A preprint analysis has identified three HLA-II haplotypes, or inherited genetic patterns, that showed contrasting associations with Alzheimer’s disease, healthy longevity and the amount of HLA-II detected in microglial tissue. Hap-B was more frequent among cognitively healthy centenarians and controls than among Alzheimer’s patients. Hap-R was less frequent in controls than in Alzheimer’s patients and less frequent in centenarians than controls at several genetic resolution levels. Hap-Y was enriched in controls relative to Alzheimer’s patients only for two specific HLA allele features, and it was associated with higher microglial HLA-II load.
The researchers asked whether Alzheimer’s disease and longevity-associated HLA-II signals resolve into distinct haplotypes and whether those haplotypes relate to microglial immune states and quantitative or qualitative neuropathology. After quality control, the primary genetic analysis included 6,053 people: 443 cognitively healthy centenarians, 3,219 middle-aged Dutch population controls and 2,391 Alzheimer’s disease patients.
How the signals were separated
To separate signals that might be inherited together, the researchers calculated pairwise linkage disequilibrium, or LD, grouped correlated variants through LD-based clumping and entered lead SNPs together in a conditional joint analysis. The analysis identified seven candidate haplotypes that remained independently associated with Alzheimer’s disease after false discovery rate correction. Hap-B, Hap-R and Hap-Y overlapped previously reported Alzheimer’s disease or longevity signals and were carried forward, while the other four showed no such overlap.
The team then used HIBAG imputation to connect the three prioritized patterns with HLA-DRB1, HLA-DQA1 and HLA-DQB1 allele combinations. Of 18,159 imputed genotypes, 17,774 were retained, or 97.9 percent. Each prioritized haplotype had a unique combination of those HLA alleles in moderate-to-high LD, allowing comparisons at several levels of genetic detail.
The three patterns diverged
Hap-B was more frequent in cognitively healthy centenarians and controls than in Alzheimer’s patients. At the SNP level, the reported odds ratio for controls versus Alzheimer’s was 0.81, with a reported interval of 0.72 to 0.91. At one-field DRB1*04 resolution, the cognitively healthy centenarian-versus-Alzheimer’s comparison had an odds ratio of 0.77, with an interval of 0.61 to 0.97 and a P value of 0.025. These odds ratios compare groups, and the study did not report absolute risk.
Hap-R showed a different distribution. It was less frequent in controls than in Alzheimer’s patients at all seven resolution levels, and less frequent in cognitively healthy centenarians than in controls at four of seven levels. At the DR1 level, the reported odds ratio was 1.35 for centenarians versus controls, with an interval of 1.05 to 1.73 and a P value of 0.021. The DR1 comparison of centenarians versus Alzheimer’s patients had an odds ratio of 1.54, with an interval of 1.19 to 1.99. At one-field DRB1*01 resolution, the controls-versus-Alzheimer’s comparison had an odds ratio of 1.16, with an interval of 1.03 to 1.32.
Hap-Y’s association was narrower. It was enriched in controls relative to Alzheimer’s patients only at the two-field DRB1*15:01 and DQB1*06:02 levels. Both comparisons had an odds ratio of 0.79. The reported intervals were 0.67 to 0.93 for DRB1*15:01 and 0.68 to 0.93 for DQB1*06:02. Other genetic features had similar frequencies in Alzheimer’s patients and cognitively healthy centenarians.
A different signal in brain tissue
The researchers also examined post-mortem temporal cortex from the 100-plus Study. The tissue analysis included 89 cognitively healthy centenarians and 7 Alzheimer’s disease patients. Microglial marker load was measured as the percentage of immunopositive area. Linear-regression models also adjusted for sex, the first five genetic principal components, age at death and the delay before post-mortem inclusion, with false discovery rate control.
The tissue results did not mirror one another. Hap-Y was associated with higher HLA-II load, with the strongest association at DRB1*15:01 and a beta value of 1.67. Hap-R was associated with lower HLA-II load at the SNP level, with beta -0.57 and a false discovery rate of 0.029. Hap-B showed no consistent microglial association.
Useful leads, unresolved biology
Together, the findings map different associations, but they do not settle how the biology works. This was an observational case-control analysis, and the COJO-derived haplotypes were exploratory LD-based proxies rather than established causal variants. The results therefore cannot show that any of the haplotypes causes Alzheimer’s disease, healthy longevity or changes in microglial HLA-II load.
The brain-tissue result also came from a much smaller subset than the genetic analysis: 89 centenarians and 7 Alzheimer’s patients, compared with 6,053 people in the primary genetic cohort. The study did not resolve which structural or non-coding variants account for the associations, and the tissue assay was not allele-resolved. That leaves open whether the Hap-Y signal is specific to Alzheimer’s disease or relates to broader neurodegeneration. For now, the work offers candidate associations for further study, not evidence of a treatment, prevention strategy or clinical test.
Paper data and sources
Original title: Fine-mapping HLA-II haplotypes in Alzheimer’s disease and healthy longevity reveals distinct associations with microglial HLA-II load and neuropathology
Authors: Sirvent DÁ, Luimes MC, Tesi N et al.
Journal/Repository: Not provided
Status: Preprint, not yet peer-reviewed
First online: 2026-08-27
DOI: 10.64898/2026.08.24.26361177
Original paper · Full text