Tau-PET adds a regional genetic readout to Alzheimer’s biology
Alzheimer’s disease is defined by extracellular β-amyloid plaques and intracellular hyperphosphorylated tau tangles, but tau carries the tighter biological link to neuronal loss and cognitive decline. That distinction is central to the new review of genetic determinants of tau-PET imaging, because the imaging phenotype captures where tau is accumulating in the brain rather than simply whether tau pathology exists in tissue. Regional tracer uptake gives investigators a more granular endophenotype than cerebrospinal fluid measures, which can signal a systemic biochemical shift without revealing the anatomic pattern of deposition.
This review organizes the literature around genome-wide and candidate-gene studies that tried to explain why some brains accumulate more tau in specific regions than others. The practical value lies in the question the field has struggled to answer for years: whether the genetics of tau spread can be separated from the genetics of clinically diagnosed Alzheimer’s disease. The evidence suggests partial separation, but the case remains incomplete and should be treated as provisional rather than settled.
CYP1B1-RMDN2 stands out, but the field is still sparse
The most convincing locus in the review is CYP1B1-RMDN2, the only genetic signal described as formally replicated in independent cohorts. The biological appeal is straightforward. CYP1B1 participates in oxidative stress pathways, and RMDN2 has been tied to microtubule dynamics, giving the locus a plausible route to influence tau aggregation and intracellular handling. In a disease where tau misfolding, trafficking failure, and oxidative injury can reinforce one another, a locus that sits near both stress biology and cytoskeletal regulation is more than a statistical curiosity.
A single replicated locus is not a genetic architecture. It is a foothold.
Other transcriptional regulators, including ZBTB20 and EYA4, were linked to tau-PET signal, but both associations rested on a single unreplicated analysis. Common MAPT variants, which one might expect to dominate any tau-related genetic story, showed only nominal associations. That pattern matters because it suggests the regional imaging phenotype may not be controlled primarily by the most obvious tau gene, or at least not in a simple way that current sample sizes can detect. It also hints that tau deposition in vivo could be shaped by a wider regulatory network involving transcription, trafficking, and cellular stress rather than a single canonical tau pathway.
Phosphorylation, immune signaling, and trafficking appear in the same signal field
Genes involved in tau phosphorylation were also part of the review’s map. PPP2R2B was implicated, although the association was reported without a replication sample, limiting how much biological confidence can be placed in the finding. FYN was associated only nominally, with a p value of 0.048 in a candidate-gene study of 146 participants. Small samples of that size are fragile in genetics, where weak signals can disappear once the cohort changes or the phenotype is measured slightly differently. Still, the pathway is biologically coherent: tau phosphorylation is not a decorative feature of the disease, but one of the core molecular steps that allows pathology to spread and become toxic.
Immune- and trafficking-related genes such as CR1, CLU, and BIN1 also appeared repeatedly enough to suggest that neuroinflammation and endosomal transport contribute to regional tau accumulation. Those findings fit the broader Alzheimer’s literature, where microglial biology and vesicular trafficking have become central to explaining why pathology advances in some individuals and not in others. In tau-PET studies, however, the signal is still indirect. The tracer measures uptake, not neuropathologic tau itself, and each association must be interpreted through that imaging constraint.
APOE ε4 remains the dominant predictor, but its effect is conditional
Across the studies summarized in the review, APOE ε4 was the most frequently reported predictor of tau-PET burden across brain regions. Even so, its effect weakened in population-based cohorts composed largely of cognitively unimpaired participants. That attenuation is important because it suggests APOE ε4 may not drive tau-PET burden in a fully amyloid-independent way. Instead, its effect may depend heavily on amyloid context, consistent with the idea that APOE sits upstream in a broader cascade that begins with amyloid biology and only later amplifies tau deposition.
The clinical implication is not that APOE becomes irrelevant in unimpaired people, but that its meaning changes with the cohort. In memory-clinic populations enriched for symptomatic disease, APOE ε4 can look like a strong global determinant. In community cohorts, where many participants do not yet have overt cognitive impairment, the same allele may exert a smaller and more conditional effect. This is one reason the review resists overclaiming. A genotype that appears dominant in one setting may lose force once disease stage, amyloid burden, and sample composition are accounted for.
Why the imaging phenotype matters more than a standard AD diagnosis
The most interesting feature of this body of work is its attempt to move beyond diagnosis and toward spatial biology. Clinically diagnosed Alzheimer’s disease is a broad endpoint that compresses multiple biological routes into a single label. tau-PET preserves regional detail, which makes it better suited for detecting genetic effects that influence where tau accumulates, how quickly it spreads, and which circuits are exposed first. A variant that alters tau signal in a specific cortical network may never show up cleanly in a case-control GWAS of Alzheimer’s disease itself.
That is the real promise of tau-PET genetics: not a replacement for disease genetics, but a more refined phenotype for disentangling mechanisms. The review argues that the architecture of tau-PET signal appears at least partly separable from clinically diagnosed Alzheimer’s disease. Yet no formal genetic correlation has been estimated, the cohorts overlap substantially, and the sample sizes are too small for a null result to carry much weight. Those limitations leave the current literature with a familiar problem in biomarker genetics: the phenotype is sophisticated, but the evidence base is still thin.
Replication gaps limit therapeutic translation
The therapeutic implications are less direct than many genetics papers imply. A locus associated with tau-PET signal does not automatically nominate a drug target, and a statistically significant imaging association does not prove a pathway can be modified to slow neurodegeneration. The review is careful on this point. It treats tau-PET as an endophenotype worth studying, while also making clear that the current associations do not yet establish a druggable map for Alzheimer’s treatment.
Replication remains the bottleneck. Some associations rest on one analysis only, others are nominal, and several of the cohorts are small or overlapping. In a field this underpowered, the absence of an effect means very little unless the study is large enough to detect it. Until independent cohorts reproduce the stronger signals and formal correlation analyses are performed, the genetic determinants of tau-PET should be read as promising leads rather than settled biology.
The broader takeaway is narrower than the title of the literature might suggest. Tau-PET genetics is beginning to separate regional tau biology from generic Alzheimer’s diagnosis, but the clearest signal so far belongs to a single replicated locus and a handful of weak or unreplicated candidates. The next step is not more speculation about mechanism; it is larger, better harmonized imaging cohorts with replication built in from the start.