J Neurovirol 2000 May;6 Suppl 2:S5-9
University of Cambridge Neurology Unit, Addenbrooke's Hospital, Hills Road, Cambridge CB2 2QQ, UK.
Epidemiological studies implicate an interplay between genetic and environmental factors in the aetiology of multiple sclerosis. The classical genetic observations suggest that multiple sclerosis is a complex trait in which susceptibility is determined by several genes acting independently or epistatically. The main dividend from understanding the genetic basis of susceptibility in multiple sclerosis will be an improved understanding of the pathogenesis. To date, candidate gene approaches have proved relatively unrewarding other than in establishing the association with alleles of the major histocompatibility complex (MHC). In common with most other complex traits, no major susceptibility gene has been identified through full genome screens but regions of interest have provisionally been identified. An important part of future studies in the genetics of multiple sclerosis will be to resolve the question of disease heterogeneity.
Progress in characterising the extent to which genetic factors contribute to the aetiology of multiple sclerosis has depended on classical studies involving the analysis of pedigrees and recurrence risk within families, and molecular investigation either of candidate genes or the whole genome using microsatellite and single nucleotide polymorphisms.
Families can be used in several ways to explore the aetiology of multiple sclerosis. In planning a genetic study in multiple sclerosis, the usual practice is to retrieve potentially informative cases from those already known to be affected. This requires the co-ordination of lists from various sources. For surveys testing biological features, such as genetic susceptibility in families, the trend is toward inclusion of individuals who have the disease process even if this does not yet meet standard criteria for clinically definite diagnosis. For example, in surveying the concordance of multiple sclerosis in twins the aim is identify all individuals having the disease process irrespective of whether this has reached some arbitrary point of clinical expression. However, if the criteria for inclusion are too relaxed, errors of inclusion will contaminate the results. It is necessary to be aware of potential ethnic differences within or between populations when planning or comparing studies of genetic susceptibility.
Genetic epidemiological studies providing morbidity statistics for multiple sclerosis have been used over many years to generate hypotheses for the aetiology of the disease (for review, see Compston, 1998). Multiple sclerosis is common in northern Europeans but not Orientals, Asians or Africans. There is a familial recurrence rate of approximately 15%. A meta-analysis of recurrence risk amongst 44 177 relatives of 2163 probands from three population based series (Sadovnick et al, 1988; Robertson et al, 1996; Carton et al, 1997) shows that the age-adjusted risk is highest for siblings (3%), then parents (2%) and children (2%) with lower rates in second and third degree relatives. Recurrence in monozygotic twins is around 35% (Sadovnick et al, 1993; French Research Group on Multiple Sclerosis, 1992; Mumford et al, 1994). Conversely, the frequency of multiple sclerosis in non-biological parents, siblings and children is more or less identical to the population lifetime risks for Europeans, and significantly lower than those expected from the study of recurrence in the biological relatives of index cases (Ebers et al, 1995). The age-adjusted risk for half-siblings is also less than for full siblings and no difference is seen between half-siblings reared together and apart (Sadovnick et al, 1996). Recurrence is higher in the children of conjugal pairs with multiple sclerosis (6%; age-adjusted 20%) than the offspring of single parents with multiple sclerosis (2%: Robertson et al, 1997).
Strategies for identifying susceptibility genes
Originally, the search for genes determining susceptibility to complex disease was organised by comparing the frequency of polymorphisms at a given genetic locus in groups of unrelated individuals with and without the trait. A change in strategy for studying the genetics of complex traits became routine over the last decade with the re-application of previously described family based methods (Risch, 1990). The biased inheritance of genomic regions presumed to encode susceptibility genes for multiple sclerosis in affected individuals forms the basis for family-based linkage and association studies.
Family based linkage methods have been applied most energetically to sibling pairs. The probability that, for a given locus, each affected sibling will share neither, one or both parental alleles is 0.25, 0.50 and 0..25, respectively. Linkage is present if, in a sufficient sample size, that distribution is distorted in favour of one or two sharing. The evidence relating to a given locus can be expressed as the logarithm of the odds (lod) for a gene being encoded at that site (the maximum lod score) or, conversely, as the probability that such a gene is not there encoded. Linkage is most efficient for detecting genes exerting major effects.
Many markers at adjacent loci across the genome are still in linkage disequilibrium. This means that association studies can be informative even when the marker is up to 0.5 centiMorgans (cM) from a disease susceptibility locus. Transmission disequilibrium testing (TDT) uses trios consisting of a single person with multiple sclerosis and both parents (who usually are but do not need to be unaffected). The aim is to show that in a sufficient sample size, alleles encoded at a particular locus are transmitted to the affected individual more (or less) often than expected by chance. The efficiency of TDT is determined by the extent of linkage disequilibrium.
Linkage and association provide types of information. Each has its limitations and advantages. Assuming a reasonable sample size, any marker which is well distant from the true susceptibility locus will be linked but not associated. A marker which is close to the susceptibility locus will not be linked in sibling pairs if the sample size is inadequate (and to date most sibling pair studies are not) but it may be associated. Only the combination of a close marker tested in an adequate sample will demonstrate both linkage and association. One strategy is first to screen the genome with linkage to establish regions of interest and then to refine the map with transmission disequilibrium testing.
Population studies demonstrate an association between the class II MHC alleles DR15 and DQ6 and their corresponding genotypes DRB1*1501, DRB5*0101 and DQA1*0102, DQB2*0602 (Olerup and Hillert, 1991) and the gene for TNF-alpha is usually also associated (Coraddu et al, 1998) since this is encoded within the same linkage group. For example, in Sardinians, multiple sclerosis is associated with DR4 (DRB1*0405-DQA1*0301-DQB1*0302) (Marrosu et al, 1992) although the DR15 association is seen amongst patients from all other parts of Italy in which adequate family studies have been performed La Mantia et al. 1990). In the Canaries, the disease is primarily associated with DR15 and DQ6 but a secondary association exists with DR4 (Coraddu et al, 1999). In Turkey, there is also an allelic association with both DR2 (DBR1*1501, DQA*0102, DQB1*0602) and DR4 (DRB1*04, DQA1*03, DQB1*0302: Saruhan-Direskeneli et al, 1997).
Extensive searches, using association and linkage studies over many years, have only yielded additional putative candidate genes in the VH2-5 immunoglobulin heavy chain and the T cell receptor beta chain variable regions. Recent contributions to this aspect of the genetics of multiple sclerosis have tended to illustrate the difficulty which exists in confirming weak associations (VH2-5; Wansen et al, 1997; Ligers et al, 1997: and TCR-beta: Droogan et al, 1996; Wansen et al, 1997). Studies of linkage or association with the genes encoding ICAM-1 (Mycko et al, 1998, IL-1ra (De La Concha et al, 1997; Wansen et al, 1997), IL-1beta and IFN-gamma (Wansen et al, 1997), IL-4 (Vanderbroeck et al, 1997) and CCR5 (Bennets et al, 1997) have all been negative. Except in the isolated population of Finns where there is both an association and linkage to the gene for myelin basic protein (Tienari et al, 1992, 1998), studies of structural genes of myelin have also been uninformatiive (Nellemann et al, 1995; Thompson et al, 1996; Price et al, 1997; Rodriguez et al, 1997; He et al, 1998a). Some investigators have explored new candidates. Using a variety of individuals (34 families, 147 cases and 95 controls), a Scandinavian group failed to implicate any one of several candidates encoding cytokines (IFN-gamma, IL-2, IL-4, IL-10, TGF-beta1 and -beta2, and IL-4R) or myelin proteins (PLP, MAG, OMGP, CNPase) - although IFN-gamma (12q24.1), IL-4R (16p12.1) and TGF-beta2 (1q41) could not be fully exclused (he et al, 1998b). In France, efforts have been directed at growth factors which determine oligodendrocyte development, and their receptors (Reboul et al, 1999). Again, no evidence for linkage emerges but these may also be type 2 erros due to the small sample size.
Four groups of investigators have undertaken a systematic search of the genome in an attempt to locate additional susceptibility genes using affected family members - using identity by descent analysis in sibling pairs. Genotyping was completed on cohorts each of between 21-225 families, together involving in excess of 1000 individuals, for each of between 257-443 microsatellite markers. These were cosen to have an average spacing of around 10 centMorgans, giving enough power to identify regions encoding a major susceptibility gene; and they are sufficiently polymorphici to make a high proportion of the available families fully informative. Although linkage analysis has revealed several new genomic regions which may encode genes conferring susceptibility to multiple sclerosis, some of these will be true and others false positives. Superficially, the results show a disappointing lack of overlap. In common with most other complex traits, no major susceptibility gene has been identified: the possible reasons are that no such gene exists, it has been missed by all three groups, or genetic complexity (i.e. heterogeneity) has obscured the picture. The importance of the major histocompatability complex is confirmed but of the other new putative susceptibility loci, several are clearly unique to each screen and so may be false positives. The regions of interest emerging from the UK genome screen (Sawcer et al, 1996; Chataway et al, 1998) weree Icen, 5cen, 6p, 7p, 14q, 17q, 19q, Xp. They were 2p, 3p, 5p, 11q and Xp in the Canadian series (Ebers et al, 1996). The US/French consortium identified 6p, 7q, 11p, 12q and 19q (Multiple Sclerosis Genetics Group, 1996). There were no statistically significant regions of interest in the relatively small Finnish screen (Kuokkanen et al, 1997) although positive lod scores were obtained for 6p21 (MHC) and 5p14-p12; increasing the density of markers raised the lod scores in several other regions (4cen, 11tel and 17q) whereas others (2q32 and 10q21) were unchanged; when all 21 families were typed across the regions of interest, the highest lod score (2.8) was at 17q22-q24 - as in the previously reported UK screen.
In a subsequent study from Scandinavia (Xu et al, 1999), 15 markers from regions of interest identified in one or more full genome screens were assessed using affected relative pairs (from 46 multiplex families, 28 trios, and 190 cases and 148 unrelated controls): depending on the method of analysis positive results were obtained for 5p15, 6p21, 7ptr15 and 12q23; the latter two also showed transmission distortion in families and an association in the case-control series. The remaining markers (2p23, 5p, 5q, 6q22-25, 7q, 11q21-23, 12q24-qtr, 13q33-34, 16p12, 18p11 and Xp21) did not identify regions of interest in the Scandinavian series. An Italian group used 67 markers covering regions of interest from the three full genome screens and candidates (HLA-DRb1, CTLA-4, IL-9, CSF-1R, APOE, Bcl-2 and TNF-R2) in 69 multiplex families (28 from Sardinia and 41 from mainland Italy). Some positive scores were obtained especially for the region at 5q14 in the Sardinian subset. Chromosome 5 has now featured in all the family studies but the region is broad and it may be difficult to assess in more detail. Three of these regions (2p11.2, 7p15.2 and 17q12) also showed positive TDT scores (D'Alfonso et al, 1999).
These genetic analyses are predicated on the assumption that multiple sclerosis is one disease. Mutations of mitochondrial DNA are responsible for a multiple sclerosis-like illness characterised by disproportionate involvement of the anterior visual pathway (Harding et al, 1992; Riordan Eva et al, 1995) although mitochondrial genes do not contribute generally to susceptibility to multiple sclerosis (Kellar Wood et al, 1994; Kalman et al, 1995; Nishimura et al, 1995). Conditioning the United Kingdom genome screen DR15 (or an extended DR15 linked haplotye also encoding alleles of TNF and the DQ locus) shows that the regions of interest on 1p, 17p, 17q and X cluster in familes which are identical by state for DR15 whereas the non-sharing group associated with 1cen, 3p, 5cen, 7p, 14q and 22q (Coraddu et al, 1999); in addition new regions of interest are found at 5q and 13p (DR15 sharing) and 16p and 20p (DR15 non-sharing). A major part of future studies in the genetics of multiple sclerosis will be to resolve the question of disease heterogeneity.
The contribution to susceptibility made by the genes which have provisionally been identified only accounts for a proportion of the increased risk of multiple sclerosis implicated by family studies; clearly there is more to be learned concerning susceptibility to multiple sclerosis and the interplay between genes and environmental factors. The application of this knowledge for improved understanding of the pathogenesis, in couselling and in designing novel treatments is potentially considerable.