|Home | About | Journals | Submit | Contact Us | Français|
Since BACE1 was reported as the β-secretase in Alzheimer's disease (AD) over ten years ago, encouraging progress has been made toward understanding the cellular functions of BACE1. Genetic studies have further confirmed that BACE1 is essential for processing amyloid precursor protein (APP) at the β-secretase site. Only after this cleavage can the membrane-bound APP C-terminal fragment be subsequently cleaved by γ-secretase to release so-called AD-causing Aβ peptides. Hence, in the past decade, a wide variety of BACE1 inhibitors have been developed for AD therapy. This review will summarize the major historical events during the evolution of BACE1 inhibitors designed through different strategies of drug discovery. Although BACE1 inhibitors are expected to be safe in general, careful titration of drug dosage to avoid undesirable side effects in BACE1-directed AD therapy is also emphasized.
AD, a progressive neurodegenerative disorder, is the most common form of dementia that affects 4–8% of the elderly population worldwide. The neuropathological features of AD are the presence of extracellular amyloid plaques and intracellular neurofibrillary tangles (NFT) in the hippocampus and cortical grey matter of the AD brain. The amyloid plaques, also called senile plaques, result from the deposition of aggregated Aβ peptides [1, 2]. Although the naturally produced forty amino acid Aβ40 is the predominant product (~90% abundance), the slightly longer product Aβ42 tends to aggregate faster and is the major form of Aβ peptides that are linked to AD pathogenesis [3, 4]. This notion is further supported by the finding that most familial AD (FAD) mutations lead to overproduction of Aβ42 years before AD symptoms appear . Formation of NTF is via a totally different route, and NTF are composed of hyperphosporylated microtubule binding protein tau [6, 7]. The occurrence of NTF is likely a downstream event to amyloid deposition , an observation that further highlights the importance of reducing brain levels of Aβ peptides, especially of Aβ42, as a promising strategy for AD therapy. β-secretase acts on the rate-limiting first step in Aβ generation and has therefore been a focus of intense research as a prime drug target. The gene encoding β-secretase was discovered a decade ago, and is now more commonly known as BACE1 for Beta-site APP-Cleaving Enzyme 1, Asp2 or memapsin 2 [9-13]. The most exciting finding is from the studies using BACE1-null mice, which demonstrate that genetic deletion of BACE1 abolishes the production of Aβ [14-16]. Hence, inhibition of BACE1 activity could reduce the production of all forms of Aβ, thereby diminishing plaque formation.
To inhibit BACE1 effectively, knowledge of the structure and function of BACE1 is essential. BACE1 is one of the only two known aspartyl proteases that are localized on the cell membrane due to the presence of a membrane-anchoring domain. Although BACE1 and BACE2 share 59% homology and adopt a similar structural organization, BACE2 preferentially cleaves APP within the Aβ region between F19 and F20 and between F20 and A21 [17, 18]. Mice with a deficiency in BACE2 are fertile and healthy , making cross inhibition of BACE2 by BACE1 inhibitors less of a concern.
Like other aspartyl proteases, BACE1 has a classical bilobal structure, with each of its two active aspartate motifs (D93TG and D289SG) located in a separate lobe, and mutation of either aspartate residue renders the enzyme inactive. BACE1 is first synthesized in the endoplasmic reticulum (ER) as a 501-aa immature precursor protein (proBACE1). Interestingly, proBACE1 exhibits considerable enzymatic activity, unlike other aspartic zymogens, because its prodomain fails to block the entrance of its substrate into the active pocket of BACE1 [20, 21]. This unique property of proBACE1 suggests that inhibition of its prodomain shedding would be ineffective in reducing APP processing and Aβ production in AD brains.
As a type I transmembrane protein, BACE1 becomes mature in the Golgi compartment by extensive posttranslational modifications which include glycosylation , the removal of the prodomain by furin-like proprotein convertases [23-26], reversible acetylation on seven Lys residues  and palmitoylation . Even though BACE1 is transported to all secretory compartments after maturation, its transmembrane domain preferentially promotes the retention of this enzyme in the late Golgi and trans-Golgi network (TGN) compartments , where luminal BACE1 protease domain will encounter and cleave the type I transmembrane APP at the β-site.
BACE1 enzymatic activity is affected by the environmental pH, with its maximum proteolytic activity at ~pH 4.5 as shown by in vitro assays [29, 30]. Thus, BACE1 likely processes APP more efficiently in the relatively more acidic subcellular compartments (i.e., TGN and endosomes) than in the more neutral ER compartment [31-33]. Our recent studies suggest that increased expression of the BACE1-interacting protein reticulon 3 (RTN3) significantly increases retention of BACE1 in the ER, and this altered cellular trafficking can partially contribute to the reduced cleavage of APP by BACE1 [34, 35]. On the other hand, BACE1 trafficking to the endosomes is regulated via the phosphorylation state on Ser498  or dileucine residues (L499L500) in its short cytoplasmic tail [36, 37]. Golgi-localized γ-ear containing ADP ribosylation factor-binding (GGA) family proteins appear to mediate the retrograde transport of BACE1 from endosomes to TGN [38-42]. Depletion of GGA proteins by RNAi or disruption of phosphorylation on Ser498 increases retention of BACE1 in endosomes, whose favorable acidic environment enhances BACE1 activity [43, 44]. Palmitoylation of BACE1 facilitates its localization in the lipid rafts [45, 46], where APP is also enriched and efficiently processed by BACE1 [47, 48].
The ultimate evidence supporting BACE1 as a superior AD drug target is the finding that BACE1 knock-out mice do not produce Aβ [14-16]. BACE1-null mice are fertile and exhibit relatively mild phenotypes such as hypomyelination [49, 50] and schizophrenia-like behaviors . This is in sharp contrast to mice with deficiencies in components of the γ-secretase complex . For example, mice deficient in presenilin-1 , nicastrin , or APH-1  all exhibit serious developmental defects and die at early embryonic stages because γ-secretase is required to process Notch, a signaling molecule that controls cell fate in embryonic and adult myelopoiesis [56-58]. Therefore, compared with γ-secretase, BACE1 is viewed as a better drug target whose inhibition should cause less serious biological dysfunction.
Developing potent BACE1 inhibitors in a hope to identify suitable AD drug candidates has been fiercely pursued for the past decade. Several strategies of drug discovery have been explored in the search for potent BACE1 inhibitors, e.g., substrate-based design, high-throughput screening, and fragment-based lead generation approaches. In the following text, we will briefly review the history and current preclinical situations of BACE1 inhibitors being developed by these different approaches.
Substrate-based methods have often been used as the starting point for developing aspartyl protease inhibitors. The first substrate-based BACE1 inhibitor, P10-P4’ StatVal (Table 1, #1), was developed by Elan Pharmaceuticals in order to purify β-secretase from human brain homogenates . This peptidic inhibitor is a P1 (S)-statine substituted substrate analogue with an in vitro half-maximal inhibitory concentration (IC50) of ~30 nM. Shortly after the molecular cloning of memapsin 2 (BACE1), Tang (Oklahoma Medical Research Foudation) and Ghosh (University of Illinois at Chicago/Purdue University) teamed up and reported the octapeptidic, hydroxyethylene (HE) isotere-based transition-state analogue inhibitor OM99-2 (Table 1, #2) and the crystal structure of BACE1 with OM99-2 bound to its active site . While OM99-2 showed excellent inhibitory potency in vitro (Ki = 1.6 nM), its bulky peptidic structure precluded its application in vivo. Nonetheless, the BACE1/OM99-2 co-crystal structure provided coveted molecular insight into the ligand binding interactions in the enzyme active site and significantly advanced the BACE1 inhibitor design by researchers around the globe. In parallel, Kiso's group (Kyoto Pharmaceutical University) developed their own octapeptidic BACE1 inhibitor KM I-008 (cellular IC50 = 413 nM) using a hydroxymethylcarbonyl (HMC) isostere as a transition-state mimic . Further chemical modification of KMI-008 yielded more potent pentapeptidic BACE1 inhibitors KMI-420 (in vitro IC50 = 8.2 nM) and KMI-429 (in vitro IC50 = 3.9 nM) (Table 1, #3) . KMI-429 appears to significantly reduce brain Aβ production when directly injected into the hippocampus of both wild-type mice and APP transgenic mice (Tg2576 mice) . During this period of time, various substrate-based peptidomimetic inhibitors were also developed by large pharmaceutical companies and other academic research groups [63-65]. Although these peptidomimetic BACE1 inhibitors are highly potent in vitro, their inherent poor drug properties (i.e., large size, poor brain permeability, short half-life in vivo, and low oral availability) have made them unsuitable drug candidates. On the other hand, guided by a structure-based approach, these first-generation inhibitors have laid the foundation for the rational design of later generations of smaller, non-peptidic BACE1 inhibitors that have significantly improved drug properties.
The BACE1 inhibitors 5d and 24 (Table 1, #4 and #5), developed via chemical modifications of OM99-2, are typical examples of less-peptidic BACE1 inhibitors. Both of them exhibited stronger potency (cellular IC50 equal to 39 nM and 1 nM, respectively) and impressive in vivo efficacy (reduction of plasma Aβ level by 30% and 65%, respectively) when intraperitoneally (i.p.) injected into Tg2576 mice [66, 67]. Further modifications of KMI-420 and KMI-429 produced tetrazole ring-containing compounds KMI-570 (Table 1, #6; in vitro IC50= 4.8 nM) and KMI-684 (Table 1, #7; in vitro IC50= 1.2 nM), both of which also display improved brain permeability .
Encouragingly, GlxoSmithKline reported the first orally available BACE1 inhibitor GSK188909, a small non-peptidic compound developed from substrate-based design (Table 1, #8). GSK188909 displayed a cellular IC50of 5 nM and showed excellent selectivity over other aspartic proteases. When orally administered in TASTPM mice, which express both Swedish mutant APP and PS1 mutant M146V, GSK188909 effectively reduced brain Aβ levels . Subsequently, Schering-Plough also reported an orally effective 4-phenoxypyrrolidine-based BACE1 inhibitor named compound 11 with good phar-macokinetics and selectivity (Table 1, #9; Ki = 0.7 nM, cellular IC50 = 21 nM) . The most exciting news in the race of BACE1 drug discovery was perhaps the debut of CoMentis’ CTS-21166 (Table 1, #10; cellular IC50 = 1.2-3.6 nM), which is the only BACE1 inhibitor that has passed Phase I clinical trial thus far. In 2008, CoMentis revealed this small compound as a transition-state analog inhibitor (structure is currently undisclosed) with excellent properties in brain penetration, selectivity, metabolic stability, and oral availability; all of these have met the requirements of an ideal oral drug candidate . When i.p. injected (4 mg/kg over six weeks) into an aggressive APP transgenic mouse (expressing both the Swedish and London mutations), CTS-21166 reduced brain Aβ levels by over 35% and plaque load by 40%. The data from human Phase I studies suggested that this compound appeared safe at dose as high as 225mg, and when intravenously (i.v.) injected into AD patients, it caused a dose-dependant reduction of plasma Aβ levels for an extended period of time (http://www.alzforum.org/new/detail.asp?id=1790). More thorough clinical evaluation of CTS-21166 is underway by CoMentis in a partnership with Astellas Pharma in Japan. Following this trend, several companies such as Merck, Eli Lilly, and Takeda are also considering Phase I human testing with their own BACE1 inhibitors (http://www.alzforum.org/new/detail.asp?id=1790). Interesting clinical data will likely be available for these inhibitors in the near future.
Though many BACE1 inhibitors were successfully designed as substrate-based analogues, the majority of non-peptidic BACE1 inhibitors were developed using high-throughput screening (HTS) or a fragment-based lead generation approach. HTS has been conducted by many pharmaceutical companies to identify BACE1 inhibiting “hit” compounds from various chemical libraries. In 2001, Takeda reported the first series of non-peptidic BACE1 inhibitors with an in vitro IC50 of 0.35-2.93 μM using this approach . Subsequently, Wyeth reported their hit compound WY-25105 (in vitro IC50 = 3.7 μM, cellular IC50 = 20 μM) containing an acylguanidine moiety that can form key interactions with the two catalytic Asp residues (Asp93 and ASP289) at the BACE1 active site . Structure optimization of this compound has led to the development of a more potent BACE1 inhibitor named 10d with an in vitro IC50 of 0.11 μM (Table 2, #11), and this compound showed an 85-fold selectivity over another aspartic protease, cathepsin D . Using the similar approach, Johnson & Johnson developed the BACE1 inhibitor 3a with a stronger potency (Ki = 11 nM; Table 2, #12) . More importantly, this compound exhibits excellent brain permeability and oral availability. When orally administered at 30 mg/kg to rats, 3a was able to lower plasma Aβ40 levels by 40–70% 3 hours post-dosing. Despite all of these successes, none of these HTS-based inhibitors has entered a clinical trial.
Recently, “fragment-based drug discovery (FBDD)” has emerged as a novel alternative to the traditional HTS method in identifying potent BACE1 inhibitory drugs. In contrast to HTS, which uses libraries of relatively high-molecular-weight compounds, the FBDD approach takes advantage of libraries comprising more diverse and smaller-sized compounds (fragments) to identify hits that can be efficiently developed into potent leads with drug-like properties. After the hits are identified, medicinal chemistry will come into play to further optimize these leads into suitable drugs. Compared with the traditional HTS approach, a significantly higher hit ratio can be obtained using a FBDD approach even though the binding affinity between the lower-molecular-weight fragments and the large active site of BACE1 will be weaker. A variety of biophysical techniques (NMR, X-ray crystallography, surface plasmon resonance (SPR), etc.), computational tools and biochemical assays can be coupled in fragment screening. By a fragment screening approach assisted with NMR followed by X-ray crystallography and SPR assays, AstraZeneca reported a FBDD-based BACE1 inhibitory compound 29 with a cellular IC50 near 470 nM (Table 2, #13) [75, 76]. Astex, in a partnership with AstraZeneca, reported several BACE1 inhibitors developed by fragment screening through X-ray crystallography . Because this approach provides detailed binding mode of the fragments to the BACE1-active pocket, further optimization of these inhibitors by medicinal chemistry can be easily performed. Huang et al. (University of Zürich) reported non-peptidic BACE1 inhibitors using fragment screenings by a computer-assisted docking simulation method [78, 79]. Very recently, Godemann et al. (Evotec AG) identified a series of BACE1 inhibitors by a novel fluorescence-polarization-assay-based fragment screening method in conjunction with X-ray crystallography . The integration of various techniques in the fragment screening has made FBDD an increasingly popular method for designing potent small-molecule BACE1 inhibitors.
It should be emphasized that each of the aforementioned strategies of drug discovery has its pros and cons. While substrate-based BACE1 inhibitors usually show high potency and selectivity, their poor oral availability and permeability across the blood-brain barrier frequently make them unsuitable drug candidates. By contrast, the HTS method has the advantage of generating “hits” with high diversity, smaller size, and more drug-like properties (i.e., oral availability and brain penetration). However, the hit ratio of HTS tends to be extremely low and the “hits” generally have lower potency and selectivity than substrate-based inhibitors. Compared with the traditional HTS method, however, fragment-based screening enjoys much higher hit ratios, and like HTS, can identify leads with favorable drug properties. On the other hand, fragment leads are too small to exhibit satisfactory potency and selectivity, thus requires considerable subsequent modifications by medicinal chemists. Overall, a combinatorial approach that carefully integrates the strengths of different design strategies may find its successful application in the future design of more applicable BACE1 inhibitory drugs.
At the current stage, whether strong inhibition of BACE1 activity will cause untoward mechanism-based toxicity is less of a concern due to the mild phenotypes of BACE1-null mice. However, it has become increasingly clear that this enzyme also cleaves membrane-bound substrates other than APP. One of these demonstrated substrates is neuregulin-1 (Nrg-1) [49, 50], an essential neurotrophic factor that signals through the ErbB family of receptor-tyrosine kinases to regulate various neurological processes, including synaptic functions, neuronal migration, myelination, and neurotransmitter functions, among many others [81-83]. In BACE1-null mice, abolished cleavage of Nrg1 by BACE1 disrupted BACE1-dependant Nrg1/ErbB signaling, causing hypomyelination (thinner myelin) of both peripheral nerves [49, 50] and central nerves  as well as schizophrenia-like phenotypes in these mice . Furthermore, we showed that BACE1-null mice exhibit impaired remyelination in adult mice . Another important BACE1 substrate is the voltage-gated sodium channel β2 (Navβ2) subunit [85, 86]. Voltage-gated sodium channel proteins are composed of a pore-forming α subunit and auxiliary P su bun its . We recently demonstrated that BACE1-null mice display epileptic seizures presumably arising from elevated neuronal surface expression of sodium channel proteins and the resulted increased neuronal excitability . It is highly plausible that the abolished cleavage of Navβ2 by BACE1 leads to these observed alterations, because the disruption of this cleavage in BACE1-null mice will potentially affect both the expression and cellular trafficking of sodium channel proteins. The absence of BACE1 will reduce the shedding of Navβ2 intra-cellular domain that can regulate gene expression of the sodium channel α subunit ; disruption of BACE1-dependant cleavage of Navβ2 will also increase the level of full length Navβ2 protein, which is capable of binding to the α subunit and mediating its trafficking to the membrane surface . Moreover, aged BACE1 -null mice were found to exhibit hippocampal neurodegeneration that is likely attributable to the sustained asynchronous neuronal stimulation in these mice . Together, these important findings strongly suggest that BACE1 performs diverse physiological functions through processing of different substrates. Therefore, careful titration of drug dosage may still be a needed precaution to avoid the potential mechanism-based toxicity incurred by BACE1-directed AD therapy.
In the past decade, massive research efforts have been directed toward understanding BACE1 as a critical target for AD therapy. The crystal structure of BACE1 and our increasing knowledge of the cell biology of this enzyme have paved the way for the development of a wide variety of BACE1 inhibitors. During the preparation of this manuscript, several new classes of BACE1 inhibitors were reported. However, this article is not meant to be an all-inclusive summary of BACE1 inhibitors revealed to date. Instead, we have selectively presented some BACE1 inhibitors that have made pioneering contributions to the field. Although some orally effective BACE1 inhibitors have been reported and at least one (CTS-21166 from CoMentis) of these inhibitors have entered a human clinical trial, it remains a daunting challenge to find small-molecule BACE1 inhibitors with satisfactory pharmacological properties. Due to the inherent disadvantages of each drug design method, combinatorial strategies that integrate the strengths of individual methods might boost the success rate in the design of future generations of more viable BACE1 drugs. Finally, the accumulating knowledge regarding the physiological functions of BACE1 further supports that BACE1 is still a highly viable AD therapeutic target. The optimal inhibition of BACE1 in humans will likely answer whether a so-called “amyloid hypothesis"-based strategy is the best way for improving the cognitive function of AD patients.
R Yan is supported by the grant (AG025493) from the National Institutes of Health as well as grants from Ralph Wilson Foundation, American Health Assistance Foundation, Alzheimer's Association and National Multiple Sclerosis Society. The authors declare no conflict of interest.