The Alzheimer's disease Beta-secretase enzyme, BACE1

AD is the most prevalent form of dementia, and current indications show that twenty-nine million people live with AD worldwide, a figure expected rise exponentially over the coming decades. It has been recently estimated that the worldwide costs for dementia care are $315.4 billion annually (US; Alzheimer's Association). Clearly, blocking disease progression or, in the best-case scenario, preventing AD altogether would be of benefit in both social and economic terms. However, current AD therapies are merely palliative and only temporarily slow cognitive decline, and treatments that address the underlying pathologic mechanisms of AD are completely lacking. While familial AD (FAD) is caused by autosomal dominant mutations in either amyloid precursor protein (APP) [ 1 , 2 ] or the presenilin (PS1, PS2) [ 3 , 4 ] genes, the underlying cause (s) of the remaining ~98% of so-called sporadic AD (SAD) cases remain elusive. However, specific risk factors for AD have been recently identified and include aging, the presence of the apolipoprotein E4 (ApoE4) allele [ 5 ] and vascular diseases such as stroke and heart disease [ 6 - 10 ]. AD pathology Pathologically, AD is characterized by the accumulation of amyloid beta peptide (Aβ), as fibrillar plaques and soluble oligomers in high-order association brain regions. The presence of intracellular neurofibrillary tangles, neuroinflammation, neuronal dysfunction and death further characterizes this disease. Mounting evidence suggests that Aβ plays a critical early role in AD pathogenesis, and the basic tenant of the amyloid (or Aβ cascade) hypothesis is that Aβ aggregates trigger a complex pathological cascade which leads to neurodegeneration [ 11 ]. A strong genetic correlation exists between FAD and the 42 amino acid Aβ form (Aβ42; reviewed in [ 12 - 14 ]). Aβ is derived from APP and mutations in APP and PS increase Aβ42 production and cause FAD with nearly 100% penetrance. Down's syndrome (DS) patients, who have an extra copy of the APP gene on chromosome 21, and FAD families with a duplicated APP gene locus [ 15 ], exhibit total Aβ overproduction and all develop early-onset AD. In FAD, the Aβ42 increase is present years before AD symptoms arise, suggesting that Aβ42 is likely to initiate AD pathophysiology. The robust association of Aβ42 overproduction with FAD argues strongly in favor of a critical role for Aβ42 in the etiology of AD, including in SAD. Fibrillar and oligomeric forms of Aβ appear neurotoxic in vitro and in vivo. Importantly, in specific transgenic (Tg) mouse models of AD the lack of Aβ correlates with the absence of neuronal loss and improved cognitive function [ 16 - 18 ]. Such data provides direct evidence for the amyloid hypothesis in vivo, and also indicates that Aβ is directly responsible for neuronal death. Consequently, strategies to lower Aβ42 levels in the brain are anticipated to be of therapeutic benefit in AD.

Prior to its identification, numerous studies were undertaken to define the characteristics of β-secretase activity. Although the majority of body tissues exhibit β-secretase activity [ 35 ], highest activity levels were observed in neural tissue and neuronal cell lines [ 36 ]. Indeed, β-secretase appeared to predominate in neurons, with the level of β-secretase activity appearing lower in astrocytes [ 37 ]. Data showing that β-secretase efficiently cleaved only membrane-bound substrates [ 38 ] indicated that the enzyme was likely membrane-bound or closely associated with a membrane protein. Furthermore, as highest β-secretase activity was detected at acidic pH [ 28 , 39 - 41 ], within the subcellular compartments of the secretory pathway, including the trans-Golgi network (TGN) and endosomes [ 42 , 43 ], it was predicted that the active site of this enzyme is within the lumen of acidic intracellular compartments. The sequence preference for β-secretase was determined from site-directed mutagenesis of the amino acids surrounding the cleavage site in APP [ 38 ]. Substitutions of larger hydrophobic amino acids (such as Leu found in the Swedish FAD mutation) for the Met residue at P1 improve the efficiency of β-secretase cleavage. Conversely, substitution of the smaller hydrophobic amino acid Val at the same position inhibits cleavage. Many other substitutions at this site and at surrounding positions decrease cleavage, and indicate that the β-secretase is highly sequence-specific. Radiosequencing demonstrated that Aβ isolated from amyloid plaques, as well as that produced in cell lines, predominantly begins at the Asp+1 residue of Aβ [ 44 ], although minor Aβ species begin at Val-3, Ile-6, and Glu+11 [ 35 ]. Inhibitor studies suggest that the Val-3 and Ile-6 species are generated by a protease that is different from the β-secretase [ 45 ]. However, the Glu+11 species is produced in parallel with Asp+1 Aβ [ 46 ], suggesting that β-secretase is responsible for cleaving at both these positions. Interestingly, the Glu+11 species is the predominant form of Aβ made in rat primary neuron cultures [ 46 ]. Finally, β-secretase activity is insensitive to pepstatin, an inhibitor of many (but not all) aspartic proteases. During 1999–2000, five teams concluded that the novel transmembrane aspartic protease BACE1 (also named memapsin and Asp2) was the β-secretase [ 21 - 25 ]. Indeed, BACE1 exhibited all the known characteristics of the β-secretase. The 501 amino acid sequence of BACE1 bears the hallmark features of eukaryotic aspartic proteases of the pepsin family. BACE1 has two aspartic protease active site motifs, DTGS (residues 93–96) and DSGT (residues 289–292), and mutation of either aspartic acid renders the enzyme inactive [ 21 , 47 ]. Like other aspartic proteases, BACE1 has an N-terminal signal sequence (residues 1–21) and a pro-peptide domain (residues 22–45) that are removed post-translationally, so the mature enzyme begins at residue Glu46 [ 47 ]. Importantly, BACE1 has a single transmembrane domain near its C-terminus (residues 455–480) and a palmitoylated cytoplasmic tail [ 48 ]. Thus, BACE1 is a type I membrane protein with a luminal active site, features predicted for β-secretase. The position of the BACE1 active site within the lumen of intracellular compartments provides the correct topological orientation for cleavage of APP at the β-secretase site. As observed with other aspartic proteases, BACE1 has six luminal cysteine residues that form three intramolecular disulfide bonds and several N-linked glycosylation sites [ 49 ]. The pattern and level of BACE1 expression is largely consistent with those of β-secretase activity in cells and tissues [ 23 , 24 , 50 ]. The levels of BACE1 mRNA are highest in brain and pancreas and are significantly lower in most other tissues. Moreover, BACE1 mRNA is highly expressed in neurons but little is found in resting glial cells of the brain, as expected for β-secretase. The protein is abundant in both normal human and AD brain [ 23 , 50 ]. Given the low levels of β-secretase activity in the pancreas, the high pancreatic mRNA expression was initially confusing [ 22 ]. However, subsequent reports indicated that BACE1 mRNA transcripts in pancreas largely consist of a splice variant missing the majority of exon 3 [ 51 , 52 ]. This splice variant encodes a BACE1 isoform devoid of β-secretase activity, thus reconciling the paradoxically high BACE1 mRNA levels with the low β-secretase activity found in the pancreas. The functional relevance of this pancreas-specific splice variant remains unclear. BACE1 induces a dramatic increase in β-secretase activity when transfected into stable APP-overexpressing cell lines. The immediate products of β-secretase cleavage, APPsβ and C99, are increased several fold over levels found in untransfected cells, and Aβ production is also elevated. Interestingly, APPsα levels are reduced upon BACE1 transfection, indicative that α- and β-secretases

BACE2 is a homologue of BACE1 that is mapped to the DS critical region on chromosome 21 [ 72 ]. The amino acid sequences of BACE1 and BACE2 are ~45% identical and 75% homologous [ 73 , 74 ]. As with BACE1, the BACE2 C-terminal domain is significantly larger than other aspartic proteases, although overall the BACE2 structure contains features typical of this protease family. Unlike BACE1, pro-BACE2 requires autocatalytic pro-domain processing for enzymatic activation [ 75 ]. BACE1 and BACE2 have distinct transcriptional regulation and function [ 73 ]. BACE2 mRNA has been observed at low levels in most human peripheral tissues. However, unlike BACE1, which is enriched in neuronal populations [ 23 ], human adult and fetal whole brain express very low or undetectable levels of BACE2 mRNA [ 50 , 74 ]. In vitro , BACE2 can cleave APP at the β-secretase cleavage site [ 76 , 77 ] and BACE2 appears to be primarily responsible for Aβ production in Flemish mutant APP transfected cells [ 76 ]. However, other studies have demonstrated that BACE2 functions as an alternative α-secretase and as an antagonist of BACE1 [ 78 , 79 ]. BACE2 does not get upregulated to compensate for a lack of BACE1 in knockout mice [ 57 ]. Further evidence that BACE2 functions not as a β-secretase comes from a recent study which identifies BACE2 as a novel theta (θ)-secretase. Radiosequencing clearly demonstrated that the major BACE2 cleavage site is between Phe+19 and Phe+20 sites of APP, thus BACE2 cleaves APP at a novel θ-site downstream of the α-secretase cleavage site [ 80 ]. Cleavage of APP by BACE2 at this site abolishes Aβ production. Furthermore, overexpression of BACE2 reduced Aβ production in primary neuronal cultures derived from APP Tg mice [ 80 ].

The cell biology of BACE1 was investigated in order to further understand BACE1 regulation and to identify other potential therapeutic targets in the β-secretase pathway. BACE1 is initially synthesized in the endoplasmic reticulum (ER; Fig. 2 ) as an immature precursor protein (proBACE1) with a molecular mass of ~60 kDa [ 49 , 105 - 107 ]. ProBACE1 is short-lived and undergoes rapid maturation into a 70 kDa form in the Golgi, which involves the addition of complex carbohydrates and the removal of the propeptide domain. Figure 2 The intracellular trafficking of BACE1 . (1) Following synthesis in the ER, pro-BACE1 traffics to the Golgi. Here, the propeptide is removed and BACE1 undergoes significant post-translational modifications. The trafficking and localization of BACE1 appear dependent on the ACDL motif, composed of DISLL residues, in its C-terminal tail. Importantly, phosphorylation of the serine 498 residue (S498) is important for binding of BACE1 to the GGA monomeric adaptor proteins which are implicated in the sorting of cargo between the TGN and endosomes and vice versa [127-131], in addition to the Golgi complex to the endosomes [130]. (2) BACE1 traffics to the plasma membrane and while the re-internalization of BACE1 from the cell surface appears to be independent of the phosphorylation state of S498 [131], it has been reported that the di-leucine residues are important for BACE1 endocytosis [126]. Non-phosphorylated BACE1 may recycle from the endosome back to the cell surface [131]. However, BACE1 phosphorylated at the S498 site interacts with GGA1 in the endosome and traffics back to the TGN [131], and (3) the interaction of phosphorylated BACE1 with GGA3 appears to direct BACE1 to the lysosome, where it is degraded [135]. In addition to S498 phosphorylation, the di-leucine residues are also important for the sorting of BACE1 into lysosomal compartments [140]. BACE1 is also degraded in the proteasome [133]. The precise subcellular localization (early verses late endosomes) of the BACE1-GGA complex requires determination, and while it appears likely that BACE1 can traffic from the TGN to the endosome directly, it remains to be demonstrated experimentally. BACE1 undergoes extensive post-translational modifications, being glycosylated at four N-linked sites [ 49 ]. Mature N-glycosylated BACE1 is sulfated and three cysteine residues within the cytosolic tail of BACE1 become palmitoylated [ 48 ]. While glycosylation appears important for enzymatic activity [ 108 ], palmitoylation may influence the trafficking and localization of BACE1 [ 48 ]. The majority of BACE1 in cells is produced as an integral membrane protein. However, a small fraction of BACE1 undergoes ectodomain shedding, a process that is suppressed by palmitoylation [ 48 ]. Inhibition of shedding does not influence processing of APP at the β-site [ 109 ]. However, co-expression of APP and the soluble ectodomain of BACE1 in cells increases the generation of Aβ, suggesting the enhanced BACE1 ectodomain shedding may raise amyloidogenic processing of APP [ 48 ]. In addition, Murayama and colleagues reported the release of detectable levels of BACE1 holoprotein in vitro [ 110 ], although the physiological relevance of this event remains to be clarified. Interestingly, active, soluble BACE1 has been detected in human CSF, [ 111 ], a finding which raises the potential use of BACE1 detection in an easily accessible biological fluid, such as CSF, in future diagnostic or prognostic applications. The BACE1 propeptide domain is removed within the Golgi apparatus by cleavage between Arg45 and Glu46 [ 47 , 74 ]. Unlike the majority of aspartic proteases, including BACE2, which cleave the propeptide domain autocatalytically, BACE1 propeptide removal involves intermolecular cleavage by a different protease. Golgi-localized proprotein convertases (PC) are capable of cleaving the BACE1 propeptide and furin, a ubiquitous PC, appears to be the major protease regulating this process [ 48 , 74 , 107 ]. In contrast to other zymogens, proBACE1 exhibits robust β-secretase activity, indicating that the BACE1 propeptide domain does not significantly suppress protease activity [ 48 , 107 ]. These findings suggest that inhibiting the removal of the BACE1 propeptide would not be an effective therapeutic strategy for reducing Aβ levels in AD. However, proBACE1 may cleave APP early in the biosynthetic pathway leading to the generation of an intracellular pool of Aβ in the ER, thought by some investigators to be particularly neurotoxic [ 107 ]. The activity of the BACE1 enzyme is influenced by various post-translational and cell biological events. Mature BACE1 localizes largely within cholesterol-rich lipid rafts [ 112 , 113 ] and replacing the BACE1 transmembrane domain with a glycosylphosphatidylinositol (GPI) anchor exclusively targets BACE1 to lipid rafts and substantially increases Aβ production [ 114 ]. Indeed, various types of lipid stimulate BACE1 activity and th

Given the fact that the majority of APP molecules are cleaved by α-secretase moieties, with only a small fraction being BACE1 substrate, it is highly likely that other BACE1 substrates exist. In addition to APP, it is known that BACE1 also cleaves the APP homologues, amyloid precursor-like proteins, APLP1 and APLP2 [ 144 , 145 ]. While the Aβ sequence is absent in both APLP1 and APLP2, relatively little else is known about APLP1 and -2, although it is known that the APLPs can be processed by γ-secretase generating intracellular fragments with potential transcriptional activity [ 146 , 147 ]. While the essential cellular functions of BACE1 under normal conditions have proved somewhat elusive, recent findings have started to define a physiological function for this enzyme. As discussed earlier, given the apparent phenotypic alterations observed in BACE1 deficient mice, the identification of other BACE1 substrates and an understanding of their biological function is essential. BACE1 is enriched in neuronal populations, and β-secretase processing of APP is modulated by the interaction of BACE1 with neurite growth inhibitor NOGO, a member of the reticulon protein family and a component of myelin [ 123 ]. Recently, a role for BACE1 in axonal growth and brain development has been proposed, whereby BACE1 regulates the myelination process [ 148 , 149 ]. The neuronal type III isoform of the epidermal growth factor (EGF)-like factor neuregulin 1 (NRG1) regulates myelination and is a known initiator of peripheral nervous system (PNS) myelination and a modulator of myelin sheath thickness in both the CNS and PNS. BACE1 is transported to axons by a kinesin-1-dependent pathway [ 150 ] and the highest levels of BACE1 expression are observed when myelination occurs during the early postnatal stages [ 149 ]. Co-expression of BACE1 with type III NRG1 was observed within sensory and motor neurons whose axons project within peripheral nerves. Importantly, data generated from BACE1 -/- mice showed that the absence of BACE1 was associated with hypomyelination of both central and peripheral nervous system axons. In line with previous observations indicating that BACE1 deficient mice exhibit altered synaptic plasticity and decreased cognitive function [ 16 , 18 , 61 ] reviewed in [ 62 ], specific neurological impairments were also associated with the genetic ablation of BACE1 in this myelination study [ 148 ]. When compared to wild type brain, BACE1 -/- accumulated full length, uncleaved NRG1 and exhibited reduced levels of NRG1 cleavage fragments, findings consistent with a role for BACE1 in the proteolysis of NRG1 [ 148 , 149 ]. It remains undetermined as to whether NRG1 cleavage is required for maintenance of the mature myelin sheath. It should be noted that complete deletion of BACE1 was necessary in order to alter the signaling events and cause hypomyelination. In agreement with previous findings that partial BACE1 inhibition may be without effect on normal learning and memory processes, [ 18 ] reviewed in [ 62 ], BACE +/- mice did not display any of the biochemical and morphological features observed in the BACE1 null mice. Synaptic dysfunction precedes overt neurodegeneration during AD progression, and data indicates that BACE1 cleaves APP to generate Aβ in the synaptic terminal [ 138 , 139 ]. Indeed a role for BACE1 in the maintenance of synaptic function has been proposed, and several additional, non-APP BACE1 substrates have recently been identified, some of which may be localized at the terminal, suggesting that BACE1 cleavage of particular substrates may be required for normal function at the synapse. Voltage-gated sodium channels (VGSC [ 151 ]; Na v 1 [ 152 ]) are abundant ion channels responsible for the initiation and propagation of action potentials. These channels are large complexes consisting of an α subunit and at least one β subunit. The VGSCβ subunits are important auxiliary units, although they are not essential for the basic functioning of the VGSC. However, expression of both subunit types is required for full functionality of the VGSC, as well as for the modification of channel properties and intracellular localization. In a manner analogous to APP processing, VGSCβ subunits are substrates of both BACE1 and γ-secretase. BACE1 cleavage generates a membrane-tethered β-CTF [ 151 , 152 ], and the VGSCβ subunits are processed further by γ-secretase, which generates an β2-intracellular domain (β2-ICD; [ 152 ]). Recently, the functional ramifications of these cleavage events have been elucidated [ 152 ]. β2-ICD regulates expression of the α subunit, Na v 1.1. Importantly, the increased pool of Na v 1.1 is maintained within the cell and the BACE1 cleavage of the β2 subunit actually leads to loss of functional membrane channels, a reduction in Na + current and alterations in membrane excitability [ 152 ]. Importantly, the processing of VGSCβ2 and VGSCβ4 by BACE1 has been demonstrated in vivo and elevated β2-CTF and

AD pathogenesis is highly complex. Perhaps as a consequence of this complexity, many major questions about this disease remain open. Indeed, apparently basic questions including those regarding the initiating events in AD, and whether such events cause elevate Aβ, remain to be determined. Given the number of pathologies that characterize AD, it is highly possible that there are several events that culminate to trigger AD. We know that Aβ elevation occurs early in the disease and plays a central role in AD pathogenesis. More recently, other pertinent questions have come to light and a major subject for current debate is the cause of BACE1 elevation in AD and whether this elevation could be responsible for the initiation of AD pathology. Indeed, whether the BACE1 elevation in AD promotes Aβ generation and disease progression remains to be determined. However, an AD feedback loop has long been proposed and our data are suggestive of a positive feedback loop, whereby Aβ42 deposition in AD causes BACE1 levels to rise in nearby neurons. Increased Aβ production may follow, initiating a vicious cycle of additional amyloid deposition followed by further elevated BACE1 levels. Interestingly, a recent in vitro study [ 175 ] demonstrated that a relatively small increase in BACE1 expression results in sharply elevated Aβ production, until a plateau is reached, whereby further increases in BACE1 expression and activity have no further effect of Aβ genesis. Given the observation that BACE1 elevation occurs around Aβ42 plaque cores it seems possible that Aβ42 somehow triggers the BACE1 increase. It is well established that Aβ42 is neurotoxic and such toxicity may induce the BACE1 elevation. Indeed, as previously mentioned, the BACE1 promoter contains regions for transcription factors known to be activated in response to cellular stress. Furthermore, proinflammatory cytokines appear to modulate BACE1 expression [ 91 , 92 ] and following Aβ exposure a functional NFkβ in the BACE1 promoter region was stimulated [ 89 ]. However, it remains to be determined as to which is the initiating event, Aβ elevation and deposition or increased BACE1 activity. Interestingly, Tamagno and colleagues have recently proposed that Aβ acts via a biphasic neurotoxic mechanism, which is conformation-dependent, with Aβ oligomers inducing oxidative stress while fibrillar Aβ increases BACE1 expression and activity [ 168 ]. Nevertheless, whether such a biphasic mode of action occurs in vivo remains to be determined. Given the complexity of AD pathogenesis it is highly likely that, in addition to Aβ, a number of other factors could impact BACE1 levels. Aging is the strongest risk factor in AD. Given that BACE1 activity increases with age and to an even greater extent in SAD, it is plausible that AD may reflect an exaggeration of age-related changes in BACE1 activity. It is interesting to note that the vast majority of cardiovascular events occur in older people and there appears to be a close relationship between cardiovascular (and cerebrovascular) disease and AD. Individuals with AD and cerebrovascular pathologies exhibit greater cognitive impairment than those exhibiting either pathology alone [ 176 - 180 ]. Various types of heart disease (for example, atrial fibrillation, congestive heart failure, coronary heart disease, among others) and stroke (ischemic, hemorrhagic) are intimately associated with elevated AD risk (reviewed in [ 181 ]). AD patients exhibit more severe atherosclerosis in the cerebral arteries at the base of the brain (the circle of Willis) when compared to age-matched controls [ 182 ]. Indeed cerebral blood supply is limited by the vascular narrowing caused by these lesions [ 183 - 185 ] and many vascular-related AD risk factors have an established association with cerebral hypoperfusion. Importantly, there is increasing evidence from epidemiological [ 186 - 189 ] and neuroimaging studies [ 190 - 192 ] which suggests that vascular risk factors, and the ensuing reduced cerebral blood flow (CBF) and chronic brain hypoperfusion (CBH) are key factors in AD development and may even play a causative role in dementia (reviewed in [ 181 , 193 ]). Evidence suggests that the subcellular changes that are crucial to neurodegeneration development are provoked by CBH brought on by the cardio-cerebral damage associated with cardiovascular events (reviewed in [ 181 ]). Indeed, CBH is a pre-clinical condition of mild cognitive impairment (MCI), a condition thought to precede AD, and is an accurate indicator for predicting the development of AD [ 194 - 198 ]. Interestingly, an association between impaired functionality of microvessels and unfavorable evolution of cognitive function in AD patients has been recently reported [ 199 ]. CBH can cause hypoxia, in addition to ischemic episodes. The later involves both hypoxia and reoxygenation, which represent forms of cellular stress, and such episodes have also been reported to increase AD risk. Interes

Mitochondrial dysfunction in neurons can be caused by many cellular insults including transient hypoxia. Indeed, energy-depleted hippocampal neurons unable to cope with prolonged CBH undergo both oxidative and cellular stress (reviewed in [ 181 ]). Importantly, several reports document BACE1 as a stress-induced protease and in view of the apparent importance of metabolic dysfunction and amyloidosis in AD, it is noteworthy that BACE1 upregulation has been observed under a variety of experimental conditions likely involving energy disruption and/or mitochondrial stress [ 136 , 212 - 215 , 228 , 229 ]. Indeed, increased BACE1 levels and activity have been reported in both in vitro [ 210 - 212 ] and in vivo [ 213 , 215 ] under conditions of altered energy metabolism and cellular stress. In many aspects of AD, particularly at the molecular level, a chicken-and-egg scenario exists: does Aβ cause BACE1 elevation or vice versa? Does Aβ cause cerebrovascular dysfunction or vice versa? It appears likely that in most cases both scenarios might be true and may operate in a vicious circle of events. This appears to likely be the case with Aβ and oxidative stress. While Aβ accumulation may lead to oxidative stress [ 230 ] reviewed in [ 231 ], it has also been demonstrated that oxidative stress may lead to Aβ accumulation. As previously indicated, HNE has been observed in AD brain [ 200 , 201 ]. Tamagno reported that HNE addition to cultured NT2 neurons facilitated an increase in BACE1 activity [ 210 ]. Interestingly, oxidative stress in vitro resulted in significant increases in BACE1 promoter activity [ 212 ] and Tamagno and colleagues reported an increase in both BACE1 mRNA and protein levels, with increased BACE1 activity, as determined by enhanced Aβ production, being observed following HNE exposure [ 210 , 211 ]. Inhibitors of mitochondrial respiration are generally considered to cause oxidative stress [ 232 , 233 ] and thus the differentiation of energy inhibition from oxidative stress with regards to the pharmacological blockage of mitochondrial energetic processes in vivo is practically impossible. Nevertheless, following acute energy inhibition (and/or oxidative stress) in vivo [ 213 ], we observed a significant elevation in BACE1 protein. This effect was long-lasting, with BACE1 levels remaining elevated for seven days post-injection, and appeared to correspond with a significant increase in cerebral Aβ40 load in treated mice. In agreement with the previous in vitro studies, these data suggest that energy inhibition is potentially amyloidogenic. Support for these observations come from the recent findings that mitochondrial respiratory inhibition and oxidative stress increased BACE1 levels [ 215 ]. Moreover, as we previously hypothesized [ 97 ], evidence in support of a role for Aβ in elevating BACE1 was provided from data showing that the intravitreal application of fibrillar Aβ42 appeared to exert extremely potent effects and caused enhanced BACE1 expression and activity [ 215 ]. The mechanism of the energy-induced BACE1 increase in vivo is unclear and will likely be challenging to elucidate. While studies have demonstrated increases in BACE1 mRNA level following oxidative stress and hypoxic insults [ 211 , 214 ], our data indicates that a post-translational mechanism may be implicated in the in vivo BACE1 elevation during times of energy depletion [ 213 ]. BACE1 mRNA levels did not significantly increase in Tg2576 brain following energy inhibition treatment and the BACE1 protein half life is too long to account for the rapid rise in BACE1 levels observed. As previously discussed reports indicate that the 5'UTR of BACE1 mRNA influences translational efficiency [ 102 - 104 ]. Clearly, further, detailed analysis is required to solve this important issue, but full understanding of such cellular mechanisms may shed light on novel therapeutic approaches for AD.

BACE1 remains a prime drug target for inhibiting the production of Aβ. Not only does BACE1 initiate Aβ formation, but the observation that BACE1 levels are elevated in AD [ 97 , 166 - 170 ] provides a direct and compelling reason to develop therapies directed at BACE1 inhibition thus reducing Aβ and its associated toxicities. Early on, data indicated that BACE knockout mice were phenotypically normal although more recent, detailed analyses show that complete abolishment of BACE1 activity may have potentially deleterious effects. Therefore, a partial, rather than full, inhibition of this enzymatic activity may be beneficial, although the percentage of BACE1 inhibition required to significantly delay amyloid pathology and the associated cognitive changes, remains to be determined. The physiological role of BACE1 remains to be established. However, several non-APP BACE1 substrates have been recently identified and the inclusion of NRG1 and VGSCβ subunits as putative substrates for this enzyme, indicates a potential role for BACE1 in the regulation of neuronal function. BACE1 levels are elevated under a variety of conditions, including those cellular changes evoked under the stressful conditions of cardio- and cerebrovascular disease ([ 97 , 166 , 173 , 209 - 215 ]). Indeed, evidence suggests that the BACE1 substrates identified so far may play a role in the responses to stress and/or injury, such as axonal growth and the regulation of glial cell survival (NRG1; [ 240 ]), recovery from excitotoxicity (Aβ [ 65 ]), Aβ clearance (LRP; [ 241 ]), synapse formation (APP, APLP1, APLP2; [ 242 , 243 ]), neuroprotection (secreted APP ectodomain; reviewed in [ 244 ]) and immune functions (PSGL-1, ST6Gal I and IL-1R2; [ 160 , 161 , 164 ]). We speculate that BACE1 may be involved in the response to stress and/or injury and that the elevation in BACE1 levels facilitates recovery after acute stress/injury. Indeed, it appears plausible that cleavage of specific BACE1 substrates is necessary for this function. However, it may well be the case that chronic stress/injury results in pathologic BACE1 levels and deleterious amyloid formation. Indeed, chronic stress/injury could cause long-term BACE1 elevation and have harmful effects attributable to cerebral Aβ overproduction. Furthermore, we suggest that increased BACE1 activity may affect normal synaptic functioning, given that, in addition to Aβ, other BACE1 cleavage products, including specific Na v 1s (such as Na v 1.6) are localized to the synapse. Changes in the normal function of the synapse may result in the neurochemical deficits and behavioral abnormalities that have been reported in BACE1 transgenic models [ 16 , 18 , 61 , 70 ]. AD etiology remains elusive and the cause of the elevated Aβ levels observed in SAD remain to be determined. Indeed, whether increased BACE1 activity is sufficient to induce AD pathogenesis, and the initial cause (s) of the BACE1 elevation remains unknown. Several lines of evidence suggest that the observed BACE1 elevations may occur throughout the course of AD development. While our data indicates that the BACE1 elevation in AD is actively involved in AD progression and may be detected prior to the appearance of overt neurodegeneration [ 97 ], increases in BACE1 activity have also been demonstrated in vivo occurring in response to the induction of apoptosis; an event associated with the later stages of neurodegeneration [ 136 ]. It is plausible that different factors cause elevations in BACE1 levels and activity at different stages of the disease. The association of elevated BACE1 with overt amyloid pathology, in both Tg mouse models of AD and AD brain, indicates that Aβ may potentially cause the observed elevations in BACE1. Indeed, accumulating evidence suggests that BACE1 is a stress response protein [ 136 , 210 - 213 , 215 , 228 , 229 ] and Aβ is a known neurotoxin. However, while elevated Aβ levels may facilitate positive feedback loop whereby increased Aβ facilitates enhanced BACE1 activity and auto-potentiates its own production, there is accumulating data which links several well-established AD risk factors (cardiovascular diseases, stroke, ischemia and TBI), and their associated molecular changes (hypoperfusion, hypoxia, metabolic dysfunction, energy inhibition and oxidative stress), with increases in BACE1. The molecular mechanisms governing these associations are being intensely investigated. It is apparent that BACE1 expression is tightly regulated in a complex manner at both transcriptional (HIF; [ 214 , 224 ] and post-transcriptional levels including translational efficiency [ 102 ], lysosomal targeting [ 134 ], and proteasomal degradation [ 141 ]. Furthermore, additional regulation of BACE1 levels may be attributed to alterations in the levels of proteins known to associate with and affect the intracellular trafficking of the enzyme (GGA3; [ 136 ]). Clearly understanding the molecular mechanisms of BACE1 elevation during SAD may f

This work was supported by National Institutes of Aging Grants PO1 AG021184 and RO1 AG02260 (Dr. Vassar) and a John Douglas French Alzheimer's Foundation Fellowship (Dr. Cole). The authors would also like to sincerely thank Mrs. Eloise Goodhew Barnett, the John Douglas French Alzheimer's Foundation sponsor of Dr. Sarah L. Cole, for her continued support and interest in the research carried out in our laboratory. The Figures in this review were the careful work of Leslie Bembinster. Finally, Drs. Cole and Vassar would also like to thank the many scientists involved in AD and β-secretase research whose dedicated work made this review possible.