The results provide preliminary evidence that these anthocyanins can modulate the PI3K/Akt/GSK3β signaling pathway in a vertebrate model, but they do not establish a treatment for Alzheimer’s disease in people. The researchers induced selected Alzheimer’s-like features by exposing adult wild-type zebrafish to aluminum chloride in water for four consecutive days. On Day 4, fish also received an intraperitoneal aluminum chloride injection to reinforce model induction.

Donepezil, an acetylcholinesterase inhibitor used in Alzheimer’s treatment, served as the positive control. C3R, C3S, and C3X were each administered at 2 milligrams per kilogram by intragastric gavage once daily for five days, beginning with the first day of aluminum exposure. That common dose was selected from preliminary screening at 0.5, 1, and 2 milligrams per kilogram to standardize the comparison rather than to identify a separate maximum effective dose for each compound.

The principal experiment included 15 fish per group: untreated controls, aluminum-exposed model animals, donepezil controls, and one group for each cyanidin glycoside. Behavioral testing showed that the model reproduced several of the changes the investigators sought to study. In the novel object recognition test, aluminum-exposed fish spent less time exploring a novel object and showed less preference for it than untreated controls, indicating impaired recognition memory.

In the color-biased T-maze, model fish took longer to enter the rewarded green chamber and spent less time there after learning the task. All three cyanidin glycosides increased exploration of the novel object and improved T-maze performance relative to the model group. C3R and C3X also shortened the latency to enter the upper region of a novel tank, where exploration was reduced in the model group, indicating attenuation of anxiety-like behavior.

These memory effects were not attributable to a simple increase in general movement: total distance traveled and average velocity did not differ significantly between the model and treatment groups in the relevant novel object recognition and T-maze comparisons. Brain biochemical analysis indicated that the compounds partly reversed the cholinergic and oxidative abnormalities associated with aluminum exposure. C3R, C3S, and C3X shifted these measurements toward control values, as did donepezil.

The glycosides also increased superoxide dismutase activity and glutathione content and reduced malondialdehyde, which the investigators used as a measure of lipid peroxidation. Together, these observations are consistent with reduced oxidative stress and partial restoration of cholinergic function. The experiment did not establish whether changes in oxidative stress preceded the behavioral or cellular effects, and it did not exclude contributions from other mechanisms.

Histological and ultrastructural examinations supported the idea that the treatments protected brain tissue. Hematoxylin and eosin staining showed loose cellular organization, reduced neuronal density, and vascular congestion in the model group, whereas treated fish displayed more intact tissue and greater cellular density. Nissl staining revealed fewer and more disorganized neurons and fewer Nissl bodies after aluminum exposure.

Treatment with each glycoside or donepezil significantly increased the amount of Nissl-positive material. Transmission electron microscopy showed disrupted membrane boundaries, cytoplasmic vacuolization, cellular edema, swollen mitochondria, and fewer identifiable organelles in model animals. C3R produced the clearest preservation of membranes, mitochondria, and other organelles, while C3S and C3X also produced improvements that left some edema and vacuolization.

These images provide morphological evidence of neuronal protection, but they cannot determine whether mitochondrial fission, fusion, or caspase-dependent apoptosis was directly affected. Gene-expression measurements showed that aluminum exposure altered molecular markers associated with cholinergic function, the amyloid precursor protein family, neurotrophic signaling, inflammation, and PI3K/Akt/GSK3β signaling. The model brain had increased expression of ache, aplp2, il6, il1b, tnfa, and rela, along with reduced chata, bdnf, and ntrk2b.

Expression of pik3r1 and akt2 decreased, whereas gsk3ba increased. C3R, C3S, C3X, and donepezil moved these measures toward control levels, with significant reductions in the elevated inflammatory transcripts and restoration of pathway-related transcripts. Western blotting produced a parallel protein-level pattern: treated fish had lower acetylcholinesterase and higher choline acetyltransferase, while phosphorylation of Tau at Ser396 declined.

Immunofluorescence imaging likewise showed weaker acetylcholinesterase signals and stronger choline acetyltransferase signals after treatment, with C3R producing relatively pronounced recovery. Aluminum exposure reduced the phosphorylation of PI3K p85 at Tyr458 and Akt at Ser473 while increasing the ratio of phosphorylated to total GSK3β at Tyr216. C3R and C3X significantly reversed these alterations, and donepezil produced similar effects.

C3S increased PI3K Tyr458 phosphorylation and reduced GSK3β Tyr216 and Tau Ser396 phosphorylation, but it did not significantly restore Akt Ser473 phosphorylation, indicating a partial pathway response. The interpretation requires care because Akt canonically influences GSK3β through inhibitory phosphorylation at Ser9 rather than Tyr216, and the study did not directly measure GSK3β kinase activity. The measurements therefore indicate coordinated pathway-associated regulation rather than proof that Akt directly controls GSK3β phosphorylation at Tyr216.

A separate experiment tested the functional involvement of PI3K signaling by administering the PI3K inhibitor LY294002 two hours before the cyanidin glycosides. Blocking PI3K attenuated the treatment-related increases in choline acetyltransferase, phosphorylated PI3K, and phosphorylated Akt. At the same time, it increased acetylcholinesterase, phosphorylated GSK3β, and phosphorylated Tau at Ser396.

The inhibitor result supports participation of PI3K/Akt signaling in the biochemical effects observed in the model, but it does not identify the direct molecular targets of the glycosides. Total PI3K, Akt, and GSK3β abundance did not change significantly among the relevant groups. Molecular docking was used to explore possible structural differences among the compounds.

C3R received the most favorable predicted scores for human PI3Kα and AKT1, followed by C3X and C3S for PI3Kα, while C3S ranked ahead of C3X for AKT1. These calculations are hypotheses about possible accommodation at the proteins, not evidence of direct binding or functional regulation, and they cannot fully account for C3R’s stronger biological effects in the experiment. Among the three compounds, C3R emerged as the most consistently effective candidate under the shared dose and short treatment schedule.

Its advantages appeared across behavioral testing, biochemical markers, histological and ultrastructural observations, cholinergic protein measurements, and PI3K/Akt/GSK3β-associated changes. The comparison does not show that C3R’s rutinoside-linked structure caused its stronger performance; the source presents the distinct sugar groups as a possible basis for further investigation rather than a demonstrated explanation. The results instead identify C3R as a priority for systematic follow-up.

Controlled gavage also ensured that the fish received measured compound doses, but it did not reproduce the gastrointestinal conditions, food matrix, and exposure patterns associated with eating cyanidin-rich foods. Translation of the findings remains limited in several ways. The model lasted only five treatment days and reproduced selected features such as cognitive impairment, oxidative stress, cholinergic dysfunction, inflammation, and increased Tau Ser396 phosphorylation.

It did not reproduce the chronic, age-dependent development of Alzheimer’s disease. The experiment also did not determine whether behavioral benefits persist after treatment ends, whether general antioxidant effects or interactions with aluminum contributed to the phenotype, or how the compounds behave after longer exposure. The gastrointestinal stability, bioavailability, metabolism, pharmacokinetics, and brain exposure of C3R, C3S, and C3X were not established, and species-specific differences limit direct extrapolation from zebrafish to mammals.

The researchers identified mammalian work, including tests in APP/PS1 mice, as a necessary next step for evaluating pharmacokinetic properties, blood–brain barrier penetration, efficacy, and direct targets. For now, the study offers a controlled in vivo screening result with broad but preliminary implications. Food-derived anthocyanins can produce measurable behavioral, biochemical, histological, ultrastructural, and molecular effects in an adult vertebrate exposed to a toxin-induced Alzheimer’s-like phenotype.

C3R performed best overall among the three equal-dose treatments, and the inhibitor experiment linked part of its biochemical profile to PI3K/Akt signaling. However, the work does not show that dietary supplementation could treat or prevent Alzheimer’s disease in humans.

Subject of Research: Neuroprotective effects of three food-derived cyanidin glycosides in an aluminum-induced Alzheimer’s-like adult zebrafish model

Article Title: Neuroprotective Effects of Food‐Derived Cyanidin Glycosides in an Alzheimer's Disease‐Like Zebrafish Model Through Modulation of PI3K/Akt/GSK3β Signaling

Article References: Wu, Y., Tie, F., Gao, Y., Li, J., Wang, R., Hu, N., Dong, Q., & Wang, H. (2026). Neuroprotective Effects of Food‐Derived Cyanidin Glycosides in an Alzheimer's Disease‐Like Zebrafish Model Through Modulation of PI3K/Akt/GSK3β Signaling. Food Science & Nutrition, 14(10), Article e72486. https://doi.org/10.1002/fsn3.72486

Image Credits: AI Generated

DOI: 10.1002/fsn3.72486

Keywords: cyanidin glycosides, cyanidin-3-O-rutinoside, zebrafish model, Alzheimer’s-like pathology, PI3K/Akt/GSK3β, cholinergic function, oxidative stress, Tau phosphorylation, anthocyanins, neuroprotection, Neuroprotective, Effects

The rationale for the work rests on a well-documented but incompletely understood connection between metabolism and neurodegeneration. Obesity and type 2 diabetes are established risk factors for Alzheimer’s disease, and midlife obesity is associated with roughly a doubling of dementia risk. Insulin resistance, systemic inflammation, and impaired cerebral glucose uptake all appear to feed the pathological cascade that produces amyloid-beta plaques and neurofibrillary tangles. Dietary interventions such as caloric restriction, protein restriction, and intermittent fasting have each shown benefits in mouse models, but their translation to humans has been hampered by poor long-term adherence. Incretin-based drugs targeting the GLP-1 and GIP receptors, meanwhile, have produced mixed results in preclinical studies and, in recent phase 3 trials of semaglutide in early-stage Alzheimer’s disease, failed to demonstrate cognitive benefit.

Bariatric surgery occupies a different position in this landscape. Unlike diets that demand daily willpower or drugs that require indefinite adherence, sleeve gastrectomy is a one-time intervention with lifelong effects, achieving rapid and durable weight loss along with resolution of obesity-related metabolic disease. Human studies have hinted at neurological dividends: neuroimaging reveals structural and functional brain changes after surgery, cognitive improvements have been documented as early as twelve weeks postoperatively, and a recent retrospective cohort study linked bariatric surgery to reduced incidence of mild cognitive impairment and Alzheimer’s-related dementias. What remained unknown was whether the surgery directly alters Alzheimer’s-specific pathology, and through what mechanisms.

To answer that question, the team turned to the 3xTg-AD mouse, an animal that carries familial human variants of three Alzheimer’s-related genes—APP, tau, and presenilin—and develops the cardinal features of the human disease, including amyloid plaques, tangles, and progressive cognitive decline. Female mice were fed a Western diet for two months to induce obesity and glucose intolerance, mimicking the metabolic profile of patients seeking bariatric surgery, and were then randomized to receive either sleeve gastrectomy or a sham operation. After surgery, one cohort was maintained on a standard control diet while a second remained on the obesogenic Western diet for life, allowing the researchers to test the surgery’s effects both under baseline conditions and under sustained metabolic stress.

The metabolic results were striking for what they did not involve. Across twelve months of follow-up, operated mice gained less weight than sham controls in both dietary contexts, an effect driven almost entirely by reduced fat mass, with lean mass actually preserved or increased. Yet cumulative food intake was identical between groups, and measurements in metabolic chambers showed no differences in energy expenditure or substrate utilization. The body-composition shift, in other words, was not the product of eating less or burning more. In the control-fed cohort, the surgery also significantly improved insulin sensitivity on a tolerance test—an effect seen in the transgenic mice but not in wild-type littermates—even though glucose tolerance itself was unchanged, suggesting compensatory adaptations specific to the Alzheimer’s-prone background.

When the researchers examined the brains of fifteen-month-old animals, the neuropathological signature was unmistakable. Sleeve gastrectomy significantly reduced the percentage of cortical area occupied by amyloid plaques, with a parallel trend toward fewer plaques, and the effect was accompanied by reduced immunoreactivity for GFAP, a marker of reactive astrocytes, and IBA1, a marker of activated microglia—the two cellular faces of Alzheimer’s neuroinflammation. Notably, these changes were confined to the cortex; the hippocampus showed no significant differences. The regional pattern is consistent with the well-described spatiotemporal progression of Alzheimer’s disease, in which pathology first emerges in neocortical regions before spreading to medial temporal structures, raising the possibility that the surgery intervenes at an early stage of pathological propagation.

Tau phosphorylation told a more nuanced story. The surgery produced epitope-specific effects: a strong trend toward reduced phosphorylation at threonine 231, an early and clinically relevant event mediated by the kinase GSK-3β, alongside a significant increase in the AT8 epitope, which is associated with more established tangle pathology. The authors interpret this divergence as evidence that the surgery is most effective at modulating early-stage tau phosphorylation, before mature paired helical filaments form, though they acknowledge that fully powered comparisons across genotypes will be needed to resolve the pattern.

Perhaps the most mechanistically revealing findings came from the autophagy analysis. Autophagy, the cellular machinery that clears misfolded proteins and damaged organelles, is known to fail in Alzheimer’s disease, contributing to the accumulation of plaques and tangles. In the cortex of operated mice, where plaque burden had already fallen, the initiation proteins ATG5, ATG7, and ATG16L1 were downregulated and the cargo receptor p62 was reduced—consistent with a homeostatic dialing-back of autophagic demand once pathological load had been cleared. In the hippocampus, where plaque burden remained elevated, the opposite occurred: ATG7 and ATG16L1 were upregulated, LC3A/B increased, and p62 dropped sharply, indicating enhanced autophagic flux and active cargo clearance. The opposing patterns tracked inversely with phosphorylated AKT, an upstream inhibitor of autophagy initiation through the mTORC1 pathway, suggesting that restored insulin signaling after surgery may extend to the brain and drive region-specific modulation of the insulin–autophagy axis.

Cognition added a final layer of complexity. In the control-fed cohort, behavioral testing at five and eight months after surgery revealed no differences between operated and sham mice on the Barnes maze, fear conditioning, or novel object recognition—but neither did sham mice differ from wild-type animals, indicating that the standard diet had failed to produce a measurable cognitive deficit for the surgery to rescue. Under lifelong Western diet exposure, however, the picture changed dramatically. Sham-fed transgenic mice showed clear spatial memory impairment, and sleeve gastrectomy substantially rescued it: operated mice found the escape box faster, with higher success rates and shorter search paths, in both short- and long-term memory probes. The same animals also showed significantly reduced plaque numbers, reduced phosphorylated tau, visually attenuated glial activation, and—extending the benefits beyond the brain—significantly lower frailty scores across the study period, a measure of accumulated age-associated deficits that was also reduced in wild-type operated mice.

The authors are careful about limitations. The study used only female mice, because males in the available colony do not reliably develop the disease phenotype; the Western diet cohort suffered substantial age-associated mortality that reduced group sizes; autophagic flux was inferred from protein markers rather than measured directly; and survival analysis was underpowered, with no difference detected between surgical groups in either diet. Even so, the convergence of evidence—reduced adiposity, improved insulin sensitivity, attenuated cortical amyloid and glial activation, region-specific autophagy remodeling, rescued spatial memory, and reduced frailty—constitutes a compelling proof of concept. Critically, the neuroprotective effects are unlikely to be mediated solely through the GLP-1 axis: fasting GLP-1 rose in wild-type but not transgenic operated mice, and prior work has shown that sleeve gastrectomy’s metabolic benefits persist in GLP-1 receptor knockout animals. That mechanistic breadth may explain why a single surgical intervention succeeded where incretin pharmacotherapy has so far fallen short in Alzheimer’s trials. Whether the benefits translate to humans remains an open question, but the study establishes metabolic surgery as a serious candidate for disease modification in obesity-associated Alzheimer’s disease—and points the field toward the gut–brain pathways that make the stomach, unexpectedly, a target organ for the brain.

Subject of Research: Effects of sleeve gastrectomy on metabolic health, cognition, and Alzheimer's disease pathology in a transgenic mouse model

Article Title: Sleeve Gastrectomy Improves Metabolic Health, Cognition, and Alzheimer's Disease Pathology in 3xTG Mice

Article References: Babygirija, R., Illiano, J. A., Sonsalla, M. M., Zhu, G., Jabbarpour, C. R., Mathew, T., Molkentin, T., Peterson, M., Winder, C., Nair, A. R. K., Xiao, F., Wolter, J. M., Lamming, D. W., & Harris, D. A. (2026). Sleeve Gastrectomy Improves Metabolic Health, Cognition, and Alzheimer's Disease Pathology in 3xTG Mice. Aging Cell, 25(10), Article e70767. https://doi.org/10.1111/acel.70767

Image Credits: AI Generated

DOI: 10.1111/acel.70767

Keywords: sleeve gastrectomy, Alzheimer's disease, bariatric surgery, 3xTg-AD mice, amyloid plaques, tau phosphorylation, autophagy, insulin sensitivity, neuroinflammation, Western diet, GLP-1, spatial memory

The problem the researchers set out to solve is one of the most stubborn technical bottlenecks in cell biology. Antibodies are superb molecular recognition machines, capable of distinguishing protein conformations, aggregation states, and individual post-translational modifications with extraordinary precision. But the plasma membrane is an impenetrable barrier to these Y-shaped proteins. Traditional workarounds each carry significant drawbacks. Microinjection and electroporation can introduce antibodies into cells but are slow, laborious, and damaging, making them poorly suited for treating large populations. Genetically encoded intrabodies can be expressed inside cells, but they demand viral vectors or transfection, require days to weeks of expression, and typically perturb all forms of a target protein at once rather than a single modified species. In many cases, antibodies that do get taken up remain trapped in endosomes, never reaching their cytoplasmic targets.

The Northwestern team, led by Kirsten Viola, Brett Chromy, Mary Lambert, and William Klein of the Department of Neurobiology, turned to a cationic lipid with an unusual pedigree. The molecule, 1,2-dioleoyl-sn-glycero-3-ethylphosphocholine, abbreviated DOPC+, was originally developed as a gentle vehicle for DNA transfection. The CLD procedure is remarkably simple and takes less than an hour from start to finish. First, the lipid is dried into a thin film inside a glass tube under a stream of nitrogen or argon. Next, a small volume of antibody solution is added to resuspend the film, and the mixture is vortexed into a milky suspension of multi-lamellar vesicles. A brief sonication step breaks these large vesicles down into small unilamellar liposomes roughly 50 nanometers in diameter, as confirmed by atomic force microscopy. Finally, the antibody-lipid complexes are diluted into serum-free culture medium and applied to the cells, where delivery occurs in under thirty minutes.

To validate the platform, the researchers first tested whether CLD could ferry small molecules across the membrane. Fluorescent BODIPY-phallacidin, a probe that binds actin filaments, was delivered into cultured 3T3 fibroblasts and B103 rat neuroblastoma cells, where it lit up stress fibers and actin bundles at the tips of filopodia within twenty minutes. A synthetic peptide derived from the N-terminal domain of the tyrosine kinase Fyn was likewise delivered into B103 cells and dispersed homogeneously through the cytoplasm. Dextrans ranging from 10 to 70 kilodaltons also crossed successfully. The critical control came from mixing unencapsulated antibody with empty liposomes: essentially no intracellular labeling resulted. The antibody must be incorporated into the lipid complex itself; simply bathing cells in antibody alongside empty vesicles accomplishes nothing.

The decisive experiments involved full-length antibodies. When N1E-115 mouse neuroblastoma cells were incubated with liposomes loaded with Fyn3 antibody, the resulting staining pattern across the population was nearly identical to that obtained by conventional immunocytochemistry of fixed, permeabilized cells. Antibodies against vinculin and paxillin—two components of focal adhesions, the protein complexes that anchor cells to their substrate—were delivered into B103 cells and correctly localized to discrete puncta at the tips of growth cones, exactly where these proteins are known to reside. Because focal contacts have such a characteristic, spatially restricted distribution, they served as a rigorous test of whether delivered antibodies retain their targeting function in the living cytoplasm. They did, without exception.

With the delivery system validated, the team deployed their Alzheimer’s-relevant cargo. PHF-1 is a monoclonal antibody raised against paired helical filament tau, the hyperphosphorylated species that accumulates in the neurofibrillary tangles that define Alzheimer’s disease progression. It binds tau phosphorylated at serines 396 and 404, epitopes widely used as pathological biomarkers in tissue staining and in emerging fluid diagnostics. Yet the same epitope is not exclusively pathological. Earlier work had shown that PHF-1-reactive tau appears transiently in the developing nervous system—both in human fetal brain and in embryonic chick retina—where tau phosphorylation is dynamically regulated during axon formation. The B103 and N1E-115 cell lines used in the study constitutively express the PHF-1 epitope, making them convenient models.

When encapsulated PHF-1 was delivered into B103 cells, it produced filamentous staining patterns identical to those seen with traditional fixation-and-permeabilization labeling, confirming that the antibody had escaped the endosomal route and accessed its cytoskeletal target. Delivery into primary chick retinal neurons—an important demonstration, since these post-mitotic cells cannot be expanded and are far more fragile than immortalized lines—was similarly efficient, with virtually every neuron in the culture taking up the antibody. Crucially, cells tolerated the procedure well: cultures incubated with the complexes for more than twenty-four hours showed no detectable toxicity, and up to 100 micrograms of DOPC+ per milliliter of medium was tolerated by rat hippocampal neurons, chick retinal neurons, and the B103 line alike.

The surprise came when the researchers monitored living B103 cultures by time-lapse imaging after PHF-1 delivery. Within approximately three hours, cells receiving active PHF-1 exhibited dramatic neurite retraction and cell-body rounding—their slender processes withdrew, and the neurons collapsed into rounded shapes. Cultures receiving heat-inactivated PHF-1, empty vehicle liposomes, or other intracellular antibodies showed no such response. This specificity rules out a nonspecific consequence of liposomal delivery or of antibody accumulation inside cells. Instead, it points to a direct functional consequence of antibody binding: engaging the PHF-1-reactive tau pool apparently disrupts processes that maintain neuronal architecture. The authors are careful about interpretation. They note that the mechanism remains unknown—binding might interfere with physiological interactions between phosphorylated tau and the cytoskeleton, alter tau’s conformation, trafficking, or subcellular organization, or, because antibodies are multivalent, crosslink tau species in ways that perturb their function. Nor do the data imply that phosphorylation at Ser396/404 is itself uniquely required for neurite maintenance; the phenotype implicates the PHF-1-reactive population as a whole.

What the finding does suggest is provocative: a tau phospho-form celebrated as a hallmark of neurodegeneration may also perform a physiological job in healthy or developing neurons, one that becomes dysregulated in disease. This dual identity fits a broader theme in modern proteomics. Proteoforms—distinct versions of a protein generated by post-translational modifications such as phosphorylation, acetylation, truncation, or ubiquitination—can carry opposite biological activities, and recent mass-spectrometry efforts have mapped increasingly detailed tau proteoform landscapes in human brain. But defining what each proteoform actually does in a living cell has lagged far behind cataloguing what it is. Genetic approaches perturb the entire protein, not a single modified species.

CLD offers a way to close that gap. Because it delivers pre-formed, full-length antibodies acutely and broadly across a culture, it converts antibodies from passive imaging reagents into active perturbation tools, allowing researchers to interrogate specific endogenous proteoforms in their native cellular context within hours rather than weeks. In the present study, PHF-1 was initially little more than model cargo; its introduction revealed a robust and previously unappreciated phenotype. The same strategy could now be applied to disease-associated species across a range of neurodegenerative disorders. The authors caution that key questions remain—whether the neurite retraction is reversible, what downstream cytoskeletal changes mediate it, and whether the same response occurs in intact neural circuits in vivo. But the demonstration that a simple cationic lipid can deliver functional antibodies safely into populations of living neurons, and that doing so can expose an unexpected function of an Alzheimer’s-linked tau species, marks CLD as a versatile new window onto the hidden workings of the neuronal interior.

Subject of Research: Cationic liposomal delivery of functional antibodies into living neurons and the role of phosphorylated tau in neurite integrity

Article Title: Safe and efficient intracellular delivery of functional antibodies into cultured neural cells shows a role for Alzheimer’s-relevant tau phosphorylation in cell integrity

Article References: Viola, K. L., Chromy, B. A., Nowar, R., Lambert, M. P., & Klein, W. L. (2026). Safe and efficient intracellular delivery of functional antibodies into cultured neural cells shows a role for Alzheimer’s-relevant tau phosphorylation in cell integrity. Frontiers in Dementia, 5, Article 1737082. https://doi.org/10.3389/frdem.2026.1737082

Image Credits: AI Generated

DOI: 10.3389/frdem.2026.1737082

Keywords: cationic liposomes, intracellular antibody delivery, tau phosphorylation, PHF-1, Alzheimer's disease, neurons, neurite retraction, proteoforms, neurodegeneration, DOPC+, cytoskeleton, primary neuronal cultures

The technical core of the review concerns the kinase-phosphatase equilibrium that governs tau’s phosphorylation state. Tau is a natively unfolded protein that stabilizes microtubules in axons, and its function is tuned by the addition and removal of phosphate groups along its sequence, particularly in the proline-rich region and the microtubule-binding repeat region. When kinases such as glycogen synthase kinase-3 beta, cyclin-dependent kinase 5, microtubule affinity-regulating kinase 4, tau-tubulin kinase 1, tyrosine kinase 2 and cyclin-dependent kinase 3 outpace the phosphatases that oppose them, chiefly protein phosphatase 2A and protein phosphatase 1, hyperphosphorylated tau detaches from microtubules, misfolds and begins to self-assemble. The review emphasizes that this equilibrium is not a simple seesaw: individual kinases act at distinct sites and disease stages, which helps explain why broad kinase inhibition has proven difficult to translate into safe, effective therapies.

Equally important is the emerging picture of cross-talk among post-translational modifications. Phosphorylation does not act alone. The review details how O-GlcNAcylation, acetylation, ubiquitination, methylation, glycation and truncation by enzymes such as asparagine endopeptidase compete for overlapping residues and reshape tau’s conformational landscape. O-GlcNAcylation at serine and threonine residues, for instance, can sterically block phosphorylation at neighboring sites and appears protective in several models, suggesting that enhancing this sugar modification could be a viable strategy. Conversely, truncation of tau can expose aggregation-prone segments and accelerate the transition from soluble tau assemblies into paired helical filaments and straight filaments, the structural components of neurofibrillary tangles that cryo-electron microscopy has now resolved to near-atomic precision in patient brain tissue.

The review also highlights two systemic factors that have rarely been integrated into tau-centric models: nuclear envelope damage and iron metabolism. Hyperphosphorylated tau compromises the nuclear envelope and disrupts nuclear pore complexes, and interactions with the lamin B receptor contribute to DNA damage and transcriptional dysregulation in vulnerable neurons. Meanwhile, disturbed iron homeostasis pushes tau pathology toward ferroptosis, an iron-dependent form of cell death driven by lipid peroxidation and reactive oxygen species. Advanced glycation end products further stabilize pathological tau species, creating a feed-forward loop in which oxidative stress, metal imbalance and protein misfolding reinforce one another. These mechanisms position tau as a node where metabolic stress and proteostatic failure converge, rather than a standalone culprit.

Perhaps the most conceptually ambitious section of the review extends tau pathology beyond neuronal boundaries. Pathological tau does not stay confined to the neurons that produce it. The authors describe how soluble tau species spread through tunneling nanotubes, extracellular vesicles and heparan sulfate proteoglycan-mediated uptake, seeding aggregation in recipient cells and propagating along anatomically connected networks, a pattern that resting-state functional MRI studies have begun to visualize in living patients. The low-density lipoprotein receptor-related protein 1 emerges as a key mediator of tau uptake and clearance, offering a potential point of therapeutic intervention. This network-level spread helps explain why tau burden follows predictable, stage-specific routes through the entorhinal cortex, hippocampus and association cortices rather than appearing randomly.

The review then situates tau within the broader ecosystem of Alzheimer’s disease, where it synergizes with amyloid-beta deposition to ignite neuroinflammation. Activated microglia and astrocytes release inflammatory mediators that both promote tau phosphorylation and impair its clearance. The blood-brain barrier, whose breakdown is increasingly recognized as an early event in the disease, permits peripheral immune molecules and plasma proteins to infiltrate the brain, further amplifying the inflammatory cascade. Strikingly, the authors extend this systems view to the gut-brain axis, citing evidence that gut dysbiosis and lipopolysaccharide-driven inflammation can modulate tau pathology through circulating immune signals. In this framework, tau phosphorylation behaves like a barometer of whole-body physiological state rather than an isolated intracellular defect.

On the translational front, the review surveys a new generation of biofluid biomarkers that are transforming how tau pathology is detected and tracked. Cerebrospinal fluid measures of phosphorylated tau have long served as diagnostic anchors, but plasma assays based on single-molecule array technology now achieve comparable diagnostic performance, opening the door to population-scale screening. Brain-derived tau, identified through immunoprecipitation-mass spectrometry, offers improved specificity for brain-derived over peripheral tau, while assays targeting the microtubule-binding repeat region may distinguish Alzheimer’s disease from other tauopathies, including chronic traumatic encephalopathy. The authors also point to emerging technologies such as organic electrochemical transistors that could enable point-of-care tau testing, potentially pairing fluid biomarkers with PET-based standardized uptake value ratios and resting-state functional MRI to stage disease with unprecedented resolution.

Therapeutically, the review maps a pipeline that extends well beyond the first generation of anti-tau antibodies. Monoclonal antibodies targeting specific phospho-epitopes are progressing through clinical trials, while antisense oligonucleotides aim to reduce tau production at the source. Small-molecule kinase inhibitors continue to be refined with attention to isoform selectivity. The most conceptually novel entry is a class of molecules the authors describe as dephosphorylation-targeting chimeras, or DEPTACs, which recruit phosphatases to phosphorylated tau in a manner analogous to the phosphorylation-targeting chimeras that recruit kinases. Related strategies include tauopathy-homing nanoassemblies designed to deliver payloads selectively to tau-bearing tissue. The review is careful to note that many of these modalities remain in early clinical investigation, and that target engagement, dosing windows and patient selection remain unresolved challenges.

The strategic implication of the review is that Alzheimer’s disease drug development may need to abandon the search for a single decisive target in favor of precision approaches aimed at the tau interactome. Because tau phosphorylation sits at the intersection of kinase signaling, proteostasis, nuclear integrity, metal metabolism, neuroinflammation and barrier function, interventions will likely need to be matched to the dominant pathogenic mechanisms operating in each patient, guided by fluid and imaging biomarkers that can stage both molecular pathology and network dysfunction. The authors frame their integrated model as a roadmap for such stratified trials, in which biomarker-defined subgroups receive mechanism-matched therapies.

For a field scarred by repeated late-stage failures, the reframing offered by this review is both sobering and energizing. It acknowledges that amyloid-centric therapies, including the recently approved antibody treatments, address only one strand of a tangled web, and that tau, with its stronger correlation to cognitive deterioration and synaptic dysfunction, offers a more direct line to the symptoms that matter to patients. Whether the multi-scale network view of tau phosphorylation can be converted into therapies that genuinely alter the course of Alzheimer’s disease will depend on the clinical performance of the biomarker and therapeutic platforms now moving through the pipeline, but the conceptual groundwork for that effort has, with this synthesis, been laid out in unusually complete detail.

Subject of Research: Tau phosphorylation mechanisms, network spread and therapeutic targeting in Alzheimer's disease

Article Title: Tau phosphorylation in Alzheimer’s disease: emerging mechanisms, network interactions, and therapeutic implications

Article References: Li, X., Zhang, L., Zhao, L., Qin, L., Bai, Q., Zeng, C., Gao, Z., Ni, J., Zeng, G., Saidova, A., Qing, H., & Li, H. (2026). Tau phosphorylation in Alzheimer’s disease: emerging mechanisms, network interactions, and therapeutic implications. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-08889-2

Image Credits: AI Generated

DOI: 10.1186/s12967-026-08889-2

Keywords: Alzheimer's disease, tau phosphorylation, neurofibrillary tangles, kinase-phosphatase equilibrium, post-translational modifications, neuroinflammation, blood-brain barrier, gut-brain axis, biomarkers, DEPTACs, antisense oligonucleotides, precision medicine

The research team, led by investigators at Nanjing Drum Tower Hospital affiliated with Nanjing University Medical School, took an unusually comprehensive approach, weaving together human proteomics, a prospective surgical cohort, and mechanistic experiments in mice and cultured neurons. Their starting point was the Alzheimer’s Disease Neuroimaging Initiative (ADNI), a large longitudinal study in which cerebrospinal fluid from thousands of participants has been profiled using the SomaScan 7K platform, an aptamer-based technology capable of measuring more than 7,000 molecular analytes simultaneously. From this resource, the researchers defined an exploratory phenotype they call neurodegeneration-enriched cognitive vulnerability, or NECV: individuals with mild cognitive impairment who also had a documented history of surgery. By comparing the proteomic fingerprints of 152 NECV participants against 85 cognitively normal controls, and applying stringent Bonferroni correction across more than 6,000 unique proteins, they found that 14-3-3γ ranked among the top candidates, occupying a central position in the co-expression network of disease-relevant proteins.

The signal held up under scrutiny. Cerebrospinal fluid levels of 14-3-3γ were dramatically elevated in the NECV group compared with controls, a difference that reached an extraordinary statistical significance of p = 7.8 × 10⁻³². The finding replicated independently in the Parkinson’s Progression Markers Initiative (PPMI) cohort and was further confirmed in autopsy-verified Alzheimer’s disease samples. Crucially, the protein was not merely a static marker. In longitudinal survival analyses, participants with higher baseline CSF 14-3-3γ faced an increased hazard of progressing from normal cognition to prodromal Alzheimer’s disease, and those already in the vulnerable group progressed more rapidly to frank dementia. Elevated 14-3-3γ also tracked with the classic molecular hallmarks of neurodegeneration: lower CSF amyloid-beta 42, higher phosphorylated and total tau, greater cortical amyloid burden on PET imaging, smaller hippocampal volumes, reduced cerebral glucose metabolism, and steeper declines in memory, executive function, and global cognition over time.

What elevates this work beyond biomarker correlation is its translation into the surgical setting. The team prospectively enrolled 213 patients aged 65 and older undergoing major elective cardiac, orthopedic, or aortic surgery at their tertiary center, a study registered with the Chinese Clinical Trial Registry. Blood drawn on the first postoperative day, before any delirium had been documented, revealed that patients who subsequently developed postoperative delirium or POCD had significantly higher circulating 14-3-3γ levels than those who recovered uneventfully. In this cohort, 22 patients (10.3 percent) developed delirium within the first week, while 11 (5.2 percent) met criteria for POCD at day seven, all consistent with delayed neurocognitive recovery under the 2018 nomenclature consensus. When plasma 14-3-3γ was added to models built from routine clinical variables, predictive performance improved substantially: the area under the receiver operating characteristic curve for delirium rose from 0.819 to 0.921, and for POCD from 0.705 to 0.864, with statistically significant gains confirmed by DeLong tests. Decision curve and precision-recall analyses suggested the combined models offered genuine net clinical benefit, though the authors are careful to frame these findings as exploratory given the modest number of outcome events.

One of the most intriguing puzzles the study surfaces is a striking compartmental divergence. After surgery, 14-3-3γ levels fell markedly inside hippocampal neurons in mice, yet rose in the blood and cerebrospinal fluid of both mice and humans. The researchers propose that the extracellular elevation reflects a spillover phenomenon, a leakage of intracellular proteins from injured neurons, a pattern long recognized in conditions such as Creutzfeldt-Jakob disease, stroke, and traumatic brain injury, where CSF 14-3-3 proteins serve as markers of neuronal damage. The intracellular depletion, by contrast, may be the pathologically decisive event. The 14-3-3 family constitutes roughly one percent of total soluble brain protein, and its loss would destabilize a web of protein-protein interactions governing tau phosphorylation, mitochondrial function, and synaptic maintenance. In other words, the protein leaking into the bloodstream may be a useful warning signal, but the deficit left behind inside vulnerable neurons is what drives the damage.

To establish causality, the team turned to a murine model of perioperative brain injury in which mice undergo tibial fracture with intramedullary fixation under isoflurane anesthesia, a procedure that recapitulates key clinical features of major orthopedic surgery, including tissue trauma, inflammation, and postoperative pain. Within days, surgically treated mice showed impaired performance in the Y-maze spontaneous alternation test and contextual fear conditioning, both hippocampus-dependent tasks, while locomotor activity and cued memory remained intact, pointing to a specific rather than global deficit. Electron microscopy revealed fragmented microtubules, the intracellular highways that tau normally stabilizes. Golgi staining showed a loss of mature dendritic spines, the tiny protrusions where synapses form. Levels of the synaptic proteins PSD95 and synaptophysin dropped, synaptic clefts widened, and postsynaptic densities thinned. Biochemically, the surgery produced a selective and sustained hyperphosphorylation of tau at the threonine 205 residue, detectable and pronounced at postoperative day seven, while other phosphorylation sites examined, including Thr231, Ser396, and Ser262, remained unchanged at that timepoint.

The mechanistic thread connecting 14-3-3γ loss to tau damage runs through the enzymes that control tau’s phosphorylation state. When the researchers silenced 14-3-3γ in HT22 hippocampal cells, they observed activation of CDK5, reflected in an increased p25-to-p35 ratio; activation of GSK3β, seen as reduced inhibitory phosphorylation at Ser9; and inhibition of the phosphatase PP2A, marked by increased phosphorylation at Tyr307. The net effect was a shift toward kinase-dominated regulation of tau, and tau Thr205 phosphorylation climbed accordingly. Pharmacologically damping CDK5 with roscovitine, blocking GSK3β with lithium chloride, or activating PP2A with DT061 each reversed the hyperphosphorylation. Co-immunoprecipitation experiments added a structural dimension: loss of 14-3-3γ reduced the association of phosphorylated tau with the protective phosphatase PP2A while increasing its association with the kinases CDK5 and GSK3β. Immunoprecipitation followed by mass spectrometry independently identified tau as a binding partner of 14-3-3γ, and the interaction proved sensitive to the Thr205 phosphorylation state, with a phospho-mimetic T205E tau mutant binding more strongly and a phospho-deficient T205A mutant binding more weakly.

The therapeutic implications emerged from two elegant intervention experiments. First, when the researchers delivered a gene therapy vector carrying 14-3-3γ directly into the hippocampal CA1 region and dentate gyrus of mice via adeno-associated virus, the overexpression suppressed surgery-induced tau Thr205 hyperphosphorylation, restored microtubule integrity and synaptic architecture, recovered dendritic spine density, and rescued working memory and contextual fear memory. Second, and perhaps most provocatively, the team tested fusicoccin-A, a small molecule known to stabilize 14-3-3 protein-protein interactions. Administered intraperitoneally before surgery, fusicoccin-A enhanced the binding between 14-3-3γ and phosphorylated tau, restored their colocalization in hippocampal neurons, rescued synaptic protein levels, reduced tau Thr205 phosphorylation without altering total tau, and improved cognitive performance in the Y-maze and fear conditioning tests. Molecular dynamics simulations running 500 nanoseconds per system showed that fusicoccin-A binding shortened the center-of-mass distance between the two proteins, reduced structural fluctuation, and rendered the binding free energy more favorable, providing a physical basis for the stabilization. Critically, when the researchers expressed a phosphomimetic TauT205E mutant in the hippocampus, the mutant alone was sufficient to impair cognition, and it abolished the protective effect of fusicoccin-A, demonstrating that tau Thr205 phosphorylation sits downstream in the protective pathway.

The authors are candid about the limitations that temper these exciting results. The ADNI-based discovery analysis is hypothesis-generating rather than direct evidence about POCD, since that cohort lacks detailed perioperative data. The prospective surgical cohort, though consecutively enrolled, produced few outcome events, limiting the robustness of the prediction models. The mouse experiments used young adult male animals rather than the aged populations most at risk clinically, and only a single small-molecule modulator was tested, leaving dose-response relationships, chronic effects, and safety profiles unexplored. Species differences in 14-3-3 regulation and tau dynamics may also affect translational relevance. Nevertheless, the study represents a genuine conceptual advance: it elevates 14-3-3γ from a passive correlate of neurodegeneration to an active, druggable regulator of acute cognitive vulnerability, and it points toward a strategy for protecting the surgical brain that is fundamentally different from conventional anti-amyloid or anti-tau therapies. If larger, multicenter trials confirm that a simple postoperative blood test for 14-3-3γ can flag patients at risk, and if stabilizers of the 14-3-3γ-tau interaction can be developed into safe drugs, the era of passively watching surgical patients lose their cognitive edge may finally give way to one of prediction and prevention.

Subject of Research: The role of 14-3-3γ in regulating tau Thr205 phosphorylation and synaptic integrity in postoperative cognitive dysfunction

Article Title: 14‐3‐3γ Protects Against Postoperative Cognitive Dysfunction by Regulating Tau Thr205 Phosphorylation and Synaptic Integrity

Article References: Zhu, S., Han, X., Zhang, H., Ye, C., Xia, T., & Gu, X. (2026). 14‐3‐3γ Protects Against Postoperative Cognitive Dysfunction by Regulating Tau Thr205 Phosphorylation and Synaptic Integrity. Advanced Science, Article e77480. https://doi.org/10.1002/advs.77480

Image Credits: AI Generated

DOI: 10.1002/advs.77480

Keywords: postoperative cognitive dysfunction, 14-3-3γ, YWHAG, tau phosphorylation, Thr205, synaptic integrity, biomarker, postoperative delirium, fusicoccin-A, CSF proteomics, hippocampus, neurodegeneration

A new study published in the Journal of Advanced Research offers a strikingly different approach to attacking tau at its upstream source. Rather than blocking the activity of one of tau’s key dysregulating kinases, the research team designed a small molecule that eliminates the kinase protein itself. The target is death-associated protein kinase 1, or DAPK1, an enzyme that has accumulated an impressive pathogenic dossier in Alzheimer’s research. DAPK1 activates MARK1/2 to trigger abnormal tau phosphorylation, directly phosphorylates tau at disease-relevant residues, and through phosphorylation of SENP1 promotes tau SUMOylation, exacerbating tau pathology and cognitive dysfunction. Genetic knockdown and pharmacological inhibition of DAPK1 have both been shown to reduce tau phosphorylation and mitigate neurodegeneration in cell and animal models. Yet traditional small-molecule kinase inhibitors suffer from well-known drawbacks: short-lived efficacy, compensatory signaling from parallel pathways, and substantial off-target liability.

To overcome these limitations, the team turned to targeted protein degradation, a next-generation pharmacological strategy that hijacks the cell’s own proteostasis machinery. Their specific platform is hydrophobic tagging, or HyT, in which a bifunctional molecule couples a ligand that binds the protein of interest to a bulky hydrophobic moiety, typically adamantane, that mimics the exposed greasy surfaces of misfolded proteins. The cell’s quality-control machinery is fooled into ubiquitinating the tagged protein and destroying it via the ubiquitin-proteasome system. Adamantane-based hydrophobic tags carry distinct advantages over the better-known PROTAC degraders: simpler molecular design, lower molecular weight, favorable ADME characteristics, and high degradation efficiency. The approach has already been used to degrade tau, TDP-43, and mutant huntingtin, but it had never been applied to DAPK1.

From a library of candidate degraders built by conjugating a previously characterized DAPK1-binding ligand to adamantane- or fluorene-based hydrophobic tags through alkylamine linkers of varying length, one compound emerged as the clear leader. Named CJ1, the molecule reduced DAPK1 protein levels in mouse primary cortical neurons to roughly half of control levels at a screening concentration of 1 micromolar, outperforming all other compounds in the series. Molecular docking suggested that CJ1’s DAPK1-binding moiety occupies a distinct hydrophobic pocket on the kinase surface, with favorable electrostatic complementarity at the binding interface. Critically, cytotoxicity testing showed that CJ1 did not compromise the viability of primary neurons even at concentrations up to 20 micromolar, providing a wide margin between functional activity and cellular harm.

The mechanistic characterization of CJ1 was rigorous. In primary cortical neurons, the compound degraded DAPK1 in a dose-dependent fashion, with a half-maximal degradation concentration of approximately 0.18 micromolar, and a time course showing detectable degradation beginning four hours after treatment and sustained suppression lasting up to 48 hours. Quantitative PCR confirmed that DAPK1 mRNA levels were unchanged, indicating a posttranslational mechanism. When cells were pretreated with the proteasome inhibitor MG-132, CJ1-induced degradation was abolished, and ubiquitination assays demonstrated markedly increased polyubiquitinated DAPK1 in the presence of the compound. Together, these experiments established that CJ1 marks DAPK1 for ubiquitin-dependent proteasomal destruction, precisely the event-driven behavior expected of a hydrophobic tagging degrader.

With the mechanism confirmed, the investigators asked whether eliminating DAPK1 would tame tau. In primary cortical neurons, SH-SY5Y neuroblastoma cells, and SH-SY5Y cells expressing the P301L mutant human tau, CJ1 treatment reduced levels of tau phosphorylated at threonine 231, serine 262, and serine 396, along with total tau. These phosphorylation sites are deeply meaningful in Alzheimer’s biology: phosphorylation at Ser262 within the microtubule-binding repeat domain critically weakens tau’s grip on microtubules, Thr231 primes tau for further phosphorylation by GSK3beta and contributes to microtubule destabilization, and the Ser396/404 epitope marks advanced pathological tau. Several control experiments reinforced the specificity of the effect. CJ1 did not alter the levels of the major tau kinases GSK3beta and CDK5, the phosphatase PP2A, or the related kinase DAPK3. In DAPK1 knockout neurons, CJ1 lost its ability to reduce tau or phospho-tau, and a control compound lacking the DAPK1-binding moiety was inert. Overexpressing DAPK1 in cells also rescued the phenotype, strongly arguing that CJ1 acts specifically through DAPK1 degradation rather than through incidental off-target effects.

The in vivo results were the most consequential. A central hurdle for any Alzheimer’s therapeutic is crossing the blood-brain barrier, and liquid chromatography-mass spectrometry confirmed that intact CJ1 was detectable in mouse brain after systemic administration. In a tauopathy model created by injecting an adeno-associated virus carrying human P301L tau into the hippocampal CA3 region, DAPK1 was upregulated and CJ1 treatment lowered it in a dose-dependent manner, with 10 milligrams per kilogram nearly restoring baseline levels. CJ1 reduced human tau accumulation and phosphorylation at Thr231 and the AT8 epitope in the hippocampus. In aged hTau transgenic mice, which express all six human tau isoforms and develop genuine neurofibrillary tangles, five weeks of CJ1 treatment significantly reduced tangle burden in the cortex and lowered total tau and phospho-tau at multiple AD-relevant sites in both sarkosyl-soluble and, importantly, sarkosyl-insoluble fractions, the biochemical signature of aggregated, pathological tau.

Pathology translated into preserved brain structure and function. CJ1-treated tauopathy mice retained more NeuN-positive neurons, showed restored MAP2 immunoreactivity indicating healthier dendrites, and displayed dose-dependent recovery of dendritic spine density on Golgi-Cox staining. Neuroinflammation also receded: astrocytic activation marked by GFAP and microglial activation marked by Iba1 were both suppressed, pro-inflammatory iNOS decreased, and anti-inflammatory Arg-1 increased. Functionally, in the Morris water maze, CJ1-treated mice learned the platform location faster, spent more time in the target quadrant during the probe trial, and crossed the former platform location more often, with all groups swimming at comparable speeds, ruling out motor confounds. Notably, only the higher dose produced cognitive improvement, suggesting that partial biochemical correction of tau pathology is insufficient and that robust suppression is required for behavioral benefit.

Safety data were encouraging. CJ1 showed negligible hemolysis at concentrations up to 800 micrograms per milliliter, serum creatinine, ALT, and AST remained within normal ranges, and histological examination of the heart, liver, lung, and kidney revealed no treatment-related abnormalities. The authors are candid about limitations: the behavioral findings derive from an acute, region-specific model; the selectivity screen covered only a small panel of proteins; and the detailed brain distribution of CJ1 remains unmapped, as do potential differential effects on 3R versus 4R tau isoforms. Nonetheless, the study establishes CJ1 as a first-in-class DAPK1 degrader and provides a compelling proof of concept that hydrophobic tagging can eliminate an upstream driver of tau pathology inside cells, something neither conventional kinase inhibitors, which transiently block catalytic activity, nor tau immunotherapies, which target extracellular aggregates, can accomplish. By abolishing all of DAPK1’s functional domains and scaffolding roles in a catalytic, substoichiometric manner, this degrader strategy offers a mechanistically distinct and potentially more durable route to slowing the neurodegeneration that lies at the heart of Alzheimer’s disease.

Subject of Research: A hydrophobic tagging degrader that selectively eliminates DAPK1 to attenuate tau pathology in Alzheimer's disease

Article Title: Selective degradation of DAPK1 via a novel hydrophobic tagging attenuates tau pathology in Alzheimer’s disease

Article References: Li, R., Wu, X., Yao, J., Chen, J., Shui, X., Zheng, X., Tian, W., Wang, L., Zhou, Y., Zhang, T., Chen, D., Liu, Y., & Lee, T. H. (2026). Selective degradation of DAPK1 via a novel hydrophobic tagging attenuates tau pathology in Alzheimer’s disease. Journal of Advanced Research, 87, 1027-1043. https://doi.org/10.1016/j.jare.2025.12.037

Image Credits: AI Generated

DOI: 10.1016/j.jare.2025.12.037

Keywords: Alzheimer's disease, DAPK1, tau pathology, hydrophobic tagging, targeted protein degradation, neurofibrillary tangles, proteasome, tau phosphorylation, blood-brain barrier, neurodegeneration, drug discovery, tauopathy