Synaptic Compensation as a Probable Cause of Prolonged Mild Cognitive Impairment in Alzheimer’s Disease: Implications from a Transgenic Mouse Model of the Disease

Animal and housing

The homozygous 3xTg-AD mice express two mutated human transgenes, APP_Swe and tau P301L mutations, and PS1 knock-in. Breeding pairs of these mice were obtained from Dr. Frank LaFerla through the Jackson Laboratory (New Harbor, ME, USA). Wild type (WT) age-matched controls from the same background strain, a hybrid: 129/Sv×C57BL/6, were also obtained from the Jackson Laboratory. Animals were housed and bred according to approved protocols from our Institutional Animal Care and Use Committee (IACUC), according to the PHS Policy on Human Care and Use of Laboratory animals (revised January, 2013). Animals were housed as 5 animals/cage with a 12:12 h light/dark cycle and with free access to food and water. They were given a period of acclimatization of 45 min to 1 h before any behavioral test was performed. Only young adult female 3xTg-AD mice and age-matched WT control mice were used in this study because it was shown previously that female 3xTg-AD have more aggressive and consistent pathology compared to males, and female mice have more Aβ deposition, worse cognitive performance, and a higher deficit in neurogenesis than males [33–35]. To exclude any effect of behavioral tests on neurogenesis, animals other than those used for behavioral studies were employed for immunohistochemical/biochemical studies. Twenty-six 3xTg-AD and 28 WT control mice were used for immunohistochemistry and western blots. For behavioral studies, 30 3xTg-AD and 20 WT animals were employed.

Behavioral procedures: Morris water maze task

The Morris water maze task is a test of reference memory, which is a measure of hippocampal function. The test was a modification of the original procedure by Morris et al. [36]. A total of 30 3xTg-AD mice and 20 age-matched WT controls were divided into two batches of 15 3xTg-AD and 10 WT in each batch. The first batch of animals was tested for reference memory by the Morris water maze task at 12–13 weeks and the second batch was tested at 13–14 weeks. The procedure was performed in a circular pool tank, 180 cm in diameter. A 13-cm escape platform was submerged 1 cm under opaque water in the northwest quadrant. The water was made opaque by the addition of a white non-toxic chalk to make the escape platform invisible to mice. Spatial cues were set up in the test room to help the animals locate the place of the platform. The temperature of the water was kept at 21±1°C. In each trial, the animal was given 90 s to find the escape platform. If the animal did not find the escape platform, it was gently guided to it. Each animal was left for 20 s on the escape platform and then returned to its cage. In each trial, the mouse was released with its back toward the center of the tank from a different quadrant and the starting point was changed each day so that the animal would not learn the place of the platform without using the spatial cues. The test was comprised of three acquisition trials for four consecutive days. Each animal performed 12 acquisition trials, which is considered a partial learning for the spatial reference memory task. Twenty-four hours after the last acquisition trial, the retention test or probe trial was conducted in which the escape platform was removed and the animal was allowed to swim freely for 60 s.

The measure of acquisition was the time and distance swum to reach the escape platform. The measures of retention were the percent of time spent in the target quadrant and the number of platform crossings. Behavior of the animal in the Morris water maze was recorded using a Samsung digital camera (SDC 4304) mounted on the ceiling and a SMART version 2.0.14 software (Pan Lab/San Diego Instruments) was used for tracking and timing of each trial.

Tissue processing

Right after completion of the Morris water maze task, six groups of animals (10-, 12-, 13-, 14-,15-, and 16-week-old mice) were chosen for immunohistochemical and biochemical studies. Animals were anesthetized using overdose of avertin and transcardially perfused using 0.1 M PBS. The brain was immediately removed from the skull, the right hemisphere was immersion fixed with 4% paraformaldehyde in 0.1 M PBS for at least 24 h at 4°C, then transferred to 30% sucrose; the left hemisphere was dissected into hippocampus, fore-, mid-, and hind-brains, and stored at –80°C for biochemical studies. After fixation and cryoprotection, the right hemisphere was sectioned into 40 um sagittal sections then saved at –20°C in antifreeze solution (ethylene glycol, glycerol, and 0.1 M PBS in 3:3:4 ratio) until further processing.

Immunohistochemistry

For immunohistochemistry, 3–5 animals/group from the six different groups were chosen randomly and 5 to 6 sections/mouse from every tenth section were used for staining intensity scanning analysis. Sections used for a specific marker from the six groups were immunostained and analyzed by confocal microscopy under identical conditions. The following antibodies were used: mouse anti-synaptophysin (1:500; Chemicon, Temecula, CA, USA); mouse monoclonal anti-SMI52 to MAP2 (1:1000; Covance, Emeryville, CA, USA); rabbit polyclonal anti-β-III tubulin (1:1000; Covance); rabbit polyclonal anti-GluR1 (1:300; Millipore, Temecula, CA, USA). The following secondary antibodies were used: Alexa 488-conjugated goat anti-mouse IgG antibody (1:500, Molecular Probes, Carlsbad, CA, USA) and CY3-conjugated goat anti-rabbit antibody (1:500, Jackson Laboratory, Bar Harbor, Maine, USA). Sections were double-stained either with anti-synaptophysin and anti-GluR1 or with anti-MAP2 and anti-βIII-tubulin.

For all the stainings, maximum projection images were generated employing the average of 15 Z-stacks using a 20X objective with a small pinhole of a Nikon 90i fluorescent microscope equipped with Nikon C1 three-laser confocal system and a Nikon DS U1 digital camera and the entire area of DG, CA1 (stratum radiatum and Stratum pyramidae), CA3 (stratum oriens, stratum radiatum and stratum pyramidae), parietal association, and frontal cortices (all six layers) were analyzed. Mean pixel intensity was measured for each brain region using the Image J software (NIH). Each picture to be analyzed was converted to grey scale and each area to be analyzed was selected, the threshold was adjusted and the mean pixel intensity of the region of interest was quantified.

Western blots

Western blot procedures were done using 10% SDS-PAGE as previously described [37, 38]. The blots were developed using mouse monoclonal anti-synaptophysin antibody (1:3,000; Chemicon, Temecula, CA, USA), the rabbit polyclonal anti-GluR1 antibody (1:1000; EMD Millipore, Billerica, MA, USA) and rabbit monoclonal anti-GAPDH (1:1,000; Santa Cruz Biotechnology, Santa Cruz, CA, USA) was used as a loading control. The corresponding horseradish peroxidase-conjugated affinity-purified goat anti-mouse IgG (H+L) and anti-rabbit IgG (H+L) secondary antibodies were used (1:5,000; Jackson Immuno Research Laboratories, Inc., West Grove, PA, USA). The blots were visualized using the enhanced chemiluminescence (ECL) reagents, (Pierce, Rockford, IL, USA). The ECL films of the blots were scanned and analyzed using Image J software (NH). Each protein’s band intensity was quantified and then analyzed using image J. Each protein’s band intensity quantification was the result of the subtraction of the background intensity from the original intensity of the band. Finally, the intensity of each band minus the background was normalized to the intensity of the GAPDH band minus the background.

Statistical analysis

Statistical analyses were conducted using Stata and GraphPad Prism software package, version 5.0 (GraphPad Software Inc., La Jolla, CA, USA). Tests of scalar predictors were done with regression models and were performed in Stata 13 (StataCorp, College Station, TX, USA). Statistical significance of the changes in indicator levels over age was performed in regression models. Instead of comparing each week with each other week without regard for the natural ordering of the ages (as an ANOVA would), the regression analyses determined the change in each indicator per week of age across all the ages at once. As the rate of change could vary across ages, we computed models including quadratic and cubic terms for age to see if the rate at which indicators varied could best be predicted by them. When these higher-order terms are included, the results show whether, for example, a model in which indicators rise, then fall, then rise with age best fits the data. We compared the adjusted R-squared for the models to determine which model best fit the data. Table 1 and Table 2 list, for the best-fitting models, the overall F score and degrees of freedom (as in an ANOVA), the p-value that shows the overall statistical significance of the model, the adjusted R-squared (which shows how much of the variance in the indicator over age is explained by the model, adjusted for the number of predictors in the model), and the type of model. A cubic model uses age, age-squared, and age-cubed to predict the indicator. A quadratic model uses age and age-squared, and a linear model uses only age itself. We compared the adjusted R-squared for each model to determine best fit. For multiple group comparisons, one-way or two-way ANOVAs followed by post-hoc tests were used. T-tests were used for all other inter-group comparisons. Data are presented as mean±SEM. For all analyses p < 0.05 was considered to be significant.

At 12-13 weeks, synaptic deficit was accompanied by impairment in spatial reference memory in 3xTg-AD mice

The analysis of the swim speed revealed no significant difference between the WT and the 3xTg-AD mice (Fig. 1a, p = 0.84). For this reason, the learning performance of the animals was analyzed as the latency time and the distance to reach the escape platform. We found that both 3xTg-AD and WT mice learned the task well, which was evident from the decrease in the latency to escape across training sessions (Fig. 1a, p < 0.0001). However, the regression of the latency to escape across time showed that the 3xTg-AD mice spent more time to find the escape platform compared to WT (Fig. 1a, p = 0.0085). The analysis of the distance traveled to find the escape platform showed that both groups learned the task well (Fig. 1a, p < 0.0001). However, during the training days, the 3xTg-AD mice had to travel longer distance to find the escape platform (Fig. 1a, p = 0.0261) compared to WT. These data revealed that at 12-13 weeks synaptic deficit was accompanied by impairment in spatial reference memory in 3xTg-AD mice.

To evaluate the memory retention in the 3xTg-AD and WT mice, a probe trial 24 h after the last training session was conducted. The 3xTg-AD mice spent significantly less time in the target quadrant (Fig. 1b; p = 0.0004) and visited the target zone less (Fig. 1b; p < 0.007) than WT animals. These results suggest that memory retention/retrieval was compromised in 3xTg-AD mice and further corroborated that these mice were cognitively impaired.

At 13-14 weeks, synaptic compensation was able to improve learning but not retention and memory in 3xTg-AD mice

The analysis of batch 2 (13-14 weeks) animals showed that there was no difference in the swim speed between the WT and 3xTg-AD mice (Fig. 1a, p = 0.95) and the latency to escape went from highly significant in batch 1 to only a trend in batch 2 (Fig. 1a, p = 0.06). Furthermore, the difference in the distance traveled between WT and 3xTg-AD changed from significant in batch 1 (Fig. 1a, p = 0.0261) to non-significant in batch 2 (Fig. 1a, p = 0.297). Both groups learned the task well since the latency to escape and distance decreased for both groups during the training days (Fig. 1a, p < 0.0001 and p < 0.001, respectively). These data show that synaptic compensation ameliorated learning impairment during the training days in the Morris water maze task.

To determine the effect of synaptic compensation on memory retention we analyzed the percent of time in target quadrant and the number of crossings of the previous platform place. We found that the 3xTg-AD mice spent significantly less time in the target quadrant compared to WT (Fig. 1b, p = 0.0002) while there was no difference between both groups in the number of crossings to the previous place of the platform (Fig. 1b, p < 0.238). This shows that memory retention stayed impaired in 3xTg-AD mice even with the presence of synaptic compensation.

When we pooled the two batches together aged 12–14 weeks, we found that the 3xTg-AD mice were impaired compared to WT in spatial reference memory. The analysis of the swim speed revealed no significant difference between the wild type and the 3xTg-AD mice (Supplementary Figure 1a). For this reason, the learning performance of the animals was analyzed as the latency time to reach the escape platform. We found that animals from the two different groups learned the task well as seen by a decrease in the latency to escape across training sessions (Supplementary Figure 1a, p < 0.0001). However, the regression of the latency to escape across time showed that the 3xTg-AD mice learned more slowly than the WT across the training days (Supplementary Figure 1b, p = 0.048).

The evaluation of the memory retention of the two batches combined in the 3xTg-AD and WT mice was conducted 24 h after the last training session. We found that the 3xTg-AD mice spent significantly less time in the target quadrant (Supplementary Figure 1c, p < 0.001) and visited less the previous place of the platform (Supplementary Figure 1d, p < 0.005) than WT animals. These results showed that the 3xTg-AD mice were impaired in memory retention/retrieval and that the 3xTg-AD mice were impaired in spatial reference memory.

We found enhanced motor learning coordination as reported previously [37, 40] and normal level of anxiety in the 10–14 weeks 3xTg-AD mice (Supplementary Figure 2a-f).

Impairment in spatial reference memory is accompanied by a transient increase in neuronal and synaptic markers in 3xTg-AD mice

Synaptic deficit is an established phenomenon and an early event in AD. Synaptophysin loss in the frontal cortex strongly correlates with dementia and cognitive impairment [15]. To examine synaptic changes at very early stages of the disease, we evaluated immunohistochemical staining of brains from 3xTg-AD mice aged 10, 12, 13, 14, 15, and 16 weeks using four different antibodies: anti-synaptophysin, anti-GluR1, anti-MAP2, and anti-β-III tubulin.

Cornu ammonis 1 (CA1)

Across time changes in mean pixel intensity showed that the synaptophysin expression in CA1 in 3xTg-AD mice from 10–16 weeks of age was best modeled by a cubic regression model. In this model, synaptophysin expression diminished at 12 weeks of age and was then compensated for from 13 to 16 weeks of age (Fig. 2a; Table 1). Similarly, western blots of synaptophysin expression were best fit by a cubic model, with a significant decrease in synaptophysin level in the hippocampus at 12 weeks compared to 10 weeks, followed by an increase by 13-14 weeks and then a decrease by 15-16 weeks in 3xTg-AD mice (Fig. 3a, b; F (3, 18) = 5.47, p = 0.0075).

The examination of GluR1 expression in the CA1 region showed no significant difference across time (Fig. 2b; Table 1). Similarly, western blots showed no change of the expression level in the hippocampus (Fig. 3c, p = 0.64; Supplementary Table 1).

The expression of the dendritic marker MAP2 was best fit by a linear model (Fig. 2c; Table 1).

The expression of β-III tubulin dropped at 12 weeks compared to 10 weeks and then increased from 13–16 weeks (Fig. 2d; Table 1), though again the best fit was obtained with a linear model.

Cornu ammonis 3 (CA3)

The expression of synaptophysin in the CA3 region was best fit by a cubic model (Supplementary Figure 3a; Table 1). As in the CA1 region, the expression of synaptophysin dropped at 12 weeks compared to 10 weeks and increased from 13 to 15 weeks of age. No significant change was detected in the GluR1 and MAP2 expressions in the CA3 region (Supplementary Figure 3b,c; Table 1). β-III tubulin was best modeled by a cubic regression with a decrease in the expression at 12 weeks compared to 10 weeks and increase from 13–16 weeks of age (Supplementary Figure 3d; Table 1).

Dentate gyrus (DG)

In the DG, the expression of synaptophysin dropped at 12 weeks of age compared to 10 weeks and then increased from 13 to 16 weeks. The changes followed a cubic model (Fig. 4a; Table 1). No change was found in the expression of GluR1 (Fig. 4b; Table 1). The expression of MAP2 followed a quadratic model (Fig. 4c, Table 1); β-III tubulin expression was best fit by a quadratic model with a drop at 12 weeks compared to 10 weeks and increase from 13–16 weeks (Fig. 4d; Table 1).

Parietal cortex

Expression of synaptophysin, MAP2, and β-III tubulin showed a drop in expression across time at 12 weeks compared to 10 weeks and an increase from 13–16 weeks. A cubic model best fit the expression of synaptophysin, while β-III tubulin and MAP2 followed a quadratic model (Supplementary Figure 4a, c, d; Table 1). The expression of GluR1 did not change significantly (Supplementary Figure 4b; Table 1).

Frontal cortex

The expressions of synaptophysin and β-III tubulin changed across time. Synaptophysin followed a cubic model (Supplementary Figure 5a; Table 1), while for β-III tubulin, a linear model fit best (Supplementary Figure 5d; Table 1). Both markers decreased at 12 weeks of age compared to 10 weeks and increased from 13–16 weeks. No significant change was detected in the expression of GluR1 (Supplementary Figure 5b; Table 1) or MAP2 (Supplementary Figure 5c; Table 1).

Cross-sectional comparison between 3xTg-AD and control mice confirms synaptic compensation in the former

To understand if the changes in the expression of the synaptic, dendritic, and neuronal markers are relevant to the poor cognitive performance in the 3xTg-AD mice we analyzed these data as a percent of control (WT) for different ages from the age of 10–16 weeks in different brain regions. Figure 2 and Figure 4 and Supplementary Figure 3–Figure 5 show the pattern of expression of synaptophysin, GluR1, MAP2, and β-III tubulin in the 3xTg-AD alone across time. However, the expression patterns of the four markers also fluctuate in the WT across time.

CA1

Cross-sectional comparison of 3xTg-AD as a percent of control (WT) mice revealed a synaptic deficit at 12 weeks followed by compensation in different brain regions at different ages. In the CA1 region, change in synaptophysin expression was not significant (Supplementary Figure 6a; Supplementary Table 2). Western blots showed changes in synaptophysin expression best fit by a quadratic model with synaptic deficit at 12 weeks compensated from 13 weeks (Fig. 3a-b; F (2, 7) = 12.56; p < 0.005).

GluR1 expression showed no difference between 3xTg-AD mice and WT for all time points studied (Supplementary Figure 6b; Supplementary Table 2).

Similarly, MAP2 expression was comparable in 3xTg-AD and WT mice (Supplementary Figure 6c; Supplementary Table 2).

Level of β-III tubulin did not change significantly but the lowest level was at 13 weeks of age with a slight compensation from 14–16 weeks of age (Supplementary Figure 6d; Supplementary Table 2).

CA3

Cross-sectional comparison of 3xTg-AD as a percent of control mice revealed that synaptophysin expression did not differ significantly between 3xTg-AD and WT (Supplementary Figure 7a; Supplementary Table 2).

GluR1 expression did not change significantly between 3xTg-AD and WT in all the time points studied (Supplementary Figure 7b; Supplementary Table 2).

MAP2 showed a trend toward decrease at 12 weeks, compensated for at 15 weeks, then dropped significantly at 16 weeks in 3xTg-AD compared to WT (Supplementary Figure 7c; Supplementary Table 2).

β-III tubulin expression in the CA3 did not change significantly between 3xTg-AD and WT in the different time points studied (Supplementary Figure 7d; Supplementary Table 2).

DG

Cross-sectional comparison of 3xTg-AD as a percent of control mice showed that synaptophysin expression in the DG did not differ significantly between WT and 3xTg-AD across time (Fig. 5a; Supplementary Table 2).

Overall, there was no significant difference in the expression of GluR1 between 3xTg-AD and WT (Fig. 5b; Supplementary Table 2).

MAP2 expression decreased at 12 weeks, increased from 13–15 weeks, and then dropped again at 16 weeks in 3xTg-AD mice compared to WT (Fig. 5c; Supplementary Table 2).

Overall β-III tubulin expression did not change significantly between 3xTg-AD and WT (Fig. 5d; Supplementary Table 2).

Parietal cortex region

Cross-sectional comparison of 3xTg-AD as a percent of control mice showed that synaptophysin expression decreased significantly at 12 weeks in 3xTg-AD compared to WT and increased from 13–16 weeks (Supplementary Figure 8a; Supplementary Table 2).

GluR1 expression did not change significantly between 3xTg-AD and WT (Supplementary Figure 8b; Supplementary Table 2).

Change in MAP2 expression was not significant (Supplementary Figure 8c, Supplementary Table 2).

Change in β-III tubulin expression was not significant (Supplementary Figure 8d. Supplementary Table 2).

Frontal cortex

Cross-sectional comparison of 3xTg-AD as a percent of control mice showed that synaptophysin decreased at 12 weeks and increased at 14 weeks. The compensation for synaptic deficit occurred from 13 to 15 weeks (Supplementary Figure 9a; Supplementary Table 2).

GluR1 expression showed no significant change (Supplementary Figure 9b; Supplementary Table 2).

MAP2 expression did not change significantly between 3xTg-AD mice and WT (Supplementary Figure 9c, Supplementary Table 2).

β-III tubulin expression in 3xTg-AD mice was lowest at 13 weeks and increased non-significantly at 14 weeks of age. Compensation for synaptic deficit happened mainly at 14 weeks (Supplementary Figure 9d; Supplementary Table 2).

Small changes in the expressions of neuronal and synaptic proteins in some brain regions in WT mice

Cross-sectional comparison between 3xTg-AD and control mice shows that synaptic compensation happens at different time points in different brain regions

A comparison of changes in the expression of different synaptic and neuronal markers described above in the 3xTg-AD mice as % of WT showed that synaptic compensation happened in different brain regions at different time points (Fig. 7, Supplementary Table 2). Indeed, for the synaptic marker synaptophysin, for example, synaptic compensation in the CA3 region happened at 14 weeks, while in the parietal and frontal cortices, compensation happened at 13 weeks. Supplementary Table 2 shows the significances of regional changes for synaptic and neuronal markers at different ages.