The medial temporal lobe, encompassing the entorhinal cortex and hippocampus, is where Alzheimer’s pathology takes root. Tau tangles appear here earlier than anywhere else in the brain, and the density of these misfolded filaments tracks closely with the episodic memory failures that define the disease. Yet while the link between neocortical tau and amyloid-β is well established, the relationship within the medial temporal lobe has remained murky. Positron emission tomography, the workhorse of in vivo amyloid and tau imaging, struggles in this deep, folded region because of limited spatial resolution, off-target tracer binding, and reduced sensitivity. Autopsy tissue, by contrast, allows direct, region-specific quantification of plaques and tangles, which is precisely why the researchers turned to two of the world’s most rigorously characterized brain donation programs.

The main cohort came from the Religious Orders Study and the Rush Memory and Aging Project, longitudinal studies that have followed older Catholic clergy and Chicago-area residents for decades, with 1,587 participants donating their brains at death. The replication cohort drew from the Arizona Study of Aging and Neurodegenerative Disorders, adding 963 more donors. In the main cohort, amyloid-β load was measured as the percentage of tissue area occupied by immunostained deposits, while tau was quantified as neurofibrillary tangle density using the AT8 antibody across eight cortical regions. The Arizona cohort used semi-quantitative CERAD-style grading with thioflavin-S, Campbell-Switzer, and Gallyas staining. Crucially, both datasets allowed the team to compute separate amyloid-β and tau composites for the medial temporal lobe and for the six remaining neocortical regions, and to adjust for co-pathologies including TDP-43, Lewy bodies, and vascular lesions.

The starting point was unsurprising: in both cohorts, carrying at least one APOE-ε4 allele and being female were each associated with significantly greater medial temporal tau tangle burden, with all p-values below 0.001. APOE-ε4 carriers also showed markedly higher amyloid-β in both the neocortex and the medial temporal lobe. The effects of female sex on amyloid were weaker and less consistent, reaching significance in the main cohort only. The pivotal question was what would happen when amyloid-β entered the statistical models as a potential mediator of the tau associations.

Adjusting for global neocortical amyloid-β alone attenuated both associations but did not eliminate them. The turning point came when the researchers jointly accounted for global and regional medial temporal amyloid-β. Under that fuller specification, the association between APOE-ε4 and medial temporal tau dissolved entirely, falling to non-significance in both cohorts with p-values of at least 0.19. The female sex effect, however, remained robust, with p-values of 0.004 or lower. Models incorporating both spatial scales of amyloid pathology explained substantially more variance in medial temporal tau than global measures alone, and the results held when the team additionally adjusted for diagnostic group, clinical stage, or neocortical tau burden.

Path analysis, a form of structural equation modeling that traces how effects flow through intermediate variables, sharpened the picture. The researchers ordered the pathology variables a priori according to known neuropathological staging: diffuse neocortical plaques precede neuritic plaques, neocortical amyloid precedes medial temporal amyloid, and medial temporal amyloid precedes tangles. In both cohorts, roughly three quarters of the total APOE-ε4 effect on medial temporal tau was mediated by amyloid-β, 79.5 percent in the main cohort and 74.5 percent in the replication sample, and the residual direct effect was statistically indistinguishable from zero. For female sex, less than half of the effect was amyloid-mediated, just 44.0 percent and 14.4 percent respectively, and a significant amyloid-independent component persisted in both cohorts.

Sensitivity analyses reinforced the dissociation. Among donors with no significant amyloid-β pathology on the CERAD scale, 499 in the main cohort and 211 in the replication set, APOE-ε4 showed no association with medial temporal tau at all. Female sex, meanwhile, remained significantly linked to higher tau in the main cohort and showed a strong trend in the replication sample. When the team repeated the analyses separately for the entorhinal cortex and hippocampus, and when they tested for APOE-ε4 by sex interactions, the same pattern emerged: after accounting for amyloid-β, women carried more medial temporal tau regardless of genotype.

The authors propose plausible, if unproven, mechanisms for the sex-specific vulnerability. Endocrine transitions, particularly the loss of estrogen around menopause, have been linked to tau burden in prior imaging studies, and sex differences in neuroimmune biology, including microglial activation, have been tied to both amyloid and tau pathologies. The persistence of the sex effect in amyloid-free brains and after adjustment for disease-stage proxies suggests that female-associated tau accumulation in the medial temporal lobe is at least partly an amyloid-independent process, potentially accelerating progression and amplifying tau-mediated cognitive decline. The study’s design, being cross-sectional autopsy data, cannot establish causality or temporal ordering, and the cohorts lack racial, ethnic, and socioeconomic diversity, limitations the authors acknowledge candidly.

For the APOE-ε4 story, the findings carry a more encouraging message. If the gene variant amplifies tau largely through its strong effects on amyloid-β accumulation and distribution, then the new generation of anti-amyloid therapies might, in principle, blunt the downstream tau consequences that matter most for cognition, particularly in early disease stages when localized amyloid and tau deposits fall below PET detection thresholds. The study also highlights a methodological blind spot: most in vivo studies quantify amyloid with a global neocortical composite, potentially obscuring the local amyloid-tau interplay within the medial temporal lobe that the autopsy data reveal. Whether plaque removal alters tau trajectories in a genotype-dependent manner will require longitudinal and interventional evidence that this design cannot provide.

The practical upshot is a call for precision in trial design. Post hoc analyses of some anti-amyloid trials have hinted at sex differences in treatment outcomes, though such exploratory subgroup findings were never powered for definitive interaction testing. The new results argue that sex and APOE genotype should be built into prespecified analytic plans for therapies targeting amyloid, tau, or APOE biology, rather than treated as afterthoughts. They also frame a pressing open question for the field: identifying the biological basis of women’s heightened medial temporal tau vulnerability could clarify why Alzheimer’s prevalence and progression differ by sex and open therapeutic avenues that current amyloid-centric strategies leave untouched. Until then, the autopsy bench has delivered a clear verdict, APOE-ε4 works through amyloid, and female sex, at least in part, works around it.

Subject of Research: Amyloid-β-mediated versus amyloid-independent pathways linking APOE-ε4 and female sex to medial temporal tau pathology in Alzheimer's disease

Article Title: Global and regional amyloid-β in explaining APOE-ε4 and sex associations with medial temporal tau: a neuropathology study

Article References: Salvadó, G., Groot, C., Wisse, L. E., Vogel, J. W., Beach, T. G., Serrano, G. E., Reiman, E. M., Schneider, J. A., Bennett, D. A., Ossenkoppele, R., & Hansson, O. (2026). Global and regional amyloid-β in explaining APOE-ε4 and sex associations with medial temporal tau: a neuropathology study. Acta Neuropathologica, 152(1), Article 49. https://doi.org/10.1007/s00401-026-03095-2

Image Credits: AI Generated

DOI: 10.1007/s00401-026-03095-2

Keywords: Alzheimer's disease, APOE-ε4, tau tangles, amyloid-β, medial temporal lobe, neuropathology, sex differences, hippocampus, entorhinal cortex, brain donation, dementia risk, clinical trials

GSK-3β has occupied a central place in Alzheimer’s research for decades. The enzyme is a serine/threonine kinase that phosphorylates a wide array of substrates, including the tau protein whose hyperphosphorylated tangles are a hallmark of the disease. In a mouse model engineered to overexpress GSK-3β, previous work had already documented a decrease in dendritic spine density in the dentate gyrus, the gateway through which cortical information enters the hippocampus. What remained unknown was how the rest of the trisynaptic circuit—the entorhinal cortex, CA3, and CA1—responds to elevated levels of this kinase. The new study, led by María Llorens-Martín and colleagues at the Centro de Biología Molecular Severo Ochoa in Madrid, set out to fill that gap.

To capture the fine structure of neurons, the researchers turned to a classical but technically demanding technique: intracellular injection of Lucifer Yellow, a fluorescent dye that fills individual neurons in fixed tissue and reveals their dendritic trees and spines in exquisite detail. Using this method, the team analyzed dendritic spine morphology and density on both apical and basal dendrites of pyramidal neurons in the entorhinal cortex, CA3, and CA1, comparing wild-type mice with mice overexpressing GSK-3β. Spines—the tiny protrusions studding dendrites where most excitatory synapses reside—come in distinct morphological classes, including mushroom, thin, and stubby forms, each thought to reflect different stages of synaptic maturity and stability. The head diameter and length of each spine type were quantified, alongside the overall density of spines.

The first striking result came from the wild-type animals themselves. Even in healthy brains, the study found intrinsic differences in spine density and in the distribution of spine types between apical and basal dendrites across the different regions of the trisynaptic circuit. This baseline regional heterogeneity matters, because it establishes that apical and basal dendrites— which receive different complements of afferent inputs—are not interchangeable substrates. Any pathological process acting on the circuit will therefore be superimposed on a landscape already patterned by region- and compartment-specific wiring.

When the researchers compared the GSK-3β-overexpressing mice to their wild-type littermates, a clear hierarchy of vulnerability emerged. The entorhinal cortex, despite being the region typically hit earliest by Alzheimer’s pathology in humans, showed only subtle morphological changes in this model, a finding the authors interpret as a form of partial structural resilience. The CA3 region, likewise, presented minimal alterations overall—although here the basal dendrites were more affected than their apical counterparts, a pattern the authors attribute to the differential inputs these two dendritic compartments receive. In sharp contrast, CA1 stood out as the most affected region of the entire circuit, a result consistent with the well-documented vulnerability of CA1 to Alzheimer’s disease pathology in both human patients and animal models.

Beyond spine shape, the study probed the functional architecture of the circuit by examining excitatory innervation. The researchers stained for vesicular glutamate transporter 1, or VGlut1, a marker of excitatory presynaptic boutons, and measured the percentage of area occupied by VGlut1-positive terminals around the dendritic compartments. In both CA1 and the basal dendrites of CA3, the GSK-3β-overexpressing mice showed a reduced percentage of area occupied by VGlut1-positive boutons, indicating a loss of excitatory innervation. Because excitatory drive is the currency of synaptic plasticity—the cellular substrate of learning and memory—this loss suggests that GSK-3β overexpression does not merely remodel the shape of spines but actively disrupts the connectivity needed for plasticity to occur.

The regional selectivity of these alterations carries significant implications for how scientists think about the progression of Alzheimer’s disease. Rather than a diffuse, homogeneous degeneration of the hippocampal formation, the data support a model in which specific subregions and even specific dendritic compartments are differentially susceptible to a single molecular insult. The vulnerability of CA1 is particularly noteworthy: this region serves as the major output hub of the hippocampus, relaying processed information back to the cortex. Structural compromise at this node, combined with lost excitatory innervation, could plausibly undermine the coherence of the entire memory circuit even while upstream regions such as the entorhinal cortex and CA3 appear comparatively preserved.

The study also underscores the value of apical–basal comparisons. In CA3, basal dendrites were more affected than apical dendrites, likely reflecting the different sources of input they receive—apical dendrites in CA3 are heavily targeted by perforant path fibers from the entorhinal cortex and associational inputs, whereas basal dendrites sample a different mix of local and long-range afferents. That GSK-3β overexpression left a compartment-specific fingerprint on the circuit suggests that the enzyme’s effects are mediated not only by intrinsic neuronal properties but also by the identity of the inputs each dendrite integrates. This level of resolution would have been impossible with bulk tissue measurements and highlights why single-cell anatomy remains indispensable in neuroscience.

Methodologically, the work relied on carefully characterized mouse cohorts, with the GSK-3β-overexpressing model validated by β-galactosidase and NeuN staining to confirm transgene expression and neuronal counts across the entorhinal cortex, CA3, and CA1. Excitatory and inhibitory balance was further assessed by quantifying GAD65-positive inhibitory boutons alongside VGlut1-positive excitatory terminals, allowing the team to compute ratios that capture the excitatory–inhibitory landscape around each dendritic compartment. Statistical comparisons were performed across regions, genotypes, and dendritic compartments, with sample sizes of four to five animals per group for the various analyses, lending statistical weight to the observed regional differences.

Taken together, the study paints a nuanced picture of GSK-3β as a driver of regionally selective synaptic deterioration within the memory circuit. By demonstrating that the entorhinal cortex retains partial structural resilience, that CA3 is largely spared except for its basal compartment, and that CA1 is the most affected node, the researchers provide a cellular framework for understanding why certain hippocampal areas fail early in Alzheimer’s disease while others endure. The loss of excitatory innervation in CA1 and basal CA3 further implicates disrupted synaptic plasticity as a downstream consequence of kinase overexpression. As the field continues to pursue GSK-3β as a therapeutic target, these findings caution that interventions will need to account for the striking regional and compartmental heterogeneity of the hippocampal circuit—and they offer a detailed anatomical benchmark against which future protective strategies can be measured.

Subject of Research: Regional dendritic spine and innervation changes caused by GSK-3β overexpression in the hippocampal trisynaptic circuit of a mouse model relevant to Alzheimer's disease.

Article Title: Regional differences in dendritic spine morphology and innervation of the trisynaptic circuit of a mouse model of GSK-3β overexpression

Article References: Alonso-Moreno, M. C., Terreros-Roncal, J., Rodríguez-Moreno, C. B., Ávila, J., & Llorens-Martín, M. (2026). Regional differences in dendritic spine morphology and innervation of the trisynaptic circuit of a mouse model of GSK-3β overexpression. Cellular and Molecular Life Sciences. https://doi.org/10.1007/s00018-026-06304-z

Image Credits: AI Generated

DOI: 10.1007/s00018-026-06304-z

Keywords: Alzheimer's disease, GSK-3β, dendritic spines, hippocampus, trisynaptic circuit, entorhinal cortex, CA1, CA3, pyramidal neurons, VGlut1, synaptic plasticity, mouse model