Science
The mechanisms underneath the headlines.
Lecanemab and donanemab clear amyloid and slow early Alzheimer's decline by a fraction of a point on an 18-point scale. Here is what that fraction looks like from a waiting room, and what the MRI schedule is for.
Assays for p-tau217 are now FDA-cleared and sitting in clinic order sets. The evidence says they are very good at spotting brain amyloid in people with symptoms, and says much less about everything else a worried family wants to know.
Mice with more of a secreted brain enzyme called PLA2G2F kept more synapses, calmed more inflammation and performed better on memory tests as they aged. In human brain tissue examined after death, the same enzyme tracked with people who stayed sharp despite amyloid plaques.
A scoping review in Frontiers in Dementia went looking for every study in the past decade that fed magnetoencephalography data to machine learning for dementia risk. The count came back at fourteen. Here is what those papers reach, and where they stop.
P3 was the leftover nobody studied, the scrap from the safe cut of amyloid chemistry. A UC Santa Cruz biochemist says the test tube evidence always said otherwise, and that the amyloid cascade needs a bigger cast list.
An 863-person imaging study published in Nature reports that the substantia nigra and all three major deep brain stimulation targets wire preferentially into a network that fuses thinking, body state and action, skipping the cortical strips that move a hand or a foot
Scientists took tau filaments from the brains of people who had died with Alzheimer's disease or corticobasal degeneration, injected them into ordinary mice, and watched the new tau that grew there copy the original's atomic shape down to the fold. It is the first direct structural check on a long-standing idea: that these misfolded proteins spread the way a stamp spreads its pattern, not the way a rumor changes with each retelling.
A phase 3 trial of a once-daily tablet built specifically for people who carry two copies of the APOE4 gene, the highest-risk and most predictable form of Alzheimer's, did not slow cognitive decline overall, but a prespecified subgroup with mild cognitive impairment showed a statistically nominal benefit, researchers reported.
Wave Life Sciences has stopped developing WVE-004, an antisense drug built to silence the C9orf72 gene mutation behind both frontotemporal dementia and ALS, after a mid-stage trial found no clinical benefit despite clear biological activity.
A 32-page review in Cell takes the last cycle's scattered discoveries, on amyloid, tau, microglia and APOE, and sets them in causal order. It is a synthesis, not new data, and most of the machinery it describes was traced in mice and dishes.
A repeat survey of North American movement disorder clinicians finds barriers down and referrals up since the PD GENEration program began giving away panel testing. What it cannot show is that the program caused the change.
A Nature Communications study in mice finds that a chemical tag on tau, a protein best known for forming Alzheimer's tangles, has a normal job: helping recruit the specific neurons that store a memory and helping those memories survive over weeks, not just days.
A test-tube and worm study reports that tau fibrils behave like catalysts for amyloid-beta, and that the shape of the tau fold decides how strong the effect is. It is bench work, not people.
A Nature Neuroscience study reports that in older adults with tau in the frontal cortex, the big slow waves of deep sleep stay put instead of sweeping across the brain, and that the stalling tracks with weaker overnight memory. It is an association, not a diagnosis.
A new preprint from a Berkeley structural biology lab used cryo-electron tomography on cultured neurons and astrocytes and found no evidence that tau fibrils tear their way out of lysosomes. The compartments stayed intact. The leak, the authors argue, is something subtler. This is laboratory cell work, not peer reviewed, and not a treatment.
A laboratory study in Communications Biology reports that disease-linked chemical changes expose a normally hidden sticky region of alpha-synuclein, letting it grab other proteins indiscriminately, a mechanism the authors call a 'Multifactorial Random Disorder Model.'
A Drosophila study published in Cell Death & Disease reports that a mutation in CHCHD10, a gene linked to frontotemporal dementia, releases a brake on an endo-lysosomal protein called Rab7A and makes tau pathology worse in flies. The mechanism differs from other lysosomal pathways already described for FTD, and the findings have not yet been shown in humans.
A new mathematical model simulates how amyloid-beta and blood flow interact across the whole brain, suggesting the two feed each other in a self-reinforcing loop that can tip a region toward runaway pathology once a large enough seed of protein takes hold.
A phase 1 study delivered a single viral vector carrying both tyrosine hydroxylase and AADC genes into the striatum, aiming for cells that manufacture dopamine on their own rather than depending on oral levodopa. Twelve-month safety data are now published; the trial was not designed to prove the therapy works.
A January 2026 review in the Journal of the American College of Cardiology lays out the biology that vascular dementia therapies are aiming at, and is candid that most of the aiming has not yet hit anything
The World Stroke Organization's first vascular dementia fact sheet, published in the International Journal of Stroke, pools autopsy studies, exposes how wildly diagnostic criteria disagree, and places small vessel damage in the causal chain. It is a map of what is known, and of how much is not.
A study of patients with mild cognitive impairment linked to Lewy bodies finds that abnormal brain wave patterns, smaller insula volume, and higher blood tau and GFAP together tracked a shorter dementia-free survival, though the finding needs testing in larger groups before it could guide care.
A laboratory study finds that tau, best known for stabilizing microtubules in neurons and for forming the tangles seen in Alzheimer's disease, also condenses on centromeric DNA and helps mitotic spindles grab hold of chromosomes, a role distinct from its function in mature nerve cells.
A study published March 26 in Cell Reports Medicine ties a malfunctioning form of the enzyme GRK2 to nerve cell death and amyloid buildup, and reports that an experimental compound blocking the clumping slowed disease markers and lengthened survival in mice.
In 756 adults followed at Vanderbilt for up to 11 years, lower baseline cardiac output predicted faster medial temporal change on MRI in carriers of the Alzheimer's risk gene, and not in non-carriers. The study is observational and its authors say it needs replication.
Three years after cryo-EM revealed TAF15 filaments in patient brains, a Nature Communications paper reports that brain-derived TAF15 aggregates can template new aggregates in cultured cells, and pins the behaviour on a handful of short sequence motifs. All of the work is in test tubes and cells, not in people.
A Nature report on 38 donated hippocampi maps three stages of neurogenesis across young adults, healthy agers, SuperAgers, preclinical pathology and Alzheimer's. It is cross-sectional tissue work, not proof that new neurons make memory.
A reanalysis of more than 1,600 postmortem RNA sequencing samples reports that the aberrant splicing left behind by TDP-43 concentrates in oligodendrocytes in ALS and in neurons in FTD, a distinction with consequences for biomarkers and drug targets.
In post-mortem tissue from 30 people with ALS, 20 with ALS-FTD and 32 controls, an Ulm and DZNE team sorted neuronal nuclei by whether TDP-43 had left the nucleus, then read the genes and chromatin of the same cells. Five excitatory subtypes stood out.
A Stanford preprint traces frontotemporal dementia's biggest gene and its biggest modifier to a single measurable lesion: a TMEM106B fragment piling up inside neurons' waste-disposal compartments. The work is in mice, stem cell derived neurons and donated brain tissue, and it has not been peer reviewed.
A preprint posted on 6 March 2026, not yet peer reviewed, followed 450 people with Parkinson's and no dementia for a mean of 32.7 months. Baseline visual performance separated those who declined; APOE4 on its own did not, and the GBA1 signal rested on 24 carriers.
A preprint posted on 28 July 2026, not yet peer reviewed, clusters 225 newly diagnosed patients on 22 fluid markers. The groups differ in biology but not in how the disease looked or progressed over five years.
A Nature Medicine study of 464 people in the UK and Italy finds that the gut bacteria of GBA1 variant carriers without symptoms sit halfway between healthy controls and people with Parkinson's. It is a snapshot, not a prophecy, but it is the clearest map yet of a question nobody has answered: why only about one carrier in five ever converts.
A hypothesis-generating platform called XunZi flagged CHK2 and IRAK4 as candidate targets across Parkinson's disease models. Blocking Chk2 protected dopamine neurons and improved movement in mice. No human testing has been done.
Bench work with scrapie-positive sheep tissue finds the inhibition comes from damage to the assay's own protein substrate, not destruction of the prion seeds, and that whole blood is one of the few sample types still near inhibitory levels after a 1,000-fold dilution.
Laboratory work from the MRC Prion Unit at UCL, published in PNAS on 11 June 2026, reports a dividing cell system that propagates variant Creutzfeldt-Jakob disease prions from patient brain while remaining refractory to sporadic CJD isolates. No patients were treated or tested.
A UCL team reports a dividing human-prion cell line, EKV, that keeps sporadic CJD prions growing indefinitely and measures how much infectivity a sample carries in weeks rather than years. It works only for some patients' prions, and it is laboratory work, not treatment.
A post mortem analysis of 1,938 older adults, published in Brain, maps cerebral amyloid angiopathy against age, APOE genotype and Alzheimer's pathology. The tightest link to plaques and tangles turned up where researchers least expected it: in APOE e2 carriers without Alzheimer's disease.
A purpose-built mouse carrying amyloid, tau and alpha-synuclein together shows amyloid amplifying the other two, but only when the plaques got there first. This is animal work, published 3 July 2026 in Alzheimer's & Dementia.
A Mayo Clinic team read gene activity spot by spot across the temporal cortex of ten postmortem brains. Layer 5 carried the most Lewy bodies and the highest SNCA expression, but several disturbed signals sat where the aggregates did not.