Investigational BACE inhibitors for the treatment of Alzheimer’s disease

Bruno P. Imbimbo & Mark Watling

To cite this article: Bruno P. Imbimbo & Mark Watling (2019): Investigational BACE inhibitors for the treatment of Alzheimer’s disease, Expert Opinion on Investigational Drugs, DOI: 10.1080/13543784.2019.1683160
To link to this article: https://doi.org/10.1080/13543784.2019.1683160

Published online: 29 Oct 2019.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ieid20



Investigational BACE inhibitors for the treatment of Alzheimer’s disease
Bruno P. Imbimboa and Mark Watlingb
aResearch & Development, Chiesi Farmaceutici, Parma, Italy; bCNS & Pain Department, TranScrip Partners, Reading, UK

Introduction: The amyloid hypothesis of Alzheimer’s disease (AD) states that brain accumulation of amyloid-β (Aβ) oligomers and soluble aggregates represents the major causal event of the disease. Several small organic molecules have been synthesized and developed to inhibit the enzyme (β-site amyloid precursor protein cleaving enzyme-1 or BACE1) whose action represents the rate-limiting step in Aβ production.
Areas covered: We reviewed the pharmacology and clinical trials of major BACE1 inhibitors.
Expert opinion: In transgenic mouse models of AD, BACE1 inhibitors dose-dependently lower Aβ levels in brain and cerebrospinal fluid (CSF) but the evidence for attenuation or reversal cognitive or behavioral deficits is very scanty. In AD patients, BACE1 inhibitors robustly lower plasma and CSF Aβ levels and reduce brain plaques but without cognitive, clinical, or functional benefit. To date, seventeen BACE1 inhibitors have failed in double-blind, placebo-controlled clinical trials in patients with mild-to- moderate or prodromal AD, or in cognitively normal subjects at risk of developing AD. Several of these studies were prematurely interrupted due to toxicity or cognitive and behavioral worsening compared to placebo-treated patients. Elenbecestat, the last BACE1 inhibitor remaining in late clinical testing for AD, was recently discontinued due to safety concerns.
Received 3 September 2019
Accepted 17 October 2019
BACE inhibitors; BACE1; BACE2; Alzheimer’s disease

⦁ Introduction
Alzheimer’s disease (AD) is a chronic neurodegenerative disease that usually starts slowly and gradually worsens over time. It accounts for around 70% of dementia cases. The commonest early symptom is difficulty remembering recent events and as it progresses, symptoms can include language problems, disorienta- tion in time and space, mood disturbances, apathy, inability to perform activities of daily living and behavioral issues. Eventually, even basic functionalities are lost, ultimately leading to death. Although variable, the typical life expectancy following diagnosis is three to eight years [1]. The cause of AD is poorly understood. The risk of AD is partially driven by genetics; a recent meta-analysis identified 29 risk loci, implicating 215 potential causative genes. Gene-set analyses indicated biological mechanisms involved immunity processes, lipid metabolism, tau binding proteins, and amyloid precursor protein metabolism [2]. Histologically, the AD brain is characterized by extraneuronal plaques of β-amyloid (Aβ), intraneuronal neurofibrillary tangles of hyperphosphorylated-tau, dysfunctional microglia, reactive astrocytes and dystrophic neur- ites. No treatments stop or reverse the progression of AD, though cholinesterase inhibitors or the NMDA-antagonist memantine may temporarily improve symptoms.

⦁ The β-amyloid hypothesis of Alzheimer’s disease
The Aβ cascade hypothesis of AD assumes that the brain accu- mulation of this 40–42 amino acid peptide represents the initial event of the pathological process and starts 15–20 years before
clinical symptoms become apparent. Point mutations of the amyloid precursor protein (APP) and the enzymes involved in its processing (PSEN1 and PSEN2) alter Aβ production and cause the autosomal dominant familial forms of AD. A specific muta- tion of APP (Ala673Thr) is known to protect against AD onset and cognitive decline in cognitively healthy elderly individuals (Figure 1). The mutant amino acid is adjacent to the β-site amyloid precursor protein cleaving enzyme-1 (BACE1) cleavage site and produces a ~ 40% reduction in Aβ production in vitro [3]. In late-onset sporadic AD, accumulation of brain Aβ has variously been attributed to faulty clearance [4] or increased BACE1 activ- ity [5]. Aβ accumulation has also been linked to the apolipopro- tein E ε4 (APOE-ε4) allele, the most important genetic risk factor associated with sporadic AD [6]. These observations underpin the amyloid hypothesis of AD and during the last 20 years inten- sive efforts have been made to identify compounds which can antagonize Aβ accumulation in the AD brain.

⦁ β-site amyoid protein cleaving enzyme (BACE) 1 and 2
BACE1 is a single-transmembrane aspartyl protease expressed mainly in the central nervous system and localized to presynaptic terminals [7]. BACE1 cleaves the extraneuronal portion of APP, generating a large soluble extracellular fragment called sAPPβ and a transmembrane C-terminal fragment (β-CTF), known also as C99. C99 undergoes intramembrane cleavage by the γ- secretase complex (presenilin, nicastrin, anterior pharynx-

CONTACT Bruno P. Imbimbo [email protected] Research & Development Department, Chiesi Farmaceutici, Largo Francesco Belloli 11/a, Parma 43122, Italy
© 2019 Informa UK Limited, trading as Taylor & Francis Group


Article highlights

⦁ Several potent BACE1 inhibitors have been developed to decrease Aβ generation and lower subsequent Aβ oligomer formation and hence decrease neuronal and glial cytotoxicity and slow cognitive decline in AD patients.
⦁ Evidence for positive behavioural effects of BACE1 inhibitors in ani- mal models of AD is very scanty. Few studies have shown BACE1 inhibitors to reverse or attenuate memory and behavioural deficits in transgenic mouse models of AD.
⦁ Despite strong target engagement and achieving up to 80–90% cerebrospinal fluid Aβ reductions in humans, BACE1 inhibitors trialled in patients with mild-to-moderate or prodromal AD, or in cognitively healthy subjects at risk of developing AD, have failed to demonstrate
significant slowing of cognitive decline.
⦁ Moreover, recent full disclosure of the results of such trials has revealed significant cognitive and clinical worsening in subjects receiving BACE1 inhibitors that has led to premature interruption of such trials.
⦁ The reasons for these disappointing results are unclear, but probably include BACE1 vs BACE2 selectivity, the nonselective inhibition of BACE1 activity on substrates besides APP (which are known to be important for diverse biological functions) or even the accuracy of
the theory that Aβ accumulation is responsible for AD.
⦁ BACE1 is not a viable therapeutic target in AD.
This box summarizes key points contained in the article.

defective 1 and presenilin enhancer 2) releasing the APP intracel- lular domain (AICD or C59) into the cytoplasm and Aβ peptides of varying length into the extraneuronal space (Figure 1). The 42- aminoacid form of Aβ (Aβ1-42) is highly prone to oligomerization and amyloid plaque formation. Since its initial identification and isolation in 1999, BACE1 has become one of the most important therapeutic targets for AD [8] because its action represents the rate-limiting step in the production of Aβ in the brain.

BACE2 is a close homologue of BACE1 and also cleaves APP within the Aβ region. In addition, BACE2 cleaves the melano- cyte protein PMEL in pigment cells of the skin and eye, gen- erating melanin pigments, and appears to have other roles such as glucose homeostasis [9].
Insulin resistance and impaired glucose homeostasis are important indicators of Type 2 diabetes and are early risk factors of AD. Genetic depletion of BACE1 decreases body weight, pro- tects against diet-induced obesity and enhances insulin sensitiv- ity in mice [10] suggesting that BACE1 is a potential regulator of glucose homeostasis. Studies using a neuron-specific human BACE1 knock-in mouse model (PLB4) have shown that centrally expressed human BACE1 induces systemic glucose intolerance, a fatty liver phenotype and impaired hepatic glycogen storage [11]. Since expression and activity of BACE1 is increased in the AD brain [5], these observations provide a potential mechanism for the high prevalence of metabolic disturbance in AD and support the hypothesis that AD may represent a central form of diabetes (Type 3 diabetes).

⦁ Testing of BACE1 inhibitors in transgenic mouse models of ad
In the last 20 years vigorous research targeted the identification of BACE1 inhibitors. After initial difficulties, several groups success- fully synthesized potent small organic molecules with different selectivity for BACE1 and BACE2 and characterized their pharma- cology in transgenic murine models of AD. Several compounds displayed good brain penetration and were shown to lower dose- dependently brain Aβ levels in animals [12]. However, very few studies have shown BACE1 inhibitors to reverse or attenuate memory and behavioral deficits in transgenic mouse models

Figure 1. APP processing and mutations affecting β-secretase cleavage.
(a) APP is sequentially cleaved by two aspartic proteases to generate Aβ. During the first step, BACE1 cuts APP to create the N-terminus of Aβ. Two APP fragments are produced: C99 and sAPPβ. During the second step, C99 is cleaved by the γ-secretase enzyme to generate Aβ (yellow) that is then released into the lumen and secreted into the extracellular space. An intracellular domain, AICD or C59 (pink), is also produced.(b) The aminoacids in and around the Aβ region of APP are represented in different colors based on BACE1 and γ-secretase cleavages. Aminoacids that affect BACE1 processing of APP are shown in circles with capital letters, within which the wildtype residue is identified by the single-letter aminoacid code. The Lys670Asn/Met671Leu (Swedish) and Ala673Val mutations cause familial Alzheimer disease (FAD) by increasing the rate of BACE1 cleavage and Aβ production, whereas the Ala673Thr mutation protects against AD by doing the opposite. All three mutations occur at or within one aminoacid of the BACE1 cleavage site. Scissors show cleavage sites of the various secretases. APP = amyloid precursor protein. Aβ = amyloid-β peptides. sAPPβ = soluble peptide APPβ.

of AD. One study featured a noncompetitive BACE1 inhibitor (TAK- 070) orally administered for 9 days to 18-week old Tg2576 trans- genic mice [13]; another featured a centrally delivered BACE1 inhibitor (Compound C) administered intracerebroventricularly for 8 weeks to 18-month old Tg2576 transgenic mice [14]; a third studied a brain penetrant BACE1 inhibitor (NB-360) admi- nistered in the chow for 6 weeks to 6–8-month old double- transgenic mice (APP23xPS45) [15]. Surprisingly, none of the major BACE1 inhibitors advanced to clinical testing were docu- mented to show behavioral effects in animal models of AD [16]. The effects of genetic BACE1 ablation are complex and depend on the genotype of the animals. In transgenic mice overexpressing human APP (Tg2576) or human APP and PS1 (5xFAD), knocking out the BACE1 gene attenuates or abolish memory deficit [17]. On the other hand, transgenic BACE1−/- normal mice show impair- ment in both spatial and working memories and decreased anxi- ety [18]. These data suggest that intact BACE1 is required for some normal hippocampal memory processes [19] and emotional func- tion [20].

⦁ Major clinical trials with BACE1 inhibitors
In recent years, several BACE1 inhibitors have undergone clin- ical evaluation. Initially, several BACE1 inhibitors were termi- nated during Phase 1 or early Phase 2 studies for liver (LY2886721), ocular (LY2811376), or cardiac (AZD-3839) toxi- city. Other BACE1 inhibitors (verubecestat, atabecestat, lana- becestat, LY3202626, umibecestat) reached late stages of clinical development but failed to show cognitive or functional benefit in large placebo-controlled studies. Late-stage clinical failures occurred in mild-to-moderate AD patients, patients with early stages of AD and even cognitively healthy subjects at risk of developing AD (Table 1). Elenbecestat was the last BACE1 inhibitor in Phase 3 to be abandoned. Instead of describing all of these BACE1 inhibitors, we will review the most recent major Phase 2/3 clinical trials.
In February 2017, an 18-month trial of verubecestat (12 or 40 mg/day) in 1,958 mild-to-moderate AD patients (EPOCH) was stopped for lack of efficacy [21]. From a pharmacodynamic per- spective, verubecestat efficiently lowers Aβ levels in cerebrosp- inal fluid (CSF) in a dose-dependent manner; nevertheless, it has not improved either cognition or activities of daily living in patients. In February 2018, a two-year trial of verubecestat (12 or 40 mg/day) in 1,454 people with subjective memory decline and positive amyloid positron emission tomography (PET) (APECS) reached futility criteria and was halted [22]. Recipients of 40-mg verubecestat performed significantly worse than those on placebo on clinical global performance, as measured with the Clinical Dementia Rating – Sum of Boxes (CDR-SB). Similarly, patients on verubecestat showed worse functionality than to those on placebo as measured with the Alzheimer’s Disease Cooperative Study Group-Activities of Daily Living (ADCS-ADL) scale. These detrimental effects were seen after 3 to 24 months and 9 months of treatment, respectively. Compared to placebo, the drug accelerated (at both doses) the rate of conversion from prodromal AD to overt dementia (25% vs 20% per year). Subjects treated with verubecestat complained more frequently of anxi- ety, depression and disturbed sleep and mean Neuropsychiatric Inventory (NPI) scores worsened in a dose-dependent way,
reaching statistical significance in the 40-mg dose group com- pared to placebo (p < 0.05). Verubecestat recipients also showed a faster reduction in brain and hippocampal volumes compared to placebo-treated subjects after 3 months of treatment, i.e. on the same time-scale as cognitive worsening [22].
In May 2018, a 54-month trial of atabecestat (5 or 25 mg/ day), initially planned in 1,650 asymptomatic amyloid-positive subjects at risk of developing AD (EARLY), was halted during its recruitment phase due to liver toxicity and an unfavorable benefit-risk ratio [23]. The EARLY study targeted a prodromal population, with participants having a CDR staging of 0 and confirmed amyloid positivity by PET or CSF. At the time of study interruption, 557 participants had taken atabecestat for
≤18 months and about half of them for only three. Equal numbers of participants received 5 mg, 25 mg or placebo. The high-dose group had a statistically significant cognitive decrement compared to placebo after both 6 and 12 months of treatment on a composite cognitive scale (ADCS-PACC) and after 3 months on a neuropsychological test battery (RBANS). There were no differences in ADCS-ADL in this study, but more depression, anxiety, and sleep or dream-related problems were observed in atabecestat recipients [23].
In June 2018, a 2-year trial of the nonselective BACE1/2 inhibitor lanabecestat in 2,218 subjects with early AD (AMARANTH) and a 3-year trial in 1,722 mild AD patients (DAYBREAK-ALZ) were discontinued for futility [24]. In both studies, subjects were treated with placebo or lanabecestat at 20 or 50 mg/day. The primary efficacy measure in both studies was the 13-item form of the Alzheimer’s Disease Assessment Scale – Cognitive Subscale (ADAS-Cog13). In the AMARANTH study the high dose of lanabecestat produced more adverse events and withdrawals compared to placebo. The common- est concerning events were psychiatric events and weight loss of nearly 2 kg.
In July 2018, a 1-year trial of LY3202626 (3 or 12 mg/day) in
316 amyloid-positive mild AD patients (NAVIGATE-AD) was terminated after meeting futility criteria. No clear effect on cognition was found, although there were hints of a worsened deficit in the 3 mg group at 24 weeks on the ADAS-Cog13 and at 52 weeks on the Mini Mental State Examination (MMSE) [25].
In July 2019, two 5-year, placebo-controlled trials of umi- becestat, a selective BACE1 inhibitor, in 1,626 cognitively nor- mal subjects at risk of developing AD (APOE ε4 allele and brain Aβ-positivity), were discontinued because the drug worsened cognition, brain atrophy and weight loss [26].
The last BACE1 inhibitor abandoned in advanced clinical development was elenbecestat. A small 18-month, double- blind, placebo-controlled study of elenbecestat evaluated the preliminary safety and tolerability of elenbecestat in 70 subjects with MCI-to-moderate AD [27]. Patients received either elenbe- cestat at 5 mg/day (n = 17), 15 mg/day (n = 19) or 50 mg/day (n = 17), or placebo (n = 17). During the study the protocol was modified and patients in the 5 mg and 15 mg groups switched to 50 mg. The mean duration of treatment at 50 mg/day was 11 months (n = 38). Twenty-seven subjects (39%) discontinued the study: 5/17 on placebo (29%) and 22/53 on elenbecestat (42%), with nightmares as a notable side effect on active drug. At 18 months, elenbecestat dose-dependently decreased Aβ


Table 1. Main phase 2 and/phase 3 double-blind, placebo-controlled studies of BACE1 inhibitors in mild-to-moderate AD, prodromal AD, early AD and in subjects at risk of developing AD.

Study Code and Acronym
Subject Population Number of Subjects and Treatment Duration
Verubecestat (MK-8931) Merck Sharp & NCT01739348 EPOCH Mild-to-moderate AD 1,454 subjects 18 months Discontinued
Verubecestat (MK-8931) Merck Sharp & NCT01953601 APECS Prodromal AD 1,454 subjects 24 months Discontinued
Atabecestat (JNJ-54,861,911) Janssen/Shionogi NCT02569398 EARLY Asymptomatic subjects at risk of developing AD 596 subjects 54 months Discontinued
Atabecestat (JNJ-54,861,911) Janssen/Shionogi NCT01760005 DIAN-TU Pre-symptomatic APP or PSEN1 or PSEN2 carriers 490 subjects 4 years Atabecestat arm
Lanabecestat (AZD3293, AstraZeneca/Eli NCT02245737 AMARANTH Early AD 2,218 subjects 24 months Discontinued
LY3314814) Lilly
Lanabecestat (AZD3293, AstraZeneca/Eli NCT02783573 DAYBREAK- Mild AD 1,722 subjects 36 months Discontinued
LY3314814) Lilly ALZ
LY3202626 Eli Lilly NCT02791191 NAVIGATE-AD Mild AD 316 subjects 12 months Discontinued
Umibecestat (CNP520) Amgen/Novartis NCT02565511 Generation Cognitively normal, homozygous APOE ε4 carriers with age between 60 481 subjects 5 years Umibecestat arm
Study 1 and 75 years discontinued
Umibecestat (CNP520) Amgen/Novartis NCT03131453 Generation Cognitively normal, homozygous APOE ε4 carriers with age between 60 1,145 subjects 5 years Discontinued
Study 2 and 75 years
Elenbecestat (E2609) Eisai/Biogen NCT02956486 MissionAD1 Early AD 950 subjects 24 months Discontinued
Elenbecestat (E2609) Eisai/Biogen NCT03036280 MissionAD2 Early AD 950 subjects 24 months Discontinued
APP: amyloid precursor protein; PSEN1: presenilin1; PSEN2: presenilin2; APOE: apolipoprotein E
* The status of the study is based upon that reported in: ClinicalTrials.gov (https://clinicaltrials.gov/). Last accessed: 14 October 2019

Table 2. BACE1 substrates and their physiological role.
BACE1 substrate Physiological role

APP Regulation of neurite outgrowth, synapse formation, and synaptic plasticity. Also regulates metal homeostasis
APLP1 Regulation of neurotransmission and plasticity in CNS synapses
APLP2 Regulation of synaptic function and plasticity in CNS
Contactin-2 Regulates axon guidance, cell adhesion, neurite outgrowth
Jagged-1 Balances astrogenesis and neurogenesis, notch signaling influences neural plasticity, long-term memory, synapse remodeling transmitter release through astrocytes
CHL1 Regulates axon guidance, cell adhesion, neuronal migration, and neurite outgrowth
Neurexin 1α and 3β Regulates synapse assembly and maintenance
Neuroligin-1 Regulates myelination, neuronal migration, and oligodendrocyte differentiation. Regulates synaptic transmission and plasticity via neurotransmitter receptors
Sez6 Regulates dendritic arborization and affects excitatory synapse development and maintenance and formation of neuronal circuits
Sez6L Regulates synapse maturation, tumor suppressor function, and free cholesterol levels

β (β1-4) auxiliary subunits of the VGSC subtype Nav1
Voltage-gated potassium channel accessory subunits KCNE1 and KCNE2
Modulates cell surface expression of Nav1 sodium channels and thus controls excitability and propagation of action potentials in the neuronal membrane
Regulation of cardiac and brain potassium channel subunit trafficking and maintenance of membrane excitability

levels in CSF and the 50 mg/day dose produced a significant lowering of brain Aβ load on PET (n = 35). However, placebo an- d elenbecestat groups did not differ on either the Alzheimer’s Disease Composite Score (ADCOMS, p = 0.38) or CDR-SB (p = 0.55). Recently, two 24-month, double-blind, placebo- controlled studies of elenbecestat (MissionAD1 and MissionAD2) in approximately 1,900 subjects with early AD (MCI due to AD and a subset of very mild ADwith positive biomarkers for brain amyloid pathology) were discontinued for ‘unfavourable risk-benefit profile’. Patients were being treated with elenbecestat (50 mg/day) or placebo daily during the 24- month treatment period, and the primary endpoint was the CDR- SB [28].

⦁ Selectivity and toxicity of BACE1 inhibitors
Why do BACE1 inhibitors worsen memory? There are more than 40 known BACE1 substrates (Table 2), and BACE inhibi- tors may block one or more of them with neurodegenerative consequences. In addition to generating Aβ, BACE1 is impor- tant for various functions such as axon growth and pathfind- ing, astrogenesis, neurogenesis, hyperexcitation, and synaptic plasticity [29,30]. Therefore, the effects of BACE1 inhibitors on the processing of some of the numerous sub- strates of BACE1 might mask a potential cognitive benefit. Substrates like seizure protein 6 (SEZ6) [31], close homolog of L1 (CHL1) [32] and neuregulin-1 [33] have been identified as potential culprits. Indeed, several studies using BACE1- and BACE2-deficient mice demonstrated that these two pro- teases affect a wide range of physiological substrates and functions within and outside the nervous system. For BACE1 this includes axon guidance, neurogenesis, muscle spindle formation, neuronal network functions and myelination, whereas BACE2 was been shown to be involved in pigmen- tation and pancreatic β-cell function [34]. Interestingly, stu- dies have indicated that prolonged treatment with BACE1 inhibitors may negatively affect spine formation and bone density, hippocampal long-term potentiation and cognition in wild-type mice [35]. The ideal candidate could be a specific BACE1 inhibitor that avoids substrates other than APP, but it is not clear whether APP-selectivity is possible.
The clinical relevance of BACE1 inhibitor APP-specificity is unclear. Long-term toxicity studies with umibecestat in rats have shown that six months’ treatment with doses up to 200 mg/kg/day (a 10-fold excess over the highest dose used in humans) significantly reduced brain Aβ levels but did not affect length and organization of mossy fibers in the hippo- campus [19], a toxic effect observed in adult conditional BACE knockout mice that is believed to account for the detrimental cognitive effects of BACE inhibitors in AD patients. In addition, umibecestat shows a 3-fold higher affinity for BACE1 over BACE2 [36]. In spite of its biochemical and pharmacological selectivity, umibecestat was found to worsen cognitive decline compared to placebo in subjects at risk of developing AD (APOE4 carriers between 60 and 75 years of age) in the sec- ondary prevention trials of the API GENERATION program. This may suggest that cognitive worsening with BACE1 inhibitors may be independent of specific hippocampal toxicity or selec- tivity vs BACE2 but depends only upon the extent of brain Aβ reduction.

⦁ Other anti-Aβ drugs accelerating AD decline
Detrimental effects on cognition and clinical global status have been observed not only with BACE1 inhibitors but also with other anti-Aβ drugs. An 18-month, double-blind, pla- cebo-controlled study with the γ-secretase inhibitor sema- gacestat (100 and 140 mg/day) in 1,537 mild-to-moderate AD was discontinued prematurely because of deterioration in patient cognition [37]. Patients receiving the highest dose (140 mg/day) performed significantly worse than placebo- treated patients for cognition (as assessed with MMSE), functionality (as assessed with ADCS-ADL), psychiatric symp- toms (as assessed with NPI) and clinical global performance (as assessed with CDR-SB). This dose of semagacestat had been shown to significantly decreased the production of CNS Aβ by 52% in a previous study [38]. Avagacestat, another γ-secretase inhibitor, worsened cognition (MMSE) compared to placebo in a 6-month study involving 209 mild-to-moderate AD patients [39]. Tarenflurbil, a γ- secretase modulator, significantly worsened CDR-SB com- pared to placebo in an 18-month study involving 1,684

mild AD patients [40]. CAD106, an active anti-Aβ vaccine, tended to worsen MMSE compared to placebo (p = 0.052) in a 90-week study involving 121 patients with mild AD [41]. Another Aβ antigen, AD02, worsened cognition (ADAS-Cog) and clinical global status (CDR-SB) versus control treatment in an 18-month study involving 332 patients with early AD [42]. Scyllo-inositol, an Aβ aggregation inhibitor, dose- dependently increased mortality in an 18-month study involving 353 mild-to-moderate AD patients [43].

⦁ Conclusions
Despite strong target engagement and reducing CSF Aβ by up to 80–90% in humans, BACE1 inhibitors have failed to demonstrate significant slowing of cognitive decline in clinical trials in patients with mild-to-moderate AD, prodromal AD, or in cognitively healthy subjects at risk of developing AD. Moreover, recent full disclosure of the results of such trials has revealed significant cognitive and clinical worsening in subjects receiving BACE1 inhibitors that led to premature trial discontinuation. The types of adverse effects observed in patients treated with BACE1 inhi- bitors (cognitive, clinical, functional and psychiatric) are quite similar of those observed with γ-secretase inhibitors. Acceleration of hippocampal atrophy observed with the BACE inhibitor verubecestat [22] is also similar to the significant brain volumetric loss observed in AD patients treated with semagace- stat [37] or avagacestat [44]. These elements of toxicity and acceleration of decline with BACE1 and γ-secretase inhibitors have been ascribed to lack of selectivity and accumulation of abnormal APP fragments. For γ-secretase inhibitors, lack of selec- tivity for Notch cleavage and accumulation of C99 (β-CTF) were mainly advocated to cause detrimental effects in AD patients. For BACE1 inhibitors, lack of selectivity for SEZ6 or CHL1 or neure- gulin-1 have been mainly proposed to explain clinical detrimen- tal effects. In terms of APP processing, BACE1 inhibitors reduce C99 (β-CTF) accumulation [45] but still adversely affect cognition. Thus, it is hard to explain why BACE1 and γ-secretase inhibitors produce similar cognitive and clinical detrimental effects in AD patients, particularly in those at the early stages of the disease. There is only one factor common to BACE1 and γ-secretase inhibitors, which is their ability to reduce Aβ production, the primary purpose for which they were synthesized and devel- oped. The logical question is, therefore: Is the dramatic reduction in Aβ actually the cause of their toxicity? The answer appears to us to be ‘yes’, taking into account that similar detrimental cog- nitive effects have also been observed with other types of anti- Aβ drugs. Aβ is a widely-expressed peptide with a physiological role in several brain functions [46,47] and its dramatic reduction with BACE1 inhibitors could be generating multiple brain dys- functions. Studies have documented roles for Aβ in regulating neuronal electrophysiology [48,49], synaptic plasticity and mem- ory [50–52], long-term potentiation [53], neuronal transmission [54], learning and memory [55], hippocampal and memory con- solidation [51], neurogenesis [56] and neuronal survival [57]. Thus, in the AD brain, Aβ overproduction could represent an adaptive response to unknown upstream pathological events (one or more of chronic inflammation, tau-accumulation, viral or bacterial infection, metabolic failure, abnormal microglial
functioning, oxidative stress and cholesterol metabolism derangement). In this case, the use of drugs that depress produc- tion of native Aβ (BACE1 and γ-secretase inhibitors) from APP could accelerate the cognitive decline of AD patients, rather than slow it. Such detrimental effects of BACE1 and γ-secretase inhi- bitors would be particular evident in very early stages of the disease (patients with prodromal AD or cognitively normal sub- jects with brain Aβ accumulation) because at this stage physio- logical hyperproduction of Aβ may still be compensating for dysfunction due to the initial onset of neuronal death.

⦁ Expert opinion
Multiple clinical failures of BACE1 inhibitors and the consistent cognitive and clinical worsening observed in AD patients, espe- cially those at earlier stages, put an end at this class of anti-Aβ drugs. The defensive argument that lower doses of BACE1 inhibi- tors may work if given during the initial stages of the disease process is not supported by the evidence. A large study (EARLY) on atabecestat at low (5 mg/day) and high (25 mg) doses in cognitively normal subjects at risk of developing AD was inter- rupted because the drug worsened cognition compared to pla- cebo. Similarly, a large study (Generation Study 1) on umibecestat at low (15 mg/day) and high (50 mg/day) doses was prematurely terminated because of accelerated cognitive decline. The early interruption of these long-term trials is of particular significance since they occurred in cognitively normal subjects who are expected to show a very slow rate of cognitive decline. Interestingly, participants in the Generation Study 1 study were not requested to have elevated brain amyloid to enter the study. This means that BACE1 inhibitors may worsen cognition even at very earlier stages of the disease when brain Aβ deposition is not yet measurable. We believe that the recent discontinuation of elenbecestat, the last BACE1 inhibitor still in late stage clinical development, was predictable based on the results of a small Phase 2 study in 70 subjects with MCI-to-moderate AD. Indeed, 22 out of 53 patients on elenbecestat (42%) did not complete the study and nightmares were a notable side effect of active drug. There were no reasons why elenbecestat should be different from the other BACE1 inhibitors that have been trialed. Interestingly, the chemical structure of elenbecestat, a pyrimidine carboxamide, is very similar to LY3202626, which was also discontinued for poor clinical results (Table 1).
Indeed, the multiple clinical failure of BACE1 inhibitors and many other anti-Aβ drugs, including potent Aβ-oligomeric specific monoclonal antibodies (aducanumab and crenezu- mab) capable of clearing brain amyloid plaques to undetect- able levels, tell us that something is wrong with the Aβ amyloid hypothesis. Although sporadic AD may be pathophy- siology different from familial AD, we know that any challenge to the Aβ cascade hypothesis of AD must propose alternative explanations for the relationship between point mutations of the PSEN and APP genes and the occurrence of autosomal dominant AD. It is generally believed that PSEN and APP mutations in familial AD have a gain-of-function effect. However, a large study found that about 90% of 138 patho- genic PSEN1 mutations led to reduced production of Aβ1-42 and Aβ1-40 [58]. Some APP mutations increase total Aβ levels, while others decrease them. Variants such as A2V, H6R, D7N

produce Aβ1-42/Aβ1-40 ratios similar to wild type APP, whereas APP variants E22G, E22K, and E22Q lower this ratio [59]. A recent study in a large panel of isogenic knock-in human induced pluripotent stem cell (iPSC) lines carrying APP and/or PSEN1 mutations showed that APP and PSEN1 mutations had discordant effects on Aβ production but similar effects on APP β C-terminal fragments (β-CTFs), which accumulated in all mutant neurons. Endosomal dysfunction correlated with accu- mulation of β-CTFs, not Aβ [45]. Thus, the Aβ amyloid theory of AD may be more complex than previously thought. Current trials with monoclonal antibodies in pre-symptomatic carriers of pathogenic APP or PS mutations will tell us if the Aβ amyloid hypothesis should also be revisited in autosomal dominant familial AD.
The high failure rate of the anti-AD therapies may be due to the complexity of the pathogenic mechanisms of the disease and our incomplete understanding of the relation- ships between the numerous pathways involved in the development of AD and subsequent neurodegeneration. It has been proposed that attacking one specific pathogenic pathway may be not enough to change the course of the disease. Thus, a potentially effective strategy could be com- bining different therapeutic agents that target different pathogenic mechanisms (Aβ, tau, inflammation, ApoE, etc.) [60]. This approach appears attractive in theory, but practi- cal hurdles are represented by considerations related to pharmacology (how to select the proper doses of the dif- ferent drugs to be combined), statistics (the need to adopt factorial design), regulatory precedent (no specific guide- lines are available) and business (the development of agree- ments between different pharmaceutical companies). Nevertheless, initial attempts at combining drugs are evi- dent – such as the Phase 3 ALZT-OPT1 trial testing a combination regimen of cromolyn and ibuprofen. The AMBAR study has also just tested plasma exchange therapy to deliver human albumin in combination with intravenous immunoglobulin, with encouraging results. A Phase 2 study is testing anti-Aβ antibodies (LY3002813) in combination with a BACE1 inhibitor (LY3202626) [60], and a Phase 1 study is testing a combination of ABvac40 (Aβ active immu- nization agent), LY3002813 (passive Aβ immunization agent) and ACI-35 (tau active immunization agent) in patients with early to moderate AD [60].
Since the AD pathological process starts 15–20 years before clinical change occurs, the right approach should be to carry out primary prevention trials before the development of brain pathology (Aβ plaques, tau tangles, neuroinflammation) or cognitive symptoms [61]. This approach is being pursued in cognitively healthy subjects carrying dominant inherited APP or PS mutations causing AD. People with these mutations develop Aβ plaques in their brains in their 20 s and 30 s, and their children have a 50% chance of inheriting the condi- tion. Symptoms appear at a predictable time, based on the age of disease onset in their parents. Thus, experimental therapy (anti-Aβ monoclonal antibodies) is being started many years before symptoms are expected to appear. Exactly how the results from rare genetic forms of AD will translate to the more common sporadic form is uncertain. However, the fundamental processes that lead to dementia
appear quite similar [61]. Trying to identify who in the general population will get sporadic AD, predicting when they will develop Aβ pathology, and treating them early enough to make a clinical difference would require trials involving many thousands of people over long periods of time (probably decades). Even if this were possible, the duration of therapy required would significantly raise the safety threshold for the drugs being used.
An alternative option is to address the modifiable risk factors for developing AD. These include hypertension, dysli- pidaemia and obesity at midlife, diabetes mellitus, smoking, physical inactivity, depression and low levels of education [62]. Given the multifactorial etiology of sporadic AD, multidomain interventions that simultaneously target several risk factors may have more chance of success. Three large multidomain studies (FINGER, MAPT and PreDIVA) have been recently com- pleted and one of them, the FINGER study, showed that a multidomain lifestyle intervention can benefit cognition in elderly people with an elevated risk of dementia. Other ran- domized clinical trials are being conducted or organized to investigate whether lifestyle interventions can reduce the risk of cognitive decline and dementia in elderly adults, but such studies are methodologically and logistically challenging [62].

This paper was not funded.

Declaration of interest
BP Imbimbo is an employee at Chiesi Farmaceutici. He is listed between the inventors of numerous patents by Chiesi Farmaceutici on anti-Alzheimer drugs. M Watling is a consultant at TranScrip Partners. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, con- sultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures
Peer reviewers on this manuscript have no relevant financial or other relationships to disclose

Papers of special note have been highlighted as either of interest (•) or of considerable interest (••) to readers.
⦁ Brookmeyer R, Corrada MM, Curriero FC, et al. Survival following a diagnosis of Alzheimer disease. Arch Neurol. 2002;59(11):1764–1767.
⦁ Jansen IE, Savage JE, Watanabe K, et al. Genome-wide meta- analysis identifies new loci and functional pathways influencing Alzheimer’s disease risk. Nat Genet. 2019;51:404–413.
⦁ Jonsson T, Atwal JK, Steinberg S, et al. A mutation in APP protects against Alzheimer’s disease and age-related cognitive decline. Nature. 2012;488:96–99.
⦁ Mawuenyega KG, Sigurdson W, Ovod V, et al. Decreased clearance of CNS β-amyloid in Alzheimer’s disease. Science. 2010;330:1774.
⦁ Yang LB, Lindholm K, Yan R, et al. Elevated β-secretase expression and enzymatic activity detected in sporadic Alzheimer disease. Nat Med. 2003;9:3–4.
⦁ Liu CC, Liu CC, Kanekiyo T, et al. Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat Rev Neurol. 2013;9:106–118.

⦁ Kandalepas PC, Sadleir KR, Eimer WA, et al. The Alzheimer’s β- secretase BACE1 localizes to normal presynaptic terminals and to dystrophic presynaptic terminals surrounding amyloid plaques. Acta Neuropathol. 2013;126:329–352.
⦁ Yan R, Vassar R. Targeting the β-secretase BACE1 for Alzheimer’s disease therapy. Lancet Neurol. 2014;13:319–329.
⦁ Ghosh AK, Brindisi M, Yen YC, et al. Highly selective and potent human β-secretase 2 (BACE2) inhibitors against type 2 diabetes: design, synthesis, X-ray structure and structure-activity relationship studies. ChemMedChem. 2019;14:545–560.
⦁ Meakin PJ, Harper AJ, Hamilton DL, et al. Reduction in BACE1 decreases body weight, protects against diet-induced obesity and enhances insulin sensitivity in mice. Biochem J. 2012;441:285–296.
⦁ Plucińska K, Dekeryte R, Koss D, et al. Neuronal human BACE1 knockin induces systemic diabetes in mice. Diabetologia. 2016;59:1513–1523.
⦁ Coimbra JRM, Marques DFF, Baptista SJ, et al. Highlights in BACE1 inhibitors for Alzheimer’s disease treatment. Front Chem. 2018;6:178.
⦁ Fukumoto H, Takahashi H, Tarui N, et al. A noncompetitive BACE1 inhibitor TAK-070 ameliorates Aβ pathology and behavioral deficits in a mouse model of Alzheimer’s disease. J Neurosci. 2010;30:11157–11166.
⦁ Thakker DR, Sankaranarayanan S, Weatherspoon MR, et al. Centrally delivered BACE1 inhibitor activates microglia, and reverses amyloid pathology and cognitive deficit in aged Tg2576 mice. J Neurosci. 2015;35:6931–6936.
⦁ Keskin AD, Kekuš M, Adelsberger H, et al. BACE inhibition- dependent repair of Alzheimer’s pathophysiology. Proc Natl Acad Sci U S A. 2017;114:8631–8636.
⦁ Panza F, Lozupone M, Logroscino G, et al. A critical appraisal of amyloid-β-targeting therapies for Alzheimer disease. Nat Rev Neurol. 2019;15:73–88.
⦁ Cole SL, Vassar R. The Alzheimer’s disease β-secretase enzyme, BACE1. Mol Neurodegener. 2007;2:22.
⦁ Laird FM, Cai H, Savonenko AV, et al. BACE1, a major determinant of selective vulnerability of the brain to amyloid-β amyloidogen- esis, is essential for cognitive, emotional, and synaptic functions. J Neurosci. 2005;25:11693–11709.
⦁ Vassar R. Editorial: implications for BACE1 inhibitor clinical trials: adult conditional BACE1 knockout mice exhibit axonal organiza- tion defects in the hippocampus. J Prev Alzheimers Dis. 2019;6:78–84.
⦁ Plucińska K, Crouch B, Koss D, et al. Knock-in of human BACE1 cleaves murine APP and reiterates Alzheimer-like phenotypes. J Neurosci. 2014;34:10710–10728.
⦁ Egan MF, Kost J, Tariot PN, et al. Randomized trial of verubecestat for mild-to-moderate Alzheimer’s disease. N Engl J Med. 2018;378:1691–1703.
•• Phase 3 trial of the BACE1 inhibitor verubecestat in patients with mild-to-moderate Alzheimer’s disease terminated early for futility.
⦁ Egan MF, Kost J, Voss T, et al. Randomized trial of verubecestat for prodromal Alzheimer’s disease. N Engl J Med. 2019;380: 1408–1420.
•• Phase 3 trial of the BACE1 inhibitor verubecestat in patients with prodromal Alzheimer’s disease terminated early for accel- erated cognitive, functional and behavioural deterioration in verubecestat-treated patients.
⦁ Henley D, Raghavan N, Sperling R, et al. preliminary results of a trial of atabecestat in preclinical Alzheimer’s disease. N Engl J Med. 2019;380:1483–1485.
•• Phase 3 trial of the BACE1 inhibitor atabecestat in cognitively normal subjects at risk of developing Alzheimer’s disease ter- minated early for accelerated cognitive decline in the high dose atabecestat-treated subjects.
⦁ Strobel G. BACE inhibitors: postmortem on one, live updates on two. Alzheimer Res Forum. 2019 May;17.
⦁ Lo AC, Duggan Evans C, Mancini M, et al. Results from the phase 2 NAVIGATE-AD clinical trial evaluating LY3202626 BACE inhibitor in
patients with mild Alzheimer’s disease dementia. J Prev Alzheimers Dis. 2018;5(Suppl 1):S37.
⦁ Shugart J, Strobel G. Cognitive decline trips up API trials of BACE inhibitor. Alzheimer Res Forum. 2019 July;12.
⦁ Lynch SY, Kaplow J, Zhao J. et al. Elenbecestat, E2609, a BACE inhibitor: results from a phase-2 study in subjects with mild cogni- tive impairment and mild-to-moderate dementia due to Alzheimer’s disease. Alzheimers Dementia. 2018;14(Suppl):P1623.
⦁ Caouette D Eisai and Biogen to discontinue phase III clinical studies of BACE inhibitor elenbecestat in early Alzheimer’s disease. Eisai’s press release;2019 Sept 13.
⦁ Yan R. Physiological functions of the β-Site amyloid precursor protein cleaving enzyme 1 and 2. Front Mol Neurosci. 2017;10:97.
⦁ Zhu K, Peters F, Filser S, et al. Consequences of pharmacological BACE inhibition on synaptic structure and function. Biol Psychiatry. 2018;84:478–487.
⦁ Zhu K, Xiang X, Filser S, et al. Beta-site amyloid precursor protein cleaving enzyme 1 inhibition impairs synaptic plasticity via seizure protein 6. Biol Psychiatry. 2018;83:428–437.
⦁ Hitt B, Riordan SM, Kukreja L, et al. β-site amyloid precursor protein (APP)-cleaving enzyme 1 (BACE1)-deficient mice exhibit a close homolog of L1 (CHL1) loss-of-function phenotype involving axon guidance defects. J Biol Chem. 2012;287:38408–38425.
⦁ Hu X, Fan Q, Hou H, et al. Neurological dysfunctions associated with altered BACE1-dependent neuregulin-1 signaling. J Neurochem. 2016;136:234–249.
⦁ Vassar R, Kuhn PH, Haass C, et al. Function, therapeutic potential and cell biology of BACE proteases: current status and future prospects. J Neurochem. 2014;130:4–28.
⦁ Filser S, Ovsepian SV, Masana M, et al. Pharmacological inhibition of BACE1 impairs synaptic plasticity and cognitive functions. Biol Psychiatry. 2015;77:729–739.
⦁ Neumann U, Ufer M, Jacobson LH, et al. The BACE-1 inhibitor CNP520 for prevention trials in Alzheimer’s disease. EMBO Mol Med. 2018;10(11):e9316.
⦁ Doody RS, Raman R, Farlow M, et al. A phase 3 trial of semagacestat for treatment of Alzheimer’s disease. N Engl J Med. 2013;369:341–350.
⦁ Bateman RJ, Siemers ER, Mawuenyega KG, et al. A γ-secretase inhibitor decreases amyloid-β production in the central nervous system. Ann Neurol. 2009;66:48–54.
⦁ Coric V, van Dyck CH, Salloway S, et al. Safety and tolerability of the γ-secretase inhibitor avagacestat in a phase 2 study of mild to moderate Alzheimer disease. Arch Neurol. 2012;69:1430–1440.
⦁ Green RC, Schneider LS, Amato DA, et al. Effect of tarenflurbil on cognitive decline and activities of daily living in patients with mild Alzheimer disease: a randomized controlled trial. JAMA. 2009;302:2557–2564.
⦁ Vandenberghe R, Riviere ME, Caputo A, et al. Active Aβ immu- notherapy CAD106 in Alzheimer’s disease: a phase 2b study. Alzheimers Dement. 2016;3:10–22.
⦁ Schneeberger A, Hendrix S, Mandler M, et al. Results from a phase II study to assess the clinical and immunological activity of AFFITOPE® AD02 in patients with early Alzheimer’s disease. J Prev Alzheimers Dis. 2015;2:103–114.
⦁ Salloway S, Sperling R, Keren R, et al. A phase 2 randomized trial of ELND005, scyllo-inositol, in mild to moderate Alzheimer disease. Neurology. 2011;77:1253–1262.
⦁ Coric V, Salloway S, van Dyck CH, et al. Targeting prodromal Alzheimer disease with avagacestat: a randomized clinical trial. JAMA Neurol. 2015;72:1324–1333.
⦁ Kwart D, Gregg A, Scheckel C, et al. A large panel of isogenic APP and PSEN1 mutant human iPSC neurons reveals shared endosomal abnormalities Mediated by APP β-CTFs, not Aβ. Neuron. 2019 Aug 12. pii: S0896-6273(19)30636-1. doi: ⦁ 10.1016/j.neuron.2019.07.010. [Epub ahead of print].
⦁ Bishop GM, Robinson SR. Physiological roles of amyloid-β and implications for its removal in Alzheimer’s disease. Drugs Aging. 2004;21:621–630.

⦁ Puzzo D, Gulisano W, Arancio O, et al. The keystone of Alzheimer pathogenesis might be sought in Aβ physiology. Neuroscience. 2015;307:26–36.
⦁ Kamenetz F, Tomita T, Hsieh H, et al. APP processing and synaptic function. Neuron. 2003;37:925–937.
⦁ Esteban JA. Living with the enemy: a physiological role for the
β-amyloid peptide. Trends Neurosci. 2004;27:1–3.
⦁ Parihar MS, Brewer GJ. Amyloid-β as a modulator of synaptic plasticity. J Alzheimers Dis. 2010;22:741–763.
⦁ Puzzo D, Privitera L, Fà M, et al. Endogenous amyloid-β is necessary for hippocampal synaptic plasticity and memory. Ann Neurol. 2011;69:819–830.
⦁ Lawrence JL, Tong M, Alfulaij N, et al. Regulation of presynaptic Ca2+, synaptic plasticity and contextual fear conditioning by a N-terminal β-amyloid fragment. J Neurosci. 2014;34: 14210–14218.
⦁ Palmeri A, Ricciarelli R, Gulisano W, et al. Amyloid-β peptide is needed for cGMP-induced long-term potentiation and memory. J Neurosci. 2017;37:6926–6937.
⦁ Abramov E, Dolev I, Fogel H, et al. Amyloid-β as a positive endo- genous regulator of release probability at hippocampal synapses. Nat Neurosci. 2009;12:1567–1576.
⦁ Morley JE, Farr SA, Banks WA, et al. A physiological role for amyloid- β protein: enhancement of learning and memory. J Alzheimers Dis. 2010;19:441–449.
⦁ López-Toledano MA, Shelanski ML. Neurogenic effect of β-amyloid peptide in the development of neural stem cells. J Neurosci. 2004;24:5439–5444.
⦁ Plant LD, Boyle JP, Smith IF, et al. The production of amyloid β peptide is a critical requirement for the viability of central neurons. J Neurosci. 2003;23:5531–5535.
⦁ Sun L, Zhou R, Yang G, et al. Analysis of 138 pathogenic mutations in presenilin-1 on the in vitro production of Aβ42 and Aβ40 pep- tides by γ-secretase. Proc Natl Acad Sci U S A. 2017;114:E476–85.
⦁ Kepp KP. Alzheimer’s disease due to loss of function: a new synth- esis of the available data. Prog Neurobiol. 2016;143:36–60.
⦁ Cummings JL, Tong G, Ballard C. Treatment combinations for Alzheimer’s disease: current and future pharmacotherapy options. J Alzheimers Dis. 2019;67:779–794.
⦁ McDade E, Bateman RJ. Stop Alzheimer’s before it starts. Nature. 2017;547:153–155.
⦁ Kivipelto M, Mangialasche F, Ngandu T. Lifestyle interventions to prevent cognitive impairment, dementia and Alzheimer disease. Nat Rev Neurol. 2018;14:653–666.