https://orcid.org/0000-0003-2552-6546
Abstract
ATRX inactivation occurs with IDH1R132H and p53 mutations in over 80% of Grades II/III astrocytomas. It is believed that ATRX loss contributes to oncogenesis by dysregulating epigenetic and telomere mechanisms but effects on anti-glioma immunity have not been explored. This paper examines how ATRX loss contributes to the malignant and immunosuppressive phenotypes of IDH1R132H/p53mut glioma cells and xenografts.
Isogenic astrocytoma cells (+/−IDH1R132H/+/−ATRXloss) were established in p53mut astrocytoma cell lines using lentivirus encoding doxycycline-inducible IDH1R132H, ATRX shRNA, or Lenti-CRISPR/Cas9 ATRX. Effects of IDH1R132H+/−ATRXloss on cell migration, growth, DNA repair, and tumorigenicity were evaluated by clonal growth, transwell and scratch assays, MTT, immunofluorence and immunoblotting assays, and xenograft growth. Effects on the expression and function of modulators of the immune microenvironment were quantified by qRT-PCR, immunoblot, T-cell function, macrophage polarization, and flow cytometry assays. Pharmacologic inhibitors were used to examine epigenetic drivers of the immunosuppressive transcriptome of IDH1R132H/p53mut/ATRXloss cells.
Adding ATRX loss to the IDH1R132H/p53mut background promoted astrocytoma cell aggressiveness, induced expression of BET proteins BRD3/4 and an immune-suppressive transcriptome consisting of up-regulated immune checkpoints (e.g., PD-L1, PD-L2) and altered cytokine/chemokine profiles (e.g., IL33, CXCL8, CSF2, IL6, CXCL9). ATRX loss enhanced the capacity of IDH1R132H/p53mut cells to induce T-cell apoptosis, tumorigenic/anti-inflammatory macrophage polarization and Treg infiltration. The transcriptional and biological immune-suppressive responses to ATRX loss were enhanced by temozolomide and radiation and abrogated by pharmacologic BET inhibition.
ATRX loss activates a BRD-dependent immune-suppressive transcriptome and immune escape mechanism in IDH1R132H/p53mut astrocytoma cells.
ATRX loss activates BRD-dependent immune escape in IDH1 mutant astrocytoma.
The immune-suppressive transcriptome is augmented by chemo-radiation.
Immunotherapy is effective for a number of cancers, but primary brain tumors remain unresponsive. We describe a previously unrecognized BRD-dependent immune escape mechanism in IDH1R132H/p53mut/ATRXloss astrocytoma cells. This mechanism and its augmentation by standard-of-care chemo-radiation likely contribute significantly to the immunosuppressive tumor microenvironment and evasion of anti-tumor immunity. Their inhibition should improve the efficacy of emerging immunotherapeutics for WHO Grades II/III astrocytomas.
Low and intermediate grade IDH1 mutant astrocytomas (WHO Grades II/III) primarily occur in young adults and are associated with prolonged survival relative to glioblastoma and other WHO Grade IV astrocytomas. However, these lower grade malignancies have a high propensity to recur and progress to higher grade despite surgical resection and ionizing radiation +/− chemotherapy.1 Since treating recurrent malignant astrocytoma has been frustratingly ineffective, it is imperative to design more effective therapies that can ideally be incorporated with current standard-of-care therapy before recurrence. Immunotherapy offers this opportunity since the ability of astrocytoma to evade immunity plays a critical role in resistance to current therapies and recurrence. A more complete understanding of the mechanisms employed by glioma cells to escape anti-tumor immunity is needed.
Whole-genome sequencing has shown that ~80% of Grades II/III astrocytomas contain IDH1R132H mutation, p53 mutation, and loss-of-function of the chromatin remodeler ATRX.2 IDH1 mutation is recognized as an early event in astrocytoma development prior to the accumulation of the p53 and ATRX mutations.3 This sequence of mutational events indicates that p53 and ATRX mutations play pivotal roles in astrocytoma initiation and/or progression and immune evasion that are required for the formation of clinically significant tumors. Recent studies have demonstrated that ATRX loss, when combined with IDH1R132H in glioma drives the alternative lengthening of telomere (ALT) phenotype,4 enhances DNA repair efficiency and resistance to ionizing radiation.5 ATRX loss in combination with p53 loss was found to promote murine neuroepithelial progenitor cell migration and astrocytic lineage differentiation.6 Whether and how ATRX loss promotes astrocytoma immune evasion has not been explored.
Emerging evidence indicates that IDH1 mutant gliomas contain fewer tumor-infiltrating CD4+ and CD8+ T cells than IDH1 wild-type gliomas, suggesting that the IDH1 mutation represses anti-tumor immunity.7–9 The molecular basis for the immune suppression associated with IDH1 mutant astrocytoma remains poorly understood due in part to its association with other oncogenic mutations such as ATRXloss. In this paper, we specifically address the roles of ATRX loss using transgenic IDH1R132H/p53mut+/−ATRXloss astrocytoma models. Introducing ATRX loss to the IDH1R132H/p53 mutation background promoted colony growth in soft agar, migration and invasion, chemo-radiation resistance, and tumor xenograft growth in vivo. Comprehensive screening of immune checkpoint, immunosuppressive cytokine/chemokine expression profiles, and epigenetic mediators showed that ATRX loss induces immunosuppressive ligands (e.g., PD-L1, PD-L2), an immunosuppressive cytokine/chemokine profile (e.g., IL33, CXCL8, CXCL3, IL6, and CXCL9) and BET proteins BRD3/4. Furthermore, we show that adding ATRXloss to the IDH1R132H/p53mut background increases PD-L1–dependent T-cell apoptosis and induces the expansion of pro-tumorigenic/anti-inflammatory M2-like macrophages. These transcriptional and immune-suppressive responses to IDH1R132H/p53mut/ATRXloss are enhanced by standard-of-care temozolomide (TMZ)/radiation (RT) and inhibited by BET inhibitors.
Materials and Methods
Preparation of Transgenic Astrocytoma Cells (+/−IDH1R132H/+/−ATRXloss)
See Supplemental Materials and Methods for details regarding the generation of lentiviruses. U373 or Snb19 cells were infected with lentivirus encoding IDH1R132H with either sh-ATRX, CRISPR/Cas9 ATRX or control virus. Cells were cultured in puromycin (1 µg/mL for U373 and 5 µg/mL for Snb19) and hygromycin B (500 µg/mL) to establish stable cell lines U373 and Snb19 expressing IDH1R132H and ATRXloss. For the generation of Snb19 ATRX KO cells, 2 independent ATRX sgRNAs were used. Single clones were sorted into 96-well plates and tested for KO by DNA-PCR using primers (F: GGGTTTGTGGAGTTATAGGTATTG; R: GAACTCATGTCAACCACTATTCTG) and immunoblot analysis. For the generation of GL261 expressing ATRXloss, three shRNA-mATRX sequences were evaluated for mATRX knockdown efficiency (sequence 1: CCTTCTAACTACCAGCAGATT; sequence 2: CCCACGGATGAGAATGTAAAT; sequence 3: CCACTAACACTCCTGAGGATT). Only sequence 2 was found to sufficiently knockdown mATRX (>80%) and sequence 2 was used for generating GL261-ATRXloss cells.
T-cell Apoptosis, Proliferation, IL2 Production Assays and Tumor Cell Killing Assay
4 × 105 astrocytoma cells expressing IDH1R132H/p53mut+/−ATRXloss were treated +/− 600 µM TMZ in culture medium containing doxycycline for 48 h. Cells were then washed to remove TMZ and co-cultured with 4 × 104 activated Jurkat cells or activated CD8 T cells overnight. Jurkat cell or CD8 T-cell apoptosis was measured using the Caspase-Glo 3/7 assay (Promega). IL2 levels in co-culture supernatants were measured by ELISA (BioLegend). To evaluate the effects of PD-L1 inhibition, IDH1R132H/p53mut+/−ATRXloss cells were pretreated with 20 mg/mL purified anti-PD-L1 antibody (29E.2A3, BioLegend) for 24 h prior to co-culture with activated Jurkat cells. For CD8 T-cell-mediated tumor cell killing assay, U373 cells expressing IDH1R132H/p53mut+/−ATRXloss were cultured in 96-well plates overnight prior to co-culture with activated CD8+ T cells (1:5 tumor cell: CD8+ T-cell ratio) for 48 h. Non-adherent cells were removed by washing with PBS and the remaining adherent U373 IDH1R132H/p53mut+/−ATRXloss tumor cells were photographed and then quantified by CCK-8 assay. To evaluate the effects of macrophages on CD8 T-cell proliferation, the number of CFSE-labeled CD8 T cells (carboxyfluorescein succinimidyl ester, CFSE) was quantified by flow cytometry after 72 h co-culture with macrophages.
Xenograft and GL261 Mouse Models
All animal protocols were approved by the Johns Hopkins Animal Care and Use Committee. Tumor growth in vivo was examined using a subcutaneous xenograft model and intracranial glioma mouse model as previously described with some modifications.10 For subcutaneous xenografts, U373 cells expressing +/−IDH1R132H/p53mut+/−ATRXloss (4 × 106n = 6) were mixed with an equal volume of matrigel prior to implant to the flanks of 8-week-old female immunodeficient mice (SCID, The Jackson Laboratory, Bar Harbor, ME, USA). After 21 weeks, the animals were sacrificed by perfusion with 4% paraformaldehyde. Tumor sizes were measured using calipers, and volumes were estimated by the formula: volume = (length × width2)/2. The expression of IDH1-R132H and ATRX were quantified by immunohistochemistry (see Supplementary Figure 1C and D). For orthotopic tumors, GL261 cells expressing IDH1R132H/p53mut+/−ATRXloss (1 × 105/2 μL PBS, n = 10) were stereotactically implanted into the right caudate/putamen of the 8-week-old female C57BL/6j mice (Jackson Laboratory). When the first animal showed neurological symptoms (~27 days post-implantation), all animals were sacrificed and perfused. Brains were removed, sectioned, and select sections stained with hematoxylin/eosin. Expression of IDH1-R132H and mATRX were quantified by immunohistochemistry (see Supplementary Figure 5B). Tumor sizes were quantified by measuring the maximum tumor cross-sectional area on hematoxylin/eosin-stained coronal sections using computer-assisted image analysis (MCID software) and volumes estimated as previously described.10
Statistical Analysis
All experiments were preformed in >2 biological replicates and all in vitro experiments performed three times. Results are expressed as means ± standard error of the mean (SEM). The significance of differences was determined using GraphPad Prism software (GraphPad Software, La Jolla, CA, USA). Means were compared using analysis of one-way ANOVA. Post-hoc tests included either Student’s t-test, Tukey, or Mann–Whitney test as required by experimental design. Significant difference from the corresponding control is indicated as *, **, or *** for P < .05, .01, or .001, respectively.
Results
ATRX Loss Promotes Aggressive Phenotypes in IDH1 Mutant Astrocytoma Cells
To determine if ATRX loss cooperates with IDH1 and p53 mutations to promote astrocytoma malignant progression, we generated transgenic human astrocytoma models with p53 mutation +/−IDH1R132H mutation and +/−ATRX knockdown and knockout. U373 and Snb19 cells with endogenous p53 mutant were sequentially transfected first with doxycycline (Dox)-inducible IDH1R132H11 and then ATRX-targeting shRNA (or control shRNA)12 or ATRX-specific CRISPR-Cas9 KO via lentivirus.13 Western blot analysis and PCR gel electrophoresis analysis of these cell models confirmed IDH1R132H expression upon doxycycline treatment, and ATRX expression inhibition or knockout (Figures 1A, Supplementary Figure 1A). Gas chromatography/mass spectrometry analysis of culture medium from cells treated +/−Dox showed >50-fold increase of the oncometabolite 2-hydroxyglutarate (2-HG) expected from IDH1R132H expression (Figure 1B). The 2-HG levels in these cells treated with Dox are comparable to that reported in heterozygous cells.14,15
ATRX loss promotes aggressive phenotypes in IDH1R132H/p53mut astrocytoma cells. P53 mutant U373 and Snb19 cells +/−IDH1R132H, +/−ATRXloss were established using lentivirus encoding either doxycycline (DOX) inducible IDH1R132H or shRNA against ATRX (or control) as described in Materials and Methods. A. Immunoblots showing induction of IDH1R132H by DOX and ATRX expression inhibition by sh-ATRX and CRISPR/Cas9 ATRX. B. 2-Hydroxyglutarate (2-HG) levels in conditioned medium of U373 and Snb19 cells expressing +/−IDH1R132H measured by mass spectrometry. C. Clonogenicity of U373 cells (IDH1R132H/p53mut +/−ATRXloss) in soft agar. Images of colony densities are shown (left panel). Colonies >100µm diameter were quantified by computer-assisted image analysis (right panel). D. Invasion capacities of U373 IDH1R132H/p53mut+/−ATRXloss cells measured by transwell invasion assay. E. Immunoblot showing effects of TMZ, γ-radiation or TZP on γH2AX expression in U373 IDH1R132H/p53mut+/−ATRXloss cells. F. Immunofluorescent staining and quantification of γH2AX expression in IDH1R132H/p53mut/ATRXloss cells treated +/−TMZ. G. Viability measured by MTT assay of U373 IDH1R132H/p53mut+/−ATRXloss treated with TMZ or PARP inhibitor TZP for 72 h. *P < .05, **P < .01, and ***P < .001.
We assessed the impact of ATRX loss on multiple parameters linked to glioma malignancy. Adding ATRX loss to the IDH1R132H/p53mut background increased colony formation in soft agar (Figure 1C), invasion in transwell assays (Figure 1D), and migration using wound-healing assays (Supplementary Figure 1B). ATRX loss substantially decreased DNA double-strand breaks (DSBs) in IDH1R132H/p53mut cells treated with TMZ, γ-radiation or the PARP (poly ADP-ribose polymerase) inhibitor TZP (Talazoparib) as determined by γ-H2AX phosphorylation and γ-H2AX foci formation (Figure 1E and F). The addition of ATRX loss to the IDH1R132H/p53mut background also decreased cell sensitivity to temozolomide (TMZ) and to TZP (Figure 1G). Consistent with published reports,16 ATRX knockdown enhanced the growth of tumor xenografts derived from IDHwt/p53mut and IDH1R132H/p53mut astrocytoma cells (Supplementary Figure 1C–E) and increased tumor Ki67 proliferative index (Supplementary Figure 1H). Together, these results show that ATRX loss promotes a more malignant phenotype in IDH1R132H/p53mut astrocytoma cells and enhances tumor propagating capacity in p53mut cells with either wild-type or mutant IDH1.
ATRX Loss Induces Immune Checkpoints and Alters Cytokine/Chemokine Profiles in IDH1R132H/p53mut Cells
We examined the effects of ATRXloss on immunomodulatory pathways in IDH1R132H/p53mut cells by first performing a comprehensive screen of immune checkpoint molecules (total of 22 genes) by qRT-PCR. Statistically significant expression changes >2-fold (increase or decrease) in both biological replicates (i.e., U373 and Snb19 models) was required for further study. Based on these criteria, PD-L1 and PD-L2 were up-regulated and CEACAM1 was down-regulated in IDH1R132H/p53mut/ATRXloss cells compared with IDH1R132H/p53mut/ ATRXwt cells (Figures 2A, Supplementary Figure 2A). PD-L1 was most prominently up-regulated (2.5–3.5-fold) by ATRXloss in our experimental biological replicates. Western blotting and immune fluorescence analysis showed that PD-L1 was markedly increased in the triple mutant cells (IDH1R132H/p53mut/ATRXloss) compared with either double mutant (IDH1wt/p53mut/ATRXloss or IDH1R132H/p53mut/ATRXwt) or single mutant (IDH1wt/p53mut/ATRXwt) cells (Figures 2B and C, Supplementary Figure 2B), indicating that ATRX loss in the IDH1R132H/p53mut background drives PD-L1 induction.
ATRX loss induces PD-L1 and an immunosuppressive cytokine/chemokine profile in IDH1R132H/p53mut cells in vitro and in vivo. A. Expression quantification by qRT-PCR of 22 immune checkpoint ligands in IDH1R132H/p53mut+/−ATRXloss U373 cells; inhibitory immune checkpoint molecules are labeled red and co-stimulatory ligand is labeled green. B. Immunoblots showing PD-L1 expression in IDH1R132H/p53mut/ATRXloss, IDH1R132H/p53mut, ATRXloss/p53mut, and p53mut U373 and Snb19 cells. C. Immunofluorescence showing membrane distribution of PD-L1 in IDH1R132H/p53mut+/−ATRXloss U373 cells. D. Expression quantification by qRT-PCR of 21 cytokines and chemokines in IDH1R132H/p53mut+/−ATRXloss U373 cells (left panel). E. Immunoblots showing IL33 level in whole cell extracts. F. and G. Immunohistochemistry and quantification of PD-L1 and IL33 in IDH1R132H/p53mut+/−ATRXloss U373 tumor xenografts. H. IL-33 and CXCL3 expression in TCGA IDH1mut/non-1p19q co-deleted ATRX-mutant vs. ATRX-WT clinical samples (http://gliovis.bioinfo.cnio.es). *P < .05, **P < .01, and ***P < .001.
Chemokines and cytokines have important roles in modulating glioma immune escape through multiple mechanisms. We performed a comprehensive screen of 21 cytokines/chemokines in IDH1R132H/p53mut/+/−ATRXloss cells. These cytokines/chemokines have multiple functions including the recruitment of effector CD8+ cells and macrophages, myeloid-derived suppressor cells (MDSCs), and regulatory T cells (Tregs) into the tumor microenvironment.17–19 qRT-PCR analysis, requiring statistically significant expression changes >2-fold (increase or decrease) in both biological replicates, identified five up-regulated (i.e., IL-33, CXCL8, IL-6, CXCL3, CSF2, labeled red) and one down-regulated (i.e., CXCL9, labeled green) cytokines/chemokines in IDH1R132H/p53mut/ATRXloss cells compared with IDH1R132H/p53mut/ATRXwt cells (Figures 2D, Supplementary Figure 2C). Immunoblot of cellular protein confirmed high expression of IL33 in IDH1R132H/p53mut/ATRXloss cells (Figure 2E). Consistent with in vitro findings, immunohistochemistry staining of tumor xenografts confirmed a substantial increase in expression of PD-L1 and IL33 in IDH1R132H/p53mut/ATRXloss-derived tumors compared with IDH1R132H/p53mut/ATRXwt-derived tumors (Figure 2F and G). Analysis of TCGA-IDH1mut/non-1p19q co-deletion glioma samples show trends associating higher IL33 and CXCL3 expression with clinical ATRX-mutant glioma samples compared with ATRX-WT glioma samples (P = .09 and P = .07, respectively) (Figure 2H). TCGA analyses revealed no statistically significant associations between ATRX status and IL6, IL8, CXCL9 (Figure S2E).
Collectively, these results indicate that IDH1R132H/p53mut/ATRXloss promotes adaptive and innate immunosuppressive mechanisms by modulating immunosuppressive checkpoints (PD-L1 and PD-L2) and cytokine and chemokine expression (IL33, CXCL3).
Effects of ATRXloss on Immune Modulator Expression are Enhanced by Chemo-radiation
TMZ and γ-radiation are standards-of-care for astrocytoma, and PARP inhibitors are undergoing clinical testing in patients with IDH1 mutant tumors.20 It is important to understand the effects of therapeutics on the immune microenvironment of IDH1R132H/p53mut/ATRXloss astrocytoma in order to optimally integrate immunotherapeutics with standard-of-care therapies.21 PD-L1 mRNA and protein levels were significantly increased in IDH1R132H/p53mut/ATRXloss cells following their treatment with either temozolomide (TMZ), γ-radiation (IR), or the PARP inhibitor TZP (Figures 3A and B, Supplementary Figure 3A). Similarly, PD-L1 expression determined by flow cytometry was up-regulated in response to TMZ and radiation in MGG119 and MGG 152 cells derived from a clinical IDH1R132H/p53mut/ATRXmut astrocytoma (Figure 3C). Effects of TMZ and radiation on the expression of key cytokines and chemokines (i.e., IL33, CXCL8, CSF2, IL6, and CXCL9) regulated by IDH1R132H/p53mut/ATRXloss as shown in Figure 2 were also examined. IDH1R132H/p53mut/ATRXloss cell models showed that CXCL8 and IL6 are significantly increased by TMZ and radiation. TMZ and radiation increased IL33 expression in two IDH1R132H/p53mut/ATRXloss cell models (Figures 3D and E, Supplementary Figure 3B) and CSF2 was induced by radiation and minimally induced by TMZ (Figure 3D). CXCL9 expression was significantly decreased by IR but not by TMZ in IDH1R132H/p53mut/ATRXloss cells compared with IDH1R132H/p53mut/ATRXwt cells (Figures 3D and Supplementary Figure 3C). Thus, standard-of-care chemo-radiation therapy activates multiple responses with potential immune-suppressive implications in IDH1R132H/p53mut/ATRXloss astrocytomas.
Chemo-radiation augments the immunosuppressive transcriptome of IDH1R132H/p53mut/ATRXloss astrocytoma cells. A. and B. PD-L1 expression in IDH1R132H/p53mut+/−ATRXloss U373 and Snb19 cells treated with TMZ (600 μM) and radiation (5 Gy) for 5 h (qRT-PCR, panel A) and for 48 h (immunoblot, panel B). C. Cell-surface PD-L1 expression determined by flow cytometry in patient-derived IDH1R132H/p53mut/ATRXloss MGG119 and MGG152 astrocytoma cells treated with TMZ (600 μM) and radiation (5Gy) for 48 h. D. qRT-PCR analysis measuring expression of key cytokines/chemokines in IDH1R132H/p53mut+/-ATRXloss U373 cells treated with TMZ or IR. E. qRT-PCR analysis measuring expression of key cytokines/chemokines in IDH1R132H/p53mut/ATRXloss MGG119 cells treated with TMZ and radiation. *P < .05, **P < .01, and ***P < .001.
Loss of ATRX in IDH1R132H/p53mut Cells Suppresses T cells and Induces Pro-tumorigenic/Anti-inflammatory M2-like Macrophages
The PD-L1 and chemokine/cytokine expression changes induced by ATRXloss predicted coincident immune-suppressive effects. These were examined in vitro and in vivo. CD8 T-cell apoptosis determined by caspase-3/7 activation was increased >1.8-fold when anti-CD3/anti-CD28 activated human CD8+ T cells were co-cultured with IDH1R132H/p53mut/ATRXloss vs IDH1R132H/p53mut/ATRXwt cells (Figure 4A, left panel). Consistently, the CCK-8 assay showed enhanced survival of ATRXloss cells compared with ATRXwt cells indicating that ATRXloss imparts relative resistance to T-cell-mediated tumor cell killing (Figure 4B). Similarly, Jurkat cell apoptosis determined by caspase-3/7 activation was increased >5-fold when co-cultured with IDH1R132H/p53mut/ATRXloss vs. IDH1R132H/p53mut/ATRXwt cells (Figures 4C, left panel and S4A). Jurkat cell apoptosis was increased even further when co-cultured with TMZ pretreated IDH1R132H/p53mut/ATRXloss cells (~11-fold) vs. IDH1R132H/p53mut/ATRXwt cells (~2-fold) (Figure 4C, left panel). Pretreating tumor cells with anti-PD-L1 antibody prior to co-culture prevented Jurkat cell apoptosis confirming apoptosis induction by PD-L1 (Figure 4C, left panel). Jurkat cell activation measured by IL2 production was also inhibited by IDH1R132H/p53mut/ATRXloss cells and this effect was also abrogated by anti-PD-L1 (Figure 4C, right panel). PD-L1-dependent inhibition of IL2 production in co-cultures was also significantly enhanced by TMZ (Figure 4C, right panel).
ATRX loss in IDH1 mutant glioma cells suppresses T cells and induces immune-suppressive M2 macrophages. A. IDH1R132H/p53mut+/-ATRXloss U373 cells were co-cultured with activated CD8+ T cells (10:1) and CD8 T-cell apoptosis measured by caspase 3/7 activation. B. IDH1R132H/p53mut+/−ATRXloss U373 cells were co-cultured with activated CD8+ T cells (1:5) for 24 h, remaining viable adherent cells shown in phase-contrast photomicrographs (left panel) and quantified by CCK-8 assay (right panel). C. IDH1R132H/p53mut+/−ATRXloss U373 cells were treated +/−TMZ for 48 h prior to culture with activated Jurkat cells +/−anti-PD-L1 antibody and Jurkat cell apoptosis measured by caspase 3/7 activation (left panel). IL2 production as a marker of Jurkat cell activation was measured by ELISA (right panel). D. Activated, CFSE-labeled PBMC-derived CD8+ T cells were co-cultured with macrophages pretreated with conditioned medium from ATRXwt or ATRXloss cells and T-cell proliferation quantified by flow cytometry (left panel). Histogram shows CD8+ T-cell numbers (right panel). E. Expression levels of CD8A and CD68 in clinical ATRXmut vs. ATRXwt TCGA-IDHmut-non-1p19q co-deletion astrocytoma datasets (http://gliovis.bioinfo.cnio.es). F. TIMER datasets showing infiltration of CD8+ T cells and macrophages in ATRXmut vs. ATRXwt glioma samples (http://cistrome.org/TIMER). *P < .05, **P < .01, and ***P < .001.
Tumor-associated macrophages (TAMS) are the most abundant tumor-infiltrating cells in malignant astrocytoma.22 These TAMs predominantly exhibit a pro-tumorigenic/anti-inflammatory M2-like phenotype shown to promote suppression of innate anti-tumor immunity, tumor invasion, and metastasis.23 To determine if ATRXloss influences M2-like TAM polarization, THP-1 monocytes were differentiated into M0 macrophages using PMA (phorbol 12-myristate 13-acetate) and then co-cultured with either IDH1R132H/p53mut/ATRXloss or IDH1R132H/p53mut/ATRXwt cells. qRT-PCR showed a significant increase in expression of M2-like markers (i.e., CD206, CD163) and a significant decrease in expression of proinflammatory M1-like markers (TNFα, IL6, CD80, HLA-DRA) in response to ATRX loss (Supplementary Figure 4B). To further evaluate the macrophage response to IDH1R132H/p53mut/ATRXloss glioma cells, macrophages were generated using conditioned medium (CM) derived from either IDH1R132H/p53mut/ATRXloss or IDH1R132H/p53mut/ATRXwt tumor cells supplemented with cytokines IL4, IL10, M-CSF for 7 days as described by Benner et al.24 Flow cytometry showed a significant increase in CD206+ macrophages in response to ATRXloss CM (Supplementary Figure 4C). Moreover, flow cytometry showed significant inhibition of T-cell proliferation by macrophages pre-incubated with ATRXloss CM compared to macrophages pre-incubated with ATRXwt CM (Figure 4D). Consistent with these in vitro responses, tumor xenografts derived from IDH1R132H/p53mut/ATRXloss cells contained many more Iba-1+ TAMs and many more cells expressing the M2-like markers CD206+ and Arg1+ than IDH1R132H/p53mut/ATRXwt tumors (Supplementary Figure 4D). These results show that ATRXloss activates pathways that increase the number of tumor-associated immunosuppressive tumor-promoting macrophages.
TCGA-IDH1mut/non-1p19q co-deletion glioma datasets were used to evaluate the relationships between ATRX mutation and immune infiltrates in clinical specimens. Clinical IDH1mut/non-co-deletion/ATRXmut gliomas showed strong trends associating lower expression of CD8A (CD8+ cell marker, P = .05) and higher expression of CD68 (macrophage marker, P = .08) compared with IDHmut/non-co-deletion/ATRXwt gliomas (Figure 4E). Moreover, publicly available TIMER datasets (Tumor Immune Estimation Resource, http://cistrome.org/TIME) show that ATRX-mutant clinical samples contain significantly more macrophages and fewer CD8+ T cells compared with their ATRXwt samples (Figure 4F).
ATRXloss Induces an Immunosuppressive Microenvironment and Growth of Orthotopic Astrocytoma
To further determine if ATRX loss induces a biologically relevant immunosuppressive microenvironment in astrocytoma, murine IDH1R132H/p53mut+/−ATRXloss GL261-LUC glioma cells were established (Figure 5A, Supplementary Figure 5A and B). Immunoblot analysis confirmed IDH1R132H induction by doxycycline with concurrent PD-L1 induction and qRT-PCR analysis demonstrated induction of PD-L1, mIL6, mCXCL8, mIL33 in the ATRXloss GL261 cells compared to ATRXwt controls (Figure 5A and B). Brains from Dox-treated immune-competent C57BL/6j mice bearing orthotopic IDH1R132H/p53mut+/−ATRXloss tumors were harvested and examined for immune cell infiltration and tumor size. ATRX loss tumors were ~3 fold larger than controls (24.9 ± 6.99 vs. 8.3 ± 5.95 mm3) (Figure 5C). Immunofluorescence and IHC staining showed a significant increase in Iba-1+ TAMs and Arg1+ M2 macrophages, FOXP3+ T cells and PD-L1 expression in the IDH1R132H/p53mut/ATRXloss GL261 tumors compared to IDH1R132H/p53mut/ATRXwt tumors (Figure 5D and F). Tumor-infiltrating CD8+ T cells were not detected regardless of ATRX status (not shown) consistent with Kohanbash et al. showing few detectable CD8+ T cells in GL261 tumors expressing transgenic IDH1R132H.8
ATRX loss promotes an immunosuppressive microenvironment in an immune-competent model of orthotopic astrocytoma. A. Immunoblots showing induction of IDHR132H by DOX, loss of ATRX expression by sh-ATRX and PD-L1 induction in murine GL261 astrocytoma cells. B. qRT-PCR analysis measuring expression of six cytokines and chemokines in IDH1R132H/p53mut+/−ATRXloss GL261 cells. C. Representative H&E staining and size quantification of xenografts derived from IDH1R132H/p53mut+/−ATRXloss GL261 cells. D. and E. Immunofluorescence staining and quantification of Iba-1+ (green), M2 marker Arg1+ (red) and PD-L1(red) in orthotopic tumor xenografts derived from IDH1R132H/p53mut+/−ATRXloss GL261 cells. F. Immunohistochemistry staining and quantification of FOXP3+ T cells (brown) in orthotopic tumor xenografts derived from IDH1R132H/p53mut+/−ATRXloss GL261 cells (n = 8 random fields per group). *P < .05, **P < .01, ***P < .001. See online supplementary material for a color version of this figure.
BRD-dependency of Immunosuppressive Transcriptome and Phenotype Induced by ATRX Loss
ATRX loss promotes oncogenesis through global epigenomic remodeling.6,25,26 Therefore, we hypothesize that epigenetic mechanisms have a major role in the immune escape mechanisms of IDH1R132H/p53mut/ATRXloss astrocytoma. To identify potential epigenetic regulators of PD-L1 and cytokine/chemokine induction in IDH1R132H/p53mut/ATRXloss cells, we evaluated a panel of 12 small-molecule inhibitors that target epigenetic regulators in IDH1R132H/p53mut+/−ATRXloss astrocytoma cells. The pan-BET domain inhibitor JQ1 (200 nM) potently inhibited PD-L1 and PD-L2 induction in the IDH1R132H/p53mut/ATRXloss cells (Figure 6A, Supplementary Figure 6A). Inhibition of PD-L1 induction by the BET inhibitors JQ1 and OTX015 was further validated by qRT-PCR and western blotting (Figure 6B and Supplementary Figure 6B). JQ1 also significantly inhibited the induction of several relevant cytokines and chemokines (e.g., CXCL8, IL33, IL6, and CSF2) in IDH1R132H/p53mut/ATRXloss cells (Figure 6C). Pretreating ATRXloss cells with JQ1 under conditions that inhibited the immunosuppressive transcriptome significantly increased their susceptibility to CD8+ T-cell-mediated cell death (Supplementary Figure 6C). qRT-PCR analysis revealed induction of BRD3 and BRD4 in response to ATRX loss (Figure 6D) and western blot analysis further confirmed markedly increased BRD3 protein in the IDH1R132H/ATRXloss cells compared with either IDH1R132H/ATRXwt or IDHwt/ATRXloss cells (Figure 6E). TCGA datasets revealed that BRD3 expression is significantly higher in WHO Grades II and III gliomas compared to Grade IV gliomas, the majority of which are IDH1wt/ATRXwt (Figure 6F, right panel). Moreover, TCGA-IDH1mut/non-1p19q co-deletion glioma dataset analysis shows that BRD3 expression trends higher (P = .12) in clinical ATRXmut glioma samples compared with their ATRXwt counterparts (Figure 6F, left panel, Supplementary Figure 6D). These findings support BRD-dependent mechanisms in the induction of immune-suppressive gene expression in response to ATRXloss.
Pan-BET bromodomain inhibitors attenuate immunosuppressive transcriptome and immunosuppressive responses to IDH1R132H/p53mut/ATRXloss. A. qRT-PCR quantification of PD-L1 expression in IDH1R132H/p53mut/ATRXloss glioma cells treated with the indicated epigenetic inhibitors for 72 h. B. immunoblot analysis of PD-L1 expression in IDH1R132H/p53mut+/−ATRXloss glioma cells treated with JQ1(JQ+) or inactive control (JQ−) for 72 h; C. qRT-PCR quantification of key cytokines/chemokines in IDH1R132H/p53mut/ATRXloss glioma cells treated with JQ1(+/−) for 72 h. D. qRT-PCR quantification of BRDs-1,2,3,4 in IDH1R132H/p53mut+/−ATRXloss glioma cells. E. Immunoblot of BRD3 expression in +/−IDH1R132H/p53mut+/−ATRXloss U373 and SNB19 glioma cells. F. BRD3 expression levels in clinical Grades II, III and IV TCGA glioma datasets (left panel) and in IDHmut/non-1p19q co-deleted ATRX wild-type and ATRX-mutant TCGA glioma datasets (right panel) (http://gliovis.bioinfo.cnio.es/). G. Model of immune escape mechanisms promoted by ATRX loss in IDH1R132H/p53mut astrocytoma. *P < .05, **P < .01, and ***P < .001.
Discussion
ATRX mutation or loss occurs in multiple tumor types including WHO Grades II and III astrocytoma and secondary glioblastoma, consistent with a “driver” role in these malignancies.27 Abundant prior research documents the effects of ATRX inactivation, either alone or in combination with other mutations (e.g., IDH1R132H and p53mut, H3.3K27M expression and p53mut) on DNA damage,28 alternative lengthening of telomeres (ALT),27 copy number alteration (CAN),29 genomic instability30 and tumor growth.16,31 However, whether and how ATRX inactivation influences anti-tumor immune mechanisms in astrocytoma had yet to be demonstrated. A few recent reports have linked IDH1 mutation with reduced infiltration of anti-tumor immune cells (i.e., CD8+ T cells).8,9,32 These studies mainly use IDH1 mutant cells and murine models that did not harbor concomitant mutations in ATRX and as a result do not replicate the mutational profile of most IDH1 mutant astrocytomas that also have ATRX mutation or loss. In this paper, we demonstrate for the first time that ATRX loss cooperates with IDH1R132H and p53mut to promote a BRD-dependent immunosuppressive transcriptome, its link to the immunosuppressive tumor microenvironment and enhancement by chemo/radiation (see Figure 6G).
PD-L1 is expressed in Grades II–IV gliomas. Moreover, the expression levels of PD-L1 are positively correlated with increased tumor grade. Different molecular markers and mechanisms appear to contribute to high levels of PD-L1 in glioma. Loss of phosphatase and tensin homolog (PTEN) and activation of the phosphatidylinositol-3-OH kinase (PI(3)K) pathway have been shown to post-transcriptionally increase PD-L1 expression in glioma.33 Multiple cytokines such as IFN-γ, TNF-a, cell growth factors (e.g., EGF), hypoxia, and exosomes in the tumor microenvironment have been reported to induce PD-L1 expression via signaling pathways involving transcription factors such as Stat1/3, HIF-1, IRF1, and NF-κB.34–36 2-Hydroxyglutarate, the product of the IDH1mut neoenzyme, was found to increase PD-L1 promoter methylation and suppress PD-L1 expression.37,38 Here, we show that the expression of PD-L1 and several relevant immune-suppressing cytokines/chemokines are induced by ATRX loss in IDH1mut/p53mut cells and inhibited by pharmacologic pan-BET inhibitors JQ1 and OTX015 (Figure 6A–C). This implicates a critical role for BRDs in driving the expression of multiple effectors of the immune-suppressive microenvironment in IDH1R132H/p53mut/ATRXloss astrocytoma, a role supported further by the induction of BRD3/4 in ATRXloss cells and increased expression of BRD3 in clinical ATRXmut glioma samples (Figure 6E–F). Identifying the full spectrum of mechanisms responsible for the immune-suppressive transcriptome in IDH1R132H/p53mut/ATRXloss astrocytoma requires further study. For example, unbiased global ChIP-seq and RNA-seq will identify genome-wide BRD-dependent and -independent immune effector regulation in IDH1R132H/ATRXloss astrocytoma. Notably, we found that AGI-5198 and GSK-864, specific inhibitors of mutant IDH1, did not affect PD-L1 expression in IDH1R132H/p53mut/ATRXloss (Supplementary Figure 2D), suggesting that inhibitors of mutant IDH enzymatic activity and checkpoint immunotherapy target distinct mechanisms and might synergize in IDHmut/ATRXloss astrocytoma.
Pro-tumorigenic/anti-inflammatory M2-like macrophages are known to promote glioma cell survival and angiogenesis but molecular mechanisms underlying M2-like TAM polarization in glioma remain incompletely defined. We now show that adding ATRXloss to the IDH1R132H/p53mut background promotes M2-like TAM polarization and up-regulates IL33, IL6, CXCL8, cytokines/chemokines that facilitate M2 macrophage polarization.17,18 These cytokines/chemokines are also highly expressed in gliomas, and modulate glioma cell survival, migration, and invasion.39,40 However, it remains unclear which of these cytokines/chemokines serve as the principal trigger to evoke M2 macrophage polarization. In addition to up-regulating cytokines and chemokines, adding ATRXloss to the IDH1R132H/p53mut background also inhibited expression of Th1-type chemokine CXCL9 that attracts effector CD8+ T cells.41 Consistent with constitutive PD-L1 expression and loss of recruiting chemokine CXCL9 in IDH1R132H/p53mut/ATRXloss cells, TCGA-ATRX-mutant glioma samples show fewer tumor-infiltrating CD8+ T cells relative to ATRX-WT glioma samples (Figure 4E).
A critical finding in this study is the enhancement of immune-suppressive effects of ATRXloss by standard-of-care ionizing radiation (IR) and temozolomide (TMZ). This finding identifies previously unrecognized immune-suppressive effects of radiation and temozolomide on the tumor microenvironment distinct from their well-recognized systemic immune-suppressive actions. It has recently been reported that IDH1R132H expression in the context of p53 and ATRX gene inactivation promotes up-regulation of the ATM/ATR/Chk1 kinase pathway and enhances DNA damage responses.5 The ATM/ATR/Chk1–dependent kinase pathway has also been implicated in PD-L1 induction in response to IR and chemotherapy42 and our preliminary findings show increased p-ATM in the IDH1R132H/p53mut/ATRXloss glioma cells compared with the IDH1R132H/p53mut/ATRXwt cells following either IR and or TMZ treatment (Supplementary Figure 7). Thus, inhibiting the DDR pathway might inhibit PD-L1 induction by IR and/or TMZ in IDH1R132H/p53mut/ATRXloss cells. Understanding how these multiple mechanisms along with BRD inhibitors alter tumor microenvironment responses to chemo-radiation should inform novel therapeutic approaches for enhancing current standard treatments and their combination with immunotherapeutics.
Immunotherapy is effective for a number of cancers, but brain tumors have remained poorly responsive. We describe previously unrecognized immune escape mechanisms and their augmentation by standard-of-care therapy in IDH1R132H/p53mut/ATRXloss astrocytoma cells. These mechanisms are likely to contribute significantly to the ability of WHO Grades II/III astrocytomas to evade antitumor immunity and their inhibition offer promising new approaches for improving the efficacy of emerging immunotherapeutics. Our findings provide a foundation for future pre-clinical and clinical studies designed to rigorously determine if molecular or pharmacologic BRD/BET inhibition alters the IDH1R132H/p53mut/ATRXloss tumor immune microenvironment and enhances anti-tumor immune responses to anti-tumor immune therapies (e.g., checkpoint blockade, CAR-T, vaccines) to provide a survival benefit.
Funding
This work was supported by the National Institutes of Health (NIH) grants NS096754, NS073611, NS110087 (J.L.), and NS091165 (S.X.).
Conflicts of interest statement. The authors disclose no potential conflicts of interest.
Authorship statement. Y.L., and J.L. designed research; C.H., K.W., C.D., L.K., Y.L., Y.F., and B.L. performed experiments; C.H., K.W., Y.L., L.K., T. M., S.X., M.Y., D.C., C.J., and M.L. analyzed experiments, and produced figures; and Y.L. and J.L. wrote the paper. All authors commented on the manuscript. The authors declare no conflict of interest.
