PP2Ac Deficiency Enhances Tumor Immunogenicity by Activating STING–Type I Interferon Signaling in Glioblastoma

Despite ongoing improvements in surgery, radiation, and chemotherapy, GBM has a 5-year survival less than 5% and a median survival less than 15 months (1). More effective therapies are needed. Immune-checkpoint blockade, including anti-PD1/PD-L1 and anti-CTLA4, has been approved for multiple cancers but has not demonstrated clinical efficacy in GBM (2, 3). Low immunogenicity of GBM with minimal MHC-I expression and paucity of T-cell infiltration are major barriers for effective immune-checkpoint blockade (4–7).

PP2A is composed of a catalytic (C), regulatory (B), and scaffolding (A) subunit. PP2A is a major protein phosphatase that accounts for 50% to 70% of the total serine/threonine phosphatase activity in eukaryotic cells to counterbalance the regulatory effects of kinases in modulating signaling pathways that underlie normal physiology and pathobiology of cancer and other diseases (8, 9). We and others have demonstrated that inhibiting PP2Ac in tumor cells can impair DNA damage response (10–12). However, prior studies using immunocompromised models failed to address if PP2Ac inhibition alters antitumor immunity. Our group was the first to report that pharmacologic inhibition of the PP2Ac enhances the efficacy of immune-checkpoint blockade in multiple PD-1–resistant mouse tumor models (13, 14). However, given the ubiquity of PP2Ac expression in many cell types and the involvement of PP2Ac in many cellular pathways, the mechanisms of how PP2Ac regulates antitumor immunity are unclear.

cGAS is a critical sensor for cytosolic double-strand DNA (dsDNA) to activate innate immunity through STING–type I IFN signaling (15, 16). cGAS binding to dsDNA in tumors cells leads to the formation of 2′,3′-cyclic GMP-AMP (cGAMP). cGAMP can activate STING in tumor cells or can be exported as an “immunotrasmitter” to activate STING in dendritic cells (DC) and tumor-associated myeloid cells (17, 18). STING activation induces phosphorylation of interferon regulatory factor 3 (IRF3) to promote type I IFN production and antitumor immune responses (19). Here, using SB28, a glioma model with low tumor mutational burden (TMB), MHC-I expression, and resistance to immune-checkpoint blockade (20), we found that PP2Ac deficiency can unleash a CD8⁺ T cell–dependent antitumor immune response and sensitizes tumor to checkpoint immunotherapy. Mechanistically, glioma-specific PP2Ac deficiency promotes the accumulation of cytosolic dsDNA and cGAMP production, which in turn stimulate IFN signaling in both tumor and immune cells to remodel the immunosuppressive tumor microenvironment. Our results add to accumulating evidence that PP2Ac inhibition represents a promising novel strategy for antitumor immunotherapy. However, this study is the first to establish a role for PP2Ac in regulating cytosolic dsDNA production to mediate STING signaling activation.

Mouse glioma cell line SB28 (RRID: CVCL_A5ED) was kindly provided by Dr. Hideho Okada of University of California, San Francisco (San Francisco, CA). Mouse GL261 (RRID: CVCL_Y003) glioma cell line was kindly provided by Dr. Zhengping Zhuang of the NIH. Human SF8628 (RRID:CVCL_IT46) diffuse intrinsic pontine glioma (DIPG) cell line was purchased from Millipore Sigma. Human medulloblastoma cell line D425 (RRID:CVCL_1275) was provided by Dr. Samuel Cheshier of the University of Utah (Salt Lake City, UT). All cell lines were regularly examined (every 6–12 months) for Mycoplasma contamination (IDEXX, Clongen Laboratories, or Invivogen), last tested March 2023. SB28 and SF8628 are maintained in DMEM with 10% FBS. GL261 cells are maintained in RPMI-1640 with 10% FBS. D425 cells were plated in a neural stem cell expansion medium consisting of Neurobasal−A (Invitrogen), B27 (Invitrogen), human-bFGF (20 ng/mL; Shenandoah Biotech), human-EGF (20 ng/mL; Shenandoah Biotech), human recombinant LIF (Shenandoah Biotech), and heparin (10 ng/mL). All cells were maintained at 37°C under 5% CO₂. Cells were passaged for not longer than 3 weeks from thawing to collection for experiments.

All animal work was approved by the Institutional Animal Care and Use Committee at the University of Texas at Austin and University of California San Francisco. Mice of both sexes, between the ages of 6 and 10 weeks of age, were used for the study. C57BL/6 (RRID:MGI:2159769) were obtained from The Jackson Laboratory. All mice were maintained under pathogen-free conditions.

For orthotopic brain tumor models, 8- to 10-week-old C57BL/6 mice (male and female in equal numbers) were used for intracranial studies. Cell lines were suspended in DMEM for inoculation. Mice were anesthetized with isoflurane, and tumor cells were injected orthotopically in 3 μL. Using a stereotactic frame, a burr hole was formed on the skull via 0.7-mm drill bit 1.5 mm laterally to the right and 1.5 mm rostrally from the lambda, and a noncoring needle (Hamilton 7804-04, 26s gauge) was used to inject the cells at a depth of 3 mm into the brain from the burr hole. Skin incision was sutured. Mice were then monitored daily. Survival endpoint was defined as weight loss greater than 20% relative to baseline, body conditioned score less than 2, or presence of focal neurologic deficits. For radiotherapy, 4 days after implantation, mice head was focally irradiated with an X-ray radiator (MultiRad 350, Precision), with 4 Gy daily for 2 consecutive days (4 × 4 Gy). Lead shield was applied to cover the mice except the cranium to achieve local radiation to the brain.

Cells were trypsinized and washed with FACS buffer (PBS, 2% FBS, 1 mmol/L EDTA). Surface staining was performed by adding the surface antibodies to the cell suspension in 100 μL FACS buffer. After incubating for 30 minutes, cells were washed with FACS buffer and analyzed using the Cytek Aurora cytometer and analyzed using SpectroFlo (Cytek Bioscience) and FlowJo software (RRID:SCR_008520). For intracellular phospho-staining, cells were washed and resuspended in 4% formaldehyde for 15 minutes and then in ice for 10 minutes with chilled 100% methanol. Cells were then stained for phospho-protein.

Bone marrow–derived dendritic cells (BMDC) were derived as previously described (21). Briefly, mouse bone marrow cells were isolated from C57BL/6 mice and cultured in complete RPMI-1640 media with 50 μmol/L of β-mercaptoethanol (BME) and recombinant murine GM-CSF (20 ng/mL; PeproTech) for 8 days. 50K BMDC were plated with 250K of either WT or PP2A KO SB28 OVA cells in a 12-well plate. After 24 hours of incubation, the DCs were collected for flow cytometry analysis.

The Click-iT EdU Alexa Fluor 647 Flow Cytometry Assay Kit (Thermo Scientific, C10424) was used per the manufacturer's instruction. EdU was incubated with cells for 2 hours prior to analysis by flow cytometry. DNA content was counterstained using FxCycle Violet (Thermo Scientific, F10347).

Electrical impedance was measured every 30 minutes over 72 hours and normalized to the starting value (normalized cell index). For the data analysis, the time-points from when the signal is stable (after cell seeding or drug treatment) to 72 hours was used on the xCELLigence RTCA Software Pro (ACEA Biosciences Inc.).

BMDC was differentiated in the same manner as above. Splenocytes from OT-I mouse were harvested and CD8⁺ T cells were isolated using column-based magnetic cell isolation (Miltenyi Biotec). CD8⁺ cells were stained with CellTrace Violet (CTV; Thermo Scientific) before being cultured with BMDC and either WT or PP2A^KO-OVA tumor at a ratio of 5:1:1 (CD8⁺:BMDC:tumor). After 48 hours, suspension cells were collected for flow cytometry analysis.

Splenocytes from OT-I mouse were harvested and CD8⁺ T cells were isolated using column-based magnetic cell isolation (Miltenyi Biotec). A total of 2 × 10⁶ CD8⁺ cells were then plated with IL2 (50 ng/mL) and 25 μL of CD3/CD28 Dynabeads (Gibco). After 3 days, CD8 cells were plated with the WT or PP2A^KO-OVA SB28 tumors at a ratio of 1:1, 1:5, and 1:10 (tumor: CD8). After 1 hour, brefeldin A (Invitrogen) was added and after another 5 more hours, the cells underwent surface staining for CD8⁺ and were then washed and resuspended in IC Fixation Buffer (Thermo Scientific) at room temperature for 20 minutes and then washed with Perm/Wash buffer (Thermo Scientific) before being stained for IFNγ and collected for flow cytometry analysis.

The 50K of SB28 or GL261 cells were plated in a 96-well plate in 100 μL of the medium. After 24 hours, the supernatant was collected. For LB-100 treatment, cells were first treated with a dose titration of LB-100 and 3 hours later supernatant was collected. HMGB in the supernatant was quantified using the Lumit HMGB1 Human/Mouse Immunoassay kit (Promega, W6110) per the manufacturer's instruction.

The PP2A Phosphatase Assay Kit (Millipore) was used according to the manufacturer's instructions. Briefly, using the same amount of starting protein lysate for each condition, PP2A was immunoprecipitated using Anti-PP2A, C subunit (clone 1D6, Millipore) and protein A agarose slurry. The slurry was then washed with TBS before a standard amount of threonine phosphopeptide, a substrate of PP2A, was added to the mixture. Phosphate was released as a product of the reaction. The absolute amount of phosphate released was quantified with a malachite green solution, which was used as a measure of PP2A activity. Experiments were performed in triplicate, and the data are presented as a percentage mean of relative PP2A activity compared with control ± SEM.

Total RNA was extracted using the PureLink RNA Mini Kit (Invitrogen, 12183025) according to the manufacturer's instructions. cDNA synthesis was performed with 0.5 to 1 μg of total RNA using the High-Capacity cDNA Reverse Transcription Kit (Invitrogen, 4374967). mRNA levels were measured with gene-specific primers using the SYBR Green PCR Master Mix (Bio-Rad, 1725120). The results were normalized to GAPDH or OAZ in human or mouse samples, respectively. The primers used are shown in Supplementary Table S1.

For CD8 and CALR visualization in mice tissue, SB28 tumors were collected at survival endpoint. Frozen tissue sections on slides were set at room temperature, then fixed with 4% paraformaldehyde followed by washing with PBS ×2, and then permeabilized with PBST (0.25% TritonX-100) and blocked with 10% normal goat serum/PBST for 30 minutes at room temperature. Samples were then incubated with primary antibodies at 4°C overnight in a humid box. After washing the samples with PBST for 5 × 10 minutes, secondary antibodies in PBST were added to the sample and incubated for 1 hour. 4,6-diamidino-2-phenylindole (DAPI; 300 nmol/L) was used to stain nuclei. Samples were covered with a coverslip with mounting medium and sealed with nail polish and then subjected for imaging using a confocal microscope (Zeiss). CD8⁺ cells were quantified by counting the number of CD8 within each field of view. CALR expression was quantified by the mean fluorescent intensity of the red channel of the region of interest. Analysis was performed on 3 independent samples per group (n = 3) and 3 regions of interest per sample.

For immunoblot analysis, whole-cell lysates were prepared in RIPA lysis buffer (Thermo Scientific, 89900) containing Halt Protease Inhibitor Cocktail (Thermo Scientific, 78429). The protein concentrations were determined by BCA Protein Assay Kits (Pierce, 23227). A total of 20 to 30 μg protein samples were mixed with 4× Laemmli buffer (Bio-Rad, 1610747) and denatured at 95°C for 10 minutes. The sample was separated by SDS-PAGE and transferred to a nitrocellulose membrane (Bio-Rad, 1704270). Membranes were blocked with 3.5% BSA (Fisher, BP1600) and incubated with primary antibodies overnight at 4°C, followed by HRP-conjugated secondary antibodies for 2 hours at room temperature. Signal was detected using Clarity Western ECL (Bio-Rad, 1705061) and captured using the ChemiDoc Imaging System (Bio-Rad). Antibodies used are shown in Supplementary Table S2.

No statistical methods were used to predetermine the sample size. For cell-based experiments, biological triplicates were performed in each single experiment in general, unless otherwise stated. Animal experiments were performed in C57BL/6 mice. Animals were randomized into different groups after tumor cell inoculation; at least 9 to 10 mice were used for each group, unless otherwise indicated. Survival functions were estimated by the Kaplan–Meier methods and compared using the log-rank test or Wilcoxon rank tests as indicated. Two-tailed t tests were used to compare treatment versus control groups. ANOVA models were used to compare continuous outcomes across multiple experimental groups, unless otherwise indicated in each figure legend. Statistical analysis was performed using GraphPad Prism8 software (GraphPad Software, Inc.; RRID:SCR_002798).

Data including total RNA-seq, single-cell RNA-seq are deposited at NCBI GEO GSE213309. Whole-exome sequencing has been deposited to SRA and released under access SUB12057901 (PRJNA973640). Codes, QC results, and supplementary files can be found at Figshare (DOI: 10.6084/m9.figshare.21094390). Publicly available data generated by others were used by the authors including “Merged Cohort of LGG and GBM” (TCGA, Cell 2016, https://www.cbioportal.org/study/summary?id=lgggbm_tcga_pub). All other raw data generated in this study are available upon request from the corresponding author.

We previously showed that pharmacologic inhibition of PP2Ac synergizes with PD1 blockade in the GL261 murine glioma model (13). However, the underlying mechanism or cell type responsible for the PP2Ac-mediated effect on antitumor immunity is unclear. To investigate the effect of PP2Ac deficiency in glioma cells on antitumor immune response, we performed genetic KO of the predominant alpha-isoform of PP2Ac (Ppp2ca) by CRISPR in murine SB28. A clone with confirmed protein-level knockout by Western blot and frameshift mutations in both alleles of Ppp2ca by whole-exome sequencing (WES) was selected for further experiments (Supplementary Fig. S1A and S1B). Compared with GL261, SB28 more consistently recapitulate the immune-silent nature of human glioma with low mutational burden, minimal MHC-I expression, and resistance to immune-checkpoint blockade (20). We obtained the global gene-expression profile of WT and PP2Ac^KO cultured SB28 cells using RNA-seq. Gene ontology (GO) pathway analysis using differentially upregulated genes (log₂FC > 0.5, FDR < 0.01) in PP2Ac^KO SB28 demonstrated that the top enriched biological pathways are related to type I IFN signaling and MHC-I antigen processing/presentation (Fig. 1A). Gene set enrichment analysis (GSEA) confirmed that type I IFN response pathway was enriched in PP2Ac^KO SB28 (Fig. 1B). Additionally, we confirmed the transcription of IFNβ and IFN response-related genes to be highly upregulated in PP2Ac^KO SB28 using RT-PCR (Fig. 1C). Phosphorylation of IRF3, which controls type I IFN transcription, and phosphorylation of STAT1, which is the direct downstream signaling of IFNα/β receptor (IFNAR) were both enhanced in PP2Ac^KO (Fig. 1D; Supplementary Fig. S1C). Given type I IFN has been shown to inhibit glioma proliferation (22), we examined if PP2Ac deficiency will affect tumor proliferation using the xCELLigence platform that provides continuous real-time measurement of cellular viability. PP2Ac^KO SB28 grew at a significantly lower rate than WT (Fig. 1E). We then performed cell-cycle analysis that showed PP2Ac^KO cells have decreased S and increased G₀ and G₂–M phases (Fig. 1F), suggesting that PP2Ac^KO cells have enhanced cell-cycle arrest. We then asked if increased cell death or apoptosis also contributed to decreased viability. We found no significant change in the frequency of an apoptosis marker, Annexin V (Fig. 1G). Given the enhanced IFN signaling, a known marker of immunogenic cell death (ICD), we asked if PP2Ac^KO tumors had an increase in other markers of ICD (23). Calreticulin (CALR) when exposed on plasma surface facilitates phagocytosis by DCs and in turn antigen uptake and presentation (23). CALR expression was significantly enhanced in PP2Ac^KO SB28 (Fig. 1H and I). Another ICD marker, high-mobility group box 1(HMGB1), when secreted, activates innate immune cells via TLR4-MYD88 signaling (23). We quantified HMGB1 level in tumor-conditioned medium and found a marked elevation of HMGB1 secretion by PP2Ac^KO cells (Fig. 1J). Downregulation of MHC-I in cancer cells is a common mechanism of immune evasion (24) and type I IFN is known to positively regulate MHC-I expression (25). Consistently, transcription of genes encoding an essential component of MHC-I (β2m), MHC loading machinery (Erap1, Tap 1, and Tap2), and transcriptional coactivator of MHC-I (NLRC5) was upregulated in PP2Ac^KO SB28 (Fig. 1K). The surface presentation of MHC-I was also consistently enhanced in PP2Ac^KO (Fig. 1L). We confirm the generalizability of this observation by generating CRISPR KO of PP2Ac in GL261 murine glioma, SF8628 human DIPG, and D425 human medulloblastoma cell lines, all of which demonstrated increased MHC-I expression (Fig. 1M; Supplementary Fig. S1D). Next, we asked if the observed increase in MHC-I expression in PP2Ac^KO cells is dependent on enhanced IFN signaling. Treatment of PP2Ac^KO SB28 with anti-IFNAR, which antagonizes type I IFN receptor, IFNAR (Fig. 1N) or ruxolitnib, a janus kinase (JAK) inhibitor (Supplementary Fig. S1E) that blocks JAK/STAT1 activity downstream of IFNAR, reduced MHC-I expression to level close to WT. In addition, the enhanced expression of IFN-related genes in PP2Ac-deficient tumors was also reversed by anti-IFNAR treatment (Fig. 1O). To confirm these mechanistic findings are not limited to SB28, we showed that PP2A deficiency in GL261 similarly resulted in decreased cellular proliferation (Supplementary Fig. S2A), increased pIRF3 and pSTAT1 (Supplementary Fig. S2B), enhanced IFN-stimulated genes (Supplementary Fig. S2C) that is dependent on IFNAR signaling (Supplementary Fig. S2D).

LB-100, a small-molecule inhibitor of PP2Ac, is currently under clinical investigation (NCT04560972 and NCT03027388). We have previously demonstrated in the GL261 orthotopic model that LB-100 synergized with anti–PD1 blockade (13). However, the ubiquity of PP2Ac expression precluded the understanding of what cell type(s) is responsible for the immunogenic effect of LB-100. Our data suggest that tumor-intrinsic PP2Ac deficiency is sufficient to promote IFN signaling in vitro. To show the translation relevance of this effect, we asked if LB-100 can reproduce the phenotypes observed in PP2Ac^KO glioma cells. First, we confirmed the dose-dependent inhibition of PP2Ac activity in WT SB28 by LB-100 relative to PP2Ac^KO cells (Fig. 2A). The observed “residual” PP2Ac activity in PP2Ac^KO SB28 is likely attributed to the persistence of the beta isoform of the PP2Ac (Ppp2cb) that is not distinguished by the immunoprecipitation technique used for the assay. Using the xCELLigence platform, we found that WT SB28 was sensitive to LB-100 treatment, resulting in a dose-dependent decrease in cell viability (Fig. 2B and C) within several hours of drug treatment. We also found that LB-100 induced ICD in a dose-dependent fashion, with increased expression of CALR (Fig. 2D) and secretion of HMGB (Fig. 2E). Similarly, the phosphorylation of STAT1 (Fig. 2F), transcription of IFN-stimulated genes (Fig. 2G), and surface expression of MHC-I (Fig. 2H) were all enhanced by LB-100. We confirmed these are similarly observed in GL261 cells (Supplementary Fig. S2D–S2G).

Type I IFN is known to promote DC activation (26, 27). Given the increased expression of IFN-stimulated genes in PP2Ac^KO glioma cells, we asked if they are more efficient in activating DC. We generated WT and PP2Ac^KO SB28 cells expressing ovalbumin (OVA) and cocultured WT or PP2c^KO SB28 cells with BMDCs. BMDCs cocultured with PP2Ac^KO SB28 compared with WT SB28 have increased expression of activation markers, including MHC-II, MHC-I, and CD86 (Fig. 3A). Moreover, they also had increased expression of OVA-H2Kb (SIINFEKL; Fig. 3A), which is an OVA peptide fragment presented on MHC-I of DC after phagocytosis of the tumor. Expression of SIINFEKL on DC is an indication of cross-presentation, suggesting that tumor PP2Ac deficiency can promote DC activation of tumor-associated antigen (TAA)–specific CD8⁺ T cells. Expression of activation markers was similarly enhanced in BMDC cocultured with non-OVA expressing SB28 or GL261 glioma cells (Supplementary Fig. S3). Next, we asked if this enhanced activation of DC is cell-contact dependent by performing coculture in a transwell insert. We similarly found an increase in the activation of DC cells (Fig. 3B), suggesting soluble factors secreted by tumor cells contributed to the promotion of DC activation.

Next, we investigated whether enhanced MHC-I expression in PP2Ac-deficient glioma affects tumor killing by TAA-specific CD8⁺ T cells. We first confirmed that PP2Ac^KO SB28 cells have increased expression of OVA-H2Kb (SIINFEKL; Fig. 3C). Then, we tested if PP2Ac deficiency-mediated enhanced MHC-I expression is functionally relevant for TAA-specific CD8⁺ T-cell cytotoxicity by incubating WT or PP2Ac^KO SB28-OVA cells with OT-I T cells preactivated by CD3/CD28 beads (Fig. 3D). We found a significant increase in OT-I T cells mediated killing against PP2Ac^KO SB28-OVA cells (Fig. 3E) across a range of effector:target (E:T) ratios. The percentage of IFNγ producing OT-I T cells was also significantly enhanced when OT-I T cells were cocultured with PP2Ac^KO SB28 (Fig. 3F), suggesting that PP2Ac^KO SB28 cells were more susceptible to TAA-specific T-cell cytotoxicity. Given the enhanced activation of BMDC primed by PP2Ac-deficient tumors (25, 26), we tested the ability of tumor-primed BMDC to activate naïve CD8⁺ OT-I T cells by coculturing BMDC, WT or PP2Ac^KOSB28-OVA tumors and naïve CD8⁺ OT-I T cells labeled with CTV dye. BMDC primed by PP2Ac^KO tumor significantly enhanced OT-I CD8⁺ cell proliferation (Fig. 3G). Collectively, these data show that PP2Ac deficiency in glioma cells enhances TAA presentation on MHC-I, promotes DC maturation and cross-presentation of TAA-specific antigen to activate naïve CD8⁺ T cells, and increases susceptibility to TAA-specific T cell–mediated killing.

We next examined the effect of PP2Ac deficiency in glioma cells in vivo by orthotopically implanting WT or PP2Ac^KO tumors in C57BL/6 mice. Survival was significantly prolonged in mice bearing PP2Ac^KO compared with WT SB28 in both SB28 and GL261 (Fig. 4A and B). We first examined the degree of CD8⁺ T-cell infiltration in the tumor microenvironment. At the survival endpoint, SB28 tumors were stained for CD8 T cells by immunofluorescence. We found a significant increase in tumor-infiltrating CD8⁺ T cells in PP2Ac^KO tumors (Fig. 4C). Given the increased abundance of tumor-infiltrating CD8⁺ T cells, we asked if they are required or contributed to the prolonged survival of PP2Ac^KO glioma bearing mice. We performed systemic CD8⁺ T-cell depletion by treating mice with isotype or anti-CD8 depleting antibody prior to tumor implantation and throughout the study. CD8 depletion significantly diminished, but did not completely abolish, the survival benefit of PP2Ac deficiency, suggesting that CD8-mediated adaptive immune response contributed to but is not completely responsible for the beneficial effects of PP2Ac deficiency (Fig. 4D). As we have previously shown that tumor-intrinsic type I IFN signaling promoted MHC-I expression and other IFN-stimulated genes, we tested if PP2Ac deficiency-mediated enhanced survival is dependent on IFN signaling by systemic blockade of IFNAR. Abrogation of type I IFN signaling completely reversed the survival benefit of PP2Ac deficiency (Fig. 4E). Given we observed a significant increase in ICD in vitro (Fig. 1H–J), we asked if this is similarly observed in vivo. Tumors harvested from mice in Fig. 4E at the survival endpoint were stained for CALR by immunofluorescence. We found a significant increase in CALR expression in PP2Ac^KO tumors (Fig. 4F), which is abolished in mice treated with anti-IFNAR (Fig. 4F). This suggests that PP2Ac deficiency promoted ICD in a type I IFN-dependent manner. Next, we asked if PP2Ac deficiency sensitizes mice to checkpoint blockade. We treated tumor-bearing mice with isotype or combined anti-CTLA4/PD1 blockade. Although dual checkpoint blockade marginally increased the median survival of mice bearing WT SB28, it dramatically prolonged survival of mice bearing PP2Ac^KO SB28. In addition, the rate of mice with complete response (CR) was strikingly increased from 10% to 80% (Fig. 4G). We then asked if CR mice from PP2Ac^KO tumors carry immunologic memory against WT and PP2Ac^KO SB28 tumors by rechallenging CR mice with WT and PP2Ac^KO SB28 tumors. We simultaneously implanted WT and PP2AcKO SB28 cells in the left and right flank, respectively (Fig. 4H). We found that CR mice had protective immunity against PP2Ac^KO SB28 and partial immunity against WT SB28. PP2Ac^KO SB28 was unable to grow in cured mice, whereas WT SB28 grew to smaller sizes in CR mice than in naïve mice (Fig. 4H; Supplementary Fig. S4). These data indicate that PP2Ac-deficient glioma cured by checkpoint blockade can elicit adaptive memory antitumor immunity. However, the fact that memory immunity against WT tumors was incomplete suggests that WT tumors can evade immunosurveillance despite shared neoantigens with PP2Ac^KO tumors.

PP2A is known to play a role in modulating DNA damage repair (DDR; refs. 11, 12, 28, 29). A recent study demonstrated that inactivation of a PP2A scaffolding subunit promotes microsatellite instability in colon cancer (30). We asked if increased antigenicity by enhanced neoantigen production contributed to increased immunogenicity in PP2Ac^KO SB28 by analyzing the mutational landscape of WT and PP2Ac^KO SB28 using WES. We identified a significant increase in TMB attributed to new missense mutations in PP2Ac^KO cells (Fig. 5A). Frequency of frameshift and inframe deletion was unchanged (Supplementary Fig. S5A). In addition, we ran algorithms to predict peptide binding to MHC-I or MHC-II to estimate potential neoepitope formation and found a significant increase in neoepitopes for MHC-I and MHC-II in PP2Ac^KO SB28 (Fig. 5B; Supplementary Fig. S5B). These data suggest that PP2Ac deficiency, likely through its impact on DDR, promotes TMB and in turn neoantigen formation to enhance tumor antigenicity. We confirmed that PP2Ac-deficient glioma cells harbor increased unrepaired dsDNA in the nucleus by quantifying phosphorylated γH2AX (H2AX-P) in SB28 (Fig. 5C and D) and GL261 cells (Supplementary Fig. S6A and S6B). However, our in vitro data using the OVA-OTI model showed that PP2Ac deficiency could enhance DC priming and OVA-specific T-cell cytotoxicity, suggesting that tumor immunogenicity is enhanced not solely by the promotion of neoantigen formation. Our group and others have previously shown that PP2Ac is critical in regulating G₂–M transition, and PP2A inhibition in combination with DNA-damaging treatment (such as chemotherapy or radiation) can promote aberrant entry into mitosis, resulting in cell death by mitotic catastrophe (10–12, 28, 31, 32). It has also been shown that mitotic progression following DNA damage can promote micronuclei formation to induce cGAS–STING–IFN activation (33). We, therefore, hypothesize that PP2Ac deficiency can promote activation of cGAS–STING–type I IFN by promoting the accumulation of cytosolic dsDNA. We measured the level of dsDNA in the cytoplasmic fraction of WT and PP2Ac^KO SB28, GL261, and D425 cells and found a significant increase in dsDNA accumulation in PP2Ac^KO tumor cells (Fig. 5E). We confirmed this using immunofluorescence staining of dsDNA in SB28 (Fig. 5F) and GL261 (Supplementary Fig. S6C). Next, we asked if enhanced cytosolic accumulation of dsDNA translates into increased production of cGAMP, which can be exported by tumor cells as “immunotransmitter” and be taken up by host antigen-presenting cells, such as macrophages and DCs, to activate STING to produce type I IFN (17, 18). We measured cGAMP in whole-cell lysate of WT and PP2Ac^KO SB28, GL261, and D425 cells and found cGAMP production to be enhanced in PP2Ac-deficient tumors (Fig. 5G). This is consistent with our coculture experiment that showed PP2Ac deficiency could promote DC activation in a contact-independent fashion (Fig. 3B), as cGAMP can be exported by tumor to activate immune cells. To determine if the increase in IFN signaling in PP2Ac^KO tumors is dependent on cGAS, we generated PP2Ac/cGAS double KO (dKO) SB28 cells. We found the promotion of IFN signaling in PP2Ac^KO SB28 as assessed by protein expression of pIRF3 and pSTAT1 was markedly reduced in dKO SB28 compared with PP2A^KO SB28 cells (Fig. 5H). We also found the host of overexpressed IFN-simulated genes in PP2Ac^KO to be significantly decreased in dKO SB28 (Fig. 5I). These data suggest that PP2Ac deficiency promotes type I IFN signaling in a cGAS-dependent fashion.

Previous studies demonstrated that radiation can promote antitumor immunity through activation of cGAS–STING signaling by enhancing dsDNA production (34). We therefore asked if PP2Ac deficiency can promote radiation induced dsDNA accumulation and further enhance radiation-induced antitumor immunity. We radiated WT or PP2Ac^KO SB28 with increasing dosage (0–20 Gy) and found that radiation complemented the effect of PP2Ac deficiency in promoting dsDNA production (Fig. 5J). To test this effect in vivo, we orthotopically implanted WT or PP2Ac^KO SB28 and administered focal brain radiation on post-implant day 4 (4 Gy on days 4 and 5 for a total dose of 8 Gy). PP2Ac^KO SB28 tumors were significantly more sensitive to radiation, achieving a 50% durable response compared with 0% in WT (Fig. 5K). These data suggest that PP2Ac deficiency in tumor cells can synergize with radiotherapy.

To explore how glioma PP2Ac deficiency reshapes the tumor microenvironment in vivo, we performed scRNA-seq of WT and PP2Ac^KO SB28 murine glioma after orthotopic implantation. Tumors were harvested on post-implant day 18 and disaggregated. Tissues from 3 mice were pooled for each condition and the whole-cell content was analyzed by scRNA-seq. Unsupervised clustering and uniform manifold approximation and projection (UMAP) analyses were performed on 15,448 cells (Fig. 6A). Immune cells were segregated by the expression of CD45⁺ and tumor cells by expression of GFP. We utilized the Seurat package to perform fine clustering (Supplementary Fig. S7 and Supplementary Data S1). PP2Ac deficiency in glioma cells distinctly remodeled both the tumor and immune compartments (Fig. 6B). We first focused on the effect of glioma PP2Ac deficiency on immune cells. We utilized gene expression of canonical markers to classify CD8⁺ T cells (CD3⁺ CD8b1); CD4⁺ T cells (CD3⁺ CD4⁺); natural killer (NK) cells (NKG7⁺CD3⁻); macrophages (TAM; CD11b⁺CD11c⁻); DCs (CD11b⁺CD11c⁺; Supplementary Fig. S8A). The DC population was further subclassified into conventional type 1 DC cDC1 (CD11c⁺, CD103⁺), which is critical for antigen cross-presentation. We found the frequency of CD8 T cells, NK cells, and cDC1 cells increased 3.8- to 5.5-fold in PP2Ac^KO tumors (Fig. 6C; Supplementary Fig. S8B). The infiltration of T cells, NK and cDC1s in human cancers has each been associated with improved prognosis and response to immunotherapy (35–37). Consistent with PP2Ac-deficient tumors being more sensitive to checkpoint blockade, we found an increase in expression of PD1 in T cells and PD-L1 in tumor and myeloid cells of PP2Ac^KO tumors (Supplementary Fig. S8C; ref. 38).

Next, we examined the impact of glioma PP2Ac deficiency on myeloid cells, which are the predominant immune cells. We identified the differentially expressed genes (DEG; −log₁₀(adjusted P) > 20, log₂(FC) > 0.5 or log₂(FC) < −0.5) in myeloid cell clusters (TAMs + DC: clusters 0, 11, 12, 13, 14, 16) between WT and PP2Ac^KO tumors (Fig. 6D). Consistent with our in vitro finding of PP2Ac deficiency enhancing tumor cGAMP production, which can activate STING-IFN signaling in myeloid cells, we found the top upregulated DEGs to be known IFN response genes (ISG15, IRF7, STAT1, CXCL19, RSAD2, and Mx1). In addition, components of MHC-I (B2m, H2-Q1, H2-Q7, and H2-Q6), MHC-I loading machinery (TAP1 and TAPBP), MHC-II (H2-DMa and H2-Aa), and CD74—an invariant MHC-II chain implicated in trafficking of MHC-II proteins—were all highly enriched (Fig. 6D). An unbiased GO pathway analysis demonstrated the top enriched biological pathways of myeloid DEGs in PP2Ac^KO tumors were related to type I IFN signaling and MHC-II antigen processing/presentation (Fig. 6E and F). On the other hand, among downregulated DEGS in myeloid cells of PP2Ac^KO tumors were Mrc1(CD206) and Arg1 (Fig. 6D and F), markers frequently associated with anti-inflammatory, immunosuppressive macrophages (39–41). Next, we further characterized the heterogenous TAMs populations. Seurat analysis identified 5 TAMs clusters (0, 11, 13, 14, 16). We examined the expression of known myeloid markers among these clusters (Fig. 6G and H). Cluster 11 expressed markers of classic bone marrow–derived monocytes (Chil3⁺), whereas cluster 0 had low Chil3 expression and high expression of mature macrophage markers such as C1Q. Clusters 13 and 14 expressed TMEM119, indicating they are likely resident microglia. Cluster 16 exclusively expressed S100A9, which is a marker of immunosuppressive monocytic myeloid-derived suppressor cells (M-MDSC)-derived macrophages (42). We calculated the fold change in cluster frequencies between WT and PP2Ac^KO tumors and found that clusters 0 and 11 were enriched in PP2Ac^KO tumors, whereas clusters 13, 14, and 16 were enriched in PP2Ac^WT tumors (Fig. 6I). To delineate the functional significance of these clusters, we performed GSEA using cluster-defining gene lists (Supplementary Data S1) and compared the relative expression of selected pathways across the five clusters (Fig. 6J). Clusters 0 and 11, which were enriched in PP2Ac^KO tumors, had increased type I/II IFN signature in addition to NFkB activation. This is consistent with increased STING activation as STING is known to stimulate both IFN and NFkB downstream signaling (43). They also expressed high antigen processing/presentation activity. Clusters 13, 14, 16, which were enriched in WT tumors, all had suppressed IFN signature. Cluster 14 was notable for a distinctly high oxidative phosphorylation signature, which is associated with immunosuppressive “M2”-like macrophages (44). S100A9-positive M-MDSC–derived macrophages (cluster 16) are associated with shorter survival in patients with head and neck cancer and poor response to PD-1 antibody treatment in patients with metastatic melanoma (42). In summary, our data demonstrated that PP2Ac deficiency in glioma cells significantly remodels the immune tumor microenvironment. There was a significant increase in IFN-activated macrophages with enhanced markers of activation and antigen presentation. There was also a decrease in immunosuppressive macrophages. cDC1 DC infiltration was also enhanced, which presumably led to an influx of cytotoxic CD8 T cells and NK cells. These observations are consistent with PP2Ac deficiency-mediated dsDNA accumulation and cGAMP production that can activate STING signaling in antigen-presenting cells (TAMs and DC).

Finally, we examined the impact of glioma PP2Ac deficiency on GFP⁺ SB28 cells. Seurat analysis identified 9 clusters (1, 2, 3, 4, 5, 6, 7, 8, 17). We analyzed the DEGs [−log₁₀(adjusted P) > 20, log₂(FC) > 0.5 or log₂(FC) < −0.5] between WT and PP2Ac^KO tumors and found the top overexpressed genes in PP2Ac^KO glioma cells were IFN response elements (ISG15, STAT1, CXCL10, and IRF7) including components of MHC-I (B2m and H2-D1) and MHC-I loading machinery (TAP1 and TAPBP; Fig. 7A). GO pathway analysis using upregulated DEGs in glioma cells of PP2Ac^KO tumor demonstrated that the top enriched biological pathways were predominately related to type I IFN signaling (Fig. 7B and C). We calculated fold change in cluster frequencies between WT and PP2Ac^KO tumors and found that clusters 17 and 1 were enriched in PP2Ac^KO, whereas clusters 7, 8, and 5 were enriched in WT tumors (Fig. 7D). GSEA was performed to compare the relative expression of gene signatures of selected pathway across the 5 clusters (Fig. 7E). Cluster 17, which was markedly enriched in PP2Ac^KO tumors (by >30-fold), was defined by high type I/II IFN signature. Clusters 5, 7, and 8, which were enriched in WT tumors, had enhanced signature of oxidative phosphorylation. In addition, cluster 8 had increased expression stem cell makers, whereas clusters 5 and 7 had increased MYC target expression. To establish if the downregulated DEG (log₂FC < −0.5, FDR < 0.01) signature in GFP⁺ tumor clusters (1, 2, 3, 4, 5, 6, 7, 8, 17) in PP2Ac^KO tumors (PP2ADN; Supplementary Table S3) has clinical significance in human samples, we obtained bulk RNA-seq data set from The Cancer Genome Atlas (TCGA) with combined low-grade glioma (LGG) and glioblastoma (GBM). We calculated the average normalized expression of PP2ADN genes and found its expression to be significantly higher in GBM compared with LGG (Fig. 7F). Survival is also worse for patients with higher PP2ADN expressions in the combined GBM/LGG cohort (Fig. 7G). When grading is stratified, PP2ADN expressions continue to significantly predict worse survival in LGG patients (Supplementary Fig. S9A), whereas there is a trend toward worse survival in GBM patients (P = 0.08) with limited sample size (Supplementary Fig. S9B). In summary, the scRNA-seq data demonstrated that glioma PP2Ac deficiency resulted in a marked increase in IFN activation and MHC-I expression in an immune-silent glioma model. In addition, the gene signature downregulated in PP2Ac-deficient glioma cells correlates with higher grading and worse prognosis in human clinical data. Therefore, this study strongly argues for PP2A inhibition as an appealing strategy for novel treatment of GBM.

Our group previously reported that LB-100, a small-molecule inhibitor of PP2Ac currently in phase Ib and II clinical trials (NCT03027388 and NCT04560972), can synergize with anti-PD1 checkpoint blockade in preclinical models of non-CNS (14) and CNS tumors (13). However, as PP2Ac is ubiquitously expressed in many cell types, from tumor to immune cells, our previous work using LB-100 failed to delineate what cell type is responsible for the effect of enhanced tumor immunogenicity. In addition, LB-100 also has an antagonistic effect against another phosphatase family, PPP5C, thereby raising the question of whether PP2Ac is the target responsible for promoting tumor immunogenicity (45). In this study, we found that tumor-specific PP2Ac deficiency can promote tumor immunogenicity and sensitize checkpoint blockade in vivo using a nonimmunogenic glioma model SB28. Mechanistically, we found that PP2Ac deficiency promotes ICD, dsDNA accumulation, cGAMP production, and TMB. scRNA-seq demonstrated that glioma-specific PP2Ac^KO can remodel the immune microenvironment by enhancing the infiltration of IFN-activated TAMs. Tumor infiltration of cross-presenting cDC1 cells, NK cells, and cytotoxic CD8 T cells are all concomitantly increased. These data suggest that the downregulation of PP2Ac in tumor cells can promote antitumor immunity through enhanced cGAS–STING–IFN signaling.

PP2Ac is known to be required in various steps of DNA damage response (46). A recent study showed that the inactivation of a scaffolding subunit of PP2A, PPP2r1a, can induce microsatellite instability by increasing phosphorylation of retinoblastoma protein and histone deacetylase 2, which trigger neoantigen production and in turn promote immunogenicity of tumors in non-CNS cancer models (30). Consistently, we found an increase in TMB and predicted neoepitope bound to MHC-I and MHC-II in PP2Ac^KO glioma. However, we demonstrate in this study that PP2Ac inactivation can enhance cGAS–STING–IFN signaling in both tumor and immune cells, providing a complementary mechanistic rationale to inhibit PP2Ac to enhance antitumor immunity. The connection between tumor PP2Ac and cGAS–STING–IFN stimulation has not been reported and is the principal novelty and significance of the current study. The precise mechanism of how PP2Ac deficiency leads to the accumulation of cytosolic dsDNA remains to be elucidated. In macrophages, PP2A has also been implicated in the dephosphorylation of IRF3 as a deactivation mechanism that contributes to the termination of IRF3-type I IFN signaling in response to Toll-like receptor and STING (47, 48). Whether this is a relevant regulatory mechanism in tumor cells, specifically in response to cGAS-STING activation, has not been explored. Therefore, beyond the increase in cytosolic dsDNA, we cannot rule out the possibility that PP2Ac deficiency can contribute to enhanced IFN signaling through downstream regulation of the STING–IFN pathway.

Our in vitro data demonstrated that PP2Ac deficiency in tumor cells promoted MHC-I expression in an IFNAR and JAK–STAT-dependent manner, suggesting that KO of PP2Ac promotes autocrine activation of tumor IFN signaling. Systemic IFNAR blockade in vivo abolished the survival benefit of tumor PP2Ac deficiency. Although activation of the IFNAR-JAK/STAT pathway induces expression of interferon-stimulated genes and acutely promotes the efficacy of cancer immunotherapies (25–27), it is also recognized that chronic tumor IFNAR-JAK/STAT stimulation could be counterproductive (49, 50). The role of type I IFN in the GBM microenvironment is controversial as both beneficial (51, 52) and detrimental effects (53, 54) of IFN stimulation had been suggested. A recent study in melanoma showed that the impact of type I IFN signaling on antitumor immunity depends on the balance between immune and tumor type I IFN stimulation (49). Type I IFN signaling in immune cells is beneficial, whereas tumor type I IFN stimulation can have opposing effects. On one hand, it carries the benefit of promoting MHC-I expression, on the other, it induces “adaptive resistance” mechanisms such as upregulation of PD-L1 or IDO that can be immunosuppressive (50). Our scRNA-seq data suggest that in PP2Ac-deficient tumor, IFN signaling is enhanced in both immune and tumor cells. It is possible that in the context of a highly immune-silent tumor such as glioma with low baseline MHC-I expression, the benefits of IFN-induced MHC-I expression outweigh the detriment of adaptive resistance.

Not only is PP2Ac expressed in many cell types, but it is also implicated in many signaling pathways. The diverse function of PP2A holoenzyme is dictated by the heterogenous regulatory B subunits. Therefore, there is reservation in systemically inhibiting the conserved PP2Ac subunit for fear of eliciting unacceptable side effects or antagonizing specific PP2A complexes that function as tumor suppressors (55). In fact, some groups advocate for the development of allosteric PP2A activators of specific PP2A holoenzyme (56) involved in positively regulating tumor-suppressor pathways. Our study is significant in providing the rationale that, at least in the context of glioma, even though inhibition of PP2Ac in tumor cells may perturb various pathways, on balance it results in a favorable therapeutic outcome of increased immunogenicity and sensitivity to checkpoint blockade. It is also possible that PP2Ac deficiency can have a nonimmune mediated, tumor-intrinsic effect on glioma cells. Our in vivo CD8 depletion studies demonstrated a residual benefit of PP2Ac deficiency with CD8 depletion in the absence of checkpoint blockade, which can be attributed to other immune cells (NK, CD4, and DC) or a direct effect on tumor growth. Our scRNA-seq data showed that the downregulated gene signature in PP2Ac-deficient glioma cells is correlated with higher histologic grade of glioma and predicts worse survival in patients, which suggests that PP2Ac deficiency has the potential to directly affect tumor growth independent of its effect on immunogenicity. The translational significance of this study is the implication that if systemic pharmacologic inhibition (i.e., with LB-100) cannot achieve sufficient PP2A inhibition in glioma, either due to the blood–brain barrier or systemic intolerance, local delivery of LB-100, such as via convection-enhanced delivery can be a viable alternative strategy.

I. Mondal reports grants from NIH and Department of Defense (DOD) during the conduct of the study. O. Das reports grants from NIH and the DOD during the conduct of the study. Z. Meng reports grants from NIH/NIGMS, the Elsa U. Pardee Foundation, and NIH/NINDS during the conduct of the study and from the Stanley J. Glaser Foundation outside the submitted work. R.O. Lu reports grants from the NIH and DOD during the conduct of the study; in addition, R.O. Lu has a patent for oxabicycloheptanes for modulation of immune response issued and with royalties paid. W.S. Ho reports grants from NIH/NINDS, Rally Foundation, and Matthew Larson Pediatric Brain Tumor Foundation during the conduct of the study; in addition, W.S. Ho has a patent for oxabicycloheptanes for modulation of immune response issued and with royalties paid. No disclosures were reported by the other authors.

I. Mondal: Data curation, formal analysis, writing–review and editing. O. Das: Data curation, formal analysis, writing–review and editing. R. Sun: Data curation. J. Gao: Data curation. B. Yu: Formal analysis. A. Diaz: Supervision. J. Behnan: Data curation. A. Dubey: Data curation. Z. Meng: Writing–review and editing. E. Eskandar: Writing–review and editing. B. Xu: Software, formal analysis, supervision, visualization, methodology, writing–review and editing. R.O. Lu: Conceptualization, data curation, formal analysis, supervision, funding acquisition, methodology, writing–original draft, writing–review and editing. W.S. Ho: Conceptualization, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing.

This study was supported by Department of Defense PRCRP W81XWH-20-1-0428 (R.O. Lu), NIH R01NS126501 (R.O. Lu), NIH R01NS131545 (W.S. Ho), and NIH R35GM142504 (Z. Meng). RNA-seq and scRNA-seq were performed by the Genomic Sequencing and Analysis Facility at UT Austin, Center for Biomedical Research Support. RRID#: SCR_021713.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.