The elite journal Nature Cell Biology (NCB) has a very commendable policy on data integrity: the deposition of raw data, in particular original scans of western blots and other gel analyses. This is to avoid dishonest authors playing tricks with loading controls, antibody specificity and stealthy gel splicing or even band duplications. The rule however starts applying only after the paper passed peer review and was accepted for publication, and because of that, some authors deposit any odd gel picture and count on nobody bothering to check, or to take action.
This was exactly what the Spanish star cancer researcher Carlos López-Otín, professor of Biochemistry and Molecular Biology at the University of Oviedo, did. For his lab, the disregard for loading controls is standard practice: he even managed to reuse same picture of one no less than 23 times (read here). Recently, some of those papers, all at Journal of Biological Chemistry (JBC), had to be corrected.
The embarrassment for the Nature family journal NCB is the bigger because the flagship Nature issued in 2017 a Mentoring Award to Lopez-Otin (read here), where one really wonders what exactly Nature aimed to reward here. The skill of publishing in their own high impact journals using a sod-all, who-cares attitude to research integrity, which Lopez-Otin taught his mentees?
Clara Soria-Valles , Fernando G. Osorio , Ana Gutiérrez-Fernández , Alejandro De Los Angeles , Clara Bueno , Pablo Menéndez , José I. Martín-Subero , George Q. Daley , José M. P. Freije , Carlos López-Otín
NF-κB activation impairs somatic cell reprogramming in ageing
Nature Cell Biology (2015) doi: 10.1038/ncb3207
The first author Clara Soria Valles is presently postdoc with another significant co-author of that paper, George Q Daley. Do you know who Professor Daley is? Merely the Dean of Harvard Medical School and one of the most important and influential people worldwide in the field of stem cell research. According to Author Contributions statement, “G.Q.D. provided critical materials and participated in the preparation of the manuscript“. It is not clear if that participation involved the checking of data integrity and if those “crucial materials” Daley provided were already published (if they were, he should not have claimed co-authorship). In any case, with Daley and his postdoc on board, the correction of that paper will become a rather sensitive and political issue. If it ever happens, that is. With that paper and the help of Daley, the cancer researcher Lopez-Otin was able to conquer another piping hot and highly lucrative research field: that of stem cells and regenerative medicine. It earned him an ERC grant right away, worth of €2.5mn from 2017 to 2022, to work on continuation of his NCB study, on “Deconstructing Ageing: from molecular mechanisms to intervention strategies“. It is very convenient that the elite European public funder ERC does not seem to care a bit about research integrity, and is rather proactive in that regard.
The paper in Nature Cell Biology which made all this possible, postulated the decisive role of the transcription factor protein NF-kB in cellular senescence as barrier to reprogramming of cells to pluripotency, and offered a potential cure to child patients suffering from the deadly premature ageing syndrome, which is caused by genetic and cell nucleus defects, specifically the Néstor–Guillermo progeria and the Hutchinson–Gilford progeria.
Lopez-Otin’s impactful publication of 2015 must have given hopes to the families of sick children. If only they knew how sloppily the data was assembled.
Much of the sleuthing came from the pseudonymous Clare Francis, who comments and shares image evidence on my site as “Zebedee”, several others also contributed on PubPeer. There are many issues with that NCB paper, the PubPeer thread went over 120 comments, despite moderation. Some of the criticism appears misguided, when two or more western blots photographed together on one film were confused for a single gel. But most of the PubPeer evidence has its validity. For brevity, I will present below only select evidence.
The debate started almost exactly one year ago, in September 2017 with this comment, where a mismatch between figure and raw data in 3 figures was reported:
Figure 6b shows an array of 6 Western blots, with b-actin being the loading control, and no two of these panels came from the same gel. They all, p-IKK2, p-ATM, IkBa and the loading control b-actin came from a separate gel each, presented in the final figure as a single blot. Further PubPeer debate proved that also NF-kB and Oct-1 signals came from two more gels. Note that the lanes of original blot scans are not labelled, there is also no legend.
The signal for NF-kB protein in that figure and in another one, Figure 3c, does not really match the original scan. Basically, the authors failed to provide correct gels for their NF-kB data. And the ones they did provide show fat and rather specific looking bands underneath. If the authors cropped this image data, it was against Nature‘s own instructions for authors. But then again, those do not seem the correct blots anyway.
The NF-kB is the central analysis of the study, it is the hub protein of the provided key mechanism and even part of the paper’s title. Surely the NF-kB data should be the most clear and reliable? Yet also in Figures 3c and 6b, the data in the figure doesn’t match what the authors provided as raw data. From Oct-1 analyses it also becomes rather obvious that the Lopez-Otin lab cannot be bothered with proper loading controls. Some 14 samples were loaded on a gel and used for each of the 3 separate experiments in need of some Oct-1 data. The bands don’t seem to match. Is this even the Oct-1 signal the gel shows, or maybe a forgotten loading control from yet another experiment?
The authors did same again, with yet another “Oct-1” and “NF-kB” gels for Supplementary Figure 8, as commented on PubPeer. In fact, what are those published NF-kB bands anyway? After a PubPeer user boosted the contrast of raw gel scans, more bands appeared. Because the authors did not bother to provide molecular weigh marker scale as requested by the publisher, we cannot know which of the smear (if any) might correspond to the correct NF-kB band.
This is what Nature Publishing Group itself has to say as their expectations towards the authors about the data integrity of western blots (highlights mine):
Electrophoretic gels and blots
Positive and negative controls, as well as molecular size markers, should be included on each gel and blot—either in the main figure or an expanded data supplementary figure. […]
The display of cropped gels and blots in the main paper is encouraged if it improves the clarity and conciseness of the presentation. In such cases, the cropping must be mentioned in the figure legend. (Some journals require full-length gels and blots in supplementary information wherever possible [as does Nature Cell Biology, -LS].)
- Quantitative comparisons between samples on different gels/blots are discouraged; if this is unavoidable, the figure legend must state that the samples derive from the same experiment and that gels/blots were processed in parallel. Vertically sliced images that juxtapose lanes that were non-adjacent in the gel must have a clear separation or a black line delineating the boundary between the gels. Loading controls (e.g. GAPDH, actin) must be run on the same blot. Sample processing controls run on different gels must be identified as such, and distinctly from loading controls
- Cropped gels in the paper must retain important bands.
- Cropped blots in the body of the paper should retain at least six band widths above and below the band.
- High-contrast gels and blots are discouraged, as overexposure may mask additional bands. Authors should strive for exposures with gray backgrounds. Multiple exposures should be presented in supplementary information if high contrast is unavoidable. […]
Of course, those rules might not apply to friends of Daley’s like Lopez-Otin. This is also why three physically different gels were used to make the small Figure 2a. The gel for Lamin B1 has 9 lanes (plus molecular weight marker), the gels for p16 and b-actin 6 lanes. Yet also those latter two must be separate gels, because the p16 signal is in different orientation. Either that, or the authors didn’t like the result of their p16 experiment and flipped it around so the progeria samples (NGPS) would show more of the senescence-associated protein p16.
Figure 4i has another flip, the gel bands for the methylated histone protein H3K79me2 are reversed. Why would one do that? The reverse order to load a gel like that would be highly unusual and illogical. What was going on there? The original loading control seems to have been already used in Figure 3C. It definitely was a different gel from the methylated histone analysis from Figure 4i shown below. Never mind all that, the figure was good enough to be quantified into a bar diagram. Normally one needs triplicates for that, but the authors forgot to upload them, or maybe simply couldn’t find something which might match distinctly.
Yet here it is again, this loading control, reused in the best Lopez-Otin manner. Yes, admittedly those are same samples, according to figure legends. But not from the same gels, as raw data indicates. Why was it such a problem for the authors to probe each of these gels for equal loading and to provide correct loading controls? Why do we have to trust them blindly to have loaded all samples in equal amounts on different gels (if interested, here is more on importance of loading controls)? I am not sure NCB‘s impact factor and elite status of peer review is any argument on why controls should be suspended.
There was even more band flipping, in Figure 7a (a similar event took place also in Figure 6b). It seems to be same gel, stripped of the previous antibody signal and re-probed. So why the flip? A mistake? Was one of authors bored, and wished to troll? Or was something wrong with the results, did they need some fixing? Of course the loading control doesn’t match, and it is anyone’s guess what “NF-kB” and “Oct-1” bands might have been in their other life.
Also in Figure 8B, the authors apparently probed the same gel, but maybe once again found themselves unhappy with the result. The methylated histone lanes used were samples 2, 3, 4, but the b-actin loading controls were 3, 2, 1, in this orientation. From the same gel, mind you. Maybe the loading control was “unusable”, because the differences in H3k79me2 signal would be less prominent if b-actin was matched correctly?
Lopez-Otin and his co-authors made a mockery of good scientific practice and the concept of loading controls by uploading as raw data whatever they had at hand. The publisher must now decide if it wishes to defend its guidelines of “Loading controls (e.g. GAPDH, actin) must be run on the same blot” and “Cropped gels in the paper must retain important bands” or in fact, that “the final image must correctly represent the original data and conform to community standards“.
And of course maybe the scientific community and the progeria patients’ families should be cautioned through a correction notice how the data which gave them hope was assembled.
If you are interested to support my work, you can leave here a small tip of $5. Or several of small tips, just increase the amount as you like. Your generous patronage of my journalism, however small it appears to you, will greatly help me with my legal costs.