When Moderna and BioNTech demonstrated mRNA vaccines at scale in 2020–2021, the default assumption in much of the biotech community was that lipid nanoparticle delivery had been solved. You encapsulate your mRNA in an ionizable lipid system, you inject it, it gets taken up, protein is expressed. The COVID vaccine lipid nanoparticles are liver-tropic, but surely tweaking the formulation could redirect them anywhere you needed — including the brain.
The reality of CNS-targeted mRNA delivery is considerably more complicated. After spending time working through the literature and running preliminary formulation experiments, we want to lay out what makes the blood-brain barrier such a specific and persistent challenge for LNP-based mRNA — not to be pessimistic about the field, but because understanding the problem precisely is the only way to make progress on it.
What the Blood-Brain Barrier Actually Is
The blood-brain barrier (BBB) is a specialized interface between the cerebrovascular endothelium and brain parenchyma. Unlike peripheral capillary endothelium, which permits paracellular transport of many small molecules, brain endothelial cells are connected by tight junctions built primarily from occludins and claudins (notably claudin-5 and claudin-3) that essentially eliminate the aqueous paracellular pathway. The transendothelial electrical resistance (TEER) of brain endothelium is typically reported in the range of 1500–2000 Ω·cm² in vivo — compare this to peripheral endothelium at roughly 2–20 Ω·cm².
Beyond the tight junctions, the BBB is maintained by an ensemble of cell types: astrocytic endfeet ensheathing capillaries, pericytes embedded in the basement membrane, and microglia in proximity. The term "neurovascular unit" better captures the actual biology. This ensemble does not simply block transport passively — it actively maintains CNS homeostasis by controlling ionic concentrations, clearing metabolic waste, and regulating immune surveillance.
For drug delivery, the consequence is that passive diffusion across brain endothelium is essentially restricted to small, lipophilic molecules below roughly 400–500 Da. LNPs are on the order of 80–150 nm in diameter. They are not getting through by diffusion.
The LNP Size and Charge Problem
Standard COVID-vaccine-lineage LNPs are formulated to particles in the 80–120 nm range with a slightly negative or near-neutral surface charge (zeta potential around -5 to -10 mV) at physiological pH. The ionizable lipid component carries a positive charge at endosomal pH (roughly 5.5–6.0) to facilitate endosomal escape, but is largely neutral at blood pH of 7.4.
This formulation is well-optimized for hepatic uptake. ApoE from serum adsorbs onto the LNP surface and the ApoE-decorated particle is recognized by LDL receptor and LDLR-related proteins on hepatocytes, driving efficient liver uptake. The problem is that this hepatotropism is not incidental — it is the result of deliberate physicochemical selection. When you try to re-route the same particle type to brain endothelium, you are fighting against serum protein adsorption patterns (the protein corona) that direct the particles away from your intended target.
Reducing LNP size to sub-60 nm improves some aspects of distribution, but creates stability and encapsulation efficiency tradeoffs. Very small LNPs tend toward inverse hexagonal or cubic lipid phases under certain conditions, which affects mRNA release kinetics. The relationship between particle size and brain penetration is not monotone — there are physical reasons why particles in certain size ranges might exploit endocytic pathways at the BBB while larger particles cannot, but moving to very small sizes introduces its own formulation problems.
PEGylation: Necessary but Double-Edged
PEGylation — incorporating PEG-lipid conjugates like PEG2000-DMG or PEG2000-DSPE into the LNP shell — is standard practice for extending circulation half-life and reducing opsonization. Without PEG-lipid, LNPs are cleared by the mononuclear phagocyte system within minutes. With appropriate PEGylation (typically 1.5–3 mol% PEG-lipid), circulation times extend to hours.
But PEG creates two problems for BBB crossing. First, the steric barrier that prevents opsonins from binding also prevents the specific protein interactions (like ApoE loading) that you might need for receptor-targeted uptake at the BBB. Second, there is documented PEG-antibody induction after repeat dosing — the anti-PEG immune response can accelerate clearance of subsequent doses, which is particularly relevant for a chronic neurological application that would require repeat administration.
Various alternatives — PEG chains with faster shedding kinetics, alternative steric polymers like poly(2-oxazoline), or surface presentation of targeting ligands through PEG-conjugated antibody fragments — are being explored in the literature, but each introduces new complexity into the formulation and manufacturing process.
P-glycoprotein, BCRP, and Active Efflux
Even if an LNP could traverse the brain endothelium, the BBB has an active efflux system designed to pump out lipophilic molecules that do manage to cross. P-glycoprotein (P-gp, encoded by ABCB1) and Breast Cancer Resistance Protein (BCRP, encoded by ABCG2) are ATP-binding cassette transporters expressed at the luminal face of brain endothelial cells. They recognize a broad range of substrates and actively pump them back into the bloodstream.
P-gp and BCRP are critical reasons why many small molecule drugs that reach every other tissue still fail to achieve therapeutic brain concentrations. For LNPs, the direct efflux risk is lower because intact nanoparticles are not substrates for these transporters — they are too large. However, the ionizable lipid components themselves, after LNP disassembly in the endosome, can potentially be recognized as efflux substrates, and there is evidence that LNP uptake activates innate immune signaling in endothelial cells that upregulates efflux pump expression over time. This is understudied and we're not saying it definitively limits CNS mRNA delivery — but it is a confound that peripheral (liver-targeted) LNP development has not had to grapple with.
Transcytosis vs. Paracellular: The Two Paths Across
For particles that cannot use passive diffusion or paracellular routes, the remaining option is transcytosis — vesicular transport across the endothelial cell from lumen to abluminal face. Brain endothelial cells do conduct constitutive transcytosis at low levels, primarily clathrin-mediated endocytosis followed by transcytotic routing. The problem is that most endocytic events in brain endothelial cells are directed toward lysosomal degradation rather than transcytosis. The transcytotic capacity in brain endothelium is substantially lower than in peripheral endothelium, which is why large protein drugs have poor CNS penetration even when they are receptor-targeted.
Receptor-mediated transcytosis (RMT) is the most studied approach for exploiting the transcytotic pathway. Transferrin receptor (TfR1, encoded by TFRC) is highly expressed on brain endothelium and mediates iron transport into the CNS — it is a genuine transcytotic receptor, not just an endocytic one. LNPs conjugated with transferrin or anti-TfR1 antibody fragments have shown enhanced brain uptake in rodent studies, with some reports of 5–10× improvement over untargeted controls. LRP1 (low-density lipoprotein receptor-related protein 1) is another validated RMT receptor, which is part of why ApoE-mimetic peptides on LNP surfaces have been explored.
The translational caveat here is substantial: most of the RMT receptor targeting data comes from rodent models, and the expression levels and transcytotic capacity of TfR1 and LRP1 differ between mouse brain endothelium and human brain endothelium. The failure of some transferrin-conjugate therapeutics to translate from rodent to human CNS delivery is a well-documented pattern in the field. We are not saying RMT targeting doesn't work — we are saying that rodent efficacy data for BBB-crossing LNPs should be interpreted with explicit acknowledgment of the species difference.
Alternative Routes: Focused Ultrasound and Intranasal
Given the difficulties with systemic LNP delivery to the brain, two alternative approaches are active research areas.
Focused ultrasound (FUS) combined with circulating microbubbles can transiently open the BBB at targeted brain regions. The acoustic radiation force and microbubble oscillation cause transient widening of tight junctions and increase transcytotic activity, creating a window of perhaps 4–6 hours where peripherally administered nanoparticles can access brain parenchyma at the sonicated location. FUS-BBB opening has advanced to clinical evaluation for drug delivery in glioblastoma and Alzheimer's disease. The approach is mechanistically compelling and the spatial targeting is impressive — you can open the BBB in a sub-centimeter volume.
The limitation for chronic neurological indications is procedural burden. A patient requiring repeat LNP-mRNA dosing for a neurodegenerative condition would need repeat FUS sessions, each requiring MRI guidance and anesthesia in some protocols. This is not a blocker for a serious indication, but it constrains the patient population and administration logistics substantially. For longevity-type applications where the target patient is not yet severely impaired, the risk-benefit calculus is different than for ALS or severe Parkinson's.
Intranasal delivery exploits the olfactory and trigeminal nerve pathways, which provide a non-vascular route from nasal mucosa into the brain. LNPs administered intranasally can reach olfactory bulb and, to a lesser extent, deeper brain structures via axonal transport along olfactory receptor neurons. The absolute brain concentrations achievable by intranasal delivery are low — typically a small fraction of the administered dose reaches target structures — but for some applications where target cells are accessible near olfactory entry points, it may be sufficient.
We are currently evaluating intranasal administration as one route for our CNS-targeted programs, specifically for targets where expression in olfactory bulb or hippocampus would be sufficient for initial proof-of-concept studies. We are not claiming this solves the systemic CNS delivery problem — it does not, and for many brain regions it is not a viable route at all.
What This Means for mRNA Neurodegeneration Programs
The practical consequence of the BBB challenge is that mRNA neurodegeneration programs require explicit delivery solutions, not just mRNA sequence optimization. You cannot take a peripherally validated LNP formulation, design an mRNA for a longevity-associated brain protein, and expect meaningful CNS expression.
At caVos, when we identify a target through our comparative genomics pipeline that requires CNS expression — Klotho, for instance, which is expressed in the choroid plexus and relevant to neurological aging — the delivery problem is co-equal in importance with the mRNA sequence design problem. This is different from how some early-stage mRNA groups are structured, where delivery is treated as a downstream concern. The constraint shapes the target selection: we preferentially advance targets where either (a) peripheral organ expression is therapeutically sufficient (as it is for circulating Klotho), or (b) there is a defined delivery strategy with animal data supporting brain access.
The LNP field for CNS delivery is genuinely improving. The combinatorial lipid screening approaches developed originally for liver delivery are being applied to CNS, and there are now several published LNP formulations with documented CNS activity in rodents and non-human primates. But the gap between liver delivery efficiency and brain delivery efficiency remains at least two orders of magnitude for most formulations. That is not a minor engineering problem — it is a fundamental biological constraint that any honest CNS mRNA program has to confront directly.
