Recent Insights into PDGF‐Induced Gliomagenesis

Abstract

Gliomas are aggressive and almost incurable glial brain tumors which frequently display abnormal platelet‐derived growth factor (PDGF) signaling. Evidence gained from studies on several in vivo animal models has firmly established a causal connection between aberrant PDGF signaling and the formation of some gliomas. However, only recently has significant knowledge been gained regarding crucial issues such as the glioma cell of origin and the relationship between the transforming stimulus and the cellular characteristics of the resulting tumor. Based on recent evidence, we propose that PDGF can bias cell‐fate decisions, driving the acquisition of cell type‐specific features by the progeny of multipotent neural progenitors, thus determining the shape and direction of the transformation path. Furthermore, recent data about the cellular mechanisms of PDGF‐driven glioma progression and maintenance indicate that PDGF may be required, unexpectedly, to override cell contact inhibition and promote glioma cell infiltration rather than to stimulate cell proliferation.

PLATELET‐DERIVED GROWTH FACTOR (PDGF) SIGNALING

In vertebrates, the PDGF family of ligands consists of four members (PDGF‐A to ‐D), while two different receptors exist [platelet‐derived growth factor receptor (PDGFR) alpha and beta], which belong to a receptor tyrosine kinase superfamily comprising also the receptors for the vascular endothelial growth factor (VEGF) family of ligands. PDGF‐A/B ligands occur as disulfide bridged homo‐ and heterodimers, while PDGF‐C and ‐D homodimerize (Figure 1A). The two receptors bind various different ligand dimer combinations and can, in turn, heterodimerize (Figure 1A), activating similar but distinct signaling pathways [4, 42, 46) for comprehensive reviews of PDGF signaling functions]. In agreement with this, the phenotypic defects observed upon loss of either receptor are rescued in knock‐in mice expressing chimaeric receptors in which the intracellular domain of the knocked‐out receptor form has been replaced by that of the other (53). This strongly indicates that the different phenotypes displayed by the knockout mice for the two PDGFRs 83, 84 are mostly because of changes in the spatiotemporal patterns of gene expression rather than of differences in intrinsic biochemical activities. However, signaling specificities exist and different signaling pathways have been shown to regulate distinct processes, in vitro and in vivo. Knockout and knock‐in mice, harboring PDGFRα forms with mutations specifically affecting distinct signaling pathways, revealed specific and overlapping functions for phosphatidylinositol 3‐kinase (PI3K)‐, Rous sarcoma oncogene (Src)‐ and phospholipase‐C gamma (PLCγ)‐modulated pathways in various aspects of mouse embryonic development, like oligodendrocyte lineage and skeletal development [see below, 54, 66].

Figure 1.

Platelet‐derived growth factor (PDGF) ligands and receptors and their effects on the glial lineage. A. Known members of the PDGF families of ligands and receptors; arrows indicate their known in vitro dimerization properties and their interaction with receptors. B. PDGF can induce neural progenitor cells (NPCs) to acquire an oligodendrocyte progenitor cell (OPC) identity. C. PDGF exerts its effects on OPC migration and proliferation by activating different signaling pathways in a concentration‐dependent manner. D. Neuronal and oligodendroglial generation by subventricular zone (SVZ) neural stem cells are regulated by PDGF signaling. CC = corpus callosum; LV = lateral ventricle; OB = olfactory bulb; PDGFR = platelet‐derived growth factor receptor; PI3K = phosphatidylinositol 3‐kinase; PLCγ = phospholipase‐C gamma. References are present for data summarized in B, C and D. The figure was drawn on the bases of the data contained in the cited papers and represents an original artwork.

PDGF ROLES IN CENTRAL NERVOUS SYSTEM (CNS) DEVELOPMENT AND REPAIR

PDGF signaling exerts multiple functions during CNS development. PDGF‐A and PDGFRα are responsible for glial progenitor proliferation and survival, during oligodendrocyte development. PDGF‐A and PDGFRα knockout mouse lines show defective CNS myelination, due to defective survival, migration and differentiation of oligodendrocyte progenitor cells [OPCs; 10, 17, 34, 54]. The down‐regulation of PDGFRα expression, due to the loss of activity of the putative transcriptional activators Sox9/10, has a similar effect (31). The contributions of different signaling pathways to PDGFRα activity in regulating the development of the oligodendrocyte lineage have been analyzed in vitro and in vivo by studying various mutants showing defective pathway activation. Activation of both Src and PI3K is necessary for proper CNS myelination (54). In vitro, PI3K and PLCγ have been shown to relay PDGFRα signals triggered by distinct concentrations of PDGF, with PI3K responding to the low concentrations and PLCγ responding to the high concentrations. This is reflected by the differential effects of PDGF on key aspects of OPC biology in vivo (66); while proliferation is promoted by low and high PDGF concentrations, other aspects, like migration and differentiation, are stimulated preferentially by low PDGF concentration via the activation of the PI3K pathway [Figure 1C, (66)].

While PDGFs are mostly known for their mitogenic effects on embryonic, perinatal and adult OPCs 88, 90, embryonic mouse and rat telencephalic neural progenitor cells (NPCs) can also respond to PDGF signaling by acquiring glial progenitor features in vitro[Figure 1B; 6, 47], suggesting PDGF may also act instructively, rather than just as a mitogen, in modulating the size of the embryonic OPC pool. Similarly, PDGF can bias the progeny's fate of PDGFRα‐expressing cells in the subventricular zone (SVZ) which are a subset of bipotent adult neural stem cells (NSCs). These cells can generate both oligodendrocytes and neurons, and PDGF administration in vivo induces a shift toward oligodendrocyte production, at the expense of neuron generation [Figure 1D; 50, 67], suggesting a cell‐fate modulatory role for PDGF also in the postnatal CNS. Furthermore, studies on rodent models of acute and chronic demyelination have shown that myelin repair relies in part on PDGF‐induced expansion of the OPC pool 33, 67, 71, 91, demonstrating a role for PDGF in many aspects of adult CNS homeostasis and further highlighting its relevance in the control of OPC proliferation and differentiation. Many other signaling molecules influence the oligodendrocyte lineage, such as fibroblast growth factor family members [FGFs; 7, 8, 13, 32, 69, 70, 73, 74] and epidermal growth factor [40, 49)]. Moreover, PDGF signaling may crosstalk with pathways activated by these factors. Indeed, cooperation between PDGF and FGF2 sustains the in vitro expansion and inhibition of differentiation of neonatal and adult OPCs 13, 90, while FGF2 affects PDGFRα expression by OPCs (65).

PDGF IN GLIOMAS

PDGF signaling is altered in a wide spectrum of human tumors, and its relevance to the biology of gliomas will be the main focus of this article. Prior to discussing the roles of PDGF signaling alteration in gliomagenesis, we will briefly try to outline the current understanding of the classification of gliomas and of the molecular lesions associated with them.

Gliomas are the most aggressive tumors originating in the CNS 51, 52, 75 and are traditionally classified according to their histopathology and the exhibition of cell type‐specific features resulting in a very complex classification scheme. Current classification recognizes several groups of gliomas, among which the most prominent are astrocytic, oligodendroglial, oligoastrocytic and ependymal tumors. Gliomas are further subdivided into grades I–IV with grade IV being the most malignant [see (60) and references therein for a review of the current classification scheme]. Currently, only astrocytic tumors are recognized to reach grade IV even if the presence of grade IV oligodendroglial forms is increasingly appreciated (60). Despite the common practice of diagnosing gliomas according to these criteria, increasing awareness is being drawn to the poor correspondence between histopathological classes and biological proprieties. A recent analysis based on genome wide expression profiling data revealed, for instance, that tumors histopathologically classified as astrocytomas randomly distribute across classes defined by very different biological signatures and the same is observed for oligodendrogliomas (57). This indicates that distinct tumors are often confused and considered as a uniform entity. The histopathological classification of gliomas is further weakened by not taking in due consideration recent findings in developmental biology showing that the barrier between neural cell lineages is not as rigid as thought previously. Lineage analysis in the murine brain showed the existence of progenitor cells that are able to generate astrocytes and oligodendrocytes, at least at perinatal stages 24, 64, 92. In support of this view, primary astrocytomas and oligodendrogliomas display frequent mutations at the same loci, as in the case of the Isocytrate dehydrogenase 1 (IDH1) gene 39, 75, 89, suggesting that some of these tumors may actually share their cell of origin.

The incurable nature of gliomas is in part the result of their poor accessibility to surgical and pharmacological treatments, but is mostly a consequence of their infiltrative activity, which makes complete surgical resection almost impossible. The molecular bases for such properties are currently being intensely investigated, but the majority of results have been obtained using established human or rodent cell lines for in vitro studies or for xenotransplant assays. Because of the strong differences that these cell lines display, when compared with primary human glioma samples, in terms of gene expression profiles and some general biological properties 55, 56, some caution is necessary when interpreting these results, as they may be influenced by the experimental system. Recent studies have, however, made more frequent use of primary human gliomas, and knowledge is rapidly accumulating. Currently, little is known about the cellular targets of transformation within the human brain, even if these have tentatively been deduced from the molecular phenotypes displayed by the different glioma subtypes. However, inferences about the identity of the cells of origin of the various forms can be made by analyzing the molecular phenotypes of gliomas generated in vivo by targeting specific neural cell populations in animal models. Indeed, the targeting of specific cell populations with oncogenic stimuli is significantly improving our understanding of how various neural cell lineages are affected by individual or combined molecular alterations, as further discussed below 2, 9.

Different molecular lesions have been identified as enriched in specific glioma subtypes 35, 38. Although an exhaustive coverage of all the molecular alterations reported in gliomas is far beyond the scope of this article, we summarized in Table 1 the most common alterations reported to be associated with human gliomas. However, we draw the attention of the reader to the correlative nature of these data which critically rely on histopathological classifications whose validity is poorly supported by recent analyses as mentioned above 57, 58. This might explain why different reports provide divergent estimates of lesion frequencies.

Table 1.

Common genetic lesions in human gliomas. From 35, 39, 75 asterisks denote presence of indicated alterations in the various tumor groups.

As shown in Table 1, a very common alteration in gliomas is the overexpression of PDGF ligands or their receptor PDGFRα which occurs frequently in primary and secondary glioblastoma, anaplastic astrocytomas and even low‐grade oligodendrogliomas 1, 35, 38. Typically, PDGF pathway overactivity in gliomas results from the amplification‐dependent overexpression of the receptors or the ligands. Receptor‐activating mutations are comparatively rare events. Both receptors and ligands can be coexpressed by tumor cells 44, 72, endothelial and mural cells around blood vessels, indicating that autocrine/paracrine signaling may play an important role in the establishment and maintenance of these tumors (43). The broad range of human gliomas displaying altered PDGF pathway activity strongly suggests that this signaling axis plays central roles in the events underlying gliomagenesis. However, not much is known, in general, about the in vivo mechanisms of PDGF signaling during the initiation and progression of human tumors 25, 76. PDGF ligands are potent mitogens for many cell types, such as neural progenitors, members of the oligodendrocyte lineage and adult NSCs. Furthermore, PDGF signaling functionally interacts with signals evoked by other secreted molecules, such as VEGF and FGF2, in the modulation of angiogenic processes (pericyte activation and recruitment), which is reflected by the high expression of both PDGFRα and vascular endothelial growth factor receptor 2 (VEGFR2) in many human gliomas (77), and by the ability of PDGF signaling to up‐regulate VEGF/VEGFR expression 37, 42. It seems, therefore, likely that in the context of human brain tumor biology, PDGF signaling is relevant for both tumor expansion and survival, stimulating proliferation and indirectly promoting nutrient supply to the tumor mass (28). As a possible example of how PDGF signaling may promote glioma cell survival, it was recently shown that PDGFR signaling triggers autophagy in tumor cells in response to severe hypoxia. Little is, however, known about the functional significance of PDGF signaling in supporting human glioma survival and growth, and the use of cell lines may be of limited utility because of the deep differences they show when compared with the modeled cell types 55, 56. Data obtained using animal models, though, have shed some light on the role of PDGF in driving and supporting glial tumorigenesis, with potential implications for the human pathology, as detailed below.

PDGF‐DRIVEN GLIOMAGENESIS IN ANIMAL MODELS

As first demonstrated by Uhrbom and colleagues, in vivo PDGF overexpression in neonatal mouse NSCs efficiently induces the formation of gliomas [Figure 2A; 26, 45, 87]. In these experiments, diverse sets of proliferating neural cell types could potentially be targeted, comprising late radial glial cells and various glial progenitors residing in the SVZ or close to it, which can generate oligodendrocytes, astrocytes and neurons 15, 16, 22, 36, 64, 68. These tumors, albeit histopathologically heterogeneous, were consistently glial fibrillary acidic protein (GFAP) negative, supporting their identification as nonastrocytic gliomas. Similarly, different adult neural cell populations can be transformed in vivo by PDGF signaling. Adult white matter OPCs (9) and SVZ NSCs (50) respond to intense PDGF signaling by extensively proliferating and by generating gliomas (Figure 2B). Transduction of adult forebrain white matter glial progenitor cells with PDGF‐B quickly produces glioblastomas with clear oligodendroglioma features, and the same happens if white matter‐derived oligodendrocyte progenitors are transduced ex vivo, followed by intracranial transplantation, confirming the ability of this specific cell population to be transformed by PDGF (9). Analogously, continuous intraventricular perfusion of PDGF‐AA induces proliferation of PDGFRα‐positive adult NSCs that populate the SVZ, eventually resulting in the formation of glioma‐like lesions (50). Altogether, these data show that gliomas arising as a consequence of aberrant PDGF signaling to different neural precursors invariably show oligodendroglial features while lacking astrocytic characteristics 9, 26, 50. The data gained from in vivo studies on adult rodents suggest that the abundant proliferating Ng2+ adult human oligodendrocyte progenitors (79) may be the preferential targets of PDGF‐induced transformation leading to oligodendroglioma formation.

Figure 2.

Summary of in vivo models of platelet‐derived growth factor (PDGF)‐induced gliomagenesis; their basic experimental settings, targeted cell populations and tumorigenic effects. A. PDGF‐B transduction of neonatal neural precursors using the RCAS‐tva avian retroviral system. B. PDGF‐B/GFP transduction of adult white matter progenitors (upper cartoon) and continuous PDGF‐A intraventricular perfusion of the adult brain (lower cartoon). C. In utero transduction of telencephalic neural progenitors with PDGF‐B/GFP retroviral vectors. References are present for each model. The figure was drawn on the bases of the data contained in the cited papers and represents an original artwork. GFP = green fluorescent protein; RCAS = Replication‐Competent ASLV long terminal repeat with a Splice acceptor; SVZ = subventricular zone.

The consistent generation of pure oligodendrogliomas following PDGF overexpression may be caused by the selective transformation of a PDGF‐responsive subset of oligodendrocyte‐restricted neural progenitors or may reflect an instructive activity inducing an oligodendroglial identity. This issue is critical to understand the diversity of the basic processes underlying glial tumorigenesis. While the generation of oligodendrogliomas following the transduction of white matter progenitors provides little information in this regard, as the targeted cells already show oligodendroglial features (9), the generation of the same kind of tumors from the more immature progenitors residing in the SVZ suggests that PDGF acts instructively by inducing an oligodendroglial fate. This view is further substantiated by the evidence that oligodendrocytes are induced from SVZ progenitors at the expense of olfactory bulb neurons, suggesting that PDGF acts by unbalancing the normal output of adult NSCs. It must be noted that the diversity and real potency of the SVZ precursor cells are still not completely characterized, so that the lack of astrocytic tumors could, in principle, reflect fate restrictions of the targeted cells rather than an instructive effect.

Some attempts to clarify the relationship between the developmental potential of the cells of origin and the characteristics of the PDGF‐induced gliomas have been made by the retroviral targeting of cell subpopulations defined by the expression of an avian retroviral receptor under the control of the nestin‐ or human GFAP (hGFAP) promoters [n‐tva and g‐tva mouse line, respectively; (26)]. In this paradigm, while both targeted populations gave rise to oligodendrogliomas, the hGFAP‐expressing cells generated fewer tumors also comprising rare mixed oligoastrocytomas following infection with a Replication‐Competent ASLV long terminal repeat with a Splice acceptor (RCAS) retrovirus carrying PDGF‐B.

The interpretation of this result, however, is not obvious at least for two reasons. First, abundant reactive astroglial infiltrates may represent a confounding factor. Infiltrating astrocytes would make it difficult to distinguish between oligodendrogliomas and mixed oligoastrocytomas if a clear discrimination between tumor cells and infiltrating cells was not possible, as is the case in most of the cited studies which do not use a direct reporter of transduction. Indeed, the infiltrative nature of the abundant GFAP‐positive population within the PDGF‐B‐induced tumors becomes evident when a transduction reporter is employed 6, 18. Second, the actual developmental fate potentials of the targeted cells in the n‐tva and g‐tva forebrains largely overlap. This is suggested by the finding that the hGFAP‐expressing population comprises radial glial cells and various neonatal neural precursor cell populations 5, 16, 62, 63. Furthermore, at late embryonic/perinatal stages, the expression of both nestin and the hGFAP promoter marks ventricular zone multipotent radial glial cells and more committed glial progenitors 15, 68. Astrocytes outside the SVZ stem cell niche may therefore account for only a minority of the transduced cells in g‐tva mice, suggesting that the n‐tva and g‐tva models may be almost equivalent. The difference in tumor incidence following the infection of n‐tva and g‐tva could therefore be caused by lower transduction efficiency rather than by differences in intrinsic characteristics of the two populations.

More recent experiments have hit a precursor population which surely harbors multipotent progenitor cells, namely the radial glia cells of the embryonic telencephalon. This was achieved by means of in utero injections of retroviral vectors expressing PDGF‐B [Figure 2C; 6, 18]. Remarkably, despite transducing such a highly heterogeneous progenitor population, capable of giving rise to all mature CNS cell types (63), this approach led to the formation of pure oligodendrogliomas, as shown by microarray analysis. This analysis showed a surprisingly uniform gene expression profile among PDGF‐induced gliomas and revealed a closest similarity to OPCs when compared with that of a broad set of neural cell types 6, 18.

Albeit suggestive, these data do not provide evidence for the ability of PDGF‐B to bias cell‐fate choices. This aspect has been directly addressed by in vitro retroviral lineage tracing on cultured neural progenitors. Following clonal transduction with a PDGF‐B retrovirus, individual multipotent neural progenitors, typically generating a mixed neuronal and glial progeny, were found to give rise to OPCs more frequently than the control transduced cultures. Notably, the frequency of OPC‐containing clones increased without a variation in the total number of the infected progenitors (6). This shows that PDGF‐B promotes the formation of OPCs by inducing an OPC identity in immature neural progenitors or in their progeny. This was also supported by other experiments indicating that the abundance of OPCs increases significantly in cultures of embryonic neural progenitors treated with PDGF [47, 78; our unpublished observations]. Altogether, these data suggest that neural precursor fate modulation may be a relevant process in the course of glioma initiation by PDGF, possibly representing a common theme during diverse gliomagenic events. Further support to the view that PDGF may bias cell‐fate choices toward the oligodendrocyte lineage, is lent by observations showing that PDGF signaling promotes proliferation of neonatal cultured neural precursors and suppresses their astrocytic differentiation. This correlates with low levels of Akt activity, whereas astrocyte generation from these cells can be promoted by PDGFR inhibition or Akt constitutive activation 27, 86. Moreover, Akt signaling was suggested to modulate the effects of PDGF on neonatal neural progenitors, by promoting the acquisition of astroglial features in vitro and in vivo (27). In vivo coinfection with Akt and PDGF‐B retroviruses promoted the formation of mixed gliomas in which, however, only Akt‐transduced cells showed astrocytic features (27). Given the absence of a reporter for PDGF‐transduced cells, it is hard to say if the same cells had also been transduced by PDGF, but, in any case, it can be concluded that PDGF can, paracrinally or autocrinally, promote the formation of astrocytic gliomas only in conjunction with Akt signaling. On a similar note, it was shown that suppression of the PI3K/Akt/mTOR pathway causes KRas/Akt‐induced astrocytomas to lose astrocytic features gaining Olig2 expression (48). These experiments reinforce the view that Akt signaling may actively suppress an oligodendroglial program, as further suggested by the observation that Akt‐induced Olig2 nuclear extrusion is necessary for astrocytic differentiation (81). However, it is also possible that selective survival and expansion of the oligodendroglial component of mixed tumors partially account for those in vivo observations (48). In agreement with the possibility that PDGF signaling is alone unable to transform astrocytes, Jackson and colleagues (50) showed that intraventricular PDGF‐AA perfusion caused the expansion of the adult NSC compartment, leading to the generation of oligodendrogliomas, rather than astrocytomas. Also in line with this is the recent observation that PDGF overexpression is alone unable to transform GFAP‐expressing cells in an in vivo transgenic model driving PDGF‐B expression from the hGFAP promoter, which is active in adult astrocytes and SVZ NSCs (41). This could be explained by the ability of PDGF to drive NSCs toward the oligodendroglial lineage, in which hGFAP promoter may no longer be transcribed, thereby shutting off PDGF expression, in a negative feedback manner.

As discussed above, the data gained from the analysis of in vivo models of PDGF‐induced gliomagenesis suggest that the main outcome of the aberrant activation of PDGF signaling is the generation of oligodendroglial tumors. In the case of our model 6, 18, this is well supported by gene expression profile comparisons between tumor cells and primary neural cell types comprising cells of both the astrocytic and oligodendroglial lineages. This observation is seemingly at odds with the reported occurrences of PDGF signaling alterations also in human gliomas with astrocytic features. At least two nonmutually exclusive explanations can, however, reconcile these observations. First, alterations in PDGF pathway activity may represent a secondary event during the progression of at least some astrocytomas. In this scenario, despite the inability of PDGF overexpression to transform astrocytes or to give rise to astrocytomas from adult NSCs, increased PDGF signaling may still represent an advantageous secondary molecular lesion for already initiated astrocytic gliomas. Yet, it is also possible that some gliomas classified as astrocytomas actually represent oligodendroglial tumors. Such misclassification may either result from the rather unreliable nature of histopathological evaluations (as discussed above) and from the presence of abundant reactive astrocytic infiltrates. Mouse oligodendrogliomas composed of fluorescently tagged tumor cells provide a strong argument in favor of the latter possibility 6, 9, 18. Cells with astrocytic features often represent a relevant fraction of the cells within tumors, yet the ability to visualize single tumor cells clarifies that these are only recruited/activated astrocytes. It is, therefore, possible that misidentification of some gliomas as astrocytomas has contributed to the discrepancy between observations made in human and model brain tumors.

TUMOR PROGRESSION DURING PDGF‐INDUCED GLIOMAGENESIS

Overexpression of PDGF‐B in the very heterogeneous (63) NPC population present during mouse embryonic telencephalic development (embryonic day 14; E14) results in a surprisingly uniform group of oligodendrogliomas, which progress rapidly toward highly malignant stages 6, 18. Histopathological features (ie, tumor grade) correlate with both tumor onset (early/low grade vs. late/high grade) and with the ability to initiate de novo tumor formation in vivo. Early‐onset PDGF‐induced tumors are devoid of any sizeable tumor‐initiating cell subpopulation, if assayed in in vivo tumor cell reinjection experiments, while later‐arising tumors show high‐grade glioma features and can be serially propagated in vivo (Figure 3). This suggested that progression occurs in vivo in this model, in contrast to the possibility that high‐grade tumors might represent slowly growing gliomas possessing tumorigenic potential since the early steps of their growth. Notably, during the initial stages of tumor growth, all PDGF‐B‐induced tumors are equivalent in lacking any sizeable tumor‐propagating population, therefore demonstrating that a progression process underlies the acquisition of the tumorigenic potential at later stages. These observations strongly indicate that, albeit likely to be responsible for the early hyperproliferative phase, PDGF is alone insufficient to provide these cells with a full‐blown tumorigenic potential. This model of PDGF‐induced oligodendroglioma formation is therefore a useful system for investigating the genetic lesions responsible for glioma progression and, at the same time, is a good system for deciphering the bases of oncogene addiction, a still poorly understood phenomenon.

Figure 3.

Platelet‐derived growth factor (PDGF)‐B induced gliomas only gradually acquire a fully malignant phenotype. Glioma can be observed in an in vivo model of gliomagenesis induced by the embryonic transduction of neural progenitors with PDGF‐B retroviruses. Experimental setting and survival curve show two main tumor groups that can be distinguished based on symptom onset and tumor‐propagating potential. The figure was drawn on the bases of the data contained in the cited papers and represents an original artwork. GFP = green fluorescent protein.

PDGF‐INDUCED GLIOMA MALIGNANCY AND ONCOGENE ADDICTION

Ligand‐bound PDGFRs can activate a multitude of different signaling pathways, influencing cell survival, proliferation and motility, and any PDGF‐secreting cell will likely affect neighboring cells. In diverse tumor settings, PDGF signaling sustains tumor growth at least in part by promoting the establishment and maintenance of an intratumoral blood vessel network and by recruiting stromal cellular components to the tumor 30, 59, 76, 82. Furthermore, glioma cells typically show extraordinary infiltrative activity, a feature faithfully reproduced in mouse models of PDGF‐induced gliomagenesis, where PDGF‐overexpressing cells extensively migrate through the brain in a way reminiscent of the migratory activity of neural progenitors 6, 9, 18. Indeed, recent observations have suggested that glioma cells exploit the molecular machinery used by normal neural progenitors 11, 20.

Interestingly, there is remarkable correlation between the in vitro behavior of cultured high‐grade PDGF‐B‐induced gliomas and their in vivo infiltrative and tumorigenic potentials (18). The ability of cells derived from these gliomas to recapitulate tumor formation upon in vivo transplantation correlates with their capacity to rapidly infiltrate the brain and with the ability to override cell–cell contact inhibition in vitro. PDGF‐induced high‐grade gliomas indeed strictly depend on PDGF, and lose their ability to propagate as tumors in vivo after the loss of PDGF overexpression [(18), see below]. In vitro, these cells become contact inhibited, whereas PDGF‐overexpressing cells normally grow past confluence and form massive cell foci 9, 18. Furthermore, the loss of PDGF overexpression impairs the infiltrative potential of glioma cells, which become unable to infiltrate the brain after intracranial transplantation. However, the reexpression of PDGF‐B readily restores the tumorigenic potential of these cells, and this is associated with the reestablishment of both in vitro focus formation and in vivo infiltration. This indicates that the central role that PDGF signaling plays in regulating the tumorigenic potential of PDGF‐B‐transformed cells may be dependent on its impact on tumor proliferation and migration.

NEURAL PRECURSOR RECRUITMENT BY PDGF

An important and almost unexplored issue in brain tumor development and progression is represented by the interactions between tumor and normal untransformed brain cells. The mammalian brain contains a diverse set of broadly distributed glial progenitors with still incompletely understood roles in oligodendrocyte homeostasis and reactive astroglial generation in response to brain injury 14, 19, 21, 29, 80, 85. Glioma‐derived factors or tumor‐induced brain damage may influence these cells, with possible impacts on tumor growth. In vivo observations performed on PDGF‐induced glioma models have indeed provided evidence for the abundant presence of recruited oligodendroglial progenitors and reactive astrocytes within growing tumors. Assanah and colleagues showed that gliomas arising from PDGF‐B‐transduced adult white matter glial progenitors comprise an abundant recruited (untransduced) proliferating OPC population (Figure 4). Given their active proliferative activity, these cells likely contribute to tumor growth, suggesting that normal untransformed glial progenitors can be recruited into expanding tumor masses. In further support of this hypothesis, it was recently shown that gliomas induced by the transduction of PDGF‐B into embryonic neural progenitors also comprise abundant untransduced proliferating cells with the characteristics of OPCs [Figure 4; (6)]. Interestingly however, these recruited cells differed radically from PDGF‐B‐overexpressing transduced tumor cells, because they completely lacked the ability to reinitiate tumor growth in vivo following flurescence‐activated cell sorting (FACS)‐mediated purification and intracranial reinjection into the adult mouse brain. These findings show that PDGF‐induced gliomas can potently recruit proliferating resident glial progenitors to the tumor mass and that these cells likely contribute to tumor growth. Despite the lack, to the best of our knowledge, of similar observations in human gliomas, it is likely that nontransformed neural progenitors recruited by glioma‐secreted factors significantly contribute to tumor growth. Proliferating NG2/Olig2‐expressing glial progenitors are an especially abundant cell population within the human brain (79), making their recruitment a realistic possibility. It will therefore be interesting to see this issue explicitly addressed by analyzing primary human glioma samples or their xenotransplants, to evaluate the presence of recruited cells and to evaluate their possible contribution to tumor growth. A second class of abundantly infiltrating cells is represented by reactive astrocytes. These cells may actually represent multiple distinct populations, as lineage‐tracing experiments in the injured brain have yielded conflicting results, suggesting astrocytic 14, 21 or OPC 61, 85 origins for these cells. Regardless of the source, the presence of reactive astroglia within gliomas may have important implications, as they may not simply represent a response to glioma‐induced brain damage. Evidence for an active role of infiltrating astrocytes in modulating glioma growth has been obtained by studying PDGF‐B‐induced gliomas, in which increased Sonic hedgehog (Shh) pathway activity was shown to correlate with increased glioma malignancy (12). In this model, infiltrating astrocytes where shown to be a source of Shh, even if no in vivo evidence was provided for the ability of Shh inhibition to impair tumor growth. However, based on previous evidence for the role of Shh pathway activity in maintaining the tumor‐propagating cells from human gliomas (23), it is possible that astrocyte‐derived Shh may contribute to sustain glioma growth both in human and in model gliomas. The recent observation that astrocyte‐secreted Shh affects the proliferation of Olig2‐positive OPCs in response to brain injury (3) further supports the view that interaction between oligodendroglial tumor components and reactive astrocytes may affect tumor growth.

Figure 4.

Resident glial progenitors are recruited to the tumor mass and contribute to the growth of platelet‐derived growth factor (PDGF)‐induced gliomas without achieving an autonomous tumorigenic potential, as revealed by reinjection experiments following dissociation. The figure was drawn on the bases of the data contained in the cited papers and represents an original artwork. FAC‐sorting = fluorescence‐activated cell‐sorting; GFP = green fluorescent protein.

CONCLUSIONS AND FUTURE DIRECTIONS

PDGF signaling exerts multiple functions during mammalian development. Oligodendrocytes and their progenitors rely on PDGF signaling for proper expansion, survival, migration and differentiation. Alterations of PDGF signaling can negatively affect this cell population with severe impacts on CNS development and homeostasis. During postnatal life, deregulation of PDGF signaling leading to aberrant receptor activity results in various hyperproliferative diseases, among which are glial brain tumors. Human gliomas of different grades display altered PDGF signaling, indicating that PDGF may have broad relevance to human gliomagenesis, either by initiating or by supporting the progression of these tumors. In support of this, experiments using rodent models have established a clear role for PDGF signaling in initiating the formation of oligodendroglial tumors. PDGF overexpression in embryonic and neonatal NPCs, as well as into adult glial progenitors and NSCs, almost invariably leads to oligodendroglioma formation. This depends, at least in part, on the ability of PDGF to influence cell‐fate choices during the differentiation of neural stem/progenitor cells at all developmental stages. In turn, the developmental plasticity of the target cell population will likely concur with the specific stimulus to shape the “gliomagenic landscape,” that is, to determine the type of tumor(s) generated.

The clinical implications of this type of observations are not obvious, but a deepening of our understanding of the interplay between oncogenic stimuli, tumorigenesis and differentiation is likely to represent a relevant step forward in our ability to classify, study and hopefully target human gliomas in a more efficient way.

Despite the fact that PDGF has been conclusively associated with the formation of gliomas for many years now, thanks to the many studies performed using in vivo models of gliomagenesis, little is currently known regarding the identities of the transformed cell populations and, even more importantly, the mechanisms underlying PDGF‐mediated transformation. Recent works have, however, begun to address these issues, yielding important insights into both areas. It is now clear that diverse neural cell types can respond to aberrant PDGF signaling, resulting in the formation of gliomas. Remarkably, data from in vivo studies using rodent models strongly indicate that, when representing the initiating stimulus, PDGF mostly induces the formation of gliomas with oligodendroglial features. This reflects the fact that at least some PDGF‐responsive adult neural progenitors seem to be already quite lineage restricted, as is the case of adult OPCs. Another important cell population that seems to respond to oncogenic PDGF stimulation is represented by adult SVZ NSCs. In this case, PDGF‐driven gliomagenesis may be exploiting a physiological mechanism of cell‐fate decision which normally would regulate the differentiation of a subset of NSCs into either neurons or OPCs. To the exclusion of SVZ astrocyte‐like NSCs, astrocytes seem instead to be relatively refractory to PDGF‐induced transformation, consistent with their lack of PDGFR expression (21). This would imply that PDGF signaling deregulation by human astrocytic tumors may represent a secondary lesion, rather than an initiating one. These findings, albeit apparently conflicting with the common observation of PDGF/PDGFR overexpression in human astrocytomas and glioblastomas, may be reconciled considering that PDGF‐induced murine gliomas, although surely constituted by OPC‐like cells, histopathologically often resemble human glioblastomas. Indeed, only the use of reporter genes to reliably identify the transformed component within the experimental murine gliomas made it possible to classify them as oligodendrogliomas. Therefore, some human gliomas, classified as astrocytic tumors, may actually be of oligodendroglial nature in their transformed component while possessing a large fraction of recruited GFAP‐positive astrocytes.

Despite these advances, we are still in need for better animal models. Specifically, it is urgent to gain a deeper understanding of basic neurodevelopmental issues, among which are the origins, diversity and interrelationships of the diverse neural cell types found within the developing and mature vertebrate CNS. Clarifying these issues will allow the targeting of specific neural cell types with precise combinations of oncogenic insults, possibly in a spatiotemporally regulated manner. Furthermore, cell type‐specific isolation and gene expression profiling of normal and transformed cell types will add to our knowledge of the molecular mechanisms underlying transformation and progression, in vivo.

The molecular bases for the in vivo transforming ability of PDGF are also still poorly characterized, even if drawing parallels with its physiological roles in the regulation of proliferative and migratory activities of normal neural progenitors are already providing useful insights. On the other hand, understanding of the roles played by PDGF signaling in the maintenance and progression of gliomas will benefit from the accumulating evidence gained from the analysis of different physiological and tumoral systems. In gliomas and other tumors, PDGF ligands have well‐established roles as potent regulators of angiogenic processes. Together with recent observations suggesting that PDGF signaling may also promote the infiltrative activity of glioma cells, these observations provided a good rational basis for the design of efficient therapeutic strategies directed at inhibiting PDGF signaling in both tumor and nontumor cells, in vivo. Supporting the validity of this approach, we have recently documented a case of oncogene addiction, whereby PDGF‐induced high‐grade fully progressed gliomas remain strictly dependent on their initiating stimulus (ie, PDGF) for maintaining their tumor‐propagating potential. Clearly, the ability to efficiently deliver inhibitors to the tumor mass will remain a pressing priority in the field of neuro‐oncology.

Lastly, an issue which has recently emerged from in vivo studies on animal models is that of the diversity and roles of nontumor cells recruited to gliomas. Despite their abundance, very little is known about the possible functional roles of reactive astrocytes and recruited resident OPCs. Recent insights suggest that, akin to many nontransformed cell populations recruited to other tumor types, these reactive astrocytes and OPCs may actively or passively promote tumor growth, making them possible targets for future therapeutic intervention strategies.