Highlights from the Literature

Evolution of non-GCIMP glioblastoma from a common subtype

Primary, non-GCIMP glioblastoma in adults have been classified into 4 gene expression-based subtypes: proneural (PN), neural (NL), classical (CL), and mesenchymal (MES). Motivated by the unclear boundaries between the subtypes that have been emphasized in recent analyses of tumor heterogeneity, the authors investigated the molecular relationship among the subtypes. Recurring to analysis of the temporal molecular evolution in GBM (Retracing Evolutionary Steps in Cancer; RESIC), gain of trisomy of chromosome 7 and loss of one copy of chromosome 10 emerged on top of the evolutionary chain of events, followed by deletion of CDKN2A, before TP53 mutations, or loss of NF1. In order to identify genes driving gain of CHR7 and loss of CHR10, which is a difficult task given the large number of genes affected by gene dosage on both chromosomes, an original computational strategy was developed. The procedure ranked genes affected in downstream pathways, of genes on CHR7 and CHR10 affected by gene dosage, and genes that were associated with patient survival. PDGF and PTEN emerged as the top-ranked candidate genes for driving gain of CHR7 and loss of CHR10, respectively. These observations in the human tumors were then followed up by the construction of genetically engineered mouse brain tumor models. Overexpression of PDGF was sufficient to induce brain tumors, which was enhanced by knocking down PTEN, while knock-down of PTEN was not. By constructing brain tumor models for the different combinations of gene alterations characteristic for human PN and MES GBM and comparing their respective expression profiles, the authors showed that PDGF alone or in combination with knock-down of TP53 was classified most closely to the PN subtype, whereas the combination of PDGF with knock-down of NF1 more closely resembled the MES subtype. From these results the authors concluded that PDGF is sufficient to induce PN GBM, while the additional loss of NF1 leads to the conversion to the MES subtype. Taking together these mouse data and the observation of temporal acquisition of genetic alterations in human GBM, the authors made the provocative conclusion that most non-GCIMP GBM evolve from PN-like precursor GBM.

There are certainly some limitations to the computational procedures and statistical interpretations identifying driver genes for gain of CHR7 and loss of CHR10. Further, the expression levels of PDGF inducing the murine brain tumors were likely substantially above the fold change of less than a factor 2 observed in the human GBM. Nevertheless, this is a refreshing new view with impact on treatment strategies. Certainly, other genes or combinations thereof on CHR7 or CHR10 merit a closer look as does integrating the epigenetic aspects associated with the tumor-driven copy number changes.

MYCN: Drugging the undruggable

MYC proteins are transcription factors that lay claim to two infamies. Firstly, they are potent oncogenes amplified or overexpressed that drive initiation and progression of several brain tumors including medulloblastoma and neuroblastoma. Secondly, designing direct inhibitors of MYC is challenging and has yielded less-than-promising results. Therefore, novel strategies for identifying MYC inhibitors would be beneficial both clinically and from a research perspective. MYCN has been reported to be stabilized by Aurora kinase A and is an attractive therapeutic target: it is a negative prognostic marker in neuroblastoma and plays a key role in the cell cycle and proliferation of several cancers. Gustafson et al rationalized that the kinase-independent function of Aurora kinase A may require a special conformation and that inhibitors of that conformation may destabilize MYCN. Thirty-two predicated compounds were screened to identify conformational changes in Aurora kinase A that would then lead to MYCN degradation. The authors identified that one compound (CD532) was able to inhibit phosphorylation of histone H3.3, a known target of Aurora kinase A, and diminish MYCN protein. In contrast, previously established Aurora kinase inhibitors had only a modest effect on MYCN protein levels. Inhibition of the MYCN-Aurora kinase A interaction by CD532 specifically led to proteosomal degradation of MYCN by the ubiquitination pathway.

Using elegant crystal structure studies the authors were able to demonstrate that the compound CD532 was able to stabilize an inactive form of Aurora kinase A and that degradation of MYCN by CD532 requires a conformation-specific inhibition of Aurora kinase A. When compared to well-established Aurora kinase A inhibitors, only CD532 was able to inhibit phosphorylation of Aurora kinase A, inhibit MYCN, and prevent cells from entering S-phase.

Having the dual ability to block both Aurora kinase A activity as well as MYCN stability, CD532 could serve as a promising MYC-directed treatment in cancers where MYCN is amplified or overexpressed. Thus, to determine if MYCN could serve as a biomarker to sensitivity, the authors screened 169 tumor-derived cell lines, 87 with accompanying mRNA expression data. Interestingly, CD532 had activity in most cell lines tested, and drug activity correlated with MYCN expression. Furthermore, tumor cell lines with MYCN amplification were the most sensitive. In both a neuroblastoma xenograft model and a Sonic Hedgehog (SHH) subgroup model of medulloblastoma, CD532 was able to diminish MYCN protein levels, reduce tumor volume, and increase survival in vivo.

To date, MYCN is the most characterized oncogenic event in childhood neuroblastoma and is a potential target in several other cancers including the SHH subgroup of medulloblastoma. Gustafson et al using a comprehensive and elegant series of experiments have identified an “amphosteric compound”, one that inhibits Aurora kinase A activity (orthosteric) and one that destabilizes its protein interactions, leading to MYCN degradation (allosteric). While further studies examining the clinical efficacy of CD532 are warranted, it is possible that the seemingly “invincible” MYCN oncoprotein has deficiency in its armor of “undruggability”.

Hominoid-specific enzyme GLUD2 and IDH-mutant glioma

Mutation of isocitrate dehydrogenase 1 (IDH1), or the homologous gene IDH2, is thought to be the initiating event for a majority of secondary glioblastoma (GBM) and lower-grade diffuse gliomas. The most common mutation in this gene is IDH1^R132H, which alters the activity of this enzyme, redirecting metabolic flux to generate 2-hydroxyglutarate (2-HG) at the expense of α-ketoglutarate (α-KG). The mechanism by which this mutation contributes to tumor growth, however, is not well understood.

Chen et al investigated the mechanism of tumor growth induction by IDH1^R132H using a model system of p53-deficient murine neural stem cells competent to generate proneural gliomas. Infection of these cells with RCAS vector encoding PDGF resulted in glioma within 2-4 weeks, whereas infection with IDH^R132H/RCAS failed to generate glioma up to 6 months later. Growth and clonogenicity was rescued upon transfection with IDH1^WT, despite the presence of 2-HG, suggesting that growth inhibition is not likely due to 2-HG toxicity; rather, it may be due to diversion of cytosolic α-KG from IDH1^WT. Furthermore, the concentration of α-KG in IDH1^R132H gliomas is generally maintained at normal levels, suggesting that certain metabolic enzymes may be overexpressed to compensate for the diversion of α-KG. Comparison of expression profiling data from 3 series of IDH1^R132H and IDH1^WT gliomas, along with RNAseq data from TCGA, revealed that the only genes consistently up-regulated in IDH1^R132H tumors were glutamate dehydrogenase 1 (GLUD1) and glutamate dehydrogenase 2 (GLUD2).

Knockdown of GLUD1/2 using shRNA resulted in reduced tumor volume and cell density relative to grafts of cells transduced with control vector. GLUD1/2 are mitochondrial enzymes that catalyze the conversion of glutamate to α-KG in the reductive glutaminolysis pathway critical for lipogenesis. Expression of these enzymes in glioma progenitor cells harboring IDH1^R132H or IDH1^WT showed that GLUD1 does not promote growth of IDH^R132H murine glioma progenitor cells, whereas GLUD2 rescued this effect only in IDH1^R132H but had no effect on parental cultures. Furthermore, fractional labeling analysis showed reduced conversion of glucose and glutamine to lipids in IDH^R132H cultures and this effect was rescued when cultures were transfected with GLUD2 but not GLUD1.

Based on these data, the authors suggest that the effect exerted by IDH1^R132H on lipogenesis is inhibited by GLUD2. Additionally, GLUD2 antagonized the growth inhibitory effects of IDH1^R132H in vivo, completely abrogating the inhibitory effects of IDH1^R132H on tumor growth and aggressiveness. Their results show that expression of GLUD2 in IDH1^R132H -bearing cells diverts glutamine-derived α-KG towards oxidative generation of citrate in the TCA cycle of mitochondria at the expense of flux to 2-HG in the cytosol. While these findings suggest that specific therapeutic approaches targeting glutamate metabolism or availability of α-KG may be applicable to IDH1^R132H gliomas, further in-depth investigation is needed to determine whether other α-KG-dependent processes play a role in growth-promoting effects of IDH1^R132H.

Mitigating toxicity of whole-brain radiotherapy with hippocampal avoidance

Fractionated radiation therapy is a modestly effective therapy against brain metastases; however, cognitive decline is a well-recognized potential side effect. Since the subgranular zone of the hippocampal gyrus is a key source of radiation-sensitive neural stem cells, this region may be an important mediator of radiation-induced cognitive damage. Thus arose the concept of using intensity-modulated radiation therapy to minimize the dose to the hippocampal neural stem cell niche during the administration of otherwise whole-brain radiation therapy (WBRT), termed hippocampal avoidance WBRT (HA-WBRT).

Radiation Therapy Oncology Group (RTOG) 0933 was a single-arm multi-institution phase II study of HA-WBRT for patients with brain metastases. The study included rigorous and appropriate inclusion and exclusion criteria as well as careful quality assurance of technical capacities and radiation plans to ensure that the radiation dose to the hippocampus did not exceed specified criteria to qualify as HA-WBRT. This approach reduced the dose to the neural stem cell compartment of the hippocampus by > 80%. A memory test, the Hopkins Verbal Learning Test-Revised Delayed Recall (HVLT-R DR), represented the primary endpoint, with a planned comparison to results from a previous RTOG WBRT brain metastasis study not incorporating hippocampal avoidance. HVLT-R DR, quality of life, and functional assessments were performed at baseline and at 2, 4, and 6 months from initiation of treatment, with patients serving as their own controls. The historical control group had manifested a 30% mean decline in HVLT-R DR from baseline to 4 months, so RTOG 0933 targeted a ≤15% mean relative decline after HA-WBRT.

The study met its primary endpoint: mean relative decline at 4 months with HA-WBRT was only 7%, significantly lower than the 30% mean decline in the historical control group. Moreover, only three (4.5%) of 67 patients with intracranial progression of brain metastases failed in the hippocampus, supporting the safety of this approach. Age ≥60 and higher hippocampal dose predicted greater drop in HVLT-R DR, suggesting possibilities for further refinement of this technique and patient selection. Median survival was 6.8 months, with only 7% of patients dying from brain metastases compared with 62% from systemic cancer, re-emphasizing that overall survival is a poor endpoint for brain metastasis clinical trials.

The authors concluded that HA-WBRT is associated with significant memory preservation compared to standard WBRT in historical controls, with acceptably small risk of failure of tumor control in the hippocampus. Limitations of the study include the single-arm design with a historical control group, suboptimal compliance with HVLT-R DR testing (71% at 4 months and 54% at 6 months), and inability to assess effects of HA-WBRT beyond 6 months given the sample size and high death rate. Confirmatory evidence for the benefits of HA-WBRT may come from two planned follow-up phase III studies - one comparing WBRT to HA-WBRT in prophylactic cranial irradiation for small cell lung cancer and another comparing HA-WBRT versus standard WBRT for established brain metastases.