Olivier Voinnet, a high-profile plant scientist at ETH Zurich, has earned a mega-correction. It wrapped up a rough year for the biologist, which included his seventh retraction, and a CNRS investigation that found evidence of misconduct.
This latest correction, to a paper on the mechanisms behind RNA silencing in Arabidopsis, was published in RNA. The 2007 paper has been cited 101 times, according to Thomson Scientific’s Web of Knowledge. The corrigendum modifies three figures in total.
The notice is long, so we’re not going to post the whole thing here. The first error in “Transitivity in Arabidopsis can be primed, requires the redundant action of the antiviral Dicer-like 4 and Dicer-like 2, and is compromised by viral-encoded suppressor proteins” is a clarification to a legend:
The original legend for Figure 1, panel C could potentially be misleading and is therefore corrected as follows…
The rest results from “mounting mistakes:”
A comparison of the published figures and the original raw data has revealed several mounting mistakes, which have been shared with the editor-in-chief and systematically corrected based on the original files provided.
The note explains how each figure was affected:
- In Figure 2, panel D a “blot used to mount this panel has been spliced to remove unnecessary data but this operation was not disclosed.”
- In Figure 2, panel E a “GFP control for this figure comes from a separate blot, which was not specified in the published figure, creating an offset in the rRNA loading control.”
- Figure 3, “was originally believed to depict the same blot as in Figure 2E, stripped and rehybridized for various endogenous small RNAs. This led to the use of the same rRNA loading control as in Figure 2E…However, inspection of the original data revealed that none of the VSRs used in the experiments depicted in Figure 3 were in the GFFG–GFP background used in Figure 2E.”
- In figure 4, the “blot used to mount this panel has been spliced to remove unnecessary data but this operation was not disclosed.”
The correction also notes:
All of the authors have approved these corrections and apologize for any resulting inconvenience.
The editor in chief of RNA, Timothy Nilsen, told us that the decision to correct the paper
was made by me and only me without consultation with anyone at CNRS or ETH. Dr. Voinnet requested the opportunity to correct misrepresentations that were present in his 2007 paper. I granted that request because I believe that these sorts of things should be aired publicly in front of the scientific community and not anonymously or behind closed doors. It is up to the scientific community to make their own judgments regarding the credibility of the correction. The decision by RNA (me) in no way was intended to either explicitly or implicitly condone the actions of Dr. Voinnet.
He also told us:
A full retraction of the work was neither proposed nor discussed.
In the meantime, we’ve uncovered two other corrections Voinnet issued earlier this year. The first paper, “Isoprenoid biosynthesis is required for miRNA function and affects membrane association of ARGONAUTE 1 in Arabidopsis,” was published in PNAS and cited 37 times, had problems with a figure. Here’s more from the correction note, issued in May:
It has been brought to our attention that a loading control panel in Fig. 5C had been incorrectly assembled. We have verified that mistakes were indeed made in assembling the loading control panels in the upper half of Fig. 5C. We have retrieved the original exposures of the Western blots and Coomassie-stained membranes. The original data show that the conclusion drawn on the basis of the data inFig. 5C, that in inflorescences, ago1-38 mutant protein is less abundant specifically in microsomal fractions, remains valid.
The other unearthed correction is for “Selective autophagy degrades DICER and AGO2 and regulates miRNA activity,” published in Nature Cell Biology and cited 69 times. Here’s the correction note, which was published in July:
In the version of this Letter originally published, the TUBA immunoblotting panel and the Coomassie-stained panel was used in Figure 5c (in the dataset corresponding to let-7 antagomir treatment, top set of panels), and reused in Figure 5e, without appropriate acknowledgement. The Coomassie-stained gel was vertically flipped in Figure 5e but the alignment of the lanes was maintained. The TUBA immunoblot and Coomassie-stained membranes represent experimental controls. The p62 panel in Figures 2j and 2k (CQ-treated) was also reused without appropriate attribution. In all cases of reuse of blots between panels, the samples were obtained within one representative experiment and processed in parallel.
The authors confirm that all instances of vertical splicing of lanes, for example in Figs 1 and 3, were carried out in full compliance with the journal guidelines. All spliced samples were collected and processed in a single experiment.
The original publication was missing Supplementary Fig. S3 containing the uncropped scans of the blots; this has now been included online.
We’ve also found three corrections on two papers from years past:
-
“Multivesicular bodies associate with components of miRNA effector complexes and modulate miRNA activity,” published in Nature Cell Biology, and was corrected in 2009
-
“Small RNA duplexes function as mobile silencing signals between plant cells,” published in Science; corrected in 2010, and in 2011
By our count, Voinnet has a total of 18 corrections for 17 papers. One of his papers was corrected twice; our correction count also includes three papers that were corrected and later retracted.
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.
Two blogs that recently discuss the Voinnet case:
http://science21.blogs.courrierinternational.com/archive/2015/12/29/l-affaire-d-olivier-voinnet-et-le-fonctionnement-de-la-reche-51306.html
https://forbetterscience.wordpress.com/2016/01/21/olivier-voinnet-case-correcting-the-uncorrectable/
A correction for a Voinnet paper appears in Science today.
Original:
Hierarchical action and inhibition of plant Dicer-like proteins in antiviral defense.
Angélique Deleris, Javier Gallego-Bartolome, Jinsong Bao, Kristin D Kasschau, James C Carrington, Olivier Voinnet, Science, 313 (2006)
http://science.sciencemag.org/content/313/5783/68
Erratum:
Science 22 Jan 2016:
Vol. 351, Issue 6271, pp.
DOI: 10.1126/science.aaf2336
http://science.sciencemag.org/content/351/6271/aaf2336
“In the Research Article “Hierarchical action and inhibition of plant Dicer-like proteins in antiviral defense,” the following mistakes were made by Olivier Voinnet, the corresponding author, during the final stages of figure mounting. The original lab book data provided to the editors of Science showed that these errors did not alter the data in any material way that could be construed to benefit the results and their conclusions. Olivier Voinnet takes the full responsibility for the mistakes. All the authors have approved the following corrections and wish to apologize deeply for the inconvenience caused.
Figure 1D depicts TRV-PDS infections of various Arabidopsis silencing mutants. The published Panel 1D was inappropriately mounted with the wrong ribosomal RNA loading control and this has created confusion in its final layout, in which the left-hand side of the blot, devoid of RNA, was wrongly set as a mock (i.e., non-infected) track. However, a cognate mock had been purposely prepared from a separate virus infection experiment alongside its corresponding siRNA blot loading control. We have retrieved all the original data intended to mount Panel 1D and have produced a corrected panel. The correction has no bearing on the original conclusions that distinct single Dicer mutations affect specifically the accumulation of some virus-derived siRNAs, but that these changes in siRNA accumulation cause no noticeable effects on virus (TRV-PDS) RNA accumulation.
[INSERT CORRECTED Fig. 1D HERE]
Figure 1E depicts TRV-PDS infections of various combination Arabidopsis silencing mutants. The published panel was inappropriately mounted with the wrong ribosomal RNA loading control. In the mounting, this has caused a dcl4 mutant track to be mistakenly used in place of a wild-type track. A vertical line to signal gel splicing did not separate this lane. We have retrieved all the original data intended to mount Panel 1D and have produced a corrected panel. The correction has no bearing on the original conclusions that distinct double Dicer mutations affect the accumulation of specific virus-derived siRNA species, revealing a hierarchical action among the Dicer enzymes. In particular, the dcl2-dcl4 double mutation markedly enhances TRV-PDS accumulation and concomitantly eliminates virus-induced gene silencing of PDS (panel 1L), highlighting the key, redundant contribution of DCL2 and DCL4 to antiviral RNAi through the production of 22-nt and 21-nt virus-derived siRNA, respectively.
[INSERT CORRECTED Fig. 1E HERE]
Figure 2A is a composite of several separate Northern blots and was found to be erroneously assembled:
(i) The published left panel of Fig. 2A is a section of an original TCV siRNA Northern Blot for which the wrong rRNA loading control was used. We have retrieved the correct loading control for this panel and have amended it accordingly, also specifying the correct separation of samples according to loading control preparation. The amendment does not alter the original conclusion that TCV-derived siRNAs are mostly 22-nt in length, and produced in a strict DCL2-dependent manner.
(ii) The published right panel of Fig. 2A was incorrect as it contained the wrong rRNA loading control and incorporated tracks from the original siRNA blot not corresponding to the labels indicated on the panel. We have retrieved all the original data intended to mount this panel and its cognate loading control and have produced a corrected panel. Despite the nature of the mistake, the amended panel does not contradict the original conclusions drawn from the analysis of single dicer mutants (Fig. 2A, left panel) and confirms that DCL2 is the sole contributor to TCV-derived siRNA production. In particular, the dcl3-dcl4 double mutant, where DCL2 is still active, does accumulates high levels of 22-nt siRNAs, the diagnostic products of DCL2.
(iii) The published bottom panel of Fig. 2A, which depicts a Western analysis of the TCV P38 protein, was initially split to accommodate the wrong loading controls of each upper panel. We have retrieved the original Western blot and loading controls, and have produced a corrected panel. The data from the amended P38 Western blot concur with the original conclusions that none of the single or double dicer mutations affects TCV accumulation noticeably.
[INSERT CORRECTED Fig. 2A HERE]
Three panels in the Supplementary Materials were also affected. A revised SM file has been posted with the corrected figures, described here.
Figure S1D had a mounting error in which a cognate siRNA panel (upper panel, comparing TRV-derived siRNA species in various rdr mutants) was mistakenly combined with the wrong viral RNA Northern blot and loading controls. We have retrieved all the original data intended for panel D and have produced a corrected panel. The results in the amended figure show that none of the single rdr mutations of Arabidopsis affects the accumulation of TRV-PDS, which is in agreement with the original statement made in reference to Figure S1D. Thus, the mounting error and its correction have no bearing on the original conclusions that single rdr mutations do not overtly affect TRV-PDS accumulation or virus-derived siRNA production.
Figure S1F did not depict the Northern blot intended for publication. We have retrieved all the cognate data prepared originally to mount panel F and have produced a corrected panel. The correction has no bearing on the original conclusion that TRV-PDS accumulates more in the dcl2-dcl4 double mutants, in which virus-induced gene silencing of PDS is concomitantly alleviated. Virus accumulation is even more pronounced in the dcl2-dcl3-dcl4 triple mutant but sporadic VIGS is observed in that case, again agreeing with the original statements.
Figure S2A was assembled with the wrong Northern blot. We have retrieved all the cognate data prepared originally to mount panel A relating to the absence of detectable effects of single rdr mutations on TCV siRNA accumulation. These data were part of the original blots used to correct Fig. 2A (left and right). The correction has no bearing on the original conclusions.”
A correction for a second Voinnet paper, this time from 2010, appears in Science today.
Original
Small RNA duplexes function as mobile silencing signals between plant cells
Patrice Dunoyer, Gregory Schott, Christophe Himber, Denise Meyer, Atsushi Takeda, James C Carrington, Olivier Voinnet, Science, 328 (2010)
http://science.sciencemag.org/content/328/5980/912
Science 14 May 2010:
Vol. 328, Issue 5980, pp. 912-916
DOI: 10.1126/science.1185880
Erratum
Science 22 Jan 2016:
Vol. 351, Issue 6271, pp.
DOI: 10.1126/science.aae0387
http://science.sciencemag.org/content/351/6271/aae0387
“In the Report “Small RNA duplexes function as mobile silencing signals between plant cells,” the following mistakes in Figs. 1E and 3A were made during the final stages of figure mounting. The original lab book data provided to the editors of Science showed that these errors did not alter the results and their conclusions. Patrice Dunoyer, who assembled the figures, takes full responsibility for these mistakes. The figures have been corrected in the PDF and HTML versions of the Report online.
Figure 1E depicts @HA immunoprecipitation experiments performed on SUC:P19HA transgenic lines in the SUC:SUL RNAi reporter system. The published low–molecular-weight Northern blot depicted in this figure panel was inappropriately mounted with the wrong rRNA loading control, which was mistakenly reused from a previous publication [P. Dunoyer et al., Nat. Genet. 39, 848 (2007)]. In addition, a divider should have been added to clearly indicate splicing of the original blot between the depicted “Total RNA” and “IP@HA” samples. The authors have retrieved all of the original data intended to mount Fig. 1E and have produced a corrected figure panel. The correction has no bearing on the original conclusions that the extent of SUL-silencing suppression correlates with the extent to which the 21-nt SUL siRNAs are sequestered by P19.
Figure 3A depicts @AGO1 immunoprecipitation experiments performed on SUC:P19HA or SUC:P19 transgenic lines in the SUC:SUL reporter system. The published low–molecular-weight Northern blot depicted in this figure panel (left side; SS, Suc-P19#10 samples) was inappropriately mounted with the wrong rRNA loading control. In addition, a divider should have been added to clearly indicate splicing of the original blots between the depicted “Total RNA” and “IP@AGO1” samples. The authors have retrieved all of the original data intended to mount Fig. 3A and have produced a corrected figure panel. The correction has no bearing on the original conclusions that approximately half of the 21-nt SUL siRNA pool is sequestered away from AGO1 by phloem-specific P19 and that this sequestration is sufficient to prevent SUL-silencing movement.
Two figure panels in the Supplementary Materials were also incorrect. A revised SOM has been posted with the corrected figures, described here.
In fig. S2B, the published low–molecular-weight Northern blots were inappropriately mounted with the wrong rRNA loading controls. In addition, for the “@SUL” and “@159” low–molecular-weight Northern blots depicted on the right side, a divider should have been added to clearly indicate splicing of the original blots between the “SS” and “SS/SUC-P21HA#3” samples. We have retrieved all of the original data intended for fig. S2B and have produced a corrected figure panel. The correction does not alter our results, as the original rRNA data provide equally good support for equal loading.
In fig. S2C, the published low–molecular-weight Northern blot contains a duplicated section of the original blot, made in order to create an artificial spacer between the “Total RNA” and “IP@HA” samples. We have retrieved the original data intended for fig. S2C and have produced a corrected figure panel. The correction has no bearing on our original conclusions that, as seen with P19, phloem-specific expression of P21 prevents SUL-silencing movement through sequestration of the 21-nt SUL siRNA pool.”
A third Science paper, also from 2006, is corrected today.
Original
A plant miRNA contributes to antibacterial resistance by repressing auxin signaling”
Lionel Navarro, Patrice Dunoyer, Florence Jay, Benedict Arnold, Nihal Dharmasiri, Mark Estelle, Olivier Voinnet, Jonathan D G Jones, Science, 312 (2006)
http://science.sciencemag.org/content/312/5772/436
Science 21 Apr 2006:
Vol. 312, Issue 5772, pp. 436-439
DOI: 10.1126/science.1126088
Erratum
http://science.sciencemag.org/content/351/6271/aae0382
Science 22 Jan 2016:
Vol. 351, Issue 6271, pp.
DOI: 10.1126/science.aae0382
“In the Report “A plant miRNA contributes to antibacterial resistance by repressing auxin signaling,” a loading control panel in Fig. 2A was incorrectly assembled. The authors have retrieved the original small RNA blots with their cognate rRNA loading controls. The original data show that the Report’s biological conclusion (miR393 is induced in response to the active flagellin-derived peptide flg22) remains valid. Patrice Dunoyer, who assembled Fig. 2A, takes full responsibility for this mistake. The corrected figure panel and caption follow.
Northern analysis of miR393 (left panels) and miR171 (right panels) upon treatment with flg22 (upper panels) or flg22A.tum (bottom panels). Col-0 seedlings were treated with either 10 μM flg22 or 10 μM flg22A.tum. The same membrane was first probed with a miR393 probe (left panels) and subsequently stripped and reprobed with a miR171 probe (right panels). rRNA, ethidium bromide staining of ribosomal RNA; R, miRNA signal ratio between flg22-treated versus flg22A.tum-treated samples at each time point.”