Clonal Hematopoiesis of Indeterminate Potential (CHIP) and Incident Type 2 Diabetes Risk

Clonal hematopoiesis of indeterminate potential (CHIP) is characterized from DNA sequencing of peripheral blood as the presence of an expansion of a somatic mutation acquired in a progenitor blood stem cell. Prevalence of one or more CHIP mutations increases notably at older ages (≥65 years) (1), and its occurrence has been associated with an approximately twofold greater risk of developing coronary heart disease (CHD), particularly for carriers of somatic mutations in DNMT3A, TET2, ASXL1, and JAK2 (2,3). Cross-sectionally, among patients without prevalent heart disease, having CHIP was associated with a threefold higher coronary artery calcification score, underscoring a potential role in atherosclerotic progression (3). Although mechanisms underlying these relationships are unknown, potential pathways include increased inflammation or impaired immune function (4).

Given its strong correlation with age and CHD, CHIP may potentially be implicated in type 2 diabetes (T2D), but prior research in humans is sparse (4). Cross-sectionally, among individuals with obesity, baseline predictors of their subsequent clonal expansion rate (increase of variant allele frequency [VAF]) included insulin, HOMA of insulin resistance, and lower HDL cholesterol levels, after accounting for BMI level, suggesting a potential role of poor cardiometabolic health per se in CHIP progression, although directionality is unclear (5). Experimental mouse models of induced clonal expansion of TET2 mutation demonstrated an acceleration of aging-induced cardiometabolic dysfunction, including greater atherosclerosis, insulin resistance, and impaired fasting glucose (6). Separately, TET2 expansion in mice with diet-induced obesity significantly enhanced progression to insulin resistance (6), while in another case it was observed that a mouse model of obesity-driven inflammation led to greater CHIP expansion (7).

We therefore prospectively evaluated the relationship of CHIP with incident T2D in the Trans-Omics for Precision Medicine (TOPMed) program. TOPMed benefits from cohorts with high-coverage whole-genome sequencing (WGS) of stored blood samples (8), phenotyping for T2D risk factors, and longitudinal follow-up for incident T2D (9). In addition to evaluating baseline CHIP carrier status, we investigated CHIP on five previously identified CHD-related genes to identify their potential overlap with T2D risk.

In our analyses we included 17,637 participants from six TOPMed cohorts that met criteria of having at least 1,000 genotyped participants with derivation of CHIP status and longitudinal follow-up for incident T2D, including the Coronary Artery Risk Development Study in Young Adults (CARDIA), Cardiovascular Health Study (CHS), Framingham Heart Study (FHS), Jackson Heart Study (JHS), Multi-Ethnic Study of Atherosclerosis (MESA), and the Women’s Health Initiative (WHI) (Supplementary Fig. 1). Participants with prevalent T2D were excluded to avoid potential for reverse causation from blood draws occurring after diagnosis. We also excluded those with a history of cardiovascular disease (coronary artery disease, CHD, myocardial infarction, stroke [acute ischemic stroke, cerebrovascular accident, or acute hemorrhagic stroke]) or cancer diagnosed before blood collection, other than nonmelanoma skin cancers, based on available data in each cohort. See Supplementary Material for additional cohort-specific information. This study was approved by the respective institutional review boards of individual cohorts, and informed consent was obtained from all participants.

Blood DNA-derived high-coverage WGS was performed at the Broad Institute of MIT and Harvard as part of previous TOPMed research projects (10). WGS data were analyzed with GATK Mutect2 somatic variant caller, as previously described in detail (3,8); additional CHIP calling on freeze 9 was used for this publication with similar methods. CHIP was previously defined in TOPMed as the prevalence of a somatic VAF ≥2% on ≥1 of 74 prespecified driver mutations of hematopoietic stem cell expansion, and the cut point of VAF ≥10% was observed to be associated with elevated risk of developing CHD, while clonal expansion of VAF <10% carried only nominal risk (2). Thus, we adapted CHIP prevalence of ≥10% as our primary exposure of interest, with sensitivity analyses to examine the impact of VAF.

Details of the cohort-specific methods for ascertainment of T2D and related covariates have previously been published (11) and are summarized in Supplementary Material. Briefly, participants without T2D at genotyping blood draw were followed until date of T2D diagnosis, meeting one or more of the following criteria: fasting glucose ≥7 mmol/L, HbA_1c ≥6.5%, ≥11.1 mmol/L on 2-h oral glucose tolerance test, nonfasting glucose ≥11.1 mmol/L, physician-diagnosed T2D, self-reported T2D, or use of an antidiabetes medication. We harmonized phenotype data and additional covariate information from TOPMed cohorts, including age at blood draw, sex, BMI (weight in kilograms divided by the square of height in meters), waist circumference, smoking status, usual diet and alcohol intake, self-reported race/ethnicity (12), educational attainment, prediabetes status (not meeting diagnostic criteria for T2D with at least one of the following: fasting glucose between 5.6 and 7.0 mmol/L, HbA_1c between 5.7 and 6.5%, or between 7.8 and 11.1 mmol/L on 2-h oral glucose tolerance test), and baseline use of blood pressure or cholesterol-lowering medications. Among the cohorts a diet quality score was derived according to the Alternative Healthy Eating Index (AHEI), with collection of information on usual diet from food-frequency questionnaires (13). For cohorts with repeated assessments for any variables we used the data ascertained on or closest to blood draw.

Because a CHIP mutation can occur on more than one driver gene, we defined the exposure status as no CHIP versus at least one CHIP variant (yes/no). We further defined CHD-related CHIP as having a CHIP on one or more a priori selected genes that were previously related to incident CHD: DNMT3A, TET2, ASXL1, JAK2, or TP53 (14). We also categorized according to the total number of CHIP mutations: noncarriers versus carriers with one CHIP gene variant or with two or more. In a sensitivity analysis, we applied a more stringent definition of CHIP, limited to variants with VAF ≥15% and ≥20%. Participants without CHIP served as the exposure reference group for all analyses.

Baseline for the analyses was date of blood draw from which CHIP was derived. Participants’ follow-up was included until date of incident T2D diagnosis or last available visit—whichever came first. We used multivariable-adjusted Cox proportional hazards regression models to estimate the association of CHIP with incident T2D with stratification on age as the underlying timescale. Multivariable-adjusted models included sex (male/female), BMI (continuous), smoking status (never smoker, former, current), education (less than high school, high school or equivalent, some college, college degree or higher), and self-reported race/ethnicity (White, Black, Native American, Asian American, Hispanic, other). We adjusted for alcohol use except in the case of CHS, where this information was not available (nondrinker, light [women 1–14 g/day, men 1–28 g/day], moderate [women 15–28 g/day, men 29–42 g/day], heavy [women >28 g/day, men >42 g/day]). Indicator categories were used for missing categorical covariate data, including smoking status (<5% all cohorts) and alcohol (<1% to 40% across cohorts). BMI was missing for ∼1% and imputed as the cohort-specific median value. We cohort analyzed data separately and then combined the cohort-specific hazard ratios (HRs) and SEs using inverse variance–weighted fixed-effects meta-analyses to obtain the combined summary statistics and confirmed minimal between-study heterogeneity with I² values and P value for heterogeneity. We stratified the multivariable models by baseline characteristics to evaluate whether the association of CHIP with T2D incidence varied by baseline risk, including sex, age <60.0 vs. ≥60.0 years, BMI <30.0 vs. ≥30.0 kg/m², and self-reported race/ethnicity.

There were 17,637 participants eligible from 20,776 with data across six TOPMed cohorts (Supplementary Fig. 1). Mean (SD) for cohort participants in aggregate was 63.4 (11.5) years, ranging from 44.3 (6.5) years in CARDIA to 72.9 (5.2) years in CHS. Mean BMI was 28.4 (5.9) kg/m² overall, with lowest mean in CHS (26.6 [SD 4.6] kg/m²) and highest in JHS (31.4 [7.2] kg/m²). Of participants, 76% were female and 35% non-White, including 24% Black and 6% Hispanic.

CHIP was identified in 1,055 total participants (6.0%) overall, ranging in prevalence from 1.6% in CARDIA to 11.5% in CHS. Only 9% of these participants carried more than one CHIP mutation. Among CHIP carriers, most participants (n = 919 [87.1%]) carried a mutation on at least one of the five a priori defined CHD-related genes. The prevalence of CHIP accumulation was higher across older age categories, with 3.8%, 9.4%, 15.8%, and 23.1% for <70, 70–79, 80–89, and ≥90 years, respectively. Cohort-specific baseline demographics and health and lifestyle factors by CHIP status (none vs. at least one variant) are given in Table 1. Briefly, participants with CHIP were on average older but there were minimal trends for differences in lifestyle factors such as smoking status, alcohol use, BMI, or waist circumference.

Mean (SD) follow-up time from baseline blood draw was 9.8 (5.5) years, ranging from FHS with 6.2 (2.3) years to WHI with 12.2 (6.8) years. In combination, 2,467 cases of incident T2D were reported for the cohorts. The incidence rates for T2D in the non-CHIP reference groups were lowest for FHS (8.2 per 1,000 person-years) and highest for JHS (23.5 per 1,000 person-years). Results for the age- and multivariable-adjusted models of CHIP status with T2D risk are shown in Table 2. With adjustment for age, the meta-analyzed cohort estimates indicated a 22% higher risk of developing T2D (95% CI = 1.03, 1.44) for CHIP versus no CHIP. Results were similar after we additionally adjusted for sex, BMI, smoking, alcohol, race/ethnicity, and education (HR 1.23; 95% CI 1.04, 1.45), and there was minimal statistical heterogeneity between cohort estimates (I² = 27%, P = 0.23).

Among participants with prevalent CHIP at baseline, 88% were carriers of at least one a priori mutation implicated in CHD, and the relationship between CHD-CHIP and T2D was similar to that for the overall multivariable-adjusted results (HR 1.23; 95% CI 1.03, 1.46) (Fig. 1 and Supplementary Table 1). Individually, prevalence of CHD-related mutations indicated a higher T2D risk for TET2 carriers (HR 1.48; 95% CI 1.05, 2.08) and ASXL1 carriers (HR 1.76; 95% CI 1.03, 2.99), and possibly for DNMT3A carriers (HR 1.15; 95% CI 0.93, 1.43). JAK2 and TP53 mutations were relatively uncommon, and statistical power to assess T2D risk was low.

The multivariable-adjusted estimates for having one and two or more CHIP mutations with incident T2D risk were HR 1.23 (95% CI 1.03, 1.46) and HR 1.50 (0.80, 2.81), respectively (Supplementary Table 2). In a sensitivity analysis we implemented a higher threshold of VAF for defining CHIP and observed similar results for CHIP with VAF ≥15% compared with no CHIP (HR 1.36; 95% CI 1.11, 1.67); however, increasing the threshold to VAF ≥20% drastically reduced sample size, with only 44 total T2D cases, without indication of an association with T2D (HR 1.03; 95% CI 0.76, 1.39). In the multivariable models with stratification by baseline characteristics, we did not observe effect modification by sex, age <60.0 vs. ≥60.0 years, BMI <30.0 vs. ≥30.0 kg/m², or self-reported race/ethnicity, as shown in Supplementary Table 4.

We analyzed 17,637 participants across six TOPMed cohorts with data available for genotyping, CHIP derivation, and prospective follow-up for incident T2D. Our analyses were conducted in large longitudinal cohorts with a wide range of ages, self-reported race/ethnicities, and other demographics contributing to T2D risk status. CHIP prevalence was higher with older age at blood draw, consistent with findings of previous epidemiologic analyses (2). Participants with CHIP had a modest but significant 23% higher risk of developing T2D over nearly a decade of follow-up. Among those with a priori defined CHD-related CHIP mutations, results were similar, owing to these representing 88% of overall CHIP.

Although it is established that risk of developing T2D increases with age, reasons for deterioration in insulin sensitivity and β-cell function and mass with aging are largely unknown. In prior studies, the prevalence of clonal expansion of somatic mutations in hematopoietic stem cells was found to increase sharply at older ages, implicating CHIP in aging-related chronic diseases (2,15). Indeed, the accumulation of CHIP variants is positively associated with aging-related cancers, cardiovascular diseases, and all-cause mortality (2). Bonnefond et al. (16) also reported a higher prevalence of clonal mosaicism among patients with prevalent T2D versus without T2D, although the cross-sectional design of the study precluded the ability to delineate the temporal direction of this association. In a recent analysis in a retrospective cohort of older adults in Korea, investigators reported a positive association between CHIP and T2D incidence among 92 Korean older adults with VAF>10% in comparison with no CHIP, but analyses were unadjusted for T2D risk factors and other potential confounders (17). Overall, we observed that the presence of CHIP, with adjustment for age and other T2D risk factors, was related to higher risk of developing T2D over follow-up, particularly for CHIP on TET2 and ASXL1, previously related to atherogenic disease.

Our analysis identifies a potential shared pathophysiology of CHD and T2D that had not previously been characterized from longitudinal data. The link between T2D and CVD is well-known, as they share several upstream risk factors including age, obesity, smoking, diet, and other lifestyle factors; therefore, it is plausible that an accumulating burden of clonal expansion for certain variants precipitates both outcomes. Carriers of somatic clonal expansion mutations in DNMT3A, TET2, ASXL1, JAK2, and TP53 genes have up to twofold higher risks of incident CHD and higher coronary artery calcification scores than noncarriers (2). Further, an animal model of TET2 hematopoietic clonal expansion had significantly larger atherosclerotic lesions induced in comparison with controls. Additionally, experimental evidence in mice indicated that TET2 loss-of-function mutations in bone marrow cells exacerbated obesity-related insulin resistance (6), and CHIP-enhanced IL-1β expression in white adipose tissue may have mediated these effects. Other mechanistic research also suggests that clonal expansion may promote atherosclerosis through a number of local and systemic inflammatory pathways (2,6,16); thus, it is plausible that inflammation serves as one potential mechanism for CHIP as a driver of aging-related chronic diseases including CHD and T2D.

Strengths of this study are the inclusion of large well-phenotyped cohorts representing diversity in age, sex, and self-reported race/ethnicity. Long-term follow-up for incident T2D outcomes allows us to establish temporality, with CHIP preceding T2D. Repeated assessments in CHIP carriers have shown that CHIP progresses over time; thus, using only a single measure we may misclassify those with VAF <10% at baseline. We also speculate that the CHD-related genes act upstream of T2D development, although the reverse mechanism is also plausible, whereby deterioration of glycemic control increases the likelihood of CHIP occurring; however, we carefully excluded participants with T2D, CVD, or history of cancer at baseline, and cases were identified over long-term follow-up of median ∼10 years from blood draw. Statistical power to detect associations with T2D risk is relatively limited for the less common driver mutations. Further, as we did not adjust for multiple comparisons in the individual gene analyses, there may be associations with T2D due to chance.

CHIP mutations located on genes previously implicated in CHD risk, but not overall CHIP, were associated with higher T2D risk. CHIP may reflect a shared pathophysiology of CHD and T2D that had not previously been characterized from longitudinal data. CHIP overall including on driver mutations previously associated with CHD was associated with development of T2D, implicating CHIP as a mediator of T2D risk through atherosclerosis-related pathways. Mechanistic research is warranted to identify the precise causal pathways underlying these observations. Further, whether CHIP or its downstream effects on atherosclerosis are modifiable is unknown. Addressing these gaps will inform potential therapies and determine whether CHIP represents a clinically targetable pathway of cardiometabolic risk.

Acknowledgments. The authors thank the studies and participants who provided biological samples and data for TOPMed. A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org. FHS acknowledges the dedication of the FHS study participants without whom this research would not be possible. The authors thank the staff and participants of JHS. The authors thank the other investigators, the staff, and the participants of MESA for their valuable contributions. A full list of participating MESA investigators and institutes can be found at https://www.mesa-nhlbi.org.

L.A.C. is deceased.

Funding. J.M., J.D., D.D., and P.W. are supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant U01 DK078616, and J.M. and D.D. are supported by the NIDDK grant U01 DK105554. J.C. is supported by the National Heart, Lung, and Blood Institute (NHLBI) grant T32 HL129982. R.V. is supported in part by the Evans Medical Foundation and the Jay and Louis Coffman Endowment from the Department of Medicine, Boston University School of Medicine.

Collection of molecular data for the TOPMed program was supported by the NHLBI: genome sequencing for “NHLBI TOPMed: Whole Genome Sequencing and Related Phenotypes in the Coronary Artery Risk Development in Young Adults (CARDIA) Study” (phs001612) was performed at the Baylor College of Medicine Human Genome Sequencing Center (HHSN268201600033I); genome sequencing for “NHLBI TOPMed: Whole Genome Sequencing and Related Phenotypes in the Cardiovascular Health Study (CHS)” (phs001368) was performed at Broad Institute of MIT and Harvard Genomics Platform (HHSN268201600034I) and supported by the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium (HL105756; genome sequencing for “NHLBI TOPMed: Whole Genome Sequencing and Related Phenotypes in the Framingham Heart Study (FHS)” (phs000974) was performed at Broad Institute of MIT and Harvard Genomics Platform (3U54HG003067-12S2); genome sequencing for “NHLBI TOPMed: Whole Genome Sequencing and Related Phenotypes in the Jackson Heart Study (JHS)” (phs000964) was performed at New York Genome Center (HHSN268201100037C); genome sequencing for “NHLBI TOPMed: Whole Genome Sequencing and Related Phenotypes in the Multi-Ethnic Study of Atherosclerosis Study (MESA)” (phs001416) was performed at Broad Institute of MIT and Harvard Genomics Platform (3U54HG003067-13S1); genome sequencing for “NHLBI TOPMed: Whole Genome Sequencing and Related Phenotypes in the Women’s Health Initiative Study (WHI)” (phs001237) was performed at Broad Institute of MIT and Harvard Genomics Platform (HHSN268201500014C); core support including centralized genomic read mapping and genotype calling, along with variant quality metrics and filtering, was provided by the TOPMed Informatics Research Center (3R01HL-117626-02S1, contract HHSN268201800002I); core support including phenotype harmonization, data management, sample-identity quality control, and general program coordination were provided by the TOPMed Data Coordinating Center (R01HL-120393, U01HL-120393, contract HHSN268201800001I).

CARDIA was conducted and supported by the NHLBI in collaboration with the University of Alabama at Birmingham (HHSN268201800005I and HHSN268201800007I), Northwestern University (HHSN268201800003I), University of Minnesota (HHSN268201800006I), and Kaiser Foundation Research Institute (HHSN268201800004I). CARDIA was also partially supported by the Intramural Research Program of the National Institute on Aging (NIA) and an intra-agency agreement between NIA and NHLBI (AG0005). CHS was supported by NHLBI contracts (HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, and 75N92021D00006), NHLBI grants (U01HL080295, U01HL130114), and additional support from the National Institute of Neurological Disorders and Stroke (NINDS) and the NIA (R01AG023629). FHS was supported by NHLBI contracts (NO1-HC-25195, HHSN268201500001I, and 75N92019D00031) and grant supplement (R01 HL092577-06S1) for this research. JHS was supported by contracts with the NHLBI and the National Institute on Minority Health and Health Disparities in collaboration with Jackson State University (HHSN268201800013I), Tougaloo College (HHSN268201800014I), the Mississippi State Department of Health (HHSN268201800015I), and the University of Mississippi Medical Center (HHSN268201800010I, HHSN268201800011I, and HHSN268201800012I). The authors thank the staff and participants of JHS. MESA was supported by contracts and grants with the NHLBI and NIDDK (75N92020D00001, HHSN268201500003I, N01-HC-95159, 75N92020D00005, N01-HC-95160, 75N92020D00002, N01-HC-95161, 75N92020D00003, N01-HC-95162, 75N92020D00006, N01-HC-95163, 75N92020D00004, N01-HC-95164, 75N92020D00007, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168, N01-HC-95169, UL1-TR-000040, UL1-TR-001079, UL1-TR-001420, UL1TR001881, DK063491, and R01HL105756). WHI was funded by contracts with the NHLBI (75N92021D00001, 75N92021D00002, 75N92021D00003, 75N92021D00004, and 75N92021D00005).

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. L.M.R. is a consultant for the TOPMed Administrative Coordinating Center (through Westat). No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.K.T. analyzed data and wrote the manuscript. A.K.M. analyzed data and reviewed and edited the manuscript. J.W. analyzed data and reviewed and edited the manuscript. S.R. reviewed and edited the manuscript. K.E.W. reviewed and edited the manuscript. A.G.B. reviewed and edited the manuscript. D.D. reviewed and edited the manuscript. E.A.W. reviewed and edited the manuscript. J.C. reviewed and edited the manuscript. A.C. reviewed and edited the manuscript. L.A.C. reviewed and edited the manuscript. J.D. reviewed and edited the manuscript. M.O.G. reviewed and edited the manuscript. X.G. reviewed and edited the manuscript. B.H. reviewed and edited the manuscript. L.A.L. reviewed and edited the manuscript. S.L. reviewed and edited the manuscript. L.M.R. reviewed and edited the manuscript. A.P.R. reviewed and edited the manuscript. S.S.R. reviewed and edited the manuscript. K.D.T. reviewed and edited the manuscript. L.T. reviewed and edited the manuscript. J.G.W. reviewed and edited the manuscript. P.W. reviewed and edited the manuscript. A.P.C. reviewed and edited the manuscript. R.S.V. reviewed and edited the manuscript. M.F. reviewed and edited the manuscript. B.M.P. reviewed and edited the manuscript. C.K. contributed to discussion and reviewed and edited the manuscript. J.I.R. contributed to discussion and reviewed and edited the manuscript. J.M. contributed to discussion and reviewed and edited the manuscript. J.E.M. contributed to discussion and reviewed and edited the manuscript. D.K.T. 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 presented in abstract form at the 81st Scientific Sessions of the American Diabetes Association, New Orleans, LA, 25–29 June 2021, and EPI|Lifestyle 2023, Boston, MA, 1–3 March 2023.