Introduction

Early hopes for BACE1

As a target for drugs to prevent or treat Alzheimer's disease, β-secretase seemed like a sure thing. Currently, the main theory of AD pathogenesis—the "amyloid hypothesis"—attributes the cognitive decline and brain damage seen in this neurodegenerative disorder to the buildup of extracellular amyloid-β peptide (Aβ), which over time forms plaques and oligomers in the brain. Amyloid-β generation is initiated by the cleavage of amyloid precursor protein (APP) by β-secretase, now known to be a transmembranic aspartic protease. Stop the activity of this enzyme, and it might be possible to prevent amyloid-β from accumulating in the brain.

BACE1 has proven to be a remarkably challenging target.

Beta-site APP cleaving enzyme (BACE1, also known as memapsin-2 and Asp-2) was first cloned and identified as the brain β-secretase enzyme in 1999 by five separate groups. It was an exciting find. The enzyme seemed to be the rate-limiting step in Aβ production, implying that shutting it down would efficiently reduce brain levels of the pathogenic peptide. An aspartic protease, BACE1 was expected to have important structural and functional similarity to the HIV proteases that had been so successfully targeted by protease inhibitors earlier in the decade. The picture looked even rosier two years later, when the first BACE1 knockout mouse was apparently normal and healthy, a promising sign that blocking the enzyme would not result in toxic effects.

Despite all the early promise, said organizer and speaker David Riddell of Wyeth Research, BACE1 has nonetheless proven to be a "remarkably challenging" target. Ten years after its discovery, only one BACE1 inhibitor is now in clinical trials. At the March 24, 2009, meeting of the Biochemical Pharmacology Discussion Group, four out of five of the researchers who discovered BACE1 reviewed the state of the art of drug development focused on this enzyme, describing a few bright spots and some unexpected new findings about its role in the Aβ production pathway.

BACE1: How far we've come

The processing of amyloid precursor protein into the highly fibrillogenic amyloid-β-42 was described in a backgrounder on BACE1 by Robert Vassar of Northwestern University, who detailed the history of the discovery and validation of the enzyme. Glucose deprivation elevates BACE1 levels; since glucose metabolism is impaired in early AD, it is possible that increased BACE1 may be an important early step in the pathogenesis of the neurodegenerative disorder. Vassar's research points to the translation initiation factor elF2α, which is phosphorylated during conditions of glucose deprivation, as a mediator of increase in BACE1 mRNA translation.

When the first BACE-1 knockout mouse was generated in 2001, the animals appeared to be completely normal, but more recent studies have documented subtle adverse phenotypes, warned Lisa McConlogue of Elan Pharmaceuticals. These include hypomyelination and behavioral patterns associated with schizophrenia. Encouragingly, studies with BACE1 heterozygous knockout animals suggest that even a partial suppression of BACE1 activity can result in a dramatic reduction in Aβ plaques over the long term.

Even partial suppression of BACE1 may result in a dramatic reduction of amyloid plaque.

Many candidate BACE1 inhibitors have fatal flaws—typically, they are too big to get into the brain. Jordan Tang of the Oklahoma Medical Research Foundation described his team's efforts to develop a BACE1 inhibitor and presented data from one candidate that successfully inhibits Aβ production in a widely used transgenic mouse model of the disease. He briefly described results from the first phase 1 clinical trial of a BACE1 inhibitor developed by CoMentis. Ishrut Hussain of GlaxoSmithKline followed with an update on GSK's BACE1 inhibitor program, focusing on a compound that does not readily reach the brain but has good drug properties such as selectivity and potency, providing proof of mechanism.

Wyeth's efforts to develop small-molecule BACE1 inhibitors led to an interesting discovery described by David Riddell—the suggestion that contrary to general assumption, BACE1 may not be the rate-limiting step in amyloid production, at least in the one commonly-used model for the disease, the transgenic Tg2576 mouse.

One implication of this newer research is that targets upstream of BACE1 may prove to be more tractable. On the other hand, the hopes for this enzyme may yet be rewarded: some findings indicate that even partial inhibition of BACE1 or normalization of BACE1 levels can substantially reduce amyloid accumulation.

BACE1 in Drug Development

Speakers:
Jordan Tang, Oklahoma Medical Research Foundation
Ishrut Hussain, GlaxoSmithKline R&D
David R. Riddell, Wyeth Discovery Neuroscience

Highlights

• Small, potent BACE1 inhibitors have been identified with good drug properties.
• A BACE1 inhibitor has been developed that rescues age-related cognitive deficits in transgenic 2576 mice.
• In a phase 1 clinical trial, a BACE1 inhibitor developed by CoMentis was safe and significantly reduced plasma Aβ in healthy volunteers.
• BACE1 inhibitors are 4- to 10-fold less potent against APPsw than APPwt in cell preparations.
• Chronic BACE1 inhibitor dosing leads to significant decrement in Aβ plaque in the brain.
• BACE1 appears not to be rate-limiting under conditions of APPsw overexpression.

BACE1 reaches the clinic

The search for effective BACE1 inhibitors has not been completely fruitless. Jordan Tang of the Oklahoma Medical Research Foundation described the strategy his group has used to identify a suitable inhibitor. They worked toward a transition-state mimic, a strategy that had been successful in development of other protease inhibitors, and they knew the challenges: A successful BACE1 inhibitor must be potent, with a Ki in the low nanomolar range, small enough to cross the blood-brain barrier (between 550 and 650 Da is a "realistic" size), and selective with respect to other aspartic proteases such as memapsin 1 and cathepsin D.

A first generation BACE1 inhibitor, here shown in green in the active site cleft, is potent but too large to be effective in vivo.

Lacking a large library of compounds to screen, the group turned to structure-based design. OM99-2, designed so that the scissile bond is substituted by a hydroxyethylene transition-state isostere, was an early key BACE1 inhibitor. It was potent but large.

The group determined the crystal structure of the catalytic domain, using that structure-activity information to design new inhibitors by cutting outside residues while increasing protein-ligand interactions in the central residues. Through an iterative process of refinement, using enzyme interaction data and information about Aβ inhibition in cells, they gradually reduced the size and maintained the potency. "A lot of this is trial and error," said Tang. They settled on lead compounds with good selectivity with respect to other aspartic proteases, most importantly cathepsin D, which is nearly 1000 times more prevalent than BACE1 in the body.

Continuous delivery of inhibitor GRL-8234 in Tg2576 mice reduced soluble Aβ by about 50% in plasma, and about 40% in the brain. It did not rescue older mice with pre-established amyloid plaque, but if given before plaque onset began and continued for several months, it reduced plaque size by 50% and rescued cognitive decline. In this animal model, "even though we inhibit only 50% of Aβ production and still have robust plaque at the end, we don't need to eliminate all the plaques to rescue cognitive decline," Tang said. With BACE1 inhibited, APP processing apparently shifts to the β-secretase enzyme, known to be less pathogenic.

In collaboration with CoMentis, which he cofounded, Tang's group has launched the first clinical trial of a BACE1 inhibitor. Healthy volunteers given six intravenous doses of the drug tolerated the treatment and showed a robust, sustained reduction of plasma Aβ without a subsequent rebound.

The inhibitor inhibited

Recent findings from GlaxoSmithKline's program in BACE1 inhibitors were described by Ishrut Hussain. For this team as well, rational drug design proved more fruitful than the high-throughput screening approach, and again, the team sought a compound that would replace the scissile bond with a stable transition-state mimic. GSK 188909 was potent and selective, with a molecular weight of about 600. The compound effectively lowered Aβ in cells transfected with either wild-type or Swedish-mutant APP (APPsw), but unfortunately was also a good substrate for P-glycoprotein (P-gp), limiting its efficacy: in vivo, only 5% of the compound reached the brain. Predosing young transgenic mice with a P-gp inhibitor to facilitate brain entry revealed that the compound was indeed effective in lowering soluble Aβ. Radioligand assays confirmed that the compound binds to BACE1. Combining binding and in vivo data, the team estimated that greater than 90% receptor occupancy was required for efficacy. Encouragingly, acute dosing studies for a maximum of five days without the P-gp inhibitor also reduced brain Aβ.

Ultimately, "suboptimal pharmacokinetics" and the necessity for high doses render this compound less than ideal for development, Hussain said, but the fact that it is selective and potent provides important proof of mechanism. Such a compound could potentially be given to patients along with a P-gp inhibitor, much as CYP3A4 inhibitors are given along with certain protease inhibitors to treat HIV. However, Hussain pointed out, this approach was not ideal for elderly AD patients, who are frequently treated with polypharmacy, due to the potential for adverse drug-drug interactions.

An unexpected finding

Unique among these groups, researchers at Wyeth found promising leads from high-throughput screening for potential BACE1 inhibitors, as David Riddell described. These small molecules, unlike the peptimimetics described by other groups, more easily achieve good brain penetration along with selectivity and reasonably good potency.

BACE1 may not always be the rate-limiting step in amyloid production.

Their work with a series of inhibitors led the team to question the assumption that BACE1 is the rate-limiting step in Aβ production, at least when APPsw is the substrate, as it is in the Tg2576 mouse. Their findings suggested that with APPsw as the substrate, BACE1 needs to be inhibited by greater than half to have an effect on Aβ.

"It seems to suggest that in wild-type animals, BACE1 is potentially rate limiting, whereas in transgenic mice, it is no longer rate-limiting," said Riddell. The crucial implication: Where BACE1 inhibitors are concerned, the Tg2576 mouse may be a misleading model. These BACE1 inhibitors may still be useful; as McConlogue pointed out, reducing A-β by even a small amount can have powerful effects if sustained over time.

The first rush of enthusiasm over BACE1 as a therapeutic target for AD has certainly dimmed, as the last 9 years presented challenges that were unforeseen in those early days. Nonetheless, given the key role of this enzyme in the amyloid cascade, BACE1 inhibition may yet prove to be a good way to delay the onset or limit the progress of amyloid accumulation. In any case, drug development efforts have yielded important insights into the pathogenesis of this neurodegenerative disease.