Charged multivesicular body protein 7 was identified as a prognostic biomarker correlated with metastasis in colorectal cancer

Abstract

BACKGROUND

Metastasis is the main reason leading to death in colorectal cancer (CRC) and about 25% of CRC patients developed metastasis when first diagnosed. Thus, unveiling biomarkers of CRC metastasis is of great significance.

AIM

To reveal biomarkers of CRC metastasis.

METHODS

Weighted gene co-expression network analysis was conducted to identify metastatic biomarkers in CRC through a systematic analysis of the GSE29621 dataset. Comprehensive validation was performed subsequently using publicly available datasets from The Cancer Genome Atlas and Gene Expression Omnibus and supplemented with experimental verification in CRC cell lines. Moreover, the identified hub gene charged multivesicular body protein 7 (CHMP7) was further subjected to clinical correlation analysis via Kaplan-Meier survival curves and Gene Set Enrichment Analysis to assess its prognostic significance and potential mechanistic involvement in CRC progression.

RESULTS

CHMP7 was identified as a key metastatic biomarker of CRC which displayed lower expression in CRC tissues, especially in CRC patients with metastasis and CRC cell lines with high metastasis potential. The expression of CHMP7 was significantly correlated with normal, metastatic tumor, pathologic stage, and lymphatic invasion (P < 0.05). CRC patients with higher expression of CHMP7 exhibited better overall survival. Besides, Gene Set Enrichment Analysis results showed that CHMP7 might be involved in metastatic related pathways.

CONCLUSION

Our results indicate that CHMP7 might be a prognostic biomarker correlated with CRC metastasis.

Key Words: Colorectal cancer; Weighted gene co-expression network analysis; Charged multivesicular body protein 7; Metastasis biomarker; Prognosis

Core Tip: Tumor metastasis is the main reason leading to death in colorectal cancer (CRC). Unveiling CRC metastasis related biomarkers is of great significance. In this study, charged multivesicular body protein 7 (CHMP7) was identified as a prognostic factor correlated with metastasis in CRC. CHMP7 showed a lower expression in CRC patients with metastasis and CRC cell lines with high metastasis potential. The expression of CHMP7 was significantly correlated with normal, metastatic tumor, pathologic stage, and lymphatic invasion (P < 0.05). CRC patients with higher expression of CHMP7 exhibited better overall survival. Besides, Gene Set Enrichment Analysis results showed that CHMP7 might be involved in metastatic related pathways.

INTRODUCTION

Colorectal cancer (CRC) is a common digestive tract malignant tumor which ranks the third and second for morbidity and mortality, respectively[1]. Although patients benefit from the standard treatment, surgery combined with chemotherapy, there are still some patients suffered local recurrence and distant metastasis after surgery[2]. Tumor metastasis is the leading cause of death in patients with CRC. About 25% of CRC patients developed metastasis by the time they were first diagnosed who exhibited poor prognosis with less than 5% of the 5-year survival rate[3]. Therefore, it is of great significance to uncover biomarkers related to CRC metastasis.

Weighted gene co-expression network analysis (WGCNA) can be applied to reveal the relationship between gene expression profile and phenotype[4]. Utilizing a systemic network analysis approach, WGCNA operates through pairwise correlation calculations of gene expression profiles. This methodology clusters co-expressed genes into biologically meaningful modules based on topological overlap measures, followed by rigorous correlation analysis between module eigengenes and clinical phenotypes to identify functionally relevant gene clusters. The unbiased module construction enables systematic discovery of gene networks associated with specific sample characteristics, while subsequent trait-module correlation analysis facilitates the identification of key molecular signatures underlying disease progression. By performing WGCNA, multiple disease biomarkers and therapy targets have been identified. For example, Guan et al[5] identified replication factor C subunit 4 as a key regulator of cell proliferation in nasopharyngeal carcinoma via WGCNA and in vitro and in vivo experimental validation. Huang et al[6] analyzed expression array data via WGCNA and discovered that NPNT was a metastatic associated target of liver cancer. Jiao et al[7] discovered potential biomarkers for early detection of CRC via bioinformatics analysis including WGCNA. Zheng et al[8] identified CHAF1B as one of the most important driver gene in lung squamous-cell carcinoma through WGCNA. Ma et al[9] identified 83 hub genes via WGCNA and further found cyclin dependent kinase inhibitor 3 as a key regulator of G2/M cell cycle. These studies imply that metastatic associated genes could be explored by performing WGCNA. In the present study, we applied WGCNA to reveal biomarkers involved in CRC metastasis. CHMP7 was identified as a biomarker of CRC metastasis via WGCNA, protein-protein interactions, gene expression analysis, gene enrichment, and survival curve analysis (Figure 1).

Figure 1 Flow chart of the design. WGCNA: Weighted gene co-expression network analysis; CRC: Colorectal cancer; PPI: Protein-protein interaction; CHMP7: Charged multivesicular body protein 7; TCGA: The Cancer Genome Atlas; N: Normal; T: Tumor; M: Metastatic tumor; GSEA: Gene Set Enrichment Analysis.

MATERIALS AND METHODS

Data collection

The microarrays and RNA-HTSeq data of CRC were downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) and The Cancer Genome Atlas (TCGA, https://cancergenome.nih.gov/), respectively. GSE29621 was used for WGCNA. GSE88839, GSE110223, and TCGA_GTEx_COAD from TCGA datasets were used to analyze the different expression of CHMP7 between normal tissues and cancer tissues. GSE14333, GSE72970, and TCGA_COAD datasets were used to verify different expression of CHMP7 between metastatic and non-metastatic CRC tissues.

WGCNA

R package of WGCNA was applied to construct gene co-expression network. Data from GSE29621 consists of 65 samples with 20460 genes. The genes with the top 25% variance were screened. A total of 5115 genes were obtained for WGCNA. To construct the net, a proper soft thresholding power was selected to establish an adjacent matrix according to the degree of connectivity. The proximity matrix and topological matrix were obtained according to the soft thresholding power. The topological overlap matrix was constructed on the basis of the topological matrix, a soft thresholding power of 4, and a minimal module size of 30. The obtained topological matrix was clustered by heterogeneity, and finally, the clustering tree was cut into different modules by using the dynamic shear method. The correlations between the modules and clinical traits were subsequently analyzed to identify the most significant module.

Module–trait relationship constructing

The eigengenes adjacency based on their correlation was calculated to assess the co-expression similarity of the identified gene modules. The correlation between module and clinical traits were analyzed by calculating the correlation between module eigengene and clinical traits. In brief, the modules constructed by WGCNA were first sequenced to calculate the module eigengene of each module and correlated with clinical traits[10]. P < 0.05 was statistically significant. Gene modules with high correlation with clinical traits were selected for further study.

Hub gene screening

The modules in the WGCNA that were highly correlated with the clinical traits of metastasis were selected, and the genes in the modules were imported into Cytoscape software for identifying hub genes. First, we selected the cluster with the highest score through MCODE plug-in base Cytoscape, and then adopted all the 12 algorithms in CytoHubba application, a plug-in of Cytoscape. Through the calculation and analysis of all 12 algorithms, the top 10 genes in at least 10 algorithms were selected as Hub genes.

The mRNA expression of CHMP7 in TCGA and GEO databases

Four GEO datasets were preprocessed by robust multi-array average method using R software as described previously[11]. In brief, background correction and standardization were performed firstly. Then, the expression value was calculated and the missing value was replaced to obtain the gene expression matrix. The expression matrix obtained from GEO and TCGA databases were analyzed using the R packages limma and DESeq2, respectively, to identify differentially expressed genes (DEGs) with a criterion of P < 0.05 and |fold change| > 2.

Copy number variation analysis of CHMP7 in CRC public datasets

The copy number variation of CHMP7 in CRC samples were detected via using cBioPortal. Data from the Colorectal Adenocarcinoma (TCGA, Firehose Legacy) study cohort was applied to analyze the copy number variation of CHMP7.

Protein expression of CHMP7 in public datasets and CRC cell lines

The protein expressions of CHMP7 in CRC public datasets were evaluated by using the University of Alabama at Birmingham cancer data analysis (UALCAN) online database with default settings (https://ualcan.path.uab.edu/index.html). Meanwhile, the protein expressions of CHMP7 were also detected in four CRC cell lines with different metastatic potential[12]. Cells were harvested by centrifugation and lysis buffer containing protease inhibitors was added to lyse cells. Then, equal amounts of samples were separated by sodium-dodecyl sulfate gel electrophoresis, transferred to nitrocellulose and immunoblotted with anti-CHMP7 (Cat No. 16424-1-AP) and anti-actin antibody (Cat No. 66009-1-Ig) from Proteintech Group, Inc in Wuhan, China. The blots were cut prior to hybridization with antibodies during blotting.

Survival analysis of CHMP7 in CRC patients

Gene Expression Profiling Interactive Analysis database was used for survival analysis. In the datasets selection module, colon adenocarcinoma (COAD) was selected. Samples were divided into high and low expression groups according to the CHMP7expression with median as the group cutoff. P < 0.05 was statistically significant.

Gene Set Enrichment Analysis

Gene Set Enrichment Analysis (GSEA) is a computational method that determines whether an a priori defined set of genes shows statistically significant, concordant differences between two biological states. GSEA was obtained from the Broad Institute (http://software.broadinstitute.org/gsea/index.jsp). The CRC related TCGA molecular profile data was input with the Gene Set Databases annotation file “c2. cp. kegg. V 7.4. symbols gene sets”. The cut-off values were predefined as false discovery rate < 0.25 and P < 0.05.

Nomogram predictive model

A nomogram is a prognostic model that is widely applied in oncology research[13]. On the basis of multiple regression analysis, the nomogram integrates multiple predictive indicators using scaled line segments and plots them on the same plane in a certain proportion to express the interrelationships between various predictive variables in the predictive model. By integrating various prognostic factors, nomogram can predict an individual’s probability of a clinical event, which fulfills our drive towards personalized medicine. In brief, the glm function was applied to construct a binary logistic model after data cleaning. The R package “rms” was subsequently utilized to construct nomogram related models.

Statistical analysis

Statistical analyses were performed with R software 4.0.2 and GraphPad Prism 8.3.0 software. Two-tailed Student’s t-test to analyze the data between groups. A P value < 0.05 was used to demonstrate a significant difference.

RESULTS

Construction of gene co-expression network and module detection with WGCNA

Gene expression matrix of GSE29621 was obtained through R software processing. Genes with top 25% variance (a total of 5115 genes) were obtained for WGCNA, including sample dendrogram and the heatmap of trait indicators (Figure 2A). Gene co-expression networks were subsequently constructed by using the WGCNA package on the basis of these 5115 genes. To construct a scale-free network, 4 was chosen as the soft threshold (β = 4, scale free R² = 0.85) (Figure 2B). A total of 21 modules were obtained with grey module representing non-coexpressed genes through one step construction and module detection (Figure 2C). Next, according to the relationships between different modules and metastatic traits, the modules related to CRC metastasis were selected for further study.

Figure 2 Network visualization. A: Samples dendrogram cluster and the trait heatmap; B: The appropriate soft threshold power = 4 was selected; C: Gene cluster tree.

Module–clinical trait relationships and validation

WGCNA can be used to analyze correlations between modules and clinical parameters. The module-clinical trait relationship was determined by summarizing the gene expression profile of the module eigengene, and then calculating the correlation between the module eigengenes and clinical parameters. To identify the module-clinical trait relationship of gene modules, we first categorized genes into the corresponding module according to the modules constructed. Modules with correlation coefficients (cor) > 0.2 and P values < 0.05 were identified as significant modules. The results revealed that the module significantly positively correlated with the metastatic tumor (M) stage was greenyellow (cor = 0.42, P = 5e-04) whereas the modules notably negatively correlated with the M stage included turquoise (cor = -0.34, P = 0.006), salmon (cor = -0.3, P = 0.02), and magenta (cor = -0.26, P = 0.04) (Figure 3A). The relationships between module membership and gene significance (GS) were subsequently analyzed to determine the modules related to metastatic clinical traits. As shown in Figure 3B, the absolute values of GS in the green yellow and salmon modules were the highest among all the modules, which had positive and negative relationships with CRC metastasis, respectively (Figure 3B). To identify key genes negatively associated with CRC metastasis, the salmon module was chosen for the analysis between module membership and GS with P = 1.5e-07 and cor = 0.53 (Figure 3C).

Figure 3 Salmon module was identified as the most negatively correlated module of colorectal cancer metastasis. A: Heat map of the correlation between modules eigengenes and clinical traits. The numbers in the upper part of the square represent the correlation coefficient. While the numbers in parentheses represents the P value. Red color represents the positive correlation, and blue represents the negative correlation; B: Gene significance across modules; C: A scatter plot of the gene significance for colorectal cancer M stage the module membership in the salmon module.

CHMP7 was identified as the hub gene of CRC metastasis

The salmon module was chosen as the pivot module which was significantly negatively correlated to the metastatic clinical traits. Subsequently, all the 86 genes in the salmon module were input into software Cytoscape to visualize their interactions (Figure 4). Two genes, CHMP7 and N-acetyltransferase 1 (NAT1), were identified as hub genes by using the plug-in CytoHubba. NAT1 has been reported as a prognostic biomarker and could restrain cell proliferation via regulating phosphatidylinositol 3-kinase/Akt/mechanistic target of rapamycin kinase signaling pathway in COAD[14,15]. While the role of CHMP7 in COAD has not been reported. Therefore, CHMP7 was selected as the hub gene in the following analysis.

Figure 4 The interaction network diagram of all the 86 genes in salmon module. Nodes represent genes and a darker color indicates a higher number of connected nodes for the gene.

CHMP7 exhibited a decreased expression in metastatic CRC

Subsequently, the expression pattern of CHMP7 in CRC was investigated. Firstly, the mRNA expression of CHMP7 in CRC and normal tissue was evaluated via analyzing GEO and TCGA data. The results displayed that CHMP7 was down-regulated in tumor tissues compared with normal tissues (Figure 5A). Especially, the expression of CHMP7 in metastatic CRC was much lower than that in non-metastatic CRC (Figure 5B). Moreover, CRC patients in advanced stage showed a much lower expression of CHMP7 than patients in early stage (Figure 5C). Furthermore, to explore the potential mechanism of low mRNA expression of CHMP7 in CRC, copy number variation of CHMP7 in TCGA COAD was analyzed by utilizing Tumor Immune Estimation Resource online database. The results indicated that 6% of CRC patients displayed a copy number deletion of CHMP7 (Figure 5D).

Figure 5 The mRNA of charged multivesicular body protein 7 was down-regulated in colorectal cancer. A: The expression of charged multivesicular body protein 7 (CHMP7) between tumor and normal tissues in colorectal cancer datasets from Gene Expression Omnibus and TCGA_GTEx_COAD; B: The expressions of CHMP7 between metastatic and non-metastatic colorectal cancer samples in datasets from Gene Expression Omnibus and TCGA_COAD; C: The correlation between CHMP7 expression and tumor stage; D: Copy number variations of CHMP7 in TCGA_COAD samples. ^aP < 0.05, ^cP < 0.001, ^dP < 0.0001. CHMP7: Charged multivesicular body protein 7; TPM: Transcripts per million; N: Normal; T: Tumor; M: Metastatic tumor; TCGA: The Cancer Genome Atlas; COAD: Colon adenocarcinoma.

Correspondingly, the protein expression of CHMP7 in CRC was detected by using public datasets and CRC cell lines with different metastatic potential. By analyzing proteomic data from CPTAC samples, we found that the protein expression of CHMP7 was also down-regulated in CRC tumor compared with normal tissues (Figure 6A). Patients in advanced stages showed a decrease of CHMP7 protein expression (Figure 6B). Interestingly, it demonstrated that CRC patients with WNT pathway alteration showed a significant lower expression of CHMP7 (Figure 6C). Especially, CHMP7 showed a lower expression in CRC cell lines with high metastatic property (HCT116 and SW620) than cell lines with low metastatic potential (HCT-15 and SW480) (Figure 6D).

Figure 6 The protein expression of charged multivesicular body protein 7 was down-regulated in colorectal cancer. A: The protein expression of charged multivesicular body protein 7 (CHMP7) between tumor and normal tissues in colorectal cancer datasets from TCGA_COAD; B: The correlation between protein expression of CHMP7 and tumor stage; C: The protein expression of CHMP7 between patients with WNT pathway alteration and non-alteration; D: The protein expression of CHMP7 in colorectal cancer cell lines with different metastatic potential. ^aP < 0.05, ^cP < 0.001. CHMP7: Charged multivesicular body protein 7; N: Normal; T: Tumor.

Correlations between CHMP7 expression and clinical characteristics of CRC patients

To further verify the correlation between CHMP7 expression and clinical trait of CRC, RNA-seq data with clinical information of CRC from TCGA was analyzed. We found that CHMP7 expression was significantly correlated with pathologic stage, normal (N), M stage, and lymphatic invasion (Table 1, P < 0.05). Compared with patients in stage I and II, CHMP7 was significantly down-regulated in stage III and IV. Meanwhile, CHMP7 was significantly down-regulated in patients with N1 and N2 compared with N0. Moreover, CHMP7 was down-regulated in patients with metastasis and lymphatic invasion compared with patients without metastasis and lymphatic invasion. Subsequently, the relationship between CHMP7 expression and patients’ survival was estimated. The results indicated that patients with higher expression of CHMP7 had a prolonged survival time compared with patients with lower expression of CHMP7 (Figure 7A). The effects of CHMP7 expression and M stage on patients’ survival in CRC were assessed via nomogram by constructing a cox proportional risk regression model (P < 0.05). The results showed that patients with higher expression of CHMP7 and lower grade of metastasis have better survival (Figure 7B). These results implied that CHMP7 might be a prognostic biomarker for CRC.

Figure 7 The correlations between charged multivesicular body protein 7 expression and colorectal cancer overall survival. A: The high expression of charged multivesicular body protein 7 was positively correlated with better survival in colorectal cancer patients; B: A nomogram was established to predict the risk score and survival probability of colorectal cancer patients. CHMP7: Charged multivesicular body protein 7; TPM: Transcripts per million; M: Metastatic tumor; HR: Hazard ratio.

Table 1 Correlations between charged multivesicular body protein 7 expression and patient clinical characters.

Characteristic	n	Expression of CHMP7	P value
Age, years			0.709
≤ 65	194	3.29 ± 0.54
> 65	284	3.28 ± 0.49
Gender			0.126
Female	226	3.32 ± 0.49
Male	252	3.25 ± 0.52
Pathologic stage			< 0.001
I and II	268	3.39 ± 0.48
III and IV	199	3.14 ± 0.50
N stage			< 0.001
N0	284	3.37 ± 0.50
N1 and N2	194	3.16 ± 0.49
M stage			< 0.001
M0	349	3.30 ± 0.49
M1	66	3.03 ± 0.49
Lymphatic invasion			0.005
No	266	3.35 ± 0.50
Yes	168	3.21 ± 0.48

CHMP7: Charged multivesicular body protein 7.

GSEA analysis

To obtain further insights into the potential mechanism of CHMP7 in CRC, GSEA was performed to explore pathways that CHMP7 might be involved. CRC samples from the TCGA database were divided into high and low CHMP7 expression groups. The DEGs between these two groups were subsequently obtained via the R package DESeq2. Furthermore, the DEGs were input into GSEA for pathway enrichment analysis. The results showed that CHMP7 might be involved in epithelial to mesenchymal transition in CRC and in canonical WNT signaling and CRC (Figure 8).

Figure 8  Potential mechanism of charged multivesicular body protein 7 on colorectal cancer tumorigenesis based on Gene Set Enrichment Analysis.

DISCUSSION

Due to the lack of obvious symptoms and effective detection methods in the early stage of CRC, a large proportion of CRC patients are already in advanced stages when they were diagnosed for the first time. About half of the CRC patients progressed to metastasis for their initial visits[2]. Therefore, unveiling metastatic biomarkers of CRC is of great importance. WGCNA is a powerful, versatile tool for gene expression analysis, offering robust network modeling, integration with diverse data types, and biologically meaningful insights. Its emphasis on modular structures, hub genes, and clinical correlations makes it widely applicable in genomics and systems biology research. In the present study, WGCNA was performed to explore genes associated with CRC metastasis (Figure 2). A total of 21 modules were uncovered to be correlated with CRC metastasis among which the salmon module exerted as the strongest negatively correlated module with CRC metastasis (Figure 3). Further insights into the salmon module led to the discovery of 86 genes (Figure 4). Among these 86 genes, 12 genes were identified to be associated with CRC metastasis, including potassium channel tetramerization domain containing 9[16], glutathione-disulfide reductase[17], elongator acetyltransferase complex subunit 3[18], NAT1[19], damage specific DNA binding protein 2[20], claudin 23[21], coiled-coil domain containing 25[22], solute carrier family 6 member 14[23], PDZ binding kinase[24], frizzled class receptor 3[25], four and a half LIM domains 2[26], and chromogranin A[27]. Our results confirmed the previous studies which implies that our strategy is reasonable.

CHMP7 is an endoplasmic reticulum-localized protein. By forming a complex with endosomal sorting complex required for transport-III, CHMP7 could regulate the endosomal sorting pathway and nuclear envelope reformation[28,29]. Besides, the oligomerization of CHMP7 could mediate three-way ER junctions and the interactions between endoplasmic reticulum and mitochondria[30]. CHMP7 was also reported to regulate the neurodevelopment of ADHD[31]. Recently, the role of CHMP7 in tumor progression was explored. Genomic analysis identified CHMP7 as a therapeutic and prognostic biomarker for endometrial carcinoma[32]. CHMP7 was also identified as a biomarker for immunotherapy and chemotherapy in CRC patients[33]. Downregulation of CHMP7 via siRNA transfection inhibited the cell proliferation, migration, and invasion of esophageal cancer cells[34]. In the present study, we found that CHMP7 was significantly downregulated in CRC tumor tissues (Figures 5A and 6A), especially in metastatic CRC tumors (Figure 5B). The expression of CHMP7 was significantly negatively associated with tumor metastasis (N and M stage) and lymphatic invasion (Table 1). CRC cell lines with high metastatic potential showed a decreased CHMP7 protein expression (Figure 6D). These results suggest that CHMP7 is a biomarker of CRC metastasis.

WNT signaling is crucial for maintaining the normal function of the intestinal epithelium. Many key components of WNT signaling, such as APC, WNT3A, and β-catenin, have been confirmed to be involved in CRC metastasis[35]. Interestingly, we found that CRC patients with WNT pathway alterations presented lower protein expression of CHMP7 than patients without WNT pathway alterations (Figure 6C). The GSEA results also indicated that CHMP7 was involved in canonical WNT signaling and CRC (Figure 8). These results implied that the expression of CHMP7 was negatively correlated with WNT signaling pathway. Subsequently, based on the expression profiles of four CRC cell lines (SW620, SW480, HCT-15, and HCT116), we conducted an expression correlation analysis between CHMP7 and key WNT pathway proteins. The results indicated that CHMP7 was positively correlated with the expression of WNT pathway-inhibiting proteins (glycogen synthase kinase 3 beta, APC, beta-transducin repeat containing E3 ubiquitin protein ligase) and negatively correlated with WNT pathway-activating proteins (WNT5A, catenin beta 1) (Supplementary Table 1). Especially, CHMP7 showed a significant positive correlation with the expression of glycogen synthase kinase 3 beta. These results further validate the negative correlation between CHMP7 and the WNT pathway. There are also limitations for the present study. The expression of CHMP7 needs to be validated in local CRC specimens. Especially, CRC tumor tissues with or without metastasis were needed.

CONCLUSION

Our results indicate that CHMP7 is a prognosis biomarker in CRC and might be involved in tumor metastasis.