|Home | About | Journals | Submit | Contact Us | Français|
The normal BBB (blood-brain barrier) consists of a series of structures collectively known as neurovascular units, or NVU, that are composed of endothelial cells and astrocyte end feet separated by a basal lamina at their interface. The integrity of the BBB and specifically endothelial tight junctions is maintained by interactions between these different components and the local microenvironment of the NVU. Central nervous system cancers such as gliomas disrupt the integrity of the BBB and this compromise is associated with increased tumor growth and invasion of the surrounding brain parenchyma. Because the relationship between glioma-induced BBB breakdown and glioma invasion remains poorly understood, and the host microenvironment can influence tumor cell migration, we used immunohistochemical techniques to characterize tumor associated BBB remodeling. Using an orthotopic xenograft model of glioma, we demonstrate that tumor cells induce specific changes in the composition of the basal lamina and in astrocytic components of the NVU. We suggest that these changes may be essential to understand the capacity of gliomas to regulate BBB integrity and as such, glioma invasion into brain parenchyma.
Gliomas are associated with aggressive invasion of the surrounding brain parenchyma. While resection reduces the primary tumor burden, extensive migration of glioma cells away from the primary tumor mass prevents the complete removal of tumor cells and represents a significant challenge in the treatment of glioma. In vivo experimental models of glioma have been used to demonstrate that tagging tumor cells can be used to monitor the infiltration of tumor cells in the parenchyma and/or perivascular spaces even at a single cell level (Lampson et al., 1993; Lund et al., 2006; Winkler et al., 2009; Zhang et al., 2002). However, the relationship between glioma-induced BBB dysregulation and glioma invasion remains poorly understood.
Gliomas disrupt the integrity of the blood-brain barrier (BBB), and this compromise is associated with increased glioma tumor growth and infiltration in human and rodents (Lund et al., 2006; Neuwelt and Rapoport, 1984). Whereas normally, the BBB consists of endothelial cells held together by tight junctions, it also contains a basal lamina surrounding the capillary walls and adjacent astrocyte end feet that form a structure collectively known as the neurovascular unit, NVU (Iadecola, 2004; Simard and Nedergaard, 2004). The interactions between the different cell types of NVU are disrupted in the microenvironment of the glioma (Abbott et al., 2006; Abramsson et al., 2007; Virgintino et al., 2007).
While many studies with malignant glioma focus on the dysregulated signaling pathways in tumor cells as a potential target for therapeutics (Furnari et al., 2007), our previous studies have shown the importance of the host microenvironment to influence the migration of the tumor cells (Criscuoli et al., 2005; Lund et al., 2006). For example, the presence of specific host factors such as src in the brain endothelium is essential for tumor-mediated BBB breakdown and glioma invasion (Eliceiri et al., 1999; Lund et al., 2006). In this study, we compared tumor-induced changes in the NVU of primary vs secondary tumors (as defined by their distance from the primary tumor), and suggest that changes in basal lamina composition may mediate the capacity of gliomas to regulate BBB integrity.
In the normal brain, a subset of astrocytes extends projections to the basal lamina that surrounds the microvascular endothelium. Immunohistochemical staining with anti- glial fibrillary acidic protein (GFAP) demonstrates the difference between arterioles/venules and the capillary bed and highlights the specific association of GFAP-positive astrocytes with larger vessels that have a minimal role in regulating vascular permeability (Fig 1A). Immunostaining of 60μm thick tissue sections with an anti-GFAP antibody reveals astrocytes along arterioles and venules (Fig 1A) while lesser staining for GFAP was observed on branching capillaries (arrow, Fig 1A). To define the endothelial-astrocyte junction in the normal mouse brain and characterize the specificity of the various astrocyte markers associated with blood vessels, we examined the distribution of astrocytes using three different astrocyte markers; 1) GFAP as a generic marker of astrocytes (Eng, 1985); 2) aquaporin-4 (AQ4), a water channel protein expressed in astrocyte endfeet (Warth et al., 2004), and 3) aldehyde dehydrogenase 1 family, member L1 (AldH1L1), a protoplasmic astrocyte marker (Cahoy et al., 2008; Oldham et al., 2008) (Fig 1B). As shown in Fig 1, astrocytes at the astrocyte-endothelial junction of the BBB are AQ4-positive but GFAP-negative. The anti-AQ4 antibody labels the astrocytic endfeet that surrounds the CD31-positive capillaries while GFAP-positive astrocytes are not associated with CD31-positive capillaries (Fig 1B and 1C). In contrast, the GFAP-positive astrocytes that are associated with blood vessels generally define arterioles or venules. In contrast, AldH1L1 immunoreactivity was detected in both subclasses of astrocytes and is associated with capillaries and arterioles/venules (Fig 1D). These studies define a subset of AQ4+, GFAP− astrocytes that constitute the primary population of astrocytes that are associated with the normal brain capillary endothelium.
Although cell culture studies have characterized the interactions between astrocytes and brain endothelial cells (Abbott et al., 2006), these in vitro approaches do not fully recapitulate the complexity of the brain microenvironment. Therefore, we developed an orthotopic xenograft model of glioma growth and invasion in an immuno-deficient background to assess the dysregulation of BBB induced by human malignant glioma. DBTRG malignant gliomas (Kruse et al., 1992) maintain the characteristic infiltrative and pseudopalisading phenotype of gliomas and induce vascular permeability in the brain when injected in vivo into mice, thus serving as a pathophysiologically relevant model to study glioma (Lund et al., 2006). To facilitate the characterization of glioma-induced remodeling of astrocyte-basal lamina interface, we used reporter genes to enable noninvasive localization of DBTRG glioma cells in vivo. DBTRG-luc cells were used to localize tumors noninvasively and DBTRG-RFP cells were used for high resolution because of its compatiblity with indirect immunofluorescence microscopy. These complementary labeling techniques enable the non-invasive quantitation of glioma growth while also facilitating localization of individual tumor cells by fluorescence. Furthermore, tumor growth throughout the cortex of the brain and extensive infiltration of the tumor cells can be readily detected ex vivo. A representative set of brain slices of these tumor bearing-mice is shown in Fig 2C and 2D to localize the tumor. Tumor progression was also monitored noninvasively at 0, 4, 7, 14, and 21 days after implantation (Fig 2A and 2B) by measuring luciferase activity in the tumor cells. Representative brain slices showing luciferase activity confirm the presence of tumor cells and were then evaluated by immunohistological analyses for markers of BBB integrity.
We used antibodies for specific ECM markers such as laminin, agrin and tenascin and an endothelial cell marker (CD31) to characterize glioma-induced remodeling of the basal lamina of the NVU. Tumor margins were identified by indirect fluorescence of the tumor cells and their characteristic dense nuclei pattern from DAPI staining. Ipsilateral and contralateral control areas (with no tumor) in tumor- bearing brains slices showed similar staining pattern to the normal brain, and were used as internal controls for immunostaining in this study. In this study, primary tumors are identified by its size, (i.e. the main bulk of the tumor) and tumors located at least 50 μm distal to the primary tumor margin that are smaller in size in the brain were defined as secondary invasive tumors. Using CD31 staining, intratumoral vasculature had decreased vessel density, higher intervascular distance, but increased microvascular diameter when compared to normal blood vessels (Fig 3A). In primary tumors, laminin staining was restricted to the basal lamina of the endothelial perivascular space (Fig 3B). However, in secondary tumors, laminin was observed in a strong, multilayered staining pattern at the leading edge of secondary tumors and not with any apparent association with endothelial cells (Fig 3B). Whether the laminin detected is tumor-derived or host-derived remains unknown. To evaluate the effects of tumor-induced remodeling on proteoglycans deposited into the basal lamina, we assessed agrin immunoreactivity in the tumor microenvironment. Agrin is an extracellular heparan sulfate proteoglycan that is expressed by both the astrocyte and the endothelial cells (Barber and Lieth, 1997; Rascher et al., 2002; Warth et al., 2004). As shown in Fig 3C, agrin staining was decreased in the primary tumor, and like laminin, was localized to the surface of endothelial cells, surrounds the leading edge of the secondary tumor, and is not associated with endothelial cells. In contrast, tenascin, a high molecular weight, multifunctional, extracellular matrix glycoprotein that is upregulated in malignant glioma in vivo and promotes migration of glioma cells in vitro (Deryugina and Bourdon, 1996), shows a significantly different pattern of staining. Tenascin staining was increased within the solid tumor mass and a wide spread network of tenascin immunoreactivity was observed in the tumor core but not in secondary tumors (Fig 3D). Although in vitro studies predict a role for tenascin in glioma-cell migration, tenascin staining was decreased at the tumor margins and in secondary tumors; regions where it was predicted to be elevated if it were involved in secondary glioma infiltration in vivo (Fig 3D).
Astrocytes are widely recognized components of the BBB that confer brain endothelial cells with their characteristic barrier tight junctions (Abbott et al., 2006), yet the capacity for astrocytes to undergo tumor-induced redistribution is poorly understood. To this end, we used an anti-GFAP antibody as a general astrocyte marker, an anti-AQ4 antibody as a marker for the astrocyte-endothelial interface and an anti-Aldh1L1 antibody as a protoplasmic astrocyte marker (Cahoy et al., 2008; Oldham et al., 2008) to assess the effects of gliomas on BBB astrocytes. Elevated GFAP staining was found at the margin of the primary tumor, however, within the solid tumor mass, GFAP was decreased (Fig 4B). AQ4 immunoreactivity was increased in the primary tumor where astrocytic contact with endothelial is lost (Fig 4C). Secondary tumors showed elevated AQ4 immunoreactivity at the leading edge of the tumor (Fig 4C), but not in intratumoral blood vessels. This is in contrast to the localization of AQ4, which is observed at the glial endfeet around the vessels in normal brain (Fig 1B) and in control area in tumor-bearing brain (Fig 4C). Aldh1L1 staining was also increased at the margin of primary tumors but decreased in the tumor core (Fig 4D). These findings suggest that glioma induce differential remodeling in different subclasses of astrocytes.
In this study, we show that tumor induces specific changes in endothelial cells, remodeling of the ECM and changes in endothelial-associated astrocytes. We observed a decrease in laminin and agrin, and an increase in tenascin in the primary tumor (Fig 3). In secondary tumors, both astrocytes and blood vessels remain separated by a basal lamina that contains laminin and agrin (Fig 3 and Fig 4). The changes in staining observed in the primary tumor are in general agreement with observations from rodent models and clinical samples of glioma (Chekhonin et al., 2007; Rascher et al., 2002; Warth et al., 2004; Warth et al., 2007). For example, tenascin upregulation is accompanied with loss of agrin expression in human glioma tissues (Rascher et al., 2002). Increased expression of laminin, fibronectin, and vitronectin was also shown at the infiltration zone of human glioma cells implanted into rat brain (Mahesparan et al., 2003). Here, we provide evidence for a differential glioma-mediated remodeling of the NVU in primary vs secondary tumors. Understanding the temporal-spatial control of these differences may be important in terms of identifying novel therapeutic targets.
Changes in ECM composition influence rigidity and structure (i.e., fibrillary collagen, fibronectin) and suggest that a malignant infiltration of glioma may stimulate the development of distinct provisional ECM. Indeed, tumor cells induce changes in ECM by secretion of growth factors and angiogenic factors such as VEGF, bFGF, PDGF, and TNF-α (Gesa Rascher-Eggstein, 2004). In turn, the different components of the provisional ECM influence tumor cells, either by changing the mechanical rigidity (i.e. stiffness) of the environment (Huang and Ingber, 2005; Paszek et al., 2005) or by modulating the activity of different growth factors and creating a positive-feedback loop in the tumor-host microenvironment.
The xenograft model described here is well suited to distinguish the role of host/tumor in tumor growth and progression and is associated with BBB breakdown : First, tumor margins can be easily defined using RFP-tagged tumor cells (Lund et al., 2006). Second, the leading edge of primary and secondary tumors are readily defined using confocal microscopy on thick brain sections. Finally, the cellular origin of various ECM components can be assessed using species-specific antibodies. For example, a mouse-specific antibody to tenascin can be used to demonstrate that the high levels of tenascin detected in primary tumors are produced by the host, even though the levels of immunoreactivity decrease at the leading edge of the tumor (Fig 3).
The findings described here also suggest that changes in the astrocytic endfeet (i.e. AQ4) of the NVU and/or changes in the basal lamina (i.e. agrin) in brain tumors may influence the function of astrocytes. These changes presumably lead to dysregulation of the BBB. Maintenance of their integrity is necessary in order to restrict ionic and fluid movements between the blood and the brain. The water channel protein AQ4 has a crucial role in maintaining the integrity of BBB by forming a functional complex with dystrophin and dystroglycan to form orthogonal arrays of intramembranous particles at the astrocytic endfoot membranes (Wolburg et al., 2009). The stability of this complex is dependent on extracellular matrix components, for example agrin (Noell et al., 2007). Although the molecular mechanisms underlying the changes in the endothelial BBB remain unclear, understanding how the integrity of BBB is maintained in the normal brain has a wider impact on other pathologies of the central nervous system including stroke, traumatic brain injury, multiple sclerosis and neurodegenerative diseases (e.g. Alzheimer's disease) which like glioma are all associated with functional alterations of the BBB (Abbott et al., 2006).
Rag2 knockout mice were used for xenograft studies based on previous studies using rag2−/−, src+/+ or rag2−/−, src+/− hosts (Eliceiri et al., 1999; Lund et al., 2006). All animal handling procedures were approved by the University of California San Diego Institutional Animal Care and Use Committee.
Early passages of patient-derived human glioma cells, DBTRG (a kind gift from Dr. C. Kruse) were used for xenograft studies. Cells were maintained in Dulbecco's modified Eagle's minimum essential medium supplemented with 10% fetal bovine serum, penicillin, streptomycin, nonessential amino acids, and glutamine in a humidified atmosphere containing 5% CO2 at 37°C.
For in vivo bioluminescence imaging of orthotopic brain tumors, early passages of DBTRGs were transduced with lentivirus expressing firefly luciferase (DBTRG-luc) or red fluorescent protein (DBTRG-RFP). To engineer lentivirus expressing luciferase, the open reading frame of firefly luciferase was excised from pGL4.14 (Promega) and cloned into the Xho I site and Xba I sites of the lentiviral vector, pLVX (Clontech). Infectious lentiviral particles were prepared in a 293T packaging cells (ATCC) using the manufacturer's protocol (Lenti-X Lentiviral Expression Systems, Clontech). Transduced cells were selected by either flow cytometry (DBTRG-RFP) or by adding 2μg/ml puromycin in the culture medium for at least 2 weeks (DBTRG-luc).
After immobilizing the mice in a rodent stereotactic frame, an incision was made in the skin, and a burr hole made in the skull. One million tumor cells or same volume of saline were injected at a rate of 1−2 microliter/minute using a microsyringe (Hamilton, Reno, NV) mounted on a stereotactic frame (Kopf Instruments, Tujunga, CA) using coordinates of 1 mm lateral and 1mm posterior to the Bregma and 1.5 mm below the dura. The incision was closed with veterinary adhesive and topical lidocaine administered. After 21 days, mice were subjected to systemic intracardiac perfusion with 1 USP unit/ml of heparin dissolved in saline, and serial-cryosections of brains were processed as indicated in the figure legends to yield 20μm or 60 μm brain slices. Immunostaining of tissue sections were imaged with an Olympus Fluoview 1000 (ASW 1.7b) laser scanning confocal microscope (Olympus, Melville, NY) equipped with 20x/0.7N.A.(dry), or 860x/1.4N.A.(oil) objective lenses on a BX61 microscope (Olympus).
Three weeks after implantation, the animals were perfused with heparin/saline by intracardiac injection, and brains or 1mm sections of the brains were harvested and cryoembedded in OCT. Standard immunohistochemistry was performed on cryosections (20 μm) of tumor samples using the following primary antibodies at the following dilutions. Anti-CD31(553370, BD Biosciences, 1:100), rabbit polyclonal anti-laminin(L9393, Sigma, 1:1200), mouse monoclonal anti-GFAP (C9205, Sigma, 1:200), anti-aquaporin 4 (Chemicon,1:1000), monoclonal anti-mouse tenascin (T3413,Sigma,1:100), anti-Aldh1L1 (N103/39 NeuroMab, 1:100), anti-agrin (rabbit polyclonal antibody, a kind gift from Dr. M.A. Ruegg, 1:100). Alexa-fluor-conjugated secondary antibodies were used at a 1:200 dilutions and were purchased from Molecular Probes (Eugene, OR). All sections were counterstained with 4'-6-Diamidino-2-phenylindole (DAPI).
Hair was removed from mice with electric clippers and Nair before imaging at each time point. Mice were injected intraperitonealy with D-luciferin (150μl of 15mg/ml stock) to image tumor burden and monitor tumor growth immediately after and at 4, 7, 14, and 21 days after implantation. Bioluminescent signals were assessed 10 minutes after luciferin injection over an integration time of 10 seconds and 1 minute using a cooled charge-coupled device (CCD) camera (Lumina; 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. Pseudocolor images representing the intensity of the bioluminescent signal (blue is the least intense and red is the most intense) were superimposed over the grayscale reference image and displayed with the scales for the pseudocolor intensity (Unit = photons/sec/ cm2/steradian). Images were analyzed using Living Image software version 3.1 (Caliper Life Sciences). Following non-invasive imaging over 21 days, mice were sacrificed and images from 1mm sections of fresh mouse brain were collected to identify tumor-bearing sections and for immunohistochemical analyses.
We acknowledge the expert technical assistance of Montha Pao.
These studies were supported by grant from the NHLBI (B.E.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.