Fat cell paper earns unusually detailed retraction

jcbA pair of cell biologists have taken responsibility for extensive figure errors that scuttled their paper in the Journal of Cell Biology.

While there were five authors, first and last authors Eva Szabo and Michal Opas took responsibility in the notice. A number of figures “contain incorrect data and/or presentation errors,” and the original data isn’t available for verification. The notice is unusually clear about which figures and data are compromised.

The paper was published in 2008, and retracted on January 12, 2015. It has been cited 32 times, according to Thomson Scientific’s Web of Knowledge.

Here’s the notice for “Calreticulin inhibits commitment to adipocyte differentiation’:

The editors of The Journal of Cell Biology have been notified by Dr. Michal Opas that he and the other authors of the paper referenced above retract the paper. As a result of this retraction, no data in this paper should be cited in the scientific literature. University of Toronto supports this retraction.

The authors provided the following statement:

Due to errors in image placement and data presentation during figure preparation, four figure parts (Fig. 1 C, Fig. 3 C, Fig. 7 A, and Fig. 7 E) contain incorrect data and/or presentation errors, as articulated below. Due to the inability of the authors to locate the correct data for Fig. 7 E, the quantification in Fig. 7 F cannot be validated. Due to an error in experimental design, the Western blot data in Fig. 4 D are not sufficient to support the quantification in Fig. 4 D. The other figures in the paper and the other parts of the figures listed above (Fig. 1, A, B, and D–F; Fig. 2; Fig. 3, A, B, D, and E; Fig. 4, A–C and E–H; Fig. 5; Fig. 6; Fig. 7, B–D and G–I; Fig. S1; and Fig. S2) were not affected by these errors.

Authors Eva Szabo and Michal Opas take full responsibility for these errors. Authors Yuanyuan Qiu, Shairaz Baksh, and Marek Michalak did not participate in data collection or figure preparation for any of the figures for which errors have been identified.

The errors identified were the following:

(1) The text fails to note the intentional duplication of the top GAPDH panel in Fig. 1 C with the top GAPDH panel in Fig. 7 A.

(2) Two panels in Fig. 1 C do not accurately represent the original data. Specifically, the PPARγ2 panel and the corresponding GAPDH panel do not properly show that lanes 1 and 2 are derived from a different gel than lanes 3 and 4. In addition, the image for the left two GAPDH lanes is flipped horizontally in the figure relative to the original data.

(3) Two panels in Fig. 3 C do not accurately represent the original data. Specifically, lanes 1 and 2 of the bottom GAPDH panel in the untreated dataset, which are identical to lanes 3 and 4 of the top GAPDH panel in the untreated dataset, are incorrectly presented with a splice line between the lanes. Lanes 1 and 2 of the C/EBPα panel in the untreated dataset, lanes 3 and 4 of the C/EBPα panel in the untreated dataset, and lanes 3 and 4 of the bottom GAPDH panel in the untreated data also are incorrectly presented with splice lines between the lanes. Lastly, lanes 1 and 2 of the bottom GAPDH panel should be presented as lanes 3 and 4 and vice versa.

(4) The text fails to note the intentional duplication in Fig. 3 C of the WT and G45crt−/−data in the top and bottom GAPDH panels for the untreated dataset, once corrected as per point 3 above.

(5) Three panels in Fig. 3 C do not contain the correct data. Specifically, all three L32 panels are placeholder images that were not replaced with the corresponding experimental data before publication.

(6) The +BAPTA-AM C/EBPα panel in Fig. 4 D is a duplicate of the +BAPTA-AM PPARγ2 panel in the same figure. The correct C/EBPα data were derived from a separate gel than the untreated C/EBPα data and thus cannot be used to support the quantification in Fig. 4 D.

(7) Three panels in Fig. 7 A do not properly represent the original data. Both GAPDH panels and the CaMK II Thr286 panel fail to contain the necessary splice lines between lanes 2 and 3. In addition, the data in lanes 3 and 4 of the top GAPDH panel are flipped horizontally relative to the original data.

(8) Five panels in Fig. 7 E do not contain the correct data. Specifically, all four L32 panels are placeholder images that were not replaced with the corresponding experimental data before publication. In addition, the data shown in the top GAPDH panel in the untreated dataset, which is identical but flipped 180 degrees relative to the bottom GAPDH panel in the KN-93–treated sample, are not the correct data. The correct GAPDH data corresponding to the PPARγ2 panel in the untreated dataset could not be located. As a result, the quantification in Fig. 7 F cannot be validated.

(9) The text incorrectly states that quantification of aP2 levels is shown in the graph in Fig. 7 F.

As a result of these errors, the conclusion that BAPTA-AM treatment increased C/EBPα expression in all cell lines and was indicative of restored adipogenic potential in G45[P+C] cells (Fig. 4 D) cannot be validated. In addition, the conclusions that inhibition of CaMK II in CGR8+/+, G45−/−, L7+/−, or L7−/− cells decreased PPARγ2 expression (Fig. 7, E and F); that treatment with KN-92 had no effect on PPARγ2 expression in CGR8+/+, G45−/−, L7+/−, or L7−/− cells (Fig. 7, E and F); and that these findings were indicative of an important role for the calmodulin–CaMK II pathway during adipogenesis in embryonic stem cells cannot be validated.

The authors apologize for any confusion these errors may have caused to the research community.

When we reached out to Opas, he forwarded us to the PR department at the University of Toronto, who said:

The University is aware of, and supports, the retraction.

They declined to answer further questions. We’ve also reached out to the journal, and will update if we hear anything.

Update, 2:15 p.m. EST, 01/16/15: We’re still not sure what prompted the retraction, but we’ve learned the journal received an email describing many of these problems in February 2013.

Update, 7:45 p.m. EST, 01/16/15: To clarify, the journal declined to provide more details yesterday, saying:

The published retraction notice includes all of the pertinent details regarding this paper and its current status.

The journal also declined to say whether they had received the February 2013 email mentioned above, which was sent by Clare Francis and which we reproduce here with email addresses redacted:

———- Forwarded message ———-
From: clare francis <redacted>
Date: Fri, Feb 15, 2013 at 1:53 AM
Subject: concerns image confusion J Cell Biol. 2008 Jul 14;182(1):103-16.
To: Liz Williams <redacted@mail.rockefeller.edu>
Cc: redacted@rockefeller.edu

concerns J Cell Biol. 2008 Jul 14;182(1):103-16. doi: 10.1083/jcb.200712078. Epub 2008 Jul 7.
Calreticulin inhibits commitment to adipocyte differentiation.

Source

Department of Laboratory Medicine and Pathobiology, Institute of Medical Sciences, University of Toronto, Toronto, Ontario M5S 1A8, Canada.

Figure 1.

http://www.ncbi.nlm.nih.gov/core/lw/2.0/html/tileshop_pmc/tileshop_pmc_inline.html?title=Click%20on%20image%20to%20zoom&p=PMC3&id=2447897_jcb1820103f01.jpg

Figure 1C. PPARgamma2 panel. Light, vertical streak between the bands in the middle lanes.

L32 panel. Splicing between the middle lanes.

No splicing in the aP2 panel.

Figure 1E. CRT panel. The band in the left lane looks like it is one its own rectangle of background.

Figure 3.

http://www.ncbi.nlm.nih.gov/core/lw/2.0/html/tileshop_pmc/tileshop_pmc_inline.html?title=Click%20on%20image%20to%20zoom&p=PMC3&id=2447897_jcb1820103f03.jpg

Figure 3C. Left-most PPARgamma2 panel. Suspect that the band in the right-most lane has been spliced in. The band has a vertical, straight left edge.

I think that the 3rd and 4th GAPDH panels in the left set of indiviual GAPDH panels are vertically compressed versions of the bands in the middle GAPDH panel above and to the right of them.

Left L32 Panel. Splicing between the 1st and 2nd lanes. Vertical dark streak. Possible splicing between 3rd and 4th lanes.

I think that the band in the L7crt-/- lane of the +BAPTA-AM L32 panel is very likely the same as the band in the L7 lane of the Untreated L32 panel. Note the dark area just above the left ends of the bands.

Figure 4.

http://www.ncbi.nlm.nih.gov/core/lw/2.0/html/tileshop_pmc/tileshop_pmc_inline.html?title=Click%20on%20image%20to%20zoom&p=PMC3&id=2447897_jcb1820103f04.jpg

Figure 7D. I suspect that the bands in the right C/EBPalpha panel are vertically compressed versions of the bands in the right PPARgamma2 panel. I am aware that bands of proteins about the same molecular weight may be similar, but not the same. The distribution of signal with the C/EBPalpha and PPARgamma 2 bands is very similar.

The GAPDH panels that accompany the C/EBPalpha and the PPARgamma2 panels contain quite different bands so it is odd that the bands in the PPARgamma2 and E/CBPalpha panels should be similar even.

Figure 7E. 

The left-most (Untreated) figure 7E L32 panel is highly reminiscent (likely the same) as the left (Untreated) L32 panel in figure 3E, yet the genotypes are different. Note the dark area just above the left ends of in the bands in the left lanes of both panels and the general distribution of light areas with the other bands.

I think that the band in the G45-/- lane of the KN-62 treated figure 7E L32 panel is the same as the band in the L7 lane of the +BAPTA-AM L32 panel in figure 3E. Note the white squiggle, almost like a signature above and just to the right of the middle of the bands.

L7+/- lane 7E is same as L7crt-/- lane 3E.

L7-/- lane 7E is same as WT lane 3E.

Within figure 7E L32 panels. 

KN-62 treated L7+/- lane is same as Untreated CGR8+/+ lane.

KN-62 treated CGR8+/+ lane is same as Untreated G45-/- lane.

I think that the images are at the point where they do not make sense.

 

10 thoughts on “Fat cell paper earns unusually detailed retraction”

  1. This is the way a retraction should be: complete explanation on what is incorrect, why it cannot be trusted. It is also a good way to show what should not be done. Bravo on the detailed notice.

  2. JCB is the most serious journal I know. They were pioneer on definitions of what image manipulation is, followed by many other journals;. They take very seriously any misconduct allegations.

  3. I side with DT, this detailed retraction notice seems more like enforced by an honest journal. I may be too cynical, but I have difficulty believing that any even most penitent author would undergo such public self-flagellation. Thus, well done, JCB!

    1. The funny thing is that JCB states it routinely screens images for manipulation, but these errors contained a lot of splicing. One would think that JCB would have caught this before publication if they really screen their images.

  4. I have to admit I do not understand this phenomenon. Pubpeer seems to be full of comments pointing out photoshopped imagery, so it seems to be a common occurrence. But the resulting images are often so badly done they are easily detected. So why do it? Is the pressure to publish really so great that hordes of scientists would risk exposure as frauds or incompetents?

  5. I thought that the 2015 retraction of the J Cell Biol paper would had led to an investigation of the authors’ work. Apart from stating that they support the retraction did U Toronto have any more to say?

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.