TCF7L2 Genetic Variants Do Not Influence Insulin Sensitivity or Secretion Indices in Autoantibody-Positive Individuals at Risk for Type 1 Diabetes

The TCF7L2 locus is the common variant signal with the largest effect on type 2 diabetes (T2D) (1,2). TCF7L2 encodes a transcription factor that plays a key role in the Wnt signaling pathway and has been implicated in blood glucose homeostasis. TCF7L2 variants are associated with abnormalities in insulin secretion, incretin and glucose-induced glucagon secretion, insulin processing, β-cell development, and insulin sensitivity in people with T2D and healthy controls (3–5).

We have previously reported that, among individuals with recent-onset type 1 diabetes (T1D), those who carried T2D-associated TCF7L2 gene variants compared with individuals who did not had lower area under the curve (AUC) glucose and higher AUC C-peptide in response to a mixed meal tolerance test and higher frequency of a single positive islet autoantibody (6). This suggests that individuals with T1D who have this single nucleotide polymorphism (SNP) may have abnormalities of glucose metabolism that are less typical of T1D than individuals without the T2D-associated genetic variant. In addition, T2D-associated TCF7L2 gene variants were more common in children with single compared with multiple autoantibody positivity at the onset of T1D (7) and in individuals with T1D who did not have T1D-associated HLA genotypes (8). Overall, these findings support the hypothesis that TCF7L2-associated pathogenic mechanisms could drive individuals to cross the diabetes threshold earlier than expected based on their degree of autoimmune destruction of β-cells. Therefore, diabetes in these individuals would have a mixed pathogenesis (9) that could be leveraged to inform prevention and treatment strategies. However, it is unclear if the TCF7L2 SNP may influence glucose metabolism in nondiabetic high-risk autoantibody-positive relatives of individuals with T1D. If nonautoimmune diabetogenic factors play a role in progression to T1D in this subset of relatives at risk, therapies addressing the underlying mechanisms could be tried to prevent or delay development of T1D.

Here, we aimed to test whether T2D-associated TCF7L2 genetic variants affect insulin sensitivity or secretion in autoantibody-positive relatives without clinical T1D (i.e., stage 1 or 2 of the disease).

TrialNet, established in 2000, is a National Institutes of Health–funded international network of centers that aims to prevent T1D and stop disease progression (10). The TrialNet Pathway to Prevention (PTP) study is an observational study that prospectively follows nondiabetic, islet autoantibody–positive individuals (first- or second-degree relatives of patients with T1D) for the development of islet autoimmunity and progression to clinical T1D (11). Of 4,041 TrialNet PTP participants enrolled to date, we included those autoantibody-positive participants who had available data for baseline oral glucose tolerance test (OGTT) and TCF7L2 SNP information (n = 1,093). We examined OGTT results and excluded participants who had fasting or 2-h glucose values that met the criteria for diabetes diagnosis (n = 4) (Supplementary Figure 1). In addition, we excluded 28 individuals who were younger than 2 years of age at baseline and therefore their BMI data could not be used in the analysis. The final study sample included 1,061 participants. All study participants provided informed consent prior to screening and enrollment, and the study was approved by the responsible ethics committee at each site.

All participants were screened for autoantibodies to GAD65 (GAD65A), insulin (microinsulin autoantibody assay [miAA]), and IA-2 antigen (IA-2A). If any of these were positive at screening, autoantibodies to zinc transporter 8 and islet cell antibodies (ICAs) were also tested. Once identified as being autoantibody positive, subsequent evaluations of autoantibodies for participants included testing for all five T1D-associated autoantibodies (11). Islet autoantibody (12) and C-peptide (13) assays have been previously described. HLA genotyping was performed at the T1D Genetics Consortium Laboratories. The Illumina Immunochip was used to genotype TCF7L2 SNPs at the Center for Public Health Genomics at the University of Virginia. Immunochip is a custom array of 186,000 SNPs selected from regions of the genome robustly associated with autoimmune diseases. Because TCF7L2 rs7903146 was not included in Immunochip, we studied rs4506565 and rs7901695, which are in strong linkage disequilibrium (r² between 0.780 and 0.912 and between 0.780 and 0.980, respectively, in populations of European ancestry) with rs7903 146 (https://ldlink.nci.nih.gov).

We tested the effect of TCF7L2 genotype on indices of insulin sensitivity and secretion calculated with fasting and OGTT-stimulated glucose, insulin, and C-peptide levels obtained at baseline enrollment. Fasting and OGTT-derived insulin indices have been validated against the gold-standard hyperinsulinemic-euglycemic and hyperglycemic clamp studies in children (14,15) and proven useful for large cohort studies, including youth with T2D, where more intensive measures of insulin sensitivity and secretion are not feasible (16).

For insulin sensitivity, we evaluated 1/fasting insulin (1/I_F), an index validated against the hyperinsulinemic-euglycemic clamp in children (17) and used in studies of T1D (18) and T2D (16,19). As an index of insulin secretion reflecting the first phase of insulin response, we studied the 30-min C-peptide index (ΔCpep₃₀/ΔG₃₀) (14,15,20,21), which is based on the increment in C-peptide plasma concentration relative to the increment in plasma glucose over the first 30 min after glucose challenge, according to the following formula: 30-min C-peptide index = (ΔC₃₀)/(ΔG₃₀), where ΔC₃₀ = (C₃₀ − C₀) ng/mL ΔG₃₀ = (G₃₀ − G₀), and C₃₀ indicates C-peptide at 30 min and C₀ C-peptide at 0 min.

We also assessed the OGTT-derived disposition index or oral disposition index (oDI), which reflects β-cell function relative to insulin sensitivity, is related to glycemic regulation (16), and predicts the development of diabetes (22). The original formula for the oDI is as follows: oDl = (1/I_F)(30-min C-peptide index).

The relationship between log (1/I_F) and log (ΔC₃₀/ΔG₃₀) had a slope of −0.298 (95% CI −0.35 to −0.24), statistically different from 0, which has been deemed acceptable as quasi hyperbolic (23). However, given that the slope was different from −1 (P < 0.001), to satisfy the prerequisite of a slope of −1 and verify the validity of the use of the oDI in our study population, we applied a correction to the oDI formula (oDIc), as follows: oDIc = (1/I_F)³ (30-min C-peptide index).

The relationship between log (1/I_F)³ and log (ΔC₃₀/ΔG₃₀) has a slope of −0.894 (95% CI −1.05 to −0.73) and is not significantly different from −1 (P = 0.198). Of note, these OGTT-derived indices have been similarly used in assessment of insulin sensitivity and β-cell function in large clinical studies in adults and youth with T2D (16,19,24).

Characteristics were summarized across the overall cohort as well as by group using descriptive statistics. Continuous measures were compared across groups using Kruskal-Wallis and Wilcoxon rank sum tests for pairwise comparisons between any two groups of interest. The Spearman rank correlation test was used for age and BMI z score, Wilcoxon rank sum test for ethnicity and sex, and Kruskal-Wallis test for TCF7L2 SNPs. The χ² test was used to evaluate differential distributions of characteristics and factors across groups. We tested the difference in BMI z score between TCF7L2 allele groups using ANOVA. For each index, a generalized linear regression model was used to compare TCF7L2 allele groups, adjusting for known confounders such as age and BMI z score. The effect of the interaction between TCF7L2 SNP and BMI z score on oDIc (primary outcome) and insulin sensitivity and secretion indices (secondary outcomes) was studied. To determine if the analyses were sensitive to those with an abnormal baseline OGTT, all analyses were repeated limiting the population to just those with a normal OGTT. Because of the nonnormal distribution of the indices, for all parametric testing and modeling, the indices were log transformed. All analyses were performed using the statistical program R (version 3.5.1 for Windows; https://www.r-project.org/) and SAS (version 9.4). To account for multiple comparisons (multiple indices and tests examined), Bonferroni correction was used to maintain a family-wise error rate of 0.05. P values <0.0086 were considered statistically significant.

The study population consisted of 1,061 participants with a mean age of 16.3 years (SD 13.0) (72.4% were <18 years old) and age- and sex-adjusted BMI z score of 0.6 (SD 1.1); 47.7% were male and 91.1% were non-Hispanic White. Seventy percent (n = 743) had a normal OGTT, and the remaining participants (n = 318; 30%) had dysglycemia. Two T2D-associated alleles were present in 8.0% of the participants, one allele in 43.9%, and none in 48.1%. Other characteristics of the study cohort are listed in Supplementary Table 2.

In univariate analysis, 1/I_F was inversely correlated with BMI z score (r = −0.20; P < 0.001) and was associated with being non-Hispanic White (P = 0.002). The C-peptide index correlated with age (r = 0.37; P < 0.0001) and BMI z score (r = 0.28; P < 0.0001) but not with ethnicity. The oDIc, a measure of β-cell function relative to insulin sensitivity, was negatively correlated with age (r = −0.20; P < 0.0001), BMI z score (r = −0.36; P < 0.0001), and ethnicity (higher in non-Hispanic individuals) (P = 0.008) but not with sex. None of the indices of insulin sensitivity or secretion were associated with TCF7L2 SNPs (zero vs. one or two alleles) or sex (Table 1). To better understand the relationship between TCF7L2 variants and BMI z score, we compared BMI z score by TCF7L2 genotype. The mean (SD) BMI z score in participants with zero, one, and two alleles were 0.63 (1.04), 0.60 (1.12), and 0.64 (1.03), respectively, with no statistical difference (P = 0.912) (Supplementary Figure 2).

In multivariable analysis models including age, BMI z score, ethnicity, sex, and TCF7L2, 30-min C-peptide index was significantly increased with higher BMI z score (P < 0.001) and older age (P < 0.001), 1/I_F with lower BMI z score (P < 0.001), and oDIc with lower BMI z score (P < 0.001) (Table 2).

When limiting the multivariable analysis to adults (≥18 years old) (n = 293), the only relationship that persisted as significant was BMI z score with 1/I_F (P < 0.001) and oDIc (P = 0.007) (Table 3). In children (n = 768), both age and BMI z score were independent positive predictors of 30-min C-peptide index (P < 0.001) and negative predictors of 1/I_F (P < 0.001). While age and BMI z score were negative predictors for oDIc (both P < 0.0001), male sex was positively associated with 1/I_F (P < 0.004) and oDIc (P < 0.0001) (Table 4). Supplementary Table 3 summarizes the interaction between BMI and TCF7L2 genetic variants.

Restricting the analysis to participants who did not have dysglycemia, as reflected by a normal OGTT (n = 743; 70% of the cohort), yielded similar results. Analysis of unrelated participants (n = 1,022) only, after exclusion of 39 individuals from 35 multimember families, still indicated that none of the indices of insulin sensitivity or secretion were associated with TCF7L2 SNPs, comparing zero versus one or two alleles (Supplementary Table 4A–D) or comparing zero versus two alleles (Supplementary Table 5A–D) in univariable or multivariable analysis with adjustment for age, BMI z score, ethnicity, and sex in the overall cohort, children (<18 years old), and adults (≥18 years old) at the 0.0086 significance level. We conducted similar analyses separately in non-Hispanic White (n = 839) (Supplementary Table 6A–D), Hispanic (n = 85), and African American (n = 72) participants and found no statistical differences between those with zero and those with one or two T2D-associated TCF7L2 alelles at the 0.0086 significance level.

In this large cohort of autoantibody-positive participants followed in the TrialNet PTP study (N = 1,061), we did not observe an effect of T2D-associated TCF7L2 genetic variants on indices of insulin sensitivity or secretion after accounting for BMI z score, age, sex, and race/ethnicity. The same results were obtained when the analyses were restricted to participants with a normal OGTT and when stratifying by age (adults vs. children).

Previous studies have demonstrated that TCF7L2 variants are associated with abnormalities in insulin secretion, proinsulin processing, β-cell development in animal and human cellular models, and insulin sensitivity in individuals with T2D as well as in healthy controls (3–5,25–27). We previously demonstrated that TrialNet participants with new-onset T1D and T2D-associated TCF7L2 variants had higher AUC C-peptide and lower AUC glucose in mixed meal tolerance tests (6). Furthermore, in the DPT-1 study, Greenbaum et al. (28) identified a subset of participants with T1D diagnosed by 2-h OGTT criteria alone, with normal fasting glucose. These individuals with T1D had normal glucose homeostasis in the fasting state but showed abnormalities in islet function in response to glucose challenge, with low first phase of insulin release and low insulin response to arginine, low incretin effect, and decreased suppression of glucagon after i.v. and oral glucose (29), suggesting the presence of metabolic defects typically seen in T2D. Here, we aimed to test whether T2D-associated TCF7L2 genetic variants could be linked with abnormalities in insulin secretion and sensitivity in autoantibody-positive relatives of those with T1D. However, in this cohort of individuals at risk for T1D, we did not find differences in insulin sensitivity or secretion indices by TCF7L2 genotype.

A potential explanation for these negative findings in autoantibody-positive relatives is that subclinical abnormalities in insulin secretion that precede clinical onset (30,31) may minimize differences in insulin secretion and/or sensitivity that could be determined by TCF7L2 genetic variants. Furthermore, heterogeneity in the degree of glucose metabolism deterioration within the population of autoantibody-positive individuals could mask differences among groups; however, restricting the analysis to participants with a normal OGTT did not yield different results. In this study, we were not able to study glucagon or incretins, which could be affected in carriers of T2D-associated TCF7L2 SNP variants, as has been shown in individuals with T2D (25), obese adults (26), and obese adolescents with normal or impaired glucose tolerance (27). Finally, the group with two T2D-associated TCF7L2 SNP alleles was relatively small, because only 8% of participants were homozygous for the variant, which is consistent with previously published data in the general population (32).

In our study population, we did not observe an association between TCF7L2 SNP variants and BMI z score. An interaction between TCF7L2 and increased BMI has been previously reported, where the TCF7L2 risk allele and being overweight increased the risk of T2D (33), but this study was conducted in an exclusively adult population, which could explain the difference.

We observed that several factors affected the insulin indices under study. Older age was associated with higher insulin secretion (C-peptide index) and higher oDIc. In children, the relationships between age and insulin sensitivity, insulin secretion, and oDIc persisted after adjustment for BMI, sex, and ethnicity. In adults, after adjustment for age, sex, and ethnicity, BMI z score remained negatively related to insulin sensitivity and oDIc. Sex and ethnicity did not influence any of the insulin indices in the multivariable analysis, except in children, where boys had higher insulin sensitivity and oDIc than girls. These expected relationships of BMI and age with insulin sensitivity and secretion indices are in contrast to the absence of a relationship of TCF7L2 genetic variants with these indices in our study population.

Important strengths of the study are the large overall sample size with data on SNP genotyping and OGTT. Among the limitations of the study are those inherent to the surrogate estimates of insulin sensitivity and secretion, which are not validated in populations at risk for T1D, although some studies have relied on fasting indices of insulin resistance in individuals with T1D (18). OGTT-derived indices had reasonable prognostic performance in predicting the development of T1D in moderate-risk individuals with a positive family history of T1D and positive antibodies but normal glucose tolerance (34). In obese youth with T1D, fasting and OGTT-derived indices of insulin sensitivity were more strongly related to clamp-derived measures than oral disposition index measures (35). This may be due to the fact that, with autoimmune β-cell dysfunction, the feedback loop between insulin sensitivity and secretion is disrupted. However, our sensitivity analysis that included only individuals with a normal OGTT showed similar results. Nevertheless, it is important to assess β-cell function relative to insulin sensitivity, and future studies must address these relationships more rigorously in individuals with or at risk for T1D, along with measures of incretin effect. It is possible that more direct measurements of in vivo insulin sensitivity and secretion in individuals with T1D will be able to uncover differences that were not captured by the use of OGTT-derived indices in the current study.

In summary, TCF7L2 genetic variation was not related to indices of insulin sensitivity or secretion in autoantibody-positive relatives of individuals with T1D. Additional studies are required to understand whether this genetic factor may have an effect on glycemia through its influence on glucagon and incretin action.

Funding. This trial was sponsored by the Type 1 Diabetes TrialNet Study Group, a clinical trials network funded by the National Institutes of Health through the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Allergy and Infectious Diseases, and the Eunice Kennedy Shriver National Institute of Child Health and Human Development through cooperative agreements U01 DK061010, U01 DK061034, U01 DK061042, U01 DK061058, U01 DK085465, U01 DK085453, U01 DK085461, U01 DK085466, U01 DK085499, U01 DK085504, U01 DK085509, U01 DK103180, U01 DK103153, U01 DK085476, U01 DK103266, U01 DK103282, U01 DK106984, U01 DK106994, U01 DK107013, U01 DK107014, and UC4 DK106993 and JDRF.

The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health or JDRF.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.J.R. contributed to study design, as well as data analysis and interpretation, and wrote the first draft of the manuscript. M.V.W., L.E.B., D.C., and S.G. contributed to data analysis. I.M.L., A.P., A.K.S., C.E.-M., D.B., and J.M.S. contributed to data interpretation. F.B. contributed to study design, data analysis and interpretation, and writing. All authors revised and edited the manuscript. M.J.R., M.V.W., I.M.L., L.E.B., D.C., S.G., A.P., A.K.S., C.E.-M., D.B., and J.M.S. were members of the Type 1 Diabetes TrialNet Study Group at the time of the study (Supplementary Table 1). M.J.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were communicated as an oral presentation at the 80th Scientific Meeting of the American Diabetes Association, San Francisco, CA, 12–16 June 2020.