The major histocompatibility complex (MHC) HLA region on chromosome 6p21 contains the major locus of type 1 diabetes (IDDM1). Common allelic variants at the class II HLA-DRB1, -DQA1, and -DQB1 loci account for the major part of IDDM1. Previous studies suggested that other MHC loci are likely to contribute to IDDM1, but determination of their relative contributions and identities is difficult because of strong linkage disequilibrium between MHC loci. One prime candidate is the polymorphic HLA-DPB1 locus, which (with the DPA1 locus) encodes the third class II antigen–presenting molecule. However, the results obtained in previous studies appear to be contradictory. Therefore, we have analyzed 408 white European families (200 from Sardinia and 208 from the U.K.) using a combination of association tests designed to directly compare the effect of DPB1 variation on the relative predisposition of DR-DQ haplotypes, taking into account linkage disequilibrium between DPB1 and the DRB1, DQA1, and DQB1 loci. In these populations, the overall contribution of DPB1 to IDDM1 is small. The main component of the DPB1 contribution to IDDM1 in these populations appears to be the protection associated with DPB1*0402 on DR4-negative haplotypes. We suggest that the HLA-DP molecule itself contributes to IDDM1.

A positive association of the DPB1*0301 allele with type 1 diabetes has previously been reported in a study analyzing 42 Mexican-American type 1 diabetic families and ethnically matched control subjects. Analysis of the linkage disequilibrium patterns in Mexican-Americans indicated that this association was not explained by the linkage disequilibrium of DPB1*0301 with high-risk DR-DQ haplotypes (1). In a subsequent study of 180 white European-derived nuclear families largely from the U.S., Noble et al. (2) found that after stratifying a number of DR-DQ haplotypes according to DPB1 type, the DPB1*0301 allele was more frequent in type 1 diabetic patients than in affected family-based control subjects (AFBACs). Another allele, DPB1*0402, was decreased in the DR3 haplotypes of the type 1 diabetic patients compared with those of the family-based control subjects (2). Importantly, the frequency of DPB1*0402 was also decreased in the Mexican-American families (1). These positive results were not replicated in a case-control study of Norwegians. Comparing the frequencies of DPB1 alleles in 237 patients and 287 control subjects matched for the same high-risk DR3/DR4 and DR4/DR4 genotypes, no significant independent association of DPB1 alleles was found (3). Most recently, Noble et al. (4) extended their initial observations by analyzing an additional 89 type 1 diabetic families (total n = 269 families). Their analyses suggested that DPB1*0301 and DPB1*0202 appeared to be primarily predisposing, whereas DPB1*0402 and DPB1*0401 showed possible protective effects. They suggested that DPB1 might primarily contribute susceptibility to, rather than protection from, type 1 diabetes.

In the present study, we analyzed the association of the HLA-DPB1 alleles in a collection of 408 type 1 diabetic white European families, of which 200 were from Sardinia and 208 were from the U.K. These families did not overlap with those considered in previous studies (1,2,3,4). First, we compared the overall association of the DPB1 locus with the DRB1 and DQB1 loci in these two sets of families in a single-point analysis using the extended transmission/disequilibrium test (ETDT) (5). Without taking into account linkage disequilibrium between the three class II loci, the overall association of HLA-DPB1 with type 1 diabetes in these families was strong (P = 5.3 × 1010). However, this association did not approach the level of significance of the association of HLA-DRB1 and -DQB1 (P = 2.1 × 1090 and 8.5 × 1083, respectively). These associations did not vary significantly between the two populations (data not shown). However, the vast majority of the overall DPB1 association was due to linkage disequilibrium with the DQB1 and DRB1 loci. Using a modified version of the ETDT—the conditional ETDT (6)—we found that after taking into account linkage disequilibrium with the HLA-DQB1 and –DRB1 loci, the association of DPB1 was only weakly significant (P = 1.0 × 102 and 1.9 × 102, respectively). Conversely, after taking DPB1 into account, the overall associations of DRB1 and DQB1 yielded P values of 7.9 × 1041 and 1.5 × 1039, respectively. Although these results suggest that DPB1 might contribute to the association of the HLA region to type 1 diabetes, they also indicate that its overall genetic effect is considerably smaller than that of DRB1 and DQB1 in these two populations.

Next, we considered the single-point association of the different DPB1 alleles using the transmission/disequilibrium test (TDT) (7) (Table 1). The most significant results were the increased transmission frequencies of the putative predisposing DPB1*0202 and DPB1*0301 alleles (P = 4 × 104 and 1.1 × 106, respectively) and the decreased transmission of the putative protective allele DPB1*0402 (P = 1 × 107). Based on these results and on previous findings (1,2,4), we evaluated the relative transmission or predisposition of the DPB1*0301, DPB1*0202, and DPB1*0402 alleles according to which DRB1-DQA1-DQB1 haplotypes they were on. To carry out these analyses, we used the haplotype method (HM) (8), which we modified by incorporating TDT into the test (HM-TDT). This modified test evaluates the association of specific alleles of DPB1, taking into account linkage disequilibrium with alleles of the DRB1 and DQB1 loci as well as the association of an allele of any locus in linkage disequilibrium with alleles at another locus (see research design and methods). DPB1*0301 was significantly and consistently more frequently transmitted than DPB1*0402 on DR3 (DRB1*0301-DQA1*0501-DQB1*0201) (P = 2.3 × 102) (Table 2) and DR1 (DRB1*01-DQA1*0101-DQB1*0501) haplotypes (P = 2.1 × 102) (Table 3). A similar finding, albeit not significant at the 5% level, was found for DR16 (DRB1*1601-DQA1*0102-DQB1*0502) haplotypes (P = 0.13) (Table 4). For example, the DPB1*0301 allele is associated with a 5.4-fold greater disease risk than the DPB1*0402 allele on DR3 haplotypes, as estimated using the odds ratio for transmission (ORT) (Table 2). Also, DPB1*0202 was significantly more frequently transmitted than DPB1*0402 on DR3 haplotypes (P = 2.0 × 103) (Table 2). In addition, DPB1*0202 was also significantly more frequently transmitted on DR3 haplotypes than were DPB1*0401 and DPB1*0201 (P = 3.1 × 102 and 4.4 × 102, respectively) (data not shown), two other DPB1 alleles. The heterogeneity in transmission of DPB1*0401 and DPB1*0201 was not observed on other DR-DQ haplotypes (data not shown). No significant heterogeneity was detected at the DPB1 locus between the transmitted and nontransmitted chromosomes of the DR4 haplotypes (data not shown). There was consistency in the trends shown by the DPB1*0301 and DPB1*0402 alleles between the U.K. and Sardinian populations. The DPB1*0202 allele was virtually absent in the Sardinian families, and thus the putative permissive effect in disease susceptibility of this allele on DR3 haplotypes could not be evaluated in this population.

The relative contributions of DPB1 alleles to IDDM1 were delineated by studying an even larger data set—which included 176 U.S. families that have been previously studied for DP (2) (total n = 582)—and by determining the effect of each DPB1 allele on the relative association of DQB1- DQA1-DRB1 haplotypes (see research design and methods) (Table 5). Overall, the most important component of the DPB1 association with type 1 diabetes in this mixed sample set seemed to be the “protective” effect of the DPB1*0402 allele. For instance, the ORT drops from 1.2 for the DR1 haplotype with DPB1*0301 to 0.1 for DR1 with DPB1*0402 (P = 5.1 × 103 in a pairwise comparison of the two haplotypes) (Table 5). When the DR3 haplotype with DPB1*0301 was compared with the DR3 haplotype with DPB1*0402, the ORT values were 5.8 and 1.5, respectively (P = 1.4 × 102 in a pairwise comparison of the two haplotypes) (Table 5). The negative association of DPB1*0402 was further illustrated by evaluating the net effect of allelic variation at DPB1 on the transmissions of DRB1-DQA1-DQB1 haplotypes to affected children (Fig. 1). Most strikingly, DPB1*0402-positivity converted neutral DR-DQ haplotypes, such as DR1 and DR16, into protective haplotypes. The percentage transmission of the grouped DR1 + DR16 haplotypes was 36.6% when DPB1 alleles were not taken into account; this decreased to 6.1% when DPB1*0402 was considered. Only a small increase toward a positive association of DR1 + DR16 was observed in the presence of DPB1*0301 (43.1% transmission to affected children). The protection associated with DPB1*0402 was even able to reduce, but not to completely override, the predisposition conferred by DR3 haplotypes; in the presence of DPB1*0402, the DR3 haplotypes had a neutral association (47.4% transmission to affected children) instead of the highly positive type 1 diabetes association (78.8%) normally seen. In contrast, the inclusion of DPB1*0301 only marginally increased the transmission of DR3 haplotypes (80.9%) (Fig. 1).

As shown in Table 5, the DPB1*0202-DR3 haplotype may be more predisposing than the DPB1*0301-DR3 haplotype (ORT = 31.3 and 5.8, respectively), but the difference in transmission between the two DR3 haplotypes was not significant (P = 0.14 in a pairwise comparison of the two haplotypes). Furthermore, in northern European populations, DPB1*0202 very frequently occurs on the extended and predisposing DR3-B18 haplotype. This makes it difficult to define its individual predisposing effect within this extended haplotype. Further studies from other populations in which DPB1*0202 is included within different extended haplotypes might clarify whether DPB1*0202 is independently predisposing to type 1 diabetes.

Finally, allelic variation at the DPB1 locus did not affect the transmission of the highly predisposing DRB1*0405/*0401-DQA1*0301-DQB1*0302 haplotypes (Table 5). Because Lie et al. (3) only analyzed DR4-positive individuals, this is likely to be the explanation for their failure to detect an effect of DP in their Norwegian data set. Contrast these data with the protection against type 1 diabetes provided by the DR15 (DRB1*1501-DQA1*0102-DQB1*0602), DR14 (DRB1*1401-DQA1*0101-DQB1*0503), or DR7 (DRB1*0701-DQA1*0201-DQB1*0303) haplotypes, which are independent of allelic variation at DPB1 (F.C. and J.A.T., unpublished data).

Taken together, our results provide consistent and significant evidence that haplotypes that are identical at the DRB1, DQA1, and DQB1 loci but different at the DPB1 locus have different associations with type 1 diabetes. These conclusions are in agreement with those of Erlich and colleagues (1,2,4). We cannot conclude, however, that the DPB1 effects described here or elsewhere are directly attributable to polymorphisms in the DPB1 locus itself. Evidence that DR and DQ are primary etiological determinants of IDDM1 and not just in linkage disequilibrium with another locus includes the correlation of polymorphic amino acids in the peptide-binding active site of the molecules with susceptibility and resistance to disease (9,10) as well as data from biochemical (11), structural (12), transgenic (13), and mechanistic studies (14). Hence, we compared the exon 2–encoded amino acid sequences of the positively associated DPB1*0301 and *0202 alleles with that of the protective DPB1*0402 allele. No simple amino acid residue disease-risk correlation was evident (data not shown). The DPB1 effect might result from a complex interaction owing to the joint action of multiple residues at different peptide-binding pockets, including P9, P4, and P1. Alternatively, the association of DPB1 alleles may not be caused by residue variation in the second exon of this locus at all but instead may result from other non–DP-DR-DQ polymorphism(s) in strong linkage disequilibrium with it. However, given its function in antigen presentation and its homology to DR and DQ, we favor a model in which DPB1 contributes in a primary way to type 1 diabetes predisposition/resistance. That the association of specific DPB1 alleles was consistently observed on different DRB1-DQA1-DQB1 haplotypes and even in distantly related populations is consistent with a primary role for the products of the DPB1 locus. Nevertheless, the contribution of DPB1 to IDDM1 is small because DPB1*0402-positive DR3, DR1, and DR16 haplotypes have relatively low frequencies in these populations (4% in the total data set, according to the AFBAC frequencies of the different haplotypes). However, the DPB1 locus could have a larger effect in populations in which these DPB1*0402-positive haplotypes are more frequent.

The data set consisted of 200 Sardinian, 208 U.K., and 176 U.S. families (total affected children = 212 from Sardinia, 412 from the U.K., and 352 from the U.S.). The average age (means ± SD) at disease onset was 10.7 ± 7.2 years (females 10.1 ± 6.8, males 12.5 ± 8.6) in the U.K., 11.5 ± 7.9 years (females 10.2 ± 7.1, males 11.1 ± 7.3) in the U.S., and 8.4 ± 4.7 years (females 8.0 ± 4.0, males 8.6 ± 5.1) in the Sardinian patients. The U.K. families were part of the British Diabetic Association Warren Repository (15). The 176 U.S. multiplex families were from the Human Biological Data Interchange (16). The 200 Sardinian families were typed using polymerase chain reaction (PCR) amplification of the polymorphic second exon of the HLA-DRB1, -DQB1, and -DPB1 genes and dot blot analysis of amplified DNA with sequence-specific oligonucleotide (2). The 208 U.K. families were typed for the HLA-DRB1 and -DQB1 loci using a combination of serological and PCR–sequence-specific primer (SSP) methods by the Transplant Unit in Oxford, U.K. (16). The HLA-DPB1 locus was typed in these families using a PCR-based dot blot assay and PCR-SSP. Alleles at the DQA1 locus in the Sardinian and U.K. sample sets were inferred based on their known patterns of linkage disequilibrium with the DRB1-DQB1 haplotypes. Typing data for the U.S. families was obtained through the Human Biological Data Interchange, from which DNA samples from family members were purchased (17). The HLA data from the U.S. families reported in this article overlap with those previously reported (2).

Single-point analysis of the HLA-DRB1, -DQB1, and -DPB1 loci was performed using the ETDT (5). This test takes into account the transmission or nontransmission of alleles of a marker relative to the alleles of the marker present on the other parental chromosome. The ETDT takes multiple alleles into account and obtains a global P value indicative of the degree of significance of the association with the disease at each individual locus. To distinguish primary associations from those due to linkage disequilibrium at the established disease predisposing loci, we used a variant of the ETDT called conditional ETDT (6). This test allows us to analyze the overall effect of one locus while taking into account the association of other linked loci. The conditional ETDT compares the transmission of haplotypes constructed from all the loci against the null hypothesis that all haplotypes identical at the conditioning loci have equal transmission weights. The single-point association of individual DPB1 alleles was evaluated using the TDT (7). To study the transmission of specific DPB1 alleles conditioned on alleles or haplotypes at other loci, we used an HM-TDT (8). The HM was originally designed as a test for homogeneity of relative allele frequencies at a test locus on haplotypes identical for alleles at another locus. In this study, we applied the same concepts contained in the original description of the HM (8) to test the null hypothesis of equality of transmission of marker haplotypes identical at one variant but different at another closely linked variant. If there is heterogeneity in the transmission of two marker haplotypes that are identical at a predisposing marker (variant A) but different at a putative predisposing marker at another site (variant B), then this is evidence that variant A does not entirely explain disease predisposition and that variant B itself or another marker in linkage disequilibrium with variant B is influencing the transmission of variant A and thus the disease susceptibility. Specifically, the transmission and nontransmission counts for the two haplotypes evaluated by TDT may be arranged in a 2 × 2 contingency table and tested by Fisher’s exact test or Pearson’s χ2 test. To maintain the independence of these data, we must exclude individual parents having both of the haplotypes being considered, but the transmission data for those parents may be analyzed by standard TDT, and the statistic may be added to that of the 2 × 2 table to give an overall χ2 test with two df. The heterogeneity in transmission between the two haplotypes can be quantified by the ORT calculated from the 2 × 2 contingency table of TDT transmission counts. The transmission data from the individual parents carrying both of the haplotypes being compared have also been discarded in computing the ORTs. We applied the following formula: [(a × d)/(b × c)], where a is the number times a given haplotype is transmitted to affected children, d is the number times another haplotype that is identical to the previous haplotype at a predisposing marker (variant A) but different at the test locus (variant B) is not transmitted to affected children, b is the number times the first haplotype is not transmitted to affected children, and c is the number times the second haplotype is transmitted to affected children. When one element of this equation was 0, we used the following formula: [(2a + 1) (2 days +1)]/[(2b +1) (2c +1)].

Confidence intervals were calculated using the following formulas: variances were computed by taking the inverse of the sum of the inverses of a, b, c, and d. The standard deviation was calculated by taking the square root of the inverse of the variance. The standard deviation was then multiplied by 1.96, and this quantity was added to and subtracted from the mean to give a 95% confidence interval. We used the mathematical framework applied in the HM-TDT to rank the four locus (DRB1-DQA1-DQB1-DPB1) haplotypes around a “reference” haplotype; in this case we refer to the method as the pairwise (PW)-TDT.

Although it may be less powerful in comparison with the original HM (8), we applied the HM-TDT in this study because it has the advantage of not being sensitive to even recent population stratification and as such makes possible analysis and meta-analysis of mixed data sets. A requirement for the validity of HM-TDT is that the parental genotypes be in Hardy-Weinberg equilibrium. Measured directly by the exact test using the Markov-chain approach (18), the parental genotypes did not show any significant deviation from Hardy-Weinberg equilibrium (data not shown). To further exclude this possibility, in the Sardinian and U.K. sample sets, the DPB1 alleles were also analyzed conditional on DRB1-DQA1-DQB1 using another variant of the ETDT, the pairwise ETDT (PETDT), which is not sensitive to deviation from Hardy-Weinberg in the parental genotypes (6); the results obtained with the PETDT and the HM-TDT were fully consistent (data not shown). Both the HM-TDT and PETDT assume multiplicative allele effects for the genotype relative risks at the conditioning loci, as defined by Schaid (19), which is consistent with the genotype relative risks of the HLA region and implies that the haplotypes from both parents represent independent data points. The frequencies of alleles and haplotypes in the Sardinian, U.K., and U.S. populations were deduced from the AFBAC frequencies, calculated as described by Thomson (20). Haplotypes were established following the cosegregation of alleles within families and using computer programs written by F. Dudbridge. Only haplotypes certain from parental genotype data (and in the absence of intercrosses) were considered in the analyses shown in this report. Only the probands were evaluated in all the families with more than one affected sibling.

FIG. 1.

Effect of DPB1 alleles on the transmission of DRB1-DQA1-DQB1 haplotypes of interest in the Sardinian/U.K./U.S. data set. N, neutral; NT, not transmitted; P, protective; S, susceptible; T, transmitted.

FIG. 1.

Effect of DPB1 alleles on the transmission of DRB1-DQA1-DQB1 haplotypes of interest in the Sardinian/U.K./U.S. data set. N, neutral; NT, not transmitted; P, protective; S, susceptible; T, transmitted.

Close modal
TABLE 1

Transmission of DPB1 alleles in a combined data set of 408 type 1 diabetic families from Sardinia and the U.K.

DPB1Sardinia
U.K.
Total
Percent-age TP < 0.05
TNTTNTTNT
0202 14 15 93.8 4 × 10–4 
0301 129 53 65 54 194 107 64.5 1.1 × 10–6 
1501 61.5 — 
0601 11 11 61.1 — 
0101 40 28 44 29 60.3 — 
0501 12 10 14 11 56.0 — 
1001 10 11 19 19 50.0 — 
1101 50.0 — 
0401 65 75 105 120 170 195 46.6 — 
0201 60 89 49 42 109 131 45.4 — 
0901 10 14 12 18 40.0 — 
1301 12 12 21 36.4 — 
1401 10 28.6 — 
0402 31 21 46 24 77 23.8 1 × 10–7 
Others 10 — — 
DPB1Sardinia
U.K.
Total
Percent-age TP < 0.05
TNTTNTTNT
0202 14 15 93.8 4 × 10–4 
0301 129 53 65 54 194 107 64.5 1.1 × 10–6 
1501 61.5 — 
0601 11 11 61.1 — 
0101 40 28 44 29 60.3 — 
0501 12 10 14 11 56.0 — 
1001 10 11 19 19 50.0 — 
1101 50.0 — 
0401 65 75 105 120 170 195 46.6 — 
0201 60 89 49 42 109 131 45.4 — 
0901 10 14 12 18 40.0 — 
1301 12 12 21 36.4 — 
1401 10 28.6 — 
0402 31 21 46 24 77 23.8 1 × 10–7 
Others 10 — — 

Data are n or %. Only alleles present in at least 10 heterozygous parents in the total data set were included. T, transmitted; NT, not transmitted.

TABLE 2

The relative predisposition of DPB1 alleles on DR3 (DRB1*301-DQA1*0501-DQB1*0201) haplotypes in 408 Sardinian and U.K. families

DRB1DQA1DQB1DPB1Sardinia
U.K.
Total
ORT95% CIP HM-TDT
TNTTNTTNT
0301 0501 0201 0202 13 14 16.9 (1.8–158.1) 2.0 × 10−3 
0301 0501 0201 0301 105 21 20 125 28 5.4 (1.7–17.2) 2.3 × 10–2 
0301 0501 0201 0402 — — 
DRB1DQA1DQB1DPB1Sardinia
U.K.
Total
ORT95% CIP HM-TDT
TNTTNTTNT
0301 0501 0201 0202 13 14 16.9 (1.8–158.1) 2.0 × 10−3 
0301 0501 0201 0301 105 21 20 125 28 5.4 (1.7–17.2) 2.3 × 10–2 
0301 0501 0201 0402 — — 

Data are n. The HM-TDT P values as well as the ORT were calculated with pairwise comparisons using BRB1*0301-DQA1*0501-DQB1*0201-DPB1*0402 as the reference haplotype. T, transmitted; NT, not transmitted.

TABLE 3

The relative predisposition of DPB1 alleles on DR1 (DRB1*01-DQA1*0101-DQB1*0501) haplotypes in 408 Sardinian and U.K. families

DRB1DQA1DQB1DPB1Sardinia
U.K.
Total
ORT95% CIP HM-TDT
TNTTNTTNT
01 0101 0501 0301 14 10.9 (1.2–97.1) 2.1 × 10–2 
01 0101 0501 0402 17 19 — — 
DRB1DQA1DQB1DPB1Sardinia
U.K.
Total
ORT95% CIP HM-TDT
TNTTNTTNT
01 0101 0501 0301 14 10.9 (1.2–97.1) 2.1 × 10–2 
01 0101 0501 0402 17 19 — — 

Data are n. The HM-TDT P values as well as the ORT were calculated with pairwise comparisons using DRB1*01-DQA1*0101-DQB1*0501-DPB1*0402 as the reference haplotype. T, transmitted; NT, not transmitted.

TABLE 4

The relative predisposition of DBP1 alleles on DR16 (DRB1*1601-DQA1*0102-DQB1*0502) haplotypes in the Sardinian and U.K. family sets

DRB1DQA1DQB1DPB1Sardinia
U.K.
Total
ORT95% C.I.P HM-TDT
TNTTNTTNT
1601 0102 0502 0301 11 16 11 16 6.4 (0.7–57.0) 0.13 
1601 0102 0502 0402 — — 
DRB1DQA1DQB1DPB1Sardinia
U.K.
Total
ORT95% C.I.P HM-TDT
TNTTNTTNT
1601 0102 0502 0301 11 16 11 16 6.4 (0.7–57.0) 0.13 
1601 0102 0502 0402 — — 

Data are n. The HM-TDT P values as well as the ORT were calculated with pairwise comparisons using the DRB1*1601-DQA1*0102-DQB1*0502-DPB1*0402 haplotype as the reference. T, transmitted; NT, not transmitted.

TABLE 5

The relative transmission of DRB1-DQA1-DQB1-DPB1 haplotypes in a combined data set of 584 Sardinia, U.K., and U.S. families

DRB1DQA1DQB1DPBITNTPercentage TP TDT <0.05ORT95% CIP PW-TDT <0.05
0301 0501 0201 0202 22 95.7 1.2 × 10–5 31.3 (4.0–245.8) 3.8 × 10–5 
0405 03 0302 0201 18 94.7 9.6 × 10–5 29.4 (3.7–231) 3.2 × 10–5 
0401 03 0302 0301 26 89.7 1.9 × 10–5 13.3 (3.7–48.0) 8.2 × 10–5 
0401 03 0302 0402 17 89.5 5.7 × 10–4 13.9 (3.0–64.4) 1.1 × 10–4 
0401 03 0302 0201 19 82.6 1.8 × 10–3 7.8 (2.4–25.0) 8.7 × 10–4 
0401 03 0302 0401 103 25 80.5 5.4 × 10–12 6.2 (3.2–11.8) 8.1 × 10–8 
0405 03 0302 0301 33 78.6 2.1 × 10–4 6.6 (2.7–16.3) 2.4 × 10–4 
0301 0501 0201 0301 136 32 81.0 9.1 × 10–15 5.8 (3.2–10.9) 4.5 × 10–7 
0404 03 0302 0301 13 76.5 3.9 × 10–2 4.8 (1.4–16.3) 1.7 × 10–2 
0301 0501 0201 0201 53 25 67.9 1.5 × 10–3 3.8 (1.9–7.4) 3.8 × 10–3 
0301 0501 0201 0401 122 61 66.7 8.1 × 10–6 3.6 (2.0–6.4) 1.1 × 10–4 
0301 0501 0201 0101 61 28 68.5 5.4 × 10–4 3.1 (1.6–6.1) 9.9 × 10–2 
0404 03 0302 0601 60.0 — 3.0 (0.9–10) — 
0402 03 0302 0401 60.0 — 2.5 (0.8–7.6) — 
0404 03 0302 0401 13 61.9 — 2.2 (0.8–6.0) — 
0301 0501 0201 0402 10 47.4 — 1.5 (0.5–4.0) — 
01 01 0501 0301 14 17 45.2 — 1.2 (0.5–2.9) — 
01 01 0501 0201 11 14 44.0 — 1.1 (0.5–2.9) — 
1601 0102 0502 0401 10 13 43.5 — 1.1 (0.4–2.9) — 
1601 0102 0502 0301 11 16 40.7 — 1.0 (0.4–2.5) — 
01 01 0501 0401 30 49 38.0 3.2 × 10–2 1.0 — 1.0 
0701 0201 0201 0401 15 37.5 — 1.0 (0.4–2.5) — 
1601 0102 0502 0201 22 36 37.9 — 0.9 (0.5–1.9) — 
0401 03 0301 0401 23 25.8 7.0 × 10–7 0.5 (0.2–1.4) — 
0701 0201 0201 0201 13 13.3 4.5 × 10–3 0.3 (0.1–1.3) — 
11-12 0501 0301 0301 14 12.5 2.7 × 10–3 0.2 (0.0–1.1) 5.1 × 10–2 
11-12 0501 0301 0401 40 11.1 5.7 × 10–7 0.2 (0.1–0.7) 13 × 10–2 
11-12 0501 0301 0201 27 10.0 1.2 × 10–5 0.2 (0.1–0.7) 6.4 × 10–3 
01 01 0501 0402 23 8.0 2.7 × 10–5 0.1 (0.0–0.6) 4.2 × 10–3 
11-12 0501 0301 0402 29 3.3 3.2 × 10–7 0.1 (0.0–0.4) 3.7 × 10–4 
1501 0102 0602 0401 64 3.0 2.3 × 10–14 0.1 (0.0–0.2) 3.1 × 10–6 
DRB1DQA1DQB1DPBITNTPercentage TP TDT <0.05ORT95% CIP PW-TDT <0.05
0301 0501 0201 0202 22 95.7 1.2 × 10–5 31.3 (4.0–245.8) 3.8 × 10–5 
0405 03 0302 0201 18 94.7 9.6 × 10–5 29.4 (3.7–231) 3.2 × 10–5 
0401 03 0302 0301 26 89.7 1.9 × 10–5 13.3 (3.7–48.0) 8.2 × 10–5 
0401 03 0302 0402 17 89.5 5.7 × 10–4 13.9 (3.0–64.4) 1.1 × 10–4 
0401 03 0302 0201 19 82.6 1.8 × 10–3 7.8 (2.4–25.0) 8.7 × 10–4 
0401 03 0302 0401 103 25 80.5 5.4 × 10–12 6.2 (3.2–11.8) 8.1 × 10–8 
0405 03 0302 0301 33 78.6 2.1 × 10–4 6.6 (2.7–16.3) 2.4 × 10–4 
0301 0501 0201 0301 136 32 81.0 9.1 × 10–15 5.8 (3.2–10.9) 4.5 × 10–7 
0404 03 0302 0301 13 76.5 3.9 × 10–2 4.8 (1.4–16.3) 1.7 × 10–2 
0301 0501 0201 0201 53 25 67.9 1.5 × 10–3 3.8 (1.9–7.4) 3.8 × 10–3 
0301 0501 0201 0401 122 61 66.7 8.1 × 10–6 3.6 (2.0–6.4) 1.1 × 10–4 
0301 0501 0201 0101 61 28 68.5 5.4 × 10–4 3.1 (1.6–6.1) 9.9 × 10–2 
0404 03 0302 0601 60.0 — 3.0 (0.9–10) — 
0402 03 0302 0401 60.0 — 2.5 (0.8–7.6) — 
0404 03 0302 0401 13 61.9 — 2.2 (0.8–6.0) — 
0301 0501 0201 0402 10 47.4 — 1.5 (0.5–4.0) — 
01 01 0501 0301 14 17 45.2 — 1.2 (0.5–2.9) — 
01 01 0501 0201 11 14 44.0 — 1.1 (0.5–2.9) — 
1601 0102 0502 0401 10 13 43.5 — 1.1 (0.4–2.9) — 
1601 0102 0502 0301 11 16 40.7 — 1.0 (0.4–2.5) — 
01 01 0501 0401 30 49 38.0 3.2 × 10–2 1.0 — 1.0 
0701 0201 0201 0401 15 37.5 — 1.0 (0.4–2.5) — 
1601 0102 0502 0201 22 36 37.9 — 0.9 (0.5–1.9) — 
0401 03 0301 0401 23 25.8 7.0 × 10–7 0.5 (0.2–1.4) — 
0701 0201 0201 0201 13 13.3 4.5 × 10–3 0.3 (0.1–1.3) — 
11-12 0501 0301 0301 14 12.5 2.7 × 10–3 0.2 (0.0–1.1) 5.1 × 10–2 
11-12 0501 0301 0401 40 11.1 5.7 × 10–7 0.2 (0.1–0.7) 13 × 10–2 
11-12 0501 0301 0201 27 10.0 1.2 × 10–5 0.2 (0.1–0.7) 6.4 × 10–3 
01 01 0501 0402 23 8.0 2.7 × 10–5 0.1 (0.0–0.6) 4.2 × 10–3 
11-12 0501 0301 0402 29 3.3 3.2 × 10–7 0.1 (0.0–0.4) 3.7 × 10–4 
1501 0102 0602 0401 64 3.0 2.3 × 10–14 0.1 (0.0–0.2) 3.1 × 10–6 

Data are n or %. Only DRB1-DQB1-DQA1-DPB1 haplotypes detected in at least 15 informative meioses were included. The PW-TDT P values as well as the ORT values were calculated with pairwise comparisons using the DRB1*01-DQA1*0501-DQB1*-0501-DPB1*0401 haplotype as reference. T, transmitted; NT, not transmitted.

This study was supported in part by the Regione Autonoma Sardegna (L.R.11, 30–4-90), Diabetes U.K. (formerly the British Diabetic Association), the Juvenile Diabetes Foundation International, the Medical Research Council of U.K., Eli Lilly U.K., and the Wellcome Trust.

F.C. and J.A.T. are recipients of a Wellcome Trust Biomedical Research Collaboration Grant, J.A.T. was a Wellcome Trust Principal Research Fellow, and F.D. is a U.K. Medical Research Council Special Training Fellow in Bioinformatics.

We thank A. Cao, M. Silvetti, and Michael Whalen for continuous help and support; D. Clayton for statistical advice; M. Chessa, P. Frongia, and R. Ricciardi for DNA collection from Sardinian patients; the Human Biological Data Interchange for U.S. family DNA samples and HLA typing data; and the British Diabetic Association for provision of U.K. families.

1.
Erlich HA, Rotter JI, Chang JD, Shaw SJ, Raffel LJ, Klitz W, Bugawan TL, Zeidler A: Association of HLA-DPB1*0301 with IDDM in Mexican-Americans.
Diabetes
45
:
610
–614,
1996
2.
Noble JA, Valdes AM, Cook M, Klitz W, Thomson G, Erlich HA: The role of HLA class II genes in insulin-dependent diabetes mellitus: molecular analysis of 180 Caucasian, multiplex families.
Am J Hum Genet
59
:
1134
–1148,
1996
3.
Lie BA, Akselsen HE, Joner G, Dahl-Jorgensen K, Ronningen KS, Thorsby E, Undlien DE: HLA associations in insulin-dependent diabetes mellitus: no independent association to particular DP genes.
Hum Immunol
55
:
170
–175,
1997
4.
Noble JA, Valdes AM, Thomson G, Erlich HA: The HLA class II locus DPB1 can influence susceptibility to type 1 diabetes.
Diabetes
49
:
121
–125,
2000
5.
Sham PC, Curtis D: An extended transmission/disequilibrium test (TDT) for multi-allele marker loci.
Ann Hum Genet
59
:
323
–336,
1995
6.
Koeleman BPC, Dudbridge F, Cordell HJ, Todd JA: Adaptation of the extended transmission/disequilibrium test to distinguish disease associations of multiple loci: the Conditional Extended Transmission/Disequilibrium Test.
Ann Hum Genet
64
:
215
–221,
2000
7.
Spielman RS, McGinnis RE, Ewens WJ: Transmission test for linkage disequilibrium: the insulin gene region and insulin-dependent diabetes mellitus (IDDM).
Am J Hum
  Genet
52
:
506
–516,
1993
8.
Valdes AM, Thomson G: Detecting disease-predisposing variants: the haplotype method.
Am J Hum Genet
60
:
703
–716,
1997
9.
Todd JA, Bell JI, McDevitt HO: HLA-DQ β gene contributes to susceptibility and resistance to insulin-dependent diabetes mellitus.
Nature
329
:
599
–604,
1987
10.
Cucca F, Lampis R, Frau F, Macis D, Angius E, Masile P, Chessa M, Frongia P, Silvetti M, Cao A, De Virgiliis S, Congia M: The distribution of DR4 haplotypes in Sardinia suggests a primary association of type I diabetes with DRB1 and DQB1 loci.
Hum Immunol
43
:
301
–318,
1995
11.
Nepom BS, Nepom GT, Coleman M, Kwok WW: Critical contribution of β chain residue 57 in peptide binding ability of both HLA-DR and -DQ molecules.
Proc Natl Acad Sci U S A
93
:
7202
–7206,
1996
12.
Corper AL, Stratman T, Apostolopoulos V, Scott CA, Garcia KC, Kang AS, Wilson IA, Teyton L: A structural framework for deciphering the link between I-Ag7 and autoimmune diabetes.
Science
288
:
505
–511,
2000
13.
Wen L, Wong FS, Tang J, Chen NY, Altieri M, David C, Flavell R, Sherwin R: In vivo evidence for the contribution of human histocompatibility leukocyte antigen (HLA)-DQ molecules to the development of diabetes.
J Exp Med
191
:
97
–104,
2000
14.
Schmidt D, Verdaguer J, Averill N, Santamaria P: A mechanism for the major histocompatibility complex–linked resistance to autoimmunity.
J Exp Med
186
:
1059
–1075,
1997
15.
Bain SC, Todd JA, Barnett AH: The British Diabetic Association—Warren Repository (Editorial).
Autoimmunity
7
:
83
–85,
1990
16.
Lernmark A, Ducat L, Eisenbarth G, Ott J, Permutt MA, Rubenstein P, Spielman R: Family cell lines available for research.
Am J Hum Genet
47
:
1028
–1030,
1990
17.
Bunce M, Barnardo MC, Welsh KI: The PCR-SSP Manager computer program: a tool for maintaining sequence alignments and automatically updating the specificities of PCR-SSP primers and primer mixes.
Tissue Antigens
52
:
158
–174,
1998
18.
Guo SW, Thompson EA: Performing the exact test of Hardy-Weinberg for multiple alleles.
Biometrics
48
:
361
–372,
1992
19.
Schaid DJ: Case-parents design for gene-environment interaction.
Genet Epidemiol
16
:
261
–273,
1999
20.
Thomson G: Mapping disease genes: family-based association studies.
Am J Hum Genet
57
:
487
–498,
1995

Address correspondence and reprint requests to the following authors:

Francesco Cucca, Dipartimento di Scienze Biomediche e Biotecnologie, University of Cagliari, Via Jenner, Cagliari 09121, Italy. E-mail: fcucca@mcweb.unica.it.

John Todd, Wellcome Trust Centre for Molecular Mechanisms in Disease, Wellcome Trust/MRC Building, Addenbrooke’s Hospital, Hills Road, Cambridge CB2 2XY, U.K. E-mail: john.todd@cimr.cam.ac.uk.

Received for publication 30 June 2000 and accepted in revised form

AFBAC, affected family-based control subject; ETDT, extended transmission/disequilibrium test; HM, haplotype method; MHC, major histocompatibility complex; ORT, odds ratio for transmission; PCR, polymerase chain reaction; PW, pairwise; SSP, sequence-specific primer.