Reply to Ricciotti et al.: Evidence for vascular COX isoforms

We welcome interest in our work and encourage debate. As we have published previously (1), we agree with FitzGerald and coworkers (2) that those drugs that inhibit COX-2 have potential deleterious effects upon cardiovascular health, although we disagree with their conclusions that COX-2 drives prostacyclin release in healthy vessels or that we have omitted to cite relevant literature.

Our paper in a recent issue of PNAS (3) contained references that represent the breadth of literature in this area, including two reviews from FitzGerald and coworkers and their most recent relevant study (4), and we encourage readers to read widely.

We are aware that FitzGerald and coworkers believe urinary 2,3-dinor 6-keto PGF_1α (PGIM) to be a definitive marker of endothelial cell production of prostacyclin. However, this is not universally accepted, as has been reviewed at length by Flavahan (5) and ourselves (1). We would argue that a direct way to characterize the COX isoform generally responsible for the production of endothelial PGI₂ is to determine enzyme expression and activity in fresh blood vessels. Here, results are overwhelmingly in favor of COX-1 (1, 5, 6). Notably, no researchers have provided direct evidence of the dominant expression and enzyme activity of COX-2 in healthy blood vessels, although COX-1 has been readily found. In their most recent study, FitzGerald and coworkers showed expression of COX-2 in vascular cells in culture, conditions known to induce COX-2 in some settings, but, unlike our study, provided no evidence from fresh vessels (4). Indeed, we showed that culture of vessels induced COX-2 protein and activity after just 2 h (3). With regard to PGI₂ production, in our studies, we used immunoassays to measure prostacyclin’s breakdown product, 6-ketoPGF_1α. This is general practice, and our results were in accordance with multiple other published studies, including those using mass spectrometry-based approaches, in that we found the release of PGI₂ from fresh, healthy blood vessels to be dependent upon COX-1 (1, 5, 6).

We note an error for 6-ketoPGF_1α in fig. 6 of our paper (3), where “ng/mL” should have read “pg/mL.” However, with respect to our use of bradykinin as a selective endothelial cell activator, this approach is entirely in accordance with earlier studies mentioned in our paper, and those of FitzGerald and coworkers (3). We did not measure bradykinin’s effects on urinary prostacyclin metabolites and, so, did not comment on that aspect. We would remind readers that in our study, we found that the ability of i.v. bradykinin to elevate circulating 6-ketoPGF_1α was absent in COX-1 knockout mice but normal both in COX-2 knockout mice and in wild-type mice treated with a selective inhibitor of COX-2. It was also prevented in wild-type mice by diclofenac, which inhibits both COX-1 and COX-2.

Finally, we would emphasize that we, in no way, distance ourselves from the conclusion that nonsteroidal anti-inflammatory drugs (NSAIDs) are associated with an increased risk for myocardial infarction (1). We reiterate our conclusion that whereas urinary PGIM levels are driven, in part, by COX-2 activity, vascular prostacyclin production in healthy blood vessels is generally driven by COX-1. We believe that this conclusion can help focus research in this important area.