This article in 30 seconds
- Even after 24 hours of ultracentrifugation, roughly 70% of the RNA in bovine serum remained. In the actual measurements, only 19–33% of total RNA was removed.
- The cause is the principle of the method, not the performance of the machine. Ultracentrifugation separates by making heavy things sink, so RNA bound to proteins — a form that does not sink — cannot be scooped out.
- This is a measurement problem, not a safety problem. The authors state explicitly that no data show this RNA to be functional in recipient cells; the issue is interference with analysis.
1. What was measured — a test under harsher-than-usual conditions
In essence, this study is a test of how well the “remove it later” procedure actually performs, run under harsher conditions than usual. It appeared in Scientific Reports in 2016[1]. The procedure had three steps.
- Ultracentrifuge fetal bovine serum (FBS) at 100,000 g and 4℃, for 80 minutes, 5 hours, or 24 hours.
- After spinning, separate the pelleted fraction from the supernatant.
- Measure the total RNA in each and compare how much was removed.
In the field of conditioned medium (the whole of the liquid that remains after culturing cells and removing the cells), protocols that use serum pre-depleted of vesicles (EV-depleted FBS) are well established. This study directly measured how much such a protocol actually removes.
2. The result — 19–33% removed, more than two-thirds left behind
Twenty-four hours of ultracentrifugation removed 19–33% of the total RNA. Put the other way around, 67–81% stayed on the supernatant side — that is, in the EV-depleted FBS. The paper’s abstract sums the result up as follows.
“The majority of them (>70%) are retained even after extended ultracentrifugation in the preparations of vesicle-depleted FBS”
The majority of the FBS-derived RNA — more than 70% — is still there in the vesicle-depleted FBS preparations, even after the extended spin.
Wei et al. (2016), abstractTwo premises matter when reading this number.
First, 24 hours is a harsher condition than usual. According to the paper, most researchers prepare EV-depleted serum by ultracentrifugation at 70,000–110,000 g for 2–18 hours; 24 hours exceeds that. The reasonable reading, then, is that a standard protocol would leave even more behind.
Second, only total RNA was measured. Proteins, endotoxin (a fever-inducing substance originating from bacteria), and other contaminants were not measured. This is not a claim that 70% of impurities in general remain — the figure concerns RNA alone.
3. Why it cannot all be removed — what does not sink cannot be sunk
The conclusion is simple: this is a limit of the method’s principle, not of the machine’s performance. Ultracentrifugation separates particles by making them sink under their own mass. Membrane-enclosed vesicles sink; material in other forms does not.
The key point
Leave muddy water standing and the sand settles to the bottom — but the salt dissolved in the water never settles, no matter how long you wait. Ultracentrifugation is that settling process made vastly more powerful. It can sink the “sand” — the vesicles — but RNA dissolved in complex with proteins, the “salt,” will not sink however long you spin.
The authors, too, infer that much of the residual RNA exists as ribonucleoprotein complexes (RNA bound to proteins). Not being vesicles, they are inherently hard to bring down with a method built to sink vesicles.
In the actual measurements, 5–40 ng/mL of total RNA was recovered from untreated FBS, and preparations that had gone through EV depletion still contained an average of 34 ng/mL.
4. What happens when it stays — interference with analysis
The residual bovine RNA finds its way into your cell-side measurements. In a phrase: you lose the ability to tell what the cells secreted apart from what was in the medium to begin with. The paper demonstrated the following three points.
| Phenomenon | Details |
|---|---|
| Misread as human | Bovine RNA is rich in evolutionarily conserved sequences that match the human genome. Even under conditions allowing no mismatches at all, 13.2% of supernatant-side reads mapped to the human genome |
| Mistaken for cell-derived | miR-122, characteristic of the liver, was detected at high levels in conditioned medium from glioma cells, which are not hepatic. Fresh, unused medium, however, contained comparable levels — the source was the medium, not cellular secretion |
| Creeps into intracellular measurements too | miR-1246, which has no homologous sequence in the mouse, was reproducibly detected in mouse cell lines cultured with 10% FBS. Switching to EV-depleted medium significantly reduced the signal |
All three phenomena come down to that one point: the distinction is lost. When reading analytical data on conditioned medium, you have to ask whether a number really is of cellular origin.
At the same time — and with equal weight — it needs to be kept in view that this study did not demonstrate a safety problem. In the discussion, the authors state:
“As of today, there are no data suggesting the functional activity of this low-level bovine RNA in the recipient cells”
As of today, no data suggest that this low-level bovine RNA has any functional activity in recipient cells.
Wei et al. (2016), discussionThe issue is interference with the measurement system, not safety. That is: no longer knowing what you are measuring. Drop this distinction and you misread what the study means.
5. How far this generalizes — what the study does not show
The short answer: what this study supports is a claim about the limits of ultracentrifugation — no more. Three caveats deserve to be stated plainly.
- Only ultracentrifugation was tested. Subsequent work reported that an ultrafiltration-based protocol, which separates by membrane filtration, depleted vesicles more efficiently than overnight ultracentrifugation or commercial EV-depleted serum[2]. Generalizing to “no method can remove it” would be wrong
- The size of the effect varies with the experimental system. Because the amount of RNA cells secrete varies widely, interference is conspicuous in some systems and barely matters in others
- The results come from a single site and three lots. Differences in serum brand, lot, and centrifuge rotor remain as confounding factors
On the second point, the authors themselves write:
“In some cases, therefore, the anticipated contribution of vdFBS-containing media would be negligible, whereas in others, it may skew the results significantly”
The contribution of media containing EV-depleted FBS may therefore be negligible in some cases — and may skew the results significantly in others.
Wei et al. (2016), discussion6. Implications for design — remove it later, or never bring it in
What the study shows is that the “remove it later” approach carries a method-specific blind spot as a matter of principle. With ultracentrifugation, whatever does not sink remains. Switch methods and you merely change what slips through — the need to keep measuring how much was actually removed never goes away.
The other design is not to bring it in at all. Without bovine serum, the route by which bovine RNA gets in simply does not exist — and there is no removal rate to argue about.
The key point
Wiping spilled ink completely off the floor is hard — but if the ink never enters the room, there is nothing to wipe up. That is the difference between a design that keeps verifying how much was removed afterwards and a design that never lets it in.
Guidance from the Japanese Society for Regenerative Medicine likewise states that media free of animal-derived components should be used, and points out that when fetal bovine serum is used, vesicles of bovine serum origin may carry over and exhibit unexpected biological activity[3].
The full picture of the “leave it out” design is covered in Medium Design by Exclusion; how residual components differ by manufacturing approach is in Stem cell conditioned medium vs. secretome.
Frequently Asked Questions
Does “70% remains” refer to impurities in general?
No — the figure refers to total RNA only. RNA was the only thing measured in this study; proteins, endotoxin, and other contaminants were not measured. In the actual measurements, 19–33% was removed and 67–81% remained, which the abstract expresses as “more than 70%.”
Is the residual bovine RNA harmful to the body?
The study does not show that. The authors state explicitly that, as of today, no data indicate that this low-level bovine RNA is functional in recipient cells. The issue is not safety but interference with measurements — losing the ability to tell components the cells secreted apart from components that came with the medium.
Is bovine RNA impossible to remove by any method?
That is not the right conclusion. This study tested only ultracentrifugation. Subsequent work reported that an ultrafiltration-based protocol, which separates by membrane filtration, depleted material more efficiently than overnight ultracentrifugation or commercial EV-depleted serum. The accurate reading is that each method misses different things.
References
- [1] Wei Z, Batagov AO, Carter DR, Krichevsky AM. Fetal Bovine Serum RNA Interferes with the Cell Culture derived Extracellular RNA. Sci Rep. 2016;6:31175. doi:10.1038/srep31175
- [2] Kornilov R, Puhka M, Mannerström B, et al. Efficient ultrafiltration-based protocol to deplete extracellular vesicles from fetal bovine serum. J Extracell Vesicles. 2018;7(1):1422674. doi:10.1080/20013078.2017.1422674
- [3] Japanese Society for Regenerative Medicine (in cooperation with the Japanese Society for Extracellular Vesicles), Guidance on the Clinical Application of Extracellular Vesicles and Related Products, 1st Edition, April 30, 2024
The quotations in English are verbatim excerpts from reference [1]; the emphasized restatements are our own.
This article presents the findings of published research. Our stem cell secretome (the totality of components the cells secreted during culture) is supplied as a research-use reagent and is not intended to demonstrate efficacy against any specific disease. The studies discussed did not test our products.
View stem cell secretome specifications Medium Design by Exclusion