All animal experiments in the present study were approved by the Institutional Animal Care and Use Committee of Guangdong Laboratory Animals Monitoring Institute (Guangzhou, China; IACUC no. 2015023). All procedures were performed in accordance with the AAALAC guidelines (14).

The Slit2-Tg mice overexpressing human Slit2 were from Guangdong Pharmaceutical University (Guangzhou, China), as previously described (15). The heterozygous transgenic mice were crossed with C57BL/6 mice (Stock no. 000664; Jackson Laboratory, Ben Harbor, ME, USA) to generate Slit2-Tg mice and wild-type littermates (WT mice). Unless otherwise noted, the animals used in the present study defined as aging were 15-month-old adult male mice. All mice were provided with water and a standard chow diet ad libitum. The mice were housed in a specific pathogen-free facility with a 12 h light/dark cycle at 23±2°C and 50±10% humidity.

The transgenic offspring were identified by polymerase chain reaction (PCR) using the following primer sequences: Slit2 forward 5′-CCC TCC GGA TCC TTT ACC TGT CAA GGT CCT-3′ and Slit2 reverse 5′-TGG AGA GAG CTC ACA GAA CAA GCC ACT GTA-3′ (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA); the product size was 645 bp. In all experiments, the animals were anesthetized with chloral hydrate (4.2%, 0.01 ml/g).

Following CO₂ euthanasia, mouse brains were removed and total RNA extraction using TRIzol (Invitrogen; Thermo Fisher Scientific, Inc.) and RT was performed using the PrimeScript™ RT reagent kit (Takara Bio, Inc., Otsu, Japan) at 37°C for 30 min and 85°C for 1 min, according to the manufacturer's protocol. The primers used for Slit2 were provided by Invitrogen; Thermo Fisher Scientific, Inc. and were as follows: Forward, 5′-AGC CGA GGT TCA AAA ACG AGA-3′ and reverse, 5′-GGC AGT GCA AAA CAC TAC AAG A-3′ (Invitrogen; Thermo Fisher Scientific, Inc.). The qPCR analysis was performed as follows: 95°C for 30 sec followed by 40 cycles of 95°C for 5 sec and 60°C for 34 sec, using an ABI-7500 Fast Real-Time PCR system (Applied Biosystems; Thermo Fisher Scientific, Inc.) in a total volume of 25 µl containing 2 µM primers, 80 µg cDNA and 12.5 µl of the SYBR^® Premix Ex Taq™ (Takara Bio, Inc.). The results were quantified using the 2^−ΔΔCq method (16).

The fluorescent CSF tracer penetrating into the intact living brain was visualized through a closed cranial window using a custom-built 2-photon laser scanning microscope (Leica Microsystems GmbH, Wetzlar, Germany), as described previously (17,18). Briefly, the animals were anesthetized and securely placed on a stereotaxic device with the skull level between the bregma and lambda. The stereotaxic coordinates of the right parietal cortex were 2 mm caudal from the bregma, and 1.7 mm lateral from the midline (19). A thin cranial window over the parietal area was prepared for 2-photon in vivo imaging. As described by Ren et al (20), FITC-conjugated dextran (40 kDa; D-1841; Thermo Fisher Scientific, Inc.) was first reconstituted with artificial CSF into a concentration of 0.5%. This was injected it into the subarachnoid CSF via cisterna magna puncture at a rate of 2 µl/min for a period of 5 min with a microsyringe pump controller, following which 200 µl of dextran rhodamine B (70 kDa; D-1841; Thermo Fisher Scientific, Inc.) was injected intravenously immediately prior to imaging to visualize the vasculature. The mice were immediately transferred to the Leica DM6000 CFS stage (Leica Microsystems GmbH) and their head immobilized.

Imaging was performed using a Leica SP5 2-photon imaging system (Leica Microsystems GmbH) equipped with a Ti:Sapphire laser (Coherent Chameleon Ultra II; Coherent, Inc., Santa Clara, CA, USA) and 25X/0.95 NA water-immersion objective and controlled with Leica LAS X 3.3.0 software (Leica Microsystems GmbH). Images of the cerebral vasculature were first captured with 512×512 pixel frames from the surface to a depth of 300 µm with 2-µm z-steps, to avoid the influence of laser irradiation. Quantification of the mean pixel intensity of the tracer was processed from 40 µm (xyz) of the 3-dimensional image stacks obtained 5, 15, 30, 45, and 60 min following intra-cisternal injection. To examine the dissipation of tracer in the paravascular space of penetrating arterioles, the 2-photon in vivo image of 100-µm below the surface was captured. Donut-shaped region of interests (ROIs) were defined and the mean pixel intensity of the tracer within these ROIs in the paravascular space was quantified. The para-arteriolar ROIs at each time point in each animal were separately averaged to generate values for a single parameter.

For histological evaluation, the mice were transcardially perfused with 50 ml of ice-cold saline followed by 200 ml of 4% (w/v) formaldehyde in PBS. The brains were removed, immersed in 20–30% sucrose solution overnight and embedded in Tissue-Tek Optimal Cutting Temperature compound (Sakura Finetechnical, Co., Ltd., Tokyo, Japan) at −20°C. Brain tissue blocks were consecutively cryosectioned into 10-µm thick cross sections with a cryostat microtome (Leica Microsystems GmbH).

For immunofluorescence staining, the frozen sections were treated with 0.3% triton and 10% anti-donkey serum (Abcam, Cambridge, UK) for 1 h at room temperature. Subsequently, the sections were incubated overnight at 4°C in the dark with the following primary antibodies: Mouse anti-glial fibrillary acidic protein (GFAP; cat. no. MAB360), rabbit anti-AQP4 (cat. no. AB3068), mouse anti-Aβ1-40 (cat. no. MABN11), rabbit anti-Aβ1-42 (cat. no. AB5078P; all 1:500; EMD Millipore, Billerica, MA, USA). Alexa Fluor^® 488-conjugated immunoglobulin G (heavy and light chain), F(ab)₂ Fragment antibodies were used as secondary antibodies and incubated with the membrane at 37°C for 1 h in the dark. These secondary antibodies were anti-mouse (cat. no. 4409) for GFAP, anti-rabbit (cat. no. 4412) for AQP4 and Aβ1-40, and anti-rabbit (cat. no. 4413; all 1:300; Cell Signaling Technology, Inc., Danvers, MA, USA) for Aβ1-42. All sections were mounted with DAPI as a nuclear stain. A Leica TCS SP5 Spectral one-photon microscope (Leica Microsystems GmbH) was used to acquire immunofluorescent staining data. The excitation powers were 5 mW for IgG Alexa Fluor 488 and 0.1 mW for IgG Alexa Fluor 555. The photomultiplier tube value was 800 V without offset. All immunofluorescence staining was repeated three times. All images were captured at the same exposure time.

The polarization of astrocytic AQP4 was evaluated in accordance with a previous study (21). The color channels in the histological sections labeled for GFAP and AQP4 were separated, and each image was uniformly captured at two levels (high and a low stringency thresholds). The low-stringency threshold defined the overall area of AQP4-immunoreactivity, whereas the high-stringency threshold defined the area of intense AQP4-immunoreactivty that was localized to perivascular endfeet. The ratio of the low stringency area:high stringency area was defined as 'AQP4 polarity'. A higher AQP4 polarity represented a greater proportion of immunoreactivity restricted to perivascular regions, whereas a lower proportion indicated that the distributed immunoreactivity was between the perivascular endfeet and the soma.

The Morris water maze experiment was performed according to the protocols in a previous report by our group (17). The investigators were blinded during the experiment. The maze consisted of a circular tub (120 cm in diameter, 50 cm in height) and a white circular platform (10 cm). The tub was surrounded by a curtain, which was located ~1 m from the tub wall and painted with distinct geometric cues, the water (24±1°C) was rendered opaque with white tempera paint to conceal the platform. Over 4 consecutive days, the platform was submerged 1 cm under the surface of the water in the center of one of the pool quadrants. The mice were subjected to four trials (up to 60 sec) per day from each of the four start locations. Animals that failed to locate the platform within the allotted 60 sec were gently guided to the platform. All mice remained on the platform for 10 sec at the end of each trial. On day 5, the platform was removed and a single 60 sec probe trial was performed. The swim paths were recorded using an overhead video camera and tracked by ANY-maze 6.0 (San Diego Instruments, San Diego, CA, USA). The velocity during the probe trial, the number of times the target area (former platform) was crossed and the time spent in each quadrant during the probe trial were recorded.

All data are presented as the mean ± standard deviation or standard error of the mean. An independent sample t-test or Wilcoxon rank sum test was used for comparison between two groups. One-way analysis of variance (ANOVA) or Kruskal-Wallis test and LSD t-test or Bonferroni test were used for comparison of mean pixel intensity with the PVS and the latency to the platforms during the water maze training. SPSS 20.0 (IBM SPSS, Armonk, NY, USA) software was used for the statistical analysis. Images and sections were analyzed by an investigator, who was blinded to the experimental conditions. ImageJ 1.50i (National Institutes of Health, Bethesda, MD, USA) software was applied for analysis of the immunohistochemical results. The histology data were analyzed according to a previous study (22). Briefly, four locations per sample (three fields per section; six sections per mouse) were used for analysis. Differences in fluorescent CSF tracer, perivascular GFAP and polarization of AQP4, Aβ1-40 and Aβ1-42 immunofluorescence between the Slit2-Tg mice and WT mice were compared using an unpaired t-test. Differences in the Morris water maze results were evaluated by one-way ANOVA followed by Tukey's post hoc test for multiple comparisons. P<0.05 was considered to indicate a statistically significant difference.

Impairment of paravascular pathway function in the aging brain has an adverse effect on glymphatic CSF recirculation (3). To investigate the effect of Slit2 on paravascular pathway function in the aging brain, the present study verified whether Slit2 was expressed in the mouse brain using RT-qPCR analysis, the results of which showed the overexpression of Slit2 in the brain of the Slit2-Tg mice, compared with the WT mice (Fig. 1A). Following this, the dynamics of the paravascular CSF-ISF exchange in vivo were evaluated by 2-photon microscopy and the intra-cisternal injection of fluorescent CSF tracer (FITC-conjugated dextran, MW 40 kDa). The cerebral vasculature was visualized through a thinned-skull window over the parietal area following caudal vein injection of Rhodamine B. As shown in Fig. 1B, the intra-cisternal injection of FITC tracer was followed by a distinct paravascular influx, which moved rapidly into the cortex along penetrating arterioles and entered the interstitium of the parenchyma. One-way ANOVA indicated that the quantification of mean pixel intensity of the 3D image stacks (Fig. 1C) was significantly different at different time points in the WT group (F=9.927, P<0.001). The LSD-t test showed that interstitial accumulation of the tracer appeared in the parenchyma within 5 min (29.22±12.53) and increased at 15 min (31.34±3.65), although there was no significant difference from that at 5 min (P>0.05). The mean pixel intensity of the CSF tracer peaked at ~30 min (58.50±5.66, P<0.001) following injection in the aging WT mice, and gradually reduced at 45 min (45.84±8.85, P<0.05) and at 60 min (41.16±4.41, P>0.05). In the Slit2-Tg mice, interstitial accumulation of the CSF tracer was also observed within 5 min (41.11±12.66), and peaked at ~15 min (60.75±6.90). Subsequently, the mean pixel intensity was significantly decreased at 30 min (39.73±7.77), 45 min (32.60±4.98) and 60 min (19.61±5.22). However, one-way ANOVA indicated that the mean pixel intensities were not significantly different from each other (F=1.385, P>0.05). The independent sample t-test indicated no significant difference in the pixel intensity at 5 min post-CSF tracer injection (t=1.492, P>0.05) or between the peak pixel intensity of the CSF tracer between the WT mice and Slit2-Tg mice (t=0.563, P>0.05). However, there was significant attenuation of the pixel intensity of CSF tracer accumulation in the parenchyma of the Slit2-Tg mice compared with that in the WT mice at 45 min (t=2.917, P<0.05) and 60 min (t=7.051, P<0.001).

The CSF tracer was analyzed in the perivascular space of penetrating arteries 100 µm below the cortical surface (Fig. 1D-a). In the aging brain of the WT mice, one-way ANOVA indicated that the accumulation of CSF tracer along perivascular spaces was significantly different at different time points (F=8.643, P<0.001). The LSD-test showed that the CSF tracer penetrating into the brain parenchyma was observed within 5 min (56.03±15.18), increased at 15 min (72.98±17.68, P<0.05) and peaked at 30 min (96.98±26.53) (Fig. 1D-b and E, P<0.01). No significant decrease in the fluorescence intensity of the CSF tracer was observed at 45 min (90.20±23.20; t=0.667, P>0.05) or 60 min (91.67±14.27). By contrast, the Kruskal-Wallis test indicated that the accumulation of CSF tracer along perivascular spaces was significantly different at different time points in the Slit2-Tg mice (P<0.001). It was present at 5 min (66.83±23.36), but decreased at 15 min (49.89±20.43) (Fig. 1D-b and E). The fluorescence intensity of CSF tracer in the paravascular space gradually decreased at 30 min (34.60±15.29), 45 min (30.21±13.48) and 60 min (22.96±11.36). Notably, the peak intensity of CSF tracer in the WT mice was significantly higher than that in the Slit2-Tg mice (t=0.243, P<0.001). An independent t-test showed that the fluorescence intensity of the CSF tracer was significantly increased at 60 min, compared with that at 5 min (t=0.276, P<0.001) in the aging WT mice, whereas the Wilcoxon rank sum test on the fluorescence intensity was significantly decreased at 60 min, compared with that at 5 min (P<0.001) in the aging Slit2-Tg mice. These results indicated that the overexpression of Slit2 accelerated paravascular CSF-ISF exchange in the aging brain.

The depolarization of AQP4 in reactive astrocytes is closely associated with impairment of the paravascular pathway in the aging brain (3). To understand why the overexpression of Slit2 restores the function of the paravascular pathway, the activation of astrocytes in the brain parenchyma and the polarization of AQP4 were evaluated. As shown in Fig. 2A, the GFAP-positive astrocytes were widespread in the cortex and hippocampus of the aging brain in WT and Slit2-Tg mice. An independent sample t-test indicated that the mean fluorescence intensity of GFAP-positive cells was significantly decreased in the Slit2-Tg mice, compared with that in the WT mice in the cortex (43.21±8.16, vs. 54.21±8.58; t=0.814, P<0.05; Fig. 2B-a) and hippocampus (40.02±4.28, vs. 59.08±9.89; t=0.069, P 0.01; Fig. 2B-b).

As a main component of water channel proteins expressed by astrocytes, AQP4 is polarized in the perivascular astrocytic endfeet in the healthy young brain, but not in the aging brain. AQP4 delocalization from the endfeet to the soma of astrocytes is, in part, associated with the failure of the paravascular pathway (3). Therefore, the present study investigated the polarization of AQP4 in the aging brain of WT and Slit2-Tg mice (Fig. 2C-a). In the Slit2-Tg mice, the expression of AQP4 was well distributed around the perivascular region, where AQP4 sheathed the astrocytic endfeet, and was less widespread in the astrocytic soma (Fig. 2C-b), whereas AQP4 was mis-located in the soma of astrocytes in the WT mice (Fig. 2C-b). According to a previous report (21), the value of AQP4 polarity was analyzed, which was defined as the low stringency area (overall area of AQP4-immunoreactivity in the image): High stringency area (area of intense AQP4-immoreactivty localized to the perivascular endfeet in the image) in the WT mice (Fig. 2D-a) and Slit2-Tg mice (Fig. 2D-b). An independent sample t-test indicated that astrocytic AQP4 polarity was significantly increased in the aging Slit2-Tg mice (0.88±0.10), compared with that in the WT mice (0.50±0.15; t=0.368, P<0.001, Fig. 2E). This result suggested that the improved paravascular pathway function in the aging brain induced by the overexpression of Slit2 was accomplished by the enhancement of astroglial water transport.

The disruption of the BBB caused by aging results in loss of vasomotion and decreases the efficiency of paravascular pathway clearance of Aβ (3,23), In the present study, the dynamic change of BBB function was evaluated by in vivo 2-photon microscopy and intravenous injection of dextran rhodamine B (MW 40 kDa). The 3D image stacks (Fig. 3A) showed that intravenous injection of dextran rhodamine B rapidly leaked from blood vessels into the brain parenchyma of WT mice. However, rhodamine B was restricted inside the blood vessels of the brain and minimal leakage was observed in the brain parenchyma of the Slit2-Tg mice.

To quantify the leakage of rhodamine B from the BBB, the total fluorescence intensity in the extravascular compartment was analyzed (24) (Fig. 3B). Two-way repeated ANOVA indicated no significant interaction between group and time factors (P>0.05). The main effect of the group and time factors were significant (F=4.152, P<0.05 and F=41.52, P<0.001, respectively). Bonfferoni's post hoc test was used to analyze the fluorescence intensity to examine the BBB permeability. No significant difference between the WT and Slit 2-Tg mice was observed at 5 min (598.50±162.11, vs. 414.41±153.84 AU, P>0.05) or 15 min (864.48±299.30, vs. 460.78±159.32 AU, P>0.05). The fluorescence intensity in the extravascular compartment was significantly decreased in the Slit-Tg mice, compared with that in the WT mice at 30 min (443.08±85.49, vs. 1,004.13±310.60 AU, P<0.05), 45 min (1,077.08±420.20, vs. 489.39±104.72 AU, P<0.01) and 60 min (1,174.16±427.65, vs. 536.12±148.46 AU, P<0.01) (Fig. 3C). These results indicated that the overexpression of Slit2 maintained the integrity of the BBB in the aging brain.

The paravascular pathway and interstitial waste removal are suppressed with aging, which may contribute to the accumulation of Aβ leading to the pathogenesis of neurodegenerative diseases, including AD (3). To evaluate the effect of Slit2 on the accumulation of Aβ, immunofluorescent staining was performed to analyze the deposition of Aβ1-42 and Aβ1-40 in the brain parenchyma of aging mice. It was found that increased Aβ1-40 moved out from the blood vessels of the WT mice than that in the Slit2-Tg mice in the cortex and hippocampus (Fig. 4A). An independent sample t-test indicated that the overall fluorescence intensity was significantly decreased in the cortex of the Slit2-Tg mice (13.65±2.57), compared with that of the WT mice (33.70±8.18; t=5.726, P<0.001) (Fig. 4B). The accumulation of intra-neuronal Aβ1-42 was further analyzed in the two regions (Fig. 4C), and the mean fluorescence intensity of Aβ1-42 was also significantly decreased in the cortex of the Slit2-Tg mice (14.88±2.47), compared with that of the WT mice (23.98±4.70; t=4.194, P<0.01) (Fig. 4D). These results indicated that the over expression of Slit2 restored the function of the paravascular pathway and reduced Aβ deposition in the aging brain.

To investigate the effect of the overexpression of Slit2 on the spatial memory cognition of aging mice, Morris water maze tests were performed. As shown in Fig. 5A, the Wilcoxon rank sum test revealed no significant difference in the latency to reach platforms between the WT mice and Slit2-Tg mice (42.097±3.842, vs. 44.739±2.194; P>0.05) on day 1. The Kruskal-Wallis test showed that the latency on different training days was significantly different (P<0.001) in the WT group. The Bonferroni test showed that the latency on day 4 (20.593±1.800) and day 5 (18.499±3.238) was significantly decreased, compared with that on day 1 (P<0.001; P<0.001); the latency on day 5 was also significantly decreased, compared with that on day 2 (31.788±2.554) (P<0.05). However, the latency on day 3 (31.402±4.214) did not differ significantly from that on the other days (all P>0.05). In the Slit2-Tg mice, the Kruskal-Wallis test showed that the latencies on different training days were significantly different (P<0.001). The Bonferroni test was used to compare with the latency on day 1. The latency to reach the platform was significantly lower on day 2 (19.326±2.995; P<0.001), day 3 (15.780±3.102; P<0.001), day 4 (19.108±3.436; P<0.001) and day 5 (12.327±1.479, P<0.001), however, no significantly difference was observed among these consecutive 3 days. The latency was significantly decreased in the Slit2-Tg mice, compared with that in the WT mice on days 2 and 3 (t=3.166 and 2.985, P<0.01); whereas no significant differences were observed on days 4 and 5 between the two groups (t=0.383 and 1.734, P>0.05). In the probe trial (Fig. 5B), the independent sample t-test indicated no significant difference in velocities between the WT mice (30.03±1.30 cm/s) and Slit2-Tg mice (33.308±1.34 cm/s; t=1.753, P>0.05; Fig. 5C), whereas the time to the target area (previous platform) was significantly increased in the Slit2-Tg mice (8.20±0.59), compared with that in the WT mice (5.10±0.433; t=4.223, P<0.001; Fig. 5D). Finally, the time spent in the target quadrant (%) was analyzed (Fig. 5E), independent sample t-test indicated that the time spent in the target quadrant (%) was significantly increased in Slit2-Tg mice (53.417±5.287), compared with that in WT mice (38.982±3.215; t=2.333; P<0.05). These data collectively suggested that the overexpression of Slit2 restored the function of the paravascular pathway, which assisted in improving spatial memory cognition in the aging mice.