I am a fan of PubPeer, as it provides a forum for discussion between authors and the wider community, something I have discussed in a number of posts (two examples being here and here). Two days ago, My colleague Mike Cross came by my office, having just delivered a pile of exam scripts for second marking (it’s exam and marking season), asking if I had seen a comment on our paper on PubPeer. I had not – too many e-mails and too busy to look at incoming!
So I looked at the question, which relates to panels in two figures being identical in our paper on neuropilin-1 and vascular endothelial growth factor A (VEGFA) – indeed they are labelled as being identical.
However, the reason for this and the underlying experimental design is perhaps less obvious, unless you read the paper very carefully and are a fully paid up member of the VEGF, neuropilin and endothelial cell communities.
So a quick e-mail to Kat, who did this work as part of her PhD, elicited that evening some files. To make these accessible, I signed up, long overdue, at Figshare. A minor struggle, as ever, learning another new system and I need to go back there and tidy up some of the false starts.
On Figshare we have the experimental design, which in essence is that each experiment is pretty massive, with everything done with the same passage of cells. This ensures maximum comparability.
So why did we put the same panels in two figures? As I point out in my response on PubPeer, this is to strengthen the key finding of the paper, that neuropilin-1 is a pure VEGF receptor 2 (VEGFR2) agonist. In the field, this is new. Indeed, when Kat was doing the initial pathfinding experiments, the neuropilin-1 alone condition was a negative control, since according to the literature, neuropilin-1 only potentiates the activity of VEGFA towards VEGFR2. I was certain that the initial result was an artefact, because the endothelial cells are cultured in medium with VEGFA165. VEGFA165 is a heparin-binding isoform of VEGFA. Thus, washing the cells and replacing the medium with one lacking VEGFA165 might not remove effectively VEGFA165, since much of this might be bound to pericellular matrix heparan sulfate, rather than be in the bulk medium. Indeed, accumulation in matrix followed by slow delivery to tyrosine kinase receptors is a general property of most heparin-binding growth factors. Happily there was a control, the VEGFA121 isoform, which activates VEGFR2 like the other isoforms, but does not bind heparin, heparan sulfate or indeed neuropilin-1. So the final experimental design incorporated cells cultured for a passage in VEGFA165 (the ‘standard’ isoform, which everyone uses, and which binds heparin and heparan sulfate) and cells that had been cultured in VEGFA121. The latter would definitely be removed by washing the cells, just as one can remove epidermal growth factor (EGF) this way. Figure 4, which reports the VEGFA121 data, needed the VEGFA165 data alongside to allow readers to see that culturing the cells in either of the two isoforms was equivalent and that the neuropilin-1 response was also equivalent. This formally excludes the artefact, and also shows that the cells effectively consume the VEGFA165 before we put on the neuropilin. Hence we deliberately used data from the same experiment in Figs 1 and 4 to hammer the point home.
In hindsight, what was perhaps lacking was making this more explicit in the text. A contributing factor to muted language is that the paper challenged the existing dogma regarding how neuropilin-1 exerts its angiogenic activity. At least some of the reviewers would be from labs that had established this dogma and we did not want to upset people more than necessary.
Ferniglab Blog 