Modulation of in vitro activity of zymogenic and mature recombinant human β-secretase by dietary plants


L. Bennett, CSIRO Food and Nutritional Sciences, Private Bag 16, Werribee, VIC 3030, Australia
Fax: +613 9731 3390
Tel: +613 9731 3200


The in vitro activity of human recombinant β-secretase (BACE1) was studied using a fluorogenic substrate based on the cleavage site for the enzyme in the Swedish mutation of amyloid precursor protein. The enzyme was inhibited by a control peptide inhibitor with good repeatability. The enzyme preparation comprised a mixture of pro-enzyme or zymogen and mature enzyme whereby the pro-enzyme sequence forms a ‘flap’ that can obstruct the binding site. ‘Open flap’ forms of the zymogen and mature enzyme are active, but the ‘closed flap’ form of the zymogen is inactive. This mixture of enzyme populations permitted apparent stimulation of enzyme activity under particular conditions, presumably due to facilitating flap-opening of the zymogen. As reported for heparin, enzyme activation was stimulated in the presence of low concentrations of Tween 20 and dimethylsulfoxide before becoming inhibited at higher concentrations. Dietary plant extracts either consistently inhibited (e.g. clove, tea, cinammon) or consistently stimulated (e.g. mushroom, parsley, asparagus) BACE1. Common structural features identified by Fourier transform infrared spectroscopy revealed that BACE1 activity could be explained by differential interactions of either small molecule or polymeric species with mature versus zymogen forms of the enzyme, respectively. Further, enzyme activity could be reversed by mixtures of high and low mass species. These results may have implications for the regulation of β-secretase activity in vivo by either endogenous or possibly dietary factors and for a potential role of BACE1 in stimulation of the production of amyloid beta peptide in sporadic Alzheimer’s disease.


Alzheimer’s disease


amyloid precursor protein


β-site APP-cleaving enzyme




epigallocatechin gallate


Alzheimer’s disease (AD) is the most common type of dementia which ranks sixth in all causes of death in the USA and is increasing as the population ages [1]. In spite of intensive research focus on AD pathogenesis and efforts towards development of potential therapeutic interventions, there are currently no known effective treatments.

AD is characterized by the presence of senile amyloid plaques and neurofibrillary tangles in the brain with tangle assembly of hyperphosphorylated tau proteins occurring downstream of amyloid aggregation [2]. Amyloid plaques are formed by overproduction and subsequent aggregation of the 40–42 amino acid peptide, β-amyloid (Aβ). β-amyloid is produced by cleavage of the membrane-anchored amyloid precursor protein (APP) by the β-site APP-cleaving enzyme (BACE1, EC, also known as β-secretase. This forms the peptide identified as C99, which is subsequently cleaved by γ-secretase to form either Aβ40 or Aβ42 [2]. The longer form of Aβ is implicated as causative in AD through association with early onset AD. In vitro studies have shown that Aβ42 aggregates into fibrils faster than Aβ40 and it is considered to be more toxic than Aβ40 [3]. The alternative non-toxic pathway for APP processing, whereby APP is cleaved by α-secretase and then γ-secretase to produce p3, a peptide not involved in AD [2], is the pathway that predominates in non-neuronal cell types.

The inhibition of enzymes responsible for β-amyloid production, BACE1 and γ-secretase, representing the ‘toxic’ pathway of APP processing, are therefore considered as possible therapeutic targets for AD [4]. BACE1 is the rate-limiting enzyme for Aβ42 production and is upregulated under conditions of induced cerebral hypometabolism generating energy insufficiency and oxidative stress in a transgenic mouse model, but the reason for the upregulation of Aβ42 production in AD is unknown [5]. Inhibitors of BACE1 based on inhibitors of cathepsin B inhibited the processing of wild-type APP into Aβ42, lowered plaque load and improved memory in mouse studies [6]. Inhibition of these enzymes could reduce the levels of β-amyloid protein in the brain and might retard or stop progression of the disease.

Recent studies have shown BACE1 to have substrates and physiological roles in neuronal function other than APP cleavage with specific functions yet to be fully elucidated [7]. Nevertheless, apart from minor behavioural changes [7], BACE1 knockout mice show no phenotype with minimal toxicity expected from its inhibition [8]. Inhibition of BACE1 has been shown to prevent Aβ production and prevent neuronal loss in transgenic mice [9]. Conversely, γ-secretase is known to have over 50 substrates [10] and knockout mice lacking one of the subunits required for activity do not survive to birth [11].

BACE1 is a transmembrane aspartic protease, which together with BACE2 forms a novel branch of the pepsin family [8]. BACE1 comprises 501 amino acids with 64% sequence homology to BACE2, and <30% sequence homology to human pepsin family members [8]. Only BACE1 occurs significantly in the brain [12]. Unlike other members of the aspartyl protease family, the presence of the pro-domain on BACE1 does not abolish activity [13]. The zymogen is partially active and the pro-domain has the function of assisting in protein folding [14]. The mature enzyme is released following cleavage of the pro-domain by furin and other members of the pro-enzyme convertase family [15,16]. The zymogen exhibits moderate activity compared with the mature enzyme [17] that can be inhibited in the presence of the pro-domain peptide [17].

A range of in vitro inhibitors of BACE1 have been identified and testing has progressed into in vivo animal models in order to optimize with respect to metabolic stability, oral bioavailability and brain uptake [18,19]. Inhibitors based on peptidic, natural and synthetic chemical species have been developed with interest in development of non-peptidic therapeutic agents growing significantly in recent years [20,21]. For example, the synthetic inhibitor KMI-429, a peptide spanning the BACE1 binding site of APP containing a hydroxymethyl isostere linkage, successfully downregulated Aβ production in cell and APP transgenic mouse studies [22]. Novel molecular inhibitors of BACE have also been based on transition state isostere analogues of isonicotinamide [23], and inhibitors of acetyl cholinesterase were found to exhibit potent BACE1 inhibition activity [24]. BACE1 inhibition was also achieved with anti-sense oligodeoxynucleotides directed to the Swedish mutation of APP associated with inherited AD [25], and antibodies targeted to the APP substrate at the BACE1 binding site exerted downregulation of Aβ40 expression in neuroblastoma and astrocytoma cells [26].

In vitro BACE1 inhibitory activity has also been discovered in a range of natural products (Table 1), particularly by metabolites of polyphenolic structure [27,28]. Analogues of green tea catechins including epigallocatechin gallate (EGCG) exhibited non-competitive inhibition of BACE1 with IC50 values in the micromolar range [29]. Other non-competitive inhibitors of BACE1 were identified from natural products, including transcis resveratrol and related compounds [29] and deacetylated and aminoethylated chitosans [30] and polyphenolics: ellagic acid and punicalagin from pomegranate husk [31]. A subsequent feeding study with pomegranate juice to AD transgenic mice demonstrated significant improvement in memory and lowering of plaque deposition that could be attributed to the BACE1 inhibitory and probably anti-oxidant activity [32].

Table 1.   Summary of IC50 values of BACE1 inhibitors found in plants, showing structures. NC indicates that inhibition was non-competitive.
Resveratrol oligomersPaeonia lactiflora seed0.41 × 10−6 minline image (monomer)[71]
Resveratrol (cistrans mixture)Smilax Rhizoma15 × 10−6 m (NC)inline image[72]
OxyresveratrolSmilax Rhizoma7.6 × 10−6 m (NC)inline image[72]
VeraphenolSmilax Rhizoma4.2 × 10−6 m (NC)inline image[72]
Cis-scirpusin ASmilax Rhizoma10 × 10−6 m (NC)inline image[72]
LuteolinPerilla frutescans var. acuta5.0 × 10−7 m (NC)inline image[27]
Rosmaranic acidPerilla frutescans var. acuta2.1 × 10−5 m (NC)inline image[27]
Ellagic acidPomegranate husk3.9 × 10−6 m (NC)inline image[31]
PunicalaginPomegranate husk0.41 × 10−6 m (NC)inline image[31]
HeparinPorcine intestinal mucosaConcentration-dependent effect on enzyme activityinline image[33]
HispidinPhellinus linteus4.9 × 10−6 minline image[73]
1,2,3,4,6-Pentagalloyl-b-d-glucopyranosideSanguisorbae Radix3.76 × 10−6 m (NC)inline image[74]
Tellimagrandin IISanguisorbae Radix (edible herb)3.1 × 10−6 m (NC)inline image[74]
(−) Epigallocatechin gallate (EGCG)Green tea1.6 × 10−6 m (NC)inline image[29]
(−) Epicatechin gallateGreen tea4.5 × 10−6 m (NC)inline image[29]
(−) Gallocatechin gallateGreen tea1.8 × 10−6 m (NC)inline image[29]
Catechin (+ or −)Green tea∼ 3.2 × 10−5 m (NC)inline image[29]
Epicatechin (+ or −)Green tea∼ 2.5 × 10−5 m (NC)inline image[29]

In addition to inhibitory activity, the activation of BACE1 by heparin has also been reported [33]. Observations of both inhibitory and activation activity following the screening of a large library of dietary plants (> 200 species and variants) led us to studies aimed at understanding this behaviour. The aim of this study was to investigate the compositional factors in dietary plants responsible for variability in the activity of recombinant BACE1. The baseline variability in the assay was characterized before investigating the effects of chemical environment and compositional profile of selected dietary plants on enzyme activity. The results were interpreted by considering differential interactions of agents with either the zymogen or the mature enzyme.

Results and Discussion

Structural and chemical determinants of BACE1 activity

The BACE1 inhibition assay was adapted to 96 well plate format and optimized with respect to substrate and enzyme concentrations for optimal between-day coefficient of variation for assay repeatability (8.3%). The typical fluorescence changes of assay reagents in the absence and presence of the control inhibitor showed that background levels of fluorescence from the enzyme and substrate were stable under standard assay conditions and could be arithmetically accounted for (Fig. 1). The suppression of enzyme activity by the positive inhibitor control is evident as a lowering of fluorescence due to suppression of substrate cleavage (Fig. 1).

Figure 1.

 Fluorescence intensity monitored over 2 h, showing contributions from reagents and positive control and the effect of added DMSO (0.9 m, 8% v/v) on enzyme activity. Error bars represent the standard deviation from = 2–20 replicates. The amino acid sequence of the positive control and location of the uncleavable statine residue (3S,4S)-4-amino-3-hydroxl-6-methylheptanoic acid are included.

The recombinant enzyme preparation comprised a mixture of unspecified proportions of pro-enzyme (zymogen) and mature enzyme. Interactions of solutes with the zymogen ‘flap’ that promoted access of the substrate to the active site were manifested as ‘enzyme activation’ as shown to occur in the presence of 8% (0.9 m) dimethylsulfoxide (DMSO) (Figs 1 and 2). Batch-wise variation in the zymogen content of the enzyme mixture would be expected to lead to variation in capacity for enzyme activation (i.e. negative inhibition) while activity and inhibition of activity of the mature enzyme should be approximately constant, assuming minor variation in absolute concentration of mature enzyme used under assay conditions. The phenomenon of activation of BACE1 has been previously reported and attributed to zymogen-specific effects [33].

Figure 2.

 Effects of the presence of (A) Tween 20 and (B) DMSO on the inhibition of BACE1 showing dose–responses for Tween 20 or DMSO alone and for increasing concentrations of the positive control in the absence or presence of 0.0005% of Tween 20 or 5.0% (0.56 m) of DMSO, respectively. Values on the x axis refer to concentrations of either Tween 20 or DMSO alone, or control inhibitor for mixture experiments. (C) Concentration dependence of the effects of heparin on inhibition of BACE1. Results represent the mean of duplicate determinations with error bars calculated from half the range.

Application of the assay to analysis of dietary plants revealed examples of enzyme activation or ‘negative inhibition’ which was further studied. Possible redox chemical reactivity of plants and enzyme were evaluated by studying the effects of treatment with various oxidants and reductants. The results revealed that the presence of either glutathione or dithiothreitol increased the fluorescence or cleavage rate of the substrate but did not affect the enzyme (not shown). Likewise, BACE1 activity was unaffected by treatment with polyphenol oxidase (not shown). These effects suggested that redox chemical reactivity of plant sample components might contribute to either positive or negative changes in substrate fluorescence, representing enzyme activation or inhibition, respectively, but would not be expected to transition between positive and negative fluorescence changes as a function of concentration.

Effects of Tween 20 and DMSO on BACE1 activity were also studied (Fig. 2). In both cases, the concentration-dependent activity of BACE1 was U-shaped exhibiting transitional enzyme activation before inhibition at higher concentrations. Maximum negative inhibition was observed at 13 nm Tween 20 (Fig. 2A) and 1 m DMSO (Fig. 2B), respectively. When a fixed level of either Tween 20 (0.0005%, 4.1 nm) or DMSO (5% or 0.5 m) was included with the standard inhibitor, the enzyme stimulation (negative inhibition) effect was abolished in both cases (Fig. 2). For Tween 20, the inhibitory activity of the control inhibitor was potentiated by Tween 20 with decrease of the EC50 (curve shift to the right, Fig. 2A). This was in agreement with Pocari et al. [34] using a time-resolved BACE1 inhibition fluorescence assay, who showed ∼ 8-fold activation of BACE1 by 0.1% Tween 20 (814 nm). For DMSO, the EC50 of the standard inhibitor was unaffected by the presence of DMSO at 0.56 m (Fig. 2B). In order to suppress the tendency for activation of the enzyme, the assay was subsequently conducted using 5% DMSO as a standard condition in agreement with that used by Klaver et al. [35]. Others have shown that BACE1 inhibitor EC50 values were not affected by levels of DMSO up to 15% [36].

Interactions of BACE1 with heparin have been described that also involve concentration-dependent enzyme activity that transitions from activation to inhibition with increasing concentration [37]. In the BACE1 activity assay using 5% DMSO, heparin significantly activated BACE1 at concentrations above 1 μg·mL−1 but there was no evidence of enzyme inhibition up to 500 μg·mL−1 (Fig. 2C). Using a recombinant preparation containing only mature BACE1 and no zymogen, heparin sulfate was previously shown to inhibit BACE1 cleavage of APP in vitro through binding directly to the enzyme, which was strongly dependent on the average molecular mass of the heparin and also the nature of the heparin-bound saccharide [38]. Furthermore, BACE1 activity was elevated when cellular synthesis or sulfation of heparin sulfate was inhibited and heparin was identified as a ‘natural regulator’ of BACE1 [38]. Subsequent studies showed that heparin activated BACE1 at low concentrations (1 μg·mL−1) and inhibited the enzyme at higher concentrations (10–100 μg·mL−1) [33,37]. Differences between molecular forms of heparin could account for the different effects on BACE1 as a function of concentration.

As for other aspartic protease inhibitors, BACE1 contains a secondary alcohol residue that acts as a transition state mimetic forming a hydrogen bond with the catalytic aspartic acid groups [39,40]. The integrity of the hydrogen bond or a substituted water molecule [41] is necessary for activity and explains the strong dependence of activity on pH. The optimal pH for activity of the mature enzyme is 4.5, above which activity diminishes due to closing of the substrate binding site and below which activity diminishes due to release of the Asp-bound water molecule or interruption of the Asp-linked hydrogen bond. These results show that Tween 20, DMSO and heparin appear to exert a common capacity to activate the BACE1 activity at specific molecular ratios.

BACE1 is expressed as a pro-enzyme or zymogen with a 47-residue sequence that is cleaved by a protease into its most active form [36]. The activity of BACE1 is regulated by both its pro-enzyme sequence and perturbations in its structural configuration. As such, BACE1 may be susceptible to specific ligands and solvation factors present in either in vitro or in vivo environments. The pro-enzyme sequence forms a ‘flap’ with capacity to block the active domain and prevent substrate access. As such, the zymogen possesses low activity in the closed flap configuration and partial activity in the open flap configuration, while the mature enzyme with the pro-enzyme sequence cleaved is fully active under optimal conditions [36]. The recombinant human BACE1 thus exists as three distinct enzyme populations at room temperature [39] of increasing activity: closed flap zymogen; open flap zymogen; and mature enzyme. Recombinant BACE1 has been reported to exhibit poor affinity to known substrates, e.g. APP, which may reflect the presence of zymogen forms in which access to the active site is restricted [42].

A model of the relative activities of respective molecular forms of BACE1 has been proposed to account for the effects of exogenous species on enzyme activity (Fig. 3). The model suggests that inhibitors binding in the vicinity of but not in the binding site are adequate to open the flap of the zymogen and activate the enzyme. Inhibitor binding may either act directly by blocking the binding site or elicit a cooperative response leading to closure of the pro-sequence flap [40]. An unusual property of BACE1 appears to be the capacity for an exogenous species to stimulate enzyme activity by inducing opening of the zymogen flap or displacement of the equilibrium towards the open form, which involves only a small change of free energy [43]. This schematic depicts inhibitors competing with the substrate at the binding site, in the open flap enzyme state for illustrative purposes. However, it is more likely that inhibitors of the open flap act in non-competitive mode (Table 1). The inclusion of 5% DMSO in the assay was intended to suppress observed activation of the enzyme at low concentrations of selected inhibitors, i.e. those proposed to interact with the zymogen flap. By working at the level of DMSO where enzyme activity was maximal (∼ 5% or 0.56 m, Fig. 2B), this ensured that the flap of the zymogen was substantially ‘open’. However, further ‘activation’ of the enzyme by compounds tested in the presence of DMSO (e.g. heparin, beta glucan), suggested even further capacity of these compounds to activate the enzyme by zymogen flap interactions.

Figure 3.

 Schematic representation of the proposed mechanism of interaction of positive or negative inhibitors (activator) on either closed flap or open flap forms of BACE1 zymogen or the mature enzyme, proposing that activating inhibitors can interact with both the zymogen flap and binding site to produce concentration-dependent effects. Binding interactions leading to opening of the zymogen flap can promote substrate access and enzyme activity. This schematic depicts inhibitors competing with the substrate at the binding site, in the open flap enzyme state for illustrative purposes although the non-competitive mode of inhibition is more likely to occur (Table 1).

Modulation of BACE1 activity in vitro

With this understanding of the capacity of BACE1 to be either activated or inhibited as a result of interactions with multiple domains of the enzyme (Fig. 3), the assay was applied to a selection of dietary plants and specific phytochemicals. While several other researchers have noted the phenomenon of ‘activation’ of BACE1 under the influence of either chemical environment or particular additives, the effect on assigning a lower baseline of enzyme activity as required for determination of inhibitor EC50 has not been addressed. Dose–response curve fitting tended to compute an average through enzyme activity values obtained at low inhibitor concentrations producing some lower baselines that were significantly non-zero, based on standard errors (not shown). Nevertheless, values of EC50 were determined without forcing fitted curves through zero or other adjustment. It is likely that activation of BACE1 activity evident in some inhibitor dose–response curves could affect fitted parameters and exert a small effect on the accuracy of values of EC50, within the standard error range. However, upper baseline values reflecting 100% enzyme inhibition are not expected to be affected. The EC50 values cited therefore reflect the standard conditions of the assay and dose–response profiles arising from interactions of samples with the mixed enzyme population. All plant preparations involved processing methods intended to inactivate endogenous enzymes such as polyphenol oxidase and to maximize solubility by the use of heating and ultrasonication.

Black tea extract inhibited BACE1 in a dose-dependent manner which could be partly explained by the inhibitory activity of EGCG (Fig. 4A). The apparent EC50 of EGCG was higher (less inhibitory) than a reported value of 1.6 μm (Table 1) presumably reflecting the effect of DMSO in the assay. Likewise, pomegranate husk inhibited BACE1 with an EC50 of 1.19 μg·mL−1 (Fig. 4B, Table 2) which could be partly explained by the inhibitory activity of ellagic acid with an EC50 value of 0.85 μg·mL−1 (2.8 μm, Fig. 4B, Table 2), in reasonable agreement with a reported value of 3.9 μm (Table 1). EGCG and ellagic acid are both heterocyclic polyphenolics with extensive peripheral substitution by hydroxyl groups, which may account for inhibitory activity.

Figure 4.

 Effects of selected dietary plants and related major metabolites on activity of BACE1 showing dose responses for (A) black tea and EGCG and (B) pomegranate husk and ellagic acid. All samples were tested in the presence of 5% DMSO and the results represent the mean of at least duplicate determinations with error bars calculated from half the range or standard deviation.

Table 2.   Summary of activity of selected processed plant products and individual compounds in the BACE1 inhibition assay, showing values of EC50 for positive inhibitors. Results are computed from the mean of at least two determinations of each sample concentration, showing the standard error associated with the calculated EC50. ND, not determined, low or negative activity.
Plant or compoundEC50 (μg·mL−1)R2(adj)BACE1 inhibitory activity (%)
AsparagusND −59.3 ± 2.8
Beta d-glucanND  
Cinnamon (ground)41.5 ± 12.00.94 
Clove64.0 ± 170.96 
Ellagic acid0.85 ± 1.30.88 
Mushroom – buttonND −26.3 ± 0.02
Mushroom – Honey BrownND −18.1 ± 0.1
Mushroom – EnokiND −18.6 ± 2.4
Mushroom – ShiitakeND −18.7 ± 4.4
Parsley (continental)ND −20.4 ± 1.5
Pomegranate husk1.19 ± 0.110.99 

Clove extract also exhibited dose-dependent inhibition of BACE1 (Fig. 5A) with an EC50 of 64.0 μg·mL−1 (Table 2). However, the major essential oil of clove, eugenol [44], did not account for BACE1 inhibitory activity in clove (Fig. 5A). In contrast, button mushroom (Agaricus bisporus) exhibited capacity for BACE1 activation (negative inhibition, Fig. 5B) over a broad concentration range, as was previously observed for heparin (Fig. 2C), but the major polysaccharide in button mushroom, beta d-glucan, did not account for enzyme activation (Fig. 5B). Furthermore, consistent activation of BACE1 was observed for Honey Brown, Enoki and Shiitake mushrooms suggesting the presence of a common bioactive component (Table 2).

Figure 5.

 Effects of selected dietary plants and related major metabolites on the activity of BACE1 showing dose–responses for (A) clove and eugenol and (B) button mushroom and beta d-glucan. All samples were tested in the presence of 5% DMSO and results of three independent assays are shown for button mushroom, each representing the mean of at least duplicate determinations with error bars calculated from half the range or standard deviation.

Beckman et al. [33] explained the concentration-dependent effect on recombinant BACE1 activity observed for heparin by the interaction of heparin with the zymogen flap of BACE1 that drove an open flap configuration and increased substrate accessibility. This interaction subsequently permitted auto-catalysis of BACE1 which manifested as apparent inhibition of enzyme activity. It is not known if the progressive inhibition of BACE1 activity in the presence of higher concentrations of Tween 20, DMSO or EGCG reflects true enzyme inhibition or the process of auto-catalysis but extended incubation experiments (16 h) did not indicate any decline in enzyme activity rate (not shown). The results reported here for bovine lung heparin (Fig. 2C) therefore suggested interactions of proteoglycan structures with the zymogen flap that stimulated activity but did not permit auto-catalysis. The structural properties of glycosaminoglycans necessary for zymogen binding and enzyme activation have recently been described [45] and suggest that other sources of branched polysaccharide such as mushroom [46] might explain the activation of BACE1 by button mushroom reported here. Others have noted that the zymogen ‘flap’ of BACE1 in addition to the ‘10s’ loop can be targeted for regulating activity of the enzyme [47].

A summary of plant metabolites and other reported inhibitors of BACE1 is given in Table 1, showing that the majority of inhibitors of polyphenolic type exhibited inhibition by a non-competitive mode. This signifies that enzyme activity was moderated by a mixed mode of inhibition involving affinity of the inhibitor for both the free enzyme and the enzyme–substrate complex but the inhibitor does not compete by direct competition for the substrate binding site. From this extensive precedent, it is reasonable to assume that the metabolites of a polyphenolic nature present in the dietary plants studied here also inhibited BACE1 by a non-competitive mode.

Characterization of BACE1 modulators by FTIR

Modulation of BACE1 activity by dietary plants and metabolites implicated a range of polyphenolics and polysaccharide structures. The complex plant extracts were analysed by FTIR spectroscopy using the fingerprint region (1600–800 cm−1) to characterize the chemical classes of dominant polymeric backbone and side chain residues.

The set of plants and compounds associated with strong inhibition of BACE1 including cinnamon, black tea, pomegranate, clove and ellagic acid revealed common spectral features (Fig. 6A). The broad peak from 3348 to 3070 cm−1 is due to the O–H stretching region and is common to most samples, with the C–H stretching region evident at 2923–2849 cm−1. The complex mixtures of carbohydrates present in the dietary plants masked signals from individual metabolites except for the broad peaks evident around 1620 and 1030 cm−1. Other putative identifications were aromatic =C–O–C vibrations in clove and cinnamon at ∼ 1030 cm−1 and of eugenol in clove at ∼ 1514 cm−1 [48]. A higher degree of fine structure is evident for pure ellagic acid, showing a major carbonyl peak at ∼ 1694 cm−1, which was also detected in pomegranate husk (Fig. 6A). The profile of black tea reflected the oxidized state of polyphenolics [49] as a result of tea processing [50].

Figure 6.

 FTIR profiles comparing a series of (A) inhibitors of BACE1 including cinnamon, black tea, pomegranate husk, clove and ellagic acid and (B) activators of BACE1 including button mushroom, parsley, asparagus and heparin.

In comparison, the set of plants and compounds associated with activation of BACE1 including button mushroom, parsley, asparagus and heparin (Fig. 6B) revealed structural features that were common to the inhibitors but also peaks that were distinct from the set of BACE1 inhibitors (cf. Fig. 6A,B). As for the inhibitors, hydroxyl and C–H stretching peaks were evident at 3433–3248 cm−1 and 2925–2850 cm−1 respectively, of similar intensity to the inhibitor set (Fig. 6A). In contrast to the inhibitor set, peaks at 1620–1628 cm−1 and 1581 cm−1, mainly due to amide I and amide II stretching, suggested the presence of proteins, while profile similarities with heparin [51] also suggested that either proteoglycan or glycosaminoglycan structures were present in these plants. A broad peak at 1400 cm−1 appears to be exclusive to the set of activating plants and compounds, and was associated in mushroom with C–H and C–O–H (pyranose ring) bending [52]. The FTIR profile for mushroom is in good agreement with published data [53].

The FTIR technique provided an informative approach to characterizing plants and seeking patterns that permitted speculation regarding the compositional differences that could account for capacity to either inhibit or activate BACE1. We hypothesize that the macromolecular structures in the ‘activator’ plants (Fig. 6B) have capacity to interact with the zymogen flap and promote substrate access to the binding site, thereby stimulating enzyme activity. In contrast, plants containing an abundance of polyphenolic metabolites (Fig. 6A) are proposed to inhibit by direct blocking at the active site. This hypothesis can be further tested with additional samples and fractions separated by size.

The results indicated more stringent structural requirements for either activation or inhibition of BACE1. That is, mono-phenolic eugenol was not inhibitory whereas a range of hydroxyl-rich polyphenolics were inhibitory. Furthermore, beta glucan polysaccharide was not activating whereas the glycosaminoglycan heparin structure was able to activate the enzyme presumably through interaction with the zymogen flap. In support, the screening of a library of 1400 small molecular weight compounds (< 500 Da) using the Perkin Elmer Victor BACE1 activity kit indicated that only two compounds (0.1%) promoted enzyme activation (Liu et al., unpublished data, International Conference on Alzheimer’s Disease, 2008), thus demonstrating that molecular size is an important factor in manipulation of the conformation of the zymogen flap.

The zymogen-interacting plants mushroom and asparagus have each been reported to contain N-linked glycans with mushrooms containing exclusively oligomannosidic structures and asparagus containing complex structures with an abundance of Fuc, Xyl and Lewis A residues [54]. The glycan content of parsley has not been reported but contains 13% dietary fibre which includes polysaccharides. Although plant sample preparation involving microwave heating followed by ultrasonication is expected to solubilize metabolites and oligosaccharides, it is also possible that ultrasonication promoted hydrolysis accompanied by oxidative polymerization of sugars and oligosaccharides, as was shown for mushroom polysaccharides [55]. It is therefore possible that polysaccharide structures that appeared to influence conformation of the zymogen flap were the result of chemical processes occurring during processing and were modified from the fresh plant.

Significance for modulation of BACE1 activity in vivo

Collectively, these data indicate that BACE1 activity is highly dependent on chemical environment and further suggest that, in its natural environment at the membrane surface [56], it may be subjected to passing molecular traffic and factors influencing membrane fluidity and permeability. A mixture experiment of beta glucan (no effect when tested alone) and ellagic acid (strong inhibitor when tested alone) demonstrated the modulation of enzyme activity towards an activated state at highest concentrations of ellagic acid and lowest concentrations of beta glucan (Fig. 7). Thus, even low concentrations of beta glucan reversed the inhibitory activity of ellagic acid, presumably due to a cooperative effect between ellagic acid and beta glucan that enhanced the interaction with the zymogen flap (Fig. 3). This result may exemplify capacity for activation or inhibition of BACE1 as regulated by the cellular membrane environment. It is difficult to speculate upon the specific endogenous factors in vivo that might cooperatively modulate the activity of BACE1 via interactions with the zymogen flap. However, this behaviour of BACE1 invites hypotheses in relation to the effects of molecular species that may become upregulated during the onset of AD, and that in stimulating BACE1 activity, also promote Aβ42 production.

Figure 7.

 Three-dimensional plot showing the effect of combinations of beta glucan and ellagic acid on the activity of BACE1. The combination of polyphenolic and polysaccharide species activated the enzyme in contrast to the concentration dependence observed for individual species. All samples were tested in the presence of 5% DMSO and the results represent the mean of duplicate determinations with error bars calculated from half the range.

For example, cell and transgenic mouse studies with EGCG produced a lowering of Aβ42 production by opposing pathways [57–59]. Specifically, EGCG stimulated the non-amyloidogenic α-secretase APP processing pathway while also appearing to upregulate BACE1 activity in SweAPPN2a cells [60]. This might be explained by cooperative interactions of EGCG and cellular glycans with the zymogen population of BACE1 as modelled in Fig. 3. EGCG also exerted protection against AD pathology via inhibition of Aβ42 fibril self-assembly and anti-oxidant activity [61], indicating a pleiotropic behaviour of this compound.

High cellular levels of cholesterol have been shown to promote APP cleavage and Aβ42 production as demonstrated in cell lines [62–64], showing increase in Aβ42 production as a result of diet-induced hypercholesterolaemia and decrease in Aβ42 production following treatment with cholesterol-lowering statins. This is supported by epidemiological studies linking a reduction in the incidence of AD in patients taking cholesterol-lowering medications [65,66]. It has also been suggested that Aβ42 regulates cholesterol production in a feedback system [67] by differential inhibition exerted by Aβ40 and Aβ42 of enzymes in the cholesterol synthesis pathway. However, it also appears that inadequate cholesterol can increase Aβ42 production levels [68] with positive correlation of cholesterol and BACE1 activity at high membrane cholesterol levels and negative correlation at low membrane cholesterol levels [69]. Thus it appears that BACE1 activity is influenced by cholesterol in a similar U-shaped manner to that demonstrated for Tween 20, DMSO, heparin and components of mushrooms. Apart from cholesterol, the implications for health or relevance to AD of dietary factors with cholesterol-like properties are unknown and unlikely to be bioavailable to brain in the chemical state associated with the undigested product.

The activity of BACE1 in the brain appears to be crucial in determining the processing of APP into Aβ42, and factors responsible for the regulation of activity of zymogenic and mature populations of this enzyme may be important in disease pathogenesis, particularly in sporadic AD where BACE1 levels are elevated. The capacity for food-related macromolecular versus small molecular structures to interact exclusively with BACE1 zymogen versus mature enzyme, respectively, may provide a model that assists understanding of BACE1 activity in vivo. The results may also have implications for identification of dietary factors or other therapies to regulate the activity of BACE1, where it seems necessary to inhibit both zymogen and mature forms while avoiding zymogen activation.

Materials and methods

BACE1 (His tag, human, recombinant), β-secretase fluorescent substrate IV and APP β-secretase inhibitor were obtained from Calbiochem (San Diego, CA, USA). DMSO, Tween 20, EGCG (95%, from green tea), eugenol, beta glucan from barley and heparin (sodium salt) from bovine lung were obtained from Sigma (St Louis, MO, USA). Ellagic acid (> 96%) was obtained from Fluka (St Louis, MO, USA). Ascorbic acid was obtained from Melbourne Food Ingredient Depot (Melbourne, Australia), and ethanol was from CSR (North Ryde, New South Wales, Australia).

BACE1 inhibition assay

Recombinant human BACE1 (amino acids 1–460, EC with a C-terminal His tag, expressed in NSO murine myeloma cells, was supplied as a mixture of the 68–70 kDa pro-enzyme or zymogen and the 65–67 kDa mature enzyme. The ratio of zymogen to mature enzyme was not specified and possibly varied between batches. Standard conditions of the assay were as follows. A test sample aliquot (25 μL in assay buffer: 100 mm sodium acetate, pH 4.5, containing 20% DMSO) was pre-incubated (15 min at 22 °C) with BACE1 enzyme (25 μL, 4 μg·mL−1 in assay buffer, no DMSO) before adding the fluorescent substrate (50 μL, 25 μm in assay buffer, no DMSO) and commencing monitoring of the reaction in a 96 well plate for 2 h at 22 °C. The microplate fluorescence reader (Biotek FL600; Bio-tek Instruments, Winooski, VT, USA) was set at excitation and emission wavelengths of 360 and 485 nm, respectively. Enzyme activity was observed as a linear increase in fluorescence over the monitoring period and was expressed as percentage inhibition [1 − (Fsample/Funinhibited control)] × 100, after correction for appropriate reagent controls. The positive control was β-secretase inhibitor peptide (EC50 = 242 nm) used at 300 nm, which was included in each assay to monitor between-day variability. For dose–response experiments, the concentration of the sample was the only variable altered, with all other conditions remaining constant. The assay was consistent between days with an average coefficient of variation of 8.3%.

Dried plant samples prepared as described below were dispersed in 10% ethanol at approximately 50 mg·mL−1 total solids by vortexing and were agitated during extraction (30 min at 22 °C). The supernatants recovered by centrifugation (2000 g for 15 min) were diluted 1/100 in assay buffer containing 20% DMSO before a quarter dilution in the assay to a final nominal solids concentration of 50 μg·mL−1 and 5% DMSO (0.64 m) in assay buffer. Other test samples were prepared in 10% ethanol at stock concentration ranges required for determination of half-maximal effective inhibitory concentration (EC50). Deviations from these standard conditions are described in the text and figure captions.

The values of EC50 were computed using the standard curve analysis function of sigma plot for windows Version 11.0 (Systat Software Inc., Chicago, IL, USA), by a four-parameter logistic equation (Eqn 1).


In semi-log plot format, the curve follows a sigmoidal shape which models a typical dose–response curve with a variable slope parameter [70]. The four parameters obtained from this type of fitting are min and max, the bottom and top respectively of the curve in units of inhibitor concentration; EC50 the half-maximal effective inhibitory concentration; and hillslope which characterizes the slope of the curve at EC50. The method of calculation of EC50 values complies with that described by Sebaugh [70] for determining ‘relative’ values of EC50. Values of EC50 are only reported if the experimental data provided adequate data points above and below computed ‘bend’ points of fitted curves calculated according to Sebaugh [70].

Dietary plant sample preparation

Plant foods were obtained from local retail suppliers in either a fresh form (pomegranate husk isolated from seed flesh, button mushroom, continental parsley, asparagus) or dried form (Madura black tea, clove, cinnamon). Fresh plants were washed before discarding inedible parts (except for pomegranate husk, which was processed for testing). Plant material was chopped and homogenized in a food processor before dispersing to 50% total solids (v/v) in water and processing (5 min on high setting). Dispersed products were heated by microwave (high power for 10 min) and cooled in an ice bath before adding ascorbic acid (0.1% final concentration) and ethanol (1% v/v final concentration) for microbial stabilization. Samples were finally ultrasonicated (ultrasonics using Hielscher Focus 1 kW probe at 100% power) for 5 min, in order to maximize the solubility of solutes, before freeze drying and storage at −18 °C.

FTIR spectroscopic analysis

The FTIR spectra were recorded in a Shimadzu 8400S FTIR spectrophotometer (Rydalmere, New South Wales, Australia) in transmittance mode equipped with a deuterated triglycine sulfate detector. Samples were analysed in the dry state held in a ZnSe crystal in horizontal attenuated total reflectance mode under constant pressure by performing 128 scans per sample of eight independent replicates, at 8 cm−1 resolution. After applying 10-point smoothing, profiles were auto-corrected with a background spectrum (air) recorded under the same conditions, and corrected spectra were compared for repeatability. Spectra were analysed using ir prestige-21 software (Shimadzu Scientific Instruments, Columbia, MD, USA) and a representative spectrum was selected for reporting purposes.


This work was funded within the CSIRO Preventative Health National Research Flagship and we are also grateful for helpful discussions with colleagues from CSIRO Material Science and Engineering, Parkville, Victoria.