Cerebral vascular and blood brain-barrier abnormalities in a mouse model of epilepsy and tuberous sclerosis complex

Abstract

Objective:

Tuberous sclerosis complex (TSC) is a genetic disorder, characterized by tumor formation in the brain and other organs, and severe neurological symptoms, such as epilepsy. Abnormal vascular endothelial growth factor (VEGF) expression may promote angiogenesis in kidney and lung tumors in TSC and has been identified in brain specimens from TSC patients, but the role of VEGF and vascular abnormalities in neurological manifestations of TSC is poorly defined. In this study, we investigated abnormalities in brain VEGF expression, cerebral blood vessel anatomy, and blood-brain barrier (BBB) structure and function in a mouse model of TSC.

Methods:

Tsc1^GFAPCKO mice were used to investigate VEGF expression and vascular abnormalities in the brain by western blotting and immunohistochemical analysis of vascular and BBB markers. In vivo two-photon imaging was used to assess BBB permeability to normally impenetrable fluorescently-labeled compounds. The effect of mechanistic target of rapamycin (mTOR) pathway inhibitors, VEGF receptor antagonists (apatinib), or BBB stabilizers (RepSox) were assessed in some of these assays, as well as on seizures by video-EEG.

Results:

VEGF expression was elevated in cortex of Tsc1^GFAPCKO mice, which was reversed by the mTOR inhibitor, rapamycin. Tsc1^GFAPCKO mice exhibited increased cerebral angiogenesis and vascular complexity in cortex and hippocampus, which were reversed by the VEGF receptor antagonist, apatinib. BBB permeability was abnormally increased and BBB-related tight junction proteins, occludin and claudin-5, were decreased in Tsc1^GFAPCKO mice, also in an apatinib- and RepSox-dependent manner. The BBB stabilizer (RepSox), but not the VEGF receptor antagonist (apatinib), decreased seizures and improved survival in Tsc1^GFAPCKO mice.

Significance:

Increased brain VEGF expression is dependent on mTOR pathway activation and promotes cerebral vascular abnormalities and increased BBB permeability in a mouse model of TSC. BBB modulation may affect epileptogenesis and represent a rationale treatment for epilepsy in TSC.

1. Introduction

Tuberous sclerosis complex (TSC) is a genetic disorder, primarily characterized by the development of tumors or other pathological lesions in the brain and other organs ^{1, 2}. Neurological involvement often accounts for the most severe and disabling symptoms of TSC, including epilepsy, intellectual disability, and autism ³. Dysregulation of the mechanistic target of rapamycin (mTOR) pathway due to mutation of either the TSC1 or TSC2 gene promotes tumor growth in various organs, but the mechanisms leading to the primary neurological manifestations of TSC are poorly understood and usually appear to be independent of overt tumor growth per se.

Vascular endothelial growth factor (VEGF) is a family of cytokines or growth factors that primarily promotes angiogenesis, though may have a number of other developmental, inflammatory or metabolic functions ^{4, 5}. In TSC, VEGF levels are increased in the serum and peripheral organs of TSC patients, and are hypothesized to serve as a biomarker and probable driver of vascular tumors in TSC, such as renal angiomyolipomas and pulmonary lymphangioleimyomatosis ^{6, 7}. In the brain, increased VEGF expression has also been identified in cortical tubers and other pathological brain specimens from TSC patients, as well as in a mouse model of TSC ⁸. However, the potential role of abnormal VEGF expression in the neurological manifestations of TSC has not been well-defined or thoroughly investigated.

VEGF can promote cerebral angiogenesis, but may also have other effects in the brain, such as on neuronal and glial proliferation, differentiation, migration, and synaptic plasticity ^{4, 5}. Furthermore, along with its angiogenic properties, VEGF also increases vascular permeability, potentially promoting leakage of the blood-brain barrier (BBB) ^{9, 10}. Vascular density is increased in TSC mouse models and human focal cortical dysplasia specimens ^11–13, but the dependence of these vascular abnormalities on VEGF or the downstream functional consequences of either abnormal VEGF expression or vascular abnormalities have not been established. In this study, we further characterized abnormalities in brain VEGF expression, vascular structure, and BBB permeability and investigated their mechanistic relationship and possible functional effects on seizures in a mouse model of TSC.

2. Materials and Methods

2.1. Animals

Care and use of all mice were conducted according to an animal protocol approved by the Washington University School of Medicine (WUSM) Animal Studies Committee, and consistent with National Institutes of Health (NIH) guidelines on the Care and Use of Laboratory Animals. In addition, NIH guidelines on Rigor and Reproducibility in Preclinical Research were followed, including use of randomization, blinding, both sexes, and statistical/power analyses. Tsc1^flox/flox-GFAP-Cre knock-out (Tsc1^GFAPCKO) mice with conditional inactivation of the Tsc1 gene in both neurons and glia were generated as described previously ^{14, 15}. Tsc1^flox/+-GFAP-Cre and Tsc1^flox/flox littermates have previously been found to have no abnormal phenotype and were used as control animals in these experiments. No sex differences were detected in any experiment and the sex of mice data have been identified by color-coding in dot plots for all figures (male – black symbol; female – gray symbol).

2.2. Drug Treatment

In some experiments, three-week-old Tsc1^GFAPCKO mice were randomized to treatment with the mTOR pathway inhibitor, rapamycin (6 mg/kg/d, i.p) or vehicle for one week, with brain tissue then harvested for western blot analysis. Rapamycin (LC Labs) was initially dissolved in 100% ethanol, stored at −20°C, and diluted in a vehicle solution containing 5% Tween 80, 5% PEG 400 (Sigma), and 4% ethanol.

In other experiments, three-week-old Tsc1^GFAPCKO mice were randomized to treatment with 1) apatinib (Selleck Chemicals), a tyrosine kinase inhibitor that inhibits VEGF receptor-2, (75 mg/kg/d, i.p.) ¹⁶, 2) RepSox (STEMCELL Technologies), a TGF-beta inhibitor that induces claudin/tight junction expression and improves BBB function (10mg/kg/d, i.p.)^{17, 18}, or 3) vehicle (dimethyl sulfoxide, DMSO, Sigma) for at least one week. In some mice, brain tissue was then harvested for western blot or immunohistochemistry, or transcardial perfusion of dextran and immunofluorescence analysis. Other mice underwent craniotomy for in vivo two-photon imaging after one week of apatinib, RepSox, or vehicle treatment. For video-EEG studies of seizure and survival analysis, in two separate studies, apatinib, RepSox, or vehicle was administered starting at three weeks (prior to onset of seizures and video-EEG recording) until death. Apatinib or RepSox was initially dissolved in DMSO, stored at −20°C, and diluted in saline before injection.

2.3. Western blotting

Western blot analysis was used to measure protein levels of VEGF, hypoxia inducible factor alpha (HIF-1α), tight junction proteins (occludin, claudin-5, ZO-1), and PS6 and S6 (as assays of mTOR activity) in the brains of Tsc1^GFAPCKO and control mice, using standard methods as described previously ¹⁵. Briefly, brains were dissected and homogenized with the cortex being isolated. Equal amounts of total protein extract were separated by gel electrophoresis and transferred to nitrocellulose membranes. β-actin was used as a loading control. Membranes were incubated with primary antibodies to VEGF (#sc-7269, Santa Cruz Biotechnology, 1:1000), HIF-1α (#14179, Cell Signaling Technology, 1:1000), occludin (#ab167161, Abcam, 1:1000), claudin-5 (#35-2500, Invitrogen Life Technologies, 1:1000), ZO-1 (#61-7300, Invitrogen Life Technologies, 1:1000), PS6 (#2215, Cell Signaling Technology, 1:1000), S6 (#2217, Cell Signaling Technology, 1:1000) or β-actin (#4970, Cell Signaling Technology, 1:1000), followed by reaction with a peroxidase-conjugated secondary antibody anti-rabbit IgG (#7074, Cell Signaling Technology, 1:1000) or anti-mouse IgG (#7076, Cell Signaling Technology, 1:1000). Signals were detected by enzyme chemiluminescence (Cytiva-Amersham) and quantitatively analyzed with ImageJ software (NIH, Bethesda, MD).

2.4. Immunohistochemistry

Histological analysis was performed in a blinded fashion to assess CD31 expression, a marker of vascular endothelial cells. In brief, brains were perfusion-fixed with 4% paraformaldehyde and cut into 45 μm sections with a freezing microtome. Immunohistochemistry was performed by labeling with primary antibody, anti-CD31(# 77699, Cell Signaling technology, 1:100), followed by labeling with secondary antibody Alexa-488 conjugated goat anti-rabbit IgG (#A11034, Invitrogen Life Technologies, 1:300). In addition, sections were counterstained with TO-PRO-3 Iodide (#T3605, Invitrogen Life Technologies, 1:1000) for the nonspecific nuclear staining of all cells.

Images were acquired with a Zeiss LSM PASCAL confocal microscope (Zeiss Thornwood, NY). In images from coronal sections at approximately 2 mm posterior to bregma, from two sections per mouse, from a total of six mice per group, the length of CD31 positive cells in each group was measured by outlining the positive staining area using ImageJ software (NIH, Bethesda, MD).

To assess glial proliferation, after 1 week treatment of Apatinib, starting from 3 weeks of age, immunohistochemistry analysis was performed by standard methods. Briefly, brains were perfusion-fixed with 4% paraformaldehyde and cut into 45 μm sections with a freezing microtome. Some sections were stained with GFAP anti-mouse antibody (1:500; #3670, Cell Signaling Technology) followed by labeling with secondary antibody, Cy3 conjugated goat anti-mouse IgG (1:500; Jackson Immuno, West Grove, PA). In addition, some sections were counter-stained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI, 1:1000, Life Technologies), for the nonspecific nuclear staining of all cells. Images were acquired with a NanoZoomer HT Scan system 2.0 (Hamamatsu, Shizuoka, Japan). Regions of interest were marked in sections of neocortex and hippocampus by a 600 μm × 600 μm wide box and GFAP-immunoreactive cells were counted by an investigator blinded to the treatment of the mice.

2.5. Cerebral blood vessel analysis with conjugated dextran immunofluorescence

Vehicle or apatinib-treated mice were transcardially perfused with Alexa Fluor 488-conjugated dextran (Dextran-AF488,100ul/mouse, 0.5% 3000 MW, Thermofisher, #D22910) for 3 minutes. Brains were removed and fixed in 4% PFA for 24 hours at 4°C, and then cut into 45 μm sections with a freezing microtome. After mounting sections on slides, images were acquired with a Zeiss LSM PASCAL confocal microscope (Zeiss Thornwood, NY). In images from coronal sections at approximately 2 mm posterior to bregma (two sections per mouse from a total of six mice per group), vessels were analyzed by using the AngioTool software (Version 0.5, https://ccrod.cancer.gov/confluence/display/ROB2/Home), including assessment of total vessel length, vessel density, junction density, and lacunarity ^{11, 19} (Supplemental Fig. S1).

2.6. In Vivo Two-Photon Imaging of BBB Permeability

After one week treatment of vehicle, apatanib, or RepSox starting at 3 weeks of age, Tsc1^GFAPCKO mice underwent surgery at 4 weeks of age for cranial window placement and in vivo two-photon imaging of BBB permeability and vascular or astrocytic end-feet structure. Mouse craniotomy surgery was performed using similar methods as previously reported ^{20, 21}. Briefly, mice were anesthetized with 2% isoflurane anesthesia and held in a custom-made stereotaxic device which could be mounted to a microscope stage. A heating pad was used to maintain body temperature while under anesthesia. A round cranial window was made by removing part of the skull (4 mm in diameter) over the left somatosensory cortex while leaving the dura intact. For in vivo labeling of cortical astrocytes, sulforhodamine 101 (SR101, #S7635, Sigma) was dissolved in saline (50μM) and briefly applied directly to exposed left cortical surface. After surface application (5 min), the exposed cortical surface was rinsed repeatedly with pre-warmed saline (5 min). A glass coverslip (#1.5, 5 mm) was centered over the cranial window and attached to the skull with dental acrylic. For vascular architecture and BBB permeability labeling, 200 μl fluorescein isothiocyanate–dextran (FITC, 4KDa, Sigma) was injected intraperitoneally (i.p., 10mg/kg) 15 min before starting the imaging experiment.

Images of astrocytes and capillaries in the cortex or extracellular dextran (as an indicator of BBB leakage) were obtained through the cranial window with a two-photon microscope (LSM 510, Zeiss, Thornwood, NY) and a water immersion objective (Zeiss, 40×0.8 numerical aperture (NA), IR-adjusted, Zeiss). A Titanium-Sapphire pulsed infrared laser (Coherent, Santa Clara, CA) was used to stimulate SR101 (red) and FITC (green) at 860 nm. Low-magnification images 50 −100 μm below the neocortical surface were first obtained to identify regions of interest (ROI) containing SR101 positive astrocytes and FITC filled capillaries. At higher magnification (3x digital zoom), z-stacks of 6 −10 images with 1 μm - step increment were taken. Individual image was acquired at 12 bits with frame averaging (2–4 times). The same excitation laser power and acquisition settings (e.g., detection gain, amplifier offset, amplifier gain) were maintained in all animals for standardized comparison.

Post hoc image analysis was performed using LSM 5 Image Examiner software (Zeiss) and Image J software (NIH) in a blinded fashion to evaluate the changes in astrocyte structure (astrocyte end-feet) and blood-brain barrier (BBB) permeability in apatinib, RepSox and vehicle treated mice. BBB permeability was reflected by extravascular fluorescent signal intensity of FITC. For measuring the extravascular fluorescent signal intensity of FITC, 4-6 areas of 150 μm ×150 μm were chosen as the ROI for each mouse. For measuring the ratio of astrocyte end-feet to the capillary vascular length, 5-7 areas of 40 μm ×100 μm with only one single capillary were chosen as the ROI for each mouse (Supplemental Fig. S1).

2.7. Video-EEG Monitoring

Vehicle, apatanib, or RepSox-treated Tsc1^GFAPCKO mice underwent video-EEG monitoring starting at 4 weeks of age, using established methods for implanting epidural electrodes and performing continuous video-EEG recordings, as described previously ^{15, 22}. Briefly, three-week old mice were anesthetized with isoflurane and placed in a stereotaxic frame. Epidural screw electrodes were surgically implanted and secured using dental cement for long term EEG recordings. Four electrodes were placed on the skull: one right and one left central-parietal electrodes (1 mm lateral to midline, 2 mm posterior to bregma), one frontal electrode (0.5 mm anterior and 0.5 mm to the right or left of bregma) and one occipital electrode (0.5 mm posterior and 0.5 mm to the right or left lambda). A referential recording montage involved two EEG channels with the right and left central-parietal “active” electrodes being compared to either the frontal or occipital “reference” electrode. Video and EEG data were acquired simultaneously with an AD Instruments PowerLab video-EEG system. The entire EEG record was analyzed for electrographic seizures by a blinded reviewer, with video confirmation to rule out artifacts and confirm clinical seizure activity. Electrographic seizures were identified by their characteristic pattern of discrete periods of rhythmic spike discharges that evolved in frequency and amplitude lasting at least 10 s, typically ending with repetitive burst discharges and voltage suppression. On video analysis, the behavioral correlate to these seizures typically involved head bobbing, rearing with forelimb clonus, and occasional generalized convulsive activity. Seizure frequency (number of seizures per week period, based on analysis of the entire EEG record) was calculated from each weekly epoch. The natural history of epilepsy in Tsc1^GFAPCKO mice have been well-described^{14, 15, 22, 23}, with onset of seizures around 4 weeks and progressive increase in seizure frequency and premature death over the following 4-6 weeks, including increased clustering of seizures with decreased mice numbers that cause increased variability.

2.8. Statistics

All statistical analysis was performed using GraphPad Prism 10.0 (GraphPad Software). For western blot studies, immunohistochemistry/histological studies, and in vivo imaging studies of extravascular FITC intensity and astrocyte end-feet length, quantitative differences between groups were analyzed by one-way ANOVA with Tukey’s multiple comparisons post hoc tests for comparisons between three groups and two-way ANOVA with Tukey’s multiple comparisons post hoc tests for comparisons with multiple groups and treatments. Comparable non-parametric tests were used when data did not fit a normal distribution. For analysis of seizure frequency over time, as missing values from premature death prevented two-way ANOVA with repeated measures, multiple t-tests were used. The Log-rank (Mantel-Cox) test was used for survival analysis. Quantitative data are expressed as mean ± SEM. Statistical significance was defined as p < 0.05.

3. Results

3.1. VEGF expression is increased in Tsc1^GFAPCKO mice and is mTOR-dependent.

VEGF-A expression was measured by western blotting in four week old Tsc1^GFAPCKO mice. VEGF expression was increased in cortex of Tsc1^GFAPCKO mice compared with control mice (Fig. 1), consistent with previous studies of VEGF immunohistochemistry in Tsc1^GFAPCKO mice ⁸. Furthermore, this difference was reversed with rapamycin, indicating the increased VEGF expression was mTOR-dependent. Similarly, HIF1α, an upstream regulator of VEGF, was elevated in Tsc1^GFAPCKO mice and was also reversed by rapamycin.

Figure 1. VEGF and HIF1α are upregulated in Tsc1^GFAPCKO mice and inhibited by rapamycin.

Protein expression of VEGF and HIF1α were assessed by Western blotting in the cortex of four weeks old Tsc1^GFAPCKO mice. (A) Representative Western blots of VEGF, HIF1α and β-actin. Vehicle-treated Tsc1^GFAPCKO mice (KO + Veh) have significantly increased VEGF (B) and HIF1α (C) levels in the cortex, compared with control mice (Cont + Veh). Rapamycin treatment (6 mg/kg/d, i.p. for seven days; KO + Rap) significantly inhibited the upregulated-VEGF and HIF1α in Tsc1^GFAPCKO mice. * p<0.05, ** p<0.01, by one-way ANOVA (n = 6 mice/group; male – black symbol, female – gray symbol). Cont = control, KO = Tsc1^GFAPCKO, Veh = vehicle, Rap = rapamycin.

3.2. Tsc1^GFAPCKO mice exhibit hypervascularization which is VEGF-dependent

To get an initial assessment of angiogenesis, immunohistochemistry of the vascular endothelial cell marker, CD31, was performed in four week old Tsc1^GFAPCKO mice. CD31 expression was increased in cortex and hippocampus of Tsc1^GFAPCKO mice, compared with control mice (Fig. 2). Quantitative analysis of linear measurements of CD31 expression along cerebral blood vessels confirmed an increase in blood vessel staining, indicating increased angiogenesis in Tsc1^GFAPCKO mice. The VEGF-2 receptor antagonist, apatinib, blocked the increased CD31 expression in Tsc1^GFAPCKO mice, suggesting that the increased VEGF expression was driving the increased angiogenesis.

Figure 2. CD31 is upregulated in Tsc1^GFAPCKO mice and inhibited by Apatinib.

Protein expression of CD31 was assessed by immunohistochemistry in the brains of four week old Tsc1^GFAPCKO mice. (A) Representative confocal images of immunohistochemical staining of CD31 (green) and TO-PRO-3 Iodide (blue). TO-PRO-3 Iodide was used as an optimal fluorescence dye for nuclear counterstaining. Relatively stronger CD31 expression was detected in both cortex (upper panel) and hippocampus (lower panel) in the brain sections of Tsc1^GFAPCKO mice (KO + Veh), compared with control mice (Cont + Veh). Apatinib (75 mg/kg/d, i.p. for one week, KO + Apa) inhibited the CD31 expression. (B) Quantitative analysis confirmed an increase in CD31-positive staining in vehicle-treated Tsc1^GFAPCKO group (KO + Veh) compared with vehicle-treated control group (Cont + Veh), by using measurement of the length of CD31 positive staining. Apatinib (KO + Apa) significantly decreased CD31 expression. *** p<0.001 vs. vehicle-treated control mice by one-way ANOVA (n = 6 mice/group; male – black symbol, female – gray symbol). Scale bar = 100 μm. Cont = control, KO = Tsc1^GFAPCKO, TO PRO3 = TO-PRO-3 Iodide, Veh = vehicle, Apa = apatinib.

To more directly assess vascular anatomy, cerebral blood vessels were imaged after transcardial perfusion with Alexa Fluor 488-conjugated dextran in four week old Tsc1^GFAPCKO mice. The density of blood vessels in the cortex and hippocampus, as well as the total length and branching of vessels, was increased in Tsc1^GFAPCKO mice compared with control mice (Fig. 3). These vascular abnormalities were blocked by apatinib, again indicating a dependence on VEGF. While Tsc1^GFAPCKO mice exhibit megalencephaly and increased cortical thickness, it is important to note that the above vessel abnormalities were significant after normalizing for cortical area, so were not simply due to having increased cortex.

Figure 3. Hypervascularization displayed in Tsc1^GFAPCKO mice and reversed by Apatinib.

(A) Representative images of Dextran-AF488-filled blood vessels in coronal sections of cortex and hippocampus (dentate gyrus) of four-week old vehicle-treated control mice (Cont + Veh), vehicle-treated Tsc1^GFAPCKO mice (KO + Veh), and Apatinib-treated Tsc1^GFAPCKO mice (75 mg/kg/d, i.p. for one week, KO + Apa). (B-E) Quantitative analysis of different vessel parameters, including vessel density (B), normalized junction density (C), normalized total vessel length (D), and normalized lacunarity (E). *p < 0.05, ** p < 0.01, versus vehicle-treated Tsc1^GFAPCKO mice by one-way ANOVA (n = 6 mice/group; male – black symbol, female – gray symbol). Scale bar = 100 μm. Cont = control, KO = Tsc1^GFAPCKO, Veh = vehicle, Apa = apatinib.

3.3. Tsc1^GFAPCKO mice have increased BBB permeability, which is VEGF-dependent

As VEGF can increase vascular permeability, BBB permeability was assessed in Tsc1^GFAPCKO mice using in vivo two-photon imaging to monitor for the leakage of FITC-conjugated dextran into the cerebral extravascular space. In control mice, cerebral blood vessels were filled with FITC, but there was no obvious extravascular leakage of fluorescence (Fig. 4Ab, h, n). In contrast, Tsc1^GFAPCKO mice displayed extravascular fluorescence detected in focal areas surrounding cerebral blood vessels (Fig. 4Ae, B). The increased extravascular fluorescence in Tsc1^GFAPCKO mice was blocked by the VEGF antagonist, apatinib, again indicating VEGF dependence (Fig. 4Ak, B). Furthermore, RepSox, a TGF-beta antagonist that stabilizes BBB function, also inhibited this extravascular fluorescence increase in Tsc1^GFAPCKO mice (Fig. 4Aq, B).

Figure 4. In vivo imaging of abnormal blood-brain barrier function and structure in Tsc1^GFAPCKO mice.

(A) Two-photon in vivo imaging of systemically-injected FITC-conjugated dextran (green) and SR101-labeled astrocytes (red) assessed BBB permeability. In vehicle-treated, Apatinib-treated and RepSox-treated control mice (Cont + Veh, Cont + Apa, Cont + Rep), cerebral blood vessels were filled with FITC (green), but there was no evidence of extravascular fluorescence (Ab, h, n). In vehicle-treated Tsc1^GFAPCKO mice (KO + Veh), extravascular FITC leaked around the capillaries (Ae), which was reversed by apatinib treatment (KO + Apa) (Ak) and RepSox treatment (KO + Rep) (Aq). Two-photon in vivo imaging of astrocytes end-feet (red) surrounding cortical cerebral capillaries demonstrated that end-feet structure was disrupted around capillaries in KO + Veh mice (Af), as compared with control mice (Cont + Veh, Cont + Apa, Cont + Rep)(Ac, i, o). Apatinib and RepSox treatment had no effect on the recovery of astrocytes end-feet structure (Al, r), compared with vehicle-treated Tsc1^GFAPCKO (KO + Veh) mice (Af). (B) Quantitative analysis of extravascular fluorescence demonstrates increased BBB permeability in vehicle treated Tsc1^GFAPCKO mice (KO + Veh), compared with vehicle, Apatinib or RepSox treated control mice (Cont + Veh, Cont + Apa and Cont + Rep) mice and Apatinib or RepSox treated Tsc1^GFAPCKO mice (KO + Apa, KO + Rep). * p<0.05 by two-way ANOVA with Tukey’s multiple comparison test. Cont + Veh: n = 35 ROIs in 8 mice; Cont + Apa: n=41 ROIs in 8 mice; Cont + Rep: n=42 ROIs in 7 mice; KO + Veh: n=35 ROIs in 8 mice; KO + Apa: n=32 ROIs in 8 mice; KO + Rep: n=36 ROIs in 7 mice. (C) Quantitative analysis of the ratio of astrocyte end-feet to vascular length reveals a significant decrease in KO + Veh mice, KO + Apa and KO + Rep mice compared with control mice (Cont + Veh, Cont + Apa, Cont + Rep). No significant difference was found among KO + Veh, KO + Apa and KO + Rep mice. *p<0.05 by two-way ANOVA with Tukey’s multiple comparison test. Cont + Veh: n=47 capillaries in 8 mice; Cont + Apa: n=51capillaries in 8 mice; Cont + Rep: n=45 capillaries in 7 mice; KO + Veh: n=46 capillaries in 8 mice; KO + Apa: n=46 capillaries in 8 mice; KO + Rep: n=47 capillaries in 7 mice. See also Supplemental Fig. S2 for dot-plots and comparison of analysis based on individual mice.

As astrocyte end-feet may also contribute to BBB maintenance and Tsc1^GFAPCKO mice have a number of astrocyte abnormalities, in vivo two-photon imaging was also used to assess the relationship of astrocyte end-feet, as visualized by the locally-applied astrocytic stain SR101, to surround FITC-labeled cerebral capillaries. Compared with control mice, Tsc1^GFAPCKO mice exhibited more disruption of astrocyte end-feet structure surrounding blood vessels, with a decreased astrocyte end-feet to vascular length ratio (Fig. 4Ac, f, C). However, apatinib or RepSox had no significant effect in reversing this structural abnormality (Fig 4Al, r, C). For both the BBB permeability and astrocyte end-feet analyses, multiple samples (n=4-7/mouse) were taken per mouse. When analysis was re-done based on mouse averages (n=7-8 mice/group) to assess if significant differences were attributable to individual mice, the variability between mice was relatively low and the main statistical differences between knock-out and control groups were still significant, though there was a non-significant trend in the effects of apatinib and RepSox in reversing these differences (Supplemental Fig. S2).

3.4. Tsc1^GFAPCKO mice have decreased BBB-related tight junction proteins, which are VEGF-dependent.

Tight junction proteins linking cerebral vascular endothelial cells are also key structural components of the BBB and potential targets for BBB modulation. The protein expression of the tight junction protein, occludin, was decreased in cortex or hippocampus of Tsc1^GFAPCKO mice (Fig. 5A, C). Another tight junction protein, claudin-5, was also decreased in cortex, but not hippocampus in Tsc1^GFAPCKO mice (Fig. 5B, D). In contrast, ZO-1 expression was unchanged in both hippocampus and cortex of Tsc1^GFAPCKO mice (Supplemental Fig. S3). The VEGF antagonist, apatinib, reversed the decreased expression of occludin and claudin-5 in Tsc1^GFAPCKO mice (Fig. 5A, B). In addition, RepSox, the BBB stabilizer, also reversed the decreased expression of occludin and claudin-5 in Tsc1^GFAPCKO mice (Fig. 5C, D).

Figure 5. Expression of tight junction proteins, occludin and claudin-5, is decreased in Tsc1^GFAPCKO mice and reversed by the VEGF antagonist, Apatinib and the BBB modulator, RepSox.

Occludin and claudin-5 expression in the hippocampus and cortex was assessed by Western blotting. Occludin levels in both hippocampus and cortex (A, C) and claudin-5 levels in cortex, but not hippocampus (B, D), were significantly decreased in vehicle-treated Tsc1^GFAPCKO (KO + Veh) mice compared with control mice (Cont + Veh and Cont+Apa). The decrease in occludin and claudin-5 levels were reversed by Apatinib (75 mg/kg/d, i.p. for one week) treatment of Tsc1^GFAPCKO mice (KO + Apa) (A,B). Similarly, the decrease of occludin and claudin-5 levels were also reversed by RepSox (10mg/kg/d, i.p. for one week) treatment of Tsc1^GFAPCKO mice (KO + Rep) (C, D). * p<0.05 by two-way ANOVA with Tukey’s multiple comparison test for A,B and one-way ANOVA for C,D. A: Cont + Veh: n=9 mice; Cont + Apa: n=9 mice; KO + Veh: n=10 mice; KO+ Apa: n=9 mice. B: Cont + Veh: n=11 mice; Cont + Apa: n=11 mice; KO + Veh: n=12 mice; KO+ Apa: n=11 mice. C and D: Cont + Veh: n=8 mice; KO + Veh: n=9 mice; KO+ Rep: n=10 mice. male – black symbol, female – gray symbol.

3.5. Effect of VEGF and BBB modulators on epilepsy and survival of Tsc1^GFAPCKO mice.

To investigate the potential functional consequences of theses vascular and BBB abnormalities, the effect of treatment with modulators of VEGF and BBB was assessed on seizures and survival in Tsc1^GFAPCKO mice, which have a well-described development of seizures around 4 weeks of age that increases in frequency over the subsequent 4-6 weeks and is associated with premature death ^{15, 22, 23}. Treatment with the VEGF receptor antagonist, apatinib, starting at 3 weeks of age (prior to onset of seizures) at the same dose effective in inhibiting the vascular abnormalities had no significant effect on seizure frequency, duration, latency to first seizure (i.e., age at first seizure), or survival in Tsc1^GFAPCKO mice compared with vehicle-treated Tsc1^GFAPCKO mice (Fig. 6B–E). However, the direct BBB stabilizer, RepSox increased the latency to first seizure, decreased seizure frequency and improved survival compared with vehicle-treated Tsc1^GFAPCKO mice, but had no effect on seizure duration (Fig. 6F–I). To evaluate whether RepSox might have off-target effects, such as on the mTOR pathway, in contrast to the effects of rapamycin¹⁵, RepSox was found to have no effect on PS6 expression (Supplemental Fig. S4), gross brain weight/pathology and body weight (Supplemental Fig. S5), and GFAP expression (Supplemental Fig. S6), indicating that the effects of RepSox on seizures were likely related to its documented effects on BBB function.

Figure 6. The BBB stabilizer, RepSox, but not the VEGF antagonist, Apatanib, decreases seizures and improves survival in Tsc1^GFAPCKO mice.

Tsc1^GFAPCKO mice underwent continuous video-EEG monitoring from four weeks of age. (A) Representative seizure on EEG of a Tsc1^GFAPCKO mouse. (B-E) Apatinib treatment (75 mg/kg/d, i.p., KO + Apa), starting at three weeks of age, had no effect on seizure frequency (B), seizure duration (C), and latency to first seizure/age at first seizure (D) and did not improve survival of Tsc1^GFAPCKO mice (E) (p>0.05 by t-test in B-D, p>0.05 by Log-rank Mantel-Cox test in E, KO + Veh: n=18 mice; KO + Apa: n=15 mice). (F-I) In contrast, RepSox treatment (10 mg/kg/d, i.p.; KO + Rep), starting at three weeks of age, decreased seizure frequency at 4-7 weeks of age (F), increased latency to first seizure (H), and improved survival (I), but had no effect on seizure duration (G), compared with vehicle-treated Tsc1^GFAPCKO mice (KO + Veh). (*p < 0.05, by t-test in F-H *p<0.05 by Log-rank Mantel-Cox test in I; n = 8 mice per group). Note that group size decreased over time as indicated in B, F, E, and I. Male – black symbol, female – gray symbol.

4. Discussion

In this study, we have investigated the role of VEGF and cerebral vascular abnormalities in a mouse model of TSC. VEGF expression was increased in the cortex of

Tsc1^GFAPCKO mice and was reversed by rapamycin, indicating its mTOR-dependence. Tsc1^GFAPCKO mice exhibited excessive cerebral angiogenesis, with increased vascular density and branching. These vascular abnormalities were blocked by a VEGF receptor antagonist, supporting that the elevated VEGF is involved in driving these vascular changes. Furthermore, Tsc1^GFAPCKO mice exhibited increased BBB permeability and decreased BBB-associated tight junction protein expression, which were also dependent on VEGF. Finally, a BBB stabilizer, RepSox, reversed the BBB abnormalities, decreased seizures and improved survival in Tsc1^GFAPCKO mice.

Elevated VEGF expression has previously been reported in cortical tuber specimens from TSC patients, as well as in immunohistochemical studies in cortex of Tsc1^GFAPCKO mice ⁸. The western blot studies in the present study confirm these results in Tsc1^GFAPCKO mice and extend the previous findings by performing more quantitative analysis and demonstrating their dependence on mTOR with reversal by rapamycin. As mTOR hyperactivation due to TSC gene mutations is implicated as a central driver of tumor growth and mTOR inhibitors are effective treatments for tumors in TSC ^{24, 25}, it is perhaps not surprising that VEGF expression in the brain is also mTOR-dependent. However, the precise cellular source of this brain VEGF expression is not entirely clear. Previous immunohistochemical studies indicate that both neurons and astrocytes express VEGF in Tsc1^GFAPCKO mice ⁸, which is consistent with the fact that Tsc1 inactivation and mTOR hyperactivation occurs in both neurons and astrocytes in Tsc1^GFAPCKO mice.

As a major function of VEGF is to stimulate angiogenesis, it is also rational that Tsc1^GFAPCKO mice exhibit increased cerebral vascular density and branching. Similar vascular abnormalities were previously reported in other TSC mouse models^{11, 12} and were reversed by rapamycin¹¹, again indicating mTOR dependence, though the relationship to VEGF was not investigated . The reversal of these vascular abnormalities in Tsc1^GFAPCKO mice by a VEGF antagonist demonstrates for the first time that cerebral blood vessel anomalies are VEGF-dependent in a TSC brain model. mTOR- and VEGF-dependent stimulation of angiogenesis likely drives the vascular components of TSC-related tumors, such as renal angiomyolipomas and pulmonary lymphangioleimyomatosis ^{6, 7}. However, the functional or pathophysiological consequences of vascular proliferation in the brain are not known, as discrete vascular structural lesions are not as obvious in the brain of TSC patients.

Besides stimulating angiogenesis per se, VEGF has a number of other functions in the brain, such as regulation of neuronal and glial proliferation, differentiation, migration, and synaptic plasticity ^{4, 5}. In addition, independent of angiogenesis, VEGF may increase vascular permeability and decrease endothelial tight junction proteins, which may lead to breakdown of the BBB ^{9, 10, 26}. mTOR activation in itself has also been shown to stimulate BBB breakdown and mTOR inhibition can decrease BBB leakage ²⁷. Breakdown of the BBB has been shown to trigger a cascade of events that can trigger epileptogenesis and other neurological sequelae ^{28, 29}. The current study demonstrates BBB dysfunction in a TSC mouse model for the first time, as well as structural and molecular changes that potentially underlie BBB dysfunction, in astrocyte end-feet structure and vascular endothelial tight junction protein expression. Furthermore, the BBB stabilizer, RepSox, decreased seizures and improved survival (possibly related to the decrease in seizures and associated SUDEP). As BBB breakdown promotes epilepsy in other models ^{18, 28, 29}, the present findings indicate that BBB dysfunction contributes to epileptogenesis and BBB stabilizers may serve as rational treatment for epilepsy in TSC.

Limitations of this study include potential off-target effects or inconsistences of the pharmacological modulators used. VEGF, and VEGF antagonists, may have numerous other effects that are independent of the BBB. As apatinib reversed vascular abnormalities, but did not affect seizures, in Tsc1^GFAPCKO mice, the precise relationship between vascular structure and BBB function to seizures may not be completely uniform or overlapping. Similarly, the TGF-β antagonist, RepSox, could have other effects independent of BBB regulation, that could contribute to its antiseizure effect and improved survival. However, while upstream inhibition of the mTOR pathway by rapamycin has dramatic effects in preventing epilepsy, reversing pathological abnormalities/megalencephaly, and prolonging survival in Tsc1^GFAPCKO mice ^{15, 30}, RepSox did not inhibit mTOR activity or have gross pathological effects on brain weight, indicating that its antiseizure effects are distinct from rapamycin and likely involve BBB modulation. Besides distinctive mechanisms of action, differences between apatinib and RepSox in effects on seizures could also be related to pharmacokinetic properties or compensatory downstream or upstream effects. Future studies will investigate other pharmacological or genetic methods for specifically targeting and manipulating vascular and blood-brain mechanisms in this mouse model to provide additional evidence of mechanistic specificity.

Key Points:

• Vascular endothelial growth factor is increased in cortex of a mouse model of TSC and blocked by a mechanistic target of rapamycin inhibitor

• Increased angiogenesis and structural vascular abnormalities are prominent in cortex and hippocampus of Tsc1^GFAPCKO mice

• Blood-brain barrier permeability is increased and BBB-related tight junction proteins are correspondingly decreased in Tsc1^GFAPCKO mice

• A VEGF receptor antagonist reverses the vascular and BBB abnormalities in Tsc1^GFAPCKO mice, but has no effect on seizures or survival

• A BBB stabilizer, RepSox, reverses BBB abnormalities, decreases seizures and improves survival in Tsc1^GFAPCKO mice