Dr. Cazeneuve is the recipient of a fellowship from INSERM.
Familial Mediterranean fever among patients from Karabakh and the diagnostic value of MEFV gene analysis in all classically affected populations
Version of Record online: 1 AUG 2003
Copyright © 2003 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 48, Issue 8, pages 2324–2331, August 2003
How to Cite
Cazeneuve, C., Hovannesyan, Z., Geneviève, D., Hayrapetyan, H., Papin, S., Girodon-Boulandet, E., Boissier, B., Feingold, J., Atayan, K., Sarkisian, T. and Amselem, S. (2003), Familial Mediterranean fever among patients from Karabakh and the diagnostic value of MEFV gene analysis in all classically affected populations. Arthritis & Rheumatism, 48: 2324–2331. doi: 10.1002/art.11102
- Issue online: 1 AUG 2003
- Version of Record online: 1 AUG 2003
- Manuscript Accepted: 14 APR 2003
- Manuscript Received: 30 JAN 2003
- NATO. Grant Number: LST.CLG.978883
- AFM/INSERM (Réseau de Recherche sur les Maladies Rares)
- Calouste Gulbenkian Foundation (Lisbon, Portugal)
Familial Mediterranean fever (FMF) is an autosomal-recessive disorder that is common in Armenian, Turkish, Arab, and Sephardic Jewish populations. Its clinical diagnosis is one of exclusion, with the patients displaying nonspecific symptoms related to serosal inflammation. MEFV gene analysis has provided the first objective diagnostic criterion for FMF. However, in the absence of an identified mutation (NI/NI genotype), both the sensitivity of the molecular analyses and the involvement of the MEFV gene in FMF are called into question. The present study was designed to further evaluate the diagnostic value of MEFV analysis in another population of Mediterranean extraction.
The MEFV gene was screened for mutations in 50 patients living in Karabakh (near Armenia) who fulfilled the established criteria for FMF. In addition, we analyzed published series of patients from the above-mentioned at-risk populations.
The mutation spectrum in Karabakhian patients, which consisted of only 6 mutations (with 26% of NI alleles), differed from that reported in Armenian patients. Strikingly, among patients from Karabakh and among all classically affected populations, the distribution of genotypes differed dramatically from Hardy-Weinberg equilibrium (P = 0.0016 and P < 0.00001, respectively). These results, combined with other population genetics–based data, revealed the existence of an FMF-like condition that, depending on the patients' ancestry, was shown to affect 85–99% of those with the NI/NI genotype.
These data illuminate the meaning of negative results of MEFV analyses and show that in all populations evaluated, most patients with the NI/NI genotype had disease that mimicked FMF and was unrelated to the MEFV gene. Our findings also demonstrate the high sensitivity of a search for very few mutations in order to perform a molecular diagnosis of MEFV-related FMF.
One major aim of the first molecular studies performed in populations affected by familial Mediterranean fever (FMF) was to test the diagnostic value of the analysis of MEFV (1, 2), a strong candidate gene for this autosomal-recessive condition (MIM 249100) that primarily affects Armenian, Sephardic Jewish, Turkish, and Arab populations (3). The diagnosis of FMF is indeed one of exclusion, based on established clinical criteria (4). FMF patients display recurrent episodes of fever and serosal inflammation (manifested by nonspecific signs, such as sterile arthritis, peritonitis, and/or pleurisy) associated with an increased erythrocyte sedimentation rate and increased serum levels of acute-phase proteins. However, establishing the diagnosis of FMF is essential, mainly because the signs and symptoms of FMF may lead to unnecessary exploratory surgery. Moreover, daily administration of colchicine controls the symptoms in most patients with FMF (5, 6) and, most important, prevents the occurrence of renal amyloidosis (7), the main complication of FMF, whose occurrence is influenced by MEFV genotype and MEFV-independent modifying genetic factors (8).
In the course of the first molecular investigations performed in patients with FMF, we showed that characterization of mutated MEFV alleles represents the first objective diagnostic criterion for this disease. Indeed, we identified 2 MEFV allele mutations in 89% of a large and unselected cohort of unrelated Armenian patients (9), who fulfilled the established clinical criteria for the diagnosis of FMF. It was noteworthy, however, that in 11% of the patients, either no mutation was detected or only 1 FMF allele was characterized, even after the screening of all 10 MEFV coding regions and intron boundaries for mutations by denaturing gradient gel electrophoresis (DGGE), a highly sensitive technique for the detection of mutations (10). This latter observation left open the possibility that, in some cases, the signs and symptoms of FMF documented in patients of Armenian ancestry did not result from mutations in the MEFV gene. Thus, an investigation of other affected populations might provide information of critical importance in confirming or excluding this hypothesis. Except for our previous study (9), however, none of the studies performed thus far in large homogeneous populations was based on exhaustive screening for mutations of all MEFV exons. Indeed, the latter studies focused on the search for few known mutations, and, since all of the studies described patients with uncharacterized FMF alleles, the possibility that MEFV mutations remained undetected cannot be ruled out.
In the present study, we further tested the diagnostic value of MEFV gene analysis and delineated the mutation spectrum in another population of FMF patients of Mediterranean extraction. We evaluated a cohort of unrelated patients living in Karabakh, which is near Armenia and is inhabited mainly by people of Armenian ancestry. As in Armenia, the prevalence of FMF in this area is high (Atayan K, Sarkisian T: unpublished observations). Considering the frequency of uncharacterized MEFV alleles, the unexpectedly low number of Karabakhian patients with only 1 identified MEFV mutation led us to test whether the MEFV genotype distribution was consistent with Hardy-Weinberg equilibrium (i.e., binomial distribution). The data obtained prompted us to perform a detailed analysis of 4 previous reports on various populations classically affected by FMF: subjects of Armenian (9), Sephardic Jewish (11), Turkish (12), and Arab (13) extraction. The results of this evaluation challenge both the diagnostic value of MEFV gene analysis in all tested populations and the molecular basis of this hereditary inflammatory disorder.
PATIENTS AND METHODS
We evaluated 50 consecutive FMF patients from 39 unrelated families living in Karabakh, representing a total of 84 independent alleles. All of the patients belong to families in which there was no known consanguinity. None of the families was selected through genetic linkage analyses using MEFV gene markers. The age range of the patients was 4–71 years (mean 25.0 years). The ratio of males to females was 0.61. Clinical features were recorded by the same investigator (KA) using a standardized form. The diagnosis of FMF was made according to established clinical criteria (4). Informed consent was given by all patients or their parents.
The mutation analysis was performed as described previously (9). Briefly, the M694V and V726A mutations were searched for by restriction enzyme analyses. MEFV gene analysis was completed in patients who had, at most, only 1 of these 2 mutations, by DGGE screening of all 10 coding exons and flanking intron sequences. Samples displaying a shift in mobility were directly sequenced.
The spectrum of MEFV gene mutations in Karabakh and in Armenia (9) was compared with the chi-square test (using the Yates' adjustment when appropriate). We calculated p1 and q1, the frequency of identified (I) and nonidentified (NI) mutations among Karabakhian patients, using the classic counting method.
where nI/I is the number of patients carrying an I/I genotype, nI/NI the number carrying an I/NI genotype, nNI/NI the number carrying an NI/NI genotype, and NTotal is the total number of patients (NTotal = nI/I + nI/NI + nNI/NI).
The expected distributions of I/I, I/NI, and NI/NI genotypes were then calculated according to the assumption of Hardy-Weinberg equilibrium. The observed and expected distributions of the 3 genotypes were compared using the chi-square test. Similar calculations were performed from the data in 4 published reports that focused on the investigation of patients belonging to populations classically affected by FMF (i.e., Armenians , Sephardic Jews , Turks , and Arabs ).
The results of these statistical analyses, combined with population genetics–based data (see Discussion), led us to conclude that the signs and symptoms of FMF presented by several patients did not result from mutations in the MEFV gene. To evaluate the number of these latter patients (NOther), we first calculated the number of patients whose phenotype did result from mutations in MEFV (NMEFV). To this end, we considered that the symptoms were related to MEFV in patients carrying 2 mutated alleles (I/I genotype). In addition, assuming that some mutated alleles might have escaped our molecular screening procedure, we postulated that the symptoms were also related to the MEFV gene in all patients carrying an I/NI genotype and in an undetermined number of patients carrying an NI/NI genotype; the NI alleles in these patients therefore correspond to uncharacterized mutated MEFV alleles. Designating p2 as the frequency of identified mutations and q2 as the frequency of nonidentified mutation among the NMEFV patients, with q2 = 1 – p2 and p1 ≠ p2, then
Given 1 and 2, we have the following:
Thus, NMEFV is equal to
and NOther is equal to
The spectrum of MEFV gene mutations in patients from Karabakh is detailed in Table 1. Six different mutations and several polymorphisms (data not shown) were identified in our population sample. These mutations accounted for 74% of the 84 independent FMF alleles. M694V, M680I, and V726A substitutions represented 68%, 16%, and 5% of the 62 characterized alleles, respectively. Thirteen different genotypes were characterized among 45 of the 50 patients (Table 2). Both FMF alleles were characterized in 30 of the 45 patients (67%). Both FMF alleles remained uncharacterized in 7 patients, whereas 1 FMF allele was uncharacterized in 8 patients (Table 2).
|Mutation||No. (%) of independent alleles|
|Genotype*||No. (%) of patients|
We then compared the spectrum of MEFV gene mutations among Karabakhian patients with the spectrum reported among Armenian patients (9). The frequency of nonidentified MEFV alleles was much higher in patients from Karabakh than in patients from Armenia (26% [22 of 84] versus 7% [12 of 163]; P = 0.00005) (Figure 1A). The spectrum of identified mutations in Karabakhian and Armenian patients also differed (Figure 1B). Indeed, the M694V and R761H mutations were more frequent among Karabakhian patients (P = 0.01 and P = 0.04, respectively). Conversely, the V726A mutation represented only 5% of the identified alleles in Karabakhian patients, whereas it accounted for 26% of the characterized FMF alleles in Armenian patients (P = 0.0003). Only 1 mutation (M680I) was found at a similar frequency in the 2 populations. No new MEFV mutation was identified among the Karabakhian patients.
One striking observation was the relatively low number of Karabakhian patients with only 1 identified MEFV mutation and the relatively high number of patients with 2 uncharacterized MEFV alleles (Table 2), with respect to the high frequency of uncharacterized FMF alleles (Table 1). This unexpected finding prompted us to test whether the distribution of I/I (2 identified MEFV mutations), I/NI (1 identified MEFV mutation), and NI/NI (no identified MEFV mutation) genotypes complied with Hardy-Weinberg equilibrium.
This analysis revealed that the distribution of genotypes among Karabakhian patients differed significantly from a binomial distribution (P = 0.0016) (Figure 2). A similar analysis performed in the neighboring Armenian population of patients demonstrated the same departure from Hardy-Weinberg equilibrium (P < 0.00001) (Figure 2), whereas the distribution of the most common MEFV mutations in these 2 populations did not significantly differ from Hardy-Weinberg expectations (P = 0.053 and P = 0.796 for M694V and M680I in Karabakhian patients; P = 0.092, P = 0.446, and P = 0.616 for M694V, V726A, and M680I, respectively, in Armenian patients).
To determine whether this departure in the distribution of I/I, I/NI, and NI/NI genotypes from Hardy-Weinberg equilibrium was specific to the Karabakhian and Armenian populations of FMF patients or was also present among the other affected populations, we used a similar approach to test the distribution of MEFV genotypes in published series of relatively homogeneous populations of patients (11–13). Strikingly, in all tested series, the observed distribution of genotypes differed dramatically from Hardy-Weinberg expectations (P < 0.00001) (Figure 2). Importantly, in each of these populations of different ethnic origins, the disequilibrium was due to an important paucity of patients with the I/NI genotype and, conversely, a relative excess of patients with no identified mutation and with 2 characterized MEFV mutations, although to a lesser extent.
This finding reveals the existence of an FMF-like condition that is unrelated to the MEFV gene (see Discussion). Thus, we were prompted to estimate the number of patients whose phenotype did not result from mutations in the MEFV gene (see Patients and Methods). Our calculations showed that, depending on the series being evaluated, this proportion varied from 7% to 21%, reaching 14% among Karabakhian patients (Table 3). Moreover, the proportion of patients with an NI/NI MEFV genotype whose phenotype could not be explained by mutations in the MEFV gene ranged from 85% among Turkish patients to 99% among Arab patients (Table 3).
|Origin of the patients||Author, year (ref.)||nI/I||nI/NI||nNI/NI||NTotal||NMEFV||NOther||NOther||NOther|
|Armenian||Cazeneuve et al, 1999 (9) and present study†||126||11||10||147||137.2||9.8||0.07||0.98|
|Sephardic Jewish||Livneh et al, 1999 (11)||133||31||14||178||165.8||12.2||0.07||0.87|
|Turkish||Akar et al, 1999 (12)||139||55||36||230||199.4||30.6||0.13||0.85|
|Arab||Shinawi et al, 2000 (13)||45||6||14||65||51.2||13.8||0.21||0.99|
In this study, which was designed to further test the diagnostic value of MEFV gene analysis in patients with FMF, we evaluated a cohort of patients from Karabakh and analyzed the genetic data obtained in populations classically affected by FMF. In all cases, we observed a departure from Hardy-Weinberg equilibrium in the MEFV genotype distribution, which resulted essentially from a paucity of patients with only 1 characterized mutation and an excess of patients with no identified mutation. This finding, in combination with other population genetics–based data, reveals the existence of an FMF-like condition that is unrelated to the MEFV gene, and sheds new light on the diagnostic value of MEFV analysis.
Delineation of the MEFV mutation spectrum in Karabakhian patients revealed a high percentage of NI alleles (26%), as compared with that reported among Armenian patients (7%) (9). Since no particular MEFV haplotype was overrepresented among the NI alleles (data not shown), we excluded the existence of a unique and rare MEFV mutation resulting from a founder effect, which would have escaped our mutation screening procedure. When considering the characterized mutated alleles, 3 of the 6 MEFV mutations identified in Karabakhian patients (i.e., M694V, V726A, and R761H) were found at different frequencies in the Armenian patients. Together, these results show that the MEFV mutation spectrum differs significantly between these 2 populations living in neighboring areas, an observation that may result from a founder effect and the relatively independent evolution of each population since the beginning of the twentieth century.
Most important, the distribution of I/I, I/NI, and NI/NI genotypes at the MEFV locus differed dramatically from a binomial distribution among Karabakhian patients, as well as among patients of Armenian, Turkish, Arab, or Sephardic Jewish ancestry (Figure 2). At least 3 hypotheses may account for this observation: frequent consanguineous unions in these populations, a possible recruitment bias, and the existence of an FMF phenotype unrelated to MEFV in several patients with no identified mutation.
Many lines of evidence strongly oppose the first 2 hypotheses. There was no known consanguinity in the Armenian and Karabakhian patient populations. Most important, the distribution of the most common MEFV mutations in these 2 populations did not differ significantly from Hardy-Weinberg expectations. This observation is evidence against a selection bias that would have favored the recruitment of severely affected patients into the study. It also argues against the hypothesis of an increased embryonic death of zygotes with 2 severe MEFV mutations; indeed, under such a hypothesis, the number of patients carrying these mutations would be much lower than would be expected from Hardy-Weinberg equilibrium, a situation reported in other conditions, such as congenital disorders of glycosylation type Ia (14). The binomial distribution of the most common MEFV mutations, therefore, reflects random inclusion of FMF patients in the above-mentioned studies and further supports the fact that in these population samples, the requirements for Hardy-Weinberg equilibrium were indeed fulfilled.
It is important to emphasize that the deviations from Hardy-Weinberg equilibrium documented in all the population samples evaluated herein cannot be explained by the existence of unidentified MEFV mutations that would have escaped the mutation search. Indeed, in this latter situation, whatever the molecular diagnostic procedure (i.e., complete screening for all MEFV exons [ref.9 and the present study] or search for a limited number of known mutations [11–13]), the NI alleles should be normally distributed. This was not the case (Figure 2). Taken together, these data allow us to reject the first two hypotheses. We therefore conclude that, according to the third hypothesis, the departure from Hardy-Weinberg equilibrium documented in these populations of FMF patients reveals the existence of an FMF-like phenotype that is unrelated to MEFV in several patients with no identified MEFV mutation.
Although the patients evaluated in each series fulfilled the established clinical criteria for a diagnosis of FMF, the implication of MEFV in their disease phenotype can be ascertained only in those who had 2 identified MEFV mutations. Most of the remaining patients may therefore have other conditions, including conditions that possibly result from mutations at other loci. The family history of FMF among the relatives of several Karabakhian patients who had no identified MEFV mutation (data not shown) further supports the latter hypothesis. The existence of a locus heterogeneity in FMF has already been suggested from the findings of evaluations of 2 Turkish families (15). However, that hypothesis was based on a segregation analysis of polymorphic markers of the 16p13 region, and not on the analysis of the MEFV gene, which was unknown at that time. Under these experimental conditions, an incomplete penetrance of the disease phenotype as well as intrafamilial MEFV allelic heterogeneity—two well-known features of FMF that have been recently documented in all classically affected populations (9, 16–19)—could easily account for the absence of linkage between those markers and the disease.
Having shown that the Hardy-Weinberg disequilibrium documented in the present study indeed reflects the existence of a group of patients whose phenotype does not result from mutations in the MEFV gene, we estimated the number of these patients. Depending on the patients' ancestry, the estimated proportion of NI/NI patients whose phenotype did not result from MEFV mutations was found to vary from 85% to 99% (i.e., 7–21% of the patients with a clinical diagnosis of FMF). It must be kept in mind that in each series, this proportion is underestimated. Indeed, given the high frequency of healthy individuals with an MEFV mutation in the heterozygous state in the 4 at-risk populations (13, 16, 20, 21), the phenotype of several patients with an I/NI genotype at the MEFV locus probably does not result from mutations in the MEFV gene.
What are the consequences of these findings for the diagnosis of FMF? First, our results should be considered in light of the efficiency of the molecular strategies for detecting 2 mutated MEFV alleles that were used by the different groups of investigators. Although the MEFV mutation spectrum differs among classically affected populations, the routine diagnostic procedure is based on a search for a limited number of known mutations. As shown in Table 4, this strategy leaves 7.9–21.5% of the patients with 2 uncharacterized FMF alleles, depending on the ancestry of the patients. In this regard, the complete screening of all coding sequences and intron boundaries performed in the Armenian patients appears to be particularly efficient, since the 2 FMF alleles remained uncharacterized in only 3.5% of those patients (9).
|Origin of the patients||Author, year (ref.)||Diagnostic strategy||% of patients|
|Two identified MEFV mutations (I/I)||No identified MEFV mutations (NI/NI)|
|Armenian||Cazeneuve et al, 1999 (9)||DGGE*||89.4||3.5|
|Sephardic Jewish||Livneh et al, 1999 (11)||Search for 4 mutations†||74.7||7.9|
|Turkish||Akar et al, 1999 (12)||Search for 7 mutations‡||60.4||15.7|
|Arab||Shinawi et al, 2000 (13)||Search for 4 mutations§||69.2||21.5|
At first glance, this observation may encourage the use of diagnostic procedures based on the analysis of all 10 MEFV exons. However, using this strategy, we showed that up to 15.6% of the Karabakhian patients had no identified MEFV mutation. Furthermore, the a posteriori analysis of the MEFV mutation spectrum in the Armenian (9) and Karabakhian populations indicates that the search for 6 mutations (i.e., M694V, V726A, M680I, F479L, R761H, and E148Q) would have provided a similar rate of identified mutations. Taken together, these data do not support the time-consuming and costly diagnostic strategy based on a complete screening of all coding sequences and intron boundaries. Second, and most important, in each affected population, the estimation of the number of patients whose phenotype did not result from mutations in MEFV (NOther) allowed us to calculate the proportion of such patients among individuals with an NI/NI genotype (Table 3). Strikingly, whatever the ancestry of the populations and whatever the diagnostic procedure, this ratio (NOther/nNI/NI) is particularly high, ranging from 0.85 to 0.99. The proportion reaches 0.99 in the Arab population; such a high proportion in this population is all the more since striking, first, only 69.2% of the patients carried mutated MEFV alleles—as identified by the search for only 4 mutations (Table 4)—and second, according to our calculations, at least 21% of these patients presented with an FMF-like phenotype that did not result from mutations in MEFV (Table 3).
From a more general viewpoint, the consequences of this study are 2-fold. The complete screening of the MEFV gene in patients presenting with a clinical diagnosis of FMF (according to the established clinical criteria ) is expected to identify only a few patients with 2 mutated alleles that would not have been identified by means of a diagnostic procedure designed to search for a limited number of known mutations. In other words, the probability that a patient indeed carries 2 MEFV mutations despite the negative results of a limited search for mutations is particularly low, ranging from 1% to 15%, depending on the patient's ethnic background (Table 3). Thus, the present study discloses the existence of an FMF-like disorder in all classically affected populations that is not due to mutations in the MEFV gene. The identification of the pathophysiologic mechanisms underlying this newly recognized disorder will be essential in clarifying the nosology of a heterogeneous group of diseases that, to date, are phenotypically indiscernible.
We are indebted to the patients and their families for their willingness to participate in this study.
- 19Familial Mediterranean fever: the segregation of 4 different mutations in 13 individuals from one inbred family: genotype-phenotype correlation and intrafamilial variability. Am J Med Genet 2002; 109: 198–201., , , , .
- 20A population-based survey reveals an extremely high FMF carrier frequency in Armenia, suggesting heterozygote advantage [abstract]. Am J Hum Genet 1999; 65 Suppl: A400., , , , , , et al.