PLXDC1 May Serve As A Target For Combining Antiangiogenic Therapy and Immunotherapy In Gastric Adenocarcinoma

Introduction Based on the immunosuppression of traditional antiangiogenic agents in the treatment of tumors and the newly proposed concept of antiangiogenic therapy combined with immunotherapy, this paper will mainly explore the prospects of PLXDC1 in stomach adenocarcinoma (STAD) regarding antiangiogenic therapy and immunotherapy. Methods First, the transcriptional and translational levels of PLXDC1 in STAD were analyzed using the Oncomine, The Cancer Genome Atlas (TCGA) and Human Protein Atlas databases and then univariate and multivariate Cox regression analyses were performed using TCGA data. Next, we explored the correlation between PLXDC1 and STAD immunity from multiple aspects. Finally, based on the acquisition of immunomodulators associated with PLXDC1 expression from TISIDB, we constructed PLXDC1-related immune prognostic signatures of four genes (NT5E, CTLA, TGFBR1, and CSF1R) and constructed a nomogram for predicting survival to analyze the clinical utility of PLXDC1 in immunotherapy. Results Our results demonstrated that PLXDC1 was highly expressed in STAD and that its high expression was associated with poor prognosis in STAD. Multivariate Cox analysis suggested that PLXDC1 could be used as an independent prognostic risk factor for STAD. The high-risk group for which we constructed PLXDC1-related immune prognostic signatures showed poorer prognosis compared to low-risk group, and the risk score of our model could be used as an independent risk factor for STAD prognosis. Moreover, the nomogram survival prediction system showed good accuracy of the constructed immune signatures. Conclusions In conclusion, PLXDC1 can serve as a biomarker for the diagnosis and treatment of STAD and it may also be a new target for STAD immunotherapy. Therefore, PLXDC1 can combine antiangiogenic therapy with immunotherapy. cancer cases . Patients with early STAD limited to the mucosa and submucosa have ve-year survival rates of 70–95% with surgical while western surgical and population-based series show ve-year survival rates of 20–30% for with advanced tumors that have the submucosa . searching for biomarkers for STAD, earlier great importance for


Introduction
Stomach cancer has the fth highest incidence and the second highest mortality rate of all cancers worldwide 1 . According to the pathological staging of gastric cancer, stomach adenocarcinoma (STAD) accounts for 95% of all gastric cancer cases 2 . Patients with early STAD limited to the mucosa and submucosa have ve-year survival rates of 70-95% with surgical treatment 3 . while western surgical and population-based series show ve-year survival rates of 20-30% for most patients with advanced tumors that have penetrated the submucosa 4 . Therefore, searching for biomarkers for STAD, aiding in the earlier detection of cancer, is of great importance for patient prognosis.
Immunotherapy is a landmark discovery in cancer treatment that kills tumor cells mainly by modulating the immune system and altering the tumor immune microenvironment 5,6 . Additionally, an increasing number of studies have shown that immune cell in ltration plays a crucial role in tumor prognosis 7,8 . Currently, immunotherapy is widely practiced in the clinical treatment of non-small cell lung cancer, and programmed death-ligand 1 (PD-L1) agents developed to target programmed death-1 (PD-1) on the tumor surface have improved the survival of tumor patients by disrupting the tumor's immune evasion mechanism and killing tumor cells [9][10][11] . The remarkable success of immunotherapy in non-small cell lung cancer provides a new direction for immunotherapy in STAD 12 . A study found that PD-L1 was expressed in tumor cells and immune stroma across all stages and histologies of STAD, while patients with higher CD8 + T-cell densities also had higher PD-L1 expression, suggesting the occurrence of adaptive immune resistance mechanisms 13 . However, immunotherapy targeting PD-1 is only applicable to some STAD patients, and effective targets of immunotherapy for most patients still need to be explored, so identifying suitable immunotherapy targets is of great signi cance for STAD patients.
Plexin domain containing 1 (PLXDC1) is a vascular protein associated with angiogenic states 14 . It was demonstrated that pigment epithelium derived factor binds to PLXDC1 on the cell surface through its extracellular structural domain, thus achieving anti-neoangiogenesis and inhibiting tumor growth 15 . Currently, PLXDC1 is reported to be involved in the development of various cancers. For example, glioblastoma endothelium drives bevacizumab-induced in ltrative growth through the regulation of PLXDC1 16 , and PLXDC1 promotes migration and invasion in gastric cancer 17 . However, there are no relevant studies on PLXDC1 in STAD and its immunity, and this study will focus on these aspects.

Acquisition of PLXDC1 expression pro les in gastric cancer
First, we used the Oncomine database (https://www.oncomine.org/resource/login.html.) and selected the "Gene Differential Expression" module to explore the expression of PLXDC1 across cancers. Then, we chose gastric cancer as the subject of our study and further analyzed the differential expression of PLXDC1 in gastric cancer and its subtypes with normal tissues using the dataset on the website (P-value < 0.05). Next, we downloaded the STAD data (normal = 32, cancer = 375) from TCGA database (https://www.cancer.gov/.) and compared the expression of PLXDC1 in STAD and normal tissues using the "limma" R package (P-value < 0.05; t test). Finally, using the Human Protein Atlas database (https://www.proteinatlas.org/.), we analyzed the expression of PLXDC1 in STAD at the protein level.

Survival analysis of PLXDC1 in gastric cancer
Using the Kaplan-Meier Plotter online website (https://kmplot.com/analysis/.), we selected the "Gastric Cancer" module to classify high-and low-risk groups according to the median value of PLXDC1 expression. Then, we analyzed the correlation of PLXDC1 expression with overall survival (OS) (n = 875), rst progression (FP) (n = 640), and post-progression survival (PPS) (n = 498) in gastric cancer (P-value < 0.05; log-rank test). Next, we downloaded STAD data (normal: n = 32, tumor: n = 375) from TCGA and then performed univariate Cox analyses of PLXDC1 with clinical factors such as age, gender, grade, stage, T stage, N stage and M stage (P-value < 0.05). Then, we used receiver operating characteristic (ROC) curves to assess the accuracy of the prognostic value of these clinical factors, and the larger the area under the curve (AUC) was, the more accurate the prediction. Finally, using multivariate Cox analysis, we identi ed independent risk factors for prognosis.(P-value < 0.05)

Enrichment of immune-related pathways
We used GSEA version 4.1.0, a method for the analysis of genome-wide expression pro ling microarray data that compares genes with prede ned gene sets, to enrich for immune pathways associated with PLXDC1 in STAD. The gene expression matrix in STAD was processed by Perl software to obtain the input le related to the target gene. Then, the upregulated gene set was de ned as phenotype h (h = 3649/5093), the downregulated gene set was de ned as phenotype l (l = 1444/5093), and the "h-versus-l" and "immune signature" patterns were selected to enrich for PLXDC1-related immune pathways.

Analysis of the immune microenvironment of PLXDC1 in STAD
First, we utilized the R package obtained from the Cell Type Identi cation by Estimating Relative Subsets of RNA Transcripts (CIBERSORT) (https://cibersort. stanford.edu/.) website, which can qualify and quantify 22 types of immune cells in the tissue, to visualize immune cell in ltration in STAD and to analyze the mRNA expression matrix for further differential analysis of immune cells in ltrating in STAD and normal tissues 18 (STAD vs. normal = 375 vs. 32). Moreover, we performed correlation analysis of immune cells in the STAD immune microenvironment using the "corrplot" R package.
Subsequently, we used TIMER 2.0 (http://timer.cistrome.org/.), a comprehensive resource for systematic analysis of immune in ltrates across diverse cancer types, and selected the "Immune Association" panel to analyze the correlation between the copy number of PLXDC1 and the number of immune cells (P-value < 0.05). The results of the analysis were supplemented with the results of CIBERSORT for the missing data from TIMER 2.0.

Immunomodulators in the immune microenvironment of STAD
Using the "Immunomodulators" panel in the TISIDB, we analyzed the correlation between 23 immunoinhibitors and 44 immunostimulators produced in the STAD immune microenvironment and PLXDC1 expression (P-value < 0.05). Then, the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (https://www.string-db.org/online.) site was used to construct a protein interaction network of immunomodulators associated with PLXDC1 expression. Next, we performed Gene Ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of immunomodulators associated with PLXDC1 expression in the STAD immune microenvironment using the WebGestalt online tool (http://www.webgestalt.org/.), a functional enrichment analysis web tool that supports three well-established and complementary methods for enrichment analysis, including overrepresentation analysis, gene set enrichment analysis, and network topology-based analysis.

Construction of PLXDC1-related immune prognostic signatures and performance evaluation
Based on 48 immunomodulators associated with PLXDC1 expression, we used the least absolute shrinkage and selection operator (LASSO) regression algorithm with penalty parameter tuning conducted by 10-fold cross-validation with the "glmnet" and "survival" packages. Then, the screened immunomodulators from LASSO were subjected to stepwise multivariate Cox proportional hazards regression analysis to obtain the optimal candidates and construct an immune prognostic model of immunity. The formula for calculating the risk score was as follows: where "coe " and "Xi" represent the coe cient and expression level of each PLXDC1-related immunomodulator, respectively. Patients with STAD were classi ed into high-risk and low-risk groups based on the median score as the risk cutoff point. The survival curves of the two groups were plotted using the Kaplan-Meier method and compared by the log-rank test using the "survival" and "survminer" packages in R, with a P-value < 0.05 indicating signi cance. ROC curve analysis was conducted, and the AUC values were obtained to evaluate the prognostic model's predictability using the "survivalROC" package. Finally, we integrated the risk score obtained from the model with clinical data such as age, gender, grade, stage, T stage, N stage, and M stage downloaded from TCGA (n = 163) for univariate and multivariate Cox analyses to explore whether the risk score was an independent prognostic risk factor.

Construction a nomogram prognostic prediction model
Based on the results of the multivariate Cox analysis, we combined the risk score and clinical factors including age, gender, grade, stage, T stage, N stage, and M stage. Each factor was scored in the nomogram scoring system and then the scores were summed to obtain an overall score to predict survival at 1, 3, and 5 years. The concordance index (C-index) and calibration plot were used to assess the predictive performance and discrimination ability of the nomogram scoring system (C = 0.5, no predictive capability; C = 0.51 ~ 0.70, low accuracy; C = 0.71 ~ 0.90, medium accuracy; and C > 0.90, high accuracy).

Statistics
We have implemented all statistical analyses using R (version 4.1.0) and some online databases. Differences between two groups of continuous variables were analysed using t-tests, and differences in continuous variables over two groups were analysed using kruskal-wallis tests. Moreover, survival analysis between the two groups was performed using log-rank test. P value < 0.05 was considered statistically signi cant.

Expression of PLXDC1 in gastric cancer
First, the pan-cancer analysis in Oncomine showed that PLXDC1 was highly expressed in multiple cancers, including gastric cancer (Fig. 1a). Then, analysis of gastric cancer and its subtypes showed that PLXDC1 was overexpressed in gastric cancer compared to normal tissue according to the data of Wang 20 . Based on the data of Cho 21 , PLXDC1 was highly expressed in diffuse gastric adenocarcinoma, gastric mixed adenocarcinoma and gastric intestinal type adenocarcinoma (Table 1). Next, we downloaded the data of STAD in TCGA, and the analysis revealed that PLXDC1 was overexpressed in STAD compared to normal tissue (P-value < 0.05; t test) (Fig. 1b, Fig. 1c). Moreover, we used the Human Protein Atlas to con rm the conclusion that PLXDC1 was highly expressed in STAD at the protein level (Fig. 1D). Overall, PLXDC1 is always highly expressed in gastric cancer and its subtypes.

Survival analysis of PLXDC1 in gastric cancer
First, the survival analysis on the Kaplan-Meier Plotter online site showed that high expression of PLXDC1 in gastric cancer was associated with OS, FP, and PPS (P-value < 0.05; log-rank test), representing poor prognosis (hazard ratio (HR) > 1) (Fig. 2a). Then, we analyzed the prognosis of PLXDC1 in STAD.
Univariate Cox analysis showed correlations of PLXDC1, age, gender, grade, stage, T stage, N stage and M stage with OS (P-value < 0.05), representing poor prognosis (HR > 1) (Table 2). Furthermore, the ROC curves showed high predictive accuracy for age, grade, stage, N stage and M stage in the univariate Cox prognostic analysis (Fig. 2b). Finally, multivariate Cox analysis indicated that PLXDC1 and age were independent prognostic risk factors for STAD (P-value < 0.05) (Fig. 2c). Overall, high expression of PLXDC1 is associated with poor prognosis in gastric cancer and STAD. Moreover, it may serve as an independent prognostic risk factor in STAD.  Fig. 3a, Fig. 3b).
Overall, PLXDC1 may be involved in STAD immune regulation.

Analysis of the immune microenvironment in STAD
First, we visualized the immune in ltration of 22 immune cells in STAD using CIBERSORT (Fig. 4a). Differential analysis of in ltrated immune cells revealed that "B cells naive, T cells CD4 naive, Macrophages M0 and Macrophages M1" showed high in ltration levels in STAD compared to normal tissues, while "B cells memory, Plasma cells, T cells CD8, T cells CD4 memory resting, T cells gamma delta, Monocytes and Mast cells resting" showed low in ltration levels (P-value < 0.05) (Fig. 4b).
"Macrophages M0" and "T cells CD8" were the most signi cantly negatively correlated in the STAD immune cell in ltration correlation analysis, while "T cells CD4 memory activated" and "T cells CD8" were the most positively correlated (Fig. 4c). Overall, there is a complex immune microenvironment in STAD.

Correlation analysis of PLXDC1 and the STAD immune microenvironment
TISIDB online website analysis revealed that the expression of PLXDC1 was closely associated with the immune microenvironment across cancers (Fig. 5a). In the immune microenvironment of STAD, we found that PLXDC1 expression was positively correlated with Tcm_CD8, Tem_CD8, Tcm_CD4, Tem_CD4, Tfh, Tgd, Th1, Th17, NK, MDSC, NKT, Act_DC, pDC, iDC, macrophages, eosinophils, mast, and neutrophil (rho > 0, P-value < 0.05) (Fig. 5b), while it was negatively correlated with Act_CD4 ( Fig. 5b) (rho < 0, P-value < 0.05). In addition, STAD immune subtype phenotyping (C1/2/3/4/6) was correlated with PLXDC1 expression, and the highest gene expression was found in the C6 phenotype (P-value < 0.05) (Fig. 5c). Finally, we explored the correlation between PLXDC1 copy number and the number of immune cells in the STAD immune microenvironment using the TIMER 2.0 database. The results indicated that there was a correlation between PLXDC1 copy number and multiple immune cells (P-value < 0.05) (Fig. 5d). In summary, our analysis from multiple perspectives revealed a close relationship between PLXDC1 and the STAD immune microenvironment.

Construction and validation of a nomogram for the survival prediction of patients with STAD
To improve the model's clinical applicability, we constructed a statistical nomogram model in the training set to predict 1-, 3-, and 5-year patient survival by integrating the model's risk score and clinical factors such as age, gender, grade, stage, T stage, N stage, and M stage (Fig. 8a). To assess the accuracy of the clinical prediction model, we plotted the calibration curves of the 3-year and 5-year survival rates of STAD patients based on the model (Fig. 8b). At the same time, we assessed the overall prediction accuracy of the model using the C-index, and the results indicated that the prediction accuracy of our nomogram reached 0.722 (Fig. 8c).

Discussion
In 1971, Folkman et al. proposed a hypothesis that tumors, to promote tumor cell growth, provide nutrition to their cells by creating new blood vessels 23 . Subsequently, this hypothesis was con rmed as one of the hallmarks of tumors 24 . Until now, the antiangiogenic agents developed have achieved great success in treating many types of cancer, such as breast cancer [25][26][27] , lung cancer 28,29 , colorectal cancer 30 and gastric cancer 31 . Previous studies found that PLXDC1 was closely associated with angiogenesis 14 . In this study, we used the Oncomine, TCGA and Human Protein Atlas databases and found that PLXDC1 was highly expressed in STAD in terms of gene transcription and translation. Simultaneously, survival analysis showed that high expression of PLXDC1 in STAD was associated with poor prognosis. Moreover, multivariate Cox analysis showed that PLXDC1 could be an independent prognostic risk factor for STAD. These ndings suggest that PLXDC1 can be used as a biomarker for the diagnosis and treatment of STAD.
Despite the achievements of antiangiogenic drugs in the treatment of gastric cancer, some intractable problems have emerged 32 . Inhibition of angiogenesis provides only a transient survival bene t to patients because of the presence of escape mechanisms in tumors, including the upregulation of compensatory pathways 33 , vasculogenic mimicry 34 and the recruitment of bone marrow-derived cells 35 . Moreover, several studies have found that angiogenesis inhibitors increase the aggressiveness and metastasis of tumors 36,37 . For example, treatment with short-term sunitinib led to accelerated tumor metastasis in a mouse model 38 . Notably, scientists have discovered that the use of antiangiogenic agents may cause tumors to develop immunosuppression 39,40 . This is because the in ltration and function of immune cells depend on normal vascular channels and their provision of oxygen and nutrients [41][42][43][44] . However, antiangiogenic agents disrupt the blood vessels of the tumor, preventing immune cells from in ltrating and functioning. Based on the immunosuppression and multiple de ciencies of antiangiogenic agents, researchers recently proposed a program that combines antiangiogenic agents with immunotherapy 39,45,46 , which may bring a breakthrough in tumor treatment.
In this study, we found that PLXDC1 is closely associated with the immune microenvironment of STAD from immune pathway enrichment analysis, immune cell in ltration analysis and correlation analysis of immune cells, immune subtypes, and copy number with PLXDC1. In addition, we found 48 immunomodulators in the immune microenvironment of STAD associated with PLXDC1 expression through the TISIDB website. Immunomodulators are an essential component of the tumor microenvironment, and changes in them have a signi cant impact on patient survival 22 . Then, we constructed PLXDC1-related immune prognostic signatures containing four genes (NT5E, CTLA, TGFBR1, and CSF1R) based on these prognosis-related immunomodulators using LASSO regression and stepwise multivariate Cox proportional hazards regression analyses. The results indicated that the high-risk group of our constructed model was associated with the poor prognosis of patients, and the accuracy of the model reached 0.722. Furthermore, multivariate Cox analysis showed that the risk score of the model was an independent risk factor for the prognosis of STAD patients. Therefore, we can conclude that PLXDC1 may be a new target for STAD immunotherapy.
In summary, PLXDC1 can be used as a biomarker for the diagnosis and treatment of STAD, while immune analysis suggests that it may serve as a new target for STAD immunotherapy. Because PLXDC1 can inhibit tumor angiogenesis while playing a role in immunotherapy, we speculate that PLXDC1 might represent a breakthrough in tumor therapy 39 . However, there are some shortcomings in our study. Most of our data came from authoritative databases on the web, and the authenticity of the data relies heavily on these open databases. For this reason, we used multiple databases for validation to verify our ndings from multiple perspectives simultaneously, which supports the robustness of our ndings.

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