Why novel dsRNA molecules in GM food are of little to no concern

Written by Alexander Huszagh

Recently, concerns were raised about the potential risks of dietary double stranded RNA (dsRNA) and microRNA (miRNA) molecules silencing human genes, after research by Zhang et al. showed the presence of plant miRNA in human blood plasma, as well as providing evidence that this plant miRNA enters the system by dietary uptake in mice. The group then demonstrated that this plant miRNA could silence genes in the mice, leading other researchers to separately raise concerns that diets consisting of genetically modified organisms could lead to the uptake of novel dsRNA molecules that could silence human genes.
Gene-silencing by RNA interference, or RNAi, typically occurs by way of short sequences of RNA which bind to a target messenger RNA sequence (mRNA) and inhibit it, either signaling the mRNA for deletion or inhibiting its expression. This occurs through perfect or near-perfect base pairing of the sequence to a short segment on the mRNA strand. Uptake of dsRNA or miRNA at levels that would lead to gene silencing would therefore be an important consideration in food safety, including the safety of GMOs.

Diagram detailing the creation of miRNA. From Wikimedia Commons.

However, recent studies have failed to reproduce these results and questioned the likelihood of dietary uptake of plant miRNA molecules in mammals. Witwer and other scientists from Johns-Hopkins showed that the plant miRNA discovered in human cells (miR160 and miR166) by Zhang et al. did not reflect the concentrations of miRNA in rice. In rice, miR166 levels were higher than miR160 levels, however, in humans, miR160 levels were substantially higher than miR166 levels. Witwer et al. showed that blood levels of these plant miRNAs were not appreciably changed after eating, and never resembled the levels present in the dietary sources. Reported levels of plant miRNA levels were also significantly lower and more variable than previously reported.
This raises questions of dose. Responses in biological systems vary dramatically with respect to dose, such as how drinking sea water can induce vomiting while a pinch of salt in a glass of water may be refreshing. Previous studies have established that levels of miRNA below 100 copies per human cell would have no impact on gene expression and therefore no regulatory capacity. Zhang et al. found 853 copies of miR160 per cell in mice, low compared to native miRNA levels but at levels with gene-silencing capability. The lower levels of miR160 reported by Witwer et al., along with the fact that plasma miRNA levels did not reflect the dietary sources and the high variation in observed plant miRNA levels makes it unlikely that dietary sources impact gene expression through miRNA uptake.
In addition, the amount of rice in the diet of the mice in the Zhang et al. study were abnormally high. The amount of rice fed daily to the mice would equate to a human eating 33 kg of rice/day, 50 times the daily amount of rice consumption in Brunei, which has the highest rice consumption per capita. Rice consumption per capita also sharply falls in other countries, making the diet in the study 70 times the daily consumption in Vietnam, the next country in terms of rice consumption. Since we can reasonably expect miRNA uptake levels to be proportional to the amount of rice consumed if diet is the primary source of these miRNA molecules, the reported 853 copies/cell would fall to an estimated 17 and 12 copies/cell in the average Bruneian and Vietnamese diets, respectively. The levels reported by Zhang et al. would there fall well outside the range of significant biological activity in a typical diet, and therefore should not impact gene expression under realistic circumstances.
Frank is looking for elusive effects of dsRNA in rice

The lower concentrations of reported by other researchers compound these concerns, and make it highly unlikely that any dietary sources of dsRNA would have any impact on gene expression. Furthermore, the risks presented by dsRNA and other forms of RNAi are not exclusive to transgenic organisms, nor are they substantially changed by the presence of transgenes. Non-genetically modified plants, such as in the study by Zhang et al., also produce dsRNA. The addition of one or multiple transgenes and the resulting additional dsRNA would not significantly change the amount of dsRNA our system digests daily, nor would these transgenes be more likely to inhibit human gene expression than genes native to a dietary source.
Overall, these studies show that it is highly unlikely that dietary sources of RNA could have biologically relevant impacts on human gene expression. Further studies may elucidate whether RNA sequences are actually assimilated into blood plasma from dietary sources, however, it is improbable that these RNA molecules would be present in sufficient quantities to have any detrimental (or beneficial) effect on humans. Ultimately, although some researchers may maintain concerns about the physiological role of dietary sources of RNA in humans, these claims are not well-established and can currently be considered improbable.


Zhang, L.; Hou, D.; Chen, X.; Li, D.; Zhu, L.; Zhang, Y.; Li, J.; Bian, Z.; Liang, X.; Cai, X.; Yin, Y.; Wang, C.; Zhang, T.; Zhu, D.; Zhang, D.; Xu, J.; Chen, Q.; Ba, Y.; Liu, J.; Wang, Q.; Chen, J.; Wang, J.; Wang, M.; Zhang, Q.; Zhang, J.; Zen, K.; Zhang, C.-Y. Exogenous plant MIR168a specifically targets mammalian LDLRAP1: evidence of cross-kingdom regulation by microRNA. Cell research 2011, 22, 107–26.
Heinemann, J. A.; Agapito-Tenfen, S. Z.; Carman, J. A. A comparative evaluation of the regulation of GM crops or products containing dsRNA and suggested improvements to risk assessments. Environment International 2013, 55.
Witwer, K. W.; McAlexander, M. A.; Queen, S. E.; Adams, R. J. Real-time quantitative PCR and droplet digital PCR for plant miRNAs in mammalian blood provide little evidence for general uptake of dietary miRNAs: Limited evidence for general uptake of dietary plant xenomiRs. RNA Biology 2013, 10.
AE, Isaacs SK, Baggish AL, Chan
SY. Ineffective delivery of diet-derived microRNAs to recipient animal organisms. RNA Biology
2013; 10:1107-16; PMID: 23669076; http://dx.doi.
Brown, B.; Gentner, B.; Cantore, A.; Colleoni, S.; Amendola, M.; Zingale, A.; Baccarini, A.; Lazzari, G.; Galli, C.; Naldini, L. Endogenous microRNA can be broadly exploited to regulate transgene expression according to tissue, lineage and differentiation state. Nature biotechnology 2007, 25, 1457–67.

Written by Guest Expert

Alexander Huszagh is a C++/Python developer with a Masters in cellular and molecular biosciences from UC, Irvine. He applies computational biology and mass spectrometry to the study of protein machines.


  1. have a look at the FSANZ response
    Response to Heinemann et al on the regulation of GM crops and foods developed using gene silencing
    (May 2013)
    Key points:
     A recent scientific article (Heinemann et al, 2013) claims that small double-stranded RNAs (dsRNAs) generated in GM plants as a result of using gene silencing techniques can create biosafety risks that are not being adequately assessed by regulators such as Food Standards Australia New Zealand (FSANZ). They suggest changes to the safety assessment process to address their concerns.
     FSANZ has carefully examined the arguments put forward in the article, and has thoroughly researched the scientific literature on gene silencing. The weight of scientific evidence published to date does not support the view that small dsRNAs in foods are likely to have adverse consequences for humans.
     In formulating their hypothesis, the authors have not taken into account the fact that small dsRNAs are ubiquitous in the environment and in the diverse range of organisms we consume as food, including plants and animals. This establishes a long history of safe human consumption which pre-dates the use of such techniques in GM plants.
     The authors failed to adequately acknowledge that developing oral therapies based on small dsRNAs targeted against human viruses and other diseases such as cancer has so far been unsuccessful because of the barriers that exist to their uptake, distribution and targeting within the body.
     The authors have also underestimated the strengths of the GM food safety assessment to detect possible unintended effects, including those that could arise from the use of gene silencing.
     There is no scientific basis for suggesting that small dsRNAs present in some GM foods have different properties or pose a greater risk than those already naturally abundant in conventional foods.
     The current case-by-case approach to GM food safety assessment is sufficiently broad and flexible to address the safety of GM foods developed using gene silencing techniques. This approach enables additional studies to be requested should that be necessary to further inform the safety assessment of a particular GM food.
     FSANZ will continue to monitor the scientific literature for any new developments which may be relevant to GM food safety assessment.

  2. Nice job on explaining the dosage issue. Most of the pieces I’ve seen didn’t go there.
    Stability is also important. How long can those molecules really persist? I know they claimed there was a special mechanism, other people were pretty skeptical of that too. Emily Willingham wrote a nice piece at the time that covered that:

    They are “very fragile,” he said. The idea that these RNAs would not be immediately destroyed by the digestive enzymes of the gut “is controversial, to say the least,” he added. “Hundreds of millions of dollars have been spent trying to stabilize single-stranded nucleic acids in the bloodstream, let alone the digestive tract, and unmodified RNA has never been found to survive.”

  3. I wanted to give the Zhang et al. the benefit of the doubt, since they did showed that there are 2′-methylated ends in human blood in 5/10 samples as well as the blood of rats fed rice. The 2′-methylation could give these miRNAs some meta-stability, and the exogenous miRNA levels in humans did not reflect the levels of miRNA expression in plants as well as stayed relatively constant before and after gavage (in terms of the number of quantification cycles needed until detection) in Witwer at al., so I did not feel it was a fair analysis to discount their results of exogenous miRNA uptake to contamination or instability. Therefore, there is some validity to the claim that exogenous miRNA can be absorbed into blood plasma by some mechanism, despite the obvious concerns raised by Witwer et al. and Snow et al.
    Therefore, with those parameters, I decided to consider if Zhang et al.’s conclusions were well established, what would be the impact on human gene expression under real conditions. Under real conditions, dose is of a primary concern, and the extremely low copies/cell under real dietary conditions makes it extremely unlikely that dietary sources of miRNA would impact human physiology. In short, I chose to consider an effectively worst case scenario, and yet there is little to no biologically relevant concern at these levels.
    Questions of stability and uptake are highly relevant to a discussion of the uptake, but I chose to focus primarily on the dose since 2 studies have found miRNA168a levels in mice. However, the dose would almost certainly be below biologically active levels, and therefore makes accidental gene silencing from a typical diet extremely unlikely even with all those assumptions.

  4. I like that you chose to explore the worst case scenario, but there is one generalization that I’m concerned with. While seems unreasonable that someone might eat 33kg of rice each day, might it be possible that a small subset of the population is 50 times more responsive to some particular miRNAs? Perhaps people taking drugs or with a pre-existing conditions that affect the same pathway?
    If plant regulatory RNAs are getting into people the effects would be unpredictable and could affect only a small number of people who, for whatever reason, respond differently than most people. The effects would be hard to identify with modern medical technology. If it were discovered, and the RNAs are native to the plant, It should be no different for the industry than discovering a new allergy. For GMOs, because of the hype around them, it would be nice to have a statistic like “this is unlikely to affect more than one person per million,” but I don’t know how one could screen for rare cases.

  5. I don’t think Zhang’s article should have passed peer review. For one, evaluate the corrigendum of figure 5. It severely undermines one of the main points, while the published western seems to show an effect (although it is presented without an adequate control, what is the baseline?), the full series undermines the figure. Also, the other westerns in the paper show an increase from baseline with the chow of LDLRAP levels. In particular look at Figure 6E, 6f, what is the control for change from chow? Worse, look at 6j, with nanomolar inhibition of this supposedly femtomolar effective miRNA, what is the return of function? Minimal, barely above noise. Totally unconvincing.
    The test diet increased the plasma levels of the mi168 sequence from 3.5 femtomolar to 6 femtomolar. No matter what the ultimate concentration was per cell, their test was showing a baseline of 50% of the concentration when supposedly exposed to the transgenic miRNA. What is up with their assay? I think they’re looking at noise. Then they;re adding picomoles of this miRNA to an animal that weighs 30gm and supposedly showing an effect when we know from actual studies on miRNA/siRNA efficacy that you need wildly higher concentrations to affect cells in culture. These are homeopathic doses of miRNA. This paper makes no sense.
    My read is poor review led to the publication of noise.

  6. Hi:
    In the rice/mice experiment, what was the effect of the silenced genes on the organism? and is the effect claimed to be permanent?
    apart from all the excellent criticisms, these questions seem pertinent.

  7. additional clarification:
    when I inquired if the silenced gene would stay silenced I should have said at what dose. so, if the dose is heavily reduced or discontinued?

  8. No, the effect is not permanent but considering if it was dietary due to a staple cereal, this could be a perpetuating phenomenon. In this study, they showed that mice and humans could have exogenous miRNAs in their blood serum, that these levels were at a biologically active level, and that these were regenerated when anti-miRNAs were used to inhibit the miRNAs present in the blood serum. They then showed that this miRNA could inhibit a cholesterol synthesis feedback mechanism by inhibiting LDL removal from the blood by rice feeding (the main packaging version of cholesterol to the cells), and this would lead to high cholesterol levels in the blood serum. They showed that this could be reversed with anti-miRNA (in short, that this is temporary).
    However, the diets are unnaturally high in rice, as I mentioned. There is also the question of replication, and their signal to noise ratio (as mentioned before) is very high. They had unreliable results, with many populations of humans showing no exogenous miRNA. Although I do feel the paper rightfully made it through peer-review (see this chart:
    http://www.nature.com/cr/journal/v22/n1/fig_tab/cr2011158f4.html#figure-title ), I feel their results are an anomaly. These effects would easily fall out of the biologically active realm with realistic dietary conditions, and the lack of replication further questions the validity of their results. It’s an interesting study, but as of now it is on shaky ground.

  9. I think it should have, but solely for the results in Figure 4 and Figure 5 (and likely should not have been published in Nature). Zhang et al. less so showed the uptake of exogenous miRNA from dietary sources reliably and more so showed an in vitro inhibition of endogenous gene expression from an miRNA found in human and mouse blood serum. Those are still interesting results.
    However, I agree with the rest being noise, and that the central tenant of this paper that exogenous miRNA can be absorbed into the bloodstream by diet is not supported by the data, which is also corroborated by Snow et al., 2013 and Witwer et al., 2013. Although the doses they showed are lower than that of miR16, they were still in “biologically active ranges”, however, this also came with the caveat of high signal-to-noise, unnaturally rice consumptions, among others.
    It is technically possible that the differences between Witwer et al. and Zhang et al. are via the mode of miRNA uptake, but that runs counter to the central theme of this paper. This paper argues it is via dietary sources, but they used a fresh rice type feed as opposed to a slurry. However, it is unlikely that significant quantities of the miRNA would transfer the cell membrane solely due to transient exposure, such as through the skin, or that this would penetrate to significantly raise levels in the blood serum.
    Basically, I am overall of the general opinion that this paper does not present a convincing argument for substantial uptake of miRNA into the bloodstream from dietary sources, nor of realistic gene silencing from exogenous sources. However, the other results are intriguing. Basically, I feel it should have been published in a lower journal, because it still posed interesting results (I’m of the general opinion that as long as there are not obvious, glaring flaws, or ethical dilemmas, publish first, interpret later is a good philosophy. It allows free diffusion of high-quality studies that vets for the base line, and then allows analysis from there. I feel this study makes the base line, but is hardly convincing beyond there).

  10. It wouldn’t be sensible that certain populations would be more sensitive to particular RNAs at the doses present. Although with imperfect base pairing (greater than 95% base pairing for 20+ nucleotides, or the near-perfect case), it is possible that SNPs (single nucleotide polymorphisms) could lead to different sensibilities to RNAi in different individuals. However, it is unlikely that drugs, or environmental factors, would play a substantial role in compounding these effects. This is due to the law of mass action: simply, there is not enough RNAi at these concentrations to inhibit endogenous gene expression. This is since there is not enough to cause probabilistic interactions with the mRNA sequences, or pre-mRNA sequences, or that there is not enough to cause a substantial decrease in gene expression. The overall effect of the RNAi would be unlikely to change with decreasing the amount of the target mRNA sequence in the first case, and still hypothetical in the second case. By reducing the amount of one reagent, the reaction slows. Likewise, interactions between the two with less of one reagent would be less likely than before, and therefore we can consider it improbably that at extremely low miRNA levels to have an unusual impact on certain individuals rather than others. Obviously, if the concentrations were above or around 100 copies/cell, we might have a different scenario, but one that would still have to be verified experimentally for validity.
    It is technically possible, but it isn’t very likely.

  11. just out of curiosity (non scientist speaking here), how different is the effect in this experiment on both the organism (mice) and the cholesterol synthesis feedback mechanism compared to bad diet–of the obesity producing variety?
    I’m wondering if temporary gene silencing leading to temporary ill health is not pretty typical with unhealthy diets? wondering if it’s only in the context of discussions of GMOs that what may be relatively common is turned into something creepy sounding.
    sorry for the potentially idiotic question, but… someone has to have this job.

  12. I will say this: the results are similar (in a way, but milder) to familial hypercholesterolemia. Basically, a receptor leads to lower LDL uptake into the cells, which means more in the bloodstream and overall more cholesterol synthesis. In short, it would be worse than diet, which has no effect up until a certain plateau before having (very) detrimental effects for increased cholesterol intake afterwards. THAT BEING SAID: The results do not support any of that as an actually biologically relevant result in vivo. They do not suggest that dietary RNAi can inhibit gene expression under in vivo conditions. In short, even if you choose to eat 33 kg of rice/day (115 thousand calories per day), it is unlikely that there will be any biologically active effect on you. Even in the worst case scenario, if all of Zhang’s results are corroborated, eating under 4 kg of rice/day (14 thousand calories) would lead to no biologically active effect.

  13. Just out of curiosity, what is the mouse equivalent of human eating 33kg of rice in a day? Indeed I doubt eating that much is feasible for a human so how was it (the equivalent amount) feasible for mice?

Comments are closed.