Genome sequencing unveils mutational landscape of the familial Mediterranean fever: Potential implications of IL33/ST2 signalling

Abstract Familial Mediterranean fever (FMF) is the most common auto‐inflammatory disease. It is transmitted as autosomal recessive trait with mutations in MEditerranean FeVer (MEFV) gene. Despite a typical clinical expression, many patients have either a single or no mutation in MEFV. The current work is aimed to revisit the genetic landscape of FMF disease using high‐coverage whole genome sequencing. In atypical patients (carrying a single or no mutation in MEFV), we revealed many rare variants in genes associated with auto‐inflammatory disorders, and more interestingly, we discovered a novel variant ( a 2.1‐Kb deletion) in exon 11 of IL1RL1 gene, present only in patients. To validate and screen this patient‐specific variant, a tandem of allele‐specific PCR and quantitative real‐time PCR was performed in 184 FMF patients and 218 healthy controls and we demonstrated that the novel deletion was absent in controls and was present in more than 19% of patients. This study sheds more light on the mutational landscape of FMF. Our discovery of a disease‐specific variant in IL1RL1 gene may constitute a novel genetic marker for FMF. This finding suggesting a potential role of the IL33/ST2 signalling in the disease pathogenicity highlights a new paradigm in FMF pathophysiology.

Familial Mediterranean fever (FMF) is the most common Mendelian auto-inflammatory disease, characterized by uncontrolled activation of the innate immune system, resulting in recurrent brief episodes of fever and serositis with chest, abdominal, joints and muscles pain. 4 Predominantly, FMF affects people from Mediterranean and Middle Eastern ethnic origins (1/200-1/1000). 5 The causing gene of FMF is the MEditerranean FeVer (MEFV) gene. 5,6 The MEFV gene encodes 781 amino acids pyrin (or marenostrin) protein, which is mostly expressed in neutrophils, eosinophils, monocytes, dendritic cells and fibroblasts. 7,8 The exact physiological role of pyrin protein is not clear; however, it is suggested to play a role in apoptosis, inflammation, cytokine production and innate im-  10 and p.Glu148Gln in exon 2, together they represent more than 80% of the disease-causing mutations. 9 The analysis of the typical FMF patients revealed an autosomal recessive model of inheritance. 5 The disease can segregate either in homozygous or in a compound heterozygous modality. However, it is observed that a substantial number of FMF patients are either heterozygous or carry no MEFV mutation. The possibility of pseudo-dominance is considered in rare cases but it is yet to be proven and it could not explain the large number of clinical FMF cases. 10,11 The hypothesis of digenic or oligogenic inheritance is gaining attention and could explain the divergence of clinical FMF with single or no mutation in MEFV gene from the typical paradigm of recessive inheritance. 12,13 The presence of mutations in modifier genes associated with inflammation or interactions between MEFV mutation and modifying allele in genes involved in known auto-inflammatory diseases, as reported in a limited number of studies, could also be responsible for the large spectrum of FMF phenotypes. 14,15 The lack of comprehensive genetic analyses of FMF patients with single or no mutated allele in MEFV gene is prompting us to investigate the genetic landscape of FMF disease in a large cohort of FMF patients with different MEFV mutational profiles using both Sanger and whole genome sequencing (WGS).

| Patients and controls
The study population consisted of 402 unrelated Lebanese subjects including 184 FMF patients (102 males and 82 females with median age 17 ± 5 years) recruited from several medical centres in Beirut, Lebanon, and 218 gender and ethnicity matched healthy controls recruited among subjects visiting the hospitals for routine health check-up and who were free from any chronic inflammatory and autoimmune disease. Blood sample collection and storage was managed by the Medical Center CEMEDIPP and the American University of Science and Technology in Beirut, Lebanon. The diagnosis of FMF in our patients was made according to the established criteria of both Sohar (Tel Hashomer criteria) 5 and Livneh. 16 More stringent clinical diagnosis criteria were used to establish the diagnosis of FMF in patients with a single disease-causing MEFV variant or with no identified MEFV variants. The 184 FMF patients were randomly selected from a large cohort of patients for whom Sanger sequencing of 10 exons of MEFV gene was performed, and who, based on copies of MEFV mutated allele, were stratified into three groups: (a) zero mutation: patients without any mutation in MEFV gene; (b) single mutation: patients with only one mutation in MEFV gene; and (c) double mutation: patients with two MEFV mutations. In order to increase the chance to identify novel and/or modifier genes for FMF, we purposely enriched our study cohort with more patients with a single or no variant in MEFV gene. We performed WGS on 50 patient samples (11 patients with double MEFV mutation, 19 patients with a single mutation and 20 patients with no MEFV mutation) randomly selected from the 3-Sanger sequencing-defined subcategories and that of 26 healthy control subjects.
The study protocol was approved by ethics committee of Sidra Medicine, Doha, Qatar (Protocol number # 1511002018). All study subjects signed a written informed consent prior to be enrolled in the study.

| Sample preparation and whole genome sequencing (WGS)
Peripheral blood samples were collected from patients and controls in EDTA tubes and genomic DNA was extracted by standard saltprecipitation methods. 17  Excess adapters and enzymes were removed using AMPure beads (Beckman Coulter Genomics). Indexed libraries were size-selected to the 350 bp range using bead-based capture, and the concentration of amplifiable fragments was determined by qPCR, relative to sequencing libraries with a known concentration. Normalized libraries were clustered on a c-BOT machine, and 125 bp paired-end sequencing was performed on the HiSeq 2500 system.

| WGS data analysis
Paired-end raw fastq files were mapped to the reference human genome, build GrCh37, using BWA-MEM aligner: 0.7.12-r1039, 18   We have submitted all the variants reported here to LOVD website (https://www.lovd.nl).  To confirm the outcome of the AS-PCR, the qRT-PCR was performed using two sets of pair of primers; one set was used to amplify a DNA fragment within the 2.1-Kb deletion (forward  Chi-square test was used to compare the frequency between the two groups of patients (patients with a single or no MEFV mutation vs patients with 2 MEFV mutations), and the Phi coefficient was used to generate the effect size of this novel variant in patients.

| Characterization of MEFV mutations in patients with FMF
The 184 FMF patients of the present study were randomly selected from a large cohort for which Sanger sequencing of coding sequence of MEFV gene was performed. In order to increase the chance to un-

| Mutational Spectrum of genes associated with other AIDs in FMF patients
After filtering out variants which had high prevalence in the general population (allele frequency > 1%) or were present in controls, we first examined variants in known AID-associated genes. More than 50 genes associated with auto-inflammatory disorders were selected from Systemic autoinflammatory disease (SAID; http://www.autoi nflam mator y-search.org/)) and Infever database. 32 The list of the novel variants of genes associated with AIDs found in the 50 FMF patients is shown in Table 2. We observed that 10 out of the 50 FMF

| Identification of novel variants in inflammatory genes in FMF patients
Variants in known AID-associated genes identified in our cohort were not sufficient to completely draw the genetic variation pattern in our FMF patients. We further looked for the predicted pathogenic variants in inflammatory genes either interacting with known genes associated to AIDs or involved in auto-inflammation processes, using knowledge base of Ingenuity variant analysis. The list of variants in inflammatory genes found in the 50 FMF patients is shown in Table 3.

| Copy number variant (CNV) analysis in FMF
Beside point mutations and small indels, we also looked for the structural variants in the whole genome of the 50 FMF cases. Variant  Figure S1.

| The IL1RL1 gene deletion in familial Mediterranean fever patients
To validate the finding revealed by WGS and CNV analysis, a

| D ISCUSS I ON
In the present report, we showed significant genetic heterogeneity cytokine signalling and (f) macrophage activation syndromes. 37 We filtered our WGS data for the variants in more than 50 genes associated to AID belonging to one or another of the above-mentioned AID categories and investigated for potential pathogenic variants common to FMF cases and absent in controls. Although no single RAB27A encodes small Rab GTPase, which plays a role in exocytosis of cytotoxic vesicles, while STXBP2 is involved in the release of cytotoxic granules by natural killer cells. 38 Mutations in these genes are supposed to impair their normal function and could lead to increased macrophages activation and cytokine production. 39 Other AIDassociated gene variants were identified in our patients, but they were not frequent and were present only in one or two cases each.
We further investigated variations in novel genes, which are reported to interact with known auto-inflammatory genes or which may have a role in auto-inflammation process. Replication of the present findings in other populations will be useful to determine whether the association between these genetic markers and FMF can be generalized. We believe that our find-

ACK N OWLED G EM ENTS
This work was supported by Sidra Medicine, Qatar and Open Access funding provided by the Qatar National Library.

CO N FLI C T S O F I NTE R E S T
The authors confirm that there are no conflicts of interest.