Major histocompatibility complex class II alleles and the course and outcome of MS

The temporal profile and the rate of accumulation of disability from MS vary considerably between individuals.¹ Genetic factors determine the outcome as well as susceptibility to spontaneous autoimmune disease in nonhuman mammals, at least in part.^2-5 Others have made similar observations in human autoimmune disease. For example, the severity of rheumatoid arthritis and propensity to extra-articular disease is greater in individuals homozygous for the DR4 allele.⁶

In population studies, various major histocompatibility complex (MHC) alleles, particularly class II alleles, are associated with susceptibility to MS.⁷ Recently, two independent genomic mapping studies^8,9 have reported linkage of the MHC locus to MS in multiplex families, although a third study¹⁰ did not show convincing linkage to MHC. The pivotal function of class II MHC molecules in antigen presentation to CD4-positive T cells has led others to examine the putative association between the MHC and the variation in the course and outcome of MS. Other investigators have reported associations between the temporal course and the severity of MS and MHC class II alleles, but few studies have been conducted in a population-based, completely ascertained sample.¹

The population of persons with MS in Olmsted County, Minnesota, has been tracked over nine decades, principally through periodic prevalence studies. A recent cross-sectional analysis of disability¹¹ provided an excellent source of clinical information about the temporal course and disability profile of MS in this community. We report our findings, which examine the putative association between course and outcome of MS and class II MHC DR and DQ alleles.

Methods. Patient ascertainment. Patients were ascertained from a geographically based population of patients with MS in Olmsted County, Minnesota. The overall sex-and age-adjusted prevalence of MS in Olmsted County is 167.5 per 100,000 (adjusted to the 1950 US white population). The 119 patients evaluated constitute 73.4% of the population of persons with clinically definite MS evaluated for a cross-sectional analysis of disability in 1991.¹¹ The duration of MS was 17.9± 11.7 years (mean ± SD). We attempted to collect blood from all patients, including institutionalized and severely affected patients. All patients were white, and most were of mixed northern European ancestry, most commonly Norwegian, German, Danish, and Swedish. Ten percent reported part east European ancestry. Two patients were of part Native American ancestry; one was of part and another of entirely French Canadian ancestry. One was of Ashkenazi Jewish ancestry. Ancestry was unknown for six percent of patients. Controls were derived from 100 consecutive normal blood donors in Olmsted County. We analyzed the differences in demographic features and outcome variables between patients genotyped and those not typed in this study.

Classification of MS and grading of disability. Contemporary thought suggests that primary progressive (PP) MS is unique and differs from relapsing-remitting (RR) and secondary progressive (SP) MS, both of which appear to be part of a continuum; most patients with RR-MS develop SP-MS.¹ Accordingly, we divided patients into two groups, PP or "bout onset" (BO); the latter category includes patients with RR- and SP-MS.

Expanded disability status scale (EDSS) and duration of disease were used to derive the progression index (EDSS/duration of disease from onset). Severity of disability indexed to duration by the progression index was ranked from 1 to 10, as described previously¹²; briefly, patients with the lowest score had the highest EDSS scores relative to duration. Only patients with duration of MS ≥5 years were ranked because the progression index is less meaningful in patients with disease of shorter duration. Patients with similar duration of MS, within 5 years, were ranked separately according to progression index to minimize the effect of variation in the denominator.

MHC class II typing. DNA was isolated from buffy coat cells by the method of Miller et al.¹³ The first 81 patients were typed by restriction fragment length polymorphism (RFLP), and the remaining 38 patients were typed by PCR using sequence-specific oligonucleotide primers. RFLP was performed by 3-hour digestion of 15µg of genomic DNA with Taq I restriction endonuclease followed by electrophoresis on a 0.7% agarose gel and Southern blotting to nylon membrane. The membranes were hybridized overnight with biotinylated cDNA probes specific for DQB1 and DRB1 genes (Oncor, Gaithersburg, MD) and visualized with streptavidin-biotin-alkaline phosphatase-nitroblue tetrazolium color reaction. The RFLP patterns were interpreted by one of us(P.J.S.), as described by Noreen et al.¹⁴ Most but not all genotypes were adequately discerned by this method; for example, DR13-DQ6 is difficult to separate from DR13-DQ7 haplotypes (both are DR6). DQ8 and DQ9 cannot be distinguished (both are DQ3).

Sequence-specific primers for known DRB1, DRB3, DRB4, DRB5, and DQB1 alleles were used for PCR typing reactions.¹⁵

Assignment of some DQ antigens and assignment of haplotypes was based on established patterns of linkage disequilibrium. For some analyses, patients were grouped according to serologic classifications, in part because numbers of patients in certain allelic subclassifications by DNA-based methods did not yield sufficient numbers of patients for analysis. For example, patients with DR15 and DR16 were classified as DR2 because very few patients who are DR2 positive were DR16 positive. Similarly, patients with DR17 or DR18 were classified as DR3 for some analyses. Also, this type of grouping seemed appropriate considering the difficulty in unambiguous separation of some patterns by RFLP as described above.

We separately analyzed patients who were homozygous for given DRB haplotypes as well as specific heterozygotes where sufficient numbers of patients were available for analysis.

DQ alleles were grouped as follows: DQ1 = DQ5 and 6; DQ2 = DQ2; DQ3 = DQ7, 8, and 9; DQ4 = DQ4, in large part because of limitations in the definitive separation of genotypes, as discussed above. Because DQ1 subspecificities could be distinguished, and because significant associations with these subspecificities were observed, we analyzed DQ5 and DQ6 separately.

Statistical methods. We analyzed the association of each haplotype and each DR and DQ allele, the frequency of which was at least 3% in the population with MS or 5% in the control population. We also analyzed the association of genotypes (e.g., DR2 homozygote, DR2-DR3 heterozygote) to determine whether an interaction between alleles or a "dose effect" was evident. We compared the association of discrete variables (MS versus control; BO versus PP) with chi-square test for 2 × 2 tables or Fisher's exact test when the expected counts were small.

The odds ratios (ORs) for alleles or haplotypes and their 95% CIs were calculated for MS versus controls and for PP- versus BO-MS. The CIs provide an interval of ORs that are consistent with our observations and, hence, give an impression of the magnitude of ORs that we can exclude.

Within each category (e.g., haplotype, genotype), a Bonferonni correction to the p value was performed for multiple comparisons, and those p values retaining statistical significance after this correction are so indicated. Findings were judged significant if a two-sided p value was less than 0.05.

Because a positive association was found when comparing MS patients versus controls, the relative predispositional effects (RPEs) method¹⁶ was used to test other haplotype frequencies in sequence. When the overall chi-square statistic for the comparison of the distribution of MHC haplotypes (each person contributing two haplotypes) was significant, the haplotype with the largest contribution was removed, and the comparison was repeated after normalizing the expected frequency distribution of the remaining haplotypes. Haplotypes showing the strongest positive or negative association were sequentially removed until the overall comparison of the remaining haplotypes was not statistically significant.

For ordinal variables, such as ranked severity score, differences in the mean score were assessed using Wilcoxon's rank sum test. The distribution of ranked severity scores was inspected to determine if differences in the distribution were not revealed by differences in the means. Furthermore, differences in the mean slope of change of EDSS as a function of duration, assuming EDSS = 0 at disease onset, were assessed by linear regression to minimize any potential loss of information by the ranked severity score.

Results. Analysis of patients typed and not typed. The patients genotyped in this study did not differ significantly from those assessed in the cross-sectional disability study¹¹ but not genotyped for the following variables: sex (p = 0.76), age at diagnosis (p = 0.08), EDSS at time of cross-sectional prevalence study (p = 0.61), ranked severity score (p = 0.79). The patients who were genotyped had a slightly earlier age at onset (28 versus 30.5 years, p = 0.04).

Distribution of disability. The frequency distribution of ranked severity scores approximates a normal distribution with a slight skew to the right (data not shown). The skew would be expected by the existence of a subset with "benign" MS.¹

Distribution of MHC class II alleles and susceptibility to MS. The ORs for haplotypes and the more common genotypes of MHC class II alleles in MS patients and controls are shown in the table. (The raw frequencies and p values for all DR and DQ allotypes, haplotypes, and genotypes have been filed with the National Auxiliary Publications Service [NAPS]. See Note at end of text.) Haplotypes and genotypes are listed in descending order of frequency in the MS population.

The DR15-DQ6 haplotype, which constitutes 71 of 75 (95%) DR2-positive patients, was associated with MS (OR = 3.3; 95% CI = 9.1 to 5.8;p = 0.00002). Homozygosity for DR2 was not significantly more frequent in persons with MS than it was in the general population (OR = 2.6; 95% CI = 0.7 to 10.1; p = 0.23). However, there was a significant association with the DR2-DR3 genotype, which was present in 11.8% of MS patients compared with 1.0% of controls (OR = 13.2; 95% CI = 1.7 to 102.3; p = 0.002).

MS was significantly associated with DR13-DQ7 (OR = 11.1; 95% CI = 1.4 to 87.0; p = 0.004) and negatively associated with DR1-DQ5 (OR = 0.3; 95% CI = 0.1 to 0.7; p = 0.003). DQ6 was also associated with MS (p = 0.009), but this association was explained by the association with the DR15-DQ6 haplotype. DQ5 was negatively associated with MS(p = 0.008), but this association was explained by the negative association with the DR1-DQ5 haplotype (p = 0.003).

Distribution of MHC class II alleles and temporal course of MS. The ORs for having PP-MS versus BO-MS according to MHC class II haplotypes and genotypes are shown in the table. (Complete data on the frequency of allotypes, haplotypes, and genotypes in patients with PP-MS versus BO-MS have been filed with NAPS. See Note at end of text.) None of the haplotypes were significantly associated with the temporal course, although the OR for the DR4-DQ8 haplotype was 3.1 (95% CI = 0.9 to 10.8) for PP-versus BO-MS (41.7% versus 18.7%), and the OR was 4.1 (95% CI = 0.9 to 18.3) for the DR1-DQ5 haplotype (25.0% versus 7.5%). The OR for the DR17-DQ2 (DR3) MS haplotype was 0.2 (95% CI = 0.03 to 1.9) for PP-MS versus BO-MS (8.3% versus 28.0%). The small number of PP patients (n =12) limits the statistical confidence of our measured associations as reflected in the width of the CIs.

Severity of MS. (Data showing the mean ranked severity scores according to carriage of haplotype, DR and DQ allotypes, and genotype for the most common allelic pairs have been filed with NAPS. See Note at end of text.) We also analyzed the differences in the rates of change of EDSS score between individuals by linear regression, assuming that the EDSS = 0 for all patients at onset of MS. No significant associations between MHC and disability indexed to disease duration were uncovered by any of the methods of data analysis.

Discussion. The class II MHC proteins, the major restriction elements for cell-mediated immunity, have a pivotal role in development of the T-cell repertoire during ontogeny and subsequently in immune recognition. Hence, MHC genes are plausible candidate genes that could influence the outcome of an immune-mediated disease such as MS. The consistent association of one MHC haplotype, DR2 or DRB5*0101-DQA1*0102-DQB1*0602, with MS⁷ and the recent demonstration of linkage in multiplex families between the MHC locus and MS susceptibility^8,9 provide further impetus to consider its potential as a prognosis-determining gene. Indeed, the MHC appears to influence not only susceptibility to other immune-mediated diseases such as rheumatoid arthritis but also several measures of disease severity, including severity of erosive articular disease and the propensity to extra-articular involvement, which are generally considered to be markers of more severe disease.⁶

Numerous investigators have addressed the putative association between class II MHC and the course of MS since 1973. The conclusions are diverse and have been reviewed and summarized.¹ The conclusions are sufficiently conflicting to preclude any consensus based on the available literature. The most common MHC haplotypes are those associated with the DR2, DR3, and DR4 alleles. DR2 has been associated with severe outcome in four studies,^17-20 a remitting course and benign course in one study,²¹ and has not been found to be associated with either course or severity in four other studies.^22-25 Similarly, DR3 has been associated with a favorable course in two studies,^19,20 a poor outcome in one study,²⁴ and BO, as opposed to PP, course in one study.²⁶

One recent study by Runmarker et al.²⁴ is unique in that the ascertainment was population-based and relatively complete, and the outcome information was based on longitudinal disability data spanning 25 years. They found no association between DR2 and the course or outcome of MS. DR17-DQ2 (DR3) and DR1-DQ5 haplotypes were associated with a poor outcome, as defined by the rates at which EDSS score 6 was reached using survival analysis.

DR2 was significantly more common in patients with MS compared with a control group from Olmsted County. Others have also reported an association with the DR15-DQ6 (DR2) haplotype in patients with MS of northern European ancestry.⁷ Additionally, the DR13-DQ7 (DR6) haplotype was associated with MS, being present in 12 of 119 MS patients (10.1%) and 1 of 100 controls (1.0%). This association, which remained significant in the RPEs analysis and after correction for multiple comparisons, has not been observed by others. DR1-DQ5 was less common in the MS patients; it was present in 11 of 119 patients with MS (9.2%) and 25 of 100 controls (25%). Runmarker et al.²⁴ reported that DR1-DQ5 was present in 7% of 121 patients with MS and 18% of controls; the difference was significant, but not after correction for multiple comparisons. In our study, the positive associations with DR15-DQ6 and DR13-DQ7 were significant using the RPEs method for analyzing significance of secondary MHC associations. The DR1-DQ5 association was not significant by this method.

MHC class II haplotype was not associated with the temporal course or severity of MS. However, DR4-DQ8 (DR4) was present in 5 of 12 patients with PP-MS (41.7%) compared with 20 of 107 patients (18.7%) with BO-MS (OR = 3.1, 95% CI = 0.9 to 10.8). DR17-DQ2 (DR3) was present in 30 of 107 patients(28.0%) with BO-MS compared with 1 of 12 patients (8.3%) with PP-MS (OR = 0.2, 95% CI = 0.03 to 1.9). The small number of cases of PP-MS significantly limit the power of this analysis, as indicated by the broad 95% CI.

Olerup et al.²⁶ previously reported an association between DR3 and RR course and between DR4 and PP course. They initially reported an association with the DQB1 restriction fragment pattern seen in DR4-DQ8, DR7-DQ9, and DR8-DQ4 haplotypes. Subsequently, in a different data set, they confirmed an association between PP-MS and DR4-DQ8, which was present in 39% of 36 patients with PP-MS compared with 18% of 143 patients with BO-MS.⁷ Furthermore, they found that BO-MS was associated with DR17-DQ2 (DR3); this haplotype was found in 27% of 143 patients with BO-MS compared with 6% of 36 patients with PP-MS.⁷ Our results are similar to those of Olerup and Hillert.⁷ However, others have not confirmed these associations. Francis et al.²⁵ reported that 7 of 19 patients (37%) with PP-MS were DR4 positive compared with 15 of 52 (29%) with BO-MS. Runmarker et al.²⁴ reported that 28% of 14 patients with PP-MS were DR4 positive compared with 32% of 102 patients with RR-MS.

Three of 12 patients (25%) in our study with PP-MS had the DR1-DQ5 haplotype compared with 8 of 107 BO-MS patients (7.5%) (OR = 4.1, 95% CI = 0.9 to 18.3). Francis et al.²⁵ found only 1 of 71 patients with MS (1.4%) had the DR1-DQ5 haplotype. Runmarker et al.²⁴ reported that DR1-DQ5 was present in 7% of 14 patients with PP-MS and 7% of 102 patients with BO-MS. However, they did report an association of DR1-DQ5 with a severe course of MS based on a longitudinal analysis of disability, an association that we did not find based on our disability analysis.

Aside from immunogenetic background, PP-MS may differ in other ways from RR- or SP-MS. PP-MS has been associated with relatively fewer lesions on MRI of the brain²⁷ and relatively less inflammation on postmortem examinations.²⁸

Similarly, other subtypes of idiopathic inflammatory demyelinating disease may have distinctive clinical manifestations and immunogenetic differences compared with MS. Devic's disease (neuromyelitis optica) is associated with recurrent severe relapses of inflammatory demyelinating disease, usually confined to the spinal cord and optic nerves, often associated with clinical or immunologic evidence of B-cell autoimmunity.^29,30 Kira et al.³¹ have recently shown that DR2, which is present in 41.2% of Japanese patients with "Western-type MS" and 14.2% of healthy Japanese controls, is not present in any patients with neuromyelitis optica ("Asian-type MS").

For some autoimmune diseases, the primary MHC susceptibility allele may be DQ rather than DR. For example, in collagen-induced arthritis in recombinant mice, it has recently been shown that the H-2E (DR-equivalent) genes are protective in the context of H-2A (DQ-equivalent) susceptibility genes.³² DQ8 transgenic mice in MHC class II-deficient animals are highly susceptible to collagen-induced arthritis.³³ Linkage studies may fail to uncover this relation because protection by some DR molecules is dominant over DQ-mediated susceptibility. Although one study by Francis et al.³⁴ suggested that the primary association was with the DQ1 gene, rather than DR2, subsequent studies by the same investigators²⁵ and by others⁷ did not confirm this conclusion. It is impossible epidemiologically to study the associations of DR independent of DQ, given the degree of linkage disequilibrium. The association of DR2 with MS susceptibility is stronger than the DQ association in our study.

Our study has many strengths, but some limitations. The strengths are as follows:

1. Completely ascertained population, which allowed us to address the putative role of MHC across the spectrum of MS-related disability. Those genotyped did not differ from those not genotyped in terms of our outcome variables.

2. DNA-based genotyping, which allows for subclassification of serologic specificities.

3. Temporal course and disability data for each patient.

4. Broad range of disability scores.

5. Relatively long median duration of disease thereby permitting accurate classification of longterm disability.

6. Analysis of association according to zygosity, which allowed us to address the effect of dose dependence of MHC effect, and allelic interactions where sample size was sufficient.

Limitations of our study include the following:

1. Limited population size and a relatively large number of genotypes, which compromise the power of the study. The relatively small number of PP-MS versus BO-MS patients compromises the statistical analysis of significance; this has been a problem in most studies because PP-MS constitutes only 15 to 20% of MS patients. Our ability to detect an association is estimated by the reported ORs, which are acceptable for the more common genotypes. For example, we can be certain with 95% confidence that the DR15-DQ6 haplotype does not pose a risk of PP-MS greater than an OR= 2.1. We found a trend to a negative association with DR17-DQ2 (OR = 0.2, p = 0.18); the lack of statistical significance must be viewed in light of the lower limit of the confidence interval, which indicates that the OR could be as low as 0.03.

2. Reliance on the EDSS as a marker of disease severity, which may miss occasional patients whose disease course is not adequately reflected in the EDSS score. The EDSS is the current gold standard for quantitating MS-related disability and is a composite based on all major dimensions affected by the disease, even though it is not equally sensitive to all.

3. Ranking of disability based on cross-sectional as opposed to longitudinal data. The ranking scheme, although it may lose some information, is a conservative approach. The variation in the denominator (duration of disease) on the range of the progression index was limited by grouping patients into 5-year cohorts according to duration of MS. Furthermore, linear regression of EDSS as a function of duration failed to reveal any association of any MHC allotype or haplotype with disease severity. Progression of disease was assumed to progress in a linear fashion in the regression analysis.

4. Assignment of haplotypes based on linkage disequilibrium pattern rather than by direct determination of haplotype by typing other relatives and segregation analysis. Linkage dysequilibrium is a standard and accepted method for assigning haplotypes with MHC studies given the strength of linkage disequilibrium in the DR, DQ region, but misclassification of haplotypes could mask an association.

5. Inability to dissect possible independent or interacting immunologic effects of DR and DQ.

Our study does not demonstrate a clear association between MHC and the course and severity of MS. Given the trend toward an association of PP-MS with the DR4-DQ8 (DR4) haplotype and BO-MS with the DR17-DQ2 (DR3) haplotype, which was also reported by another group,^7,26 additional studies of this putative association may be valuable. However, we found no other association between MS course and severity and class II MHC alleles. Our results and the largely negative and inconsistent results among the different studies may result from the limited power of the studies and from ethnic or other regional variations in MS; however, they likely indicate that the MHC class II genes at most modulate the course of MS or, in other words, account for only a small part of the variance. Transgenic mouse models in MHC knockout mice may clarify associations not evident from epidemiologic studies due to interaction of MHC influences (e.g., DR and DQ).

Acknowledgments

We gratefully acknowledge Dr. A. Siva for participating in the cross-sectional assessment of disability. Deborah Falbo and Steven DeGoey were instrumental in the DNA extraction and genotyping. Theresa Hanson typed the manuscript.

Note. Readers can obtain 6 pages of supplementary material from the National Auxiliary Publications Service, 248 Hempstead Turnpike, West Hempstead, NY 11552. Request document no. 05454. Remit with your order, not under separate cover, in US funds only, $15.00 for photocopies or $5.00 for microfiche. Outside the United States and Canada, add postage of $4.50 for the first 20 pages and $1.00 for each 10 pages of material thereafter, or$1.75 for the first microfiche and $1.00 for each fiche thereafter. There is a $25.00 invoicing charge on all orders filled before payment.