Global epidemiology of Familial Mediterranean fever mutations using population exome sequences

Familial Mediterranean fever (FMF) is an inherited disorder characterized by recurrent episodes of fever accompanied by sterile peritonitis, arthritis, and pleuritis. Many mutations in the MEFV gene have been identified as causing FMF. However, accompanying epidemiological information remains quite scarce except in some Mediterranean countries, and the degree of penetrance has been a subject of controversy. Here, I established a genetic epidemiology of full FMF mutations using two population exome studies. Of 57 mutations associated with FMF, 22 were detected in a total of 9007 individuals from two exome sequences. Exome-based epidemiology revealed the carrier rates of FMF in 28 populations in 19 countries by individual mutation and showed strong population specificity for the MEFV mutations. Unexpectedly high carrier rates suggested that some mutations are benign variants with no pathological significance and highlighted the need for caution in analyzing MEFV mutations. Similar approach could be used to uncover the incomplete or no penetrance of mutations in most inherited disorders.

The causative gene for FMF, MEFV, was first identified by two independent groups in 1997 (The French International FMF consortium 1997;The International FMF Consortium 1997). The MEFV gene is located on chromosome 16p13.3 and encodes a 781-amino acid protein (10 exons) known as pyrin (The French International FMF consortium 1997;The International FMF Consortium 1997). Previous studies on FMF patients and animal models suggest that MEFV mutations lead to gain of pyrin function, resulting in increased IL-1b secretion by monocytes and a prolonged inflammatory response when stimulated with lipopolysaccharide (Chae et al. 2011;Omenetti et al. 2014).
To date, over 50 MEFV mutations have been identified in FMF patients (Touitou 2001;Giancane et al. 2015). The five founder mutations, E148Q, M680I, M694I, M694V, and V726A, were reported to account for approximate 70% of cases of FMF from the Mediterranean origin and the nonfounder mutations would constitute the remaining proportion (Touitou 2001) although the pathogenic role of some mutations including E148Q remains debatable (Tchernitchko et al. 2003(Tchernitchko et al. , 2006Giancane et al. 2015). Draft guidelines for the genetic diagnosis of FMF were also prepared based on the current practice and data from the literature (Shinar et al. 2012). However, despite the recent advances in FMF studies, the epidemiological information is still insufficient and inconclusive, except in some Mediterranean countries. In addition, the degree of penetrance of many mutations has not been a subject of research for many years. The aim in this study was to establish the global epidemiology of autosomal recessive FMF (MIM# 249100) mutations using the population exome sequences and to evaluate the penetrance of each MEFV mutation.

Pairwise proportion tests of data consistency between two different exome resources
To project the performance of risk prediction based on analyses of exome sequence studies, exome-based estimates were statistically compared with the clinical prevalence survey. Evidence of data consistency was based on significant differences in pairwise comparisons between populations if two estimates differed significantly (two-sample test for equality of proportions with continuity correction). The standard hypothesis test was H 0 : p 1 = p 2 against the alternative (two-sided) H 1 : p 1 6 ¼ p 2 . The pairwise prop test can be used to test the null hypothesis that the proportions (probabilities of success) in two groups are the same. This test is referred to as a z-test because the statistics were as follows: wherep ¼ ðp 1 þ p 2 Þ=ðn 1 þ n 2 Þ and indices (1, 2) refer to the first and second line of the table. In a two-way con-tingency table where H 0 : p 1 = p 2 , this should yield comparable results to those of the ordinary v 2 test.

Results
Strategy for epidemiological research on MEFV mutations using population exome sequences As a first step toward determining the genetic epidemiology of recessive FMF, genotyping datasets were collected from the 1000G and NHLBI projects for the causative variations (Fig. 1) (Milhavet et al. 2008), but only a subset is found in patients with typical FMF disease and others are benign polymorphisms including synonymous and nonsynonymous variants. Of 57 MEFV mutations for FMF, 22 were detected in one or both of the two population exome datasets ( Fig. 1) (the list of MEFV mutations was retrieved from NCBI OMIM, INFEVERS, and PubMed; see also supplementary references). These mutations were present above the minimum variant allele frequency threshold (5.55E-05). MEFV mutations were classified by mutation type, derived allele frequency (DAF), countries, racial group, and clinical impact ( Fig. 1 and Table 2). The other 35 mutations were not detected owing to their low DAF in a total of 9007 individuals (Fig. 1).

Carrier rate variability by mutation type and ethnicity
Among the 22 MEFV mutations detected, the most common was E148Q, with a frequency of 1 in 22.7 (4.40%) ( Table 2). The allele frequencies of the other four founder mutations were as follows: M680I, 0.00555%; M694I, 0%; M694V, 0.0222%; and V726A, 0.139% ( Table 2). The prevalence rates are theoretically estimated to be equivalent to the squares of these carrier rates. Previous studies demonstrate that E148Q is not likely to be a disease-causing mutation (Tchernitchko et al. 2003(Tchernitchko et al. , 2006Giancane et al. 2015) and the high carrier rate of E148Q in this research (Table 2) was consistent with the previous data. Thus, E148Q is excluded from following analysis.
Carrier frequencies for disease-causing variants vary significantly by racial and ethnic groups (Lazarin et al. 2013). Figure 2 shows a global map of the DAF distribution of the 21 MEFV mutations for FMF except E148Q. Among the other four founder mutations, M680I was Caucasian-specific, and V726A was much more prevalent in Caucasian (Table 2). In East Asian populations, the average DAF was 18.7% (CDX, 19.4%; CHB, 17.5%; CHS, 17.6%; JPT, 19.7%) ( Fig. 2 and Table 2). Also, the major variant in Europeans (NHLBI: 1.52%; 1000G: 1.79%) and Hispanics (1000G: 0.432%) was I591T. A recently available database, ExAC, with more than 60,000 exomes was additionally searched for 57 MEFV mutations (Table 3) although this database does not provide the detailed information about country and ethnic group of individuals. A total of 31 MEFV mutations were detected (Table 3) and the similar DAF distribution was obtained between ExAC and 1000 Genome + NHLBI data (  A flowchart used to study epidemiology shows the process of mutation detection using the 1000G and NHLBI datasets. A total of 18,014 alleles were screened for 57 MEFV mutations linked to FMF. A schematic representation of FMF-associated mutations in the MEFV gene is shown below. MEFV mutations most frequently found in FMF patients reside in exon 10 (depicted in green), which encodes the C-terminal PRYSPRY domain. Exons 2, 3, and 5 contain a substantial number of rare mutations (depicted in green). The five founder mutations are indicated in red. # R408Q and P369S mutations had been reported in cis as a single allele resulting in a variable clinical symptoms and R408Q is not likely to be a disease-causing variant (Bell et al. 2011).

Consistency of data between two different exomes
To determine the validity of this methodology, I examined the extent of differences in the two exome-based prevalence rates by comparing DAFs in African and European ancestries between the 1000G and NHLBI datasets (Table S1). A pairwise proportions test was used to test the null hypothesis that the proportions in the two estimates were identical. This formula is referred to as a Z test because the statistic is as follows: wherep ¼ ðp 1 þ p 2 Þ=ðn 1 þ n 2 Þ and the indices (1, 2) refer to the first and second lines of the table. A pairwise proportion test between the two exome resources showed no significant differences between the two different exomes (43 cases; P ≫ 0.05), except in one case (P < 0.05) (Table  S1). This finding suggests that exome-based predictions are free of most confounding factors (such as diagnostic criteria, instrument, skill, and screening rate) and may be a more objective indicator.

Discussion
Although genetic research into FMF began in 1990s (The French International FMF consortium 1997; The International FMF Consortium 1997), we still lack a complete picture of its genetic variation, carrier frequency, and penetrance. Exome-based epidemiology is a promising alternative method for genetic epidemiology because it provides information on both common and rare mutations in large numbers of individuals. Using exome data from a total of 9007 individuals from the two largest population exomes, I established a reliable epidemiology of FMF mutations with a small margin of error.
Estimated prevalence rates are considerably higher than those seen in clinical practice There are only little data of carrier frequencies for comparison with this study because previous studies analyzed individuals mainly from some Mediterranean countries. Hofer et al. (2006) reported FMF prevalence rates in individuals from Western countries on a mass scale. The relative frequencies of the mutations found in this study [NHLBI EA, 4.98% (3.87%; except E148Q); 1000G EUR, 5.27% (4.38%; except E148Q) ( Table 2)] are considerably higher than the results of Hofer et al. (2006) who reported a prevalence of as 2.5 per 100,000 (0.004%) people even if E148Q is excluded in this research. In another study in Japanese populations, carrier frequencies [1000G JPT, 41.3% (19.7%; except E148Q) ( Table 2)] were much higher than the clinical incidence rate (0.000417%; http://www.nanbyou.or.jp/entry/3238). The variant T267I was frequently detected in Bangladesh (0.581% Table 2), where FMF patients were not reported. These results suggest that some MEFV mutations, including E84K, L110P, E148Q, T267I, P369S, R408Q, S503C, and I591T, are polymorphisms, not disease-causing mutations. It has been proposed that E148Q is likely to be a polymorphism, not a disease-causing mutation, and has low penetrance (Tchernitchko et al. 2003(Tchernitchko et al. , 2006 and my findings agree with these research results. Furthermore, the result here suggested the nonpenetrance of some other mutations. There is often the case where the causative mutations are determined too easily without analyzing potential effect of mutations (Cooper et al. 2013;van Rheenen et al. 2014;Siemiatkowska et al. 2014). Discordance between DAF and incidence rate may be caused by another infrequent mutation closely linked to these mutations. Another promising hypothesis is that FMF is a multifactorial and polygenic genetic disorder associated with the effects of multiple genes in combination with lifestyle and environmental factors.
In conclusion, exome-based epidemiology revealed the country-by-country carrier rates of FMF with respect to each mutation and provided a clue to understand the penetrance and screening priority of each mutation. An unexpectedly high carrier rate of FMF in Europeans and Asians raises the strong possibility that some MEFV mutations may be benign variants with few or no pathological significance. This study highlights the need for caution in interpreting genetic tests in FMF patients. Similar method could be used to uncover the incomplete or no penetrance of mutations in genetic disorders.