Abstract
Alzheimer's disease (AD) is characterized by formation of neuritic plaque primarily composed of a small filamentous protein called amyloid-β peptide (Aβ). The rate-limiting step in the production of Aβ is the processing of Aβ precursor protein (APP) by β-site APP-cleaving enzyme (BACE1). Hence, BACE1 activity plausibly plays a rate-limiting role in the generation of potentially toxic Aβ within brain and the development of AD, thereby making it an interesting drug target. A phase II trial of the promising LY2886721 inhibitor of BACE1 was suspended in June 2013 by Eli Lilly and Co., due to possible liver toxicity. This outcome was apparently a surprise to the study's team, particularly since BACE1 knockout mice and mice treated with the drug did not show such liver toxicity. Lilly proposed that the problem was not due to LY2886721 anti-BACE1 activity. We offer an alternative hypothesis, whereby anti-BACE1 activity may induce apparent hepatotoxicity through inhibiting BACE1's processing of β-galactoside α-2,6-sialyltransferase I (STGal6 I). In knockout mice, paralogues, such as BACE2 or cathepsin D, could partially compensate. Furthermore, the short duration of animal studies and short lifespan of study animals could mask effects that would require several decades to accumulate in humans. Inhibition of hepatic BACE1 activity in middle-aged humans would produce effects not detectable in mice. We present a testable model to explain the off-target effects of LY2886721 and highlight more broadly that so-called off-target drug effects might actually represent off-site effects that are not necessarily off-target. Consideration of this concept in forthcoming drug design, screening, and testing programs may prevent such failures in the future.