Malignant Glioma Microenvironment

Interactions of Malignant Gliomas With the Neural Extracellular Matrix

The neural ECM is the molecular scaffold that fills the extracellular space (ECS) in the CNS and forms the immediate physical support of all cell types. The presence of a significant neural extracellular volume was disputed until the early 1980s owing to artifacts in the preparation of neural tissue for microscopy and deficiencies in the histologic detection of proteoglycans and mucopolysaccharides. ^, However, nonmorphologic methods applied to fresh tissues, based on the diffusion of charged or fluorescent molecules, demonstrated that the ECS occupies up to 20% of total CNS volume, which was later confirmed with improved fixation techniques and immunohistochemistry. ^, Measurements in live animals put the CNS among the tissues with largest extracellular volume when compared with other highly vascularized tissues such as kidney or skeletal muscle. Measurements of the ECS have been performed in freshly resected gliomas, revealing changes in the ECS volume and its tortuosity compared with normal brain tissue. Both low- and high-grade astrocytomas have larger ECS volume than normal brain, with volume increases that correlate with grade. Even in highly cell-dense glioblastomas, this acellular volume can reach up to 40% of the total tumor volume (and 58% of total volume in the necrotic regions of these tumors).

Extracellular tortuosity (not to be confounded with the tortuosity of blood vessels) is a measurement of how much the extracellular diffusion of small molecules is limited in real tissues compared with a perfect homogeneous medium. Changes in the amount and composition of the ECM can increase the extracellular tortuosity, revealing important barriers to molecule diffusion. Extracellular tortuosity is similar between low-grade astrocytomas and normal brain, but it increases significantly in anaplastic astrocytomas and glioblastomas. This increase does not correlate with tumor cell density, but with the accumulation of ECM forming fine “molecular nets” in high-grade gliomas. Interesting to note, neither the ECS volume nor the extracellular tortuosity are different in grade II oligodendrogliomas compared with normal brain; this is thought to be due to the fact that these tumors are able to intermix with brain tissue without causing significant disruption of the neural architecture. ^,

The normal stromal ECM of the CNS, found in the white matter and gray matter neuropil, is a soft gel rich in proteoglycans, glycoproteins, and hyaluronic acid, but notoriously devoid of large fibrillar proteins (e.g., collagens, elastins, fibronectin). ^{, ,} A different type of ECM, forming a well-defined basal lamina (BL) rich in fibrillar proteins (in particular collagen type IV, laminin type I, and fibronectin), is tightly restricted to the choroid plexus and the perivascular and subpial spaces in the CNS. The major components of the typical neural ECM and the perivascular BL in the CNS, as well as their modifications by malignant gliomas, are listed in Tables 138.1 and 138.2 , respectively.

Malignant gliomas show very different interactions with both types of ECM: tumor cells usually adhere well to fibrillar proteins of the BL in vitro and migrate along the BL of perivascular and subpial surfaces, which form major anatomic routes for tumor dispersion. ^, However, glioma cells do not seem to be able to cross this BL in vivo, which would lead to intravasation or intrameningeal spread, both of which are extremely rare in these tumors. ^, On the other hand, adhesion of isolated glioma cells in vitro can be attenuated by stromal chondroitin sulfate, hyaluronic acid, or myelin-associated glycoprotein. However, individual glioma cells readily traverse the ECM of the neural parenchyma, dispersing along white matter tracts and through the gray matter neuropil. It is worth noting that the interaction of glioma cells with the BL and stromal neural ECM seem to be essentially reversed in nonneural tumors that metastasize to the CNS. Metastatic cells cross the BL and readily extravasate into the brain perivascular space, but they rarely invade the inhibitory neural ECM and instead grow close to their extravasation sites as contained masses.

The motility of individual glioma cells appears to involve locally restricted degradation of the ECM ^, and may be largely facilitated by “squeezing” of the cell body through ECM spaces (see Fig. 138.1B–C ), which is driven by myosin II–dependent cell contraction ^, and rapid changes in total cellular volume ^, ( Video 138.2 ). The resulting amoeboid movement has been largely compared with similar migratory behavior observed in neural cell precursors, reinforcing the concept that glioma cells may derive from a motile neural precursor. Cell motility is, in addition, facilitated by additional mechanisms of ECM remodeling, such as incorporation of glioma-secreted matrix molecules into the neural ECM and covalent modification of the ECM scaffold. The resulting changes open intercellular spaces and increase the stiffening of the ECM, facilitating tumor cell adhesion and migration.

Most of the glioma-secreted proteases that degrade and remodel the ECM belong to the large families of matrix metalloproteases (MMPs) and ADAMTS (a disintegrin and metalloprotease with thrombospondin motifs). MMPs have been classically implicated in glioma invasion owing to their ability to cleave and degrade fibrillar proteins such as collagens and fibronectin. ^, The collagenases MMP-2 and MMP-9 have been identified numerous times as typical MMPs upregulated in gliomas compared with normal brain tissue, and have been shown to be required to promote tumor invasion. Although these metalloproteases can also degrade chondroitin sulfate proteoglycans in vitro, ^, cleavage of these proteoglycans by MMPs in vivo is less common and possibly less functionally relevant. ^, Instead, neural proteoglycans are primarily cleaved by ADAMTS enzymes, in particular the aggrecanases ADAMTS-4 and ADAMTS-5, with minor contribution of ADAMTS-1. Of these, ADAMTS-5 is particularly elevated in malignant gliomas and correlates with tumor grade. ^, Finally, it is worth noting that although studies of MMPs and ADAMTS have dominated the protease-oriented research for malignant glioma, other proteases such as the cysteine cathepsins also play a significant role in degrading the neuropil and BL ECM. ^, Although these proteases are mostly active in the acidic lysosomal environment, their secretion into the acidified tumor stroma makes them a relevant factor in ECM remodeling that may contribute to glioma invasion and angiogenesis. ^{, ,}

As indicated earlier, the dispersion of single glioma cells through the neural ECM is largely dependent on their ability to open intercellular spaces and squeeze through them in an amoeboid manner. ^, This process may also be facilitated by the secretion of proteoglycans and hyaluronic acid by glioma cells. These high-molecular-weight, hygroscopic molecules retain water and increase the hydrated space around the tumor cell, facilitating cell growth and movement. Brevican and versican are the two major chondroitin sulfate proteoglycans secreted by glioma cells and have been shown to promote tumor growth and invasion. ^{, ,} Both proteoglycans exhibit specific isoforms upregulated in glioma ^, and are, in addition, substrates for the ADAMTS enzymes. Work with glioblastoma cells in culture has shown that specific domains of these proteoglycans may be released from the ECM scaffold by ADAMTS cleavage, acting as signals that activate integrin and epidermal growth factor receptor–mediated signaling to promote tumor cell proliferation (versican) ^, and invasion (brevican). ^,

Finally, the ECM of the tumor microenvironment can change not only by the addition or degradation of ECM molecules but also by the chemical (covalent) modification of these molecules. For example, glioma cells secrete enzymes of the lysyl oxidase family, which are copper-dependent enzymes that cross-link neighboring lysine residues in collagen and elastin. In normal tissues, this crosslinking is critical to provide strength and stability to major fibrillar proteins, and absence or inhibition of lysyl oxidases results in overall weakness of bone, ligaments, and skin. In solid tumors, lysyl oxidase increases the firmness of the ECM, which facilitates cell migration and contributes to tumor invasion and metastasis. The same role has been postulated for lysyl oxidases in gliomas, although their specific molecular targets in the glioma ECM have not yet been identified.

Interactions of Malignant Gliomas With Astrocytes and Neurons

Significant advances have been made in recent years in uncovering the ways in which malignant glioma cells integrate with the major CNS cell types—astrocytes and neurons. These interactions are revealing important potential therapeutic developments, as well as the concept that malignant glioma cells integrate with functional neuronal networks, and indeed that these interactions play an important role in tumor cell survival and growth.

Glial cells of the astrocytic lineage are the most abundant cell type in the adult neural parenchyma and the major source of molecules that form the stromal ECM of the CNS. Astrocytes are, in addition, a key cell type in the perivascular niche where most gliomas arise by transformation of neural precursors of the glial lineage into glioma initiating cells. Astrocytes are therefore in intimate contact with neoplastic cells and contribute to the metabolic and functional support of nascent gliomas through the delivery of nutrients and cytokines and removal of metabolic byproducts. It has been hypothesized that newly formed glioma cells can functionally “co-opt” surrounding astrocytes, which would help in the initial stages of tumor dispersion. ^, A recent experimental model has confirmed this co-optation by showing that glioma cells secrete molecules of the tumor necrosis factor (TNF) superfamily that do not promote their own invasion but activate astrocytes instead, which in turn respond with their own proinvasive factors in vivo.

Histologic analysis of the tissue surrounding human gliomas reveals a broad band of astrocytes with the characteristic reactive phenotype that is observed in other neural injuries: hypertrophic and highly ramified bodies, high expression of intermediate filament proteins such as glial fibrillary acidic protein (GFAP) and vimentin (see Fig. 138.1A ), and formation of a glial scar rich in chondroitin sulfate proteoglycans, which extends several hundreds of microns around the diffuse borders of the tumor. ^, The glial scar isolates damaged neural tissue and restricts cell motility and axonal extension into or through the injury epicenter. However, current evidence suggests that reactive astrogliosis around malignant gliomas does not limit the dispersion of the tumor cells and may even contribute to tumor invasion through the accumulation of astrocyte-secreted trophic factors in the periphery of the tumor. ^{, ,}

The communication between glioma cells and the reactive astrocytes found within and around the tumor is complex and bidirectional. Soluble factors secreted by glioma cells promote astrocyte proliferation and likely contribute to exacerbating the peritumoral gliosis. These soluble factors have not been completely elucidated but are known to include the transforming growth factors α and β (TGF-α and TGF-β), platelet-derived growth factor (PDGF), TNF molecules (mentioned earlier), and cytokines such as CXCL12. ^, Astrocytes and astrocyte precursors not only proliferate and increase their GFAP immunoreactivity in the presence of glioma cells but also exhibit changes in their expression of molecules that regulate ECM structure. For example, astrocytes co-cultured with glioblastoma cells show increased expression of MMP-2 ^, and decreased expression of the MMP inhibitor TIMP-2, both of which may contribute to increased peritumoral ECM degradation and therefore to tumor expansion and invasion. Similarly, GFAP-reactive astrocytes recovered from PDGF-induced glioblastomas have been shown to upregulate MMP-10 (a protease from the stromelysin family), tenascin-C (an ECM protein in the perivascular niche), and basal lamina–associated collagen type VI, all of which can also contribute to increased tumor cell adhesion and migration. Indeed, glioma-initiating cells co-cultured with astrocytes or their conditioned medium exhibit increased expression of promitotic and proinvasive genes, as well as increased migration toward glial cells.

Interesting to note, glioma cells and astrocytes interact not only through a two-way paracrine communication mediated by soluble factors, but also by direct physical association through the formation of gap junctions between normal and malignant cells. These gap junctions, identified by expression of connexin-43, have been demonstrated as functional connections that permit coordinated propagation of intercellular calcium signaling, and may increase the invasive potential of glioblastoma cells through the neural parenchyma. ^, More recently, co-culture experiments have shown that astrocytes increase the resistance of glioma cells to the standard chemotherapeutics temozolomide, doxorubicin, and vincristine. The increased apoptotic resistance is, at least in part, mediated by intercellular communication through gap junctions between glioma cells and astrocytes.

It is known that neuronal signaling can produce factors that promote glioma growth—for example, via the secretion of neuroligin 3 (NLGN3) —and data show that direct synaptic interactions between glioma cells and normal neurons may play an important role in tumor biology, providing a fundamental new concept in the field. Glioma cells in mouse models can integrate into neuronal networks via synapses, and postsynaptic currents mediated by glutamate receptors of the AMPA subtype can drive tumor progression. ^,