1 INTRODUCTION

Complement-mediated atypical hemolytic uremic syndrome (aHUS) is an ultra-rare disease caused by a dysregulated activation of the alternative complement pathway. It approximately afflicts one individual per two million person-years [1] and it predominantly occurs in a relapsing fashion. If not promptly treated, aHUS is associated with a high degree of renal morbidity and overall mortality [2]. However, the disease penetrance and severity of renal outcome rest on a complex synergy of several factors: (1) coinciding complement-amplifying conditions (e.g., concomitant autoimmune diseases or infections), (2) present risk haplotypes and polymorphisms, (3) copy number variations and complex genomic rearrangements, (4) the potential presence of factor H-specific antibodies, and (5) rare variants residing in genes encoding complement activators, inhibitors, and thrombosis-associated proteins [3]. Consequently, manifestation of disease is generally conditioned by the specific combination of predisposing factors. Rather than being causative, the rare genetic variants mainly appear to be risk factors for clinical breakthrough [4] which illustrates the genetic heterogeneity of disease.

The phenotypical resemblance with more prevalent diseases that manifest with microangiopathic hemolytic anemia, thrombocytopenia, and acute renal failure makes the diagnostic assessment ambiguous and time to diagnosis is often delayed. These diseases are primarily typical HUS (caused by enterohemorrhagic Escherichia coli or Shigella dysenteriae type 1) and thrombotic thrombocytopenic purpura (confirmed by the finding of ADAMTS13 activity <5%) [1]. Furthermore, several conditions causing non-complement-mediated aHUS have been described and these too must be tentatively ruled out before genetic testing is conducted [5]. In addition, some of the latter conditions have been suggested to potentially contribute to flare-ups of complement-mediated aHUS by means of triggering a complement response already genetically prone to dysregulation. Ultimately, this renders the diagnostic work-up difficult without the genetic testing [6]. However, the genetic work-up is complicated, time-consuming and not always accessible in the routine clinical setting. Moreover, it is recommended that the genetic results are interpreted by a laboratory with specific expertise in aHUS [1]. In Sweden, one clinical expert genetic laboratory is established [7].

In this study, we present a diagnostic scheme for the assessment of potential cases with complement-mediated aHUS that is compliant with the American College of Medical Genetics and Genomics (ACMG) guidelines for the interpretation of sequence variants [8]. We applied it to 20 subjects, originally identified in a previously described Swedish retrospective inpatient cohort of clinically suspected, but unconfirmed aHUS patients [9]. We aimed to investigate characteristics and outcomes of subjects with clinically suspected aHUS and potential differences in clinical features between cases of complement-mediated versus non-complement-mediated disease. We hypothesized that less severe or less frequently relapsing cases of complement-mediated aHUS were not subjected to genetic work-up in the routine clinical setting and that they consequently remained undiagnosed.

2 SUBJECTS AND METHODS 2.1 Study population

The inclusion process was performed as previously described [9]. Briefly, subjects referred nationwide to the Coagulation Unit, Skane University Hospital, Sweden, during the years 2007–2012, with a confirmed ADAMTS13 activity of >5% and routine laboratory findings of (1) hemolytic anemia (increased lactate dehydrogenase and hypohaptoglobinemia and/or reticulocytosis and/or unconjugated hyperbilirubinemia), (2) thrombocytopenia, and (3) renal failure (increased creatinine and/or cystatin C) were included in a retrospective cohort of clinically suspected aHUS patients. Subjects with a positive microbiological testing for enterohemorrhagic Escherichia coli were excluded. Results from peripheral blood smears and direct antigen tests (DAT) were not required for inclusion. Included subjects were categorized into two clinical groups: suspected aHUS with (n = 103), respectively, without (n = 31) potential triggers or causes. The prior study indicated that complement-mediated aHUS potentially was underdiagnosed in the cohort at the time of discharge [9].

Subjects fulfilling the inclusion criteria of the prior study were asked for participation in the present investigation which was composed of blood sampling (whole blood and EDTA plasma) for whole genome sequencing and complement analyses, medical record reviewing and inquiries regarding current diagnoses and renal status.

2.2 Laboratory investigations 2.2.1 Nephelometry

Complement component 3 (C3) and 4 (C4) concentrations were determined using BN Prospec Systems (Siemens Healthineers). Reference range based on healthy controls was 0.86–1.73 g/L for C3 and 0.13–0.31 g/L for C4. The threshold for lower range protein detection was 10 mg/L.

2.2.2 Enzyme-linked immunosorbent assay

Concentrations of factor I (FI), factor H (FH), and FH-specific antibodies were determined using enzyme-linked immunosorbent assays (ELISA). Factor I and FH-specific antibodies were determined as previously described [10, 11]. Details on the FH ELISA are provided in Appendix S1.

Fifty-three unmatched control subjects were randomly selected in a Swedish cohort of healthy subjects. Concentrations were calculated relative to a logarithmic standard curve using a nonlinear regression sigmoidal model. Samples and controls were analyzed in duplicates and the final concentration was calculated as the mean of the two independent assays.

The FI and FH concentrations were expressed as the percentage of meanControls (normal range 100 ± 30%). The concentration of FH-specific antibodies was considered significantly increased if titers were > (meanControls + 3 SD). The cut-off was calculated to 289 AU/mL.

2.2.3 Immunoblotting

Presence of factor H-related protein 1 (FHR1) was determined as previously described [9].

2.2.4 Flow cytometric determination of CD46 surface expression

The surface expression of CD46 (membrane cofactor protein) on monocytes, lymphocytes, and neutrophils was determined to investigate whether genetic variants identified in the CD46 gene were correlated with absent or deficient expression of CD46 on host cells. Normal expression was defined as mean fluorescence intensity ±3 SD of 18 healthy control subjects. Details are provided in Appendix S1.

2.2.5 Whole genome sequencing

The following genes were screened for rare genetic variants: (1) genes encoding complement factors: C3, CFB, Properdin (CFP; enhancer of the alternative pathway); (2) complement inhibitors: CFI, CFH, CFHR1/3/5, CD46, Vitronectin (VTN), and (3) genes directly or indirectly influencing thrombus formation: Plasminogen (PLG), Thrombomodulin (THBD), Diacylglycerol kinase ɛ (DGKE).

Genomic DNA from whole blood was purified using AllPrep DNA/RNA Mini Kit (Qiagen) and library prepared using Nextera DNA Flex Library Prep Kit (Illumina). Paired-end whole genome sequencing was performed on the NovaSeq6000 sequencer (Illumina). Raw fastq files were aligned to the human reference genome (hg19). Quality thresholds for sequencing were ≥ 30-fold average sequencing depth and ≥98% of the genome sequenced at minimum 10-fold. Alignment file pre-processing and germline variant calling was performed by GATK v4.1 analysis software (Broad Institute). The Ingenuity Variant Analysis tool (Qiagen) was used to identify potentially causative variants. Variants were kept for further analysis if variant call quality was ≥20 and if variants were outside the top 5% most exonically variable 100-base windows in healthy public genomes. Variants which had ≥5% allele frequency in 1000 Genomes Project, the Exome Aggregation Consortium, the Genome Aggregation Database (gnomAD) or the National Heart Lung and Blood Institute Exome Sequencing Project were excluded. Variants that were exonic or no more than 10 bases into intron as well as predicted pathogenic or likely pathogenic according to the CentoMD or the Human Gene Mutation Database were kept. Finally, missense variants and variants that were associated with a loss-of-function of a gene were kept, that is, frameshifts, inframe insertion/deletions (InDel), or start/stop codon changes.

In addition, c.3572C>T,p.S1191L and c.3590T>C,p.V1197A are frequent pathogenic variants residing in the CFH gene. They arose through gene conversion between CFH and CFHR1 [12] and are at risk of aligning to CFHR1 in bioinformatics. Thus, exon 23 of CFH and exon 6 of CFHR1 were manually curated in the Integrative Genomics Viewer software [13] and a p.S1191L/p.V1197A positive control was provided for reference in the assessment.

2.3 Variant classification

The detected variants were classified in compliance with the established ACMG guidelines with modifications to allele frequency-related criteria and some exceptions to the Rules for Combining Criteria.

2.3.1 In-silico predictions

For missense variants the established SIFT [14], PROVEAN [15], PolyPhen-2 [16], and CADD [17] prediction tools were applied. Cutoffs for damaging predictions were set to alignment score < 0.05, alignment score < −2.5, Naïve Bayes posterior probability score > 0.15 and c-score > 15, respectively. For splicing defects, we applied CADD, NNSplice [18], dbscSNV [19], and MaxEntScan [20]. Cutoffs for damaging predictions were set to c-score > 15, mutation ÷ wildtype score < 0.85, RF score > 0.6, and mutation ÷ wildtype score < 0.80, respectively. The output of the tools needed to be unanimous to provide supporting evidence of pathogenic or benign impact.

2.3.2 Variant allele frequencies

Data were retrieved from the gnomAD [21]. The minor allele frequencies (MAF) for detected variants in the nonFinnish European population, the global population, and the highest MAF regardless of ethnic origin (Hi_Freq) were collected.

MAF-related criterion for benign variant classification

(1) 0.1% ≤ Hi_Freq MAF < 1% provided supporting evidence of benign impact, (2) 1% ≤ Hi_Freq MAF < 5% provided strong evidence of benign impact, (3) Hi_Freq MAF ≥ 5% provided stand-alone evidence of benign impact.

MAF-related criterion for pathogenic variant classification

(1) Hi_Freq 0.01% ≤ MAF < 0.1% provided supporting evidence of pathogenic impact for single-nucleotide variants (SNV) in CFH, CFI, CD46, C3, DGKE, and VTN [22], (2) Hi_Freq MAF < 0.01% provided moderate evidence of pathogenic impact for SNVs and InDel variants in CFH, CFI, CD46, C3, and DGKE [23].

2.3.3 Exceptions to the ACMG rules for combining criteria Reclassification of variants of unknown significance (VUS)

If the MAF-related criterion constituted the only provided evidence in favor of pathogenic impact, variants with conflicting evidence (both benign and pathogenic criteria) were classified as likely benign.

Reclassification of VUS for CFB, CFP, and C3 variants

Gain-of-function alterations in the CFB, CFP, and C3 genes result in an excessive complement activation. However, in-silico predictions potentially do not recognize gain-of-function alterations as damaging. Moreover, loss-of-function alterations are potentially predicted as damaging in-silico although the absence of CFB, CFP, and C3 gene products in principle does not result in an excessive complement response [24]. Due to the lack of functional evaluations, classifications heavily rely on in-silico predictions. Therefore, null variants (loss-of-function) were classified as likely benign if (1) pathogenic in-silico prediction output resulted in conflicting evidence or if (2) available evidence was insufficient to determine a classification according to the Rules for Combining Criteria to Classify Sequence Variants. However, if the criterion for functional studies supportive of pathogenic impact (unlikely) was met, the variant was defaulted to VUS granted that it was unqualified for a likely pathogenic/pathogenic classification. Non-null variants with conflicting evidence were regarded VUS without exception, that is, including cases where the MAF-related criterion constituted the only provided evidence in favor of pathogenic impact.

2.3.4 Clinical classification Definite complement-mediated aHUS

(1) The subject harbored ≥1 disease-contributing genetic variant (pathogenic or likely pathogenic) and/or (2) the subject featured significant titers of FH-specific antibodies.

Highly suspected complement-mediated aHUS

(1) The subject harbored ≥1 likely disease-contributing genetic variant (VUS located in a mutational hotspot and/or a critical functional domain) and/or (2) the subject featured distinct alternative pathway complement consumption during the acute episode and thrombotic microangiopathy (TMA) was confirmed in a renal biopsy.

Non-complement-mediated aHUS

(1) The subject harbored no disease-contributing/likely disease-contributing genetic variants and (2) the subject was afflicted by a condition acknowledged to cause phenotypical HUS by means of other mechanisms (Table 1).

TABLE 1. Triggers and causes of atypical hemolytic uremic syndrome Infections Drugs Malignancy Autoimmune disease

Neuraminidase-mediated aHUS

Influenza A

Streptococcus pneumoniae

 

Viruses

HHV6, VZV, CMV, EBV

HCV, HAV

HIV

Coxsackie B virus

Parvovirus B19

Dengue virus

Norovirus

Bacteria

Haemophilus influenzae

Bordetella pertussis

Clostridium difficile

Campylobacter upsaliensis

Fusobacterium necrophorum

 

Parasites

Plasmodium falciparum

 

 

Potentially all infections resulting

in sepsis with MOF and DIC

Calcineurin inhibitors

Cyclosporine

Tacrolimus

VEGF inhibitors

Bevacizumab

Sunitimib

mTOR inhibitors

Sirolimus

Everolimus

CD52 inhibitors

Alemtuzumab

General cytotoxics

Mitomycin C

Cisplatin

Vincristine

Gemcitabine

Antibiotics and antiparasitics

Ciprofloxacin

Quinine

Platelet inhibitors

Clopidogrel

Interferon alpha/beta

Contraceptives

Illicit drugs

Heroin

Ecstasy

Cocaine

Metastatic microvascular neoplasms

Lymphoma

HUS phenotypes have also been reported

in association with the following malignancies

Prostatic

Gastric

Hepatic

Pancreatic

Breast

Ovarian

Lung

Colon

Systemic lupus erythematosus

Anti-phospholipid syndrome

Renal scleroderma crisis

Dermatomyositis

Syndromes of HSCT Pregnancy-mediated Vasculitis Renal cortical necrosis Unspecified glomerular disease Malignant hypertension

Radiation

Graft-versus-host disease

CMV infections

Associated drugs

Pregnancy

Postpartum period

Preeclampsia

HELLP syndrome

Polyarteritis nodosa

Infectious vasculitis

(e.g. Rickettsia rickettsii)

Regardless cause Regardless cause Regardless cause Heparin-induced thrombocytopenia and thrombosis Paroxysmal nocturnal hemoglobinuria Pancreatitis Combined methylmalonic aciduria and homocystinuria Disseminated intravascular coagulopathy Hepatitis B vaccination Notes: The table is adapted and modified from Akesson et al. [25]. The conditions presented may be triggers of complement-mediated aHUS, causes of non-complement-mediated aHUS or both. Abbreviations: CMV, cytomegalovirus; DIC, disseminated intravascular coagulation; EBV, Epstein–Barr virus; HAV, hepatitis A virus; HCV, hepatitis C virus; HELLP, hemolysis, elevated liver enzymes, low platelets syndrome; HHV6, human herpes virus 6; HIV, human immunodeficiency virus; HSCT, hematopoietic stem cell transplantation; MOF, multiple organ failure; mTOR, mammalian target of rapamycin; VEGF, vascular endothelial growth factor; VZV: varicella zoster virus. HUS-like phenotype

(1) The subject harbored no disease-contributing/likely disease-contributing genetic variants and (2) the clinical presentation and/or additional laboratory results deemed aHUS unlikely, for example, if the hemolysis criterion in a subject presenting with septic shock was met due to increased lactate dehydrogenase and slightly elevated unconjugated hyperbilirubinemia alone, without further evidence of microangiopathic hemolytic anemia.

2.4 Statement of ethics

The study was conducted in accordance with the Declaration of Helsinki. The Regional Ethics Committee at Lund University approved the study (diary number: 2013/514) and the use of healthy control samples for the laboratory assays (diary number: 2017/582). Written informed consent was obtained from all subjects (or by proxy legal guardians).

2.5 Statistical analyses

Continuous variables with a non-Gaussian distribution were descriptively presented with medians and lower and upper quartiles (q1–q3). Statistical differences for continuous non-normally distributed variables between two groups were evaluated using the Mann–Whitney U test. Statistical differences between categorical variables were evaluated using Fisher's exact test. All tests were exact and all p-values were two-tailed. The p-values <0.05 were considered significant. Statistical analyses were performed using IBM SPSS Statistics version 25.0 (IBM). The ELISA curve fits were executed using GraphPad Prism version 8.0.0 (GraphPad Software).

3 RESULTS 3.1 Mortality ratio and study inclusion

The present study was conducted in median 63 months (q1–q3: 46–77) subsequently to the acute episode onset. At the time of study inclusion, 52% (n = 54/103) and 19% (n = 6/31) of the subjects of the initial cohort of clinically suspected aHUS patients with respectively without potential triggers or causes were deceased.

Three subjects were excluded due to untraceable provisional identification numbers and 59% (n = 42/71) of the remaining subjects declined or did not respond to the study participation inquiry. Nine additional subjects were lost due to nonresponse after request of rebleed following incomplete blood sampling by local laboratories. Ultimately, 20 subjects were included in the present study for the assessment of cases with complement-mediated aHUS: those with (n = 16) and those without (n = 4) potential triggers or causes. Three out of the four subjects without association to potential triggers or causes were categorized as definite or highly suspected complement-mediated aHUS. Thrombotic thrombocytopenic purpura was confirmed for the fourth subject subsequently to discharge. Thus, it was categorized as an HUS-like phenotype. In total, four subjects were attributed an HUS-like phenotype, 10 were categorized as non-complement-mediated aHUS, and six were categorized as definite or highly suspected complement-mediated aHUS (Figure 1).

Study inclusion process. Numbers are presented within brackets. Additional percentages are presented with reference to deceased subjects per group and the proportion of potential study subjects that declined study participation. By means of combining clinical and whole genome sequencing data, the study cohort was partitioned into three groups as presented

3.2 Variant classifications

Twenty unique genetic variants were identified in 14 subjects. All variants were heterozygous and missense variants were the most prevailing (Table 2). Additional information on variant specifics, reference database identifications, MAFs and the ACMG classifications are provided in Tables S1, S2, S3, and S4. In total, one (disease-contributing) pathogenic and four (likely disease-contributing) VUS were identified and distributed among five out of six subjects in the complement-mediated aHUS group (Table 3).

TABLE 2. Variant type specifics and classification Variant IDa Subject (#) Gene (transcript variant) Variation type Translation effect Protein domain Protein variant In-silico predictions ACMG classification 1 A, B C3 (c.193A>C) SNV Missense MG 1 p.K65Q Damaging Pathogenic 2 F C3 (c.4030-4C>T) SNV Splice acceptor site — — Benign Likely benign 3 J C3 (c.4850+3G>A) SNV Splice donor site — — Benign Likely benign 4 C, K CD46 (c.1058C>T) SNV Missense TM p.A353V Benign Benign 5 D CFB (c.1697A>C) SNV Missense SP p.E566A Benign Likely benign 6 G, M CFH (c.-307CT) SNV Promoter — — N/A Likely benign 7 B, L CFH (c.2634C>T) SNV Synonymous CCP 15 p.H878H N/A Likely benign 8 J CFH (c.3133+8G>T) SNV Splice donor site — — Benign Likely benign 9 E CFH (c.3148A>T) SNV Missense CCP 18 p.N1050Y Conflicting Likely benign 10 B CFH (c.3450A>G) SNV Missense CCP 19 p.I1150M Conflicting

VUSb

11 D CFHR5 (c.485_486dupAA) Insertion Frameshift CCP 3 p.E163fs*10 N/A

VUSb

12 I CFI (c.982G>A) SNV Missense Linker 2 p.G328R Damaging

VUSb

13 H CFI (c.1322A > G) SNV Missense SP p.K441R Benign Benign 14 A CFI (c.1534+5G>T) SNV Splice donor site — — Conflicting Likely benign 15 E CFI (c.1547G>T) SNV Missense SP p.G516V Damaging VUSb 16 F DGKE (c.35C>T) SNV Missense Signal peptide p.P12L Conflicting Likely benign 17 J PLG (c.185+4T>C) SNV Intronic — — Conflicting Benign 18 E PLG (c.266G>A) SNV Missense PAN p.R89K Conflicting Likely benign 19 G PLG (c.1567C>T) SNV Missense Kringle 5 p.R523W Conflicting Likely benign 20 S PLG (c.2356C>T) SNV Missense SP p.R786C Conflicting VUS Notes: In-silico predictions were assessed for missense and splice site variants only. The prediction was conflicting if the output of the implemented annotation tools were not unanimous. Abbreviations: SNV, single-nucleotide variant; VUS: variant of unknown significance. a All detected variants were heterozygous. b Likely disease-contributing. TABLE 3. Subject characteristics Subject (#) Variant ID (classification) Time to study (months) Dialysis or transplant Diagnosis Medication and interventions HUS-like phenotype P — 94 — TTP Plasmapheresis, rituximab T — 65 — SLE Azathioprine, hydroxychloroquine, prednisone J 3 (LB), 8 (LB), 17 (B) 87 — SLE Azathioprine, prednisone M