Overstating results, eLife, and “distributed viruses”

Ed Yong, a journalist who other scientists usually regard highly, just published a story “A New Discovery Upends What We Know About Viruses”, describing a paper published in eLife, that claims to show that “infection can operate at a level above the individual cell level, defining a viral multicellular way of life.”

The authors of the paper, the eLife Digest (it’s version of a lay summary), and Ed Yong all claim that this is a revolution in our understanding of viral infections, and even evolution. This is not true.

Now, there are articles like this published all the time in science, and journalists often fall for the claims made in press releases and overblown articles. But the reason I’m writing a blog post about this particular case is that it comes close on the heels of another article in eLife, an op-ed by Deputy Editor Eve Marder, claiming that scientists often overstate their findings, and that “rookie errors” like these are why ECR authored papers have lower acceptance rates at eLife. In another blog post, ECR authors overselling results was a complaint made by quite a few other eLife editors/reviewers. I and others have also heard eLife editors make this (as far as I know, evidence-free) case in person. On that occasion this was used to partly explain away the idea that name recognition biases publication success. In this post I try to show why this argument is flawed, taking the example of one recent article in eLife.

FBNSV is a nanovirus. (Source: ViralZone)

Back to the paper. First, I’m not disputing the results or importance of this paper, just the conclusion made by the authors that they present a new paradigm, and the mischaracterisation of prevailing thought in virology. The study looked at Faba bean necrotic stunt virus (FBNSV), a virus that’s comprised of 8 sub-genomes, each carried in separate viral capsids. Now, any high-school/undergraduate student of biology should know that viruses are basically infectious DNA or RNA that code for some proteins but also use a lot of host proteins to replicate. In lay terms, “viruses” refers to the entire DNA/RNA and protein complex, though more properly this is called the “virion”, while the word “virus” has a more complicated definition in science. I personally use the term virus interchangeably, though as a molecular biologist, more often to refer to just the DNA or RNA components within a host cell. Anyway, semantics.

FBNSV, as the name suggests, infects plants, and in this study the authors essentially wanted to answer the question of how a virus with seven independent, and yet co-dependent genomes, can infect its host. This is a fascinating question, though not nearly as unresolved as the authors make out. Indeed the key focus of the paper and its media coverage seems to be an antiquated view of viruses, suggesting that they infect individual cells one at a time. Obviously if you’re a virus that needs 8 different particles (or even 2–4 particles for that matter) to infect the same host cell, the odds of your survival are low, and yet such viruses do exist, and have been known about for decades.

Here, I’m first going to dissect the paper and then come back to my overall point. Feel free to skip down if you aren’t interested in the details of this specific study.

The authors begin by saying that virologists have, and still believe, that viruses infect individual host cells sequentially. They cite several papers, but the most recent citation is over 30 years old, and nevertheless doesn’t support the point they make.

The experiment they perform is to infect a host plant with the virus, and then probe the tissues of the plant with fluorescent-labelled probes specific to the 8 different DNA segments, and later with fluorescent-labelled antibodies to detect the presence of the different proteins encoded by each of the 8 different DNAs. It’s important to note that they do not at any point probe for the location of more than 2 DNA segments at a single time. So they’re actually visualising the presence of pairs of different DNA segments concurrently, but are blind to the location of the other 5 segments. They also admit that they did not test all of the 28 possible pairs, and instead only visualised 7 pairs in all. Their microscopy appears very well performed, and well-controlled. But that’s not the point of my critique.

They then probed cells that contained only the S segment of the viral DNA, to see if these cells also contained the M-Rep protein (coded for by the R segment), and as expected, 85% of them did. The M-Rep protein is essential for replication of all the DNA segments, so it’s unsurprising that it was present in cells that contained other DNA segments individually. How else would these DNA segments replicate?

In the discussion the author’s claim that they have discovered a new “concept in virology”, i.e. that some viruses can lead a “a multicellular way of life”, whatever that means. In my view, all they’ve done is shown visually something a lot of virologists already knew was likely happening.

What the evidence for this? Research over the last 20 years into a different group of plant viruses, the Geminiviruses (or more specifically the Begomovirus genus of family Geminiviridae). These viruses, like the nanovirus studied in the eLife paper, infect plants, and have a multipartite genome ( 2 different DNAs in their case). They are far more economically important, and in general are a larger family of viruses. As a result, they’ve been studied in far more detail.

Now in the geminivirus field it’s been relatively common knowledge that the two different genomes (A and B, with A coding for replicative proteins and B coding for movement proteins) do not exist in stochastically equivalent amounts in the host. I always assumed that virologists knew that the two individual virus genomes could be present independently in different tissues, i.e. that viruses could be “distributed” as Ed Yong put it (Edit: Dr. Ralf Reski, a senior professor at the University of Freiburg in Germany, echoed this sentiment on Twitter). In fact, four years ago, researchers from Holger Jeske’s group (a pioneer in geminivirus biology) performed a very similar experiment as that published in eLife with a model geminivirus showing this exact result. Here’s the paper: https://jvi.asm.org/content/89/23/11926.

The paper was published in the Journal of Virology, a less selective journal than eLife, and probably as a result the authors did not feel the need to overstate their results.

Results from Weigel et al., 2015, Journal of Virology, examining the two different genomes of a begomovirus. Note the “distributed nature of the two DNA segments”

In this earlier paper the authors present an intriguing and somewhat novel population genetic framework for understanding the replication of multi-partite viruses. In fact they even describe a similar study performed by another group that looked at interdependence in viral replication between two different viruses, and explain why their study is different.

This is how science is supposed to be written up, in my opinion, and hopefully also that of eLife’s editors.

Instead in eLife, the authors do not cite this paper, or other papers that have looked at the the question of multipartite virus replication. In fact the idea that all viral genomes do not have to be present in the same cell for DNA-virus infectivity seems rather established looking at earlier papers (RNA viruses are a different matter). And their comments are then further exaggerated by a famous science journalist, leading the public to believe that virologists have misunderstood the nature of viruses until today.

As those of us who write for the public know, this perception that science is in a constant state of revolution and that it is impossible to arrive at certainty in science are major hurdles in science communication. They are the same sentiments that lead to anti-vaccine movements and climate-change denialism.

And is my critique of this paper new? Actually no.

Thankfully eLife publishes the review reports and a quick glance at that section shows that my concerns about overstatement of the findings (though not as concretely as I have done here), were echoed by the reviewers. As the editor’s summary of the review process says, “All three reviewers, though, also consider that major claims, particularly proposing a new paradigm in (plant) virology based on a multicellular way of life, are not fully justified by the observations, and that the presentation and interpretation of the results should be toned down. This would not diminish the importance of the results, but it should help to highlight the observations that are new here.” However, the final paper remains an exercise in exaggeration and misinformation. And the Atlantic article covering it is a terrible example of science journalism (I suspect that in this case the journalist simply ignored the Reviewer comments which are placed at the end of every eLife article).

(Edit: I’m adding this paragraph as clarification in response to some specific comments from virologists on Twitter). In summation, earlier papers had already shown that one of the two DNAs that make up a multipartite virus (a Begomovirus) can move and replicate independently. This new paper shows that the other DNAs do too (or at least when comparing pairs, in a different family of multipartite viruses), and shows that the Rep protein moves from cell-to-cell. And yet the second paper claims to be the first to present a multicellular conception of life. This is incorrect in my opinion.

At the end of the day this is just one paper out of the hundreds that eLife publishes. The reason I pick on this however is that it shows that the perception of ECRs being prone to overstating their findings is untrue. Instead based on this example, it appears that every author should overstate their findings to publish in selective journals like eLife. It also shows how dangerous this pressure to exaggerate conclusions can be for public understanding of a scientific field.

I’m tired of hearing this excuse (that ECRs oversell results) by editors at eLife (and probably other journals) to explain the lower acceptance rate of papers from ECRs authors, and I for one want editors to think of this case the next time they feel the need to selectively disparage ECR authors and downplay the effects of name recognition and seniority in publishing.

Edit: Dr. Siobhan Duffy, one of the experts quoted in the Atlantic article, points out on Twitter that the eLife study is the first to demonstrate movement of the virus Rep protein, and the first to convincingly show DNA-B movement independent of DNA-A (in the previous Jeske study this was found only 8% of the time and not considered important). I agree with Dr. Duffy that these are important points of novelty of the paper, but they do not show that the authors are the first to conceptualise multicellular infection. In Dr. Duffy’s words, “The Montpellier virologists might not be the first to conceptualize a multicellular way of life for viruses, but I do think they provide the first strong evidence for it!”

Disclaimer: I have not published, nor tried to publish in eLife as yet. I’m a member of eLife’s Early Career Advisory Group but the opinions above are purely my own. My PhD involved studying ways to develop immunity to geminiviruses and hence my familiarity with the topic.