GPNMB expression impacts prognosis and immune inltration in cancers

Background Glycoprotein non-metastatic gene B (GPNMB) can regulate tumor progression by interacting with T cell function. However, the association between GPNMB and tumor-inltrating immune cells and prognosis of various cancers is poorly understood. We use the Oncomine and TIMER database to investigate GPNMB expression in multiple tumors. The PrognoScan database, Kaplan-Meier plotter are used to analyze tumor prognosis of GPNMB. R packages are used to performed multivariable cox regression analysis. We use TIMER and GEPIA database to explore the association between GPNMB expression and tumor immune inltration levels, and immune cell markers. GPNMB related transcription factors and transcription-target networks are investigated via TTRUST database and GeneMANIA . A high level of GPNMB expression was signicantly associated with poor prognosis in stomach adenocarcinoma (STAD). While, a high level of GPNMB expression was signicantly associated with favorable prognosis in lung adenocarcinoma (LUAD). Besides, GPNMB expression levels can impact the prognosis in STAD and LUAD patients with lymph node metastasis. Moreover, GPNMB expression level has signicant relationships with B cells, CD8 + T cells, CD4 + T cells, macrophages, neutrophils, and DCs inltrating levels in STAD and LUAD. Besides, various immune gene markers of STAD and LUAD are signicantly related to GPNMB expression. In addition, the GPNMB related transcription factors are MITF and TP53. The transcript-target networks are mainly responsible for signal transduction in response to DNA damage, DNA damage response, signal transduction by p53 class mediator, mitotic G1 DNA damage checkpoint, G1 DNA damage checkpoint. These results indicate that GPNMB is signicantly associated with prognosis and immune inltrating levels in various cancers patients, especially in STAD, LUAD patients. Multiple immune gene markers of STAD and LUAD are signicantly related to GPNMB expression, especially monocyte, macrophage polarization, and functional T cells gene markers. Our study signies that GPNMB plays an essential role in (OS HR = 0.69, P = 0.035; PPS HR = 0.6, P = 0.014; PPS HR = 0.24, P = 0.016). Interestingly, high expression of GPNMB was associated with poor OS in M0 stage (OS HR =1.42, P = 0.015). High expression of GPNMB was associated with poor OS in N1, N1+2+3 stage and poor PPS in N1, N2, N1+2+3 stage (OS HR =1.42, P = 0.015; OS HR =1.5, P = 0.0038; PPS HR =1.81, P = 0.013; PPS HR =0.5, P = 0.0086; PPS HR =1.58, P = 0.003). Overexpression of GPNMB was associated with poor OS in intestinal and diffuse Lauren classication but favorable OS in mixed Lauren classication (OS HR =1.58, P = 0.0042; OS HR =1.5, P = 0.019; OS HR =0.21, P = 0.0022). Overexpression of GPNMB was associated with poor PPS in diffuse Lauren classication(PPS HR = 1.9, P = 0.0019). High expression of GPNMB was correlated with poor OS in well and moderate differentiated type but favorable OS in poorly differentiated type (OS HR =0.59, P = 0.011;OS HR =2.41, P = 0.0083; OS HR =6.43, P = 0.0044). Besides, high expression of GPNMB was correlated with favorable PPS in poorly differentiated type (PPS, HR =0.49, P = 0.039) (Table II). These results indicate that GPNMB expression level is related to prognosis in gen-der, stage1-2, T3-4 stage, N1-3 stage, M0 stage, Lauren classication and differentiation of gastric cancer patients. In addition, we enrolled patients from TCGA explore the value of and other factors.The multivariable Cox regression indicates that the OS of LUAD patients is signicantly associated with GPNMB expression, T stage, N stage and stage (all P<0.05, gure 4). Specically, N stage signies regional lymph node metastasis, and the present study reveal that GPNMB expression level can impact the prognosis in STAD, and LUAD patients with lymph node metastasis. signies in response to DNA

ONCOMINE database ONCOMINE database (www.oncomine.org) is an online cancer microarray database, including DNA or RNA sequences analysis [29]. In our study, we collect GPNMB transcription expressions of different cancer tissues and adjacent normal tissues from the ONCOMINE database. We use p-value: 0.001, fold change: 1.5, gene rank: 10%, data type: mRNA as cut-off of p-value and fold change.
Timer TIMER (https://cistrome.shinyapps.io/timer/) from TCGA, is an open website to investigate the correlation between genes and immune cell in ltration [30]. In our study, we use the "Gene module" to estimate the relationship between GPNMB expression and immune cell in ltration. Spearman's relation and statistical signi cance are used to evaluate the correlation between GPNMB expression and tumor-in ltrating immune gene markers. Moreover, the correlation standard is evaluated using the value: very weak (0.00-0. 19), weak (0.20-0.39), moderate (0.40-0.59), strong (0.60-0.79), very strong (0.80-1.0). P-value < 0.05 is considered to be signi cantly different.

PrognoScan
PrognoScan database (http://www.abren.net/PrognoScan/) is a cancer microarray datasets that can be used to investigate the relationship of GPNMB expression and survival prognosis in multiple cancers types with a Cox P-value< 0.05 [31].

Kaplan-Meier Plotter
We used the Kaplan Meier plotter (http://kmplot.com/analysis/) to analyze the prognostic value of GPNMB expression in cancers such as breast, ovarian, lung, liver, and gastric cancer samples [32]. In Kaplan-Meier plotter, patients are divided into high and low expression group according to median values of mRNA expression and KM survival curves is used to validate the survival status. P-value < 0.05 is considered to be signi cantly different.
GEPIA Dataset GEPIA (http://gepia.cancer-pku.cn/index.html) provides mRNA expression data of tumors and normal samples according to Cancer Genome Atlas (TCGA) and Genotype tissue Expression dataset projects [33]. In this study, we use GEPIA to con rm the relationship between GPNMB expression and immune gene markers in TIMER.
GeneMANIA GeneMANIA (http://www.genemania.org) is a web that could construct a protein interaction (PPI) network, gene interaction, co-expression, gene enrichment, and gene co-localization [35]. In this study, we use GeneMANIA to construct GPNMB related transcription factor-target network.

Statistical analysis
Multivariable Cox regression analysis was conducted with R package (version 3.6.3) using a signi cant level of 0.05.

GPNMB mRNA expression in Different Cancer Types
We used the oncomine database (www.oncomine.org) to explore the expression of GPNMB expression in tumors and normal tissues of various cancer types.
This study revealed that the GPNMB expression was higher in brain and CNS cancer, cervical cancer, esophageal cancer, gastric cancer, head and neck cancer, kidney cancer, liver cancer, lung cancer, leukemia, lymphoma, melanoma, pancreatic cancer, prostate cancer, and sarcoma compared to normal tissues ( Figure  1A). Moreover, lower expression was detected in bladder, breast, colorectal, esophageal cancer, leukemia, lung, ovarian, and sarcoma cancers in some data sets. We summarized the detailed results of GPNMB expression in various cancer types in table I. Then, TIMER was used to further explore GPNMB expression in various cancer types and normal tissues in Figure 1B. GPNMB expression was signi cantly lower in BLCA (bladder urothelial bladder carcinoma), BRCA (invasive breast carcinoma), COAD (colon adenocarcinoma), READ (rectum adenocarcinoma), LUAD (lung adenocarcinoma), and UCEC (uterine corpus endometrial carcinoma) compared to adjacent healthy tissues. Besides, GPNMB expression was signi cantly higher in CHOL (cholangiocarcinoma), ESCA (Esophageal carcinoma ), HNSC (head and neck cancer), KICH (kidney chromophobe), KIRC(kidney renal clear cell carcinoma), KIRP(Kidney renal papillary cell carcinoma), LIHC (liver hepatocellular carcinoma), LUSC (Lung squamous cell carcinoma), STAD(stomach adenocarcinoma) compared to adjacent normal tissues.
In addition, we enrolled 410 LUAD patients from TCGA to explore the prognostic value of GPNMB and other clinical factors.The multivariable Cox regression analysis indicates that the OS of LUAD patients is signi cantly associated with GPNMB expression, T stage, N stage and stage (all P<0.05, gure 4).
Speci cally, N stage signi es regional lymph node metastasis, and the present study reveal that GPNMB expression level can impact the prognosis in STAD, and LUAD patients with lymph node metastasis.

GPNMB expression is associated with immune in ltration level in cancers
We used TIMER to explore the association between GPNMB expression and immune in ltration levels in 39 cancer types. Our study reveals that GPNMB expression is signi cantly related to tumor purity in 27 cancer types and signi cantly related to B cell in ltration levels in 25 cancer types. Besides, GPNMB expression is signi cantly related to CD8+ T cells in ltrating levels in 21 cancer types, CD4+ T cells in ltrating levels in 23 cancer types, macrophages in 29 cancer types, neutrophils in 28 cancer types, and dendritic cells in 32 cancertypes.
Then, we further investigated the relationship between GPNMB expression and immune in ltration in distinct cancers. Interestingly, our study reveals that  Figure 5F). The results above displayed that GPNMB is tightly related to immune in ltration in Breast, COAD, LUAD, LIHC, STAD. (Table III and Figure 6). Gene markers of monocyte such as CD86 and CD115 have strong relationships with GPNMB expression, and gene markers of TAM such as CCL2, CD68, IL10 have a moderate or weak correlation with GPNMB, and M1 macrophages gene markers such as IRF5 has a weak association with GPNMB expression, and gene markers of M2 macrophages such as CD163, VSIG4, MS4A4A have a moderate and strong association with GPNMB expression in STAD, and LUAD (Tables III, IV). Our study indicated that GPNMB expression level was signi cantly related to most of the immune cell's immune markers in STAD, and LUAD, especially monocytes, TAMs, M1, and M2 macrophages markers. Nevertheless, the GPNMB expression level was signi cantly associated with only 20 immune markers in LUSC (Table III).Specially, we revealed CD86, CD115 of Monocyte, CCL-2, CD68, IL10 of TAMs, IRF5 of M1 Macrophage, CD163, VSIG4 and MS4A4A of M2 phenotype are signi cantly correlate with GPNMB expression in STAD and LUAD (P < 0.05; Figure 6A-L). We further used the GEPIA database to verify the relationship between GPNMB expression and the immune markers of monocytes, TAMs, M1, and M2 macrophages markers in STAD, LUAD and LUSC, and the results were similar to those of TIMER (Table  IV). Our results indicate that GPNMB is associated with macrophage polarization in STAD and LUAD.

The relationship betweenGPNMB expression and immune markers T cells are comprised of T cells CD8+ lymphocytes, helper T cells, memory T cells, and T regulatory cells (FOXP3+). And functional T cells are comprised of T helper cells, Treg, and T cell exhaustion. Our study used TIMER and GEPIA databases to investigate the relationship between GPNMB and immune cells immune markers included CD8+ T cells, T cells (general), B cells, monocytes, TAMs, M1 and M2 macrophages, neutrophils, NK cells, DCs functional T cells and exhausted T cells in STAD and LUAD with the control of LUSC
Besides, GPNMB expression has a positive relationship with the DC in ltration level in STAD, and LUAD. DC immune markers such as HLA-DPB1,HLA-DQB1, HLA-DRA, HLA-DPA1, BDCA-1, BDCA-4, and CD11c are signi cantly related to GPNMB expression. In addition, our study signi ed that GPNMB had a positive relationship with T helper cells (Th1, Th2, Tfh, and Th17), Treg and T cell exhaustion immune markers such as T-bet, STAT4, STAT1, IFN-γ, TNF-α, GATA3,  STAT5A, IL13, BCL6, IL21, STAT3, FOXP3, CCR8, STAT5B, TGFb, PD-1, CTLA4, LAG3 and TIM-3 in STAD, and LUAD ( P<0.0001, Table III). In summary, our study signi es that GPNMB plays an essential role in the immune microenvironment of STAD, and LUAD.
Enrichment analysis of GPNMB related transcription factorsfunctional networks TTRUST database revealed that GPNMB related transcription factors were MITF and TP53 (Table V). Also, we used GeneMANIA to construct the transcriptiontarget network( gure 7). The functional analysis of the transcription-target network is mainly responsible for signal transduction in response to DNA damage,DNA damage response, signal transduction by p53 class mediator, mitotic G1 DNA damage checkpoint, G1 DNA damage checkpoint.

Discussion
GPNMB is a type I transmembrane glycoprotein and correlated with cancer growth and metastasis. GPNMB elevated in in ltrating immune cells such as macrophages, dendritic cells, CD14+ monocytes [16][17][18][19]. Besides, GPNMB can elevateT cells in the tumor microenvironment to promote cancer progression and metastasis [24][25][26]. In the present study, we reveal the association between GPNMB expression and various cancers prognosis. A higher level of GPNMB expression was associated wih poor prognosis in gastric cancer. A higher level of GPNMB expression was associated wih favorable prognosis in lung cancer. Also, our study signi es that GPNMB expression is associated with tumor in ltration level and immune cell markers of gastric cancer and lung adenocarcinoma.
In the present study, we used Oncomine and TIMER to explore the GPNMB expression in various cancer types and normal tissues. According to the oncomine database, GPNMB expression elevated in brain and CNS cancer, breast cancer, cervical cancer, esophageal cancer, gastric cancer, head and neck cancer, kidney cancer, leukemia, liver cancer, lung cancer, lymphoma, melanoma, pancreatic cancer, prostat-e cancer, and stroma; In contrast, GPNMB expression reduced in bladder, breast, colorectal, esophageal cancer, leukemia, lung, ovarian, and sarcoma cancers in some data sets ( Figure 1A). Based on TIMER which includes TCGA data, GPNMB expression signi cantly reduced in BLCA, BRCA, COAD, READ, LUAD, and UCEC compared to adjacent healthy tissues; while GPNMB expression elevated in CHOL, HNSC, KICH, KIRC, KIRP, LUSC, LIHC, and STAD compared to adjacent normal tissues( Figure 1B). Then, we used PrognoScan and Kaplan-Meier Plotter to analyze the correlation between GPNMB expression and cancer prognosis. PrognoScan showed that high GPNMB expression was signi cantly associated with poor prognosis in blood, brain, colorectal, breast, skin cancer, prostate cancers (Figures 2/3). Besides, the Kaplan Meier plotter database indicated that a higher level of GPNMB expression correlated with poor prognosis in stomach adenocarcinoma, LIHC cancer, breast cancer, and ovarian cancer, and correlated with favorable prognosis in lung adenocarcinoma. In summary, our study reveals that GPNMB could be a prognostic biomarker in blood, brain, colorectal, breast, LUAD, melanoma, prostate, ovarian, LIHC, and gastric cancer.
Besides, low expression of GPNMB is associated with prognosis in gender, stage1-2, T3-4 stage, N1-3 stage, M0 stage, Lauren classi cation, and differentiation of gastric cancer (Table II). The multivariable Cox regression analysis indicates that the OS of LUAD patients from TCGA is signi cantly associated with GPNMB expression, T stage, N stage and stage (all P<0.05, gure 4). We can infer that GPNMB expression level can impact the prognosis in STAD and LUAD patients with lymph node metastasis.
Previous studies reported that tumor-in ltrating lymphocytes' density and location were signi cantly related to sentinel lymph node status and impacted survival status in colorectal cancer and melanoma [36][37]. In the present study, we investigated the correlations between GPNMB expression and tumorin ltrating immune cells. The results indicate that GPNMB expression was correlated moderately or strongly with macrophages and DC and positively related to CD8+ T,CD4+ T cells, and neutrophils in BRCA, LIHC, LUAD, COAD, and STAD ( Figures 3A-D, F). Furthermore, we investigated the association between GPNMB expression and tumor-in ltrating immune cells markers in STAD and LUAD with control of LUSC. Gene markers of monocyte such as CD86 and CD115 had strong relationships with GPNMB expression, and gene markers of TAM such as CCL2, CD68, IL10 had a moderate or weak correlation with GPNMB, and M1 macrophages gene markers such as IRF5 had a weak association with GPNMB expression, and gene markers of M2 macrophages such as CD163, VSIG4, MS4A4A had a moderate and strong association with GPNMB expression (Tables III, IV). All the results suggest that GPNMB expression plays a critical part in regulating tumor-in ltrating monocyte and macrophages in STAD, and LUAD. Moreover, Our study also signi es that GPNMB has a positive relationship with T helper cells (Th1, Th2, Tfh, and Th17), Treg and Tcell exhaustion immune markers in STAD, and LUAD(P<0.0001, Table III). In summary, Our study signi es that GPNMB plays a vital role in the immune microenvironment of STAD and LUAD. In addition, we used TTRUST to explore GPNMB related transcription factors, which are MITF and TP53. GeneMANIA was used to construct a transcriptor-target network, which is mainly responsible for signal transduction in response to DNA damage, DNA damage response, signal transduction by p53 class mediator, mitotic G1 DNA damage checkpoint, G1 DNA damage checkpoint.
In melanoma, breast cancer, gliomas, prostate cancer, lung cancer, and bladder cancer, It was reported that high expression of GPNMB could promote cancer cell progression and invasion, and downregulate cancer cell apoptosis [24][25][38][39][40][41][42][43]. In colon cancer, tumor-in ltrating immune cells such as T cells could reduce tumor progression and the density, distribution of immune cells could haveadvantages in predicting survival status than TNM or Duke's classi cation  [20]. The previous study reported thatTh2 cells, T helper cell in ltration could produce antitumor immunity and impacted the survival status of cancer patients [48]. Therefore, the correlation between GPNMB and tumor immune in ltration could impact the prognosis in STAD and LUAD.

Conclusions
In summary, our study reveals that GPNMB expression could impact the prognosis of various cancer types such as lung adenocarcinoma, gastric cancer patients, and is positively related to immune cell ltrating level in various cancer types, especially in STAD, and LUAD. Moreover, GPNMB expression plays a vital part in regulating monocyte, macrophage polarization, and functional T cells in STAD, and LUAD. Therefore, GPNMB could be a potential biomarker for the tumor immune in ltration and prognosis prediction in STAD, and LUAD.

Declarations Acknowledgment
We sincerely thank the researchers for providing their databases information online, and it is our pleasure to acknowledge their contributions.
Funding Not applicable

Con icts of Interest
The authors have no con icts of interest to declare.
Ethical Statement: The current study received approval from the Xiangya Hospital of Central South University according to the Declaration of Helsinki. The retrieval of every data was collected from the open web, con rming that all of the written informed consents were attained. The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Author Contributions
Yang Wang conceived, collected, analyzed the data. Qiong Chen conceptualized, developed an outline for the manuscript, revised the manuscript and approved the nal version to be published. Yang Wang, Qiong Chen wrote the manuscript and approved the nal manuscript.

Consent for publication
This manuscript is approved by all authors for publication

Availability of data and materials
The datasets generated during our study are not publicly available but available on reasonable request.      Association between GPNMB expression and survival status of different cancer types in Prognoscan database.

Figure 4
Multivariable Cox regression analysis in LUAD patients from TCGA.  The LUSC acts as the control group, and GPNMB expression is not signi cantly related to macrophage polarization of LUSC.

Figure 7
Protein-protein interaction network of GPNMB related transcription factors networks (GeneMANIA). Protein-protein interaction (PPI) network and functional analysis, which indicate the gene enrichment in the target network of GPNMB related transcript factors. Different colors in the net-work edge signify the bioinformatics methods: physical interactions, co-expression, predicted, co-localization, pathway, genetic interactions and shared protein domains. The different colors of lines signify the enrichment genes biological functions.