The hereditary periodic fever syndromes are a group of Mendelian disorders characterized by episodic fever and serosal or synovial inflammation. Familial Mediterranean fever (FMF) and the hyperimmunoglobulinemia D and periodic fever syndrome are both recessively inherited, while three dominantly inherited syndromes have been described, the best-characterized of which is familial Hibernian fever (FHF). The last year has seen two major developments in this field: the FMF gene was identified on chromosome 16p by positional cloning, and a second major periodic fever locus was mapped to distal chromosome 12p. The FMF gene (MEFV) encodes a novel 781 amino acid protein; to date, eight different missense mutations and a number of polymorphisms have been described. Seven of the eight mutations occur within a region of 82 amino acids near the C-terminus. Computational analysis of the conceptual protein reveals five different domains/motifs compatible with a nuclear effector function. MEFV is expressed preferentially in granulocytes and myeloid bone marrow precursors, giving rise to speculation that the protein may serve as a transcriptional regulator of inflammation in granulocytes. The second periodic fever locus was mapped by two different groups: one studying FHF, the other studying a similar dominantly inherited syndrome designated familial periodic fever. Both genes map to the same 19 cM region on distal chromosome 12p, strongly suggesting a common locus. The molecular characterization of the periodic fever genes should provide important new insights into the regulation of inflammation in general.
The hereditary periodic fever (HPF) syndromes are a group of Mendelian disorders characterized by self-limited episodes of fever accompanied by inflammation of the serosal or synovial membranes, without apparent infectious etiology (1). Although periodic diseases have been noted since ancient times, hereditary periodic inflammatory disorders have only been recognized for the last half-century, largely owing to refinements in laboratory methods to exclude microbial causes. These disorders hold a special fascination, not only because of their episodic nature, but because they may provide a window for understanding the regulation of inflammation in a much broader spectrum of clinical settings.
The HPF syndromes have been classified largely on clinical grounds (reviewed in refs 1,2). Two of these disorders, familial Mediterranean fever (FMF; MIM 249100) and the hyperimmunoglobulinemia D with periodic fever syndrome (HIDS, also known as Dutch-type periodic fever; MIM 260920), are inherited as autosomal recessive traits. FMF is the longest recognized of the HPF syndromes, having been first described by Siegal (3) as benign paroxysmal peritonitis in 1945. It is seen most often in populations of eastern Mediterranean ancestry (non-Ashkenazi Jews, Armenians, Turks and Arabs); carrier frequencies as high as 1:5 have been estimated in these groups (4–6), raising the possibility of a selective advantage for heterozygotes. Abdominal attacks with acute peritoneal inflammation are the most common manifestation, occuring in >90% of patients, with 30–85% of patients experiencing pleural attacks and 35–75% of patients experiencing arthritic attacks, depending on the series (7–9). A characteristic skin rash usually confined to the foot, ankle or lower leg may also occur (10). The acute episodes are characterized by the massive influx of neutrophils into the affected anatomical compartment. Over time, some patients also develop amyloidosis, due to the deposition of a cleavage product of one of the acute phase reactants in a number of tissues, most notably the kidneys. Daily oral colchicine has been demonstrated to prevent both the acute attacks and the development of amyloidosis (11–15). The FMF susceptibility locus maps to chromosome 16p13.3 in all affected populations (16–19).
HIDS was first described in several Dutch patients by van der Meer et al. in 1984 (20). As in FMF, attacks may manifest with abdominal pain or arthritis, but HIDS attacks tend to be somewhat longer (3–7 days in HIDS versus 1–3 days in FMF) (21,22). Moreover, pleural involvement is rare in HIDS, and the distribution of skin and joint involvement tends to be different. Amyloidosis has not been reported in HIDS. Elevated IgD levels both during and between attacks is the hallmark of this disorder, although small percentages of patients with other HPF syndromes may have marginal IgD elevations (22). Family studies suggest Mendelian recessive inheritance, but linkage has been excluded for the FMF region of chromosome 16 and for the immunoglobulin heavy chain region of chromosome 14 (23). Insofar as HIDS attacks do not correlate with IgD levels, and HIDS patients have a number of immunological abnormalities (24,25), it is likely that the HIDS gene acts by a mechanism other than just the regulation of IgD levels.
The nosological classification of the dominantly inherited HPF syndromes remains provisional, in part because most of the literature on this subject derives from clinical descriptions of isolated families. Familial Hibernian fever (FHF; MIM 142680) was first described in 1982 in a large family of Irish-Scottish ancestry (26); an updated pedigree of the original family (27) and two additional families (28) have since been published. FHF attacks tend to last longer than the attacks of the recessive HPF syndromes, and can also be distinguished by a high frequency of conjunctivitis, periorbital edema, myalgia, scrotal pain and inguinal hernia. Amyloidosis has only been documented in one of the 24 FHF cases reported, and patients tend to respond to corticosteroids rather than colchicine. Until several months ago, the only linkage data on FHF excluded the FMF region of chromosome 16 (29).
Recently, a large Australian family of Scottish descent with a dominantly inherited HPF was described (30). In addition to the ancestry and mode of inheritance, this syndrome resembles FHF with respect to duration of attacks, myalgia, scrotal pain, absence of amyloidosis and therapeutic response to corticosteroids but not colchicine. While acknowledging the similarities to FHF, the authors provisionally designated this disorder familial periodic fever (FPF). There are also reports in the literature of Finnish (31) and Austrian (32) families with dominantly inherited HPF syndromes without amyloidosis.
Since there appear to be genetic and environmental factors influencing the occurrence of amyloidosis in FMF, it is not clear that dominantly inherited HPF with amyloidosis (MIM 134610) is actually a different disease from those cases without amyloidosis. Others have described families of Swedish (33) and German-American (34) ancestry, while we recently have reported two families, one of Puerto Rican and one of Northern European ancestry (35). Like the dominant HPF syndromes without amyloidosis, these families tend to have longer attacks, and their attacks appear relatively resistant to colchicine.
Here we review recent developments in the molecular characterization of the HPF syndromes. One major milestone reached during the last year was the identification of the gene causing FMF (36,37). Our discussion will include an updated tabulation of mutations, an analysis of the likely function of the protein and data on lineage-specific gene expression. We will then turn our attention to recent linkage data indicating a second major HPF locus on distal chromosome 12p (30,38).
Identification of the Gene Causing Familial Mediterranean Fever
The FMF gene, designated MEFV (for Mediterranean fever), was identified independently by two positional cloning consortia in the summer of 1997 (36,37). The gene encodes an ∼3.7 kb transcript expressed predominantly in granulocytes; the predicted product is a novel 781 amino acid protein with homology to several transcription factors. The positional cloning proved particularly challenging for several reasons: (i) there were no translocations or large-scale deletions to guide the mapping; (ii) the FMF candidate interval, which lies between PKD1 and RSTS (CBP) on chromosome 16p, proved to be unstable in large-insert yeast artificial chromosomes (YACs) (39); (iii) MEFV is a novel gene that was not present in any of the expressed sequence tag (EST) databases; (iv) the expression of the gene is relatively tissue specific; and (v) all four of the mutations in the original reports are conservative missense mutations.
There was a strict association of each of these four mutations with disease-associated microsatellite haplotypes (36,37). One mutation, in which valine is substituted for methionine at position 694 (M694V), was seen in all chromosomes bearing the haplotype designated ‘A’ by our Consortium and ‘Med’ by the French FMF Consortium. This haplotype is seen in a very high percentage of North African Jewish FMF carrier chromosomes, as well as smaller percentages of Iraqi Jewish, Arab, Armenian and Turkish FMF chromosomes. This mutation was also observed in carrier chromosomes bearing three other apparently distinct microsatellite haplotypes, which converge to the same single-nucleotide polymorphism (SNP) haplotype as the A haplotype within MEFV (36). The substitution was not seen in any of an aggregate of almost 600 normal control chromosomes examined by the two Consortia.
A second mutation (36,37), with alanine substituted for valine at codon 726 (V726A), was observed in carrier chromosomes of several ethnic backgrounds, all bearing one specific microsatellite/ SNP haplotype. Two less common mutations, M680I (36,37) and M694I (37), were found on specific Armenian and Arab haplotypes, respectively. As was the case for M694V, none of these substitutions was found on a combined panel of almost 600 normal chromosomes. The observation of M694V and V726A mutations and their associated haplotypes in populations that may have been separated for many centuries (i.e. North African and Iraqi Jews, Ashkenazi Jews and Arab Druze) suggests that these mutations may be quite old. Moreover, the fact that more than one MEFV mutation occurs at high frequency in several eastern Mediterranean populations supports the speculation that these mutations may confer a selective advantage on heterozygotes.
Table 1 lists the eight missense mutations that we have observed to date (I. Aksentijevich et al., manuscript in preparation; 40). It is noteworthy that seven of the eight mutations are clustered within 82 amino acids of one another near the C-terminus of the protein. In a series of 100 periodic fever referrals recently studied at the National Institutes of Health (NIH), the most common MEFV mutations were V726A, E148Q and M694V in that order. We also observed a complex allele bearing E148Q and V726A on the same carrier chromosome. Somewhat unexpectedly, we found a substantial number of both Ashkenazi Jewish and Italian-American patients with typical clinical histories and MEFV mutations. Table 1 also indicates that there is a large number of polymorphisms in MEFV, which makes mutational screening methods like single strand conformation polymorphism (SSCP) less useful for this gene.
The percentage of carrier chromosomes accountable by the current list of mutations varies among studies. In the original reports identifying the gene, both consortia found that ∼85% of carrier chromosomes could be attributed to one of the four initially described mutations. However, it should be noted that these study populations were skewed towards subjects of North African Jewish background with relatively severe disease. Another early report on Turkish FMF patients (41) found a similarly high percentage of carrier chromosomes subsumed by the M694V, V726A and M680I mutations. However, our more recent experience with a more ethnically diverse and clinically heterogeneous tertiary American cohort indicates that there is still a substantial fraction (perhaps 40%) of carrier chromosomes in this population with unidentified mutations (40). Thus, as is the case for many other genetic diseases, clinical judgement is still very important, particularly in those cases without any of the currently known mutations.
The role of specific mutations in determining susceptibility to systemic amyloidosis in FMF is still unclear. M694V is very common in the North African Jewish population, where amyloidosis is also frequent (42), while V726A is more common in populations with a lower incidence of amyloidosis. This has led to the hypothesis that M694V homozygotes might be more prone to amyloidosis, and that V726A might have a lower risk or even be protective. Preliminary data from non-Ashkenazi Jewish patients tend to support both an increased risk of amyloidosis, and a more severe phenotype in other respects, in M694V homozygotes (43,44). Nevertheless, it will be important to obtain confirmatory data from other populations. Our own experience with American patients is difficult to interpret, because both amyloidosis and M694V homozygosity were relatively uncommon in our American cohort. At this point, it is clear that V726A is not protective, since we and others have observed individuals with amyloidosis who have at least one copy of V726A (40,45).
The Pyrin/Marenostrin Protein
With seven of the eight mutations shown in Table 1 localized within the B30.2 domain, it would appear that this region must be critical to the function of pyrin. The B30.2 domain originally was defined as a region of sequence similarity between an orphan exon identified in the human MHC, designated B30.2, and the C-terminal domain of a diverse group of intracellular, membrane-bound and secreted proteins (52). Several B30.2 family members, such as butyrophilin (55) and stonustoxin (56), are homologous to pyrin only in this domain. In the case of butyrophilin, a major substituent of the milk fat globule that happens also to be encoded in the MHC, the B30.2 sequence, functioning as an intracellular cytoplasmic domain, is linked to a transmembrane segment followed by two extracytoplasmic immunoglobulin-like domains (53,54). Two other B30.2 proteins, the α- and β-subunits of stonustoxin, are secreted as a lethal substituent of the stonefish venom (53).
To connote its relationship to fever, our Consortium named the protein product of MEFV ‘pyrin’, while the French Consortium, alluding to the Latin name for the Mediterranean Sea (Mare Nostrum), has called it ‘marenostrin’. At the time of writing there is no convention on nomenclature. Much of our present understanding of this protein is based on computational analyses of the predicted sequence for homologies and functional motifs (36,46). Figure 1 summarizes the results of this type of analysis, identifying five different motifs/conserved domains: (i) a bZIP basic domain (amino acid 266–270) (47); (ii) a B-box zinc finger (amino acids 375–407) (48); (iii) an a-helical or potential coiled-coil domain (amino acids 408–594) (49); (iv) two nuclear localization signals (amino acids 157–163 and 420–437) (50,51); and (v) a B30.2 (rfp) domain (amino acids 598–774) (52–54).
However, perhaps more germane to the possible function of pyrin, several B30.2 proteins also contain a B-box zinc finger and an α-helical (coiled-coil) domain (54). These include the amphibian nuclear proteins XNF7 (57) and PwA33 (58), which direct pattern formation in the developing embryo, and MID1, a recently identified putative transcription factor that is mutated in Opitz G/BBB syndrome (59). Two other members of the group, EFP [estrogen-responsive finger protein, encoded by ZNF147 (60)] and AFP [acid finger protein, encoded by ZNF173 (61)], are thought to be DNA-binding proteins, based primarily on their predicted structures. Also included in this group are several proteins of immunological interest: Staf50 is a human interferon-induced protein that attenuates transcription directed by the human immunodeficiency virus long terminal repeat (HIV LTR) (62); Ro52/SS-A is a ribonucleoprotein targeted by autoantibodies present in the sera of patients with Sjögren's syndrome and systemic lupus erythematosus (63–65); and rpt-1 is a murine protein that suppresses gene expression directed by the interleukin-2 receptor α-chain and HIV LTR promoters (66). Finally, three other pyrin homologs containing the B-box, coiled-coil and B30.2 domain are fusion partners in chimeric oncoproteins: rfp (ret finger protein) (67), PML (promyelocytic leukemia) (68–71) and TIF (transcriptional intermediate factor) (72). Unlike pyrin, nearly all of the aforementioned proteins also have a second cysteine-rich metal-binding domain, the RING finger (73,74), located N-terminal to the B-box.
Three activities, homomultimerization, nuclear translocation and subnuclear targeting, are common to several of these proteins. The functional roles of individual domains have been probed by site-directed mutagenesis. In various experimental systems, B-box integrity appears important both in homomultimerization (75) and subnuclear targeting (58,76,77). There appears to be general agreement that the coiled-coil domain mediates protein dimerization (49,75). However, to date, there are no site-directed mutagenesis data for the B30.2 domain. In the case of butyrophilin, cross-linking and immunoprecipitation experiments suggest that the B30.2 domain interacts with xanthine oxidase (78,79), but the relevance of this observation for pyrin is unclear, given that butyrophylin lacks both a B-box and coiled-coil domain.
Based on the preponderance of the evidence, it is thus likely that pyrin is a nuclear factor that undergoes homomultimerization. The presence of nuclear localization signals (NLS) (50,51) further buttresses this view. These sequences include a basic residue cluster, PLSKREE, beginning at amino acid residue 157, and a bipartite NLS motif, beginning at residue 420. The bipartite NLS consists of two basic domains separated by a 10 amino acid spacer region.
A bZIP basic domain is also present in pyrin. The complete bZIP domain consists of a 14–20 amino acid basic region followed six residues downstream by a leucine zipper (47). In the proteins where it has been analyzed, the leucine zipper mediates protein dimerization, while the basic region functions as a sequence-specific DNA-binding domain. The presence of the basic domain (though not the leucine zipper) further suggests that pyrin functions as a nuclear effector molecule.
Lineage-Specific Expression of MEFV
The tissue distribution of expression of a gene can be an important clue regarding its function. Our initial survey, using commercial multiple-tissue northern blots, demonstrated expression in peripheral blood leukocytes (PBLs), but not in a panel of RNAs taken from several other tissues, including heart, brain, placenta, lung, liver, muscle, kidney, pancreas, prostate, testis, ovary, small intestine and colon (36). As shown in Table 2, we also found no expression in spleen, thymus and lymph node, tissues comprised predominantly of lymphocytes. This suggested that message expression in peripheral blood might be restricted to a subpopulation of leukocytes. To test this hypothesis, we have studied various subpopulations of PBLs, first on northern blots and later by RT-PCR. By both methods, we have found MEFV expression in granulocytes but not in lymphocytes or macrophages/monocytes.
We have also explored the question of when myeloid cells begin to express MEFV. Although expression is difficult to demonstrate with mRNA derived from unfractionated bone marrow, message is detectable in bone marrow leukocytes. To begin to examine expression in hematopoietic precursor cells, we have studied the HL-60 promyelocytic leukemia cell line (Table 2; M. Centola et al., manuscript in preparation). When cultured without any stimuli, these cells do not express MEFV. However, when they are cultured with retinoic acid, which drives the cells to differentiate towards mature granulocytes, MEFV expression is induced. Retinoic acid-treated HL-60 cells produce primary but not secondary granules, and thus these results suggest that MEFV expression comes on before the milepost of secondary granule expression. It is also noteworthy that phorbol 12-myristate 13-acetate (PMA), which induces HL-60 to differentiate along the monocytic lineage, does not induce MEFV expression. Moreover, two other leukemic cell lines, the K562 erythro-leukemia and the MOLT4 lymphoblastic leukemia, also do not express MEFV. In contrast, we did find MEFV expression in the SW480 colon adenocarcinoma cell line, which produces the granulocyte growth factor GM-CSF (80).
Since MEFV appears to be expressed preferentially in granulo-cytes, it is tempting to speculate that pyrin acts as a transcription factor regulating inflammatory responses of myeloid cells. Given the recessive inheritance of FMF and the dramatic inflammation seen during FMF attacks, pyrin might act either as a transcrip-tional activator of an anti-inflammatory mediator, or a transcrip-tional repressor of a pro-inflammatory mediator. The relatively conservative missense mutations observed to date might permit adequate function of pyrin under baseline conditions, but would lead to decompensation with subclinical tissue injury or other forms of stress. Possibly, more profound loss-of-function mutations could lead to a much more severe, possibly even lethal, inflammatory phenotype. Still unexplained is the question of why FMF attacks seem to have a strong preference for serosal and synovial membranes. We have not observed MEFV expression in synovial tissue; the French Consortium has reported finding expression by RT-PCR in a synovial specimen from a rheumatoid arthritis patient (37), but it is unclear if this was due to expression in synoviocytes or infiltrating leukocytes. Expression was not detected in one peritoneal fibroblast cell line (81). Possibly, the tissue specificity of attacks may be due to a pyrin-mediated regulation of granulocyte adhesion molecules for serosal or synovial vascular beds.
Chromosome 12P13 Harbours a Second Periodic Fever Locus
Mapping of the FMF locus to chromosome 16p13.3 permitted testing of families with other HPF syndromes for linkage to this region, on the chance that some of these clinical variants might be allelic variants. For FHF and FPF, the two dominantly inherited periodic fever syndromes in which family sizes were sufficient to permit such analysis, linkage was excluded for chromosome 16 markers (29,30). Genome-wide searches subsequently were undertaken in the UK and Australia to map FHF and FPF, respectively.
The Australian group was the first to report linkage (30), after having tested 236 of a planned 330-marker search. In a single large family, a pairwise LOD score of 3.98 was observed at D12S77, with a maximal multipoint LOD score of 6.14 at D12S356 (Fig. 2). Recombinants placed FPF between D12S314 and D12S364, a rather substantial ∼19 cM interval. Meanwhile, the UK group had excluded approximately one-third of the genome for FHF, having looked at 78 markers on nine chromosomes, before also finding evidence for linkage on distal chromosome 12p (38). In a panel comprised of the only three families with FHF currently in the literature, a maximal pairwise LOD score of 3.11 was observed at D12S77 (θ = 0.12). Multipoint linkage analysis placed FHF between D12S93 and D12S77, with a peak multipoint LOD score of 3.79 (Fig. 2). Thus, the susceptibility genes for both of these dominantly inherited HPF syndromes map to the same chromosomal region.
Assuming that these two clinical entities are actually caused by mutations in the same gene, the number of families currently available would not permit an efficient positional cloning effort. However, it is possible that at least some of the other dominantly inherited HPF families, including those that are of non-Irish/Scottish ancestry and/or those that present with amyloidosis, may also show linkage to this region and allow further interval narrowing. Pursuing a unifying hypothesis for the HPF syndromes, perhaps the FHF/FPF gene may be a homolog ofMEFV, or it may encode a protein that binds pyrin or participates in the same biochemical pathway. Alternatively, there are a number of other genes with immune function in the candidate interval, such as TNFRRP, CD4 and CD69 (38).
Studies of the HPF syndromes promise to shed important new light on the molecular basis of inflammation. In the case of FMF, the recent identification of a novel gene invites a wide range of new studies. A more complete mutational inventory will contribute not only to clinical diagnosis, but may direct studies of the protein to certain functional domains. With the development of anti-pyrin antibodies, it will be possible to test the hypothesis that the pyrin protein is a nuclear factor, and to monitor the movement of this protein within the cell in a variety of physiological settings. Experiments are also warranted to determine the identity of molecules that interact with pyrin (whether they be other proteins or nucleic acid sequences), and to define the other elements in the pyrin pathway of inflammation. Our laboratory recently has cloned the mouse homolog of pyrin, and we are developing lines of knockout and knockin mice to probe pyrin physiology in experimental animals.
Similarly, a positional candidate approach to the FHF/FPF gene will provide important insights, whether in the pyrin pathway or some other branch of the inflammatory cascade. Linkage studies in unclassified HPF families should either add to the chances of finding the FHF/FPF gene (if these families are linked to chromosome 12p13), or help to establish them as independent clinical entities. In the end, it is our hope that the HPF syndromes will help to define the molecular basis of inflammation, much as the childhood immunodeficiency diseases helped to define the cellular basis of immunity 30 years ago.
The authors thank the many patients who, through their participation in NIH clinical studies, have taught us much. We also thank all of the other members of the International FMF Consortium for their ideas, their effort and their camaraderie in the positional cloning project. Finally, we thank Dr Henry Metzger, Scientific Director of NIAMS, for his unfailing support during the course of this work.
- transcription, genetic
- positional cloning
- amino acids
- familial mediterranean fever
- maximal expiratory flow-volume curves
- missense mutation
- bone marrow
- inflammatory disorder
- hyperimmunoglobulin d with periodic fever
- mendelian disorders
- hereditary autoinflammatory diseases