L-Norvaline Reverses Cognitive Decline and Synaptic Loss in a Murine Model of Alzheimer’s Disease

Homozygous 3×Tg mice harboring a mutant APP (KM670/671NL), which is a human mutant PS1 (M146V) knock-in, and tau (P301L) transgenes (B6;129-Psen1^tm1Mpm Tg(APPSwe, tauP301L)1Lfa/J) were obtained from Jackson Laboratory (Bar Harbor, ME) and bred in our animal facility. These mice exhibit a synaptic deficiency, with both plaque and tangle pathology [34]. Four-month-old homozygous 3×Tg mice and age-matched male C57BL/6 mice were used as nontransgenic controls (non-Tg) in all the experiments. C57BL/6 mice are commonly used as non-Tg controls of 3×Tg mice [35]. The average weights of the 3×Tg and non-Tg animals were marginally different at the beginning of the experiment. However, the difference was not statistically significantly different, and the difference was negligible at the end of the experiment.

Randomly chosen animals were divided into 4 groups (14–15 mice in each group). The animals were housed in cages and provided with water and food ad libitum. The dose administered was about 40–50 mg/kg/day, which was similar to that employed in a previous study in which L-norvaline was used to treat artificial metabolic syndrome in a rat model [32]. In the earlier study, the animals received norvaline dissolved in water at a dose of 50 mg/kg/day for 6 week via force gavage feeding. In the present study, L-norvaline (Sigma) was dissolved in the animals’ water (250 mg/L) supplied in the animals’ cages. For 3×Tg mice aged 5 months with an average weight of about 30 g (33 g by the end of the experiment), the dose was about 1.5 mg/day, taking into account that the average consumption of water by mice under laboratory conditions equals about 5–6 mL/day [36].

The mice underwent appropriate treatment. The experimental design is presented in Fig. 1. The weights of the animals were measured every week during the experiment, and no significant treatment-related effects on the weights were observed.

All animal housing and procedures were performed in compliance with the guidelines established by the Israeli Ministry of Health’s Council for Experimentation on Animals and with Bar-Ilan University guidelines for the use and care of laboratory animals in research. The experimental protocol was approved by the ethics committee for animal experiments of Bar-Ilan University (Permit Number: 82-10–2017).

After the behavioral tests, the mice were deeply anesthetized with sodium pentobarbital (60 mg/kg) administered intraperitoneally and decapitated. Their brains (4 from each group) were sliced (0.5 mm thick) immediately in a mouse brain slicer matrix. Sections between 1.7 and 2.2 mm posterior to bregma according to the atlas of Franklin and Paxinos were used for sampling [40]. The hippocampi were punched in the region of the dentate gyrus in each hemisphere using a 13-gauge microdissection needle, frozen, and stored at − 80 °C.

The assessment of “hit” proteins’ expression was performed using a Kinex KAM-1150E antibody microarray (Kinexus Bioinformatics) in accordance with the manufacturer’s instructions. The analyses were done with hippocampal lysates as described on Kinexus’ web page (www.​kinexus.​ca). Briefly, lysate protein from each sample (100 μg) was labeled covalently with a fluorescent dye combination. Free dye molecules were then removed via gel filtration. After blocking nonspecific binding sites on the array, an incubation chamber was mounted onto the microarray to permit the loading of the samples. After incubation, unbound proteins were washed away. Two 16-bit images from each array were then captured using a ScanArray Reader (Perkin-Elmer). An antibody array was performed in parallel on 4 different chips. The output of the array consisted of the average normalized net signals (i.e., the average of 4 normalized net signal values of each antibody on the microarray).

The standard deviation and percent standard deviation of 4 separate measurements of globally normalized signal intensity values for each different antibody on the microarray were calculated. The data are presented as change from control (CFC%). A positive value corresponded to an increase in signal intensity in response to the treatment, with a value of 100% corresponding to a 2-fold increment in signal intensity. A negative CFC value indicated the degree of reduction in signal intensity from that of the control.

Each parameter has its Student’s t test p value, which is the probability (p) value that there is no difference between the control and test samples. A p value determined with N = 4 measurements in each set, which were paired and 2-tailed in distribution. A p value of ≤ 0.05 was accepted as statistically significant.

For Golgi staining, 3 mice in each group were perfused transcardially with 0.1 M phosphate-buffered saline (pH 7.4), and their brains were processed using a superGolgi Kit according to the manufacturer’s protocol (Bioenno Lifesciences, Santa Ana, CA, USA). Briefly, the brains were immersed in impregnation solution for 11 days, followed by incubation for 2 days in a postimpregnation solution. Once the impregnation of neurons was complete, the brain samples were coronally sliced (150 μm) using a vibrating microtome (Campden Instruments, Lafayette, IN) and collected serially in a mounting buffer. They were then mounted upon 1% gelatin-coated glass slides, stained with a staining solution, and coverslipped using Permount (Fisher Scientific, Houston, TX, USA). Two corresponding sections between 1.7 and 2.0 mm posterior to bregma according to the atlas of Franklin and Paxinos per animal were chosen for stereological analysis [40].

An upright ApoTome (Quorum Technologies, Lewes, UK) microscope was used for imaging. To assess dendritic morphology, low magnification (× 10/0.3, × 20/0.8, and × 40/0.75 lens) images (Z-stack with 0.5 μm intervals) of pyramidal neurons, with cell bodies located in the CA1 region of the hippocampus, were captured using an ORCA-Flash4.0 V3 camera. High-magnification (× 100/1.4 oil objective with digital zoom 3) images (Z-stack with 0.25 μm intervals) were also obtained.

The captured images were coded, and an investigator blinded to the experimental conditions measured the numbers of spines per dendritic length using Zen Blue 2.5 (Zeiss, Oberkochen, Germany) and Neurolucida 360, version 2018.1.1 (MBF Bioscience, Williston, VT, USA). software, which provides an automatic unbiased quantitative 3D analysis of identified neurons [41]. The spine detection threshold was set an outer range of 2.5 μm, the minimal height was 0.3 μm, the sensitivity was 100%, and the minimum count was 10 voxels.

The spine density was analyzed in each hemisphere of each murine brain. Two hippocampal pyramidal cells, with soma located in the center of 150-μm corresponding sections, were selected for the analysis (24 neurons per experimental group). The spine density on a secondary oblique dendritic branch localized in the stratum radiatum of the CA1 hippocampal pyramidal neurons was also quantified.

Seven animals from each group were deeply anesthetized and transcardially perfused with 30 mL of phosphate-buffered saline (PBS), followed by 50 mL of chilled paraformaldehyde 4% in PBS. The brains were carefully removed and fixed in 4% paraformaldehyde for 24 h. The brains from 3 mice per group were then transferred to 30% sucrose in PBS at 4 °C for 24 h and frozen at − 80 °C. The brains from another 4 mice (including non-Tg mice) in each group were dehydrated and paraffin embedded.

The brains were sliced on a Leica (Wetzlar, Germany) CM3050 S cryostat to produce 30-μm floating sections. For β-amyloid, 1-16 (6E10) antibody staining, corresponding sections (between 1.7 and 2.2 mm posterior to bregma) were blocked and incubated overnight with primary purified anti-β-amyloid, 1-16 antibody 6E10 (1:100, Biolegend) at 4 °C. This was followed by washing and incubation with goat anti-mouse Immunoglobulin G secondary antibody Alexa555 (1:200, Invitrogen) at room temperature for 1 h and Hoechst (1:5000, Sigma) for 3 min.

The paraffin-embedded tissue blocks were chilled on ice and sliced on a Leica (Leica Biosystems, Nussloch, Germany) RM2235 manual rotary microtome at a thickness of 4 μm. The sections were mounted onto gelatin-coated slides, dried overnight at room temperature, and stored at 4 °C inside storage boxes.

For the quantitative histochemical analysis of microglia (Iba1), astrocytes (GFAP), ARGI, and ARGII immunopositivity, 5 brain coronal sections per mouse (1.9–2.0 mm posterior to bregma) were used. Five serial sections at a thickness of 4 μm were cut at 20-μm intervals throughout the brain. Immunohistochemistry (IHC) was carried out on the plane-matched coronal sections.

Staining was performed on a fully automated Leica Bond III system (Leica Biosystems Newcastle Ltd., UK). The tissues were pretreated with an epitope-retrieval solution (Leica Biosystems Newcastle Ltd., UK), followed by incubation with primary antibodies for 30 min. The Iba1 antibody (Novus Biologicals, #NB100-1028), GFAP (Biolegend, #835301), ARGI (GeneTex, GTX113131), and ARGII (GeneTex, #GTX104036) dilutions were 1:500, 1:1000, 1:500, and 1:60, respectively. A Leica Refine-HRP kit (Leica Biosystems Newcastle Ltd., UK) was used for detection and counterstaining with hematoxylin. The omission of Iba1, GFAP, ARGI, and ARGII primary antibodies served as a negative control (Supplementary Fig. S1). For a positive control of ARGI, murine liver tissue and kidney were stained for ARGII.

The sections were mounted and viewed under an Axio Scan.Z1 (Zeiss, Oberkochen, Germany) fluorescent and bright-field slide scanner with a × 40/0.95 objective. Images were taken with a Z-stack of 0.5 μm. Also, an Axio Imager 2 Upright ApoTome microscope was used to capture images with a × 100/1.4 oil immersion objective. Immunolabeling was performed in the corresponding hippocampal areas, and image analysis was carried out using Zen Blue 2.5 (Zeiss, Oberkochen, Germany) and Image-Pro® 10 (Media Cybernetics, Rockville, MD) software with a fixed background intensity threshold for all sections representing a single type of staining.

Resting microglia exhibit a ramified phenotype and adopt different morphologies, ranging from a highly ramified to an amoeboid-like phenotype upon activation [42]. To quantify morphological changes in resting microglia, we used morphological parameters that are accepted in the literature and expected to capture the shift from resting to activated phenotype [43]. The parameters included the cell perimeter length, Feret’s diameter, soma size, and sphericity (4π × area/perimeter²). To assess the staining density, the integrated optical density (IOD), which was defined as the mean density per pixel area (lum/pix²), was calculated [44]. The IOD values of the Iba1-stained microglia from CA3 regions with a surface area of 0.1 mm² were calculated as log₁₀ of the values acquired using Image-pro® 10. Two appropriate sections (bregma − 1.8 and − 2.0 mm) from each mouse were included in the analyses (N = 125 cells in the untreated groups; N = 90 cells in the L-norvaline-treated groups).

To determine the level of β-amyloidosis, the immunoreactivity of the A11 antibody and anti-amyloid fibril OC antibody in hippocampal lysates was examined. The A11 antibody detects the conformation of amyloid oligomers, irrespective of their amino acid sequence [45], and the anti-amyloid fibril OC antibody recognizes fibrillary forms.

In addition, the relative differences in the expression levels of various neuroplasticity-related proteins that showed significant changes in the antibody microarray assay were analyzed and validated. The setup and the antibody nomenclature are presented in Supplementary Table 1.

The protein concentration was determined using the Bradford assay. Hippocampal lysates (15 μg) obtained from brain samples of the 3×Tg mice treated with L-norvaline or vehicle (4 tissue punches from each group) were analyzed using a Kinetworks™ Custom Multi-Antibody screen 1.0 (Kinexus Bioinformatics, Vancouver, Canada) in accordance with the instructions of the manufacturer. Briefly, the Kinetworks™ analysis involves resolution of a single lysate sample by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and subsequent immunoblotting using validated antibodies. Antibodies bound to their target antigen on the nitrocellulose membrane were detected using an enhanced chemiluminescence detection system.

Statistical analyses were conducted using SPSS version 22 (IBM, Armonk, NY, USA) for Windows. The significance was set at 95% of confidence. All the results are presented as mean with standard error. The Shapiro–Wilk test showed that the data fit a normal distribution, and Levene’s test was performed to confirm equal variance between the groups being compared.

For the Y-maze, visible platform test, and probe trials (MWM), the means were compared between groups using a one-way analysis of variance (ANOVA), with Tukey’s multiple comparison test used for post hoc analyses. For the hidden platform test, the escape latencies in the 2 trials were averaged for each mouse each day and then analyzed across the 5 days of testing [37]. An ANOVA was applied, with day as the repeated measure and latency as the dependent variable. For comparisons between groups, Tukey’s multiple comparison test was conducted when the main analysis revealed significance. The Student t test was performed to compare the means between 2 groups. All data are presented as mean values. Throughout the text and in bar plots, the variability is indicated by the standard error of the mean (SEM). In the figures with box and whisker graphs, the boxes extend from the 25th to 75th percentiles. The line in the middle of the box is plotted at the median. The whiskers denote the smallest and largest values.

To investigate whether L-norvaline treatment affected learning and memory in AD pathogenesis, a set of behavioral tests was administered following L-norvaline treatment for 2 months, beginning when the mice were aged 4 months.

Short-term working memory was assessed in a Y-maze spontaneous alternation test. The results of the one-way ANOVA test revealed a significant treatment-related effect on the “percentage of alternations” (F(3,54) = 4.525, p = 0.0067). The post hoc Tukey multiple comparison test indicated that the percentage of alterations in the control 3×Tg mice was lower than that in the L-norvaline-treated mice and wild-type (WT) controls (p < 0.05) (Fig. 2a). The L-norvaline treatment had no significant effect on the alternative behavior of the non-Tg mice. The treatment also did not affect the total distance traveled by the animals during the test (p = 0.86) (Fig. 2b).

The MWM was used to determine the effect of L-norvaline upon spatial memory. In the visible platform test, there were no significant between-group differences (p = 0.99), which indicated that the treatment had no influence on murine motility or vision (Fig. 2c).

In the hidden platform swimming test, there were significant differences between the groups according to the repeated measures ANOVA (F(3,12) = 7.725, p = 0.0039). Tukey’s multiple comparison test revealed a significant (p < 0.01) effect of the treatment on memory acquisition in the 3×Tg group (Fig. 2d).

In the probe trial on the last day of testing, the platform was removed. The relative time spent by the mice in the target quadrant, where the hidden platform had been previously placed, was significantly longer in the L-norvaline-treated group as compared with that of the control group (34.8 ± 8.06 s vs 24.5 ± 7.88 s; p < 0.05; Fig. 2e). To visualize spatial preferences in the probe test, a heat map was used (Fig. 2g). The plots demonstrated that the treated 3×Tg animals spent more time than the control mice did in the target quadrant. The treatment did not affect the swim speed of the animals (Fig. 2f).

Overall, the results of the behavioral experiments suggested that L-norvaline improved spatial memory acquisition in 3×Tg mice.

Recent evidence pointed to a role for soluble amyloid oligomeric and fibrillar species in synaptic dysfunction, neuronal apoptosis, and brain damage [46]. To investigate the impact of L-norvaline on the amount of toxic prefibrillar Aβ oligomers and fibrillary forms, we examined A11 and OC immunoreactivity of hippocampal lysates. Pooled equal amounts (10 μg) of total lysates from the hippocampi of each group of mice were run on sodium dodecyl sulfate–polyacrylamide gels and incubated with appropriate antibodies (Supplementary Table 1 and Supplementary Fig. S2). The L-norvaline treatment resulted in a substantial (about 30%) reduction in the levels of A11-reactive oligomers and OC-reactive fibrillar species (Table 1).

To examine the effect of the L-norvaline treatment on the total amyloid burden in the brains of the 3×Tg mice, coronal brain sections were stained with human APP/Aβ-specific antibody. By the time they were 6 months old, the 3×Tg mice exhibited enhanced intracellular deposition of Aβ in the IV–V layers of the prefrontal cortex (PFC) (Fig. 3a–c) and hippocampus (Fig. 3d–f).

We did not detect significant differences in the level of Aβ deposition in the hippocampi of the 2 groups. However, there was a general trend toward a slight reduction in 6E10 positivity in the treated group. There was a significant (p = 0.032) reduction in the level of 6E10 positivity in the cortex of the 3×Tg mice treated with L-norvaline (Fig. 3g). These data pointed to an effect of L-norvaline on the level of β-amyloidosis in the 3×Tg mice.

Dendritic spines are specialized structures on neuronal processes, which are involved in neuronal plasticity [47]. A strong correlation between dendritic spine density and memory acquisition was demonstrated in rodents using different behavioral paradigms [48]. Furthermore, 3×Tg mice showed progressive dendritic spine deficiency as compared with WT mice [49], and this deficiency can be reversed by the treatments [50].

In the present study, dendritic spines were examined by Golgi staining in the vehicle- and L-norvaline-treated groups. Dendritic spine density was quantified on a secondary oblique dendritic branch localized in the stratum radiatum of the CA1 hippocampal pyramidal neurons (Fig. 4a, c, d). The results of the one-way ANOVA test revealed a significant effect of the treatment on dendritic spine density (F(3,92) = 3.338, p = 0.0173). The post hoc Tukey multiple comparison test pointed to a significantly lower spine density in the control 3×Tg mice as compared with that in the L-norvaline-treated mice (p < 0.05) (Fig. 4b). The average increase following the treatment was about 20% (1.326 ± 0.059 vs 1.10767 ± 0.06 spines per micrometer). The treatment did not affect dendritic spine density in the non-Tg groups.

To gain additional molecular insight into the behavioral phenotype, the effect of the L-norvaline treatment on the levels of neuroplasticity-associated proteins in the 3×Tg mice was investigated using a set of proteomics assays. First, we tested the protein levels of 12 known important neuroplasticity-related proteins in the hippocampus by a Western blot analysis (Supplementary Table 1). Second, we performed an antibody array analysis to detect changes in protein levels at the whole throughput level (approximately 1000 proteins). The Western blot analysis revealed an increase in synaptic- and neuroplasticity-related proteins in the L-norvaline-treated group (Fig. 5a). Remarkably, the expression of vesicular glutamate transporters 1 and 3 in the brain increased by 458% and 349% respectively, potentially increasing vesicular glutamate levels.

The protein microarray analysis revealed elevated levels of other neuroplasticity-related proteins following the L-norvaline treatment. The expression level of postsynaptic protein neuroligin-1, which mediates the formation and maintenance of synapses, was amplified by 252% following the treatment. However, the p value of the change detected by the antibody microarray was 0.07. Furthermore, the levels of synaptophysin, which is a synaptic vesicle glycoprotein, increased by 50% in the hippocampi of the 3×Tg mice treated with L-norvaline, and this finding was statistically significant (p = 0.039).

Notably, the expression of endothelial NOS increased by 68% (p = 0.041). Moreover, there was a substantial significant (p = 0.015) increase in cyclin E protein, which is a central component of the cell cycle machinery and known to regulate synaptic plasticity and memory formation [51].

Microglia are the first line of active immune defense in the brain, and an increase in microglial density is indicative of elevated pathogenic insults [52]. Previous studies reported increased microglial density in the hippocampi of AD patients [53] and hippocampi of 3×Tg mice [54]. Microglia densities in the CA3 area (0.2 mm²) of the 3×Tg (Fig. 6a, b) and non-Tg (Fig. 6c, d) mice were analyzed using image-processing software and a manually set threshold. To guarantee the selection of only cells present in the acquired field, for Iba1 staining, only cells with an area greater than 25 μm² were considered in the analysis, as widely accepted in the literature [55].

The results of the one-way ANOVA test revealed a significant effect of the treatment on microglial density and the “Iba1-positive area fraction” (F(3,52) = 4.64, p = 0.006 and F(3,52) = 4.098, p = 0.01, respectively). The post hoc Tukey multiple comparison test indicated a significant reduction (p < 0.05) in microglial density in the area of interest in the L-norvaline-treated 3×Tg group (63.84 ± 7.43 vs 83.66 ± 5.11 cells/mm², on average) (Fig. 6e) and a significant (p < 0.05) decrease in the Iba1-positive area fraction (0.48 ± 0.04 vs 0.67 ± 0.05%) (Fig. 6f).

Microglial morphology and function are closely related. Microglia are sensitive to the surrounding microenvironment, and various stimuli regulate their morphology [56]. Neuroinflammation is characterized not only by an increase in microglial density but by diverse morphological changes, including de-ramification [57]. To address this issue, several morphological features of stained microglia were analyzed using image analysis software Image-Pro® 10 (Media Cybernetics, Rockville, USA) and ZEN 2.5 (Zeiss, Oberkochen, Germany). Iba1 is known to be expressed weakly by ramified microglia but upregulated and strongly expressed by activated microglia [58]. We used the IOD index to characterize Iba1 staining intensity [59]. The results revealed a significant treatment-related effect on the microglial IOD (Fig. 7a). In the treated group, this parameter was reduced by 30%.

A detailed analysis of microglial somata revealed several other treatment-related effects. First, the average size (surface area) of the measured cells in the treated group contracted by 17% (50.92 ± 1.69 μm² in the treatment group vs 60.93 ± 1.06 μm² in the untreated group; p < 0.0001) (Fig. 7c). Second, the roundness or sphericity of the somata (4π area/perimeter²), which reflects microglial reactivity, increased significantly in the L-norvaline-treated group (0.763 ± 0.012 in the treatment group vs 0.71 ± 0.005 in the untreated group; p < 0.0037) (Fig. 7d). Additionally, we measured the Feret diameter or max–min caliper. The analysis revealed a significant reduction in Feret’s diameter in the treated group (6.82 ± 0.182 μm vs 8.61 ± 0.26 μm; p < 0.0001) (Fig. 7b).

Furthermore, the distribution analyses revealed a left-shifted distribution of soma size and right-shifted distribution of circularity in the treated group (Fig. 7e, f). Together, these findings suggested that the number of microglia with a small, round cell body (resting microglia) increased after the treatment. The cells in the control group had a bigger and more irregular soma shape (activated microglia) as compared with those in the treated group. The observations herein were in accord with the common classification of microglia morphology, in which resting cells are considered small, round cells with elaborate ramifications, and activated cells are considered amoeboid, with retracted processes [60].

Astrocytes modulate and control synaptic activity [61]. GFAP is highly expressed in astrocytes [62]. We stained the brains and observed GFAP immunoreactive cells with a typical stellate shape (Fig. 8e, f). Of note, the shape of the astrocytes in brain sections from the 3×Tg mice differed from those in non-Tg animals, with the astrocytes of the non-TG mice displaying a highly ramified star-shaped configuration (Supplementary Fig. S4). The astroglia in brain sections from the control 3×Tg mice showed fewer processes and a reduced volume (Fig. 8a, c) as compared with astroglia in brain sections from the animals treated with L-norvaline (Fig. 8b, d). We quantified the levels of GFAP immunoreactivity (reflected by the volume density) within the hilus area. The results of a comparative analysis using a one-way ANOVA (F(3,51) = 9.263, p < 0.0001), followed by Tukey’s multiple comparison tests between the experimental groups, revealed significantly increased (p < 0.001) glial somatic volumes in the 3×Tg animals treated with L-norvaline (Fig. 8g). The analysis did not reveal any significant changes in the number and surface area of GFAP-positive objects in the non-Tg mice following the treatment.

In addition, we measured the astrocyte density within the hilus. According to the literature, the average surface area of an astrocyte is about 250 μm² [63]. To count the cells, we filtered out all GFAP-positive objects less than 70 and more than 500 μm². The results revealed no significant differences (p = 0.55) in the density of GFAP-positive astrocytes between the control and experimental groups (Fig. 8h).

We examined ARGI protein spatial expression within the hippocampi of the 3×Tg mice and its spatial relationship with Aβ deposition using immunocytochemistry (Fig. 9a, b). ARGI preferentially accumulated in the CA1–CA4 areas of the hippocampus (Fig. 9c–f) and was mainly detected inside the cells (Fig. 9g, h). The L-norvaline treatment led to a significant (p = 0.0073) reduction (more than 2-fold) in the overall ARGI-positive cell surface area (Fig. 9i) and IOD (p = 0.0004) (Fig. 9j), which reflected the decrease in the levels of the ARGI protein in the brains of the treated animals.

In the non-Tg mice, hippocampal pyramidal neurons expressed a significantly lower level of the ARGI protein as compared with that in hippocampal pyramidal neurons of the 3×Tg mice (Supplementary Fig. S5). No differences in the level of immunopositivity were detected in the treated and control groups using the same software and the same threshold. Of note, several ARGI-positive star-shaped cells with dark nuclei were seen in the hippocampus (Fig. S5 insets).

We detected augmented ARGII immunopositivity of cells in the CA2 hippocampal area of the 3×Tg mice (Fig. 10a–d) but not in that of the non-Tg animals (Supplementary Fig. S6). The same pattern of staining was observed in the treated and control groups. At × 40 and higher magnification, mitochondria were observed, most clearly in the dentate gyrus, without significant differences in the numbers and intensity between the groups (Fig. 10e, f). We did not detect any significant effect of the treatment on the “ARGII-positive objects surface area” (Fig. 10g). However, the IOD index was significantly reduced (p < 0.05) in the L-norvaline-treated 3×Tg mice (Fig. 10h).