Conditional deletion of the focal adhesion kinase FAK alters remodeling of the blood-brain barrier in glioma

Abstract

Gliomas generally infiltrate the surrounding normal brain parenchyma, a process associated with increased vascular permeability (VP) and dysregulation of the blood-brain barrier (BBB). However, the molecular mechanisms underlying glioma-induced VP in the brain remain poorly understood. Utilizing a conditional, endothelial-specific deletion of the focal adhesion kinase FAK in the mouse (FAK CKO), we show that FAK is critical for destabilization of the tumor endothelium in tumor-bearing mice, with mutant mice exhibiting a relatively stabilized vasculature to wild-type mice (FAK WT). Tumor vessels in the FAK CKO mice displayed reduced VP compared to FAK WT mice, resulting in reduced tumor growth. Additionally, FAK CKO mice displayed partial restoration of cell-cell junction proteins in the tumor vessels and astrocyte-endothelial interactions in tumors, revealing an additional role of astrocytes in mediating tumor-induced VP. Together, these results provide genetic evidence that FAK is a mediator of tumor-induced VP in the brain. Our findings may help understand how therapeutics might be used to regulate cell type-specific interactions to restore BBB structure/function in cancer and perhaps other pathological conditions.

Introduction

In the normal brain, many different cell types interact with each other to maintain a tight barrier between the blood and brain parenchyma, which is also known as the BBB. The BBB consists of tight inter-endothelial cell junctions, surrounded by astrocyte endfeet, and separated by a basal lamina. The integrity of normal BBB integrity is compromised in pathological conditions, such as it occurs in the growth and progression of malignant glioma (1). Several studies have shown that gliomas alter specific cell-type interactions, leading to dysregulation of the BBB (i.e. BBB breakdown) leading to an increase in VP (2, 3). For example, our previous studies have shown that gliomas induce remodeling of the BBB in a cell-type specific manner (4) and that alterations of specific host components regulates tumor-mediated VP and tumor invasion(3, 5).

To identify more selective mediators of the BBB in terms of function and cell-type, we have focused on the role of a key downstream target of Src, FAK, (6) in normal adult brain and tumor-bearing brain. FAK is enriched in brain blood vessels compared to surrounding cell types, phosphorylated in response to VEGF(7–11), and forms signaling complexes with Src and integrins in VEGF-induced signaling (8, 12, 13). Standard knockout approaches have determined that FAK-deleted embryos have severely impaired blood vessel development (14, 15). Subsequent endothelial-cell specific deletion of FAK also led to an embryonic lethal phenotype, indicating a functional role of FAK in the development of the vascular endothelium in particular (14, 16–18). More recently, PYK2 has been shown to have a compensatory role for FAK (19), while a differential role of kinase-independent and dependent functions of FAK in vascular development of embryogenesis have been reported (20, 21). Clearly there remains a need to better understand cell- and tissue-type specific mechanisms regulating vascular integrity and the FAK conditional knockout mouse is one such model to define the functional relevance of FAK in vivo.

Materials and Methods

Mice

FAK deletion in adult endothelia was achieved by tamoxifen administration to FAK^flox/flox mice (a kind gift from Dr. H. Beggs (22)) crossed with Tie2-Cre/ER^T2 (tamoxifen-inducible Cre activation under the control of endothelial-specific Tie2 promoter) mice (a kind gift from Dr. J. Esko (23)) and backcrossed into Rag2^−/−, immunodeficient background (FAK CKO). Littermates that were null for Cre gene were used as controls (FAK WT). 10-week old FAK CKO and FAK WT mice were injected intraperitonealy with 2mg tamoxifen (Sigma, St.Louis, MO) for five consecutive days to induce recombination. All animal handling procedures were approved by the University of California San Diego Institutional Animal Care and Use Committee.

Glioma tumor cells

Early passages of patient-derived human glioma cells, DBTRG (a kind gift from Dr. C. Kruse)(24) were used for xenograft studies. DBTRGs transduced with lentivirus expressing firefly luciferase (DBTRG-luc) or red fluorescent protein (DBTRG-RFP) were generated as described earlier (4).

In vivo bioluminescent imaging

Fur was removed from mice with electric clippers and Nair (Church & Dwight Co., Inc., Princeton, NJ) before imaging at each time point. Bioluminescent signals were assessed 10 minutes after D-luciferin injection (at the steady state of luminescent signal) using a cooled charge-coupled device (CCD) camera (Spectrum; Caliper Life Sciences, Hopkinton, MA) capable of in vivo imaging. Tumor growth was monitored by quantitation of light emission from a region of interest drawn over the tumor region at each time point (Unit = radiance). Images were analyzed using Living Image software version 3.1 (Caliper Life Sciences, Hopkinton, MA).

Biochemical analyses in isolated endothelial cells

Mouse brain microvessels were isolated from tamoxifen treated FAK WT and FAK CKO mice based on previously published protocols (25–28). Mice were injected with 1mg of FITC-lectin (FL 1101-5, Vector Labs, Burlingame, CA) 20 minutes prior to sacrifice to validate the integrity of isolated vessels. Genomic DNA was purified from isolated microvessels (QIAGEN, Valencia, CA) and used as templates for PCR analysis as described earlier (29). The following antibodies were used to validate downregulation of FAK in isolated microvessels preps; anti-FAK (C-20, BD Transduction Laboratories, San Jose, CA, 1:1,000), anti-β-actin (Cell Signaling Technology, Danvers, MA, 1:500), anti-synaptophysin (Abcam, Cambridge, MA, 1:1,000), anti-AQ4 (Millipore, Bedford, MA, 1:500), anti-P-glycoprotein (Abcam, Cambridge, MA,1:250).

Vascular Permeability Assay

Mice were injected with FITC-dextran 70kDa (Sigma, St.Louis, MO), 20mins later mice were subjected to systemic intracardiac perfusion with 1 USP unit/ml of heparin dissolved in saline. 1mm thick brain sections were imaged to measure vascular permeability (FITC) and tumor burden (RFP) with a deep cooled CCD imaging system equipped with appropriate fluorescence filter cubes with background subtraction and images were analyzed using Living Image software version 3.1 (Lumina, Caliper Life Sciences, Hopkinton, MA).

Immunohistochemistry

Standard immunohistochemistry was performed using the following primary antibodies at the following dilutions. Anti-CD31 (553370, BD Biosciences, San Jose, CA, 1:100), rabbit polyclonal anti-laminin(L9393, Sigma, St.Louis, MO, 1:1200), mouse monoclonal anti-GFAP (C9205, Sigma, St.Louis, MO, 1:200), rabbit polyclonal anti-aquaporin 4 (Millipore, Bedford, MA,1:100), rabbit polyclonal ZO-1 (Zymed, San Francisco, CA, 1:100), rabbit polyclonal occludin (Invitrogen, Carlsbad, CA, 1:100). Alexa-fluor-conjugated secondary antibodies were used (Molecular Probes, Eugene, OR, 1:200). All sections were counterstained with DAPI. Immunostaining of tissue sections were imaged with an Olympus Fluoview 1000 (ASW 1.7b) laser scanning confocal microscope equipped with 10×/0.4N.A. or 20×/0.7N.A. dry objective lenses on a BX61 microscope (Olympus, Melville, NY).

Quantitation of vessel morphology

Tumor-bearing mice were injected with 1mg of FITC-lectin (FL 1101-5, Vector Labs, Burlingame, CA) 20 minutes prior to sacrifice to quantify vessel diameters of FAK WT and FAK CKO mice. Vessel diameters were measured using Olympus Fluoview 1000 (ASW1.7b) software. To measure vascular density, immunohistochemistry was performed on brain sections with tumor using anti-CD31 antibody and Alexa-flour-conjugated secondary antibody as described above. Numbers of CD31-positive vessels were counted per unit area (630µm×630µm)

Results

Validation of the conditional and endothelial specific knockout of FAK in endothelium

To achieve conditional and endothelial specific knockout of FAK, mice with the second kinase domain exon of FAK locus flanked by lox P sites (FAK^flox/flox) mice (22) were crossed with transgenic mice expressing tamoxifen-inducible Cre under the regulation of the Tie2 promoter (Tie2-Cre/ER^T2, i.e. endothelial-specific)(23). These mice were then backcrossed onto an immunodeficient background for the purpose of xenografting human brain tumor cells. Using these FAK^flox/flox; Tie2-Cre/ER^T2 ;Rag2^−/− (FAK CKO) mice, we first validated efficient recombination following a timecourse of tamoxifen administration based on previous studies (29). In consideration of the potential effects of tamoxifen in this model, control mice were also subjected to tamoxifen but lacked the Cre transgene with a genotype of FAK^flox/flox;Rag2^−/− and are referred to as FAK WT. FAK deletion in this model was assessed by immunoblotting isolated brain microvessels from FAK CKO and FAK WT mice using an established technique for the biochemical isolation and analysis (26–28). Perfusion with fluorescent-labeled lectin 20 minutes prior to sacrifice, followed by imaging on a fluorescent microscope verifies the integrity of isolated microvessels (Figure 1A). Compared to the whole brain lysates, we observed an enrichment of an endothelial cell marker, P-glycoprotein, and an astrocyte endfoot marker, Aquaporin 4, whereas a decrease in a neuronal marker, Synaptophysin (Figure 1B)(28). Using isolated microvessels from FAK CKO and FAK WT mice, we observed that recombination occurred as early as 3 days after tamoxifen treatment (data not shown). Recombination of FAK gene was detected in Cre-expressing FAK CKO mice, while no recombination was detected in mice lacking the Cre transgene (Figure 1C) or mice treated with vehicle only (data not shown). The expression of FAK protein is downregulated more than 5.8 fold at 7 days after tamoxifen treatment in isolated microvessels (Figure 1D). We did not observe complete deletion of FAK gene (Figure 1C upper band) or complete removal of the FAK protein (Figure 1D), which can be partially explained by the isolation of capillary-associated astrocytes endfeet (AQ4, Figure 1B) together with the microvessels. Importantly, we did not observe downregulation of FAK expression in whole brain lysates, indicating endothelial-specific decrease of FAK (Figure 1D). FAK-deletion in the adult endothelium of FAK CKO mice did not result in any apparent morphological changes of the vessels or changes in vascular permeability in non-tumor bearing brains (Supplementary Data, Figure 1). Moreover, we have not noticed any changes of the vessels in non-tumor bearing brains of FAK CKO mice or FAK WT mice compared to normal brain (with no tamoxifen treatment), indicating no side-affect of tamoxifen treatment in our model (Supplementary Data, Figure 1A).

Figure 1. Validation of the conditional and endothelial specific knockout of FAK in endothelium.

(A) Brain microvessels from tamoxifen-treated mice were isolated to verify efficient recombination of FAK locus. Brains were perfused with FITC-lectin before isolation of microvessels to validate their integrity. Size bar = 20µm

(B) The purity of isolated microvessels was verified by immunoblotting for the enrichment of an endothelial cell marker (i.e. P-glycoprotein) and an astrocyte endfoot marker (i.e. Aquaporin 4) and decrease of a neuronal marker (i.e. Synaptophysin) in lysates from microvessels compared to their levels detected in whole brain. Equal amount of total protein (20µg) was loaded per lane.

(C) Efficient Cre-mediated recombination and FAK deletion was assessed by PCR analysis of genomic DNA from microvessels isolated from tamoxifen treated brains. Results from 7 days after tamoxifen administration is shown. Amplification of smaller 328bp product (FAK CKO, arrow) indicates successful recombination where control mice show amplification of endogenous FAK gene (1.6kb, arrowhead) with no detectable recombination (FAK WT).

(D) A decrease in FAK protein in lysates of brain microvessels of FAK CKO mice was verified by Western blot analysis. A 6-fold decrease in FAK protein levels in FAK CKO brain microvessels was observed compared to those in FAK WT mice (n=4–6 each) 7 days post-tamoxifen treatment. We did not observe downregulation of FAK expression in whole brain lysates, indicating endothelial-cell specific deletion of FAK. Western blotting for β-actin was used as a loading control. (p<0.05, Wilcoxon rank sum test, two-sided, error bar = standard deviation)

Effect of FAK deletion in host compartment (i.e. endothelium) on tumor growth

In vitro and in vivo studies have demonstrated FAK as a mediator of VEGF signaling in the endothelium (7–11), however whether FAK is necessary for VEGF-mediated VP in brain tumor growth is unknown. To determine whether an absence of FAK in the host compartment vascular endothelium affects the growth of orthotopic tumor cells (i.e. tumor compartment), we stereotaxically injected firefly luciferase-tagged human glioma cells (DBTRG-luc) that are known to express VEGF (3) into the brains of FAK CKO and FAK WT mice. Mice were non-invasively imaged to detect the bioluminescence of the luciferase-labeled tumor cells as described in the Materials and Methods. Earlier studies have reported the tight correlation between the noninvasive, bioluminescent signal of luciferase-labeled tumor cells with caliper measured tumor volume in different organs including the brain (30–32). While rapid tumor growth was observed in FAK WT host from two to four weeks, tumor growth in the FAK CKO mice was reduced (Figure 2). By day 28 post-injection, we observed a 15-fold reduction of tumor burden in FAK CKO mice compared to FAK WT mice (p=0.002). We observed similar results with ex vivo measurements of bioluminescence from 1mm tumor-bearing brain sections or by standard hematoxylin and eosin staining of tumor tissue (Supplementary Data, Figure 2). In most cases, tumors of FAK CKO mice eventually grow to the same size of the FAK WT mice, but only with a longer incubation time (data not shown). These results suggest that FAK is essential in the endothelium of tumor-associated blood vessels in vivo.

Figure 2. Effect of FAK-deletion in host compartment (i.e. endothelium) on tumor growth.

(A) A schematic diagram of the time course of tamoxifen treatment and non-invasive imaging of the implanted tumor is shown.

(B) DBTRG-luc tumor cells were implanted by stereotaxic intracranial injection into FAK WT and FAK CKO mice. Tumor growth was monitored non-invasively for four weeks post-implantation. Exposure matched, representative images taken at each time point is shown.

(C) Quantitation of the DBTRG-luc tumor growth over a 28-day timecourse is shown from FAK WT vs. FAK CKO mice. Total tumor luciferase signal (radiance) at each time point is normalized to the signal measured at day 0 immediately after tumor implantation (set as 100). Brain tumors implanted in FAK WT mice demonstrate rapid growth, whereas the tumor growth rate in FAK CKO hosts is significantly reduced over 14–28 days. (p=0.002, n=10–11 per group, Wilcoxon rank sum test, two-sided)

(D) Tumor growth at day 28 of each group is shown, demonstrating a significant decrease (15-fold) in tumor burden in FAK CKO vs. FAK WT hosts. (p=0.002, n=10–11 per group, Wilcoxon rank sum test, two-sided)

FAK-mediated remodeling of brain tumor-associated blood vessels

To determine whether the reduced tumor growth in FAK CKO mice was associated with changes in the tumor vasculature, we examined morphological and functional changes in the vessels of FAK CKO vs FAK WT mice. In non-tumor bearing mouse brains, the appearance of the vessels was generally similar in FAK CKO and FAK WT in terms of vascular density, vascular permeability, and immunohistochemical staining of BBB compartments (i.e. localization of endothelium, basal lamina and astrocyte endfeet) (Supplementary Data Figure 1). In contrast, the tumor-associated blood vessels FAK CKO mice had a distinct phenotype compared to FAK WT mice (Figure 3). Consistent with previous observations in both CNS and non-CNS tumors (33–35), intra-tumoral vessels in FAK WT mice exhibited distorted, hyper-dilated and tortuous tumor vessel architecture as assessed by perfusion labeling with the fluorescently-labeled lectin (Figure 3A, open arrowheads in tumor region) (33). In contrast, tumor-associated blood vessels in FAK CKO mice appeared to have a relatively ‘normalized’ phenotype in terms of the increased number of identifiable intra-tumoral vessels (Figure 3C). We have noted that vessels in the brains of FAK CKO mice tended to be longer and thinner (i.e. less tortuous)(Figure 3B) and less branched in three dimensions (Figure 3A, closed arrowheads in tumor region) compared to intra-tumoral vessels in FAK WT mice. Yet, they are still relatively hyper-dilated when compared to vessels that were not associated with tumor (data not shown). We also observed similar morphological changes of the intra-tumor vessels of FAK CKO or FAK WT mice at tumor margins and in smaller satellite tumors (data not shown).

Figure 3. FAK-mediated remodeling of brain tumor-associated blood vessels.

(A) FAK WT and FAK CKO mice were injected intravenously with 1mg of FITC-lectin to image vessels. Tumor-associated vessels (i.e. dilated, disordered and tortuous) of FAK WT mice were observed as expected (left, open arrowhead in tumor area), however tumor-associated brain blood vessels in FAK CKO mice demonstrated morphologically distinct, ‘normalized’ structure (i.e. longer, thinner, less tortuous, and less branched in three dimensions)(right, closed arrowheads in tumor area). Tumor margins were identified based on the fluorescence of their RFP label (DBTRG-RFP) or distinct DAPI stained tumor pattern of luciferase labeled cells by tracing back to the injection site (DBTRG-luc). Size bar = 100µm

(B) FAK WT and FAK CKO mice were injected intravenously with 1mg of FITC-lectin to image vessels. The average diameter of tumor-associated vessels was measured as described in Materials and Methods. We observed smaller tumors in FAK CKO mice when compared to FAK WT mice. Data was measured from at least four independent fields per group. (p=0.03, N=4, Wilcoxon rank sum test, two-sided)

(C) Vascular density of FAK WT and FAK CKO was quantified by counting CD31-positive vessels per field. The number of tumor-associated vessels was higher in FAK CKO mice than that of FAK WT mice. (p=0.04, N=6–9, Wilcoxon rank sum test, two-sided)

Role of FAK in regulation of brain tumor-induced VP of the BBB

To assess the leakiness of tumor-associated blood vessels, FAK CKO and FAK WT mice injected with fluorescently labeled glioma cells (i.e. red fluorescent protein, RFP) were subjected to intravenous administration of fluorescently labeled 70kDa dextran (i.e. FITC). In FAK WT mice, tumor vessels were found leaky and extravasation of dextran was observed, whereas in FAK CKO mice, tumor vessels were less leaky and less extravasation of the dye was observed (Figure 4A). To quantitatively assess VP of tumor-associated vessels, FAK CKO and FAK WT mice injected with RFP-labeled glioma cells were injected intravenously with 70 kDa FITC-dextran followed by systemic saline perfusion. In FAK WT mice, we observed a statistically significant correlation between tumor size (i.e. fluorescent intensity of RFP-labeled glioma cells) and VP (i.e. fluorescent intensity of FITC-dextran) (Figure 4C, p=0.002). In contrast, there was no correlation between tumor size and VP in FAK CKO mice (Figure 4D, p=0.78). Following a four-week incubation period, the tumors in FAK CKO host were generally smaller compared to those of FAK WT mice as shown in Figure 2, so we let some FAK CKO tumors grow larger to ensure that there was a difference in permeability in size-matched tumors. After an additional two-week incubation period, on average we observed a 6.5 fold reduction in VP in FAK CKO mice as compared to FAK WT mice even in size-matched tumors (Figure 4B, p<0.05). We did not observe any differences of VP in FAK CKO or FAK WT mice brain sections with no tumor (used as internal controls, Supplementary Figure 1) indicating efficient perfusion of the brains. This observation suggests that decreased VP in tumor-associated FAK CKO vessels may be associated with the relatively ‘normalized’ morphology of tumor-associated vessels in FAK CKO mice (Figure 3).

Figure 4. Role of FAK in regulation of brain tumor-induced VP of the BBB.

(A) Tumor-bearing FAK CKO and FAK WT mice were injected with 70kDa FITC-dextran to image leakiness of the vasculature. FAK WT mice exhibit leaky vessels shown by extravasation of the FITC-dextran tracer (left). Tumor vasculature of FAK CKO mice demonstrate relatively ‘normalized’ morphology and less extravasation of the FITC-dextran tracer (right).

(B) VP of tumor-associated blood vessels in FAK CKO vs. FAK WT mice was quantitated in size-matched tumors (tumor in FAK CKO mice were incubated longer to assess FAK-dependent changes in VP in size-matched tumors),as described in the Materials and Methods. FAK CKO mice demonstrated a 6.5 fold reduction of VP compared to FAK WT mice (p<0.05, Wilcoxon rank sum test, two-sided, error bar = standard deviation).

(C) Following a 28-day incubation post-tumor implantation, FAK WT mice were injected with 70kDa FITC-dextran permeability tracer, tumor burden and vascular permeability was measured. A representative image of a 1mm thick brain section is shown. Tumor burden (upper panel) and vascular permeability from the same section were quantified (lower panel). Vascular permeability correlates with tumor size (p=0.002, n=17, Kendall’s rank correlation test, two-sided).

(D) Four to six weeks post-tumor implantation (tumor in FAK CKO mice were incubated longer so that the tumor size was large enough to measure VP), tumor burden and VP of tumor-associated blood vessels in FAK CKO vs. FAK WT mice were quantified. No correlation between tumor size and VP was observed (p=0.78, n=14, Kendall’s rank correlation test, two-sided).

Immunohistochemical assessment of FAK-mediated changes in the BBB

To understand the mechanism underlying reduced vascular permeability that was observed in tumors grown in the FAK CKO host (Figure 4), we have examined the expression of cell-cell adhesion molecules in the tumor vasculature. In particular, we have focused on ZO-1 and occludin, for the expression of these molecules are often associated with intact endothelial barrier integrity in normal vessels (ref). While tumor vessels in FAK WT mice have reduced ZO-1 and occludin expression, the expression of these molecules were relatively restored in intratumoral vessel of FAK CKO mice (Figure 5A,B).

Figure 5. Immunohistochemical assessments of FAK-mediated changes in the endothelial-astrocyte interface.

(A,B) Immunohistochemical stainings of cell-cell junction protein ZO-1 (A) and occludin (B) in tumors of FAK WT and FAK CKO mice are shown. The immunoreactivity of ZO-1 and occludin is downregulated in tumors of FAK WT mice while the expression is restored (arrow) in intratumoral vessels of FAK CKO mice (exposure-matched images are shown). Quantitation of ZO-1 and occludin immunoreactivity in tumor area of FAK WT and FAK CKO mice is shown on right. ZO-1, occludin immunoreactivity is 3 and 2-fold higher in FAK CKO mice compared to FAK WT mice, respectfully. (p<0.05, Wilcoxon rank sum test, two-sided, error bar = standard deviation) Size bar = 100µm

(C) Immunohistochemical staining of tumor-bearing brain sections using an anti-CD31 antibody for endothelial cells and anti-AQ4 antibody to localize vessel-associated astrocytic endfeet is shown. Note the increased AQ4 immunoreactivity in FAK CKO mice compared to exposure-matched images of FAK WT mice. The interaction between AQ4 positive astrocytes and tumor vessels are lost in tumor of FAK WT mice as reported earlier (left), however the interaction is partially restored in tumors of FAK CKO mice (right, arrowhead and arrow). Examples of AQ4-positive astrocytes in close contact with CD31-postive tumor vessels (arrow) are shown in higher magnification (lower). Size bar = 100µm

(D) Quantitation of AQ4 immunoreactivity in tumor area of FAK WT and FAK CKO mice is shown. AQ4 immunoreactivity is 5-fold higher in FAK CKO mice compared to FAK WT mice (p<0.05, Wilcoxon rank sum test, two-sided, error bar = standard deviation)

Astrocyte endfeet have been shown by several groups to be closely associated with brain endothelium (1), while there are specific changes to these associated astrocyte structures in endothelium of tumor (4, 36). Because AQ4-positive astrocyte endfeets are implicated in the maintenance of normal BBB (1, 36, 37), we next examined tumor-induced remodeling of specific BBB compartments. Relatively normal AQ4 distribution (i.e., restored expression and contact between astrocytes with intratumoral vessel (Figure 5C and 5D) as found in WT brain with no tumor) was found in FAK CKO host while astrocytic contact with endothelial cells are lost in FAK WT host as described earlier (Figure 5C,D) (4).

In both FAK WT and FAK CKO mice, FAK-deletion did not mediate any further changes of the BBB with the markers we have tested to this date (Supplementary Figure 3A). But we observed reactive astrocytes that were expressing GFAP and AQ4 at the tumor margin in tumors of both FAK WT and FAK CKO mice (Figure 5C upper row and Supplementary Figure 3A). and differential immunostaining of laminin inside the tumor bed vs adjacent normal area as described earlier (4).

These results suggest that the restoration of cell-cell adhesion molecules in the endothelium and restoration of astrocyte-endothelial interaction in tumor vessels may contribute together to reduced VP in FAK CKO mice (Figure 4). It also demonstrates that specific changes in the endothelial cells influence distinct, yet related cellular mediation of BBB integrity (i.e. astrocytes).

Discussion

In this study, we have characterized the effect of endothelial-specific deletion of FAK in the brain with a focus on a VEGF-expressing brain tumor-mediated vascular remodeling. We identify a novel role for FAK, a target of Src, in the regulation of the endothelial component of the adult BBB. We observed that FAK CKO mice had a relatively stabilized tumor vasculature (Figure 3) with reduced vascular permeability (Figure 4) and reduced tumor growth (Figure 2). Consistent with a role of FAK in VP, our previous studies in Src knockout mice have demonstrated a leakage-resistant phenotype in the brain, leading to reduction of glioma infiltration (3). Together, these studies demonstrate genetic evidence for a role for the FAK signaling pathway in the maintenance of BBB integrity.

Based on the central role for FAK in mediating signaling from the extracellular matrix through integrin in migration (38), tumor progression and downstream responses (39), there has been a major effort to better understand the functional role of FAK in knockout animals. Standard FAK-deleted (FAK KO) embryos exhibit developmental defects, including severely impaired blood vessel development (14). Despite the integral role of FAK in the formation of focal adhesions and migration (6, 40), FAK-null fibroblasts did not prevent the formation of focal adhesions, suggesting a role for FAK in the turnover of focal adhesions (41). Endothelial-cell specific deletion of FAK also led to an embryonic lethal phenotype, thus functionally demonstrating the requirement for FAK in vascular development (14, 16–18). More recently, differential role of kinase-independent and dependent functions of FAK in vascular development of embryogenesis has been reported (20, 21). These studies propose that the kinase activity of FAK is required to maintain the integrity of endothelial cells whereas the structural functions of FAK are likely to be important for endothelial cell survival (20, 21). In addition, we have previously shown that the phosphorylated form of FAK is associated with Src in VEGF-mediated activation of endothelial cells (8). Based on these studies, it is likely that the kinase activity of FAK is essential in mediating tumor-induced VP shown in our CNS model.

When using knockout mouse models, consideration of mechanism of compensation is essential, and in the case of FAK CKO mice, evidence for such compensation by PYK2 has been described in vitro and in vivo (19, 42–44). While we did not observe an increase in PYK2 in our FAK CKO mice (data not shown), further examination of such models may yield important insights into tissue-specific compensatory pathways.

The orthotopic xenograft brain tumor model used in this study is well suited for identifying the tumor-induced changes in the host compartment where the focus is host (i.e. endothelia and indirectly astrocytes) rather than tumor cells. By utilizing different genetic backgrounds of the host (i.e., Src KO and endothelial-specific FAK KO), we can examine the functional relevance of specific signaling intermediates in the BBB that affect glioma progression. While other cell types in the brain participate in maintaining the BBB (i.e. through interaction with astrocyte endfeet), in vivo studies to understand the mechanism regulating the interaction between endothelial cells and astrocytes remains poorly understood. Our study demonstrates that specific changes in the endothelial cells mediates changes in astrocytes as observed by the remodeling of the structure and function of endothelial-associated astrocyte endfeet in the BBB in vivo (Figure 5). To this end, we observed a correlation between restoration of AQ4 and down regulation of VP, however it would be necessary to further investigate the role of AQ4 in vascular permeability using other functional models, for example, AQ4-knockout mice. However, these studies provide the basis for the systematic dissection of the role of specific cell types, which through targeted gene knockout strategies can be used to determine the functional relevance of endothelial:astrocyte interactions in vivo. Alterations in these interactions may be useful to better understand in the development of therapeutics that crosses the BBB.