BASIC RESEARCH www.jasn.org

Genetic Analysis of 400 Patients Refines Understanding and Implicates a New Gene in Atypical Hemolytic Uremic Syndrome

Fengxiao Bu,1,2 Yuzhou Zhang,2 Kai Wang,3 Nicolo Ghiringhelli Borsa,2 Michael B. Jones,2 Amanda O. Taylor,2 Erika Takanami,2 Nicole C. Meyer,2 Kathy Frees,2 Christie P. Thomas,4 Carla Nester,2,4,5 and Richard J.H. Smith2,4,5

1Medical Genetics Center, Southwest Hospital, Chongqing, China; and 2Molecular Otolaryngology and Renal Research Laboratories, 3College of Public Health, 4Division of Nephrology, Department of Internal Medicine, Carver College of Medicine, and 5Department of Pediatrics, Carver College of Medicine, University of Iowa, Iowa City, Iowa

ABSTRACT Background Genetic variation in complement genes is a predisposing factor for atypical hemolytic uremic syndrome (aHUS), a life-threatening thrombotic microangiopathy, however interpreting the effects of genetic variants is challenging and often ambiguous. Methods We analyzed 93 complement and coagulation genes in 400 patients with aHUS, using as controls 600 healthy individuals from Iowa and 63,345 non-Finnish European individuals from the Genome Aggre- gation Database. After adjusting for population stratification, we then applied the Fisher exact, modified Poisson exact, and optimal unified sequence kernel association tests to assess gene-based variant burden. We also applied a sliding-window analysis to define the frequency range over which variant burden was significant. Results We found that patients with aHUS are enriched for ultrarare coding variants in the CFH, C3, CD46, CFI, DGKE,andVTN genes. The majority of the significance is contributed by variants with a minor allele frequency of ,0.1%. Disease-related variants tend to occur in specific complement protein domains of FH, CD46, and C3. We observed no enrichment for multiple rare coding variants in gene-gene combinations. Conclusions In known aHUS-associated genes, variants with a minor allele frequency .0.1% should not be considered pathogenic unless valid enrichment and/or functional evidence are available. VTN,which encodes vitronectin, an inhibitor of the terminal complement pathway, is implicated as a novel aHUS- associated gene. Patients with aHUS are not enriched for multiple rare variants in complement genes. In aggregate, these data may help in directing clinical management of aHUS.

J Am Soc Nephrol 29: 2809–2819, 2018. doi: https://doi.org/10.1681/ASN.2018070759

Atypical hemolytic uremic syndrome (aHUS) complement-mediated disease, aHUS develops in defines a spectrum of thrombotic microangiopa- people carrying predisposing genetic abnormalities thies (TMAs) characterized by hemolytic anemia, thrombocytopenia, and acute renal injury not Received July 25, 2018. Accepted September 12, 2018. Escherichia coli caused by Shiga toxin-producing Published online ahead of print. Publication date available at 1,2 or ADAMTS13 deficiency. It is ultrarare, with www.jasn.org. an incidence of approximately 0.5 per million per Correspondence: Dr. Richard J.H. Smith, Molecular Otolaryn- year and until the introduction of eculizumab, a gology and Renal Research Laboratories, University of Iowa, 285 humanized mAb against C5 that blocks the termi- Newton Road, 5270 CBRB, Iowa City, IA 52242. Email: richard- nal pathway of the complement cascade, it carried a [email protected] very poor prognosis.3,4 As the quintessential Copyright © 2018 by the American Society of Nephrology

J Am Soc Nephrol 29: 2809–2819, 2018 ISSN : 1046-6673/2912-2809 2809 BASIC RESEARCH www.jasn.org in complement genes after exposure to a host of triggering/ Significance Statement causal events that include infection, drugs, malignancy, trans- plantation, and pregnancy.2 Although atypical hemolytic uremic syndrome (aHUS) is caused by Genetic studies in patients with a clinical diagnosis of complement dysregulation, in half of patients, mutations are not fi aHUS identify mutations in alternative pathway-related genes identi ed in complement genes. The authors screened 400 patients with aHUS for variation in 93 complement and coagulation genes, 5–9 in up to half of cases. The list of extensively reported aHUS finding that patients with aHUS are more likely than controls to carry genes includes CFH (implicated in approximately 25% of rare coding variants in CFH, C3, CD46, CFI,andDGKE, but not in patients), CD46 (approximately 10%), C3 (approximately CFB, PLG, and THBD. They also demonstrate VTN (a gene not 6%), CFI (approximately 6%), CFB (approximately 2%), previously identified as aHUS-related) as enriched in patients, and fi THBD (approximately 2%), and a noncomplement exception, highlight speci c protein domains in CFH, C3, and CD46 as aHUS- related. They propose a minor allele frequency threshold of 0.1% DGKE 2 (approximately 3%). Autoantibodies against factor H for a variant to be considered as possibly disease relevant. These (FHAA) account for 5%–13% of cases and are associated with data may help in directing clinical management of patients with the absence of both copies of CFHR1.10 If a genetic variation is aHUS. found, it is often considered a predisposing factor rather than a direct cause of aHUS. This distinction reflects the landscape of aHUS is changing and other complement genes like high variability in disease penetrance, with the notable excep- C4BPA,19 C7,20 and CFHR2,21 and noncomplement genes like tion being pathogenic variants in DGKE, which follow an au- CBL, INF2,22–24 MMACHC,25–27 CLU,28 PLG,29 and F12,30 have tosomal recessive inheritance pattern.11–14 been implicated in pathogenesis, we sought to analyze rare cod- The term “primary aHUS” has been proposed by some cli- ing variant burden in a large aHUS cohort in which we control nicians to designate patients with aHUS who carry a genetic for population stratification, integrate two control cohorts, and abnormality in complement genes; however, the distinction be- study a large number of genes. tween primary and secondary aHUS is challenging for several reasons. First, significant complement variants are not identified in a large portion of the patients with aHUS who respond to METHODS terminal complement-blocking treatment.3 Second, in many persons who carry genetic variants in complement genes, the Participants disease does not develop in the absence of triggering events. Patients referred to the Molecular Otolaryngology and Renal Third, aHUS shows variable penetrance, making it difficult to Research Laboratories at the University of Iowa (UI) for a ge- interpret the role many genetic variants play in disease. Fourth, netic evaluation for TMAs were enrolled in this study. Atypical crosstalk between the complement and coagulation pathways HUS was diagnosed by the referring physicians on the basis of makes it challenging to provide an integrated interpretation of the presence of hemolytic anemia, thrombocytopenia, and re- genetic results. Lastly, a TMA lesion is a component of many nal injury, absence of Shiga toxin-producing E. coli,and diseases, which confounds the diagnosis of aHUS.1,2,5 ADAMTS13 activity .10%. Patients were screened for vari- When genetic variants are identified in a patient with aHUS, ants in 93 TMA-related genes (Supplemental Table 1) using a it is critically important to determine their clinical significance, targeted genomic enrichment panel known as CasCADE/ as it has a bearing on long-term anticomplement therapy. Pur- GRP.29,31 The UI control group comprised 600 unrelated Eu- ported disease variants are often defined as such because they ropean Americans (300 males and 300 females) screened by are not detected in a few hundred healthy controls. This ap- the SeqCap EZ whole-exome panel plus custom targeted re- proach has been challenged by publications showing that ul- gions (v1/v2; Roche Sequencing, Pleasanton, CA).32 Both ca- trarare but benign variants are not uncommon.15 For example, ses and UI controls are genetically of European descent, with Marinozzi et al.16 demonstrated experimentally that nine out other ethnicities removed to decrease population stratification of 15 reported CFB gene mutations were unrelated to aHUS (Supplemental Figures 2 and 3). Relatedness analysis was done pathogenesis. Novel variants have also been identified in CFH to eliminate close relatives. A second control cohort of 63,345 that may be unrelated to aHUS.17 Although these reports sup- individuals was accessed by utilizing the non-Finnish Euro- port the value of functional studies to assess variant impact, pean (NFE) population extracted from the Genome Aggrega- these studies are labor intensive and difficult, making func- tion Database (gnomAD).15 The study was approved by the tional assessment impractical in every instance. Institutional Review Board of Carver College of Medicine at Phenotypic variability in presentation adds another layer of UI (IRB ID# 201502804). complexity. A recent collaborative, multi-institution study failed to find enrichment for rare genetic variants in the CFB, THBD, Sequencing and Bioinformatics and PLG genes in patients with aHUS compared with controls Genomic DNA was extracted from whole blood using the Gentra from the Exome Aggregation Consortium database.18 That Puregene Kit (QIAGEN, Valencia, CA) or Chemagic 360 instru- study, however, did not control for population stratification, ment (PerkinElmer Inc., Waltham, MA). Targetedgenomic enrich- which affects rare variant burden (Supplemental Figure 1). In ment was automated using the customized SureSelect Target addition, only a few genes were considered. Because the genetic Enrichment System (Agilent Technologies Inc., Santa Clara, CA)

2810 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2809–2819, 2018 www.jasn.org BASIC RESEARCH and the Zephyr Workstation (PerkinElmer), as described.29,31 En- using the SALSA MLPA Reagent Kit (MRC-Holland, Amsterdam, riched libraries were pooled and sequenced on HiSeq 2500 or The Netherlands), as described.29 Five normal and three positive MiSeq Sequencers (Illumina Inc., San Diego, CA). Using GATK controls were included in each run. best practices, the in-house workflow integrated multiple tools including Trimmomatic (v0.36) for adaptor sequence removal,33 Detection of FHAA BWA mem (v0.7.15) for alignment to the human reference ge- FHAA were detected by ELISA, as described.42 nome GRCh37/hg19,34 Picard Tools (v2.7.1; www.broadinstitute. github.io/picard) MarkDuplicates for PCR duplication removal, Statistical Analyses GATK (v3.7) BaseRecalibrator for recalibration, HaplotypeCaller Statistical analyses were completed using R software (v3.3.2). for single nucleotide variations (SNVs) and insertions/deletions Population stratification was evaluated using EIGENSOFT (indels) calling, and GenotypeGVCFs for joint genotyping on (v7.2.1).43 Relatedness analysis was performed using VCFtools gVCF files.35 FastQC (www.bioinformatics.babraham.ac.uk/ (v0.1.14).44 The association between aHUS and rare coding var- projects/fastqc) and Picard CollectHsMetrics were used to assess iant burden by gene was tested using several complementary sequencing quality. Variants were annotated using Variant Effect methods. One-sided variant burden comparisons between pa- Predictor (v84).36 The Human Gene Mutation Database tients with aHUS and NEF controls were performed using the (HGMD, v2016r4)37 and ClinVar (v20170228)38 were used to Fisher exact and modified Poisson exact tests. The Poisson test query reported disease mutations. Variant frequency in popula- compared expected and observed ratios of summed variant allele tions was annotated using gnomAD. Other operations on BAM numbers in cases to the summed number in both cases and NFE and VCF file were performed using SAMtools,39 BCFtools controls (Equation 1). The expected ratio was derived from all (v1.5),40 and in-house scripts. variants regardless of functional impact, whereas the observed Sequencing quality was controlled at the sample and variant ratio was derived from the subset of rare coding variants. This levels. Low quality was defined as (1) coverage under 1 million ratio is more robust to potential bias within datasets caused by reads, (2) mean depth ,303,(3) theoretical sensitivity of inconsistencies in experimental and analytic workflows (Supple- heterozygous SNP detection ,95% (Supplemental Figure 4, mental Figure 1).45 After adjusting for multiple testing, P,0.05 Supplemental Material), or (4)significant drift in the ratio of and P,0.001 were considered suggestive and significant, alternative/reference allele depth (Supplemental Figure 5). Us- respectively. ing Sanger sequencing data, low-quality SNVs were defined as any of QD (quality-by-depth),4.5, FS (Fisher strand . . score) 60, SOR (strand odds ratio) 3, MQ (root mean SumðACcaseÞ , ,2 Equation 1 : R ¼ square of the mapping quality) 39.5, MQRankSum 28, SumðACcaseÞþSumðACcontrolÞ and ReadPosRankSum,25.0; low-quality small indels were defined as any of QD,4.5, FS.200, and ReadPosRank- Optimal unified sequence kernel association test (SKAT-O), Sum,220.0. CR1, C2, CFHR1, CFHR3,andC4A/C4B were which combines burden and variance-component analyses, removed from analysis because of ambiguous read alignments was performed using the SKAT package (v1.3.0) to compare caused by high sequence homology. patients with aHUS and UI controls.46 Linear weighted kernel, missing cut-off of 0.9 and b-weights were used to calculate the Definition of Variants permutation P-value, with adjustment of covariates including On the basis of sequence ontology, variant functional impact was age, sex, and principal components of population stratifica- defined as: (1) “high” for non-sense, canonical splice-site SNVs tion. P,0.05 was considered significant. and frameshift indels; (2) “moderate” for missense SNVs and A sliding window was applied to identify MAF ranges over in-frame indels; (3) “low” for synonymous SNVs; and (4) which variant enrichment occurs. A window size of 0.001 was “modifier” for noncoding variants (Supplemental Table 2).41 moved from 0 to 0.01 along the MAF axis in a step of 1e26. Rarity was on the basis of minor allele frequency (MAF) in the Variants within the window were collapsed and applied to bur- NFE population, with common, rare, and ultrarare variants de- den tests during each step, generating a set of P-values that re- fined by MAFs of $1%, ,1%, and ,0.01%, respectively. Rare flected enrichment within specific MAF ranges (see animated coding variants with high or moderate function impact were Supplemental Material). Significance was crossvalidated be- included in subsequent analyses. Variant pathogenicity was de- tween NFE and UI control cohorts, as well as across the four termined by absence in public databases, enrichment in the pa- burden tests. tient cohort, and well studied functional impacts (described in Enrichment of gene-gene combinations for rare coding detail in the Supplemental Material). Computational prediction variants was tested by a permutation analysis under the null of variant effect was not applied (Supplemental Figure 6). hypothesis that variant combination is random and indepen- dent. Genotypes were randomly shuffled among samples to Multiplex Ligation-Dependent Probe Amplification generate the number of gene-gene combinations per permu- CopynumbervariationintheCFH-CFHRs genomic region was tation. Empirical P-values were calculated after 100,000 per- evaluated by multiplex ligation-dependent probe amplification mutations. P,0.05 was considered significant.

J Am Soc Nephrol 29: 2809–2819, 2018 Refining Understanding of Hemolytic Uremic Syndrome Genetics 2811 BASIC RESEARCH www.jasn.org

With TMA condition 623 cases With diagnosis of TTP, HUS, and other TMAs 113 cases In house controls (UI control) With diagnosis of aHUS 600 unrelated health subjects 510 cases Failed sequencing QC filtering 7 cases, 1 control Passed sequencing QC filtering 503 cases 599 controls Removed due to stratification 93 outliers Passed population stratification test 410 cases 599 controls Removed due to relatedness 10 related cases Passed relatedness check 400 cases 599 controls

With <0.1% nonsynonymous/splicing variation With homozygous deletion or rearrangement Without genetic variation in CFH, CD46, C3, CFI and CFB in the CFH-CFHRs region in CFH, CD46, C3, CFI, CFB and CFHR1-5 105 patients 46* patients 255 patients

CFH: 36 cases CFHR3-CFHR1 del: 27 cases DGKE: 4 cases (P: 18, LP: 8, VUS: 10) (homozygous/compound het) FHAA: 9 cases CD46: 24 cases PLG: 6 cases (P: 6, LP: 13, VUS: 5) CFHR1 del: 4 cases THBD: 2 cases C3: 18 cases FHAA: 2 cases (P: 9, LP: 5, VUS: 4) FHAA: 5 cases CFHRs fusion: 15 cases CFI: 9 cases Unknown: 238 cases (VUS: 9) CFH+CFHR1: 5 cases DGKE: 2 het cases CFB: 6 cases CFHR1+CFH: 3 cases (P: 1, VUS: 5) CFHR3+CFHR4: 5 cases Combined: 12 cases (P: 11, VUS: 1) Complex: 2 cases

Figure 1. A total of 400 patients and 599 in-house controls were included in this study on the basis of diagnosis and after filtering for quality, population stratification and relatedness. *Five patients carried rare coding variants in complement genes and a homozygous deletion in CFHR3-CFHR1, and one patient carried a variant in CFB, a homozygous deletion in CFHR3-CFHR1 and FHAAs. No other overlap among CFHR fusion genes, FHAAs and rare variants in complement genes was observed (het, heterozygous; LP, likely pathogenic; P, pathogenic; TTP, thrombotic thrombocytopenic purpura; VUS, variant with uncertain significance).

Data Sharing and 599 UI controls (299 males and 300 females). Male pa- Rare variants identified in this patient cohort are publicly avail- tients showed lower median age than females (14.8 versus able in the Database of Complement Gene Variants (http:// 27.3; P=0.01; Supplemental Figure 7). www.complement-db.org/home.php). A total of 2183 variants were identified and passed quality control filtering in the aHUS cohort (Table 1). Included in this number were 637 rare and 95 novel coding variants. There RESULTS were 672 rare and 98 novel coding variants in UI controls. Several common variants in the F5, CFH, CFHR2, CD46, Summary of Genetic Variation and MASP1 genes were significantly associated with aHUS Atypical HUS was diagnosed in 510 out of 623 referrals. In this (Supplemental Table 3). group of 510 patients, there were seven sequencing quality A total of 93 variants with MAF,0.1% were identified on 127 control failures, 93 ethnic outliers, and ten related samples alleles in 105 patients in the known aHUS complement genes (Figure 1). One control was excluded because of low sequenc- (CFH, CD46, C3, CFI,andCFB). CFH carried the highest num- ing quality. Analyses were completed on 400 patients with ber of variants (36 patients), followed by CD46 (24 patients), C3 aHUS (228 males and 182 females, median age of 21.6 years) (18 patients), CFI (nine patients), and CFB (six patients); 12

2812 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2809–2819, 2018 www.jasn.org BASIC RESEARCH

Table 1. Variants identified in patients with aHUS (n=400) other patients. Fifteen different FHR fusion genes were identified and UI controls (n=599) (Supplemental Table 4). With the exception of six patients who SNVs Indels carried variants in complement aHUS genes and were homozy- Filtering Steps CFHR3-CFHR1 Case Control Case Control gous-deleted of , there was no overlap across patient groups. Pass quality control filter 2063 1898 120 113 Nonsynonymous/splice site 756 791 36 29 MAF,1% 607 647 30 25 Rare Variant Burden by Fixed MAFs MAF,0.1% 470 497 27 21 Rare variant burden was examined across genes. Because of MAF,0.01% 329 306 21 15 platform bias on indel calling, only SNVs were analyzed. Sig- Not reported in gnomAD 83 90 12 8 nificant enrichment was confirmed in patients for rare coding variants in CFH, CD46, C3,andCFI genesacrossmultiple burden tests (Table 2). CFB and PLG were significant only patients (11.4%) carried variants in more than one gene. Of the with MAF,0.01 as the threshold (Table 2). Significant enrich- remaining 295 patients, four were homozygous or compound ment in DGKE was confirmed by three tests, with the excep- heterozygous for mutations in DGKE, six carried variants in tion being the SKAT-O test with adjustment for population PLG, and two carried variants in THBD. Homozygous deletion principle components, which may reflect subpopulation strat- of CFHR3-CFHR1 or CFHR1 alone was found in 31 patients; 11 ification in the DGKE group. Patients were not enriched for of these patients had FHAAs. FHAAs were also detected in five rare coding variants in CFHR5 and THBD,norinthe

Table 2. Comparison of rare coding SNV burden in patients with aHUS, UI controls, and NFE controls MAF Adjusted Unadjusted Gene aHUS RV aHUS CN UI RV UI CN NFE RV NFE CN PFisher PPoisson Threshold PSKAT-O PSKAT-O CFH 0.0001 42 800 4 1198 6.3e210a 5.6e213a 512 126,720 1.3e231a 1.8e226a CFH 0.001 48 800 8 1198 3.8e209a 2.6e213a 1188 126,720 3.8e223a 4.0e218a CFH 0.01 54 800 27 1198 5.6e205a 8.0e206a 2486 126,720 1.8e214a 5.7e210a CD46 0.0001 24 800 2 1198 1.2e205a 2.9e208a 215 126,724 1.6e221a 3.3e215a CD46 0.001 25 800 3 1198 8.5e206a 3.3e208a 376 126,724 2.4e217a 2.5e211a CD46 0.01 25 800 3 1198 8.5e206a 3.3e208a 376 126,724 2.4e217a 2.5e211a C3 0.0001 18 800 12 1198 0.008a 0.001a 900 126,728 2.9e205a 3.1e205a C3 0.001 22 800 18 1198 0.07 0.004a 1490 126,728 3.7e204a 4.0e204a C3 0.01 34 800 30 1198 0.06 0.06 3646 126,728 0.026a 0.027a CFI 0.0001 10 800 3 1198 0.01a 0.006a 355 126,712 1.2e204a 2.8e204a CFI 0.001 12 800 4 1198 0.01a 0.006a 695 126,712 0.002a 0.004a CFI 0.01 26 800 15 1198 0.005a 0.003a 1757 126,712 1.0e204a 4.7e204a CFB 0.0001 3 800 2 202 0.95 0.49 342 126,720 0.48 0.50 CFB 0.001 6 800 2 538 0.57 0.24 559 126,720 0.18 0.29 CFB 0.01 30 800 2 538 4.0e24a 2.5e207a 1975 126,720 1.9e205a 8.2e205a THBD 0.0001 1 800 2 1198 0.46 0.78 272 126,612 ,0.99 ,0.99 THBD 0.001 2 800 2 1198 0.92 0.48 489 126,612 0.78 ,0.99 THBD 0.01 10 800 12 1198 0.05 0.10 1804 126,612 0.88 ,0.99 DGKE 0.0001 7 626 2 1198 0.29 0.01 238 126,682 2.4e204a 4.1e204a DGKE 0.001 8 626 4 1198 0.25 0.03 340 126,686 3.7e204a 0.001a DGKE 0.01 13 630 20 1198 0.41 0.34 1824 126,686 0.18 0.26 CFHR5 0.0001 2 800 6 1198 0.36 0.62 422 126,692 ,0.99 ,0.99 CFHR5 0.001 5 800 10 1198 0.49 0.85 985 126,692 0.84 0.84 CFHR5 0.01 15 800 39 1198 0.18 0.11 3425 126,692 0.19 0.19 PLG 0.0001 6 800 2 1198 0.38 0.16 489 126,728 0.14 0.14 PLG 0.001 10 800 5 1198 0.57 0.33 930 126,728 0.09 0.09 PLG 0.01 37 800 40 1198 0.007a 0.002a 3812 126,728 0.01a 0.01a VTN 0.0001 6 800 2 1198 0.22 0.05a 357 126,690 0.03a 0.04a VTN 0.001 9 800 3 1198 0.07 0.02a 619 126,690 0.02a 0.03a VTN 0.01 14 800 12 1198 0.15 0.20 1547 126,690 0.19 0.27 aHUS RV, aggregated allele count of rare coding variants in patients with aHUS; aHUS CN, total chromosome number of patients with aHUS; UI RV, aggregated allele count of rare coding variants in UI controls; UI CN, total chromosome number of UI controls; Adjusted PSKAT-O, P-value of SKAT-O test, correcting for population stratification; Unadjusted PSKAT-O, P-value of SKAT-O test with no adjustment; NFE RV, allele count of rare coding variants in gnomAD NFE controls; NFE CN, total chromosome number of gnomAD NFE controls; PFisher, P-value of burden test using Fisher exact test; PPoisson, P-value of burden test using modified Poisson exact test. aP,0.05.

J Am Soc Nephrol 29: 2809–2819, 2018 Refining Understanding of Hemolytic Uremic Syndrome Genetics 2813 BASIC RESEARCH www.jasn.org

SKAT-O, UI Burden, NFE SKAT-O, UI Burden, NFE 12.5 20 8 10.0 30 6 15 7.5 20 CFH CD46 4 10 5.0 −log10(P) −log10(P) −log10(P) −log10(P) 10 2.5 2 5

0.0 0 0 0

0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 MAF MAF MAF MAF

3 8 3 8

6 6 2 2 C3 4 CFI 4 −log10(P) −log10(P) −log10(P) −log10(P) 1 1 2 2

0 0 0 0

0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 MAF MAF MAF MAF

3 8 3 8

6 6 2 2 CFB 4 DGKE 4 −log10(P) −log10(P) −log10(P) −log10(P) 1 1 2 2

0 0 0 0

0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 MAF MAF MAF MAF

3 8 3 8

6 6 2 2 THBD 4 PLG 4 −log10(P) −log10(P) −log10(P) −log10(P) 1 1 2 2

0 0 0 0

0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 MAF MAF MAF MAF

3 8 3 8

6 6 2 2 CFHR5 4 VTN 4 −log10(P) −log10(P) −log10(P) 1 −log10(P) 1 2 2

0 0 0 0

0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 0 0.005 0.01 MAF MAF MAF MAF

Figure 2. Sliding windows analysis indicates significant enrichment for novel/ultrarare coding variants in CFH, CD46, C3, CFI, DGKE and VTN in patients with aHUS. A window size of 0.001 was moved from 0 to 0.01 in steps of 1e-6. Variants within each window were tested in patients with aHUS and the two control cohorts using four algorithms. Four sets of P-values are shown: black curve, SKAT-O test adjusting for population stratification in UI controls; yellow curve, SKAT-O test without this adjustment; red curve, Fisher exact test in NEF controls; blue curve, modified Poisson exact test in NEF controls; dashed horizontal lines, P50.05 and P50.001. purported aHUS genes C4BPA, C7, MMACHC, CLU, CFHR2, suggestive when compared with NEF controls after adjusting and F12. Of other tested genes, only VTN was enriched for rare for multiple testing. coding variants, nine of which were identified in 400 patients (2.25%; MAF,0.1%), compared with three variants in 599 UI Burden Analysis with MAF Sliding Windows controls (0.5%), and 619 variants in 63,345 NFE controls Sliding-window analysis defined MAF enrichment boundaries (0.97%). These differences for VTN were significant when (Figure 2). Ultrarare/novel coding variants in the CFH, CD46, the patient cohort was compared with UI controls and C3, CFI,andDGKE genes predominantly contributed to the

2814 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2809–2819, 2018 www.jasn.org BASIC RESEARCH

Table 3. aHUS variants reclassified from pathogenic to likely benign on the basis of MAF and lack of differential enrichment in patients Gene HGVS dbSNP aHUS MAF UI Control MAF NFE MAF Max MAF Max Pop CFH c.3148A.T; p.N1050Y rs35274867 0.0288 0.0217 0.0198 0.0271 AFR CD46 c.1058C.T; p.A353V rs35366573 0.0163 0.0150 0.0197 0.0580 FIN CFI c.1534+5G.T rs114013791 0.0088 0.0167 0.0156 0.0156 NFE CFH c.2307C.T rs74842824 0.0125 NA 0.0105 0.0364 ASJ CFH c.2634C.T; p.H878H rs35292876 0.0113 0.0134 0.0100 0.0101 SAS CFH c.2850G.T; p.Q950H rs149474608 0.0063 0.0075 0.0059 0.0178 ASJ C3 c.2203C.T; p.R735W rs117793540 0.0013 0.0017 0.0025 0.0124 ASJ CFI c.1322A.G; p.K441R rs41278047 0.0075 0.0050 0.0024 0.0477 ASJ CFH c.3019G.T; p.V1007L rs534399 0.0000 0.0025 0.0014 0.2715 ASJ CD46 c.1148C.T; p.T383I rs146803767 0.0000 0.0017 0.0010 0.0018 FIN HGVS, standard variant nomenclature by the Human Genome Variant Society; dbSNP, variant identification number from the database for Single Nucleotide Polymorphisms; Max AF, maximum allele frequency across all subpopulations; Max Pop, subpopulation carrying the maximum allele frequency. significance burden. In the CFB and PLG genes, significance carrying variants in complement aHUS genes, 12 patients peaks were contributed by rs45484591 (CFB: p.Glu566Ala, (11.43%) carried variants in gene pairs. A permutation test aHUS MAF=3%, NFE MAF=0.99%, UI MAF=1%) and gives 14.20 as the expected total variant combinations rs4252128 (PLG:p.Ala494Val,aHUSMAF=1.75%,NFE (P=0.92). In all gene pairs (the CFH-CD46 gene pair was MAF=0.39%, UI MAF=0.33%). Significance burden in VTN most frequently observed, n=6), the observed variant combi- was contributed predominantly by eight variants (nine alleles) nations did not differ from the expected variant combinations with MAF,0.022%. On the basis of these data, variants pre- (Table 4). disposing to aHUS had an MAF,0.1%, with few exceptions.

Reclassification of Reported Disease Mutations DISCUSSION Of 329 pathogenic aHUS variants listed by HGMD, 61 were replicated in this study. However, we downgraded ten (3.04%) In this study, we sequenced 93 complement and coagulation variants in CFH, C3, CD46,andCFI reported as pathogenic to genes to test the hypothesis that patients with aHUS are en- likely benign, on the basis of high MAF in the NFE group and riched for rare coding variants in specific genes after correcting other populations and absence of enrichment in patients with for population stratification. We used two control cohorts, a aHUS (Table 3). small UI cohort and a large NFE cohort. The UI cohort enabled us to test for noncausal variants and variant effects in different Enrichment of Disease Variants in Featured Domains directions (i.e., protective versus risk) using the SKAT-O test. Disease variants reported in HGMD and identified in this study The NFE cohort provided power to uncover small genetic showed a distinct pattern of distribution across specific comple- effects in the same direction (i.e., all risk) using the Fisher ment protein domains compared with variants in NFE controls. and Poisson exact tests. Significant enrichment for aHUS-related variants was found in Results from both comparisons agreed that CFH, C3, CD46, short consensus repeat (SCR) 19 (Fisher exact P=0.004; odds CFI,andDGKE are enriched for rare coding variants in pa- ratio [OR], 2.71 with 95% confident interval [95% CI] 1.34 to tients with aHUS (Figure 2). However, in other reported 5.25) and SCR 20 (P53.91e-12; OR, 7.53; 95% CI, 4.16 to 13.57) aHUS-related genes, there was no enrichment, and for two of FH, the thioester-containing domain of C3 (P51.86e-5; OR, genes, CFB and PLG,thesignificance was driven by two asso- 3.66; 95% CI, 2.00 to 6.58), and the vWf type A in FB (P50.002; ciated variants. These results are consistent with the recently OR, 4.58; 95% CI, 1.56 to 14.30) (Figure 3). In the CD46 protein, reported collaborative study looking at 3128 patients with aHUS-related variants localized mainly to the extracellular SCR aHUS from six centers.18 In that study, CFB, THBD, CFHR5, domains (P50.001; OR, 5.84; 95% CI, 1.73 to 30.69). Several and PLG showed insignificant burden for ultrarare variants other protein domains, such as SCR10 in FH (P50.11; OR, 1.98; compared with Exome Aggregation Consortium controls, 95% CI, 0.83 to 4.36) and macroglobulin 6a in C3 (P50.21; OR, suggesting that if these genes make any contribution to 2.08; 95% CI, 0.39 to 7.30), were possibly enriched for aHUS- aHUS, it is small. In the absence of segregation and functional associated variants but further confirmation is required. data, pathogenic mutations in CFB, THBD, CFHR5,andPLG should be reported with caution. No Enrichment for Multiple Rare Coding Variants in In CFH, CD46,andDGKE, it is the extremely rare coding Patients with aHUS variants that contribute to the aHUS-associated enrichment Enrichment for multiple rare coding variants was tested in we observed. These variants show unique distribution pat- CFH, CD46, CFI, C3,andCFB (Table 4). Of 105 patients terns across protein domains compared with the distribution

J Am Soc Nephrol 29: 2809–2819, 2018 Refining Understanding of Hemolytic Uremic Syndrome Genetics 2815 BASIC RESEARCH www.jasn.org

FH 0.0001

0.004 1e-05

MAF 0.002 Density

1e-06 0.000 HGMD Current study gnomAD

SCR1 SCR2 SCR3 SCR4 SCR5 SCR6 SCR7 SCR8 SCR9 SCR10 SCR11 SCR12 SCR13 SCR14 SCR15 SCR16 SCR17 SCR18 SCR19 SCR20

CD46 0.0001 0.0050 1e-05

MAF 0.0025 Density

1e-06 0.0000 HGMD Current study gnomAD

SCR1 SCR2 SCR3 SCR4

C3 0.0001

0.002 1e-05 MAF Density

1e-06 0.000 HGMD Current study gnomAD

MG1 MG2 MG3 MG4 MG5 MG6a LNK ANA MG6b MG7 CUBg TED CUBf MG8 C345C

FI 0.0001

1e-05 0.002 MAF Density

1e-06 0.000 HGMD Current study gnomAD LDL-receptor LDL-receptor Kazal-like SRCR Peptidase S1 class A 1 class A 2

FB 0.0001 0.0050 1e-05

MAF 0.0025 Density

1e-06 0.0000 HGMD Current study gnomAD

SCR1 SCR2 SCR3 VWFA Peptidase S1

Figure 3. Rare coding variants (MAF,0.1%) associated with aHUS show specific distribution patterns across protein domains in complement genes. Reported aHUS mutations from HGMD and coding variants from the current aHUS cohort have been mapped to their corresponding protein domain in each complement gene. As compared with the NFE cohort from the gnomAD database, aHUS- related variants accumulate in the last two SCRs of FH, four SCRs of CD46, the TED of C3, and the vWf type A domain of FB. Density curves (red for HGMD and variants identified in this study; blue for NFE variants) were approximated using Gaussian kernel density estimation.

2816 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2809–2819, 2018 www.jasn.org BASIC RESEARCH

Table 4. aHUS gene-gene pairs carrying rare coding variants More than one rare coding variant was identified in 11.43% Observed Expected of patients (expected rate, 13.5%), with lower rates reported by et al. 6 et al. CFH Gene 1 Gene 2 G1/G2 G1/G2 G1/G2 both Bresin (7.14%) and Noris (7.92%, only , G1 G2 P-Value CD46 CFI 11 Neither Both Both ,and considered), meaning that patients with aHUS are not enriched for multiple rare variants in comple- CFH CD46 40 26 33 6 3.1 0.08 5–9,11 CFH CFI 44 10 49 2 2.8 0.6 ment genes (Table 4). In addition, of the 329 variants CFH C3 44 19 40 2 4.1 0.93 previously implicated in aHUS, ten variants above an MAF CFH CFB 46 6 53 0 0.7 ,0.99 of 0.1% showed no enrichment in patients compared with CD46 CFI 31 11 62 1 1.6 0.65 controls, and are likely benign (Table 3). CD46 C3 31 20 53 1 2.3 0.89 By screening all complement and coagulation genes, we CD46 CFB 32 6 67 0 1.2 0.3 identified significant enrichment for rare coding variants in CFI C3 12 21 72 0 1.3 ,0.99 VTN. VTN encodes vitronectin, a multifunctional protein CFI CFB 12 6 87 0 0.2 ,0.99 abundant in serum and the extracellular matrix that plays a , C3 CFB 21 6 78 0 0.4 0.99 wide role in regulating complement activation, coagulation, G1, number of patients only carrying coding variant in gene 1; G2, number of fibrinolysis, wound healing, and cell adhesion.49,50 It inhibits patients only carrying coding variant in gene 2. the terminal complement cascade by blocking the formation of membrane attack complex and has been reported in im- of all rare variants from NFE controls (Figure 2). The affected mune-complex GN.51–53 We identified nine alleles carrying domains include SCRs 19 and 20 in FH, the thioester-containing eight rare missense variants in eight patients; all variants had domain in C3, vWf type A in FB, and all SCRs in CD46 (Figure 2). MAFs,0.025% (Table 5). Of these eight patients, one also Variant enrichment in the SCRs of CD46 is likely due to dy- carried FHAAs and another carried a fusion protein. namic alternative splicing in exons coding C-terminal do- An intuitive hypothesis in the pathophysiology of aHUS is mains, which increases the tolerance of this region of the that variants in vitronectin impair its function and thereby gene to truncating variations.47,48 Examples of ultrarare repli- compromise negative regulation of the terminal complement cable variants in our aHUS cohort include CFH c.3644G.Ap. pathway and increase the risk of aHUS. VTN variants identified R1215Q (six patients), C3 c.188C.T p.P63L (three patients), in controls included 35 non-sense and 587 missense variants and CD46 c.725T.G p.F242C (three patients), none of which (one non-sense and two missense variants in the UI controls, was present in .138,000 gnomAD controls. Variants with an and 34 non-sense and 585 missense variants in the 63,345 NFE MAF.0.1% are unlikely to be disease related. controls), consistent with tolerance to inferred loss of function Rare variants in CFH, C3, CD46, CFI,andCFB were iden- (from Exome Aggregation Consortium, pLI (loss-of-function tified in 105 patients (26.3%), which is lower than the antic- intolerance)=0.00).15 The mouse homozygous for the targe- ipated rate based on other reports.6,7,18 This difference reflects ted deletion of vitronectin appears phenotypically normal,54 the inclusion of “secondary” aHUS, the implementation of but minor changes are seen in thrombi stabilization, delay of relatedness analysis, and the use of a lower MAF (,0.1%). vessel occlusion, and platelet aggregation.55 The vitronectin In a study by Osborne et al., 39% of patients with aHUS har- variants from our aHUS cohort distributed across multiple bored variants with an MAF,1.0%.18 Applying this MAF to domains, which interact with many serum proteins.49 As our cohort increases the rare variant carrier rate to 36% (144 such, it is possible that VTN variation increases the risk of patients), suggesting that lowering the MAF had a major im- aHUS rather than directly causing the disease, and that the pact. We recommend reviewing archived data and applying an association with aHUS may underlie a mechanism that is more MAF of 0.1% to refine the genetic diagnosis of aHUS. complex than merely enhancing complement dysregulation.

Table 5. Rare coding variants in VTN identified in patients with aHUS Chromosome Position HGVS dbSNP AF aHUS AF NFE AF All Domain chr17 26,694,473 c.1354C.T; p.Arg452Trp rs560780885 0.0025 2.22e24 1.12e24 Hemopexin chr17 26,694,960 c.1100A.G; p.Lys367Arg 0.00125 0.00 4.42e26 chr17 26,695,917 c.802C.T; p.Arg268Trp rs564459012 0.00125 2.37e25 7.59e25 Hemopexin chr17 26,696,628 c.429T.A; p.His143Gln 0.00125 0.00 0.00 chr17 26,696,693 c.364G.T; p.Ala122Ser rs2227741 0.00125 1.81e25 1.23e25 chr17 26,696,830 c.227C.T; p.Thr76Met rs150757499 0.00125 1.51e24 1.01e24 chr17 26,696,968 c.164C.T; p.Thr55Met rs147146251 0.00125 1.58e25 7.58e25 Somatomedin B chr17 26,696,972 c.160T.C; p.Tyr54His 0.00125 0.00 3.23e25 Somatomedin B HGVS, standard variant nomenclature by the Human Genome Variant Society; dbSNP, variant identification number from the database for Single Nucleotide Polymorphisms; AF aHUS, allele frequency in aHUS patients; AF NFE, allele frequency in non-Finnish European controls; AF All, allele frequency in allGenome Aggregation Database (gnomAD) subjects.

J Am Soc Nephrol 29: 2809–2819, 2018 Refining Understanding of Hemolytic Uremic Syndrome Genetics 2817 BASIC RESEARCH www.jasn.org

In conclusion, after adjusting for population stratification syndrome: A nationwide French series comparing children and adults. in patients and controls, we demonstrate domain-specificen- Clin J Am Soc Nephrol 8: 554–562, 2013 richment for ultrarare coding variants in the CFH, C3, CD46, 8. Loirat C, Fakhouri F, Ariceta G, Besbas N, Bitzan M, Bjerre A, et al.: HUS In- CFI DGKE fi ternational: An international consensus approach to the management of atypical ,and genes. As a guide to variant classi cation, we hemolytic uremic syndrome in children. Pediatr Nephrol 31: 15–39, 2016 show that pathogenicity is unlikely when variants have an 9. Jokiranta TS: HUS and atypical HUS. Blood 129: 2847–2856, 2017 MAF.0.1%. The impact of more common variants should 10. Durey MA, Sinha A, Togarsimalemath SK, Bagga A: Anti-complement- be interpreted with caution. We also show that patients with factor H-associated glomerulopathies. Nat Rev Nephrol 12: 563–578, 2016 aHUS are not enriched for rare variants in multiple comple- 11. Noris M, Caprioli J, Bresin E, Mossali C, Pianetti G, Gamba S, et al.: VTN Relative role of genetic complement abnormalities in sporadic and ment genes. Lastly, our results identify as another gene familial aHUS and their impact on clinical phenotype. Clin J Am Soc associated with aHUS. In aggregate, these data should help to Nephrol 5: 1844–1859, 2010 refine the long-term clinical management of patients with 12. Esparza-Gordillo J, Goicoechea de Jorge E, Buil A, Carreras Berges L, aHUS. López-Trascasa M, Sánchez-Corral P, et al.: Predisposition to atypical hemolytic uremic syndrome involves the concurrence of different sus- ceptibility alleles in the regulators of complement activation gene cluster in 1q32. Hum Mol Genet 14: 703–712, 2005 13. Frémeaux-Bacchi V, Miller EC, Liszewski MK, Strain L, Blouin J, Brown ACKNOWLEDGMENTS AL, et al.: Mutations in complement C3 predispose to development of atypical hemolytic uremic syndrome. Blood 112: 4948–4952, 2008 We are grateful to the many clinicians who have entrusted us with 14. Liszewski MK, Leung MK, Schraml B, Goodship TH, Atkinson JP: fi genetic and complement function testing of their patients. Modeling how CD46 de ciency predisposes to atypical hemolytic uremic syndrome. Mol Immunol 44: 1559–1568, 2007 F.B. and R.J.H.S. conceived the study and wrote the manuscript. 15. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, F.B., N.G.B., M.B.J., A.O.T., E.T., and K.F. performed sequencing- et al.: Exome Aggregation Consortium: Analysis of protein-coding related experiments. Y.Z. and N.C.M. performed functional assays. genetic variation in 60,706 humans. Nature 536: 285–291, 2016 F.B. performed statistical analysis, with contributions by K.W. 16. Marinozzi MC, Vergoz L, Rybkine T, Ngo S, Bettoni S, Pashov A, et al.: R.J.H.S., C.N., and C.P.T. performed clinical and genetic diagnoses, Complement factor B mutations in atypical hemolytic uremic syndrome- disease-relevant or benign? J Am Soc Nephrol 25: 2053–2065, 2014 with contributions by F.B., Y.Z., and N.G.B. 17. Merinero HM, García SP, García-Fernández J, Arjona E, Tortajada A, This study was supported in part by the Foundation for Children Rodríguez de Córdoba S: Complete functional characterization of dis- with Atypical HUS and an unrestricted award by the Navikas Family ease-associated genetic variants in the complement factor H gene. Renal Research Fund. Kidney Int 93: 470–481, 2018 18. Osborne AJ, Breno M, Borsa NG, Bu F, Frémeaux-Bacchi V, Gale DP, et al.: Statistical validation of rare complement variants provides in- sights into the molecular basis of atypical hemolytic uremic syndrome DISCLOSURES and C3 glomerulopathy. J Immunol 200: 2464–2478, 2018 None. 19. Blom AM, Bergström F, Edey M, Diaz-Torres M, Kavanagh D, Lampe A, et al.: A novel non-synonymous polymorphism (p.Arg240His) in C4b- binding protein is associated with atypical hemolytic uremic syndrome REFERENCES and leads to impaired alternative pathway cofactor activity. JImmunol 180: 6385–6391, 2008 20. Poolpol K, Gadner B, Neururer S, Mellmann A, Karch H, Orth D, et al.: Do 1. Nester CM, Barbour T, de Cordoba SR, Dragon-Durey MA, Fremeaux- complement factor H 402Y and C7 M allotypes predispose to (typical) Bacchi V, Goodship TH, et al.: Atypical aHUS: State of the art. Mol haemolytic uraemic syndrome? Int J Immunogenet 38: 383–387, 2011 Immunol 67: 31–42, 2015 21. Zhang T, Lu J, Liang S, Chen D, Zhang H, Zeng C, et al.: Comprehensive 2. Goodship TH, Cook HT, Fakhouri F, Fervenza FC, Frémeaux-Bacchi V, analysis of complement genes in patients with atypical hemolytic ure- Kavanagh D, et al.: Conference Participants: Atypical hemolytic uremic syn- mic syndrome. Am J Nephrol 43: 160–169, 2016 drome and C3 glomerulopathy: Conclusions from a “Kidney Disease: Im- 22. Lemaire M, Frémeaux-Bacchi V, Schaefer F, Choi M, Tang WH, Le proving Global Outcomes” (KDIGO) controversies conference. Kidney Int Quintrec M, et al.: Recessive mutations in DGKE cause atypical he- 91: 539–551, 2017 molytic-uremic syndrome. Nat Genet 45: 531–536, 2013 3. Legendre CM, Licht C, Muus P, Greenbaum LA, Babu S, Bedrosian C, et al.: Terminal complement inhibitor eculizumab in atypical hemolytic- 23. Seaby EG, Gilbert RD, Andreoletti G, Pengelly RJ, Mercer C, Hunt D, fi uremic syndrome. N Engl J Med 368: 2169–2181, 2013 et al.: Unexpected ndings in a child with atypical hemolytic uremic 4. Rathbone J, Kaltenthaler E, Richards A, Tappenden P, Bessey A, syndrome: An example of how genomics is changing the clinical di- Cantrell A: A systematic review of eculizumab for atypical haemolytic agnostic paradigm. Front Pediatr 5: 113, 2017 uraemic syndrome (aHUS). BMJ Open 3: e003573, 2013 24. Challis RC, Ring T, Xu Y, Wong EK, Flossmann O, Roberts IS, et al.: 5. Kavanagh D, Goodship TH, Richards A: Atypical hemolytic uremic Thrombotic microangiopathy in inverted formin 2-mediated renal dis- – syndrome. Semin Nephrol 33: 508–530, 2013 ease. JAmSocNephrol28: 1084 1091, 2017 6. Bresin E, Rurali E, Caprioli J, Sanchez-Corral P, Fremeaux-Bacchi V, 25. Koenig JC, Rutsch F, Bockmeyer C, Baumgartner M, Beck BB, Kranz B, Rodriguez de Cordoba S, et al.: European Working Party on Comple- et al.: Nephrotic syndrome and thrombotic microangiopathy caused by ment Genetics in Renal Diseases: Combined complement gene mu- cobalamin C deficiency. Pediatr Nephrol 30: 1203–1206, 2015 tations in atypical hemolytic uremic syndrome influence clinical 26. Kömhoff M, Roofthooft MT, Westra D, Teertstra TK, Losito A, van de Kar phenotype. J Am Soc Nephrol 24: 475–486, 2013 NC, et al.: Combined pulmonary hypertension and renal thrombotic 7. Fremeaux-Bacchi V, Fakhouri F, Garnier A, Bienaimé F, Dragon-Durey microangiopathy in cobalamin C deficiency. Pediatrics 132: e540– MA, Ngo S, et al.: Genetics and outcome of atypical hemolytic uremic e544, 2013

2818 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2809–2819, 2018 www.jasn.org BASIC RESEARCH

27. Bouts AH, Roofthooft MT, Salomons GS, Davin JC: CD46-associated 42. Zhang Y, Meyer NC, Wang K, Nishimura C, Frees K, Jones M, et al.: atypical hemolytic uremic syndrome with uncommon course caused by Causes of alternative pathway dysregulation in dense deposit disease. cblC deficiency. Pediatr Nephrol 25: 2547–2548, 2010 Clin J Am Soc Nephrol 7: 265–274, 2012 28. Ståhl AL, Kristoffersson A, Olin AI, Olsson ML, Roodhooft AM, 43. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich Proesmans W, et al.: A novel mutation in the complement regulator D: Principal components analysis corrects for stratification in genome- clusterin in recurrent hemolytic uremic syndrome. Mol Immunol 46: wide association studies. Nat Genet 38: 904–909, 2006 2236–2243, 2009 44. Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, 29. Bu F, Maga T, Meyer NC, Wang K, Thomas CP, Nester CM, et al.: et al.: 1000 Genomes Project Analysis Group: The variant call format Comprehensive genetic analysis of complement and coagulation and VCFtools. Bioinformatics 27: 2156–2158, 2011 genes in atypical hemolytic uremic syndrome. J Am Soc Nephrol 25: 45. Neale BM, Rivas MA, Voight BF, Altshuler D, Devlin B, Orho-Melander 55–64, 2014 M, et al.: Testing for an unusual distribution of rare variants. PLoS Genet 30. Matsukuma E, Gotoh Y, Kuroyanagi Y, Yamada T, Iwasa M, Yamakawa 7: e1001322, 2011 S, et al.: A case of atypical hemolytic uremic syndrome due to anti- 46. Lee S, Emond MJ, Bamshad MJ, Barnes KC, Rieder MJ, Nickerson DA, factor H antibody in a patient presenting with a factor XII deficiency et al.: NHLBI GO Exome Sequencing Project—ESP Lung Project Team: identified two novel mutations. Clin Exp Nephrol 15: 269–274, 2011 Optimal unified approach for rare-variant association testing with ap- 31. Bu F, Borsa NG, Jones MB, Takanami E, Nishimura C, Hauer JJ, et al.: plication to small-sample case-control whole-exome sequencing High-throughput genetic testing for thrombotic microangiopathies studies. Am J Hum Genet 91: 224–237, 2012 and C3 glomerulopathies. JAmSocNephrol27: 1245–1253, 2016 47. Manchester M, Liszewski MK, Atkinson JP, Oldstone MB: Multiple 32. Goes FS, Pirooznia M, Parla JS, Kramer M, Ghiban E, Mavruk S, et al.: isoforms of CD46 (membrane cofactor protein) serve as receptors for Exome sequencing of familial bipolar disorder. JAMA Psychiatry 73: measles virus. Proc Natl Acad Sci U S A 91: 2161–2165, 1994 590–597, 2016 48. Russell SM, Sparrow RL, McKenzie IF, Purcell DF: Tissue-specificand 33. Bolger AM, Lohse M, Usadel B: Trimmomatic: A flexible trimmer for allelic expression of the complement regulator CD46 is controlled by Illumina sequence data. Bioinformatics 30: 2114–2120, 2014 alternative splicing. Eur J Immunol 22: 1513–1518, 1992 34. Li H, Durbin R: Fast and accurate short read alignment with Burrows- 49. Leavesley DI, Kashyap AS, Croll T, Sivaramakrishnan M, Shokoohmand Wheeler transform. Bioinformatics 25: 1754–1760, 2009 A, Hollier BG, et al.: Vitronectin--master controller or micromanager? 35. McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, IUBMB Life 65: 807–818, 2013 et al.: The Genome Analysis Toolkit: A MapReduce framework for an- 50. Preissner KT: Structure and biological role of vitronectin. Annu Rev Cell alyzing next-generation DNA sequencing data. Genome Res 20: 1297– Biol 7: 275–310, 1991 1303, 2010 51. Sheehan M, Morris CA, Pussell BA, Charlesworth JA: Complement in- 36. McLaren W, Gil L, Hunt SE, Riat HS, Ritchie GR, Thormann A, et al.: The hibition by human vitronectin involves non-heparin binding domains. ensembl variant effect predictor. Genome Biol 17: 122, 2016 Clin Exp Immunol 101: 136–141, 1995 37. Stenson PD, Mort M, Ball EV, Evans K, Hayden M, Heywood S, et al.: 52. Milis L, Morris CA, Sheehan MC, Charlesworth JA, Pussell BA: Vi- The Human Gene Mutation Database: Towards a comprehensive re- tronectin-mediated inhibition of complement: Evidence for different pository of inherited mutation data for medical research, genetic di- binding sites for C5b-7 and C9. Clin Exp Immunol 92: 114–119, 1993 agnosis and next-generation sequencing studies. Hum Genet 136: 53. Podack ER, Preissner KT, Müller-Eberhard HJ: Inhibition of C9 poly- 665–677, 2017 merization within the SC5b-9 complex of complement by S-protein. 38. Landrum MJ, Lee JM, Riley GR, Jang W, Rubinstein WS, Church DM, Acta Pathol Microbiol Immunol Scand Suppl 284: 89–96, 1984 et al.: ClinVar: Public archive of relationships among sequence variation 54. Zheng X, Saunders TL, Camper SA, Samuelson LC, Ginsburg D: Vi- and human phenotype. Nucleic Acids Res 42: D980–D985, 2014 tronectin is not essential for normal mammalian development and 39. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, et al.: 1000 fertility. Proc Natl Acad Sci U S A 92: 12426–12430, 1995 Genome Project Data Processing Subgroup: The Sequence Alignment/ 55. Reheman A, Gross P, Yang H, Chen P, Allen D, Leytin V, et al.: Vi- Map format and SAMtools. Bioinformatics 25: 2078–2079, 2009 tronectin stabilizes thrombi and vessel occlusion but plays a dual role in 40. Li H: A statistical framework for SNP calling, mutation discovery, asso- platelet aggregation. J Thromb Haemost 3: 875–883, 2005 ciation mapping and population genetical parameter estimation from sequencing data. Bioinformatics 27: 2987–2993, 2011 41. Eilbeck K, Lewis SE, Mungall CJ, Yandell M, Stein L, Durbin R, et al.: The Sequence Ontology: A tool for the unification of genome annotations. This article contains supplemental material online at http://jasn.asnjournals. Genome Biol 6: R44, 2005 org/lookup/suppl/doi:10.1681/ASN.2018070759/-/DCSupplemental.

J Am Soc Nephrol 29: 2809–2819, 2018 Refining Understanding of Hemolytic Uremic Syndrome Genetics 2819 Supplemental information

Table of Contents

SUPPLEMENT METHODS ...... ERROR! BOOKMARK NOT DEFINED. Theoretical Heterozygous SNP Sensitivity (THS) ...... 2

Variant Pathogenicity Interpretation ...... 2

Reference ...... 3

SUPPLEMENT TABLES ...... ERROR! BOOKMARK NOT DEFINED. Supplement Table S1. Genes sequenced ...... 4

Supplement Table S2. Sequence ontology terms and functional impact definition ...... 6

Supplement Table S3. Association analysis for common variants ...... 7

Supplement Table S4. FH-FHRs fusion proteins identified in 400 aHUS patients ...... 8 Supplement Table S5. Rare variants with 0.1% < MAF (NFE) < 1% identified in CFH , CD46 , C3 , CFI and CFB in 400 aHUS patients ...... 9

SUPPLEMENT FIGURES ...... ERROR! BOOKMARK NOT DEFINED. Supplement Figure S1. Rare variants in the CFH gene are significantly more abundant in the non- Finnish European (NFE) subpopulation as compared to the Finnish (FIN) subpopulation from gnomAD due to population stratification...... 10 Supplement Figure S2. Cluster analysis shows that UI controls and aHUS cases are mostly similar to NFE...... 11 Supplement Figure S3. Population stratification within aHUS cases and UI controls was used to remove outliers...... 12 Supplement Figure S4. Mean coverage is correlated with theoretical heterozygous SNP sensitivity (HET SNP sensitivity) in patients and controls...... 13 Supplement Figure S5. Six samples with a shifted ratio of ref/alt reads were excluded from the study cohort...... 14 Supplement Figure S6. Prediction score distribution of neutral and pathogenic variants on CFH gene...... 15 Supplement Figure S7. Median age of female patients is significantly lower than that of male patients ...... 16 Supplement Figure S8: Enrichment of ultra-rare variants “contaminates” the result of association analysis with higher MAF cut-off...... 17

SUPPLEMENTAL METHODS

Theoretical Heterozygous SNP Sensitivity (THS)

THS is a quality metric that estimates the theoretical sensitivity to detect heterozygous variants based on coverage distribution and base quality distribution from massively parallel sequencing data. 1,2 Under the assumptions: 1) DNA is diploid; 2) at a HET site with genotype AB, the only possible calls are A and B; 3) there is no reference bias; and 4) coverage distribution P(n) and base quality distribution P(q) are known and statistically independent, the model of HET detection is based on Bernoulli distribution as:

1 2 > ∏ ∏(1 − ) where, n is the depth from the coverage distribution P(n) ; m is the number of true alternate −qj/10 alleles from m ~ binomial (n, 0.5) covering the HET site; ej = 10 is the probability of error, and qj is from the base quality distribution P(q) .

Variant Pathogenicity Interpretation

Pathogenicity of variant was based on absence in large health populations, presence/enrichment in aHUS patients, and functional association.

1. Large health populations refer to the gnomAD database (138,632 subjects).

2. Presence in aHUS patients is defined as: 1) reported in the literature, 2) reported in an aHUS disease mutation database, or 3) observed in our patient cohort.

3. Enrichment in aHUS patients is determined by association analysis in patients and controls with adjustment for population stratification.

4. Functional association is defined as: 1) well-studied functional changes that contribute to aHUS development, 2) truncating protein where loss of function is a known disease mechanism, 3) known disruption of protein structure (e.g. cysteine-related missense variants in SCRs of CFH and CD46 ), or 4) localization in well-defined aHUS-related domains.

Pathogenic is defined as: 1) absent in gnomAD and reported at least once in the literature or an aHUS database and observed at least once in our patient cohort; OR 2) absent in gnomAD and observed at least twice in our patient cohort and with a functional impact.

Likely pathogenic is defined as: 1) absent in gnomAD and observed at least twice in our patient cohort; OR 2) absent in gnomAD and observed at least once in our patient cohort and with a functional impact; OR 3) significantly enriched in our patient cohort compare to the corresponding gnomAD population and with a functional impact. Likely benign is defined as: frequency > 0.1% in any gnomAD population and not enriched in patients and lacking a functional effect.

Benign is defined as: frequency > 1% in any gnomAD population and not enriched in patients.

Reference

1. Yossi Farjoun Jon Bloom: Theoretical HET Sensitivity. https://www.broadinstitute.org/files/shared/mia/theoretical_HET_sensitivity.pdf

2. Kylee Degatano, David Benjamin, Jonathan M. Bloom, Maura Costello, Jason Rose, Kathleen TibbeFs, CharloFe Tolonen, Yossi Farjoun: Optimizing Delivered Sequencing Data with a Theoretical Sensitivity to Heterozygous SNPs. ASHG, 2016. http://www.genomics.broadinstitute.org/data- sheets/POS_OptimizingDeliveredSequencingDataTheoreticalSensitivityHeteroSNPs_ASHG_20 16.pdf

SUPPLEMENTAL TABLES

Supplement Table S1. Genes sequenced

Gene Full Name RefSeq ID

1 A2M Alpha -2-Macroglobulin NM_000014 2 ABCD4 ATP Binding Cassette Subfamily D Member 4 NM_005050 3 ADAMTS13 ADAM Metallopeptidase With Thrombospondin Type 1 Motif 13 NM_139025 4 ADM Adrenomedullin NM_001124 5 ADM2 Adrenomedullin 2 NM_001253845 6 APCS Amyloid P Component, Serum NM_001639 7 C1QA Complement C1q A Chain NM_015991 8 C1QB Complement C1q B Chain NM_000491 9 C1QC Complement C1q C Chain NM_172369 10 C1R Complement Component 1, R Subcomponent NM_001733 11 C1S Complement Component 1, S Subcomponent NM_201442 12 *C2 Complement C2 NM_000063 13 C3 Complement C3 NM_000064 14 C3AR1 Complement C3a Receptor 1 NM_004054 15 *C4A Complement C4A NM_007293 16 *C4B Complement C4B NM_000715 17 C4BPA Complement Component 4 Binding Protein Alpha NM_000715 18 C4BPB Complement Component 4 Binding Protein Beta NM_000716 19 C5 Complement C5 NM_001735 20 C5AR1 Complement C5a Receptor 1 NM_001736 21 C5AR2 Complement C5a Receptor 2 NM_018485 22 C6 Complement C6 NM_000065 23 C7 Complement C7 NM_000587 24 C8A Complement C8 Alpha Chain NM_000562 25 C8B Complement C8 Beta Chain NM_000066 26 C8G Complement C8 Gamma Chain NM_000606 27 C9 Complement C9 NM_001737 28 CD46 Membrane Cofactor Protein NM_002389 29 CD55 Decay Accelerating Factor For Complement NM_000574 30 CD59 Membrane Attack Complex Inhibition Factor NM_000611 31 CFB Complement Factor B NM_001710 32 CFD Complement Factor D NM_001928 33 CFH Complement Factor H NM_000186 34 *CFHR1 Complement Factor H Related 1 NM_002113 35 CFHR2 Complement Factor H Related 2 NM_005666 36 *CFHR3 Complement Factor H Related 3 NM_021023 37 CFHR4 Complement Factor H Related 4 NM_006684 38 CFHR5 Complement Factor H Related 5 NM_030787 39 CFI Complement Factor I NM_000204 40 CFP Complement Factor Properdin NM_002621 41 CLU Clusterin NM_001831 42 COLEC11 Collectin Subfamily Member 11 NM_199235 43 CPN1 Anaphylatoxin Inactivator NM_001308 44 *CR1 Complement C3b/C4b Receptor 1 NM_000651 45 CR2 Complement C3b/C4b Receptor 2 NM_001006658 46 CRP C-Reactive Protein NM_000567 47 DGKE Diacylglycerol Kinase Epsilon NM_003647 48 F10 Coagulation Factor X NM_000504 49 F11 Coagulation Factor XI NM_000128 50 F12 Coagulation Factor XII NM_000505 51 F2 Coagulation Factor II, Thrombin NM_000506 52 F2RL2 Coagulation Factor II Thrombin Receptor Like 2 NM_004101 53 F3 Coagulation Factor III, Tissue Factor NM_001993 54 F5 Coagulation Factor V NM_000130 55 F7 Coagulation Factor VII NM_000131 56 F8 Coagulation Factor VIII, Procoagulant Component NM_000132 57 F9 Coagulation Factor IX NM_000133 58 FCN1 Ficolin 1 NM_002003 59 FCN2 Ficolin 2 NM_004108 60 FCN3 Ficolin 3 NM_003665 61 FGL2 Fibrinogen Like 2 NM_006682 62 IFNG Interferon Gamma NM_000619 63 INF2 Inverted Formin, FH2 And WH2 Domain Containing NM_022489 64 ITGAM Integrin Subunit Alpha M NM_000632 65 KLKB1 Kallikrein B1 NM_000892 66 LMBRD1 LMBR1 Domain Containing 1 NM_018368 67 MAP3K5 Mitogen-Activated Protein Kinase Kinase Kinase 5 NM_005923 68 MASP1 Mannan Binding Lectin Serine Peptidase 1 NM_001879 69 MASP2 Mannan Binding Lectin Serine Peptidase 2 NM_006610 70 MBL2 Mannose Binding Lectin 2 NM_000242 71 MBTPS1 Membrane Bound Transcription Factor Peptidase, Site 1 NM_003791 72 MMACHC Methylmalonic Aciduria CblC Type, With Homocystinuria NM_015506 73 MMADHC Methylmalonic Aciduria CblD Type, With Homocystinuria NM_015702 74 MTR 5-Methyltetrahydrofolate-Homocysteine Methyltransferase NM_000254 75 MTRR 5-Methyltetrahydrofolate -Homocysteine Methyltransferase Reductase NM_002454 76 PHB Prohibin NM_002634 77 PLAT Plasminogen Activator, Tissue NM_000930 78 PLAU Plasminogen Activator, Urokinase NM_002658 79 PLG Plasminogen NM_000301 80 PROC Protein C, Inactivator Of Coagulation Factors Va And VIIIa NM_000312 81 PROS1 Protein S alpha NM_000313 82 PTX3 Pentraxin 3 NM_002852 83 SERPINA1 Serpin Family A Member 1 NM_000295 84 SERPINA5 Serpin Family A Member 5 NM_000624 85 SERPINC1 Serpin Family C Member 1 NM_000488 86 SERPIND1 Serpin Family D Member 1 NM_000185 87 SERPINE1 Serpin Family E Member 1 NM_000602 88 SERPINF2 Serpin Family F Member 2 NM_000934 89 SERPING1 Serpin Family G Member 1 NM_000062 90 THBD Thrombomodulin NM_000361 91 VSIG4 V-Set And Immunoglobulin Domain Containing 4 NM_007268 92 VTN Vitronectin NM_000638 93 VWF von Willebrand Factor NM_000552 Genes marked with asterisk were not included in burden analysis due to ambiguous read mapping Supplement Table S2. Sequence ontology terms and functional impact definition

SO accession SO term IMPACT SO:0001893 transcript_ablation HIGH SO:0001574 splice_acceptor_variant HIGH SO:0001575 splice_donor_variant HIGH SO:0001587 stop_gained HIGH SO:0001589 frameshift_variant HIGH SO:0001578 stop_lost HIGH SO:0002012 start_lost HIGH SO:0001889 transcript_amplification HIGH SO:0001821 inframe_insertion MODERATE SO:0001822 inframe_deletion MODERATE SO:0001583 missense_variant MODERATE SO:0001818 protein_altering_variant MODERATE SO:0001630 splice_region_variant LOW SO:0001626 incomplete_terminal_codon_variant LOW SO:0001567 stop_retained_variant LOW SO:0001819 synonymous_variant LOW SO:0001580 coding_sequence_variant MODIFIER SO:0001620 mature_miRNA_variant MODIFIER SO:0001623 5_prime_UTR_variant MODIFIER SO:0001624 3_prime_UTR_variant MODIFIER SO:0001792 non_coding_transcript_exon_variant MODIFIER SO:0001627 intron_variant MODIFIER SO:0001621 NMD_transcript_variant MODIFIER SO:0001619 non_coding_transcript_variant MODIFIER SO:0001631 upstream_gene_variant MODIFIER SO:0001632 downstream_gene_variant MODIFIER SO:0001895 TFBS_ablation MODIFIER SO:0001892 TFBS_amplification MODIFIER SO:0001782 TF_binding_site_variant MODIFIER SO:0001894 regulatory_region_ablation MODERATE SO:0001891 regulatory_region_amplification MODIFIER SO:0001907 feature_elongation MODIFIER SO:0001566 regulatory_region_variant MODIFIER SO:0001906 feature_truncation MODIFIER SO:0001628 intergenic_variant MODIFIER High and moderate variants were included in the analysis

Source: http://www.ensembl.org/info/genome/variation/predicted_data.html Supplement Table S3. Association analysis for common variants

MAF MAF MAF

Chr dbSNP position Ref Alt Gene Function aHUS UI Control P1 P1 adj OR 1 gnomAD P2 OR 2 1 rs9287090 169510380 A G F5 p.Leu1316Leu 15.58% 28.00% 1.13E-10 6.41E-08 0.47 21.50% 3.85E-05 0.68 1 rs3753396 196695742 G A CFH p.Gln672Gln 28.88% 15.33% 4.32E-13 6.12E-10 2.24 16.71% 1.86E-17 2.02 1 rs1065489 196709774 T G CFH p.Glu936Asp 28.75% 15.33% 8.82E-13 6.25E-10 2.23 16.84% 8.34E-17 1.99 1 rs3828032 196920178 T C CFHR2 intronic 39.38% 28.17% 1.87E-07 5.30E-05 1.66 30.00% 1.85E-08 1.52 1 rs11118580 207959070 C T CD46 intronic 30.00% 21.17% 8.57E-06 1.87E-03 1.60 21.37% 1.29E-08 1.58 3 rs3733001 186938956 T C MASP1 intronic 30.50% 22.83% 1.42E-04 2.36E-02 1.48 25.18% 6.90E-04 1.30 MAF: minor allele frequency

Supplement Table S4. FH-FHRs fusion proteins identified in 400 aHUS patients

Patient ID Fusion Protein (inferred based on MLPA) 1 FH SCR 1-18 + FHR1 SCR 4-5 2 FH SCR 1-18 + FHR1 SCR 4-5 3 FH SCR 1-18 + FHR1 SCR 4-5 4 FH SCR 1-19 + FHR1 SCR 5 5 FH SCR 1-19 + FHR1 SCR 5 6 FHR1 SCR 1-2 + FH SCR 18-20 7 FHR1 SCR 1-2 + FH SCR 18-20 8 FHR1 SCR 1-3 + FH SCR 19-20 9 FHR3 SCR 1-4 + FHR4 SCR 4-5 10 FHR3 SCR 1-4 + FHR4 SCR 4-5 11 FHR3 SCR 1-4 + FHR4 SCR 4-5 12 FHR3 SCR 1-4 + FHR4 SCR 4-5 13 FHR3 SCR 1-4 + FHR4 SCR 4-5 14 Complex (FH and FHR1 involved) 15 Complex (FH, FHR3, FHR1, FHR4 and FHR2 involved)

Supplement Table S5. Rare variants with 0.1% < MAF (NFE) < 1% identified in CFH , CD46 , C3 , CFI and CFB in 400 aHUS patients

aHUS Control NFE Max Max Gene HGVS dbSNP Pathogenicity MAF MAF MAF MAF Pop CFH c.2850G>T; p.Gln950His rs149474608 0.63% 0.75% 0.59% 1.78% ASJ B CFH c.2867C>T; p.Thr956Met rs145975787 0.13% 0.33% 0.17% 0.17% NFE LB CFI c.1657C>T; p.Pro553Ser rs113460688 0.38% 0.33% 0.27% 0.27% NFE LB CFI c.1322A>G; .Lys441Arg rs41278047 0.75% 0.50% 0.24% 4.77% ASJ B CFI c.782G>A; p.Gly261Asp rs112534524 0.13% 0.00% 0.19% 0.46% ASJ LB CFI c.782G>A; p.Gly261Asp rs112534524 0.13% 0.00% 0.19% 0.46% ASJ LB CFB c.1697A>C; p.Glu566Ala rs45484591 3.00% 0.00% 1.00% 2.10% ASJ VUS C3 c.4855A>C; p.Ser1619Arg rs2230210 0.63% 0.25% 0.22% 0.22% NFE VUS C3 c.2203C>T; p.Arg735Trp rs117793540 0.13% 0.17% 0.25% 1.24% ASJ B C3 c.463A>C; p.Lys155Gln rs147859257 0.75% 0.58% 0.54% 0.54% NFE LB

SUPPLEMENTAL FIGURES

Supplement Figure S1. Rare variants in the CFH gene are significantly more abundant in the non-Finnish European (NFE) subpopulation as compared to the Finnish (FIN) subpopulation from gnomAD due to population stratification.

Three algorithms are used to test for enrichment and demonstrate that the modified Poisson exact test is least sensitive to population stratification. P values are shown as curves: red curve, Fisher’s exact test; green curve, Poisson exact test; blue curve, Chi-square test; red dashed line, P value=0.05.

Supplement Figure S2. Cluster analysis shows that UI controls and aHUS cases are mostly similar to NFE.

For clustering, Euclidean distance was calculated based on allele frequencies of variants in each population. Hierarchical cluster analysis was applied using Ward's clustering criterion. Based on this cluster analysis, we used the NFE subpopulation as an additional control.

Supplement Figure S3. Population stratification within aHUS cases and UI controls was used to remove outliers.

Left panel, distribution of patients and controls prior to sample removal; right panel, distribution of cases and controls after removing outliers.

Supplement Figure S4. Mean coverage is correlated with theoretical heterozygous SNP sensitivity (HET SNP sensitivity) in patients and controls.

Random down sampling demonstrated a quick drop of HET SNP sensitivity when mean coverage is below 30X. Most raw sequencing data (purple) from patients (dots) and controls (triangles) is good. Two low quality samples were excluded. Blue line, 95% of HET SNP sensitivity, Red line, 30X of mean coverage.

Supplement Figure S5. Six samples with a shifted ratio of ref/alt reads were excluded from the study cohort.

High quality samples are expected to have peaks at 0.5 and 1.0.

Supplement Figure S6. Prediction score distribution of neutral and pathogenic variants in CFH .

Neutral variants from gnomAD and pathogenic variants from the literature were both filtered by MAF < 0.1%. 18 different tools were used to perform in-silico prediction on these variants to compare neutral and pathogenic variants across the entire gene (white background) or restricting the analysis to SCR19-20 (grey background). Note the heavily mixed distribution of neutral and pathogenic variants for all tools, indicating their ineffectiveness in predicting variant effect in CFH for aHUS.

Supplement Figure S7. Median age of female patients is significantly lower than that of male patients

Female: left panel, 27.3 years; Male: right panel, 14.8 years; Mann-Whitney U testP = 0.011

Supplement Figure S8: Enrichment of ultra-rare variants ‘contaminates’ the result of the association analysis when MAF thresholds are set higher.

The minor allele frequency threshold (cut off) was increased in a stepwise fashion to select variants for the analyses. Sets of p values are shown as curves: black curve, SKAT-O test adjusting for population stratification in UI controls; yellow curve, SKAT-O test without adjusting in UI controls; red curve, Fisher’s exact test in NEF controls; blue curve, Poisson exact test in NEF controls; grey dashed line, P<0.05; green dashed line, P<0.0005.