The brain swaps out its own immune cells after 50, four weeks of different food made older adults biologically younger, a blood test for early pancreatic cancer — and four other things this week.

Week of September 21st, 2026

This week's research keeps returning to the same uncomfortable gap: the distance between when something changes in your body and when anyone can see it. We open with a single-cell map of the aging human hippocampus showing that between roughly 50 and 75, the brain quietly swaps out the immune cells it was born with for more inflammatory replacements. From there: a four-week randomized feeding trial in Australians aged 65 to 75 in which three of four diets produced measurably younger biomarker profiles, the molecule that finally explains why APOE4 carriers develop leaky blood-brain barriers, a large and cleanly negative trial showing that opening a narrowed carotid artery does nothing for cognition, an immune sensor that mistakes your own broken DNA for a virus and sabotages the repair crew while it inflames the tissue, the lipid that lean fat tissue makes to kill breast cancer cells and stops making once it becomes obese, and a blood test that caught 86.8% of stage I-II pancreatic cancers. Seven findings, graded honestly. Here's what's worth your attention.

1. Your Brain Starts Replacing Its Own Immune Cells After 50

The brain's immune cells were supposed to be the one population you are born with and keep. A team led by Bing Ren at the New York Genome Center and Columbia, with Nathan Zemke at UC San Diego and Xiangmin Xu at UC Irvine, published a single-cell map of gene regulation and three-dimensional genome organization in the human hippocampus across a wide span of adult ages, in Science. Between roughly ages 50 and 75, microglia that originated during embryonic development declined sharply and were progressively replaced by cells whose molecular features resembled immune cells from the blood, carrying stronger inflammatory signatures than the ones they replaced. The same window showed a substantial drop in the cell populations that maintain the blood-brain barrier, and across several brain cell types, a broad erosion of the genome's three-dimensional folding, the architecture that determines which genes stay reachable and which get switched off. The paper is one of six published in Science out of the NIH's decade-long 4D Nucleome program.

Grade it as a landmark descriptive map rather than anything you act on. This is postmortem human tissue compared across ages, not the same brains followed over time, so "a shift between 50 and 75" is an inference from a cross-section rather than a trajectory anyone watched happen. What it changes is the mental model. Brain aging has been described for decades as gradual attrition; this looks more like coordinated remodeling, with immune cells, blood vessels, neurons, and genome architecture turning over together inside a defined midlife window. If the replacement microglia really are more inflammatory and less competent at housekeeping, that is a plausible mechanistic bridge between "age" and "why age is the dominant risk factor for Alzheimer's." The practical residue is unglamorous and familiar: the levers you actually have over neuroinflammation and vascular integrity in your fifties and sixties are sleep, blood pressure, glucose control, and training.

Zemke NR, Lee S, Mamde S, et al. Epigenetic and 3D genome reprogramming during the aging of the human hippocampus. Science. 2026;393(6809). DOI: 10.1126/science.adt8307. https://doi.org/10.1126/science.adt8307

2. Four Weeks of Different Food Moved Biological Age in People Aged 65 to 75

Most diet-and-aging evidence comes from mice eating one thing for a lifetime, or from humans reporting what they think they ate. The Nutrition for Healthy Living study at the University of Sydney's Charles Perkins Centre did something narrower and cleaner. Caitlin Andrews and Alistair Senior randomized 104 Australians aged 65 to 75, with BMI between 20 and 35, to one of four controlled diets for four weeks. Protein was held constant at 14% of energy in every arm. Two arms were omnivorous, two were semi-vegetarian with 70% of protein from plants; within each pair, one was higher in fat and lower in carbohydrate and the other the reverse. The team then combined 20 biomarkers, including cholesterol, insulin, and C-reactive protein, into a single biological age score. Three of the four groups came out with younger biomarker profiles after four weeks. The omnivorous high-fat arm, which was closest to what participants had been eating already, did not move at all. The strongest statistical signal came from the omnivorous high-carbohydrate arm: 14% protein, 28 to 29% fat, 53% carbohydrate. Published in Aging Cell.

Grade it as a real randomized human trial with a surrogate endpoint and a short leash. The design is the strength: controlled feeding, randomization, and a protein level deliberately held identical across arms so the comparison isolates the fat-versus-carbohydrate and animal-versus-plant axes instead of confounding them with protein. The limits are equally clear. Four weeks, 104 people, one narrow age band, and a biological age score built from blood chemistry, which is a snapshot of physiology rather than a measure of aging itself. A lower calculated biological age is not a reversed aging process, and nobody has shown these shifts persist or translate into fewer diseases, which the authors say plainly. The interesting part is direction rather than magnitude: holding protein fixed, both cutting fat and shifting protein toward plants moved the needle, and the diet that moved it most was the higher-carbohydrate one, which cuts against a decade of low-carb messaging aimed at exactly this age group.

Andrews CJ, Ribeiro RV, Gosby A, et al. Short-Term Dietary Intervention Alters Physiological Profiles Relevant to Ageing. Aging Cell. 2026;25(5). DOI: 10.1111/acel.70507. https://doi.org/10.1111/acel.70507

3. The Leaky Blood-Brain Barrier in APOE4 Carriers Finally Has a Culprit

APOE4 is the strongest common genetic risk factor for late-onset Alzheimer's, and one of its earliest visible effects is a leaky blood-brain barrier, which has been a description rather than a mechanism for years. Caghan Kizil's group at Columbia's Taub Institute, working with Mayo Clinic Florida, Mount Sinai, Bordeaux, and Dresden, named the missing molecule in Nature Aging: fibronectin, made by astrocytes. Across postmortem human brain tissue, three-dimensional human vascular models, and living animals, APOE4 together with amyloid-beta 42 and inflammatory signals drove astrocytes to upregulate fibronectin and deposit it in excess around blood vessels. Fibronectin accumulation on its own was sufficient to cause barrier leakage, and it did so by disrupting VEGF, HB-EGF, and IGF-1 signaling through integrin-mediated focal adhesion kinase activity. Reducing fibronectin, or restoring the growth factor signaling it had disrupted, rescued barrier function both in engineered human vasculature and in vivo.

Grade it as an unusually well-triangulated mechanism with a therapeutic target attached and no therapy yet. It carries more weight than the average pathway paper because the same finding holds across human autopsy tissue, engineered human blood vessels, animal models, and clinical datasets, which is four different ways of being wrong that would all have to fail in the same direction. What it does not do is help an APOE4 carrier this year: there is no fibronectin-lowering drug, and lowering fibronectin in a dish is several orders of difficulty removed from doing it safely in a human brain over decades. Its significance is strategic. It moves the vascular side of Alzheimer's from a correlate to a causal chain with named links, and that is the state a problem has to reach before anyone can drug it.

Bhattarai P, Yilmaz E, Cakir EÖ, et al. Fibronectin mediates APOE4-driven blood-brain barrier dysfunction in Alzheimer's disease. Nature Aging. 2026;6(9):1901-1927. DOI: 10.1038/s43587-026-01204-0. https://doi.org/10.1038/s43587-026-01204-0

4. Opening a Narrowed Carotid Artery Did Not Make Anyone Think Better

A narrowed carotid artery with no symptoms is one of medicine's most tempting problems: the plumbing is visibly wrong, the fix exists, and asymptomatic stenosis is associated with worse cognition. The cognitive results of the CREST-2 trials, published in JAMA by Ronald Lazar at the University of Alabama at Birmingham and James Meschia at Mayo Clinic Jacksonville, tested whether fixing the plumbing fixes the thinking. Across 146 US sites, 2,366 patients aged 35 and older with 70% or greater asymptomatic carotid stenosis were enrolled between December 2014 and July 2024; 2,165 of them (91.5%) completed cognitive testing, mean age 70.1 years, 37.8% women. Two parallel trials compared carotid stenting plus intensive medical management against intensive medical management alone (n=1,078), and carotid endarterectomy plus intensive medical management against medical management alone (n=1,087). Five tests covering learning, memory, executive function, attention, and processing speed were administered by telephone, scored blind to treatment, annually for up to four years. Nothing moved. The adjusted between-group difference in composite z score at four years was 0.03 for stenting (95% CI, -0.09 to 0.15) and 0.02 for endarterectomy (95% CI, -0.09 to 0.14), against a prespecified minimal clinically important difference of 0.50.

Grade it as a high-quality prespecified negative result, which is more useful than it feels. The confidence intervals are tight enough to rule out a clinically meaningful cognitive benefit rather than merely fail to find one, and that distinction separates a trial that settles something from a trial that was underpowered. Two honest caveats. This is about cognition, not stroke, and CREST-2's stroke results are a separate question with a separate answer. And everyone in every arm received intensive medical management, which is itself an active and effective treatment, so the comparison is procedure-plus-good-medicine against good medicine, not against nothing. The practical read for anyone told they have a narrowed carotid without symptoms: the reason to consider a procedure is stroke prevention, discussed with a vascular specialist on that basis alone. It is not going to sharpen your memory, and that hope should not be part of the arithmetic.

Lazar RM, Meschia JF, Edwards LJ, et al. Revascularization of Asymptomatic Carotid Artery Stenosis: Cognitive Results of the CREST-2 Randomized Trials. JAMA. 2026. DOI: 10.1001/jama.2026.14703. https://doi.org/10.1001/jama.2026.14703

5. The Immune Sensor That Mistakes Your Own DNA for a Virus

For thirty years the story about DNA damage and aging has been that the damage itself does the harm. An international team led by Marva Bergman and Itamar Harel at the Hebrew University of Jerusalem, with Bérénice Benayoun at USC, argues in Genes & Development that the body's reaction is doing much of it. In DNA repair disorders such as Ataxia-Telangiectasia and Bloom syndrome, where repair machinery fails and broken DNA accumulates, fragments escape into the cell's cytosol. There they hit cGAS, an immune sensor whose job is detecting viral DNA and which cannot reliably tell your own genetic material from an invader's. The result is chronic sterile inflammation, inflammation with no infection behind it. The team also caught cGAS doing something worse: moving into the nucleus and directly disrupting DNA repair, so the same molecule both inflames the tissue and sabotages the crew sent to fix it. In a short-lived vertebrate model that compresses aging into months, lowering cGAS activity improved neuroinflammation, tissue degeneration, and loss of reproductive capacity.

Grade it as a genuine conceptual shift confined, for now, to rare disease models. The claim that matters is Bergman's: the body can cope with more DNA damage than we assumed, provided the inflammatory response is kept in check. That reframes a class of disorders from unfixable to potentially manageable, because you do not have to repair every lesion if you can quiet the overreaction to them. Three things keep it honest. These are rare genetic syndromes with catastrophic repair failure, not ordinary aging, and the authors explicitly note that reversing disease-driven degeneration is not the same as slowing the rate of aging. The model is a fish, not a person. And cGAS is load-bearing: it is how you detect viruses, so any drug that blunts it has to thread a needle between calming sterile inflammation and leaving you undefended. The reason to care anyway is that chronic inflammation and genomic instability are not rare at all, and a mechanism that ties them together is worth knowing about.

Bergman M, Goshtchevsky U, Atlan T, et al. A dual role for cGAS in shaping cellular and organismal responses to genomic instability. Genes & Development. 2026;40(11-12):852. DOI: 10.1101/gad.352760.125. https://doi.org/10.1101/gad.352760.125

6. Lean Fat Tissue Makes a Molecule That Kills Breast Cancer Cells

Obesity raises breast cancer risk and makes the tumors more aggressive, a link that has been solid epidemiologically and vague mechanistically for decades. Keren Hilgendorf's lab at the University of Utah and the Huntsman Cancer Institute, with Alana Welm, put a molecule in the gap in Science. Healthy mammary adipocytes secrete an oxylipin called 9S-HODE, and 9S-HODE pushes breast cancer cells into ferroptosis, a form of cell death driven by disrupted iron handling. Adipocytes in obese mammary tissue make less of it, and tumors in obese mice showed correspondingly less ferroptosis. The team then ran the logic in both directions: blocking ferroptosis accelerated tumor growth in lean mice, and restoring 9S-HODE suppressed tumor growth in obese mice. In humans, mammary 9S-HODE content was inversely correlated with body mass index, and 9S-HODE inhibited the growth of patient-derived breast cancer organoids.

Grade it as a satisfying mechanistic explanation for a well-known risk, with a therapeutic idea attached and a long road ahead. What makes it more than a mouse story is the shape of the evidence: a molecule identified, its loss mapped onto obesity in human tissue, and the effect reversed in both directions in animals and in human-derived organoids. The reframing is worth sitting with. Fat tissue here is not a passive reservoir that raises estrogen and background inflammation; lean adipose tissue is actively producing a tumor-suppressing signal that obese adipose tissue stops making. That is a loss of protection rather than an addition of risk, and it is a different thing to target. The caveats are standard: this is preclinical, there is no 9S-HODE drug, and nobody should read it as a reason to buy a supplement. The actionable version has not changed and did not need to. Holding onto lean mass and a healthy body composition through midlife is, among other things, a way of keeping your own fat tissue doing this job.

Curtin MC, Jackson AE, Lee MD, et al. Lean adipocyte oxylipin signaling restrains breast cancer through ferroptosis. Science. 2026;393(6816):eaea4287. DOI: 10.1126/science.aea4287. https://doi.org/10.1126/science.aea4287

7. A Blood Test That Caught Five of Six Early Pancreatic Cancers

Pancreatic cancer kills mostly because it is found late. By the time symptoms appear the window for curative surgery has usually closed, and there has never been a blood test good enough to change that. Ajay Goel's group at City of Hope's Beckman Research Institute reports one in Nature Medicine. They built and validated a ten-microRNA blood signature, combined it with the existing tumor marker CA19-9 into a composite score called PANXEON, and tested it in an international, multicenter, prospective study of 1,785 people with and without pancreatic ductal adenocarcinoma across four countries. The microRNA signature alone reached an area under the curve of 88.6% in the testing cohort, with 83.8% sensitivity for early-stage disease and minimal cross-reactivity with other gastrointestinal cancers. Combined with CA19-9, sensitivity for stage I-II disease was 86.8%, with a false-positive rate of 3.2% in low-risk controls and 15.6% in high-risk controls. In people with high-risk pancreatic cysts, the score flagged 64.3% of high-grade dysplasia. In a small group of 19 patients, signature levels fell during neoadjuvant chemotherapy and after surgery, then rose again before the disease came back.

Grade it as a strong biomarker development study that is not yet a screening test, and that distinction is the whole story. Catching 86.8% of stage I-II pancreatic cancers would be a real change in a disease where most diagnoses arrive too late for surgery, and the fact that levels track treatment response and recurrence hints at a second use as a monitoring tool. The number to hold onto is the 15.6% false-positive rate in high-risk controls. Pancreatic cancer is rare enough in the general population that a test with this profile, applied broadly, would generate far more false alarms than cancers found, each one leading to imaging, anxiety, and sometimes a procedure. That is exactly why the authors position it for high-risk groups, meaning people with pancreatic cysts, a strong family history, or new-onset diabetes later in life, and why they call for larger prospective studies rather than declaring victory. This is not something to request at your next physical. It is something to watch, and a reason for anyone with a known high-risk profile to stay inside a surveillance program.

Xu C, Mannucci A, Han H, et al. Liquid biopsy for early detection of pancreatic ductal adenocarcinoma. Nature Medicine. 2026. DOI: 10.1038/s41591-026-04625-x. https://doi.org/10.1038/s41591-026-04625-x

Have a great week,

Dr. Smith and Winston

Medical disclaimer

The content above is for informational and educational purposes only and does not constitute medical, nutritional, or professional advice. It is not intended to diagnose, treat, cure, or prevent any disease and is not a substitute for personalized guidance from a licensed physician or other qualified healthcare provider. Individual needs and risks vary. Do not begin, stop, or modify any exercise regimen, dietary plan, medication (including any change in dietary fat, carbohydrate, or protein intake, any decision about carotid revascularization, or any cancer screening or surveillance schedule), or supplement based on this article alone. Consult your doctor first — particularly if you have or suspect a cardiovascular, metabolic, kidney, gastrointestinal, or neurological condition, take prescription medications, or are pregnant or breastfeeding. Studies summarized here vary in design and quality; preclinical and animal findings often do not translate to humans, and associations reported in observational research do not prove cause and effect.