Authors withdraw already-corrected JBC paper questioned on PubPeer

An assistant professor of neuroscience at Tufts has withdrawn a paper published last year after one round of corrections wasn’t enough to fix all of the study’s problems.

The study, “Sirtuin 2 (SIRT2) enhances 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced nigrostriatal damage via deacetylating forkhead box O3a (Foxo3a) and activating Bim protein,” by Gizem Donmez and colleagues, had already been subject to an extensive correction in May:

In Figs. 2D, 3A, and 3E, the samples were run on gels with 18 wells, which contain repetition of some lanes and also combination of different experiments. The samples of interest were run on different parts of the same gel. To make it easy for the readers to interpret, the samples of interest were spliced together. However, the splicing was not made clear by insertion of dividing lines. Corrected versions of the figures are shown below with the insertion of dividing lines between the spliced lanes.

In addition, in Figs. 2D and 3A, the labels “NRS” and “SIRT2” on the top are supposed to show the NRS or the antibody that was used to coat the beads in the immunoprecipitation assay, respectively. The beads were coated with NRS or Foxo3a antibody. “Foxo3a” should have been used as a label and not “SIRT2.” This error has been corrected.

Due to an error during the preparation of Fig. 3F (right panel), a “no MPTP” sample was shown on lane 4 instead of an “MPTP” sample. This is now corrected in the figure shown below. The replacement figure is the original gel.

The corrected images in no way affect the conclusions of the paper or the original interpretation of the results. The authors apologize for any confusion caused by this error.

That correction raised some eyebrows at PubPeer:

..the “correction” would appear to be rather ill-conceived. Briefly, the problem with the original paper involved western blots which appeared to be spliced together, as indicated by sharp steps along the top and bottom edges of some blots. In addition, it appeared that the spliced-in bands were duplicated from elsewhere in the images.

In the correction, the undisclosed splicing is now shown by a series of black lines in the blots. However, these corrections completely fail to address the problem that the western blot bands on either side of the splice line are still identical. As such, the same band is used to represent biological samples of different origin. A key example would be the “corrected” Figure 2D, left panel, Foxo3a blot, right two lanes (spliced together). The right band appears to be a clone of the one immediately to its left. They are seemingly identical right down to the last pixel. If you don’t believe me, blow up the resolution and look at the series of small white spots underneath each band (tilting the screen back at an angle can help with contrast sometimes, to make accessory features more visible).

Donmez gave us this statement:

After the correction was published, it came to our attention that a few errors were made during the assembly of a few panels. We requested the withdrawal of the paper since a second correction could not be issued. These unintentional errors do not invalidate the results and conclusions of our paper. The results will be submitted for a publication at a future date.

The paper has been cited seven times, according to Thomson Scientific’s Web of Knowledge, including once by the correction and twice by other papers written by Donmez.

The PDF of the paper is marked “WITHDRAWN July 17, 2013,” but there’s no withdrawal notice. The JBC tells us that the notice — a disappointing ““This article has been withdrawn by the authors” will appear online on Friday, August 16th, when the abstract will also include a link to it. At that point, it will also be picked up by PubMed:

…but it will not be linked to the original article in PubMed right away.  I was told that a staff shortage at NLM is delaying data review to allow such linking.