Year: 2021
Type: article
Abstract: Although immunotherapies have become a strong therapeutic option against a number of solid cancers, the number of patients with glioblastoma (GBM) who show long-term improvement after immune interventions remains small 1. Accordingly, it has become critical to determine why only some GBM patients respond well to immunotherapy and how those patients can be pre-emptively identified. To identify potential markers of immunoresponsiveness, a recent study from Gromeier and colleagues 2 analyzed the molecular features of GBM in patients treated with different immunotherapies after tumor recurrence, including immune checkpoint-blocking antibodies as well as therapy with an oncolytic virus that induces innate immune inflammation 3. In all these cases, the investigators found that a low or very low frequency of mutations in the recurrent GBMs was a key feature correlated with improved patient survival after immunotherapy. Importantly, this low tumor mutational burden (TMB) did not affect tumor recurrence or the survival of patients who did not receive immunotherapy, indicating that the mutational load did not have an intrinsic effect on tumor progression but simply reflected the ability of the recurrent tumor to respond to immunostimulatory treatments. To further understand how the TMB could reflect immunoresponsiveness, the investigators analyzed immune signatures in the tumors and observed that the TMB of primary GBMs had negligible impact on their immune status. In sharp contrast, in patients that had recurrence after receiving standard chemoradiation, a low TMB in the recurrent tumor correlated with a clear signature of immune inflammation and T cell reactivity. The investigators interpreted their results by proposing that standard-of-care therapy causes some degree of immune activation, which only persists in tumors with low mutational burden and can be detected when these tumors recur. This smoldering immune response is insufficient to stop the growth of the recurring GBMs; however, when external immunotherapy is applied the ongoing immune response is boosted to the point that it can damage the tumor and achieve a therapeutic effect, therefore improving patient survival. It is important to note that this study did not define absolute measures of mutational load or immune response that could be used as prognostic criteria of individual tumors. Nevertheless, these compelling results set the stage to further refine the use of TMB combined with other immunological markers to identify patients who respond to immunotherapy and to evaluate treatments that could drive patients towards an immunoresponsive status. Chimeric antigen receptor (CAR)-based immunotherapy is a promising therapeutic approach for non-solid tumors but faces many challenges in glioblastoma (GBM), including a highly immunosuppressive tumor microenvironment (TME), and antigen heterogeneity. CAR-T cell efficacy has been shown to improve with the use of IL-12 in preclinical models of hematologic malignancies. IL-12 is a pro-inflammatory cytokine with potent tumor-suppressive activity that can affect the adoptively transferred CAR-T cells, the endogenous T cell compartment, and can skew the myeloid compartment towards a pro-inflammatory phenotype within the TME. Although systemic delivery of IL-12 has been reported to increase toxicity, this can be mitigated by local administration of IL-12. Agliardi et al reported that combining EGRFvIII targeting CAR T cells with local delivery of IL-12 reshapes the TME to sustain anti-tumor immunity in two preclinical orthotopic models: a highly immunogeneic (GL261) GBM and a less immunogeneic (B16.F10) melanoma models.1 A parental GL261 line was transduced with the extracellular portion of EGFR containing the variant III mutation fused with the transmembrane domain of the murine EGFR. A second-generation CAR was constructed, and murine T cells were transduced to express the EGFRvIII-specific CAR to efficiently kill EGFRvIII+ GL261 but not parental GL261 in vitro. They tested efficacy of EGFRvIII-directed CAR-T cells in vivo in mice bearing orthotopic EGFRvIII+ GL261 tumors. CAR-T cell administration resulted in tumor clearance and long-term survival in 50% of treated mice. They then developed a ‘stress’ model using a reduced dose of CAR-T cells administered at day 11 (early stage) or day 17 (late stage) post-tumor implantation. Early administration of CAR-T cells resulted in control of tumor growth and significantly improved survival, but late administration of CAR-T cells resulted in poor tumor control, suggesting that CAR-T cells are more efficacious against newly established tumors and less effective on mature tumors. Importantly, the latter more closely resembles the stages at which CAR-T cells will be used in the clinic. Tumor volumes as measured by MRI showed that, while both single therapies were able to delay tumor growth, the combination of systemic EGFRvIII-specific CAR infusion and local IL-12:Fc administration eliminated tumors and showed a synergistic effect on the overall survival. GL261 is a highly immunogenic model. This does not represent the immunologically “cold” human gliomas. Thus, a second intracranial tumor model, B16.F10 was also evaluated to recapitulate brain metastases of melanoma. In this aggressive but less immunogenic intracranial model, combined IL-12 and CAR-T cell therapy also promoted an effective and persistent anti-tumor response. The results of this study will need confirmation in models closer to human GBM. The variable expression of EGFRvIII will also need to be addressed in studies targeting EGFRvIII using CAR-T cell approaches. For an IL-12:Fc/CAR-T cell therapy to be successful in GBM, targeting of multiple antigens will be key. Combination approaches with checkpoint inhibitors or agonistic antibodies may need to be evaluated in future combinations with cell-based therapies. The discovery of the K27M and G34R/V mutations in histone H3 genes as driving alterations in pediatric high-grade glioma (pHGG) was made all the more intriguing by the characteristic spatial segregation of affected tumors within the neuraxis. Specifically, K27M-mutant tumors exclusively arise in midline structures extending from the spinal cord to the diencephalon, whereas G34R/V-mutant tumors arise in the cerebral hemispheres, with no hindbrain involvement. In a recent Cell Stem Cell article, Bressan et al. explored the foundations of this geographic distinction, employing neural stem cell (NSC) lines derived from both fetal forebrain and hindbrain.1 They first confirmed that their forebrain and hindbrain NSCs maintained appropriate marker expression after passaging and exhibited strong transcriptional correlations with G34R/V and K27M-mutant pHGGs, respectively. They then stably expressed wild-type and mutant H3.3 genes in both NSC cell types, assessing phenotypic outcome. In this context, they found that K27M-mutant H3.3 exclusively transformed hindbrain NSCs, yielding increased proliferation and clonogenicity, along with reduced senescence, consistent with earlier literature.2 By contrast, G34R-mutant H3.3 induced overt senescence in hindbrain NSCs, whereas its expression in proliferating forebrain NSC cultures was largely tolerated. That being said, the failure of G34R H3.3 to transform even forebrain NSCs signaled a pathogenic requirement for coinciding molecular alterations, echoing recent work on PDGFRA amplification in supratentorial, G34R/V-mutant pHGG.3 Indeed, coincident PDGFRA overexpression and TP53 knockout successfully induced proliferation and clonogenicity and repressed senescence in G34R-expressing forebrain NSCs, but not in hindbrain counterparts. The authors then ablated the existing G34R mutation in a patient-derived pHGG cell line, documenting reduced proliferation and clonogenicity and increased senescence in a manner effectively reverted by G34R-mutant, but not wild-type, restoration. Surprisingly, G34R ablation did not result in widespread shifts in transcription or the distribution of histone marks, reminiscent of prior work indicating that G34R/V mutations only impair H3K36 trimethylation on the mutant histones themselves.4 As such, specific forebrain-related genes enriched for G34R occupancy were more selectively impacted. Finally, the authors demonstrated that CRISPR ablation of the transcription factor FOXG1, a master regulator of forebrain identity, hampered the growth of G34R-mutant patient-derived glioma cells in vivo, providing further evidence that specific developmental contexts are required for G34R-mediated tumorigenesis. This findings are highly significant in that they further clarify the unique molecular and histiogenic context driving G34R/V-mutant pHGG, further establishing therapeutically relevant links between developmental state and oncogenic potential in brain cancer. A wealth of transcriptional profiling data has informed our understanding of glioblastoma. Yet, transcriptomic classification of glioblastoma into molecular subtypes has, so far, failed to inform prognosis or treatment. Garofano et al. opted to take an integrated approach, instead of a reductionist approach, to interrogate the transcriptional profile of glioblastoma 1. Using pathway-extrapolation from published single-cell RNAseq data, the authors classified glioblastoma into glycolytic/plurimetabolic (GPM), mitochondrial (MTC), neuronal (NEU), and proliferative/progenitor (PPR). The GPM subgroup was characterized by concurrent activation of multiple energy-producing programs, including glycolysis/hypoxia-related, lipid, amino acid, steroid, iron, and sulfur metabolism. Mesenchymal and immune-related functions were also a feature of this new subgroup. Mitochondrial/OXPHOS activities, however, were not a characteristic of GPM subgroup but instead associated with the MTC subgroup. Additionally, MTC was correlated with fatty acid oxidation. The NEU subgroup was defined by neuronal functions such as axonogenesis and synaptic transmission. The PPR subgroup was characterized by cell cycle progression, DNA replication, mitosis, DNA damage, and neural stem/progenitor cell markers. This new classification has parallels with previously postulated schemes 2: NEU and PPR subgroups resemble NPC1, NPC2, and OPC states. The GPM subgroup resembles the MES state, and MTC the AC-like state. The GPM subgroup featured PTEN, RB1, and NF1 mutations and amplification of MDM4. The MTC subgroup presented with NRAS mutations and CDK4/MDM2 amplifications. The NEU subgroup displayed ATRX and TET1 mutations. The PPR subgroup was characterized by the presence of PDGFRA and EZH2. EGFR amplification and mutations were observed in the MTC and PPR subtypes. Differences in subgroup enrichment between tumor rim and core were also observed. The NEU features were more prevalent (41%) in peripheral areas of the tumor, whereas PPR (38%), MTC (26%), and GPM (23%) were enriched in the tumor core. Importantly, the dominant state of the tumor, classified as MTC according to this new paradigm, correlated with a reduced risk of death, independent of age, sex, or MGMT methylation status for patients with IDH wildtype glioblastoma. Furthermore, the glucose-proton symporter SLC45A1 was predicted to be a tumor suppressor in the MTC subgroup. The identification of a metabolic gene expression module specific to the MTC subgroup not only offers a new classification scheme for glioblastoma with prognostic value, but also represents a potential therapeutic vulnerability of this lethal disease. Inhibitors of mitochondrial complex I, an inhibitor of mitochondrial protein synthesis and an inducer of mitochondrial ROS, reduced the viability of MTC patient-derived cell viability but not of GPM cells, offering hope that this classification scheme may lead to disease stratification with therapeutic implications. In recent years, single cell RNA-seq (scRNA-seq) revolutionized our understanding of intra-tumoral heterogeneity in gliomas. The common theme that emerged from these studies is that glioblastoma (GBM) cells exist in diverse gradients of transcriptomic states. Rather than depending on a hierarchical organization of cancer stem cells and their progeny, GBMs display high plasticity, where tumor cells undergo reversible state transitions towards phenotypes adapted to their surrounding microenvironment1,2. In a recent paper in Nature Cancer3, Richards et al. expand scRNA-seq analysis to 28 GBM stem-like cell cultures (GSCs) and 7 patient tumors. The authors confirm high inter-patient and intra-tumoral transcriptomic diversity and focus their analysis on cultured GSCs. They describe two major axes of variation: (i) a gradient of cells expressing neurodevelopmental processes (i.e. Developmental state) to cells with active inflammatory wound response programs (i.e. Injury Response state); and (ii) an orthogonal gradient recapitulating astrocyte maturation from stem-like to differentiated states. The first program was very pronounced in GSCs in vitro and was less visible in primary GBM cells isolated from patient tumors. GSCs with a high proportion of Injury Response states expressed pathways of inflammation and resembled mesenchymal-like states and reactive astrocytes. At the other end of the gradient, Developmental states showed programs reminiscent of neural development and gliogenesis, resembling previously described NCP, OPC, and AC-like states.1 Although Developmental-type GSCs showed increased sphere formation capacity, all GSCs self-renewed and gave rise to tumors in vivo at similar latency. Plasticity between these states was mediated by inflammatory cytokines, such as tumor necrosis factor (TNF)-α and IL-1α, suggesting a role for tumor-associated macrophages in shaping these transitions. The final composition of transcriptomic states varied among parental tumors and derived cell lines, indicating that a large proportion of states were patient specific, partially depending on the genetic background. CRISPR-Cas9 dropout screens performed on representative GSCs identified genes characterizing Developmental and Injury Response states that were essential for their growth. Developmental GSCs relied more on aerobic metabolism, whereas Injury Response cells were more dependent on glycolysis, cell cycle, and DNA repair. Based on the very low proportion of GSC-like cells detected in patient samples, the authors suggest that the differentiation axis observed in patient tumors supports a hierarchical organization of GBM. This, however, does not consider the evident transcriptomic plasticity of GBM cells, also meaning that cells put into culture will likely adapt towards phenotypic states most adapted to the culture conditions. This study was published in parallel with two papers describing additional gradients around neurodevelopmental and metabolic axes4,5. It will be interesting to determine how the various transcriptomic programs relate to each other and how they can be harnessed for therapy design. Clearly, the traditional concept of uniform cancer stem cells as therapeutic target needs adjustment. Instead, future therapies need to take into account inherent heterogeneity and plasticity of GBM cells.
Source: Neuro-Oncology
Language: English
Subfield: Genetics
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4213226840