CA2246487A1 - Dna assay for the prediction of autoimmune diabetes - Google Patents

Abstract

The present invention relates to a novel DNA
assay for the diagnosis and/or prediction of autoimmune diabetes. The present invention relates to a DNA assay for the prediction of autoimmune diabetes in human subjects, which comprises the steps of a) obtaining a DNA sample from the subject and PCR
amplification of class III alleles of variable number of tandem repeats (VNTR) located upstream of the insulin gene (INS); and b) electrophoretic co-migration of amplified fragment to identify class III
alleles of the insulin VNTR which are identical to S1 or S2, from alleles which differ from S1 and S2;
whereby identification of S1 or S2 allele is indicative of at least 1% risk of autoimmune diabetes.

Description

DNA A S F R E PR TI N OF I D E
HACKGRO(I~ OF TINVEIyITION
(a) Fiead of the Invention The invention relates to a novel DNA assay for the diagnosis and/or prediction of autoimmune diabetes.
(b) Des rigtion of Prior Art Diabetes is a major cause of morbidity and mortality in industrialized societies. It has been estimated that one of every seven health-care dollars goes to treating diabetes and its complications. Type 1 diabetes (also called insulin-dependent or juvenile diabetes, henceforth referred to in this document as "diabetes") is due to the autoimmune destruction of the insulin-producing pancreatic (3-cells. Type 1 diabetes is less common than type 2, accounting for only 10-20~
of cases in Caucasians. However, because it is much more severe and starts much earlier in life, it accounts for a large proportion of diabetes-related morbidity and mortality.
Type 1 diabetes involves autoimmune destruction of the insulin-producing pancreatic ~i-cells. Insulin, an autoantigen in this process, is expressed in human thymus at levels dependent on alleles at the upstream INS VNTR, to which the IDDM2 susceptibility locus has been mapped. Chromosomes carrying the dominantly protective (Bennett ST et al., 1995, Nat. Genet.
9(3):284-292), long INS VNTR alleles (class III) produce 2-3 times higher levels of insulin gene (INS) mRNA than those with predisposing, short class I
alleles (Vafiadis P et al., 1997, Nat. Genet.
15(3):289-292; Pugliese A et al., 1997, Nat. Genet.
15 (3) :293-297) .
Preveatio~ of diabetes It is estimated that by the time symptoms of diabetes appear, more than 95°s of the (3-cell mass has been destroyed. Given the irreversibility of this destruction, the most promising approach to the disease is prevention. This will require an intervention, at some time before symptoms appear, aimed at modulating the immune system to prevent the antigen-specific autoimmune reaction. Although the specific causative autoantigen(s) in diabetes is (are) not known, insulin, the main product of the a-cell, appears to be an autoantigen of major importance.
Based on this, the DPT (diabetes prevention trial) a large study, is now underway in the United States.
Insulin injections or oral insulin administration is used in the hope of helping individuals at risk for diabetes acquire immune tolerance to insulin. The results of this study will not be known for a number of years. It is possible that it will be superseded by other studies, which will be based on more precise scientific rationale provided by current spectacular advances in immunology. There is a tangible likelihood that in the coming decade a safe and effective method of preventing or reversing the diabetes autoimmune process will have been found.
Identification of individuals at risk An effective intervention to prevent diabetes is very unlikely to be inexpensive, safe and convenient enough to be applied to the general population (i.e. in the fashion of infectious disease vaccines). The intervention will most likely need to be targeted to individuals that can be identified as being at a substantial risk for diabetes.
Autoantibodv t~sting The DPT and similar trials have focused on first-degree relatives of diabetics that are positive for autoantibodies known to predict diabetes. These autoantibodies against protein components of the ~-cell become positive at least a year or two before the onset of clinical diabetes. However, even if diabetes is prevented with 100~s efficacy in all first-degree relatives of diabetics, this will only abolish less than 10% of new cases of diabetes, as the majority of individuals with diabetes do not have a previously affected first-degree relative. To be meaningful, an effective prevention will have to be applicable to the general population. Screening the general population for antibodies has been shown to be feasible in practice, and predictive of diabetes. However, since antibodies only become positive at a finite time point before the onset of diabetes, most people destined to become diabetic at some future point will be negative when tested early in life. Therefore, antibody screening of the general population will need to be repeated at intervals of 1-2 years, a totally impractical proposition.
Genetic testing An alternative way of predicting high risk for diabetes is through DNA testing. It is known that genetic predisposition plays a major role in diabetes.
Identical twins of individuals with diabetes have a risk that ranges from 30 to >60~ in various studies, as opposed to fraternal twins, that are concordant only in 5-10~ of the cases (Bennett ST et al., 1996, Ann. Rev.
Genet. 30:343-370). The predisposition is not inherited in a Mendelian fashion, which means that this complex phenotype requires predisposing genetic material involving more than a single gene. Linkage studies have identified as many as 16 different genetic loci that are potentially linked to diabetes (in genetics, the term "locus" is used instead of "gene", when only a location in the gename is known, but the gene[s]
involved remain [s] to be identified) . Although many of these loci will probably turn out to be statistical artifacts, it is clear that in order to define the "diabetes genotype", markers at more than one loci will have to be typed. A good definition of the "diabetes genotype" would be: a combination of alleles that correctly predicts diabetes with a probability that approaches that predicted from being the identical twin of a diabetic. Obviously, a risk of at least 30~ would justify preventive intervention.
So far only two of the several loci have been defined precisely enough to be used to this end. They are termed IDDM1 and IDDlul2, and can be used already to obtain a certain degree of risk estimation (Bennett ST
et al., 1996, Ann. Rev. Genet. 30:343-370). Using similar approaches several more loci can be likewise defined, that together predict diabetes with specificity that approaches that of the idealized diabetes genotype. Specificity in a diagnostic test is the percentage of positive tests that turn out to be true. The twin studies show that 30-60$ is the best specificity that can be achieved with DNA testing, but even a specificity of 10~ or less can be very useful depending on how safe, simple and inexpensive future preventive interventions turn out to be.
Once defined, the loci could be genotyped on a few drops of blood obtained at birth as part of neonatal screening programs currently in place for other diseases. Those individuals who exceed a certain threshold of risk can then be followed with antibody testing or treated prior to the appearance of autoantibodies, depending on what the optimal strategy will be determined to be.
Conventfona~ testing tie ~DD1~2 locus It involves determination of the VNTR by Southern blotting, a cumbersome technique that is not easily amenable to miniaturization and automation (Bennett ST et al., 1995, Nat. Genet. 9(3):284-292).
Polymerase chain reaction (PCR) has been used to identify class I alleles, but there is no published report of successful amplification of class III
alleles.
More importantly, the conventional Southern blotting approach only allows classification of individuals as having no class III allele (genotype:
I/I) or having at least one class III allele (genotype I/III or III/III). Individuals in the latter category will be assigned a risk that is approximately 4-fold less (Bennett ST et al., 1996, Ann. Rev. Genet. 30:343-370), and this estimate will be entered in a formula along with information from other IDDM loci, in order to calculate risk. The risk assignment can be erroneous in individuals carrying the specific alleles S1 or S2.
It would be highly desirable to be provided with a novel DNA assay for the diagnosis and/or prediction of autoimmune diabetes which assay overcomes the drawbacks of the prior art.
SUbB~iARY "OF TIC ~~1V8NTION
One aim of the present invention is to provide a PCR-based method to distinguish not only between class I and class III alleles of the insulin VNTR, but also between different alleles within class III.
Specifically, the method of the present invention can reliably distinguish alleles in class III
that are identical to S1 or S2, from those that are different from either of them. This distinction is very important, as the present genetic studies show that S1 and S2 are predisposing to diabetes, while all other class III alleles are protective.

_ 6 ComQ.nercial applications of the DNA assay The DNA assay of the present invention can be easily adapted to a miniaturized, automated genotyping approach utilizing fluorescent labeling. This genotyping will be an important part of a panel of genotypes that will determine diabetes risk.
Advantages and iaaprovemwents over existing technology The assay of the present invention is PCR-based, which makes it ideal for miniaturization and automation. The applicants are the first to develop a protocol for the successful amplification of class III
insulin VNTR, technically a challenging task, as it involves amplification of highly repetitive GC-rich fragments of 2-3 kb length, well beyond the sizes handled by conventional PCR.
It distinguishes S1 and S2 alleles among class I I I , and assigns a higher rather than a lower risk to them, which will improve both the sensitivity and the specificity of the method. This is the main advantage of the assay of the present invention.
To illustrate the importance of the assay of the present invention, the following must be considered:
The applicants have examined 167 diabetic children. Of these, 16 had a paternally transmitted S1 or S2 allele (high-risk class III). The conventional method, unable to distinguish S1 and S2 alleles, would have assigned to them the four-fold lower risk associated with class III as a whole. The assay of the present invention assigns them the true risk conferred by these alleles, which is 4.6-fold higher.
Thus, 16 of 167 children, almost 10%, would have been erroneously given a risk sixteen-fold lower than appropriate. Ten per cent of the population may not seem to be much, but it must be born in mind that in a complex disease like diabetes, a universal risk determination must be pieced together from evaluation of small effects at each locus. For this reason, the assay of the present invention is very likely to become an indispensable part of any attempt to predict diabetes.
In accordance with the present invention there is provided a DNA assay for the prediction of autoimmune diabetes in human subjects, which comprises the steps of a) obtaining a DNA sample from the subject and PCR
amplification of class III alleles of variable number of tandem repeats (V'NTR) located upstream of the insulin gene (INS); and b) electrophoretic co-migration of amplified fragment to identify class III alleles of the insulin VNTR which are identical to S1 or S2, from alleles which differ from S1 and S2;
whereby identification of S1 or S2 allele is indicative of at least 1~ risk of autoimmune diabetes, which is >10 fold higher than that erroneously predicted by existing methods of the prior art.
The INS VNTR is composed of a variable number of tandem 14-15 by repeat sequences, with the consensus repeat unit ACA GGGG TGT GGGG.
The amplification of step a) is effected using at least one primer pair selected from the group consisting of VNTR5 (TCAGGCTGGACCTCCAG!GTGCCTGTTCTG)/
VNTR6 (GCTGGTCCTGAGGAAGAGGTGCTGACGA) and VNTR7 (GGCATCT
TGGGCCATCCGGGACTG)/VNTR8 (GCAGGGCGGGGCTCTTTGCGCTG).
The sample is selected from the group consisting of blood, saliva, urine and hair follicle.
The electrophoretic co-migration of step b) is effected using PAGE.

_ g _ The primer pair for PCR amplification of class III alleles of variable number of tandem repeats (VNTR) located upstream of the insulin gene (INS) which comprises VNTR5 (TCAGGCTGGACCTCCAGGTGCCTGTTCTG)/ VNTR6 (GCTGGTCCTGAGGAAGAGGTGCTGACGA) or VNTR7 (GGCATCTT
GGGCCATCCGGGACTG)/VNTRB (GCAGGGCGGGGCTCTTTGCG CTG).
BRIEF D~SC~'=IPTION OF TF~E ,~lRANINQS
Fig. 1 illustrates the amplification products of the INS VNTR PCR protocol for all classes are shown for samples with monoallelic (m) or biallelic (b) INS
expression in thymus;
Fig. 2A illustrates the 5' sequences of the cloned S1, E1 and 814 alleles;
Fig. 2B illustrates the same sequences in a different notation wherein each individual repeat is represented by a letter, as previously described (Bennett ST et al., 1996, Ann. Rev. Genet. 30:343-370), and compared to the sequence of repeats in alleles cloned by Owerbach & Gabbay from a protective haplotype (III-G) and a very protective haplotype (III-A) as previously described (Bennett ST et al., 1996, Ann.
Rev. Genet. 30:343-370); and Fig. 3 illustrates the identification of class III alleles as S1 or S2 type.
DETAILF~? DESCRIPTION OlE fE ~N~T ION
Type 1 diabetes involves autoimmune destruction of the insulin-producing pancreatic (3-cells. Insulin, an autaantigen in this process, is expressed in human thymus at levels dependent on alleles at the upstream INS VNTR, to which the IDDM2 susceptibility locus has been mapped (Bennett ST et al., 1995, Nat. Genet.
9(3):284-292; Hennett ST et al., 1996, Ann. Rev. Genet.
30:343-370). Chromosomes carrying the dominantly protective, long INS ~'N'TR alleles (class III) produce _ g _ 2-3 times higher levels of insulin gene (INS) mRNA than those with predisposing, short class I alleles (Vafiadis P et al., 1997, Nat. Genet. 15(3):289-292;
Pugliese A et al., 1997, Nat. Genet. 15(3):293-297).
Higher thymic INS expression may promote better induction of immune tolerance to (3-cells. However, a few specific class III INS VNTR alleles are associated with complete silencing of thymic INS expression (Vafiadis P et al., 1997, Nat. Genet. 15(3):289-292;
Pugliese A et al., 1997, Nat. Genet. 15(3):293-297).
Our hypothesis predicts that such alleles are predisposing rather than protective. To test this prediction, we examined the distortion of transmission of two such silencing alleles (S1 and 52) from non-diabetic parents to 167 diabetic children, using a novel PCR-based method of restriction fingerprinting.
Transmission of S1+S2 from heterozygous fathers was significantly more frequent than expected, while maternal S1+S2 were transmitted significantly less often, as expected of class III alleles, and as was observed in all other class III alleles from either parental sex. This finding can best be explained by genomic imprinting that silences thymic INS only on paternal chromosomes carrying specific class III
alleles (allele-restricted imprinting).
Early reports suggested the presence on an insuli-like factor in the thymus. The applicants and others demonstrated that the insulin gene (INS) was transcribed and translated in human fetal (Vafiadis P
et al., 1997, Nat. Genet. 15(3):289-292) and postnatal thymus (Pugliese A et al., 1997, Nat. Genet. 15(3):293-297). In the majority of thymus samples selected to be INS VI~J'I'R class I/III heterozygotes, thymic INS
expression was 2-3 times higher from the gene copy linked to the class III allele as compared to the class I allele. In 5 of 22 samples, however, INS expression was completely silenced from the copy linked to the class III allele, while the class I linked copy was expressed. Monoallelic expression has been observed in genes subject to genomic imprinting, which involves silencing of either the paternal or the maternal allele. We showed that the difference in thymic INS
expression levels between class I and III alleles is independent of parental origin, thus ruling out partial imprinting as the cause of this difference (Vafiadis P
et al., 1997, Nat. Genet. 15(3):289-292). However, parental DNA corresponding to the thymus samples with "silencing" class III alleles was not available, so it could nqt be determined whether genomic imprinting may explain this observation.
Insulin is the only known (3-cell specific type 1 diabetes autoantigen. An autoimmune reaction against INS-encoded epitopes may thus result in specific targeting of the pancreatic (3-cells for destruction, and increased thymic INS expression could explain the dominantly protective effect of class III alleles through better tolerance induction, a thymic process that is known to be dose-dependent.
If the above hypothesis that thymic INS
expression has a dose-dependent effect on susceptibility to type 1 diabetes is correct, then alleles associated with silencing of thymic INS
expression would be more predisposing than all other alleles. To test this hypothesis, we have compared the transmission frequency of class III alleles matching two class III alleles associated with silencing of INS
expression (which we call S1 and S2) to all other alleles, from heterozygous mothers and fathers, to diabetic offspring. We predicted that silencing-associated class III alleles of the S1 and S2 type would be transmitted to diabetic offspring more often than other alleles. Furthermore, we predicted that i~
silencing of thymic INS expression is due to genomic imprinting, the transmission distortion of S1 and S2 type class III alleles would be confined to transmissions from only one parental sex. Thus, silencing of INS in the thymus would be due to allele-specific genomic imprinting, similar to that we have described for the human IGF2R gene.
The known diabetes susceptibility loci This is the major diabetes locus, accounting for almost half of the genetic component. It is located on chromosome 6p21, and involves the cluster of class II HLA histocompatibility genes. PCR-based testing exists to identify several alleles, some of which are predisposing to diabetes while others are protective or neutral. Individuals with the protective alleles are extremely unlikely to have the disease, but the predisposing alleles are common in the general population, and the specificity of predicting diabetes based on their presence is low. Additional loci must be determined for meaningful prediction.

This locus has been mapped to chromosome 11p15.5, to a polymorphism consisting of variable number of tandem repeats (VNTR), 0.5 kb upstream of the insulin gene (INS). The number of repeats in each of the two copies of an individual's chromosome 11 can range from 40 to several hundred. Short VNTR alleles (40-60 repeats) are the most common and are classified together as class I. Long alleles (>120 repeats) are called class III and are found in about 20~ of Caucasian chromosomes, while intermediate class II
alleles (60-120 repeats) are extremely rare.

Individuals who have inherited a chromosome with a class III allele from at least one parent have a 3-5 fold less probability of diabetes than those who have a class I allele on both chromosomes. Therefore, class III alleles, as a group, can be considered dominantly protective.
Recently, we demonstrated a putative mechanism for the protective effect of the class III VNTR: We found that the human thymus produces small amounts of insulin, and chromosomes with a class III allele make 2-3 times more insulin mRNA in the thymus than those with class I.
However, in a small number of cases the class III chromosome instead of producing more than the class I, produced no insulin at all! We predicted that those specific class III alleles associated with this paradoxical phenomenon (called S alleles, for "silencing") will not be protective (as expected of class III alleles) but must be predisposing. Indeed studies on 167 diabetic children and their parents confirmed that, although class III alleles as a whole were inherited less often than expected by chance alone, the S alleles were actually transmitted much more often.
To test our hypothesis, we first developed a PCR based method of identifying INS VNTI~ alleles. The INS VNTR is composed of a variable number of tandem 14-15 by repeat sequences, with the consensus repeat unit ACA GGGG TGT GGGG (Bennett ST et al., 1996, Ann. Rev.
Genet. 30:343-370). In Caucasians, most alleles are either in the class I (26-63 repeats) or the longer class III (140-210 repeats) category (Bennett ST et al., 1996, Ann. Rev. Genet. 30:343-370). Initially, these alleles were studied by Southern blot, and later the class I alleles were studied by PCR (Bennett ST et al., 1995, Nat. Genet. 9(3):284-292; Bennett ST et al., 1996, Ann. Rev. Genet. 30:343-370). Due to the high GC-rich content and repetitive nature of this sequence, resulting in a highly stable intramolecularly folded structure, amplification of the long class III alleles has not been possible by PCR. Here, we present the first report, to our knpwledge, of PCR amplification for all classes of IN5 VNTR alleles, including class III and the intermediate-sized class II alleles, by a single PCR protocol (Fig. 1). The first four simples have a class I/III genotype, the next has a class I/II
genotype, and the final sample has two class III
alleles distinguishable by size. The number of repeat units in each class I allele is shown. Co-dominant segregation of class TII alleles within families confirmed the high fidelity of the method and the stability of these alleles within families. This technique represents an important advancement in the study of the IDDM2 susceptibility locus.
We used this technique to clone the class III
allele from one of our two informative thymus samples with complete silencing of INS expression from the class III chromosome (allele S1) (Vafiadis P et al., 1997, Nat. Genet. 15(3):289-292). The identity and integrity of the cloned allele, S1 was repeatedly demonstrated by showing electrophoretic co-migration of PCR amplification products from both genomic and cloned DNA templates (n>15). This indicates that the two products are the same size, and thus no major recombir~ational event resulting in loss of sequence has occurred in the cloned allele. Similarly, both cloned and genomic DNA demonstrated the same RFLP band pattern after MspI digestion (MspI recognizes an uncommon variant of the repeat unit (Bennett ST et al., 1996, Ann. Rev. Genet. 30:343-370)), suggesting the same arrangement of internal repeat sequences. To further verify the identity of the S1 clone, we attempted to sequence the insert. Sequencing from the 3' end of the insert proved unsuccessful due to a number of strong stops. However, we were able to sequence about 340 by of 5' sequence, including about 20 by immediately upstream of the INS VNTR and the first approximately 18 repeat units. These appear to be identical to those obtained for the previously cloned class III-G allele (Bennett ST et al., 1996, Ann. Rev. Genet. 30:343-370), although certain nucleotides could not be determined with absolute certainty (Fig. 2). The insert sequence begins at position 14 for S1, 10 for El and 1 for 814.
The INS VNTR sequence begins at position 37 for S1, 32 for E1 and 22 for allele 814 (Fig. 2A). If the S1 allele has a different effect on thymic INS expression than the published class III-G allele, then the difference is unlikely to be due to sequence variation in this 5' region. Using the same methods, we similarly cloned and confirmed a class III allele associated with enhanced expression of thymic INS as compared to a class I allele (E1) and a class I allele with 42 repeat units (corresponds to allele 814 in mobility units) (Fig. 2). Repeats represented by bold type could not be ascertained with complete certainty, but the most likely sequence is presented.
To test our hypothesis regarding the mechanism of 117,DM2 encoded susceptibility, 167 families consisting of mother, father and type 1 diabetes affected child were genotyped for the INS VNTR by this PCR method. In Caucasians, about 75~ of INS VNTR
alleles are class I and about 25~ are class III
alleles. The parental genotypes in our mostly Caucasian population follow this distribution with 59~
class I/I, 37~ I/III and 4~ III/III as compared to expected frequencies of 56% I/I, 38% I/III and 6%
III/III. As expected, the genotype frequencies in affected offspring were skewed towards a higher percentage of predisposing class I/I genotypes: 68%
I/I, 28~ I/III and 4% III/III. From this group, 139 parental class III alleles could be unambiguously determined to have been transmitted or not-transmitted to diabetic offspring from mothers or fathers. We used cloned and genomic DNA to PCR amplify the S1 allele.
Similarly, we used genomic DNA to amplify the class III
allele associated with silencing of thymic INS
expression in a second sample (which we call S2) and S2-identical alleles (see Methods). These were then used as markers against which all parental class III
alleles were compared and identified as S1 or S2 type alleles or as non-S1/ non-S2 type alleles.
Each allele was determined to either be the same size or a different size than alleles S1 or S2 by PAGE on a long gel (38.5 cm). All class III alleles analyzed were either larger than both S1 and S2, smaller than S1 and S2 or equal in size to either S1 or S2 (Fig. 3).
The class III INS VNTR alleles in families with a type 1 diabetes offspring (a to n) are compared in terms of size to the class III alleles associated with silencing of thymic INS expression (S1 and S2) (Fig. 3A). These include father-affected offspring (b & c) and mother-affected offspring (m & n) pairs. The class III INS VNTR alleles in families with a type 1 diabeter> offspring (a to m, not the same samples as in A) are compared in terms of MspI RFLP to the class III
alleles associated with silencing of thymic INS
expression (S1 and S2) (Fig. 3B). These include a mother-affected offspring pair (a & b). The size in by of molecular weight standards are shown at the left.

Thus, the S1 allele is only slightly larger than the S2 allele, and there are no allele sizes between them that can be resolved by our method.
Furthermore, each class III allele was loaded into two separate wells of the polyacrylamide gel for comparison to S1 and S2 separately. Those that were equal in size to S1 were always larger than the S2 allele, and those that were equal in size to S2 were always smaller than S1 allele. This indicates that the same alleles behave in a consistent manner between separate loadings in PAGE and that our method can consistently identify small differences in migration distance.
In addition to size, we attempted to distinguish alleles by repeat unit composition through comparison of their MspI RFLP band pattern (fingerprint). Class III alleles were classified as either the same as or different from S1 or S2 (Fig. 3).
Most alleles had two large digestion fragments that were well above the largest digestion fragments from any of the class I alleles. Thus for class I/III
heterozygotes, we were able to distinguish between class I and III digestion fragments in this size range (see Methods). The size of the largest digestion fragment did not vary much between alleles. Similarly, the second largest fragments from most alleles were all within a narrow size range. This indicates that most class III alleles have a very similar sequence composition. The exception to this were the class III
alleles present in an uncommon haplotype, previously referred to as the very protective haplotype (VPH) (Bennett ST et al., 1995, Nat. Genet. 9(3):284-292), which were much more thoroughly digested than those from the common protective haplotypes (PH) (Bennett ST
et al., 1995, Nat. Genet. 9(3):284-292), resulting in much smaller digestion fragments. Thus, class III

alleles found in the VPH are smaller and likely contain more copies of the repeat sequence recognized by Mspl than the more common class III alleles found in the PH.
The size and MspI analyses were also combined such that class III alleles were considered as S1 or S2 type alleles only if they matched the S1 or S2 allele, respectively, in terms of size and MspI RFLP pattern.
Alleles were thus designated as S1 or S2 type alleles based on size only, MspI RFLP only, or based on both size and MspI RFLP.
The transmission frequency of S1 and S2 alleles to diabetic offspring is affected by parental origin.
S1 and S2 alleles, as determined by size, are transmitted to diabetic offspring significantly more often from fathers (16 transmitted/9 not-transmitted) than from mothers (10 transmitted/24 not-transmitted;
p<0.017). As expected, there was no significant transmission distortion between maternal and paternal non-S1/non-S2 alleles (Table 1).
Transmission analy*is of non-Sl and non-S2 class III alleles Alleles Transmitted Not p-value transmitted x test By size only:
Maternal non-S1 or non-S2 14 22 0.96 Paternalnon-S1 non-S2 18 26 or By MspI RFLP only:

Maternalnon-S1 non-S2 18 40 0.23 or Paternalnon-S1 non-SZ 24 31 or By size and MspI
RFLP:

Maternalnon-S1 non-S2 21 43 0.27 or Pat~rnalnon-S1 non-S2 26 33 or The parental class III alleles from 167 families with a type 1 diabetes affected offspring were determined to have either been transmitted or not-transmitted to affected offspring from mothers or fathers. These alleles were identified to be either the same or different than alleles S1 or S2, which are associated with silencing of INS expression in thymus, by either size only, MspI RFLP only or by a combination of both size and Mspl RFLP. The transmission distortions are as expected from previous studies for class III alleles, and no different between maternal and paternal non-S1/non-S2 alleles.
There was also no significant difference in the transmission distortion of S1 and S2 type alleles from mothers to diabetic offspring as compared to all non S1/non-S2 alleles (Table 2).
Table 2 Transmission aaalysis of maternal S1 and S2 class III
alleles Alleles TransmittedNQt p-value transmittedx2 test By size only:

Maternal S1 or S2 10 24 0.38 Maternal and Paternal non-S132 48 or non-S2 By MspI RFLP only:

Maternal Sl or S2 7 6 0.38 Maternal and Paternal non-S142 71 or non-S2 By size and MspI RFLP:

Maternal S1 or S2 5 3 0.32 Maternal and Paternal note-3147 76 or non-9~:2 The possible effects of maternal S1 or S2 class III alleles on type 1 diabetes susceptibility are evaluated. The transmission rate of maternal S1 or S2 alleles to diabetic offspring does not differ significantly from transmission of all non-S1/non-S2 alleles. Maternal S1 or S2 alleles are thus protective of type 1 diabetes.
Thus, all alleles that are NOT paternal S1 or paternal S2 alleles, can be grouped together as they all demonstrate the same transmission distortions to diabetic offspring, as previously described for class III as a whole. Furthermore, paternal S1 and S2 alleles are transmitted significantly more often to diabetic offspring than all other alleles (Table 3).

Trans~anission analysis paternal S1 and S2 class III

alleles Alleles Transmitted Not p-value transmittedx2 test .err By size only:

Not-pate7rnal Sl or S2 42 72 0.023 Paternal S1 or S2 16 9 By MspI RFLP only:

Not-pate~'nal Sl or S2 49 77 0.06 Paternal S1 or S2 9 4 By size and MspI

RFLP:

Not-paternal S1 or S2 52 79 0.11 Pat~rnal S1 or S2 6 2 There was no significant transmission distortion between maternal S1 or S2 alleles, maternal non-S1 or non-S2 alleles and paternal non-S1 or non-S2 alleles and so these were grouped together (not-paternal S1 or S2). There is a significant transmission distortion of S1 or S2 alleles from fathers in a direction opposite to that seen with all other class III alleles.

This suggests that S1 and S2 class III alleles are predisposing for type 1 diabetes, but only when paternally inherited. Furthermore, most individuals are class I/III heterozygotes, thus paternally transmitted S1 or S2 type alleles are preferentially transmitted to diabetic offspring in comparison to class I alleles, suggesting that they are more predisposing than class I alleles. These results support our a priori hypothesis.
In order to ascertain if all S1 and S2 type alleles, as determined by the above methods, are associated with silencing of thymic INS expression, we characterized the class III alleles in 16 class I/III
samples which express both INS copies in the thymus.
Most of the class III alleles (15/16) were different from S1 or S2 alleles by size and/or MspI RFLP
analysis, while one was found to be identical to Sl by size and MspI RFLP pattern. Despite this, it is associated with enhanced thymic INS expression relative to the class I allele. There may two likely explanations for this. First, since parental origin in this sample was unknown, the class III allele could be of maternal origin, and therefore not expected to silence thymic INS expression. Alternatively, it could be a paternal allele that is different from S1 and S2 in terms of small differences in size and/or sequence variation not differentiated by MspI digestion. Note that the El allele which we have cloned is from this individual, thus E1 is an S1-type allele by size and MspI analysis, although it behaves like most class III
alleles as an enhancer of INS expression in thymus relative to class I alleles.
We have undertaken an analysis of the INS VNTR
role in type 1 diabetes. The results of this study provide further evidence that IDDM2 contributes to the pathogenesis of type 1 diabetes through an immunoregulatory mechanism. The INS VNTR may affect the level of thymic INS expression and thus the efficiency of immune tolerance induction to this (3-cell restricted autoantigen. This may have important implications for clinical trials of insulin prophylaxis in type 1 diabetes, which may be of particular benefit to individuals with class III alleles that silence thymic INS, as inefficient immune tolerance induction to insulin antigens may be a particularly important pathogenic mechanism in this group. Furthermore, our data are compatible with allele-restricted imprinted INS expression in the thymus, with repression of the paternal INS copy only when linked to particular class III INS VNTR alleles.
Methods DNA samples DNA was extracted using a phenol-chloroform method from blood samples collected from patients of the Montreal Children's Hospital diabetes clinic and their parents, and from participants of the Minneapolis branch of the multicenter study of the natural history of diabetic nephropathy in type 1 diabetes. Samples were obtained with signed and informed. Human fetal thymus tissues were obtained at the time of pregnancy termination with written consent from the mother. The tissue was pulverized under liquid nitrogen and DNA and RNA were extracted using phenol-chloroform under neutral and acid conditions, respectively.
INS VNTR PCR for all claatses The PCR reaction for amplification of all classes of INS VNTR alleles contained approximately 100-200 ng genomic DNA or 0.0000235 ng of PvuII
digested cloned DNA, 0.2 mM of each dNTP, l~Ci P32-dCTP, 1 mM MgCl2, 2.5 ~.1 lOX NH4 PCR Reaction buffer (ID Labs), 0.375 units ID-Zyme thermostable DNA
polymerase (ID Labs), 100 ng of sense primer and 100 ng of antisense primer. The primers used were either VNTR5 (TCAGGCTGGACCTCCAGGTGCCTGTTCTG) and VNTR6 (GCTGGTCCTGAGGAAGAGGTGCTGACGA) previously described by Bennett et al (Vafiadis P et al., 1997, Nat. Genet.
15(3):289-292) or the newly designed VNTR7 (GGCATCTTGGGCCATCCGGGACTG) and VNTRB
(GCAGGGCGGGGCTCTTTGCGCTG) primers that directly flank the INS VNTR and are present in the cloned DNA. The PCR was carried out for 25-26 cycles of: 94°C/30 sec denaturation, 62°C/30 sec annealing and 70°C/3 min 30 sec plus a 4 sec extension per cycle. Products were visualized by polyacrylamide gel electrophoresis (PAGE) and autaradiagraphy.
Cloning of INS VRTR alleles The INS VNTR alleles to be cloned were PCR
amplified using the VNTR5 and VNTR6 primers. The PCR
product was double digested with NcoI (recognizes a site 33 by 5' to the INS VNTR), and Pstl (recognizes a site 28 by 3' to the INS VNTR). The NCOI/PstI double digestion product was desalted with the Wizard DNA
Clean-up System (Promega Corp.), concentrated by evaporation and purified from a low melting agarose gel following electrophoresis using the Wizard PCR Preps DNA Purification System (Promega Corp.) Similarly, the Promega T vector was digested in the polylinker region with NcoI and PstI, dephosphorylated with CIAP (Gibco-BRL), and purified from a low melting agarose gel using the Wizard PCR Preps DNA Purification System (Promega Corp.). The INS VNTR allele was then directionally ligated into the T-vector using T4 DNA Ligase (Gibco-BRL). The ligation reaction was transformed into JM109 competent cells (Promega Corp.) and clones containing the INS VNTR-T vector construct were identified by restriction analysis.
DNA sequencing of cloned INS 'VNTR alleles The INS VNTR alleles were sequenced by automated fluorescent sequencing (Applied Biosystems Inc., Perkin Elmer) using the T7 primer. Sequencing of the C-rich strand was unsuccessfully attempted using the Sp6 primer.
Identification of class III alleles as S1 or S2 type alleles To determine if particular class III alleles were the same size as either the S1 or S2 allele, they were amplified by PCR using the VNTR7/VNTR8 primer pair and electrophoresed on a 38.5 cm long 8~ polyacrylamide gel for approximately 6 hours at 60 watts. To distinguish small differences in size, PCR product from the S1 and S2 alleles were loaded in every 2 to 4 lanes, interspersed throughout the alleles whose size was to be determined. The migration distance of the unknown allele was then compared to the adjacent S1 or S2 allele. The S1 allele was amplified from cloned and/or genomic DNA, and the migration distance of the unknown alleles were compared to that of the S1 allele.
Almost every gel had at least one PCR product (many had several) amplified from S1 genomic DNA loaded in a well next to PCR product from cloned DNA. In all cases, the PCR product from cloned S1 and genomic S1 DNA migrated to the exact same distance, indicating that PCR product from the cloned S1 allele DNA template is identical to the genomic DNA template. Similarly, the S2 allele was amplified from genomic DNA and compared in size to other class III alleles. As this allele was not cloned and genomic DNA quantities of the S2 allele were limited, an allele found to be identical in size to S2 was used as a size marker for the S2 allele. PCR

product from this marker allele was electrophoresed with PCR product from genomic DNA containing the S2 allele, and in all cases they migrated to the same distance, indicating that they are the same size by our method. Furthermore, most gels had multiple loadings of PCR product from the S1 allele genomic DNA next to the marker S1 allele and again, they migrated to the same distance in all cases. Thus, PCR product of class III alleles was loaded onto a polyacrylamide gel, adjacent to PCR product from S1 or S2 alleles, electrophoresed and determined to be either the same size or a different size than S1 and S2 alleles.
The same PCR products were digested with MspI
and electrophoresed in a 38.5 cm long 8~ polyacrylamide gel for 4 hours at 60 watts. The same type of analysis and controls were used as for the size analysis described above. For most samples, the two largest MspI digestion fragments were compared between the unknown class III allele and the S1 and S2 alleles. In class I/III individuals, these fragments are always located above even the largest class I allele digestion fragments and are of much lower abundance due to preferential amplification of the smaller class I
alleles over the class III alleles. This was determined by comparing the MspI digestion fragments in a number of class I/III individuals obtained from PCR
product where only the class I allele is amplified (using the previously described PCR conditions (Bennett ST et al., 1995, Nat. Genet. 9(3):284-292)) and PCR
product from our PCR protocol for all classes which amplifies both class I and III alleles.
The present invention will be more readily un derstood by referring to the following examples which are given to illustrate the invention rather than to limit its scope.

~X~MPL,E I
A child is screened for diabetes risk, in a future implementation of a genetic screening to identify candidates for a preventive intervention.
Based on the cost, inconvenience, and risk of the intervention it has been decided that it is well worth treating five individuals exceeding the risk threshold, even though only one will develop diabetes if untreated. The threshold is therefore set to a value that requires a specificity of 20~.
The particular child tests positive for a paternally transmitted Sl allele, using the method of the present invention. The maternal allele is an ordinary class I. By existing technology, the assigned I/III genotype will be deemed to confer an approximately four-fold less risk than I/I, the most common Caucasian genotype (Bennett ST et al., 1996, Ann. Rev. Genet. 30:343-370). By the method of the present invention, it will be assigned a four-fold higher risk than all other IDDM2 genotypes. Thus, without the present invention, the child will have been assigned a risk estimate of an order of magnitude lower than the correct value. No matter what the genotype in other loci is, this error is likely to result in misclassification of most individuals in this situation.
Based on the frequency of the untransmitted alleles (Tables 2 and 3) this mis-classification will apply to 7~ of all individuals in a population screen.
While the invention has been described in con-nection with specific embodiments thereof, it will be understood that it is capable of further modifications and this application is intended to cover any varia-tions, uses, or adaptations of the invention following, in generalA the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains and as may be 5 applied to the essential features hereinbefore set forth, and as follows in the scope of the appended claims.

Claims (6)

1. A DNA assay for the prediction of autoimmune diabetes in human subjects, which comprises the steps of c) obtaining a DNA sample from said subject and PCR amplification of class III alleles of variable number of tandem repeats (VNTR) located upstream of the insulin gene (INS); and d) electrophoretic co-migration of amplified fragment to identify class III alleles of the insulin VNTR which are identical to S1 or S2, from alleles which differ from S1 and S2;
whereby identification of S1 or S2 allele is indicative of at least 1% risk of autoimmune diabetes.

2. The DNA assay of claim 1, wherein INS VNTR is composed of a variable number of tandem 14-15 by repeat sequences, with the consensus repeat unit ACA
GGGG TGT GGGG.

3, The DNA assay of claim 1, wherein the amplification of step a) is effected using at least one primer pair selected from the group consisting of VNTR5 (TCAGGCTGGACCTCCAGGTGCCTGTTCTG)/VNTR6 (GCTGGT
CCTGAGGAAGAGGTGCTGACGA) and VNTR7 (GGCATCTTGGGCCATCC
GGGACTG) /VNTR8 (GCAGGGCGGGGCTCTTTGCGCTG).

4. The DNA assay of claim 1, wherein said sample is selected from the group consisting of blood, saliva, urine and hair follicle.

5, The DNA assay of claim 1, wherein said electrophoretic co-migration of step b) is effected using PAGE.

6. A primer pair for PCR amplification of class III alleles of variable number of tandem repeats (VNTR) located upstream of the insulin gene (INS) which comprises VNTR5 (TCAGGCTGGACCTCCAGGTGCCTGTTCTG)/
VNTR6 (GCTGGT CCTGAGGAAGAGGTGCTGACGA) or VNTR7 (GGCATCTTGGGCCATCCGGGACTG)/VNTR8 (GCAGGGCGGGGCTCTTTGCG
CTG).