three major toxicities were encountered at various stages of development: pulmonary hemorrhage (exhibited during early-stage development), bone physeal abnormalities (exhibited during first-time-in-human studies), and heart valve lesions (exhibited during 10-day dose range–finding studies prior to first-time-in-human studies). The investigation of each posed different challenges and demanded different techniques.


The first problem encountered in the ALK5 program was pulmonary hemorrhage. Histopathology of lung tissue in treated rats showed diffuse alveolar damage characterized by fibrin exudation, alveolar septal necrosis, and inflammatory cells; the damage was present with a number of test compounds. Because the researchers were looking at several compounds from a series, they wanted to determine whether they would encounter this problem with every compound.

Frazier’s group hypothesized that the lung damage was being caused by free radical production and reactive oxygen species. Therefore, they decided to look at the different compounds, see which ones caused free radical production, and then determine whether this superoxide production correlated with the alveolar damage. Using an in vitro model that employed an A549 lung adenocarcinoma cell line, they incubated the cells with either the compound or a control for 4 hours, exposed the cells to a hydroethidine dye, and then ran them through a flow cytometer with a 488 nm argon laser. When reactive oxygen species are present, hydroethidine dye turns into ethidium bromide, which fluoresces when exposed to 488 nm light. Therefore, this approach made it possible to determine quickly whether free oxygen radicals were present in a given set of lung cells.

The group looked at 150 compounds originating from three separate programs at GSK, all of which had encountered the same kind of pulmonary hemorrhage. A clear dose–response relationship was found, with higher doses leading to greater superoxide production. Some compounds led to much greater superoxide production than others, while some showed few reactive oxygen species at all (see Figure 2-1). Furthermore, the superoxide production correlated with the histopathology results: the compounds that showed increased superoxide production were the same as those that showed increased lung damage. Finally, after determining which compounds were causing superoxide production, the group reexamined those compounds’ biochemical structures and found that most had a similar side chain. Frequently, one chemical series tends to be highly prone to reactive oxygen, and in this case the effects of the reactive oxygen had nothing to do with the fact that the ALK5 signal was occurring. Thus the group concluded that the lung damage was not due

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement