The approach described here should significantly advance the efficiency and speed with which we discover and develop treatments for rare cancers and cancer subtypes. Studies of genetic mouse models have uncovered treatments for some brain tumors (Romer et al., 2004
; Rudin et al., 2009
); but a lack of ependymoma models has precluded similar efforts for this disease. We describe an integrated, multi-platform, in vitro
and in vivo
drug screen of an accurate mouse model of ependymoma (mEPEphb2
); identifying the IGF, EGF and centrosome cycle pathways as key candidate regulators of subtype-D tumors, as well as several treatment leads for the disease. We believe that these treatments are significantly enriched for drugs that will ultimately prove effective in patients because mEPEphb2
tumors reproduce the histology, ultrastructure, and transcriptome of subtype-D ependymoma with remarkable fidelity (Johnson et al., 2010
Ependymomas retain many of the biological properties of their parent NSCs (Johnson et al., 2010
). Therefore, many potent anti-ependymoma drugs may also be neurotoxic. This notion is supported by the results of our HTS, in which the majority of active agents, including molecular targeted therapies, were ‘equipotent’ against ependymoma and NSCs. Although similarities between ependymoma and NSCs may mean that cancer-selective agents are rare, our HTS system can identify these drugs. This represents a significant advance for clinical trial design. For example, ependymoma-selective agents (e.g., 5-FU) may be paired with effective but less selective drugs, to maximize efficacy while minimizing overlapping toxicity. Preclinical optimization of appropriate drug combinations for clinical trial in rare cancers is especially important, since the small patient populations place severe constraints on the number of clinical studies that can be conducted.
It is important to note that mEPEphb2
tumors model just one (subtype-D) of nine possible subtypes of human ependymoma (Johnson et al., 2010
). Therefore, some drugs identified in the current study may retain ‘subtype-D’ specificity in humans, rather than displaying broad anti-ependymoma activity; however this is a strength rather than a weakness of our approach. Although many human cancers have been carved up into distinct subtypes, these are not usually considered in the clinical trial design because we have lacked the means to predict their sensitivity to drugs. Consequently, clinical trials that fail to recruit adequate numbers of treatment-sensitive patients run the risk of rejecting useful drugs. Accurate mouse models of genomic subtypes of cancers present the exciting opportunity to model human cancer heterogeneity during pre-clinical drug development. To this end, we are using the same methodology employed to produce mEPEphb2
tumors to generate mouse models of the remaining eight ependymoma subtypes. Together with our integrated in vitro
and in vivo
screening approach, this battery of models should allow us to complete in months, numerous single and multi-drug pre-clinical trials that would take decades to conduct in the clinic. Drugs prioritized through this approach could then be passed to definitive clinical trials.
The active agents identified in our study not only provide therapeutic leads for the clinic but also insights into disease biology. Specifically, our combined HTS and kinome-wide binding assays of active kinase inhibitor scaffolds, have unmasked the IGF signaling and centrosome cycle pathways as regulators of subtype-D ependymoma. Further work will be required to define the aberrant function and clinical significance of these pathways, but studies of normal NSCs strongly suggest that these pathways are likely to be important in ependymoma. Lehtinen and colleagues showed recently that the CSF provides a proliferative niche for forebrain NSCs that includes IGF2 as a major constituent. Specifically, they demonstrated that IGF2 present in the CSF is bound by IGF1R expressed in the apical membranes of NSCs, stimulating cell proliferation (Lehtinen et al., 2011
). The vast majority of ependymomas arise directly adjacent to the ventricular system, most likely from NSCs of the SVZ (Johnson et al., 2010
; Kleihues et al., 2002
; Taylor et al., 2005
). Furthermore, we show here that IGF1R is upregulated in mEPEphb2
cells relative to their parental NSCs, and that IGF2-IGF1R signaling in these cells is blocked by drugs that inhibit their proliferation.
Considerable evidence also points to the centrosome cycle as a critical regulator of NSC polarity, self-renewal and proliferation (Lesage et al., 2010
). In this regard, Wang and colleagues showed that the centrioles are preferentially inherited during the asymmetric division of mouse NSCs, such that the renewed daughter inherits the maternal centriole while the new centrosome is inherited by the differentiating daughter neuroblast (Wang et al., 2009
). Many of the inhibitors the we identified as active against mEPEphb2
cells and NSCs, target kinases that regulate critical steps in this process including centrosome duplication e.g., CDK2 and TTK; spindle orientation and centrosome separation e.g., PLK1; and spindle maintenance e.g., GAK. Thus the centrosome cycle is intimately related to NSC fate and may well be disrupted and targetable in ependymoma. We are currently working to determine the roles of PIP5K1C and YSK4 kinases in NSCs and mEPEphb2
cells that were also identified in our kinome-wide binding assays but are less well understood. Notably, PIP5K1C that bound three of four kinase scaffolds in our screen, maps to a focal amplicon that we observed in human cerebral ependymomas (19p13.3), and has been reported to maintain stem cell proliferation and inhibit neuronal differentiation (Johnson et al., 2010
; Wang et al., 2007
; Yu et al., 2011
). Thus this kinase is of particular interest as a potential therapeutic target in ependymoma.
Our study also emphasizes the value of HTS for drug retooling. 5-FU is active against glioblastoma but this drug has never been tested formally in patients with ependymoma (Grunda et al., 2010
). Indeed, in the absence of the evidence provided in the current study, it is unlikely that an early generation chemotherapeutic like 5-FU would ever be selected for clinical trial against ependymoma. Therefore, HTS approaches provide us with the opportunity to rationally re-purpose cancer drugs. Well characterized and accurate models should improve the efficiency and the confidence with which we do this. When these studies are coupled with comprehensive pharmacokinetic studies, these models can also help determine the most appropriate mode of administration of re-purposed drugs. Concurrent measures of brain tumor and plasma 5-FU levels in our model suggest that bolus administration might be a more effective way to deliver the drug to ependymoma. Importantly, we are now using these preclinical response and pharmacokinetic models to test combinations of drugs identified as most active in our single agent HTS and in vivo
studies. This includes combinations of centrosome cycle or IGF inhibitors, with conventional cytotoxic agents such as 5-FU. We envisage that continued pharmacokinetic and response assessments of these combination therapies will allow us to design the optimal schedule for translation to the clinic.
In summary, we describe an integrated, multi-platform in vitro and in vivo HTS of ependymoma that identifies a series of biological insights and potential therapies for clinical development. Our mouse model allows both testing of drug efficacy in a specific genomic subtype of the disease, and concurrent assessment of toxicity to the parental normal NSC. The development of additional mouse models of ependymoma subtypes should allow further comprehensive preclinical assessment of therapies for clinical trials tailored to all disease subtypes.