Reduced MEFV messenger RNA expression in patients with familial Mediterranean fever




Familial Mediterranean fever (FMF) is the most common inherited periodic syndrome. The disease phenotype and the almost exclusive expression of the causative gene, MEFV, in leukocytes suggest that this gene plays an important role in the inflammatory cascade. Since most of the known mutations are conservative, we sought to determine how minor DNA defects can give rise to the dramatic phenotypic features seen in FMF.


To address whether the molecular basis of the phenotype–genotype correlation could be related to altered MEFV messenger RNA (mRNA) expression, we quantified the relative abundance of MEFV transcripts in peripheral blood leukocytes from patients with FMF, healthy carriers of a single MEFV mutation, and healthy control subjects.


We found significantly lower expression of MEFV mRNA in genetically ascertained FMF patients than in healthy controls (0.7 versus 1.1; P = 0.00001). In healthy carriers, the mRNA levels were intermediate, suggesting a true dose-response relationship between the number of mutations and the abundance of MEFV transcripts. The difference between healthy controls and healthy carriers was significant (1.1 versus 0.8; P = 0.008), demonstrating that the decrease in mRNA expression is related to a molecular defect independent of FMF symptoms. MEFV mRNA expression was also found to be a function of the type of mutations. The lowest MEFV levels were found in healthy carriers and patients with M694V. Moreover, we observed an inverse correlation with the clinical severity score (r = −0.6, P = 0.04 and r = −0.6, P = 0.004 in patients with 1 and 2 M694V mutations, respectively).


Our results demonstrate that MEFV message levels are related to both the genotype and the phenotype, and suggest that the pathophysiology of FMF relies on a quantitative defect of MEFV mRNA expression.

Familial Mediterranean fever (FMF) is a recessively inherited disease that is mainly prevalent in Arabs, Armenians, Jews, and Turks. It is the first hereditary periodic fever syndrome whose gene (MEFV) has been identified (1, 2). FMF is an autoinflammatory disease characterized by acute self-resolving attacks of fever and serositis, during which levels of acute-phase reactants are especially elevated. Between attacks, the patients are completely asymptomatic, although many of them have significant residual inflammatory activity (3).

Because the FMF gene is predominantly expressed in granulocytes (2, 4, 5), it is commonly accepted that these cells are the major participants in the disease. Challenging this hypothesis, recent studies have demonstrated that several inflammatory mediators, including cytokines, modulate MEFV messenger RNA (mRNA) levels in monocytes but do not change message levels in neutrophils. The MEFV-encoded protein called marenostrin/pyrin has been localized in the cytoplasmic perinuclear area by green fluorescent protein fusion assays (5–8); however, nothing is yet known of its function. A recent study has provided indirect evidence suggesting that the protein may regulate inflammatory responses at the level of leukocyte cytoskeletal organization (8).

The symptoms of FMF range from isolated fever to renal failure due to AA amyloid deposits in the kidneys. The genetic basis accounting for this large spectrum of clinical features was first expected to be found through the identification of the spectrum of MEFV mutations. Since the cloning of the gene, however, no gross rearrangement in the MEFV gene has been detected, and the vast majority of disease-causing sequence alterations result in relatively conservative amino acid changes (9–12). Yet, the disease severity does correlate with the patient's genotype: homozygosity for M694V, the most frequent mutation, is associated with a severe form of FMF (13–17), whereas homozygosity for E148Q is associated with mild disease expression and low penetrance (18, 19). It is still unexplained how relatively conservative amino acid substitutions can dramatically affect the phenotype. One hypothesis that possibly accounts for these observations is that such point mutations result in an alteration in the abundance of MEFV transcripts.

To address the molecular mechanisms that underlie the relationship between the phenotype and genotype in FMF, we extensively defined the pattern of MEFV mRNA expression according to the clinical and genetic status of the patients. Our findings are presented herein.


FMF patients, healthy carriers of a single MEFV mutation, and healthy controls.

The patient group comprised 96 consecutive patients who met the clinical criteria for FMF (20) and who had been referred to us for genetic diagnosis between October 1998 and August 2000. All blood samples were obtained when the FMF was in remission, and the patients had not had an attack of FMF for weeks and sometimes months (the question was systematically asked at the time of blood drawing). An aliquot of blood was saved for prospective RNA expression analysis.

The diagnostic strategy, based on screening the 2 mutational hot spots of the MEFV gene (exons 2 and 10), has been described elsewhere (9, 21) and covers all mutations more frequent than 1% (22). We identified mutations on both inherited copies of MEFV in 52 patients (genetically ascertained FMF; mean age 24), on 1 inherited copy of MEFV in 22 patients (mean age 25), and no mutation in 22 patients (mean age 40). The ethnic distribution of the 96 patients was as follows: 42% were non-Ashkenazi Jews, 18% were Arabs from North Africa, 11% were Turks, 6% were Armenians, and the others were of miscellaneous origins, mostly Europeans of other ancestries.

Clinical information was computerized from the answers to a uniform questionnaire that included all items that make up the FMF severity score, as defined previously (23). These items and their scores are as follows: age at onset, 0–3 points; frequency of attacks, 1–3 points; articular signs, 0–3 points; erysipelas-like erythema, 0–2 points, and amyloidosis, 0–4 points. The colchicine parameter was not included since only 59% of the patients were receiving colchicine at the time of blood sampling. We set up an automatic calculation of the scores to yield a clinical severity score, which ranged from 2 to 10 in our patients.

The healthy carrier group consisted of 37 subjects who were identified through family studies (e.g., asymptomatic parents of patients with 2 mutations) and were confirmed by DNA testing as carrying a single MEFV mutation. Their mean age was 42 years.

The control group comprised 31 healthy volunteers (mean age 40), most of whom were staff members, whose blood was also drawn between 1998 and 2000. None of the healthy controls were related to the patients. All were screened for the MEFV gene mutation, and no mutation was found.

Cell separation.

Blood was drawn into tubes containing heparin. For the main study, whole peripheral blood leukocytes (PBLs) were purified 24–48 hours later by lysis of the red cells with blood lysis buffer containing 155 mM NH4Cl, 10 mM KHCO3, and 1 mM EDTA (30 minutes at 4°C). After a centrifugation step (1,200g for 10 minutes at 4°C), cells were washed twice with blood lysis buffer, counted, and stored at −80°C until used.

Studies of separated leukocytes were also conducted on a few subjects (12 healthy controls and 12 M694V-homozygous FMF patients). For those studies, PBLs were purified by standard Ficoll-Hypaque density-gradient centrifugation (Sigma Diagnostics, St. Quentin Fallavier, France) 4 hours after the blood was drawn. Monocytes were purified from the mononuclear cell fraction by adherence to plastic. Polymorphonuclear cells (PMNs) were affinity-purified by pan–human class II HLA–coated magnetic beads (Dynabeads; Dynal, Compiègne, France) after hypotonic lysis of erythrocytes to withdraw the residual erythrocytes and enhance the yield and degree of purification of the PMNs. The purity of each preparation was verified by investigating 2 cell-specific genes: neutrophil antigen (NA) alloantigens (24), which are expressed in PMNs but not in monocytes, and the interleukin-10 (IL-10) gene (25), which is only expressed in monocytes.

RNA extraction.

Total RNAs were prepared from PBLs by TRIzol extraction according to the manufacturer's protocol (Gibco BRL, Cergy-Pontoise, France). Briefly, PBLs were lysed in 1 ml TRIzol reagent, and RNA was extracted with 0.2 ml of chloroform. After centrifugation at 12,000g for 15 minutes at 4°C, total RNAs were precipitated from the aqueous upper phase with isopropanol (30 minutes at 4°C), washed with 75% ethanol, and vacuum-dried. The pellet was resuspended in 30 μl of water containing RNase inhibitor (RNasin; Promega, Charbonnières, France), and RNA concentrations were determined by measuring absorbance at 260 nm.

Reverse transcription (RT).

One microgram of total RNA was denatured for 10 minutes at 70°C, then reverse-transcribed in a total volume of 25 μl containing 50 mM Tris HCl (pH 8.3), 75 mM KCl, 3 mM MgCl2, 10 mM dithiothreitol, 0.5 mM dNTPs, 15 μg/ml of random primers (Promega), 40 units of RNasin (Promega), and 200 units of Moloney murine leukemia virus reverse transcriptase (Gibco BRL). Samples were incubated for 10 minutes at room temperature, 45 minutes at 42°C, and 3 minutes at 95°C, and then finally cooled to 4°C.

Semiquantitative RT–polymerase chain reaction (RT-PCR).

General conditions.

Complementary DNA (cDNA) was amplified with AmpliTaq Gold (PE Biosystems, Foster City, CA) under standard PCR conditions. The oligonucleotide PCR primers we used were as follows: for MEFV (576 bp), 5′-TCCATTTCTGAACGCAGG-3′ (exon 8) and 5′-ACGATGAGCCCATCTGCC-3′ (exon 3); for HPRT (213 bp), 5′-TGTAATGACCAGTCAACAGGG-3′ and 5′-TGGCTTATATCCAACACTTCG-3′; for NA antigen (831 bp), 5′-TCTTTGGTGACTTGTCCA-3′ and 5′-GCCACTGCTCTTATTACT-3′; and for IL-10 (328 bp), 5′-AAGCTGAGAACCAAGACCCAGACATCAAGGCG-3′ and 5′-AGCTATCCCAGAGCCCCAGATCCGATTTTGG-3′.

The amplification reactions were carried out in a DNA thermocycler (MJ Research, Watertown, MA), and after an initial denaturation step for 9 minutes at 95°C, the samples were cycled for 1 minute at 94°C, 1 minute at 60°C, and 2 minutes at 72°C, then elongated for 4 minutes at 72°C. MEFV and HPRT were amplified for 28 and 30 cycles, IL-10 was amplified for 30 cycles, and NA was amplified for 35 cycles.

The amplified products were separated by electrophoresis on an 8% acrylamide gel and visualized after ethidium bromide staining. The images were directly captured with a phosphorimager system (Perfect Image; Clara Vision, Orsay, France), and the intensity of the DNA fragments was densitometrically scanned with the corresponding software (Gel Analyst). A typical experiment is shown Figure 1.

Figure 1.

Expression of MEFV mRNA, as measured by reverse transcription–polymerase chain reaction analysis. A typical experiment is shown. The amplified products (MEFV and HPRT) were separated by electrophoresis on an 8% acrylamide gel and visualized after ethidium bromide staining, as described in Patients and Methods. Lanes 3 and 8, healthy controls; lanes 2 and 7, familial Mediterranean fever (FMF) patients with no identified mutations; lanes 1 and 9, healthy carriers of a single MEFV mutation; lanes 5 and 10, FMF patients with 1 mutation; lanes 4 and 6, FMF patients with 2 mutations. MW = molecular weight markers (Marker 9; Eurogentec, Luik, Belgium).

Control steps for validation of the quantification strategy.

The primer sequences were chosen to span 2 exons of the studied genes to eliminate the risk of amplifying any contaminating genomic DNA. The HPRT housekeeping gene, which is moderately expressed in mammalian cells (26), was coamplified with MEFV as an internal control. The relative expression of MEFV (MEFV:HPRT ratio) was systematically calculated to normalize possible variations in the yield of the RT-PCR step. The optimum number of cycles was determined by a time-course assay ranging from 24 to 32 cycles using RNA samples with different levels of expression of the studied genes. This test confirmed that for all samples, the MEFV:HPRT ratios, which were evaluated at different PCR rounds, were constant and allowed coamplification of MEFV and the HPRT internal control in the initial linear phase between 28 and 30 amplification cycles (results not shown). All subsequent analyses were thus performed in this range of PCR coamplification efficiencies.

Expression levels were determined based on 2 independent sets of RT-PCR. One reference subject was included in each PCR round and served as a control for between-assay reproducibility (range of MEFV:HPRT ratios 0.98–1.02). The specificity of the amplified products was verified by hybridization with 5′-biotinylated probes: 5′-GTACACCTGCTCCAGCTT-3′ (exon 5) for MEFV and 5′-CCTTGGTCAGGCAGTATA-3′ for HPRT (data not shown).

Patients and healthy subjects were chronologically assigned a code number. The samples were handled blindly in this order by a single investigator (CN), who did not know their corresponding genetic and clinical status.

Real-time quantitative RT-PCR.

A real-time RT-PCR method using LightCycler technology (Roche Diagnostics, Meylan, France) suitable for quantitative determination of MEFV mRNA expression was developed on a limited number of samples (n = 50). MEFV and HPRT amplifications were achieved independently with the following primers: for MEFV (215 bp), 5′-GGCTAAGACAGTGCCT-3′ (exon 7) and 5′-GGTAAGCGGTTTCTGC-3′ (exon 10); and for HPRT (256 bp), 5′-CTGACCTGCTGGATTACA-3′ and 5′-GCGACCTTGACCATCTTT-3′.

Reactions were performed in a 20-μl volume with 2 μl of diluted cDNA sample, 0.5 μM primers, and 4 mM MgCl2. Nucleotides, Taq polymerase, and buffer were included in the LightCycler–DNA Master SYBR Green I mixture (Roche Diagnostics). A typical protocol took ∼45 minutes to complete and included a 10-minute denaturation step followed by 40 cycles with denaturation for 10 seconds at 95°C, annealing for 5 seconds at 60°C (for MEFV) or at 55°C (for HPRT), and extension for 10 seconds at 72°C. Detection of the fluorescent product was carried out at the end of the 72°C extension period.

To confirm amplification specificity, the PCR products from each primer pair were subjected to a melting curve analysis. Standard curves for the value of MEFV and HPRT mRNA were generated using a control cDNA sample diluted at 2-fold intervals. All standards and samples were assayed in triplicate. The mRNA values for all unknown samples were determined by LightCycler analysis software (version 3.1), according to the standard curves. Data are represented as the mean level of expression of MEFV relative to HPRT expression (MEFV:HPRT ratio).

Statistical analysis.

For mean comparison experiments, we used the Mann-Whitney and the Kruskal-Wallis tests, and for correlation studies, we used the Spearman correlation coefficient. Differences between groups were considered significant when the P value was less than 0.05. In cases of multiple comparisons, a Bonferroni correction was applied. SAS version 6.12/UNIX software (SAS Institute, Cary, NC) was used in all cases.


Decreased MEFV mRNA expression in patients with FMF-related symptoms.

As an initial step in searching the molecular key to FMF, we compared the abundance of MEFV mRNA between healthy controls and patients with FMF-related symptoms (all referrals have definite FMF according to clinical criteria [20]). The mean MEFV:HPRT ratio (relative expression) was clearly decreased in the patients compared with the healthy controls (0.8 versus 1.1; P = 0.0001) (Figure 2). This decrease was even more significant (0.7 versus 1.1; P = 0.00001) when the comparison was restricted to patients ascertained through genetic diagnosis (homozygotes or compound heterozygotes) (Figure 3).

Figure 2.

Relative levels of MEFV mRNA expression in healthy controls and patients with familial Mediterranean fever (FMF)–related symptoms. MEFV mRNA expression was assessed in 31 healthy control subjects (open bar) and in the entire study group of 96 patients with FMF-related symptoms (solid bar). MEFV mRNA expression was significantly lower in the FMF patients. Values are the mean and SD expression relative to HPRT.

Figure 3.

Comparative levels of MEFV mRNA expression according to the number of MEFV mutations identified. Five study groups were evaluated, as follows (as shown from left to right in the figure): 31 healthy control subjects, 22 familial Mediterranean fever (FMF) patients with no MEFV mutations, 37 healthy carriers of a single MEFV mutation (ascertained through family studies), 22 FMF patients with 1 mutation, and 52 FMF patients with 2 mutations. MEFV levels varied significantly between the groups (P < 0.0001 by Kruskal-Wallis test), and the observed expression decreased with an increase in the number of mutations identified. After Bonferroni adjustment, this decrease was significant for comparisons between the healthy controls versus the genetically ascertained FMF patients (n = 52), the healthy controls versus the healthy carriers (n = 37), and the FMF patients with no mutations (n = 22) versus the FMF patients with 2 mutations (n = 52). The number of MEFV mutations identified by our diagnostic strategy (see Patients and Methods) and the presence and absence of FMF symptoms are shown across the bottom. Values are the mean and SD expression relative to HPRT.

Decreased MEFV mRNA expression with an increased number of mutations.

We next assessed the mRNA levels of the gene according to both the number of mutations identified by our routine screening strategy and the absence or presence of FMF-related symptoms (Figure 3). In asymptomatic individuals, the levels of MEFV message in the healthy carrier group were lower than those in the healthy control group (0.8 versus 1.1; P = 0.008). In individuals with FMF symptoms, the genetically ascertained patient group with 2 mutations displayed lower mean expression than did the group of patients with no identified mutations (0.7 versus 0.9; P = 0.005). The MEFV mRNA expression in patients with 1 mutation (0.8) was between that of patients with 2 mutations and patients with no mutations.

Decreased MEFV mRNA expression according to the type of mutation.

We also examined the MEFV mRNA levels according to the type of mutation (Figure 4). M694V was evaluated separately from other MEFV mutations because it is the most severe and the most frequent mutation in FMF. M694V was significantly associated with the lowest MEFV mRNA levels in all subjects. The difference in MEFV mRNA expression between M694V-homozygous patients and healthy controls was extremely significant (0.6 versus 1.1; P < 10−5) (data not shown).

Figure 4.

Comparative levels of MEFV mRNA expression according to the type of MEFV mutations. Among the healthy carriers of a single MEFV mutation (lightly shaded bars), those carrying M694V (n = 22) expressed lower mRNA levels than did those with other MEFV mutations (n = 15; 1 with E148Q, 4 with M680I, 5 with M694I, 1 with K695R, 2 with V726A, 1 with A744S, and 1 with R761H). Among familial Mediterranean fever (FMF) patients with 1 mutation (darkly shaded bars), those carrying M694V (n = 12) expressed lower mRNA levels than did those with other MEFV mutations (n = 10; 5 with E148Q, 2 with Y688X, 1 with M694I, 1 with K695R, and 1 with A744S). Among FMF patients with 2 mutations (solid bars), there was a difference in mRNA levels (P = 0.02 by Kruskal-Wallis test) between those with no M694V mutations (n = 10; 2 with M694I/E148Q, 5 with M680I/M680I, 1 with M680I/V726A, and 2 with M694I/M694I), those who were M694V compound heterozygotes (n = 14; 4 with M694V/E148Q, 2 with M694V/M680I, 1 with M694V/M694I, 1 with M694V/K695R, 5 with M694V/V726A, and 1 with M694V/R761H), and those who were M694V homozygotes (n = 28), and the observed expression decreased with an increased number of M694V alleles. After Bonferroni correction, this decrease was significant between the patient group with no mutations and those with 2 M694V mutations. The number of MEFV mutations identified by our diagnostic strategy (see Patients and Methods) and the presence and absence of FMF symptoms are shown across the bottom. Values are the mean and SD expression relative to HPRT.

In healthy carriers, the MEFV:HPRT ratio was lower in those with the M694V mutation (0.7 versus 1.1; P = 0.004). In patients with FMF symptoms and only 1 identified mutation, the MEFV message levels were lower in those who carried the M694V mutation than in those with any other mutation (0.6 versus 1.1; P = 0.04) (Figure 4), especially if the other mutation was E148Q (0.6 versus 1.4; P = 0.004) (data not shown). Interestingly, the MEFV:HPRT ratio for this mutation, which is known to be mild in FMF, was even slightly higher than that observed in healthy controls (1.4 versus 1.1; P not significant) (data not shown). In the patient group with 2 mutations, the MEFV mRNA expression decreased with an increased number of M694V alleles. The MEFV:HPRT mRNA ratios were 0.9, 0.8, and 0.6 in patients with 0, 1, or 2 M694V alleles, respectively. The difference in expression between the patients with 0 and 2 M694V mutations was significant (P = 0.015).

Inverse correlation between MEFV mRNA expression and the FMF clinical severity score.

To further investigate the molecular mechanisms underlying the relationship between the phenotype and the genotype in FMF, we determined clinical severity scores in our patients. This scoring system includes the major clinical features of the FMF phenotype, and the score increases with increasing severity of the disease (23). The M694V homozygous genotype was associated with a more severe phenotype. The clinical severity score of M694V-homozygous patients was higher than that of patients with other genotypes (6.7 versus 5.1; P = 0.002).

When we plotted the MEFV mRNA levels against the clinical severity score, we found no significant correlation in the entire group of patients or in patients subgrouped according to the number of MEFV mutations identified (data not shown). However, when we focused the analysis on the MEFV genotype, we confirmed the implication of the severe M694V mutation. Indeed, in the patient group with only 1 M694V mutation, we found an inverse correlation between MEFV mRNA expression and the clinical severity score (r = −0.6, P = 0.04) (Figure 5A). This result was more significant (r = −0.6, P = 0.004) in patients who inherited a second M694V mutation (Figure 5B), whereas no correlation could be individualized in the mixed patient group that inherited a second different mutation (M694V compound heterozygotes) (Figure 5C).

Figure 5.

Correlation between the level of MEFV mRNA expression and the clinical severity score for familial Mediterranean fever (FMF). The MEFV mRNA levels in patients with at least 1 M694V mutation were plotted against their clinical severity scores (calculated as described in Patients and Methods). A, FMF patients with 1 M694V mutation (n = 12); B, FMF patients with 2 M694V mutations (n = 28); C, FMF patients who were M694V compound heterozygotes (n = 14). A significant correlation was found in patients who were apparently true M694V heterozygotes and in patients who were M694V homozygotes, but not in patients who were M694V compound heterozygotes.

Decreased MEFV mRNA expression in FMF patients according to cell type.

MEFV was reported to be expressed in neutrophils and, to a lesser extent, in monocytes. Our experiments were performed with total RNA prepared from PBLs. To verify that the observed differences in mRNA levels in whole blood were not due to differences in the survival of these cell lineages, we investigated the expression of MEFV in isolated monocytes and PMNs (Figure 6). We fractionated freshly collected leukocytes from an additional group of 12 M694V-homozygous FMF patients and 12 healthy control subjects. We found a statistically significant difference in MEFV expression in isolated monocytes (0.7 in FMF patients versus 1.1 in controls; P = 0.04), which is the almost exclusive population of remaining cells at the time we receive the samples after blood is drawn. Furthermore, we also separated PMNs from the same fresh blood samples, and documented a parallel decrease in MEFV mRNA expression between M694V-homozygous patients and healthy controls. This difference was not statistically significant.

Figure 6.

Levels of MEFV mRNA expression in purified monocytes and polymorphonuclear cells. Blood samples were obtained from an additional 12 patients with familial Mediterranean fever (FMF) patients and an additional 12 healthy control subjects. White blood cells were purified as described in Patients and Methods, and the MEFV mRNA levels relative to HPRT were compared in monocytes and polymorphonuclear cells from the 2 study groups. An FMF-associated decrease in MEFV mRNA expression in monocytes was significant. Values are the mean and SD.

Validation of mRNA quantification by real-time RT-PCR.

To further confirm the decrease in MEFV mRNA levels in FMF patients, a real-time RT-PCR was performed on a limited number of samples (n = 50) from subjects of the same 5 groups as presented in Figure 3. Using this more sensitive and precise approach, we obtained the following MEFV:HPRT ratios: 1 in healthy controls, 0.8 in patients with no mutations, 0.7 in healthy carriers of a single MEFV mutation, 0.6 in patients with 1 mutation, and 0.4 in patients with 2 mutations. The P values were similar to those for the analyses shown in Figure 3 (P = 0.00001 for healthy controls versus patients with 2 mutations, P = 0.015 for healthy controls versus healthy carriers, and P = 0.004 for patients with no mutations versus patients with 2 mutations).


The main finding of the present study is the demonstration of reduced MEFV mRNA expression in FMF patients and in healthy carriers. Shortly after the cloning of the gene responsible for FMF, we showed that patients who were homozygous for M694V displayed a more severe disease than patients with other genotypes (13). This observation was confirmed by other investigators (14–17). In contrast, patients homozygous for E148Q displayed a milder disease or were asymptomatic (18, 19). To understand how these missense mutations give rise to such variable clinical features, we undertook a prospective study and established an RNA bank from whole blood samples of consecutive patients whose FMF was in remission.

Patients with FMF-related symptoms, especially those with genetically ascertained FMF, expressed lower levels of MEFV mRNA than did healthy controls. Message levels in healthy carriers identified through family studies were intermediate, which supports a true dose-response effect. The difference between healthy controls and healthy carriers was significant (1.1 versus 0.8; P = 0.008), demonstrating that the decrease in MEFV mRNA expression is related to a molecular defect that is independent of symptoms. Our results are also consistent with those in a previous report from another group of investigators, who showed enhanced acute-phase reactants in healthy carriers (3). It is important to note that while FMF could be definitely proved by the detection of 2 mutations, the actual disease status remains uncertain for some of the patients with fewer than 2 identified MEFV mutations. These patients could have either true FMF with rare MEFV mutations (which occurs in <1% of patients) (22), a rare form of FMF with dominant inheritance (27), or a condition that resembles FMF.

MEFV mRNA expression was also related to the type of mutation. The most severe M694V mutation was associated with the lowest MEFV mRNA levels, especially when the patient was homozygous, while the mildest E148Q mutation was associated with the highest MEFV mRNA levels. We could not specifically assess patients with rare genotypes, and we cannot exclude the possibility that such patients might express even less or more MEFV. Nevertheless, since MEFV mRNA expression was dependent on both the number and the type of mutations, it is likely that the MEFV genotype is directly or indirectly involved in the regulation of its own expression.

Because there is a correlation between the FMF phenotype and the MEFV genotype and, as shown here, between the MEFV genotype and MEFV mRNA expression, we also determined whether there was a correlation between the FMF phenotype and MEFV mRNA expression. We thus plotted the clinical severity scores against the MEFV:HPRT ratios and demonstrated that the disease severity was related to low levels of MEFV mRNA in M694V homozygotes and in apparently true M694V heterozygotes. The fact that MEFV mRNA expression was both decreased and inversely correlated with the clinical severity score in patients with only 1 M694V mutation, as in M694V homozygotes, was unexpected and might account for the recently reported cases of dominance associated with mutations at this codon (27). However, the clinical severity score was not correlated with MEFV mRNA levels in M694V compound heterozygotes, which probably reflects a combined effect with the second mutation. We could provide a rational explanation for this, at least with regard to E148Q. We observed that patients with E148Q had higher MEFV mRNA levels than did the other patients. It has been suggested that some low-penetrant mutations, such as V726A and E148Q, may have a protective effect (for review, see ref. 22). The other less frequent mutation could not be singly correlated with the clinical severity score. It can be anticipated that a similar relationship would hold for all or most of the mutations.

We also studied other conditions that might influence MEFV mRNA levels, such as colchicine treatment (Table 1) and the ancestry of the patients, which could be confounding factors. We found no statistically significant difference between patients who were treated with colchicine and those who were not, a finding consistent with previous data obtained in vitro (4), or between the patients of different ethnic groups.

Table 1. MEFV messenger RNA expression levels in patients with familial Mediterranean fever, by colchicine treatment group*
 Colchicine treatment
  • *

    Values are the mean levels of expression of the MEFV gene relative to the HPRT housekeeping gene; n values are the number of patients. P values comparing colchicine treatment groups were not significant.

All patients0.74 (n = 61)0.74 (n = 35)
Patients with 2 mutations0.70 (n = 39)0.67 (n = 13)
Patients with 1 mutation0.98 (n = 10)0.66 (n = 12)
Patients with 0 mutation0.90 (n = 12)0.96 (n = 10)

FMF was classically considered a neutrophil-mediated disease because of the huge influx of neutrophils at the site of inflammation and because for some time, MEFV was thought to be exclusively expressed in these cells. A growing body of data (4, 5, 7) has established that monocytes also play a key role in this disease. The decrease in MEFV mRNA expression arose in both cell types, especially in purified monocytes, where the MEFV mRNA levels were significantly lower in FMF patients than in healthy controls (0.7 versus 1.1; P = 0.04). However, the decrease was not significant in PMNs, probably because of the limited number of samples that could be recruited shortly after blood drawing and because of the known variable expression of MEFV in these cells (4).

Among the possible mechanisms that could account for the observed differences between study groups is an alteration of MEFV mRNA stability in the presence of MEFV mutations. MEFV mRNA with mutations located close to the 3′ end, such as M694V, could be more sensitive to nucleases. Mutations could also alter the tridimensional conformation of the DNA and impair the access of transcription factors. Alternatively, mutations could alter the processing of the messages. Finally, wild-type marenostrin could lead, via a positive autoregulation pathway, to the continued expression of MEFV. If the activity of marenostrin were destroyed, as is likely in M694V mutants, the positive autoregulation pathway would no longer function correctly, and thus lead to a lower expression of MEFV mRNA. It is noteworthy that relatively moderate variations in the MEFV mRNA levels were associated with a dramatic effect on the disease phenotype. We are currently characterizing a series of rabbit polyclonal antibodies to investigate the expression of marenostrin and to determine whether changes in the abundance of MEFV transcripts are truly related to changes in the abundance of the encoded protein.

It has recently been demonstrated that MEFV mRNA expression is up-regulated by a number of inflammatory mediators, such as interferon-γ (4, 7). Such mediators activate transcription factors that interact with binding sites homologous to inflammatory promoter elements, which are present in the upstream region of the MEFV gene. In the present study, we provide additional insight into the understanding of the mechanism of MEFV mRNA expression, which is also apparently dependent on the correct coding sequence of the gene.

The same study conducted in FMF patients during disease attacks would be warranted. Centola et al (4) demonstrated that MEFV expression was stimulated in vitro by lipopolysaccharide in monocytes from healthy subjects. It can be speculated that this expression is accordingly enhanced during FMF attacks. Those authors suggested that the gene functions in a Th1-responsive negative-feedback loop during proinflammatory activation of myeloid cells and that the pathophysiologic features of FMF result from defects in this inhibitory activity. Our findings revealed that the MEFV mRNA levels were lower in FMF patients in remission than in healthy controls. During FMF attacks, the inflammatory stimuli would raise MEFV expression to a level that is insufficient to feed the negative-feedback loop and reduce the inflammatory process. Unfortunately, due to the organization of our department, which is primarily aimed at genetic diagnosis, all patients were referred to us when they were in clinical remission. For the same reason, the inflammatory activity could not be tested biologically through a determination of the patients' levels of acute-phase reactants upon blood sampling. However, these parameters are not definite markers of inflammatory remission, since they are known to sometimes remain elevated between FMF attacks (3, 28).

In conclusion, we have demonstrated that MEFV message levels are related to both the genotype and the phenotype of FMF. Our results support the idea that the MEFV transcriptional pathway is misregulated in FMF patients and suggest that the pathophysiology of the disease relies on a quantitative defect in the expression of MEFV mRNA.


We thank Dr. B. Boizet and Dr. M. Dupont for helpful discussions, and we thank Professor D. Cattan, P. Dellamonica, R. Manna, and all the clinicians for their assistance in clinical diagnosis. We are indebted to the FMF patients for their devoted contribution to this study.