Abstract
The amyloid hypothesis posits that the amyloid-β aggregates in the brain initiate a cascade of events that eventually lead to neuron loss and Alzheimer’s disease. Recent clinical trials of passive immunotherapy with anti-amyloid-β antibodies support this hypothesis, because clearing plaques led to better cognitive outcomes. Orally available small molecule BACE1 inhibitors are another approach to slowing the buildup of plaques and thereby cognitive worsening by preventing the cleavage of amyloid-β protein precursor (AβPP) into amyloid-β peptide, the major component of plaques. This approach is particularly attractive because of their ease of use, low cost, and advanced clinical stage. However, although effective in preventing amyloid-β production in late-stage clinical trials, BACE inhibitors have been associated with early, non-progressive, likely reversible, cognitive decline. The clinical trials tested high levels of BACE inhibition, greater than 50%, whereas genetics suggest that even a 30% inhibition may be sufficient to protect from Alzheimer’s disease. Aside from AβPP, BACE1 cleaves many other substrates in the brain that may be contributing to the cognitive worsening. It is important to know what the cause of cognitive worsening is, and if a lower level of inhibition would sufficiently slow the progress of pathology while preventing these unwanted side effects. Should these side effects be mitigated, BACE inhibitors could rapidly move forward in clinical trials either as a primary prevention strategy in individuals that are at risk or biomarker positive, or as a maintenance therapy following amyloid clearance with an anti-amyloid antibody.
INTRODUCTION
The amyloid hypothesis suggests that the accumulation of the amyloid-β (Aβ)1–42 peptide leads to a cascade of neuronal pathology that eventually causes dementia. The Aβ peptide is directly neurotoxic, and various mechanisms including membrane disruption and pore formation have been proposed as a means of neurotoxicity [1]. From the earliest pathological characterization of Alzheimer’s disease (AD), one of the hallmarks of pathology is neuritic dystrophy surrounding plaques. These neuritic dystrophies cause a disruption of calcium signaling and action potential propagation through affected neurons [2]. Furthermore, it is hypothesized that the dystrophic neurites might be a site of tau seeding and spreading to unaffected areas.
As of the writing of this text, there are two FDA approved anti-amyloid antibodies approved for mild cognitive impairment due to AD and early AD, aducanumab and lecanemab, with a third, donanemab, pending FDA approval. Eli Lilly’s Phase III clinical trial of donanemab, an antibody targeting an N-terminal truncated form of Aβ42, demonstrated a 35% and 36% reduction in Integrated Alzheimer’s Disease Rating Scale and Clinical Dementia Rating Scale-Sum of Boxes scores, respectively, in individuals with low and intermediate tau pathology [3, 4]. Furthermore, nearly 1/3 of individuals cleared amyloid below 24.1 centiloids within the first 6 months of treatment, which neared 80% by the study’s termination at 18 months. Eisai and Biogen’s lecanemab, which binds to soluble Aβ42 protofibrils, demonstrated in their Phase III clinical trial a > 50 centiloid reduction in amyloid PET compared to baseline after 18 months of treatment, which was accompanied by a statistically significant reduction in cognitive decline and function [5, 6]. Biogen’s aducanumab, which binds soluble oligomers and insoluble fibrils, demonstrated clinical efficacy in individuals on the highest dose of drug in a post-hoc analysis, which was accompanied by a 71% reduction in amyloid [7]. Collectively, this data supports the notion that amyloid plays a role in cognitive decline in humans, and a reduction in amyloid levels may improve cognitive outcomes.
Although promising, in contrast to orally bioavailable small molecule drugs, anti-amyloid immunotherapies presently require intravenous infusion, although there are ongoing clinical trials to improve ease of use through subcutaneous delivery. Anti-amyloid is costly, with current estimates in the tens of thousands of dollars. Furthermore, nearly all anti-amyloid immunotherapies have led to an increased risk for amyloid related imaging abnormalities (ARIA) edema or hemorrhaging, although these are symptomatic in a small proportion of patients. This ARIA is likely due to rapid amyloid clearance in the vasculature, in combination with robust microglial activation by the antibodies [8]. Last, amyloid therapies administered at a later stage of disease, in a moderate to severe population, have yet to demonstrate any slowing of cognitive decline. This may be due to the propagation of tau spreading that has already triggered an irreversible rate of degeneration. In an attempt to meet this unmet need for safe and affordable treatments that could potentially be used early on before robust disease, small molecule inhibitors of the proteases that lead to the production of plaques, including β-amyloid cleaving enzyme-1 (BACE1), were developed.
In addition to BACE inhibitors, other small molecules are also in development that aim to inhibit the formation of amyloid plaques. Gamma secretase modulators are a promising approach to shift Aβ from longer (Aβ42/43) to shorter (Aβ37/38) isoforms by modulating γ-secretase-mediated cleavage of amyloid-β protein precursor (AβPP) while sparing any effect on total Aβ or other substates, such as Notch [9]. Other small molecules are being tested for their ability to inhibit Aβ aggregation [10].
BACE
Aβ peptide is formed by the cleavage of AβPP by BACE1 in the luminal compartment of acidic vesicles, followed by cleavage of the remaining β-C terminal fragment by the γ-secretase. BACE1 has been shown to be essential in the formation of amyloid plaques, and mice overexpressing human AβPP fail to develop plaques when genetically deficient in BACE1. Mice deficient in BACE1 from birth have high early postnatal mortality rate, defects in myelination, seizures, and memory deficits (Table 1). However, these defects are lost in mice in which BACE1 is conditionally deleted in adulthood.
Phenotype of BACE1 deficient mice
The defects seen in BACE deficient mice are likely due to the fact that in addition to AβPP, BACE1 cleaves a multitude of other substrates, typically type I transmembrane proteins (Table 2). Many of these substrates have important functions in the nervous system, including dendritic spine formation, myelination, calcium channel activity, and axon guidance. Two notable substrates are seizure protein 6 (Sez6) and close homolog of L1 (CHL1). Cranial window imaging in mice revealed that treatment with a BACE inhibitor leads to lower spine density, and in particular the formation of new spines [22]. This reduction in spines is likely due to inhibition of BACE1 of Sez6, because mice deficient in Sez6 had no further reduction in dendritic spines when treated with a BACE inhibitor compared to Sez6 deficient mice treated with a vehicle control [23]. CHL1, a molecule involved in axon guidance, is a BACE1 substrate, and highly enriched in the mossy fiber pathway of the hippocampus [24]. Mice deficient in CHL1 have a defect in neuron projections in the mossy fiber pathway, which is phenocopied in BACE1-deficient mice from birth [25]. This defect is maintained when BACE1 is deleted from mice in adulthood [21]. Failure to cleave these substrates may lead to the cognitive worsening demonstrated in mice and humans treated with BACE inhibitors, described below.
Selected BACE1 substrates and their functions
CLINICAL TRIALS
Efficacy
Given the essential role for BACE in the formation of plaques, orally bioavailable brain-penetrant small molecule BACE1 inhibitors were developed and tested in both pre-clinical and large-scale clinical trials (Table 3). The inhibitors demonstrated on-target efficacy by showing a dose dependent reduction of Aβ42 in the cerebrospinal fluid. Other metrics, such as a increase in sAβPPα and decrease in sAβPPβ echoed the on-target effects. Furthermore, in three of the phase II/III clinical trials in which this metric was investigated, there was a statistically significant reduction in plaque load measured by amyloid PET scans [41, 42] (https://clinicaltrials.gov/ct2/show/results/NCT02956486). However, there were no benefits to cognition, measured by a variety of cognitive tests. In fact, individuals receiving inhibitors showed a small but significant reduction in specific metrics of cognition.
Outcomes of BACE1 inhibitor clinical trials. Reproduced from McDade et al., Nat Rev Neurol (2021) [43]
Side effects
Early molecules developed to inhibit BACE had off-target effects via inhibition of other important proteases, such as Cathepsin D [51]. The molecules that moved forward in trials were specific for BACE1, and its close homologue BACE2, which is found primarily in peripheral tissues. The molecules tested in clinical trials have various preferences for BACE1 or BACE2, and it is unknown whether side effects caused by the inhibitors are due to inhibition of BACE1 or BACE2; however, due to its high expression in the brain it is likely that the effects of cognition are due to BACE1 inhibition.
Elevation of hepatic enzymes were observed with atabecestat, but otherwise not a significant issue in other BACEis [52]. Other common side effects included an increase in falls, sleep disturbances, anxiety, and suicidal ideation. Most prominently, there was an early, non-progressive, and reversible decrease in measures of cognition and hippocampal volume. Atabecestat noted cognitive worsening in Preclinical Alzheimer’s Cognitive Composite (PACC) and Repeatable Battery for the Assessment of Neuropsychological Status (RBANS), particularly in episodic memory tasks [52]. There was also a decrease in total brain volume that was not correlated with cognitive worsening. Lanabecestat demonstrated no change on global cognition for its clinical trial AMARNTH, but a decrease in specific measures including RBANS total score, immediate memory, and visuospatial construction [46]. On verubecestat, there was no change in global cognition in the EPOCH clinical trial, except in patients with established dementia at 13 weeks in ADAS-Cog [42]. In the APECs trial, there was worse performance in the primary endpoint CDR-SB for the highest dose of verubecestat, CCD-3D total score, attention and processing, episodic memory, and digit symbol coding [46]. With umibecestat, there was worsening in delayed and immediate memory and hippocampal volume [48]. With the trial LY3202026 NAVIGATE-AD, there was no worsening of cognition, but there was a noted a reduction in hippocampal volume. Finally, with elenbecestat there was no statistically significant worsening in cognition.
Interestingly, the trials noting the greatest worsening on cognition were in cognitively unimpaired individuals or individuals at the earliest states of AD. The latter trials that did not have a notable effect tested individuals at later stages of AD.
Reversibility of cognitive worsening
Given that the negative effects on cognition were replicable across different inhibitors and clinical trials, it was important to determine if the cognitive decline was reversible. Indeed, atabecestat and umibecestat followed up with study participants after BACEi withdrawal and found a reversibility of worsening in PACC and RBANS for atabecestat, as well as delayed and immediate memory for umibecestat [52]. With atabecestat and umibecestat, participants also received follow-up MRIs after BACEi withdrawal, confirming that the decrease in hippocampal volume too was reversible.
Improvements in verbal domains of cognition
Intriguingly, there was a statistically significant improvement in language in multiple clinical trials that noticed deficits in other domains of cognition. Specifically, there was an increase in the RBANS language index on lanabecestat, better performance in category fluency and letter fluency on verubecestat, and improvements in verbal scores on umibecestat [46].
MOVING FORWARD WITH BACE INHIBITORS
Timing
Anti-amyloid immunotherapy has demonstrated that disease-modifying slowing of cognitive decline is achievable in AD and supports an important role of amyloid in the disease [53]. However, the modest improvements in cognition and activities of daily living suggest that by the time clinical symptoms manifest, a complex “cellular phase”, mediated by glia and other cells, has taken over and is leading to neurodegeneration and cognitive impairment [54]. This, coupled with the only modest reduction in amyloid in individuals on BACE inhibitors, suggests that treatment with BACE inhibitors should begin in the presymptomatic stage of AD for a primary prevention strategy. Alternatively, treating with inhibitors after clearance of amyloid with anti-amyloid immunotherapy could preserve cognition and keep from further decline.
Preclinical studies support treatment with BACE inhibitors at early stages of amyloidosis. Serial amyloid PET imaging in PS2APP mice revealed that BACE inhibitors are more effective at slowing the progression of mice with a lower plaque burden at baseline compared to a higher burden at baseline [55]. In regions where vehicle-treated mice had a low progression of plaque size increase, BACE inhibitor was able to completely prevent plaque progression. This is compared with regions in which vehicle-treated mice had a high progression of plaque size increase, where BACE inhibitor-treated mice were only able to slow progression by 40%. Additionally, in cranial window imaging, an 80% reduction of Aβ40 and Aβ42 in the forebrain by BACE inhibition slowed the buildup of new plaques by a factor of 12; however, existing plaques had only a modest reduction in plaque growth [56]. The benefit of BACE inhibition on prevention of new plaque growth was particularly pronounced in areas distant from existing plaques, and less beneficial in areas close to existing plaques. In a study monitoring spine dynamics in vivo, APPNL - G - F knock-in mice had defects in spine plasticity and in particular the formation of new spines, and this defect was rescued when mice were treated at a young age with a low plaque burden [57].
In a separate study using a γ-secretase inhibitor, treatment of Tg2576 mice for 3 months early in plaque pathology led to a reduction in plaque pathology at 15 months, whereas mice treated later on when there was already significant plaque pathology had almost identical plaque loads to untreated mice at 15 months [58]. Thus, evidence from preclinical models overwhelmingly favors the use of BACE inhibitors when levels of plaque pathology are low.
In order to treat at low levels of pathology in humans, there needs to be widely accessible biomarkers for amyloid pathology. Studies in individuals with autosomal dominant forms of AD have shed light on the relationship between amyloid alterations and cognitive decline [59]. These longitudinal studies suggest that Aβ42 in the cerebrospinal fluid starts declining as early as 25 years before expected symptom onset, likely due to its being sequestered in amyloid aggregates in the brain. Also, amyloid plaques are detectable in the brain at least 15 years before symptom onset by amyloid PET [59]. Although useful in predicting conversion to dementia, CSF and PET imaging are costly and invasive, and may be difficult to access in a rural setting. As an alternative, mass spectrometry and ultrasensitive immunoassays have been developed to detect AD biomarkers in the blood [60]. Specifically, Aβ42/Aβ40 ratios, pTau181, pTau217, and pTau231 correlate highly with amyloid pathology [61]. In many studies, these blood-based biomarkers were able to predict progression to AD and could discriminate between AD and non-AD dementia.
Early detection of amyloid-positive individuals, particularly those that are ApoE ɛ4-positive or have a family history of dementia, may enable dosing with BACE inhibitors before any symptoms arise. This would be akin to statins for heart disease, whereby individuals who are positive for the serum cholesterol biomarker and therefore are at risk for heart disease take low-dose statins for the duration of their lives as a preventative measure. However, it would be imperative for there to be no cognitive worsening caused by BACE inhibition, as this would likely be an unacceptable side effect in cognitively unimpaired individuals despite its seeming non-progressive and reversible nature.
Combination therapy
Early dosing before cognitive symptoms begin may be the most attractive strategy for BACE inhibition, but may not be the most feasible to immediately implement in the clinic. Another approach could be to completely remove amyloid by anti-amyloid immunotherapy, then followed with low-level BACE inhibition. In fact, this was the strategy employed by the GENERATION clinical trial by Eli Lily, whereby donanemab was to be dosed, followed by lanabecestat. The lanabecestat arm of the clinical trial was terminated due to discontinuation of its other trials with lanabecestat [62]. In preclinical studies, co-treatment with weekly gantenerumab and daily high-dose RO5508887 BACE inhibitor in APPLondon mice with a pre-existing plaque load led to nearly complete plaque abrogation after 5 months of treatment. Co-treatment performed better than mono-treatment with either gantenerumab or RO5508887 alone [63]. Similarly, co-treatment of plaque-bearing APPLondon mice with weekly anti-pyroglutaminated Aβ antibody BAMB31 and daily atabecestat for 11 weeks lowered Aβ42 levels and plaque load to baseline levels [62]. Although these studies support co-treatment, neither studied investigated the effect of antibody administration co-administered with a lower dose BACE inhibition, or antibody administration followed by low dose BACE inhibition.
To the latter point, it is important to know the rate at which plaques reemerge after clearance with an anti-amyloid antibody. If there is rapid reemergence of plaques, it is unlikely that a low dose of BACE inhibition would be efficacious. In the phase III clinical trial of donanemab, it was demonstrated that donanemab fully cleared amyloid by 12 months of treatment. Computational modeling of follow-up PET scans estimated that a period of 4.7 years following cessation of donanemab treatment may be necessary for individuals to reach the threshold of positivity by amyloid PET.[64] Following treatment of donanemab with low dose lanabecestat could lead to a much greater length of time than 4.7 years until amyloid positivity (Fig. 1).
Dosing
In clinical trials, the lowest dose of BACE inhibitor explored inhibited Aβ42 production by just over 50% with lanabecestat, but most inhibitors inhibited Aβ42 by over 70%. Because of the cognitive worsening experienced at the higher doses of BACE inhibitor, there is interest in determining whether lower doses of BACE inhibitor would spare individuals of the cognitive worsening, while still delaying the buildup of plaques. In the clinical trial of lanabecestat, the lower dose of 20 mg, that inhibited Aβ42 in the CSF by just over 50%, led to a statistically significant –13.7 reduction of mean centiloid change of amyloid after 104 weeks measured by amyloid PET from baseline [41].
Genetically, there is also evidence in support of a moderate reduction of Aβ42 production to protect from AD. In a cohort of Icelanders, a rare coding mutation in AβPP was found more frequently in healthy elderly controls compared to age-matched individuals with AD (0.62% versus 0.13%, odds ratio – 5.29) [65]. This AβPP mutation, A673T, is located within the BACE recognition sequence, reduces BACE cleavage of AβPP by 40% in vitro, and slightly reduces the aggregation efficiency of Aβ42 [65, 66]. Furthermore, a study conducted on healthy non-demented Finnish men demonstrated that carriers of this mutation have a 28% reduction of Aβ40 and Aβ42 in their plasma [67]. Together, these data suggest that reducing Aβ42 by ~ 30% may be the target to aim for with BACE inhibition to reduce the risk of developing AD.
Preclinical studies support that a partial inhibition of BACE may not lead to cognitive worsening. In rat primary neurons treated with varying doses of BACE inhibitors, high doses significantly decreased synaptic transmission, whereas lower doses that inhibit Aβ production by 50% or less have no detrimental effect on synaptic transmission [68]. In fact, the a low dose of BACE Inhibitor IV led to a slight improvement in synaptic transmission, potentially by promoting the production of the neuroprotective sAβPPα [69].
SUMMARY
The only FDA-approved disease-modifying drugs for AD, the Aβ immunotherapies, are costly, relatively invasive, and may lead to potentially harmful side effects. As an alternative, small molecule orally bioavailable BACE inhibitors were developed and tested in several late-stage clinical trials. These inhibitors were effective in lowering amyloid beta levels, but often led to mild non-progressive and reversible cognitive decline. This may be due to inhibition of cleavage of other BACE substrates, such as those involved in dendritic spine formation or axon guidance. If the side effects were mitigated, such as through a low dose, distinct formulation, or other pharmacological or non-pharmacological approaches, BACE inhibitors could be used as a maintenance therapy after initial clearance of amyloid plaques with an antibody, or in a primary prevention strategy in individuals who have high genetic risk before AD pathology even begins.
AUTHOR CONTRIBUTIONS
Robert Vassar, PhD (Conceptualization; Resources; Writing – review & editing); Elyse A. Watkins, PhD (Conceptualization; Writing – original draft; Writing – review & editing).
Footnotes
ACKNOWLEDGMENTS
The authors have no acknowledgments to report.
FUNDING
The authors have no funding to report.
CONFLICT OF INTEREST
The authors have no conflict of interest to report.