31 March 2011. Whatever happened to BACE inhibitors? For years after the gene for this APP-cleaving enzyme was cloned (Vassar et al., 1999), excitement was building among scientists who thought they finally had the golden target in hand for rational drug development in Alzheimer’s disease. But the target proved to be surprisingly obstreperous and its luster has dimmed. Clinical trials were a long time in coming, and periodic rumors that company X or company Y did finally have a BACE inhibitor in human tests invariably faded into the silence that means development ended before Phase 2. More recently, the field seems to have developed a second wind, but not without casualties along the way. At the 10th AD/PD International Conference, held 9-13 March 2011 in Barcelona, Martin Citron, told the story of one such setback in some detail (see brief mention in prior ARF related news story). A cautionary tale of the vagaries of drug development, Citron’s talk opened a rare window into the closed world of pharmaceutical company research.
Citron has worked on inhibiting BACE, first at Amgen, now at Eli Lilly and Co., Indianapolis, ever since his lead in cloning the gene. When the protein’s crystal structure came into view (Hong et al., 2000) and the first knockout mice proved viable (Luo et al., 2001; Cai et al., 2001), a field of contenders trying to block the enzyme was off to the races. Pharma companies had been trying to inhibit β-secretase even before 1999, without success, but these papers created more heated momentum behind BACE.
That was a full decade ago. Why so slow? Initially, the reasons were technical. FRET-based assays generated false positives, high-throughput assays failed, and a “myopic focus” on high-potency compounds led scientists astray as they pursued candidates that penetrated the brain poorly or did not clear properly. “I learned that inhibitor development is much more difficult than people think,” Citron said.
He was not alone. Many groups were spinning their wheels, and questions arose about BACE as a drug target in private conversations and in the literature. “Now there is a cloud over the target. I want to show you today that BACE1 is indeed druggable,” Citron said.
How so? The Lilly scientists, including Patrick May, Robert Dean, Stephen Lowe, and others, did eventually find a BACE inhibitor that entered the brain when taken by mouth. It robustly reduced Aβ in humans. They had started fresh, screening non-peptidic small molecules with an eye toward favorable drug properties rather than potency. Then they improved potency with medicinal chemistry. They ended up with a molecule that sits in the S3 pocket of BACE1’s active site and reacts with two catalytic sub-sites of the enzyme. It had a fair potency of 0.25 micromolar in five different cell lines. Called LY 2811376, the molecule behaved itself in vivo. It reduced brain Aβ in PDAPP transgenic mice in dose-dependent fashion. It changed both upstream and downstream CSF and plasma biomarkers in the expected ways in mice and dogs. “With this and a clinical toxicity data package, we were ready to go into the clinic,” Citron said.
There, too, things went passably for a while. A single ascending dose study in healthy volunteers yielded the pharmacokinetic and pharmacodynamic data the scientists wanted to see. Maximal plasma concentration, half-life, time course of Aβ reduction, and dose dependence—all that stuff looked good. Likewise, a 36-hour spinal catheterization study in healthy volunteers showed a decrease in CSF Aβ40 and 42, again with dose dependence, time course, and biomarkers behaving as desired. Adverse events up to this point included headache, palpitations, colds—nothing the data safety monitoring board found concerning.
“We were riding high on the biomarker data,” Citron said.
While they were preparing for Phase 2, the downfall came. Rat toxicology studies showed that a higher dose given for three months ravaged the pigment epithelium of the rat’s eye. This retinal layer had inclusions and extensive damage. Lilly ended dosing and brought people in for eye assessments, which thankfully showed no abnormalities, Citron said.
What happened? The scientists still don’t know what the compound does to the eye. BACE knockout mice do not show this phenotype, but they do develop it when treated with LY 2811376, suggesting to Citron that this is an idiosyncratic effect of this particular compound, not of BACE inhibition. This would imply that the compound is dead for AD (and any other indication, for that matter), but BACE as a target still stands. This toxic effect does not show up with other compounds of this chemical series, either, Citron said.
Asked later what it’s like to get this far only to be bounced by an off-target effect, Citron said dryly: “Welcome to drug development. This happens all the time. It just never gets published.”
So what’s to be learned? LY 2811376 is the first BACE inhibitor with profound CNS effects, and constitutes proof of concept that BACE is druggable, said Citron. He noted that Lilly has another (retina-tested) compound in the clinic, as do CoMentis (see ARF Keystone story), Eisei, Merck/Schering, and TransTech Pharma. At AD/PD 2011, AstraZeneca scientists presented preclinical and biomarker data on a new small-molecule inhibitor on three posters.
Other scientists noted, though, that it remains difficult to find BACE inhibitors that achieve effective exposure levels in the brain without also creating toxic exposure levels in the periphery. Others suggested that, rather than continuing to butt its head against the wall of BACE inhibition, scientists might be well advised to explore BACE modulation. The γ-secretase field appears to be migrating away from inhibition and toward modulation. Likewise, regulation of BACE by modulation, rather than inhibition, could be amenable to manipulation, commented Barbara Tate of Satori, Inc., in Cambridge, Massachusetts. Indeed, the U.S. Patent Office in the past year issued half a dozen patents on compounds claimed to modulate BACE, indicating that some companies are already pursuing this softer route (see, e.g., Albrecht et al., 2011; Malamas et al., 2010).
In addition, little flags about potential safety liabilities with BACE are fluttering on the basic science front. This aspartyl protease cleaves a number of substrates besides APP. Also in Barcelona, Michael Willem presented new data on neuregulin, a physiological substrate of BACE1. Working with Christian Haass at Ludwig-Maximilians-Universität, Munich, Germany, Willem had originally discovered that BACE1 cleaves neuregulin (Willem et al., 2006).
Willem showed a slide of some 16 BACE inhibitor compounds by many companies that have since been abandoned. These make handy tools for his work exploring BACE1 and its targets, both in cell culture and in mice. All BACE inhibitors Willem has used so far, including LY 2811376, slow the turnover of neuregulin-1-b1. Both the peripheral and central nervous system of young and adult mice express this isoform, Willem told the audience. The inhibition causes the unprocessed substrate to accumulate; this, in turn, could slow downstream signaling by neuregulin cleavage products. The biological activity of neuregulin is far from fully understood, but neuregulin-1 is thought to play a role in synaptic transmission and the maintenance of synapses.
“Neuregulin cleavage is dependent on BACE1. There is no redundancy for that in mice. I still do not know what neuregulin really does in the brain. But if you block BACE1, you will block neuregulin, too,” Willem said.—Gabrielle Strobel.