We used 23 wild-type (WT) and 35 homozygous transgenic B6;129 male and female 3×Tg mice. 3×Tg mice express three mutations associated with familial AD: APP (amyloid precursor protein) KM670/671NL (Swedish), MAPT (microtubule-associated protein Tau) P301L, and PSEN1 (presenilin-1) M146V (Oddo et al., 2003). All experiments were conducted in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals.

This injection method was previously used for intra-arterial delivery of cell-therapies in a mouse stroke model without ischemic injury or alterations in cerebral blood flow (Silasi et al., 2015). Mice at about 16 months old were anesthetized with 5.0% isoflurane and maintained on 1.5% isoflurane in 70% N₂O and 30% O₂ using a capillaries-animal anesthesia system. The local anesthetic bupivacaine (2 mg/kg) was injected subcutaneously at the incision site. The surgical site was shaved with hair removal lotion (Nair^®, Church and Dwight, Ewing, NJ) and cleaned with 70% ethanol and povidone iodine (Betadine^®, Avrio Health, Stamford, CT). Before surgery, the depth of anesthesia was assessed by toe pinch. The right common carotid artery was exposed through a neck incision. The external carotid artery and the pterygopalantine arteries were tied with suture. About 2000 Fluoresbrite Yellow Green microspheres at ≈20.0 um caliber (Polysciences, Warrington, PA) suspended in a 100 μL of phosphate buffer saline (PBS) were injected into the common carotid artery over approximately 1 min with a 33G needle. Microspheres traveled towards the right internal carotid artery, the circle of Willis, and then entered the right and left brain. After removing the needle, bleeding was controlled by absorbable suture knot. Unilateral common carotid artery occlusion can reduce blood flow to the brain, similar to aging conditions (Thong-Asa and Tilokskulchai, 2014).

About 18 h after surgery, mice were peritoneally injected with methyl ester of 8-[2-(2-pentylcyclopropylmethyl)-cyclopropyl]-octanoic acid (DCPLA-ME, MedChemExpress, Monmouth Junction, NJ) at 3 mg/kg body weight in sterile normal saline (3 times/week). Non-treated groups received the same vehicle volumes, mechanism of delivery, and frequency of administration as the treated groups.

Eight days after surgery, mice were moved to the test room in their home cages at least 1 h before daily trials. The maze pool had a diameter of 114 cm and height of 60 cm and was filled with 40 cm H₂O (22 ± 1°C) mixed with 200 mL of non-toxic white Tempera (BesTemp, Certified Color Corp., Orange, CA). The maze was divided into four quadrants. Mice were trained for 5 days (3 trials/day) to find a hidden platform (9 cm diameter) centered in one of the quadrants and submerged about 2 cm below the water surface. At the start of all trials, mice were placed individually in the water facing the maze wall, using different starting positions for each trial, and allowed to swim until they found the platform, where they remained for 20 s before being returned to their home cages. A mouse that failed to find the platform within 1.5 min was guided there by the investigator, with 90 s scored. The swim path was recorded with a video-tracking system, which computed latency to the platform, swim distance, and percentage of time spent in the quadrants. At 24 h after the training trials, a probe trial (a quadrant test or retention trial) was given with the platform removed to assess memory retention for its location by the distance the mouse moved in the quadrants. The video-tracking system tracked the animal’s movements in each quadrant for 1 min.

Mice were deeply anesthetized with intraperitoneal injection of 100 mg/kg body weight ketamine and 10 mg/kg xylazine. Animals were perfused through the heart with cold PBS for less than 4 min to wash out the blood and subsequently with 4% paraformaldehyde in PBS. Brains were then removed, postfixed for 20 min, and stored in PBS at 4°C. The dorsal hippocampi were sectioned with a cryostat, and 4 hippocampal sections (30 μm thickness) were selected every 400 μm for each hippocampus.

The samples were treated with Image-iT FX signal enhancer (Thermo Fisher Scientific, Grand Island, NY, United States) for 30 min at room temperature and then with 5% normal goat serum and 0.5% Triton X-100 in PBS for 50 min to block non-specific protein binding sites. Primary antibodies were: 8-OHdG (mouse monoclonal antibody; 1:100; Genox, Shizuoka, Japan, cat # N45.1); cleaved caspase 3 (rabbit polyclonal IgG; 1:100; Cell Signaling Tech, Danvers, MA, cat # 96645); MnSOD (rabbit; 1:400; MilliporeSigma, Burlington, MA, cat # 06-984); ERK1/2 (mouse; 1:500; Invitrogen, Waltham, MA, cat # 13-6200); HIF-1α (mouse; 1:250; R&D Systems/Bio-Techne, Minneapolis, MN, cat # MAB19351); TFAM (rabbit polyclonal IgG; 1:500; Invitrogen, cat # MA5-35365); VEGF (mouse monoclonal IgG; 1:50; Santa Cruz Biotechnology, cat # sc-7269); PKCε (rabbit; 1:500; MilliporeSigma, cat # 06-991); glial fibrillary acidic protein (GFAP) (rabbit; 1:1,000; Thermo Fisher Scientific, cat # J64334); neurogranin (rabbit; 1:500; MilliporeSigma, cat # AB5620); synaptophysin (mouse; 1:500; MilliporeSigma, cat # MAB5258-I); and myelin basic protein (rabbit, 1:100, Protein Tech, Rosemont, IL, cat # 10458-1-AP) at 4°C for 24 h. Tissue sections were switched to a new incubation solution and washed with PBS (3 times, 5 min each). The samples were then incubated with Alexa Fluor 488 anti-mouse IgG (1:1,000; Thermo Fisher Scientific, cat # A32731) or Alexa Fluor 568 donkey anti-rabbit IgG (1:1,000; Thermo Fisher Scientific, cat # A10042) for 3 h at room temperature. Vascular endothelia were stained at room temperature for 3 h with the DyLight fluor 594-conjugated Lycopersicon esculentum (tomato) lectin (Vector Laboratories, Burlingame, CA, 1:50). The sections were mounted using Prolong glass antifade mountant with NucBlue stain (Invitrogen) to counter stain nuclei.

The images were oriented with a Zeiss Axio Observer Z1 microscope equipped with a 710 confocal scanning system using the 10× objective lens in the DAPI channel (for staining DNA in nuclei). The random area that appeared immediately after switching to the higher magnification lens, 63X Plan-APO Chromat oil immersion objectives (1.4 NA), was imaged for appropriate fluorescence. Confocal images were acquired in line scan mode with a pinhole of approximately 1.00 Airy unit, and averaged data from several (4×) images were reported. Using range (red/blue) indicator in the software, the maximal gray level was set at the under-saturation of each fluorescence channel, except the vascular endothelial cell marker tomato lectin was imaged at saturated intensity because it was not used for quantification. Images were obtained and quantified with the NIH ImageJ program. Under the ImagJ program, the pixel gray levels on the original confocal images were not changed during the adjustment of confocal images on the computer monitor.

Due to thick tissue samples, the insufficiency of passive penetration of antibodies and dyes did not generate uniformly deep staining. Quantification of fluorescence intensities of targeted proteins or nucleoside were therefore normalized with DAPI fluorescence (DNA). Most control data were set at 100%, and other experimental data were defined as percentage of their controls. In the illustrated figures, the brightness and/or contrast of DAPI and other fluorescence channels in an individual image were manually adjusted until DAPI fluorescence among experimental groups reached the same levels, and, presumably, DNA was stable in all nuclei. Because the blood vessel marker tomato lectin was not used for quantification and imaged at saturated intensity, in the figure panels, tomato lectin fluorescence was adjusted separate from adjustments made to DAPI.

After immunohistochemical processing, we measured the densities of presynaptic axonal terminals (synaptophysin grains) and postsynaptic membranes (neurogranin grains) per 33.7 × 33.7 × 0.6 μm³ volume. Images were analyzed by particle counting using the NIH ImageJ program commands. The 8-bit gray scale images were inverted so that the dark pixels became light and vice versa. The background of the photographic negative was subtracted with a rolling ball radius at least the size of the largest object that was not part of the background. Using the threshold method, the images were converted from an 8-bit (256 shades of gray) to a 1-bit (black or white) to define an individual pixel as background or particle component. The above procedures were repeated with a different rolling ball radius and threshold method until the particles were the same sizes of synaptophysin or neurogranin particles seen in the original confocal images. Only particles within the range of particle sizes of the measured structures were defined and counted. In addition, synaptopysin intensity was analyzed from the original images (33.7 × 33.7 μm²) without the above adjustment, used for grain counting.

Protein extraction was adapted from Thacker et al. (2021). Fixed-tissue samples were sonicated in 500 μL of modified fixed tissue lysis buffer [containing: 100 mM NaCl, 25 mM EDTA, 500 mM TRIS-HCl, 1% (v/v) Triton X-100, 1% (v/v) IGEPAL CA-630 (NP40), 2% (w/v) SDS, and protease inhibitor cocktail (Thermo Fisher Scientific)]. Homogenates were incubated at 90°C for 120 min under gentle agitation (300 rpm on a shaker) followed by centrifugation (1,000 × g) at 4°C. Supernatants were collected, and protein concentrations were determined with a Bio-Rad DC protein assay kit and aliquots stored at −80°C. Proteins were separated on Nu-Page 4%–12% Bis-Tris polyacrylamide gels (Invitrogen). Using an iBlot2 (Thermo Fisher Scientific), proteins were transferred to nitrocellulose membranes. Membranes were incubated with Odyssey blocking buffer (LI-COR) for 1 h at room temperature. Membranes were then treated with primary antibodies against MnSOD (rabbit polyclonal IgG; 1:4,000; MilliporeSigma, cat # 06-984); β-Actin (rabbit polyclonal IgG; 1:25,000; Novus Biological, cat # NBP2-76367); GFAP (rabbit; 1:1,000; Thermo Fisher Scientific, cat # J64334); myelin basic protein (rabbit, 1:2,000, Protein Tech, Rosemont, IL, cat#10458-1-AP); and neurogranin (rabbit; 1:500; MilliporeSigma, cat#AB5620). Blots were then incubated with DyLight 680 anti-mouse antibody (1:1,000, Invitrogen, cat # 35518) and DyLight 800 anti-rabbit antibody (1:1,000, Cell Signaling Technology, cat # 5151). We used a LI-COR imaging system to image the blots. The densitometric value for the protein were quantified with the NIH ImageJ. The target protein normalized with β-actin was used for analysis.

For behavioral and morphological studies, data were analyzed with analysis of variance (ANOVA). Data with a significant overall difference among the groups as demonstrated with an ANOVA analysis were further analyzed for Tukey’s multiple comparison or t-test. For western blots, only t-test was performed. The confidence level was 95% (α = 0.05).

Mice were injected into right common carotid artery after neck surgery with about 2,000 yellow green microbeads at ≈20.0 um diameter (Silasi et al., 2015). Microbeads traveled through the Circle of Willis and entered both the left and right brains. The microbeads at ≈20.0 um diameter impeded local microcirculation through terminal arterioles (>6 μm in diameter; Figure 1B) (McDowell et al., 2021). Microbeads were small enough to pass cerebral arteries throughout the whole brain (Figure 1A). This resulted in microinfarct (artificial cerebrovascular disease) (Figures 1C,D).

We recently demonstrated that the PKCε activators can promote VEGF and MnSOD expression and prevent MV loss and/or spatial memory defect in aged rats and Tg2567 mouse model of AD (Millien et al., 2022). Tg2576 mice overexpressed only Amyloid beta (Aβ) peptide (Westerman et al., 2002). We then aimed to further study therapeutic effect of the PKCε activator DCPLA-ME in 3×Tg mice that expressed both amyloid plaques and neurofibrillary tangles (Oddo et al., 2003) with and without cerebral microinfarcts. After microbead injection, mice were treated with or without the PKCε activator DCPLA-ME. For spatial learning and memory studies, mice were trained to find a submerged platform in a water maze pool (Figure 1E). The training regimen (3 swims per day) was intentionally made more difficult to better reveal cognitive deficits. At 24 h after 5 days of water maze training and learning evaluation, memory retention was assessed with a probe test that allowed mice to find the target area of the removed platform. There were significant differences among all experimental groups for 5 days swim learning (F_4,449 = 8.763, p = 0.001, ANOVA) and memory (F_4,46 = 3.379, p = 0.017). Compared with WT mice, 3×Tg mice showed a significant impairment in learning (an increase in latency time to reach platform, p = 0.002) and memory (to concentrate on the removed platform area, p = 0.013) (Figures 1F,G). MI enhanced the learning impairment (p = 0.040) and memory defect (p = 0.047, Figures 1F,G).

At 24 h after the probe test, morphological change in capillaries and arterioles were further studied in situ. Mice were fixed with formaldehyde, and brains were used for histochemical and microscopic studies (Figure 2A). Vascular endothelial cells were stained with tomato lectin and imaged with confocal microscope (Figure 2B). We quantified the change in capillaries (<6 μm in diameter) and arterioles (>6 μm in diameter) (McDowell et al., 2021). ANOVA revealed significant differences among animal groups for the capillary number (F_4,294 = 3.321, p = 0.011) and perivascular space (F_4,282 = 5.213, p = 0.001) as well as for the arteriolar diameter (F_4,289 = 2.547, p = 0.040) and endothelial cell density (F_4,153 = 3.120, p = 0.017) (Figures 2C–F). For the post-hoc test, Tukey multiple comparison was performed.

In WT mice, MI did not affect the number of capillaries and arterioles per 200 × 200 μm² hippocampal area, diameter, endothelial cell number/100 μm perimeter, or the ratio of perivascular space divided by MV area of capillaries and arterioles (Figures 2C–F, compared WT + MI to WT). In 3×Tg mice, the number of capillaries, but not arterioles, decreased (p = 0.015) (Figure 2C, Tg vs. WT). In 3×Tg + MI mice, the density of capillaries was not different from WT and 3×Tg mice (Figure 2C). This suggests that MI increases angiogenesis in 3×Tg mice. The data showed that MV diameters in the hippocampal CA1 areas were smaller than the microbeads (20 μm in diameter). Most capillaries were in the range of 3–5 μm in diameter. Arterioles were mostly 7–10 μm; less than 5% were 10–20 μm in diameter. MI increased the diameter of arterioles (p = 0.037) and the number of their nuclei (p = 0.030) (Figures 2D,E; Tg + MI vs. Tg). The increased nuclei were the large oval, round, and spindle shape of endothelial cells rather than small and thin nuclei of pericytes and smooth muscle cells (Figure 2B). MI increased (p = 0.001) the ratio of perivascular space divided with its MV area for capillaries but not arterioles (Figure 2F).

DCPLA-ME treatment protected the learning (p = 0.002) and memory defect (p = 0.039) in 3×Tg mice with MI (Figures 1F,G; Tg + MI + DCP vs. Tg + MI). DCPLA-ME also prevented an increase in the diameter or wall cells of arterioles (p = 0.025) and the pericapillary space dilation (p = 0.002) in 3×Tg mice with MI (Figures 2D–F). The results indicate that DCPLA-ME prevents the effect of cerebrovascular MI.

It is well known that the targeted cell signals in the present study are not specially expressed in capillaries and arterioles. VEGF is also expressed in neurons, astrocytes, microglia etc., (Rosenstein et al., 2010; Argaw et al., 2012; Okabe et al., 2020). Therefore, double immunohistochemistry at an individual capillary or arteriole was used (Figures 3, 4). Negative controls (without primary antibody and tomato lectin) for immunohistochemical staining showed that all immunohistochemistry of target cell signal molecules (Figures 3, 4) were not specifically located in capillaries or arterioles.

Oxidative stress is involved in capillary loss as well as age-related arteriolar and artery wall thickening (Jia et al., 2019; Sharma et al., 2022; Wang Y. et al., 2023). We used immunohistochemistry to study the pathogenesis of AD complexed with mild cerebrovascular disease in microvessels double labelled with the vascular endothelial marker tomato lectin at single endothelial cell level. The interaction of hydroxyl radical (HO), the most toxic reactive oxygen species (ROS), with a nucleoside, such as deoxyguanosine, leads to the formation of 8-OHdG (Kasai, 1997). We also investigated the change in cleaved caspase-3 involved in apoptosis and change in mitochondrial MnSOD that affects oxidative stress.

ANOVA revealed a significant difference among animal groups for cytoplasmic (presumably mitochondrial) 8-OHdG (F_3,171 = 3.663, p = 0.014) and cytosolic cleaved caspase-3 (F_3,123 = 4.037, p = 0.009) in arterioles, but not capillaries, and for mitochondrial MnSOD in both capillaries (F_4,171 = 3.663, p = 0.014) and arterioles (F_3,102 = 4.656, p = 0.004) (Figures 4A–C). In 3×Tg mice, MnSOD decreased (p = 0.002) in capillaries, but not arterioles, while 8-OHdG and cleaved caspase-3 were not affected in both capillaries and arterioles (Figures 4A–C, Tg vs. WT). The data indicates that capillary loss is associated with mild oxidative stress induces in 3×Tg mice.

Although the capillary loss is not further changed, the hypoperfusion and hypoxia are promoted during early AD (Brown and Thore, 2011; Hunter et al., 2012). Hypoxia increases HIF-1α but inhibits the mitochondrial transcription marker (TFAM), mitochondrial MnSOD, and VEGF expression (Lu et al., 2010; Yeo, 2019). Therefore, we studied the change in HIF-1α stability and TFAM expression (Figures 3A, 4D; Tg vs. WT). Significant differences among animal groups were observed for HIF-1α in both capillaries (F_3,113 = 4.230, p = 0.007) and arterioles (F_3,113 = 3.067, p = 0.031) and for TFAM in both capillaries (F_3,161 = 13.064, p = 0.001) and arterioles (F_3,93 = 8.592, p = 0.001). The post-hoc Tukey multiple comparison exhibited an increase in HIF-1α stability (p = 0.018) but a decrease in TFAM expression (p = 0.047) in capillaries of 3×Tg mice (Figures 3A, 4D; Tg vs. WT). We also studied the change in VEGF/PKCε/ERK signal cascade. For capillaries, we observed significant differences among animal groups for VEGF (F_3,113 = 4.230, p = 0.007) and PKCε (F_3,152 = 2.704, p = 0.048), but not ERK (Figures 3B–D). Post-hoc Tukey multiple comparison revealed that VEGF (p = 0.012) and PKCε (p = 0.030) decreased in capillaries (Figures 3B–D; Tg vs. WT).

The data indicate that capillary loss is related with mild oxidative stress; an increase in HIF-1α; and a decrease in TFAM, VEGF, and PKCε in the hippocampi of 3×Tg mice at 16 months old.

In arterioles of 3×Tg mice, HIF-1α stability did not change, but TFAM expression decreased (p = 0.005) (Figures 3A, 4D; Tg vs. WT). This indicates a HIF-1α-independent decrease in TFAM in arterioles in the 3×Tg hippocampus. We also studied change in the VEGF/ PKCε/ERK1/2 on cell proliferation in the arterioles and observed significant differences among animal groups for VEGF (F_3,113 = 4.230, p = 0.007), PKCε (F_3,106 = 3.755, p = 0.014), and ERK1/2 (F_3,114 = 2.858, p = 0.04). However, the expression of VEGF, PKCε, and ERK1/2 in arterioles did not change in 3×Tg mice (Figures 3B–D; Tg vs. WT). These results suggest that although the arteriolar structure did not change, the function may be affected due to dysregulation of mitochondrial transcription.

Cerebrovascular MI increased the strong oxidative DNA damage marker 8-OHdG (p = 0.007) and apoptosis (p = 0.022), but decreased (p = 0.016) mitochondrial MnSOD in arterioles (Figures 4A–C; Tg + Mi vs. Tg). However, in capillaries, cerebrovascular MI reduced MnSOD but did not affect 8-OHdG and cleaved case-3 (Figures 4A–C). These results suggest that MI induces strong oxidative stress and apoptosis in arterioles, while capillary loss in 3×Tg mice with and without MI is related to mild oxidative stress that is not strong enough to induce mitochondrial oxidative DNA damage and apoptosis.

In 3×Tg mice with MI, HIF-1α stability increased in capillaries (p = 0.005) and arterioles (p = 0.019). MI increased mitochondrial TFAM in 3×Tg mice to a level not different from WT controls (Figures 3A, 4D; Tg + Mi vs. Tg). VEGF increased (p = 0.002) in both capillaries and arterioles of 3×Tg + microinfarcts compared with 3×Tg (Figure 3B). However, MI increased the downstream VEGF molecular target PKCε (p = 0.007) and ERK1/2 (p = 0.043) in arterioles but not capillaries (Figures 3C,D).

These results indicate that MI induces strong oxidative stress-associated apoptosis and sustained hypoxia (very low O₂ in the cytoplasm, as mentioned in the Introduction). The sustained hypoxia increases VEGF and its downstream ERK1/2 involved in anti-apoptosis and cell proliferation, resulting in over-proliferation of MV cells and the increased diameter of arterioles.

When we compared 3×Tg + MI mice with and without DCPLA-ME treatment, DCPLA-ME prevented an increase in HIF-1α (Figure 3A) and prevented the loss of PKCε and MnSOD (Figures 3C, 4C). DCPLA-ME also protected the reduction of VEGF and enhanced (p = 0.001) VEGF more than the WT controls (Figure 3B). These results suggest that DCPLA-ME increases PKCε, VEGF (cell proliferation), and MnSOD. An increase in MnSOD protects against oxidative stress and the HIF-1α stability in capillaries of 3×Tg mice with MI.

DCPLA-ME prevented strong oxidative DNA damage (8-OHdG), apoptosis (cleaved caspase 3), HIF-1α, VEGF, PKCε, and ERK1/2 in arterioles (Figures 4A,B, 3A–D; Tg + MI + DCP vs. Tg + MI). These results suggest that DCPLA-ME activates anti-apoptosis and prevents exaggerated cell proliferation in arterioles.

Western blots of the dorsal hippocampus revealed a decrease in MnSOD (p = 0.040) in 3×Tg mouse hippocampus (Figures 5A,B; Tg vs. WT), confirming an age-related change in AD. However, western blots showed an increase in MnSOD (p = 0.041) in 3×Tg mouse hippocampus with microinfarcts (Figures 5A,B; Tg + MI vs. Tg), confirming that MI induces sustained hypoxia in 3×Tg mouse hippocampus. DCPLA prevented the rise of MnSOD in 3×Tg mice with cerebral microinfarcts (Figures 5A,B; Tg + MI + DCP vs. Tg + MI).

The results from western blots of MnSOD at the dorsal hippocampal levels (Figures 5A,B) were different from those studied with immunohistochemistry (Figure 4C). This confirms that for targeted cell signals that were not expressed only in blood vessels, immunohistochemistry at an individual capillary or arteriole will give more accurate results than western blot analysis.

Astrocytes are ideally positioned to mediate neurovascular coupling, relaying signals from neurons to blood vessels that regulate blood flow in the brain (MacVicar and Newman, 2015). Immunohistochemistry of the astrocyte marker GFAP and the vascular endothelial cell marker tomato lectin was performed. In Figure 6A, A-V coupling areas were identified as colocalization (yellow) of astrocytic end feet (green fluorescence) and vascular endothelial cells (red fluorescence). ANOVA showed significant differences among animal groups for the number of astrocytes (F_4,72 = 6.587, p = 0.002) and A-V coupling areas (F_4,178 = 8.247, p = 0.001). In WT mice, although MI did not induce change in morphometry of microvessels (Figures 2B–E), MI induced loss of astrocytes (p = 0.004) and reduction of A-V coupling areas (p = 0.003) (Figures 6B,C; WT + MI vs. WT). In 3×Tg mice without MI, we observed the loss of astrocytes (p = 0.019) and A-V coupling areas (p = 0.001) (Figures 6B,C; Tg vs. WT). MI enhanced the loss of astrocytes (p = 0.011) and A-V coupling areas (p = 0.042) (Figures 6B,C; Tg + MI vs. Tg). Western blots of the dorsal hippocampus showed an increase in GFAP protein in 3×Tg (p = 0.042), 3×Tg + MI (p = 0.023), and 3×Tg + MI + DCP mice (p = 0.042) (Figures 6D,E). This suggests that the remaining astrocytes are enlarged (hypertrophy) to compensate for the loss of astrocytes. DCPLA-ME prevented the enhancement effect of MI on the loss of astrocytes (p = 0.006) and A-V coupling areas (p = 0.009) (Figures 6B,C).

Next, the effects of changes in capillaries and arterioles as well as A-V coupling on neurons were further investigated. We studied changes in synapses associated with MV damages in 3×Tg mice with MI. Immunohistochemical detection visualized with a confocal microscope (Figure 7A) was used to stain presynaptic axonal boutons (presynaptic vesicle membrane protein synaptophysin) and post-synaptic membranes (neurogranin). Presynaptic axon boutons (synaptophysin grains) and post-synaptic membranes (neurogranin grains) were counted in a 30 μm × 30 μm area. Synaptophysin intensity indicated presynaptic vesicle amount within the axonal boutons. Significant differences among animal groups were observed for presynaptic axon bouton density (F_4,95 = 16.285, p = 0.001), presynaptic vesicle concentration (F_4,98 = 6.470, p = 0.001), and post-synaptic membrane density (F_4,79 = 8.677, p = 0.001).

The number of presynaptic boutons decreased (p = 0.001) in 3×Tg mice (Figure 7B; Tg vs. WT). In Figure 7B, although MI induced the loss of presynaptic boutons in WT mice (WT + MI vs. WT), MI did not enhance the loss of presynaptic boutons in 3×Tg mice (Tg + MI vs. Tg). The presynaptic vesicle concentration in 3×Tg mice was not different from that in WT mice (Figure 7C; Tg vs. WT). In Figure 7C, MI induced the loss of presynaptic vesicle concentration in both WT (p = 0.001, WT + MI vs. WT) and 3×Tg (p = 0.049, Tg + MI vs. Tg) mice. Similar to changes in presynaptic vesicle concentration (Figure 7C), MI activated the loss (p = 0.001) of postsynaptic membrane density (neurogranin grains) in WT and 3×Tg mice (Figure 7D). In 3×Tg mice, although the number of postsynaptic membranes did not change, western blots demonstrated a down-regulation of the postsynaptic membrane protein neurogranin (Figures 7E,F). These data suggest the shrinkage of postsynaptic membrane size in 3×Tg mice. DCPLA-ME prevented the reduction of presynaptic vesicles and membranes and postsynaptic membranes from 3×Tg mice with cerebral MI (Figures 7B–F).

The presence and extent of white matter hyperintensities or leukoaraiosis is a radiographic marker (e.g., MRI) of capillaries’ cerebral vessel disease, cognitive impairment, and functional disability (Capizzano et al., 2004). In AD brains, more myelin loss occurs in the late Braak stage (Ihara et al., 2010; Erten-Lyons et al., 2013). Changes in the myelinated axons in the white matter were quantified in the perforated path that provides a connectional route from the entorhinal cortex to all fields of the hippocampal formation.

We used immunohistochemistry and confocal microscopy to determine the changes in the myelin basic protein (MBP) in the hippocampal perforant path, which is the principal source of cortical input to the hippocampal formation (Figure 8A). Significant differences among animal groups were observed for the number (F_4,75 = 4.148, p = 0.005) and size (F_4,75 = 37.919, p = 0.001) of MBP profiles. The number and size of MBP profiles did not change in WT mice with MI or 3×Tg mice. A combination of MI in 3×Tg mice increased (p = 0.002) the number but decreased (p = 0.001) the size of myelinated axons, indicating uneven demyelination (Figures 8B,C; Tg + MI vs. Tg). Western blots demonstrated the downregulation (p = 0.031) of MBP in the dorsal hippocampus of 3×Tg mice with microinfarcts (Figure 8E). DCPLA-ME prevented demyelination in 3×Tg mice with MI (Figures 8B–E).