Type 2 Diabetes Genetic Risk and Type 1 Diabetes Heterogeneity and Progression

Abstract

Insulin secretion varies widely in preclinical type 1 diabetes. To understand the pathogenesis of this metabolic heterogeneity, we asked whether genetic predisposition to type 2 diabetes, quantified by a type 2 diabetes genetic risk score (T2D-GRS), modulates β-cell function and disease progression in individuals at risk of type 1 diabetes. We analyzed 4,324 islet autoantibody–positive TrialNet Pathway to Prevention participants with genome-wide genotyping and oral glucose tolerance testing. Both T2D-GRS and the type 1 diabetes genetic risk score 2 (T1D-GRS2) differed significantly across five previously described groups defined by C-peptide area under the curve (AUC; a measure of insulin secretion). The highest C-peptide AUC group, compared with the lowest, had significantly higher T2D-GRS, lower T1D-GRS2, higher BMI z-score, greater insulin resistance, older age, and lower prevalence of male participants; multiple islet autoantibody positivity; and IA-2 or insulin autoantibody positivity. Progression to clinical (stage 3) type 1 diabetes was significantly associated with T1D-GRS2 across all groups and with T2D-GRS in all but the lowest C-peptide AUC group. In conclusion, type 2 diabetes genetic burden shapes metabolic heterogeneity and accelerates progression in preclinical type 1 diabetes. These results support the evaluation of type 2 diabetes–related mechanisms as targets to improve the prediction and prevention of type 1 diabetes.

Article Highlights

• Heterogeneity in β-cell function is a barrier to precision medicine in type 1 diabetes.

• We asked whether type 2 diabetes–associated genes influence insulin secretion and progression to clinical type 1 diabetes in autoantibody-positive individuals.

• A type 2 diabetes genetic risk score was associated with higher C-peptide area under the curve (AUC) and increased clinical type 1 diabetes risk in all but the lowest C-peptide AUC subgroup.

• Addressing type 2 diabetes mechanisms could improve type 1 diabetes prediction and prevention.

Graphical Abstract

Introduction

Despite advances in treatment, individuals with type 1 diabetes, their families, and society continue to face considerable burdens, with health outcomes remaining suboptimal (1,2). Individuals at increased risk of type 1 diabetes, identified by predictive models (3), are eligible for treatments that delay its onset (4). Furthermore, several agents have been shown to slow disease progression, as evidenced by an attenuated decline in insulin secretory capacity after the onset of stage 3 (clinical) type 1 diabetes (5–7). Nevertheless, responses to treatment are neither universal nor durable (8). New interventions, either alone or in combination with those proven effective so far, are warranted to elicit or prolong protective effects in individuals at risk of or diagnosed with type 1 diabetes. Heterogeneity in type 1 diabetes is manifested not only in variable responses to treatment but also in diverse demographic, clinical, immunologic, metabolic, and genetic characteristics (9,10). Some autoantibody-positive individuals exhibit classic type 1 diabetes features (e.g., young age, White race, low BMI, high prevalence of multiple islet autoantibodies, and presence of type 1 diabetes–associated genes), whereas others lack most of these traits but share a similar risk of progression to clinical diabetes (11). A deeper understanding of the pathophysiologic mechanisms underlying heterogeneity in type 1 diabetes will improve the accuracy of predictive models and create opportunities for novel interventions.

We previously stratified autoantibody-positive relatives of individuals with type 1 diabetes into metabolic zones defined by insulin secretion (measured as C-peptide area under the curve [AUC]) and glucose AUC from oral glucose tolerance tests (OGTTs) (12). We found that participants in the highest C-peptide AUC group were older and had a higher prevalence of obesity, a lower prevalence of multiple autoantibody positivity and type 1 diabetes–associated HLA-DRB1*03:01-DQA1*05:01-DQB1*02:01 (DR3-DQ2) or HLA-DRB1*04-DQA1*03:01-DQB1*03:02 (DR4-DQ8) haplotype, and slower progression to stage 3 type 1 diabetes. Here, we tested the hypothesis that autoantibody-positive individuals in the higher C-peptide AUC groups would have greater type 2 diabetes genetic risk and that this risk would influence progression to stage 3 type 1 diabetes. Uncovering the contribution of type 2 diabetes–related genes to type 1 diabetes heterogeneity and natural history will improve type 1 diabetes prediction and prevention.

Research Design and Methods

Participants

Type 1 Diabetes TrialNet is a National Institutes of Health–funded international consortium that aims to prevent type 1 diabetes and stop disease progression (13). Its observational arm, the Pathway to Prevention Study (ClinicalTrials.gov identifier NCT00097292), has screened diabetes-free relatives of individuals with type 1 diabetes (first-degree relatives aged 2–45 years and second- or third-degree relatives aged 2–20 years) for the presence of islet autoantibodies since 2004. Autoantibody-positive participants are followed for the progression of islet autoimmunity and/or the development of stage 3 type 1 diabetes, as previously described (14).

We analyzed 4,324 Pathway to Prevention participants with single confirmed or multiple autoantibody positivity who had OGTT-derived C-peptide AUC and glucose AUC data, as well as genome-wide single nucleotide polymorphism (SNP) genotyping. Genome-wide association screening was performed with the TEDDY array, an Illumina HumanCoreExome BeadArray including ∼90,000 SNPs as additional custom content, selected from regions of the genome robustly associated with autoimmune diseases and pathways relevant to type 1 diabetes initiation and progression, as previously described (15). Participants were screened for islet autoantibodies to GAD (GADA), insulin (IAA), and IA-2 (IA-2A). Autoantibody-positive individuals underwent confirmation evaluation, including testing for all five type 1 diabetes–associated autoantibodies (i.e., GADA, IAA, IA-2A, islet cell antibody, and ZnT8 autoantibody) (14). Assays for islet autoantibodies (16,17) and C-peptide (18) have been previously described. HLA genotyping was performed at the TrialNet Core Laboratory at the Barbara Davis Center for Diabetes (Aurora, CO). All study participants provided written informed consent or assent before screening and enrollment, and the responsible ethics committee at each site approved the study.

Genetic Risk Scores

The type 1 diabetes genetic risk score 2 (T1D-GRS2) (19) and type 2 diabetes genetic risk score (T2D-GRS) (20) were used to quantify genetic burden for types 1 and 2 diabetes. Both T1D-GRS2 and T2D-GRS help discern diabetes type, predict diabetes risk, and determine disease pathogenesis and heterogeneity (19–23). T1D-GRS2 includes 67 SNPs (19), of which 30 were directly genotyped. We imputed the remaining non-HLA SNPs (median R² = 0.997; range 0.858–0.999) using the Trans-Omics for Precision Medicine (TOPMed) Imputation Server (24,25). We also imputed five SNPs in the HLA region (rs72848653 [R² = 0.999], rs9266268 [R² = 0.999], rs16899379 [R² = 0.998], rs2524277 [R² = 0.995], and rs9268500 [R² = 0.925]) using the Michigan Imputation Server with a high-resolution HLA reference panel spanning five global populations (n = 21,546) based on whole-genome sequencing data (26). Code for generating the HLA interaction part of T1D-GRS2 is freely available online (https://github.com/sethsh7/PRSedm). T2D-GRS was calculated from the Polygenic Score Catalog entry PGS002308 (https://www.pgscatalog.org/score/PGS002308/) (20). Genome-wide association study genotype imputation was performed with the TOPMed Freeze 8 reference panel (built from 97,256 deeply sequenced human genomes encompassing >308 million genetic variants) via the TOPMed Imputation Server (24,25). We retained variants with minor allele frequency ≥0.5% for imputed data and with imputation quality R² ≥ 0.80. In total, 238,760 genotyped and 887,780 imputed SNPs were used to calculate T2D-GRS using the Plink 1.9 score header sum flag.

Metabolic Measures

We used C-peptide and glucose values during the OGTT conducted at the baseline visit to stratify participants into 25 prespecified metabolic zones defined by five C-peptide AUC groups (<1.16 [n = 873], 1.16 to <1.66 [n = 1,151], 1.66 to <2.15 [n = 919], 2.15 to <2.65 [n = 639], and ≥2.65 nmol/L [n = 742]) and five glucose AUC groups (<6.38, 6.38 to <7.22, 7.22 to <8.05, 8.05 to <8.88, and ≥8.88 mmol/L), as previously described (12). Using the resulting bidimensional grid, we generated heat maps of T1D-GRS2 and T2D-GRS. As measures of insulin secretion, we analyzed C-peptide index, AUC ratio (27), and Index60 (28–30), which have been validated in autoantibody-positive individuals, calculated as follows: C-peptide index = (C-peptide30 − fasting C-peptide)/(glucose30 − fasting glucose), where C-peptide is expressed in ng/mL and glucose in mg/dL, and C-peptide30 and glucose30 denote C-peptide and glucose values at 30 min during the OGTT; AUC ratio = (C-peptide AUC/glucose AUC) ∗ 1,000, where C-peptide is expressed in ng/mL and glucose in mg/dL; and Index60 = 0.36953 (log fasting C-peptide) = 0.0165 ∗ glucose60 − 0.3644 ∗ C-peptide60, where C-peptide is expressed in ng/mL and glucose in mg/dL, and glucose60 and C-peptide60 denote blood glucose and C-peptide values at 60 min during OGTT.

We calculated the glucose fractions that are dependent (dAUCGLU) or independent (iAUCGLU) of insulin secretion, as previously described (31): dAUCGLUC = 132.00 + (12.64 ∗ Index60) and iAUCGLU = glucose AUC − dAUCGLU.

As a measure of insulin resistance, we calculated the HOMA for insulin resistance (HOMA-IR), which has been shown to influence progression to stage 3 type 1 diabetes in autoantibody-positive relatives (32), with the following formula: HOMA-IR = (fasting glucose ∗ fasting insulin)/405, where fasting glucose is expressed in mg/dL and fasting insulin in μU/mL.

Statistical Analyses

For each C-peptide AUC group, we calculated the median and interquartile range (IQR) of continuous demographic variables, T1D-GRS2, T2D-GRS, and metabolic measures, as well as proportions for categorical characteristics. These measures were compared across all C-peptide AUC groups using the Kruskal-Wallis test for continuous variables and the χ² test for categorical variables. To compare the highest and lowest C-peptide AUC groups, we applied the Wilcoxon rank-sum test for continuous variables and the χ² test for categorical variables.

Univariate Cox proportional hazards models of time to stage 3 type 1 diabetes were fit for T1D-GRS2 and T2D-GRS, each standardized to 1 SD unit, within each C-peptide AUC group. Hazard ratios (HRs), 95% CIs, and survival estimates were derived. We also plotted cumulative incidence of type 1 diabetes by GRS quartile, defined as follows: quartile 1 (Q1), 25th percentile (minimum); Q2, 25th–50th percentile; Q3, 50th–75th percentile; and Q4, 75th percentile (maximum). We plotted cumulative incidence of type 1 diabetes by each GRS quartile. All analyses were performed in SAS (version 9.4), and statistical significance was set at P < 0.05.

Data and Resource Availability

Original data generated and analyzed during this study are included in this published article or in the data repositories listed in references. Resource sharing does not apply, because this research did not create resources.

Results

Characteristics of the 4,324 autoantibody-positive relatives of individuals with type 1 diabetes included in this analysis are listed in Table 1. Overall, 49.5% of participants were female, with a median age of 11.3 years (IQR 7.2, 17.4) at enrollment. Participants had a median BMI of 18.7 kg/m² (IQR 16.1, 23.8) and BMI z score of 0.23 (IQR −0.7, 1.1). The HLA-DR3-DQ2/DR4-DQ8 genotype was present in 20.5% of participants; 86.6% of participants were White, and 9.6% were Hispanic or Latino.

Table 1.

Characteristics of the overall cohort of participants and by C-peptide AUC

Parameter	All participants	C-peptide AUC, nmol/L					P
							Overall comparison	Lowest vs. highest group
		<1.16	1.16 to <1.66	1.66 to <2.15	2.15 to <2.65	≥2.65
n	4,324	873	1,151	919	639	742	—	—
Female sex	49.54	38.83	45.09	52.61	58.53	57.49	<0.001	<0.001
Age, years	11.29 (7.17, 17.35)	6.01 (3.98, 8.26)	9.21 (6.46, 13.11)	12.92 (9.39, 20.86)	15.09 (11.30, 31.06)	16.42 (12.80, 36.82)	<0.001	<0.001
BMI z score	0.23 (−0.67, 1.08)	−0.04 (−0.75, 0.56)	0.08 (−0.76, 0.76)	0.13 (−0.84, 0.99)	0.45 (−0.68, 1.41)	1.18 (0.00, 2.12)	<0.001	<0.001
BMI, kg/m2	18.71 (16.08, 23.80)	15.89 (14.93, 17.06)	17.05 (15.54, 20.05)	19.86 (17.18, 23.80)	22.64 (18.96, 26.50)	26.11 (21.43, 31.08)	<0.001	<0.001
Race							<0.001	0.163
AI/AN	0.25	0.57	0.35	0	0	0.27
Asian	1.13	1.03	0.96	0.98	0.78	2.02
Black/AA	2.54	3.89	2.09	2.07	2.03	2.70
More than one race	2.04	1.95	1.82	2.50	1.72	2.16
NH/other Pacific	0.23	0.23	0.26	0.22	8.14	0.27
Unknown	7.24	5.84	6.95	7.40	0.16	8.36
White	86.56	86.48	87.58	86.83	87.17	84.23
Ethnicity							<0.001	<0.001
Hispanic or Latino	9.62	6.19	8.51	9.25	12.36	13.48
Not Hispanic or Latino	85.89	88.77	86.71	85.31	84.19	83.42
Unknown	4.49	5.04	4.78	5.44	3.44	3.10
Single autoantibody positive	44.24	27.49	41.88	47.88	50.86	57.41	<0.001	<0.001
IAA	20.18	32.50	19.92	16.14	18.77	18.78	<0.001	<0.001
IA-2A	6.85	8.33	8.51	6.36	7.08	4.46	0.141	0.041
GADA	72.97	59.17	71.58	77.50	74.15	76.76	<0.001	<0.001
Multiple autoantibody positive	52.57	71.25	55.95	48.31	43.35	38.54	<0.001	<0.001
IAA	47.20	70.79	49.44	40.04	37.87	32.88	<0.001	<0.001
IA-2A	39.66	51.32	41.53	38.30	33.18	30.32	<0.001	<0.001
GADA	82.45	82.82	83.41	83.03	80.13	81.81	0.460	0.595
HLA DR3-DQ2/DR4-DQ8 genotype	20.51	24.40	23.11	19.80	17.37	15.50	<0.001	<0.001
T1D-GRS2	13.52 (12.07, 14.84)	13.99 (12.84, 15.18)	13.72 (12.27, 14.97)	13.40 (12.01, 14.72)	13.10 (11.64, 14.53)	13.06 (11.36, 14.47)	<0.001	<0.001
HLA T1D-GRS2	9.54 (8.19, 10.84)	9.99 (8.74, 11.11)	9.68 (8.32, 10.93)	9.46 (8.09, 10.71)	9.27 (7.86, 10.69)	9.11 (7.65, 10.42)	<0.001	<0.001
Non-HLA T1D-GRS2	4.00 (3.45, 4.53)	4.08 (3.54, 4.55)	4.03 (3.53, 4.57)	4.00 (3.38, 4.52)	3.90 (3.31, 4.48)	3.96 (3.43, 4.47)	<0.001	0.006
T2D-GRS	0.37 (0.22, 0.51)	0.34 (0.20, 0.49)	0.36 (0.22, 0.51)	0.37 (0.22, 0.52)	0.37 (0.23, 0.51)	0.39 (0.25, 0.54)	0.001	<0.001

Data are given as percentage or median (IQR) unless otherwise indicated.

AA, African American; AI, American Indian; AN, Alaska Native; NH, Native Hawaiian.

Both T1D-GRS2 and T2D-GRS were significantly different across C-peptide AUC groups (P < 0.001 and P = 0.01, respectively). Heat maps revealed clear gradients in the distributions of T1D-GRS2 (Fig. 1A) and T2D-GRS (Fig. 1B) across metabolic zones. Consistent with these patterns, participants in the highest C-peptide AUC group had lower T1D-GRS2 than those in the lowest group (P < 0.001) (Fig. 1A and Table 1). They also had significantly lower HLA and non-HLA components of T1D-GRS2 (P < 0.001 and P = 0.006, respectively) (Supplementary Figs. 1 and 2) and a lower prevalence of the HLA DR3-DQ2/DR4-DQ8 genotype (P < 0.001). Conversely, the highest C-peptide AUC group had a higher T2D-GRS than the lowest group (P < 0.001) (Fig. 1B and Table 1). In contrast, both scores increased with rising glucose AUC; the highest glucose AUC group had significantly higher T1D-GRS2 and T2D-GRS than the lowest glucose group (both P < 0.001; data not shown).

Figure 1.

Heat maps of T1D-GRS2 (A) and T1D-GRS2 (B) by metabolic zones in autoantibody-positive relatives of individuals with type 1 diabetes. Metabolic zones are defined by C-peptide AUC and glucose AUC, with glucose AUC increasing from 1 to 5 on the y-axis and C-peptide AUC increasing from A to E on the x-axis.

We examined the influence of T1D-GRS2 and T2D-GRS on progression to stage 3 type 1 diabetes (Table 2). We observed that whereas T1D-GRS2 was a significant predictor in all C-peptide AUC groups (HR 1.54–1.89; all P < 0.001), T2D-GRS predicted progression in four of the five groups (HR 1.15–1.28; P = 0.01–0.04), but not in the lowest C-peptide AUC group (P = 0.56). Furthermore, cumulative incidence curves stratified by GRS (Fig. 2) showed that T1D-GRS2 predicted the development of stage 3 type 1 diabetes in both the lowest (HR 1.6; 95% CI 1.4–1.8) (Fig. 2A) and highest C-peptide AUC groups (HR 1.7; 95% CI 1.4–2.1) (Fig. 2B). In contrast, T2D-GRS predicted stage 3 disease in the highest C-peptide AUC group (HR 1.2; 95% CI 1.0–1.5) (Fig. 2D), but not in the lowest (HR 0.97; 95% CI 0.87–1.1) (Fig. 2C).

Table 2.

Univariate Cox regressions of time to stage 3 type 1 diabetes progression using T1D-GRS2 and T2D-GRS

	Maximum likelihood estimates
	df	Parameter estimate	SE	χ2	P *	HR	95% confidence limit
T1D-GRS2
Total population	1	0.53	0.04	228.18	<0.0001	1.70	1.59–1.83
C-peptide AUC, nmol/L
<1.16 (lowest)	1	0.45	0.07	47.99	<0.0001	1.57	1.38–1.78
1.16 to <1.66	1	0.47	0.06	55.42	<0.0001	1.60	1.41–1.81
1.66 to <2.15	1	0.43	0.08	27.92	<0.0001	1.54	1.31–1.80
2.15 to <2.65	1	0.64	0.13	25.04	<0.0001	1.89	1.48–2.43
≥2.65 (highest)	1	0.53	0.11	228.18	<0.0001	1.70	1.37–2.10
T2D-GRS
Total population	1	0.09	0.03	8.15	0.004	1.09	1.03–1.16
C-peptide AUC, nmol/L
<1.16 (lowest)	1	−0.03	0.05	0.34	0.56	0.97	0.87–1.08
1.16 to <1.66	1	0.14	0.05	6.40	0.01	1.15	1.03–1.28
1.66 to <2.15	1	0.18	0.08	5.70	0.02	1.20	1.03–1.39
2.15 to <2.65	1	0.25	0.10	6.42	0.01	1.28	1.061.56
≥2.65 (highest)	1	0.19	0.09	4.24	0.04	1.21	1.01–1.45

*Pearson χ² test.

Figure 2.

Cumulative incidence of stage 3 type 1 diabetes by T1D-GRS2 (A and B) and T2D-GRS (C and D) quartile in the lowest (A and C) and highest (B and D) C-peptide AUC groups.

We compared phenotypic traits associated with type 2 diabetes across C-peptide AUC groups (Table 1). Participants in the highest C-peptide AUC group were older (median 16.4 vs. 6.0 years; P < 0.001), more likely to be female (57.5% vs. 38.8%; P < 0.001), more often single autoantibody positive (57.4% vs. 27.5%; P < 0.001), less likely to be non-Hispanic (83.4% vs. 88.8%; P < 0.001), and had higher BMI z scores (median 1.2 vs. −0.04; P < 0.001). The prevalence of IAA and IA-2A positivity was lower in the highest (32.9% and 30.3%, respectively) than in the lowest C-peptide AUC group (70.8% and 51.3%, respectively; P < 0.001 for both comparisons), whereas GADA positivity did not differ (81.8% and 82.8%, respectively; P = 0.6). Metabolic markers of insulin secretion and insulin resistance varied significantly across all C-peptide AUC groups (all P < 0.001) (Supplementary Table 1). Participants in the highest C-peptide AUC group had a higher C-peptide index, higher AUC ratio, lower Index60, higher HOMA-IR, and higher independent glucose fraction than those in the lowest group (all P < 0.001).

Discussion

In this large cohort of 4,324 autoantibody-positive individuals, we found higher T2D-GRS values among participants in the higher C-peptide AUC groups, providing a genetic explanation for the insulin-resistant phenotype (e.g., older age, higher BMI z score, and elevated HOMA-IR) that we previously observed in autoantibody-positive individuals with elevated C-peptide (12). Moreover, T2D-GRS predicted progression to stage 3 type 1 diabetes in all but the lowest C-peptide group. Accordingly, T2D-GRS2 serves as a clinically actionable biomarker for refined risk stratification and supports the evaluation of type 2 diabetes–directed therapies for type 1 diabetes prevention.

Heat map visualizations clearly reflected a pronounced genetic contrast between the lowest and highest C-peptide AUC groups. Moreover, within the highest C-peptide AUC group, the intensity of color indicative of T2D-GRS values was greater in participants with the highest glucose AUC than in those with the lowest; this gradient was absent in the lowest C-peptide AUC group. The striking metabolic heterogeneity was confirmed by formal statistical testing. Even among autoantibody-positive individuals without diabetes, the burden of type 2 diabetes–related genetic variants tracked with higher glucose levels.

In contrast with T2D-GRS, T1D-GRS2 was lower in the groups with higher C-peptide AUC. Partitioning this score showed that both its HLA and non-HLA components were lower in participants with elevated C-peptide. This finding extends our earlier observation that autoantibody-positive relatives lacking the high-risk HLA-DR4-DQ8 or HLA-DR3-DQ2 haplotype have higher C-peptide, older age, higher prevalence of obesity, and lower prevalence of White race (33). Consistent with our previous findings (11,12,31), participants in the groups with higher C-peptide AUC had fewer canonical type 1 diabetes features, including a lower prevalence of multiple autoantibody positivity and a higher independent glucose fraction. Interestingly, although IAA and IA-2A were less frequently positive in those in the higher C-peptide AUC, no significant differences were observed for GADA positivity, reinforcing evidence that GADA marks a distinct pathogenic pathway (34). Taken together, these data demonstrate that genetic differences underlie phenotypic heterogeneity in type 1 diabetes (12), strengthening the case for precision medicine (9). Tailoring interventions requires biomarkers to identify subsets of individuals, and our study supports the use of C-peptide AUC as a marker of metabolic heterogeneity.

T2D-GRS was a significantly influential factor of progression to stage 3 type 1 diabetes in all but the participants in the lowest C-peptide AUC group. This observation has implications for type 1 diabetes prediction and prevention. T1D-GRS2 (19) and T1D-GRS (a previous version containing only 30 SNPs) (21) have been shown to improve the performance of models to predict progression to type 1 diabetes in individuals at risk (23,35). Our data suggest that T2D-GRS could further improve the type 1 diabetes prediction model. Given the high prevalence of genetic (and environmental) factors that predispose individuals to type 2 diabetes in the general population, it is not surprising that autoantibody-positive individuals are not spared. Therefore, certain autoantibody-positive individuals with type 2 diabetes–associated genes would be expected to have insulin resistance and β-cell dysfunction of nonautoimmune origin. Of note, the effect of the T2D-GRS on progression to stage 3 type 1 diabetes was smaller than that of the T1D-GRS2, likely reflecting the slower course of the natural history of preclinical type 2 diabetes compared with type 1 diabetes. Our prior studies have suggested that pathogenic factors associated with type 2 diabetes are present in individuals at risk of type 1 diabetes (36). Type 2 diabetes–associated variants in the TCF7L2 gene are more prevalent in those with single islet autoantibody positivity (37) and indicators of insulin resistance (38) and are also associated with a milder loss of insulin-containing islets in human pancreas donors with type 1 diabetes (39). Therefore, type 2 diabetes genetic variants may define a distinct subset of individuals with type 1 diabetes. Insulin resistance was associated with a higher risk of progression to stage 3 type 1 diabetes in adult participants in TrialNet (40) and in participants at moderate or high risk in the Diabetes Prevention Trial-1 (41). Similarly, in individuals diagnosed with type 2 diabetes, the presence of islet autoimmunity has been associated with rapid β-cell decline and worse glycemic control (42), further supporting the concept of multiple diabetogenic mechanisms coexisting in an individual. It will be important to determine the extent to which elevated glycemia in autoantibody-positive individuals is determined by pathophysiologic mechanisms of type 2 diabetes and whether these mechanisms interact with those associated with islet autoimmunity and type 1 diabetes.

A plausible explanation for the lack of an association of T2D-GRS with progression to stage 3 type 1 diabetes in the group with the lowest C-peptide AUC is that when β-cell mass or function is already severely compromised, minimal room is left for additional influences, such as insulin resistance, to modulate further progression. This finding further reinforces the concept of heterogeneity in type 1 diabetes. Overall, our observations substantiate the presence of type 2 diabetes–related pathogenic factors among autoantibody-positive individuals at risk of type 1 diabetes and support the consideration of clinical trials to test preventive and therapeutic interventions addressing those mechanisms.

This study is limited in that it is a secondary analysis of existing data. Therefore, variables are missing, notably social determinants of health, which would broaden the perspective of our analysis. Furthermore, those in the study population with a non-European background were in the minority. Because the prevalence of type 2 diabetes is greater in non-European individuals (particularly in pediatrics), this limitation may have caused an underestimation of type 2 diabetes admixture in our findings. Future studies should examine whether type 2 diabetes genetic risk influences autoantibody conversion dynamics, such as transitions from single to multiple autoantibody positivity. In addition, studies that model the combined effects of T1D-GRS2 and T2D-GRS will be important to establish whether an integrated score improves prediction beyond the individual components.

In conclusion, our study shows that autoantibody-positive individuals at risk of type 1 diabetes with the highest C-peptide AUC have genetic and phenotypic features of type 2 diabetes. T2D-GRS was associated with an increased risk of progression to stage 3 type 1 diabetes among autoantibody-positive individuals with higher C-peptide levels. These results suggest that the development of diabetes in autoantibody-positive individuals might have a mixed etiology and involve pathogenic mechanisms typical of both types 1 and 2 diabetes. These findings have implications for the prediction and prevention of type 1 diabetes.

This article contains supplementary material online at https://doi.org/10.2337/figshare.30593381.

Funding Statement

This work was supported by National Institutes of Health (NIH) grants R01 DK121843, NIH R01 DK124395, and UC4 DK106693. T.M.T. is supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant K23 DK136931. Type 1 Diabetes TrialNet is funded by the NIH through NIDDK grant R01 DK124395; the National Institute of Allergy and Infectious Diseases; 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, UC4 DK106993, and UC4 DK117009-01; and Breakthrough T1D (formerly JDRF).