Developmental origins of brain tumors

Abstract

Brain tumors are devastating due to the high fatality rate and the devastating impact on life qualities of patients. Recent advancement of comparative transcriptome profiling tools and mouse genetic models has greatly deepened our understanding of the developmental origins of these tumors, which could lead to effective therapeutic strategies. We review recent progresses in three types of brain tumors: ependymoma, medulloblastoma, and malignant glioma. The conceptual framework established by these studies converged on three important aspects. First, subtypes in each tumor group originate from distinct cell types. Second, each cell-of-origin is uniquely sensitive to some but not other genetic mutations. Lastly, mutant stem cells may not transform until they differentiate into more restricted progenitor cell type. Overall, these findings indicate the existence of intricate interactions between gene mutations and developmental context for the formation of brain tumors.

Introduction

Brain cancers, such as glioblastoma, are often lethal because of their invasive nature and resistance to conventional surgical procedures, chemo- and radiation-therapies. The urgent need for novel therapies has led to great emphasis on the study of the cell of origin for these tumors, grounded within modern concepts on neuroglial lineage development. From the developmental perspective, the cell of origin is defined as a progenitor that gives rise to one or a few progeny cell types in a specific tissue or organ. As a deviation from normal development, the cell of origin for cancer can be viewed as a specific cell type that can become transformed to drive the formation of the tumor mass [1]. Brain tumors, including ependymoma, medulloblastoma and malignant glioma, are speculated to have distinct cellular origins, which can be discriminated on the basis of anatomical location, expression of cellular markers, and morphological resemblance to normal brain cells [2]. With the refinement of genomic analysis tools and genetic animal model systems, a surge of recent studies on brain tumors have revealed a new level of complexity and the dynamic nature of their origins. Here we provide an overview of this topic, focused on a few representative studies to illustrate how technical advancement enhanced our understanding of the developmental origins of brain tumors.

Developmental Origins of Ependymoma from Region-restricted Precursors

Ependymoma is a group of brain tumors localized on the ventricle wall along the cerebrospinal axis [2,3]. Despite histopathological resemblances, ependymomas at different regions, such as supratentorial, posterior fossa, and the spinal cord, have disparate prognoses [4,5]. Such a discrepancy could be caused either by different cellular origins or by distinct genetic mutations that led to these tumors. Using a powerful cross-species genomic comparison approach, Johnson et al [6] compared the transcriptome of human ependymoma samples with that of mouse NSCs isolated from discrete locations along the cerebrospinal axis, at both embryonic and adult stages. Remarkably, their molecular profiling data demonstrated that human supratentorial ependymomas with amplified EphB2 and deleted INK4a/ARF specifically matched the INK4a/ARF null mouse embryonic NSCs from the cerebral region but not those from other locations or at the adult age. Using a mouse model, they further showed that ependymomas resembling human supratentorial ependymoma could only be induced by over-expressing Ephb2 in embryonic cerebral NSCs. This study elegantly demonstrated that the heterogeneity of ependymomas is most likely rooted in the differences of their developmental origins. Importantly, this study also demonstrated a unique susceptibility of a subtype of NSCs to particular genetic lesions, suggesting a synergistic effect between mutations and signaling contexts in the cell of origin. Similar findings of regional expression differences have also been described in pilocytic astrocytoma [7], suggesting that the notion of region-restricted cellular origins is also supported by findings with other types of brain tumors.

Developmental Origins of Medulloblastoma from Hindbrain Precursors

Medulloblastoma (MB) is a type of primitive neuroectodermal brain tumors that originates from the hindbrain/posterior fossa area and typically involves the cerebellum in the pediatric age range. Multiple MB subtypes have been described showing different genetic alterations, pathological features, and locations [8]. For example, the desmoplastic subtype tends to localize on the surface of the brain and to carry mutations in the Shh pathway, while the classic subtype tends to reside deep in the cerebellar region and to carry mutations in the Wnt pathway. The developmental origin of the desmoplastic subtype has been linked to granule neuron precursors (GNPs), based on multiple lines of evidences: the superficial tumor location that matches GNP location, augmented Shh signaling that is critical for GNP proliferation, and resemblance in transcriptome profiles to GNPs. This notion was further solidified by mouse models that recapitulated this MB subtype after introducing activating mutations in the Shh pathway with GNP-specific Atoh1-Cre (also known as Math1-Cre) [9,10]. However, the developmental origins of the classic subtype of MB remained unclear until recently. Gibson et al [11] suggests that MBs with aberrant activities of the Wnt pathway likely originate from Zic+ cells (the presumed precursor of mossy-fiber neurons) in the dorsal brain stem but not GNPs in the EGL. They demonstrated this convincingly with a mouse medulloblastoma model induced by overexpressing the activated version of Ctnnb1 (gene encoding β-catenin) in BLBP+ radial glia cells. The dorsal brainstem location, molecular and pathological features of induced tumors all matched well with human classic subtype but not the desmoplastic subtype of MB. These studies strongly suggest that these subtypes of MBs have distinct developmental origins, each of which is uniquely susceptible to corresponding mutations.

Developmental Origin of glioma: neural stem cells or restricted progenitors?

As the most common and fatal group of brain tumors, glioma poses significant challenges to both cancer biologists and clinicians. Traditionally, gliomas have been classified as astrocytoma and oligodendrocytoma based on morphological resemblance to normal glial cell types in the brain called astrocytes and oligodendrocytes, respectively. However, the existence of a subtype termed astro-oligodendroglioma manifesting mixed morphologies of both cells suggests that the boundary of astrocytoma and oligodendrocytoma may not be clear [2]. Taking advantage of the large genomic and transcriptomic dataset on human gliomas, a recent landmark paper [12] from the Cancer Genome Atlas (TCGA) consortium re-classified malignant gliomas into four distinct subtypes based on their different molecular signatures: neural, proneural, classic and mesenchymal. This new molecular classification should not only make a strong impact on molecule-based diagnoses of glioma subtypes, but also pave the way for the identification of the developmental origins for these subtypes using mouse cancer models.

During normal brain development, neural stem cells (NSCs) give rise to all the mature neurons, astrocytes, and oligodendrocytes via intermediate, more restricted progenitor cells [13–15]. Due to the transient nature of intermediate progenitors, most of them remain elusive despite many years of research, with oligodendrocytic progenitor cells (OPCs) as an exception [16]. Here we review recent papers that support the roles of NSCs and OPCs as the origin for gliomas.

Neural stem cells (NSCs) as the origin for gliomas

The NSC is characterized by its capacity to self-renew and its plasticity to differentiate into multiple cell types [15]. Observations of the cellular heterogeneity and renewability of glioma cells [17,18] led to the hypothesis that NSCs may be the origin for gliomas.

A few recent studies using genetic mouse models provided some supportive evidences for this hypothesis. For instance, the introduction of mutations in tumor suppressor genes p53 and NF1 [19], p53 and PTEN [20], or the overexpression of a mutant form of p53 [21], into NSCs with hGFAP- or Nestin-Cre efficiently induced mouse brain tumors that recapitulate human glioma pathology. Concordantly, introducing INK4a/Arf and EGFRVIII mutations into cultured neural spheres, which has been generally believed to enrich NSC population in vitro, afforded these cells capacities to initiate gliomas after intracranial grafting into immuno-compromised mice [22,23]. Lastly, Alcantara Llaguno et al [24] delivered Cre into specific anatomical structures in the brain via virus injection to inactivate tumor suppressor genes, and showed that subventricular zone (SVZ, where adult NSCs reside) but not other brain regions is particular susceptible to glioma formation. In summary, these collective evidences provided strong support for the hypothesis that NSC is the origin for gliomas.

However, the NSC hypothesis faces an intrinsic experimental quandary to unequivocally distinguish two scenarios: do NSCs serve as a direct origin, or do lineage-restricted progenitor cells that derive from NSCs function as the actual origin for gliomas. If the latter turns out to be true, what would be the identity of such progenitor cells?

Oligodendrocyte progenitor cells (OPCs) as the origin for gliomas

OPCs, reflected by their name, were originally believed to function solely as the precursor for mature oligodendrocytes. However, many studies have shown that OPCs persist into adulthood (4–5% of total adult brain cells) as the largest proliferating pool in both rodent and human brain parenchyma [16]. An initial clue that OPCs could be involved in glioma was the observation that Olig2, a bHLH transcription factor essential for OPC development, was not only universally expressed in human gliomas and also functionally required for gliomagenesis in a mouse model [22,25].

The possible role for OPCs as the glioma origin has been further supported by recent papers. Weiss and his colleagues [26,27] reported that human S100β promoter-directed expression of v-erbBan activated variant of EGFRcould induce glioma formation in a p53-null sensitized genetic background. S100β, though widely expressed in many mature cell types, does not turn on its expression in NSCs. Thus, this study provided strong evidence that cells other than NSCs can function as the glioma origin. In parallel, by injecting PDGF-BB expressing retrovirus into the white matter tract of mice, the Canoll group demonstrated that the well-known mitogen for developing OPCs could effectively initiate glioma formation in p53or PTEN or p53/PTEN double mutant mice [28]. Further analysis of tumors in both models revealed salient features of OPC, suggesting the critical role of OPC in gliomagenesis. More specifically, Lindberg et al [29] were able to generate gliomas by utilizing an RCAS/tv-a mouse model to over-express PDGF-BB in a CNPase-positive cell population that includes OPCs, although the penetrance was lower than what the Canoll group reported [28]. Lastly, the inactivation of p53 and NF1 by OPC-specific NG2-Cre transgene [30] led to the formation of malignant gliomas [31]. All these evidences clearly demonstrated that OPC could serve as the developmental origin for gliomas.

It is worth noting that cross-species transcriptome analysis revealed OPC-originated mouse gliomas matched well with human proneural subtype of GBM [12,28,31], suggesting OPCs as the developmental origin for this particular subtype of glioma. Importantly, in these animal studies the initial mutations used to drive transformation widely embodied those most frequently found in human GBMs, including p53PTENv-erbB, Nf1 and Ras [32], suggesting that OPCs could turn malignant with a wide range of mutations.

The Cell-of-origin and the cell-of-mutation are two distinct entities

Previously, the origin of cancer was defined by a “tried-and-true” methodology, whereby oncogenic mutations were introduced into populations of cells with particular developmental or regional characteristics to test their transforming potentials. However, the advent of new mouse genetic tools allowed us to dissect the tumor development process with unprecedented temporal and spatial resolution. It became clear that cells acquiring initial mutations (cell-of-mutation) might not directly transform; rather lineage-restricted progeny cells derived from the cell-of-mutation could inherit the mutations and subsequently transform into malignancy (see Figure 1B for an example). In the case of desmoplastic medulloblastoma model, although mutations that activate Shh signaling in both NSCs and GNPs could lead to tumor formation, the malignant growth is only seen in GNPs but barely in NSCs, suggesting that GNP is the cell-of-origin while both NSC and GNP could serve as the cell-of-mutation [9,10].

Figure 1. MADM, a mouse genetic mosaic model, can be used for identifying tumor cell of origin and evaluating novel targets against such cells for tumor prevention or treatment.

(A) Scheme that shows how MADM can concurrently mutate and label cells via a Cre/loxP-mediated inter-chromosomal mitotic recombination event. Green cells are homozygous null for the gene-of-interest that is distal to the MADM cassette on the same chromosome; red cells are WT cells. For details, see Zong et al 2005.

(B) Scheme that illustrates how to use MADM-based cancer model to identify cell of origin by analyzing the aberrant growth of all progeny lineages derived from mutant stem/progenitor cells in a defined organ. Although we use the brain as an example here, the same principle should be applicable to tumors in other organs. For details, see Liu et al 2011.

(C) The MADM system can be further modified by incorporating the Tet-ON system to investigate the effects of further genetic manipulation on the cell of origin. With the temporal control of the intervention, one could model the treatment effects of any gene on tumor initiation, progression, and malignancies.

The lesson learned from medulloblastoma studies now demands one to generate the same set of mutations in NSCs and each of progeny lineages to test their ability to form brain tumors before claiming the cell-of-origin. However, this is currently unfeasible for most brain tumors since some of the lineage-restricted progenitors have not been well defined and others have no specific genetic tools. An alternative solution to tease apart cell-of-origin and cell-of-mutation is to analyze individual cell lineages along the entire developmental process of tumorigenesis by using a mouse genetic mosaic model termed MADM (Mosaic Analysis with Double Markers) [31,33,34] (see also Figure 1A). Via Cre/loxP-mediated mitotic inter-chromosomal recombination, MADM generates a small number (0.1%–1% or much lower) of GFP-labeled homozygous mutant cells, thus mimicking the sporadic loss of heterozygosity (LOH) of TSGs in human cancers. MADM provides precise control for tumor initiation in vivo since the cell type specificity and the timing of mutation induction are completely determined by the activity of Cre transgenes. Since the efficiency of Cre/loxP-induced LOH is much higher than that of spontaneous LOH, the probability of tumor initiation from cells that are not labeled by GFP is very low. Moreover, MADM also labels sibling WT cells with red fluorescent protein such as DeRed or tdTomato, allowing the direct phenotypic comparison between mutant and WT sibling cells at a single-cell resolution in vivo, with the latter as an ideal internal control. Lastly, the permanent labeling of mutant cells with GFP allows one to trace the lineage of these cells, providing direct in vivo evidence of their aberrant capacity in proliferation, migration, differentiation, and even trans-differentiation into other cell types.

Using MADM, the developmental origin of gliomas induced by p53/NF1 mutations was examined by first introducing the mutations into NSCs and then analyzing the overexpansion in each progeny lineage. Surprisingly, detailed analysis of lineage-specific expansions prior to full malignancy revealed subtle or no over-proliferation in NSCs, neurons and astrocytes, but dramatic aberrant growth in OPCs. Along with histological, transcriptomic profiling, and genetic evidences, it became clear that OPCs function as the cell-of-origin in this model, while NSCs act as the cell-of-mutation [31]. This study not only helped reconcile one of the debates in the glioma field, but also provided experimental principles that could be used to identify the origin of other tumor types (illustrated in Figure 1B). To model treatment effects of a gene of interest at distinct tumor developmental stages, one can incorporate the Tet-ON/OFF binary system into MADM by fusing one chimeric MADM cassette with Dox-inducible transcription factor rtTA or tTA [35]. The modified MADM system would allow one to turn on/off the expression of TRE-driven effector gene at desired time, including but not limited to tumor suppressor gene, oncogene, or shRNAs against such genes (Figure 1C).

The distinction between cell-of-mutation and cell-of-origin carries great significance for both medical and basic research. For clinicians, the cell-of-origin is the transforming step of tumors thus would be an ideal therapeutic target. For basic researchers, the selectivity of transforming potentials in particular cell types clearly demonstrates signaling synergies between gene mutations and signaling context in the cell-of-origin. Further studies of such synergies will not only lead to novel ideas for clinical interventions, but also provide deep insights of the robustness of normal developmental programs.

Conclusions

Recent advancement in bioinformatics analysis and mouse genetic tools has greatly enhanced our understanding for the developmental origins of brain tumors. Large-scale genomic comparison between tumors and normal cell types, sometimes cross-species, has been highly successful in many studies. However, due to the highly plastic nature of end-stage cancer cells, their “birthmark” could be tampered during tumor progression. For example, the entire transcriptome profile of GBM could change dramatically by disturbing a single gene [36,37]. Therefore, it is imperative to complement bioinformatics analyses with mouse genetic studies to consolidate the claim of the developmental origin of brain tumors.

Finally, it is important to note that recent findings brought about even more new questions. Each type of brain tumor now can be divided into subgroups based on their unique molecular signatures in addition to their pathological distinctions. This implies that distinct cell types could serve as the cell of origin for each of these subtypes, a notion best supported by studies of medulloblastoma. In the case of gliomas, it is highly possible that multiple cell types, such as NSCs, glial progenitors, and even mature astrocytes, could serve as the developmental origins for non-proneural subtypes. Although these questions can only be resolved with further studies, the breathtaking progresses in recent years gave us the confidence that these goals should be within our reach.

Highlights.

Distinct developmental origins and susceptibility to mutations contribute to subtypes of brain tumors

Gene expression profiling, especially cross-species analysis, is a powerful tool to reveal the heterogeneity of tumor origin

Mouse cancer models with lineage tracing capacity provide concrete evidences of brain tumor origins

Cell-of-mutation and cell-of-origin are distinct concepts: former refers to cells with initial mutations, while the latter refers to cells that transform into tumors.