Welcome back to Cereal Bio, where this week biology spent its time blurring every line it could find: alive versus dead, old versus young, human expert versus machine. We've got brains on life support pulling lab shifts, a longevity trial that finally left the petri dish, and gene editing that treats your DNA like a typo to fix. Then we're heading to the lake, because Juhannus is upon us and even biotech writers need a sauna. Let's get into it before the bonfires start.
Table of Contents
🧠 Disembodied human brains are doing your drug trials now
NEWS
Here is a sentence that should not be true in 2026: a Connecticut startup is pulling human brains out of recently deceased donors, plugging them into a life-support machine, and using them to test drugs. The company is Bexorg, the machine is BrainEX, and according to its own marketing, the organs "hover between life and death," which is either the most honest or the most unsettling tagline in biotech this year.
Bexorg was founded in 2021 at Yale by Zvonimir Vrselja and Nenad Sestan, the same lab that stunned everyone in 2019 by keeping decapitated pig brains alive for 36 hours. BrainEX pumps synthetic blood, oxygen, and nutrients through the brain's vasculature, keeping cells active for about a day. There's no consciousness, the company insists, and to be sure, the brains are almost devoid of coordinated neural firing and dosed with the anesthetic propofol to suppress stray electrical activity. The fact that you have to anesthetize something to be safe is, of course, not entirely reassuring.
The pitch is realism. CEO Vrselja told Science, "You get cells that have been there for 60 to 80 years," complete with decades of drugs, environment, and disease. Pitt neurodegeneration researcher Bruna Bellaver called it "a huge step up from mouse models," and she has a point: Biohaven tested a Parkinson's drug that did nothing in mice but worked in the human brains at one-twentieth the expected dose, data that has since helped push a drug into a Phase 1 trial now dosing people. After 24 hours, each brain is sliced into hundreds of pieces, with plans to process up to 1,600 a year. Bexorg hasn't published its human work yet, so for now, the biggest data point is the one giving everyone the willies.
🍄 Fungi will turn Mars into farmland (the paper does not say that)
RESEARCH
Every few months a headline promises that mushrooms will let us grow potatoes on Mars, and this is that month. The reality, as usual, is quieter than the press. A team from the US and Brazil published a paper in Frontiers in Astronomy and Space Sciences arguing that certain fungi could help future Martian crops. Note the verb. This is a proposal and literature review, not a greenhouse full of Martian tomatoes.
The problem is real enough. Martian regolith is a lousy growing medium: it lacks nitrogen, phosphorus and potassium and is laced with toxic perchlorates, plus a charming dash of aluminum and manganese. The authors suggest mycorrhizal fungi, which strike up symbiotic deals with plant roots, could improve iron uptake, soak up oxidative stress, and build better soil structure. It's plausible because some fungi are absurdly tough. Cryomyces antarcticus has been parked on the outside of the ISS and survived the vacuum, temperature swings, and radiation, which is more than most of us could manage.
The appeal is economic: shipping soil and fertilizer to Mars is the kind of cost that makes accountants weep, so growing your own is the dream. To be fair to the researchers, they are upfront about the holes. We don't know whether the crops would be safe to eat, how they'd react to radiation, or how to even validate the idea ahead of a mission. So: a real starting point, dressed up as a finished farm.
🔬 An AI just beat brain-tumor specialists 2-to-1
RESEARCH
Diagnosing a brain tumor is brutally hard. Many can only be pinned down by DNA methylation profiling, the gold standard, which needs special labs and roughly two weeks. An AI system called Hetairos, built by teams at the German Cancer Research Center and Heidelberg, wants to shortcut that using the cheap, universal thing every pathology lab already has: a stained tissue slide.
In Nature Cancer, the team reports that Hetairos predicts 102 molecular tumor subtypes straight from standard H&E slides and beat five board-certified neuropathologists in a head-to-head, scoring 0.68 to their 0.30. Counting each expert's top three guesses, the AI still led 84% to 50%. It was trained and validated on more than 11,000 slides from 9,606 patients across eleven centers on four continents, and crucially, it reports its own confidence, hitting around 87% accuracy on half to two-thirds of cases it flags as high certainty.
The headline number is speed. In a prospective run, molecular diagnosis took about twelve days while Hetairos returned an answer in twelve minutes on ordinary computer hardware. Before anyone fires their neuropathologist, the caveats matter. The developers frame it as a triage and support tool, not a replacement, and on the rarest subtypes, those with fewer than ten training cases, the human specialists held their own. Still, for clinics without a methylation lab, twelve minutes versus twelve days is the kind of gap that changes who gets diagnosed at all.
⏪ Remember the "age reversal" trial? They just dosed the first human
NEWS
Back in February, we covered the FDA greenlighting Life Biosciences to test a therapy meant to make old cells behave young. The "we'll believe it when it's in a person" part has now happened: the Boston company announced it has treated the first participant in the trial, making this the world's first attempt at cellular reprogramming as a medicine.
The approach is partial reprogramming. Flip on three of the four "Yamanaka" genes, OCT4, SOX2, and KLF4, and you can nudge an aged cell back toward a youthful state without erasing what kind of cell it is, the catch that makes full reprogramming so dangerous. Those four factors are the same ones that, at full blast, can rewind an adult cell all the way to a stem-cell-like blank slate, which is great for growing stem cells and terrible for keeping your retina a retina. The therapy, called ER-100, is injected into the eye to treat a form of glaucoma, with the hope of regrowing neurons in the optic nerve, which normally don't regenerate.
Why the eye first? Because the real worry with reprogramming is that you talk a cell into being young, and it keeps going until it's cancerous, and the eye is a contained place to find out. The proof-of-concept is encouraging but still rodent-grade: in 2020, the same trick regrew optic-nerve neurons and reversed vision loss in mice with glaucoma. One longevity researcher quoted by Nature put the upside and the risk in the same breath, calling the potential for catastrophic side effects high. One patient down, a lot of follow-up to go.
🩸 Scientists edited a cholesterol gene with one shot, maybe for good
RESEARCH
Statins work, but only if you take them, and a depressing 30% to 50% of people quit their cholesterol-lowering drugs within a year. So the obvious move is to make the edit permanent and skip the daily pill entirely. That's the bet behind VERVE-102, a base-editing therapy from Verve Therapeutics (now an Eli Lilly subsidiary), reported in the New England Journal of Medicine.
The target is PCSK9, a gene whose job is to keep LDL receptors from clearing cholesterol. People born with a broken copy have naturally low LDL and fewer heart attacks, so the therapy copies nature and switches it off. The clever bit is the tool: rather than CRISPR's molecular scissors, VERVE-102 uses a base editor packaged in a lipid nanoparticle and delivered as a single roughly four-hour infusion, changing a single DNA letter without cutting the strand.
The phase 1 numbers are quite lovely. Across an interim analysis of 35 participants, LDL cholesterol fell from 9% at the lowest dose to 62% at the highest, an absolute drop of 78 mg/dL, with reductions durable for up to 18 months in those followed longest. PCSK9 protein dropped by up to 88%. Now for some honesty: this is a small, early trial, the longest follow-up so far is on the lowest doses, and long-term safety and real clinical benefit, like actually preventing heart attacks, remain to be proven. "Permanent" is the design goal, not yet a fact. A one-and-done cholesterol fix is closer than ever, but your liver edits are forever, so "measure twice" has never meant more.
If there's a theme this week, it's that biology has stopped respecting its own categories. A brain can be neither alive nor dead and still clock in for work, a cell can be old and then allegedly young, a tumor can be diagnosed by something that isn't a doctor, and your cholesterol gene can be rewritten like a typo. Mars remains stubbornly free of mushroom farms, but give the press releases a few more months.
So which one rattled you most: team disembodied-brain or team please-no? Would you take a cholesterol edit knowing it's forever? Hit reply, and forward this to the friend who still thinks biotech is boring. We're taking a short break after this one, mostly because it's Juhannus and the entire country we live in is about to vanish to a lakeside cabin. Back once the bonfires are out.
Keep questioning everything (and watch the sauna temperature),
P.S. To our Finnish readers: hyvää juhannusta. We know the lake will be a brisk 16°C, and we know you'll get in anyway. We respect it. We will be doing the same, purely for science.