It was the last study ever published from prominent scientist Gerd Maul’s lab. And now it’s been retracted.
Maul was a highly cited cell biologist, with 30 papers cited at least 100 times, according to Thomson Scientific’s Web of Knowledge. He was also a well-known sculptor. When he died in 2010, he had a paper under submission at Cancer Research, which was published late that year.
This year, however, his co-authors found problems with the study. And as the notice — one of the most detailed we’ve ever seen — explains:
The authors wish to retract the article entitled “Sp100 as a Potent Tumor Suppressor: Accelerated Senescence and Rapid Malignant Transformation of Human Fibroblasts through Modulation of an Embryonic Stem Cell Program,” which was published in the December 1, 2010 issue of Cancer Research (1).
This was the final article from Dr. Gerd Maul’s laboratory. Dr. Maul passed away before the final review and acceptance of this article and his laboratory was subsequently closed. Dr. Louise Showe took the responsibility to complete the submission and now takes the responsibility for the retraction.
This article documents a program of accelerated senescence in the normal human fibroblast cell line, BJ, when the expression of the four major Sp100 gene isoforms is suppressed by appropriate lentiviral vectors (BJ-S cells). This senescent program is associated with a highly significant, transient expression of an embryonic stem cell signature. The article further reports that BJ-S cells lacking expression of all four major Sp100 isoforms, subsequently give rise to rapidly dividing “small” cells with malignant potential (BJsm cells). The malignant potential of the BJsm cells was confirmed in the study by their ability to form tumors in nude mice.
During the past 6 months, the authors have undertaken to repeat the studies described in Negorev and colleagues’ work and have uncovered problems that bring into question the conclusions of the publication that identify Sp100A as a tumor suppressor, which when silenced leads to the malignant transformation and the appearance of the BJsm cells. After several months of unsuccessful efforts to repeat this observation, the authors carried out microsatellite studies on various liquid nitrogen stocks of the BJsm “small malignant cells” archived over a period of approximately 2 years by the Dr. Gerd Maul’s laboratory. As reported in the article, these cells lack expression of Sp100 and in addition do not express T antigen. However, microsatellite studies conclusively show that these BJsm cells are not related to the BJ primary fibroblast line but are related to HEK293 cells. Unfortunately, the Dr. Gerd Maul’s laboratory also worked with a spontaneous HEK293 variant cell line, 293-S, described in several of their previous publications. These cells lack expression of Sp100 and T antigen. To further confirm that the BJsm cells described in Negorev and colleagues’ work were 293-S derived, we also carried out microsatellite analysis on DNA from the 9 frozen tumors derived from the BJsm cells. These studies confirmed that the BJsm cells injected into the nude mice were 293 cells.
To further confirm the identification of the BJsm cells as 293-S, the authors also compared the microarray gene expression data reported in Negorev colleagues’ work with microarray data in Gene Expression Omnibus (GEO) derived from independently published studies from other laboratories on BJ cells and 293 cells. This comparison also supports the conclusion that the BJsm microarray data are most consistent with published data derived from 293 cells.
Although the investigators took precautions to eliminate 293T cells as a possible contaminant, they did not take into consideration the contamination by the variant 293-S cells, which lack both T antigen and Sp100 expression. Although the authors can only speculate on the source of the contamination, the observation that BJsm cells appeared in only 5 of 8 experiments was likely due to the way selection was maintained and the different lentiviral stocks prepared for each experiment.
The microarray expression data associated with this publication in GEO (GEO entry: GSE20613) has been modified and the erroneous BJsm data has been removed. The microarray data for BJ-V control and the BJ-S cells remains publicly available.
All authors are aware of and agree to this retraction.
The study has been cited five times, according to Thomson Scientific’s Web of Knowledge. Kudos to Showe and her colleagues for painstakingly attempting to replicate the work, and for explaining the results in detail. The retraction — which as far as we can tell should not cast even a tiny shadow on the rest of Maul’s work — could not have been easy.
Showe tells us:
The retraction was not undertaken lightly. But…it was the right thing to do. Gerd was a friend as well as a respected colleague. As a measure of that respect for him and his work. we went through considerable effort to be sure our conclusions leading to the retraction were solid, before providing those conclusions to the scientific community.
We’ve covered several posthumous retractions, involving late researchers at Mount Sinai, the Mayo Clinic Arizona, and the University of Pittsburgh.
I wonder if it’s detailed because the “main author” is not around? That said, this is a pretty important retraction; retracting it will save some National [insert govt institute here] money that may pursue a line of inquiry that is wrong.
Sad and credit due to authors for tough decision. A minor point but it makes one wonder more generally if family of deceased authors should be consulted in such cases (perhaps they were). Closing statement should perhaps technically be “All living authors are aware of and agree…”
A good point. But when the manuscript was submitted, it usually requires a declaration that all authors have seen and approved the final version. Likewise, the declaration should be amended too? At the very least, there should be a footnote by the author’s name.
Nevertheless, no doubt a tough decision for the authors. Both for the initial submission and the subsequent retraction.
Paper with 3 of the same authors.
Genes Dev. 2002 Apr 15;16(8):919-32.
Figure 1.
http://genesdev.cshlp.org/content/16/8/919/F1.large.jpg
Figure 1C. It is only a little thing, but the left C643A colony and left W664A colony are highly reminiscent of each other.
David seems to be on to something. I just photoshopped that figure and I see what he means. I copied C643A left column from fig1C and stretched it to the same dimensions as W664A and found that they bear an uncanny resemblance. Even a slight section of dark blue at the top matches a cut and paste job…
Indeed. The right-hand P654C colony also has an awfully straight line underneath it…
Genes Dev. 2002 Apr 15;16(8):919-32.
Figure 2.
http://genesdev.cshlp.org/content/16/8/919/F2.large.jpg
Figure 2C. H3 panel. How different are the left 2 bands? How different are the right 2 bands?
Figure 2E. SETB1 Western blot. Vertical change in background between lanes 1 and 2. Vertical, light streak to right of band in middle. Between the middle and right bands there is a vertical, light streak.
Figure 2E. Histones panel (lowest panel). How different are the bands in the middle 2 lanes from the bands in the right 2 lanes?
Figure 4.
http://genesdev.cshlp.org/content/16/8/919/F4.large.jpg
Figure 4A. Western blots.
SETB1 panel. Vertical, straight streak between lanes 3 and 4 and a change in background.
HDAC2 panel. Vertical change in background between lanes 1 and 2.
AcH3 panel. Vertical change in background between lanes 1 and 2.
David
To answer your questions:
1. Figure 2C. H3 panel. How different are the left 2 bands? How different are the right 2 bands?
They are not the same.
2. Figure 2E. Histones panel (lowest panel). How different are the bands in the middle 2 lanes from the bands in the right 2 lanes?
They are not the same.
Therefore, in my opinion, despite the retraction, there is no evidence of fraud here.
I can see a few vertical changes in your other examples, this may be due to just cutting and pasting bands, but does not necessarily meant deliberate science fraud.
I think we need to be very clear when making these kind of calls – either there is very clear evidence or there is not and in this case, on the examples you cite, in my opinion, there is nothing to be concerned about.
In reply to Stewart August 7, 2013 at 4:48 pm
Those were questions. It is important to have the opinions of others.
Mol Cell Biol. 2008 May;28(10):3198-207.
http://mcb.asm.org/content/28/10/3198.long
On the point of low resolution images, which do have features.
How similar are the input Wb PRMT5 panel figure 2B and the TUBULIN panel figure 8D?
http://mcb.asm.org/content/28/10/3198/F2.large.jpg
http://mcb.asm.org/content/28/10/3198/F8.large.jpg
I agree that fig 2 looks okay. But there does appear to be stitching of WB panels together. Whilst this is incorrect to do, I doubt there is intention to deceive in this particular figure, since it is so obvious. The reviewers should have picked up on this and asked for the blots to be clearly separated or re-performed.
David wrote:
Genes Dev. 2002 Apr 15;16(8):919-32.
Figure 1.
http://genesdev.cshlp.org/content/16/8/919/F1.large.jpg
Figure 1C. It is only a little thing, but the left C643A colony and left W664A colony are highly reminiscent of each other
Yes, there are several markers that resemble one another in each colony. However there are several other markers in each image not consistent.
I agree with Stewart that in Genes Dev. 2002 Apr 15;16(8):919-32 Fig. 2 there are differences in the bands that make me think they are not related, but I agree with david hardman that the left C643A colony and left W664A in Fig. 1 are derived from the same image.
David wrote:
Those were questions. It is important to have the opinions of others.
Mol Cell Biol. 2008 May;28(10):3198-207.
http://mcb.asm.org/content/28/10/3198.long
On the point of low resolution images, which do have features.
How similar are the input Wb PRMT5 panel figure 2B and the TUBULIN panel figure 8D?
http://mcb.asm.org/content/28/10/3198/F2.large.jpg
http://mcb.asm.org/content/28/10/3198/F8.large.jpg
My answer:
Blow the images up, flip the http://mcb.asm.org/content/28/10/3198/F8.large.jpg horizontally and yes, they are identical images.
Well spotted, detective Hardman!
In reply to Stewart August 8, 2013 at 11:43 am
On the point of similarity of images.
J Clin Invest. 2007 Feb;117(2):482-91.
Figure 2.
http://www.jci.org/articles/view/29544/figure/2
How similar are the bands IgG bands in the left and right lanes of figure 2C?
Home found at Pubpeer for Mol Cell Biol. 2008 May;28(10):3198-207:
http://pubpeer.com/publications/18347060
I agree, too. These are the same colonies. But what does it mean in context of the whole publication?
Credit and trust are extremely important in science. Manipulation of figures with intention to deceive, if that is what it is here (I refer to Figure 1), brings into question the entire results of the paper.
David H wrote:
‘On the point of similarity of images.
J Clin Invest. 2007 Feb;117(2):482-91.
Figure 2.
http://www.jci.org/articles/view/29544/figure/2
How similar are the bands IgG bands in the left and right lanes of figure 2C?’
Initially they look similar…..but…..surprisingly……they are not identical when looking very close………sorry Dave!
and I hope you are not going to write ‘Kaboom, that was the sound of my head exploding 🙂 ‘
A recent publication by the director of the institute.
1. Cancer Res. 2012 Jul 1;72(13):3251-9.
Figure 2.
http://cancerres.aacrjournals.org/content/72/13/3251/F2.large.jpg
Figure 2A. The bands in the HCT116 and MDA-431 FLAG panels look quite similar, and are superimposable as a group.
The bands in the HCT116 and MDA-431RCC1 panels look quite similar, and are superimposable as a group.
Another two from not so long ago:
2. J Clin Invest. 2004 Oct;114(8):1117-27.
Figure 7.
http://www.jci.org/articles/view/22222/figure/7
Figure 7A. Cytochrome c panel. Vertical, light streak/step in level of bands between lanes 3 and 4. It is part of a time series.
Figure 7D. Survivn panel. Vertical, light streaks/vertical changes in background between the zero, 4 and 8 hour lanes. It is supposed to be a time series.
3. Proc Natl Acad Sci U S A. 2003 Nov 25;100(24):13791-6.
Figure 2.
http://www.pnas.org/content/100/24/13791/F2.large.jpg
Figure 2A. Left Hsp90 panel. Vertical, straight changes in background between all 3 lanes. It is supposed to be part of a dose response. The change in background between the left and middle lanes is subtle, but present. The change in background between the middle and right lanes is more obvious.
Figure 4.
http://www.pnas.org/content/100/24/13791/F4.large.jpg
Figure 4C. Vertical changes in background between the left and right lanes of the panels. At first I thought this was because the different lanes were using different antibodies. From the legend I do not believe this to be the case.
Legend figure 4C.
“C) Caspase activity. YUSAC-2 cells loaded with mAb 8E2 or IgG were analyzed for caspase-3/7 activity by Asp-Glu-Val-Asp expression and flow cytometry in the absence (Tet–) or presence (Tet+) of Tet. Data are mean ± SD of three independent experiments. (Inset) Proteolytic processing of caspase-9 in HeLa cells transduced with mAbs 8E2 or 58 to survivin. The positions of proform (Casp.9) or cleaved (Cleaved) caspase-9 are indicated.”
Retraction notice ends with this statement…”All authors are aware of and agree to this retraction”. How can all the authors be aware or agree for the retraction when the senior author is deceased>