Main

Glioblastoma (GBM; WHO (World Health Organization) grade IV) is the most common and highest-grade primary brain tumor and has a truly dismal prognosis.1 Following surgical resection and radiation therapy—the current therapeutic gold standard—mean survival is only 50 weeks.2 Perhaps more telling of its explosive natural behavior, survival of patients treated by surgical resection alone averages 14 weeks.3 Lower-grade astrocytomas (ie WHO grade II and III astrocytomas) are also ultimately fatal, but have substantially slower growth rates and longer survivals (3–8 years). With rare exception, patients with lower-grade tumors will eventually die from progression of their tumors to GBM, with short survival periods following this transition. Thus, whether arising de novo or from a lower-grade precursor, transition to GBM represents an abrupt turning point, with rapid clinical progression leading to death. The biologic properties that make the infiltrating astrocytomas fatal are known: their diffuse infiltration makes complete surgical resection nearly impossible; they are generally resistant to adjuvant therapies; and they eventually invade nervous system structures critical for life. In contrast, reasons for the explosive growth following malignant progression to GBM have not been fully explained. This review focuses on an emerging concept regarding mechanisms responsible for the accelerated growth in GBM, emphasizing a link between intravascular thrombosis, hypoxia, necrosis, and angiogenesis.

Accelerated growth in GBM

Neuroimaging and pathologic features that distinguish GBMs from lower-grade astrocytomas (grades II and III) provide clues to their escalated growth potential. By magnetic resonance imaging (MRI), lower-grade lesions are mildly expansile and show signal abnormalities—best seen on T2-weighted and FLAIR images—reflecting vasogenic edema generated in response to infiltrative tumor cells (Figure 1). These tumors show minimal or no enhancement following administration of intravenous contrast agents due to an intact blood–brain barrier and a lack of tumor necrosis. In tissue sections, grade II and III astrocytomas consist of neoplastic cells diffusely infiltrating brain parenchyma. Cell density, nuclear anaplasia, and proliferation indices increase as tumors progress. Measurements of the radial growth rates of grade II astrocytomas AAs in patients who have not received adjuvant therapy demonstrate a modest annual increase of 2 mm/year.4, 5 Slow, persistent expansion characterizes these grade II–III astrocytomas.

Figure 1
figure 1

AA (WHO grade III) and GBM (WHO grade IV) have distinct growth patterns by MRI. (a) A 47-year-old woman with seizures was found on MRI to have a left frontal lobe lesion, which on biopsy proved to be an AA. Axial MR images shows increased signal intensity on FLAIR imaging, representing vasogenic edema in response to infiltrating tumor cells (arrow). (b) Following the administration of contrast agents, most AAs do not show enhancement on T1-weighted images due to an intact blood–brain barrier and a lack of central necrosis (arrow). (c) After 1 year, the same patient developed worsening symptoms and was rescanned, revealing progression to GBM. GBMs also have hyperintense regions at their periphery on FLAIR imaging, indicative of diffusely infiltrating tumor cells (arrow). (d) A distinguishing feature is the emergence of a contrast-enhancing component centrally, which rapidly expands radially (arrow). Note the degree of midline shift in the GBM that was not present in the AA.

Full size image

Growth patterns change dramatically following transition to GBM. Radial growth rates can accelerate to values nearly 10 times greater than those in grade II astrocytomas.5 Moreover, a central, contrast-enhancing component emerges from within the infiltrative astrocytoma and rapidly expands outward (Figure 1). In tissue sections, the pathologic features that distinguish GBM from lower-grade astrocytomas are found near the contrast-enhancing rim and include (1) necrotic foci, usually with evidence of peripheral cellular pseudopalisading, and (2) microvascular hyperplasia.2 Pseudopalisades are dense collections of tumor cells surrounding necrosis that are pathophysiologically linked with adjacent microvascular hyperplasia (Figure 2). The latter is an exuberant form of angiogenesis that arises in response to the secretion of proangiogenic factors (VEGF, interleukin-8) by pseudopalisades.6 Pseudopalisading necrosis and microvascular hyperplasia are two of the most powerful predictors of poor prognosis among the diffuse gliomas.7 Rather than mere morphologic markers, these structures are almost certainly mechanistically instrumental in the accelerated growth properties that characterize the grade III to IV transition.

Figure 2
figure 2

Pathologic features of GBM. (a) Pseudopalisades are characterized by an accumulation of tumor cells around central necrosis and typically have internal, peripheral zones of fibrillarity. (b) Microvascular hyperplasia (arrow) is a form of angiogenesis that is induced by hypoxic pseudopalisading cells and is usually present in regions adjacent to necrosis. (c) Low and high (inset) magnification of immunohistochemistry for HIF-1α performed on a human GBM demonstrating increased nuclear staining in pseudopalisades (arrow) around necrosis (asterisk) but not adjacent astrocytoma. (d), (e) Intravascular thrombosis (arrows) within pseudopalisades at different stages of development. (d) A central vessel occluded by thrombus (arrow) is dilated proximal to the occlusion and surrounded by delicate fibrillarity and a low density of tumor cells, most likely due to migration of pseudopalisading cells (arrowhead) away from perivascular hypoxia. (e) Necrosis is present adjacent to a thrombosed vessel (arrow) within a developing pseudopalisade (arrowhead). (f), (g) Fibrin within an unorganized intravascular thrombus in a GBM stains with PTAH (arrow), while aggregated platelets (arrow) and vascular walls (arrowhead) are immunoreactive for factor VIII.

Full size image

Angiogenesis supports peripheral tumor growth

While necrosis has long been recognized as a marker of aggressive behavior in diffuse gliomas, by itself it does not explain rapid tumor progression. Indeed, tumor cell death is the goal of most adjuvant therapies. Rather, pseudopalisades that surround necrosis are intimately related to microvascular hyperplasia, which provides a new vasculature for rapid neoplastic expansion. Pseudopalisading cells are hypoxic and express high levels of hypoxia-inducible regulators of angiogenesis including vascular endothelial growth factor (VEGF).8, 9 VEGF concentrations have been found to be 200–300-fold higher in the cystic fluid of human GBMs than in serum.10 Such high levels result from hypoxia-inducible factor (HIF)-mediated transcription of the VEGF gene by hypoxic pseudopalisades.11, 12 Inhibition of this pathway suppresses tumor growth.13 Once expressed and secreted, extracellular VEGF binds to its high-affinity receptors, VEGFR-1 and VEGFR-2, which are upregulated on endothelial cells of high-grade gliomas, but not in normal brain.14 Receptor binding to VEGF, in turn, leads to angiogenesis in hypoxic regions adjacent to pseudopalisades, eventually leading to a vascular density in GBMs that is among the highest of all human neoplasms. The precise type of angiogenesis, microvascular hyperplasia, is characterized by numerous, enlarged, rapidly dividing endothelial cells, pericytes and smooth muscle cells that form tufted microaggregates at the leading edge of sprouting vessels.15 In its most florid form, angiogenesis takes the shape of ‘glomeruloid bodies’—a feature that is most characteristic of GBM, but is also an independent marker of poor prognosis in prostate cancer, breast cancer, and melanoma.16 Since necrosis and hypoxia are located in GBM's core and near the contrast-enhancing rim, the most biologically relevant hypoxia-induced angiogenesis occurs futher peripherally, favoring neoplastic growth outward. The permissive nature of the central nervous system parenchymal matrix to diffuse infiltration by individual glioma cells allows for this burst of peripheral expansion.17

Necrogenesis in GBM

The relationship between pseudopalisades and angiogenesis has been fairly well established, yet mechanisms initiating cell death (necrosis and/or apoptosis) and pseudopalisade formation have not. Cell death is fundamental to GBM, occurring as apoptosis, programmmed death of individual cells, and coagulative necrosis, the sheet-like coalescence of dying tissue caused by extrinsic factors. Apoptosis can be initiated through mechanisms that include death receptor (DR) ligation by members of the tumor necrosis factor (TNF) family. Best characterized in GBM are FasL (CD95L)/Fas (CD95) and TNF-related apoptosis-inducing ligand (TRAIL)/DR4/5. TRAIL induces apoptosis most effectively by binding to DR5 and signaling through its cytoplasmic death domains, ultimately activating caspase-8.18, 19 Both Fas and FasL messenger RNA (mRNA) levels are higher in astrocytomas than normal brain and correlate with tumor grade.20, 21 Most Fas expression in GBM is within pseudopalisades corresponding to their higher levels of apoptosis.22, 23 Contact between tumor cells expressing FasL and those expressing Fas may facilitate apoptosis.24 Overall, however, levels of apoptosis are generally low in malignant gliomas; apoptosis accounts for a slim minority of cell death; and apoptotic rates do not correlate with prognosis.25, 26

In contrast, coagulative necrosis represents the majority of cell death in GBMs and its degree is inversely related to patient survival.2, 27, 28 The mere presence of focal necrosis remains one of the strongest independent markers of poor prognosis.7, 29, 30 Coagulative necrosis is seen pathologically as either microscopic foci with surrounding pseudopalisading cells (‘pseudopalisading necrosis’) or as large confluent expanses of dying tumor and nontumor elements. Pseudopalisades are characteristic of GBM and likely due to tumor and stromal characteristics. When small (<100 μm), they often surround fibrillar zones (without frank necrosis); when larger (>500 μm), they always contain central necrosis and most show peripheral fibrillarity.2 Their morphologic spectrum suggests they evolve over time to incorporate larger areas. Establishing mechanisms underlying pseudopalisade formation and its role in necrosis is critical since they potentially could precede all forms of coagulative necrosis in GBM.

Classic infarction could be involved, in which tumor necrosis would result from the metabolic demands of an enlarging tumor exceeding its nutrient supply.31 Reduced oxidative phosphorylation and intracellular ATP culminate in breached membrane integrity and proteolysis. Infarction could also arise secondary to a compromised vascular supply. Increased interstitial pressure due to vasogenic edema is typical of GBM and could promote vascular collapse. Neoplastic cells themselves could alter vascular integrity by invading perivascular spaces and vascular walls.32 Lastly, intravascular thrombosis—the formation of a fibrin-platelet clot—occurs frequently in GBM and could cause vaso-occlusion and infarction. Procoagulation factors such as tissue factor (TF) and plasminogen activator inhibitor 1 (PAI-1) are increased in GBM and correlate with the extent of necrosis.33, 34 Levels of tissue type plasminogen activator (tPA), which lyses fibrin clots, are decreased.35 An environment favoring intravascular thrombosis could result in compromised blood flow, ischemia, and necrosis.

What is the pathogenesis of pseudo-palisades?

Recent studies have shed light on the initial events of pseudopalisade formation, which may in turn explain necrogenesis.36 These investigations entertained and then formally excluded other potential explanations that might result in an increased tumor cell density around a central area of necrosis. First, pseudopalisading cells were found to be less proliferative than adjacent astrocytoma cells, indicating that they are not rapidly dividing cells that have ‘outgrown their blood supply’ as has been widely believed. Second, they were not due to an influx of other, non-neoplastic cells such as inflammatory cells, but rather composed predominantly of astrocytic tumor cells. Third, pseudopalisades showed increased levels of apoptosis compared to adjacent astrocytoma, indicating that they do not accumulate due to a survival advantage. Instead, pseudopalisades are most likely formed by a wave of actively migrating tumor cells that are moving away from a central area of hypoxia. Supporting this contention, cells in pseudopalisades exhibit an adaptive response to hypoxia as evidenced by their nuclear overexpression of HIF-1α (Figure 2c);36, 37 hypoxic GBM cells migrate more rapidly than normoxic cells in vitro; and both hypoxic cells in vitro and pseudopalisades in GBM tissue sections show expression of extracellular matrix proteases associated with invasion, including MMP-2 and uPAR.38, 39, 40, 41, 42, 43 Secretion of MMP-2 activity by gliomas is markedly increased under hypoxic conditions. Together, the evidence suggests that pseudopalisades are formed by hypoxic, actively migrating neoplastic cells that have imposed themselves on a less mobile population.

At least two mechanisms could explain hypoxia-induced migration that leads to pseudopalisades around necrosis. One possibility is that cells at greatest distance from arterial supplies become hypoxic after a critical point in tumor growth due to increased metabolic demands (Figure 3, Mechanism 1). Migration outward towards nearby vessels would leave a central zone filled with fibrillar processes as well as nonmigrated cells that eventually become necrotic. An alternative explanation is that blood vessels within the neoplasm become occluded or collapse, resulting in perivascular tumor hypoxia and cellular migration away, both at the site of occlusion, and proximal and distal to it (Figure 3, Mechanism 2). In favor of the latter, many pseudopalisades have a long, narrow, and winding (‘serpiginous’) pattern when viewed in longitudinal tissue sections, suggesting an underlying vascular substrate associated with their emergence. Chief among arguments opposing vaso-occlusion as a mechanism of necrosis is that pseudopalisades do not always appear to contain a central blood vessel, occluded or not. However, the morphologic identification of such vascular structures is impossible in many instances given the extensive necrosis within pseudopalisades. In addition, tissue sampling most likely accounts for an under-representation of dysfunctional vascular structures within pseudopalisades. The idea that vascular pathology precedes pseudopalisading necrosis in GBMs deserves further attention because both experimental models and pathologic observations support such a mechanism.

Figure 3
figure 3

Potential mechanisms of pseudopalisade formation due to hypoxia-induced migration. (1) One possibility is that tumor cells at greatest distance from arterial supplies become hypoxic (shaded in blue) after a critical point in tumor growth due to increased metabolic demands and migrate toward peripheral vessels, leaving a central clear zone. (2) Alternatively, vascular occlusion or collapse within the neoplasm could lead to central, perivascular hypoxia, followed by tumor cell migration toward a viable blood supply. Illustration by Mica Duran.

Full size image

Mechanisms of vascular pathology in GBM

Vascular Regression

Experimental animal models of gliomas have suggested that the initial, preangiogenic stages of neoplastic growth actually involve regression of native blood vessels. New tumor cells first gain access to a vascular supply by ‘co-opting’ the host's blood vessels.44, 45 In response to co-option, vascular endothelial cells upregulate angiopoietin-2 (Ang-2) by mechanisms that are poorly understood. Ang-2 acts in an autocrine fashion on tumoral blood vessels as a Tie-2 receptor antagonist. In the absence of VEGF, Tie-2 receptor blockage leads to vascular destabilization, endothelial cell apoptosis, and vascular regression. Only neoplastic foci that can induce angiogenesis in this setting remain viable and grow. Typically, this involves the HIF-dependent expression of VEGF. In human specimens, Ang-2 is expressed by endothelial cells of high-grade gliomas but not low-grade gliomas or normal brain.46, 47 Its expression precedes endothelial apoptosis and the disruption of endothelial–extracellular matrix interaction, indicating that it could cause vascular injury.45 Other arguments hold that Ang-2 induces vascular changes, but does not induce apoptosis. For example, based on intravital epifluorescence microscopy of VEGF, VEGFR-2, and Ang-2 in growing gliomas, Ang-2 expression was found to correlate with vascular remodeling and sprouting, but not with apoptosis.48 Indeed, it is clear that Ang-2 mediates vascular remodeling in the presence of VEGF.49 However, the weight of evidence indicates that Ang-2 antagonism of Tie-2 in the absence of VEGF has more severe effects on endothelium, causing apoptosis and regression.50, 51 Ang-2-mediated vascular pathology could initiate the hypoxic conditions that precede tumor cell migration and pseudopalisade formation in GBM.

Intravascular Thrombosis and Vaso-occlusion

Intravascular thrombosis, either through engagement of intrinsic or extrinsic pathways of coagulation, could substantially contribute to vaso-occlusive mechanisms underlying necrosis in GBM, whether secondary to Ang-2-mediated vascular events or independently. A strong relationship between abnormal blood clotting and human malignancy is well established.52 Patients with GBM are at high risk for developing deep vein thrombosis in the lower extremities as well as pulmonary thromboembolism, indicating a profound systemic disturbance in coagulation.53 Perhaps less appreciated, intravascular thrombosis is a frequent finding both macroscopically and microscopically within GBM tissue, noted histologically in over 90% of well-sampled cases.33, 36, 54 Importantly, the frequencies of intravascular thrombosis in neoplastic tissue and at distant sites are much higher in patients with GBM than AA. Critical prothrombotic events must occur in this transition.

A recent morphologic survey of vascular pathology and intravascular thrombosis within human GBM specimens found that over 50% of pseudopalisades had evidence of a central vascular lumen, either viable, degenerating, or thrombosed.36 In all, 20% of pseudopalisades contained a vessel with intravascular thrombosis as evidenced by complete luminal occlusion by an unorganized platelet and fibrin clot (Figure 2d–g). Analysis of pseudopalisade sizes and shapes led to the conclusion that tissue sampling and tangential sections could account for an underestimation of the true frequency of intravascular thrombosis within pseudopalisades, resulting in a historic underappreciation of its relevance to necrosis in GBM. Thus, we believe a substantial subset of pseudopalisades around necrosis represents hypoxic tumor cells migrating from vaso-occlusion secondary to thrombosis. There is a strong possibility that this mechanism is more generally relevant to coagulative necrosis in GBM.

What might cause increased thrombosis within GBM? At the transition of AA to GBM, numerous tumoral and vascular changes likely promote blood clotting. Normal vessels of the central nervous system show minuscule levels of protein diffusion though their walls due to a highly restrictive blood–brain barrier. The latter is formed primarily by endothelial tight junctions, but also has contributions from astrocytic foot plates, extracellular matrix, and endothelial–pericytic interactions.55, 56 This barrier becomes breached in GBM, which can be visualized radiologically by their high degree of contrast enhancement. Such enhancement is due to the increased permeability of vessel walls to both intravenously injected contrast agents and to proteins that bind them, such as albumin.57, 58 As a result, contrast agents leak into tumoral but not normal brain tissue. Microscopically, there is increased fenestration of vessel walls, pericyte detachment, and alteration of the extracellular matrix.59 All events that initate this increased permeability have not been completely defined, but VEGF expression by neoplastic cells is a known cause of increased vascular leakage.60, 61 The end result is that plasma coagulation factors enter tissue spaces where they are activated, leading to the formation of a fibrin plug and platelet aggregation.

Tissue factor

A critical regulator of tissue hemostasis and a potent stimulant of thrombosis is TF, a 47 kDa transmembrane glycoprotein receptor.62 In normal tissue, TF is expressed almost exclusively by stromal cells. A disruption of vascular integrity is necessary for TF to bind to its activating ligand from the plasma, factor VII/VIIa, in order to initiate thrombosis. The end result of TF/factor VIIa activation is the generation of thrombin from prothrombin, causing platelet aggregation, fibrin deposition, and local hemostasis. The tight regulation of TF expression that is present in normal tissue is lost in pathologic conditions including neoplasia. A variety of cancers show increased expression of TF by neoplastic cells, stroma, and endothelium. A direct correlation between TF levels and tumor grade has been noted for multiple tumor types,63, 64, 65 including gliomas.34 TF protein is expressed in over 90% of malignant astrocytomas, but only in 10% of grade I and II astrocytomas. In situ hybridization localized its expression to neoplastic cells. Similarly, a reverse transcriptase-polymerase chain reaction study found that, on average, TF mRNA levels were 500 × greater in astrocytomas than normal brain tissue and that its levels correlated with histologic grade.66 Precise mechanisms responsible for increased TF expression by gliomas are not known. Tissue hypoxia leads to increased rates of thrombosis as well as increased TF expression by mononuclear phagocytes. The latter depends on Egr1-mediated transcription rather than HIF.67, 68, 69 Whether tumor hypoxia leads to increased TF expression in gliomas is unknown. Nonetheless, its high expression in GBM could explain the strong tendency toward thrombosis.

Activation of plasma coagulation factors, thrombin in particular, has biologic significance beyond clot formation. Thrombin is a potent physiologic activator of protease-activated receptors (PARs), a family of G-protein-coupled transmembrane receptors. PAR1, the family's prototype, is activated when thrombin cleaves its amino-terminal extracellular domain and unmasks a new N-terminus, which then serves as the receptor's ligand.70 Activated PAR1 tranduces intracellular signals by coupling to G-proteins, predominantly Gαi, Gαq, and Gα12/13.71 Secondary signals are generated through Rho, phospholipase C (IP3 and diacylglycerol), and inhibition of adenyl cyclase. Although PAR1 is expressed at low levels in most normal epithelia, it is aberrantly overexpressed by a variety of carcinomas including those of breast, colon, lung, and stomach.72, 73 PAR1 activation can transform cells and is able to enhance tumorigenicity, in large part by signaling through Gαq and Gα13.74, 75 It is also clear that PAR1 activation promotes invasive and metastatic properties of malignant cells.72, 76 Mechanisms of increased invasion include its ability to direct cytoskeletal actin rearrangements, phosphorylation of focal adhesion kinases, and recruitment of αvβ5 integrin to contact sites.77 PAR1 protein is present in both human and mouse brain, mostly in astrocytes, where it can be activated by thrombin.78, 79 More recent investigations of human GBM cell lines and short-term cultures derived from human specimens have demonstrated PAR1 expression on tumor cells. Its activation by thrombin and PAR1 agonists causes increased phospho-inositol (PI) hydrolysis and calcium mobilization, presumably coupling through Gαq.79 More work will be required to determine the biologic relevance of TF activation of thrombin and consequent PAR1 signaling in gliomas.

Conclusion

Pseudopalisades and the ensuing microvascular hyperplasia that are associated with accelerated growth in GBM may result from the following sequence (Figure 4): (1) vascular occlusion, possibly related to Ang-2-mediated endothelial apoptosis and often associated with intravascular thrombosis; (2) hypoxia in regions surrounding vascular pathology; (3) outward migration of glioma cells away from hypoxia creating a peripherally moving wave (pseudopalisade); (4) death of nonmigrated cells leading to central necrosis; (5) secretion of soluble proangiogenic factors (VEGF, IL-8) by hypoxic pseudopalisading cells; (6) an exuberant angiogenic response creating microvascular proliferation in regions peripheral to central hypoxia; (7) enhanced outward expansion of infiltrating tumor cells toward a new vasculature. Since both necrosis and glomeruloid vascular proliferation are also markers of poor prognosis in other types of cancer,15, 16 the identification of their underlying mechanisms may have more general implications for tumor angiogenesis and malignant progression. Once identified, the pathophysiologic triggers underlying vaso-occlusion will become attractive, novel targets for antitumor therapy.

Figure 4
figure 4

Schematic representation of pseudopalisade formation in GBM. Vaso-occlusion/collapse and intravascular thrombosis lead to tissue hypoxia in the perivascular region. Hypoxic tumor cells then migrate away along cellular processes, leaving a fibrillar center initially. Hypoxia-induced cellular migration also occurs at sites proximal and distal to vaso-occlusion and thrombosis due to a lack of functional blood flow. Tumor cells that do not migrate become hypoxic and undergo apoptosis or necrosis, eventually leaving an enlarging central necrotic zone. Vessels distant to the vascular occlusion become necrotic and degenerate. Hypoxic pseudopalisading cells show increased HIF-mediated transcription and VEGF secretion, leading to microvascular hyperplasia nearby. Illustration by Mica Duran.

Full size image