The endothelial cells lining the blood vessels of your brain are stitched together so tightly that the gaps between them measure roughly 4 nanometers — about the width of a strand of DNA, and small enough to exclude more than 98% of every small-molecule drug ever designed to treat a neurological disease. Glucose slips through. Ketones slip through. Oxygen slips through. The antibody a pharmaceutical company spent a billion dollars engineering to clear amyloid plaques? It bounces off.
That mismatch — a filter so selective it feeds the brain sugar while starving it of medicine — is the single biggest reason neurology has lagged decades behind oncology and cardiology. And in March 2026, for the first time, the FDA approved a biologic specifically engineered to sneak past it.
A wall that isn’t a wall
The blood-brain barrier is not a membrane in the anatomical sense. It is a property of the endothelial cells that line the capillaries threading through the human brain. In every other organ, the cells lining blood vessels have small gaps between them — fenestrations — that let fluid, nutrients, and drugs leak out into the surrounding tissue. In the brain, those gaps are sealed shut.
The seals are protein complexes called tight junctions, made of occludin, claudins, and junctional adhesion molecules. They fuse adjacent endothelial cells so completely that the paracellular space between them narrows to roughly 4 nanometers. For reference: a water molecule is about 0.3 nanometers across. A glucose molecule, about 0.9. A typical antibody, around 10 to 15. The math is unforgiving.
Peter Searson, a professor at Johns Hopkins University who leads a team building tissue-engineered replicas of brain capillaries, describes the structure bluntly. According to research from Johns Hopkins University, the blood-brain barrier acts as a security system that maintains a tightly controlled biochemical environment for the brain. Researchers at Johns Hopkins note that the blood-brain barrier’s effectiveness means that apart from a few very small molecules, most drugs cannot penetrate into the brain.
What gets through, and how
The barrier is selective, not absolute. Glucose crosses through a dedicated transporter called GLUT1, which the brain expresses at high density on capillary endothelium precisely because neurons consume large amounts of glucose and cannot store their own. Ketone bodies — beta-hydroxybutyrate and acetoacetate — cross via monocarboxylate transporters, which is why a fasting brain can keep working when glucose runs low. Amino acids ride the LAT1 transporter. Iron piggybacks on transferrin. Oxygen and carbon dioxide diffuse straight through the lipid membrane.
Small, fat-soluble molecules can also slip through by dissolving into the endothelial cell membrane and out the other side. That is why caffeine reaches the brain within minutes, why alcohol crosses almost instantly, and why nicotine reaches the brain within seconds of a cigarette pull. The barrier is engineered against water-soluble large molecules — the exact category most modern drugs fall into.
The exclusion is severe. The endothelial monolayer blocks passive diffusion of essentially all large molecules and more than 98% of small-molecule drugs, according to Denali Therapeutics chief executive Ryan Watts, who laid out the numbers at the opening plenary of the Alzheimer’s Association International Conference in London this month. Every antibody, enzyme, and oligonucleotide ever aimed at a brain disease faces the same problem before it reaches its target: it cannot get in.
The barrier Paul Ehrlich stumbled into
The phenomenon has been known since the 1880s, when the German physician Paul Ehrlich injected water-soluble dyes into animals and noticed that every organ turned blue except the brain and spinal cord. His student Edwin Goldmann later injected the same dye directly into cerebrospinal fluid and got the opposite result: the brain stained blue, the body stayed clear. Something between the blood and the brain was acting as a one-way filter.
It took another century to work out the molecular architecture — the tight junctions, the efflux pumps like P-glycoprotein that actively kick foreign molecules back into the bloodstream, the astrocytic end-feet that wrap around capillaries and instruct the endothelium to stay sealed. Together these features make the brain the most pharmacologically isolated organ in the body. Blood-borne infections of the brain are rare precisely because the barrier works so well. The cost is that the same barrier blocks the medicines meant to fix the brain when something goes wrong.
The workarounds have always been crude
For decades, neurologists had three bad options. Inject drugs directly into the cerebrospinal fluid via lumbar puncture — the route used for certain tau-targeting oligonucleotides, which require periodic spinal taps. Blow the barrier open temporarily with focused ultrasound or hyperosmotic mannitol, a blunt instrument that lets both the drug and everything else in the blood cross at once. Or reformulate the molecule to be small enough and fat-soluble enough to diffuse through — a constraint that rules out the entire class of protein therapeutics and gene-editing tools.
Chemotherapy for brain tumors runs headlong into this problem. As a Frontiers review of CNS pharmacokinetics notes, most systemic cancer drugs achieve brain concentrations a fraction of what they reach in the rest of the body, which is why pediatric brainstem tumors remain among the most challenging cancers to treat.
Roche, the Swiss pharmaceutical giant, has spent years developing what it calls Brainshuttle — a bispecific antibody format designed to grab a ride across the barrier. As China Daily reported in May, the company is now pushing that platform into the Chinese neurology market, betting that a delivery system rather than a new drug target is the real bottleneck.
Hijacking the iron delivery route
The insight driving the current wave of blood-brain barrier drugs is that the barrier is not a wall — it is a gateway with specific doors. One of those doors is the transferrin receptor, or TfR1, which the endothelial cells express in abundance on their bloodstream-facing side. Its job is to pull iron across the barrier for myelin synthesis and neurotransmitter production. It does this millions of times a day.
If a drug can be engineered to bind that receptor without competing with the body’s iron supply, it can catch a ride. Denali Therapeutics has been engineering a half-antibody fragment tuned to grab TfR1 on the blood side, get pulled through the cell, and release its cargo — an enzyme, an antibody, an oligonucleotide — on the brain side. In non-human primates, oligonucleotides fused to the transport vehicle show more than 1,000-fold greater brain exposure than the same molecule given without it, according to Denali’s presentation at AAIC 2026.
The first drug through the door is AVLAYAH (tividenofusp alfa-eknm), an enzyme replacement therapy for Hunter syndrome, approved by the FDA in March 2026. Hunter syndrome is a lysosomal storage disorder in which children lack the IDS enzyme; without it, sugars accumulate in tissues including the brain, causing progressive neurological decline. Traditional enzyme replacement therapy could clear the sugars from the body but not the brain, because the enzyme could not cross the barrier. AVLAYAH can. It is the first FDA-approved biologic specifically engineered to shuttle a therapeutic protein into the human brain via receptor-mediated transcytosis.
Small molecules that were already small enough
The other approach is to design small molecules that slip through on their own. That is what a team at the University of Arizona did with an experimental drug called XL20, aimed at ALS. The disease is defined by clumps of a protein called TDP-43 forming inside motor neurons; the vast majority of ALS cases are sporadic, with no known genetic cause, but nearly all share this pathological signature.
Xinglong Wang, a professor at the R. Ken Coit College of Pharmacy, spent a decade with his team identifying a single 20-amino-acid stretch of TDP-43 — residues 320 to 340 — that appears to be the toxic region. Delete it in mice, and TDP-43 stops killing neurons but continues doing its normal job. The team then screened for a small molecule that would bind that region and cross the blood-brain barrier without disturbing anything else. XL20, described in a Nature Aging paper this month, extended survival in ALS model mice by about a week and reversed damage in human motor neurons in culture.
The researchers noted that current FDA-approved treatments for ALS provide only modest benefits. The researchers emphasized the urgent need for breakthrough treatments in ALS. The same TDP-43 pathology is central to a form of late-life dementia called LATE, which affects roughly one in three people over 80, and is found alongside amyloid in more than half of Alzheimer’s brains.
Nanoparticles as smugglers
A third strategy skips the receptor route entirely and uses engineered particles. Science Blog has covered work on gold nanoparticle carriers that ferry lithium — a drug used for bipolar disorder that is toxic at the doses required to reach effective brain concentrations by conventional dosing — directly across the barrier. The particles are small enough to cross via existing transport pathways and release their cargo only inside the brain, dropping the effective systemic dose by orders of magnitude.
Copper, meanwhile, appears to be a piece of the same puzzle in the opposite direction. New imaging work published this month in News-Medical shows that copper imbalance across the barrier may itself drive protein aggregation in neurodegenerative disease — meaning the barrier is not just an obstacle to treatment, but part of the pathology when it fails.
The AI accelerant
The bottleneck has always been finding molecules with the right shape, the right lipophilicity, and the right binding affinity to slip through the barrier without being flushed back out by efflux pumps. That is a chemistry search problem, and it is exactly the kind of problem machine learning has started to chew through. In July, Insilico Medicine announced a $165 million deal in which its AI platform identified a novel target for a central nervous system disease, screening compounds for brain penetration in silico before any of them touched a lab bench.
Harvard Medical School has been running parallel efforts to map the molecular signatures of drugs that successfully cross, building a predictive framework for what a barrier-penetrant molecule looks like before anyone tries to synthesize it. The combination — AI-designed molecules plus receptor-mediated shuttles plus better in vitro barrier models like the ones Searson’s Johns Hopkins lab is building from stem cells — is starting to look like a genuine pipeline rather than a series of individual moonshots.
Why this matters now
Alzheimer’s deaths have risen more than 140% since 2000, even as heart disease, stroke, and cancer mortality have all fallen. The gap is partly demographic — people live long enough to develop dementia — but partly pharmacological. The drugs that transformed cardiology and oncology in the last four decades never worked in neurology because they never reached the brain.
The first-generation anti-amyloid antibodies, lecanemab and donanemab, do cross the barrier, but only about 1 to 2% of the injected dose makes it through, which is why they must be given at high concentrations and why they cause amyloid-related imaging abnormalities — brain swelling and microbleeds — in a meaningful fraction of patients. A blood-brain barrier-crossing version that delivered 10 to 30 times more drug to the brain per injected dose could, in principle, work at lower systemic doses with less inflammation of cerebral blood vessels.
Denali’s DNL921, an amyloid antibody engineered with the transport vehicle platform, is expected to enter Phase 1 trials with initial data in 2027. Its companion drug DNL628 uses the same platform to deliver an antisense oligonucleotide that suppresses tau production — the first time such a molecule will have been given intravenously rather than by spinal tap.
The gate, not the wall
The 4-nanometer gap between endothelial cells has held for the entire history of neuropharmacology. It is still there. The tight junctions have not been widened, the barrier has not been dissolved, and the brain remains as pharmacologically isolated as it was when Ehrlich watched his dyes stop at the capillary wall in 1885.
What has changed is the recognition that the barrier was never really a wall. It was always a set of doors — transferrin, glucose, amino acid, monocarboxylate — evolved for the traffic the brain needs. Learning to knock on those doors with the right molecular key, rather than trying to break down the wall, is the shift that produced AVLAYAH in March and XL20 this month and will likely produce a dozen more candidates in the next three years.
The brain still gets its glucose. The medicine, for the first time in a century, is starting to follow it in.