Abstract Affected sib pair and linkage disequilibriumanalysis, intrafamilial and case-control association studieswere performed on 81 Danish multiplex insulin-depen-dent diabetes mellitus (IDDM) families (382 individuals)and 82 healthy Danish controls. The results confirm thelinkage of D15S107 to IDDM in these Danish IDDMfamilies (P = 0.010). When these data are combined withthose of previous studies, an even stronger case for link-age can be made (P = 0.0005). Our analyses show that theD15S107*130 allele provides increased susceptibility (P =0.02, relative risk = 3.55) and that the D15S107 locuscontributes up to 16% of the familial clustering of IDDM.The analysis of affected sib pairs suggests that HLA andD15S107 may possibly act independently of each other.Taken together with our previous findings, our results sug-gest that three causes of susceptibilities can be discernedin the IDDM patient population: (1) a major susceptibilitycaused by the HLA-DRB1 alleles; (2) a minor susceptibil-ity caused by the joint action of HLA and other non-HLAgene(s); and (3) a minor susceptibility caused by non-HLA gene(s).

Introduction

Insulin-dependent diabetes mellitus (IDDM) is a multifac-torial disease characterised by the destruction of the in-sulin-producing beta cells of the pancreas. The incidenceby the age of 20 years is about 0.4% in Caucasians of Eu-ropean descent. The peak age of onset is about 12 years.Exogenous insulin administration is required by the pa-tients. Several regions on different chromosomes have

been implicated in the development of IDDM. The majordisease locus (sometimes referred to as IDDM1) is en-coded by the HLA class II genes within the major histo-compatibility complex (MHC) on chromosome 6p21 (Ne-rup et al. 1977; Rønningen et al. 1992; Buyse et al. 1994;Davies et al. 1994; Field et al. 1994; Hashimoto et al.1994; Pociot et al. 1994; Zamani et al. 1994).

An IDDM2 locus in the insulin gene region (located onchromosome 11p15) was initially found by case-controlstudies and was confirmed with family-based associationstudies and affected sib pair analysis (Julier et al. 1991;Bain et al. 1992; Lucassen et al. 1993; Owerbach andGabbay 1994; Bennett et al. 1995).

Recently, a number of other non-HLA genes on chro-mosome 15q26 (IDDM3) (Field et al. 1994; Luo et al.1995), on 11q13 (IDDM4) (Davies et al. 1994; Field et al.1994; Hashimoto et al. 1994), on 6q (IDDM5) (Davies etal. 1994), on 2q (IDDM7) (Davies et al. 1994; Copemanet al. 1995) and on 6q25-q27 (IDDM8) (Luo et al. 1995)have been suggested to increase susceptibility to IDDM.Although linkage with affected sib pair analysis has beenshown in all these non-HLA genes, some of them are notconsistently observed in some studies and heterogeneityin the evidence presented was found between some datasets (Davies et al. 1994; Copeman et al. 1995; Luo et al.1995). These studies nevertheless support the probabilitythat many genes are involved in IDDM susceptibility andsuggest that a high degree of genetic heterogeneity existsin IDDM. This also suggests that identification and con-firmation of linkage to IDDM should be studied in differ-ent ethnic groups and geographical regions.

Previously, we have studied the association and thelinkage between IDDM and six different genes in theHLA region, in particular DRB1, in Danish IDDM fami-lies. Strong linkage was found, with the highest risk con-ferred by DRB1Lys71+ (Zamani et al. 1996). However, only33% of the familial clustering of IDDM could be ex-plained by the HLA locus in these families, suggestingthat other genes might be involved. Since IDDM in thesefamilies did not link to the locus on chromosome 2q31-q33 (IDDM7; Copeman et al. 1995), we examined the ef-

Mahdi Zamani · Flemming Pociot · Peter Raeymaekers ·Jørn Nerup · Jean-Jacques Cassiman

Linkage of type I diabetes to 15q26 (IDDM3) in the Danish population

Hum Genet (1996) 98 :491–496 © Springer-Verlag 1996

Received: 18 March 1996 / Revised: 17 May 1996

ORIGINAL INVESTIGATION

M. Zamani · P. Raeymaekers · J.-J. Cassiman (Y)Center for Human Genetics, University of Leuven, Campus Gasthuisberg, Herestraat 49, B-3000 Leuven, BelgiumTel.: +32 16 345860; Fax: +32 16 345997 e-mail: Jean-Jacques.Cassiman@med.kuleuven.ac.be

F. Pociot · J. NerupSteno Diabetes Center, Gentofte, Denmark

fect of the IDDM3 locus (15q26) in these families. In par-ticular, linkage with the D15S107 marker was tested, andits possible additive effect to an HLA effect was studied.The results confirm the linkage of D15S107 with IDDM inthe Danish families (P = 0.010), identify the D15S107*130allele as a susceptibility allele [P = 0.02, relative risk (RR)= 3.55] and show that D15S107 contributes up to 16% tothe familial clustering of IDDM.

Materials and methods

Patients

Eighty-two families (388 individuals), in whom at least two indi-viduals were affected with IDDM, were studied. The families, allof Danish Caucasian origin, were unrelated and the affected indi-viduals were diagnosed as having IDDM according to the WorldHealth Organisation criteria. For the association study, 82 IDDMprobands (first affected offspring of each family) from the familieswere selected. For linkage studies, the sibs, affected and non-af-fected, together with the parents of the original 82 IDDM pro-bands, were studied. In one family, incorrect segregation of bothD15S107 and D15S153 markers was observed. This family wasexcluded from the study. The 81 multiplex families comprised atotal of 382 individuals, 173 IDDM patients and 52 healthy sibs.For the association study, one proband who had a negative resultfor typing was excluded from the study and the genotypes of 80probands were analysed. For the haplotype relative risk (HRR)study, the families with affected and/or missing parents were ex-cluded.

The control group of unrelated Caucasians from Danish origincomprised 82 individuals who were randomly selected from anethnically matched Danish population and was used in the case-control study. All the controls were healthy and did not have anypersonal or family history of IDDM or other endocrinopathies.

Microsatellite marker typing

Microsatellite markers were typed by PCR using the followingprimer sequences (Malcolm and Donlon 1994): D15S107A, 5′-TCA-AAAGCTTGTAATTGGAG-3′; D15S107B, 5′-XTGCAATCTGTA-AGCATTCCT-3′; D15S153A, 5′-AGTACCTGAAAGGGTGGG-3; D15S153B, 5′-XGATCAGTGTAGGCTCCAAA-3 [X = fluores-cein isothiocyanate (FITC)]. One of the PCR primers in each pairwas end-labelled using FITC. PCR amplifications were performedon 20 ng of genomic DNA with 25 pmol of each primer in a 25-µlreaction volume containing 0.5 U Taq DNA polymerase (Perkin-Elmer Cetus, Norwalk, USA), 1.5 mM MgCl2, 200 µM of eachdNTP (Pharmacia LKB Biotechnology, Uppsala, Sweden), 50 mMKCl, 10 mM TRIS-HCl pH 8.3, 0.1% gelatin. The amplificationwas carried out in a DNA thermal cycler (GeneAmp PCR system9600/Perkin-Elmer Cetus). Thermocycling conditions were 94°C,3 min 20 s, followed by 25 cycles of 30 s at 94°C for denaturing,30 s at 55°C for annealing and 30 s at 70.8°C for extension.

A 3-µl aliquot of the amplification product was mixed with anequal volume of stop solution (deionised formamide with 5 ng/mldextran blue), denatured at 85°C for 3 min and cooled on ice im-mediately. The denatured DNA was then loaded on a 6% Hy-drolink Long Ranger gel containing 7 M urea and electrophoresedat 1500 V, 38 mA, 34 W with the gel temperature at 42°C for 230min on an automated DNA sequencer (A.L.F. Pharmacia). TheD15S107 and the D15S153 markers were typed on an automatedDNA sequencer which allowed us to quantify the PCR productsand two known external markers at 123 bp and 235 bp were usedin each sample for accurate sizing.

Thirty samples were typed twice for both markers in two inde-pendent experiments to control the reproducibility of the D15S107and the D15S153 assays. Identical results were obtained.

Statistical methods

Analysis of affected sib pairs

Linkage analysis on affected sib pairs was performed by the ex-tended affected sib pair analysis (ESPA) computer programme(Sandkuyl 1989). In sib pair analysis, which is a parameter-inde-pendent method to evaluate linkage between a trait and a marker,the frequencies of affected sib pairs (ASPs) that share 0, 1 and 2 al-leles identical by descent were compared with the expected valuesof 0.25, 0.5 and 0.25, respectively. If there is no linkage betweenthe marker and the proposed disease gene, these frequenciesshould approximate the expected values. A significant deviationwith an excess in shared alleles shows linkage between the markerand the disease gene.

The contribution of a gene to the familial clustering of IDDMcan also be evaluated by measuring the proportion of affected sibpairs that share no alleles or haplotypes identical by descent (IBD)(Risch 1987).

Relative risk

Relative risks (odds ratio) were calculated using the method ofWoolf (1995): [number of patients with the specific allele /numberof patients without this allele]/[number of controls with the spe-cific allele/number of controls without this allele]. The level of sig-nificance in allele or genotype frequencies was assessed by theFisher’s exact test (Fisher 1960). P-values were corrected for mul-tiple comparisons by the Bonferoni’s correction (Dunn 1958,1961). Only P-values and relative risks (RR) were calculated forthose alleles or genotypes which were observed more than 10times in the total population (patient and control).

Haplotype relative risk

Haplotype relative risk (HRR) is a reliable alternative method toRR for calculating the risk of disease in the presence of a particu-lar genotype. For HRR, probands and their parents are genotyped.The “case” alleles are the alleles which are transmitted to the af-fected probands, the “control” alleles are the non-transmitted alle-les (Falk and Rubinstein 1987; Schaid and Sommer 1994). For in-stance, if the parents have the genotypes A/a and A/a and the af-fected proband has the genotype a/a, the control will be A/A.

Transmission/disequilibrium test

The transmission/disequilibrium test (TDT), which evaluates thefrequency with which a disease allele is transmitted from het-erozygous parents to affected offspring, is calculated by the for-mula: [number of transmitted disease alleles – number of non-transmitted disease alleles]2/[number of transmitted disease alleles+ number of non-transmitted disease alleles] (Spielman et al. 1993;Schaid and Sommer 1994). The TDT can detect both linkage andassociation of markers with the disease or indeed can detect thelinkage when association exists between a marker locus and a dis-ease locus (Spielman et al. 1993; Schaid and Sommer 1994).

Results

Affected sib pair analysis

The results of ESPA between IDDM and D15S107 andD15S153 in 81 multiplex families are shown in Table 1.D15S153 was chosen as a control marker and was simul-taneously studied with D15S107. The genetic distance be-tween D15S107 and D15S153 is 51.7 cM. The ESPA

492

analysis in a total of 91 sib pairs did not show any evi-dence for linkage between IDDM and the D15S153marker. For the D15S107 marker, 113 parent/sib paircombinations were informative and 49 parent/sib paircombinations were uninformative, as the parent was ho-mozygous for D15S107. The frequency of allele sharingwas 61.1% in affected sib pairs from informative families,which is significantly higher than the expected value of50% allele sharing (P = 0.010).

When we pooled the data of the 81 Danish familieswith similar data on 49 human biological data interchange(HBDI) Caucasian families obtained by Luo et al. (1995),the P-value very significantly increased to 0.0005.

Contribution of D15S107 to familial clustering

Genetic contribution to a disease is determined by the de-gree of familial clustering of the disease (λs), which can beestimated from the risk for siblings of patients divided bythe frequency of disease in the general population (Risch1987). The λs for IDDM has been estimated to be about15 in the Caucasian population (Risch 1987). The λs forD15S107 (λs1) to IDDM can be determined by the ratio ofthe expected proportion of affected sib pairs sharing zeroalleles identical by descent, which is 0.25, and the ob-served proportion. In the affected sib pair analysis usingthe D15S107 marker, 31 sib pairs were completely infor-mative (both parents heterozygous), of which 5 sharedzero alleles IBD. Therefore, λs1 is 1.55 to IDDM. Conse-quently, the genetic contribution of D15S107 to the famil-ial clustering of IDDM will be about 16%.

Genetic heterogeneity of IDDM susceptibility in relation to HLA-DRB1

To examine the HLA-related genetic heterogeneity, fami-lies were divided into two subgroups on the basis of theirHLA-DRB1 sharing in the affected sib pairs of each fam-

ily: (1) sib pairs which share higher than or equal to 50% of their DRB1 genes (major susceptibility providedby DRB1); and (2) sib pairs which share less than 50% of their DRB1 genes (less predisposed by DRB1). In thefirst group, D15S107 gene sharing in the affected sib pairsis not elevated (sharing = 59%) but shows a borderlinesignificant P-value (P = 0.05). In the second group, whichshowed < 50% sharing for the DRB1 alleles, a highersharing for the D15S107 was observed (66.7%) comparedto the general Danish IDDM population. Despite the factthat this sample is nearly 3 times smaller than the firstgroup, a lower P-value of 0.03 was observed. This sug-gests that the susceptibility gene in the D15S107 re-gion may act independently of DRB1 and supports the hypothesis of genetic heterogeneity for IDDM suscepti-bility.

Case-control association studies

The frequencies of the D15S107 alleles in 80 patients and82 healthy controls are shown in Table 2. Four of the eightD15S107 alleles (130, 132, 134, 136) were more frequentin both the patient and the control group. Case-control as-sociation analysis revealed that there was a significant in-crease in the frequency of the D15S107*130 allele in theIDDM patients compared with the control group (P =0.02, RR = 3.55 with 95% confidence limit of 1.38–8.2).

Family-based association studies

The frequency distributions of the D15S107 alleles trans-mitted to the affected probands, and to the controls (non-transmitted alleles) are given in Table 3. The frequency ofallele 138 was decreased in affected probands comparedto controls, but not significantly. The frequency of allele130 was significantly increased in patients compared to

493

Table 1 Affected sib pairs analysis for the D15S107 and D15S153marker loci in 81 insulin-dependent diabetis mellitus (IDDM) fam-ilies (ns not significant)

Markers Not shared Shared (%) Non-in- Total P(%) formative

D15S107 44 (38.9) 69 (61.1) 49 0.010D15S153 64 (47.4) 71 (52.6) 21 n.s

Pooled data seta

D15S107 62 (37.3) 104 (62.7) 94 0.0005

HLA sharing ≥ 50%D15S107 34 (40.9) 49 (59.0) 40 0.05

HLA sharing < 50%D15S107 10 (33.3) 20 (66.7) 9 0.03

a Data from our 81 Danish families were pooled with those from 49human biological data interchange families studied by Luo et al.(1995)

Table 2 Distribution of D15S107 marker alleles in Danish IDDMpatients and healthy controls (RR relative risk, CL with a 95% con-fidence limit, n number of chromosomes)

Alleles IDDM Controls Pa RR (CL)(n = 160) (n = 164)

n Fre- n Fre-quency quency

130 19 0.118 6 0.037 0.02 3.55 (1.38–8.2)132 12 0.075 16 0.097134 27 0.168 27 0.165136 83 0.525 89 0.543138 10 0.063 12 0.073140 4 0.025 8 0.049142 3 0.019 6 0.036144 2 0.013 0 0.000

a P-value of Fisher’s exact test with correction for multiple com-parison. Only P-values were calculated and corrected for those al-leles which were observed more than 10 times in the total (patientand control) population

the controls (P = 0.03). This identifies this allele as a sus-ceptible allele with an HRR of 3.74 for developing IDDM.

These association studies did not show any associationbetween D15S153 and IDDM.

Linkage disequilibrium study

To confirm linkage and association of D15S107 to IDDMdetected by sib pair and association studies, we performedthe TDT, one of the most powerful family-based associa-tion methods. Allele 130 was transmitted from heterozy-gous parents to affected probands 17 times and was nottransmitted 5 times, which significantly deviates from theexpected transmission ratio (P = 0.010; Table 4).

Discussion

We have studied linkage and association of 15q26 withtype 1 diabetes (IDDM) by affected sib pair analysis,case-control and family-based association studies. Ouranalysis provides independent evidence for the confirma-tion of linkage between D15S107 and IDDM (P = 0.010;

Table 1), and unaffected sib pair analysis revealed signif-icant deviation from random sharing (not shown). More-over, when we combined our data set with 49 HBDI fam-ilies studied by Luo et al. (1995) to test the linkage ofD15S107 with IDDM in a larger number of families (intotal, 130 families), the P-value very significantly in-creased (P = 0.0005), while the P-value in the Luo fami-lies was only 0.020.

In the affected sib pair analysis from 31 fully informativepairs, 5 pairs did not share any allele. Based on the formulaof Risch (1987), we estimated the degree of familial clus-tering due to D15S107 (λs1) to be 1.55. This is higher thanthat found by Luo et al. (1995) (1.4) and lower than whatwe found in the same population for IDDM1 (Zamani et al.1996) (2.45). The contribution of IDDM3 to the familialclustering of IDDM (λs1) in our study is higher than IDDM2found by Copeman et al. (1995) in the Danish population.However, the value of λs varies for different non-HLAgenes between various data sets (Davies et al. 1994; Cope-man et al. 1995; Cordell and Todd 1995; Luo et al. 1995).

HLA-DRB1-related genetic heterogeneity of IDDMwas tested on the basis of DRB1 sharing. Families withless DRB1 sharing showed significant D15S107 genesharing in affected sib pairs and the frequency of genesharing increased (66.7%) compared to the total (61.6%).In contrast, DRB1-sharing families did not show this sig-nificance (Table 1). This suggests that D15S107 may actindependently of DRB1 in IDDM predisposition. It is alsopossible that the less HLA-predisposed families requiremore susceptibility factors to develop the disease.

Linkage of D15S107 with IDDM has been reported bytwo independent studies. Field et al. (1994) in the com-bined data set of three independent family groups (Cana-dian, British, American) showed linkage of D15S107 toIDDM with a P-value of 0.001 and Luo et al. (1995)found evidence for linkage with a P-value of 0.020 and of0.007 in the combined data set. We have now confirmedthe linkage of D15S107 to IDDM in 81 Danish multiplexIDDM families (P = 0.010) with strong evidence for link-age with the combined data set (P = 0.0005). The percent-age of gene sharing in our study (61.1%) was slightlyhigher than that obtained by Field et al. (1994) (54.8 ±1.6%) and Luo et al. (1995) (59.7%). These findings showthat D15S107 is truly linked to IDDM.

To identify which alleles of the D15S107 marker pro-vided susceptibility for IDDM, we applied case-controland intrafamilial association studies. Allele 130 showedsignificant association with IDDM (P = 0.02) with a RRof 3.55 (Table 2).

We also evaluated the association of D15S107 by fam-ily-based association studies. This method is not sensitiveto effects of population stratification (as seen in case-con-trol studies) because the affected and appropriate controlsare from the same population. The results confirm the as-sociation of allele 130 with IDDM (P = 0.03) with a HRRof 3.74 (Table 3). Finally, linkage and association weretested by the TDT, which again confirmed the linkage ofD15S107 and association of allele D15S107*130 toIDDM (P = 0.010; Table 4).

494

Table 3 Family-based association of D15S107 alleles with IDDMin Danish IDDM families (HRR haplotype relative risk, CL with a95% confidence limit, n.s. not significant, n number of chromosomes)

Alleles IDDM Controls Pa HRR (CL)(n = 136) (n = 136)

n Fre- n Fre-quency quency

130 17 0.125 5 0.036 0.03 3.74 (1.35–9.11)132 12 0.088 8 0.059134 22 0.162 28 0.206136 69 0.507 74 0.544138 9 0.066 17 0.125 n.s.140 3 0.022 2 0.015142 2 0.015 0 0.000144 2 0.015 2 0.015

a P-value of Fisher’s exact test with correction for multiple com-parison. Only P-values were calculated and corrected for those al-leles which were observed more than 10 times in the total (patientand control) population

Table 4 Transmission/disequilibrium test for D15S107 alleles inDanish IDDM families. Numbers of transmissions and non-trans-missions from heterozygous parents to affected probands areshown for five frequent D15S107 alleles. Only a χ2 value (1 df) atsignificant level P < 0.05 is indicated

Alleles Transmitted Non- χ2 Ptransmitted

130 17 5 6.55 0.010132 11 7134 18 24136 22 32138 7 14

We previously found that DRB1Lys71+ is an importantsusceptibility factor for IDDM in a Belgian population(Zamani et al. 1994). This was confirmed in the Danishpopulation with an even higher degree of risk (RR = 17for DRB1Lys71+ and RR = 103 for DRB1Lys71+/+; Zamani etal. 1996). In order to identify whether allele D15S107*130had an additive effect on DRB1Lys71+ in the susceptibilityfor IDDM or acted independently, we looked for the com-bined effect of D15S107 alleles and DRB1Lys71+. From 16individuals who carried allele D15S107*130, 15 individu-als had DRB1Lys71+ (10 homozygous and 5 heterozygousfor Lys71+) and one case did not carry any copy ofDRB1Lys71+. Since from the 80 patients only 4 did notcarry any copy of DRB1Lys71+, no statistical analysis couldbe performed to determine the effect of the D15S107*130allele.

In the affected sib pair analysis, our results showedpossible independent action of D15S107, which was alsosuggested by Field et al. (1994). Luo et al. (1995), how-ever, did not observe significant differences between lessand more HLA-predisposed family groups in sib pairanalysis for IDDM3. In spite of the results of the sib pairanalysis, we could not show an effect of IDDM3 in asso-ciation studies. In general, our results provide evidencethat in the IDDM patients examined at least two groupscan be identified: in the first group, a major susceptibilityfactor is contributed by the HLA gene, with very stronglinkage and association with DRB1Lys71+ (Zamani et al.1994, 1996). The results of heterogeneity testing betweenIDDM1 and IDDM4 or IDDM3 is indeed an argument forthe existence of this group (Field et al. 1994; Cordell andTodd 1995; Cordell et al. 1995; the present study). In thesecond group, containing mainly minor susceptibilities,two or more subgroups could be present; in a first sub-group(s), in addition to HLA, other susceptibility gene(s)seem to be required to develop IDDM. Indeed the jointaction of HLA (IDDM1) and IDDM2 has been clearlydemonstrated by a multiplicative model, while an hetero-geneity model for these has been rejected (Dizier et al.1994; Cordell et al. 1995). In another subgroup(s), the de-velopment of IDDM appears to be independent of theHLA gene. This group includes those cases who do notcarry any copy of DRB1Lys71+ (Zamani et al. 1994, 1996).It has indeed been shown by an heterogeneity model thatcontribution of IDDM1 and IDDM4 to the developmentof IDDM are independent of each other (Cordell et al.1995). As in other multifactorial diseases, the involve-ment of numerous non-HLA genes, together with envi-ronmental factors, makes it much more complex to studythe second group and to show linkage or association ofnon-HLA genes with IDDM. Analysis of the combinator-ial effect of these genes will indeed require very largepopulations but could lead to the further classification ofIDDM into different subtypes.

References

Bain SC, Prins JB, Hearne CM, Rodrigues NR, Rowe BR,Pritchard LE, Ritchie RJ, Hall JRS, Undlien DE, RønningenKS, Dunger DB, Barnett AH, Todd JA (1992) Insulin gene en-coded susceptibility to type 1 diabetes is not restricted to HLA-DR4-positive individuals. Nat Genet 2:212–215

Bennett ST, Lucassen AM, Gough SCL, Powell EE, Undlien DE,Pritchard LE, Merriman ME, Kawaguchi Y, Dronsfield MJ,Pociot F, Nerup J, Bouzekri N, Cambon-Thomsen A, Rønnin-gen KS, Barnett AH, Bain SC, Todd JA (1995) Susceptibilityto human type 1 diabetes at IDDM2 is determined by tandemrepeat variation at the insulin gene minisatellite locus. NatGenet 9:284–292

Buyse I, Sandkuyl LA, Zamani M, Gu XX, Bouillon R, Bex M,Dooms L, et al (1994) Association of particular HLA class IIalleles, haplotypes, and genotypes with susceptibility to insulindependent diabetes mellitus in the Belgian population. Dia-betologia 37:808–817

Copeman JB, Cucca F, Hearne CM, Cornall RJ, Reed PW, Ron-ningen KS, Undlien DE, Nistico L, Buzzetti R, Tosi R, PociotF, Nerup J, Cornèlis F, Barnett AH, Bain SC, Todd JA (1995)Linkage disequilibrium mapping of type 1 diabetes susceptibil-ity gene (IDDM7) to chromosome 2q31-q33. Nat Genet 9:80–85

Cordell HJ, Todd JA (1995) Multifactorial inheritance in type 1 di-abetes. Trends Genet 11:499–504

Cordell HJ, Todd JA, Bennett ST, Kawaguchi Y, Farrall M (1995)Tow-locus maximum lod score analysis of a multifactorialtrait: joint consideration of IDDM2 and IDDM4 with IDDM1in type I diabetes. Am J Hum Genet 57:920–934

Davies JL, Kawaguchi Y, Bennett ST, Copeman JB, Cordell HJ,Pritchard LE, Reed PW, et al (1994) A genome-wide search forhuman type 1 diabetes susceptibility genes. Nature 371:130–136

Dizier MH, Babron MC, Clerget-Darpoux F (1994) Interactive ef-fect of two candidate genes in a disease: extension of themarker-association-segregation c2 method. Am J Hum Genet55:1042–1049

Dunn OJ (1958) Estimation of the means of dependent variables.Ann Math Stat 29:1095–1111

Dunn OJ (1961) Multiple comparisons among means. Am J StatAssoc 56:52–64

Falk CT, Rubinstein P (1987) Haplotype relative risk: an easy reli-able way to construct a proper control sample for risk calcula-tion. Ann Hum Genet 51:227–233

Field LL, Tobias R, Magnus T (1994) A locos on chromosome15q26 (IDDM3) produced susceptibility to insulin-dependentdiabetes mellitus. Nat Genet 8:189–194

Fisher RA (1960) The design of experiments. Oliver and Boyd,Edinburgh, p 258

Hashimoto L, Habita C, Beressi JP, Delepine M, Besse C, Cam-bon-Thomsen A, Deschamps I, et al (1994) Genetic mappingof a susceptibility locus for insulin-dependent diabetes mellituson chromosome 11q. Nature 371:161–163

Julier C, Hyer RN, Davies J, Merlin F, Soularue P, Briant L,Cathelineau G, Deschamps I, Rotter JI, Froguel P, Boitard C,Bell JI, Lathrop GM (1991) Insulin-IGF2 region on chromo-some 11p encodes a gene implicated in HLA-DR4-dependentdiabetes susceptibility. Nature 354:155–159

Lucassen AM, Julier C, Beressi JP, Boitard C, Froguel P, LathropM, Bell JI (1993) Susceptibility to insulin dependent diabetesmellitus maps to a 4.1 kb segment of DNA spanning the insulingene and associated VNTR. Nat Genet 4:305–310

Luo DF, Bui MM, Muir A, Maclaren NK, Thomson G, She JX(1995) Affected-sib-pair mapping of novel susceptibility geneto insulin-dependent diabetes mellitus (IDDM8) on chromo-some 6q25-q27. Am J Hum Genet 57:911–919

Malcolm S, Donlon TM (1994) Report of the second internationalworkshop on human chromosome 15 mapping 1994. CytogenetCell Genet 67:2

495

Nerup J, Cathelineau C, Seignalet J, Thomsen M (1977) HLA andendocrine diseases. In: Dausset J, Svejgaard A (eds) HLA andDisease. Munksgaard, International Copenhagen, pp 149–167

Owerbach D, Gabbay KH (1994) linkage of VNTR/insulin-geneand type I diabetes mellitus: increased gene sharing in affectedsibling pairs. Am J Hum Genet 54:909–912

Pociot F, Rønningen KS, Bergholdt R, Lorenzen T, Johannesen J,Ye K, Dinarello CA, Nerup J and the Danish study group of di-abetes in children (1994) Genetic susceptibility markers inDanish patients with type 1 (insulin dependent) diabetes evi-dence for polygenecity in man. Autoimmunity 19:169–178

Risch N (1987) Assessing the role of HLA linked and unlinked de-terminants of disease. Am J Hum Genet 40:1–14

Rønningen KS, Spurkland A, Tait BD, Drummond B, Lopez-Lar-rea C, Baranda FS, Menendez-Diaz MJ, Caillat-Zucman S,Beaurain G, Garchon HJ, Ilonen J, Reijonen H, Knip M,Boehm BO, Rosak C, Loliger C, Huhnl P, Ottenhoff T, ContuL, Carcassi C, Savi M, Zanelli P, Neri TM, Hamaguchi K,Kimura A, Dong RP, Chikuba N, Nagataki S, Gorodezky C,Bebaz H, Robles C, Coimbra HB, Martinho A, Ruas MA,Sachs JA, Garcia-Pachedo M, Biro A, Nikaein A, DombrauskyL, Gonwa T, Zmijewski C, Monos D, Kamoun M, Layrisse Z,Magli MC, Balducci P, Thorsby E (1992) HLA class II associ-ation in insulin-dependent diabetes mellitus among Black, Cau-casoids, and Japanese. In: Tsuji K, Aizawa M, Sasazuki T (eds)HLA 1991. Oxford University Press, Oxford, pp 713–722

Sandkuyl LA (1989) Analysis of affected sib pairs using informa-tion from extended families. In: Elston RC, Spence MA, HodgeSE, MacCluer JW (eds) Multipoint mapping and linkage basedupon affected pedigree members: genetic analysis workshop 6.Liss, New York

Schaid DJ, Sommer SS (1994) Comparison of statistics for candi-date-gene association studies using cases and parents. Am JHum Genet 55:402–409

Spielman RS, McGinnis RE, Ewens WJ (1993) Transmission testfor linkage disequilibrium: the insulin gene region and insulindependent diabetes mellitus (IDDM). Am J Hum Genet52:506–516

Woolf B (1955) On estimating the relation between blood groupand disease. Ann Hum Genet 19:251–253

Zamani M, Spaepen M, Buyse I, Marynen P, Bex M, Bouillon R,Cassiman JJ (1994) Improved risk assessment for IDDM byanalysis of amino acids in HLA-DQ and DRb1 loci. Eur J HumGenet 2:177–184

Zamani M, Flemming P, Spaepen M, Raeymaekers P, Nerup J,Cassiman JJ (1996) Linkage and association of the HLA genecomplex with IDDM in 81 Danish families: strong linkage be-tween DRB1Lys71+ and IDDM. J Med Genet (in press)

496