Highlights from the Literature

Does Immunotherapy Increase the Risk of Radiation Necrosis in Patients With Brain Metastases?

Martin and colleagues address this question by comparing patients treated at the time of diagnosis of brain metastases (BM) with immunotherapy or not and observing an incidence rate of radiation necrosis post-initial stereotactic radiotherapy to characterize potential adverse symptomatic radiation potentiation.¹ The no immunotherapy group included 365 patients and the immunotherapy group 115 patients. Immunotherapy treatment entailed overwhelmingly the use of immune checkpoint inhibitors and primarily anti-PD-1 monoclonal antibodies. Patient groups were matched with respect to age, sex, race, primary cancer (mostly non-small cell lung cancer and melanoma with a small fraction of renal cell cancer), performance status, maximum diameter of the largest BM, neurologic symptoms, presence and anatomic location of extracranial disease, type of treatment radiation (single fraction SRS vs. multi-fraction SRT), number of prior systemic therapies, and salvage radiotherapy following initial radiotherapy. The exclusion of breast cancer, a common cancer that metastasizes to brain, reflects current usage of immune checkpoint inhibitors wherein for breast cancer there is no approved indication. Symptomatic radiation necrosis was defined as an enlarging lesion after stereotactic radiation causing neurologic symptomatology that displayed one of the following characteristics: pathology specimen showing only necrosis (if surgical resection performed) or changes consistent with necrosis on dual-phase PET–CT or serial MRI. Symptomatic necrosis occurred in 20% vs. 7% of patients who did vs. did not receive immunotherapy, respectively. Receipt of immunotherapy was associated with symptomatic radiation necrosis after adjustment for tumor histology. The authors acknowledge study limitations including the retrospective nature, the comparatively small sample size, and the lack of a standard method of radiographic interpretation as to what constitutes radiation necrosis.

An additional challenge not mentioned in the study is the confounding effect of pseudoprogression on immunotherapy.² In a recent meta-analysis of 8 multicenter clinical trials in melanoma with anti-PD-1 antibody-based therapy, 51% received anti-PD-1 antibody beyond RECIST-defined progressive disease.³ Fourteen percent of the treatment beyond progression group demonstrated a RECIST-defined response, which suggests a portion of progression on immunotherapy is a treatment-related pattern and not true progression.

Overall, the current study by Martin and colleagues is hypothesis generating as to the potentiation of radiation injury in patients with BM treated with stereotactic radiotherapy and antibody-based immunotherapy.

References

Final Analysis of EF-14: Tumor-Treating Fields (Optune) in Newly Diagnosed Glioblastoma

In 2015, Roger Stupp and colleagues published preliminary results of a planned interim analysis of EF-14, an open-label trial randomizing patients with glioblastoma to standard post-radiation temozolomide with or without tumor-treating fields (TTFields).¹ Those results, based upon the first 315 of a planned 700 patients, demonstrated benefit from the addition of TTFields to standard therapy and led to approval of the device from the U.S. Food and Drug Administration.

The newly published manuscript analyzes the outcome of all 695 patients (accrual was terminated prematurely based upon positive interim analysis results, and patients on the control arm were allowed to cross over to receive the device at that point).² The study randomized patients without post-radiochemotherapy radiographic worsening to the device versus no device in a 2:1 ratio, with stratification for extent of surgery and MGMT methylation status. The treatment arms were well balanced on all prognostic variables. TTFields were initiated between 4 and 7 weeks from completion of radiation, with 6–12 cycles of maintenance temozolomide. TTFields could be continued in the experimental arm for 24 months or until the second tumor progression; patients in the control arm were allowed to receive TTFields at tumor progression. Progression-free survival (PFS) was the primary endpoint, and overall survival (OS) a secondary endpoint. MRIs underwent blinded central radiology review for determination of tumor progression.

Results from the entire study population closely matched those from the interim analysis. PFS significantly favored the TTFields arm (median PFS 6.7 vs. 4.0 months, HR 0.63, P < 0.001). Similarly, OS was substantially improved in the device arm as well (median 20.9 vs. 16.0 months, HR 0.63, P < 0.001). The benefits of TTFields were independent of MGMT status, extent of tumor resection, age, and performance status. There were no new safety signals related to TTFields; skin irritation, which was rarely severe, was the only toxicity more common in the experimental arm.

These results confirm the findings of the published interim analysis indicating a statistically and clinically significant benefit from the addition of TTFields to standard therapy. While the lack of a sham device control arm is arguably a shortcoming, and we have but a single positive trial of this novel modality, the blinded central confirmation of PFS improvement, the extent of OS improvement, and the similarity of the outcome in the control arm to the control arm of RTOG 0525³ (which randomized patients at a similar time point) all support a true biological antitumor effect of TTFields beyond placebo.

References

GSC-Derived Pericytes in the Periscope: Torpedoing the Blood-Tumor Barrier With Ibrutinib

Transdifferentiation of cancer cells into other supporting cell types in the tumor environment was established long ago, but in recent years there has been an explosion of interest in this phenomenon in glioblastoma (GBM). Initial reports described transdifferentiation of GBM stem-like cells (GSCs) into endothelial-like cells, similar to phenomena observed in other cancers.¹ However, another report challenged this, describing that GSC transdifferentiation into the pericytes that ring endothelium-lined blood vessels was a more common and significant phenomenon.² A new report from this latter group and others revisits and expands upon this concept, identifying GSC-derived pericytes as an important barrier to chemotherapy penetration that can be targeted with ibrutinib inhibition of bone marrow and X-linked (BMX) non-receptor tyrosine kinase.³

The authors first showed that higher pericyte coverage correlated with worse prognosis in GBM, but only in patients who had received chemotherapy. Selective elimination of GSC-derived pericytes disrupted the blood-tumor barrier but not the blood-brain barrier—an important distinction—and allowed greater penetration of intravenously administered agents into the GBMs. BMX was identified as a key driver in GSC-derived pericytes, and the authors drew on the literature to identify ibrutinib as an FDA-approved agent with potent inhibition of BMX. Ibrutinib indeed proved to be able to target the pericytes in GBM but not normal brain, which in turn allowed better penetration and anti-GBM efficacy of chemotherapy agents that normally lack adequate blood-tumor barrier penetration.

These results raise some important, unanswered questions. Does higher pericyte coverage in GBM correlate with any established GBM prognostic markers, common mutations, or gene expression-based subtypes of GBM? Is BMX a particularly important target for GSC-derived pericytes, or does ibrutinib target GSCs or GBM cells in general as well as through other mechanisms? While BMX and ibrutinib have both been reported on previously in GBM, this work provides new understanding with welcome clinical implications: increasing chemotherapy penetration into GBM is a long-sought goal in neuro-oncology, and the combination of ibrutinib with chemotherapy could be moved rapidly to clinical trials.

References

Mouse Modeling for H3.3 K27M-Mutant Diffuse Midline Glioma

Recent large-scale genomic profiling has emphatically confirmed the molecular distinctions between pediatric high-grade gliomas (HGGs) and their adult counterparts. In particular, large subsets of pediatric HGG harbor gain-of-function mutations in histone H3.3 variants rarely seen in adults. Among these, the K27M mutation in either H3F3A or HIST1H3B represents the defining molecular alteration of diffuse midline glioma, a highly aggressive HGG variant formerly known as diffuse intrinsic pontine glioma. H3.3 K27M almost invariably co-occurs with TP53 mutation and frequently associates with PDGFRA amplification and/or ATRX inactivation. Its pathogenic mechanism of action appears to involve a dominant negative effect that dramatically reduces genome-wide levels of the H3K27me3 histone mark in chromatin.

The development of targeted therapies to combat K27M glioma has been hindered by a relative paucity of in vivo disease models for preclinical testing. In a recent Cancer Cell paper, Pathania et al. reported the first highly penetrant, de novo mouse model dependent on H3.3 K27M for its pathogenesis.¹ After failing in their attempts to generate such a disease model in post-natal mice, the authors found success with an in utero electroporation strategy using the piggyBac transposon-based system. In doing so, they were able to generate infiltrative HGG in both infratentorial and supratentorial distributions with only H3.3 K27M combined with Tp53 loss, distinguishing their model from earlier studies requiring the addition of activated PDGF signaling. Interestingly, controls overexpressing GFP, wild-type H3.3, and G34R-mutant H3.3, the latter found in supratentorial, hemispheric pediatric HGG, were non-tumorigenic, reflecting the precise cellular and molecular contexts required for transformation in this model. Further inclusion of ATRX knockdown appeared to reduce tumor invasiveness, while the addition of wild-type PDGFRA expression increased aggressiveness.

All H3.3 K27M-expressing models reported in this study serially engrafted in recipient mice and were pliable in drug testing paradigms. Moreover, they exhibited much reduced H3K27me3 levels and gene expression signatures similar to those of H3.3 K27M human tumors. Finally, their patterns of H3K27me3, as determined by ChIP-seq, also resembled those of H3.3 K27M-mutant diffuse midline glioma in terms of differentially impacted gene clusters.

These findings are highly significant in that they establish a working in vivo disease model for diffuse midline glioma that is pathogenically dependent on the relevant disease-defining molecular alteration. In this way, they lay the groundwork for more effective preclinical testing and drug development.

Reference

Mutant IDH1-Dependent Chromatin State Reprogramming, Reversibility, and Persistence

More than 80% of low-grade gliomas harbor mutations in IDH1 or IDH2, which has spurred intense research into the roles these mutations and associated oncometabolite 2HG play in gliomagenesis. Mutant IDH induces global epigenetic alterations involving DNA hypermethylation that appear to promote tumor growth. IDH inhibitors are currently being tested in clinical trials and in in vivo models but the impact on tumor growth is variable raising questions about the therapeutic benefit or, more precisely, how best to target this promising tumor specific marker.

In a recent paper, Turcan et al used immortalized human astrocytes and patient-derived glioma tumorspheres models to better characterize dynamic alterations in DNA methylation, histones, and transcriptional reprogramming in the setting of IDH1 (R132H) mutation.¹ They found that some epigenetic changes due to IDH1 were reversible but others were not despite withdrawing mutant IDH1, suggesting a possible mechanism for IDH inhibitor resistance. In particular, L1CAM, a molecule expressed on glioma stem cells, was persistently upregulated while CD24 expression was IDH dependent (CD24 is another putative stem cell marker).

Both repressive and activating chromatin marks were expressed in the setting of IDH mutation, and DNA methylation appeared to occur more frequently in regions of quiescent or low chromatin state. The authors postulate that this leads to a more open chromatin pattern, allowing for aberrant transcriptional activation and, thus, genomic instability. In addition, gain in H3K4me3 at the PDGFRA promoter was seen, building on the intriguing story of PDGFRA alterations in IDH mutant gliomas.²

This study comprehensively identifies dynamic changes in DNA methylation in IDH mutant gliomas and uncovers potential novel therapeutic strategies.

References