MTAP loss correlates with an immunosuppressive profile in GBM and its substrate MTA stimulates alternative macrophage polarization

Glioblastoma (GBM) is a lethal brain cancer known for its potent immunosuppressive effects. Loss of Methylthioadenosine Phosphorylase (MTAP) expression, via gene deletion or epigenetic silencing, is one of the most common alterations in GBM. Here we show that MTAP loss in GBM cells is correlated with differential expression of immune regulatory genes. In silico analysis of gene expression profiles in GBM samples revealed that low MTAP expression is correlated with an increased proportion of M2 macrophages. Using in vitro macrophage models, we found that methylthioadenosine (MTA), the metabolite that accumulates as a result of MTAP loss in GBM cells, promotes the immunosuppressive alternative activation (M2) of macrophages. We show that this effect of MTA on macrophages is independent of IL4/IL3 signaling, is mediated by the adenosine A2B receptor, and can be pharmacologically reversed. This study suggests that MTAP loss in GBM cells may contribute to the immunosuppressive tumor microenvironment, and that MTAP status should be considered for characterizing GBM immune states and devising immunotherapy-based approaches for treating MTAP-null GBM.

disease-free survival in GBM patients 21 . MTAP is a metabolic enzyme that functions in the salvage pathway of adenine and methionine, and loss of MTAP results in the accumulation of its direct substrate, methylthioadenosine (MTA), which can be released into the tumor environment [22][23][24][25] . Aberrantly accumulated MTA is known to be functionally active within cells as an inhibitor of methyltransferases 23,24 , including interfering with intracellular protein methylation in T cells 26 , as well as acting on adenosine receptors on the cell surface of melanoma cell lines 27 . Studies of pathogen-induced host inflammatory responses have linked MTA to downregulation of TNFα production by macrophages through engaging adenosine receptors 28 , and have revealed a role of MTA in controlling the host inflammatory response 29 , such that MTA has been used as an immunosuppressive drug for treating colitis, liver inflammation, brain inflammation and autoimmunity in animal models [30][31][32] .
In this study, we investigated the link between MTAP loss in GBM cells and the GBM microenvironment. We reveal that in GBM tissues, low expression of MTAP (due to deletion or epigenetic silencing) is associated with immune cell populations indicative of an immunosuppressive state. We provide evidence that MTAP lossinduced MTA accumulation stimulates M2 alternative macrophage activation. We illustrate that this effect of MTA on macrophages is independent of interleukin (IL)-4/IL-13 signaling, is mediated by the adenosine A 2B receptor, requires STAT3, and is distinct from the actions of adenosine. Finally, we show that the MTA-induced alternative macrophage activation can be pharmacologically reversed. These results provide a basis for pursuing adenosine receptor signaling as a component of tumor-mediated immunosuppression and potential target in MTAP-null GBM.

Results
Loss of MTAP expression is associated with an immunosuppressive molecular profile in GBM. To explore the association of MTAP expression with the immune cell profile in GBM we employed CIBERSORT, a method for deconvoluting gene expression data and estimating immune cell fractions in human cancers [33][34][35] . This method has been successfully used for analyzing a large number of samples across multiple cancer types 34 , is readily compatible with the microarray-based platform of the largest available GBM expression dataset, and gives unique consideration to data normalization and noise control, which is essential when analyzing heterogeneous and diffusive tumors, such as GBM 36 . We applied this analysis to the GBM dataset from TCGA (385 GBM patient samples) 11 to examine the link between MTAP loss and the GBM immune microenvironment. As MTAP alteration can occur through homozygous gene deletion or epigenetic silencing, we compared the two groups of patients with the highest and lowest MTAP expression (upper and lower quartiles, n = 96 each) ( Supplementary Fig. S1). This revealed reduced fractions of activated CD4 T cells and increased resting CD4 T cells in tumors with low MTAP expression (Fig. 1A, Supplementary Fig. S1). Additionally, two types of innate immune cells were identified as differentially represented between these two groups of tumors. There was a lower fraction of γδT cells (Fig. 1B), and a significantly higher fraction of M2 macrophages in the low-MTAP expressing tumors (Fig. 1C).
To corroborate the findings from CIBERSORT analysis, we utilized a second recently developed platform for identifying individual immune cell types, which is based on a dataset of 366 microarrays compiled from multiple independent studies 37 . This data set was previously used to generate a list of 782 genes that are specific to immune cell subpopulations (i.e. not expressed in tumor cells or normal tissue) 37 . We analyzed this list of genes in the TCGA GBM dataset and found that among the 641 of these genes for which expression data was available, 239 of them were differentially expressed between the MTAP-low and MTAP-high GBM populations (upper quartile vs lower quartile, n = 96 each group, Benjamini-Hochberg corrected p-value < 0.0185). Hierarchical clustering revealed two groups of genes that could be evaluated to compare immune cell types upregulated and downregulated in MTAP deficient GBM (104 genes upregulated in MTAP-low tumors, 135 genes downregulated in MTAP-low tumors) ( Fig. 1D; Supplementary Table 1). Using a database compiled by the Immunologic Genome Project 37 , we found that the genes upregulated in low-MTAP tumors were predominantly associated with macrophages and monocytes, while those genes downregulated in low-MTAP tumors represent activated CD4 T cells and innate lymphocytes ( Fig. 1D; Supplementary Figs. S2-S4). Collectively, the findings from various analyses of the gene expression profiles in GBM samples with low versus high MTAP expression suggest a reduction in immune-reactive T cells 38 and an increase in immunosuppressive M2 macrophages 39 in samples with low MTAP expression, opening the possibility that MTAP deficiency is linked to the immunosuppressive GBM microenvironment. Of note, CIBERSORT and Immunologic Genomic project analyses revealed differences in expression of several immune cell types including plasma cells, memory B cells, NK cells, and dendritic cells. While all these differences in the tumor immune profile may be linked or regulated by a similar mechanism and warrant additional investigation, further work presented in this study focuses on macrophage regulation as a model for MTAP deficiency regulating the immune microenvironment.
MTA stimulates alternative activation of macrophages. Macrophages are the most abundant immune cell type in GBM, representing as many as half of all cells in the tumor mass 40 , where they are known to play an immunosuppressive/pro-tumor role 39,41 . Macrophages are traditionally classified as M1 (inflammatory) or M2 (anti-inflammatory). M2 macrophages are classically stimulated by cytokines IL-4, IL-13, IL-10, and M-CSF and are associated with reduction of inflammatory signals, resolution of the immune response, wound healing, and tumor promotion 42,43 . Higher numbers of M2 tumor-associated macrophages have been linked to poor cancer prognosis 43,44 . Upregulated numbers of M2 macrophages in MTAP deficient GBM samples from our data analysis suggests a more immunosuppressive environment with reduction of inflammatory signals and decreased antigen presentation, potentially contributing to the decreased numbers of activated CD4 lymphocytes observed in these samples as well.  45 , which potentiates the effect of cytokines in promoting alternative macrophage activation through adenosine A 2 receptors 46 . This led us to hypothesize that the MTAP substrate, MTA, an adenosine analogue that is secreted by MTAP deleted cells, impacts macrophage polarization through adenosine receptor signaling. In fact, studies focusing on pathogenic mechanism and treatment of Of the 641 genes in the GBM dataset, 239 genes were expressed at significantly different levels (Benjamini-Hocbberg corrected t test P value < 0.0185) in samples with low MTAP expression compared to samples with high MTAP expression. All differentially expressed genes were included in the heatmap. In order to classify the upregulated and downregulated genes identified, each clustered gene set was analyzed using data from the immunologic genome project (https:// www. immgen. org), which uses a compendium of microarray data (366 microarrays/37 studies) from specific immune cell types to identify which immune cells express the genes of interest. Our analysis revealed two unique expression patterns. Genes expressed at higher levels in the low-MTAP samples are more representative of innate immune cells such as macrophages and monocytes, while genes downregulated in low-MTAP cells are more representative of activated CD4 T cells and innate lymphocytes. www.nature.com/scientificreports/ pathogen-induced host inflammatory responses have found that MTA can suppress lipopolysaccharide (LPS)induced expression of inflammatory response genes, such as TNF, via adenosine A 2 receptors 28,47 .
To investigate the impact of MTA on macrophage activation, we utilized the well-established murine RAW 264.7 macrophage cell line model 46,48 . As a control, RAW 264.7 cells were treated with M2 macrophage-inducing Th2-type cytokines, IL-4 and IL-13, which drove the cells toward alternative (M2) activation as expected, demonstrated by the upregulated expression of Arginase 1 (Arg1), a canonical marker for M2 macrophages 49 (Fig. 2A). When MTA was added to the IL-4/IL-13 treatment, induction of Arg1 expression was dramatically augmented (Fig. 2B). Remarkably, when MTA was used as a single agent to treat the cells, upregulated expression of Arg1 was also observed (Fig. 2C). This upregulation of Arg1 transcription was not limited to RAW 264.7 cells as it was also observed in the murine BV-2 microglial cell line, the human THP-1 monocytic cell line, and in primary human monocytes ( Fig. 2D-F). This rise in transcription was followed by a rise in Arginase-1 protein levels (Fig. 2G). Notably, despite similarly upregulating Arg1 expression, the overall effect of MTA on macrophages was distinct from that of IL-4/IL-13, as treatment of cells with MTA attenuated the IL-4/IL-13-induced expression Pparg and Mrc1, marker genes associated with the M2a macrophage subtype ( Supplementary Fig. S5A,B) 50,51 . Instead, MTA activated the expression of Vegfa, a classic feature of the M2d subtype and tumor associated macrophages (TAMs) (Fig. 2H) 51 . This phenotype was validated in human THP-1 cells and primary human monocytes differentiated into macrophage lineage (Fig. 2I,J). Prompted by this finding, we revisited the GBM gene expression data (TCGA) and confirmed that in GBM patients VEGFA expression was significantly higher in samples with low MTAP expression (Fig. 2K). Collectively, these results suggest that MTA promotes alternative macrophage activation resembling the M2d subtype and that this process is likely occurring in MTAP deficient GBMs.
To validate that the phenotype induced by exogenously administered MTA is relevant to the amount of MTA produced by tumor cells, we generated an isogenic MTAP-null derivative of the murine GBM cell line CT Together, these results support the hypothesis that MTA in the extracellular compartment of MTAP-null GBM cells can work in concert with known M2 macrophage-stimulating cytokines in altering tumor associated macrophages, and/or directly promoting the alternative activation of macrophages as a single agent.

Regulation of macrophage activation by MTA requires the adenosine A 2B receptor and STAT3.
Previous work has shown that the action of adenosine in modulating the innate immune response to cytokine signaling is achieved through adenosine A 2 receptors 51 . We tested whether the effects of MTA are also mediated through these receptors using specific antagonists of A 2A and A 2B receptors. We found that an A 2A receptor antagonist, Istradefylline, had only minimal effect on MTA-induced Arg1 expression (Fig. 3A). In contrast, an antagonist of the A 2B receptor, PSB 0788, potently attenuated the expression of Arg1 induced by MTA (Fig. 3B,C). In order to rule out potential off-target effects of the pharmacologic inhibitor and confirm involvement of the A 2B receptor in the phenomena we observed, we used CRISPR/CAS9 to genetically knock out the A 2B receptor gene, Adora2b ( Supplementary  Fig. S7). We saw decreased basal expression levels of Arg1 and Timp1 and a greatly dampened response to MTA ( Supplementary Fig. S6D,E). Collectively, these results suggest that MTA acts through the adenosine A 2B receptor and that this receptor is involved in regulating macrophage activation states.
We next asked whether a sustained contribution of the A 2B receptor signaling pathway is required for maintaining the alternative macrophage activation state. To address this question, we treated the cells with the A 2B receptor antagonist at a delayed time point following treatment with MTA. We found that treatment of the already activated macrophages with the adenosine A 2B receptor antagonist was still able to abolish Arg1 expression (Supplementary Fig. S5E), suggesting the effect of MTA can be pharmacologically reversed, an important point for determining any translational therapeutic potential.
To illuminate the downstream mediators of MTA-stimulated A 2B receptor signaling, we tested the role of transcriptional regulators CREB and STAT3, as these are reported to function downstream of adenosine receptors and are known to regulate Arg1, Vegfa, and IL10, among other genes [52][53][54][55][56] . We found that the inhibition of STAT3 completely abrogated the MTA-induced expression of all three genes. In contrast, inhibition of CREB had only marginal or no effect on MTA-induced gene expression (Fig. 3G-I; Supplementary Fig. S8A-C). We then utilized a dominant-negative construct of STAT3 to genetically reduce STAT3 activity and again observed decreased response to MTA ( Supplementary Fig. S8D) 57 . Interestingly, we were not able to reliably measure any change in STAT3 phosphorylation (p-Tyr705) levels in response to MTA at the time points measured (Supplementary Fig. S8E), suggesting such a phosphorylation change was not necessary for MTA-stimulated A 2B receptor signaling in potentiating alternative macrophage polarization. This finding is consistent with known roles of STAT3 in regulating gene expression and chromatin organization, and as an important component of cancer and immune biology as previously demonstrated 58  www.nature.com/scientificreports/ concentrations of adenosine or MTA and tested the expression of macrophage activation markers in response to each metabolite. To our surprise, we found that MTA much more potently stimulated expression of Arg1, Vegfa, and Timp1 than did adenosine (Fig. 4A-C). In addition, MTA was unique in its ability to suppress Pparg expression (Fig. 4D). Furthermore, while MTA and adenosine similarly upregulated IL10 expression (Fig. 4E), the impact of adenosine on IL10 (and Timp1) was impervious to PSB0788, suggesting adenosine was acting through a different mechanism to induce expression of these genes. were determined by RT-qPCR, error bars indicate mean ± SEM, n ≥ 5. All statistical comparisons were done using ANOVA for group comparisons or unpaired student's t-test for individual comparisons; *P < 0.05, **P < 0.005, ***P < 5 × 10 -4 , n.s. = not significant. www.nature.com/scientificreports/ We then tested the role of downstream mediators in producing the response to MTA and adenosine signaling. We found the effect of MTA and adenosine on IL10 expression were both effectively blocked by betulinic acid, an inhibitor of transcription factor C/EBP (CEBPA/CEBPB) (Fig. 4F), consistent with reports of C/EBP being a critical regulator of IL10 expression 62 and indicating a convergence of the adenosine and MTA signaling pathways through this transcription factor. Furthermore, similar to what we had observed with MTA (Fig. 3G), induction of IL10 expression by adenosine was also blocked by stattic, indicating adenosine-stimulated IL10 expression requires STAT3 (Fig. 4G). However, whereas MTA-induced expression of IL10 was independent of CREB ( Supplementary Fig. S8B), addition of CREB inhibitor 666-15 effectively blocked IL10 expression in response to adenosine (Fig. 4G), indicating that in the downstream signaling pathways there are distinct mechanisms responsible for MTA and adenosine-mediated responses in macrophages.
To further illuminate the distinct/overlapping roles of MTA and adenosine, we investigated the expression of IL6, another important regulator of macrophage activation 63,64 that has been reported to be upregulated in cells were treated with MTA or adenosine for 12 h with or without (E) A 2B receptor antagonist PSB 0788 or (F) C/EBP inhibitor betulinic acid (B. Acid) and IL10 expression was measured by RT-qPCR, n ≥ 5. (G) IL10 expression was measured by RT qPCR following adenosine treatment with or without STAT3 inhibition (Stattic), CREB inhibition (666-15), or A 2A R antagonist Istradefylline, n ≥ 5. (H,I) RAW 264.7 cells were treated with MTA or adenosine for 12 h with or without the indicated inhibitors and IL6 expression was measured by RT-qPCR, n ≥ 5. (J) RAW 264.7 cells were treated with the indicated components, denoted by color codes, and cytokines in spent media were measured using Luminex technology after 2 (IL-6) or 3 (IL-10, VEGFA) days. The diagram shows the concentrations of each cytokine in the media. All statistical comparisons were performed using ANOVA for group comparisons or unpaired student's t-test for individual comparisons; Error bars indicate mean ± SD; *P < 0.05, **P < 0.005, ***P < 5 × 10 -4 , n.s. = not significant. www.nature.com/scientificreports/ response to adenosine signaling [65][66][67][68][69] . We found that adenosine much more potently upregulated IL6 than did MTA, and that while inhibition of STAT3 decreased IL6 expression in the context of MTA treatment, STAT3 inhibition had no significant effect on IL6 expression following adenosine treatment (Fig. 4H). Furthermore, adenosine-mediated upregulation of IL6 expression was further potentiated by A 2A receptor antagonist Istradefylline, as well as by inhibition of CREB or C/EBP, while the effect of MTA was only potentiated by the C/ EBP inhibition, not CREB or A 2A receptor blockade, suggesting there are different negative feedback signaling pathways for MTA and adenosine (Fig. 4I). Thus, while MTA and adenosine both upregulate IL6 expression (adenosine to a greater extent), the dramatic differences in IL6 regulation following MTA or adenosine stimulation further demonstrate the distinctive receptors and signaling pathways activated by these two metabolites.
To confirm the cytokine production measured by gene expression analysis, we utilized multiplex cytokine quantification to measure IL-6, IL-10, and VEGFA protein levels in spent media from RAW 264.7 cells treated with MTA or adenosine. This analysis of cell protein production confirmed the gene expression findings detailed above, with increased VEGFA and IL-10 levels in response to MTA exposure and clear IL-6 and IL-10 production following adenosine treatment ( Fig. 4J; Supplementary Fig. S9).
Collectively these results suggest that MTA and adenosine each engage a distinct balance of adenosine receptor signaling, resulting in divergent pathway activation, gene expression profiles, and cytokine production. The unique effect of MTA on macrophages, which is not shared by adenosine and requires the adenosine A 2B receptor, STAT3, and C/EBP, supports an immunosuppressive, M2-polarizing effect of aberrant MTA accumulation on macrophages in MTAP-deficient GBM.

Discussion
Understanding and targeting the immuno-suppressive mechanisms of GBM is a critical step toward improving treatment for this lethal cancer type 9 . In this study, we provide multiple lines of evidence to link MTAP loss, one of the most common genetic/epigenetic events in GBM, to changes in the tumor immune microenvironment. We demonstrate using patient data that MTAP deleted GBMs contain an immunosuppressive gene expression signature, with increased numbers of M2 macrophages and decreased lymphocyte activation. Our findings resonate with a recent study showing that loss of the CDKN2A/MTAP locus confers an immunosuppressive microenvironment in human cancers 70 , raising the possibility that MTAP loss is an underlying factor. Utilizing in vitro macrophage models, we verified that the process of M2 polarization is influenced by MTA signaling through the adenosine A 2B receptor, resulting in upregulation of M2 marker genes Arg1, IL10, and Vegfa. This represents a unique mechanism of alternative macrophage activation and a potentially significant contribution to the immunosuppressive tumor microenvironment in GBM, which is known for an abundance of immunosuppressive tumor-associated macrophages 71 . It has previously been demonstrated that MTA can directly suppress the proliferation and function of T lymphocytes 26,72 . Our findings complement these studies by showing that MTA can also influence innate immune cells.
The finding that the macrophage response to MTA is distinct from the response to adenosine suggests that MTAP-deficient GBMs with aberrant MTA accumulation likely have an impact on the immune environment in a manner that is qualitatively and/or quantitatively different from tumors which simply accumulate adenosine as a mechanism of immune evasion 73,74 . The role of adenosine receptor signaling in regulating inflammation, as well as its tumor-promoting functions, are beginning to be more widely recognized 75 . Adenosine receptors and the adenosine-generating enzymes, CD39 and CD73, are being investigated as therapeutic targets, either through stimulating or countering the adenosine receptor signaling pathways to control pathogenic inflammation or treat advanced cancers [76][77][78] . In GBM, adenosine (and upregulation of CD39 and CD73) has been shown to contribute to tumor-mediated immunosuppression 79 . Additionally, A 2B receptor signaling has been shown to promote tumor metastasis and stimulate tumor angiogenesis 80,81 . Our finding that the effect of MTA on macrophage functionality is mediated by the adenosine A 2B receptor and is amenable to pharmacological intervention adds a new line of evidence to support the rationale of targeting A 2B receptor signaling in MTAPdeficient GBM. Development of MTAP-deficient, immune-competent GBM models will be necessary for further investigating the effectiveness of this strategy in vivo. We speculate that given the abundance of tumor-associated macrophages and the characteristic immunosuppressive microenvironment of GBM, altering MTA-mediated purinergic signaling has the potential to significantly impact tumor growth and augment immunotherapeutic interventions in MTAP-deleted GBM.
We note that experimental findings from the in vitro cell line models bear major limitations and may not fully explain the observed link between immunosuppression and MTAP status. In addition, those experimental results represent only a limited view of the complex time scale of the signaling mechanisms and responses involved. We suggest that further research utilizing human GBM specimens and in vivo, immunocompetent orthotopic GBM models will be necessary to assess the functional interplay between MTAP loss and the innate immune cells in GBM, and how they collectively influence the adaptive immune characteristics within the tumor.

Materials and methods
Cell lines and cell culture. RAW   Milliplex/luminex assays. To quantify cytokine levels in cell culture media, we utilized the multiplex cytokine quantification technology available from Milliplex. Custom kits were ordered that measure IL-6, IL-10, and VEGFA. Samples were processed according to manufacturer's instructions and read on a Luminex machine in the Immunology Unit of the Duke Regional Biocontainment Laboratory (RBL). www.nature.com/scientificreports/ Analysis of The Cancer Genome Atlas (TCGA) data. All TCGA data was downloaded from https:// tcga-data. nci. nih. gov/ docs/ publi catio ns/ tcga/ and through cbioportal.org 82,83 . The most recently published 2013 GBM data set was used for all analyses. For each analysis, the maximum number of complete cases available (confirmed IDH1/2 wildtype) were used unless otherwise stated, as IDH mutations are known to independently influence epigenetics and cellular differentiation. Analysis was carried out using Graphpad Prism, Microsoft excel, Genesis, and Statgraphics software. To perform this analysis, samples were ordered by MTAP expression level or equally divided into quartiles based on MTAP expression and statistical comparisons were made between the "low" and "high" expression groups to identify genes that may be impacted by MTAP status. Benjamini-Hochberg correction was performed to account for multiple testing error. We utilized MTAP expression levels to categorize patients rather than gene copy number because MTAP is known to be silenced epigenetically in a variety of cancer types [84][85][86][87] , and our analysis of DNA methylation and gene expression data in patients suggests it can also be epigenetically silenced in GBM 21 .
Statistical analysis. Statistical tests (student's t test, ANOVA) were performed using Graphpad Prism. All experiments were repeated to ensure reproducibility of results. Unless otherwise indicated, pooled data from multiple experiments was used for each figure. A P value cutoff of 0.05 was used to determine significance in all cases except where corrections were applied for larger data sets (i.e., Benjamini-Hochberg, Bonferonni).