Cut and paste and a PC crash: figure manipulations sink two papers

Two papers by an overlapping group of researchers in Italy have been retracted for manipulated figures.

In late 2013, perennial tipster Clare Francis sent their concerns about several papers, including the two that have been retracted, by authors who frequently publish together. One of the papers, in the Journal of Neurochemistry, is from a team led by Ferdinando Nicoletti; four other papers from the group have been criticized on PubPeer for image manipulation, which he addressed via email with us.

The second retracted paper, from the Journal of Immunology, has shares one author with the first: Patrizia Di Iorio of the University of Chieti, though according to Nicoletti she had no role in preparing the figures.

Here’s the April 2014 notice for “Neuroprotection mediated by glial group-II metabotropic glutamate receptors requires the activation of the MAP kinase and the phosphatidylinositol-3-kinase pathways” in the Journal of Neurochemistry. It’s behind a paywall, but the journal has assured us this is against policy and they will be fixing it shortly:

The above referenced article has been retracted by agreement between the Journal’s Chief Editor Jörg Schulz, John Wiley & Sons Ltd., and the following authors: F. Nicoletti, G. Battaglia, R.T. Ngomba, M. Storto, F. Orzi, A. De Blasi, P. Di Iorio, and V. Bruno. The authors M. D’Onofrio and L. Cuomo disagreed with the retraction. A. E. Kingston could not be tracked.

The retraction has been agreed due to concerns the Editorial Office was notified about relating to Fig. 1b (upper left panel) and Fig. 7b. In Fig. 1b, the spleen band of the immunoblot looked cut and pasted. The author confirms that the blot had been cut and pasted to show that the antibody was working.

Figure 7b. The figure shows the original immunoblot which parts of Fig. 7b of the retracted article were compiled from. The blot had been repeated twice. Lanes 1,6=Basal; 4,11= NMDA; Lane 1 of the original blot corresponds to lane 1 in Fig. 7b (Basal); Lane 11 of the original blot corresponds to lane 2 in Fig. 7b (NMDA); Lane 7 of the original blot corresponds to lane 4 in Fig. 7b (NMDA+LY379268+PD98059); Lane 5 of the original blot corresponds to lane 5 in Fig. 7b (NMDA+PD98059) Lane 3 of Fig. 7b was imported from the second blot (not available). The three NMDA + LY379268 lanes in the original blot (lanes 2, 3, 10) are clearly darker.

Further concerns related to Fig. 7b in which the middle lane (NMDA+LY379268) looked cut and pasted. The authors concluded from the immunoblots of TGF-beta, from which four of five bands were shown in Fig. 7b (see figure), that the combination between NMDA and LY379268 enhanced TGF-beta1 levels in the striatum, and that ERK/PI3-kinase-mediated induction of TGF-beta contributes to the ability of metabotropic glutamate receptors to promote neuroprotection against excitotoxic lesions. The original Western blot is shown with the lanes that correspond to the published figure. The (NMDA+LY379268) lane shown in the published figure had been imported from a different blot, the original blot of which could not be tracked.

The same held for the matching beta-actin lanes for either blot. Although the authors provided evidence that the principal findings and conclusions were confirmed in other experiments, in the figure in question the difference between the bands appears stronger than is visible in the Western blot the remaining bands had been imported from.

The senior author apologises and acknowledges the individual who brought these problems to the attention of the Journal.

This is the best image of the figure we could find. The top arrow is labeled “Band numbering in Fig 7b” and the bottom arrow is labeled “Band numbering in the original blot.”

The paper has been cited 95 times, according to Thomson Scientific’s Web of Knowledge.

Nicoletti gave us more details, including why two authors disagreed with the retraction, and what’s going on with PubPeer:

The reason why the paper has been retracted is that a single band in the representative blot shown in Fig. 7b had been imported from another blot. The blot shown in the Figure is not associated with a graph reporting densitometric data and no statistics was applied to the data. In other words, this was only a representative image. We found only one of the two blots (the one incorporating all other bands), whereas the second blot was missed (after almost 15 years). At that time, we did not think it was necessary to report in the Figure legend that data from the two blots were combined into a single Figure, and this was our mistake. The finding has been reproduced several times and we provided evidence from additional blots that data were real.

The first two Authors, who left the lab many years ago , felt that this was not sufficient to justify a retraction. My point of view (I was the lab leader, but not the corresponding Author) is that we made a mistake and, therefore, I decided to accept the retraction. I extended my apologies to the Editors and the Ethical Committee of the Journal of Neurochmistry for the time they have spent in discussing our case for many months before taking the final decision of retracting the paper. All this has been very stressful to me and I consider the case definitively closed.

He also talked about the PubPeer comments:

One of these papers is from another group (last name, Caciagli). There, we only provided drugs for the experiments. For the Plos One paper, we already replied to the Journal demonstrating that everything was fine, but we had no feed-back so far after many months. The only manuscript where there might be a mistake in mounting the Figures was the NeuroOncology manuscript. I did not check this carefully because we had no notification by the Journal.

I am sure that my collaborators are honest and, if something happened, this was by mistake.

Here’s the January 2015 notice for “Guanosine inhibits CD40 receptor expression and function induced by cytokines and β amyloid in mouse microglia cells,” in the Journal of Immunology, which was also subject to a figure correction in 2007 (also paywalled):

We wish to retract the article titled “Guanosine Inhibits CD40 Receptor Expression and Function Induced by Cytokines and β Amyloid in Mouse Microglia Cells” by Iolanda D’Alimonte, Vincenzo Flati, Mariagrazia D’Auro, Elena Toniato, Stefano Martinotti, Michel P. Rathbone, Shucui Jiang, Patrizia Ballerini, Patrizia Di Iorio, Francesco Caciagli, and Renata Ciccarelli, The Journal of Immunology, 2007, 178: 720–731.

A band in Fig. 8C appears to have been reproduced in at least five lanes. Due to a short circuit in the electrical power that irreversibly damaged various devices including PCs for data storage in their laboratories, the authors were unable to produce the original data to refute the concerns of inappropriate manipulation of the data. Nevertheless, the undersigned authors view any manipulation of data as unacceptable and regret that, without the original files, it was not possible to identify the person(s) responsible.

Hence, the article is being retracted with the consent of the authors listed here. Based on other data provided by the authors, the editors are satisfied that the findings reported in the article are nonetheless valid.

Mariagrazia D’Auro and Patrizia Ballerini did not sign the notice. The paper has been cited just once, by authors from the same group.