CPNE8 as a Novel Biomarker for the Prognostic and Immunological Role by Pan-Cancer Analysis and Cerna Network of CPNE8 in Stomach Adenocarcinoma

Calcium-dependent protein copine 8 (CPNE8), a new member of Copine family, emerges important roles in malignancies. However, the expression and potential role of CPNE8 in pan cancer remain unclear. Methods: Pan cancer analysis of CPNE8 expression in 33 types of cancers were based on data downloaded from the University of California Santa Cruz (UCSC) Xena database. The clinical signicances of CPNE8 were analyzed using Cox regression analysis and Kaplan Meier survival analysis. Correlations between CPNE8 expression and tumor immune microenvironment were assessed using ESTIMATE algorithm, CIBERSORT analysis and online database TISIDB. The correlations between CPNE8 expression and tumor mutation burden (TMB), microsatellite instability (MSI) status and response to immune checkpoint inhibitors (ICIs) were further explored. Moreover, competitive endogenous RNA (ceRNA) networks related to CPNE8 and its correlations with immune cell related genes were further studied in stomach adenocarcinoma (STAD). Results: Differential expression of CPNE8 were identied in 14 types of cancers. Differential expression of CPNE8 could predict tumor stage or prognosis of various cancer patients. Pan cancer analysis indicated that CPNE8 demonstrated oncogenic role in STAD, head and Neck squamous cell carcinoma (HNSC) and esophageal carcinoma (ESCA), and tumor suppressive role in thyroid carcinoma (THCA). These four cohorts were selected for further study due to their consistent results in expression and prognostic values. Gene set enrichment analysis (GSEA) revealed that CPNE8 might regulate several key immune-related signaling pathways. Furthermore, CPNE8 expression was signicantly correlated with the inltration of different subtypes of immune cells, TMB and MSI. Moreover, PWAR5 and LIFR-AS1 could sponge miR-222-3p as ceRNAs to regulate the expression of CPNE8 in STAD. In addition, CPNE8 was signicantly correlated with the expression of immune cell related genes, such as CD8 T cells, macrophages, B cells and dendritic cells in STAD. Conclusion: This study demonstrated could serve as a promising biomarker associated with the prognosis and immune microenvironment in several malignancies.


Introduction
Malignancies, as the major threaten of human beings, are caused by complicated events and usually underwent multiple processes. The incidences of most cancer types are increasing yearly, such as lung cancer, breast cancer, prostate cancer, colorectal carcinoma [1][2][3]. For the past decades, surgery, chemotherapy and radiotherapy strategies are the primary choices in most cancer types [4,5]. Currently, the medical practices have been changed due to the development of next generation sequencing technology [6]. The emergences of targeted therapy and immunotherapy have gained a lot of attentions [7][8][9]. But only part of patients could bene t from targeted therapies or immune checkpoint RNA-sequencing data and clinical data of pan cancer were downloaded from the University of California, Santa Cruz (UCSC) Xena database (https://xena.ucsc.edu) miRNA sequencing data was downloaded from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov).
Screening CPNE8 expression in pan cancer UCSC Xena database includes 33 kinds of TCGA cohorts. Differential CPNE8 expression was compared between tumors tissues and normal tissues using "limma" package in R.
Clinical signi cance of CPNE8 in pan cancer Clinical data of pan cancer was downloaded from UCSC Xena database and tumor stages were retrieved. Correlations between CPNE8 expression and tumor stages were analyzed using "ggpubr" package.
TISIDB, an integrated repository portal for tumor-immune system in interactions, was also used to assess the associations between CPNE8 expression and tumor stages (http://cis.hku.hk/TISIDB/index.php).

Correlations between CPNE8 expression and survival
Survival data, including survival time and survival status, was downloaded from UCSC Xena database and retrieved using "limma" package in R. The signi cances of CPNE8 expression in predicting overall survival (OS), disease speci c survival (DSS), disease free survival (DFS) and progression free survival (PFS) were analyzed using Cox regression analysis and Kaplan-Meier survival curve analysis by using "survival", "survminer", and "forestplot" packages.

ESTIMATE analysis
ESTIMATE algorithm was used to calculate immune scores and stromal scores. The stromal score represents the presence of stroma, while the immune score represents the in ltrations of various immune cells. ESTIMATE analysis was performed using "estimate" package. And the correlations between CPNE8 expression and ESTIMATE algorithm results were assessed using Spearman correlation coe cient. And when |correlation coe cient| > 0.3 and P < 0.05 was considered as statistical signi cance.

Tumor mutation burden (TMB) and microsatellite instability (MSI) correlation analyses
The TMB data was downloaded from UCSC Xena database and retrieved using Perl language. The correlations between CPNE8 expression and TMB or MSI were analyzed using "fmsb" package in R software. And the results were demonstrated using Radar graphs.

Response to ICIs
Three clinical trials data concerning the application and e ciency of anti-PD1 drugs were downloaded from GSE78220, GSE67501 and IMvigor which were involved in melanoma, human renal cell carcinoma and muscle-invasive urothelial carcinoma, respectively. Differential CPNE8 expression between response and non-response groups were assessed using "limma", "ggplot2" and "ggpubr" packages in R.
Identi cation of CPNE8 related miRNA TargetScan database http://www.targetscan.org/vert_80/ was used to predict the 3'UTR-binding miRNAs of CPNE8 transcripts. miRNA sequencing data for the STAD cohort were downloaded from the TCGA database. The correlation of these miRNAs with CPNE8 expression was analyzed using the "limma", "reshape2", "ggplot2" and "ggpubr" packages in R language. In addition, the prognostic value of CPNE8 plus associated miRNAs was analyzed using the "limma", "survival" and "survminer" packages in the R language.

Correlations with immune cell related genes
The correlations between CPNE8 expression and immune cell related gene expression were analyzed using "limma", "reshape2", "ggpubr", and "ggExtra" packages in STAD cohort.

Statistical analyses
Data retrieval was processed using Perl software. All statistical analyses were conducted using R software (version 4.1.0). P < 0.05 or 0.01 was de ned as a signi cant difference.

Clinical signi cance of CPNE8 in pan cancer
Correlations between CPNE8 expression and tumor stages were also studied in pan cancer. As shown in Figure 1B, there were positive correlations between CPNE8 expression and advanced tumor stage in Skin Cutaneous Melanoma (SKCM) (P < 0.05) and Stomach adenocarcinoma (STAD) (P < 0.05). CPNE8 expression was negatively correlated with the tumor stage in BLCA (P < 0.001), KICH (P < 0.01), and THCA (P < 0.01). Moreover, TISIDB data also veri ed that higher CPNE8 expression was correlated with advanced stage in LUSC, STAD and UCEC. While lower CPNE8 expression was positively correlated with advanced tumor stage in KICH, THCA and Uterine Carcinosarcoma (UCS) ( Figure 1C).

Prognostic value of CPNE8 in pan cancer
Moreover, the roles of CPNE8 in predicting overall survival (OS), disease speci c survival (DSS), disease free survival (DFS) and progression free survival (PFS) were determined using both Cox regression analysis and Kaplan Meier survival analysis.
Firstly, Cox regression analysis showed that high CPNE8 expression was risk factor for OS in LGG  Table 1). While high CPNE8 was protective factor for the OS of UCS (HR=0.6103, P=0.0346) ( Table 1). Kaplan Meier survival curves showed that high CPNE8 expression predicted poor OS in ACC ( Prognostic values of CPNE8 expression in predicting DSS were also analyzed by Cox regression analysis and KM analysis. As shown in Table 2 Table 2). KM curve analysis also showed that high CPNE8 expression was correlated with poor DSS in ACC ( Figure  In addition, high CPNE8 expression was risk factor correlated to DFS in STAD (HR=2.3955, P=0.0015) and ESCA (HR=2.0800, P=0.0086) ( Table 3). KM curve revealed that high CPNE8 expression predicted poor DFS in STAD (P = 0.007) ( Figure 2L).  Table 4). KM survival analysis indicated that high CPNE8 expression was correlated with poor PFS in ACC ( Figure 2M, P = 0.008), GBM ( Figure 2N, P = 0.008) and STAD ( Figure 2O, P < 0.001).

Functional annotation of CPNE8
The above results indicated that high CPNE8 expression was correlated with advanced tumor stage, poor OS, DSS, DFS and PFS in STAD. High CPNE8 expression was correlated with poor OS in HNSC. High CPNE8 expression was correlated with poor DFS in ESCA. While CPNE8 expression was a good factor in predicting PFS in THCA. Thus, GSEA was utilized to explore the potential mechanisms of CPNE8 in STAD, HNSC, ESCA and THCA. As shown in Figure 3, CPNE8 related signaling pathways were mainly enriched in Antigen processing and presentation, Cytosolic DNA pathway, Toll like receptor signaling pathway, Olfactory transduction, Calcium signaling pathway, Cell adhesion molecules cams, et al.

Correlations between CPNE8 expression and tumor immune microenvironment
GSEA results showed that the functional annotation of CPNE8 were mainly enriched in immune-related signaling pathways. Thus, we further explored the correlations between CPNE8 expression and immune microenvironment of STAD, HNSC, ESCA and THCA. Firstly, ESTIMATE algorithm was used to calculate the stromal score, immune score and ESTIMATE score. Spearman correlation coe cient analysis showed that CPNE8 expression was positively correlated with stromal score in STAD ( Figure 4A, R=0.38 P=1.8e-14). Secondly, the distribution of tumor-in ltrating immune cell (TIC) abundance was analyzed using CIBERSORT analysis. As shown in Figure 4B  Correlations between CPNE8 expression and TMB, MIS and response to ICIs The above studies revealed that CPNE8 expression was closely correlated with tumor immune microenvironment. Thus, we further investigated the correlations between CPNE8 expression and immunotherapy indicators (TMB and MSI), as well as response to ICIs. As shown in Figure  Correlations between CPNE8 expression and MSI showed that CPNE8 expression was positively correlated with MSI in BRCA (R=0.083, P=0.007) and READ (R=0.199, P=0.013). While CPNE8 expression was negatively correlated with MSI in STAD (R=-0.269, P=1.15e-07), DLBC (R=-0.490, P<0.001), LUAD (R=-0.144, P=0.001), and KICH (R=-0.265, P=0.032) ( Figure 5B). We next studied the correlations between CPNE8 expression and response to ICIs in three cohorts. GSE78220 and GSE67501 cohorts analyses showed that there were no signi cant difference in CPNE8 expression between response and non response groups in melanoma ( Figure 5C, P=0.76) and human renal cell carcinoma ( Figure 5D, P=0.23). IMvigor cohort results showed that the expression of CPNE8 was lower in response group than that in the non-response group in muscle-invasive urothelial carcinoma ( Figure 5E, P=0.0029).
Identi cation of miRNA correlated to CPNE8 in STAD The above results showed that CPNE8 expression was positively correlated with advanced tumor stage, poor OS, DSS, DFS and PFS in STAD. Moreover, the downstream mechanisms of CPNE8 were immune related. Thus, we further explored the upstream lncRNA-miRNA-CPNE8 network in STAD.

Correlations between CPNE8 expression and immune cell related genes
We further studied the correlations between CPNE8 expression and immune cells by analyzing the correlations between CPNE8 expression and immune cell related genes. As shown in Figure 8 . The clinical signi cances of CPNE8 expression were analyzed by assessing the correlations with tumor stage and patients' survival, including OS, DSS, DFS and PFS. We next selected TCGA cohorts with consistent results in differential expression and prognosis for further study. Pan cancer analysis showed that high CPNE8 expression was correlated with advanced tumor stage, poor OS, DSS, DFS and PFS in STAD. High CPNE8 expression was correlated with poor OS in HNSC. High CPNE8 expression was correlated with poor DFS in ESCA. While CPNE8 expression was a good factor in predicting PFS in THCA. Currently, there were limited studies concerning the roles of CPNE8 expression in carcinogenesis [14,15,23,24]. CPNE8 expression was initially identi ed expressed in testis and prostate [13]. Nagasawa et al. reported that knock down of CPNE8 inhibited the proliferation of clear cell carcinoma (CCC) cells and high-grade serous carcinoma (HGSC) cells [11]. Here CPNE8 expression demonstrated oncogenic roles in STAD, HNSC and ESCA. In THCA, CPNE8 could work as a tumor suppressor.
Literature searching revealed that CPNE8 worked as lipid binding protein. Reinhardt et al found that functions of CPNE8 depended on ERK activity in pankinsonian neurodegeneration [25]. However, no studies explored the working mechanism of CPNE8 in malignancies. Utilizing GSEA, we found that the underlying mechanisms concerning CPNE8 were enriched in immune-related and adhesion-related, such as antigen processing and presentation, toll like receptor signaling pathway, cell adhesion molecules cams, focal adhesion in STAD, ESCA, HNSC and THCA. Increasing studies suggested that the progression of tumors were the results of the interactions between various components of the tumor immune microenvironment (TME). TME was consisted of tumor cells, immune cells, stroma and various soluable factors. Our GSEA results revealed that CPNE8 was closely correlated to the immune-related signaling pathway. Thus, we further explored the correlations between CPNE8 expression and the tumor immune microenvironment of STAD, HNSC, ESCA and THCA. Firstly, ESTIMATE analysis was used to explore the correlations between CPNE8 expression and immune score, stromal score and ESTIMATE score. Here, CPNE8 expression was positively correlated with the stromal score of STAD. This result was consistent with the GSEA results in STAD, where the underlying signaling pathways were cell adhesion molecules cams and focal adhesion, which were closely related to tumor stroma.
Moreover, the correlations between CPNE8 expression and in ltrations of various immune cells were determined using CIBERSORT analysis. CPNE8 expression was positively correlated with T regulatory cells and CD56bright cell abundance, and negatively correlated with dendritic cell resting, CD56dim and monocyte in ESCA. CPNE8 expression was positively correlated with the abundance of macrophage, mast cell and NKT cells. The abundance of various immune cells could mediate the progression or regression of tumor cells. Immunotherapies targeting tumor immune microenvironment gain some advancements recently, such as anti-CTLA4, anti-PD-1/PD-L1 [26]. The immunotherapies could induce durable anti-tumor immunity, thus making them promising strategies in cancer treatments [27,28]. Currently, there is still lack of de nite biomarker in predicting the responses to immunotherapies. Applications of TMB and MSI could partially predict the possible e ciency of anti-PD1/PD-L1 [29,30].
Here we found that CPNE8 expression was negatively correlated with TMB and MSI status in STAD.CPNE8 expression was positively correlated with TMB in HNSC, and negatively correlated with TMB in THCA. These results might partially explain the response rate of anti-PD1 was much lower in STAD than that in HNSC. And the expression of CPNE8 was much higher in nonresponse group than that in response group according to the IMvigor cohort, which indicating that CPNE8 expression could predict the response of Atezolizumab in advanced UC. However, due to the limited number of clinical trials concerning the immunotherapies in cancer treatment, the potentials of CPNE8 expression in predicting the response to immunotherapies were restrained.
Although CPNE8 failed to demonstrate signi cant differences between normal tissues and tumor tissues in STAD. We can see that the expression of CPNE8 was higher in tumor tissues. CPNE8 expression was positively correlated with advanced tumor stage, poor OS, DSS, DFS and PFS and TME in STAD, which making it an ideal oncogene in STAD. Thus, we further explored the upstream mechanism that regulate CPNE8 expression in STAD. LncRNAs regulate mRNA expression by sponging miRNA, which is called ceRNA network. The ceRNA network of CPNE8 was not well elucidated. Liu et al found that CPNE8 was a candidate target genes of miR-375 in malignant breast cancer [23]. LncRNA RP11-396F22.1 overexpression predicted poor survival in early-stage cervical cancer. And knockdown of RP11-396F22.1 could upregulate CPNE8 expression to regulate the migration ability of cervical cancer cells [15]. Here, PWAR5 or LIFR-AS1-miR-222-5p-CPNE8 ceRNA network were constructed according to TCGA database and ENCORI database.
The effects of PWAR5 were largely unknown and still controversial. You et al identi ed PWAR5 as an oncogenic lncRNA in papillary thyroid carcinoma. And low PWAR5 expression was correlated with better OS [31]. Another study in glioma revealed that PWAR5 worked as a tumor suppressor [32]. LIFR-AS1 was identi ed as an oncogenic lncRNA in some tumors, including pancreatic cancer [33], osteosarcoma [34], glioma [35]. Studies in non small cell lung cancer[36] and breast cancer [37] showed that LIFR-AS1 might work as a tumor suppressor. Wang et al found that high LIFR-AS1 expression was correlated with poor overall survival of gastric cancers patients [38]. Pan et al revealed that LIFR-AS1 promoted the proliferation and migration of gastric cancer cells by elevating the expression of COL1A2 through sponging miR-29a-3p [39]. Here, PWAR5 expression or LIFR-AS1 expression was negatively correlated with miR-222-5p and positively correlated with CPNE8, which indicating that PWAR5 or LIFR-AS1 might sponge miR-222-5p to regulate the expression of CPNE8. However, more experiments need to be utilized to verify these results.
Moreover, CPNE8 expression was signi cantly correlated with the expression of immune cell related genes, such as T cells, B cells, macrophages, neutrophils and dendritic cells. These results strongly indicated that upregulated CPNE8 expression might result in modulating the tumor immune microenvironment. However, there are some limitations of this study, especially the veri cation in our clinical specimens, cellular functions and mice experiments. It's worth of being further explored.

Conclusion
In summary, pan cancer analysis revealed that CPNE8 could work as immune related biomarkers in predicting the survival of multiple cancer patients. Moreover, construction of PWAR5 or LIFR-AS1/miR-222-5p /CPNE8 ceRNA network further revealed the potential mechanisms of aberrant CPNE8 expression in STAD. These results strongly indicate that CPNE8 is a promising biomarker demonstrating prognostic and immunological roles in cancers.       Differential CPNE8 expression between response group and non response group in renal cell carcinoma.

Tables
(D) Differential CPNE9 expression between ICI response and nonresponse group in mUC.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.