Of mice, men, and inflammation

In PNAS, Seok et al. advocate against relying on mouse models to study human inflammatory diseases, because genomic responses in human and mouse leukocytes correlate poorly (1). We fully agree with the conclusion of the authors that it is incorrect to believe “that molecular results from current mouse models developed to mimic human diseases translate directly to human conditions” (1). Several mouse studies—although yielding solid and interesting results—have been the victim of overinterpretation; for example, by extrapolating successful pretreatment in mice to therapeutic treatment in men. However, this aspect does not render these results useless. In the field of inflammation, valuable human therapies have been derived from mouse studies (e.g., anti-TNF treatment for rheumatoid arthritis and inflammatory bowel disease). In the field of sepsis, failure of the phase III clinical trial inhibiting NO synthases, which had to be terminated because of excess mortality, was actually predicted in assorted murine shock models (2).

In this particular case, we do not agree that transcriptome changes in total white blood cells should be used as a bona fide read-out for systemic inflammatory diseases. As described in an earlier report by the same authors, there is no relation between the leukocyte transcriptome changes and clinical severity, recovery, or outcome (3). In several of our studies, we collected data indicating that parenchymal cells, rather than leukocytes, are responsible for the onset of morbidity and cardiovascular dysfunction in inflammatory shock, even for the production of systemic NO (2). Also for endotoxemia, parenchymal cell activation entirely drives systemic inflammation and mortality (4).

There are some further considerations that need to be contemplated before drawing drastic conclusions. To induce endotoxemia, the authors used doses of LPS that differed 250-times in men versus mice; in addition, the route of administration was different (intravenously in men and intraperitoneally in mice). The difference in dosing was chosen because of similar IL-6 induction patterns. At the doses used, the authors have previously shown a threefold increase in total leukocyte counts in humans, but no significant alterations in mice (5). We believe the LPS dose should have been chosen to reflect similar pathology rather than early IL-6 induction (e.g., morbidity, cardiovascular/mitochondrial dysfunction, tissue damage, induction of shock). The dose used in humans (2 ng/kg) causes a febrile, cardiovascular, and hematological response (5), and is ∼1/15 of the dose that has been reported to cause shock. In contrast, the dose used in mice (400–500 ng/kg) causes no febrile, cardiovascular, or hematological response (5), and is 1/1000–1/10,000 of a dose that would cause shock. Hence, to compare pathological responses between mice and men in the context of endotoxemia, it might have been better to challenge mice intravenously with a dose that was at least 1,000-fold higher.

The extrapolation of preclinical data to the clinic is far from straightforward, but even if mouse models do not copy the entire course or spectrum of a human disease, they are still invaluable tools to study certain pathological aspects of the disease and gain mechanistic insights.