Multiple sclerosis and the HLA-D region: linkage and association studies

Multiple sclerosis and the HLA-D region: linkage and association studies

Journal of Neuroimmunology ELSEVIER Journal of Neuroimmunology 58 (1995) 183-190 Multiple sclerosis and the HLA-D region: linkage and association st...

1MB Sizes 0 Downloads 20 Views

Journal of Neuroimmunology ELSEVIER

Journal of Neuroimmunology 58 (1995) 183-190

Multiple sclerosis and the HLA-D region: linkage and association studies ay*, N.W. Wood a, P. Holmans b, D. Clayton b, N. Robertson D.A.S. Compston a

Kellar-Wood

H.F.

a lJniversi@ of Cambridge Neurology unit, Addenbrooke’s Hospital, Hills Road, Cambridge CB2 2QQ, UK ’ MRC Biostatistics Unit, Institute of Public Health, University Forvie Site, Robinson Way, Cambridge CB2 2SR,

a,

UK

Received 24 October 1994; revised 5 January 1995; accepted 5 January 1995

Abstract Inheritance patterns of multiple genetically determined components.

sclerosis (MS) in multiplex families suggest a complex aetiology involving environmental The association

between

the HLA class II DR1.5, DQ6, Dw2 haplotype

and and MS has been

well documented in patients with ancestral origins in Northern Europe. Conversely, linkage analysis of this region in multiplex families, derived from a population base, has generated negative results. Thus, given the Dw2 specificity association, evidence implicating this locus in disease susceptibility appears contradictory. We have collected and determined the HLA-DR and -DQ haplotypes of 115 sibling pairs with multiple sclerosis, and confirm a significant association with the Dw2-associated haplotype, both in index cases and their affected siblings compared with controls. However, using a sibling pair linkage analysis that restricts haplotype sharing probabilities to defined genetic models, we have not observed linkage of this region to susceptibility in MS. We discuss the basis for association and linkage and conclude that the DR15, DQ6, Dw2 haplotype does represent a susceptibility locus but its contribution to the pathogenesis is small; although it may interact epistatically with other susceptibility genes, this haplotype is not necessary for disease expression. Keywords:

Multiple sclerosis; HLA association;

Linkage analysis -

1. Introduction Multiple sclerosis (MS) is a chronic inflammatory demyelinating disease of the central nervous system with a prevalence of approximately 130/100000 in the UK (Mumford et al., 1992). Although neither the aetiology nor the pathogenesis are well understood, the

disease is thought to be multifactorial involving genetic and environmental factors, but neither can fully account for disease expression. Genetic components are implicated by family studies, illustrating an increasing prevalence with degree of kinship to the index case (Sadovnick et al., 1988; Robertson et al., 1994) and by twin studies, in most of which a significantly higher concordance rate is observed in monozygotic (26-31%) compared to dizygotic (2.6%) twin pairs (Ebers et al., 1986; Sadovnick et al., 1993; Mumford et al., 1994). * Corresponding (1223) 336 941 0165-5728/95/$09.50

author. Phone

+44 (1223) 217 222; Fax +44

0 1995 Elsevier Science B.V. All rights reserved

SSDI 0165-5728(95)00015-l

MS was originally shown to be associated with class I alleles (A3 and B7) of the human leukocyte antigen (HLA) region and later with the class II determined specificity, Dw2 (Jersild et al., 1972; Jersild et al., 1973; Batchelor et al., 1978). More recently, application of restriction fragment length polymorphism (RFLP) and sequence-specific oligonucleotide analyses has allowed the fine sub-typing of -DR, -DQ and -DP types at the DNA sequence level, and the unique HLA haplotype associated with MS in many populations is HLA DR15 (DRBl.lSOl), DQ6 (DQA1.0102, DQB1.0602), Dw2 (Spurkland et al., 1991). This association has been confirmed in a number of studies involving Caucasian patients of northern European ancestry, but it is not universally observed; for example in Sardinians the association is with DR4 and DQB1.0302 (Marrosu et al., 1988; Marrosu et al., 1992). However, the DR15, DQ6 haplotype is still observed more often in patients than controls derived from ethnically diverse populations, despite variations in the allele and haplotype

184

H. F. Kellar- Wood et al. /Journal

of Neuroinvnunology 58 (1995) 183-I 90

frequencies; the strength of these associations can be masked by different linkage disequilibria between alleles comprising DR2 types in separate ethnic groups (Takacs et al., 1990; Sejeantson et al., 1992). At best, this association is weak, with a phenotype frequency of 50-60% in patients compared with 25% in controls; thus the associated haplotype is neither necessary nor sufficient for the expression of the disease, but may confer a relative risk of between 2 and 5 (Olerup and Hillert, 1991). Linkage and association studies can both be used to assess the genetic influence of a locus on disease susceptibility, but their differences may be confused. Association studies compare the frequency of a specific disease marker in a population of unrelated patients with controls from the same population. In order to avoid the potential bias of population stratification, associations can be considered within families using unaffected members as controls and therefore achieving very close matching. However, the degree of relatedness and hence frequency of alleles between cases and related controls diminishes the power of this method of analysis, whereas genetic transmission analysis methods (Self et al., 1991) overcome the loss of power whilst maintaining the close matching. Linkage, which determines whether there is a necessary disease gene encoded in the candidate region, is traditionally used in studies of large pedigrees containing multiple affected cases. In late onset multifactorial diseases such as MS, multiplex families are rarely observed and parents are not always available for analysis. The alternative method for studying linkage in this situation is to determine inheritance by descent (IBD) of haplotypes in affected sibling pairs; when loci are inherited in a Mendelian fashion, sibling pairs share 0, 1 or 2 haplotypes at a frequency of 0.25, 0.50 and 0.25. Evidence for linkage is provided by demonstrating a bias in favour of 1 and 2 haplotype sharing (Penrose, 1953; Risch, 1990a). This method of IBD analysis makes no assumptions regarding mode of inheritance; since patients alone but not their unaffected relatives are studied, variable penetrance at any locus and age of onset can each be ignored. The power of this method has been increased by restricting the maximisation to the set of sharing probabilities consistent with possible genetic models (Holmans, 1993). We studied 115 affected sibling pairs but did not observe linkage to the HLA-D region; conversely, we confirmed the disease association with the DR15, DQ6, Dw2 haplotype in this cohort of index cases, and with a population-based series of unrelated MS patients; comparisons were made both with unaffected siblings (using an intra-family association) and with a panel of unrelated controls. Using family based transmission analysis, we have shown that in multiplex families the MSassociated haplotype is inherited significantly more fre-

quently by affected than unaffected offspring. We conclude that the HLA-D region confers a susceptibility risk via the class II haplotype encoding DR15 and DQ6, but does not contain a gene necessary for disease expression; we discuss how the apparent contradiction between the results of association and linkage to candidate susceptibility genes in the same population can be resolved.

2. Materials and methods 2.1. Patients

Unrelated Caucasian patients with MS (n = 93) were identified as part of a prevalence study carried out in the Cambridge health district of East Anglia between 1990-1992 (Mumford et al., 1992). A second group of patients was recruited from 522 families with more than one affected member identified from throughout the United Kingdom. Of these, 291 families have affected sibling pairs and the first 109, including six with affected trios, were included in the linkage analysis. The trios were weighted 2/3 in the analysis, due to the lack of independence in the third pair, bringing the total number of sibling pairs studied for linkage to 115. The total number of index siblings used in the association study was 106. Strict clinical criteria were applied when confirming the diagnosis and both siblings had to meet criteria for clinically definite, laboratory supported definite or clinically probable disease (Poser et al., 1983). A venous blood sample (30 ml) was taken in EDTA from affected individuals and as many first degree relatives as possible (parents and siblings of affected individuals). 2.2. Controfs The control group comprised 130 unrelated caucasians, attending blood donation sessions in the Cambridge health district, matched for age, with a range of 19-73, and sex with the patient cohort identified in the Cambridge health district prevalence study (age range 17-80). 2.3. Haplotype assignment Genomic DNA was isolated from 15 ml venous blood by phenohchloroform extraction followed by ethanol precipitation (using standard methods). DNA (8-10 pg> was digested with Taq I restriction enzyme (Boehringer Mannheim) according to the manufacturers instructions, DNA restriction fragments were separated by electrophoresis (30 V for 16 h) in a 0.8% agarose gel and using the method of Southern (1975) blotted onto nylon, and optimally UV fixed. DNA

H.F. Kellar- Wood et al. /Journal

of Neuroimmunology 58 (1995) 183-190

filters were subsequently hybridised with radiolabelled DRBl and DQAl cDNA probes (kindly provided by P. Petterson, Uppsala. Sweden and J. Bidwell, Bristol, UK). The DRB probe was a 517-bp cDNA from pRTV1 (Bidwell and Jarrold, 19861, and the DQA probe was the 797-bp cDNA from pDCH1 (Auffray et al., 1984). After hybridisation, all filters were washed twice at room temperature in 2 X SSC, 5% SDS, followed by one wash with 0.2 X SSC, 5% SDS at 65” C. Filters were then exposed to X-ray film at -70” C for 1-5 days. Assignment of Taq I DRB and DQA genotypes on the basis of restriction fragment length polymorphism (RFLP) hybridisation bands was completed using standard criteria (Bidwell, 1988). DQBl analysis was not performed since linkage disequilibrium between alleles of the class II genes DQBl, DQAl and DRBl is very strong within specific ethnically defined haplotypes (Olerup and Hillert, 1991; Begovich et al., 1992). The haplotypes derived from this analysis are listed in Table 1. It should be noted that by using the Taq I DQA and DRB RFLP analyses alone it is not possible to determine a difference between DRB17(1), DQ2, (encoded by DRBl. 0301, DQA1.05011 and DQB1.0201) and DR6, DQ7 (encoded by DRB1.1402, DQA1.0501 and DQB1.0301). However, the DR6, DQ7 (14b) haplotype is extremely rare in Caucasian populations, having a frequency of 0.0 in series of 150 controls from Northern Ireland (Cullen et al., 1991); therefore all haplotypes have been classified as DR17, DQ2 for the purpose of statistical analyses.

185

2.4. Statistical methods The linkage analysis of 115 sibling pairs was based on a modification of the IBD method of maximum likelihood (Holmans, 1993). Unambiguous assignment of haplotypes is only possible when no more than one parental marker locus is homozygous; thus, when parents are heterozygous at more than one locus or not available for study, sharing probabilities for ancestral alleles were estimated using the method of maximum likelihood (Risch, 1990b). However, to ensure that these results conform to the possible constraints of Pr (0 IBD) < 0.25, Pr (1 IBD) < 0.5 and Pr (2 IBD) > 0.25, an iterative procedure (EM algorithm) was used to find the required estimates of sharing probabilities which provide the best fit to the data. In each case a likelihood ratio Chi-squared (~‘1 test for linkage has been calculated and the value of the likelihood ratio test statistic is expressed as a lod score (Holmans, 1993). Population association analyses were performed using a likelihood ratio method in which estimates for the log-likelihood for each population are maximised separately in the numerator and jointly as one sample in the denominator. The same analysis was used when making comparisons between related and unrelated individuals. The resulting test statistic has a x2 (n - 11 distribution where n represents the number of alleles. Family-based transmission analysis (Self et al., 1991) was carried out to determine whether affected individuals inherited the DR15 allele from their parents more

1 IBD

1 IBD

A

0 IBD

2 IBD

Fig. 1. Diagrammatic representation of affected sibling pair distribution of HLA-DR/DQ haplotypes within the ‘possible triangle’ of genetic models (Holmans, 1993) where the perpendicular distance of a point to the side of a triangle opposite to the j (j = 0, 1 or 2) IBD corner is proportional to Pr j IBD for that family. (A) The null hypothesis IBD (0.25, 0.5, 0.25). (B) The maximum likelihood solution IBD (0.24, 0.5, 0.26). Open circles represent a single family and closed circles more than one family.

H.F. Kellar- Wood et al. /Journal

186 Table 1 Haplotype (n = 415)

frequencies

of Neuroimmunology 58 (1995) 183-190

in all individuals from multiplex families

Haplotype

Frequency

DRl/DQS DR15/DQ6 DR16/DQ5 DR3/DQ2 DR4 DRll/DQ7 DR6/DQ5 DR6/DQ6 DR6/DQ7 DR7/DQ2 DR7/DQ9 DR8 DR9

0.0886 0.2804 0.0026 0.1937 0.1526 0.0526 0.0186 0.0401 0.0241 0.0398 0.0521 0.0428 0.0079

A total of 13 haplotypes, determined by DRB and DQA alleles was present in all 109 families (Table 11, and these provide a polymorphic information content (PIG) value of 0.81. We found no evidence for linkage of this region to MS susceptibility since the expected and observed probabilities of haplotype sharing IBD were almost identical (Table 2, Fig. 1) and generated a maximum lod score of only 0.024 (P = NS). Using the less powerful method of analysis of haplotypes shared identically by state (IBS), there was no evidence for linkage in 115 affected sibling pairs, (x2 = 0.11; df = 2, P = NS; Table 3). Conversely, a significant population association was observed between the DR15, DQ6, Dw2 haplotype and MS both in the sporadic patients (x2 = 31.2; df = 12; P < 0.01) and index cases from affected sibling pairs Table 2 Sharing of HLA class II haplotypes identical by descent (IBD) in 115 affected sibling pairs

0 1 2

0.24 0.5 0.26

Observed

Expected

0 1 2

13.7 59.2 36.0

14.3 57.6 37.1

Table 4 Significance values of presence for the DR15, DQ6, Dw2 haplotype

3. Results

Observed

Haplotypes shared IBS

x2 = 0.11; df = 2; P > 0.1 (not significant).

frequently than would be expected by chance. Such an excess would constitute evidence for an association between DR15 and MS. This type of analysis is more powerful than the intrafamily association considering unaffected siblings as matched controls, although the latter was performed for comparison purposes. Furthermore, using this method, there is no need for a separate control population thus removing a large potential source of bias.

Haplotypes shared IBD

Table 3 Sharing of HLA class II haplotypes identity by state (IBS) in 115 affected sibling pairs

Expected 0.25 0.50 0.25

Maximum lod score = 0.024. P > 0.1 (not significant).

Comparisons

x2

df

Significance value

Controls Controls Controls Sporadic

31.2 29.5 36.2 9.4

12 12 12 12

0.005 > P > 0.001 0.005 > P > 0.001 0.001 > P > 0.0001 NS

vs. sporadic cases vs. index siblings vs. all cases cases vs. siblings

29.5; df = 12; P < 0.01); haplotype frequencies between the two patient groups did not differ (x2 = 9.37; df = 13, P = NS; Table 4). The relative risk associated with the DR15 genotype was estimated for individuals in three categories: DR15/DR15; DRlS/DRX and DRX/DRX (where DRX represents any other DR type). A case/control analysis was carried out considering unrelated patients and the index case from each family as a single group (n = 199; x2 = 22.42; df = 2; P < 0.001; Table 5). The relative risk associated with an heterozygous DR15 genotype relative to DRX/DRX was 2.6, and 8.27 with a homozygous DR genotype relative to DRX/DRX, but the difference between these relative risks was not significant. In order to determine whether linkage was masked by including families which did not carry the MS-associated class II haplotype in the analysis, we analysed separately for IBD sharing those families positive for the DR15 allele (79 of 109). There was no evidence for linkage in these families alone; however, within these DRlS-positive families, 51/79 affected sibling pairs shared one DR15, DQ6, Dw2 haplotype IBS. When the affected siblings of index cases were considered as a second population and compared to unrelated controls, the DR15, DQ6, Dw2 haplotype occurred in 67 of 109 families at a frequency of 61.5% (x2 = 24.99; df = 2; P < 0.0001) confirming the association with MS. (x2 =

Table 5 Case-control comparison of DR15 genotypes Population

DRtS/DRlS

DRlS/DRX

DRX/DRX

Patients a (n = 199) Controls (n = 130)

16 2

93 37

90 93

a patient group comprises index siblings and sporadic cases. x2 = 22.4; df = 2; P < 0.0001.

H.F. Kellar- Wood et al. /Journal

of Neuroimmunology 58 (1995) 183-190

The intra family association study, which treated unaffected siblings as matched controls of the affected individuals in each family was compromised by the fact that not every family had unaffected siblings, reducing the power of this analysis by comparison with the unmatched case-control series. However, the results confirmed the population association with DR15, DQ6, Dw2 and MS (x2 = 8.21; df = 2; 0.025 > P > 0.001). The more powerful method of analysis involving transmission of haplotypes from parents to affected versus unaffected siblings showed that the MS-associated haplotype segregated to affected siblings more frequently than expected by chance (x2 = 15.65; df = 1; P < 0.001).

4. Discussion Population association studies have repeatedly documented the over-representation of the DR15, DQ6, Dw2 haplotype in patients with MS compared with controls (Swingler et al., 1987; Olerup and Hillert, 1991). We confirmed this association in comparisons of index patients with MS from multiplex families, their affected siblings and unrelated patients, with unaffected relatives and unrelated controls. Having demonstrated that our families are genetically representative of patients with MS, we have shown in the same sample that susceptibility to MS is not linked to the DR15, DQ6, Dw2 haplotype. This haplotype has a relatively high frequency in northern Europeans and their descendants, and since MS also occurs frequently in this ethnic group, the association could merely be a population marker and not causal. Nevertheless, the observation that this haplotype also occurs more frequently in patients than controls from other ethnic groups, even when contributing alleles are present at a low population frequency, supports a more significant role for HLA in disease susceptibility. Linkage analysis using IBD in MS has proved difficult in the past, owing to the requirement for a large number of families with affected sibling pairs. Since the lod score value is cumulative, studies reporting no linkage with the HLA region may be explained by the low numbers of available affected sibling pairs (Schroder et al., 1983). Data analysed from independent reports of family studies did subsequently claim linkage of MS to the HLA region (Stewart et al., 1981; Gompston and Howard, 1982; Ho et al., 1982), but these can be criticised for introducing biases of case ascertainment and not taking into consideration the possibility of variation in HLA gene frequencies between populations. In a recent reappraisal, Tienari et al. (1993) have investigated association and linkage of DQA1.0102 and DQB1.0602 alleles with MS in 21 Finnish families. Both alleles were associated with MS

187

patients and unaffected parents of MS patients and IBD analysis generated a x2 value of 6.49 (P < 0.05). If families with a parent homozygous for DQA1.0102 were excluded, this increased to 10.95 (P < 0.005); this is apparent since when parental haplotypes are homozygous for the associated allele, both haplotypes are transmitted equally to the affected offspring, providing no evidence of linkage (Julier et al., 1991). Using the same sample of Finnish families, MS also is linked with the myelin basic protein gene on chromosome 18 (Tienari et al., 1992; Tienari et al., 1994); conversely, in our British families and all other cohorts there is no evidence of linkage or association to MBP locus (Graham et al., 1993; Rose et al., 1993; Wood et al., 1994). The Finnish families (selected from a homogenous population) may have alternative susceptibility loci to British MS patients reflecting disease heterogeneity. We have analysed > 100 pairs for IBD haplotype sharing and did not obtain a lod score which suggests linkage to HLA; this finding supports the results of a population based IBD study of 40 sibling pairs which used serological HLA typing methods (Ebers et al., 1982) in addition to a non-population based IBD study of 20 sibling pairs and 2 trios (Govaerts et al., 1985). Furthermore, analysis of 79 DRlS-positive families failed to provide evidence of linkage. These results have been interpreted as providing evidence against a susceptibility gene linked to HLA and appear to conflict with the results of association studies; it is important therefore to stress the different categories of information provided by association and linkage studies. Linkage establishes whether a marker locus is located in the proximity of a necessary disease locus by showing co-segregation of the disease with any allele encoded at the marker locus. It reflects the movement of alleles between generations but it is insensitive to which allele may be co-segregating with the disease. Association determines whether a specific allele/mutation, which can be at the marker locus itself or in strong linkage disequilibrium, can be detected more frequently in patients than controls (Thomson and Bodmer, 1977; Hodge, 1993; Ebers and Sadovnick, 1994). In general, association is interpreted as representing linkage disequilibrium between the marker and disease susceptibility gene and signifies a risk for disease reflecting biological involvement of the susceptibility gene product at some level in the pathogenesis. Hodge (1993) has offered two explanations which account for weak to moderate disease associations; one based on linkage disequilibrium (LD) and the other on a susceptibility locus (S). The LD model assumes no phenocopy rate and no background incidence of disease without the disease gene; thus an incomplete association can only be explained by recombination between the marker and disease gene, implying that

188

H.F. Kellar- Wood et al. /Journal

of Neuroimmunology

there is linkage disequilibrium. However, there is a danger in adopting the requirement for a necessary gene in a multifactorial disease since this suggests a major locus, and the disease can clearly occur in individuals without the susceptibility allele. Thus, ‘necessary’ must be quantitatively dependent on relative risk conferred by the disease allele, such that a high frequency of disease may still be observed in the absence of the disease allele. Greenberg (1993) has proposed linkage analysis tests designed to distinguish S and LD models. The first approach is to determine if there is also evidence for linkage, once the association has been established; if not, the susceptibility model is favoured, since only linkage disequilibrium represents true linkage. The second approach is the partitioned association-linkage test. Here, families are partitioned into those where the index case does and does not have the associated allele; the method subsequently assesses whether there are differences in linkage analyses between these two groups. Marked deviation from the null hypothesis of IBD (0.25, 0.5, 0.25) occurs when the relative risk is fairly high, and approaches a distribution consistent with linkage. When the relative risk is low, little variation from the null hypothesis is observed and the susceptibility locus model is favoured (Greenberg and Hodge, 1992; Hodge, 1993). Indeed, when the probability of having the disease with the associated allele is less than 10 times greater than having the disease without the allele it is difficult to detect linkage. Since the relative risk conferred by the DR15, DQ6, Dw2 haplotype is low, and no marked deviation from the null hypothesis of IBD was observed in our MS families in which the index cases carried the associated haplotype (data not shown), our data would favour the S model. The majority of disease associations are incomplete - the susceptibility allele also occurring in unaffected individuals, as is the case for the MS-associated HLA class II haplotype. This observation has a number of alternative explanations; first, disease or genetic heterogeneity within the cohort under investigation may apply in late onset diseases which exhibit variable presentation and a syndrome of clinical symptoms. Secondly, if association is the result of linkage disequilibrium, there may be recombination between the disease gene and the marker in some families, resulting in only a weak or moderate association and eliminating altogether evidence for linkage. Thirdly, observed HLA haplotype frequencies within populations may disproportionately reduce the ability to demonstrate association and linkage where the MS-associated haplotype has a high population frequency; the chance that siblings will share this haplotype identically by state, but not by descent, is also high. Demonstration of linkage is also confounded in studies where the recombination fraction is closer to 0.5 than 0.0, since affected individ-

58 (1995) 183-190

uals within a family who do not carry the susceptibility allele will be interpreted as recombinants in the linkage analysis. We have demonstrated the Dw2 haplotype association in the same cohort of families in which we have not observed linkage. It has been suggested that this type of conflict results from disease heterogeneity, such that Dw2 has a role in sporadic but not in familial disease. This hypothesis must be rejected on the basis of our results since there is no evidence to show that the disease differs in clinical phenotype, age of onset, symptom expression or in other features between sporadic and familial cases (Weinshenker et al., 1990; Robertson et al., 1994). It is not possible to determine which part of the DR15, DQ6, Dw2 haplotype encodes the disease susceptibility gene since linkage disequilibrium within this haplotype is particularly strong both in the class II region, and even extends into the class I and class III regions (Tiwari and Terasaki, 1985). Additionally, most alleles present within the haplotype occur in different combinations in other non-MS-associated haplotypes; for example, DQA1.0102 occurs in the Dwl9-associated haplotype. The only DwZunique alleles are DRB5.0101 and DRB1.1501, which do not form part of any other Caucasian haplotype. However, the HLA-associated haplotype present in MS patients appears identical to that found in the general population (Cowan et al., 1991). To date, three alleles of DR15 have been observed at the sequence level, (DRB1.1501, .1502 and .1503) and > 90% of the British population, including MS patients, carry the DRB1.1501 allele (Cullen et al., 1991). Using a transmission test, we have demonstrated significant segregation of the DwZassociated haplotype to affected versus unaffected offspring, which is in accordance with the dominant action of this haplotype on susceptibility (Hillert et al., 1994). Extensive analysis of the haplotype in this region has shown no evidence for disease-specific mutations. In the last few years, efforts have been made to identify susceptibility with a specific part of the Dw2 haplotype; initially it was suggested that the DQ genes may encode an additional susceptibility locus to DR (Heard et al., 1989). Later studies demonstrated that MS is associated with DQBl genes sharing stretches of amino acid sequences, suggesting that this accounted for the ethnic variation in DR associations with MS including, DR2 in northern Europeans, DR6 in Japanese and DR4 (DQB1.0302) in Sardinians (Vartdal et al., 1989; Spurkland et al., 1991); however, these associations have not been confirmed. The same DQB genes were observed more frequently in Swedish (Olerup and Hillert, 1991) and Canadian patients (Haegert et al., 1993) than in controls, but when DRlS-positive patients were excluded from these analyses, no differences were observed between patients and controls

H. F. Kellar- Wood et al. /Journal

of Neuroimmunology 58 (1995) 183-I 90

apart from those considered secondary to the association with the DR15, DQ6, Dw2 haplotype. One explanation for the effect of susceptibility loci on disease expression is the existence of multiple interacting genes, such that in addition to linked necessary gene(s), a disease susceptibility locus may alter the penetrance of another gene by lowering the threshold for disease expression. We have investigated genetic linkage of MS with the immunoglobulin (IgG) heavy chain region and with a gene encoded at the T cell receptor p locus in the same families included in this study (Wood et al., 1995a; Wood et al., 1995b). To determine if any of these immune response genes interact epistatically we performed a linkage analysis of both the immunoglobulin (IgG) and T cell receptor (TCR) loci in 53 sibling pairs who share the MS-associated haplotype. This class II stratification did not alter the negative linkage to the IgG locus and thus provided no evidence for an epistatic interaction between the MS-associated haplotype and the IgG heavy chain; however, linkage analysis of the TCR p chain region increased TCR haplotype sharing IBD from 0.24, 0.5 and 0.26 to 0.16, 0.41 and 0.43 in 53 sibling pairs sharing the DR15, DQ6, Dw2 haplotype, generating a lod score of approximately 1 (P < 0.05). Taken together, these results indicate that both the HLA class II region and a locus linked to the TCR p chain region may contribute to susceptibility in MS epistatically due to the requirement of HLA-TCR co-operation during antigen presentation and recognition.

Acknowledgements This study was funded by the Multiple Sclerosis Society of Great Britain and Northern Ireland. We are grateful to patients and their families participating in this study; to research nurses Mary Fraser and Jackie Deans for visiting and collecting blood samples from MS patients and controls and to Julia Gray for technical assistance.

References Auffray, C., Lillie, J.W., Arnot, D., Grossberger, D., Kappes, D. and Strominger, J.L. (1984) Isotypic and allotypic variation on human class II histocompatability antigen alpha chains. Nature 308, 327-333. Batchelor, J.R., Compston, D.A.S. and McDonald, WI. (1978) The significance of the association between HLA and multiple sclerosis. Br. Med. Bull. 34, 279-284. Begovich, A.B., McClure, G.R., Suraj, V.C., Helmuth, R.C., Fildes, N., Bugawan, T.L., Erlich, H.A. and Klitz, W. (1992) Polymorphism, recombination and linkage disequilibrium within the HLA class II region. J. Immunol. 148, 249-2.58. Bidwell, J. (1988) DNA-RFLP analysis and genotyping of HLA DR and DQ antigens. Immunol. Today 9, 18-23.

189

Bidwell, J. and Jarrold, E. (1986) HLA-DR allogenotyping using exon-specific cDNA probes and application of rapid minigel methods. Mol. Immunol. 23, 1111-1116. Cowan, E.P., Pierce, M.L., McFarland, H.F. and McFarlin, D.E. (19911 HLA-DR and -DQ allelic sequences in multiple sclerosis patients are identical to those found in the general population. Hum. Immunol. 32, 203-210. Compston, D.A.S. and Howard, S. (1982) HLA typing in multiple sclerosis. Lancet ii, 661. Cullen, C.G., Middleton, D., Savage, D.A. and Hawkins, S. (1991) HLA-DR and DQ DNA genotyping in multiple sclerosis patients in Northern Ireland. Hum. Immunol. 30, l-6. Ebers, G.C. and Sadovnick, A.D. (1994) Association studies in multiple sclerosis. J. Neuroimmunol. 53, 117-122. Ebers, G.C., Paty, D.W., Stiller, CR., Nelson, R.F., Seland, T.P. and Larsen, B. (1982) HLA typing in sibling pairs with multiple sclerosis. Lancet ii, 88-90. Ebers, G.C., Bulman, D.E., Sadovnick, A.D., Paty, D., Warren, W.S., Hader, W., Murray, T.J., Seland, T.P., Duquette, P., Grey, T., Nelson, R., Nicolle, M. and Brunet, D. (1986) A population-based study of multiple sclerosis in twins. N. Engl. J. Med 315, 16381642. Govaerts, A., Gony, J., Martin-Mondiere, C., Poirer, J.C., Schmid, M., Schuller, E., Degos, J.D. and Dausset, J. (1985) HLA and Multiple sclerosis: population and families study. Tissue Ant. 25, 187-199. Graham, C., Kirk, C., Nevin, N., Droogan, A.G., Hawkins, S.A., McMillan, S.A. and McNeill, T.A. (1993) Lack of association between myelin basic protein gene microsatellite and multiple sclerosis. Lancet 341, 1596. Greenberg, D.A. (1993) Linkage analysis of ‘necessary’ disease loci versus ‘susceptibility’ loci. Am. J. Hum. Genet. 52, 135-143. Greenberg, D.A. and Hodge, S. (1992) Identity by descent (IBD) distributions in affected sib pairs (ASPS) in presence of a disease marker association can help distinguish between models. Am. J. Hum. Genet. (Suppl.) 51, A150. Haegert, D.G., Muntoni, F., Murru, M.R., Costa, G., Francis, G.S. and Marrosu M.G. (1993) HLA-DQAl and -DQBl associations with multiple sclerosis in Sardinia and French Canada: evidence for immunogenetically distinct patient groups. Neurology 43, 548-552. Heard, R.N., Cullen, C., Middleton, D., Hawkins, S.A., Francis, D.A., Hern, J.E., McDonald, W.I., Batchelor, J.R. and Lechler R.I. (1989) An allelic cluster of DQ alpha restriction fragments is associated with multiple sclerosis: evidence that a second haplotype may influence disease susceptibility. Hum. Immunol. 25, ill- 123. Hillert, J., Kail, T., Vrethem, M., Fredrikson, S., Ohlson, M. and Olerup, 0. (1994) The HLA-Dw2 haplotype segregates closely with multiple sclerosis in multiplex families. J. Neuroimmunol. 50, 95-100. Ho, H., Tiwari, J.L., Haile, R.W., Terasaki, P.I. and Morton, N.E. (1982) HLA-linked and unlinked determinants of multiple sclerosis. Immunogenetics 15, 509-517. Hodge, SE. (1993) Linkage analysis versus association analysis: distinguishing between two models that explain disease-marker associations. Am. J. Hum. Genet 53, 367-384. Holmans, P. (1993) Asymptotic properties of affected-sib-pair linkage analysis. Am. J. Hum. Genet. 52, 362-374. Jersild, C., Svejgaard, A. and Fog, J. (1972) HLA antigens and multiple sclerosis. Lancet i, 1240-1241. Jersild, C., Hansen, G., Svejgaard, A., Fog, T., Thomsen, M. and DuPont, B. (1973) Histocompatibility determinants in multiple sclerosis. Lancet ii, 1221-1224. Julier, C., Hyer, R.N., Davies, J., Merlin, F., Soularue, P., Briant, L., Cathelineau, G., Deschamps, I., Rotter, J.L., Froguel, P., Boitard, P., Bell, J.I. and Lathrop, G.M. (1991) Insulin-IGF2 region on

190

H.F. Kellar- Wood et al. /Journal

of Neuroimmunology 58 (1995) 183-190

chromosome llp encodes a gene implicated in HLA-DR4 dependent diabetes susceptibility. Nature 354, 155-159. Marrosu, M.G., Muntoni, F., Murru, M.R., Spinicci, G., Pischedda, M.P., Goddi, F., Cossu, P. and Pirastu, M. (19881 Sardinian multiple sclerosis is associated with HLA-DR4: a serologic and molecular analysis. Neurology 38, 1749-1753. Marrosu, M.G., Mutoni, F., Murru, M.R., Costa, G., Pischedda, M.P., Pirastu, M.P., Sotgiu, S., Rosati, G. and Cianchetti, C. (1992) HLA-DQBl genotype in Sardinian multiple sclerosis: evidence for a key role of DQBl.0201 and .0302 alleles. Neurology 42, 883-886. Mumford, C.J., Fraser, M.B., Wood, N.W. and Compston, D.A.S. (1992) Multiple sclerosis in the Cambridge health district of East Anglia. J. Neural. Neurosurg. Psychiatry 55, 877-882. Mumford, C.J., Wood, N.W., Kellar-Wood, H.F., Thorpe, J.W., Miller, D.H. and Compston, D.A.S. (1994) The British Isles survey of multiple sclerosis in twins. Neurology 44, 1l-15. Olerup, 0. and Hillert, J. (1991) HLA class B-associated genetic susceptibility in multiple sclerosis: a critical evaluation. Tissue Ant. 38, 1-15. Penrose, L.S. (1953) The general-purpose sib-pair linkage test. Ann. Eugenics 6, 133-138. Poser, C.M., Paty, D.W., Scheinberg, L., McDonald, WI., Davis, F.A., Ebers, G.C., Johnson, K.P., Sibley, W.A., Silberberg, D.H. and Tourtellotte, W.W. (1983) New diagnostic guidelines for multiple sclerosis: guidelines for research protocols. Ann. Neurol. 13, 227-231. Risch, N. (1990al Linkage strategies for genetically complex traits: the power of affected relative pairs. Am. J. Hum. Genet. 46, 229-241. Risch, N. (199Ob) Linkage strategies for genetically complex traits. III. The effect of marker polymorphism on the analysis of affected relative pairs. Am. J. Hum. Genet. 46, 242-253. Robertson, N.P., Fraser, M.B., Deans, J., Clayton, D. and Compston, D.A.S. (1994) Multiple sclerosis sibling pairs. J. Neurol. Neurosurg. Psychiatry 57, 250. Rose, J., Gerken, S., Lynch, S., Pisani, P., Varvil, T., Otterud, B. and Leoppard, M. (1993) Genetic susceptibility in familial multiple sclerosis not linked to the myelin basic protein gene. Lancet 341, 1179-1181. Sadovnick, A.D., Baird, P.A. and Ward, R.H. (1988) Multiple sclerosis: updated risks for relatives. Am. J. Med. Genet. 29, 533-541. Sadovnick, A.D., Rice, G.P., Bulman, D., Hashimoto, L., Paty, D.W., Hashimoto, S.A., Warren, S., Hader, W., Murray, T.J., Seland, T.P., Metz, L., Bell, R., Duquette, P., Gray, T., Nelson, R., Weinshenker, B., Brunet, D. and Ebers, G.C. (19931 A population-based study of multiple sclerosis in twins: update. Ann. Neurol. 33, 281-285. Schroder, R., Zander, H., Andreas, A. and Mauff, G. (19831 Multiple sclerosis: immunogenetic analyses of sib-pair double case families. II. Studies on the association of multiple sclerosis with C2, C4, BF, C3, C6 and GLO polymorphisms. Immunobiology 164, 160-170. Self, S.G., Longton, G., Kopecky, K.J. and Liang, K.Y. (1991) On estimating HLA/disease association with application to a study of aplastic anaemia. Biometrics 47, 53-61.

Serjeantson, SW., Gao, X., Hawkins, B.R., Higgins, D.A. and Yu, Y.L. (1992) Novel HLA-DR2-related haplotypes in Hong Kong Chinese implicate the DQB1.0602 allele in susceptibility to multiple sclerosis. Eur. J. Immunogenet. 19, 11-19. Southern, E.M. (1975) Detection of specific sequences among DNA fragments separated by gel electrophoresis. J. Mol. Biol. 98, 503-517. Spurkland, A., Ronningen, K., Vandvik, B., Thorsby, E. and Vartdal, F. (1991) HLA-DQAl and HLA-DQBl genes may jointly determine susceptibility to develop multiple sclerosis. Hum. Immunol. 30, 69-75. Stewart, G.J., McLeod, J.G., Basten, A. and Bashir, H.V. (1981) HLA family studies and multiple sclerosis: A common gene, dominantly expressed. Hum. Immunol. 3, 13-29. Swingler, R.J., Kirk, P.F., Darke, C. and Compston, D.A.S. (1987) HLA and multiple sclerosis in south east Wales. J. Neurol. Neurosurg. Psychiatry 50, 1153-1155. Takacs, K., Kalman, B., Gyodi, E., Tauszik, T., Palffy, G., Kuntar, L., Guseo, A., Nagy, C. and Petranyi, G. (19901 Association between the lack of HLA-DQw6 and the low incidence of multiple sclerosis in Hungarian Gypsies. Immunogenetics 31, 383-385. Thomson, G. and Bodmer, W. (19771 The genetic analysis of HLA and disease associations In: Dausset, J. and Svejgaard, A. (Eds.1, HLA and Disease. Munksgaard, Copenhagen, pp. 84-93. Tienari, P.J., Wikstrom, J., Sajantila, A., Palo, J. and Peltonen, L. (1992) Genetic susceptibility to multiple sclerosis linked to myelin basic protein gene. Lancet 340, 987-991. Tienari, P.J., Wikstrom, J., Koskimies, S., Partanen, J., Palo, J. and Peltonen, L. (1993) Reappraisal of HLA in multiple sclerosis: close linkage in multiplex families. Eur. J. Hum. Genet. 1, 257268. Tienari, P.J., Terwilliger, J.D., Ott, J., Palo, J. and Peltonen, L. (1994) Two-linkage analysis in multiple sclerosis (MS). Genomics 19, 320-325. Tiwari, J.L. and Terasaki, P.I. (1985) Multiple sclerosis. In: HLA and Disease Associations. Springer Verlag, New York, NY, pp. 85100. Vartdal, F., Sollid, L.M., Vandvik, B., Markussen, G. and Thorsby, E. (1989) Patients with multiple sclerosis carry DQBl genes which encode shared polymorphic amino acid sequences. Hum. Immunol. 25, 103-110. Weinshenker, B.G., Bulman, D., Carriere, W., Baskerville, J. and Ebers, G.C. (1990) A comparison of sporadic and familial multiple sclerosis. Neurology 40, 1354-1358. Wood, N.W., Robertson, N., Holmans, P., Clayton, D. and Compston, D.A.S. (19941 No linkage between multiple sclerosis and the MBP gene. J. Neurol. Neurosurg. Psychiatry 57, 1191-1194. Wood, N.W., Sawcer, S.T., Kellar-Wood, H.F., Holmans, P., Clayton, D., Robertson, N. and Compston, D.A.S. (1995a) The receptor beta locus and susceptibility to multiple sclerosis. J. Neurol., in press. Wood, N.W., Sawcer, ST., Kellar-Wood, H.F., Holmans, P., Clayton, D., Robertson, N. and Compston, D.A.S. (1995b) A susceptibility gene for multiple sclerosis linked to the immunoglobulin heavy chain variable region. Neurology, in press.