Cognitive-Enhancing Effects of a Polyphenols-Rich Extract from Fruits without Changes in Neuropathology in an Animal Model of Alzheimer’s Disease

Animals

The homozygous 3xTg-AD mouse model used here has already been described elsewhere [34, 37–39]. Briefly, this mouse model of AD harbors mutant presenilin-1 (PS1_M146V), APP_swe, and tau (P301L) transgenes. The human APP and tau-independent transgene constructs have been co-injected in embryos harvested from mutant homozygous PS1_M146V knock-in mice. Both human APP and tau transgenes are under the control of the Thy1.2 regulatory elements, co-integrated in the same locus and therefore inherited together [34]. The model progressively develops the two main pathological hallmarks of AD, namely Aβ deposits and neurofibrillary tangles (tau pathology), but also shows significant synaptic and cognitive deficits [35, 36]. It is thus an appropriate model for preclinical studies on potential drug therapies in early AD. Non-transgenic (NonTg) mice were derived from the original mouse line and were from the same genetic background. One hundred and twenty (120) non-transgenic (NonTg) and triple-transgenic (3xTg-AD) 12-month-old mice (60 females and 60 males) were assigned to this study. Seven mice (evenly distributed between groups) were found dead and were excluded from our analyses. Their death was unrelated to experimental treatment according to histopathological analyses performed by the staff of the animal facility. Moreover, an additional group of 4-month-old C57BL/6 mice (n = 10) were used as a young control group.

Drugs

The Neurophenols consortium extract is made of blueberries (Vaccinium angustifolium) and grapes (Vitis vinifera L.) extracts containing specific low-molecular-weight polyphenols (monomers, oligomers, flavonols, anthocyanins, phenolic acids, and resveratrol), formulated in a unique patented ratio of molecules. The Neurophenols consortium extract was introduced in mouse pellets in order to achieve constant plasma levels in mice. Mice were fed a control diet (named “Tpolyph”) or 500 mg of extract/kg(bw) per day (named “Polyph1”) or 2500 mg of extract/kg(bw) per day (named “Polyph2”). These doses correspond to a 2–12 g/day intake of the Neurophenols consortium extract according to U.S. Food and Drug Administration’s table for conversion between species for a human weighing 60 kg and assuming body surface area equivalence. Proportions for each polyphenol class were adjusted for both PEBG doses, which mainly contain flavan-3-ol monomers from French grape extract at respectively 129 mg/kg(bw) per day and 644 mg/kg(bw) per day, and flavonols from blueberry at respectively 4.6 mg/kg(bw) per day and 22.3 mg/kg(bw) per day. Proportions of other components of the Neurophenol consortium extract remain confidential. The dose of 500 mg/kg(bw) per day was determined based on a review of the literature from cognitive studies using grape seed or blueberry extracts in rodents and on the fact that adjustments had to be done to obtain the specific proportions for each polyphenol class mentioned above. We added a higher dose, 2500 mg/kg body weight, to evidence a dose-response effect and to allow us to perform correlative analysis between phenolic content in plasma and other endpoints.

Experimental schedule

The timeline of the study is shown in Fig. 1. For colony maintenance and to avoid interfering with developmental processes, all mice were kept on the mouse breeder chow (Teklad 2018, Harlan Laboratories, Canada) at our animal facility for the first 12 months. High dietary fat intake is widespread in our modern society and has been identified as a risk factor underlying the constant increase of AD prevalence over the last few years [40, 41]. The deleterious effect of high-fat diet on AD-like neuropathology and cognitive performance has been consistently replicated in animal models [37, 43]. Therefore, to mimic the current diet composition of what most people actually eat in industrialized nations, the animals in this study were fed a “Western diet” with 20% (w/w) of total fat. Table 1 summarizes diet composition, with fatty acid content confirmed by gas chromatography. The purified diet formula was produced by Research Diet Inc. (New Brunswick, USA) and precisely determined to avoid batch-to-batch variations commonly found in laboratory chow.

Experimental PEBG treatments were started at 12 months and lasted 4 months, as described in Fig. 1. A cognitive evaluation was performed at 15 months of age and all mice were sacrificed at 16 months of age. Thus, animals used in this study were divided into six groups: NonTg_T for non-transgenic mice fed control diet, NonTg_P1 for non-transgenic mice fed “Polyph1” diet, NonTg_P2 for non-transgenic mice fed “Polyph2” diet, 3xTg-AD_T for triple-transgenic mice fed control diet, 3xTg-AD_P1 for triple-transgenic mice fed “Polyph1” diet and 3xTg-AD_P2 for triple-transgenic mice fed “Polyph2” diet. An additional group of 4-month-old NonTg mice (NonTg_CTL) without any treatment was added as a control to evidence age-related cognitive impairment in older animals. All procedures were approved by the Laval University animal ethics committee and were performed in accordance with the Canadian Council on Animal Care guidelines.

Behavioral analysis

Samples preparation

Animals were placed under deep anesthesia with ketamine/xylazine and then, intracardiac blood samples were drawn. Mice were perfused via intracardiac perfusion with phosphate saline buffer (PBS 1X, 1 mM KH₂PO₄, 10 mM Na₂HPO₄, 137 mM NaCl, 2.7 mM KCl, pH 7.4) containing a cocktail of protease inhibitors (SIGMAFAST^TM Protease Inhibitor Tablets, Sigma-Aldrich, St. Louis, MO) along with phosphatase inhibitors (1 mM sodium pyrophosphate and 50 mM sodium fluoride). Extracts of the parieto-temporal cortex were dissected, snap frozen on dry ice, and kept at –80°C. Homogenates from cytosol fractions (TBS-soluble), membrane fractions (detergent-soluble), and detergent-insoluble fractions (formic acid-soluble) were prepared in buffers including protease and phosphatase inhibitors as already shown [35, 50]. Briefly, tissue samples were homogenized in 8 volumes of TBS containinga cocktail of protease inhibitors (Complete™, Roche Diagnostics, Indianapolis, IN, USA, and 10 μg/ml of pepstatin A) and phosphatase inhibitors [phosSTOP (Roche Diagnostics), 1 mM sodium orthovanadate and 50 mM sodium fluoride]. Samples were briefly sonicated (3×5 s) and centrifuged at 100,000× g for 20 min at 4°C to generate a TBS-soluble fraction (cytosol fraction). The TBS-insoluble pellets were sonicated (3×5 s) in 8 volumes of lysis buffer (150 mM NaCl, 10 mM NaH₂PO₄, 1 mM EDTA, 1% Triton X-100, 0.5% SDS, and 0.5% deoxycholate) containing the same protease and phosphatase inhibitor cocktail. The resulting homogenates were centrifuged at 100,000× g for 20 min at 4°C to produce a lysis buffer-soluble fraction (detergent-soluble membrane fraction). Lastly, the pellet fraction was resuspended in 99% formic acid to form the insoluble fraction. Samples used for western blot analysis were resuspended in Laemmli 1X. Protein concentrations in TBS- and detergent-soluble fractions were determined using bicinchoninic acid assays (Pierce, Rockford, IL). Molecular mechanisms were investigated in the parieto-temporal cortex because the object recognition memory is highly dependent on cortical regions comprised within [51].

ELISA

Cortical soluble (TBS fraction) and detergent-insoluble (formic acid-soluble) Aβ₄₀ and Aβ₄₂, and detergent-insoluble total human tau were analyzed separately by ELISA, as described previously [5, 52]. The detergent-soluble protein fractions were not used for these experiments. We used high sensitivity ELISA kits from Wako (Osaka, Japan) to detect both Aβ₄₀ and Aβ₄₂ and from Life Technologies/Thermo Fisher Scientific (Carlsbad, CA) to detect total human tau in the brain of 3xTg-AD mice. ELISA was performed according to the manufacturer recommendations and the plates were read at 450 nm using a Synergy^TM HT multi-detection microplate reader (Biotek, Winooski, VT).

Western blotting

Immunoblotting experiments were performed as described previously [5, 52]. Equal amounts of protein per sample (15 μg of total protein per lane) were added to Laemmli’s loading buffer, heated to 95°C for 5 min before loading in a random order, and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Proteins were electroblotted onto PVDF membranes (Millipore, Billerica, MA) before blocking in 5% nonfat dry milk and 0.5% bovine serum albumin (BSA) in PBS-Tween 20 for 1 h. Membranes were immunoblotted with appropriate primary and secondary antibodies followed by chemiluminescence reagents (Lumiglo Reserve, KPL, Gaithersburg, MD). Band intensities were quantified using a KODAK Image Station 4000 MM Digital Imaging System (Molecular Imaging Software version 4.0.5f7, Carestream Health, Rochester, NY). Refer to the summary table of the different antibodies used in this study (Table 2).

Immunohistofluorescence

Hemi-brains were postfixed with 4% PFA (4°C for 48 hours), cryoprotected with 20% sucrose-PBS, snap frozen at –80°C, and microtome-sectioned into coronal (25 μm) sections. Washes in 0.1 M PBS, pH 7.4, were performed between each step of the immunohistofluorescence protocol. Free-floating brain sections from mice were blocked for 1 h in a PBS solution containing 5 % horse serum (Invitrogen, Carlsbad, CA) and 0.4% Triton X-100. Sections were then incubated overnight at 4°C with primary antibody in the blocking solution. For neurofibrillary tangle detection, mouse anti-tau antibody (clone tau13; Covance, Berkeley, CA, USA) was used as primary antibody, followed by detection with Alexa Fluor 488-labeled secondary donkey antibody (Life Technologies, Burlington, ON, Canada). The nuclei were counterstained with 4^′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific, Waltham, MA). Finally, slices were mounted on SuperFrost Plus slides (Thermo Fisher Scientific), treated with 0.5% Sudan black (in 70% methanol) for 5 min, and placed under coverslips with Mowiol mounting medium. Immunofluorescence images were acquired using an EVOS® FL Auto Cell Imaging System (Life Technologies/Thermo Fisher Scientific) with a 20X objective. Selected photographs were taken at Bregma –2.92 mm.

Characterization of Plasma Phenolic Metabolites

Phenolic metabolites were extracted from plasma and analyzed as previously described [53]. Briefly, Waters OASIS HLB μ elution plates 2 mg –30 μm (Milford, MA, USA) were preconditioned using 250 μl of methanol and 250 μl of 0.2% acetic acid. Plasma samples (50 to 350 μl) were mixed with 4% phosphoric acid to disrupt phenol-protein binding, and loaded to the plates. The plates were then washed with 200 μl of ultrapure water and 200 μl of 0.2% acetic acid. The retained phenolic compounds were eluted with 75 μl of acetone/ultrapure water/acetic acid solution 70/29.5/0.5 v/v/v in presence of rosmarinic acid as internal standard (1 μg/ml final concentration), and then directly analyzed by ultra-high-pressure liquid chromatography tandem mass spectrometry (UHPLC-MS/MS). Their separation was performed at 30°C using an Agilent Plus C18 column (2.1 mm×100 mm, 1.8 μm) (Santa Clara, CA, USA). The mobile phase, consisting of 0.2% acetic acid in ultrapure water and acetonitrile (solvent A and B, respectively), was used with the following gradient elution: 0–8 min, 5–50% B; 8–9.10 min, 50–90% B; 9.10–10 min, 90% B; 10–10.10 min, 90–5% B; 10.10–13 min, 5% B, with a flow rate of 0.4 ml/min. The MS/MS analyses were carried out in negative mode using the following electrospray source parameters: electrospray capillary voltage: 3.01 kV, source temperature: 150°C, desolvation temperature: 400°C and cone and desolvation gas flows: 80 l/h and 800 l/h respectively. Cone voltage and collision energy parameters were optimized for each compound. The identification of metabolites was performed by comparing their retention times with those of available phenolic standards and/or analyzing their fragmentation. Data were acquired through the multiple reaction monitoring (MRM) mode, tracking the transition of parent product ion specific for each compound. The detected metabolites were quantified using standards when available. Otherwise, a relative quantitation was achieved using the calibration curve of their aglycone or most similar phenolic structure.

All chemicals analyzed (gallic acid, caffeic acid, vanillic acid, p-coumaric acid, 4-hydroxybenzoic acid, 4-hydroxyphenylpropionic acid, 3,4-dihydrox-yphenylacetic acid, 4-hydroxyphenyacetic acid, resveratrol, catechin, and epicatechin) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Liquid chromatography grade solvents acetone, methanol, and acetonitrile were purchased from EMD Millipore Chemicals (Billerica, MA, USA). Glacial acetic acid and phosphoric acid were obtained from Anachemia (Montreal, QC, Canada). Ultrapure water was obtained from a Millipore Milli-Q water purification system.

Statistical analyses

Data from all analyses were expressed as mean±SEM. Statistical comparisons of data between groups were performed with Kruskal-Wallis ANOVA analysis followed by Dunn’s multiple comparison tests when standard deviations were considered significantly different according to the Bartlett test. Otherwise, standard parametric ANOVA were used. Individual data were excluded for technical reasons or if determined as an outlier based on ROUT method using GraphPad Prism 6.0. Sex difference between male and female mice was statistically tested for each evaluated parameter with two-way ANOVA. Unless indicated otherwise, no sex difference was observed and data from males and females were pooled together. For the NOR test, the mean of each group was compared to the threshold of 50%, which represents an equal observation time between the old object and the new one. Coefficients of correlation and significance of the degree of linear relationship between parameters were determined with a simple regression model. The accepted level of significance was set at 0.05. Statistical analyses were done using Prism 6.0 or JMP Statistical Analysis Software (version 10.0).

The Neurophenols PEBG improved object recognition in 3xTg-AD mice without modifying anxiety level or motor activity

The memory of mice was evaluated using the NOR test (Fig. 2). The observation time ratio, or recognition index, was determined for each mouse and was compared with a value of 50% corresponding to the expected random selection of a mouse with equivalent interest in both objects. Younger control mice presented a recognition index of 62% (p = 0.007). In contrast, 15-month-old NonTg mice fed the non-supplemented diet did not interact with the new object more than the familiar one during the testing phase, as revealed by a recognition index of 53% (p = 0.316), highlighting a potential age effect. Next, NonTg mice fed Polyph1 or Polyph2 diets presented a recognition index significantly higher from the threshold of 50% (57%, p = 0.012 and 56%, p = 0.035, respectively). Interestingly, compared to 3xTg-AD mice under control diet, which presented a recognition index of 50%, dietary intake of PEBG led to higher recognition indexes (Polyph1 diet: 57%, p = 0.032; Polyph2 diet: 63%, p < 0.001) (Fig. 2A). Motor performance was not different between groups, regardless of diet, as revealed by the total duration of time mice spent on the rotarod device, with a tendency for 3xTg-AD to be more active than NonTg mice (Fig. 2B). These results were confirmed in the open field test where no difference of the total distance traveled by mice in one hour was observed between groups (Fig. 2C). Similarly, the dark emergence test did not reveal any effect of genotype or diet on anxiety levels (Fig. 2D), indicating that the effect of PEBG was specific to object recognition.

PEBG did not alter tau neuropathology in 3xTg-AD mice

First, we quantified total tau proteins (i.e., murine and human origin) by western immunoblotting in cytosolic fraction using a polyclonal antibody targeting both mouse and human tau C-terminus (Tau C, see Table 2). No difference was observed between genotypes and supplemented diets, regardless of the polyphenol doses (Fig. 3A). Similarly, no difference was observed for the phosphorylated forms of tau (clone CP13, tau phosphorylated at serine 202/threonine 205, see Table 2) in cytosolic fraction between groups (Fig. 3B). We confirmed these results with other specific phospho-tau antibodies (clones AT270 and PHF1; see Table 2 for phosphorylation sites and Table 4 for results). Similarly, PEBG did not significantly affect cytosolic levels (TBS-soluble) of total human tau in 3xTg-AD mice (Fig. 3C). Interestingly, despite a lack of significant difference, a strong downward trend of detergent-insoluble human tau (assessed with tau 12 antibody, see Table 2) was observed in 3xTg-AD mice fed the diet with the higher concentration of polyphenols (Fig. 4D, p = 0.077). Data obtained using clone tau 13 (see Table 4) were consistent with those gathered with tau 12 confirming the species-specificity of the polyphenols effect on human tau protein. Both monoclonal antibodies used here to detect human tau, clones tau 12 and tau 13, target epitopes in theN-terminal part of human tau protein [54, 55]. However, this tendency was not confirmed by ELISA analysis of human tau concentration in the formic acid fraction (Fig. 3E). Moreover, we investigated total human tau by immunofluorescence in the hippocampus of NonTg fed control diet and 3xTg-AD mice fed supplemented-diet or not. We first confirmed accumulation of human tau protein in hippocampus of 3xTg-AD compared to NonTg (Fig. 3F, G). However, none of the two diets with PEBG induced a significant decrease of total human tau in 3xTg-AD mice (Fig. 3G-I).

PEBG had no effect on cortical Aβ concentration in 3xTg-AD mice

We then investigated human soluble amyloid-β protein precursor (sAβPPα) and full-length human amyloid-β protein precursor (AβPP) levels (with clone 6E10, see Table 2) by western blot (Fig. 4A-C) and Aβ₄₂ and Aβ₄₀ levels by ELISA (Fig. 4D-I) in homogenates from the parieto-temporal cortex of 3xTg-AD animals. As previously reported, female 3xTg-AD mice displayed higher levels of human AβPP, soluble and insoluble Aβ₄₂ and Aβ₄₀ peptides than males, while no difference was observed on the ratio of TBS-soluble over detergent-soluble human AβPP and on the Aβ₄₂/Aβ₄₀ ratio in TBS- and formic acid-soluble fractions. Moreover, no effect of PEBG supplementation was detected on soluble and full-length AβPP levels (Fig. 4A-C) or on Aβ₄₂ and Aβ₄₀ accumulation in soluble (Fig. 4D and E, respectively) and insoluble (Fig. 4G and H, respectively) fractions, or in their respective ratios (Fig. 4F and I, respectively). Finally, PEBG, regardless of the concentration used, did not induce any change in levels of beta-secretase 1 (BACE) or of a disintegrin and metalloproteinase domain 10 (ADAM10) measured by western blotting, two proteins involved in amyloidogenic and non-amyloidogenic pathways, respectively (see Table 4 for results).

PEBG had no effect on synaptic, inflammatory, and energy metabolism proteins in the cortex of 3xTg-AD mice

Western blotting analyses revealed that PEBG supplementation did not induce any significant effect on drebrin levels in cytosolic and membrane fractions from the parieto-temporal cortex (Fig. 5A-C, respectively). Similarly, no treatment effect was observed on presynaptic proteins synaptophysin and synaptosomal-associated protein 25 (SNAP-25) (Fig. 5D and E, respectively). Moreover, levels of proteins involved in inflammatory pathway (nuclear factor kappa B) or energy metabolism (sirtuin-1) were not influenced by treatments (see Table 4).

PEBG restored mature BDNF/pro BDNF ratio in 3xTg-AD mice to NonTg level

PEBG induced an increase of the ratio of both forms of BDNF protein (mature/pro) in 3xTg-AD mice. The decrease of BDNF levels observed in 3xTg-AD mice compared to younger NonTg mice (Fig. 6; – 45%, p < 0.05) was restored in PEBG treated-mice, at both doses (Fig. 6; +57% for3xTg-AD_P1, p = 0.011 and, +59% for 3xTg-AD_P2, p = 0.026, compared to 3xTg-AD_T).

Plasma concentrations of phenolic metabolites correlated with cognitive performance

As revealed by UHPLC-MS/MS analysis, 18 phenolic metabolites were identified in the plasma, mainly intestinal and/or hepatic metabolites of catechin and resveratrol native compounds, as well as phenolic acids and microbial metabolites. Among them, 11 compounds were specifically found in the plasma of supplemented animals, with a dose-response effect (Table 3). Differences were observed between genotypes as catechins glucuronide, methyl catechins glucuronide, and 2-hydroxybenzoic acid were increased in the plasma of 3xTg-AD compared to NonTg mice on the same diet. On the other hand, hydroxyphenylvaleric acid was decreased in the plasma of 3xTg-AD compared to NonTg mice (Table 3). To probe the relationship between phenolic concentrations in plasma and cognitive performance, we performed linear regression analyses. Interestingly, levels of three important phenolic compounds found in plasma samples correlated positively with memory performance of mice evaluated in the NOR test. Indeed, regardless of genotype, three compounds displaying higher concentrations in the plasma of supplemented animals, and more specifically in the Polyph2 group, showed significant positive relationships with NOR performance of mice: 2-hydroxybenzoic acid (Fig. 7A, B; r² = 0.2229; p = 0.0032), hydroxyphenylpropionic acid (Fig. 7C, D; r² = 0.1257; p = 0.0465), and dihydroxyphenyl valerolactone (Fig. 7E, F; r² = 0.1771; p = 0.0165).

Improvement of memory performance of mice after polyphenol treatment: Does the present data fit in the literature?

In the present study, we confirmed the benefit of a diet supplemented with PEBG on cognitive performance in old mice. The NOR test used here, which is based on the fact that rodents have a tendency to interact more with a novel object than with a familiar one, has become a benchmark task to study learning and for assessing recognition memory [49, 56]. Our data show that PEBG intake prevented the age-induced decrease in performance in the NOR test observed in both NonTg and 3xTg-AD mice fed the control diet. This effect was more prominent and dose-dependent in 3xTg-AD animals. The absence of differences between groups in exploratory and motor behavior indicates that the effect of PEBG was specific for object recognition.

Accordingly, phytochemical-rich foods, particularly those rich in flavonoids, have been shown to be effective in reversing age-related deficits in memory and learning in preclinical studies. Oral administration of quercetin, a naturally occurring flavonoid, significantly improved the behavioral performance of high-cholesterol-fed old mice in the step-through test and the Morris water maze task [57]. Dietary supplementation with a grape seed phenolic extract (GSPE) prevented both Aβ and tau neuropathologies and attenuated cognitive deterioration observed in transgenic mouse models of AD [31, 32]. Moderate consumption of red wines containing high contents of grape polyphenols also reduced Aβ neuropathology and improved cognitive performance in the Tg2576 mouse model [58, 59]. More recently, a treatment with olive oil polyphenols was shown to improve contextual memory of aged C57Bl/6J mice in the step-down test to levels similar to young animals [60].

The identification of which polyphenols display sufficient bioavailability will be instrumental in our understanding of their mechanisms of action. Whether polyphenol-induced memory improvement is exclusively mediated by direct central action or through other mechanisms such as stimulation of endothelial function and peripheral blood flow [61] remains to be determined. Nonetheless, previous studies indicate that key polyphenols, such as flavonoids, may cross the blood-brain barrier and reach therapeutically significant concentrations in cerebral tissue [62, 63]. Our plasma data confirm the presence of catechin and epicatechin in the circulation after dietary supplementation with the Neurophenols PEBG. Sulfate and glucuronide forms of catechin as well as gallic, caffeic, and hydroxyphenylvaleric acids were particularly increased in a dose-dependent fashion. Interestingly, the positive correlations observed here between plasma concentration of 2-hydroxybenzoic acid, hydroxyphenylpropionic acid, and dihydroxyphenyl valerolactone with the memory performance of mice in the NOR test are consistent with cognition enhancing properties. The present data combined with previous work should lead us to consider that polyphenols may exert their behavioral effects through bioactive metabolites.

We also observed that some phenolic metabolites, namely catechins glucuronide, methyl catechins glucuronide, and 2-hydroxybenzoic acid, were increased whereas hydroxyphenylvaleric acid was decreased in the plasma of 3xTg-AD compared to NonTg mice on the same diet. Despite the great variety of natural dietary polyphenols, the diversity of polyphenol metabolites resulting from a gut microflora-dependent metabolism is relatively small due to a convergent pattern of bioconversion [64]. Although we cannot rule out a genotype-dependent variation at this point, if this was the case, one could expect differences in a larger number of polyphenol metabolites according to genotype.

The main neuropathological markers of AD are not modified by polyphenolic extracts in 3xTg-AD mice

Given the memory improvement observed in 3xTg-AD, one would expect that PEGB would have improved tau or amyloid pathologies concomitantly. Indeed, a main advantage of the 3xTg-AD model is to recapitulate key features of AD, such as Aβ and tau neuropathology as well as cognitive decline with increasing age [35, 45]. Sex difference in Aβ levels between male and female 3xTg-AD mice observed in this study is in accordance with our previous work [36, 38] and with reports from other groups [65–67]. However, the PEBG used in our study did not induce any significant effect on main neuropathological markers of AD in supplemented 3xTg-AD mice. This contrasts with previous reports in which dietary supplementation with a GSPE, or moderate consumption of red wines containing high contents of grape polyphenols, attenuated the progression of Aβ neuropathology in the Tg2576 mouse model [31, 59], and reduced tau aggregation in the brain of TMHT [32] or JNPL3 [68] mouse models of tau pathology. No effect on synaptic proteins was observed either, including drebrin, whose levels are decreased in AD [69, 70]. We previously reported that the drebrin loss in the brain cortex of 3xTg-AD was minimal compared to NonTg mice [35, 46], while no difference was observed in synaptophysin and SNAP-25 levels between NonTg and 3xTg-AD mice [35, 46]. In contrast, a previous study using a specific phenolic compound, resveratrol, already showed that APP/PS1 mice, another AD mouse model, demonstrated a 2.2-fold increase in drebrin when fed resveratrol [71]. Whereas the reasons underlying these discrepancies are unclear, our data are consistent with the growing body of evidence indicating that overt modulation of Aβ, synaptic, and/or tau neuropathology is not a prerequisite for treatment-induced cognitive enhancement [72, 73]. Neuropathological markers of AD such as Aβ, particularly soluble species, are subjected to important physiological variations (i.e., clearance mechanisms [39, 75] and the awake/sleep state [76]) and are not always directly associated with behavioral deficits [38]. A previous report suggests that functional modifications such as synaptic activity are more closely associated with behavioral expression of cognitive deficits [38]. Moreover, this lack of effect on AD markers is also consistent with the fact that the PEBG had beneficial effects on memory not only in 3xTg-AD mice but also in NonTg animals, suggesting that the Neurophenols consortium extract has a broader impact and might also be used in cases of age-related cognitive decline or cases without a definitive AD diagnosis.

Could polyphenolic extracts improve memory performance via BDNF-related pathways?

For more than two decades, BDNF has garnered significant attention as a potential therapeutic target for neurodegenerative diseases such as AD, but was not translated into clinical application due to its inability to cross the blood-brain barrier [77]. Indeed, BDNF-induced changes in neuronal spine density and morphology have been shown to dictate the efficiency of synaptic connections and, consequently, cognitive performance [78–80]. BDNF protein levels in the brain originate from the cleavage of inactive pro-BDNF into mature active BDNF by the tPA/plasmin enzyme system, which is expressed in synapses [81]. The reduced levels of active BDNF found here in 3xTg-AD mice compared to both young and old NonTg is in agreement with previous work in the same model [82], and suggest that Aβ or tau neuropathologies may progressively impair BDNF activity. Accordingly, abnormalities in BDNF activity have been reported in the brains of AD patients [83] and may represent a marker of cognitive deterioration [84]. The treatment with PEBG brought back active BDNF to levels comparable to non-transgenic animals on the control diet, and could thus underlie the observed improvement in object recognition in 3xTg-AD mice. Consistent with this hypothesis, intracerebral treatment with BNDF improved memory performance in 3xTg-AD [72] and in APP/PS1 mice [73], without altering Aβ or tau pathologies. Therefore, the PEBG-induced increase in mature active BDNF levels in the cortex of 3xTg-AD mice provides not only a potential mechanism of action of polyphenols, but also suggests that brain BDNF activity may be enhanced non-invasively using natural compounds in the diet.

This contention that polyphenols act on BDNF is fueled by a wealth of previously published evidence. Indeed, flavonoids can directly influence neuronal dendrite outgrowth in vitro as well as BDNF levels and neuronal morphology [88]. An in vivo study has shown that spatial memory performance in 12-month-old Wistar rats supplemented with blueberries (mainly anthocyanins) correlates well with the activation of CREB and with the increase of BDNF in the hippocampus [89]. Similarly, a flavonol found in strawberries (named fisetin) has also been shown to improve long-term potentiation and to enhance object recognition in male C57BL/6 mice by a mechanism dependent on the activation of BDNF and CREB [90]. More recently, a treatment with a 2% (w/w) blueberry diet resulted in significant improvement in spatial working memory that was associated with the activation of the ERK-CREB-BDNF pathway [62]. In addition, the effect of pure flavanol monomers (2)-epicatechin and (+)-catechin on spatial working memory in aged animals was paralleled by increased BDNF levels [91]. Altogether, previous results and ours are in accordance and strengthen the role of BDNF as a pivotal target pathway for the treatment of AD-related cognitive symptoms and emphasize a potential role of polyphenols as natural BDNF modulators in the central nervous system.