MEFV variant spectrum and exploratory in silico stratification of uncommon variants in 611 individuals tested for suspected familial Mediterranean fever in the Eastern Black Sea region of Türkiye

Abstract

Background

Familial Mediterranean fever (FMF) is the most common monogenic autoinflammatory disease, caused by variants in the MEFV gene. Regional MEFV variant spectra differ across Türkiye. We characterised the MEFV variant spectrum in a single-centre series of individuals tested for suspected FMF in the Eastern Black Sea region and explored in silico stratification of uncommon variants. No standardised clinical diagnostic criteria were applied; the study is therefore framed as a molecular variant-spectrum study.

Methods

We retrospectively analysed MEFV next-generation sequencing (NGS) reports from 611 consecutive individuals referred with a clinical suspicion of FMF (January 2022–October 2025). Standardised FMF classification criteria (Tel-Hashomer; Eurofever/PRINTO) were not applied; the cohort was defined by referral indication. For each variant, two metrics were reported: the proportion within the detected mutant-allele pool (n/591) and the allele frequency across all tested chromosomes (n/1222; 2 × 611). Uncommon variants were defined as all detected missense variants other than the recognised common FMF alleles (R202Q, M694V, M680I, M694I, V726A, E148Q, P369S, R408Q). Uncommon variants were submitted to the seven-tier framework of Alay (2025) as an exploratory layer supplementary to — not a replacement for — ACMG/AMP classification.

Results

At least one variant was detected in 379/611 individuals (62.0%); 591 alleles across 22 distinct missense variants were tabulated. R202Q was the most frequently detected allele in the spectrum (44.0% of detected mutant alleles; 21.3% of all tested chromosomes), exceeding M694V (22.3%; 10.8%). When restricted to high-penetrance exon 10 founder variants (M694V, M680I, M694I, V726A), the pathogenic-allele burden was substantially lower (202 alleles; 16.5% of tested chromosomes). Compound/complex genotypes were present in 25.7% of the cohort (cis/trans phase not determined). Exploratory in silico stratification flagged 7 of 10 evaluable uncommon variants toward likely-pathogenic and 3 toward likely-benign; four variants could not be mapped. A previously unreported substitution, p.Ser179Gly, was detected in one individual and is reported only as a candidate variant of uncertain significance, without pathogenicity inference.

Conclusions

In this region, R202Q predominates in the detected MEFV variant spectrum; however, its contested pathogenicity means that spectrum dominance should not be equated with regional clinical or pathogenic-allele burden. The findings describe a molecular variant spectrum, and clinical interpretation is constrained by the absence of phenotype, segregation and standardised diagnostic data. In silico stratification is exploratory and does not replace ACMG/AMP classification.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12863-026-01481-z.

Background

Familial Mediterranean fever (FMF; OMIM #249100) is the most common monogenic autoinflammatory disease, predominantly affecting populations of Mediterranean and Middle-Eastern ancestry, including Turks, Armenians, non-Ashkenazi Jews and Arabs [1]. Inherited in an autosomal recessive manner, FMF is characterised by recurrent self-limiting episodes of fever accompanied by sterile serositis, arthritis and erysipelas-like erythema [2, 3]. In Türkiye, the disease prevalence is estimated at 1 per 400–1,000 individuals, with a heterozygote carrier rate of approximately 1 in 5 [4].

The molecular basis of FMF lies in pathogenic variants of the MEFV gene on chromosome 16p13.3; this gene encodes a 781-amino-acid protein designated pyrin [2]. Pyrin is a key modulator of the pyrin inflammasome, regulating caspase-1 activation and downstream interleukin-1β maturation [5–7]. The high-penetrance mutations M694V, M680I, V726A and M694I in exon 10 and E148Q in exon 2 account for the majority of pathogenic alleles in classically affected populations [3, 8].

Multicentre studies have demonstrated that M694V is the most frequently detected mutation in Turkish FMF patients overall [1, 9]; however, regional studies reveal substantial geographical heterogeneity. Cohorts from Central Black Sea provinces have reported an unexpectedly high prevalence of R202Q (c.605G > A, p.Arg202Gln) — an exon 2 variant whose pathogenic status remains debated — as the predominant detected allele, a pattern markedly divergent from the national average [10–12].

Understanding regional variant spectra is relevant for diagnostic panel design, although this does not, on its own, establish clinical significance. As next-generation sequencing (NGS) becomes increasingly accessible, variants of uncertain significance (VOUS) are identified more frequently [13]. The 2015 ACMG/AMP guidelines [14] provide a five-tier classification but do not differentiate between VOUS calls tilted toward pathogenicity versus benignity. Earlier work demonstrated the value of gene-specific recalibration for MEFV [15, 16]. Building on this, Alay [17] recently developed a seven-tier classification framework — VOUS–, VOUS-, VOUS0, VOUS+, VOUS++, in addition to LB and LP. In this framework, “VOUS” (variation of unknown significance) is a deliberately defined umbrella category that consolidates the ClinVar “uncertain significance”, “conflicting interpretations” and uncategorised/not-provided classes and then resolves them into the five ordinal tiers above; it is therefore broader than, and not synonymous with, the ACMG/AMP “variant of uncertain significance” (VUS). Accordingly, throughout this manuscript we retain “VOUS” when referring to this framework and its tiers, and use “VUS” for the primary ACMG/AMP category reported in Table 3. This framework integrates 42 in silico tools through ensemble machine learning and offers two complementary readouts: an adaptive classifier (collapsing first- and second-tier VOUSs into LB/LP) and a more conservative rigid classifier (collapsing only second-tier VOUSs). We use this framework strictly as an exploratory in silico layer that may help prioritise uncommon variants for further evaluation, and not as a substitute for ACMG/AMP classification or for functional/clinical validation.

Table 3.

Primary ACMG/AMP category and exploratory in silico stratification of uncommon variants (framework of Alay 2025 [17])

Variant	Anchor (hg19)	ACMG/AMP	INFEVERS	7-tier	Adaptive	Rigid	Notes
p.I591T (c.1772T > C)	16:3243880 A > G	VUS	Likely benign	LB	LB	LB	Concordant LB flag (exploratory)
p.T267I (c.800 C > T)	16:3254268 G > A	VUS	VUS	VOUS-	LB	indet.	Adaptive/rigid divergent (exploratory)
p.E167D (c.501G > C)	16:3254567 C > G	VUS	VUS	LP	LP	LP	Concordant LP flag (exploratory)
p.E148V (c.443 A > T)	16:3254625 T > A	VUS	VUS	VOUS++	LP	LP	Concordant LP flag (exploratory)
p.A744S (c.2230G > T)	16:3243257 C > A	VUS	Likely benign	LP	LP	LP	Concordant LP flag (exploratory)
p.S339F (c.1016 C > T)	16:3249675 G > A	VUS	Likely benign	VOUS++	LP	LP	1 transcript VOUS, others VOUS++
p.G304R (c.910G > A)	16:3254158 C > T	VUS	Benign	LB	LB	LB	Concordant LB flag (exploratory)
p.K695R (c.2084 A > G)	16:3243403 T > C	VUS	VUS	LP	LP	LP	Concordant LP flag (exploratory)
p.R761H (c.2282G > A)	16:3243205 C > T	VUS	Pathogenic	LP	LP	LP	Concordant LP flag (exploratory)
p.F479L (c.1437 C > G)	16:3247166 G > C	VUS	Pathogenic	LP	LP	LP	Concordant LP flag (exploratory)
p.S179G (c.535 A > G)	—	VUS	Not listed	—	—	—	No exact row in deployed app; candidate variant
p.A287T (c.859G > A)	—	VUS	Not listed	—	—	—	No exact row in deployed app
p.A89T (c.265G > A)	—	VUS	Likely benign	—	—	—	No exact row in deployed app
p.L509P (c.1526T > C)	—	VUS	Not listed	—	—	—	No exact row in deployed app

ACMG/AMP = primary classification (VUS retained where functional/segregation/phenotype evidence is insufficient). The 7-tier label, adaptive and rigid columns are exploratory in silico outputs that do not replace ACMG/AMP classification. Each variant is anchored to its genomic coordinate (GRCh37/hg19) and curated c.HGVS on transcript NM_000243.3; the seven-tier outputs are reported against this anchor. LB = likely benign flag; LP = likely pathogenic flag; indet. = indeterminate. INFEVERS = classification recorded in the INFEVERS registry (https://infevers.umai-montpellier.fr; MEFV, NM_000243.3, GRCh38), as proposed by the INSAID study group and re-evaluated by the Solving MEFV Variants project (current release; last database update 2 March 2026). “Not listed” indicates that the exact substitution is not recorded in INFEVERS; for p.Ser179Gly only the neighbouring p.Ser179Ile and p.Ser179Asn are listed

Because clinical phenotyping, standardised diagnostic criteria and segregation data were not available for this referral-defined cohort, the present work is framed as a molecular variant-spectrum study rather than a clinical FMF cohort study. To date, no comprehensive study of the MEFV variant spectrum has been published specifically for the study province or its catchment area in the Eastern Black Sea region. The present study aimed to: (1) characterise the MEFV variant spectrum and frequencies — reported both as the proportion of detected mutant alleles and as the allele frequency across all tested chromosomes — in 611 consecutive individuals tested for suspected FMF; (2) describe the zygosity and compound-genotype distribution; (3) compare the spectrum with contemporary regional and national literature using metric-consistent comparisons; and (4) apply, as an exploratory layer supplementary to ACMG/AMP, the seven-tier framework [17] to the uncommon variants identified. The central message of this work is therefore descriptive and methodological: to report the regional MEFV variant spectrum using metric-consistent frequencies, to separate spectrum prominence from pathogenic-allele burden, and to provide an exploratory in silico prioritisation of uncommon variants — rather than to establish the clinical significance of any individual variant.

Methods

Study design and population

This single-centre retrospective study analysed MEFV genetic test reports from 611 consecutive individuals referred to the Department of Medical Genetics at the study institution with a clinical suspicion of FMF. The study period covered 1 January 2022 to 1 October 2025. Crucially, no standardised FMF diagnostic or classification criteria (Tel-Hashomer criteria [18]; Eurofever/PRINTO criteria [19]) were applied, and attack characteristics, colchicine response, family history and amyloidosis status were not collected. The cohort was therefore defined solely by the referring clinician’s test indication and represents individuals tested for suspected FMF rather than clinically confirmed FMF patients; the analysis is framed accordingly as a molecular variant-spectrum study. The inclusion criterion was a consecutive, interpretable diagnostic MEFV NGS report generated during the study period for an individual referred with clinical suspicion of FMF; the only exclusion criterion was a technically failed or uninterpretable report. No selection was applied on the basis of variant status, symptom severity or demographic characteristics. Consequently, the frequencies reported here describe the mutational spectrum of a referral (test-indication) population and are not intended as estimates of FMF disease prevalence, nor of the pathogenic-allele burden of the general regional population. Reporting follows the STROBE statement for observational studies; a completed STROBE checklist was provided with the submission for editorial assessment.

Ethics

The study was approved by the Ordu University Non-Interventional Clinical Research Ethics Committee (decision no. 2025/423, dated 12 December 2025) and was performed in accordance with the principles of the Declaration of Helsinki. Owing to the retrospective nature of the study and the use of fully anonymised molecular reports, the requirement for individual informed consent was waived by the committee.

Molecular testing

Genomic DNA was isolated from peripheral blood leukocytes, collected in EDTA tubes, using a column-based extraction protocol. Targeted next-generation sequencing of the MEFV gene (RefSeq NM_000243.3; Ensembl ENST00000219596; reference assembly GRCh37/hg19) was performed at an accredited external diagnostic laboratory using laboratory-specific primer/probe sets in combination with the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA, USA). Sequencing covered all protein-coding exons and the ± 10 base-pair flanking exon–intron junctions of MEFV; positions covered by fewer than 20 reads were considered to have insufficient depth and were not reported. Variant filtering and minor allele frequency annotation were performed against the gnomAD, ExAC, 1000 Genomes and dbSNP population databases, while clinical interpretation integrated the Franklin, VarSome and HGMD Public annotation resources. Variant nomenclature followed HGVS recommendations (https://varnomen.hgvs.org) on transcript NM_000243.3 (protein NP_000234.1). Zygosity of each reported variant was given in the source report with a “Heterozigot” or “Homozigot” prefix and was extracted directly from this label.

The previously unreported substitution p.Ser179Gly was detected in a single individual. Orthogonal (Sanger) confirmation and quantitative raw-read quality metrics (read depth, alternate-allele fraction, base/variant quality score) were not available for this retrospective report; accordingly, this variant is treated throughout as a candidate variant of uncertain significance and is not assigned any pathogenicity inference (see Results and Discussion). Population and clinical database interrogations (gnomAD, ClinVar, INFEVERS, dbSNP) were performed during manuscript preparation.

Classification and exploratory reclassification of uncommon variants

Each detected variant was first assigned an ACMG/AMP [14] category as the primary classification. ACMG/AMP classification was carried out by combining the automated variant-classification outputs of the Franklin (Genoox) and VarSome clinical engines with manual review of population-frequency (gnomAD, ExAC, 1000 Genomes), computational (in silico) and clinical-database (ClinVar, INFEVERS) evidence; where functional, segregation and phenotype evidence was insufficient to satisfy benign or pathogenic criteria, the variant was retained as a VUS. In the absence of functional, segregation and phenotype data, uncommon variants lacking sufficient evidence under ACMG/AMP were retained as VUS; the ACMG/AMP category is reported alongside each variant in Table 3.

In addition, all detected missense variants other than the canonical FMF founder mutations (R202Q, M694V, M680I, M694I, V726A, E148Q, P369S, R408Q) were submitted (hereafter termed the “uncommon variants”, i.e. all detected missense variants other than the recognised common FMF alleles), as an exploratory in silico layer, to the publicly deployed Shiny web application accompanying Alay (2025) [17] (https://alaymd.shinyapps.io/MEFV_app/; n = 14). The framework integrates the outputs of 42 in silico predictors (missense-pathogenicity, conservation and splice-effect tools catalogued in dbNSFP/Ensembl) through an ensemble machine-learning model trained on curated MEFV variants, and returns a seven-tier label. Two collapsing rules are provided: the adaptive classifier assigns both first- and second-tier VOUS calls to the corresponding likely-benign or likely-pathogenic category, whereas the more conservative rigid classifier collapses only the strongest (second-tier) VOUS calls and leaves first-tier VOUS calls as indeterminate (“indet.”). To avoid any ambiguity arising from differing protein-numbering conventions, each variant was anchored to its genomic coordinate (GRCh37/hg19) and its curated c.HGVS description on transcript NM_000243.3, and the seven-tier label together with the adaptive and rigid classifier outputs was recorded against this anchor. Variants without an exact row in the deployed application were flagged as “no match” and reported separately. These in silico outputs are exploratory only and do not constitute clinical reclassification.

Comparator data and statistical analysis

For each variant we report two complementary metrics: (i) its proportion within the detected mutant-allele pool (number of mutant alleles, with homozygous detections counted twice and heterozygous detections counted once, divided by 591 detected mutant alleles), describing the regional mutational spectrum; and (ii) its allele frequency across all tested chromosomes (number of mutant alleles divided by 1222 chromosomes; 2 × 611 individuals). The unqualified term “allele frequency” is reserved exclusively for metric (ii). Descriptive statistics are reported as counts and percentages.

Cross-study comparisons (Table 2) were restricted to studies reporting allele-level frequencies and were performed on a metric-consistent basis; studies reporting only genotype-level frequencies were not pooled with allele-level data and are discussed separately in the text. For each comparator, panel coverage (whether R202Q was included in the screen) is indicated, because a screen that omits R202Q cannot be compared on that allele. The full anonymised dataset underlying these counts is provided as Additional file 2 (ESM_2.xlsx). Figures were prepared as TIFF files at 300 DPI with LZW compression.

Table 2.

Comparison of R202Q and M694V across selected Turkish MEFV studies reporting allele-level data (proportion of detected mutant alleles)

Study	Region	n	R202Q in panel?	R202Q (% of detected alleles)	M694V (% of detected alleles)
Present study	Ordu (Eastern Black Sea)	611	Yes	44.0 *	22.3 *
Akbaş et al. 2021 [11]	Mid-Black Sea (Tokat)	617	Yes	27.9	13.2
Aksoy et al. 2024 [19]	Ankara (national multicentre)	2,984	Yes	39.2	25.1
Yaşar Bilge et al. 2019 [9]	National multicentre	1,719	No (not in panel)	NA	44.5

* For the present study, the true allele frequency across all tested chromosomes (n/1222) is R202Q 21.3% and M694V 10.8% (Table 1); the value shown here is the proportion of detected mutant alleles, for metric consistency with the comparator studies. NA = not assessable because R202Q was not included in the screening panel. Genotype-level studies (Çapraz & Düz 2023; Celep et al. 2019) are discussed in the text and are not included here because they report the proportion of patients carrying a variant (each patient counted once, irrespective of allele dose) — a genotype-level metric that cannot be converted to an allele/chromosome-based frequency without the underlying per-allele counts, so that pooling the two metrics in a single column would be misleading

Results

Cohort characteristics and detection rate

A total of 611 consecutive individuals tested for suspected FMF were included. MEFV analysis identified no reportable variant in 232 individuals (38.0%) and at least one variant in 379 individuals (62.0%). Of the 379 variant-positive individuals, 216 (35.4% of the total cohort; 57.0% of the variant-positive group) carried a single heterozygous variant, 6 (1.0%; 1.6%) carried a single homozygous variant, and 157 (25.7%; 41.4%) carried compound heterozygous or complex multi-variant genotypes. The cohort flow and detection summary are shown in Fig. 1.

Fig. 1.

Cohort flow and MEFV variant detection summary in the suspected-FMF cohort (n = 611). The flow diagram stratifies the cohort into variant-negative and variant-positive groups, and the variant-positive group into single-heterozygous, single-homozygous, and compound/complex genotypes. The lower panel reports allele-level summary statistics

MEFV variant spectrum and frequencies

A total of 591 alleles were tabulated across 22 distinct missense variants. R202Q (c.605G > A, p.Arg202Gln) was the most frequently detected allele within the spectrum, accounting for 44.0% of all detected mutant alleles (n = 260) but 21.3% of all tested chromosomes. M694V (c.2080 A > G, p.Met694Val) ranked second (n = 132; 22.3% of detected alleles; 10.8% of tested chromosomes), followed by E148Q (c.442G > C; n = 55; 9.3%; 4.5%), M680I (c.2040G > C; n = 43; 7.3%; 3.5%) and V726A (c.2177T > C; n = 20; 3.4%; 1.6%). Among rarer variants, P369S and R408Q were each detected in 17 alleles (they typically co-occur in cis), A744S in 6 alleles, and two further uncommon substitutions — p.Ala89Thr (c.265G > A) and p.Leu509Pro (c.1526T > C) — in one allele each (Table 1). A previously unreported missense substitution, p.Ser179Gly (c.535 A > G), was detected in one individual in apparent compound configuration with R202Q. The full distribution is presented in Table 1, with a graphical summary in Additional file 1 (ESM_1.tiff).

Table 1.

Frequency distribution of MEFV variants in the study cohort (591 detected mutant alleles; 1222 tested chromosomes)

Variant (protein)	Nucleotide	n (alleles)	% of detected mutant alleles (n/591)	Allele freq. (n/1222, %)	Homozygous / heterozygous detections (n)
R202Q	c.605G > A	260	44.0	21.3	9 / 242
M694V	c.2080 A > G	132	22.3	10.8	4 / 124
E148Q	c.442G > C	55	9.3	4.5	0 / 55
M680I	c.2040G > C	43	7.3	3.5	2 / 39
V726A	c.2177T > C	20	3.4	1.6	0 / 20
P369S	c.1105 C > T	17	2.9	1.4	0 / 17
R408Q	c.1223G > A	17	2.9	1.4	0 / 17
M694I	c.2082G > A	7	1.2	0.6	0 / 7
G304R	c.910G > A	7	1.2	0.6	0 / 7
A744S	c.2230G > T	6	1.0	0.5	0 / 6
I591T	c.1772T > C	5	0.8	0.4	0 / 5
E148V	c.443 A > T	3	0.5	0.2	0 / 3
E167D	c.501G > C	3	0.5	0.2	0 / 3
K695R	c.2084 A > G	3	0.5	0.2	0 / 3
T267I	c.800 C > T	3	0.5	0.2	0 / 3
F479L	c.1437 C > G	2	0.3	0.2	0 / 2
S339F	c.1016 C > T	2	0.3	0.2	0 / 2
R761H	c.2282G > A	2	0.3	0.2	0 / 2
A89T	c.265G > A	1	0.2	0.1	0 / 1
L509P	c.1526T > C	1	0.2	0.1	0 / 1
A287T	c.859G > A	1	0.2	0.1	0 / 1
S179G ‡	c.535 A > G	1	0.2	0.1	0 / 1
TOTAL		591	100.0	48.4	15 / 561

‡ p.Ser179Gly (c.535 A > G) detected in one individual in apparent compound configuration with R202Q; reported as a candidate variant of uncertain significance (see Discussion). ClinVar registers the neighbouring substitution p.Ser179Asn (c.536G > A; rs104895125), but no peer-reviewed clinical report of p.Ser179Gly was identified at the time of writing. “% of detected mutant alleles” describes the mutational spectrum (relative proportion within the detected mutant pool); “Allele freq. (n/1222)” is the true allele frequency across all tested chromosomes

Stratifying the detected alleles by category of pathogenic evidence clarifies the distinction between spectrum prominence and pathogenic-allele burden. Category 1 — high-penetrance exon 10 founder variants (M694V, M680I, M694I, V726A) — accounted for 202 alleles (34.2% of the detected mutant-allele pool; 16.5% of all tested chromosomes). Category 2 — E148Q, an exon 2 variant generally regarded as low-penetrance/disease-modifying — accounted for 55 alleles (9.3%; 4.5%). Category 3 — R202Q, an exon 2 variant of contested pathogenicity — accounted for 260 alleles (44.0%; 21.3%). The remaining uncommon variants together contributed 74 alleles (12.5%; 6.1%). Thus, although R202Q dominates the detected spectrum, the high-penetrance pathogenic-allele burden (Category 1) is markedly lower than its spectrum-level prominence would suggest.

Zygosity distribution

Of the 591 alleles tabulated, 561 (94.9%) derived from heterozygous reports and 30 (5.1%) from homozygous reports (15 homozygous detections × 2 alleles each). Homozygous detections were observed for R202Q (9 detections), M694V (4) and M680I (2). All homozygous detections involved either a high-penetrance exon 10 variant (M694V, M680I) or the contested exon 2 variant R202Q. Notably, no homozygous detection was observed for E148Q (0 of 55) or V726A (0 of 20) despite their substantial allele counts.

Compound and complex genotype configurations

Because allele-level zygosity ratios do not capture the biallelic architecture of the cohort, the 157 compound/complex genotypes (25.7% of the cohort; 41.4% of variant-positive individuals) are additionally considered by their highest-penetrance component. A cardinal caveat is that cis/trans phase was not determined and that two well-recognised cis haplotypes are expected within these data: M694V is frequently in linkage disequilibrium with R202Q on the same haplotype, and P369S co-occurs in cis with R408Q. Consequently, an apparent two-variant genotype may in fact reside on a single mutant chromosome, and a genotype that appears biallelically pathogenic may carry a third variant in trans. With this limitation stated explicitly, the compound/complex genotypes fall, on an unphased basis, into two groups. First, 126 of the 157 (80.3%) contained at least one high-penetrance exon 10 (Category 1) allele and would, if the alleles were biallelic and in trans, be compatible with a confirmatory (“Category 1”) FMF genotype; within this group 24 genotypes carried two or more exon 10 alleles (for example M680I + M694V or M694V + V726A), representing a strongly biallelic high-penetrance configuration even when unphased, whereas the single most frequent configuration was M694V + R202Q (n = 77) — precisely the pairing for which the recognised M694V–R202Q cis haplotype means that many may represent a single pathogenic chromosome rather than a biallelic pathogenic genotype. Second, 31 of the 157 (19.7%) were composed exclusively of exon 2 and/or uncommon variants (most frequently E148Q + P369S + R408Q, n = 6, and E148Q + R202Q, n = 5), representing a lower-penetrance, non-confirmatory configuration. Separately, three individuals were homozygous for a high-penetrance exon 10 variant as their sole finding (M680I, two; M694V, one), a clearly biallelic pathogenic genotype; the homozygous detection counts in Table 1 are reported per variant and therefore also include homozygotes occurring within compound/complex genotypes. Because phase and phenotype were unavailable, these groupings are an unphased quantitative approximation only and are not equated with clinical diagnosis; the full per-individual genotype list is provided in Additional file 2 (ESM_2.xlsx).

Comparison with selected published Turkish MEFV cohorts

Comparative frequencies of R202Q and M694V across selected published Turkish MEFV studies reporting allele-level data are summarised in Table 2; to maintain metric consistency, the present-study values in Table 2 are expressed as the proportion of detected mutant alleles (the metric most commonly reported in these comparators), while the corresponding true allele frequencies for the present study (n/1222) are given in Table 1. Panel coverage is indicated for each study, since a screen omitting R202Q cannot be compared on that allele. Among the allele-level comparators, only Aksoy et al. [20] used a denominator directly comparable to ours (a variant’s fraction among all detected mutant alleles in an FMF-referral cohort); Akbaş et al. [11] and Yaşar Bilge et al. [9] analysed clinically defined FMF patients and, in the case of Akbaş et al., computed frequencies across all screened chromosomes including variant-negative individuals. Cross-study percentages should therefore be read as broadly indicative rather than strictly normalised. Two regional studies reporting only genotype-level frequencies (Çapraz & Düz 2023 [10], with an R202Q genotype-level frequency of 43.5%, and Celep et al. 2019 [21], 50.1%) are not directly comparable on an allele basis and are therefore discussed in the text rather than pooled into Table 2. Interpretation of these patterns is presented in the Discussion.

Exploratory in silico stratification of uncommon variants

Fourteen uncommon variants were submitted to the seven-tier framework [17] as an exploratory layer. Of these, 10 were returned with a classifier output, while four — p.Ser179Gly (c.535 A > G), p.A287T (c.859G > A), p.Ala89Thr (c.265G > A) and p.Leu509Pro (c.1526T > C) — had no exact row in the deployed application and were flagged as “no match”. Results are summarised in Table 3 and visualised on the pyrin domain map in Fig. 2. Under the adaptive classifier, 7 of the 10 evaluable variants were flagged toward likely-pathogenic (E148V, E167D, S339F, F479L, K695R, A744S, R761H) and 3 toward likely-benign (T267I, G304R, I591T). Under the more conservative rigid classifier, all seven likely-pathogenic flags were retained and two of the three likely-benign flags were preserved (G304R, I591T); only T267I shifted from likely-benign (adaptive) to indeterminate (rigid). Importantly, under ACMG/AMP and in the absence of functional, segregation and phenotype data, all of these uncommon variants are retained as VUS (Table 3); the in silico flags are exploratory and may at most help prioritise variants for further evaluation. Cross-referencing the flags against the expert INFEVERS classification (Table 3) showed broad agreement for several variants — F479L and R761H, flagged likely-pathogenic here, are classified Pathogenic in INFEVERS, and G304R and I591T, flagged likely-benign, are Benign and Likely benign, respectively — but also clear discordance for A744S and S339F, which the classifier flagged likely-pathogenic yet INFEVERS lists as Likely benign, underscoring that these in silico flags require expert-panel and functional corroboration.

Fig. 2.

Pyrin (MEFV) domain map with detected variants and the exploratory seven-tier in silico stratification of uncommon variants per the framework of Alay (2025) [17]. Common variants are shown as grey circles scaled by allele count; uncommon variants flagged toward likely-pathogenic by the adaptive classifier appear as red triangles, those flagged toward likely-benign as green inverted triangles, and variants without an exact transcript-level row (p.S179G, p.A287T, p.A89T, p.L509P) as purple squares. The Hotspot 1, Tolerant and Hotspot 2 regions defined by Alay (2025) are indicated, together with the canonical PYD, B-box, Coiled-coil and B30.2/SPRY domains. The in silico flags are exploratory and do not replace ACMG/AMP classification

Discussion

Predominance of R202Q in the detected spectrum and the Eastern Black Sea pattern

The most prominent finding in this 611-individual series is that R202Q (44.0% of detected mutant alleles; 21.3% of tested chromosomes) is the most frequently detected allele within the MEFV variant spectrum, exceeding M694V (22.3%; 10.8%). At the spectrum level, this is consistent with published Black Sea regional data [10, 11]: Akbaş et al. identified R202Q as the most frequent variant (27.9%) in the Mid-Black Sea region [11], and on a genotype level, Çapraz and Düz reported an R202Q frequency of 43.5% from Amasya [10]. However, R202Q is a variant of contested pathogenicity [12, 22], and its prominence within the detected mutant-allele pool does not, by itself, indicate the regional clinical burden of FMF or the pathogenic-allele burden. When the analysis is restricted to high-penetrance exon 10 founder variants, the pathogenic-allele burden is substantially lower (16.5% of tested chromosomes). We therefore interpret R202Q predominance as a regional detection/spectrum pattern — possibly reflecting founder effects and reduced genetic admixture relative to interior Anatolian populations — rather than as evidence of a regionally dominant pathogenic mutation.

Because many experts in this field regard R202Q as a benign polymorphism rather than a disease-causing variant [12, 22], we make no recommendation that R202Q be used as a diagnostic criterion. At most, these observations suggest that reporting R202Q may be informative for regional epidemiological and variant-spectrum characterisation in individuals referred from Black Sea provinces; any diagnostic value would need to be established by adequately phenotyped, segregation-controlled studies, which the present data cannot provide.

M694V, E148Q and other major variants

Despite R202Q’s prominence in the detected spectrum, M694V remains the most clinically established MEFV mutation in the literature. Located in the B30.2/SPRY domain of exon 10, M694V disrupts pyrin’s autoinhibitory interaction with caspase-1 and modulates IL-1β maturation [5, 7]. Multi-ethnic genotype–phenotype evidence identifies M694V — particularly in the homozygous state — as the MEFV configuration most strongly associated with severe disease and AA-type renal amyloidosis [3, 23]: Shohat et al. [24] reported amyloidosis in 20.7% of M694V homozygotes versus 4.9% of compound heterozygotes, and Grossman et al. [25], using the severity score of Mor et al. [26], showed a significantly higher proportion of severe disease in 57 M694V homozygotes (p = 0.001). Reduced colchicine responsiveness in this genotype [27] is reflected in EULAR recommendations [28] advocating maximum tolerated dosing and earlier IL-1 blockade, and amyloidogenesis is further modulated by SAA1 polymorphisms [29]. By analogy with this literature, the four M694V/M694V configurations detected here might be expected to identify a higher-risk subgroup; however, because clinical phenotype, colchicine-response and proteinuria data were not available in our cohort, we present this only as literature-based context and make no cohort-derived management recommendation.

E148Q (9.3% of detected alleles) occupies an unusual position. Although E148Q was initially questioned as a benign polymorphism, its disease-causing capacity was supported by Topaloglu et al. [30]. Mechanistically, E148Q is now understood predominantly as a disease modifier that lowers the activation threshold of the pyrin inflammasome rather than as a stand-alone driver, its disease association increasing when co-inherited with non-exon-10 variants [31]. In our series, the absence of any homozygous E148Q detection despite a 9.3% spectrum proportion, together with its frequent compound configuration, is consistent with a modifier role, although phenotype data would be required to confirm this. M680I (7.3%) and V726A (3.4%) were detected at frequencies broadly consistent with prior Turkish reports [1, 9]; both reside in exon 10. P369S and R408Q typically co-occur in cis on the same haplotype and affect pyrin’s exon 3 B-box domain; carriers of this haplotype have been reported to present with atypical FMF or PFAPA-overlap phenotypes rather than classical disease [32], a consideration relevant to the 17 P369S and 17 R408Q alleles detected here [3, 22].

Compound heterozygosity and its interpretation

A notable feature of this dataset is the substantial proportion of compound heterozygous or complex multi-variant genotypes (25.7% of the cohort; 41.4% of variant-positive individuals), consistent with the higher biallelic yield of NGS-based or expanded-panel testing [33] when exon 2 variants such as R202Q and E148Q are included. A critical limitation, however, is that cis/trans phase was not determined: parental segregation testing was not available, so we cannot establish whether co-detected variants lie on the same or opposite alleles. Consequently, the clinical implications of specific biallelic configurations cannot be inferred from these data, and the following points are presented only in light of published frameworks rather than as conclusions about our cohort. The INSAID framework [13] recognises that compound genotypes containing at least one pathogenic exon 10 allele confer the highest risk of classical disease, whereas genotypes composed exclusively of exon 2 variants represent a lower-penetrance category [13, 22], and the Eurofever/PRINTO criteria [19] provide a complementary clinical framework. The recessive paradigm has also been challenged by reports of clinical FMF in single-mutation carriers [34] and proposals of non-classical recessive and atypical dominant models [35, 36]. In our referral-defined cohort, these inheritance considerations cannot be evaluated without phenotype and segregation data.

Uncommon variants and a note on p.Ser179Gly

Beyond the canonical founder mutations, the panel identified 14 uncommon variants. Several — including G304R, I591T and A744S — have been described in prior Turkish or European cohorts and carry varying levels of evidence in INFEVERS [37]. With respect to variants apparently confined to this Eastern Black Sea referral series, the candidate substitution p.Ser179Gly was not identified in the allele-level comparator cohorts or, to our knowledge, in prior peer-reviewed MEFV reports; the remaining uncommon variants have all been described previously in Turkish or international series. One variant warrants particular caution: p.Ser179Gly (c.535 A > G), detected in a single individual in apparent compound configuration with R202Q. A literature search at the time of writing did not identify a prior peer-reviewed clinical report of this specific substitution; the neighbouring substitution p.Ser179Asn (c.536G > A; rs104895125) is registered in ClinVar [38] with conflicting evidence. We make no claim regarding the pathogenicity of p.Ser179Gly. Because it was detected in only one individual, without orthogonal (Sanger) or familial validation, it is reported here strictly as a candidate variant of uncertain significance; a single previously unreported NGS call cannot be distinguished from a technical artefact without confirmatory testing, and we therefore draw no inference from it. Formal registration in INFEVERS [37] and ClinVar [38] as a VUS, together with confirmatory and segregation testing and longitudinal follow-up of the carrier, is recommended for future work.

Exploratory in silico stratification using the seven-tier framework

The 2015 ACMG/AMP framework [14] treats VOUS as a single broad bin that obscures the gradient of pathogenic likelihood, and gene-specific recalibration of in silico predictors has been shown to improve MEFV variant assessment [15, 16]. The Alay seven-tier framework [17] extends this by integrating 42 in silico tools through ensemble machine learning. We applied it strictly as an exploratory in silico layer, supplementary to and not a replacement for ACMG/AMP classification. Applied to our 14 uncommon variants, the adaptive classifier flagged 10 evaluable variants (7 toward likely-pathogenic, 3 toward likely-benign) and the rigid classifier preserved all seven likely-pathogenic flags and two of three likely-benign flags; only T267I shifted to indeterminate. We emphasise that these are exploratory in silico flags: under ACMG/AMP, and in the absence of functional, segregation and phenotype data, these uncommon variants remain VUS (Table 3). The flags may at most help prioritise variants for further evaluation (functional studies, segregation analysis, phenotype correlation) and do not constitute clinical reclassification. A practical caveat is that four variants (p.S179G, p.A287T, p.A89T, p.L509P) had no exact row in the deployed application, reflecting the framework’s dependence on the underlying dbNSFP/Ensembl catalogue [17]; to avoid ambiguity from differing protein-numbering conventions, all variants were anchored to their genomic coordinate (GRCh37/hg19) and c.HGVS on NM_000243.3 before the seven-tier outputs were recorded.

The exploratory status of these in silico flags is underscored by newly available high-throughput functional data. Using SpeckSeq — a method coupling DNA barcoding, ASC speck-based single-cell sorting and next-generation sequencing — Bronnec et al. [39] functionally stratified a large panel of MEFV variants and identified 49 gain-of-function (hypermorphic) substitutions separated into familial-Mediterranean-fever-associated and PAAND-associated groups, and used these data to support the reclassification of several previously uncertain variants, in some cases yielding new diagnoses. Because disease-causing MEFV variants act as hypermorphic, gain-of-function alleles that lower the activation threshold of the pyrin inflammasome, such functional readouts supply orthogonal evidence that in silico predictors cannot. Where the uncommon variants flagged here overlap with the SpeckSeq dataset, concordance between an in silico likely-pathogenic flag and a hypermorphic functional classification would materially strengthen the case for pathogenicity, whereas a functionally neutral result would argue against it; conversely, variants flagged as likely-benign that are absent from the hypermorphic set would be corroborated. We were unable to perform variant-level functional testing, and we therefore recommend that the flags reported here be cross-referenced against functional resources such as SpeckSeq [39] and against the curated INFEVERS classifications [37] before any clinical use.

Limitations

This study has several limitations. First and most importantly, the cohort comprises individuals tested for suspected FMF rather than clinically confirmed patients: standardised diagnostic/classification criteria (Tel-Hashomer [18]; Eurofever/PRINTO [19]), attack characteristics, colchicine response, family history and amyloidosis status were not available, which precludes genotype–phenotype correlation and substantially limits clinical interpretability. The findings should therefore be read as describing a molecular variant spectrum, not clinical disease. Second, cis/trans phase was not determined, as parental/familial segregation testing was not available; the interpretation of compound and exon 2 variant combinations is therefore limited to inference. Third, the molecular methodology — targeted Illumina sequencing of the MEFV coding exons and ± 10 bp flanking junctions with a 20-read minimum threshold — was not designed to detect deep intronic variants, large copy-number alterations or structural variants, which may be present but unascertained. Fourth, the in silico stratification is probabilistic and depends on the upstream dbNSFP/Ensembl catalogues; variants absent from those catalogues (p.S179G, p.A287T, p.A89T, p.L509P) cannot be ranked. Fifth, the previously unreported p.Ser179Gly was detected in a single individual without orthogonal or familial validation, so no pathogenicity inference can be drawn. Sixth, cross-study comparisons are complicated by inconsistencies in reporting conventions (allele-level vs. genotype-level frequencies, differing panel coverage), which we have addressed by restricting Table 2 to metric-consistent allele-level studies. Despite these limitations, this is, to our knowledge, the largest MEFV variant dataset reported from this Eastern Black Sea province and the first to characterise its variant spectrum together with an exploratory gene-specific in silico stratification.

Conclusions

In 611 consecutive individuals tested for suspected FMF in this Eastern Black Sea province of Türkiye, R202Q is the most frequently detected MEFV allele in the variant spectrum (44.0% of detected mutant alleles; 21.3% of tested chromosomes), consistent with the distinct spectrum of the Black Sea region. Because the pathogenicity of R202Q is contested, this spectrum dominance should not be equated with regional clinical or pathogenic-allele burden; the high-penetrance exon 10 founder burden is markedly lower (16.5% of tested chromosomes). The substantial proportion of compound/complex genotypes (25.7%) is consistent with the higher yield of multi-variant or NGS-based testing, although cis/trans phase was not determined. Exploratory application of the seven-tier framework [17] flagged 10 of 14 uncommon variants toward likely-pathogenic or likely-benign categories; under ACMG/AMP these remain VUS, and the in silico flags should be regarded only as a tool to prioritise variants for further evaluation, not as clinical reclassification. All clinical interpretation is constrained by the absence of phenotype, segregation and standardised diagnostic data, and the molecular findings reported here should be interpreted accordingly.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

ÇD (Çağrı Doğan) is the sole author and contributed to all aspects of the work: conceptualization, methodology, investigation, data curation, formal analysis, visualization, and writing (original draft, review and editing). The author read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The dataset supporting the conclusions of this article is included within the article and its additional file (Additional file 2, ESM_2.xlsx). The complete de-identified per-allele dataset is also openly available in the Zenodo repository: Doğan Ç. MEFV variant dataset — Eastern Black Sea region of Türkiye (611 individuals). Zenodo; 2026. https://doi.org/10.5281/zenodo.20410810.

Declarations

Ethics approval and consent to participate

The study was approved by the Ordu University Non-Interventional Clinical Research Ethics Committee (decision no. 2025/423, dated 12 December 2025) and was conducted in accordance with the Declaration of Helsinki. Given the retrospective design and use of fully anonymised molecular reports, the committee waived the requirement for individual informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.