Perspectives in Diabetes HLA and Insulin-Dependent Diabetes A Protective Perspective MICHAEL J. SHEEHY

This article presents a model for the HLA effect in insulin-dependent dlabetes mellitus (IDDM) that is almost the mirror image of a model suggested by Nepom (1). In the Nepom model, the products of certain HLA alleles are associated wlth IDDM because .

they bind and present a specific peptide or peptides so as to induce an immune response to pancreatic p-cells; certaln other alleles can protect agalnst IDDM If they compete strongly for binding of the dlabetogenic peptlde. My model focuses instead on the failure of the immune system to maintain tolerance to pancreatic p-cells. I suggest that the HLA alleles negatively associated with IDDM (e.g., DR2 and DQwl) produce products with high affinity for certain p-cell peptide or peptides needed to establish and malntain tolerance to p-cells, whereas the alleles that are common in IDDM (e.g., DR3, DR4, and DQw8) produce products that have low affinity for the tolerogenic peptide or peptldes or that bind the peptide or peptldes in the wrong orientation or configuration for establishing tolerance. I also dlscuss the multiplicity of HLA loci, alleles, and amino acids contributing to IDDM and the fact that the associations of specific loci, alleles, and even genotypes with IDDM depend not only on their intrinsic properties but also on various population parameters. Diabetes 41 :123-29, 1992.

genes, IDDM susceptibility seems to be significantly affected by at least three HLA class II genes: DQA1, DQB1, and DRB1, and probably by the DPBl locus as well.

In an earlier Perspective, Nepom (1) suggested a model wherein HLA class II alleles affect IDDM suscep- tibility as follows. 1) The array of class II dimers, encoded by any individual's HLA complex, vary in their affinities for a specific peptide that can induce autoimmunity to p-cells. 2) Only certain class II dimers, the products of "susceptibility genes," actually promote autoimmunity to p-cells after binding that peptide. 3) An individual is susceptible if the product of a susceptibility gene binds the peptide more strongly than the products of the nonsusceptibility genes present in that individual. An unstated assumption is that the concentration of the diabetogenic peptide is limiting. The mechanism whereby susceptibility and nonsusceptibility gene prod- ucts differ was not specified.

In this article, I argue for an alternative model in which immunological tolerance is the primary determinant of IDDM susceptibility. This is almost the mirror image of the Nepom model (1). In the model proposed herein, a specific peptide or peptides can potentially induce toler- . . ance to p-cells, and p-cell tolerance actually occurs if the individual has one or more class II dimers that can bind

M ost cases of insulin-dependent diabetes mellitus (IDDM) are caused by an autoim- mune process that destroys pancreatic B-cells. Like other autoimmune diseases.

autoimmune' p-cell destruction is influenced by HLA

From the Research Department, American Red Cross Blood Services, Madison; and the Department of Medicine, University of Wisconsin, Madison, Wisconsin.

Address correspondence and reprint requests to Michael Sheehy, PhD, Americar~ Red Cross Blood Services, 4860 Sheboygan Avenue, Mad~son, WI 53705.

Received for publication 28 May 1991 and accepted in revised form 27 September 1991

the tolerogen strongly and in the proper orientation. Susceptible HLA genotypes would be those that "toler- ize" ineffectively because of weak or otherwise ineffective binding of tolerogen.

HETEROGENEITY OF DQW8 HAPLOTYPES WITH REGARD TO IDDM SUSCEPTIBILITY This article originated with an attempt to explain why DQw8+ haplotypes differ greatly in their effects on IDDM susceptibility (2). Generally, in whites, DR4 haplotypes with DQw8 are associated with IDDM susceptibility, whereas DR4 haplotypes with DQw7 are not (3).

DR4 haplotypes also vary at the DRB1 locus, however,

DIABETES, VOL. 41, FEBRUARY 1992 123

with different allelic subtypes of DR4: DRB1*0401-0405 (formerly Dw4, DwlO, and Dw13-15, respectively). The DR4 subtypes also differ in their degree of association with IDDM. These associations are less well documented because the DR4 subtypes cannot be distinguished by serology or restriction-fragment-length polymorphism analysis. Until the advent of oligonucleotide typing, they could be distinguished only by reactions of bulk or cloned T lymphocytes, sophisticated biochemistry, or protein or DNA sequencing. Although the relative asso- ciations of the DR4 subtypes have varied from study to study (4- l l ) , by combining the available data, they appear to rank, from most to least associated with IDDM: 0402 > 0401 0404 and 0405 > 0403. It is least clear where allele 0405 fits into this sequence because there are relatively few data comparing this to DR4 subtypes other than 0403.

Although it was speculated that certain DR4 subtypes are associated with IDDM because of linkage disequilib- rium with DQw8, this was found not to be the case (2). Among DR4+ subjects, DRBI *0401 correlated positively with IDDM but negatively with DQw8 (2). Therefore, the DR4 subtypes are or serve as a marker for an IDDM risk factor distinct from DQw8. At least for the DR4 haplo- types, high susceptibility to IDDM seems to require a combination of IDDM-associated alleles at both DR and DQ p-chain loci (2). In that study, the DR and DQ p-chain alleles had nearly identical degrees of association with IDDM, with the DR effect actually being slightly stronger by most measures of association. Moreover, DQw8 was associated with increased risk for diabetes only when the linked DRBl allele was DRB1*0401 or 0402 (relative risk [RR] = 12), whereas DQw8 haplotypes carrying other DRBl alleles did not correlate with IDDM (RR = 0.7). Likewise, DRB1*0401 in combination with DQw7 did not correlate with IDDM (RR = 0.2) (2). Similarly, Nepom et al. (12) found a much lower risk for DQw8 haplotypes with DRB1*0404 than for DQw8 haplotypes with DRBI *0401 or 0402.

The large study (1 64 patients and 200 control subjects) by Owerbach et al. (9) has been interpreted as showing a minimal role for the DRBl alleles on DR4 haplotypes (3,9). However, the highest RR in that study was for DRBI '0402, followed by DQw8. RR values in Owerbach et al. are not directly comparable with our RR values from an earlier study (2), because Owerbach et al. calculated RR values with only the DR4+ subjects as the population base. To reflect the importance of the various alleles in the population and to allow direct comparison with our results, we recalculated their RR values in the conven- tional manner, using all patients and control subjects as the population base. The recalculated RR values were 63.3 for DRBIt0402, 11.7 for DQw8, and 3.1 for DRBlt0401. Therefore, these data do not support the contention of a predominant role for the DQBl alleles in DR4 haplotypes.

The article by Owerbach et al. (9) did not present data on combinations of DR and DQ alleles, but Owerbach shared those data in a personal communication. In Owerbach et al. (9), among Houston subjects (all control subjects were from Houston but half of diabetic patients

were from Boston), the overall RR was 10.5 for DQw8; DQw8 haplotypes with DRB1*0401 or 0402 had an RR of 13, whereas DQw8 haplotypes with DRB1*0403-0405 had an RR of only 1.9. Thus the Houston data (9), like ours (2) and those of Nepom et al. (12), show a strong effect of DRBl alleles even among DQw8' haplotypes.

ALTERNATIVE EXPLANATIONS FOR THE APPARENT DR x DQ INTERACTION IN DR4 HAPLOTYPES As summarized above, DR4 haplotypes are strongly associated with IDDM (high RR) only if they have high- risk alleles at both DR and DQ p-chain loci. This obser- vation can be explained in several ways. 1) Certain DRBl alleles (primarily 0401 and 0402) interact positively with DQw8 to promote p-cell destruction. 2) DRBl alleles 0403-0405, found on the remaining DQw8 haplotypes, are somewhat protective. 3) The different DRBl alleles mark haplotypes with distinct DQw8 subtypes. 4) The different DRBl alleles mark haplotypes with distinct alle- les at a third locus, probably in the DQ region, that interact with DQw8 to promote or prevent p-cell destruc- tion.

Hypotheses 3 and 4 seemed unlikely because of previously published results. Sequencing has not re- vealed any microheterogeneity in DQw8 alleles; more- over, all DR4 haplotypes seem to have the same DQAl allele (13), and DQA2 and DQB2 are nearly rnonomor- phic (see ref. 14). We sequenced additional examples of all these loci (14; C. Emler, M.H. Neme de Gimenez, M.J.S., American Red Cross, Madison, WI, unpublished observations) and, despite some previously undetected heterogeneity in DQA2 (1 4), found no DQ-region se- quences to account for the preferential association of certain DR4 subtypes with IDDM. In testing at the func- tional level (K. Lundin, K.E.A. Lundin, J.R. Rowe, E. Thorsby, M.J.S. Rikshospitalet, Oslo; American Red Cross, Madison, WI. unpublished observations), DQw8- specific T-lymphocyte clones could not distinguish among the DQw8 alleles on various DR4 haplotypes.

The sequence data do not rule out significant variation in regulatory sequences or posttranslational modifica- tion. Moreover, much of the sequencing, including our own, was limited to the second exon, corresponding to the putative peptide-binding pocket. Nevertheless, the existence of DRBl sequence variation and failure to find differences elsewhere among haplotypes that vary in their association with diabetes (DRBI *0401-0404) turns our attention back to the DRBl sequences themselves (hypotheses 1 and 2).

I do not favor hypothesis 1 (positive interaction of DRB1*0401 and 0402 with DQw8), because it is not clear how the DR and DQ p-chains would interact to promote diabetes other than to promote immune responses to two distinct antigenic peptides. It seems unlikely that such a mechanism would produce the strong synergy observed between DR and DQ p-chain loci (2; D. Owerbach, unpublished observations).

Hypothesis 2 (varying degrees of protection by the various DR4 subtypes) seems to me the simplest expla- nation for the DR4-DQw8 haplotype data. Certain haplo- types, such as those bearing DQw8 and DRBIt0402,

124 DIABETES, VOL. 41, FEBRUARY 1992

may confer maximum risk simply because they have minimally protective alleles at both DR and DQ loci. Dominant immune suppression by MHC class II alleles is well documented (15), and there is good evidence that HLA haplotypes bearing certain subtypes of DR2 and DQwl protect against IDDM (16,17). 1 am proposing a lesser degree of protection by certain DR4 subtypes and by DQw7. According to this "protective" hypothesis, any haplotype will be positively associated with IDDM (RR > 1) in a given population if it is less protective than the average haplotype in that population.

THE "ICEBERG" NATURE OF GENOTYPE-DISEASE ASSOCIATIONS By the same reasoning, regardless of the genetic mech- anisms involved, a given genotype will appear to be diabetogenic, neutral, or protective if it confers above- average, average, or below-average risk compared with other genotypes that exist in that population. Thus, a genotype that appears neutral in western Europe might appear diabetogenic in Japan, where the population risk for IDDM is low. If one cuts off the tip of the iceberg (the most susceptible genotypes), other genotypes will be- come the new tip of the iceberg.

OTHER EVIDENCE THAT PROTECTION FROM IDDM IS FUNCTIONALLY DOMINANT OVER SUSCEPTIBILITY Initially I and possibly many other researchers assumed that certain HLA alleles promote IDDM by actively pro- moting anti-p-cell autoimmunity. However, a "protective" model, wherein certain HLA alleles are associated with IDDM only because they are poorly protective, may be the simplest explanation for many of the data. Functional data, transgenic mouse data, and classical genetic data all reflect the major importance of protection versus lack of protection in determining who develops IDDM.

Several types of functional study suggest that the autoimmunity leading to IDDM is not due to an overly effective immune response but rather to an immune response that is deficient in some way and results in ineffective regulation. For example, humans with IDDM have subnormal rather than supranormal interleukin-2 (IL-2) production (18). Similarly, BB rats are T lym- phopenic (19), contrary to what one would expect if their immune response were excessively robust. NOD mice, although having a normal number of T lymphocytes, may be "functionally T lymphopenic" (E. Leiter, Jackson Lab- oratory, Bar Harbor, ME, personal communication). NOD mice have subnormal autologous mixed-lymphocyte re- actions and generate subnormal amounts of suppressor activity in autologous mixed-lymphocyte reactions or in response to concanavalin A (ConA) stimulation (20,21). Thymocytes from young NOD mice are deficient in their response to T-lymphocyte mitogens ConA and anti-CD3 (22; E. Leiter, unpublished observations). Diabetes inci- dence is reduced in NOD mice and BB rats by injection of tumor necrosis factor a (23,24), and in NOD mice only by injection of IL-2 or poly-IC (25), all of which are expected to upregulate the immune response.

One interpretation of certain transgenic mouse studies is that IDDM is at least partly due to specific holes in the

antigen-presenting repertoire. Like several other mouse strains, the NOD mouse lacks I-E expression because its I-Ea gene (DRa homolog) is defective (26). In all reported cases (28-30), an I-Ea transgene was protective. A specific in vitro-modified I-Ap allele (DQp homolog) was protective as a transgene (30), although other I-Ap transgenes were not protective (28,31), perhaps be- cause of inefficient pairing with the endogenous I-Aa. In two reports, protection by I-A was achieved with an d p transgene pair (31,32). In each case, the transgenes were protective despite the presence of the endogenous I-Aa and -p alleles of the NOD mouse (i.e., protection was dominant).

I suggest that the protective MHC alleles and d p allele pairs actively promote p-cell tolerance. However, these data and the classical genetic and bone marrow trans- plant data cited below are also compatible with the Nepom model (I) , which proposed that the protective H-2 molecules outcompete the NOD mouse's H-2 mole- cules for limiting amounts of a diabetogenic peptide but present the peptide in a harmless way.

In BB rat (33,34) and NOD mouse (reviewed in ref. 35) breeding studies, non-diabetes-associated MHC and non-MHC alleles were dominant or nearly dominant over the diabetes-associated alleles. Similarly, human DR3 haplotypes behave as recessive in IDDM (36). Recessive inheritance implies a deficiency, not an excess, in the amount or the functional level of a given gene product. In contrast, DR4 haplotypes seem to have additive or dominant effects on IDDM (36). However, the apparent dominance of DR4 haplotypes could possibly be an illusion if, as some studies indicate, the diabetogenic DR4 haplotype is preferentially inherited from the father (reviewed in ref. 37).

NOD mice were protected by bone marrow transplants from N O D . H - Z ~ O ~ congenic mice (0 of 10 mice diabetic at 25 wk old) but not from NOD mice (9 of 12 diabetic at 25 wk old) (38). The H-ZNoN congenic cells seem to be functionally dominant, as indicated by transplantation of a 50:50 mixture of NOD and NOD. H-ZNoN congenic cells (1 of 14 diabetic, more like the protective congenic cells than like the nonprotective NOD cells) (38).

DR PCHAIN STRUCTURAL FEATURES CORRELATING WITH IDDM SUSCEPTIBILITY/PROTECTlON Earlier in this article, I attempted to list the five DR4 subtypes in order from most to least correlated with IDDM. To reveal any possible structural correlates, we listed those alleles in that order with their first-domain amino acid sequences (the 1 st domain being the most polymorphic and probably the peptide-binding domain [39]). All sequences mentioned can be found in a review by Marsh and Bodmer (40) and other sources.

In our study (2) and that of R~nningen et al. (1 O), allele 0403 was the only DR4 subtype with an RR < 1. That allele, putatively the most protective DR4 allele, is the only one with glutamic acid instead of alanine at position 74. Thus, within the context of DR4, Glu74 may contribute to protection. The three alleles at the bottom of the list, 0403-0405, are the only ones with arginine at position 71. Therefore, Arg71 may also contribute to protection

DIABETES, VOL. 41. FEBRUARY 1992 125

within the context of DR4. Allele 0401 has a relatively conservative substitution of lysine (also basic) at position 71, whereas 0402, with the strongest IDDM association, has a more significant replacement by glutamic acid (acidic). Allele 0402 is also the only DR4 subtype with Ile instead of Leu at position 67 and Asp instead of Gln at position 70.

Thus, of the six amino acid positions where these DR4 subtypes differ, a cluster of four from positions 67-74 correlate with the degree of IDDM susceptibility. These structural features support the concept that DRBl alleles directly influence IDDM susceptibility rather than being markers for unknown genetic variation at another locus.

Unlike in white populations, in which DQw8 has con- sistently been associated with IDDM (3), DQw8 is not diabetes associated in Japan (see refs. cited in ref. 3). The lack of IDDM association with DQw8 in Japan may be due to a protective effect of the linked DRBl alleles 0403 and "DwKT2" (1 1,41).

Like the DR4 subtypes, the DR2 subtypes are easily distinguished by T lymphocytes, are difficult to distin- guish serologically, and differ greatly in their effects on IDDM. The correlation of Arg71 with protection from IDDM, discussed above for the DR4 subtypes, is also seen for the DR2 subtypes. Likewise, the correlation of lle67 with higher susceptibility recurs in the DR2 sub- types. However, these are only two of six amino acid positions where DR2 subtypes 1501 and 1502 (which correlate with strong protection from IDDM) are alike but differ from DR2 subtype 1601 (not associated with pro- tection). Any or all of the six amino acids could account for the different IDDM associations. Again, however, the structural data are consistent with a direct role for DRBl alleles in IDDM susceptibility.

An analogous situation exists for the DQBl alleles of the same DR2, DQwl haplotypes. Asp57 of DQB1, which roughly correlates with protection from IDDM (42,43), correlates with protection in the DR2 haplotypes, but it is only one of eight amino acids (Metl4, Tyr30, Asp57, Thr71, Glu74, Lys75, Arg77, Phe87) that are unique to the putatively more-protective alleles.

Because the alleles at the DQB1 and DRBl loci are in strong linkage disequilibrium on DR2 haplotypes of whites (3), one cannot distinguish between a protective effect of DR, DQ, or a combination of the two. In north India, where these DQ and DR alleles are found on separate haplotypes, both haplotypes appear to be protective (RRs of 0.1 and 0.26, respectively; 44), al- though less so than when they appear together in whites (RR = 0.04; 44). 1 suggest that specific DR2 haplotypes of whites may be strongly protective precisely because they have protective alleles at both DR and DQ loci.

Comparing the DR2 alleles as a group (strongly neg- ative to neutral associations with IDDM) with the DR4 alleles as a group (weakly negative to strongly positive associations with IDDM), one finds that the two groups of alleles differ at several sites near the amino terminus (positions 9-37).

Of the sequence features noted for the DR4 and DR2 subtypes, the correlation of Arg71 with relative protection seems to be the most consistent when one looks at the

remaining DRB1 alleles. All in all, this correlation seems to be no better or worse than that described previously for Asp57 of DQBl . It correlates with lower susceptibility in most cases; the exceptions can be rationalized in various ways, it is not an absolute but a relative effect, and other amino acids clearly play important roles as well.

For DQA1, DQBI, and DRBI, the data on IDDM susceptibility are limited to correlations and to analogies with mice. More-direct data may be obtainable if NOD mice are made transgenic with various HLA or H-2-HLA hybrid genes. As described earlier, some experiments have been done for the murine I-A loci (DQ homologs) and I-Ea (DRa homolog). Such experiments will be more difficult for DRBl or its murine homolog I-Ep, because the NOD mouse must be made I-EaC (its I-Ea gene is nonfunctional [26]), and its endogenous I-Ep must be eliminated, perhaps by homologous recombination in embryonic stem cells.

A MODEL IN WHICH THE HLA CLASS II GENES AFFECT IDDM SUSCEPTIBILITY MAINLY VIA INDUCTION1 MAINTENANCE OF TOLERANCE From human, rat, and mouse studies summarized in this article, it appears that class II dimers negatively associ- ated or weakly positively associated with IDDM are generally dominant or epistatic over those that are more positively associated with IDDM. (Epistasis is analogous to dominance but acting between alleles at different loci.) Thus, I suggest that, despite the probable role of class 11-restricted CD4+ and class I-restricted CD8+ T lym- phocytes in p-cell destruction, the genetic effect of class II alleles occurs mainly at the level of induction or maintenance of immunological tolerance. There are sev- eral levels at which this could occur.

A priori, the protection could occur 1) in the thymic medulla during negative selection of T lymphocytes reactive with self-MHC; 2) in the thymic cortex during positive selection of T lymphocytes recognizing "modi- fied self" (self-MHC plus foreign antigens); or 3) extrathy- mically, perhaps via "clonal paralysis" or stimulation of regulatory T lymphocytes. Some experimental data argue against the first two possibilities. In a transgenic mouse study, apparently normal I-E expression in thymic me- dulla was insufficient to prevent IDDM (29). Similarly, in a transplantation study (38), IDDM could not be prevented by transplantation of a thymus from a resistant strain depleted of bone marrow-derived cells, although the positive selection phase seems to be due to class II+ thymic epithelial cells. Thus, an extrathymic process seems the most probable. I favor the hypothesis of active suppression by regulatory T lymphocytes for several reasons, including the dominance of nonsusceptibility genotypes.

Thus, the genetic effects of HLA class II dimers on IDDM susceptibility might occur as outlined in Fig. 1. In this model, there is a potential tolerogen, perhaps a fragment of a p-cell-derived protein that is actually tolerogenic only if the individual has one or more MHC class II dimers that can bind and present it effectively. A class II dimer that binds the peptide strongly and in the

126 DIABETES, VOL. 41, FEBRUARY 1992

Tolerogen

STRONG PROTECTION: RR < 1

NO PROTECTION: RR > 1

WEAK PROTECTION: RR near 1

COMPETITION, NO PROTECTION: RR >> 1

FIG. 1. A model In which HLA class I1 alleles affect Insulln-dependent diabetes mellltus susceptlblllty prlmarlly via blndlng and presentation of a pptlde wlth the potentlal to Induce lmmunologlcal tolerance to pancreatic pcells. In this model, class I1 dlmers that bind the tolerogen strongly and In the proper orlentatlon (a) are strongly protectlve, whereas those that blnd It weakly (b) or not at all (c) are correspondlngly less protectlve. Flnally, some class I1 dlmers mlght blnd the tolerogen tlghtly but In the wrong orlentatlon or conformation (d). Thla last situation (d) would also be nonprotectlve; competition by such dlmers would be of functlonal slgnlflcance only If amount of tolerogen Is Ilmltlng. The text Includes a dlscusslon of posslble protective mechanisms, as well as a ratlonale for suggestlng that It Is a matter of extra-thymic peptlde presentatlon to CD4+ tolerance-malntalnlng T cells.

proper orientation (Fig. la) , stimulates the appropriate regulatory T lymphocytes, and is strongly protective. Class II dimers that bind the peptide weakly (Fig. 1 b) or not at all (Fig. 1 c) offer correspondingly lesser degrees of p-cell protection. The case in Fig. 1 dmight also exist: a dimer that competes strongly for the tolerogen but presents it in a nontolerogenic orientation or configuration. Even if this occurs, the resulting competition would only be of practical importance if the concentration of tolerogen were limiting.

This model implies a dominance hierarchy among class II dimers: negatively associated (strongly protec- tive) > "neutral" (weakly protective) > positively associ- ated (nonprotective). The various protective effects would likely be cumulative. For example, a person with a protective DR dimer and a protective DQ dimer or with two protective DQ dimers would probably be less sus- ceptible than someone with one protective dimer.

CAN A "PROTECTIVE" MODEL ACCOUNT FOR THE EXCESS OF DR314 HETEROZYGOTES? The excess of DR3/4 heterozygosity in IDDM suggests synergistic effects of DR3 and DR4 haplotypes, and the

evidence of IDDM-associated DQAl alleles (4546) and specific DQAlIDQBl combinations in cis or trans (1 0,47, 48) supports the hypothesis of trans complementation between DQa and DQp chains on DR3 and DR4 haplo- types. It is a bit simpler to explain such trans complemen- tation in terms of autoaggression (effective presentation of a diabetogenic peptide) than in terms of lack of protection. However, in the protective model, one could propose that DR3 and DR4 haplotypes are synergistic because a com- plementing DQdp dimer competes strongly for binding of a tolerogenic peptide but presents it in a nontolerogenic manner (Fig. 1 d). Another possibility, perhaps less likely, is that DQa and DQp chains from DR3 and DR4 haplotypes preferentially form trans dimers, and that those are even less protective than the cis dimers.

Rubinstein et al. (37) argued that the excess of DR3/4 heterozygotes need not be interpreted as synergy at all. They pointed out that the DR3/4 excess does not exist in all populations and suggested that an excess of DR3/4 and deficit of DR4/4 patients could be due to the sup- posed paternal bias whereby patients' DR4 haplotypes more often come from their fathers (refs. cited in ref. 37).

DIABETES, VOL. 41, FEBRUARY 1992

Whatever the mechanism, a paternal bias in DR4 trans- DR-DQ allele combinations from the Houston IDDM pop- mission as an explanation for the DR3/4 excess would be ulation study. I am also grateful for stimulating discus- equally consistent with various models. sions with Drs. Edward Leiter and Malcolm Simons.

CONCLUDING REMARKS REFERENCES Current models of the genetics of IDDM must include at 1. Nepom GT: A unified hypothesis for the complex genetics of HLA

associat~ons with IDDM. Diabetes 39:1153-57, 1990 least three HLA loci: DQA1 DQB1 and DRB1 ' and 2. Sheehy MJ, Scharf SJ, Rowe JR, Neme de Gimenez MH, Meske LM, possibly DPB1 as well (49). Considering all available Erlich HA, Nepom BS: A diabetes-susceptible HLA haplotype is data, DQAl, DQBl, and DRBl seem to have similar best defined by a combination of HLA-DR and -DQ alleles. J Clin

lnvest 83:830-35, 1989 On lDDM their relative 3, Todd JA: Genetic control of autoimmunity in type 1 diabetes,

importance varies considerably from population to pop- lmmunol Today 11 :122-29, 1990 ulation and from study to study. In my opinion, the major 4. Christy M, Green A, Cristau B, Kromann H, Nerup J, Platz P,

Thomsen M, Ryder LP, Svejgaard A: Studies of the HLA system and source of this variation is population structure (differing insulin-dependent d,abetes, Diabetes c ~ ~ ~ ~ : ~ ~ ~ - ~ ~ , 1979 linkage disequilibria, allele frequencies, and the pres- 5. Suciu-Foca N, Rubinstein P. Nicholson J, Susinno E, Weiner J,

ence or absence of specific alleles), For example, the Godfrey M, Hardy M, Rayfield E, Reemtsma K: Juvenile diabetes mellitus and the HLA system. Transplant Proc 11:1309-13, 1979

infhence of DR4 subtypes (DRBl locus) is quite evident 6, Platz P, Jakobsen BK, Morling N, Ryder LP, Svejgaard A, Thomsen in populations with substantial frequencies of DRB1'0402 M, Christy M, Kromann H, Benn J, Nerup J, Green A: HLA-D and

and 0403, which have very different effects on IDDM -DR antigens in genetic analysis of insulin-dependent diabetes mellitus. Diabetologia 21 :108-15, 1981

susce~tibility, and less evident in D O D U ~ ~ ~ ~ O ~ S lackinn 7. Bach FH. Rich S. Barbosa J. Seaall M: Insulin-deoendent diabetes either'or both of those alleles. si&ila;ly, the associated H L A ~ region encGded determinanis. Hum lmmunol

1259-64 1985 imp0rtance of DQA1 Or DQB1 On linkage dis- 8. Rowe JR,' Mickelson EM, Hansen JA, MacDonald MJ, Allen CI, equilibria and on the existence in the population of Gabbay KH, Yunis EJ, Sheehy MJ: T cell-defined DR4 subtypes as

markers for type 1 d~abetes. Hum lmmunol22~51-60, 1988 'pecific alleles at the complementa~ locus (DQB1 Or 9, Owerbach D, Gunn S, Gabbay KH: Primary association of HLA.

DQAl, respectively). Some other Sources of interstudy DQw8 with type I diabetes in DR4 patients. Diabetes 38:942-45, variation are sampling error (due to finite sample sizes) 1989

and inadequately matched patient and control groups, 10. Rnnningen KS, Spurkland A, Iwe T, Vartdal F, Thorsby E: Distribution of HLA-DRBl , -DQA1 and -DQBl alieles and DQAl -DQBl geno-

A reminder may be warranted to recall why this dis- types among Norwegian patients with insulin-dependent diabetes cussion of HLA and IDDM is biased so heavily toward mellitus. Tissue Antigens 37:105-11, 1991

11. Awata T, Kuzuya T, Matsuda A, lwamoto Y, Kanazawa Y: Contribu- degrees protection. Until my Own thinking tion of HLA-DRB1 genes to susceptibility to insulin-dependent was dominated bv susce~tibilitv alleles, with onlv a rare diabetes mellitus in Japanese (Abstract). Diabetes 40 (Suppl. afterthought to the app&ent protection by D R ~ D Q W ~ haplotypes. However, in diverse populations, the nega- tive association of IDDM with DR2 is a more common feature than the positive associations with DR3 and DR4. The shift of focus was precipitated by the DR4 haplotype data discussed earlier. Although both DQ and DR p-chain loci have alleles positively associated with IDDM, in each case, this positive association can be nullified by non-IDDM-associated alleles at the other locus (2); i.e., lack of susceptibility was dominant. In addition, all func- tional, transgenic, and classical rat and mouse genetic studies, summarized earlier, suggested that protection is functionally dominant. Although CD4+ class Il-restricted T lymphocytes seem to be important in the effector phase of p-cell destruction, I suggest that the genetic determi- nation of IDDM susceptibility by class II alleles occurs primarily at an earlier stage, affecting the individual's success at maintaining p-cell-specific tolerance.

The model presented in this paper is almost certainly an oversimplification but it may suffice to stimulate addi- tional thounht and discussion.

. . 1):153A, 1991 Nepom GT, Robinson D, Palmer J, Seyfried C, Byers P, Knitter-Jack N, Nepom B: Molecular topography of the DQ3.2 gene associated with IDDM. In Diabetes 1988. Larkins R, Zimmet P, Chisholm D, Eds. Amsterdam, Excerpta Med., 1989, p. 375-78 Gregersen PK, Shen M, Song Q-L, Merryman P, Degar S, Seki T, Maccari J, Goldberg LD, Murphy H, Schwenzer J, Wang CY, Winchester RJ, Nepom GT, Silver J: Molecular diversity of HLA-DR4 haplotypes. Proc Natl Acad Sci USA 83:2642-46, 1986 Rowe JR, Neme de Gimenez MH. Emler CA, Sheehy MJ' HLA- DQA2 (DX alpha) polymorphism and insulin-dependent diabetes. Hum lmmunol29:256-62, 1990 Sasazuki T, Kikuchi I, Hirayama K, Matsushita S, Ohta N, Nishimura Y: HLA-linked immune suppression in humans. immunology (Suppl.) 2:21-24, 1989 Thomson G: HLA-DR antigens and susceptibility to insuiin-depen- dent diabetes mellitus. ~m J Hum ~ e n e t 36:1309-17, 1984

'

Baisch JM, Weeks T. Giles R, Hoover M, Stastny P, Capra JD: Analysis of HLA-DQ genotypes and susceptibility in insulin-depen- dent diabetes mellitus. N Engl J Med 322:1836-41, 1990 Zier KS, Leo MM, Spielman RS, Baker L: Decreased synthesis of interleukin-2 (IL-2) in insulin-dependent diabetes mellitus. Diabetes 33:552-55, 1984 Mordes JP, Desemone J, Rossini AA: The BB rat. Diabetes Metab Rev 3:725-50, 1987 Serreze DV, Leiter EH: Defective activation of T suppressor cell function in non-obese diabetic mice, potential relation to cytokine deficiencies. J lmmunol 140:3801-807, 1988 Buschard K. Madsbad S. Rvaaard J: De~ressed sumressor cell - . ,., ~

activity in patients with newly diagnosed ihsulinldepe'n'dent diabe- tes mellitus. Clin Exo lmmunol 4125-32. 1980 - - - - . . -

ACKNOWLEDGMENTS 22. Ziprisb, ~azarus AH, C ~ O ~ A R , Hadzija M, Delovitch TL: Defective thymic T cell activation by concanavalin A and anti-CD3 in autoim- The leading to this discussion was supported in part mune nonobese diabetic mice. evidence for thymic T cell anergy

bv National Institutes of Health Grant DK-38209. Juvenile that correlates with onset of insulitis. J lmmunol 146 3763-71,

'Iabetes Foundation Grant 1876991 and a Matching 23. A:?ih J, Seino H, Abo T. Tanaka S. Shintani S Ohta S. Tamura K, Grant and an Established Investigator Award from the Sawai T. Nobunaaa T. Oteki T. Kumaaai K. Tovota T: Recombinant -

American Red Cross. human tumor necyosis factor a'suppr~ssesautbimmune diabetes in NOD mice. J Clin lnvest 84:1345-48 1989 I thank Dr. David Owerbach College Of Medi- 24. Satoh J, Seino H, Shintani S, ana aka S, Ohteki T. Masuda T.

cine, Houston, TX) for generous contribution of data on Nobunaga T, Toyota T: Inhibition of type 1 diabetes in BB rats with

128 DIABETES, VOL. 41, FEBRUARY 1992

recombinant human tumor necrosis factor-alpha. J lmmunol 145: 1395-99, 1990

25. Serreze DV, Hamaguchi K, Leiter EH: lmmunostimulation circum- vents diabetes in NODILt mice. J Autoimmunity 2:759-76, 1989

26. Hattori M, Buse JB, Jackson RA, Glimcher L, Dorf ME, Minami M, Makino S, Moriwaki K, Kuzuya H, lmura H, Strauss WM, Seidman JG, Eisenbarlh GS: The NOD mouse: recessive diabetogenic gene in the major histocompatibility complex. Science 231 :733-35. 1986

27. Nishimoto H, Kikutani H, Yamamura K, Kishimoto T: Prevention of autoimmune insulitis by expression of I-E molecules in NOD mice. Nature (Lond) 328:432-34, 1986

28. Uehira M, Uno M, Kurner T, Kikutani H, Mori K, lnomoto T, Uede T, Miyazaki J, Nishimoto H, Kishimoto T, Yamamura K: Development of autoimmune insulitis is prevented in Ead but not in Apk NOD transgenic mice. Int lmmunol 1 :209-13. 1989

29. Bohme J, Schuhbaur B, Kanagawa 0 , Benoist C, Mathis D: MHC- linked protection from diabetes dissociated from clonal deletion of T cells. Science 249:293-95, 1990

30. Lund T, O'Reilly L, Hutchings P, Kanagawa 0 , Sirnpson E, Gravely R, Chandler P, Dyson J, Picard JK, Edwards A, Kioussis D, Cooke A: Prevention of insulin-dependent diabetes mellitus in non-obese diabetic mice by transgenes encoding modified I-A p-chain or normal I-E a-chain. Nature (Lond) 345:727-29, 1990

31. Miyazaki T, Uno M, Uehira M, Kikutani H, Kishirnoto T, Kimoto M, Nishimoto H, Miyazaki J, Yamamura K: Direct evidence for the contribution of the unique I-ANoD to the development of insulitis in non-obese diabetic mice. Nature (Lond) 345:722-24, 1990

32. Slattery RM, Kjer-Nielsen L, Allison J, Charlton B, Mandel TE. Miller JFAP: Prevention of diabetes in non-obese diabetic I-Ak transgenic mice. Nature (Lond) 345:724-26, 1990

33. Colle E, Guttmann RD, Seemayer T: Spontaneous diabetes mellitus syndrome in the rat. J Exp Med 154:1237-42, 1981

34. Jackson RA, Buse JB, Rifai R, Pelletier D, Milford EL, Carpenter CB, Eisenbarth GS, Williams RM: Two genes required for diabetes in BB rats: evidence from cyclical intercrosses and backcrosses. J Exp Med 159:1629-36, 1984

35. Leiter EH, Serreze DV, Prochazka M: The genetics and epidemiol- ogy of diabetes in NOD mice. lmmunol Today 1 1 : 147-49, 1990

36. Thomson G: Investigation of the mode of inheritance of the HLA associated diseases by the method of antigen genotype frequen- cies among diseased individuals. Tissue Antigens 21 31-1 04, 1983

37. Rubinstein P, Walker M, Mollen N, Carpenter C, Beckerman S, Suciu-Foca N, McEvoy R, Ginsberg-Fellner F: No excess of DRt3/4

in Ashkenazi Jewish or Hispanic IDDM patients. Diabetes 39:1138- 43, 1990

38. Serreze DV, Leiter EH: Development of diabetogenic T cells from NOD/Lt marrow is blocked when an allo-H-2 haplotype is expressed on cells of hemopoietic origin, but not on thymic epithelium. J lmmunol147:1222-29, 1991

39. Brown JH, Jardetzky T, Saper MA, Samraoui B, Bjorkman PJ, Wiley DC: A hypothetical model of the foreign antigen binding site of class II histocompatibility molecules. Nature (Lond) 322:845-50, 1988

40. Marsh SGE, Bodmer JG: HLA-DR and -DQ epitopes and monoclo- nal antibody specificity. lmmunol Today 10:305-12, 1989

41, lkegarni H: Ethnic differences in IDDM class II associations (oral presentation). Int Diabetes Fed. Congr, 14th, Washington, DC, 23-28 June 1991

42 Todd JA, Bell JI, McDevitt HO: HLA-DQp gene contributes to susceptibility and resistance to insulin-dependent diabetes mellitus. Nature (Lond) 329:599-604, 1987

43. Horn GT, Bugawan TL, Long C, Erlich HA: Allelic sequence variation of the HLA-DQ loci: relationship to serology and insulin-dependent diabetes susceptibility. Proc Natl Acad Sci USA 85:6012-16, 1988

44. Fletcher J, Odugbesan 0 , Mijovic C, Mackay E, Bradwell AR, Barnett AH: Class II HLA DNA polymorphisrns in type 1 (insulin- dependent) diabetic patients of Norlh Indian origin. Diabetologia 31 :343-50, 1988

45. Todd JA, Fukui Y, Kitagawa T, Sasazuki T: The A3 allele of the HLA-DQA1 locus is associated with susceptibility to type 1 diabetes in Japanese. Proc NaH Acad Sci USA 87:1094-98, 1990

46. Todd JA, Mijovic C, Fletcher J, Jenkins D, Bradwell AR, Barnett AH: Identification of the susceptibility loci for insulin-dependent diabe- tes mellitus by trans-racial gene mapping. Nature (Lond) 338:587- 89, 1989

47. Khalil I, d'Auriol L, Gobet M, Morin L, Lepage V, Deschamps I, Park MS, Degos L, Galibert F, Hors J: A combination of HLA-DQ beta Asp57-negative and HLA DQ alpha Arg52 confers susceptibility to insulin-dependent diabetes mellitus. J Clin lnvest85: 131 5-1 9, 1990

48. Rsnningen KS, Gjertsen HA, Iwe T, Spurkland A, Hansen T, Thorsby E: Particular HLA-DQ d p heterodimer associated with IDDM sus- ceptibility in both DR4-DQw4 Japanese and DR4-DQwB/DRw8- DQw4 whites. Diabetes 40:759-63, 1991

49. Easteal S, Kohonen-Corish MRJ, Zimmet P, Serjeantson SW: HLA-DP variation as additional risk factor in IDDM. Diabetes 39: 855-57, 1990

DIABETES, VOL. 41, FEBRUARY 1992