CHIP and Aging: How Blood Stem-Cell Clones Gain Ground

Clonal hematopoiesis of indeterminate potential, or CHIP, appears to be a slow contest inside the bone marrow. Some blood stem cells acquire mutations that let their descendants gain ground over time. The result can be a detectable clone in the bloodstream long before leukemia—or without leukemia ever developing.

The most useful way to view CHIP is as age-related selection among blood-forming cells. Human studies show that clones commonly emerge and often grow over decades. Single-cell work suggests that common mutant cells tolerate inflammatory stress better than neighboring cells. Mouse experiments show how those altered immune cells can damage tissue through inflammatory signaling. That is an interesting biological story. It is not yet a reason to take metformin, rapamycin, or any other drug to suppress CHIP.

CHIP becomes more common as people age

A 2024 longitudinal ARIC study provides unusually direct evidence on emergence. Researchers sequenced blood samples from 4,187 participants at two time points separated by a median of 21 years. CHIP prevalence rose from 10.9% at the first assessment to 25.0% at follow-up.

Among 3,730 people without detectable CHIP at baseline, 735 developed it during follow-up. Of 352 clones detectable at both time points, 66% grew, 9% shrank, and 24% remained stable. Aging shifts the odds, but it does not turn every clone into a steadily rising line on a chart.

Some clones also appeared at substantial size without a previously documented smaller precursor: 37% of incident CHIP cases had a variant allele fraction of at least 10%. Variant allele fraction is the share of sequenced blood DNA carrying a mutation; it is an imperfect proxy for clone size, but larger values generally indicate that a clone occupies more blood cells.

Age was the clearest predictor in ARIC. Participants with CHIP, and those who later developed it, were older than their comparison groups. In the reported baseline comparisons, body mass index, blood pressure, diabetes, coronary disease, stroke, heart failure, LDL cholesterol, and HDL cholesterol showed weak or null predictive signals. Smoking showed a directional association at baseline. This does not make lifestyle irrelevant to health; it means conventional cardiovascular risk factors did not explain much of incident CHIP in this analysis.

A separate 2024 PNAS analysis adds an important refinement: clone size reflects both age and the number of mutations. CHIP therefore does not appear to follow a single biological clock. Mutations, the surrounding marrow environment, and time all matter.

Why some mutant stem cells gain an edge

The best mechanistic evidence here comes from a 2024 Cell study that combined human bone-marrow profiling with mouse experiments. Among 195 older adults without hematologic malignancy, clonal hematopoiesis was present in 29.2% using a variant allele fraction threshold of 0.02. DNMT3A and TET2 mutations accounted for 69% of cases.

The researchers then used TARGET-seq+ to profile 13,247 individual cells from 13 donors; 95.0% of sorted cells passed quality control. Both mutant and wild-type stem and progenitor cells in CHIP marrow showed inflammatory and quiescence-related gene programs compared with cells from people without clonal hematopoiesis. Yet mutant cells showed relatively less TNF-alpha/NF-kappaB, IL-1, interferon-alpha, and aging-related signaling than their wild-type neighbors.

That pattern supports a plausible selection model: inflammation may burden the entire marrow while leaving certain mutant stem cells comparatively better able to persist. It does not prove that inflammation causes every human clone to expand. The genotype-level sample was small—five DNMT3A cases, three TET2 cases, and one double-mutant case—so the study cannot precisely describe every CHIP subtype.

The same work also found evidence that a clone can reshape its neighborhood. In one TET2 case, immunophenotypic stem and progenitor cells expanded 2.6-fold even though 97.6% of those cells were TET2 wild type. A mutant clone may therefore affect cells that do not carry its mutation, rather than behaving as an isolated cellular island.

Inflammation may connect CHIP to tissue damage

In the Cell study's human observational arm, investigators examined 4,946 ARIC adults ages 52 to 74. CHIP prevalence was 3.9%, and 61.8% of carriers had DNMT3A mutations. After adjustment for age and sex, participants with DNMT3A CHIP had an 11.0-percentage-point higher prevalence of stage IV periodontitis than participants without CHIP.

That result is an association, not proof that CHIP caused severe periodontal disease. Confounding can remain after adjustment, and the supplied result does not establish whether reducing a clone would improve gum health.

The paired mouse experiments make the causal mechanism more plausible. Mice with 10% Dnmt3a R878H marrow showed competitive expansion of mutant cells, higher marrow and systemic inflammatory cytokines, increased gingival IL-23, IL-17, and RANKL, expansion of inflammatory immune cells, and periodontal bone loss. Femoral trabecular bone volume fraction fell only 8.9%, suggesting a tissue-specific inflammatory effect rather than generalized skeletal collapse.

Cardiovascular findings fit this inflammation-centered model, but the evidence varies by disease and population. A 2023 UK Biobank study of people with established atherosclerotic cardiovascular disease found that CHIP predicted adverse outcomes in this already high-risk population. A 2024 Circulation study linked TET2-loss CHIP with higher atrial-fibrillation risk and used mouse experiments to implicate NLRP3 inflammasome activation.

Researchers have also reported that CHIP was associated with a higher risk of new-onset myocarditis and pericarditis in UK Biobank participants without prevalent cardiovascular disease or hematologic malignancy, followed for a median of 13.6 years in a 2025 JAMA Cardiology cohort study. These observational findings warrant attention, but they do not establish that CHIP causes every inflammatory cardiac condition or that treating a clone prevents one.

The exceptions matter

CHIP does not produce a uniform aging phenotype. In the Women's Health Initiative Memory Study, a 2025 study found no association between CHIP at the conventional 2% variant allele fraction threshold and incident cognitive impairment. Larger clones, defined as a variant allele fraction of at least 8%, were associated with lower incidence of adjudicated probable dementia (hazard ratio 0.62; 95% confidence interval 0.41 to 0.94; p=0.025). The finding was statistically significant for that endpoint, but it does not mean CHIP protects the brain in every population or mutation class.

Repeated blood loss also does not simply translate into more CHIP. A 2025 Blood study found no significant difference in overall clonal hematopoiesis prevalence between 217 older men with more than 100 lifetime blood donations and 212 sporadic donors with fewer than 10 donations. Ordinary hematopoietic stress is not a universal explanation for detectable CHIP burden.

Metformin is a mouse result, not a treatment plan

The Cell study tested metformin because DNMT3A-mutant cells showed altered dependence on mitochondrial oxidative phosphorylation. In competition assays, 50 micromolar metformin suppressed the Dnmt3a R878H mutant advantage; a metformin-resistant NDI1 construct reversed that effect. In competitive transplant mice, metformin delivered at 5 mg/mL in drinking water prevented the steady rise in mutant chimerism over seven months. At the endpoint, mutant cells accounted for 84.6% of cells in untreated mice versus 56.5% in metformin-treated mice.

No randomized human trial has shown that metformin, rapamycin, or another intervention prevents CHIP, slows human clone growth, or improves clinical outcomes by suppressing a clone.

Mouse competition assays measure a biological mechanism and a surrogate outcome. They do not establish a human dose, a favorable risk-benefit balance, or a clinical benefit. For now, CHIP is best treated as a risk signal and a fast-moving research field—not a self-experimentation target.