Highlights from the Literature

Epigenomic profiling reveals distinct subclasses of AT/RT

Atypical teratoid/rhabdoid tumors (AT/RTs) are among the most malignant brain tumors of childhood. The invariable occurrence of either SMARCB1 or, much less frequently, SMARCA4 mutations in AT/RT underscores the central pathogenic role for SWI/SNF chromatin complex dysfunction. Indeed, molecular profiling studies to date have failed to document any other recurrent genomic alterations that may be driving AT/RT. However, clinically relevant heterogeneity with regard to patient survival and tumor localization implies the existence of endogenous disease subgroups.

In a recent Cancer Cell article, Johann et al probed epigenomic landscapes in a large (N = 192) sample set of AT/RTs. Global DNA methylation arrays revealed three tumor subclasses whose delineation was also supported by mRNA profiling. Their proposed nomenclature reflected these transcriptional distinctions, with AT/RT-TYR tumors exhibiting overexpression of melanosomal markers like MITF, TYR, and DCT, AT/RT-SHH tumors featuring upregulated MYCN and GLI2 reminiscent of SHH-driven medulloblastomas, and AT/RT-MYC tumors remarkable for marked overexpression of MYC. Notably, the three AT/RT subclasses were differentially correlated with supra- and infratentorial localization as well as patient age. Moreover, the manner of SMARCB1 inactivation differed by subclass, with AT/RT-TYR tumors commonly featuring broad chromosome 22 loss, whereas AT/RT-SHH and AT/RT-MYC tumors more commonly harbored focal aberrations involving the SMARCB1 gene itself.

Consistent with earlier reports, whole-genome sequencing in a subset of the larger AT/RT cohort failed to reveal additional genomic alterations. However, whole-genome bisulfite sequencing demonstrated clear distinctions in global DNA methylation patterns between the aforementioned disease subgroups. Specifically, AT/RT-TYR and AT/RT-SHH tumors exhibited genome-wide hypermethylation particularly in promoter regions with downstream effects on gene expression, whereas AT/RT-MYC tumors instead featured large “partially methylated domains” (PMDs) in association with normally inactive chromatin. Finally, the authors used H3K27acetylation and BRD4 ChIP-seq to identify enhancer regions genome wide and characterize significant differences between AT/RT subclasses. Interestingly, activated enhancers specific to each subclass demonstrated likely regulatory relationships with defining transcription factors, particularly in the case of AT/RT-TYR tumors where both OTX2 and MITF were implicated as master regulators controlling subclass-specific gene expression. Intriguingly, the authors also found that a known MITF inhibitor preferentially reduced viability in an AT/RT cell line with high MITF expression.

In summary, these findings establish clinically distinct AT/RT subclasses whose unique biological characteristics point to viable strategies for therapeutic development.

HDAC and PI3K antagonists cooperate to inhibit growth of MYC-driven medulloblastoma

Medulloblastoma (MB) is the most common malignant brain tumor in children. Recent MB molecular classifications subdivide MB into four subgroups (WNT, Sonic Hedgehog, Group 3, and Group 4) that have distinct molecular profiles and prognoses. Group 3 is partly characterized by MYC amplification or overexpression, has a worse prognosis than the Sonic Hedgehog and WNT subgroups, and has no known actionable molecular target.

A recent study by Pei et al screened and tested for new experimental therapies against MYC-driven Group 3 MB using representative mouse models. The authors used a previously developed mouse model that overexpresses myc and a dominant-negative form of p53 and that forms tumors resembling human MYC-driven MB. They first isolated cells from the mouse tumors and screened for viability using compounds from seven small-molecule libraries. Of the 3642 compounds tested, 142 inhibited cell viability by >2 fold. The 142 compounds were then tested against human MYC-driven MB cells and against normal cells (cerebellar granule neurons and astrocytes) to ensure inhibition of human tumor cells and lack of toxicity towards normal cells. Twenty-three compounds were found to kill mouse and human MB cells and spare normal cells. Among these 23 compounds were 4 histone deacetylase inhibitors (HDACI). Based on this finding, the study focused on HDACI and tested the effects of several of these compounds on human and mouse MB cells as well as normal cells. The compound with the greatest inhibitory effects in MYC-driven and other MB cells was LBH-589 (panobinostat), a pan-HDAC inhibitor that has been recently approved for the treatment of some cancers but that had not been tested in MB. To determine how LBH-589 inhibited MYC-driven tumor cell growth, they used gene expression arrays and comparisons with their previous data and uncovered transcription factor FOXO1-regulated genes as upregulated by treatment with LBH-589. They experimentally confirmed that LBH-589 upregulated the expression of FOXO1 in MB cells and showed that FOXO1 inhibits the growth of MYC-driven MB and that its upregulation mediates the anti-tumor effects of LBH-589. Since FOXO1 activity is known to be negatively regulated by PI3K/AKT, they hypothesized that LBH-589 and PI3K inhibitors might cooperate in inhibiting MB growth. To verify this hypothesis, they first experimentally showed that LBH-589 and the PI3K inhibitor BKM-120 cooperate to promote the activation of FOXO1. They then showed that combining LBH-589 and BKM-120 synergistically inhibited MB cell growth as well as the growth of mouse MB and human MB Group 3 xenografts. They concluded that combining HDACIs and PI3KIs might represent an effective therapy for MYC-driven MB.

This study used a rational screening approach and relevant mouse models to identify a new combination therapy for MB with a focus on MYC-driven tumors. Whether this therapy translates to human patients remains to be determined by clinical trials.

mTOR inhibition for patients with relapsed primary CNS lymphoma

A recent article by Korfel et al¹ reports on the trial NCT00942747, which assessed 75 mg flat-dosed weekly mechanistic target of rapamycin (mTOR) inhibition with i.v. temsirolimus in 37 patients with relapsed/refractory intermediate- to high-risk primary CNS B cell lymphoma (with one patient suffering from a T cell lymphoma). Patients in this uncontrolled Phase I/IIa trial had relatively unfavorable prognostic factors with a median age of 70 years and a median time from prior therapy of only 3.9 months. Pretreatment was standard from the standpoint that all patients had received a high-dose methotrexate-based poly-chemotherapy and only two patients had received whole-brain radiotherapy as part of the initial treatment. As temsirolimus is registered at 25 mg per week for renal cell cancer but 75 mg per week for mantle cell lymphoma, the study group decided on initiating the study with 3 patients on the 25 mg dose and escalated to 75 mg after observing limited toxicity.

The grade 3 or 4 toxicity seems relevant for low platelets (8/37 patients, 22%) and hyperglycemia (11/37 patients, 30%), although parallel steroid intake will have contributed to the latter. Importantly, all toxicities were well in line with the prior observed toxicities despite the rather elderly population.

Without engaging in the debate of whether partial responses are relevant for primary CNS lymphoma given the biology and growth kinetics of the disease, the fact that the study revealed some objective imaging responses (5/37 patients, 13.5% complete remission) is remarkable and clearly evidence of efficacy. The duration of progression-free survival of 2.1 months (95% CI of 1.1–3.0 months) indicates pre-existing and rapid development of resistance in all patients.

Given the burden of the treatment and the relevant toxicity, further escalation of the dose would not seem to yield a solution to the problem. As discussed in the article, temsirolimus might serve as a combination partner. However, given the glioblastoma experience in combination with temozolomide² or other targeted agents, toxicity will also be a potential limitation in those scientifically driven combination trials.

Taking into account the targeted nature of temsirolimus, the most important next step would be a comprehensive molecular characterization of patients treated with mTOR inhibition, as the data speak for a specific molecular determinant for resistance. As of now, the trial fails to offer conclusive or comparative data to any of the other options at relapse of a primary CNS lymphoma but clearly positions temsirolimus as one of the options that warrant further (molecular-driven) investigation.

Medulloblastoma genotype dictates blood-brain barrier phenotype

The molecular subclassification of medulloblastomas into 4 genotypes revealed that these tumors are different entities and opened the possibility for different therapeutic strategies for this childhood tumor where toxicity from treatment is a particular concern. In a series of elegant experiments, Phoenix et al evaluated why the WNT-medulloblastoma (WNT-MB) subtype was more susceptible to treatment and had a better prognosis, even with CNS dissemination, than the other medulloblastoma subtypes. The authors found that the tumor vasculature was different in WNT-MB compared with the other medulloblastoma subtypes, and this was one possible explanation for the increased response to treatment.

Using animal models, they demonstrated that the WNT-MB subtype had increased mean vessel density with increased vascular porosity. These changes were driven by the tumor genotype and were not a result of the different origin of WNT-MB, which arises from the dorsal brainstem. The increased permeability of WNT-MB allowed better penetration of vincristine, a chemotherapy not thought to pass through an intact blood-brain barrier well. Vincristine did not penetrate into the more treatment resistant SHH-medulloblastoma (SHH-MB) subtype that does not have the same abnormal vasculature.

Furthermore, the authors showed that a paracrine secretion loop by WNT-MB tumor cells suppressed WNT signaling in the adjacent endothelial cells, and by modulating WNT signaling they could restore the integrity of the vasculature or induce increased permeability in the SHH-MB subtype. These findings have clear implications for treatment selection in children with WNT-MB and raise the possibility that modulating tumor vasculature in other types of medulloblastoma (or even other tumors) may increase the response rate to treatment. With increased permeability, strategies focused on chemotherapy may be sufficient to treat these tumors and spare patients the potential side effects of aggressive resection (ex. cerebellar mutism) or radiation (ex. decreased cognition). The authors have launched an imaging clinical trial, based on the differential penetration of contrast agents through porous and nonporous vasculature, to identify WNT-MB tumors prior to surgery in order to tailor therapy – taking that important next step to confirm animal-based findings in humans. In summary, this study elegantly emphasizes the importance of understanding how tumor biology drives response to therapy and can lead to novel therapeutic strategies.

A new target for the mesenchymal glioblastoma subtype

It has long been known that some of the worst cancer behaviors—invasion, metastasis, and treatment resistance—can be driven by a mesenchymal phenotype. While in some cancers this aggressive mesenchymal phenotype arises from epithelial-to-mesenchymal transition (EMT), in glioblastoma (GBM) a mesenchymal subtype was identified in recent years from gene expression studies from initial surgical samples.^1,2 There is a dire need to identify therapies that target the mesenchymal cancer phenotype, both in GBM and in cancer as a whole, but to date there have not been clear successes on this front.

A recent report by Kim et al identifies a kinase target in mesenchymal GBM that may provide therapeutic leverage.³

Kim and colleagues began with gene expression data from GBM stem cell-like lines of the mesenchymal and proneural subtypes, as well as astrocytes and neural progenitors, to identify genes up-regulated in mesenchymal GBM lines. Mixed lineage kinase 4 (MLK4) emerged as one of the genes with expression most highly up-regulated in the mesenchymal subtype. This prompted testing of potential functional implications, and the authors found that MLK4 knockdown preferentially damaged mesenchymal GBM cell viability in vitro and diminished mesenchymal GBM growth in vivo. MLK4 knockdown in mesenchymal GBM lines also reduced mesenchymal characteristics such as invasiveness, high glycolytic activity, and expression of mesenchymal markers. Notably, a clear mechanism was identified by which MLK4 influences the mesenchymal phenotype; MLK4 binds and phosphorylates IKKα, which regulates the known mesenchymal driver NF-κB.⁴ MLK4 knockdown was able to sensitize a proneural GBM line to radiation therapy in vivo, which was consistent with the prior observation that radiation could convert GBM from the proneural to the mesenchymal phenotype.⁴ High MLK4 protein expression was noted in patient GBM samples with high expression of the mesenchymal marker CD44, correlating with shorter survival in these samples, but anti-correlated with high expression of the proneural marker OLIG2.³

There may well be therapeutic implications to these findings. While inhibitors of MLK4 have not yet been identified, such a kinase is likely to be more easily druggable than a prime mesenchymal target such as NF-κB. There have been concerns regarding the utility of targeting particular subtypes, given that individual GBMs harbor cells of each subtype and also have the potential to switch from one subtype to another. However, it may well be possible to develop combination regimens that simultaneously attack each of the subtypes, making it imperative to develop therapies that target each. The identification of MLK4 as a target for the challenging mesenchymal subtype may provide a valuable piece of this puzzle. Independent of a GBM subtype-targeting strategy, anti-mesenchymal targets may also provide dividends in sensitizing GBM to therapies such as radiation.