A type 1 diabetes subgroup with a female bias is characterised by failure in tolerance to thyroid peroxidase at an early age and a strong CTLA-4 gene association

Abstract

Aims/hypothesis:

HLA haplotypes DRB1*03-DQB1*02 and DRB1*04-DQB1*0302, allelic variation of the T cell regulatory gene CTLA4 and allelic variation of the T cell activation gene PTPN22, have been associated with type 1 diabetes and autoimmune thyroid disease (AITD). Using thyroid peroxidase autoantibodies (TPOAbs) as an indicator of thyroid autoimmunity, we assess whether the association of these loci is different in type 1 diabetes cases with TPOAbs, than in those without.

Methods:

TPOAbs were measured in 4,364 type 1 diabetes cases from across Great Britain. 67% were under 18 years. These cases and 6,973 geographically matched controls were genotyped at CTLA4, PTPN22, HLA-DRB1 and HLA-DQB1.

Results:

462 (10.8%) type 1 diabetes cases had TPOAbs. These cases had a stronger association with CTLA4 (odds ratio [OR] = 1.49 for the G allele of the single nucleotide polymorphism rs3087243; 95% confidence interval [CI] = 1.29-1.72) than in the TPOAb-negative cases (P = 0.0004; OR = 1.16; 95% CI = 1.10-1.24), or in type 1 diabetes cases overall (OR = 1.20; 95% CI = 1.13-1.27). The female:male ratio was higher (1.97:1) in this subgroup compared with type 1 diabetes patients without TPOAbs (0.94:1; P = 1.86 × 10⁻¹⁵). TPOAb status did not correlate with age-at-diagnosis of type 1 diabetes or with PTPN22 (Arg620Trp; rs2476601).

Conclusions/interpretation:

Our results identify a subgroup of type 1 diabetes cases that are sensitive to allelic variation of the negative regulatory molecule CTLA-4, and indicate that TPOAb testing could be used to sub classify type 1 diabetes cases or subjects at risk for genetic, biological or clinical studies.

INTRODUCTION

Autoimmune diseases, a failure of immune tolerance to self antigens, are strongly inherited and often co-occur in the same patients or families [1-5], probably due to sharing of susceptibility alleles across multiple loci [6] and environmental factors [7, 8]. In particular, autoimmune thyroid disease (AITD; Graves' disease and Hashimoto thyroiditis associated with hyper- and hypothyroidism, respectively) and autoimmune type 1 diabetes are often diagnosed in the same subjects [1, 2, 4, 9, 10].

In addition to associations with HLA class II alleles and haplotypes, notably, DRB1*03-DQB1*02 [10-12], and the lymphoid specific phosphatase (LYP) T cell activation gene PTPN22 (28), AITD and type1diabetes have been associated with the CTLA-4 gene[5], encoding a vital negative immunoregulatory receptor in T cell activation and expansion. The association with AITD is strong (odds ratio ∼ 1.5) with single nucleotide polymorphisms (SNPs) in the 3'UTR of the gene being the strongest candidates for the causal variant(s) [13]. The most disease associated SNP is rs3087243 (with aliases 6230G>A and CT60) [13]. However, the effect is much smaller in type 1 diabetes (OR ∼ 1.15-1.2) [14], and recently, a study of 769 Japanese type 1 diabetes cases, 345 of which were diagnosed with AITD [15] has claimed that CTLA4 is only associated with AITD, and that the association with type 1 diabetes is secondary to anti-thyroid autoimmunity such that the G allele of the CTLA4 rs3087243 SNP does not have any direct effect on susceptibility to type 1 diabetes. However, the reported differences were not statistically significant. In three other studies, the CTLA-4 gene was reported to be associated with co-occurrence of type 1 diabetes and AITD (a form of autoimmune polyendocrine syndrome (APS) [2, 16]) [5, 17, 18].

Autoantibodies to components of the thyroid gland, including thyroid peroxidase autoantibodies (TPOAbs), often precede AITD diagnosis [4, 19], and the CTLA-4 gene is associated with TPOAb production [20]. In the USA, 4.8% of individuals aged 12-19 years have thyroid peroxidase autoantibodies [21], whereas 10 to 30% of patients with type 1 diabetes have thyroid antibodies, and up to 50% of these progress to clinical AITD [16].

We have collected plasma and DNA samples from over 4,000 cases of type 1 diabetes, mostly from paediatric clinics from across Great Britain. Hence, in the present report, we have investigated the association of CTLA-4 in type 1 diabetes cases subclassified into those with TPOAbs and those without.

SUBJECTS, MATERIALS AND METHODS

Subjects

We studied 4,364 type 1 diabetes cases from Great Britain, mostly diagnosed under age 16 years with a mean (± SD) of 7.4 (± 3.9) years, in which plasma TPOAb levels were measured, and 6,973 geographically matched controls from the 1958 British Birth Cohort [22]. Approval from the relevant ethics committee was sought before DNAs were collected and written informed consent obtained from all participants or their parents in the event they were too young to give consent. This investigation was carried out in accordance with the principles of the Declaration of Helsinki as revised in 2000. TPOAbs in the GB type 1 diabetes cases was measured with the PLATO processor ELISA, supplied by Sweden diagnostics UK. Recombinant TPO antigen was used, standardised against the national institute of biological standards and controls standard sera 66/387. Individuals were considered positive for TPOAb above 85 IU/ml.

Genotyping

The rs3087243 (+6230G>A) SNP in the CTLA-4 gene was genotyped in all samples because it was the most associated variant in a recent study of type 1 diabetes and AITD [13] using TaqMan MGB chemistries (Applied Biosystems). Details of the genotyping of the SNPs of PTPN22 (Arg620Trp; rs2476601) and INS (rs689) were described previously [22, 23]. HLA-DRB1 and -DQB1 were typed in a subset of these British cases using the Dynal-Roche system.

We successfully genotyped and measured TPOAb in 3,998 type 1 diabetes cases at CTLA4, 4,219 cases at INS, 4,203 cases at PTPN22, 2034 cases at HLA-DQB1 and 1,952 cases at HLA-DRB1. We also had genotypes for 6,866 controls at CTLA4.

Statistical Methods

Logistic regression was used to test for association and to calculate odds ratios (OR), with 95% confidence intervals (95% CI), within STATA8 (www.stata.com) using the rs3087243 A/A genotype as reference at OR = 1.0. As cases and controls were collected from across Great Britain, analyses were stratified according to 12 geographical regions by putting a regional variable into the regression [24]. In this way any confounding due to variation in allele frequency across Britain was minimised. To test for an interaction effect between genotype (at rs3087243, PTPN22 or INS) or haplotype (HLA-DRB1-HLA-DQB1) and presence of TPOAbs in type 1 diabetes cases, a case-only logistic regression approach was adopted with TPOAbs status as the outcome variable. We also tested whether gender explained the variation in TPOAbs by regressing gender on TPOAbs status. Gender was found to be an important determinant of TPOAbs status (1 × 10⁻⁷) and so was included as a confounder in all analyses of TPOAb by placing a gender variable into the regression.

RESULTS

Thyroid autoantibodies in type 1 diabetes

Out of 4,364 case samples tested, 462 (10.6%) were TPOAbs positive, which is similar to previous studies [4, 5]. The female:male ratio was significantly elevated, at 1.97:1, in the TPOAbs-positive cases compared to the TPOAbs-negative cases, 0.86:1, or to type 1 diabetes overall, 0.94:1 (Table 1). Being female carries an increased likelihood of a type 1 diabetes case having TPOAbs (OR = 2.31; 95% CI = 1.87-2.85; P = 1.86 × 10⁻¹⁵; Table 1). The mean age-at-diagnosis of diabetes did not differ between the TPOAbs-positive (7.7 years) and negative cases (7.5 years; P = 0.16; Table 1; Fig 1), with median ages at type 1 diabetes diagnosis of 8 and 7 years, respectively (Table 1 and Fig 1). In the TPOAbs-positive subgroup there was no difference between females and males in the age-at-diagnosis of type 1 diabetes (means 7.72 and 7.56 years, and ranges of 1-16 and 1-21). There was no effect of duration of type 1 diabetes on the presence of TPOAbs (P = 0.73) once age at blood donation was accounted for in the regression model.

Table 1.

Phenotypes of the thyroid autoantibodies-tested subgroups in type 1 diabetes.

	All type 1 diabetes cases	TPOAbs positive cases	TPOAbs negative cases	P
Female:Male (n ;%)	0.94:1 (2112:2251; 48.4%:51.6%)	1.94:1 (305:157; 66.0%:34.0%)	0.86:1 (1767:2060; 46.2%:53.8%)	1.86 × 10−15
Age at diagnosis of type 1 diabetes mean (SD) Overall Male Female	7.5 (3.9) 7.5 (4.0) 7.5 (3.8)	7.6 (4.1) 7.5 (4.5) 7.7 (3.8)	7.5 (3.9) 7.5 (4.0) 7.4 (3.8)	0.16
Age at diagnosis of type 1 diabetes median [95% CI] Overall Male Female	7 7 7	8 7 8	7 7 7	N/A

SD: Standard Deviation

CI: Confidence Interval

N/A: not applicable

Figure 1.

Frequency distribution of age-at-diagnosis by thyroid antibodies status in one-year intervals.

Genetic control of type 1 diabetes with thyroid autoantibodies

CTLA4 showed a convincing difference between TPOAbs subgroups (P = 0.0004; Tables 2 and 3 and 4), with a 6.2% increase in the frequency of the disease-associated G allele of rs3087243 in the TPOAb-positive group (Table 2) and almost twice the risk of having both type 1 diabetes and TPOAbs, associated with the G allele or G/G genotype (odds ratio of the G/G genotype = 2.63 in the TPOAb-positive group compared to 1.37 in the TPOAb-negative group, with no overlap in the 95% confidence intervals; Table 4).

Table 2.

Genetic associations in thyroid autoantibodies-positive and -negative British type 1 diabetes cases.

	Number of casesa	Frequency (%) in TPOAbs-positive cases	Frequency (%) in TPOAbs-negative cases	Frequency (%) in femalesb	Frequency (%) in malesb	P difference
CTLA4 rs3087243 (G>A)	3998	64.8	58.6	48.6	51.4	0.0004
DRB1*03-DQB1*02	1956	38.8	33.7	47.7	52.3	0.02
DRB1*03-DQB1*02  / DRB1*03-DQB1*02	198	7.8	3.9	44.4	55.6	0.05
DRB1*0401-DQB1*0302	1956	20.3	24.9	47.0	53.0	0.72
DR3/4 (not DRB1*0403)	2046	32.0	34.1	47.5	52.5	0.52
DRB1*15-DQB1*0602	1956	0.7	0.2	36.4	63.4	0.02
INS −23HphI rs689	4219	14.9	15.4	48.4	51.6	0.63
PTPN22 rs2476601	4203	19.6	16.9	48.3	51.7	0.06

^aThese are the number of cases with TPOAbs measures who are also genotyped at the locus listed. Gender and age at donation of blood are included in tests of statistical differences between the TPOAbs-positive and -negative subgroups.

^bFor the HLA-DRB1 and -DQB1 haplotypes, percentages are based on the most likely haplotype assignment.

Table 3.

Allele and genotype frequencies at CTLA4 rs3087243 in type 1 diabetes and Graves' cases.

Allele/ Genotype	All type 1 diabetes cases n (%)	TPOAbs-positive n (%)	TPOAbs-negative n (%)	Graves' cases n (%)	Controls n (%)
G	4819 (59.3)	560 (64.5)	4174 (58.6)	835 (63.4)	7523 (54.8)
A/A	663 (16.3)	44 (10.1)	611 (17.1)	90 (13.7)	1431 (20.8)
A/G	1987 (48.9)	220 (50.7)	1732 (48.6)	301 (45.7)	3347 (48.8)
G/G	1416 (34.8)	170 (39.2)	1221 (34.3)	267 (40.6)	2088 (30.4)

Table 4.

Odds ratios with 95% confidence intervals at CTLA4 rs3087243 for all type 1 diabetes cases and by TPOAbs positivity compared to Graves' cases.

Allele/ Genotype	All type 1 diabetes cases		TPOAb positive casesa		TPOAb negative cases		Graves' disease cases b
	OR [95% CI]	P	OR [95% CI]	P	OR [95% CI]	P	OR [95% CI]	P
Gc	1.20[1.13-1.27]	3.71 × 10−10	1.49[1.29-1.72]	4.08 × 10−8	1.16 [1.10-1.24]	5.54 × 10−7	1.37 [1.19-1.57]	6.25 × 10−6
A/A	1.00 [ref]	8.66 × 10−10	1.00 [ref]	1.03 × 10−8	1.00 [ref]	2.10 × 10−6	1.00 [ref]	0.00004
A/G	1.30[1.16-1.45]		2.16 [1.55-3.01]		1.23 [1.09-1.38]		1.32 [0.99-1.74]
G/G	1.47[1.31-1.66]		2.63 [1.87-3.70]		1.37 [1.22-1.55]		1.85 [1.39-2.47]

^a2df test should be used here as the multiplicative model is inappropriate P = 0.0097.

^bIn the Graves' disease analysis, the cases, which are from a previous study [13] and were not collected from all 12 subregions of Great Britain, were matched to 1,745 controls from the same regions as the cases.

^cAllele A is used as reference.

There is one other interesting difference in the association of CTLA4 with the TPOAbs positive type 1 diabetes cases, and that is the mode of inheritance of the locus: the disease risk of the alleles does not fit with the multiplicative model, as seen in both type 1 diabetes and AITD (Table 4; P = 0.0097 against this model), and, instead, the G allele has a dominant effect with no significant difference between the G/A heterozygote (OR = 2.16) and the G/G homozygous genotype risks (OR = 2.63; Table 4). Note that the CTLA4 rs3087243 G allele does not affect age-at-diagnosis of type 1 diabetes overall (P = 0.18) or of type 1 diabetes in TPOAbs-positive cases (P = 0.67), nor is it differentially associated in females and males (P = 0.61).

We also tested whether the other established type 1 diabetes- and AITD-associated susceptibility genes, HLA-DRB1, HLA-DQB1, INS and PTPN22, were differentially associated with the two disease subgroups (Table 2). No convincing evidence of an interaction between TPOAbs-positivity and alleles, genotypes or haplotypes of these genes was obtained (Table 2).

DISCUSSION

Our results indicate that certain type 1 diabetes patients suffer from an immune tolerance defect to both the insulin-producing beta cell and the thyroid gland that can occur very early in life since TPOAbs are present in the youngest of type 1 diabetes cases (Fig 1). This form of disease is strongly regulated by allelic variation at CTLA4. This conclusion is consistent with the critical role of CTLA-4 in peripheral tolerance as an inhibitory regulator of T cell activation, T cell proliferation and apoptosis, as well as roles in the functions of antigen presenting cells and T regulatory cells [25-27]. We have proposed that a reduced function of CTLA4 associated with the rs3087243 G allele allows T cells to be more hyperactive and to respond to peripheral antigens to a greater degree than individuals carrying the SNP A/A genotype, which is associated with autoimmune disease protection and increased peripheral tolerance [13, 28]. However, our results also indicate that allelic variation of CTLA4 has a much smaller effect in approximately 90% of type 1 diabetes cases. Previous studies have reported that in excess of 15% of European type 1 diabetes cases develop anti-thyroid autoimmunity [16], and hence if we had been able to detect these within our case samples, the OR in the TPOAb-negative subgroup would probably decrease to 1 or very close to it. Our interpretation of this result is different from a recent Japanese study, that concluded that CTLA4 is only associated with AITD in type 1 diabetes cases [15], and that the association with type 1 diabetes is secondary to anti-thyroid autoimmunity such that the G allele of the CTLA4 rs3087243 SNP does not have any direct effect on susceptibility to type 1 diabetes.

We suggest that allelic variation of CTLA4 does affect type 1 diabetes directly by predisposing individuals with a G allele to reduced levels of tolerance, to a multiplicity of peripheral antigens and maybe even a form of autoimmune polyendocrine syndrome (the occurrence in patients of two or more endocrine autoimmune diseases). We also note that in patients with both type 1 diabetes and AITD, type 1 diabetes nearly always occurs more than a decade before the development of clinical AITD. However, the strongest evidence for a direct effect of allelic variation of the CTLA-4 gene on type 1 diabetes susceptibility is provided by studies in the NOD mouse [13, 29-31]. Genetic analysis of the NOD mouse model of type 1 diabetes has shown that the region of mouse chromosome 1 containing the CTLA-4 gene and the Idd5.1 susceptibility locus has a significant effect on disease susceptibility [13, 29-31]. The causal variant has been mapped to a SNP in exon 2 of Ctla4 for which the NOD allele reduces the splicing and expression of an alternative isoform of CTLA-4, the ligand-independent form (liCTLA-4) [13]. Moreover, it has been shown that liCTLA-4 has inhibitory functions in T cell activation [29].

NOD mouse studies have also shown that different combinations of alleles of multiple susceptibility genes can lead to a variety of autoimmune phenotypes, including liver disease and AITD, as well as type 1 diabetes with different rates of progression [32][Hunter K., JAT, LSW, unpublished]. In particular, the effect of resistance alleles at mouse Ctla4 is masked when resistance alleles at a separate disease locus are present in the mouse and, in a separate susceptibility gene combination, the resistance allele at mouse Ctla4 can completely mask the potent disease-causing effect of a particular susceptibility allele at another locus. It has also been shown that a third locus, Idd3, which is believed to be the interleukin-2 gene [32] [LSW, JAT, unpublished], can influence the expression of CTLA-4 [30, 33]. We, therefore, propose that AITD, type 1 diabetes and co-occurring type 1 diabetes and AITD, manifest due to different, but overlapping, combinations of susceptibility and resistance alleles of multiple loci, and that in isolated type 1 diabetes (up to 90% of cases) the constellations of alleles giving rise to this disease mask the allelically variable effect of CTLA4 to a much greater degree than the combination of alleles causing AITD or type 1 diabetes + TPOAbs as studied here. It will be interesting to try to identify genes with alleles in humans that can mask the effect of CTLA4 alleles. Nonetheless, we cannot ignore the role of the yet-to-be-defined environmental and developmental factors in determining the penetrance of susceptibility genes. Finally, the bias in the proportion of female type 1 diabetes cases with TPOAbs, is comparable to previous, small studies of 35 [5], 28 [18] and 77 [17] APS (type 1 diabetes + AITD) patients, in which the female:male ratios were 1.7:1, 2.5:1 and 2.3:1, respectively, suggesting that X-linked factors are operating. One possibility is that the skewing of X inactivation, which has been described in AITD cases [34, 35], alters expression of immune genes on chromosome X, thereby predisposing to autoimmunity in females. We note in our data that the bias may begin in cases with age-at-diagnosis less than age 10 years: female:male ratio in TPOAb-positive type 1 diabetes cases of 1.28:1, compared to 0.88:1 in the TPOAb-negative cases; therefore, before the onset of puberty, when hormonal effects are unlikely to influence the penetrance of APS in females.

We note that our study is unique in that we show a statistically significant difference in CTLA4 association between type 1 diabetes case subgroups, and in the female:male ratio, and, moreover, we have studied TPOAbs status, which is a routine clinical test, and our results show that in many type 1 diabetes cases at a young age this TPOAbs marker is present and, therefore, is a useful and early biomarker for this subclass of type 1 diabetes. In ongoing clinical trials to reverse, or retard, type 1 diabetes, inter-individual genetic variation in the balance of immune effector responses and tolerance could be partially reflected by TPOAbs status thereby making it a parameter that would positively correlate with a poor outcome to tolerance-inducing therapeutic intervention.

Abbreviations

AITD

autoimmune thyroid disease

APS

autoimmune polyendocrine syndrome

CTLA-4

cytotoxic T lymphocyte associated antigen-4

LYP

lymphoid specific phosphatase

SNP

single nucleotide polymorphism

TPOAbs

thyroid peroxidase autoantibodies