EMBO investigation yields two more retractions and three corrections for Voinnet

An investigation into the work of Olivier Voinnet by The EMBO Journal has led to another two retractions and three more corrections for the high-profile plant scientist, now suspended from the CNRS for two years.

According to the authors, Voinnet was responsible for some of the errors; all papers have been questioned on PubPeer.

The EMBO J, the flagship publication of the European Molecular Biology Organization, posted four notices earlier today and told Retraction Watch that the notice for the fifth paper would be posted by tomorrow.

This latest round brings our count for Voinnet up to 12 corrections, four retractions and one Expression of Concern.

These five papers have been cited more than 1700 times, according to Thomson Scientific’s Web of Knowledge.

Voinnet had previously promised to post retractions and corrections after they were critiqued on PubPeer and as part of the ETH investigation.

In both of the new retractions, the authors assert that the findings remain valid, either confirmed by “different laboratories” or “several independent studies.”

One of the new retractions is for a 1998 paper, “Viral pathogenicity determinants are suppressors of transgene silencing in Nicotiana benthamiana.” Commenters on PubPeer pointed out problems with the paper in January. They earned responses under the names of last author David Baulcombe and the head of the Sainsbury Laboratory, where one of the researchers used to work. The paper has been cited 827 times.

The retraction notice says the paper was “retracted by agreement between the authors, the journal Chief Editor and Head of Scientific Publications, EMBO, Bernd Pulverer, and John Wiley & Sons Ltd.” It includes an “Authors’ statement”:

The corresponding author was alerted through the PubPeer website to pixel pattern duplications in Figures 5B and 6I of this paper. The Figures address the identification of the 2b protein of cucumber mosaic virus as a viral suppressor of silencing.

The pixel pattern duplications originate from the incorrect use of image editing software to cover background marks. The image editing was done by O. Voinnet without the involvement or knowledge of other co‐authors. It involves regions of Figures 5B (Mock/GFP probe and CMV/CXMRNA2 probe) and 6I (TXMV‐2bΔ GFP probe) that are required for assessment of the data.

The identification of 2b as a suppressor of silencing was a valid interpretation of the original experiments, but the original data are not available for correction and the authors accept that integrity of the scientific literature requires retraction of the article. The silencing suppressor role of 2b has been confirmed in different laboratories. The analysis of the HCPro silencing suppressor in this paper is unaffected by the duplications in Figures 5B and 6I.

All authors concur with this statement and wish to apologize for having allowed the incorrect images to be published in The EMBO Journal.

The other retracted paper, “An endogenous, systemic RNAi pathway in plants,” was published in 2010 (and cited 148 times). Issues with this article were also highlighted on PubPeer earlier this year. First author Patrice Dunoyer said that he made the “errors” during the preparation of the study figures. Dunoyer and Voinnet (last author) took full responsibility.

The authors write in the retraction notice:

Following an investigation conducted by the The EMBO Journal, the corresponding authors were alerted of several image irregularities found in this paper. Patrice Dunoyer, who assembled the figures, acknowledges making the following errors during their preparation and that the other co‐authors were unaware of them:

Figure 3C (@GFP blot): Incorrect splicing of the blot led to a duplicated band improperly used for the final figure. Two dividers should have been also added to clearly indicate splicing of the original blot.

Figure 3C (@IR71 blot): Background duplication was done during image editing instead of putting a divider between the dcl234 samples and the col0 samples to indicate splicing of the original blot.

Figure 4A (@2039 blot): The blot shown in the right‐hand side panel of the figure (col0, O. pumila, C. rubella samples) is a duplication of the @2039 signal of the three‐first samples of the middle panel. In addition, duplicated sections in the middle panel @2039 were used to cover background noise. The first two and the last two lanes of the loading control (4A, middle panel) are duplicated and mirrored.

Figure 4D and Supplementary Figure S3: The rRNA loading controls displayed are not the cognate ones. Detection of miR159, whose accumulation is unaltered by the mutations shown in Supplementary Figure S3, provides an internal control for equal loading in these experiments.

All authors consider that the core conclusions of the published paper remain valid; graft transmission of endogenous siRNAs has been confirmed in several independent studies. However, the authors admit that integrity of the scientific literature is compromised by the data manipulation and are therefore requesting the article to be retracted.

Olivier Voinnet and Patrice Dunoyer take full responsibility for having allowed the publication of this flawed article. All authors concur with this statement and wish to apologize for the inconvenience caused.

The notice for the 2010 paper includes an “Editors’ statement,” in which they note the authors were able to provide images of the source data:

The authors provided to the journal images of source data to explain the majority of the serious image manipulations apparent in Figures 3C (GFP and IR71), 4A and 4D and Supplementary Figure S3.

In both retraction notes, the journal lists the other retraction along with an additional three papers, all of which are “subject to a corrigendum or to a retraction.” The remaining three papers have received corrections, and two of the three notices have already been posted.

In one correction notice, for a 2008 paper, “Nuclear import of CaMV P6 is required for infection and suppression of the RNA silencing factor DRB4” (also the subject of questions on PubPeer), authors outline the “errors” in figures, and note “the conclusions remain unchanged.” That notice also includes an “Editors’ statement”, in which one figure panel is retracted:

Figure 4D contained duplicated images derived from Figures 4A and 5A. Figures 4A and 5A have been verified, but since source data were not available for Figure 4D, this specific panel is herewith retracted. See author statement for further details.

The paper has been cited 83 times.

The other posted correction is for a 2002 paper, “Two classes of short interfering RNA in RNA silencing, ” which had also been flagged on PubPeer (and also received responses from commenters saying they were Baulcombe and the Sainsbury Laboratory head). This paper has been cited 670 times, according to Thomson Scientific’s Web of Knowledge.

In the correction notice, editors again retract one figure panel:

Our analysis revealed that Figure 4B contains two areas of selective background duplication, as shown in the enclosed image.

We note that Source Data for Figure 4B were not available from the authors. Since the main conclusions of this paper stand without this data, this specific panel is herewith retracted.

EMBO J will also post a correction for a 2012 article, “Differential effects of viral silencing suppressors on siRNA and miRNA loading support the existence of two distinct cellular pools of ARGONAUTE1.”

Commenters on PubPeer also raised issues with this paper in January. It has been cited 31 times, according to Thomson Scientific’s Web of Knowledge.

The journal’s editor-in-chief Bernd Pulverer told Retraction Watch that the final notice would be posted online by tomorrow, and explained more about how the issues came to light:

We first became aware of potential issues with the papers that are now subject to a corrigendum or retraction in January of this year via comments on the PubPeer website. As a matter of policy, we investigate all claims concerning image irregularities – including those made on PubPeer – through the image forensic tests applied to all manuscripts before publication at EMBO Press [for more details see recent Nature comment on this: http://www.nature.com/news/the-image-detective-who-roots-out-manuscript-flaws-1.17749].

We could identify a number of potential image manipulations in these papers and contacted the relevant authors of the papers in question in January 2015. The corresponding authors of these papers were highly cooperative and sent detailed responses. As this time, we became aware of a number of institutional investigations and contacted the relevant institutions. We also realized that a number of papers published in other journals might be affected and we contacted most of these journals in order to exchange information on policies and timelines.

He noted that the authors didn’t violate any official protocols in cases where original data were not available:

For many of the experiments in question, the authors were able to provide the original source data and in some cases the actual labbook entries. For a small number of figure panels such data was not available – it should be noted that some of the papers were published some time ago and there are no universal data retention policies – nor are these indefinite where they are available.

The journal worked with the authors to decide how to proceed with each paper, and opted for retractions when the issues appeared “widespread, likely intentional and affected key data essential for the main conclusions of the paper,” Pulverer added:

We communicated in some detail with the authors and agreed in each case on the best way to correct the scientific literature – again, the cooperation of the corresponding authors should be noted. We decided to issue a corrigendum in cases where the source data was available and the unmodified data was consistent with the conclusions made in the paper, in cases where the data in question were of relatively limited importance – that is they could be deleted without affecting the main conclusions of the paper in question. In cases where the manipulations detected were widespread, likely intentional and affected key data essential for the main conclusions of the paper, we agreed with the authors that outright retractions were appropriate as the confidence in the papers was fundamentally undermined. It should be noted that the literature has since produced data consistent with many of the findings, including those in the retracted papers. Our assessment was restricted to the papers in question and they were not influenced substantially by the wider literature. Importantly, this means that the conclusions of the retracted papers are not necessarily wrong, but rather that we have concluded that the data in the papers in question is too unreliable to be maintained in the peer reviewed quality literature….In a couple of figure panels we opted to issue a corrigendum, but we nonetheless retracted selected figure panels – in our view this is warranted in the absence of further reaching issues and since the main conclusions made in the papers were not affected.

Pulverer commended the authors and said the journal will issue an editorial that will provide more information:

In our view the authors showed exemplary cooperation during these long investigation. In our view the literature has now been corrected as best as possible in light of these most unfortunate and serious image manipulations. We will issue an editorial in the same issue as the correction and retraction statements, which will provide further details.

Voinnet won the Rössler Prize in 2013 and was awarded the EMBO Gold Medal in 2009.

We’ve reached out to Dunoyer and Voinnet for a statement and we’ll update the story with any response.

Like Retraction Watch? Consider making a tax-deductible contribution to support our growth. You can also follow us on Twitter, like us on Facebook, add us to your RSS reader, and sign up on our homepage for an email every time there’s a new post. Click here to review our Comments Policy.