Functional consequences of HLA-DQ8 homozygosity versus heterozygosity for islet autoimmunity in type 1 diabetes.

Human leukocyte antigen (HLA) class II haplotypes are established risk factors in type 1 diabetes (T1D). The heterozygous DQ2/8 genotype confers the highest risk, whereas the DQ6/8 genotype is protective. We hypothesized that DQ2/8 trans-molecules composed of α and ß chains from DQ2 and DQ8 express unique ß-cell epitopes, whereas DQ6 may interfere with peptide binding to DQ8. Here we show that a single insulin epitope (InsB13-21) within the T1D prototype antigenic InsB6-22 peptide can bind to both cis- and trans-dimers, although these molecules display different peptide binding patterns. DQ6 binds a distinct insulin epitope (InsB6-14). The phenotype of DQ8-restricted T cells from a T1D patient changed from proinflammatory to anti-inflammatory in the presence of DQ6. Our data provide new insights into both susceptible and protective mechanism of DQ, where protecting HLA molecules bind autoantigens in a different (competing) binding register leading to 'epitope stealing', thereby inducing a regulatory, rather than a pathogenic immune response.

A predictive model of short cervix at 20-24 weeks using first-trimester cervical length measurement and maternal history.

OBJECTIVES: To develop a model for the prediction of short cervix ( ≤ 15 mm) at 20-24 weeks by combining maternal history and transvaginal ultrasonographic measurement of cervical length at 11-14 weeks. To explore the value of an additional ultrasound examination of the cervix at about 17 weeks. METHODS: Longitudinal prospective study in 800 unselected pregnant women presenting for first-trimester ultrasound assessment by nuchal translucency and serum biochemistry. Cervical length was evaluated transvaginally between 11 weeks and 13 weeks and 6 days (cx1), at 16-19 weeks (cx2) and 20-24 weeks (cx3). Backward multiple logistic regression analysis with cx3 ≤ 15 mm as the dependent variable was used to identify the predictors of a short cervix at 20-24 weeks. RESULTS: Cx1 and history of preterm delivery were significant independent contributors of a short cervix at 20-24 weeks [area under the curve (AUC 0.808, p < 0.001, Model) 1]. Furthermore, the cx1/cx2 ratio was a significant independent predictor of a short cervix at 20-24 weeks (odds ratio = 58.325 p = 0.012). The addition of the cx1/cx2 ratio improved the model (AUC = 0.878, p < 0.001, Model 2). CONCLUSIONS: A short cervix at 20-24 weeks can be predicted at the 11-14 weeks scan. The addition of a cervical measurement at about 17 weeks can improve the prediction model.

Attitudes of pregnant women regarding termination of pregnancy for fetal abnormality.

OBJECTIVE: To study the attitudes of pregnant women towards termination of pregnancy for fetal abnormality. MATERIALS AND METHODS: A questionnaire was completed by all pregnant women attending routine ultrasound scan. They were asked whether they would opt for termination of the pregnancy in case the fetus was diagnosed with one of the following abnormalities: lethal anomaly, anomaly causing developmental delay, anomaly causing physical handicap, anomaly causing disfigurement and severe anomaly diagnosed after 24 weeks of pregnancy. Logistic regression analysis was used to examine the effect of a variety of demographic and socio-economic characteristics in their choices. RESULTS: A total of 533 women completed the questionnaire out of which 447 (86%) would terminate the pregnancy in case of lethal fetal anomaly. The corresponding figures for anomaly causing developmental delay, anomaly causing physical handicap and anomaly causing disfigurement were 396 (77.8%), 332 (65.9%) and 228 (45.2%). A total of 313 (64.7%) would request late termination owing to severe anomaly. The only two statistically significant factors that influenced the maternal decision on pregnancy termination were religious beliefs and the frequency of practicing religious duties (p < 0.001). CONCLUSION: The majority of pregnant women would terminate pregnancy for lethal fetal anomaly and for an anomaly causing mental or physical handicap, even in late pregnancy.

Pulsed-field gradient nuclear magnetic resonance study of transport properties of fluid catalytic cracking catalysts.

Pulsed-field gradient nuclear magnetic resonance (PFG NMR) has been applied to study molecular diffusion in industrial fluid catalytic cracking (FCC) catalysts and in USY zeolite for a broad range of molecular displacements and temperatures. The results of this study have been used to elucidate the relevance of molecular transport on various displacements for the rate of molecular exchange between catalyst particles and their surroundings. It turned out that this rate, which may determine the overall rate and selectivity of FCC process, is primarily related to the diffusion mode associated with displacements larger than the size of zeolite crystals located in the particles but smaller than the size of the particles. This conclusion has been confirmed by comparative studies of the catalytic performance of different FCC catalysts.

No specific reactivity to E. coli glutamic acid decarboxylase from sera of newly-diagnosed insulin dependent diabetic patients.

The 65 kD isoform of Glutamic Acid Decarboxylase (GAD), is one of the major autoantigens in human type 1 diabetes mellitus. This enzyme shares amino acid identity, in select regions already determined as antigenic with its counterpart from E. coli. We tested the reactivity of diabetic and normal sera and an E. coli GAD-specific monoclonal antibody (2D9) to E. coli GAD by solid phase and competition ELISA, as well as immunoblotting to check for cross-reactivity of autoantibodies to the two antigens. Specific antibodies for E. coli GAD are present in diabetics and normal subjects without any differences in frequency and titer. The reactivity of such antibodies in ELISA could be blocked in a dose-dependent manner by the addition of excess antigen in the liquid phase. Furthermore, the monoclonal antibody against E. coli GAD does not recognise human recombinant GAD65 in an ELISA. We conclude that there is no basis for cross-reactivity between the two antigens, and antibody reactivity to GAD65 in man cannot arise from cross-reactivity to the E. coli enzyme.

Modulation of T cell response to hGAD65 peptide epitopes.

Human CD4 T cell responses to an epitope of hGAD65 (GAD = glutamic acid decarboxylase), residues 555-567, are modulated by interaction with an altered peptide ligand containing modifications at TCR contact residues. Using different HLA-DR4 molecules with polymorphisms at sites corresponding to peptide binding pockets p1 and p9, we tested the effect of additional modifications in the altered peptide ligand (APL) designed to increase the avidity of the MHC-peptide interaction and therefore the efficiency of TCR signaling. Modification of the peptide or the MHC molecule which enhanced the p1 interaction also enhanced the antagonist activity of the modified APL. In contrast, modifications at p9 led to a reversal in APL function, resulting in agonist activity. Molecular homology modeling of these MHC-peptide interactions suggests a structural basis for this functional dichotomy in which topographically remote variations lead to unique interaction effects.

Novel approaches toward early diagnosis of islet allograft rejection.

BACKGROUND: The inability to diagnose early rejection of an islet allograft has previously proved to be a major impediment to progress in clinical islet transplantation. The need to detect early rejection will become even more relevant as new tolerance-inducing protocols are evaluated in the clinic. We explored three novel approaches toward development of early diagnostic markers of islet rejection after islet allotransplantation. METHODS: (a) Canine islet allograft transplant recipients were immunosuppressed for 1 month, then therapy was withdrawn. Serum glutamic acid decarboxylase antigen (GAD65), an endogenous islet protein, was monitored daily with a CO2 release assay. (b) Rodent islets were genetically engineered to express a unique foreign protein (beta-galactosidase) by using adenoviral vectors, and after allograft transplantation, the viral-specific protein was measured in serum using optical luminescence. (c) Rodents receiving islet allografts were immunosuppressed temporarily, and daily glucose tolerance tests were followed until graft failure occurred. RESULTS: (a) Although serum monitoring of GAD65 antigen demonstrated elevated levels preceding loss of graft function in preliminary studies, the effect was not reproducible in all animals. (b) Genetically engineered rodent islets demonstrated normal insulin kinetics in vitro (insulin stimulation index 2.57+/-0.2 vs. 2.95+/-0.3 for control islets, P=ns), and purified viral protein products had a stable half-life of 8 hr in vivo. After islet allotransplantation, there were two peak elevations in serum viral proteins, confirming that an intra-islet "sentinel signal" could be detected serologically during acute rejection. There was no lead-time ahead of hyperglycemia, however. (c) Daily sequential intravenous glucose tolerance (IVGT) tests demonstrated evidence of allograft dysfunction (decline in KG) with a 2-day lead time to hyperglycemia (2.58+/-0.3 vs. 1.63+/-0.2%/min, respectively, P<0.001), with an accuracy of 89%, sensitivity of 78%, and specificity of 95%. CONCLUSIONS: Of the three diagnostic tests, metabolic assessment with an abbreviated IVGT was the most effective method of demonstrating early islet dysfunction due to rejection.

Structure of celiac disease-associated HLA-DQ8 and non-associated HLA-DQ9 alleles in complex with two disease-specific epitopes.

The association of celiac disease (CD) with HLA-DQ2 and HLA-DQ8 is indicative of preferential mucosal T cell recognition of gluten fragments bound to either DQ allele. We have recently identified two gluten-derived, HLA-DQ8-restricted T cell stimulatory peptides, one each from gliadin and glutenin, recognized by specific T cell clones derived from the small intestine of CD patients. We have now performed molecular modeling and examined the fine specificity of these peptides in complex with HLA-DQ8. There is only one binding register for both peptides, with glutamine residues at the p1 and p9 anchor positions. Both T cell clones recognize substituted peptides at p1 and p9, but poorly so at p2-p8, especially the gliadin-specific clone. Contrasting patterns of recognition of p9Gln --> Glu peptide variants (both predicted as better DQ8 binders by modeling) were observed: enhancement of recognition for the gliadin peptide, yet complete absence thereof for the glutenin peptide. The double-substituted gliadin peptide variant p1/9Gln --> Glu, which can also arise by pepsin/acid/transglutaminase treatment, shows a considerable increase in sensitivity of recognition, consistent with better binding of this peptide to DQ8, as predicted by energy minimization. Surprisingly, the two native peptides are also recognized by their respective T cell clones in the context of the related molecule HLA-DQ9 (beta57Asp(+)). The p1/9Gln --> Glu gliadin peptide variant is likewise recognized, albeit with a 10-fold lower sensitivity, the first reported p9Glu binding in a beta57Asp(+) MHC II allele. Our results have important implications for the pathogenesis of autoimmune disease and the possible manipulation of aberrant responses thereof.

Modelling of the MHC II allele I-A(g7) of NOD mouse: pH-dependent changes in specificity at pockets 9 and 6 explain several of the unique properties of this molecule.

AIMS/HYPOTHESIS: We modelled the three-dimensional structure of I-A(g7), the chief genetic component of diabetes in non-obese diabetic mice, to understand the unusual properties of this molecule. METHODS: Modelling was done, in complex with established antigenic peptides, based on the structure of I-A(k). RESULTS: The selectivity of the I-A(g7) molecule changes greatly at pockets 9 and 6 but hardly at all at pockets 1, 4 and 7, between endosomal pH (5.0) and extracellular pH (7.0), in agreement with previous results. This selectivity is attributed to the unique combination of beta9His, beta56His and beta57Ser. The positive charges in and around pocket 9 at pH 5, favour binding by negatively charged residues. At pH 7 however, the uncharged alpha68, beta9 and beta56 histidines favour the accommodation of the bulky residues lysine, arginine, phenylalanine and tyrosine at pocket 9. The combination of beta9His and alpha66Glu is responsible for the pH-dependent selectivity at pocket 6. Furthermore, the lack of repulsion between beta56His and alpha76Arg at pH 7 leads to a more stable ternary complex. CONCLUSION/INTERPRETATION: These results reconcile previous conflicts over the peptide binding ability of I-A(g7) and its motif. They furthermore provide possible explanations for the short lifetime of cell-surface I-A(g7) complexes in vivo, the higher threshold of thymic negative selection and inherent self-reactivity shown by immunocytes in these mice and the protection from diabetes afforded to them by several transgenically expressed mouse class II alleles. This contributes to an understanding of the pathogenesis of Type I (insulin-dependent) diabetes mellitus in this animal.

Structural analysis of two HLA-DR-presented autoantigenic epitopes: crucial role of peripheral but not central peptide residues for T-cell receptor recognition.

Specific and major histocompatibility complex (MHC)-restricted T-cell recognition of antigenic peptides is based on interactions of the T-cell receptor (TCR) with the MHC alpha helices and solvent exposed peptide residues termed TCR contacts. In the case of MHC class II-presented peptides, the latter are located in the positions p2/3, p5 and p7/8 between MHC anchor residues. For numerous epitopes, peptide substitution studies have identified the central residue p5 as primary TCR contact characterized by very low permissiveness for peptide substitution, while the more peripheral positions generally represent auxiliary TCR contacts. In structural studies of TCR/peptide/MHC complexes, this has been shown to be due to intimate contact between the TCR complementarity determining region (CDR) three loops and the central peptide residue. We asked whether this model also applied to two HLA-DR presented epitopes derived from an antigen targeted in type 1 diabetes. Large panels of epitope variants with mainly conservative single substitutions were tested for human leukocyte antigen (HLA) class II binding affinity and T cell stimulation. Both epitopes bind with high affinity to the presenting HLA-DR molecules. However, in striking contrast to the standard distribution of TCR contacts, recognition of the central p5 residue displayed high permissiveness even for non-conservative substitutions, while the more peripheral p2 and p8 TCR contacts showed very low permissiveness for substitution. This suggests that intimate TCR interaction with the central peptide residue is not always required for specific antigen recognition and can be compensated by interactions with positions normally acting as auxiliary contacts.

Role of cytokines in the pathogenesis of anemia of chronic disease in rheumatoid arthritis.

The aim of our study was to evaluate the role of proinflammatory cytokines: tumor necrosis factor alpha (TNFalpha), interleukin-1beta (IL-1beta), and interleukin-6 (IL-6), as well as the possible contribution of interleukin-10 (IL-10) in anemia of chronic disease (ACD) of rheumatoid arthritis (RA) patients. We measured the serum levels of TNFalpha, IL-1beta, and IL-6 in 105 anemic and 127 nonanemic RA patients. We also investigated the effects of the above cytokines on the development of burst-forming units-erythroid (BFUe) and colony-forming units-erythroid (CFUe) in bone marrow cultures. Anemic patients had significantly higher serum levels of TNFalpha, IL-1beta, and IL-6 compared to nonanemics. Serum IL-10 levels were low and there was no significant difference in IL-10 concentrations between anemic and nonanemic patients. Proinflammatory cytokines inhibited proliferation of BFUe and CFUe. IL-10 did not decrease the erythroid colony growth. Proinflammatory cytokines may play a role in the pathogenesis of ACD in RA patients. Low levels of IL-10 possibly contribute to the development of ACD.

A peptide binding motif for I-Eg7, the MHC class II molecule that protects E alpha-transgenic nonobese diabetic mice from autoimmune diabetes.

The nonobese diabetic (NOD) mouse, a model of spontaneous insulin-dependent diabetes mellitus (IDDM), fails to express surface MHC class II I-Eg7 molecules due to a deletion in the E alpha gene promoter. E alpha-transgenic NOD mice express the E alpha E beta g7 dimer and fail to develop either insulitis or IDDM. A number of hypotheses have been proposed to explain the mechanisms of protection, most of which require peptide binding to I-Eg7. To define the requirements for peptide binding to I-Eg7, we first identified an I-Eg7-restricted T cell epitope corresponding to the sequence 4-13 of Mycobacterium tuberculosis 65-kDa heat shock protein (hsp). Single amino acid substitutions at individual positions revealed a motif for peptide binding to I-Eg7 characterized by two primary anchors at relative position (p) 1 and 4, and two secondary anchors at p6 and p9. This motif is present in eight of nine hsp peptides that bind to I-Eg7 with high affinity. The I-Eg7 binding motif displays a unique p4 anchor compared with the other known I-E motifs, and major differences are found between I-Eg7 and I-Ag7 binding motifs. Analysis of peptide binding to I-Eg7 and I-Ag7 molecules as well as proliferative responses of draining lymph node cells from hsp-primed NOD and E alpha-transgenic NOD mice to overlapping hsp peptides revealed that the two MHC molecules bind different peptides. Of 80 hsp peptides tested, none bind with high affinity to both MHC molecules, arguing against some of the mechanisms hypothesized to explain protection from IDDM in E alpha-transgenic NOD mice.

The destructive action of IL-1alpha and IL-1beta in IDDM is a multistage process: evidence and confirmation by apoptotic studies, induction of intermediates and electron microscopy.

Using the rat beta-cell RIN-5AH insulinoma line as a means for studying insulin-dependent diabetes mellitus (IDDM), it is shown that interleukin-1 (IL-1) induces beta-cell damage initiated by early apoptotic signals. This action is demonstrated by DNA fragmentation, as assessed by specific BrdU labeling, surface expression of Fas and nitric oxide (NO) production. In addition, the interplay between NO and Fas is shown, while scanning electron microscopy (SEM) confirms apoptosis by revealing the degree and type of cellular damage which, in the case of IL-1alpha, can be reversed by an inhibitor to NO synthesis. Apoptosis is also reconfirmed by transmission electron microscopy (TEM) by observing condensed nuclear chromatin after IL-1 exposure. Thus, treatment of insulinoma cells with IL-1alpha and IL-1beta seems to initiate a number of signals, including PKC activation as published previously, that ultimately lead to beta-cell destruction. Each IL-1 isoform, however, definitely follows a different pathway of action.

RGD sequences in several receptor proteins: novel cell adhesion function of receptors?

In the process of homology modelling of the 3-dimensional structure of alleles of the human histocompatibility protein HLA-DQ, we discovered that its RGD tripeptide (beta 167-169) forms part of a loop. A search through protein sequence data bases, revealed this cell adhesion motif in 67 integral plasma membrane proteins (in 48 extracellularly, and in the remaining 19 intracellularly), which are bona fide receptors, and none of them has thus far been considered as a cell adhesion protein. The 3-dimensional structure of one of these, the rat neonatal Fc receptor, is known and its extracellular RGD sequence is in an adhesion-like loop, a fact that went unnoticed in the original papers. In a few other cases, e.g. rat and mouse growth hormone receptor, and mouse CD40 ligand, homology modelling by ourselves and others reveals that the said sequences are part of a loop, in similarity to all RGD sequences found in proteins with established adhesion function and known 3-dimensional structure. Likewise, inspection of all known protein 3-dimensional structures containing an RGD sequence, and not having a documented cell adhesion function (total of 65 separate entries) shows that such sequence is mostly (52/65 or 80% of cases) part of a loop. We therefore call attention to these surprising findings, discuss the possible cell adhesion role of these receptor proteins, and draw an analogy from the two well characterised examples, that of soluble IGF binding protein 1 and the transcriptional activator protein Tat of HIV, where their RGD sequences have been shown by site-directed mutagenesis to participate in cell-adhesion interactions, without prior knowledge of the location of the tripeptide, or the 3-dimensional structure of the respective protein.

IL-1 beta transduces different signals than IL-1 alpha leading to class II antigen expression on beta-insulinoma RIN-5AH cells through specific receptors.

Like most cytokines, IL-1 transduces its signals for growth, differentiation and diverse cellular functions after binding to specific receptors on the cell surface. Up to now two IL-1 receptors have been reported, type I which induces signal transduction and type II which binds IL-1 but does not transduce signalling. By using the rat insulinoma RIN-5AH cell line that expresses both types of receptor mRNA, and computer-assisted binding analysis, we show that interleukin-1 beta (IL-1 beta) binds to a single class of high affinity receptors with a Kd of 155 pmol/l. The average number of receptors on adherent cell layer is calculated to be 7300 per cell. 125I-IL-1 beta binding can be competed out by unlabelled IL-1 beta. 125I-IL-1 alpha binding can be also obtained and is subject to competition by cold IL-1 alpha. Its saturation curve, however, varies within experiments due to differential receptor up-regulation. These results have also been confirmed by FACS analysis using specific antibodies to type I and II IL-1 receptors, where type I receptor antibody binds strongly to RIN-5AH cells, and type II receptor antibody shows weak staining, also due to inadequate receptor up-regulation. In order to determine whether functional signal transduction occurs via the receptors detected, it is shown that IL-1 beta is able to induce MHC class II antigen expression on the surface of the RIN cells, whereas IL-1 alpha is unable to do so, indicating different signal reception by the cells. IL-1 beta-induced class II upregulation shows moderate signs of p21ras or/and PKC dependency, whereas IL-1 alpha strongly activates both pathways that probably regulate different functions. Finally, both IL-1 alpha and beta induce nitric oxide (NO) production in a time-dependent fashion which appears to be unrelated to the signals and pathways described, but may be involved in the onset of autoimmune type 1 diabetes.

Unique peptide binding characteristics of the disease-associated DQ(alpha 1*0501, beta 1*0201) vs the non-disease-associated DQ(alpha 1*0201, beta 1*0202) molecule.

To understand the dominant association of celiac disease (CD) with the presence of HLA-DQ(alpha 1*0501, beta 1*0201), the peptide binding characteristics of this molecule were compared with that of the structurally similar, but non-CD-associated DQ(alpha 1*0201, beta 1*0202) molecule. First, naturally processed peptides were acid-extracted from immuno-affinity-purified DQ molecules of both types. Both molecules contained the Ii-derived CLIP sequence and a particular fragment of the major histocompatibility complex (MHC) class I alpha chain. Use of truncated analogues of these two peptides in cell-free peptide binding assays indicated that identical peptide frames are used for binding to the two DQ2 molecules. Detailed substitution analysis of the MHC class I peptide revealed identical side chain requirements for the anchor residues at p6 and p7. AT p1, p4, and p9, however, polar substitutions (such as N, Q, G, S, and T) were less well tolerated in the case of the DQ(alpha 1*0201, beta 1*0202) molecule. This most striking difference between the two DQ molecules is the presence of and additional anchor residue at p3 for the DQ(alpha 1*0201, beta 1*0202) molecule, whereas this residue was found not to be specifically involved in binding of peptides to DQ(alpha 1*0501, beta 1*0201). Similar results were obtained applying substitution analysis of the CLIP sequence. Molecular modelling of the DQ2 proteins complexed with the MHC class I and CLIP peptide corresponds well with the binding data. The results suggest that both CLIP and the MHC class I peptide bind DQ(alpha 1*0501, beta 1*0201) and DQ(alpha 1*0201, beta 1*0202) in a DR-like fashion, following highly similar binding criteria. This detailed characterization of unique peptide binding properties of the CD-associated DQ(alpha 1*0501, beta 1*0201) molecule should be helpful in the identification of CD-inducing epitopes.

Novel structural features of the human histocompatibility molecules HLA-DQ as revealed by modeling based on the published structure of the related molecule HLA-DR.

Structural modeling of the HLA-DQ molecules, a group of human histocompatibility antigens linked to autoimmune diseases and immunosuppression-based on the structure of the homologous molecule DR1, has revealed an overall shape typical of the class II histocompatibility molecules, yet with several novel features. These are unique to HLA-DQ and include: (1) an antigen-binding groove with a polymorphic first pocket and anchoring in the second and/or fifth pocket, (2) a polymorphic beta 49-56 dimerization patch, and (3) in many alleles a prominent Arg-Gly-Asp loop (beta 167-169), probably involved in cell adhesion, as it exhibits an architecture similar to identical sequences involved in such function. The alpha 2 beta 2 dimerisation domain and the CD4-binding region are nearly identical to their counterparts in the structure of HLA-DR1. The significance of the few substitutions in the CD-4 binding region remains to be evaluated. The polymorphic first antigen-binding pocket and the anchoring in the second and/or fifth pocket point to differences in antigenic fragment selection compared to HLA-DR antigens, while the polymorphism in the beta 49-56 homodimerization patch implies either ease of spontaneous or T lymphocyte receptor-induced homodimerization or difficulty in the latter. As homodimerization appears to be an obligatatory intermediate in the activation of cognate DQ-restricted T lymphocytes and DQ-bearing antigen-presenting cells, the dimerization properties of DQ allels signify the respective ease or difficulty of activation of these two cell types. The RGD loop confers cell adhesion possibilities to those DQ allels that possess it, yet its putative ligand cannot be defined at present. These features are suggestive of the probable mechanisms through which some of the unique immunological properties of the HLA-DQ molecules are effected.

Polymorphic structural features of modelled HLA-DQ molecules segregate according to susceptibility or resistance to IDDM.

The structural features of HLA-DQ alleles which are susceptible and resistant to insulin-dependent diabetes mellitus (IDDM) have been examined using a model of their three-dimensional structure obtained by energy minimisation, based on the published structure of HLA-DR1. The model shows DQ molecules to have an overall shape nearly identical to that of DR molecules, but with significant differences in the fine structure: 1) the antigen-binding groove of DQ molecules has a polymorphic first pocket; this pocket can be either amphiphilic or hydrophilic, 2) The beta 49-56 dimerisation domain of DQ is polymorphic: hydrophobic, or amphiphilic, or hydrophilic and positively charged, leading to spontaneous or T-cell receptor-induced homodimer formation, or T-cell receptor-induced homodimer formation, or difficulty of the formation of such dimers, respectively; 3) a prominent Arg-Gly-Asp loop is formed by some DQ alleles (beta 167-169) and probably functions in cell adhesion. There are also small differences in the residues and sequences implicated in CD4 binding (mostly in DQ beta 134-148) but the significance of these differences cannot be evaluated at present. All seven DQ alleles which confer susceptibility to IDDM possess a hydrophilic first pocket in the antigen-binding groove, a hydrophobic or amphiphilic beta 49-56 dimerisation patch that allows for spontaneous or T-cell receptor-induced dimerisation, and the Arg-Gly-Asp loop. By contrast, in the protective alleles at least one of these three features is absent. This segregation of phenotypes according to susceptibility or resistance can well explain the model of tighter autoantigen binding by the protective alleles compared to the susceptible alleles, previously proposed for the pathogenesis of IDDM.