Introduction: Clonal hematopoiesis of indeterminate potential (CHIP) is associated with an increased risk of several age-related proinflammatory diseases, possibly driven by an altered function of tissue-resident macrophages originating from the mutant clone. Surprisingly, CHIP is associated with a reduced risk of Alzheimer's disease (AD), but the mechanism is unknown. Large genome-wide association studies (GWAS) of AD have identified several single nucleotide polymorphisms (SNP) in coding or cis-regulatory regions of genes that are predominantly or only expressed in microglia, the macrophage-like cells in brain. Recently, somatic driver mutations in CHIP carriers have been detected in human microglia, suggesting that these cells could directly influence AD pathology. We hypothesized that interactions between microglial AD risk loci and CHIP could reveal specific genes that are involved in the protective effect of CHIP against AD.

Methods: The UK Biobank (UKB) cohort was used as the discovery dataset and the Alzheimer's Disease Sequencing Project (ADSP) cohort for validation. CHIP, defined as myeloid malignancy associated somatic single nucleotide variants [SNV] and insertions/deletions [indels] or mosaic chromosomal alterations [mCA], was identified from whole-exome or genome sequencing data using Mutect2 and the MoChA algorithm. Association tests were performed using Cox regression (in UKB) or logistic regression (in ADSP) with age, sex, the first 10 ancestry PCs, and genetic AD risk factors as covariates in both cohorts, with additional cardiovascular and socioeconomic risk factors in UKB.

Results: We identified 3282 incident AD cases and 437205 dementia-free controls in UKB. CHIP associated with reduced risk of incident AD (HR 0.88, 95%-CI 0.79-0.99, P = 0.03), albeit with a lower effect size than previously reported (Bouzid, Nat Med 2023). We explored statistical interactions between CHIP and five established AD GWAS risk variants located near microglial genes (CD33, TREM2, MS4A6A, SPI1, and BIN1). We identified an interaction of SNV/indel-driven CHIP with the CD33 SNP rs3865444 (interaction term: HR = 1.31 per A-allele, 95%-CI 1.06-1.61, P = 0.01). As in prior studies, rs3865444-A was protective against AD in the population overall, but, surprisingly, increased risk of AD in CHIP carriers. Once this unexpected interaction was taken into account, SNV/indel CHIP associated with protection from AD, similar to Bouzid et al. (HR 0.78, 95%-CI 0.64–0.96, P = 0.02). This effect seemed to be driven by DNMT3A CHIP (interaction HR 1.37 per A-allele, 1.06–1.78, P = 0.02), with no other CHIP classes showing a significant interaction.

To validate this finding, we analyzed data from ADSP, comprising 2719 AD cases and 6420 controls. SNV/indel CHIP and DNMT3A CHIP showed a significant interaction with rs3865444 (OR 1.31, 95%-CI 1.05-1.62, P = 0.015 and OR 1.44, 95%-CI 1.04-1.98, P = 0.026, per A-allele, respectively, in a fixed effects meta-analysis across 7 cohorts).

The A-allele of rs3865444 results in a splice isoform with reduced expression of CD33 protein. In line with this, we found rs3865444-A was strongly associated with reduced plasma CD33 levels (-1.04 SD per A-allele, 95%-CI –(1.03 – 1.06), P < 10-100) in UKB Olink proteomic data, whereas CD33 plasma levels did not associate with SNV/indel CHIP or DNMT3A CHIP. We also introduced DNMT3A mutations into CD34 cells from rs3865444 C/C, C/A, or A/A donors, but did not see any differences in macrophage forming capacity based on DNMT3A mutation status or rs3865444 genotype. CD33 expression was altered by rs3865444 but not DNMT3A mutations. Interestingly, rs3865444 showed a significant interaction between DNMT3A CHIP and peripheral blood monocyte count in UKB (interaction: 0.05 SD per A-allele, 95%-CI 0.02-0.08, P < 0.001), which was not seen for other drivers or blood cell types. We hypothesize that increased blood monocyte count in DNMT3A carriers with rs3865444-A may reflect a reduced ability of mutant monocytes to migrate to brain or other tissues in the absence of CD33.

Conclusions: We find evidence in multiple cohorts of an interaction between DNMT3A CHIP and a well-studied SNP in CD33 for risk of AD. Differences in allele frequency of this SNP across cohorts may explain some heterogeneity in the strength of the association of CHIP with AD. Furthermore, these findings implicate CD33 as functionally relevant to the association of CHIP with AD.

This content is only available as a PDF.
2025
Sign in via your Institution