Clonal hematopoiesis (CH) arises when a blood stem cell acquires a mutation that gives it a competitive edge, and its descendants gradually crowd out their neighbors. In older adults, CH is common: sequencing data put it at 6% among people aged 60-69, 10% at 70-79, and 18% after 90. The link between CH and neurodegeneration remains contested: some studies report a protective effect, others a higher risk of stroke and Parkinson's disease, and still others find no association at all. Most work relies on a single blood draw and clinical diagnoses, which makes it impossible to see any dynamics. Our study sits off to the side: the closest neighbors in the archive study Alzheimer's disease through MRI, cortical folds, or proteins, but no one had tied the longitudinal dynamics of clonal burden to changes in brain volume. LBC1936 is unique in that blood DNA methylation and structural MRI were both measured three times, letting researchers track how clonality changes and how the brain changes alongside it.
The authors used data from the Lothian Birth Cohort 1936: 1091 participants, three waves (mean ages 73, 76, and 79). Clonal burden was estimated from blood DNA methylation in two ways. The first, COMET, is an algorithm based on CpG sites whose methylation shifts predictably during clonal expansion; it returns a predicted variant allele frequency (pVAF) without knowing the specific mutation. The second, CIMS, is a weighted sum of methylation levels at 500 CpG sites associated with CHIP in an independent genome-wide study (EWAS). Brain volumes (total, gray and white matter, white matter hyperintensities, 68 cortical regions per the Desikan-Killiany atlas, and subcortical structures) came from structural MRI processed with FreeSurfer. The link between clonality trajectories and volumes was assessed using latent growth curves (LGC) within structural equation modeling (SEM): baseline levels (intercept) and rates of change (slope) were compared separately. To test regional specificity, the map of pVAF-associated longitudinal atrophy was correlated with a meta-analytic map of cortical vulnerability in Alzheimer's disease (n = 4774).
Higher clonal burden (pVAF) at baseline was associated with lower total brain volume (standardized estimate -0.270, adjusted p = 0.023) after adjusting for age, sex, and intracranial volume. For gray matter, the rate of change mattered: pVAF tracked accelerated gray matter loss over 6 years (standardized estimate -0.176, adjusted p = 0.030), though the baseline-level association was only a trend. For white matter and hyperintensities, there were no significant associations. Regional analysis revealed a left-sided pattern of longitudinal atrophy: the strongest effects were in the fusiform, lingual, and parahippocampal gyri, as well as the precuneus. This pattern correlated significantly with the Alzheimer's disease vulnerability map (r = 0.357, p = 0.003), comparable to the correlation for a family history of AD (r = 0.33). By contrast, subcortical structures, including the hippocampus, showed no significant associations, and the caudate nucleus even appeared relatively preserved in people with high clonal burden. The second measure, CIMS, was also linked to lower total brain volume at baseline (standardized estimate -0.385, adjusted p = 0.003) and to longitudinal atrophy (standardized estimate -0.187, adjusted p = 0.005), but its regional pattern differed and did not correlate with the AD map. After adjusting for blood cell composition and technical effects, the global associations held, but the regional ones weakened, and the spatial correlation with the AD map was no longer significant (r = 0.230, p = 0.059).
This work stands apart from its neighbors in the archive because it is the first to link the longitudinal dynamics of clonal burden to changes in brain volume in a cohort with repeated measurements. Most previous studies used a single blood draw and static measures, which made it impossible to separate baseline differences from progressive changes. Here, clonal expansion is tied not only to lower brain volume at baseline but also to accelerated gray matter loss, and the regional pattern partly mirrors the Alzheimer's disease map. This strengthens the hypothesis that clonal hematopoiesis contributes systemically to neurodegeneration, though the absence of an effect on the hippocampus and the weakening after adjustments suggest this is more a parallel pathway of vulnerability than a direct marker of amyloid pathology.
The field could move toward integrating mutational profiles with neuroimaging. Because COMET does not distinguish specific driver genes, adding sequencing would let researchers test which mutations (DNMT3A, TET2, ASXL1) are most strongly linked to atrophy. Multivariate approaches, such as latent models of atrophy coordination, could reveal whether clonal hematopoiesis targets particular brain networks. Comparing results with functional MRI and connectomics, as well as with PET and cerebrospinal fluid biomarkers, is another promising direction for testing links to amyloid and tau.
The findings could influence gerontology, neurology, and hematology, since they point to blood as a potentially modifiable systemic factor in brain aging. For the epidemiology of aging, this is one more argument for monitoring clonal hematopoiesis as a biomarker.
Replication in independent cohorts with repeated MRI and methylation, plus validation of COMET and CIMS against sequencing. It would also help to add mutational profiling and amyloid/tau biomarkers to separate the contribution of clonality from that of classic Alzheimer's pathology.
In the field of clonal hematopoiesis and neurodegeneration, several questions remain open: which specific mutations or clone properties (size, expansion rate, inflammatory profile) determine harm to the brain; whether the link is causal or reflects a shared aging process; and why effects differ between the cortex and subcortical structures. This work does not answer these questions, but it provides a tool (methylation as a proxy for clonality) and a cohort with a longitudinal design to study them.
The 2% variant allele frequency threshold for a CHIP diagnosis was not chosen at random: below that fraction, it is technically hard to tell true expansion from sequencing noise. But error-corrected studies show that among people over 80, clones with a frequency of 1% or more turn up in 62%, meaning most older adults carry traces of clonal expansion in their blood, even if they do not formally meet the criteria for CHIP.
The idea that mutations in blood cells can affect the brain echoes Greg Egan's novel Quarantine: there, neuron modifications and systemic changes to the body lead to unpredictable cognitive consequences. In a gentler form, the theme of blood as a carrier of aging is played out in the film In Time (2011), where life resources literally flow between people.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.