Start here.

Plain language

A human hair is about 80,000 nanometres across. A red blood cell is about 7,000. These vesicles are 30 to 150 — call it 100.

So about eight hundred of them fit across the width of one hair. Ten thousand of them in a line makes one millimetre.

The membrane wrapped around one is roughly five nanometres thick. On a 100-nanometre sphere that is proportionally thinner than the skin of a soap bubble — worth holding onto, because it is the part that has to survive being frozen.

Every comparison in this step is a comparison of size and nothing else. The hair, the blood cell, the soap bubble: analogies for how small, not for what anything does.

Full detail

Why a size is always a range and a method. Nanoparticle tracking analysis infers a diameter from Brownian motion in suspension and reports a hydrodynamic size. Transmission electron microscopy images a fixed, dehydrated, contrast-stained specimen and typically returns a smaller number. NanoFCM measures scattered light against calibrated standards.

In the study specific to this cell line, NanoFCM gave 50–150 nm and TEM gave approximately 80 nm. Same particles, two instruments, two answers. A single headline diameter with no method attached to it is not a measurement.

A geometry check worth doing once: a 100 nm sphere has a volume of about 5.2 × 10⁻¹⁹ litres. Ten billion of them occupy roughly 5.2 nanolitres — about one 191,000th of a one-millilitre fill. Whatever is being counted, it is not filling the container.

Plain language

Almost every kind of human cell sheds membrane-wrapped packets into the fluid around it. They are not debris. They are one of the ways cells pass material to each other without touching.

Follow one back to where it came from. Deep inside the cell there is a compartment — a bag, if the analogy helps, and it is only an analogy — whose own wall keeps folding inward and pinching off tiny bubbles into the space inside it. Biologists call it a multivesicular body: one bag holding dozens of much smaller ones.

Those small bubbles are built inside the cell and never touch the outside until the whole compartment is emptied out. That matters, because it is the difference between the two things a jar of these particles can contain: bubbles built inside a compartment, and bubbles pinched straight off the cell’s outer skin. Same size range, different origin, and no ordinary method can tell them apart afterwards.

Full detail

Nomenclature. “Exosome” properly denotes biogenesis rather than size: intraluminal vesicles that form inside a multivesicular body and are released when that body fuses with the plasma membrane. Most preparations cannot demonstrate that origin, so the field’s reporting guidelines prefer small extracellular vesicle for anything defined by size and density alone. A site that uses the two words interchangeably is, in strict terms, describing a size fraction.

Extracellular vesicles (EVs) are released by most cell types. Small EVs in the 30–150 nm range originate in the endosomal system: inward budding of the limiting membrane of a late endosome produces intraluminal vesicles, and fusion of that multivesicular body with the plasma membrane releases them.

Conventional surface markers are the tetraspanins CD9, CD63 and CD81. The vesicles characterised in the one published study specific to this cell line were positive for all three.

Plain language

Nothing loads these. There is no packer. The compartment wall folds inward, and whatever happens to be against that patch of wall — molecules stuck to it, molecules floating just inside it — is what gets sealed in when the bubble closes.

Some of what is inside is not cargo at all. The sorting machinery that does the folding gets caught in its own parcel, and that turns out to be useful: finding those particular proteins inside a vesicle is one of the few ways to argue it came from a compartment rather than off the cell’s surface.

There is also a hard limit set by geometry. At about 100 nanometres across, a great deal simply does not fit. A single mitochondrion is 500 to 1,000 nanometres — five to ten times the width of the whole vesicle. The nucleus is far larger again. Whatever is in there is small, and there is not much room.

Full detail

ALIX, TSG101, syntenin-1 and flotillin-1 are components of the endosomal sorting machinery that buds a vesicle inward off the wall of a multivesicular body. They end up inside the vesicle because they were part of making it — which is exactly why their presence is used as evidence of endosomal origin.

The protein fraction. Part of it is the machinery that built the vesicle in the first place and therefore travels with it. The rest is the enriched category found by proteomics on this cell source: proteins annotated to cell adhesion and extracellular matrix organisation.

The bilayer is not inert packaging. Its composition is measurably different from that of other vesicle subclasses released by the very same cells, and one of its lipids is read directly by the immune system as a recognition signal.

Plain language

The compartment travels to the edge of the cell and its wall meets the cell’s own outer wall. The two join. For a moment there is no boundary at all between the inside of the compartment and the world outside the cell — one continuous space, an open crater rather than a closed pocket.

The small bubbles are now outside. Notice that they never crossed a membrane to get there. Nothing was pushed through anything; a wall was removed.

Run that same picture backwards and you get a completely different process: a cell closing a pocket around something and drawing it in. That is what happens at the far end of this journey, in step 6. Same shapes, opposite direction — and getting the direction the wrong way round is the commonest mistake made about these particles, including by the software that draws them.

Full detail

The event is membrane fusion. The limiting membrane of the multivesicular body fuses with the plasma membrane, the lumen of the compartment becomes continuous with the extracellular space, and the intraluminal vesicles are released without ever traversing a bilayer. Once released, they are what the word exosome denotes.

The mirror-image process — inward invagination of the plasma membrane, enclosing extracellular material inside a new intracellular compartment — is endocytosis. Exocytosis and endocytosis are not variants of one another: they differ in the direction the membrane moves and in what ends up enclosed.

The sequence on this step is a model, drawn to show the geometry. It is not a micrograph and no frame of it is data. Where this page shows a measurement it names the instrument that produced it.

Plain language

Outside the cell it is crowded and wet: water, salts, sugars, loose protein, and the fibrous mesh that cells sit in. Into that goes a 100-nanometre bubble with no engine, no propeller and no sense of direction.

What moves it is collision. Water molecules strike it from every side, unevenly, millions of times a second, and it jitters. That jitter is called Brownian motion, and it is the only transport it has. Push a whole population in one direction and the push is erased in about two seconds — thermal collision beats any steering force by orders of magnitude and does not care how hard you push. Distance is not the problem out here. Direction is, and there isn’t any.

Where vesicles have been tracked in the bloodstream of laboratory animals, they disappear quickly. Most of the signal is gone within about half an hour, taken up by scavenger immune cells concentrated in the spleen and liver.

Full detail

At roughly 100 nanometres in aqueous medium, transport is diffusive. The Stokes–Einstein relation puts the diffusion coefficient of such a particle in water at the order of a few square micrometres per second — fast across a single cell, negligible across a tissue. Directed motion requires a force that exceeds thermal noise, and nothing about a passive vesicle supplies one.

In a mouse model, reported circulating half-life is on the order of 30 minutes, with roughly a 200-fold reduction in signal by 30 minutes and dominant accumulation in spleen and liver. Clearance is primarily phagocytic, mediated by macrophages of the reticuloendothelial system.

Rodent biodistribution is rodent biodistribution. There are no published human pharmacokinetic data for vesicles from this cell source.

Plain language

A vesicle drifts into a cell and the molecules on its outer surface act like a handshake, testing whether this is a cell worth interacting with.

The cell swallows the vesicle into an internal bubble called an endosome. That sounds like arrival. It is not. By default that bubble is on its way to the lysosome — the cell’s recycling compartment — where everything inside it is broken down.

For anything inside the vesicle to matter, it has to escape that internal bubble before it is digested. Most of it does not. Published estimates of successful escape are typically low single-digit percentages.

This is the biggest unsolved problem in the entire field, and it is why a number printed on a label is so difficult to interpret — the great majority of what is counted never functionally arrives anywhere.

Full detail

Internalisation proceeds through clathrin-mediated endocytosis, caveolin-dependent endocytosis, macropinocytosis, phagocytosis, lipid-raft-dependent routes and, less commonly, direct fusion with the plasma membrane. Which route dominates is cell-type and context dependent.

Internalised vesicles traffic to early and then late endosomes. The default destination is lysosomal degradation.

Functional delivery requires endosomal escape: back-fusion of the vesicle membrane with the limiting endosomal membrane, releasing luminal contents into the cytosol. The process is inefficient and is the recognised rate-limiting step for all vesicle-based delivery.

Escape efficiency is not measured on release panels. Nobody reports it per lot, because there is no standard assay for it.

Plain language

Cargo lists in this category all read the same: microRNA, proteins, lipids. True, and close to useless, because the interesting questions are which ones and how many.

One answer is specific to the placenta. There is a cluster of microRNA genes on human chromosome 19 — around 59 of them — that is switched off in essentially every human tissue except the placenta and very early embryonic cells. It exists only in primates. Mice do not have it at all.

Which means a mouse study cannot model this part of the biology. Not models it imperfectly. Cannot.

Here is the mental model nearly everyone arrives with: each vesicle is a tiny loaded syringe, packed with instructions, docking with a cell and delivering them.

It is wrong, and it is wrong by about two orders of magnitude. Before you read on, put a number down.

Full detail

C19MC — the chromosome 19 microRNA cluster: 46 genes, roughly 59 mature microRNAs, spanning about 100 kb at 19q13.41, imprinted and expressed from the paternal allele, and epigenetically silenced outside placenta and embryonic stem cells.

Note what that does to the preclinical literature in this area. Rodent work cannot be modelling a mechanism the rodent genome does not encode, so any inference drawn from it about this particular cargo is an inference about something else.

And a cargo list is not a specification. No standardised assay relates the composition of a preparation to any measurable biological activity, so “contains microRNA” on a website carries no quantitative content whatsoever.

Chevillet and colleagues quantified microRNA stoichiometry across six sources using absolute quantification against synthetic standards. The mean was 0.00825 molecules of a given microRNA per exosome. Even for the most abundant microRNA in the most concentrated preparation, occupancy stayed far below one copy per vesicle.

Stack that on the endosomal escape problem — cargo that is taken up still has to leave the endosome before it is degraded, and typically only low single-digit percentages do — and you can derive the field’s central limitation for yourself. The arithmetic of delivery does not currently work, which is a large part of why this area is still preclinical.

Plain language

Cells that stop dividing do not necessarily go quiet. Some enter a state called senescence, in which they keep secreting a set of signalling proteins into their surroundings.

In 2026 a group at the NIH National Institute on Aging, working alongside scientists from the company that developed the cell line, published what happened when trophoblast-stem-cell vesicles were added to senescent human skin cells in a dish. Measured levels of three of those proteins — CXCL1, IL8 and GDF15 — went down. Markers of DNA damage went down. NF-κB signalling was suppressed.

Now the part that matters more than the result. In vitro. Human dermal fibroblasts. One study. No animals. No people.

If you take one habit away from this page, take that one: before you believe any finding in this field, ask in what system?

Full detail

Aging Cell 2026;25(2):e70368. NIA intramural programme, grant AG000511, with five employees of the cell-line developer among the co-authors — which makes it NIH-collaborative, not NIH-independent. The distinction is worth keeping, because the second phrasing gets used loosely across this category.

Vesicle characterisation in the same paper: 50–150 nm by NanoFCM, approximately 80 nm by TEM, positive for CD9, CD63 and CD81.

What the study does not establish: anything about an animal, anything about a person, any route of administration, any dose, and any durability beyond the culture period.

Plain language

The state of it, in full:

There are no FDA-approved exosome products. Not for anything.

No clinical trial of trophoblast-derived vesicles is registered anywhere in the world. Zero.

One published study is specific to this cell line, and it was cells in a dish.

Human data elsewhere in the extracellular-vesicle field comes from entirely different cell sources and does not transfer.

No public teratoma or tumorigenicity dataset exists for these lines. What exists is karyotype stability, which is a different question answering a different thing.

The safety panel that does exist is a topical ISO 10993 series — cytotoxicity, irritation, sensitisation, phototoxicity. It says nothing about any other route.

A field can be genuinely interesting and almost entirely unproven at the same time. This one is both, and pretending otherwise is how companies in this category end up in warning letters.

Full detail

Grade: preclinical. In vitro and animal work, one in-vitro publication specific to these vesicles, early-phase human data only in the adjacent mesenchymal-stromal-cell field, no approved product anywhere, and active FDA enforcement in this space.

Borrowing evidence across cell sources is the commonest form of overclaiming here. A phase 3 programme in bone-marrow MSC vesicles is evidence about bone-marrow MSC vesicles.

Quality-control caveats that qualify every number on this page: no standardised assay for biological activity, no agreed dosing unit, substantial variability by culture format and isolation method, and frequent conflation of conditioned medium with purified vesicles.

One thing the reported release panel does not include. HLA-G is real trophoblast biology and it is described in the placental vesicle literature. It is not on the release panel for the material described here, and nothing on this page states that these vesicles carry it. The reported panel is CD9, CD63 and CD81.

The 2012 paper that much of this cell line’s early literature was built on was retracted on 11 February 2021. The grounds were image duplication, missing raw data, underpowered statistics and an undisclosed editorial conflict. It carries a retraction notice on PubMed. This page does not cite it, and the 2015 follow-up explicitly builds on it, so that is not a clean substitute either. You will not find either one anywhere in these nine steps, and that absence is deliberate.