In late 2020, lipid nanoparticles became the most famous delivery vehicle in the history of medicine. The mRNA COVID-19 vaccines — Pfizer-BioNTech and Moderna — worked in part because tiny synthetic lipid spheres could carry fragile RNA into muscle cells without it degrading before it arrived. A delivery problem, quietly solved over decades of unglamorous formulation research, suddenly went global.
The LNP didn’t come from nowhere. Its origin is usually traced to Alec Bangham, a British haematologist who noticed in the early 1960s that phospholipids spontaneously organised themselves into vesicles when placed in water — spherical shells with a bilayer structure that looked remarkably like a cell membrane. The practical logic followed: here was a container that resembled the biology it was meant to enter, capable of carrying water-soluble cargo in its aqueous interior and fat-soluble cargo within the bilayer itself. Liposomal doxorubicin — Doxil — became the first FDA-approved nanomedicine in 1995, and synthetic LNPs have been the industry’s preferred delivery workhorse since.
They work well. They can be customised — size, surface chemistry, charge, and cargo are all tunable. They can be produced at scale using well-characterised methods. Compared to most nanomedicine platforms, they’re tractable: you can make them reliably, characterise them reproducibly, and adjust them when something doesn’t work. For a field that has repeatedly struggled to translate elegant lab results into manufacturable treatments, this is not a small thing.
But the body keeps recognising them as foreign.
When synthetic nanoparticles enter circulation, plasma proteins adsorb onto the surface almost immediately — a process called protein corona formation. This coating changes the particle’s identity: it’s no longer presenting the engineered surface, it’s presenting a layer of opsonins that the immune system reads as a removal instruction. The mononuclear phagocyte system, particularly the macrophages concentrated in the liver and spleen, clears a significant fraction before the particle reaches its intended target.
PEGylation — coating the surface with polyethylene glycol — became the standard workaround. It creates a steric barrier that slows protein adsorption and extends circulation time. It works, to a degree. The problem is that repeated dosing can trigger the production of anti-PEG antibodies, which accelerate clearance on subsequent administrations and, in some patients, cause hypersensitivity reactions. The fix introduces a version of the problem it was solving.
This is the gap exosomes have stepped into.
Exosomes are nanoscale extracellular vesicles — typically 30 to 150 nanometres in diameter — released naturally by almost every cell type as part of normal intercellular communication. They carry proteins, lipids, and nucleic acids, and they’ve been doing this for considerably longer than any pharmaceutical company has existed. What makes them interesting therapeutically is that their membrane isn’t synthetic. It’s derived directly from the cell that produced them, complete with surface proteins the immune system already recognises as self.
The clearest example is CD47. This surface glycoprotein sends what researchers colloquially call a “don’t eat me” signal — it binds to inhibitory receptors on macrophages and suppresses phagocytosis. Cells expressing sufficient CD47 avoid being cleared from circulation. Exosomes inherit this signal from their parent cells, which means they can move through biological fluids without triggering the same clearance cascade that synthetic particles provoke. There’s no protein corona problem in the conventional sense, because the immune system isn’t treating exosomes as intruders — it’s receiving a signal it already knows how to read. That’s a genuinely elegant piece of biology, and the fact that it might be pharmacologically exploitable is what’s driving a lot of the current interest.
There’s also the blood-brain barrier. Most drugs can’t cross it, and most synthetic nanoparticles can’t either — which is part of why CNS drug delivery has historically been so difficult. Exosomes derived from certain cell types, particularly neural and immune cells, cross more readily. Not because they’re small enough to slip through gaps, but because the cellular uptake machinery on the other side recognises their surface markers and participates in their transport. The barrier is still a barrier; exosomes just have a legitimate reason to be on the other side.
And here is where the argument stalls.
Synthetic LNPs can be produced using microfluidic mixing — a continuous, scalable, well-characterised process. The parameters are controllable, the steps are defined, and pharmaceutical-grade batches are achievable. Exosome manufacturing is none of those things yet. Isolating high-purity vesicles from cell culture media requires techniques like differential ultracentrifugation or size-exclusion chromatography — time-consuming, low-yield processes that don’t translate easily to clinical scale. Batch-to-batch consistency is also a genuine problem: exosome composition reflects the state of the cell that produced them, which varies with culture conditions, passage number, and physiological signals. You’re not manufacturing a defined molecule; you’re harvesting the output of a living system.
Loading them with therapeutic cargo adds another layer of difficulty. Cells package their own signals into exosomes naturally, but engineering them to carry a specific drug or nucleic acid at clinically useful concentrations requires either modifying the parent cell genetically or loading the vesicles post-isolation. Both approaches have meaningful efficiency limits, and neither is fully solved at the scale needed for routine clinical production.
None of this makes exosomes a dead end. The biological advantages are real, the field is moving quickly, and there are active directions worth watching: hybrid vesicles that combine a natural membrane with a synthetic core for improved stability, engineered parent cells with modified surface profiles, and bioreactor-based production platforms designed to handle the volume. The interesting question at this point probably isn’t whether exosomes are categorically better than synthetic LNPs — it’s which class of carrier is best matched to which application. For treatments requiring repeated systemic dosing at scale, synthetic LNPs have a strong practical case. For applications involving the blood-brain barrier, or where immune evasion across multiple doses is the specific clinical problem, exosome biology starts to look like a more sensible foundation.
The field spent decades engineering particles to mimic what cells already do. The question the field is now seriously asking is whether it might be more efficient to start with the cell
