Clinical Implications of Aberrant PD-1 and CTLA4 Expression for Cancer Immunity and Prognosis: A Pan-Cancer Study

Combination therapy with inhibitors of cytotoxic T lymphocyte-associated protein (CTLA)4 and programmed death (PD)-1 has demonstrated efficacy in cancer patients. However, there is little information on CTLA4 and PD-1 expression levels and their clinical significance across diverse cancers. In this study, we addressed this question by analyzing PD-1 and CTLA4 levels in 33 different types of cancer along with their prognostic significance using The Cancer Genome Atlas (TCGA) and Cancer Cell Line Encyclopedia datasets. Liver hepatocellular carcinoma (LIHC) patients receiving cytokine-induced killer cell (CIK) immunotherapy at Sun Yat-sen University cancer center were enrolled for survival analysis. The correlation between PD-1/CTLA4 expression and cancer immunity was also analyzed. The results showed that PD-1 and CTLA4 transcript levels varied across cancer cell lines, with aberrant expression detected in certain cancer types; Kaplan–Meier analysis with the Cox proportional hazards model showed that this was closely related to overall survival in breast invasive carcinoma, glioblastoma multiforme, head and neck squamous cell carcinoma, acute myeloid leukemialymphoma, uterine corpus endometrial carcinoma, and uveal melanoma in TCGA. High serum PD-1 and CTLA4 levels predicted better survival in LIHC patients receiving CIK therapy. PD-1 and CTLA4 levels were found to be significantly correlated with the degree of tumor cell infiltration using Tumor Immune Estimation Resource, Estimating Relative Subsets of RNA Transcripts, and Estimation of Stromal and immune Cells in Malignant Tumor Tissues Using Expression Data as well as with tumor-infiltrating lymphocyte marker expression; they were also related to tumor mutation burden, microsatellite instability, mismatch repair, and the expression of DNA methyltransferases in some cancer types. Gene set enrichment analysis of 33 cancer types provided further evidence for associations between PD-1/CTLA4 levels and cancer development and immunocyte infiltration. Thus, PD-1 and CTLA4 play important roles in tumorigenesis and tumor immunity and can serve as prognostic biomarkers in different cancer types.


INTRODUCTION
Cancer is a leading cause of death worldwide, and the low efficacy of many existing therapies is a major clinical challenge (1). Molecular-level pan-cancer analyses have provided insights into the common features and heterogeneity of various human malignancies (2). For example, Cancer Cell Line Encyclopedia (CCLE) and The Cancer Genome Atlas (TCGA) were developed based on epigenomic, genomic, proteomic, and transcriptomic data from multiple human cancer cell lines and tissues (3)(4)(5). Pan-cancer analyses have also revealed the significance of specific genes and signaling pathways in cancers. For example, tumor hypoxia-associated multi-omic investigations have shown that some molecular variants are correlated with antitumor drug sensitivity or resistance, which has important implications for cancer treatment (6). The expression status of Forkhead box M1 and its relationship to etiology and outcomes of human cancers (7), as well as proteomic and genomic features related to MYC and the proximal MYC network (8) have been reported for 33 cancer types in TCGA. The expression of more than 9,000 genes in TCGA has been characterized in terms of their contribution to the immune phenotype of various cancers (9). Thus, pan-cancer analyses can be useful for the development of new combination treatments and individualized therapies. The mechanisms of immune evasion by cancer cells are the target of immunotherapies (10). Cytotoxic T lymphocyteassociated protein (CTLA)4 and programmed death (PD)-1 are receptors that attenuate the T cell response (11). Both factors are the immune checkpoint inhibitors with distinct but complementary mechanisms of action. CTLA4 is a target for monoclonal antibody-based drugs that enhance anticancer immunity such as ipilimumab, which was the first CTLA4 inhibitor to be developed and the only one to date that has been approved by the U.S. Food and Drug Administration (FDA) (12). PD-1 is a transmembrane protein that is expressed by immunocytes; blocking PD-1 signaling enhances the anticancer effect of T cells, thereby promoting cancer cell killing. The combination of nivolumab (13)-which targets PD-1-and ipilimumab increased overall survival (OS) in patients with melanoma (14), renal cell carcinoma (15)(16)(17), and advanced non-small cell lung cancer (18), and has been approved for the treatment of hepatocellular carcinoma (LIHC) patients previously treated with sorafenib (19).
Although PD-1 and CTLA4 overexpression, mutations, and gene amplification have been reported in certain cancers, the studies had small sample sizes and used different experimental approaches, making it difficult to compare the findings. Additionally, these studies focused on a single or a few types of cancer; there have been no studies comparing multiple types of cancer. To this end, the present study investigated PD-1 and CTLA4 expression profiles and their prognostic significance in various human malignancies based on large CCLE and TCGA datasets. We also examined the associations between PD-1/CTLA4 expression and the extent of tumor cell infiltration, microsatellite instability (MSI), tumor mutational burden (TMB), DNA methyltransferase (DNMT) levels, and mismatch repair (MMR) in different tumor types by gene set enrichment analysis (GSEA). The results provide important insights into the roles of PD-1 and CTLA4 in anticancer immunity.

Patient Datasets and Processing
TCGA comprises over 20,000 samples from 33 types of cancer and corresponding non-carcinoma samples. Processed level 3 RNA sequencing data and corresponding clinical annotations were obtained from TCGA using the University of California Santa Cruz cancer genome browser (https://tcga.xenahubs.net; accessed April 2020). CCLE (https://portals.broadinstitute.org/ ccle) provides genetic and pharmacologic information from a large number of human tumor models, with RNA sequencing data for over 1,000 cell lines. Ethics approval for use of human data was not required for this part of the study as only openaccess datasets were used.
A total of 122 consecutive patients with LIHC (mean age, 46.8 years; range: 22-75 years) who underwent curative resection and received adjuvant cytokine-induced killer (CIK) cell immunotherapy at Sun Yat-sen University Cancer Center between March 2004 and January 2015 were enrolled. Preparation of CIK cells and the treatment schedule are described in our previous study (20). Preoperative patient serum samples were obtained from our hospital's sample bank and analyzed using an anti-PD-1 and -CTLA4 antibody array (RayBiotech, Norcross, GA, USA; product no. QAH-ICM-1-1) according to the manufacturer's instructions. Briefly, the array was incubated with blocking buffer for 1 h before 2-fold-diluted serum samples (60 µl) were added. After overnight incubation at 4 • C followed by washes, biotin-conjugated detection antibody was added for 2 h. The array was washed and Alexa Fluor 555conjugated streptavidin was added for 1 h at room temperature. An InnoScan 300 scanner (Inopsys, Carbonne, France) was used to detect the signals (532 nm excitation); raw data were processed as images and spot intensities using Mapix 7.3.1 software (Innopsys). Serum concentrations of PD-1 and CTLA4 proteins were determined by automatic normalization and calculation. Follow-up was conducted until November 2019, with a median time of 84.3 months (range: 11.6-134.7 months). The primary endpoint was OS. This study was approved by the Ethics Committee of Sun Yat-sen University Cancer Center. All participants provided written, informed consent. The analytical workflow is shown in Supplementary Figure 1. to clinical information by patient identification number. A univariate Cox proportional hazards model was used to calculate correlations between gene expression and patient survival, where P < 0.05 was taken as the threshold for a statistically significant difference for PD-1 and CTLA4 expression in a given cancer relative to normal tissue. A Kaplan-Meier survival analysis was carried out to compare OS of patients in TCGA, which was stratified according to median PD-1 and CTLA4 expression levels with the log-rank test.

Relationship Between PD-1/CTLA4 and Tumor Immunity
Tumor Immune Estimation Resource (TIMER; https:// cistrome.shinyapps.io/timer/) allows systemic analysis of immune infiltrates in different cancer types (21) using a deconvolution statistical approach to infer tumor-infiltrating lymphocyte (TIL) counts based on gene expression data (22). Using the TIMER algorithm, we examined the associations between PD-1/CTLA4 levels and the numbers of 6 immune  (23). In this study, CIBERSORT was used to calculate the relative fractions of the 22 leukocyte types; the correlations between PD-1/CTLA4 levels and each leukocyte across 33 cancer types was then determined.
Estimation of Stromal and Immune Cells in Malignant Tumor Tissues Using Expression Data (ESTIMATE) uses gene expression profiles to predict the purity of a tumor based on infiltration of stromal cells/immunocytes (24). The ESTIMATE algorithm yields 3 scores based on GSEA of single samples, including (1) stromal score, which reflects the presence of stromal cells in tumor tissue; (2) immune score, which indicates the degree of immunocyte infiltration into tumor tissue; and (3) estimate score, which describes tumor purity. We used the algorithm to estimate both immune and stromal scores for a variety of tumor tissues, and evaluated the associations between the scores and PD-1/CTLA4 levels.
We also examined the associations between PD-1/CTLA4 levels and the expression of TIL markers (25)(26)(27). An expression heatmap was generated for gene pairs in specific cancer types and correlations were analyzed with Spearman's rank correlation test. TMB measures the number of mutations in a specific cancer genome and is used as a biomarker to identify patients that are most likely to respond to checkpoint inhibitor therapy (28). We obtained somatic mutation data of all TCGA patients (https:// tcga.xenahubs.net), calculated their TMB scores, and then analyzed the correlation between TMB and PD-1/CTLA4 level. MSI is characterized by length polymorphisms of microsatellite sequences resulting from DNA polymerase slippage. Patients with high MSI (MSI-H) cancers benefit from immunotherapy, and MSI is an index used for cancer detection (29). We computed the MSI score of each patient and performed a correlation analysis between MSI and PD-1/CTLA4. MMR, is a DNA repair mechanism in normal cells that corrects DNA replication errors. Gene mutation frequency may be increased in cancer cells as a result of downregulation of MMR genes or defective MMR (29). Here we analyzed the correlation between

Functional Analysis
We carried out GSEA using the JAVA program (http:// software.broadinstitute.org/gsea/index.jsp) to investigate the biological significance of PD-1 and CTLA4 expression levels in tumor tissues. The random sample permutation number was set as 100, and the threshold of significance was P < 0.05. The results were visualized with enrichment maps generated using Bioconductor (http:// bioconductor.org/) and R v3.6.0 software (R Foundation, Vienna, Austria).

Statistical Analysis
Survival was evaluated as OS (defined as the time from the date of diagnosis to death from any cause) and progression-free survival (PFS; defined as the time until disease progression or death from any cause). The Wilcox log-rank test was used to assess changes in the sum of gene expression z-scores of cancer tissues compared to adjacent normal tissues. Differences in PD-1 and CTLA4 levels between different tumor stages were compared with the Kruskal-Wallis test. Survival was analyzed with Kaplan-Meier curves, the log-rank test, and Cox proportional hazards regression model. Spearman's or Pearson's test was used for correlation analysis. All statistical analyses were performed using R software.
We examined PD-1 and CTLA4 expression according to age, sex, race, and tumor stage and found that older patients had higher expressions of these genes than younger patients, while no differences were observed between sexes (Figure 2). Black patients had higher PD-1 and CTLA4 levels than Caucasian patients. PD-1 level was higher whereas CTLA4 level was lower in stage I disease compared to other stages.  (A,B), the left top triangle is colored to represent the P-value, and the right bottom triangle is colored to represent the Spearman correlation coefficient. *P < 0.05, **P < 0.01, ***P < 0.001. Frontiers in Immunology | www.frontiersin.org Association Between PD-1/CTLA4 Levels and Survival PD-1 expression was identified by Cox regression analysis as a prognostic factor for OS in urothelial bladder carcinoma (BLCA), BRCA, cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), GBM, HNSC, KIRC, acute myeloid leukemia (LAML), STAD, UCEC, and uveal melanoma (UVM) ( Table 1). The Kaplan-Meier survival analysis showed that subjects with higher PD-1 levels had shorter OS than those with lower levels in GBM (P = 0.037), KIRP (P = 0.040), LAML (P = 0.002), low-grade glioma (LGG) (P < 0.001), and UVM (P < 0.001). On the other hand, subjects with higher PD-1 levels had longer OS than those with lower levels in BRCA (P = 0.014), HNSC (P = 0.006), skin cutaneous melanoma (SKCM) (P < 0.001), and UCEC (P < 0.001) (Figures 3A-I).
PD-1 expression was also a prognostic factor for PFS in BLCA, BRCA, CESC, HNSC, KIRP, SKCM, UCEC, and UVM ( Table 1). The Kaplan-Meier survival analysis showed that patients with higher PD-1 expression had shorter OS than those with lower expression in LGG (P < 0.001) and UVM (P = 0.025), whereas patients with higher PD-1 levels had longer OS than those with lower PD-1 levels in BRCA (P = 0.018), CESC (P = FIGURE 7 | Correlation between PD-1 and CTLA4 expression and cancer immunity in CIBERSORT. (A,B) CIBERSORT prediction of the association between PD-1 (A) and CTLA4 (B) levels and degree of TIL infiltration in 33 cancer types. For each pair, the left top triangle is colored to represent the P-value, and the bottom right triangle is colored to represent the Spearman correlation coefficient. *P < 0.05, **P < 0.01, ***P < 0.001.  LIHC (B,D). For each pair, the top left triangle is colored to represent the P-value, and the bottom right triangle is colored to represent the Spearman correlation coefficient. *P < 0.05, **P < 0.01, ***P < 0.001. 0.047), SKCM (P = 0.047), and UCEC (P < 0.001) ( Supplementary Figures 2A-F).

Relationship Between PD-1/CTLA4 Levels and Degree of Immune Cell Infiltration
TILs are independent predictors of sentinel lymph node status as well as survival. We examined the correlation between PD-1/CTLA4 levels and the degree of immune cell infiltration in diverse cancer types in TIMER. PD-1 and CTLA4 levels were significantly associated with tumor purity in 35 and 36 cancer types, respectively. Additionally, PD-1 and CTLA4 levels were correlated with the degree of infiltration of CD4+ T cells in 33 and 33 cancer types, respectively; of B cells in 30 and 32 cancer types, respectively; of CD8+ T cells in 32 and 34 cancer types, respectively; of macrophages in 24 and 25 cancer types, respectively; of dendritic cells in 35 and 36 cancer types, respectively; and of neutrophils in 32 and 35 cancer types, respectively. The FDA granted the accelerated approval for the use of PD-1 in combination with CTLA4 for the treatment of LIHC (19). In this study, the association of the degree of immune infiltration with the levels of PD-1 and CTLA4 within LIHC is presented as an example in the top panels of Figures 6A,B. While the pan-cancer correlations of immune infiltration level with  Table 2, respectively. CIBERSORT was used to determine immunocyte profiles in all TCGA patients, and the correlation between 22 immunocytes and PD-1/CTLA4 expression was determined for the 33 cancer types in TCGA (Supplementary Table 3). PD-1 and CTLA4 were significantly correlated with CD8+ T cell but not CD4+ naïve T cell counts in most cancers (Figure 7). Additionally, the expression of T cell markers (CD25, CD137, and human leukocyte antigen DRB1) was correlated with PD-1 and CTLA4 levels.
Immune and stromal scores for tumor tissues were calculated using ESTIMATE; we then assessed the associations between these scores and PD-1 and CTLA4 expression (Figure 8 and

Correlations Between PD-1/CTLA4 and Immune Cell Marker Expression
We examined the associations between TIL markers and PD-1/CTLA4 expression and found that PD-1 and CTLA4 levels were significantly correlated with the expression of T cell and other immunocyte markers (Figure 9), suggesting that both factors are involved in the regulation of the immune response to these cancers.

Correlation Analysis of TMB, MSI, MMR, and DNMT Expression
We next examined the associations between PD-1/CTLA4 expression and TMB, MSI, MMR, and DNMT levels (Supplementary Table 5 and Figures 10, 11) and DNMT3B) and PD-1 and CTLA4 levels are shown in Figures 11C,D, respectively.

Functional Analysis
We performed GSEA to assess the biological significance of PD-1 and CTLA4 expression in different cancers. The functional Kyoto Encyclopedia of Genes and Genomes (KEGG) and Hallmark terms for PD-1 and CTLA4 are listed in Figures 12A,B, respectively.

DISCUSSION
Combination therapy with immune checkpoint inhibitors including anti-PD-1 and -CTLA4 antibodies shows greater therapeutic efficacy than the monotherapies in several cancers (31)(32)(33)(34)(35). In the present study, we used a comprehensive pancancer approach to evaluate the clinical significance of PD-1 and CTLA4 expression in a variety of cancers. Our results showed that PD-1 and CTLA4 expression varies across cancer types and that most cancers are characterized by PD-1 and Frontiers in Immunology | www.frontiersin.org CTLA4 mutations that lead to their abnormal expression, which can serve as a prognostic biomarker. Serum PD-1 and CTLA4 levels were survival predictor in LIHC patients receiving CIK therapy. Using TIMER, CIBERSORT, and ESTIMATE, we found that PD-1 and CTLA4 overexpression was associated with TIL infiltration, immune scores, and immune marker expression. Furthermore, PD-1 and CTLA4 levels were correlated with TMB, MSI, MMR, and the expression of DNMTs. We also identified KEGG and Hallmark terms that are associated with PD-1 and CTLA4 expression.
Identifying aberrantly expressed genes in tumors is important for the development of individualized treatments, which can improve therapeutic outcomes (36). Pan-cancer analyses can reveal the functional significance of PD-1 and CTLA4 in cancers (37,38). Here we examined PD-1 and CTLA4 expression in a large and diverse set of samples from CCLE, which provides gene expression data for future experiments, and from TCGA, which provides genomic and survival data that may be useful for clinical investigations. Consistent with previous reports (39,40), we found that PD-1 was more highly expressed in older cancer patients, indicating that checkpoint inhibitor treatment may more effective in this group. Additionally, Black patients had higher PD-1 and CTLA4 levels than Caucasian or Asian patients, suggesting better outcomes following immunotherapy.
Our results showed that PD-1 and CTLA4 are implicated in cancer immunity, as evidenced by the association between PD-1 and CTLA4 levels and the degree of infiltration of immunocytes in the TIMER and CIBERSORT analyses. The ESTIMATE method has been used to assess genomic data in various cancers, including the prediction of clinical outcomes in GBM and SKCM (41,42). We determined immune and stromal scores based on TCGA data and found that PD-1 and CTLA4 levels were correlated with ESTIMATE scores as well as the expression of TIL markers.
Gene mutations are the major cause of cancer development (43), and specific mutations predict treatment response and prognosis (44,45). TMB affects the generation of immunogenic peptides, thereby affecting patients' response to immune checkpoint inhibitor treatment (46,47). Additionally, TMB and MSI reflect the production of new antibodies, with the latter being linked to increased TMB (48). In cervical squamous cell carcinoma and adrenocortical carcinoma, MSI-H was associated with an abnormally high mutation frequency (49). Thus, MSI is an important predictor of tumor development (29). MSI testing is recommended in the National Comprehensive Cancer Network guidelines for all colorectal cancer subtypes, as early detection of MSI-H can reduce mortality (50). Cancer cells with MMR deficiency (dMMR) can generate heterologous antigens that are recognized by T cells. PD-1 inhibitors are highly effective against MSI-H solid tumors (51); accordingly, the FDA has approved the use of the anti-PD-1 immunotherapy pembrolizumab for the treatment of MSI-H/dMMR solid tumors (51). Thus, TMB, MSI, and MMR can be used to predict therapeutic responses. In this study, we showed that PD-1 and CTLA4 levels were correlated with TMB and MSI in BRCA, COAD, TGCT, and UCEC. However, it remains to be determined whether the combination of PD-1 and CTLA4 inhibitors has greater efficacy than monotherapy in these cancers. Epigenetic modifications modulate gene expression and can be exploited by tumor cells to evade immune surveillance. A potential therapeutic strategy to circumvent this problem is to combine immune checkpoint and methylase inhibitors (30). PD-1 and CTLA4 have been implicated in various pathways related to immune function (52,53). We carried out GSEA to identify clinically relevant pathways that may provide clues for future research. Taken together, our findings provided clues for the roles of PD-1 and CTLA4 in cancer immunity. However, these results should be interpreted with caution since checkpoint inhibitor treatment not analyzed in our work. And more experiments are needed to demonstrate our results, such as immunohistochemistry.
In conclusion, the results of our pan-cancer analysis indicate that PD-1 and CTLA4 are useful prognostic biomarkers in some cancer types. Importantly, we found that PD-1/CTLA4 expression is associated with cancer immunity. The integrative omics-based workflow in this study can serve as a basis for developing and testing hypotheses regarding novel targets in cancer treatment.

DATA AVAILABILITY STATEMENT
All datasets presented in this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
J-NL, WL, and Q-FC conceived of and designed the study. J-NL, X-SK, TH, WL, and Q-FC performed the literature search, generated the figures and tables, and wrote the manuscript. J-NL, TH, X-SK, and RW collected, analyzed the data, and critically reviewed the manuscript. WL and Q-FC supervised the study and reviewed the manuscript. All authors contributed to the article and approved the submitted version.

ACKNOWLEDGMENTS
We thank Ms. Yue-Ming Du for encouraging Q-FC to pursue his career in medicine.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2020.02048/full#supplementary-material  Table 2 | Associations between PD-1/CTLA4 expression levels and CD4+ and CD8+ T cell, B cell, dendritic cell, neutrophil, and macrophage counts based on TIMER.
Supplementary Table 3 | Association between PD-1/CTLA4 expression levels and immunocyte marker expression as determined using CIBERSORT.