|Home | About | Journals | Submit | Contact Us | Français|
Malignant astrocytomas are infiltrative and incurable brain tumors. Despite profound therapeutic implications, the identity of the cell(s) of origin has not been rigorously determined. We previously reported mouse models based on conditional inactivation of human astrocytoma-relevant tumor suppressors p53, Nf1, and Pten, wherein through somatic loss of heterozygosity, mutant mice develop tumors with 100% penetrance. In the present study, we show that tumor suppressor inactivation in neural stem/progenitor cells is both necessary and sufficient to induce astrocytoma formation. We demonstrate in vivo that transformed cells and their progeny undergo infiltration and multi-lineage differentiation during tumorigenesis. Tumor suppressor heterozygous neural stem/progenitor cultures from pre-symptomatic mice show aberrant growth advantage and altered differentiation, thus identifying a pre-tumorigenic cell population.
Identification of the original cell that gives rise to a tumor and whether it is a limited cell type has crucial implications for understanding cancer development. This knowledge is also requisite for rigorous investigation of tumor initiation mechanisms. Using fully penetrant mouse models, we identify neural stem/progenitor cells as cancer-initiating cells and derive insight into the behavior of these tumors. We also report malignant astrocytoma mouse models wherein tumor suppressor inactivation at embryonic, early postnatal, or adult ages induces tumor formation, and demonstrates the capacity of tumor cells to differentiate within the tumor. Our studies on pre-symptomatic mutant progenitor cultures indicate that the disease could be disseminating and acquire growth advantage long before the onset of clinical manifestations.
Gliomas are the most common primary malignancies in the central nervous system (CNS). Astrocytomas, which account for the majority of these tumors, exhibit histologic resemblance to astroglial cells. The most malignant form, glioblastoma multiforme (GBM), is one of the most lethal forms of cancer, with a median survival of about one year (Maher et al., 2001; Zhu and Parada, 2002). These highly infiltrative tumors are resistant to conventional radiation and chemotherapy resulting in dismal survival outcomes that, in contrast to some forms of cancer, have improved only marginally in the past several decades (Stupp et al., 2005).
A variety of mutations have been described in human astrocytoma, and these frequently disrupt cell cycle and apoptosis regulation (INK4A, CDK4, RB, TP53), and growth factor receptor signaling (EGF, PDGF, PTEN) (Furnari et al., 2007). These genetic lesions have been exploited in the mouse to generate animal models that phenocopy the human malignancy and, thus, allow for in vivo investigation of tumor development and their use in translational studies. A number of these mouse models involve introduction of oncogenic mutations in the germline or specific cell subpopulations in the brain. These include overexpression of active forms of Ras, Akt, epidermal growth factor receptor, and platelet-derived growth factor, as well as transforming antigens such as v-src and polyoma middle T-antigen, and often in combination with mutations in tumor suppressors such as Ink4A or Arf (Fomchenko and Holland, 2006). The first endogenous genetic tumor suppressor mouse model was based on heterozygous mice carrying cis germline mutations in Nf1, a ras GTPase activating protein (rasGAP) and an effector of receptor tyrosine kinase signaling, and Trp53 (p53). Depending on genetic background, these mice develop brain tumors with varying penetrance (Reilly et al., 2000).
Further refinements were made through cre/lox technology, wherein mouse strains with germline or somatic heterozygous mutations at the p53, Nf1, and Pten tumor suppressors develop high grade astrocytomas with 100% penetrance (Kwon et al., 2008; Zhu et al., 2005a). TP53 and PTEN mutations are among the most frequent mutations reported for astrocytomas (Furnari et al., 2007; Maher et al., 2001). Patients with germline mutations in NF1, called neurofibromatosis type I, have increased susceptibility to astrocytomas (Gutmann et al., 2002), and recent detailed investigations of this 350-kb gene by the Cancer Genome Atlas Research Network indicate that along with PTEN and TP53, somatic NF1 mutations are also prevalent in sporadic GBMs. In fact, these three genes are among the top five most mutated genes in human GBMs (McLendon et al., 2008).
Mouse models harboring a heterozygous germline or conditional somatic p53 mutation combined with conditional somatic Nf1 heterozygosity develop low to high grade (secondary) astrocytomas (Zhu et al., 2005a). Tumor formation is further accelerated into high grade astrocytomas similar to primary GBM by additional loss of Pten (Kwon et al., 2008). These fully penetrant endogenous tumor suppressor-based mouse models develop tumors that are indistinguishable from the human malignancy based on known histologic and molecular criteria that define human astrocytomas.
The SVZ is an extensive germinal layer that concentrates neural and glial progenitors on the walls of the lateral ventricles of adult mammals (Alvarez-Buylla and Lim, 2004). In rodents, SVZ neural stem cells correspond to type B cells. These primary progenitors give rise to transient amplifying type C cells that undergo limited mitoses before differentiating into neuroblasts that migrate through the rostral migratory stream (RMS) and into the olfactory bulb (OB) (Doetsch et al., 1999). Neurogenesis also occurs in the subgranular zone (SGZ) of the dentate gyrus, which produces local neurons that incorporate into the granular cell layer (Gage, 2000; Zhao et al., 2008). In humans, the SVZ and SGZ have both been shown to harbor neural stem cells (Eriksson et al., 1998; Sanai et al., 2004). Recent studies have suggested the existence of additional though minor stem/progenitor niches elsewhere in the brain (Gould, 2007).
Historically, astrocytomas have been thought to arise from differentiated glia that undergo a process of dedifferentiation (Sanai et al., 2005; Sauvageot et al., 2007). However, whether mature differentiated astrocytes in their normal parenchymal environment are capable of initiating tumor formation in vivo has not been rigorously tested. The recent identification of adult neural stem cells, immature cells that divide throughout the lifetime of the individual, represents attractive targets for acquisition of tumor-causing mutations.
In the clinical setting, it is impossible to determine either the location of tumor origin or the nature of the cell type(s) capable of generating these tumors. Similarly, because of its infiltrative nature from the early onset, it is impossible to determine the in vivo fate of tumor cells. Many theories have been disseminated surrounding these issues, including an astrocytic origin, and further, that GBM contains highly heterogeneous tumor-derived cells. However, despite fundamental implications for development of therapies, direct examination or proof of these ideas remains to be established.
Physiologically relevant genetic mouse models of high grade astrocytomas provide one approach to address these critical questions. Examination of events preceding unassailable tumor development in our mouse models suggested that the tumors arise in the SVZ and possibly within stem/progenitor cells (Zhu et al., 2005a; Kwon et al., 2008). To directly examine whether neural stem/progenitors are the source of astrocytoma, we undertook a two-pronged approach. Using both genetic and stereotactic injection methods, we demonstrate that in the context of prevalent mutations and aberrant signaling pathways found in idiopathic glioma---TP53, NF1, and PTEN---embryonic, early postnatal and, importantly, adult neural stem/progenitor cells can give rise to malignant astrocytomas in vivo whereas more mature cell types cannot.
Our previous astrocytoma mouse models relied on heterozygous deletion of the tumor suppressors Nf1, p53, and/or Pten, using a human GFAP promoter-controlled cre transgenic line (hGFAP-cre) to drive recombination (Zhuo et al., 2001), followed by spontaneous loss of heterozygosity (LOH) at these loci (Kwon et al., 2008; Zhu et al., 2005a). The GFAP promoter used in these conditional knockout mice is expressed in both stem cells and white matter astrocytes. Thus, whether the tumors arose from the neural stem/progenitors cells could not be determined.
To directly examine whether tumor suppressor depletion in neural stem/progenitors was sufficient to induce astrocytoma formation, we targeted neural stem/progenitor cells in vivo using a transgene that expresses a cre recombinase-modified estrogen receptor ligand binding domain fusion protein (cre-ERT2) under the control of the nestin promoter/enhancer (Chen et al., in press). Nestin is an intermediate filament protein that is widely used to mark neural stem/progenitor cells but not differentiated astrocytes, and the second intron enhancer allows for neural precursor-specific expression in the CNS (Lendahl et al., 1990; Zimmerman et al., 1994). Tamoxifen administration induces nuclear transfer of the cre-ERT2 protein in nestin-expressing cells where it can mediate loxP-dependent recombination (Feil et al., 1996). In a Rosa26 β-galactosidase (R26-lacZ) reporter background (Soriano, 1999), neural stem/progenitor cells and all their progeny are indelibly marked by lacZ, which can be identified by X-gal staining (Figure 1A&B). Induction at embryonic day E13.5 produced a broad pattern of expression similar to the hGFAP-cre transgene that was used for the Nf1-p53-Pten conditional astrocytoma mouse models (Kwon et al., 2008; Zhu et al., 2005a), reflecting expression in embryonic telencephalic progenitors (Figure 1A). Induction at early postnatal stages exhibited marked and progressively reduced recombination in parenchyma (Chen et al., in press) and by adult ages, lacZ staining in the forebrain and midbrain almost exclusively labeled the neural stem cells in the SVZ and its progeny along the RMS and OB, as well progenitors in the SGZ and their progeny in the dentate gyrus (Figure 1B).
We bred the inducible nestin-creERT2 mice to incorporate the tumor suppressor floxed (flox) alleles (either Nf1flox/+;p53flox/flox;Ptenflox/+ or Nf1flox/flox;p53flox/flox) and injected pregnant females with tamoxifen at E13.5 or adult mice at 4 weeks of age. All tamoxifen-treated (TMX) mutant mice developed high grade astrocytomas, while vehicle-treated (Veh) mice did not (Figure 1C). E13.5-tamoxifen-treated mutant mice developed tumors at a similar rate as the previously reported hGFAP-cre; Nf1flox/+;p53−/flox;Ptenflox/+ mouse strain (Kwon et al., 2008), with a median survival of around 16 weeks. Mutant mice injected with tamoxifen at 4 weeks of age developed tumors with a median survival of around 46 weeks. Hematoxylin and eosin (H&E) staining of these tumors showed the classical features of diffusely infiltrating astrocytomas, including nuclear atypia and prominent mitoses, as well as necrosis (Figure 1D). Both E13.5- and adult-treated mutant mice developed tumors diagnosed as Grade III or Grade IV (glioblastoma multiforme) astrocytomas based on the World Health Organization classification system (Supplementary Figure 1A, Supplementary Table 1). These tumors had large numbers of Ki67+ cells, indicating robust proliferation, and were immunoreactive for Gfap, nestin and Olig2 (Figure 1E), acknowledged markers in human astrocytic tumors (Furnari et al., 2007). Consistent with activation of the Ras and Akt signaling pathways by loss of Nf1 and Pten, respectively, some tumor regions showed robust pErk and pAkt expression (Supplementary Figure 1B). We further confirmed LOH of the tumor suppressor alleles in tumors by immunostaining and genotyping (Supplementary Figure 2). Thus, cre-mediated somatic mutation of Nf1, p53, and Pten restricted to the neural/stem progenitor compartment is sufficient to replicate the high grade astrocytoma phenotype previously observed using combinations of germline and somatic mutations and a less specific hGFAP-cre driver (Kwon et al., 2008; Zhu et al., 2005a). Furthermore, loss of Nf1, p53 and/or Pten was present in all tumors and is therefore apparently required for high grade tumor induction (Kwon et al., 2008).
The presence of the R26-lacZ reporter in the context of the floxed tumor suppressors allows for lineage tracing of cells as they undergo tumorigenic transformation. While normal neural stem cells and their progeny are principally restricted to the SVZ-RMS-OB and SGZ-GL (Figure 1B), tumors arising from inducible mutant mice were found in adjacent brain regions, including the cortex and striatum, as shown by X-gal staining (Figure 2A). The cre transgene is expressed in cerebellum but only one tumor was found in cerebellum and it resembled astrocytoma and not medulloblastoma, which is the idiotypic tumor of this brain region. The β-galactosidase positive tumor cells co-stained with the astrocytoma markers Gfap, nestin, and platelet-derived growth factor α (PDGFRα; Figure 2A). Thus, mutant stem cells or their progeny migrate away from their normal niches and invade the parenchyma during tumor development.
Astrocytomas are heterogeneous tumors, with varying cellular morphologies and presence of immature and mature markers for all neural lineages. Upon examination of the tumor bulk, we found a variety of β-galactosidase-positive cells co-expressing markers of subsets of differentiated cells, such the neuronal marker Tuj1, and the glial markers S100β and adenomatous polyposis coli or APC, as shown in a thalamic tumor in Figure 2B. These immunoreactive tumor cells morphologically resembled mature neurons, astrocytes, and myelin-ensheathing oligodendrocytes. In contrast to normal CNS cells which show abundant Pten expression (Kwon et al., 2008), these marker-positive cells were Pten-negative (Supplementary Figure 2C), confirming that these “differentiated” cells are indeed cancer cells. These data provide formal evidence that tumor cells have the stem/progenitor capacity to exit the cell cycle and at least partially differentiate in situ. This may account for the heterogeneity of tumor cell types that is classically associated with high grade astrocytomas.
As an independent approach for targeting cre recombinase to the SVZ neurogenic niche, we adopted a stereotactic injection method. Numerous studies have used stereotactic targeting to study the function, lineage, and identity of neural stem cells by delivery of dyes, growth factors, or viral particles directly into the SVZ (Doetsch et al., 1999; Merkle et al., 2007; Merkle et al., 2004; Yoon et al., 1996). Injection of cre recombinase-expressing adenovirus into the SVZ of R26-lacZ reporter mice alone results in labeled neural stem cells in the SVZ and their progeny as they travel through the RMS and into the OB (Figure 3A, right panels). In contrast, cre adenovirus injection into non-neurogenic regions such as the cortex or striatum causes only localized labeling in the area of the injection and no labeling in the RMS or OB (Figure 3A, left panels). In both cases, restricted staining can also be found along the needle tract.
We injected cre adenovirus into the SVZ of three strains of tumor suppressor floxed mice (Nf1flox/flox;p53flox/flox, Nf1flox/flox;p53flox/−, or Nf1flox/+;p53flox/flox;Ptenflox/+) either at postnatal day 1–2 or adult ages (4–8 weeks). We found that viral cre-mediated tumor suppressor inactivation in the SVZ at both early postnatal and adult ages induced astrocytoma formation (Figure 3, Table 1). Injected mutant mice developed brain tumors classified as Grade III or IV astrocytomas starting six months post-injection (Supplementary Figure 3A, Supplementary Table 2). These tumors exhibited the histopathologic hallmarks of high grade astrocytomas, including diffuse infiltration, nuclear atypia, mitoses, microvascular proliferation, and necrosis (Figure 3B), and were indistinguishable from those of our previous studies (Kwon et al., 2008; Zhu et al., 2005a). The tumors were immunoreactive for Ki67, Gfap, and nestin (Figure 3C), as well as Olig2, pErk, and pAkt (Supplementary Figure 3B). We further confirmed that the tumors underwent cre recombination by β-galactosidase immunostaining (Figure 3C). An internal control and validation in these studies was a posteriori verification that intended injections into the SVZ resulted in lacZ lineage tracing of the SVZ-RMS-OB axis. We observed that only successful SVZ-RMS-OB targeted injections, as evidenced by X-gal staining in the tumor bulk, SVZ, and the olfactory bulb (Figure 4A), as well as β-galactosidase immunohistochemistry (data not shown), generated astrocytomas. We also confirmed loss of tumor suppressor alleles in tumors by PCR genotyping (Figure 4A).
Similar to the nestin-inducible mutant mice, tumors were found throughout most of the brain parenchyma (Figure 4A). Early postnatal-injected mutant mice seem to develop more extensive tumors compared to adult-injected mutant mice, which develop more restricted, albeit still invasive, high grade astrocytomas. Clear evidence of migration is shown by the presence of tumors away from the SVZ, such as in the cortex, hippocampus and thalamus. In both cases, we found intense X-gal staining of the tumor regions, as well as in the SVZ-RMS-OB (Figure 4A).
As described with the tamoxifen-inducible tumors, β-galactosidase-positive tumor cells had morphological features of differentiated CNS cells and expressed mature markers, including the astrocytic marker Gfap, oligodendrocytic marker myelin basic protein (MBP) and neuronal marker NeuN (Figure 4C). As an independent method of confirming that these marker-positive cells are indeed tumor cells, we stained these tumors for Pten, which is frequently suppressed in high grade astrocytomas. We found that a number of these “differentiated” cells are Pten-negative (Figure 4C). We also found a rare population of β-galactosidase-positive tumor cells near the cortex that express calbindin (Figure 4C), which is normally expressed by a subset of OB neurons produced by the SVZ neural stem/progenitor cells (Merkle et al., 2007). These data demonstrate that cancer-initiating cells or their progeny have the capacity to migrate throughout the parenchyma, seed tumors, and give rise to differentiated cell types during tumor development.
The above results demonstrate that progenitor cells in the SVZ have the capacity to give rise to astrocytomas. However, these studies do not rule out the possibility that additional parenchymal cells might also harbor this capacity or that tumors might arise from the few cells that are infected with the cre adenovirus along the injection track. Previous studies, using other experimental systems, which involved in vitro manipulation or oncogenic transformation, suggest that mature astrocytes can also give rise to gliomas (Bachoo et al., 2002; Dai et al., 2001; Uhrbom et al., 2002). In order to target tumor suppressor inactivation to cells outside the SVZ including astrocytes in vivo, we stereotactically delivered cre adenovirus into the cortex or striatum of 20 four to eight week-old tumor suppressor floxed adult mice injected in parallel with the previously described SVZ injections. In contrast to the successful SVZ injections, where 100% of the mice develop tumors, none of the animals injected in the cortex or striatum had evidence of tumor formation (Figure 3B, Table 1) despite clear evidence of successful cre adenovirus infection as demonstrated by X-gal staining and PCR genotyping (Figure 4B) or β-galactosidase immunohistochemistry (Supplementary Figure 4). H&E staining showed disorganization of the cortical or striatal architecture in the injection site (Figure 3B), while immunostaining showed the presence of Gfap-positive but Ki67-negative cells (Supplementary Figure 4), which is indicative of fibrosis and astrogliosis (Zhu et al., 2005b). Cells near the injection site likewise stained positive for nestin and vimentin (Supplementary Figure 4), consistent with reactive astrocytosis (Correa-Cerro and Mandell, 2007; Sofroniew, 2005), phenotypes that are similar to GFP adenovirus-injected control brains (data not shown).
Non-SVZ regions were also targeted at early postnatal ages and the majority of these mice do not develop gliomas (Table 1). However, we did find one out of twelve cases where non-SVZ-targeted neonatal floxed mice developed tumors 7 months post-injection. This is consistent with prior reports of oncogenic transformation of early postnatal brain cells. The rarity of tumor induction in non-SVZ-injected mutant mice is consistent with the rare targeting of neural progenitor cells in the cortex or striatum that still exist at early postnatal ages (Seaberg et al., 2005), or alternatively, the rare targeting of radial glia that project into the parenchyma and are the progenitors of SVZ adult stem cells (Merkle et al., 2007). Overall, these data indicate that while tumor suppressor targeting of SVZ neural stem/progenitor cells readily induces high grade astrocytoma formation, more differentiated cell types are less susceptible to malignant transformation by the tumor suppressor mutations assessed in this study.
The above studies demonstrate that neural stem/progenitor cells can spontaneously give rise to malignant astrocytomas in our tumor suppressor mouse models. This allows us to follow these cells as they evolve from normal to transformed cells, as well as to investigate the molecular events involved in tumor initiation. Hence, we went back to our original tumor mouse models, which we have previously shown to exhibit hyperplastic and migration defects in the SVZ at early stages in vivo (Kwon et al., 2008; Zhu et al., 2005a). We examined the properties of mutant SVZ cells as neurospheres, which is the classical assay for studying neural stem/progenitor behavior in vitro (Reynolds and Weiss, 1992). Interestingly, we found that cultured SVZ neurospheres from tumor-prone mutant mice (Mut3:hGFAP-cre;Nf1flox/+;p53−/+ or Mut4: hGFAP-cre;Nf1flox/+;p53−/+;Ptenflox/+) at young ages prior to any histologic evidence of hyperplasia, exhibited abnormal growth properties (Figure 5). Compared with wild type controls, heterozygous mutant SVZ neurospheres displayed increased proliferation as shown by increased neurosphere diameter (Figure 5A) and bromodeoxyuridine (BrdU) incorporation (Figure 5B), as well as decreased apoptosis, as shown by annexin V staining (Figure 5C). These cells likewise showed abnormal stem cell properties, including increased self-renewal capacity (Figure 5D), and altered differentiation potential with decreased astrocytic differentiation compared to normal cells (Figure 5E, Supplementary Figure 5). To determine whether SVZ cells underwent LOH in the young mice months prior to tumor appearance, we performed PCR genotyping of the tumor suppressor alleles in the abnormal neurosphere cultures and found the presence of wt alleles for all tumor suppressors, indicating retention of heterozygosity (Figure 5F). These data indicate that in the absence of gross morphologic abnormalities, heterozygous tumor suppressor-deficient stem/progenitor cells already display abnormal growth and differentiation properties - processes that may serve as a prelude to tumor formation.
In this report, we describe the spatial and temporal restriction of in vivo gene targeting to the neurogenic niches of the brain through the use of a tamoxifen-inducible nestin-creERT2 transgene or by stereotactic viral-mediated cre recombinase delivery to the SVZ. The results were striking in that all adult mice subjected to SVZ targeting developed astrocytomas, thus establishing that mutation of these astrocytoma-relevant tumor suppressors in the neurogenic compartment in vivo is sufficient to induce tumor formation. Importantly, all viral injections into the SVZ were validated postmortem by lacZ staining to verify effective labeling of the RMS and OB, normal destination sites for progeny derived from the SVZ. In contrast, viral targeting of adult parenchyma where the vast majority of cells are differentiated did not yield tumors but, instead, local astrogliosis and localized lacZ staining, despite demonstration of recombination in glial cells. The inducible tumor model, on the other hand, showed that specifically targeting nestin-expressing neural stem and progenitor cells induces tumor formation. These data strongly support the idea that mutations in the stem/progenitor compartment account for the majority of these tumors and identify neural stem/progenitor cells as cancer-initiating cells in our fully penetrant astrocytoma mouse models. Furthermore, astrocytoma induction occurs efficiently in embryonic, early postnatal and adult mice dependent on stem/progenitor cell targeting of the tumor suppressors.
For all of the described tumor studies, the endpoint was selected to ensure adequate incubation of all manipulated cells, whether tumor suppressor-bearing or controls. We ended the studies when specific cohorts exhibited morbidity that, in all cases, was verified to be the consequence of advanced astrocytoma. Therefore, we did not assess the natural history of tumor development or the genetic signature of the tumors since the present models are predicated on our extensively previously characterized tumor suppressor-based mouse models (Kwon et al., 2008; Zhu et al., 2005a). Moreover, we cannot rigorously distinguish between quiescent stem cells or actively dividing progenitor cells as the cells where LOH originates in our mouse models. Since the tumors arise months after targeting and in the normal course of events, transient amplifying cells reach the OB and differentiate within two weeks (Petreanu and Alvarez-Buylla, 2002), we favor the hypothesis that the tumorigenic state preexists in the stem cell population but becomes phenotypically manifest once the cells enter the transient amplifying state. Experimental investigation of these distinctions will require more refined cell type-specific promoters to drive cre-mediated recombination. It also remains to be determined whether the differential susceptibility of neural stem cells and astrocytes to transformation is dependent on local microenvironment.
The concept that some or many forms of cancer may be comprised of a subset of tumor-propagating cells and another subset of cells that cannot propagate the tumor has recently received increasing attention (Reya et al., 2001). Dirks and colleagues initially showed that human GBM xenografts into immunodeficient mice have such an identifiable subset of cancer-propagating cells or “cancer stem cells” (Singh et al., 2004). Cancer stem cells are thus technically defined in terms of their in vivo capacity for tumor initiation in serial transplantations, and rely on retrospective isolation of these self-renewing cells (Dalerba et al., 2007). It is logical then to suggest that these cancer stem cells have characteristics in common with stem cells, but whether normal stem cells are the cells of origin of these tumors remained to be experimentally established. Our data conclusively show that normal neural stem/progenitor cells are cancer-initiating cells, and can readily give rise to high grade astrocytomas.
Astrocytomas are notorious for their infiltrative capacity, a property that clinically confounds complete surgical resection. We show here that in contrast to normal adult neural stem cells that are strictly confined to the SVZ or SGZ, tumors arising from these cells or their progeny are not restricted to these niches and actually migrate away from their normal locations, thus accounting for the presence of tumors elsewhere in the forebrain, including the cortex, striatum, hippocampus, and thalamus. This can also explain the presence of tumors in regions where the hGFAP-cre transgene is not expressed in the conditional mutant mouse models (Kwon et al., 2008; Zhu et al., 2005a). Another distinct feature of human astrocytomas is the heterogeneity of cell types within these tumors. Because of their infiltrative nature, one interpretation is that “diverse” non-tumor cells are present and surrounded by tumor cells. This may be the case, to some degree, as genotyping of primary tumor tissue yields a faint wild type or non-recombined band (Figure 4A and data not shown), and the majority of cells expressing mature, differentiated markers are β-galactosidase-negative (Figure 2B, ,4C),4C), suggesting that normal cells were trapped within the tumor bulk. An additional alternative is that the tumor itself has a heterogeneous component of tumor-derived cells. This alternative has been indirectly supported by several lines of evidence including in vitro differentiation and xenografting into immunodeficient mice. However, direct demonstration that in situ, the original tumor is heterogeneous has been lacking. The power of mouse genetics permitted neural stem/progenitor compartment-specific tumor suppressor inactivation in the context of a cre-dependent lacZ reporter transgene. Thus, through morphologic assessment of lacZ-positive and Pten-negative tumor cells in conjunction with labeling with lineage-specific markers, we find that a subset of tumor-derived cells have properties of astrocytes, oligodendrocytes and neurons. We even found a rare subset of lacZ-positive tumors cells that express calbindin, which is normally expressed by a subset of OB neurons produced by the SVZ neural stem/progenitors, suggesting that the differentiation capacity of these cancer-initiating stem/progenitor cells is retained during tumor development.
The degree of differentiation is variable among individual tumor cells and between tumors from different individuals. However, since these “mature” cells may represent a less aggressive population of tumor cells, this observation suggests that differentiation therapy may provide a plausible approach to arresting tumor growth while avoiding killing bystander normal cells. In sum, the clinically relevant migration and differentiation capacity of astrocytomas fits well with a neural stem/progenitor cell origin uncovered here.
Additional astrocytoma mouse models have used combinations of oncogenic overexpression and/or tumor suppressor inactivation to induce tumor formation. Several reports have shown that nestin promoter-driven oncogenesis at early postnatal ages can give rise to astrocytomas, whereas GFAP promoter-driven oncogenesis has reduced penetrance depending on the initiating mutations (Holland et al., 2000; Uhrbom et al., 2002). The tumor cell of origin in these studies was inferred but not directly examined. Ex vivo expansion of cultured neural stem cells or neonatal astrocytes followed by transplantation into immunodeficient mice also gave rise to astrocytomas (Bachoo et al., 2002; Dai et al., 2001). Thus, to date, cells that were targeted for transformation were derived from either embryonic or early postnatal brain cells. Moreover, in vitro manipulation to establish tumorigenicity is likewise problematic because it is well established that cell culture significantly alters the normal biological behavior of cells. Studies using oncogenic mutations may also provide supraphysiologic levels of activated oncogenes. In light of the present studies, we suggest that these mouse models may be targeting embryonic precursors present in neonatal brains with the distinction that our models equally target adult stem/progenitor cells and induce high grade astrocytomas with tumor suppressor inactivation. On the other hand, the Cancer Genome Atlas project has described EGFR receptor mutations as mutually exclusive from NF1 mutations in gliomas. Thus, it is possible that other mutations in neural stem/progenitors can likewise give rise to gliomas, or that gliomas with differing genetic signatures may originate from different cell subtypes. The finding that the SVZ contains a diverse set of neural stem cells that can give rise to specific progenitor subtypes (Merkle et al., 2007) provides some clues. Whether these heterogeneous stem cell populations are susceptible to the same mutations or give rise to different tumor subtypes also remains to be examined. The experimental approaches described herein will be useful in determining the cell of origin of other models using GBM signature mutations or pathways.
The mechanisms involved in tumor initiation remain poorly understood, and identification of the cell of origin allows us to follow normal SVZ cells as they undergo transformation and full-blown tumorigenesis. We were surprised by the finding that in culture, pre-tumorigenic neurosphere-forming cells from the SVZ already show growth advantage and apoptotic resistance, while retaining heterozygosity. It is also surprising that at these early stages, the differentiation capacity of these mutant cells is altered, suggesting that the differentiation state of these immature cells is tightly coupled to self-renewal processes. This underscores the importance of the Nf1, p53, and Pten tumor suppressor pathways in regulating normal neural stem cell proliferation and self-renewal. It also has implications for therapeutic approaches should the physiological relevance of our models continue to be validated. These data imply that there remains a persistent and primed source of pre-tumorigenic cells in the neurogenic niche that will require further understanding and therapeutic targeting.
All mouse experiments were approved and performed according to the guidelines of the Institutional Animal Care and Use Committee of the University of Texas Southwestern Medical Center at Dallas. Mice with conditional Nf1, p53 and/or Pten alleles with the R26-lacZ reporter were maintained on a mixed 129Svj/C57Bl6/B6CBA background (Kwon et al., 2008). Mice that harbor the Nf1 and p53 floxed alleles in cis were generated by crossing the Nf1flox/+ strain to the p53flox/+ strain to generate Nf1flox/+;p53+/flox trans mice that were then crossed to wild type mice. Generation and maintenance of Mut3 and Mut4 mice, as well as genotyping for the flox, wild type and recombinant alleles have been previously described (Kwon et al., 2008; Zhu et al., 2005a).
Tumor suppressor floxed mice containing the nestin-creERT2 transgene were injected intraperitoneally with vehicle (9:1, sunflower oil: ethanol mixture) (Sigma) or tamoxifen (Sigma) at a working concentration of 6.7 mg/mL. Pregnant moms were injected with 1 mg tamoxifen at E13.5, whereas 4-week-old mice were injected with 83.8 mg/kg tamoxifen twice a day for 5 consecutive days (Chen et al., in press).
Tumor suppressor floxed mice at 4–8 weeks of age or postnatal day 1–2 were injected with cre- or GFP expressing adenovirus, as described previously in the literature (Doetsch et al., 1999; Merkle et al., 2004) with some modifications. Two hundred nL of adenovirus (Ad-Cre, 2.0 × 1012 pfu/mL, University of Iowa Vector Core; Ad-GFP, 1.0 × 1011 pfu/mL) was injected using a World Precision nanoinjector apparatus, according to the following coordinates - SVZ (0, 1.4, 1.6; 0.5, 1.1, 1.7; and 1, 1, 2.3), cortex (0, 3.5, 1.5), and striatum (0, 1.4, 2.6) mm anterior, lateral, and dorsal to the bregma. For early postnatal injections, postnatal day 1 or 2 pups were injected with 40 nL of Ad-Cre or Ad-GFP, as previously described (Merkle et al., 2004) with some modifications. The following coordinates for a range of neonate weights were used: 1.4–1.5g (1.5, 2.6, 1.4); 1.5–1.7g (1.6, 2.7, 1.4); 1.7–1.9g (1.7, 2.9, 1.5); >1.9g (1.7, 2.9, 1.7) mm anterior, lateral and dorsal to the bregma. All virus and tamoxifen-injected mice were followed for development of neurologic abnormalities and harvested for histologic analysis.
Mice were perfused and fixed with 4% paraformaldehyde. Five-μm sections were cut and every fifth slide was stained with H&E. Brain sections were independently examined by S.A.L. and J.C., as well as D.K.B., a certified neuropathologist, and tumor diagnosis was determined based on the World Health Organization criteria (Kleihues et al., 2002). Brains used for X-gal staining were postfixed in 2% PFA overnight. Half brains or 50-μm vibratome sections were stained in X-gal solution, and sections were counterstained by nuclear fast red, as previously described (Luikart et al., 2005). In some cases, half brains that were stained with X-gal were subsequently processed and used for immunohistochemistry. For PCR genotyping, DNA extraction and PCR was performed using tumor and non-tumor tissues, as previously described (Kwon et al., 2008).
Paraffin sections were deparaffinized, rehydrated, and subjected to citrate-based antigen retrieval. Primary antibodies against the following were used as follows: Gfap (DAKO, 1:1000; BD Biosciences, 1:200), Ki67 (Novocastra, 1:1000), nestin (BD Biosciences, 1:100), Olig2 (Chemicon, 1:1000), β-gal (ICN, 1:1000), NeuN (Chemicon, 1:500), MBP (Sternberger, 1:200), calbindin (Swant, 1:1000), vimentin (Zymed, 1:200), S100β (Sigma, 1:200), pERK (Cell Signaling, 1:400), pAKT (Cell Signaling, 1:100), Pten (Cell Signaling, 1:100), and PDGFRα (Spring, 1:50). We used both immunofluorescence staining using Cy2, Cy3, or Cy5 (Jackson Labs, 1:400) and biotin-streptavidin-Alexa Fluor-conjugated secondary antibodies (Molecular Probes, 1:1000), as well as horseradish peroxidase-based Vectastain ABC Kit (Vector Lab). Sections were examined using optical, fluorescence and confocal microscopy (Olympus and Zeiss).
Neurosphere cultures were established and maintained as described previously (Kwon et al., 2008). To measure neurosphere size, we seeded dissociated cells into 24-well plates and took all neurosphere images after 7 days of incubation. We measured their diameter using MetaMorph software (Universal Imaging Corporation). Proliferation and apoptosis in neurosphere cultures were quantified by flow cytometry using BrdU and Annexin V analysis kits (BD Biosciences). Self-renewal assay was done by plating dissociated cells with methylcellulose-containing media (final 0.8%, Sigma), as described previously (Gritti et al., 1999). For differentiation, we seeded dissociated cells in 8-chamber slides coated with Matrigel (BD Biosciences, 1:20) and cultured in Neurocult with differentiation medium (StemCell Technologies). Quantification was done in at least n=3 samples from each genotype, and student’s t-test was used for statistical analysis.
The authors thank Linda McClellan, Steven McKinnon, Alicia Deshaw, Shawna Kennedy, Jerry Chandler and Patsy Leake for technical assistance, and Parada Lab members, especially Renee McKay, for helpful suggestions and discussion. Adenovirus was provided by the University of Iowa Virus Core. This work is supported in part by the Children’s Tumor Foundation Young Investigator Award to S.A.L., Basic Research Fellowships from American Brain Tumor Association (in memory of Daniel J. Martinelli and Geoffrey J. Cunningham) to C.-H.K., and by NIH (5P50NS052606), DAMD (W81XWH-05-1-0265) and ACS (RP0408401) Grants to L.F.P. L.F.P. is an American Cancer Society Research Professor.
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.