Diabetologia

, Volume 51, Issue 12, pp 2205–2213

Gene variants influencing measures of inflammation or predisposing to autoimmune and inflammatory diseases are not associated with the risk of type 2 diabetes

  • S. Rafiq
  • D. Melzer
  • M. N. Weedon
  • H. Lango
  • R. Saxena
  • L. J. Scott
  • the DIAGRAM Consortium
  • C. N. A. Palmer
  • A. D. Morris
  • M. I. McCarthy
  • L. Ferrucci
  • A. T. Hattersley
  • E. Zeggini
  • T. M. Frayling
Article

DOI: 10.1007/s00125-008-1160-3

Cite this article as:
Rafiq, S., Melzer, D., Weedon, M.N. et al. Diabetologia (2008) 51: 2205. doi:10.1007/s00125-008-1160-3

Abstract

Aims/hypothesis

There are strong associations between measures of inflammation and type 2 diabetes, but the causal directions of these associations are not known. We tested the hypothesis that common gene variants known to alter circulating levels of inflammatory proteins, or known to alter autoimmune-related disease risk, influence type 2 diabetes risk.

Methods

We selected 46 variants: (1) eight variants known to alter circulating levels of inflammatory proteins, including those in the IL18, IL1RN, IL6R, MIF, PAI1 (also known as SERPINE1) and CRP genes; and (2) 38 variants known to predispose to autoimmune diseases, including type 1 diabetes. We tested the associations of these variants with type 2 diabetes using a meta-analysis of 4,107 cases and 5,187 controls from the Wellcome Trust Case Control Consortium, the Diabetes Genetics Initiative, and the Finland–United States Investigation of NIDDM studies. We followed up associated variants (p < 0.01) in a further set of 3,125 cases and 3,596 controls from the UK.

Results

We found no evidence that inflammatory or autoimmune disease variants are associated with type 2 diabetes (at p ≤ 0.01). The OR observed between the variant altering IL-18 levels, rs2250417, and type 2 diabetes (OR 1.00 [95% CI 0.99–1.03]), is much lower than that expected given (1) the effect of the variant on IL-18 levels (0.28 SDs per allele); and (2) estimates, based on other studies, of the correlation between IL-18 levels and type 2 diabetes risk (approximate OR 1.15 [95% CI 1.09–1.21] per 0.28 SD increase in IL-18 levels).

Conclusions/interpretation

Our study provided no evidence that variants known to alter measures of inflammation, autoimmune or inflammatory disease risk, including type 1 diabetes, alter type 2 diabetes risk.

Keywords

Autoimmune diseaseGenesGenetic epidemiologyInflammationMendelian randomisationSNPType 2 diabetes

Abbreviations

CRP

C-reactive protein

DGI

Diabetes Genetics Initiative

FUSION

Finland–United States Investigation of NIDDM

GWA

genome-wide association

IL-1RA

IL-1 receptor antagonist

MAF

minor allele frequency

MIF

macrophage migration inhibitory factor

PAI

plasminogen activator inhibitor

sIL-6R

soluble IL-6 receptor

SLE

systemic lupus erythematosus

SNP

single-nucleotide polymorphism

WTCCC

Wellcome Trust Case Control Consortium

Introduction

Inflammatory and immunological factors are associated with type 2 diabetes, but the causal direction of this association is not known. Evidence for a possible causal role of inflammatory and immunological processes in type 2 diabetes comes from several sources. These include prospective studies that show raised levels of pro-inflammatory, and lower levels of anti-inflammatory, cytokines many years before the onset of type 2 diabetes [1]. Specific examples include the presence of raised circulating IL-18 [1] and macrophage migration inhibitory factor (MIF) [2] levels in the KORA study, where average follow-up time to diagnosis of type 2 diabetes is 10 years, and similar findings observed with raised C-reactive protein (CRP) or IL-6 levels in several other prospective studies [35]. Second, mouse knockout models of certain cytokine genes, including IL18, IL1RN, IL1A and IL1B, result in adverse effects on obesity and insulin resistance in mouse models [68]. For example, the CD11B (also known as MAC-1 or ITGAM) knockout mouse is susceptible to diet-induced obesity compared with wild-type litter mates [9]. A common variant in the CD11B gene has recently been established as a risk factor for systemic lupus erythematosus (SLE) [10], suggesting a possible link between autoimmune diseases and metabolic traits. Common variants in the CDKAL1 gene are associated with type 2 diabetes, psoriasis and Crohn’s disease [11], although the associated polymorphisms are different and not correlated (in linkage disequilibrium) with each other. Finally, recent data indicating a critical role for IL-1β in beta cell function [12] have led to a small randomised controlled trial of recombinant human IL-1 receptor antagonist (IL-1RA) (anakinra) treatment in type 2 diabetes that showed improved beta cell function and reduced levels of markers of systemic inflammation in 35 anakinra-administered patients with type 2 diabetes [13].

Further evidence is needed to establish whether inflammation has a causal role in type 2 diabetes in humans. Observed changes in inflammatory protein levels/concentrations with disease progression may be secondary to early disease processes and knockout animal studies suggest that if inflammatory processes have a role it is more likely to be through obesity-induced changes. Genetics studies can add to the evidence for or against the causal role of a trait in disease. This is because genetic variants cannot be influenced by disease processes and are much less likely to be confounded than non-genetic factors. This principle of ‘Mendelian randomisation’ has been used before to indicate that raised CRP levels are likely to be a consequence rather than cause of continuous metabolic traits [14].

Detailed studies of genetic variation across genes encoding inflammatory proteins have identified an increasing number of common variants that alter circulating levels of these proteins. These include variants in or near the IL18 [1517], IL1RN [17, 18], IL6R, PAI1 (also known as SERPINE1) [17, 19], MIF [20] and CRP genes [17, 21, 22], all associated with robust statistical confidence with levels of their respective protein products. In addition the variant in IL6R is associated with IL-6 levels as well as levels of its soluble receptor (sIL-6R) [23, 24].

Recent genome-wide association (GWA) studies have resulted in a greatly increased knowledge of common gene variants that predispose to systemic or organ-specific autoimmune and inflammatory diseases. These include convincing genetic associations between variants at 30 loci, outside the HLA region, and Crohn’s disease [25, 26], type 1 diabetes [27], rheumatoid arthritis [28, 29], coeliac disease [30, 31], ankylosing spondylitis [32], SLE [10, 33, 34] and multiple sclerosis [32, 35]. Five of these loci include variants that alter the risk of more than one inflammatory disease.

It has been suggested that type 1 and type 2 diabetes share a partially overlapping aetiology [36] but there is mixed evidence to support this. Recent studies show that known type 2 diabetes variants in the TCF7L2 and FTO loci do not predispose to type 1 diabetes [37, 38] In contrast, a recent Scandinavian study indicates that individuals with adult latent autoimmune diabetes have an increased frequency of both type 1 and type 2 diabetes risk alleles [39].

In this study we aimed to provide further insight into the possible role of inflammatory and autoimmune processes in type 2 diabetes. To do this we tested the hypothesis that common gene variants known to alter either circulating levels of inflammatory proteins or autoimmune and inflammatory disease risk also alter the risk of type 2 diabetes.

Methods

Study participants

Initial type 2 diabetes GWA meta-analysis

We used the p values and ORs from a meta-analysis of three type 2 diabetes GWA scans (http://www.well.ox.ac.uk/DIAGRAM/) that had recently been carried out using 2.2 million single-nucleotide polymorphisms (SNPs) (directly typed and imputed) with 4,107 type 2 diabetes cases and 5,187 controls, to identify associations between the selected polymorphisms and type 2 diabetes. The meta-analysis included samples used in GWA scans from the Wellcome Trust Case Control Consortium (WTCCC) (1,924 type 2 diabetes cases and 2,938 population controls, all from the UK), the Diabetes Genetics Initiative (DGI) (1,022 type 2 diabetes cases and 1,075 matched controls and 326 sibships discordant for diabetes) and the Finland–United States Investigation of NIDDM Genetics (FUSION) (1,161 type 2 diabetes cases and 1,174 normal glucose-tolerant controls, from Finland) studies [40]. Each SNP passed quality control criteria in each individual type 2 diabetes GWA study, as recently described [40]. Each study analysed the data under a model that is additive on the log-odds scale. ORs from each sample, excluding the DGI sibships, were combined using a fixed-effects model [41]. Patients in the WTCCC and DGI studies positive for anti-GAD antibodies were not included [40, 42]. The FUSION GWA case sample is composed of 789 cases from FUSION and 372 cases from FINRISK 2002. FUSION cases were excluded if they were positive for anti-GAD antibodies, had a short time to onset of insulin treatment following diagnosis, or had known or probable type 1 diabetes first-degree relatives [43]. Brief details of study participants are given in Electronic supplementary material (ESM) Table 1.

Replication study participants

For replication we genotyped SNPs nominally associated with type 2 diabetes using the UK Type 2 Diabetes Genetics Consortium Collection of 3,125 cases and 3,596 controls. All study participants were recruited from the Tayside region in Dundee, Scotland and were of European origin. Cases with monogenic forms of diabetes or with a history of treatment with regular insulin therapy within 1 year of diagnosis were excluded [44]. Controls were restricted to individuals <80 years of age and participants with evidence of hyperglycaemia were excluded from the study [44]. Details are given in ESM Table 1.

SNP selection

We selected three sets of SNPs from previous studies that have either been directly genotyped [28, 3032, 35, 40, 42, 43] or imputed in the three-study diabetes meta-analysis [40]. These included: (1) Eight SNPs previously associated with circulating IL-1RA [17, 18], sIL-6R [24, 45], IL-18 [16, 17], CRP [17], plasminogen activator inhibitor (PAI)-1 [19] or MIF levels [20]. All SNPs were ‘cis’ effects in that they were present in or near the gene that codes for the protein they are associated with. This together with the statistical confidence (p ≤ 10−5) of the associations between the SNP and protein levels means the associations are very unlikely to be false-positive results. For IL1RN and IL18 genes [1618] there is evidence for two independent association signals (r2 between variants <0.5) and we therefore included both the SNPs associated with respective protein levels in each case. (2) Seven SNPs from five loci: IL23R, PTPN2, PTPN22, SH2B3 and IL2RA, which are associated with more than one inflammatory/autoimmune disease [2527, 29, 31, 32, 35]. (3) Thirty-one independently associated variants from 30 loci proven to be autoimmune disease risk variants after robust replications (overall p ≤ 10−8) and excluding the SNPs in the MHC locus. These SNPs are associated with seven diseases (Crohn’s disease, type 1 diabetes, rheumatoid arthritis, coeliac disease, ankylosing spondylitis, SLE and multiple sclerosis).

All the 46 SNPs tested for an association passed QC control criteria in each individual study, as recently described [40]. Briefly all SNPs below a minor allele frequency (MAF) of 1% were excluded from analysis and all genotyped SNPs were in Hardy–Weinberg equilibrium in each individual study (p > 10−4 for WTCCC, p ≥ 10−6 for DGI and p ≥ 10−6 for FUSION). Where genotyping data were unavailable, imputed SNPs were used for analysis as has previously been described [40]. In total 54% (25), 50% (23) and 52% (24) of the tested 46 SNPs lacked direct genotype information in the FUSION, WTCCC and DGI studies, respectively, and were therefore imputed. All imputed SNPs, except in five situations, had quality scores (r2hat in DGI and FUSION, imputation information scores in WTCCC) >80% and all, except in 17 situations, had quality scores >90%.

Individual genotyping in the replication study

We used a p value of <0.01 in the three-way meta-analysis to take SNPs forward into the replication samples. The exception to this was SNPs in the IL1RN and IL18 genes, where we used the less stringent p < 0.1. SNPs in these two genes influence circulating levels and have a stronger prior case for their products’ involvement in type 2 diabetes given previous trial and prospective studies. Genotyping was performed by KBiosciences (Hoddesdon, UK) using their own system of fluorescence-based competitive allele-specific PCR (KASPar).

Statistics

Meta-analysis of p values using an additive effects model in WTCCC, DGI and FUSION samples has been previously described [40]. The ORs were calculated excluding the related component of the DGI study. To combine the previously published meta-analysis data with replication data we performed a meta-analysis using a fixed-effects inverse variance method using the summary ORs and 95% CIs from the three-way analysis as one study and the ORs and 95% CIs from the replication samples as a second study. This meta-analysis was performed using the ‘Metan’ module in STATA version 9.1. In the three-study meta-analysis we had 80% power to detect an OR of 1.20 for a SNP with a MAF of 0.09 (the lowest MAF of all 46 SNPs) at p = 0.05. For a more common SNP (MAF = 0.30) we had 80% power to detect an OR of 1.10 at p = 0.05. Power calculations were performed using STATA version 9.1.

Sensitivity analysis using directly genotyped SNPs

For 34 of the 46 SNPs the direct genotypes, or direct genotypes at an r2 > 0.8 proxy, were available in the WTCCC and DGI studies (total 2,946 cases and 4,013 controls typed on the Affymetrix 500 K chip). For eight of these 34 SNPs the r2 > 0.8 proxy was also available in FUSION. To confirm that the use of imputed data was not substantially influencing our results, we therefore repeated the analyses using only directly genotyped variants for these 34 SNPs.

Estimating expected effect sizes of inflammatory protein variants on type 2 diabetes risk

We calculated the approximate effect size we would expect a SNP that alters inflammatory protein levels to have on the risk of type 2 diabetes, if the inflammatory protein is causal to type 2 diabetes. We used the triangulation approach outlined in Fig. 1 to (1) estimate the expected per allele effect of a SNP on type 2 diabetes risk (A in Fig. 1), given (2) the per allele effect of the SNP on inflammatory protein levels (B in Fig. 1), and (3) the correlation between inflammatory protein levels and type 2 diabetes (C in Fig. 1). We did not have measures of inflammatory-related proteins in our cases and controls and so we used data from previous studies to obtain approximate effect sizes. For example, each allele of rs2250417, in the IL18 gene, raises IL-18 levels by 0.28 SDs in the InCHIANTI study [17]. In the same study, a 0.28 SD (68 pg/ml) rise in IL-18 levels is associated with an OR of 1.15 (95% CI 1.09–1.21) for type 2 diabetes risk (L. Ferrucci, unpublished results). The prospective KORA study of type 2 diabetes showed that individuals in the top quartile of IL-18 levels, had IL-18 levels 242 pg/ml higher than those in the bottom quartile [46], and a risk ratio for type 2 diabetes of 1.96 (95% CI 1.49–2.58). Using these figures we calculated that an allele effect of 68 pg/ml will equate to an approximate expected effect of a risk ratio of 1.21 (95% CI 1.12–1.30) for type 2 diabetes. Similar calculations using the previously reported effect of rs8192284 in the IL6R gene on IL-6 levels [24] and risk estimates from prospective studies of IL-6 with type 2 diabetes [47], suggest we would expect an approximate per allele effect of rs8192284 of 1.06 (95% CI 1.04–1.08) for type 2 diabetes risk. Using the previously reported effect of rs12093699 in the CRP gene on CRP levels [17] and risk estimates from prospective studies of CRP with type 2 diabetes [5] we would expect an approximate per allele effect of rs12093699 of a risk ratio of 1.34 (95% CI 1.23–1.39) for type 2 diabetes risk (ESM Table 2). These estimates are very approximate because they use circulating measures from different studies. In the case of prospective studies, estimates are based on risk ratios rather than ORs but the incidence of type 2 diabetes (5–10%) means that risk ratios and ORs will be approximately equivalent.
https://static-content.springer.com/image/art%3A10.1007%2Fs00125-008-1160-3/MediaObjects/125_2008_1160_Fig1_HTML.gif
Fig. 1

Associations between the SNP rs2250417 and IL-18 levels, IL-18 levels and type 2 diabetes and the expected and observed effects of rs2250417 on type 2 diabetes. ORs and risk ratios for the IL-18 levels–type 2 diabetes association are those estimated for a 0.28 SD increase in IL-18 levels, the effect of rs2250417 on IL-18 levels, from the InCHIANTI [17] and KORA studies [46]. Odds ratios and risk ratios are shown with 95% CIs

Results

SNPs known to alter circulating levels of inflammatory proteins

We found no strong evidence that eight SNPs, robustly associated with circulating levels of inflammatory proteins, alter the risk of type 2 diabetes. Associations with type 2 diabetes risk for two SNPs, one in each of the IL18 and IL1RN genes, reached p < 0.1 in the three-study meta-analysis. The protein products of these two genes have both been strongly implicated in type 2 diabetes in trial and prospective studies and so we assigned a less-stringent statistical threshold for follow-up. At the time of study the SNP in each of these genes that represented the strongest association with circulating levels was uncertain. We therefore decided to increase the confidence of our genetic association findings (negative or positive) by increasing the sample size for all four SNPs in IL1RN and IL18. Meta-analysis using a total GWA and replication sample of 7,232 cases and 8,783 controls showed no strong evidence of association (Table 1). Subsequent analysis [17] indicated that rs2250417 had the strongest association with circulating IL-18 levels. The OR observed between the IL-18-altering variant, rs2250417, and type 2 diabetes is 1.00 (95% CI 0.99–1.03) (Fig. 1d). This is much lower than that expected given (1) the effect of the variant on IL-18 levels (0.28 SDs per allele [17]), and (2) estimates, based on other studies, of the correlation between IL-18 levels and type 2 diabetes risk, which are 1.15 (95% CI 1.09–1.21) per 0.28 SD increase in IL-18 levels in the InCHIANTI study (L. Ferrucci, unpublished results), and a risk ratio of 1.21 (95% CI 1.12–1.30) based on the KORA prospective study [46] (Fig. 1).
Table 1

Cytokine cis-effect SNPs and associations with type 2 diabetes: three-way meta-analysis results across WTCCC, DGI and FUSION studies

Disease SNP

Gene

Total cases

Total controls

MAF

Imputed

OR for developing type 2 diabetes given pro-inflammatory allele (95% CI)

WTCCC

DGI

FUSION

Stage 1

p value

Stage 2 (including replication samples)

p value

rs4251961

IL1RN

7,493

8,949

0.40

No

No

Yes

1.06 (0.99–1.12)

0.07

1.01 (0.99–1.03)

0.28

rs6761276

IL1RN

7,888

8,484

0.44

Yes

Yes

No

0.98 (0.92–1.05)

0.60

1.00 (0.99–1.03)

0.39

rs2250417

IL18

7,674

9,175

0.47

Yes

Yes

No

1.00 (0.95–1.07)

0.87

1.00 (0.99–1.03)

0.39

rs5744256

IL18

7,479

8,509

0.21

Yes

Yes

Yes

1.08 (1.01–1.15)

0.03

1.03 (1.00–1.05)

0.03

rs8192284

IL6R

4,549

5,579

0.35

Yes

Yes

Yes

0.96 (0.90–1.02)

0.18

  

rs12093699

CRP

4,549

5,579

0.32

Yes

Yes

No

1.02 (0.96–1.09)

0.52

  

rs2227631

PAI1

4,549

5,579

0.40

Yes

Yes

No

0.97 (0.91–1.0)

0.45

  

rs1007888

MIF

4,549

5,579

0.42

Yes

Yes

Yes

0.96 (0.90–1.02)

0.19

  

For IL1RN and IL18 genes, data from the replication study are included in the meta-analysis

All ORs and p values from the meta-analysis are given in Table 1. The upper 95% CIs for risk alleles did not exceed 1.10. There was no evidence of association when limiting the analysis to the directly genotyped proxies, which captured six of the eight SNPs through directly genotyped proxies at r2 > 0.8, in the WTCCC and DGI studies (ESM Table 3).

SNPs known to alter the risk of autoimmune diseases

We found no strong evidence that SNPs robustly associated with multiple autoimmune diseases alter the risk of type 2 diabetes. ORs and p values from the three-study meta-analysis are given in Table 2. The upper 95% CIs for risk alleles did not exceed 1.12. We found no strong evidence that SNPs robustly associated with individual autoimmune diseases, including type 1 diabetes, alter the risk of type 2 diabetes. ORs and p values from the three-study meta-analysis are given in Table 3. The upper 95% CIs for risk alleles did not exceed 1.19. There was also no evidence of association when limiting the analysis to directly genotyped proxies, which captured 28 of the 38 SNPs at r2 > 0.8 in the WTCCC and DGI studies (ESM Table 3).
Table 2

Pleiotropic autoimmune disease SNPs and their associations with type 2 diabetes: three-way meta-analysis results across WTCCC, DGI and FUSION studies

Disease SNP

Disease

Gene

Imputed

Disease risk allele frequency HapMap-CEPH

Meta-analysis OR for developing type 2 diabetes given disease risk allele (95% CI)

OR p value

WTCCC

DGI

FUSION

rs11209032

Ankylosing spondylitis

IL23R

Yes

Yes

No

0.31

0.96 (0.89–1.02)

0.16

rs11805303

Crohn’s disease

IL23R

No

No

Yes

0.29

0.93 (0.87–0.99)

0.04

rs2542151

Crohn’s disease and type 1 diabetes

PTPN2

No

No

Yes

0.19

0.94 (0.87–1.02)

0.17

rs6679677

Rheumatoid arthritis, Crohn’s disease and type 1 diabetes

PTPN22

No

No

Yes

0.14

1.02 (0.93–1.12)

0.69

rs2104286

Type 1 diabetes and multiple sclerosis

IL2RA

No

No

Yes

0.23

1.01 (0.94–1.08)

0.79

rs17696736

Type 1 diabetes and coeliac disease

SH2B3

No

No

Yes

0.35

1.03 (0.97–1.09)

0.66

rs653178

Type 1 diabetes and coeliac disease

SH2B3

Yes

Yes

Yes

0.41

1.00 (0.94–1.06)

0.99

Hapmap CEPH refers to individuals of European ancestry typed in the International Hapmap project

Table 3

Autoimmune disease SNPs and their associations with type 2 diabetes: three-way meta-analysis results across WTCCC, DGI and FUSION studies

Disease SNP

Disease group

Gene

Imputed

Disease risk allele frequency HapMap-CEPH

Meta-analysis OR for developing type 2 diabetes given disease risk allele (95% CI)

p value for OR

WTCCC

DGI

FUSION

Rs30187

Ankylosing spondylitis

ARTS1

Yes

Yes

No

0.30

0.99 (0.93–1.05)

0.81

rs6822844

Coeliac disease

IL21, IL2, TENR (also known as ADAD1), KIAA1109

Yes

Yes

No

0.79

1.02 (0.94–1.11)

0.64

rs2816316

Coeliac disease

RGS1

Yes

Yes

No

0.78

1.02 (0.94–1.10)

0.68

rs13015714

Coeliac disease

IL1RL1, IL18R1, IL18RAP, SLC9A4

Yes

Yes

No

0.23

1.05 (0.97–1.12)

0.22

rs17810546

Coeliac disease

IL12A

Yes

Yes

No

0.10

1.03 (0.94–1.13)

0.50

rs1464510

Coeliac disease

LPP

Yes

Yes

No

0.57

1.02 (0.96–1.08)

0.51

rs1738074

Coeliac disease

TAGAP

Yes

Yes

No

0.49

1.00 (0.94–1.07)

0.88

rs10883365

Crohn’s disease

NKX2–3

No

No

Yes

0.50

0.98 (0.92–1.03)

0.41

rs2241880

Crohn’s disease

ATG16L1

Yes

Yes

No

0.54

0.98 (0.93–1.04)

0.60

rs10077785

Crohn’s disease

IBD5

No

No

Yes

0.75

0.96 (0.89–1.02)

0.20

rs10801047

Crohn’s disease

Chr 1q31

No

No

Yes

0.09

1.06 (0.95–1.19)

0.27

rs6596075

Crohn’s disease

Chr 5q31

No

No

Yes

0.83

1.03 (0.95–1.17)

0.50

rs12035082

Crohn’s disease

Chr 1q24

No

No

Yes

0.37

0.99 (0.94–1.06)

0.88

rs2836754

Crohn’s disease

Chr 21q22

No

No

Yes

0.65

0.97 (0.92–1.04)

0.42

rs17221417

Crohn’s disease

CARD15 (also known as NOD2)

No

No

Yes

0.36

1.07 (0.99–1.15)

0.05

rs4958847

Crohn’s disease

IRGM

No

No

No

0.09

1.06 (0.96–1.15)

0.24

rs9858542

Crohn’s disease

Chr 3p21

No

No

Yes

0.24

0.99 (0.93–1.05)

0.66

rs17234657

Crohn’s disease

Chr 5p13

No

No

Yes

0.17

1.06 (0.97–1.15)

0.18

rs10761659

Crohn’s disease

Chr 10q21

No

No

Yes

0.45

0.96 (0.91–1.02)

0.22

rs6920220

Rheumatoid arthritis

Chr 6q23

No

No

Yes

0.17

1.05 (0.98–1.13)

0.17

rs10499194

Rheumatoid arthritis

Chr 6q23

Yes

Yes

Yes

0.82

1.03 (0.96–1.10)

0.41

rs12708716

Type 1 diabetes

KIAA0350 (also known as CLEC16A) Chr 16p13

No

Yes

Yes

0.29

1.03 (0.97–1.10)

0.31

rs763361

Type 1 diabetes

CD226

Yes

Yes

No

0.47

0.94 (0.88–0.99)

0.04

rs3788964

Type 1 diabetes

IFIH1

No

No

No

0.86

0.95 (0.87–1.02)

0.17

rs2292239

Type 1 diabetes

ERBB3

No

No

Yes

0.30

0.97 (0.91–1.04)

0.45

rs3087243

Type 1 diabetes

CTLA4

No

No

Yes

0.54

0.98 (0.92–1.04)

0.46

rs6897932

Multiple sclerosis

IL7RA

No

No

No

0.76

1.03 (0.96–1.10)

0.35

rs6445975

SLE

PXK

Yes

Yes

No

0.21

1.01 (0.95–1.08)

0.73

rs12537284

SLE

IRF5, TNPO3

Yes

Yes

No

0.16

1.07 (0.98–1.16)

0.11

rs4963128

SLE

KIAA1542 (also known as PHRF1)

Yes

Yes

No

0.65

1.00 (0.94–1.08)

0.89

rs9888739

SLE

ITGAM

Yes

Yes

No

0.11

1.03 (0.94–1.14)

0.51

Hapmap CEPH refers to individuals of European ancestry typed in the International Hapmap project

Discussion

We have found no evidence that common gene variants known to alter circulating levels of inflammatory proteins, or those robustly associated with autoimmune or inflammatory-based diseases, influence risk of type 2 diabetes. We have not corrected for multiple testing because, given that our results are consistent with no association, this would be anti-conservative, rather than conservative. We included variants that influence inflammatory protein levels that are altered many years before the onset of diabetes. Included in our study were variants known to alter IL-1RA levels. This inflammatory protein is of particular interest because administration of recombinant human IL-1RA (anakinra) in a recent parallel group trial improved beta cell function in type 2 diabetic patients. However, we found no evidence that lifetime exposure to slightly lower IL-1RA levels, as determined by IL1RN genotypes, predisposes to type 2 diabetes.

Our results are consistent with chronic inflammation being secondary to type 2 diabetes disease processes rather than being causal. This does not rule out a role for other inflammatory or autoimmune processes, for which there is no known genetic variant. There are a number of limitations to our study. The main limitation is that we have not measured inflammatory proteins in our cases and controls. This means that we cannot perform a formal Mendelian randomisation analysis, where it is preferable to have the SNP, inflammatory protein levels and disease status measured in the same study. However, using data from other studies we estimated that variants that alter IL-18, IL-6 and CRP levels should predispose to type 2 diabetes with ORs of approximately 1.15 to >1.21, 1.06 and 1.34, respectively (ESM Table 2), if these pro-inflammatory proteins are causally related to type 2 diabetes. The upper 95% CI of the effect sizes we observed (1.05, 1.02 and 1.09 for IL18, IL6R and CRP variants, respectively) are lower than the expected effect size. This is consistent with raised IL-18, IL-6 and CRP levels being secondary to, rather than causally predisposing to, type 2 diabetes. Further studies are needed with measures of inflammatory proteins in large numbers of cases and controls. It is also possible that these variants are not altering inflammatory pathways even if they are altering circulating levels.

A second limitation is that variants known to alter the risk of autoimmune and inflammatory diseases may not have an inflammatory effect in the general population. However, the identification of variants associated with inflammatory and autoimmune-related diseases indicates that the variant may be altering the expression, processing or function of a nearby gene that is critical to an inflammatory or autoimmune process prior to disease onset. Most of the autoimmune disease-associated variants have only been discovered very recently and so the disease mechanisms are not completely understood. However, early studies indicate that the variant in the IL2RA region predisposing to type 1 diabetes and multiple sclerosis alters circulating levels of IL2RA protein [48], the variant predisposing to coeliac disease in the IL18RAP locus alters mRNA levels of IL-18 receptor accessory protein in lymphocytes [31] and the variant in the IL7RA (also known as IL7R) gene predisposing to multiple sclerosis alters levels of soluble IL-7 receptor antagonist because of differential splicing of a key exon [49]. It is also suggested that the variant predisposing to ankylosing spondylitis in the ARTS1 (also known as ERAP-1) gene may have a general pro-inflammatory effect because it cleaves cell surface receptors for IL-1, IL-6 and TNF-α [32].

A final possible limitation is that we have used imputed SNP genotypes for association analysis for the majority of the 46 studied variants. However, this is unlikely to have resulted in appreciable differences in OR estimates, as has previously been shown [40, 50] and there were no positive associations when we used the strongest available directly genotyped proxies (r2 > 0.8) instead of the imputed SNPs (ESM Table 3).

A further implication of our study is that it does not support the hypothesis of a genetic overlap between type 1 and type 2 diabetes. Our results are consistent with the recent studies which show that the known type 2 diabetes- and BMI-associated variants in the TCF7L2 [37] and FTO [38] genes do not predispose to type 1 diabetes.

In conclusion, in a large case–control study of type 2 diabetes, we tested 46 SNPs associated with inflammatory diseases or inflammatory protein levels, and identified no associations with type 2 diabetes. Our study is consistent with inflammatory processes being secondary, rather than causal, to type 2 diabetes.

Acknowledgements

M. N. Weedon is a Vandervell Foundation research fellow. The FUSION study was funded by an intramural grant (NIH DK 062370) from the US National Institutes of Health (NIH). E. Zeggini is a Wellcome Trust Career Development Fellow. D. Melzer’s work is supported in part by NIH/National Institute on Aging (NIH/NIA) Grant R01 AG24233. Details of the DIAGRAM (Diabetes Genetics Replication And Meta-analysis) Consortium can be found at http://www.well.ox.ac.uk/DIAGRAM/index.html.

Duality of interest

The authors declare that there is no duality of interest associated with this manuscript.

Supplementary material

125_2008_1160_MOESM1_ESM.pdf (18 kb)
ESM Table 1Characteristics of initial scan and replication study participants (18 kb)
125_2008_1160_MOESM2_ESM.pdf (24 kb)
ESM Table 2Description of SNP effects on circulating protein levels in the InCHIANTI study [21] and measures used from previous type 2 diabetes prospective studies to derive estimates of expected effects of SNP on type 2 diabetes risk (24 kb)
125_2008_1160_MOESM3_ESM.pdf (44 kb)
ESM Table 3Results for inflammatory and autoimmune disease SNPs using the best available directly genotyped proxies (45 kb)

Copyright information

© Springer-Verlag 2008

Authors and Affiliations

  • S. Rafiq
    • 1
  • D. Melzer
    • 1
  • M. N. Weedon
    • 1
  • H. Lango
    • 1
  • R. Saxena
    • 2
    • 3
  • L. J. Scott
    • 4
  • the DIAGRAM Consortium
  • C. N. A. Palmer
    • 5
  • A. D. Morris
    • 6
  • M. I. McCarthy
    • 7
    • 8
  • L. Ferrucci
    • 9
  • A. T. Hattersley
    • 1
  • E. Zeggini
    • 7
  • T. M. Frayling
    • 1
  1. 1.Peninsula Medical SchoolUniversity of ExeterExeterUK
  2. 2.Department of Molecular Biology and Center for Human Genetic ResearchMassachusetts General HospitalBostonUSA
  3. 3.Medical and Population Genetics ProgramBroad Institute of Harvard and MITCambridgeUSA
  4. 4.Department of Biostatistics, School of Public Health, and Center for Statistical GeneticsUniversity of MichiganAnn ArborUSA
  5. 5.Population Pharmacogenetics Group, Biomedical Research Centre, Ninewells Hospital and Medical SchoolUniversity of DundeeDundeeUK
  6. 6.Diabetes Research Group, Division of Medicine and Therapeutics, Ninewells Hospital and Medical SchoolUniversity of DundeeDundeeUK
  7. 7.Wellcome Trust Centre for Human GeneticsUniversity of OxfordOxfordUK
  8. 8.Oxford Centre for Diabetes, Endocrinology and MedicineUniversity of Oxford, Churchill HospitalOxfordUK
  9. 9.Longitudinal Studies Section, Clinical Research Branch, Gerontology Research CenterNational Institute of AgeingBaltimoreUSA