Inside the body there are molecules whose job is to act as brakes, calming inflammation once it has flared up. The most prominent of them is interleukin-10 (IL-10). IL-10 has long been hoped for as a treatment candidate for rheumatoid arthritis, inflammatory bowel disease, and the neuroinflammation that follows spinal cord injury. Yet most of the clinical trials run from the 1990s into the 2000s—the Crohn’s disease trials above all—failed to show the efficacy people had expected. One of the reasons is a troublesome one: the IL-10 protein is unstable, and on top of that it disappears from the body fast.
So why not simply load it onto a “carrier” that keeps it from breaking? The carrier that has drawn attention in recent years is the extracellular vesicle (EV), a package that cells release naturally. The paper we are reading today comes from a research team at Michigan State University, published in Extracellular Vesicle: they harvested EVs from cells engineered to overexpress IL-10 and dissected, with almost obsessive granularity, what form IL-10 takes inside those vesicles and on their surface.
To give the conclusion first: IL-10 did load onto EVs. It became clearly harder to break down than recombinant IL-10 protein. It also suppressed cytokine production by inflammatory macrophages. And yet—naïve EVs, taken from cells into which no IL-10 gene had been introduced, also plainly suppressed inflammation. What is more, in primary human macrophages drawn from the blood of healthy donors, that effect was equal to the EVs loaded with IL-10. The authors do not hide this inconvenient result; they write it into the abstract itself. And this paper carries one more important premise. Every experiment described here took place in a culture dish. Not a single animal experiment was performed.
Let me settle one piece of terminology. What this article calls “naïve EVs” is what the paper writes as naïve EV—that is, EVs harvested from untouched cells that were never put through the procedure of having the IL-10 gene introduced. It does not mean “vesicles that are empty inside.” In fact, this paper detected a weak IL-10 signal even in the EV fraction derived from cells with no gene introduced, and points out the possibility that IL-10 natively associated with EVs exists. “Unmodified” and “empty” are not the same thing—and that distinction is the heart of what makes this story tricky.
Publication Details
- Paper title: Harnessing extracellular vesicles for stabilized and functional IL-10 delivery in macrophage immunomodulation
- Paper type: Original research article. It consists entirely of in vitro (cultured cell) experiments; no animal experiments, disease models, or clinical trials are included.
- Authors: Najla A. Saleh (first author), Matthew A. Gagea, Xheneta Vitija, Sadhana Kilangodi, Ahmed A. Zarea, Tomas Janovic, Jens C. Schmidt, Cheri X. Deng, and Masamitsu Kanada (corresponding author)—nine in all
- Affiliation: Multiple departments, centered on the Institute for Quantitative Health Science and Engineering (IQ) at Michigan State University, USA. Co-author Cheri X. Deng is at the University of Michigan
- Journal: Extracellular Vesicle (published by Elsevier; the official journal of the American Association of Extracellular Vesicles, AAEV), 2026, Volume 7, Article No. 100102
- DOI / Link: 10.1016/j.vesic.2025.100102
- Peer review and publication dates: Received July 8, 2025; revised manuscript received November 25; accepted December 10; published online December 12. It had already been posted in advance on bioRxiv on January 18, 2025, and two authors were added over the course of peer review, bringing the final version to nine
- Open access: Yes (CC BY 4.0. Free to reuse and redistribute with attribution)
- Impact Factor: Not yet assigned. Because this journal is not indexed in the Web of Science Core Collection, no Journal Impact Factor exists for it. On the other hand, it is already indexed in Scopus (coverage 2022–2026) and registered in DOAJ as well. It is not indexed in MEDLINE, and papers from this journal appear in PubMed only when the author manuscript has been deposited in PubMed Central under a public access policy (this paper is one such NIH-funded deposit)
- About the journal: A young society journal launched in 2022, with Professor Ke Cheng of Columbia University as Editor-in-Chief. The “citadel” of the EV field is still the Journal of Extracellular Vesicles (IF 21.7) of ISEV (the International Society for Extracellular Vesicles), but it would be fair to place this journal as “a specialist title, not yet carrying a metric, that the society of the US EV community is nurturing”
- Funding: R21-EB033554, R01-EB030565, and R01-EY016077 from the US National Institutes of Health (NIH), plus Michigan State University startup funds. The paper itself states explicitly that the work was “supported, in part”
- Conflicts of interest: The authors declare no competing interests, financial or non-financial.
About corresponding author Masamitsu Kanada: The corresponding author is an Assistant Professor at Michigan State University, running his own laboratory, the “Kanada Lab,” in IQ’s synthetic biology division (appointed July 2017). What stands out is that the entire foundation of his career lies in Japan. He earned a bachelor’s degree in biochemical engineering at Tohoku University (1999–2003) and a doctorate in cell biology at the University of Tsukuba (2003–2008), then worked on cancer drug development at a pharmaceutical company, learned intravital microscopy in Susumu Terakawa’s laboratory at Hamamatsu University School of Medicine, and mastered whole-body preclinical imaging in Christopher Contag’s laboratory in the Department of Pediatrics at Stanford University School of Medicine (2012–2017). The laboratory rests on three pillars: (1) understanding and controlling EV-mediated cell-to-cell dialogue in the tumor microenvironment, (2) remaking EVs into a new gene delivery system, and (3) searching for factors that regulate EV biogenesis and secretion. Representative work includes Proc Natl Acad Sci U S A 2015;112(12):E1433–E1442, which showed that DNA, mRNA, and protein carried by EVs each follow a different fate in the recipient cell, and—the direct technical ancestor of the present paper—Mol Cancer Ther 2019;18(12):2331–2342, which loaded minicircle DNA into microvesicles to make gene-directed enzyme prodrug therapy work. He is also one of the co-authors of MISEV2023, the international guidelines for EV research. Note that of the three NIH grants in the acknowledgments, the one on which he himself serves as principal investigator (PI) is R21-EB033554, “Extracellular Vesicle Engineering and Induced Release using ultrasound (EVEiR).” First author Najla A. Saleh is a postdoctoral researcher in the same laboratory, originally from the Federal University of Santa Catarina (UFSC) in Brazil. She has studied the antiproliferative effect that miRNA in M1 macrophage-derived EVs exerts on melanoma cells, and the theme of “macrophage × EV × immunomodulation” has run consistently through her work since her doctoral studies.
What Wasn’t Understood Before?
Inside our bodies, the cell that rushes to the scene of inflammation first and takes command is the macrophage. Macrophages are not cells with a fixed personality; they are plastic cells that change their character according to the surrounding environment, moving back and forth between two poles—broadly, the M1 type, which fans the flames of inflammation, and the M2 type, which works to bring inflammation to a close and repair tissue. Much of chronic inflammation is thought to be a state in which this balance has tipped toward M1 dominance and never comes back.
The brake that stops that runaway process is IL-10. IL-10 works as a soluble homodimer of two molecules facing each other, and binds to a complex assembled from the receptors IL-10RA (the ligand-binding side) and IL-10RB (the signal-transducing side). The enzymes JAK1 and TYK2, sitting alongside the receptors, are then phosphorylated, the transcription factor STAT3 is activated, a brake molecule called SOCS3 is induced, and the transcription factor NF-κB—the command center of inflammation—is blocked from entering the nucleus. As a result, production of inflammatory cytokines such as TNF-α, IL-1β, IL-6, and IL-12 is suppressed, and transport of MHC class II to the cell surface is suppressed as well, putting a check on T cell activation too. Just how fundamental this molecule is can be seen from the fact that mice lacking IL-10 spontaneously develop chronic colitis (Kühn R, et al., Cell 1993;75(2):263–274).
👦 Student: If there’s a brake that works that well, why not just turn it into a drug?
🧬 Dr. Exotaro: The pharmaceutical industry thought exactly the same thing thirty years ago. Recombinant human IL-10 was actually made and entered clinical trials under the name ilodecakin (trade name Tenovil). But it didn’t go well. The brake itself was genuine, yet the moment you pressed it, the brake pedal melted away—that’s the kind of drug it turned out to be.
The numbers make the situation clear. The terminal half-life of recombinant human IL-10 given intravenously to healthy volunteers was a mere 2.3±0.5 to 3.7±0.8 hours (Huhn RD, et al., Blood 1996;87(2):699–705). The clinical trial results were harsh as well. In a double-blind, placebo-controlled trial in 329 patients with treatment-refractory Crohn’s disease, the remission induction rate at every dose—1, 4, 8, and 20 µg/kg—was not significantly different from placebo (18%) (Schreiber S, et al., Gastroenterology 2000;119(6):1461–1472), and in a trial of 95 patients with mild to moderate disease, 23.5% of the 5 µg/kg group showed clinical remission and endoscopic improvement on day 29, yet the result was that the higher doses of 10 µg/kg and 20 µg/kg were, if anything, less effective (Fedorak RN, et al., Gastroenterology 2000;119(6):1473–1482). A trial for the prevention of postoperative recurrence produced no significant difference either (Colombel JF, et al., Gut 2001;49(1):42–46).
“The higher dose works less well” is bizarre behavior for a drug. Part of the reason is known. In patients given 20 µg/kg, neopterin, a marker of inflammation, was significantly elevated (Tilg H, et al., Gut 2002;50(2):191–195). Neopterin is a molecule that monocytes and macrophages make in response to IFN-γ. In other words, at high doses IL-10 flips over to the side that stimulates immunity. IL-10 is not a purely anti-inflammatory molecule; it also has the effect of enhancing the function of CD8-positive T cells and NK cells—a double-edged sword in the literal sense.
👦 Student: So the range of doses that works is extremely narrow?
🧬 Dr. Exotaro: Precisely. And since the half-life is only a few hours, just keeping the concentration inside that narrow effective window is itself an extraordinarily difficult feat. The one indication where reasonable results came out was psoriasis: there is a Phase II report of a 55.3±11.5% drop in PASI score with subcutaneous administration (Asadullah K, et al., Arch Dermatol 1999). But for the other indications, development was discontinued for lack of efficacy.
So researchers split into two camps. One line was to rebuild the protein itself to be sturdier. Natural IL-10 is a dimer in which two molecules stick together non-covalently, so under stress it comes apart into monomers. A single-chain dimer was therefore made in which the two monomers are covalently joined by a linker, and whereas the natural form loses activity within 5 minutes at 55℃, this modified form retained activity under the same conditions (Minshawi F, et al., Front Immunol 2020;11:1794).
The other line was to load it onto a carrier. Extracellular vesicles (EVs) are lipid bilayer sacs that cells release naturally; they have high biocompatibility and the property of reaching selected cells and tissues. The idea itself is not new. Work treating ischemic acute kidney injury in mice with EVs loaded with IL-10 from genetically modified macrophages (Tang TT, et al., Sci Adv 2020;6(33):eaaz0748), and work lining up decoy receptors on the EV surface to adsorb inflammatory cytokines (Gupta D, et al., Nat Biomed Eng 2021;5(9):1084–1098), have already reached the animal experiment stage.
So what was still unknown? What this paper’s introduction raises is a startlingly basic question—how a cytokine gets packaged into an EV in the first place, and how it is functionally delivered to the recipient immune cell, remain almost entirely unresolved. We are talking about carrying cargo, and yet nobody had confirmed where in the box the cargo sits, or in what form.
What Did This Paper Find?
Before anything else, let us confirm an absolute premise for reading this paper. Every experiment introduced from here on was carried out inside a culture dish. Both the producer cells and the recipient cells are cultured human (and in some cases mouse) cells. Not a single experiment involved administration to animals, so how these particles distribute inside the body, how long they remain in the bloodstream, and whether side effects occur are things this paper tells us nothing at all about.
Loading IL-10 into EVs—A Tool Called Minicircle DNA
The authors first prepared a device for making cells produce IL-10 in large amounts. What they used was a circular vector called minicircle (MC) DNA. An ordinary plasmid carries a bacteria-derived backbone sequence, and that backbone gets in the way inside mammalian cells. A minicircle is a ring made of “only the necessary parts,” with that backbone excised in advance; it has high transfection efficiency, longer-lasting expression, and low cytotoxicity. Indeed, when compared with the parental plasmid at the same DNA mass, the minicircle showed strong, sustained expression from day 2 through day 6 after transfection. This is the corresponding author’s specialty.
The cell chosen to produce the EVs was HEK293FT, an immortalized cell line derived from human embryonic kidney. The authors explicitly state their reasons: high EV yield, a track record in biomanufacturing such as lentivirus production, and less donor-to-donor variability than primary cells.
On top of that, they ran three isolation methods based on different principles in parallel on the same culture supernatant: (1) passing the medium through a 0.2 µm filter and then vacuum-filtering it through a 50 nm porous membrane (the vesicles obtained are called F-sEVs); (2) differential ultracentrifugation (which separates the material into three fractions—large EVs = lEV, small EVs = UC-sEV, and non-vesicular particles = NVEP); and (3) DEAE anion-exchange chromatography, which separates by surface charge. Splitting the same material three ways and comparing the results—this design is itself the backbone of the paper.
IL-10 Was on Both the “Surface” and the “Inside” of the Vesicles
The particle size of the F-sEVs measured by nanoparticle tracking analysis was 113.3 nm for those derived from IL-10-transfected cells and 112.7 nm for those derived from non-transfected (naïve) cells. Loading IL-10 barely changed particle size. That IL-10 was indeed on board was confirmed by three independent methods: protein (Western blot), mRNA (qPCR, n=5), and immunogold-labeled electron microscopy. The EV-positive markers CD9, CD63, CD81, and TSG101 were detected, while the negative marker Calnexin was not.
Here comes the first surprise. When IL-10 was detected by Western blot, not a single band but multiple bands appeared. In other words, IL-10 was associated with EVs not only as a monomer but also in several oligomeric (multimeric) forms. The authors add a careful caveat here: “whether these oligomers represent functionally ordered structures or merely non-functional aggregates remains undetermined.”
👦 Student: They’re inside the vesicles, right? Or are they stuck to the outside?
🧬 Dr. Exotaro: There’s a clever experiment for telling those apart. It’s called the proteinase K protection assay. When you sprinkle on a protein-degrading enzyme, only what is exposed on the outside of the sac gets eaten. What’s inside is shielded by the membrane and survives. Then you compare a condition in which you add detergent to break the membrane against one in which you don’t. In other words, you can judge that “whatever disappears only once you rip the sac open with detergent = inside” and “whatever disappears as is = outside.”
The answer from this experiment was “both.” Monomeric IL-10 and the high-molecular-weight higher-order oligomers were degraded by the enzyme alone—they are exposed on the outer surface of the vesicle. The smaller oligomers, on the other hand, were only partially degraded when detergent was used as well—at least some of them are protected inside the vesicle. The luminal marker TSG101 was properly protected and functioned as a control (though the authors honestly note that a slightly smaller minor band appeared under some conditions).
Then comes the main point: stability. When EV-incorporated IL-10 and naked recombinant human IL-10 protein were compared under three stresses—(1) 2 hours at room temperature, (2) 2 hours at 37°C, and (3) 2 freeze-thaw cycles—both the encapsulated form and the surface-bound form showed significantly higher stability than recombinant IL-10. Recalling that the single biggest reason IL-10 could never be turned into a drug was its fragility, this becomes the paper’s central claim.
The Question of “Why It Binds to the Surface” Went Unanswered
It is a strange story that something exposed on the surface does not break down. The surface of EVs carries heparan sulfate proteoglycans (HSPGs), and it has long been known that cytokines bind to their sugar chains. The authors suspected this was the adhesive and ran three experiments, but none of them reached an answer. The experiment using a drug that inhibits HSPG biosynthesis ended as “inconclusive” because IL-10 expression dropped even in the control group given the solvent DMSO, so no valid comparison could be made (the authors themselves explicitly call it inconclusive). Cleaving the sugar chains with heparinase II did not change the amount of IL-10 on the EVs, ruling out any dependence on heparan sulfate. And in an experiment attempting to artificially attach recombinant IL-10 to naïve F-sEVs after the fact, no IL-10 monomer could be detected on the F-sEV surface even after 2 hours (the authors write that they cannot distinguish whether it simply does not bind, or whether binding occurred but was too weak or too transient to detect). From this, the authors speculate that a protein “corona” containing IL-10 cannot be created after the fact, and may instead form spontaneously during the process by which EVs are made.
It Was in Every Fraction, and Every Fraction Worked
When the three fractions separated by differential ultracentrifugation—lEV, UC-sEV, and the NVEP that are not even vesicles—were examined, IL-10 was present in every fraction, mainly as a monomer. Its localization differed by fraction: in lEV it was on both the inside and the outside; in UC-sEV mainly on the outer surface; and NVEP contained both monomers and oligomers, with the monomers moreover protected from the enzyme. The exosome-enriched fraction purified by anion exchange (fractions 9-12 of the Protein-high group) was enriched in three forms in total: the monomer plus two kinds of oligomer.
Then the functional experiments. Fluorescently labeled F-sEVs accumulated markedly in the endosomes of M1-like macrophages within 17 hours. When F-sEVs at a total protein amount of 10 µg were applied to LPS-stimulated macrophages for 24 hours, the IL-10-positive F-sEVs lowered the three inflammatory cytokine genes TNFA, IL6, and IL1B to levels comparable to those of non-inflamed M0-like macrophages. Naïve F-sEVs suppressed these as well, but to a smaller degree. Measuring protein levels in the supernatant by ELISA likewise confirmed a decrease in secretion.
Up to this point, it is a clean success story. But the real highlight of this paper lies beyond it.
👦 Student: If it worked that well, isn’t that a huge success?
🧬 Dr. Exotaro: The problem starts here. The “naïve EVs”—the ones from cells that never received the IL-10 gene—suppressed inflammation solidly too. And when tested in primary human macrophages taken from the blood of healthy donors rather than in a cell line, the effects of naïve EVs and IL-10-loaded EVs came out equivalent. That said, it wasn’t a tie on every readout—for TNFA and IL1B at the mRNA level and for secreted IL-1β protein, the IL-10-loaded side did show a proper significant difference.
This is the single biggest finding of the paper, one the authors went so far as to put in the abstract. The original states: “naïve F-sEVs from non-transfected cells also exhibited anti-inflammatory effects, suggesting that cargo intrinsic to EVs contributes to their immunomodulatory activity and making it difficult to attribute the effect specifically to IL-10.”
Naturally, suspicion arises: might this simply be contamination by EVs derived from the bovine serum used in the medium? The authors went after that suspicion themselves. When they remade the EVs under completely serum-free conditions and compared them, there was no significant difference in the suppression of NF-κB activity. In other words, the anti-inflammatory property is intrinsic to the EVs themselves and cannot be explained by components mixed in from outside. It is an honest experiment that closes off their own escape route.
The inconvenient results keep coming. Blocking the IL-10 receptor α with an antibody did not block the suppression of NF-κB. In other words, the mechanism at work remains unexplained. Nor did “repolarization” from M1 to M2 occur—none of the M2 markers IL10, CD163, MRC1, or CD209 went up, and the authors describe the effect as “immunosuppressive rather than polarization-promoting.” This conflicts with an earlier study that achieved M2 conversion in mouse ischemic acute kidney injury (Tang TT, et al., Sci Adv 2020), and the authors speculate that it may stem from the difference between the environment of a culture dish and that of a living body.
Furthermore, even the NVEP, which are not even vesicles, lowered inflammatory cytokines (though the IL-10-loaded NVEP alone failed to lower IL6). Of the three fractions, only the IL-10-positive UC-sEV sustained suppression of NF-κB activity for up to 5 hours, and this was one of the few findings that can be called specific to IL-10.
👦 Student: So the results change depending on how you separate them…
🧬 Dr. Exotaro: That’s exactly it. The authors show with their own data that higher-order oligomers above 102 kDa are lost once the material goes through a process that includes serial ultracentrifugation. As for the cause, they cite earlier work showing that the strong shear forces of ultracentrifugation can disrupt EV structure, but they did not go so far as to test that directly. Even so, it means the answer to “what is loaded on the vesicle” is decided not by biology alone—half of it is decided by the experimental method. That’s a question thrown at the whole of EV research.
The authors close the paper with this single sentence: “Our findings reinforce the emerging view that EVs are not inert carriers but complex biological entities with intrinsic immunomodulatory properties, shaped by their cellular origin and by the molecular composition of both their surface-bound and internal contents.” It is a conclusion that works against their own sales pitch.
How Will the Future Change? (The Path to the Clinic)
First, let us pin down exactly where we stand. This study remains at the cultured-cell stage, and not even animal experiments have been performed. Pharmacokinetics, tissue distribution, immunogenicity, toxicity, effective dose—not one of these has been measured. To write that this is “a step forward toward clinical application” would, frankly, be contrary to fact.
What’s more, the field has already moved ahead. A study showing therapeutic effect and M2 polarization in a mouse model of ischemic acute kidney injury (Tang TT, et al., Sci Adv 2020;6(33):eaaz0748); a study that actually measured biodistribution and pharmacokinetics in mice and went as far as preventing preterm birth (Harrington B, et al., Extracellular Vesicle 2025;5:100066—the same journal as this paper); a study treating colitis with IL-10-loaded EVs wrapped in a hydrogel so they could be given orally (Liu J, et al., Small 2023;19(50):e2304023). The very idea of “carrying IL-10 with EVs to suppress inflammation” is already supported by multiple animal studies.
So where does this paper’s value lie? It lies not in advancing, but in inspecting. It is a work that stopped to check the ground the field has been sprinting across—where on the vesicle is the cytokine, and in what form? What changes when you change the isolation method? And is the “naïve EV” placed there as a control really a control that does nothing at all?
👦 Student: If naïve EVs work too, doesn’t that make loading IL-10 rather pointless?
🧬 Dr. Exotaro: Sharp question. But there are two ways to read it. One is the harsh reading: “they failed to prove any added benefit from IL-10.” The other is the constructive reading: if you can pin down the true nature of the anti-inflammatory activity that naïve EVs already possess, that itself could become a drug. The authors themselves list miRNA and a lipid called phosphatidylserine exposed on the membrane as candidates. In fact, the latter may be the more interesting door this paper has opened.
What is still missing on the path to the clinic is spelled out clearly by the authors themselves: that the structural and compositional complexity of the IL-10 associated with EVs and NVEP obstructs precise dose setting and reproducible assessment of effect; and that batch-to-batch variation arising from differences in isolation method, culture conditions, and source means a standardized manufacturing process is needed. On top of that, there is the question of whether HEK293FT can continue to be used as a source of therapeutic EVs. This cell line was immortalized with adenoviral genes and additionally expresses the SV40 large T antigen, so the regulatory hurdles concerning residual DNA and tumorigenicity are clearly higher than for mesenchymal stem cell (MSC)-derived EVs.
How to Critically Read This Research (Limitations and Paths to Improve It)
Before turning to criticism, let me state this paper’s virtues clearly. Its greatest strength is that the authors did not cut the data that break their own story. That naïve EVs had an effect; that blocking the receptor did not abolish the suppression; that M2 conversion did not occur; that the solvent control broke down and left the experiment inconclusive—all of it is written in the paper, and some of it made it into the abstract. In addition, the design that pitted three isolation methods based on different principles against each other, the willingness to functionally evaluate even the NVEP fraction that is normally discarded, and the decision to go after the confounder of serum-derived EVs themselves all deserve fair credit. The limitations of a good study and the limitations of a bad study are different things. This is the former.
With that said, here are the limitations. The limitations the authors themselves acknowledge are: (1) the effect cannot be attributed to IL-10; (2) the mechanism of action is unresolved; (3) M2 repolarization did not occur; (4) precise dose setting is not possible; (5) higher-order oligomers are lost in processes that include serial ultracentrifugation; (6) the contribution of NVEP cannot be separated out; (7) whether the oligomers are functional structures or aggregates is unknown; and (8) in the 72-hour experiments, the influence of cell death cannot be ruled out.
On the other hand, there are a number of important problems the paper does not touch on.
There is a problem with how the controls were set. The comparison control, “naïve,” consists of EVs made from cells into which no DNA was introduced at all. There is no group of cells transfected with an empty minicircle (a mock control). As a result, the difference showing that “IL-10-positive EVs work better” contains not only the effect of IL-10 but also, wholesale, the effect of the transfection procedure itself (carryover of the transfection reagent, activation of intracellular DNA sensors, stress from overexpression, changes in EV composition). Indeed, in the paper’s figures the detection patterns for CD81 and CD9 clearly differ between the naïve and transfected groups, suggesting that transfection itself changes the very character of the EV population.
How the doses were matched is also a concern. By the values shown in the paper’s figures, IL-10-positive F-sEVs were 2.0×10¹⁰ particles/mL with a total protein of 2561±431 µg/mL, while naïve F-sEVs were 1.1×10¹⁰ particles/mL with a total protein of 2085±529 µg/mL. Calculating particles per 1 µg of protein from these gives roughly 7.8×10⁶ for the IL-10-positive preparation and roughly 5.3×10⁶ for the naïve one—a gap of about 1.5-fold. In the functional experiments matched by total protein amount (Fig. 5d, Fig. 6, Fig. 7c), it follows that the IL-10-positive group received correspondingly more particles (comparisons between fractions were matched by equal volume, and the uptake experiments were matched by particle number). More serious still is the comparison between fractions: in the experiment where lEV, UC-sEV, and NVEP were each administered at 50 µL apiece, calculating from the total protein concentrations shown in the figure (118.2, 114.1, and 42608.9 µg/mL) shows that the NVEP group alone received roughly 360-fold more by protein amount and roughly 30-fold more by particle number. The conclusion that “the anti-inflammatory effect is not size-dependent” is drawn from a comparison in which dose was not matched, and as it stands it does not hold.
The difference seen at the mRNA level vanishes at the level of the protein actually secreted. The abstract states “2- to 14-fold more effective than naïve F-sEVs,” yet this multiple never appears once in the results section of the main text. Looking at the corresponding figures, at the mRNA level the difference from naïve is not significant for IL6, and at the level of secreted protein measured by ELISA, neither TNF-α nor IL-6 differs significantly from naïve (the only significant difference was for IL-1β). The same was true for the exosomes purified by anion exchange. The “2- to 14-fold” figure should not be taken at face value.
The very experiment that ruled out the mechanism cannot itself be interpreted. The finding that “the anti-IL-10Rα antibody did not block NF-κB suppression” is important, but the paper describes no isotype control, and moreover there is no positive control demonstrating that this antibody can actually block the IL-10 receptor in this experimental system. Consequently, there is no way to distinguish between “there is a receptor-independent mechanism” and “the antibody simply wasn’t working.” The same applies to the negative finding that “M2 conversion did not occur.” Because no genuine M2 macrophages treated with IL-4 or IL-13 were included as a positive control, it cannot be confirmed that the system was even capable of detecting M2 markers in the first place.
And the greatest missed opportunity is that the absolute amount of IL-10 carried on the EVs was never measured, not even once. ELISA was used only for TNF-α, IL-1β, and IL-6. How many ng of IL-10 ride on 1 µg of EVs, and what percentage of the total secreted IL-10 was recovered in the EVs (loading efficiency), remain unknown. The authors lament in the discussion that “precise dose setting is difficult,” but the biggest cause of that lies in their own failure to quantify IL-10. Incidentally, in the single-vesicle-level analysis, particles carrying IL-10 and a tetraspanin simultaneously amounted to only 2.8% for CD9, 2.6% for CD63, and 2.1% for CD81. Particles that fit the picture of an “exosome loaded with IL-10” are only a tiny fraction of the population.
So how could the quality have been raised? The cheapest and most effective single move is to include an empty-vector control. Next, knocking out the IL10RA gene in the recipient macrophages would allow receptor dependence to be judged far more reliably than antibody blockade. Doses should be matched by particle number rather than by total protein amount, and comparing EC50 values across a dose series of five or six points instead of two would replace the vague phrase “2- to 14-fold” with a single interpretable number. Simply running one IL-10 ELISA across all preparations would by itself have advanced the discussion of dose setting and potency specifications enormously. Stability should be measured not by residual bands but by biological activity after stress, and in doing so the recombinant IL-10 side must be placed in the same buffer (trehalose, HEPES, BSA), or there is no telling “whether the EVs protected it or the stabilizers did.” And comparing the proteome and small RNA of naïve EVs against IL-10-positive EVs to identify the true nature of the anti-inflammatory activity of naïve EVs—that is the most natural next move, and one the authors themselves list as their next task.
Dr. Exotaro’s Perspective
I myself have continued to work on applying small extracellular vesicles (sEVs) derived from mesenchymal stem cells (MSCs) to the treatment of spinal cord injury (SCI) and neurological disease. So I could not read this paper’s “inconvenient results” as somebody else’s problem.
In a study we reported earlier, the MSCs administered intravenously never reached the injury site themselves, yet the sEVs those MSCs released inside the body did migrate to the injury site and associated specifically with the M2 macrophages there (Nakazaki, M., et al., Journal of Extracellular Vesicles 2021;10(11):e12137). When the sEVs were administered in doses divided over 3 days, TGF-β expression within the M2 macrophages rose significantly, permeability of the blood-spinal cord barrier fell, tight junction proteins (ZO-1, occludin, N-cadherin) increased, and in the end motor function recovery equivalent to direct administration of MSCs was obtained. Myeloid cells after spinal cord injury do not sort themselves into two boxes labeled M1 and M2; they move along a continuous spectrum, and steering that movement is the key to treatment.
What strikes me here is that what we used was an unmodified, natural MSC-sEV. Vesicles with no genetic modification and no drug loading moved the character of macrophages, stabilized the barrier, and brought about functional recovery. If that is so, then the fact this paper ran into—that EVs into which no gene was introduced quiet inflammation all on their own—is, at least for me, not a surprise but something that makes perfect sense. EVs are not “boxes.” The box itself is the drug.
At the same time, this paper also throws a hard question at our field. Is your control really the right control? When you load something into EVs and report that “it worked,” did you compare against unloaded vesicles administered at the same particle number? What fraction of the effect comes from the cargo, and what fraction from the vesicle itself? Papers that answer this question head-on are, including my own, by no means numerous.
Another thing that stayed with me is that this paper never once quantified IL-10. Something that should have required no more than running a single ELISA is missing. Turned around, that also means the field of EV therapy is still young enough that we have not even reached agreement on what ought to be measured. If we mean to get closer to the clinic, then before any story about how “it worked,” I believe we must not run away from the work of counting what is loaded and how much of it there is.
Finally, I do not want to end by running this paper down. A study that deliberately wrote a conclusion weakening its own claim, and put negative data even in the abstract, deserves trust on that basis alone. More often than one might think, it is this kind of “work that inspects the ground beneath our feet,” rather than reports of flashy therapeutic effects, that is holding up the field ten years later.