Quantitative proteomic analysis showed significant differences between GC tissues and Peritoneal metastasis Tissues
GC and PM types have not been compared systematically at the tissue proteomic level, and the differentiation between PM and primary GC is of fundamental clinical importance for therapeutic stratification; we explored the possibility of identifying biomarkers for proteomic diagnosis. The experimental process of the strategy used to identify the differentially expressed proteins (DEPs) in PM tissues is shown briefly in Fig. 1A. To find meaningful protein alterations in GC and PM tissues, we selected five patients diagnosed with advanced GC who had not received any treatment before surgery, and surgically extracted their tumor tissues and PM tissues for proteome profiling. A total of 759302 spectra were generated, and 46787 peptides and 7638 proteins were identified at 1% false discovery rate (FDR). This assay identified 7638 proteins with quantitative information (Supplementary Table 1). For the experimental design with more than one replicate, proteins with a 1.5 fold change and a P-value less than 0.05 were defined as differentially expressed proteins (DEPs). Compared with the primary GC, 595 proteins in the PM tissues were upregulated among these DEPs (Supplementary Table 2), while 1050 proteins were downregulated (Supplementary Table 3). Figure 1B briefly summarizes the identification results for the samples.
For gene ontology (GO) enrichment analysis, Blast2GO software was used to evaluate the functional significance of all identified proteins. Supplementary Table S4 and Figure S1A provide detailed protein-specific information and the visualization results, respectively. GO enrichment analysis indicated that the most abundant biological processes (among 28 GO terms) mainly included: ‘cellular processes’, ‘metabolic processes’. The ‘cell’, ‘cell part’, and ‘organelle’ were the most enriched cell components (among 19 GO terms). The most enriched molecular functions (among 12 GO items) included: ‘binding’ and ‘catalytic activity’. Next, we performed GO enrichment analysis on the identified DEPs (Supplementary Table 5). We drew a GO functional classification map to illustrate all DEPs (Fig. 1C) and distinguish between upregulated and downregulated proteins (Fig. 1D). We found that both upregulated and downregulated DEPs are involved in common structural or functional processes. However, some GO biological processes and molecular functions were specifically enriched for upregulated proteins (such as cell aggregation, protein tag) or downregulated proteins (such as nitrogen utilization, hijacked molecular function) (Fig. 1D). Next, we used the database of protein orthologs classification (EuKaryotic orthologous group (KOG)), to predict the potential functions of all identified proteins (Supplementary Table 6 and Figure S1B). And among the identified DEPs (Supplementary Table 7 and Fig. 1E), the most enriched KOG category was “cellular process and signal transduction", which showed that DEPs were mainly related to signal transduction mechanisms, post-translational modification, and protein renewal. Next, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis on all identified proteins (Supplementary Table 8 and Figure S1C). Not only that, we further characterize the biologically functions involved involving these DEPs (Supplementary Table S8). The analysis showed that the upregulated proteins were related to 25 main pathways, of which ‘metabolic pathway’ was the most enriched category (Fig. 1F). Figure 1G shows the biological functions in the DEP set (Supplementary Table S9). The DEPs detected in our PM samples were mainly enriched in certain metabolic processes, especially biological processes related to lipid metabolism. Among them, ‘cholesterol metabolism’ was enriched in APOC2, LIPL, SORC2, IBP6, CD36, and APOA.
Correct protein sorting or protein targeting to the appropriate destination inside or outside the cell is essential; errors can cause illness. We predicted the subcellular localizations of the identified DEPs using bioinformatic tools (WoLF PSORT) (Fig. 1H). The analysis showed that intracellular, extracellular, and mitochondria were the most representative structures (see Supplementary Table 10 for the details). For the proteinprotein interaction network analysis, we used the STRING database (STRING 11.0) to analyze the DEPs to conduct network interaction analysis of protein-protein relationships within the first 100 confidence intervals (Figure I and Supplementary Table 11).
APOC2 was upregulated in patients with GC with PM
Through KEGG biological enrichment analysis, we found that the DEPs identified in our PM samples were mainly enriched in “regulation of lipid metabolic process”, “triglyceride catabolic process”, and “regulation of lipid catabolic process”, which ranked in the top 20 enriched biological processes. (Fig. 2A). The GSEA analysis also revealed that the DEPs identified in our PM samples were mainly enriched in “regulation of lipid metabolic process” related pathways (Fig. 2B). Therefore, we deduced that lipid metabolism might play an important role in PM. To further verify the association between the expression of lipid metabolism-related genes and the survival of patients with primary GC, we downloaded the latest transcriptome sequencing and clinical data of gastric cancer from The Cancer Genome Atlas (TCGA), including 375 cancer samples and 32 paracancerous samples. Then, we used the R package GSVA (version 1.36.2) to score a single sample of the TCGA Stomach Adenocarcinoma (TCGA-STAD) for this gene set. Next, we used the R package survminer (version 0.4.8) to determine the score to draw a survival curve (Fig. 2C). We found that patients with GC with high expression of lipid metabolism genes had a poor survival prognosis, which further indicated that lipid metabolism might influence GC progression. To search for the possible core protein of lipid metabolism in PM, we visualized the fold change of proteins in the gene set of “regulation of lipid metabolic process” using a volcano plot, and found that APOC2 is highly expressed in PM from GC (Fig. 2D).
To further explore the mechanism of PM in GCs, we first examined the expression of APOC2 in GC cell lines and found that APOC2 was highly expressed in most GC cell lines (especially MKN-45, SNU-16, BGC-823, and AGS). However, the expression of APOC2 in GES1 cells, which are epithelial cells of the gastric mucosa, was very low (Fig. 2E). Then, we performed western blotting using surgical samples to explore APOC2 levels. In GC tissue, APOC2 levels were significantly higher than in adjacent normal gastric mucosal tissues (ANTs) and even PM lesions showed higher levels of APOC2(Fig. 2F). Immunohistochemistry (IHC) staining demonstrated that APOC2 was upregulated in GC tissues compared with that in the ANTs, and notably, APOC2 was expressed at significantly higher rates in PM tissues (Fig. 2G). IHC staining was used to measure APOC2 levels in a cohort containing 111 pairs of GC samples and matched ANTs. The clinicopathological features and complete follow-up data of these patients are summarized in Table 1. The patients were divided into low (score 0–2) or high (score 3–9) APOC2 groups according to staining intensity scores. Correlation analysis showed that high APOC2 expression in GC tissues correlated significantly with more aggressive tumor phenotypes. In addition, the high expression of APOC2 in GC tissue was related to the shortened overall survival (OS) time of patients (Fig. 2H). And according to APOC2 levels, we also assessed OS based on other major clinicopathological factors. Significant differences existed in T3-T4, a higher degree of lymphatic metastasis (N2-N3) and clinical stage III + IV tumors (Fig. 2I-K).
Knockdown of APOC2 inhibited the invasion, migration, proliferation, apoptosis resistance, and EMT of GC cells
To explore APOC2’s biological functions in GC, we chose two GC cell lines, AGS and BGC-823, which express APOC2 highly, as our cellular models. AGS and BGC-823 cells were transfected with control siRNA and APOC2-siRNAs (Si1, Si2, and Si3). At 72 h after treatment, the APOC2 mRNA and protein levels were determined, and siAPOC2#1 was shown to be the most efficient in downregulating APOC2 levels in both cell lines (Fig. 3A-D). Wound-healing, migration, invasion, clonal formation, and apoptosis assays to evaluate the influence of siAPOC2#1 on AGS and BGC-823 cells showed that the knockdown of APOC2 significantly inhibited cell wound-healing (Fig. 3E, 3F), migration (Fig. 3G, 3H), invasion (Fig. 3G, 3H), clonal formation (Fig. 3I, 3J), and apoptosis resistance (Fig. 3K, 3L) compared with the siRNA control group. Moreover, APOC2 knockdown effectively increased the levels of E-cadherin, but decreased N-cadherin and vimentin, Snail, Slug, Twist1, MMP-9, and MMP-2 levels in AGS and BGC-823 cells (Fig. 3M, 3N). These results indicated that high APOC2 expression could enhance the malignant phenotype of GC cells.
RNA-seq analysis revealed that knockdown of APOC2 inhibited PI3K/AKT/mTOR signaling and reduced the metabolic activity of GC cells
To further explore the regulatory mechanism of APOC2 in GC, we downloaded the relevant GC sequencing data from the TCGA database for analysis. According to the level of APOC2 mRNA expression, we divided 375 GC tissue samples into two groups: APOC2_High (n = 188) and APOC2_Low (n = 187). Using R package limma to analyze the difference between the two sequencing datasets, we observed that APOC2 is mainly related to lipid absorption, transport, and metabolism of cells (Fig. 4A). AGS and BGC-823 cells were infected with lentivirus-shAPOC2 (to stably silence APOC2) and the control non-targeting shControl. Fluorescence (Fig. 4B) and western blotting (Fig. 4C) demonstrated the efficiency of lentiviral infection and knockdown in AGS and BGC-823 cells, respectively. We then performed transcriptome sequencing of AGS (n = 3) and AGS_shAPOC2 (n = 3) cells. The mRNA-seq analysis identified abundant gene expression changes after APOC2 knockdown. Especially, those related to metabolic pathways were downregulated, such as lipid metabolism, cholesterol efflux, HDL particle remodeling, low density lipoprotein (LDL) particle remodeling, regulation of sterol transport, and reverse cholesterol transport (Fig. 4D, 4E). We also noted changes various signaling pathways, such as cancer pathways and the PI3K-AKT signaling pathway (Fig. 4F).
To clarify the molecular mechanism by which APOC2 regulates these lipid metabolism processes, we performed Oil Red O and BODIPY 493/503 staining to test whether APOC2 affects the formation of lipid droplet (LDs) in GC cells. APOC2 knockdown significantly suppressed LD formation in AGS (Fig. 4G) and BGC-823 cells (Fig. 4H). APOC2 knockdown also markedly inhibited the TC (Fig. 3I, 3J) and TG (Fig. 3K, 3L) synthesis in GC cells. These results suggested that APOC2 plays a key role in lipid metabolism and transportation, thus we speculated that regulating its expression in GC cells would affect their bioenergetic profile. We used Seahorse XF to measure the live-cell metabolic index and found that Lenti-shAPOC2#1 cells (AGS and BGC-823) exhibit lower energetic metabolic phenotype compared with Lenti-shControl cells (Fig. 3M). There was a decrease in mitochondrial respiration rates assessed by the oxygen consumption rates (OCR) was observed in shAPOC2 AGS cells at their basal conditions (Fig. 4N). But compared with shCtrl cells, shAPOC2 cells have a lower extracellular acidification rate (ECAR) under stress conditions (Fig. 4N).
The PI3K/AKT/mTOR axis plays a positive role in EMT and promotes tumor metastasis (27, 28). The role of the mTOR complex in lipid metabolism is particularly important (29). We next demonstrated that knockdown APOC2 in AGS cells and BGC-823 cells decreased the levels of phosphorylated PI3K, AKT, and mTOR, but it barely affected total PI3K, AKT, and mTOR levels, which was consistent with the RNA-seq data (Fig. 3O, 3P). These data indicated that APOC2 is essential to regulate lipid metabolism and PI3K/AKT/mTOR signaling in GC cells.
The regulation by APOC2 of GC cell apoptosis, migration, invasion, proliferation and EMT was attributed to the PI3K/AKT/mTOR signaling pathway
Next, we explored whether APOC2 regulates the malignant phenotype through PI3K/AKT/mTOR signaling and clarified the signaling mechanism. Cell flow cytometry was used to test the effects of APOC2 knockdown and the cell-permeable PI3K pathway activator 740Y-P on AGS and BGC-823 cell apoptosis. Compared with the siCtronl group, AGS and BGC-823 apoptosis in the siAPOC2 group increased significantly, while that in the 740Y-P group decreased significantly. However, 740Y-P could partially eliminate the effect of si-APOC2 transfection on GC cell apoptosis (Fig. 5A, 5B). Wound healing, migration, invasion, and the colony formation ability of GC cells were greatly reduced in the APOC2 shRNA group and increased in the 740Y-P group, while 740Y-P restored the effects of APOC2 knockdown (Fig. 5C–H). Similarly, 3D invasion assays showed that APOC2 silencing repressed BGC-823 cell invasion significantly, and 740Y-P abolished the effect of sh-APOC2 transfection on GC cell invasion (Fig. 5I-K).
Based on the above experimental results, we preliminarily verified that APOC2 has an important regulatory role in EMT, thereby affecting the malignant characteristics of GC cells, possibly via the PI3K/AKT/mTOR signaling pathway. Next, we confirmed the relevant effectors in this pathway using western blotting. The levels of p-PI3K, p-AKT, and p-mTOR were significantly decreased in the APOC2-shRNA group compared with those in the shCtrl group. After transfection with Lenti-shAPOC2, E-cadherin, a marker of epithelial cells, was upregulated, and N-cadherin, vimentin, Snail, Slug, Twist1, MMP-9, and MMP-2 were downregulated in AGS and BGC-823 cells (Fig. 6A, 6B). But treatment with 740YP had the opposite effects (Fig. 6A, 6B).
Lentiviral transfection was used to construct a GC cell line stably overexpressing APOC2. The efficiency of infection and overexpression in AGS and BGC-823 cells were verified by fluorescence and western blotting (Fig. 6C, 6D). The levels of p-PI3K, p-AKT, and p-mTOR were enhanced in APOC2-overexpressing AGS and MGC-803 cells. Overexpression of APOC2 increased the levels of N-cadherin, vimentin, Snail, Slug, Twist1, MMP-9, and MMP-2 and decreased E-cadherin levels. GC cell treatment with LY294002, a selective inhibitor of PI3K-dependent AKT phosphorylation and kinase activity, inhibited PI3K-AKT-mTOR signaling and EMT of GC cells. Western blotting showed that APOC2 overexpression induced E-cadherin downregulation and N-cadherin, vimentin, Snail, Slug, Twist1, MMP-9, MMP-2, p-PI3K, p-AKT and p-mTOR upregulation in AGS and BGC-823 cells, which was inhibited significantly by LY294002 (Fig. 6E, 6F).