Unidirectional association of clonal hematopoiesis with atherosclerosis development

Abstract

Clonal hematopoiesis, a condition in which acquired somatic mutations in hematopoietic stem cells lead to the outgrowth of a mutant hematopoietic clone, is associated with a higher risk of hematological cancer and a growing list of nonhematological disorders, most notably atherosclerosis and associated cardiovascular disease. However, whether accelerated atherosclerosis is a cause or a consequence of clonal hematopoiesis remains a matter of debate. Some studies support a direct contribution of certain clonal hematopoiesis-related mutations to atherosclerosis via exacerbation of inflammatory responses, whereas others suggest that clonal hematopoiesis is a symptom rather than a cause of atherosclerosis, as atherosclerosis or related traits may accelerate the expansion of mutant hematopoietic clones. Here we combine high-sensitivity DNA sequencing in blood and noninvasive vascular imaging to investigate the interplay between clonal hematopoiesis and atherosclerosis in a longitudinal cohort of healthy middle-aged individuals. We found that the presence of a clonal hematopoiesis-related mutation confers an increased risk of developing de novo femoral atherosclerosis over a 6-year period, whereas neither the presence nor the extent of atherosclerosis affects mutant cell expansion during this timeframe. These findings indicate that clonal hematopoiesis unidirectionally promotes atherosclerosis, which should help translate the growing understanding of this condition into strategies for the prevention of atherosclerotic cardiovascular disease in individuals exhibiting clonal hematopoiesis.

Fig. 1. Prevalence and characteristics of CH in middle-aged individuals.

We performed deep targeted sequencing to identify somatic mutations in a custom panel of 54 CH-related genes in 3,692 individuals from the PESA cohort. a, The number of CH driver mutations identified per gene. The values above the bars indicate the percentage of mutations affecting each specific gene. b, The CH prevalence across quartiles of age. c, The number of mutations per individual across quartiles of age. d, The association between advancing age (stratified as quartiles) and CH (analyzed separately as driven by mutations in DNMT3A, TET2 or other genes) based on multivariate logistic regression analyses adjusted for sex. The bars indicate 95% confidence intervals centered in the mean value (square). e, The distribution of mutant clone size in the study population, assessed as VAF. The dashed line shows the 2% VAF threshold most typically used to identify CH. The box shows the 25th (Q1), 50th (median) and 75th (Q3) percentiles of the data. The whiskers represent Q1 − 1.5 × IQR at the minimum and Q3 + 1.5 × IQR at the maximum. f, The prevalence of CH with VAF ≥2% across quartiles of age. g, The association between gene-specific CH and female sex, based on multivariate logistic regression analyses adjusted for age. The bars indicate 95% confidence intervals centered in the mean value (square). h, The CH prevalence across quartiles of age stratified by sex. In b, f and h, CH status in individuals carrying more than one mutation was defined on the basis of the mutation with the highest VAF.

Fig. 2. Longitudinal assessment of the effects of CH on de novo femoral atherosclerosis development.

We investigated the association between CH at baseline and de novo development of femoral atherosclerosis ∼3 years and ∼6 years after enrollment among PESA participants who initially lacked detectable atherosclerosis in this region. 3DVUS imaging was used to determine the presence of femoral atherosclerosis. a, Summary of study design. b, Representative images from femoral atherosclerosis burden, assessed by 3DVUS, in an individual exhibiting de novo femoral atherosclerosis development (that is, absence of detectable atherosclerosis at the baseline evaluation (left) and plaque development at follow-up (right); scale bar, 5 mm). c,d, Incidence of de novo femoral atherosclerosis development at the 3-year follow-up (c, n = 2,347) or the 6-year follow-up (d, n = 2,214) in individuals free of CH (no CH) and in individuals exhibiting CH with VAF 0.2–2% or ≥2%. Incidence of de novo femoral atherosclerosis is also shown for myeloid CH (CHIP) or CH driven by specific CHIP genes. Statistical significance in the analyses of CH and CHIP was evaluated using univariate logistic regression models (P for trends are shown). In gene-specific analyses, statistical significance was examined through proportion tests relative to the non-CHIP carriers group (*P < 0.05, **P ≤ 0.01, ***P ≤ 0.001). e,f, The association between CH or CHIP and de novo femoral atherosclerosis development at the 3-year follow-up (e, n = 2,347) and 6-year follow-up (f, n = 2,214) based on multivariate logistic regression analyses. Statistical models were adjusted for age, sex, smoking, lipid-lowering treatment and the AUC of SBP, fasting glucose, LDL-C, BMI, CACS and global atherosclerotic plaque volume assessed by 3DVUS imaging; the bars indicate 95% confidence intervals centered in the mean value (square).

Fig. 3. Longitudinal assessment of the dynamics of CH in middle-aged individuals.

We monitored the expansion of 602 CH-related mutations through serial sequencing of blood samples from 494 individuals at baseline and ∼6 years after enrollment. Then, we investigated the association between subclinical atherosclerosis burden or related traits at baseline and the AER of the mutant hematopoietic clones. a, Summary of study design. b, The AER of mutations in any CH gene (n = 602) or specifically in DNMT3A (n = 381), TET2 (n = 80) or other genes (n = 141), compared with nondeleterious variants in CH-related genes (n = 271). Statistical significance was determined by two-sided Mann–Whitney–Wilcoxon tests (**P ≤ 0.01, ****P ≤ 0.0001). c, The AER of CH-related mutations stratified by age quartiles (n = 602). d, AER of CH-related mutations stratified by sex (n = 602). e, The correlation between AER of CH-related mutations and baseline variant allele frequency (VAF). The P value for two-sided Pearson correlation is indicated. f–i, The AER of CH-related mutations in individuals with no detectable atherosclerosis and across tertiles of plaque burden, based on CACS (f, n = 600), global plaque volume (g, n = 578), carotid plaque volume (h, n = 597) or femoral plaque volume (i, n = 581) measured by 3DVUS. j,k, The AER of CH-related mutations stratified by conventional modifiable risk factors (j, n = 602 for all variables, except for obesity, n = 601) and quartiles of hsCRP (k, n = 600). Statistical significance in c, d and f–k was determined by mixed-effects models adjusted for baseline VAF (d) or sex and baseline VAF (c and f–k). The boxes in b–d and f–k represent the 25th (Q1), 50th (median) and 75th (Q3) percentiles of the data. The whiskers represent Q1 − 1.5 × IQR at the minimum and Q3 + 1.5 × IQR at the maximum.

Extended Data Fig. 1. Basic features of clonal hematopoiesis (CH)-related mutations and subclinical atherosclerosis burden in the study population.

We performed deep sequencing to identify somatic mutations in a custom panel of 54 CH-related genes in 3,692 middle-aged individuals from the PESA cohort. a, CH-related genes included in the custom panel. Genes for which a CH mutation was found in at least one individual are underlined and in bold font. Genes typically considered as drivers of myeloid CH (frequently known as CHIP, clonal hematopoiesis of indeterminate potential) or lymphoid-CH are shown separately. b, Assay sensitivity for the detection of CH-related variants at different variant allele fractions (VAF). Dashed lines show the median sequencing depth of our study (3,712x; vertical line), and the sensitivity to detect CH variants at the minimum VAF threshold (0.2%) used to define CH (horizontal lines). c, Graphical representation of the CH-related genes found mutated in the study population; font size is proportional to the frequency of mutations. d, Proportions of different type of mutations in CH-related genes according to their functional consequence. e, Proportions of different types of single nucleotide substitutions. f, Distribution of mutant clone size, assessed as VAF, in DNMT3A (n = 657), TET2 (n = 153) and other sequenced genes (n = 362). The dashed line shows the 2% VAF threshold typically used to identify CHIP. Boxes represent the 25th (Q1), 50th (median) and 75th (Q3) percentiles of the data. Whiskers represent Q1 - 1.5*IQR at the minimum and Q3 + 1.5*IQR at the maximum. g, Proportion of CH-related variants identified after downsampling of sequencing depth to 30x, 100x, 400x and 1,000x. “base” indicates the median sequencing depth for these variants using our original sequencing approach (3,625x); n = 168 CH variants, 28 for each VAF range. h, Number of CH-related variants identified after downsampling to specific sequencing depths.

Extended Data Fig. 2. Subclinical atherosclerosis burden at baseline stratified by CH status.

We performed deep targeted sequencing to identify somatic mutations in a custom panel of 54 CH-related genes in 3,692 middle-aged individuals from the PESA cohort. CACS and 3DVUS imaging were used to determine subclinical atherosclerosis burden, assessed cross-sectionally at baseline. Graphs show the proportion of individuals with no detectable atherosclerosis and across tertiles of plaque burden, based on CAC score (a, n = 3,687), global plaque volume by 3DVUS (b, n = 3,543), carotid plaque volume by 3DVUS (c, n = 3,659) or femoral plaque volume by 3DVUS (d, n = 3,569). Statistical significance for differences in subclinical atherosclerosis burden, defined as 0 or tertiles, was evaluated using an ordinal logistic regression model.

Extended Data Fig. 3. Effects of clonal hematopoiesis (CH) on de novo atherosclerosis development in femoral arteries.

We investigated the association between CH at baseline and de novo development of femoral atherosclerosis ∼3-years after enrollment among individuals who initially lacked detectable atherosclerosis in that region, assessed by 3-dimensional-vascular ultrasound (3DVUS) imaging. a, Association between CH with VAF ≥ 2% and de novo atherosclerosis development in femoral arteries, based on multivariate logistic regression analyses adjusted for age, sex, conventional modifiable risk factors as categorical variables and the cumulative systemic burden of atherosclerosis across the timeframe of the study (AUC for CACS, assessed by CT imaging, and global atherosclerotic plaque volumes, assessed by 3DVUS imaging) (n = 1,932). b, Association between CH and de novo atherosclerosis development in femoral arteries, based on multivariate logistic regression analyses adjusted for age, sex, absolute counts of leukocytes, erythrocytes and platelets, and the cumulative systemic burden of atherosclerosis across the timeframe of the study (n = 2,353). c, Association between myeloid CH or CHIP with VAF ≥ 2% and de novo atherosclerosis development in femoral arteries, based on multivariate logistic regression analyses adjusted for age, sex, conventional modifiable risk factors and the cumulative systemic burden of atherosclerosis across the timeframe of the study (n = 1,965). d, Association between CHIP and de novo atherosclerosis development in femoral arteries, based on multivariate logistic regression analyses adjusted for age, sex, absolute counts of leukocytes, erythrocytes and platelets, and the cumulative systemic burden of atherosclerosis across the timeframe of the study (n = 2,353). Bars in each panel indicate 95% confidence intervals centered in the mean value (square).

Extended Data Fig. 4. Representative 3DVUS images from subclinical atherosclerosis burden in femoral arteries across tertiles of atherosclerotic plaque volume.

Multiplanar views (upper rows, scale bar: 2 mm) and 3D reconstruction (lower rows) from the right and left femoral arteries. Images are representative of the upper limits of each tertile of atherosclerotic plaque volume. 3DVUS, 3-dimensional vascular ultrasound.