA2R1 locus encoding Results the dominant antigen in MN6. These findings suggest that genetic Study design. Our study involved nine case-control cohorts, variation controls the immunogenicity and/or expression level of including four East Asian cohorts of 4841 individuals (1632 the PLA2R auto-antigen, as well as the production of anti-PLA2R primary MN cases and 3209 controls) and five European cohorts autoantibodies in individuals with a permissive HLA haplotype. of 7979 individuals (2150 primary MN cases and 5829 controls). However, specific causal alleles underlying GWAS associations Eight cohorts were genotyped with high density SNP arrays, have not yet been mapped at high resolution. Moreover, prior imputed using the latest whole genome sequence reference panels, GWAS was limited to Europeans, and the reported associations and meta-analyzed genome-wide, and the top 46 loci selected have not been examined comprehensively across different eth- based on P < 5 × 10−5 were tested by targeted genotyping in the nicities. Lastly, because of small sample size, the prior study might ninth cohort (Supplementary Table 1). The summary of study have missed additional disease relevant loci. cohorts, genotyping methods, and ancestry-specific imputation Herein, we report a genetic study of primary MN involving panels is provided in Table 1. 12,820 individuals (3782 biopsy-documented cases and 9038 All cases used in this study were defined by a kidney biopsy ancestry-matched controls), across nine cohorts of East Asian and diagnosis of idiopathic MN and any suspected secondary cases European ancestries. The composition of our cohorts reflects the due to drugs, malignancy, infection, or autoimmune disease were demographics of the centres that have collected DNA samples for excluded. With the exception of the German Chronic Kidney genetic studies of this rare disease over the past 15 years. By using Disease (GCKD) cohort, all controls used for discovery involved high resolution arrays with genome-wide imputation and over 7- healthy population controls and any individuals with a known fold increase in sample size compared to the prior GWAS, we diagnosis of kidney disease were excluded. The GCKD cohort was discover two previously unreported genome-wide significant risk drawn entirely from a prospective observational study of patients loci for MN and perform high resolution mapping and ethnicity- with CKD and consisted of biopsy-defined cases and controls for specific analyses of the known loci. whom CKD etiology was clearly assigned to a non-MN cause, as We describe an unusual genetic architecture of MN, with four previously described7. loci and their genetic interactions accounting for nearly one-third All genome-wide significant loci (P < 5 × 10−8) were refined by of the disease risk. Our study implicates dysregulation of NFKB1 cohort-stratified stepwise conditional analyses to define indepen- and IRF4 genes in the disease pathogenesis, providing genetic dently associated haplotypes. We also analyzed classical HLA support for potential targeting of the NF-κB and interferon sig- alleles and all common amino acid polymorphisms at class I and nalling pathways in primary MN. We also refine ethnicity-specific class II genes imputed at high resolution. We performed detailed effects at the HLA locus, defining DRB1*1501 as a major risk genomic annotations and explored epistatic effects for significant allele in East Asians, DQA1*0501 in Europeans, and DRB1*0301 loci. Based on significant GWAS loci, we designed a GRS for MN Table 1 Baseline characteristics of participants in the discovery and replication cohorts. Cohort Ancestry No. of cases No. of controls Total Genotyping platform Imputation reference population panel Asian Cohorts Chinese Discovery East Asian 561 904 1465 Zhonghua-8 chip (Illumina) 1000G Phase 3 East Asians Korean Discovery East Asian 164 708 872 MEGA chip (Illumina) 1000G Phase 3 East Asians Japanese Discovery East Asian 81 358 439 MEGA chip (Illumina) 1000G Phase 3 East Asians Chinese Replication East Asian 826 1239 2065 KASP (targeted) – All Asian 1632 3209 4841 European Cohorts European Discovery 1 European 611 1246 1857 MEGA chip (Illumina) 1000G Phase 3 Europeans European Discovery 2 European 1045 1094 2139 MEGA chip (Illumina) 1000G Phase 3 Europeans Turkish Discovery European 254 336 590 MEGA chip (Illumina) 1000G Phase 3 Europeans Sardinian Discovery European 93 1498 1591 OmniExpress (Illumina) 1000G Phase 3 Europeans GCKD Discovery European 147 1655a 1802 Omni2.5Exome (Illumina) HRC 1.1 All European 2150 5829 7979 All Participants 3782 9038 12,820 aGCKD participants with chronic kidney disease etiology assigned to a non-MN cause (see Methods). 2 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE and performed its validation in external cohorts, including the Genome-wide association. The results of combined genome- three previously published European GWAS cohorts6 and the wide meta-analyses are summarized in Fig. 1 and Table 2, with Nephrotic Syndrome Study Network (NEPTUNE) study8. more information provided in Supplementary Table 1 and Sup- The descriptions of all cohorts including ancestry analyses and plementary Figs. 3 and 4. We discovered two novel genome-wide details of statistical approaches are provided in Methods, significant loci: a locus on chromosome 4q24 encoding NFKB1 Supplementary Methods, and Supplementary Figs. 1 and 2. (rs230540, OR = 1.25, Meta-analysis P = 3.4 × 10−12) and a locus a 160 HLA 105 PLA2R1 –log10(p -value) IRF4 15 NFKB1 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Chromosome b PLA2R1 locus c NFKB1 locus d IRF4 locus 105 r2 12 r2 16 r2 0.8 rs17831251 0.8 rs230540 0.8 0.6 0.6 14 0.6 rs9405192 90 0.4 10 0.4 0.4 0.2 0.2 0.2 12 75 –log10(p -value) Recombination Rate (cM/Mb) Recombination Rate (cM/Mb) Recombination Rate (cM/Mb) 60 8 60 60 –log10(p -value) –log10(p -value) 10 60 6 8 45 40 40 40 6 4 30 4 15 20 2 20 20 2 0 0 0 0 0 0 2 2 105 r 12 r 16 r2 0.8 0.8 0.8 0.6 0.6 14 0.6 90 0.4 10 0.4 0.4 0.2 0.2 0.2 12 75 –log10(p -value) Recombination Rate (cM/Mb) Recombination Rate (cM/Mb) Recombination Rate (cM/Mb) 60 8 60 60 –log10(p -value) –log10(p-value) 10 60 6 8 45 40 40 40 6 4 30 4 15 20 2 20 20 2 0 0 0 0 0 0 MARCH7 CISD2 DUSP22 NFKB1 IRF4 LY75-CD302 PLA2R1 MIR4785 SLC9B1 HUS1B CD302 ITGB6 RBMS1 SLC39A8 MANBA EXOC2 BAZ2B LY75 UBE2D3 160.6 160.9 161.2 103.2 103.5 103.8 0.2 0.4 0.6 Chromosome 2 position (Mb) Chromosome 4 position (Mb) Chromosome 6 position (Mb) Fig. 1 Manhattan and regional plots for non-HLA loci for the combined meta-analysis of all MN cohorts. a The results of the combined meta-analysis across all cohorts; the dotted horizontal line indicates a genome-wide significance threshold (α = 5 × 10−8); the y-axis is truncated twice to accommodate large peaks over PLA2R1 and HLA loci; genome-wide-significant loci highlighted in red; b Regional plot for the PLA2R1 locus; the upper panel shows unconditioned meta-results, the lower panel depicts meta-results after conditioning for the top SNP (rs17831251). c Regional plot for the NFKB1 locus; the upper panel corresponds to unconditioned results; the lower panel shows meta-results after controlling for rs230540. d Regional plot for the IRF4 locus; the upper panel corresponds to unconditioned results; the lower panel shows meta-results after controlling for rs9405192. The x-axis denotes genomic location (hg19 coordinates), left y-axis represents –log P values for association statistics, right y-axis represents average recombination rates based on HapMap-III reference populations combined (blue line). In the conditional analyses, we conditioned on the top SNP in each individual cohort, then meta- analyzed conditioned summary statistics as described in the Methods. NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w Table 2 Effect estimates for top GWAS SNPs by ethnicity and combined across all cohorts. Locus SNP Risk allele E. Asian E. Asian E. Asian OR E. Asian P- European European European OR European Combined OR Combined case freq. control freq. (95% CI) value case freq. control freq. (95% CI) P-value (95% CI) P-value PLA2R1 rs17831251 C 0.85 0.70 2.81 (2.48–3.17) 3.5 × 10−61 0.76 0.61 1.98 (1.81–2.17) 4.7 × 10−48 2.25 (2.09–2.42) 4.7 × 10−103 NFKB1 rs230540 C 0.43 0.35 1.24 (1.14–1.36) 1.8 × 10−6 0.35 0.32 1.25 (1.14–1.36) 7.8 × 10−7 1.25 (1.17–1.33) 3.4 × 10−12 IRF4 rs9405192 G 0.51 0.42 1.40 (1.28–1.53) 8.8 × 10−14 0.73 0.69 1.18 (1.07–1.29) 6.6 × 10−4 1.29 (1.21–1.37) 1.4 × 10−14 HLA rs9271573 A 0.60 0.35 2.97 (2.69–3.28) 3.7 × 10−102 0.62 0.44 2.06 (1.89–2.25) 1.8 × 10−60 2.41 (2.26–2.57) 2.7 × 10−154 on chromosome 6p25.3 encoding IRF4 (rs9405192, OR = 1.29, (Supplementary Table 6), and their side chains map adjacent to Meta-analysis P = 1.4 × 10−14). We also confirmed strong and each other within the antigen-binding pocket of the β-chain of DR highly significant associations at the previously described loci, (Fig. 3c). including chromosome 2q24.2 encoding PLA2R1 (rs17831251, The top Asian risk allele DRB1*1501 had no significant risk effect OR = 2.25, Meta-analysis P = 4.7 × 10−103) and 6p21.32 encod- in Europeans despite its frequency being comparable between ing HLA-DQA1/DRB1 genes (rs9271573, OR = 2.41, Meta- populations (control freq. 10% vs. 8% in Europeans and East analysis P = 2.7 × 10−154). Asians, respectively). The most strongly associated European risk Conditional analyses of the three non-HLA loci revealed that allele was DQA1*0501 (OR = 2.88, Wald test P = 5.7 × 10−93, each signal is explained by a single SNP in each cohort, suggesting Fig. 2d). After conditioning the locus on DQA1*0501, DRB1*0301 a single shared risk haplotype per locus in East Asian and remained genome-wide significant (ORconditioned = 2.00, Wald test European populations (Fig. 1b–d). To further test if the causal P = 2.0 × 10−19, Supplementary Table 7) suggesting that this risk variants at these loci are likely shared between Europeans and allele is shared between Asian and European populations. We note East Asians, we performed 99% credible set analyses using that DQA1*0501 allele is twice as common in Europeans compared summary statistics for each ancestry-defined subgroup and to Asians (control freq. 30% vs. 14%). Moreover, DQA1*0501 compared them with credible sets derived from the trans-ethnic and DRB1*0301 are in imperfect LD that is stronger in Europeans meta-analysis. We confirmed that the predicted causal variants (r2 = 0.40) compared to East Asians (r2 = 0.29). Although a weak derived from the trans-ethnic analysis were largely overlapping effect of DQA1*0501 was apparent in our Asian cohorts (Fig. 2c, with ancestry-specific results (Supplementary Fig. 5). In contrast, Supplementary Table 3), this allele became non-significant after stepwise conditional analyses of SNPs at the HLA region revealed conditioning on DRB1*0301. In contrast, DQA1*0501 exhibited a a complex pattern of association, with at least three indepen- genome-wide significant risk effect after conditioning on dently genome-wide significant SNPs explaining the signal across DRB1*0301 in Europeans (ORconditioned = 2.40, Wald test P = all cohorts (Supplementary Table 2). 1.8 × 10−18). Given the complexity of the association signal at the HLA locus Given that our HLA imputation reference panels were and known differences in linkage disequilibrium (LD) patterns by considerably smaller for East Asians compared to Europeans, ancestry, we performed additional analyses of this region we sought additional validation of the observed classical HLA separately in East Asians and Europeans. In the conditional associations that were Asian-specific. We therefore created analyses of the East Asian cohorts, only two independently another reference panel based on the MHC sequence data from associated SNPs explained the entire signal at this locus Zhou et al.10 including 10,689 control individuals of Han Chinese (rs9269027 and rs1974461). In Europeans, stepwise conditional ancestry. Using SNP2HLA software, we then re-imputed classical analyses revealed three independently associated genome-wide HLA alleles for our East Asian cohorts. We used the same quality significant SNPs (rs9271541, rs9265949, and rs2858309), suggest- control filters (MAF > 0.01 and imputation R2>0.8) and methods ing a more complex pattern of association (Supplementary for association testing as described above. We observed no major Table 2). In both ethnicities, the top signal centred on HLA-DRB1 differences in the association statistics for the two Asian risk and DQA1 genes (Fig. 2a, b). alleles, DRB1*1501 (OR = 3.49, P = 3.85e−40) and DRB1*0301 (OR = 4.08, P = 6.3E−24), demonstrating that these effects do Classical HLA alleles and amino acid polymorphisms. We next not represent artifacts of smaller imputation panels. imputed classical HLA alleles at two- and four-digit resolution We next performed the analysis of HLA amino acid using ethnicity-specific reference panels (see Methods). The first substitutions in Europeans. Consistent with the association two digits specify a group of HLA alleles known as super-types as analyses of classical alleles, five bi-allelic sites in DQA1 that defined by older typing methodologies. The third through fourth correlate with the DQA1*0501 allele were most strongly digits specify nonsynonymous substitutions. Moreover, we associated with the risk of MN (75Ser-107Ile-156Leu-161Glu- imputed individual amino acid polymorphisms at class I (HLA-A, 163Ser, Wald test P = 5.7 × 10−93, Fig. 3b, Supplementary -B, and -C) and class II (HLA-DQB1, -DQA1 and -DRB1) genes. Tables 8–10). Conditioning on this haplotype in Europeans In East Asian cohorts, stepwise conditioning on classical HLA uncovered a second independent signal in HLA-DRB1, position alleles defined two independent risk alleles, DRB1*1501 (OR = 74 (Supplementary Tables 8 and 9), with Arginine representing 3.81, Wald test P = 2.0 × 10−49) and DRB1*0301 (ORconditioned = the key risk residue (OR 2.86, 95% CI: 2.54–3.23). This residue 3.88, Wald test P = 4.5 × 10−24, Fig. 2c, Supplementary Table 3). defines the European DRB1*0301 risk haplotype (Supplementary In the analysis of polymorphic amino acid sites, genetic variation Tables 9 and 10). Notably, positions 74 (Europeans) and 71 (East at only two codons encoding residues at positions 13 and 71 in Asians) are separated by a single turn along the α-helix, and their DRβ, explained the entire HLA-DRB1 signal (Fig. 3a, Supplemen- side chains are spatially close to that of position 13, located on the tary Table 4). Specifically, DRβ position 13 occupied by Arginine beta-sheet floor with its side chain oriented into the peptide- (OR = 3.68, 95% CI: 2.74–4.95) or Serine (OR = 2.76, 95% CI: binding groove (Fig. 3c). 2.06–3.71), and position 71 occupied by Lysine (OR = 3.10, 95% To confirm ethnicity-specific HLA effects, we repeated stepwise CI: 2.49–3.86) or Alanine (OR = 2.96, 95% CI: 2.55–3.45) conditioning in a joint stratified analysis of all cohorts using bi- conveyed the greatest risk (Supplementary Table 5). Consistent allelic tests of HLA alleles with formal tests of heterogeneity with a prior study in Chinese patients9, these amino acids define (Supplementary Table 11, Fig. 4a). The top classical allele the classical risk alleles DRB1*1501 and DRB1*0301 supported by all cohorts regardless of ethnicity was DRB1*0301 4 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE a 100 East Asian b 100 European HLA-DQA1 HLA-DRB1 –log10(p-value) HLA-DRB1 80 80 –log10(p-value) 60 60 40 HLA-DQA1 40 20 0 20 30 Control for HLA-DQA1 0 20 –log10(p-value) HLA-DRB1 20 10 –log10(p-value) Control for HLA-DRB1 0 10 10 Control for HLA- DQA1 & DRB1 0 0 29 30 31 32 33 34 30 31 32 33 Chromosome 6 position (Mb) Chromosome 6 position (Mb) c 0.3 Case OR = 3.81 0.4 OR = 0.43 d 0.3 Case 0.6 OR = 2.88 Control Control P = 5.7E-93 P = 2.0E-49 P = 1.3E-27 OR = 2.14 OR = 3.39 P = 1.3E-24 P = 5.2E-82 P = 9.2E-23 0.3 OR = 1.83 OR = 3.50 0.2 P = 6.8E-14 0.2 0.4 Frequency Frequency OR = 0.38 P = 6.6E-18 OR = 0.38 P = 2.3E-9 OR = 0.57 P = 6.6E-18 0.2 OR = 0.63 P = 1.3E-13 OR = 0.50 0.1 0.1 0.2 P = 2.3E-18 0.1 0.0 0.0 0.0 0.0 *0 :01 *0 :01 *0 :01 *0 :01 *0 01 01 *0 1 *0 1 *1 1 *1 1 *1 01 01 *0 01 *0 1 *0 1 *0 1 01 *0 :01 *0 :01 *1 :01 *1 01 *1 :01 *0 03 *1 2 *0 2 03 03 *1 :02 *1 02 02 *0 03 *0 02 *0 05 0 0 0 0 0 0 0 0 0 5: 6: 1: 3: 7: 2: 3: 5: 2: 1: 2: 3: 4: 5: 1: 3: 3: 5: 3: 1: 1: 8: 1: 4: 1 2 3 4 3 7 9 5 2 *0 *0 *1 *0 *0 *0 Classical HLA-DRB1 allele Classical HLA-DQA1 allele Classical HLA-DRB1 allele Classical HLA-DQA1 allele Fig. 2 Ethnicity-specific association analyses point to HLA-DRB1 and HLA-DQA1 as the top associated genes. a Regional plot for East Asian cohorts that includes association statistics for all imputed variants and classical HLA alleles; the strongest association signal was for HLA-DRB1 gene (upper panel); the results for HLA-DQA1 gene are highlighted for reference; after adjusting for all DRB1 classical alleles (red arrow), there were no residual associations across the entire 3-Mb region (lower panel). The dotted horizontal line indicates the genome-wide significance threshold (α = 5 × 10−8). b In Europeans, the strongest HLA gene association was for the HLA-DQA1 gene (upper panel); after controlling for all classical DQA1 alleles (red arrow), HLA-DRB1 gene remained genome-wide significant (middle panel); after controlling for both DRB1 and DQA1, there were no significant associations in the region (lower panel), suggesting that variation in both genes explains the entire signal. c East Asian frequency distributions of classical DRB1 and DQA1 alleles (four-digit resolution) for cases (red) and controls (black); unadjusted ORs and P-values provided for genome-wide significant alleles. d The European frequency distributions of classical DRB1 and DQA1 alleles (four-digit resolution) for cases (red) and controls (black); unadjusted ORs and P-values provided for genome-wide significant alleles. (OR = 3.71, Wald test P = 2.9 × 10−127). After conditioning for We next refined the previously reported genetic interactions DRB1*0301, the top classical allele was DQA1*0501 (ORconditioned = between the PLA2R1 locus and HLA risk haplotypes. The 1.80, Wald test P = 1.1 × 10−30), but this association was supported PLA2R1 risk genotype exhibited significant multiplicative predominantly by Europeans. After controlling for both interaction with both Asian and European HLA risk haplotypes DRB1*0301 and DQA1*0501, the top allele was DRB1*1501 (Fig. 4b, c), with the risk homozygosity at both loci associated (ORconditioned = 1.94, Wald test P = 4.7 × 10−29), but the risk effect with 89-fold increased odds of disease risk in East Asians [OR = was supported exclusively by East Asians (heterogeneity I2 = 97.5, 88.8 for double risk homozygotes (N cases/controls = 103/10) Cochrane’s Q-test P < 0.05). vs. double protective homozygotes (N cases/controls = 15/152), 95% CI: 38.0–207.3, Interaction test P = 7.8 × 10−3] and 14-fold in Europeans [OR = 14.1 for double risk homozygotes (N cases/ PLA2R1 locus and its genetic interactions. Consistent with prior controls = 291/89) vs. double protective homozygotes (N cases/ GWAS, the most significant non-HLA locus resided on chro- controls = 52/237), 95% CI: 10.0–22.1, Interaction test P = mosome 2q24.26. The top SNP was in the first intron of PLA2R1, 6.4 × 10−5]. Because the effect modification was weaker in which encodes the main podocyte autoantigen in primary MN. Europeans, we next repeated interaction testing separating This signal was supported by both ethnicities, but the effect individual HLA risk haplotypes (Fig. 4d, e). Notably, the appeared stronger in East Asians (OR = 2.81, Meta-analysis P = interaction in Europeans was driven predominantly by the 3.5 × 10−61) compared to Europeans (OR = 1.98, Meta-analysis DRB1*0301-DQA1*0501 haplotype [OR = 28.7 for double risk P = 4.7 × 10−48, Table 2). After conditioning the association on homozygotes (N cases/controls = 115/13) vs. double protective the top SNP, rs17831251, there was no residual association at this homozygotes (N cases/controls = 75/339), 95% CI: 15.1–54.4, locus, suggesting a common risk haplotype in both ethnicities Interaction test P = 2.2 × 10−3]. After removing its effect, (Fig. 1b). DQA1*0501 had no residual interaction with PLA2R1 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w a East Asian b European 80 Unconditional Conditional 1 Conditional 2 120 Unconditional Conditional 1 Conditional 2 75 107 163 13 100 60 –log10(p -value) –log10(p -value) 156 161 80 40 Control for DRB1 Control for DRB1 60 Control for DQA1 Control for DQA1 pos.13 pos.13 & pos.71 pos.75 pos.75 & DRB1 pos.74 40 20 71 20 74 0 0 0 200 0 200 0 200 0 200 0 200 0 200 0 200 0 200 0 200 0 200 0 200 0 200 DRB1 DQA1 DRB1 DQA1 DRB1 DQA1 DRB1 DQA1 DRB1 DQA1 DRB1 DQA1 Amino acid position Amino acid position c HLA-DR d HLA-DQ 161 DRα 161 163 DRα 156 163 107 156 75 DQα 75 DQα 13 13 71 107 90° 90° 71 74 DRβ DRβ 74 DQβ DQβ Fig. 3 Ethnicity-specific association analyses of DRβ1 and DQα1 amino acid sequence. a East Asian analysis of polymorphic amino acid positions within DRβ1 (blue) and DQα1 (green) molecules using conditional haplotype tests; the horizontal dash line marks the genome-wide significance level. The most strongly associated polymorphic site was position 13 in DRB1 (left panel); after controlling for this position, position 71 in DRB1 remained genome-wide significant (middle panel); after adjusting for both positions, there were no residual associations (right panel). b European analysis of polymorphic amino acid positions within DRβ1 (blue) and DQα1 (green); the DQA1 haplotype defined by amino acid positions 75, 107, 156, 161 and 163 (left panel) provided the strongest signal; after conditioning on this haplotype, position 74 in DRB1 remained genome-wide significant (middle panel); no additional independent positions were found upon further conditioning (right panel). c Protein structure of the DR molecule including α chain (pink ribbon) and β chain (blue ribbon), the side chains of amino acids at DRβ1 positions 13, 71, and 74 are located adjacent to each other in the P4 pocket of the peptide-binding groove. d Protein structure of the DQ molecule including α chain (yellow ribbon) and β chain (green ribbon); all five bi-allelic amino acid sites in DQα1 are in perfect LD with each other and define the top associated DQA1*0501 allele in Europeans. (Interaction test P = 0.1). Similarly, there was no significant genomic location of rs17831251 and rs17241973 in relationship interaction between DQA1*0501 allele and PLA2R1 locus in East to the recently published kidney compartment-specific chromatin Asians. These analyses suggest that the PLA2R1 locus interac- landscape14 (Supplementary Fig. 9); rs17241973 lies within intron tions are driven primarily by the DRB1 alleles. 1 of PLA2R1 in open chromatin that is active in both glomerular We annotated all SNPs in LD with rs17831251 for potential and tubular compartments and is contiguous with the gene impact on the structure and/or transcriptional regulation of promoter. In contrast, rs17831251 lies within a broad region of PLA2R1. We found two common missense variants in moderate increased chromatin accessibility in glomeruli, and is only 2.1-kb LD, rs35771982 (p.H300D, r2 = 0.69) and rs3749117 (p.M292V, away from a glomerulus-specific DHS that contains a high- r2 = 0.68), but the effects of these variants were considerably confidence NFKB1-binding motif. In glomerulus-specific chro- weaker compared to rs17831251, suggesting that they are unlikely matin conformation (Hi-C) data, both rs17831251 and to represent causal variants (Supplementary Table 12). Our rs17241973 make regional and distal contacts with other tissue-specific functional scoring method for non-coding variants glomerular DHS, emphasizing a composite cis-regulatory module based on the ENCODE and Roadmap Epigenetics data11 for PLA2R1 gene expression. prioritized another variant in intron 1, rs17241973 (r2 = 0.93 with rs17831251) that intersects a putative enhancer element across multiple tissues (Supplementary Fig. 6). Both rs17831251 Novel loci encoding NFKB1 and IRF4. The 4q24 locus contains and rs17241973 exhibit strong cis-eQTL effects on PLA2R1 the NFKB1 gene, which encodes an active DNA binding subunit expression wherein the MN risk alleles associate with lower of the NF-κB transcriptional complex. The top SNP, rs230540 mRNA expression of PLA2R1 across multiple tissues in GTEx12, (OR = 1.25, Meta-analysis P = 3.4 × 10−12), is an intronic variant but this effect appears reversed for the kidney tissue (Supple- predicted to have a functional effect specific to immune cells mentary Fig. 7). To further confirm these kidney-specific effects, (Supplementary Fig. 10). In agreement with our prediction, we used gene expression data from manually micro-dissected rs230540 has been associated with higher mRNA expression of human kidney compartments of 166 NEPTUNE participants13. NFKB1 in whole blood (P = 2.6 × 10−11)15 and in CD4+ T cells We detected suggestive glomerular eQTL effects that were weak, (P = 2.0 × 10−9)16. Consistent with pro-inflammatory effects of but direction-consistent with GTEx for rs17831251 (Wald test NF-κB, the MN risk haplotype at this locus determined higher P = 0.055) and rs17241973 (Wald test P = 0.024), wherein MN leukocyte counts17 and increased risk of ulcerative colitis18,19 and risk allele were associated with increased glomerular PLA2R1 primary biliary cholangitis20,21 (Supplementary Table 13, Fig. 5a). mRNA levels (Supplementary Fig. 8). NFKB1 is also expressed in human podocytes22, as well as pri- Because kidney tissue compartments are not well represented mary human glomerular and tubular epithelial cell cultures14, and in either ENCODE or Roadmap datasets, we next examined the rs230540 intersects an active glomerular DHS with several 6 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE a DRB1*03:01 Odds ratio Lower.limit Upper.limit P value East Asian 3.50 2.72 4.49 9.2E-23 European 3.39 3.01 3.81 5.2E-82 All combined 3.71 3.34 4.13 2.9E-127 Unconditional analysis 0.7 1.0 1.4 2.0 6.0 Odds Ratio DQA1*05:01 Odds ratio Lower.limit Upper.limit P value East Asian 1.29 1.07 1.56 8.9E-03 European 2.43 2.00 2.97 1.8E-18 All combined 1.80 1.63 1.99 1.1E-30 Controlled for DRB1*0301 0.7 1.0 1.4 2.0 6.0 Odds Ratio DRB1*15:01 Odds ratio Lower.limit Upper.limit P value East Asian 4.21 3.50 5.06 4.1E-53 European 1.03 0.88 1.21 0.7 All combined 1.94 1.73 2.18 4.7E-29 Controlled for DRB1*03:01 and DQA1*05:01 0.7 1.0 1.4 2.0 6.0 Odds Ratio b Asian 88.8 c European 14.1 15 p-value = 7.8E-3 p-value = 6.4E-5 80 10 Odds Ratio 60 Odds Ratio 40 5 20 0 0 GG GG TG RH+/+ TG PL RH+/+ PL A2 A2 RH+/– RH+/– R1 R1 TT TT RH–/– HLA RH–/– HLA d European 28.7 e European 30 6 5.5 p-value = 2.2E-3 p-value = 0.09 20 Odds Ratio Odds Ratio 4 10 2 0 0 GG GG TG TG PL RH+/+ RH+/+ PL A2 A2 1 RH+/– 501 RH+/– 050 A1* R1 R1 TT *0 TT Q RH–/– QA1 RH–/– ther - D 1-D 1*O aplotyp e * 030 lotype 1 DRB Hap DRB H Fig. 4 Ethnicity-specific HLA allelic effects and genetic interactions. a Stepwise conditioning of HLA risk alleles by ethnicity; Top: Unconditioned effect estimates, 95% confidence intervals, and P-values for DRB1*0301 demonstrate similar effects in East Asian and European cohorts. Middle: Effect estimates for DQA1*0501 after controlling for DRB1*0301 demonstrate significant heterogeneity, with stronger effects in the European cohorts (Cochrane’s Q P-value < 0.05). Bottom: Effect estimates for DRB1*1501 after controlling for both DRB1*0301 and DQA1*0501 demonstrate significant heterogeneity with risk effect in East Asians but no effect in Europeans (Cochrane’s Q P-value < 0.05). b PLA2R1 genotype interaction with East Asian HLA risk haplotypes, DRB1*0301 or DRB1*1501 (N cases/controls = 803/1,956, P = 7.8 × 10−3). c PLA2R1 genotype interaction with European HLA risk haplotypes, DRB1*0301 or DQA1*0501 (N cases/controls = 1880/2627, multiplicative interaction test P = 6.4 × 10−5). d PLA2R1 genotype interaction with DRB1*0301-DQA1*0501 risk haplotype in Europeans (multiplicative interaction test P = 2.2 × 10−3). e No significant interaction between PLA2R1 and DQA1 risk haplotypes other than DQA1*0501- DRB1*0301 in Europeans (multiplicative interaction test P = 0.09). RH risk haplotype. NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w a Hodgkin's lymphoma Reticulocytes C3 & C4 level Phenotype sharing single locus with MN Eosinophils+basophils Systemic sclerosis Phenotype sharing multiple loci with MN Eosinophils MN GWAS loci Nasopharyngeal ca Concordant SNP association Waist circumference Frontotemporal dementia Hematocrit Opposing SNP association IgA level Association with unclear direction of effect IBD HLA 1st RBC count PLA2R1 EBV response HLA 2nd CLL Autoimmune hepatitis IRF4 MN MS Cutanous lupus GFR SLE T1D Celiac disease Allergic disease NFKB1 HLA 3rd Tuberculosis Tonsillectomy Autism Shingles Lymphocytes Monocytes Itch from mosquito bite UC HBV PBC Drug-induced liver injury Vogt-Koyanagi-Harada syndrome Ear infections HCV cirrhosis b 0.75 c 1.5 P value = 9.0E-8 P value = 1.3E-3 0.6 1.2 Log(24-hour proteinuria level) Fraction α-PLA2R positive 0.45 0.9 0.3 0.6 0.15 0.3 0.0 0.0 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 Genetic risk score (decile) Genetic risk score (decile) Fig. 5 Pleiotropic effects of the MN loci and their clinical correlations. a The pleiotropy map was constructed based on overlapping genome-wide significant loci reported in the GWAS Catalogue: traits sharing a single locus with MN are indicated in yellow; traits sharing multiple loci are indicated in orange; arrows represent allelic associations that are identical to, or in tight LD (r2 > 0.8) with the MN risk alleles; arrow thickness is proportional to r2 between alleles; concordant effects are indicated in red and opposed effects in blue. IBD: inflammatory bowel disease, includes ulcerative colitis (UC) and Crohn’s disease (CD); PBC primary biliary sclerosis, GFR glomerular filtration rate, CLL chronic lymphocytic leukemia, HBV hepatitis B virus; b Significant positive correlation of the GRS with anti-PLA2R antibody seropositivity (N = 1114 cases, Wald test P = 9.0 × 10−8) and c log-transformed 24-h proteinuria at diagnosis (N = 1329 cases, Slope test P = 1.3 × 10−3). The x-axis depicts deciles of GRS; error bars correspond to standard errors; the P-values are adjusted for age, sex, and ethnicity. glomerular Hi-C contact sites14 (Supplementary Fig. 11). Notably, be under the transcriptional control of the NF-kB complex25–27. the MN risk allele has previously been associated with lower Unlike PLA2R1 and NFKB1, IRF4 does not appear to be estimated glomerular filtration rate in GWAS of renal expressed in human kidney cells by single nuclei RNA-seq28 function23,24, thus this locus may be more broadly associated with (Supplementary Fig. 12). We did not find functional or coding the risk of kidney disease. SNPs in LD with rs9405192, nor did we observe any cis-eQTL The top SNP on chromosome 6p25.3, rs9405192 (OR = 1.29, effects for this variant, thus the precise mechanism underlying Meta-analysis P = 1.4 × 10−14), resides upstream of IRF4 gene, this association remains unknown. However, the analysis of which belongs to the family of transcription factors regulating binding sites for individual components of the NF-κB complex in interferon-inducible genes. IRF4 is lymphocyte specific and lymphocytes29 suggested binding of the complex in close negatively regulates Toll-like-receptor signalling that is central proximity of rs9405192 (Supplementary Fig. 13). The risk allele to the activation of innate immune system; this gene is known to at this locus is also in strong LD with variants previously 8 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE associated with increased risk of inflammatory bowel disease19, cut-off, levels 2–20 U/mL are frequently considered as borderline- and in weaker LD with several risk variants for chronic negative, and levels <2 U/mL as negative31. In our cohorts, the lymphocytic leukemia (Supplementary Table 13, Fig. 5a), suggest- cut-off 2 U/mL had inadequate diagnostic specificity (range ing the pattern of pleiotropy that is similar to the NFKB1 locus. 73–92%). These results confirm the key limitation of the PLA2R Nevertheless, we detected no statistically significant genetic antibody ELISA, which has high specificity (99–100%) but low interactions of IRF4 and NFKB1 loci. sensitivity (51–60%) at the standard recommended cut-off point; In addition, we systematically annotated all other suggestive while lowering the cut-off increases sensitivity, it results in non-HLA loci defined by P < 5.0 × 10−5 and these results are inadequate specificity. Consequently, the levels in the borderline- summarized in Supplementary Table 14. To enhance potential negative range (2–20 U/mL) are difficult to interpret clinically. genetic discovery of novel podocyte antigens, we also repeated Given this limitation, we evaluated if the addition of genetic risk genome scans after conditioning for the PLA2R1 locus, but information can improve the performance of ELISA, especially in detected no additional suggestive loci. cases that fall in the borderline-negative range. First, we evaluated diagnostic properties of the GRS alone in our discovery cohorts. In SNP-based heritability and risk explained by GWAS. Using our East Asians, the genetic test had area under the receiver operating genotype data and genome-based restricted maximum likelihood characteristics curve (AUROC) of 0.80 (95% CI: 0.78–0.82), while method (GREML)30, we estimated the overall SNP-based herit- in Europeans the AUROC was 0.75 (95% CI: 0.74–0.77). ability of MN at 0.43 (SE = 0.039) in East Asians and 0.36 (SE = Combining genetic and serologic tests in the form of a CRS 0.0046) in Europeans. Remarkably, all genome-wide significant provided superior case discrimination with AUROCs of 0.96 (95% risk alleles exhibited unusually large effect sizes for GWAS. In CI: 0.95–0.98) in East Asians and 0.89 (95% CI: 0.87–0.91) in order to quantify the fraction of disease variance cumulatively Europeans (Fig. 6, Supplementary Table 18). explained by genome-wide significant SNPs and their interac- We next tested the GRS performance in several external tions, we performed ethnicity-specific GRS analyses (see Meth- validation cohorts, including three independent GWAS cohorts of ods). Each GRS was expressed as a weighted sum of risk alleles European ancestry as well as in the European-American with weights defined by their mutually adjusted effect estimates NEPTUNE participants with incident nephrotic syndrome and included the 3 independent non-HLA SNPs (rs6707458, (Supplementary Table 18C-E). Overall, the effects and the rs230540, rs9405192) as well as ethnicity-specific HLA risk alleles diagnostic performance of the GRS were comparable between and their interactions. This included rs9269027, rs1974461, and the European discovery and each validation cohort (Fig. 6c). rs9269027*rs6707458 interaction term for East Asians, and When combined with the serum antibody titer, the CRS achieved rs9271541, rs9265949, rs2858309 and rs9271541*rs6707458 AUROC of 0.96 (95% CI: 0.94–0.97) across all validation cohorts interaction term for Europeans (Supplementary Table 15). The combined (Supplementary Table 18E, Fig. 6f). GRS calculated using this method explained 32% disease risk in In the subgroup analyses, we compared the diagnostic East Asians, 25% in Europeans, and 29% of overall disease risk properties of GRS and CRS by the antibody status in all cohorts across all cohorts combined. Remarkably, the magnitude of the pooled by ancestry (Fig. 7). These analyses demonstrated that GRS effect was comparable to rare, highly penetrant mutations both GRS and CRS were predictive of case status even for causing Mendelian forms of kidney disease, with individuals in antibody-negative MN. Importantly, the CRS continued to have the top decile of GRS having 30 to 40-fold higher disease risk excellent performance in classifying the borderline-negative cases compared to the lowest decile (Fig. 6a, b). (antibody level 2–20 U/mL range), with AUROCs of 0.98 (95% CI: 0.97–0.99) in East Asians and 0.95 (95% CI: 0.93–0.96) in Europeans. Notably, among all cases for which the ELISA test was Clinical correlations of the GRS. For a subset of patients with either negative or inconclusive, adding genetic information in the available clinical data, we performed genetic correlation analyses form of CRS can establish the diagnosis in 20–37% cases with with selected clinical features reflective of disease severity. The 99% specificity. The comparison of AUROCs between GRS, CRS, GRS was positively correlated with PLA2R antibody seroposi- and serum anti-PLA2R Ab test by ancestry is provided in tivity (Wald test P = 9.0 × 10−8), and in those with detectable Supplementary Fig. 14, and the clinical implications of these antibodies, higher titers at the time of biopsy (Slope test P = 1.2 × findings are summarized in Supplementary Note 1 and 10−9, Fig. 5b). The GRS also predicted worse proteinuria at the Supplementary Table 19. time of biopsy, which represents the key marker of MN severity Lastly, we expanded our validation studies to non-European and prognosis (Slope test P = 1.3 × 10−3, Fig. 5c). Other clinical participants of the NEPTUNE study (Supplementary Tables 20 features, such as age at diagnosis, renal function, or serum and 21). Although the European risk score performance was albumin levels at the time of biopsy were not significantly cor- diminished in Hispanic Americans, the European GRS performed related with the GRS after multivariate adjustment (Supplemen- well in African Americans, and this is despite substantial tary Table 16). differences in risk allele frequencies between Europeans and Africans (Supplementary Table 22). Similar to the European Potential diagnostic implications of the GRS. Although the validation cohorts, the European GRS was superior compared to diagnosis of MN is traditionally established by a kidney biopsy, the trans-ethnic GRS when applied to the NEPTUNE minority the detection of circulating PLA2R antibodies by ELISA has populations, while the Asian GRS had relatively poor perfor- recently emerged as a useful diagnostic modality31. In this study, mance in both African-American and Hispanic/Latino cohorts we performed ELISA in sera obtained within 6 months of a (Supplementary Table 21, Supplementary Fig. 15). diagnostic kidney biopsy in a total of 2331 individuals (1488 cases, 300 healthy controls, and 543 disease controls). In East Asians, we estimated that the standard ELISA cut-off of 20 U/mL Discussion provided 100% specificity and 60% sensitivity for the diagnosis of Our study provides important insights into an autoimmune dis- MN. In the analysis of Europeans, depending on the specific ease and the genetic architecture of MN. First, we discover novel cohort, the same cut-off provided 99–100% specificity and genome-wide significant risk loci for MN with large effects 51–57% sensitivity (Supplementary Table 17). While the antibody encoding two transcriptional master regulators of inflammation, level of 20 U/mL represents the manufacturer’s recommended NFKB1 and IRF4. The association at the NFKB1 locus highlights NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w a Asian Discovery b European Discovery c External Validation 70 40 60 60 35 50 30 50 40 25 Odds ratio Odds ratio Odds ratio 40 20 30 30 15 20 20 10 10 10 5 0 0 0 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 Genetic risk score (decile) Genetic risk score (decile) Genetic risk score (decile) d 1.0 e 1.0 f 1.0 0.8 0.8 0.8 Sensitivity Sensitivity Sensitivity 0.6 0.6 0.6 0.4 0.4 0.4 0.2 0.2 0.2 CRS AUROC = 0.96 (0.95–0.98) CRS AUROC = 0.89 (0.87–0.91) CRS AUROC = 0.96 (0.94–0.97) GRS AUROC = 0.80 (0.78–0.82) GRS AUROC = 0.75 (0.74–0.77) GRS AUROC = 0.77 (0.75–0.79) 0.0 0.0 0.0 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 Specificity Specificity Specificity g h i 8 6 Spec = 0.99, Sens = 0.74 8 Spec = 0.99, Sens = 0.53 Spec = 0.99, Sens = 0.65 Spec = 0.97, Sens = 0.82 Spec = 0.97, Sens = 0.61 Spec = 0.95, Sens = 0.83 6 6 4 Z-transformed CRS Z-transformed CRS Z-transformed CRS 4 4 2 2 2 0 0 0 –2 –2 –2 Control FSGS IgAN MN Control FSGS IgAN MN Control FSGS/MCD IgAN MN (N = 56) (N = 52) (N = 47) (N = 304) (N = 99) (N = 37) (N = 88) (N = 810) (N = 145) (N = 97) (N = 14) (N = 284) Fig. 6 Diagnostic performance of the genetic risk score (GRS) and the combined risk score (CRS). Genetic effects expressed as odds ratios (OR) and 95% confidence intervals in reference to the lowest decile of the GRS distribution for a East Asian discovery, b European discovery, and c European validation cohorts combined. GRS and CRS Receiver Operating Characteristics (ROC) curves for d East Asian discovery, e European discovery, and f European validation cohorts combined; AUROC Area under the ROC curve. Distributions of the CRS between healthy controls, diseased controls, and MN cases for g East Asian discovery, h European discovery, and i European validation cohorts combined. The box plots depict medians (horizontal lines), interquartile ranges (boxes), and minimum/maximum values (whiskers). The discovery CRS cut-offs of 1.00 and 1.45 in East Asians and 2.05 and 2.72 in Europeans have 97% and 99% specificity, respectively. The same European cut-offs were applied to the validation cohorts for comparisons of specificity and sensitivity. the role of the canonical NF-κB pathway in primary MN. Upon and in the context of inflammatory diseases, including IBD33,34 activation by pro-inflammatory signals, NFKB1 undergoes pro- and MN35–37. Importantly, the MN risk allele at this locus has a teasome processing to p50, an active DNA binding subunit of the concordant effect on the risk of ulcerative colitis18,19 and primary NF-κB complex. Inappropriate activation of this pathway has biliary cirrhosis20,21. It has also been associated with increased previously been studied in progressive diabetic nephropathy32, mRNA expression of NFKB1 in cis, and reduced DNA 10 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE a GRS in East Asians b GRS in Europeans 1.0 1.0 0.8 0.8 0.6 0.6 Sensitivity Sensitivity 0.4 0.4 0.2 0.2 Ab >20 U/mL AUROC = 0.91 (0.88–0.94) Ab >20 U/mL AUROC = 0.80 (0.78–0.83) Ab 2–20 U/mL AUROC = 0.87 (0.82–0.92) Ab 2–20 U/mL AUROC = 0.74 (0.69–0.79) 0.0 Ab <2.0 U/mL AUROC = 0.78 (0.71–0.85) 0.0 Ab <2.0 U/mL AUROC = 0.69 (0.66–0.73) 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 Specificity Specificity c CRS in East Asians d CRS in Europeans 1.0 1.0 0.8 0.8 0.6 0.6 Sensitivity Sensitivity 0.4 0.4 0.2 0.2 Ab >20 U/mL AUROC = 1.00 (1.00–1.00) Ab >20 U/mL AUROC = 0.99 (0.99–1.00) Ab 2–20 U/mL AUROC = 0.98 (0.97–0.99) Ab 2–20 U/mL AUROC = 0.95 (0.93–0.96) 0.0 Ab <2.0 U/mL AUROC = 0.81 (0.74–0.87) 0.0 Ab < 2 U/mL AUROC = 0.72 (0.68–0.75) 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 Specificity Specificity e East Asians f Europeans Subgroup Cut-off Specificity Sensitivity Subgroup Cut-off Specificity Sensitivity CRS All 1.45 99% 74% CRS All 2.50 99% 60% CRS Ab- 1.45 99% 37% CRS Ab- 2.50 99% 20% GRS All 2.78 99% 15% GRS All 2.64 99% 13% Fig. 7 Diagnostic properties of the genetic risk score (GRS) and combined risk score (CRS) stratified by anti-PLA2R antibody status. Comparisons of receiver operating characteristic (ROC) curves to discriminate antibody positive (PLA2R Ab > 20 U/mL), borderline negative (PLA2R Ab 2–20 U/mL), and negative (PLA2R Ab < 2 U/mL) cases of primary MN from all available healthy and diseased controls combined for a GRS in East Asian discovery cohorts, b GRS in European discovery and validation cohorts, c CRS in East Asian discovery cohorts, d CRS in European discovery and validation cohorts. Overall sensitivities for risk score cut-offs corresponding to 99% specificities in (e) East Asian discovery cohorts and (f) European discovery and validation cohorts; CRS All: all patients with serum Ab measurements by ELISA within 6 months of a diagnostic kidney biopsy; CRS Ab-: patient subgroup with PLA2R Ab < 20 U/mL, and GRS All: all patients with GWAS data available for GRS calculation. AUROC: area under the ROC curve (95% confidence interval). methylation in trans across >400 CpGs that overlap with NF-κB- These findings suggest that different epitopes are likely presented binding sites, suggesting enhanced baseline activity of NF-κB38. to T cells to initiate the anti-PLA2R response in East Asians and The NF-κB complex is known to up-regulate IRF4 expression Europeans. We also identified specific high-risk amino acid with cross-regulatory feedback loops between NFKB1 and IRF4 substitutions, at positions 13, 71, and 74, mapping to the P4 described in several prior studies25–27. Taken together, NFKB1 pocket of DRβ1. Although the same positions contribute to the and IRF4 loci participate in a common regulatory pathway in risk of T1D39 and rheumatoid arthritis40, the effects of individual immune cells, and our genetic findings clearly establish a critical residues at each position are discordant, likely reflecting differ- role of this pathway in the pathogenesis of MN. ences in target epitopes. Second, due to the bi-ethnic composition of our cohorts, we Third, we confirm that a single haplotype at the PLA2R1 locus were able to refine ethnicity-specific effects at the HLA locus, conveys the disease risk in both East Asians and Europeans, and defining DRB1*1501 as a major risk allele in East Asians, exhibits genetic interactions with HLA-DRB1 risk alleles. Our DQA1*0501 in Europeans, and DRB1*0301 in both ethnicities. analysis supports a regulatory function of the PLA2R1 risk NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w haplotype. The candidate causal variant resides in the first intron of genetic interactions. These insights enabled formulation a pow- PLA2R1 and intersects a predicted enhancer element. While this erful genetic disease predictor that provides means to enhance a variant is normally associated with suppressed PLA2R1 transcrip- non-invasive diagnosis of MN, and can be especially useful in the tion across multiple tissues, it appears to increase expression of settings where kidney biopsy represents too great of a risk or is PLA2R1 in the kidney. This finding highlights the importance of not readily available. studying target tissues and is consistent with the findings that among CKD loci that are transcriptionally active in renal tissue, 15.8% of effects are kidney-specific41. Notably, the top variants at Methods Study design overview. We performed a genome-wide meta-analysis of eight the PLA2R1 locus also intersect a putative NF-κB binding site in discovery cohorts of East Asian and European ancestry (2956 cases and 7799 lymphocytes, although no similar data is presently available for controls), all genotyped with high resolution arrays and imputed to ~7 million podocytes. Further experimental work is thus needed to test if the common high-quality markers using ancestry-matched reference panels. The top glomerular-specific eQTL effect is under the transcriptional control signals from the meta-analysis (P < 5 × 10−5) were typed in the additional East Asian replication cohort of 826 cases and 1239 controls. Subsequently, all cohorts of NF-κB. Moreover, larger glomerular compartment-specific (3782 cases and 9038 controls) were analyzed jointly to define genome-wide sig- datasets will be needed to confirm the observed eQTL effects. nificant signals. All subjects provided informed consent to participate in genetic Another observation is that all four genome-wide significant studies, and the Institutional Review Board of Columbia University as well as local risk loci (PLA2R1, IRF4, NFKB1, and HLA) exhibit highly ethics review committees for each of the individual cohorts approved our study protocol. The individual cohorts, genotyping methods, and quality control analyses pleiotropic effects and all four lead SNPs have a concordant effect are described in the Supplementary Methods. on the risk of inflammatory bowel disease (IBD). This observation suggests shared pathogenic mechanism between IBD and MN. Considering that MN is an orphan disease without a targeted Primary association analyses and genome-wide meta-analyses. Within each cohort, primary association scans were performed for markers that were common treatment, there may now be opportunities for drug re- (MAF > 0.01) and imputed at high quality (r2 > 0.8) using logistic regression under positioning approaches from IBD, where several new anti- additive coding of dosage genotypes, and with adjustment for cohort-specific sig- inflammatory agents are currently under development. Our nificant principal components (PCs) of ancestry. To quantify potential inflation of study suggests that the NF-κB and interferon pathways may type I error due to stratification or technical artifacts, we estimated genomic represent particularly attractive drug targets. inflation factors45 for each genome-wide scan after excluding HLA and PLA2R loci. No substantial inflation was noted in any individual scan (lambda consistently Remarkably, our GWAS loci are highly predictive of the disease <1.05 for each individual cohort). Subsequently, a fixed effects meta-analysis was status and jointly explain up to one third of disease risk, an performed to combine the results of the eight discovery cohorts using METAL46. exceptionally large fraction for common alleles. This may be Genome-wide distributions of P-values were examined visually using quantile- partially explained by the fact that MN frequently occurs after the quantile plots for each individual cohort as well as for the combined analysis. The final meta-analysis quantile-quantile plot showed no global departures from the peak reproductive age, allowing the risk alleles to escape purifying expected null distribution (Supplementary Fig. 3), with the genomic inflation factor selection. Moreover, even though the risk alleles are common, our estimated at 1.03 for the overall meta-analysis. Suggestive signals were defined by interaction analysis demonstrates that specific high-risk genotype P-value < 5.0 × 10−5. To declare genome-wide significance of a novel locus, we used combinations are relatively rare in the general population, the generally accepted P-value threshold of 5.0 × 10−8. potentially explaining the low overall prevalence of MN42. The alternative hypothesis is that of balancing selection. NF-kB and Conditional analyses. To detect additional independent SNPs at genome-wide IRF4 are both involved in immune defenses against common significant loci, we performed stepwise conditional analyses of each locus using pathogens and some Phospholipase A2 ligands for PLA2R1 logistic regression. This was done by including the genotype of conditioning SNP (s) under additive coding as covariate(s) in the outcome model. The conditional represent downstream NF-kB targets with antibacterial analyses were performed individually within each cohort and with adjustments for properties43,44. Therefore, the observed high frequencies of MN cohort-specific ancestry PCs. Subsequently, the conditioned summary statistics risk alleles could be explained by their protective effects against were combined across cohorts using fixed effects meta-analysis, similar to our common infections. primary association analyses. Finally, a simple GRS based on our GWAS loci has excellent discriminant properties when combines with anti-PLA2R Credible set analyses. For each of the three genome-wide significant non-HLA ELISA test. Importantly, the combined genetic-serum test has loci, we derived 99% credible sets using the trans-ethnic and ethnicity-specific superior diagnostic properties compared to serologic test alone, meta-analysis results. First, we derived approximate Bayes factors from GWAS mitigating the key issue of low sensitivity. The GRS provides association statistics using Wakefield’s formula, as implemented in the R package gtx47. Using CAVIAR software48, we next calculated the posterior probability (PP) complementary information to the serum test and correctly re- for each SNP driving the association signal at each locus. We assumed there was classifies 20–37% of antibody-negative cases, potentially sparing only a single causal variant at each locus, since no additional independent SNPs the need for a kidney biopsy in this large subgroup of patients. were detected on stepwise conditioning analyses. We derived both trans-ethnic as In the clinical settings where neither a serum test nor a kidney well as ethnicity-specific 99% credible sets based on ranking the variants by their PPs and adding the variants to the set until cumulative PP > 99% was reached for biopsy is possible, the GRS itself can establish a diagnosis of each region. The overlaps between ethnicity-specific and trans-ethnic analyses were MN with 99% specificity in 13–15% of cases. The practical visualized in Supplementary Fig. 5. advantage of this approach is that the GRS can be readily determined at any time after birth and, unlike the serum test, it does not fluctuate with time or in relationship to the disease HLA imputation. Six discovery cohorts (Chinese, South Korean, Japanese, Eur- opean-1, European-2, and Turkish) had primary genotype data available for HLA onset, activity, or treatment. One important limitation, how- imputation and association testing. For each of these cohorts, we imputed classical ever, is that genetic effects may be population-specific and may HLA alleles at two- and four-digit resolution, as well as individual amino acid not be generalizable to populations not represented in our polymorphisms at class I (HLA-A, -B, and -C) and class II (HLA-DQB1, HLA- GWAS. The performance of the GRS is remarkably consistent DQA1 and HLA-DRB1) loci using SNP2HLAsoftware49. The European cohorts and the East Asian cohorts were imputed separately, using ethnicity-specific reference in our discovery and validation cohorts, including African- panels. For European reference, we used the pre-phased HLA reference dataset Americans, but it appears to be lower in self-reported Latino/ generated by the Type 1 Diabetes Genetics Consortium (T1DGC, 5,225 indivi- Hispanics. Therefore, future efforts extending GWAS for MN to duals)49. For our East Asian cohorts, we used the Pan-Asian HLA Reference Panel more diverse populations will be important. (268 individuals)50. For validation of classical HLA association results in East Asians, we built additional East Asian reference panel based on the MHC sequence In summary, we described a highly unusual genetic archi- data from Zhou et al. (10,689 Han Chinese)10. In the association analyses, we tecture of MN, including large effect sizes for a small number of included only common HLA alleles (MAF > 0.01) that were imputed with high common alleles and a strong evidence for ethnicity-specific certainty (R2 > 0.8). 12 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE Statistical framework for HLA association testing. Given that the frequency of genome-wide transcriptome data (Affymetrix 2.1 ST chips) performed on micro- HLA alleles can vary by ethnicity, we performed HLA association testing Eur- dissected glomerular (N = 136) and tubulointerstitial (N = 166) tissue com- opeans and East Asians separately. We used logistic regression models to test the partments from kidney biopsies of patients with nephrotic syndrome. For each additive effects of HLA allele dosages with adjustment for significant PCs of locus, we tested the index SNP and its high LD SNPs (r2 > 0.8). Testing for cis- ancestry. For multi-allelic loci, we used the following logistic regression model: eQTL effects involved all transcripts within 1-Mb region centred on each SNP using the additive linear regression with adjustments for age, sex, PEER factors X m1 X n and first 4 PCs of ancestry as described previously13. In addition to kidney tissue, logðoddsi Þ ¼ β0 þ βj xj;i þ βk Pk;i ð1Þ j¼1 k¼1 all loci were similarly interrogated for eQTL effects in the GTEx database version 8. Because NFKB1 and IRF4 both encode transcription factors, we also explored where m indicates a total number of alleles at a specific multi-allelic locus, j their potential binding in close proximity of each other or PLA2R1 gene. Based indicates a specific allele being tested, and xj,i is the imputed dosage for allele j for on the Chip-seq data for all five subunits of NFκB complex in immortalized individual i; β0 represents the intercept and βj represents the additive effect of an lymphocytes (GSE55105)29, we found that the top SNPs at PLA2R1 and IRF4 allele j; Pk,i denotes the value for kth PC of individual i, n is the total number of loci intersect potential NFκB complex binding site (Supplementary Figs. 8 and significant PCs in the dataset, βk is the effect size of principal component k. We 13). In addition, rs230492, a variant in strong LD (r2=0.94) with the top SNP at compared log-likelihoods of two nested models: the full model containing the test the NFKB1 locus intersects a potential IRF4 binding site based on the Chip-seq locus and relevant covariates with the reduced model without the test locus, but data of IRF4 (GEO: GSM803390). with the same set of covariates. The deviance was defined as −2×log likelihood ratio, which follows a X2-distribution with m − 1 degrees of freedom, from which Pleiotropy analysis. We used the latest GWAS catalogue data to perform sys- we calculated P-values. In addition to multi-allelic tests, we also performed bi- tematic cross-annotation of our top risk alleles against all other published GWAS allelic tests of association for all individual SNPs, classical HLA alleles, and indi- findings. We first identified all genome-wide significant SNPs (P < 5 × 10−8) vidual amino acid residues in HLA molecules. All analyses were performed using reported in the catalogue that resided within the genomic regions of association dosage method under additive coding. Stepwise conditioning analyses across the with MN. We assessed the extent of linkage disequilibrium (r2) between these SNPs HLA region were performed using both multi-allelic and bi-allelic coding of HLA and the top MN risk alleles based on the combined European and East Asian variants. In each round of stepwise conditioning, we first included the most sig- sequence data from 1000 Genomes (phase 3). We next defined the directionality of nificant variant as the covariate in the logistic regression model. If additional pleiotropic effects as either concordant or opposed in relationship to the MN risk independently associated markers are detected, they are included as covariates in alleles. In addition, we queried each qualifying SNP from the catalogue against our our subsequent models. We repeated these analyses until no residual associations genome-wide summary statistics to extract the odds ratios and P-values for asso- across the entire locus were observed. ciations with MN. We defined overlapping susceptibility alleles if r2 exceeded 0.2 (Supplementary Table 13). Lastly, we constructed a susceptibility overlap map that Analysis of polymorphic amino-acid sites in HLA genes. To test the effects of connects each of the MN loci to the previously associated GWAS traits and individual amino acid substitution sites within the HLA-DRB1 and HLA-DQA1 highlights associations with SNPs in high LD with the top MN signals (Fig. 5a). genes, we applied a conditional haplotype analysis using fully phased haplotypes The map was visualized with Cytoscape v.3.6 software. across the HLA region. We tested each single amino acid position by first identifying the m possible amino acid residues occurring at that position and Podocyte gene annotations of GWAS loci. To search for potential novel podo- then using m − 1 degrees of freedom test to derive P-values, with a single amino cyte antigens encoded by our suggestive loci, we interrogated each locus against a acid residue arbitrarily selected as a reference. For conditioning on individual podocyte-specific gene list predicted with in silico “nanodissection” approach amino acid sites, we used the following procedure: by adding a new amino acid (Supplementary Table 14). This computational approach used Affymetrix gene position to the model, a total of k additional unique haplotypes were generated expression data from micro-dissected glomerular and tubulo-interstitial compart- and tested over the null model (without a new amino acid site) using the like- ments of 452 renal biopsies52. In addition, we cross-annotated our positional lihood ratio test with k degrees of freedom. If the new position was indepen- candidates against the list of native podocyte proteins discovered by proteomic dently significant, we further updated the null model to include all unique profiling of mouse podocytes53. All suggestive loci were additionally tested for haplotypes created by all amino acid residues at both positons to identify multiplicative interactions with classical alleles, and the top most significant another independent position. The procedure was repeated until no statistically interactions were summarized in Supplementary Table 14. Lastly, we annotated all significant positions were observed. positional candidate genes using manual PubMed literature searches to prioritize genes with a previously established role in podocyte biology. Testing for pairwise epistasis. Multiplicative interactive effects were tested using logistic regression model; SNPs were coded under additive genotype coding (0, 1, GR analysis. To estimate the cumulative effect of independently significant 2), and interaction terms were defined as simple products of genotypes. To screen GWAS loci, we used a GRS approach. We first identified SNPs with independent for interactions, we tested a lead SNP at each of the three non-HLA loci against contributions to MN risk at each locus using stepwise conditional analyses. Next, each of the five independent HLA SNPs (three in Europeans and two in East we tested for multiplicative interactions among those independently associated Asians) resulting in a total of 15 independent tests. We additionally tested for all SNPs. We then built a logistic regression model including all independent SNPs pairwise interactions between the three non-HLA loci shared between both eth- along with their significant interaction terms to derive mutually adjusted effect nicities resulting in three additional tests. In each case, we used a likelihood ratio sizes (Supplementary Table 15). Because we observed ethnicity-specific signals at test comparing two nested models: the full model with both main effects and the the HLA locus, we generated ethnicity-specific for East Asians and Europeans interaction term to the reduced model with main effects only. Given a total of 18 separately. The ethnicity-specific risk scores were defined as a weighted sum of pairwise interaction tests, we used a Bonferroni-corrected significance threshold of independent risk alleles and their significant interaction terms, weighted by their 0.05/18 = 2.8 × 10−3. In secondary analyses, we explored if significant HLA risk mutually adjusted effect sizes. The GRS was standardized using a Z-score haplotypes interact with the PLA2R1 risk allele in the six cohorts with fully transformation based on the mean and standard deviation for the distribution of imputed classical HLA alleles. This included a total of 2759 East Asians (803 cases ethnically matched controls, so that the standardized GRS was reflective of the and 1956 controls) and 4507 Europeans (1880 cases and 2627 controls). distance between the raw score and the control mean in units of standard deviation. The final formulation of the GRS is as follows: Functional annotations of GWAS loci. We used several different approaches to  perform functional annotations of our significant loci. We first defined the East Asian GRSi ¼ 0:69173 ´ d ðrs9269027  AÞi þ 1:23685 ´ d ðrs1974461  T Þi þ0:36687 region of each locus as +/−400 kb of the index SNP. Using ANNOVAR soft- ´ d ðrs6707458  GÞi þ 0:25098 ´ d ðrs230540  CÞi þ 0:39127 ð2Þ ware51, we identified functional variants within each region that were in strong ´ d ðrs9405192  GÞi þ 0:48798 ´ d ðrs9269027  AÞi ´ d ðrs6707458  GÞi linkage disequilibrium (r2 > 0.8) with the top SNP, including all known coding,  East Asian Control Mean GRS= East Asian Control SD GRS splicing, 3′UTR and 5′UTR variants (Supplementary Table 12). To assess for potential functional variation in non-coding regions, we used our recently where GRSi = genetic risk score for individual i, di = dosage of risk allele (from 0 proposed tissue-specific functional scoring method (FUN-LDA)11. Using FUN- to 2) for individual i, East Asian Control Mean GRS = 1.6804, East Asian LDA, we estimated the posterior probability for each variant in strong LD with Control SD GRS = 1.003 the top SNP of being functional across 127 different tissues or cell types profiled  by ENCODE and ROADMAP consortia (Supplementary Figs. 6 and 10). To European GRSi ¼ 0:34945 ´ dðrs9271541  CÞi þ 0:67919 ´ d ðrs9265949  T Þi interrogate candidate variants against kidney-specific chromatin landscape, we þ 0:30707 ´ d ðrs2858309  C Þi þ 0:34601 ´ dðrs6707458  GÞi analyzed regulatory DNase-seq maps paired with RNA-seq gene expression þ 0:17450 ´ d ðrs230540  CÞi þ 0:18343 ´ dðrs9405192  GÞi ð3Þ profiles from primary outgrowth cultures of human glomeruli (composed mainly þ 0:33782 ´ d ðrs9271541  C Þi ´ dðrs6707458  GÞi of podocytes and mesangial cells) and renal cortex cultures (composed mainly of  European Control Mean GRS=European Control SD GRS tubular cells), as well as chromatin conformation (Hi-C) maps from freshly isolated human glomeruli (Supplementary Figs. 9 and 11)14. To test for eQTL where GRSi = genetic risk score for individual i, di = dosage of risk allele (from 0 effects in kidney tissue, we used gene expression data from the NEPTUNE to 2) for individual i, European GRS control mean = 1.5089, European GRS study13. This dataset is comprised of whole genome DNA sequence data and control SD = 0.8202. NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w The percentage of the total variance in disease risk explained was estimated numbers of NEPTUNE participants in other ancestral groups were too small for a using Nagelkerke’s pseudo R2 from the logistic regression model with the meaningful analysis. standardized GRS as a predictor and case-control status as an outcome. The performances of the ethnicity-specific GRS were estimated by the AUROC. We performed detailed GRS cut-off analyses in the discovery cohorts by selecting cut- PLA2R antibody testing. In total, we determined serum antibody levels in N = off points on the ROC curve that provide specificities in the range from 95 to 100%. 2331 study participants with genetic data (N = 1488 cases, N = 300 healthy con- For each cut-off point, we calculate sensitivity, specificity, positive likelihood ratio trols, and N = 543 disease controls) across all cohorts and ethnicities. The ancestry- (LR+), and negative likelihood ratio (LR-). All GRS analyses were implemented in matched diseased controls were recruited among patients commonly presenting R version 3.3.2. with nephrotic syndrome, including FSGS, MCD, IgAN. For all MN cases and disease controls, serum samples were obtained near or at the time of kidney biopsy and any samples obtained more than six months after the biopsy were excluded CRS formulation. The CRS was formulated as a weighted sum of the GRS and from the analysis. All individuals that underwent antibody testing had matched serum anti-PLA2R antibody levels in U/mL. The weight was determined using genetic data, enabling derivation and diagnostic testing of the CRS. logistic regression model, with standardized ethnicity-specific GRS (formulated as We performed a standardized measurement of serum anti-PLA2R Ab levels described above) and natural log-transformed anti-PLA2R antibody levels as two using the anti-PLA2R ELISA (IgG) test kit (EUROIMMUN Medizinische predictors and case-control status as an outcome. This resulted in the following Labordiagnostika AG), which employs the indirect ELISA methodology. The kit model: includes a 96-well microplate pre-coated with PLA2R, 5 calibrators (2, 20, 100, 500, Yi ¼ β0 þ β1 GRSi þ β2 lnðαPLA2Ri þ 0:001Þ ð4Þ and 1500 U/mL respectively), positive and negative control samples, peroxidase- labelled anti-human IgG (rabbit) enzyme conjugate, kit specific sample and wash where GRSi indicates the Z-transformed ethnicity-specific GRS for individual i, β0 buffers, Chromogen/substrate solution (TMB/H2O2), and stop solution. The assay indicates the intercept, β1 indicates the effect size of the GRS estimated based on was run as per the protocol included with the kit. A 5-point calibrated analysis was discovery cohorts; αPLA2Ri is the serum level of PLA2R antibodies in U/mL; the used to calculate the results for each assay performed. A standard curve was constant 0.001 is added to enable log transformation of undetectable (zero) levels; generated based on the spectrophotometric reading of the five calibrators included β2 represents the effect size for anti-PLA2R antibody positivity estimated based on on each microplate. As recommended by EUROIMMUN, the sample was called discovery cohorts. The weight for the In(αPLA2Ri + 0.001) term was then defined positive if antibody level was ≥20 U/mL. as follows: In the discovery stage, we analyzed sera for a total of N = 459 East Asian and N = 1034 European participants. The European cohorts included 810 cases with β2 MN, 99 healthy controls, and 125 disease controls (37 FSGS and 88 IgAN). The Weight ¼ ð5Þ β1 East Asian cohorts included 304 cases with MN, 56 healthy controls, and 99 disease controls (52 FSGS and 47 IgAN). The disease controls were not included in the Consequently, the weights for antibody levels were calculated for East Asian and GWAS discovery analysis, but were added to test for disease specificity of the European separately, which resulted in the following Crude CRS formulation: serologic test. We note that one East Asian disease control with a clinical diagnosis  GRSi þ 0:4829 ´ lnðαPLA2Ri þ 0:001Þ; if i 2 European of IgAN tested anti-PLA2R antibody positive at high titer; follow-up pathology Crude CRSi ¼ ð6Þ review found evidence of previously unrecognized sub-epithelial deposits GRSi þ 1:7712 ´ lnðαPLA2Ri þ 0:001Þ; if i 2 East Asian diagnostic of MN in addition to IgA-dominant mesangial deposits; this individual In the final step, the Crude CRS was Z-transformed using the mean and was subsequently removed from the analysis. standard deviation for ethnicity-matched healthy controls: In the validation phase, we analyzed sera for a total of N = 540 individuals  including European case-control cohorts (N = 248 cases and N = 145 healthy ðCrude CRSi  European Control Mean CRSÞ=ðEuropean Control SDÞ; if i 2 European CRSi ¼ controls) and a total of 147 European NEPTUNE participants (N = 36 cases and ðCrude CRSi  E:Asian Control Mean CRSÞ=ðE:Asian Control SDÞ; if i 2 East Asian N = 111 disease controls). In secondary analysis, we extended testing to additional ð7Þ 180 NEPTUNE participants of non-European ancestry (N = 28 cases and N = 152 controls). This included 103 NEPTUNE participants of African American ancestry where European Control Mean CRS = −1.4982, European Control SD = 1.4354, E. (16 cases and 87 disease controls) and 77 participants of Hispanic/Latino ancestry Asian Control Mean CRS = 0.3724, E. Asian Control SD = 2.7503. (12 cases and 65 disease controls). These numbers are smaller compared to the The CRS performance was estimated by the area under the receiver operating GRS validation cohorts, since not all NEPTUNE participants had sera sampled curve (AUROC). Similar to GRS, we explored different CRS cut-offs that within 6 months of biopsy. maximized specificity in the range 95 to 100%. For each cut-off, we calculated sensitivity, specificity, positive likelihood ratio (LR+), and negative likelihood ratio (LR−). The integrated discrimination improvement (IDI) and net reclassification Testing for phenotypic correlations of the GRS. Using extensive clinical infor- improvement (NRI)54 were calculated comparing the CRS test to the serum PLA2R mation available for our discovery cohorts, we investigated correlations between antibody test alone. All CRS analyses were implemented in R version 3.3.2 GRS and clinical traits from the time of kidney biopsy including age at diagnosis, (CRAN). 24-h proteinuria (P24), estimated glomerular filtration rate (eGFR), serum albumin (Alb) and serum anti-PLA2R1 antibody level (Supplementary Table 16). The eGFR European GWAS validation cohorts. For the purpose of GRS validation, we was estimated based on serum creatinine level using the Modification of Diet in utilized three previously published external GWAS cohorts of European ancestry: Renal Disease (MDRD) equation55. The proteinuria was quantified using spot the UK, French, and Dutch Validation Cohorts6. These cohorts were composed of urine protein-to-creatinine ratios. The values of proteinuria and eGFR were nor- biopsy-documented cases of primary MN and ethnicity-matched healthy controls, malized by natural log-transformation. The serum albumin and anti-PLA2R1 totalling 2,887 individuals (550 cases and 2337 controls, see Supplementary antibody levels also required natural log-transformation. For each quantitative Methods for details). For all three cohorts, we obtained primary genotype data after trait, we built a linear regression model with GRS as a predictor and each corre- quality control analysis as published previously6 and performed phasing and sponding trait as an outcome. For dichotomous traits (anti-PLA2R seropositivity or imputation using Eagle v.2.3 and Minimac3 with European populations of Phase 3 presence of nephrotic range proteinuria), we used logistic regression with a GRS as 1000 Genome Project as a reference. The cases and controls were imputed jointly. a predictor and each binary trait as an outcome. The association analysis for age at The GRS for each individual was determined using genotype dosages for the SNPs diagnosis (biopsy) was performed before and after adjustment for sex and ancestry. included in the score. The performance of GRS was then analyzed individually in The tests of proteinuria, eGFR, serum albumin and serum anti-PLA2R1 antibody each cohort, and in all cohorts combined (Supplementary Tables 17–19). levels were carried out before and after controlling for age, sex and ancestry. Statistical analyses were implemented in R version 3.3.2. NEPTUNE GRS validation cohorts. The Nephrotic Syndrome Study Network (NEPTUNE) is a prospective, longitudinal cohort recruiting participants with SNP-based heritability. We estimated the SNP-based heritability of MN in East nephrotic syndrome at the time of first kidney biopsy8; primary disease diagnoses Asians and Europeans using the GCTA-GREML algorithm30,56. For this analysis, of MN, FSGS, MCD, and IgAN were determined by a central pathology review. All we included three European cohorts (European discovery 1, European discovery 2 NEPTUNE participants underwent low-depth whole genome sequencing as and Turkish discovery) and three East Asian cohorts (Chinese, Korean and Japa- described in the Supplementary Methods. The GRS was successfully determined for nese discovery) that had primary genotype data available for joint heritability N = 475 NEPTUNE participants. This included 89 cases with the diagnosis of analysis. We first estimated pairwise genetic relationship matrix between all indi- primary MN and 386 disease controls, including 184 with FSGS, 164 with MCD, viduals using autosomal SNPs. With the GCTA software57, we next estimated the and 38 with IgAN. Our pre-specified primary GRS validation involved 180 NEP- disease variance explained by all autosomal SNPs. We transformed the estimate TUNE participants of European ancestry (46 cases and 134 disease controls, assuming an underlying liability scale and disease prevalence of 0.001. We derived Supplementary Tables 17–18). In secondary analyses, we extended our validation a standard error (SE) and 95% confidence interval for each estimate. studies to the entire multiethnic cohort of 475 NEPTUNE participants (Supple- mentary Tables 20–21), including subgroup analyses of 133 individuals of African American ancestry (18 cases and 115 disease controls) and 94 individuals of Reporting summary. Further information on research design is available in Hispanic/Latino American ancestry (18 cases and 76 disease controls). The the Nature Research Reporting Summary linked to this article. 14 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE Data availability 26. Saito, M. et al. A signaling pathway mediating downregulation of BCL6 in All genome-wide summary statistics, including those presented in Fig. 1, are freely germinal center B cells is blocked by BCL6 gene alterations in B cell available for download on our lab website: www.columbiamedicine.org/divisions/kiryluk/ lymphoma. Cancer Cell 12, 280–292 (2007). resources.php. The calculations of genetic risk score (GRS) and combined risk score 27. Boddicker, R. L. et al. The oncogenic transcription factor IRF4 is regulated by (CRS) are implemented in the form of an online risk calculator, which is also freely a novel CD30/NF-kappaB positive feedback loop in peripheral T-cell available on our lab website. The PAGE consortium control genotype data is available on lymphoma. Blood 125, 3118–3127 (2015). dbGAP under accession number phs000356.v2.p1. Primary genotype data for the 28. Lake, B. B. et al. A single-nucleus RNA-sequencing pipeline to decipher the European-1 discovery cohort is available under dbGAP accession number phs001984.v1. molecular anatomy and pathophysiology of human kidneys. Nat. Commun. p1. Our IRB determined that the use of this dataset is restricted to genetic studies of 10, 2832 (2019). kidney disease. Because of consent restrictions and/or country-specific privacy laws, we 29. Zhao, B. et al. The NF-kappaB genomic landscape in lymphoblastoid B cells. are unable to share primary genotype data on dbGAP for other cohorts. All data and Cell Rep. 8, 1595–1606 (2014). summary statistics are available from the corresponding authors upon reasonable 30. Lee, S. H., Wray, N. R., Goddard, M. E. & Visscher, P. M. Estimating missing request. heritability for disease from genome-wide association studies. Am. J. Hum. Genet. 88, 294–305 (2011). 31. Bobart, S. A. et al. Noninvasive diagnosis of primary membranous Received: 28 August 2019; Accepted: 3 March 2020; nephropathy using phospholipase A2 receptor antibodies. Kidney Int 95, 429–438 (2019). 32. Schmid, H. et al. Modular activation of nuclear factor-kappaB transcript- ional programs in human diabetic nephropathy. Diabetes 55, 2993–3003 (2006). References 33. Atreya, I., Atreya, R. & Neurath, M. F. NF-kappaB in inflammatory bowel 1. Glassock, R. J. Diagnosis and natural course of membranous nephropathy. disease. J. Intern. Med. 263, 591–596 (2008). Semin Nephrol. 23, 324–332 (2003). 34. Schottelius, A. J. & Baldwin, A. S. Jr A role for transcription factor NF-kappa 2. Debiec, H. et al. Antenatal membranous glomerulonephritis due to anti- B in intestinal inflammation. Int J. Colorectal Dis. 14, 18–28 (1999). neutral endopeptidase antibodies. N. Engl. J. Med. 346, 2053–2060 (2002). 35. Mezzano, S. A. et al. Tubular NF-kappaB and AP-1 activation in human 3. Beck, L. H. Jr. et al. M-type phospholipase A2 receptor as target antigen in proteinuric renal disease. Kidney Int. 60, 1366–1377 (2001). idiopathic membranous nephropathy. N. Engl. J. Med. 361, 11–21 (2009). 36. Mudge, S. J., Paizis, K., Auwardt, R. B., Thomas, R. J. & Power, D. A. 4. Glassock, R. J. The pathogenesis of membranous nephropathy: evolution and Activation of nuclear factor-kappa B by podocytes in the autologous phase of revolution. Curr. Opin. Nephrol. Hypertens. 21, 235–242 (2012). passive Heymann nephritis. Kidney Int. 59, 923–931 (2001). 5. Tomas, N. M. et al. Thrombospondin type-1 domain-containing 7A in 37. Liu, S. et al. Urinary messenger RNA of the receptor activator of NF-kappaB idiopathic membranous nephropathy. N. Engl. J. Med. 371, 2277–2287 (2014). could be used to differentiate between minimal change disease and 6. Stanescu, H. C. et al. Risk HLA-DQA1 and PLA(2)R1 alleles in idiopathic membranous nephropathy. Biomarkers 19, 597–603 (2014). membranous nephropathy. N. Engl. J. Med. 364, 616–626 (2011). 38. Bonder, M. J. et al. Disease variants alter transcription factor levels and 7. Wunnenburger, S. et al. Associations between genetic risk variants for kidney methylation of their binding sites. Nat. Genet. 49, 131–138 (2017). diseases and kidney disease etiology. Sci. Rep. 7, 13944 (2017). 39. Hu, X. et al. Additive and interaction effects at three amino acid positions in 8. Gadegbeku, C. A. et al. Design of the Nephrotic Syndrome Study Network HLA-DQ and HLA-DR molecules drive type 1 diabetes risk. Nat. Genet. 47, (NEPTUNE) to evaluate primary glomerular nephropathy by a 898–905 (2015). multidisciplinary approach. Kidney Int. 83, 749–756 (2013). 40. Raychaudhuri, S. et al. Five amino acids in three HLA proteins explain most of 9. Cui, Z. et al. MHC class II risk alleles and amino acid residues in idiopathic the association between MHC and seropositive rheumatoid arthritis. Nat. membranous nephropathy. J. Am. Soc. Nephrol. 28, 1651–1664 (2017). Genet. 44, 291–296 (2012). 10. Zhou, F. et al. Deep sequencing of the MHC region in the Chinese population 41. Xu, X. et al. Molecular insights into genome-wide association studies of contributes to studies of complex disease. Nat. Genet. 48, 740–746 (2016). chronic kidney disease-defining traits. Nat. Commun. 9, 4800 (2018). 11. Backenroth, D. et al. FUN-LDA: a latent dirichlet allocation model for 42. Sekula, P. et al. Genetic risk variants for membranous nephropathy: extension predicting tissue-specific functional effects of noncoding variation: methods of and association with other chronic kidney disease aetiologies. Nephrol. Dial. and applications. Am. J. Hum. Genet 102, 920–942 (2018). Transpl. 32, 325–332 (2017). 12. Consortium, G. T. et al. Genetic effects on gene expression across human 43. Nevalainen, T. J., Graham, G. G. & Scott, K. F. Antibacterial actions of tissues. Nature 550, 204–213 (2017). secreted phospholipases A2. Review. Biochim Biophys. Acta 1781, 1–9 13. Gillies, C. E. et al. An eQTL landscape of kidney tissue in human nephrotic (2008). syndrome. Am. J. Hum. Genet. 103, 232–244 (2018). 44. Morri, H., Ozaki, M. & Watanabe, Y. 5’-flanking region surrounding a human 14. Sieber, K. B. et al. Integrated functional genomic analysis enables annotation cytosolic phospholipase A2 gene. Biochem. Biophys. Res. Commun. 205, 6–11 of kidney genome-wide association study loci. J. Am. Soc. Nephrol. 30, (1994). 421–441 (2019). 45. Devlin, B., Roeder, K. & Bacanu, S. A. Unbiased methods for population-based 15. Westra, H. J. et al. Systematic identification of trans eQTLs as putative drivers association studies. Genet. Epidemiol. 21, 273–284 (2001). of known disease associations. Nat. Genet 45, 1238–1243 (2013). 46. Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and efficient meta-analysis 16. Raj, T. et al. Polarization of the effects of autoimmune and neurodegenerative of genomewide association scans. Bioinformatics 26, 2190–2191 (2010). risk alleles in leukocytes. Science 344, 519–523 (2014). 47. Wakefield, J. Bayes factors for genome-wide association studies: comparison 17. Astle, W. J. et al. The allelic landscape of human blood cell trait variation and with P-values. Genet. Epidemiol. 33, 79–86 (2009). links to common complex disease. Cell 167, 1415–1429 e19 (2016). 48. Hormozdiari, F., Kostem, E., Kang, E. Y., Pasaniuc, B. & Eskin, E. Identifying 18. Jostins, L. et al. Host-microbe interactions have shaped the genetic causal variants at loci with multiple signals of association. Genetics 198, architecture of inflammatory bowel disease. Nature 491, 119–124 (2012). 497–508 (2014). 19. Liu, J. Z. et al. Association analyses identify 38 susceptibility loci for 49. Jia, X. et al. Imputing amino acid polymorphisms in human leukocyte inflammatory bowel disease and highlight shared genetic risk across antigens. PLoS ONE 8, e64683 (2013). populations. Nat. Genet 47, 979–986 (2015). 50. Pillai, N. E. et al. Predicting HLA alleles from high-resolution SNP data in 20. Mells, G. F. et al. Genome-wide association study identifies 12 new three Southeast Asian populations. Hum. Mol. Genet. 23, 4443–4451 (2014). susceptibility loci for primary biliary cirrhosis. Nat. Genet 43, 329–332 (2011). 51. Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of 21. Cordell, H. J. et al. International genome-wide meta-analysis identifies new genetic variants from high-throughput sequencing data. Nucleic Acids Res. 38, primary biliary cirrhosis risk loci and targetable pathogenic pathways. Nat. e164 (2010). Commun. 6, 8019 (2015). 52. Ju, W. et al. Defining cell-type specificity at the transcriptional level in human 22. Wu, H. et al. Comparative analysis and refinement of human PSC-derived disease. Genome Res. 23, 1862–1873 (2013). kidney organoid differentiation with single-cell transcriptomics. Cell Stem Cell 53. Rinschen, M. M. et al. A multi-layered quantitative in vivo expression atlas of 23, 869–881.e8 (2018). the podocyte unravels kidney disease candidate genes. Cell Rep. 23, 2495–2508 23. Pattaro, C. et al. Genetic associations at 53 loci highlight cell types and biological (2018). pathways relevant for kidney function. Nat. Commun. 7, 10023 (2016). 54. Steyerberg, E. W. et al. Assessing the performance of prediction models: a 24. Wuttke, M. et al. A catalog of genetic loci associated with kidney function framework for traditional and novel measures. Epidemiology 21, 128–138 from analyses of a million individuals. Nat. Genet. 51, 957–972 (2019). (2010). 25. Grumont, R. J. & Gerondakis, S. Rel induces interferon regulatory factor 4 55. Levey, A. S. et al. Using standardized serum creatinine values in the (IRF-4) expression in lymphocytes: modulation of interferon-regulated gene modification of diet in renal disease study equation for estimating glomerular expression by rel/nuclear factor kappaB. J. Exp. Med. 191, 1281–1292 (2000). filtration rate. Ann. Intern. Med. 145, 247–254 (2006). NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w 56. Yang, J. et al. Common SNPs explain a large proportion of the heritability for Author contributions human height. Nat. Genet. 42, 565–569 (2010). K.K., N.C. and R.K. conceived the study and made the decision to publish the findings; K. 57. Yang, J., Lee, S. H., Goddard, M. E. & Visscher, P. M. GCTA: a tool for K. designed the study and provided overall supervision of the project; K.K. and L. Liu genome-wide complex trait analysis. Am. J. Hum. Genet. 88, 76–82 (2011). wrote the initial draft of the manuscript; J. Xie coordinated recruitment and genotyping of the Chinese discovery and replication cohorts; N.M. and J. Xie performed quality control, imputation and GWAS association analyses for Asian discovery cohorts; L. Liu performed quality control, imputation and GWAS association analyses for European Acknowledgements discovery and validation cohorts, and performed final statistical and bioinformatics We are grateful to all study participants across multiple nephrology centres worldwide analyses, including GWAS meta-analyses, fine-mapping, HLA analyses, functional for their contributions to this work. This work was supported by the following institu- annotations, risk score analyses, clinical correlation analyses, and generated figures and tions, grants and funding agencies in the US: Columbia University, Columbia Glomerular tables; I.I.-L. consulted on the statistics and functional annotation of GWAS loci. A.G.G. Center, National Institute for Diabetes and Digestive Kidney Diseases (NIDDK) grants consulted on the study design and critically reviewed the manuscript. J. Xie, X.Y., O.B., RC2-DK116690 (K.K., M.K.), R01-DK105124 (K.K.), R01-DK097053 (L.H.B., M.K.), L.H.B., H.D., E.H., R.A.K.S., P.B. and R.K. performed serum antibody studies. O.B., Y.L., R01-DK108805 (M.G.S.), and National Institute on Minority Health and Health Dis- J.Y.Z., P.K., N.M., R.J.R., M. Bodria, A. Khan, K. Mehl, and F.O. contributed to sample parities (NIMHD) grant R01-MD009223 (K.K., A.G.G., A.B.). M.G.S. is additionally processing, biobanking, DNA extractions, quality control, plating, clinical and genetic supported by the Charles Woodson Clinical Research Fund. The Nephrotic Syndrome data management. S.A. generated kidney compartment-specific maps of chromatin Study Network Consortium (NEPTUNE), U54-DK-083912, is a part of the National accessibility and interactions. J. Xie, N.C., M-H.Z., Z.C., L.H.R., W.W., Z.L., X.H., X.Y., Institutes of Health (NIH) Rare Disease Clinical Research Network (RDCRN), supported D.Z., J. Xu, G.L. recruited, genotyped, and clinically characterized the Chinese Discovery through a collaboration between the Office of Rare Diseases Research, National Center and Chinese Replication cohorts. A.P., C.S., M.Z., and F.C. genotyped, imputed, and for Advancing Translational Sciences (NCATS) and the National Institute of Diabetes, analyzed GWAS data for the Sardinian Discovery cohort. M.W., K-U.E., and A. Köttgen Digestive, and Kidney Diseases (NIDDK). Additional funding and/or programmatic coordinated the GCKD Study, genotyped, imputed and analyzed GWAS data for the support for this project has also been provided by the University of Michigan, the GCKD Discovery cohort. H.S. recruited and clinically characterized the Japanese Dis- NephCure Kidney International and the Halpin Foundation. The recruitment and covery cohort. H.L., J.P., B.L.C., Y.S.K., and D.K.K. recruited and clinically characterized analysis of the Chinese cohorts were supported by the National Key Research and the Korean Discovery cohort. Y.C., S.O., A.Y., N.S., H.A., M. Koc, T.B., G.K., S.U.A., and Development Program of China (2016YFC0904100), Natural Science Foundation of M.S.S. recruited and clinically characterized the Turkish Discovery cohort. M.M., P.A.C., China to the Innovation Research Group (81621092), National Natural Science Foun- A.S.B., G.B.A., S.S.-C., M. Bodria, F.Z., A.P.-P., M.D., K. Mucha, B.M., B.F., L.P., I.H., dation of China (No. 81870460, 81570598), Science and Technology Innovation Action E.A., J.B., L.M., B.V., D. Santoro, M. Bonomini, F. Londrino, L.G., J.R., V.T., C.I., S.S., Plan of Shanghai Science and Technology Committee (No.17441902200), Shanghai D. Spotti, C.M., P.M., M.G., D.R., S. Granata, G.Z., F. Lugani, G.M.G., I.P., L.A., Municipal Education Commission, Gaofeng, Clinical Medicine Grant (No.20152207), B. Sprangers, D.C.C., F.C.F., A.A., and F.S. recruited and clinically characterized patients Shanghai Jiao Tong University School of Medicine, Multi-Center Clinical Research for the European-1 Discovery Cohort. P.H., S.H., S. Gupta, C.C., S.D., N.I., R.J.P., J.C., Project (No.DLY201510), International Cooperation and Exchange Projects of Shanghai S.P., D.B., H.C.S., N.A., E.H., R.A.K.S., P.B. and R.K. participated in the recruitment, Science and Technology Committee (No.14430721000), the Outstanding Young Scholar clinical characterization and generation of genotype data for the European Discovery-2 Award for Zhao Cui (No.81622009), and Shanghai Health and Family Planning Com- and the UK Validation cohorts. B. Stengel, H.D., and P.R. contributed the French mittee Hundred Talents Program for Jingyuan Xie (No.2018BR37). The recruitment of Validation Cohort. J.M.H., M.J.C., L.A.K., and J.F.M.W. contributed the Dutch Valida- the Korean cohort was supported by the Seoul National University Hospital Human tion Cohort. M.G.S., L.H.M., L.H.B. and M. Kretzler contributed the NEPTUNE vali- Biobank, a member of the National Biobank of Korea, financed by the Ministry of Health dation cohort. R.J.F.L. and E.E.K. contributed the control genotype data for the PAGE and Welfare, Republic of Korea. P.B. and P.H. acknowledge financial support from MRC cohort. All authors have read and approved the final version of manuscript. project “Autoimmunity in Membranous Nephropathy”, grant MR/J010847/1 which funded the sample collection from MN patients across the UK. P.B., P.H. and S.H. acknowledge support from Manchester Academic Health Science Centre (MAHSC 186/ Competing interests The authors declare no competing interests. 200), the Greater Manchester Local Clinical Research Network and Kidneys for Life Charity for supporting research in MN in Manchester. We are grateful to the MENTOR study (clinical trials no. NCT01180036), for contributing blood samples of trial parti- Additional information cipants. The UK cohort was supported in part by grants from the David and Elaine Supplementary information is available for this paper at https://doi.org/10.1038/s41467- Potter Charitable Foundation (to S.H.P. and R.K.), St Peter’s Trust for Kidney, Bladder 020-15383-w. and Prostate Research (to D.B., H.C.S., S.H.P. and R.K.), Kids Kidney Research UK and Kidney Research UK (to D.B. and R.K.). The Italian cohorts were supported by the Correspondence and requests for materials should be addressed to J.X., R.K. or K.K. Italian Ministry of Health grant GR-2011-02350438 (G.Z., S.G.) and the Department of Excellence Grant 2018–2022 funded by the Italian Ministry of Education for the Peer review information Nature Communications thanks Yang Luo and the other, Department of Medical Sciences of the University of Turin (A.A.). The recruitment of anonymous, reviewer(s) for their contribution to the peer review of this work. Peer Polish cases was sponsored by the Polish Kidney Genetics Network (POLYGENES), a reviewer reports are available. collaborative effort between Columbia University and Poznań University of Medical Sciences, Poland. The full list of POLYGENES collaborators can be found in the Sup- Reprints and permission information is available at http://www.nature.com/reprints plementary Materials. The GCKD (German Chronic Kidney Disease) study was funded by grants from the German Ministry of Education and Research (BMBF, No. 01ER0804) Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in and the KfH Foundation for Preventive Medicine, with genotyping supported by Bayer published maps and institutional affiliations. Pharma AG. The list of GCKD investigators can be found in the Supplementary Materials. The work of M.W. and A.K. was funded by the CRC 1140 Initiative and by KO 3598/3–1 and CRC 992 (A.K.) of the German Research Foundation. The work of E.H. Open Access This article is licensed under a Creative Commons and R.A.K.S. was funded by the CRC 1192 from the German Research Foundation Attribution 4.0 International License, which permits use, sharing, (Projects B1 and C1). P.R. is a recipient of European Research Council ERC-2012- adaptation, distribution and reproduction in any medium or format, as long as you give ADG_20120314 grant 322947, 7th Framework Programme of the European Community appropriate credit to the original author(s) and the source, provide a link to the Creative contract 2012–305608 (European Consortium for High-Throughput Research in Rare Commons license, and indicate if changes were made. The images or other third party Kidney Diseases), and the National Research Agency grant MNaims (ANR-17- CE17- 0012-01). The Dutch studies were supported by grants from the Dutch Kidney Foun- material in this article are included in the article’s Creative Commons license, unless dation to JMH and JFW (Nierstichting Nederland grant OW08 and grant KJPB11.021). indicated otherwise in a credit line to the material. If material is not included in the We would like to thank the Population Architecture Using Genomics and Epidemiology article’s Creative Commons license and your intended use is not permitted by statutory (PAGE) consortium, funded by the National Human Genome Research Institute regulation or exceeds the permitted use, you will need to obtain permission directly from (NHGRI) with co-funding from the NIMHD, for providing population controls for this the copyright holder. To view a copy of this license, visit http://creativecommons.org/ study. For full acknowledgment of the PAGE consortium, please see Supplementary licenses/by/4.0/. Materials. The funding sources were not involved in the study design, collection, analysis, and interpretation of data, writing of the report, or in the decision to submit the paper for © The Author(s) 2020 publication. 16 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w ARTICLE Jingyuan Xie1,77 ✉, Lili Liu 2,77, Nikol Mladkova2,77, Yifu Li2, Hong Ren1, Weiming Wang1, Zhao Cui3,4,5, Li Lin1, Xiaofan Hu1, Xialian Yu1, Jing Xu1, Gang Liu3,4,5, Yasar Caliskan6, Carlo Sidore7, Olivia Balderes2, Raphael J. Rosen2, Monica Bodria2,8, Francesca Zanoni2,9, Jun Y. Zhang 2, Priya Krithivasan2, Karla Mehl2, Maddalena Marasa 2, Atlas Khan2, Fatih Ozay 2, Pietro A. Canetta 2, Andrew S. Bomback 2, Gerald B. Appel2, Simone Sanna-Cherchi 2, Matthew G. Sampson10, Laura H. Mariani11,12, Agnieszka Perkowska-Ptasinska13, Magdalena Durlik13, Krzysztof Mucha 14,15, Barbara Moszczuk14, Bartosz Foroncewicz14, Leszek Pączek14,15, Ireneusz Habura16, Elisabet Ars17, Jose Ballarin17, Laila-Yasmin Mani18, Bruno Vogt 18, Savas Ozturk19, Abdülmecit Yildiz20, Nurhan Seyahi 21, Hakki Arikan22, Mehmet Koc 22, Taner Basturk23, Gonca Karahan24, Sebahat Usta Akgul 24, Mehmet Sukru Sever6, Dan Zhang25, Domenico Santoro 26, Mario Bonomini27, Francesco Londrino28, Loreto Gesualdo29, Jana Reiterova30, Vladimir Tesar 30, Claudia Izzi31,32, Silvana Savoldi33, Donatella Spotti34, Carmelita Marcantoni35, Piergiorgio Messa9, Marco Galliani36, Dario Roccatello37, Simona Granata38, Gianluigi Zaza 38, Francesca Lugani39, GianMarco Ghiggeri39, Isabella Pisani8, Landino Allegri8, Ben Sprangers40,41, Jin-Ho Park 42, BeLong Cho42,43, Yon Su Kim44,45, Dong Ki Kim45,46, Hitoshi Suzuki 47, Antonio Amoroso 48, Daniel C. Cattran49, Fernando C. Fervenza50, Antonello Pani51, Patrick Hamilton52, Shelly Harris52, Sanjana Gupta53, Chris Cheshire53, Stephanie Dufek 53, Naomi Issler53, Ruth J. Pepper53, John Connolly53, Stephen Powis53, Detlef Bockenhauer53, Horia C. Stanescu53, Neil Ashman54, Ruth J.F. Loos 55,56,57, Eimear E. Kenny55,58,59, Matthias Wuttke 60, Kai-Uwe Eckardt61,62, Anna Köttgen 60, Julia M. Hofstra63, Marieke J.H. Coenen64, Lambertus A. Kiemeney65, Shreeram Akilesh 66, Matthias Kretzler 11, Lawrence H. Beck67, Benedicte Stengel68,69, Hanna Debiec70, Pierre Ronco 70,71, Jack F.M. Wetzels 63, Magdalena Zoledziewska7, Francesco Cucca 7, Iuliana Ionita-Laza72, Hajeong Lee45,46, Elion Hoxha73, Rolf A.K. Stahl73, Paul Brenchley74, Francesco Scolari31,75, Ming-hui Zhao 3,4,5,76, Ali G. Gharavi 2, Robert Kleta 53,77 ✉, Nan Chen1,77 & Krzysztof Kiryluk 2,77 ✉ 1 Department of Nephrology, Institute of Nephrology, Shanghai Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 2 Department of Medicine, Division of Nephrology, Columbia University, College of Physicians & Surgeons, New York, USA. 3Renal Division, Department of Medicine, Peking University First Hospital, Beijing, China. 4Institute of Nephrology, Peking University, Beijing, China. 5Key Laboratory of Renal Disease, Ministry of Health of China, and Key Laboratory of CKD Prevention and Treatment, Ministry of Education of China, Beijng, China. 6Division of Nephrology, Department of Internal Medicine, Istanbul School of Medicine, Istanbul University, Istanbul, Turkey. 7Istituto di Ricerca Genetica e Biomedica, Consiglio Nazionale delle Ricerche, Monserrato, Cagliari, Italy. 8Department of Medicine and Surgery, University of Parma, Parma, Italy. 9Nephrology Dialysis and Kidney Transplant Unit, Fundazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Università degli studi di Milano, Milan, Italy. 10Department of Pediatrics-Nephrology, University of Michigan School of Medicine, Ann Arbor, MI, USA. 11 Division of Nephrology, Department of Medicine, University of Michigan, Ann Arbor, MI, USA. 12Arbor Research Collaborative for Health, Ann Arbor, MI, USA. 13Department of Transplantology, Nephrology and Internal Diseases, Medical University of Warsaw, Warsaw, Poland. 14 Department of Immunology, Transplantology and Internal Diseases, Medical University of Warsaw, Warsaw, Poland. 15Institute of Biochemistry and Biophysics, Polish Academy of Sciences, Warsaw, Poland. 16Department of Nephrology, University Hospital of Karol Marcinkowski in Zielona Góra, Zielona Góra, Poland. 17Molecular Biology Laboratory and Nephrology Department, Fundació Puigvert, Instituto de Investigaciones Biomédicas Sant Pau, Universitat Autònoma de Barcelona, REDINREN, IISCIII, Barcelona, Spain. 18Department of Nephrology and Hypertension, Bern University Hospital, University of Bern, Bern, Switzerland. 19Nephrology Clinic, Haseki Training and Research Hospital, Istanbul, Turkey. 20 Department of Nephrology, Uludag University Faculty of Medicine, Bursa, Turkey. 21Division of Nephrology, Department of Internal Medicine, Cerrahpasa Medical Faculty, Istanbul University - Cerrahpasa, Istanbul, Turkey. 22Division of Nephrology, Department of Internal Medicine, Marmara University School of Medicine, Istanbul, Turkey. 23Department of Nephrology, Sisli Hamidiye Etfal Training and Research Hospital, Istanbul, Turkey. 24Department of Medical Biology, Istanbul School of Medicine, Istanbul University, Istanbul, Turkey. 25Department of Nephrology, Xin Hua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China. 26Department of Cinical and Experimental Medicine, Unit of Nephrology, University of Messina, Messina, Italy. 27Department of Medicine, University of Chieti-Pescara, SS. Annunziata Hospital, Chieti, Italy. 28S. Andrea Hospital, La Spezia, Italy. 29University of Bari, Bari, Italy. 30Department of Nephrology, 1st Faculty of Medicine and General University Hospital, Charles University, Prague, Czech Republic. 31Second Division of Nephrology, ASST-Spedali Civili di Brescia Presidio di Montichiari, Brescia, Italy. 32Department of Obstetrics and Gynecology, ASST Spedali Civili di Brescia, Brescia, Italy. 33Unit of Nephrology and Dialysis ASL TO4, Cirié, Turin, Italy. 34San Raffaele Hospital, Milan, Italy. 35Cannizzaro Hospital, Catania, Italy. 36Sandro Pertini Hospital, Rome, Italy. 37San Giovanni Bosco Hospital (ERK-net Member) and University of Turin, Turin, Italy. 38Renal Unit, Department of Medicine, NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications 17 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15383-w University of Verona, Verona, Italy. 39Division of Nephrology, Dialysis, Transplantation, IRCCS Giannina Gaslini, Genoa, Italy. 40Department of Microbiology and Immunology, Laboratory of Molecular Immunology, Rega Institute, KU, Leuven, Belgium. 41Department of Nephrology, University Hospitals Leuven, Leuven, Belgium. 42Department of Family Medicine, Seoul National University College of Medicine and Seoul National University Hospital, Seoul, Korea. 43Institute on Aging, Seoul National University College of Medicine, Seoul, Korea. 44Biomedical Sciences, Seoul National University College of Medicine, Seoul, Korea. 45Kidney Research Institute, Seoul National University College of Medicine, Seoul, Korea. 46Internal Medicine, Seoul National University College of Medicine, Seoul, Korea. 47Department of Nephrology, Juntendo University Faculty of Medicine, Tokyo, Japan. 48Department of Medical Sciences, University of Torino and Immunogenetics and Transplant Biology Service, University Hospital “Città della Salute e della Scienza di Torino”, Turin, Italy. 49Department of Nephrology, University of Toronto, Toronto General Hospital, Toronto, ON, Canada. 50Division of Nephrology and Hypertension, Mayo Clinic, Rochester, MN, USA. 51Department of Nephrology and Dialysis, G. Brotzu Hospital, Cagliari, Italy. 52Manchester Institute of Nephrology and Transplantation, Manchester University Hospitals NHS Trust, Manchester, UK. 53 Department of Nephrology, Division of Medicine, University College London, London, UK. 54Renal Unit, Royal London Hospital, Barts Health, Whitechapel, London, UK. 55The Charles Bronfman Institute for Personalized Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA. 56The Genetics of Obesity and Related Metabolic Traits Program, The Icahn School of Medicine at Mount Sinai, New York, NY, USA. 57The Mindich Child Health and Development Institute, Icahn School of Medicine at Mount Sinai, New York, NY, USA. 58Department of Genetics and Genomic Sciences, Mount Sinai Health System, New York, NY, USA. 59Center for Population Genomic Health, Icahn School of Medicine at Mount Sinai, New York, NY, USA. 60Institute of Genetic Epidemiology, Dep. of Biometry, Epidemiology, and Medical Bioinformatics, Faculty of Medicine and Medical Center – University of Freiburg, Freiburg, Germany. 61Department of Nephrology and Medical Intensive Care, Charité – Universitätsmedizin Berlin, Berlin, Germany. 62Department of Nephrology and Hypertension, Friedrich-Alexander-Universität, Erlangen, Germany. 63 Department of Nephrology, Radboud University Medical Center, Nijmegen, The Netherlands. 64Department of Human Genetics, Radboud University Medical Centre, Radboud Institute for Health Sciences, Nijmegen, The Netherlands. 65Department of Epidemiology, Biostatistics & HTA, Radboud University Medical Center, Nijmegen, The Netherlands. 66Department of Anatomic Pathology, University of Washington, Seattle, USA. 67 Department of Medicine, Renal Section, Boston University School of Medicine and Boston Medical Center, Boston, MA, USA. 68Institut National de la Santé et de la Recherche Médicale, Centre for Research in Epidemiology and Population Health, Villejuif, France. 69University Paris-Sud, Villejuif, France. 70Sorbonne Université, Pierre and Marie Curie University Paris 06, Paris, France. 71Institut National de la Santé et de la Recherche Médicale, Unité Mixte de Recherche (UMR) 1155, Paris, France. 72Department of Biostatistics, Mailman School of Public Health, Columbia University, New York, NY, USA. 73III Department of Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany. 74Faculty of Biology, Medicine, Health, University of Manchester, Manchester, UK. 75University of Brescia, Brescia, Italy. 76Peking-Tsinghua Center for Life Sciences, Beijing, China. 77These authors contributed equally: Jingyuan Xie, Lili Liu, Nikol Mladkova, Robert Kleta, Nan Chen, Krzysztof Kiryluk. ✉email: nephroxie@163.com; r.kleta@ucl.ac.uk; kk473@columbia.edu 18 NATURE COMMUNICATIONS | (2020)11:1600 | https://doi.org/10.1038/s41467-020-15383-w | www.nature.com/naturecommunications