Molecular and clinical characterization of PARP9 in gliomas: A potential immunotherapeutic target

Abstract Background Glioma is a primary malignancy of the central nervous system (CNS). As biomedicine advances, an efficient molecular target is urgently needed for the diagnosis and treatment of glioma. Meanwhile, several studies have demonstrated that glioma development is closely related to immunity. PARP9 is an inactive mono‐ADP‐ribosyltransferase belonging to the poly‐ADP ribosyltransferase (ARTD) family. In this article, we aimed to reveal the relationship between PARP9 and glioma and explore the potential prognostic value and immunotherapeutic targetability of PARP9 in glioma. Methods PARP9 transcript levels were analyzed with TCGA and GEO databases. The clinicopathological information of patients with glioma in the TCGA database and gene expression profiles were analyzed to determine the relationship between the expression of PARP9 and clinicopathologic characteristics. Kaplan‐Meier survival analysis, univariate Cox regression analysis, and multivariate Cox regression analysis were used for survival analysis. Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) were used for bioinformatics analysis. Correlation analysis explored the relationships between PARP9, infiltrating inflammatory immune cells, and immune checkpoint molecules. Results PARP9 is highly expressed in glioma, and high expression of PARP9 is associated with poor prognosis and advanced clinicopathological features. Bioinformatics analysis showed that some immune‐related pathways were closely associated with high expression of PARP9. Correlation analysis indicated that PARP9 was closely related to inflammatory and immune responses, high immune cell infiltration, and immune checkpoint molecules. Conclusions PARP9 may serve as an unfavorable prognosis predictor for glioma and a potential immunotherapeutic target.


| INTRODUC TI ON
Glioma is a primary malignancy of the central nervous system (CNS), accounting for 80% of all CNS malignancies. 1 According to their histological features, gliomas are classified into astrocytomas, ependymomas, oligodendrogliomas, mixed gliomas, and brainstem gliomas. 2 Despite advances in a variety of treatments, overall survival (OS) improvements in glioma patients have been limited. 3 With advances in biomedical technology, a variety of biomarkers and molecular classifications of glioma have been established. However, most of the markers are of limited use in the diagnosis and treatment of glioma.
Therefore, it is urgent to identify effective markers and therapeutic targets for glioma.
Immune checkpoint inhibitors (ICIs) perform well in the treatment of solid tumors, including melanoma, 4 non-small-cell lung cancer, 5 and renal cell carcinoma. 6 Moreover, the discovery of lymphatic vessels in the central nervous system makes the immunotherapy of glioma feasible. 7 However, immunotherapy currently does not work well for most gliomas. In previous studies, PD-L1 (programmed death ligand 1), 8 TIM3 (T-cell immunoglobulin mucin-3), 9 and IDO1 (indoleamine 2,3-dioxygenase 1) 10 transcript levels were strongly correlated with immune responses and prognosis in gliomas. A large amount of research has led to a growing recognition of the relationship between glioma and immunity.
PARP9 is an inactive mono-ADP-ribosyltransferase belonging to the intracellular diphtheria toxin-like glutamate/aspartate-specific mono-and poly-ADP ribosyltransferase (ARTD) family (also known as PARPs). 11 However, PARP9 lacks PARP activity, despite possessing carboxy-terminal amino acid sequences similar to those of other members of the PARP family. 12 Previous studies have shown that PARP9 is overexpressed in a series of solid tumors, such as breast tumors, 12 prostate tumors, 11 diffuse large B cell lymphomas, 13 and cervical tumors, 14 and PARP9 may promote metastasis, recurrence and chemotherapy resistance in these tumors. Previous studies have also shown that PARP9 can regulate macrophages, 15 which are the major type of immune cell within brain tumors, often comprising up to 30% of the tumor mass, 16 indicating that PARP9 may influence tumor immune infiltration in glioma. However, the expression of PARP9, its clinical significance and its relationship with immune infiltration in glioma remain elusive. Thus, this study aimed to reveal the relationship between PARP9 and glioma and explore the potential prognostic value and immunotherapeutic targetability of PARP9 in glioma.

| Gene expression profile data and clinical data analysis
Microarray data of glioma patients were obtained from the Gene Expression Ominibus (GEO) 17 under the accession number GSE50161 18 ; the GPL570 platform (Affymetrix Human Genome U133 Plus 2.0 Array) was used. The RNA-seq datasets and clinical data from patients with glioma from The Cancer Genome Atla (TCGA) (https://cance rgeno me.nih.gov/) 19 were downloaded and analyzed. The TCGA glioma datasets comprised low-grade glioma (LGG) and GBM datasets (HTSeq-FPKM), and they were further analyzed for associations between PARP9 expression, clinicopathologic characteristics and immune cell infiltration.

| Bioinformatics analysis
Gene set enrichment analysis (GSEA) was performed to identify differentially enriched biological pathways between the high PARP9 expression and low PARP9 expression groups. In addition, gene set variation analysis (GSVA) was performed to transform the gene expression patterns of all samples in the TCGA database into scores for inflammatory response metagenes, and correlograms were used to further verify the correlations between PARP9 and these metagenes.

| Statistical analysis
In this study, SPSS 25.0, R software 3.6 and GraphPad Prism 7.0 statistical software were used to conduct the analysis. The variables distribution was checked by the Shapiro-Wilk test. Student's t-test was performed to evaluate data that follows a normal distribution, for other variables the Mann-Whitney test was used. Descriptive statistics were used to summarize the molecular and clinical information from the TCGA database. Logistic regression tests were used to analyze the relationship between PARP9 and clinicopathological features. Kaplan-Meier survival analysis, univariate Cox regression analysis, and multivariate Cox regression analysis were used to compare the effects of PARP9 expression and other clinical variables on overall survival in patients. Classical correlation analysis was used to detect the correlation between PARP9 expression, inflammatory type, immune cell type, and immune checkpoint molecules. The heat map, circos, and corrgram functions were conducted in the R software. A P value <.05 was considered statistically significant.

| Expression of PARP9 in glioma and normal samples
By analyzing the RNA sequencing data of glioma and normal brain tissue from the TCGA database, we found that PARP9 was significantly upregulated in glioma compared to nontumor tissues ( Figure 1A, P < .001). Moreover, we also obtained the GSE50161 dataset from the GEO database to verify the results ( Figure 1B, P < .001). In addition, ROC curves were generated based on PARP9 expression and sample type for two datasets. As shown in Figure 1C,D, the area under the curve (AUC) reached 92.7% and 93.2% in the GEO and TCGA datasets, respectively. These results suggested that PARP9 may be a potential biomarker in glioma.

| Glioma patient characteristics
The analysis of the relationship between PARP9 expression and clin-
Logistic regression analysis revealed that increased PARP9 was associated with poor prognosis (

| PARP9 predicted poor prognosis in glioma
To analyze the prognostic value of PARP9 in glioma, Kaplan-Meier curves were constructed with data from the TCGA database. As shown in Figure 3, patients with high PARP9 expression had significantly poorer survival than patients with low PARP9 expression (P < .001). The results indicated that PARP9 was a poor prognosis marker in glioma (

| PARP9-related biological process and signaling pathways
We explored and verified the biological processes and signaling pathways associated with PARP9 expression using gene set enrichment analysis (GSEA). As shown in Figure 4 and Table 4, several biological processes and signaling pathways were enriched in patients with high PARP9 expression, such as antigen processing and presentation, the B cell receptor signaling pathway, cytokine-cytokine receptor interactions, Fc gamma R-mediated phagocytosis, the JAK-STAT signaling pathway, natural killer cell-mediated cytotoxicity, pathways in cancer, the T-cell receptor signaling pathway and the Toll-like receptor signaling pathway. Most of these biological processes and signaling pathways are involved in immune and inflammatory responses.

| Correlation between PARP9 and inflammatory activities
To thoroughly understand PARP9-related inflammatory activities, seven clusters containing 104 genes (File S1) representing different types of inflammation and immune activities were used for study. 20 As shown in Figure 5A, the expression of PARP9 was positively related to most of the gene clusters, such as the HCK, LCK, interferon, STAT1, MHC I, and MHC II clusters, yet it was negatively correlated with the IgG cluster, which represented B cells. To verify our above analysis, GSVA was performed to convert gene expression data into enrichment scores for metagenes. Correlograms were generated to visualize the relationship between PARP9 and seven metagenes, and the result was consistent with our above findings ( Figure 5B; Table 5).

| Correlation of PARP9 with infiltrating immune cells
Previous research has shown that tumor-infiltrating immune cells play a critical role in regulating tumor progression and prognosis. 10,21,22 We investigated the relationship between the expression of PARP9 and six immune cells that frequently infiltrate the tumor, including expression was positively correlated with 6 immune cell-specific markers, which indicated that patients with higher PARP9 expression were more likely to have more infiltrating immune cells than patients with lower PARP9 expression in glioma ( Figure 6A).  which supported our results. 29 In our study, we firstly performed the comprehensive research to identify the expression pattern and distribution of PARP9 in 1,114 glioma samples, which include, but are not limited to, lower-grade gliomas. Increased PARP9 expression was found to be significantly related to advanced clinicopathologic characteristics and poor prognosis in glioma patients. These results suggested that PARP9 may be a valuable therapeutic target for glioma gene therapy.

| D ISCUSS I ON
In previous studies, glioma cells were found to be able to produce During the last few years, remarkable advances have been made in the field of cancer immunotherapy, largely driven by the considerable success of ICIs in other tumor types. 32,33 However, glioma has remained largely refractory to current immunotherapies. 34,35 Enhanced response rates to a combination of two different ICIs have been reported in patients with melanomas compared with the responses seen with either agent as monotherapy. 36 Previous studies have reported that PD-L1, TIM-3, and IDO1 transcriptional levels are highly associated with immune response and prognosis in patients with glioma. 8-10 Furthermore, experimental results have revealed that combination anti-PD-1 plus anti-TIM-3 treatment can achieve a longer overall survival than anti-TIM-3 monotherapy in murine gliomas. 37 Targeting TAMs via the inhibition of the colony-stimulating factor-1 receptor (CSF-1R) pathway has emerged as an attractive approach for anticancer therapy. 38 Dual PD-1 and CSF-1R blockade could promote antitumor activity in preclinical models. 39

| CON CLUS IONS
In summary, this study demonstrates that PARP9 is overexpressed in glioma samples. High PARP9 expression is associated with advanced clinicopathological parameters and predicts much worse survival for glioma patients. We also found that PARP9 was involved in the inflammatory and immune responses and was correlated with checkpoint molecules. Therefore, taken together, PARP9 may serve as an unfavorable prognosis predictor for glioma and a potential immunotherapeutic target, which, when used in combination, may improve the therapeutic efficacy of ICIs.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.