Quantitative proteomics identified 3 oxidative phosphorylation genes with clinical prognostic significance in gastric cancer

Abstract The aim of the present study was to explore the underlying mechanisms involved in gastric cancer (GC) formation using data‐independent acquisition (DIA) quantitative proteomics analysis. We identified the differences in protein expression and related functions involved in biological metabolic processes in GC. Totally, 745 differentially expressed proteins (DEPs) were found in GC tissues vs. gastric normal tissues. Despite enormous complexity in the details of the underlying regulatory network, we find that clusters of proteins from the DEPs were mainly involved in 38 pathways. All of the identified DEPs involved in oxidative phosphorylation were down‐regulated. Moreover, GC possesses significantly altered biological metabolic processes, such as NADH dehydrogenase complex assembly and tricarboxylic acid cycle, which is mostly consistent with that in KEGG analysis. Furthermore the higher expression of UQCRQ, NDUFB7 and UQCRC2 were positively correlated with a better prognosis, implicating these proteins may as novel candidate diagnostic and prognostic biomarkers.


| INTRODUC TI ON
Globally, gastric cancer (GC) ranks the third among all types of malignancies in terms of cancer-associated mortality. 1 Despite the recent advances in diagnostics and therapeutics, the survival of patients with GC shows no significant improvement. GC is a multifaceted disorder and involves multiple factors, complicated biological processes and unpredictable outcomes. 2 It is unreasonable that a single molecular marker could be accurately used for the prediction, prevention and personalized medicine (PPPM) practice in GC.
Moreover, numerous molecular alterations at various levels, including DNA (genome), RNA (transcriptome), proteins (proteome) and metabolites (metabolome), have been demonstrated to participate in the progression of GC and are intertwined with diverse pathway networks. Recent studies have revealed that high-throughput proteomics renders a promising approach to understand the progression of complex tumorigenesis and to identify potent cancer biomarkers. 2 Proteomics is considered as a state-of-the-art, largescale and systematic analytic approach, enabling the identification and quantification of proteins in specific biological specimens. 3 Additionally, the difference between normal and lesion samples can be quantified using quantitative proteomics and this offers valuable evidence for identifying novel biomarkers and elucidating mechanisms underlying complicated biological processes. 4 A comprehensive molecular characterization of gastric adenocarcinoma has been studied in 2014. 5 Moreover, Mun et al 6 and Ge et al 7 reported the proteomic landscape of early-onset and diffuse-type GC, respectively. However, there are few studies of GC by using data-independent acquisition mass spectrometry (DIA-MS).
Data-independent acquisition technology accurately quantifies proteins and is suitable specifically for biomarker research. To the best of our knowledge, the major objective of cellular biology is to understand the gene expression strategy under changing microenvironment in cancer development and progression. In the present study, we aimed to discover effective biomarkers of GC.
Over the past decade, many studies have elucidated the function of mitochondria in tumour cells, Zong et al 8 which is suggestive that compromised mitochondrial bioenergetic function, along with changed phenotype, is a distinct feature of carcinogenesis. 9 Moreover, the above outcomes have been validated in numerous malignancies in the past few years. 8 Nevertheless, the property and origin of respiratory and metabolic alterations in GC cells remain unclear.

| Patients and tissue samples
For the proteomics and western blot analysis, 10

| Sample preparation
Samples were ground and lysed in the lysis buffer (8 mol/L Urea, 100 mmol/L Tris Hydrochloride, pH 8.0) supplemented with protease and phosphatase inhibitors (ThermoFisher Scientific) followed by 2 minutes of sonication (3 seconds on and 3 seconds off, amplitude 25%). Samples were centrifuged at 15 000 g for 10 minutes to collect total protein, followed by the assessment of protein concentration by Bradford protein assay (ThermoFisher Scientific).

| Protein digestion
Protein (100 μg per sample) and 8 mol/L urea (100 μL) were transferred to a new Eppendorf tube, followed by the addition of 2 μL of 0.5 mol/L trichloroethylene (TCE) and incubated at 37°C for 1 hours; next, 4 μL of 1 mol/L iodoacetamide was added to the tubes and incubated at 24°C for an additional 40 minutes. Five volumes of pre-chilled (−20°C) acetone were then added for protein precipitation overnight at −20°C.
The next day, precipitates were rinsed twice with 1 mL pre-chilled 90% acetone aqueous solution, followed by dissolution in 100 μL of 100 mmol/L tetraethylammonium bromide (TEAB). Sequence grade modified trypsin (Promega, Madison, WI) was added at a weight ratio of

| Quantification of proteins using DIA mass spectrometry
The Spectronaut X based on the extensive mass calibration. The ideal extraction window was determined by Spectronaut X, which dynamically relied on iRT calibration and gradient stability. A Q value cut-off of 1% was applied on precursor and protein levels. Decoy generation was set to mutate to apply a random number of AA position swamps (min = 2, max = length/2). All selected fragment ions passing the filters were used for quantification. The average top 3 filtered peptides, which passed the 1% Q value cut-off, were used to calculate the major group quantities.

| Validation of differentially expressed proteins (DEPs) by TCGA data sets
The prognostic value of key proteins was confirmed by Kaplan-Meier (KM) survival analysis, 10  F I G U R E 3 WGCNA analysis of the GC proteome profiling. A, Samples clustering were conducted to detect outliers between the gastric cancer (T) and adjacent tissues (P). B, Cluster dendrogram was generated by hierarchical clustering to show the modules of highly interconnected groups of genes between T and P groups. C, Heatmap was used to shown the correlation coefficient of module-traits. D, KEGG pathway analysis of proteins in turquoise

| Western blot
First, the tissues were sufficiently ground and total protein were extracted by RIPA buffer (Applygen) with a protease inhibitor cocktail and a phosphatase inhibitor cocktail and followed by western blot analysis.The following primary antibodies were used for western blot analysis: Anti-UQCRQ antibody (ab241991; 1:1000), Anti-NDUFB7 antibody (ab188575; 1:1500), Anti-UQCRC2 antibody (ab203832; 1:800). GAPDH was used to be the control of cytoplasm proteins. The above antibodies were purchased from Abcam. After primary antibody detection, membranes were incubated with the appropriate goat anti-rabbit or anti-mouse secondary antibody (Abcam). All uncropped scans for blots were presented in corresponding Source Data file.

| Statistical analysis
Unpaired Student's t test was employed to perform two-group comparison. Welch's ANOVA Test was utilized to analyse the difference between the T and P groups. A P-value < .05 as well as |fold change| >2 were used to filter DEPs.

| DIA proteomics profiling and DEPs in GC
To identify the candidate biomarker of GC, we performed a proteome-wide screening experiment. In this study, 20 samples, including ten GC tissues and ten adjacent non-tumourous tissues, were selected and subjected to DIA proteome analysis ( Figure 1A).
DDA analysis was performed for establishing a spectral library.

| Functional classification of DEPs
All DEPs, including PLXDC2 and GPX8, were analysed using bioinformatics. GO enrichment analysis was performed for these DEPs in three aspects, including biological process (Figure 2A), cellular component ( Figure 2B) and molecular function ( Figure 2C). As shown in the bubble diagrams, relevant DEPs mostly altered biological metabolic processes, such as mitochondrial electron transport, mitochondrial respiratory chain complex I assembly, NADH dehydrogenase complex assembly and tricarboxylic acid (TCA) cycle (Figure 2A), and these processes were significantly altered in GC. Interestingly, these biological processes are mainly related to oxidative phosphorylation (OXPHOS). In terms of cellular component, majority of these DEPs are located in mitochondria, including the mitochondrial membrane, mitochondrial protein complex and respiratory chain complex ( Figure 2B). The GC tissues are involved in electron transfer activity, coenzyme binding and oxidoreductase activity ( Figure 2C).

| Weighted gene co-expression network analysis of GC proteome profiling
To search the co-expression modules of genes and the relationship between modules and pathogenesis of GC, WGCNA was applied on our proteome landscape of T and P through the R package WGCNA. Firstly, sample clustering was performed to detect variation and outliers of all 20 data sets. As shown, no outlier was identified and all of the samples were used for next step analysis ( Figure 3A). Then the Pearson's correlation coefficients were calculated for genes in a pairwise manner, when threshold was set as 5, respectively. Using the matrix above, average linkage hierarchical clustering was performed on the 20 data sets to identify the densely interconnected gene modules ( Figure 3B). As a result, one module labelled with turquoise was significantly co-expressed (P = .001), and the correlation coefficient between tumour/normal was −0.67 ( Figure 3C), which rendered an important function of genes for GC process in the module. To detect the fundamental function of the turquoise module, KEGG pathway analysis was used. As a result, genes in turquoise module were mainly related to oxidative phosphorylation, citrate cycle (TCA cycle) and so on ( Figure 3D).

| Pathway enrichment along with network analysis displayed relevant pathways of GC
KEGG database was searched for pathway enrichment analysis.
By analysing DEPs, 38 pathways were significantly changed in GC tissues compared to those in adjacent non-tumourous tissues.
The significantly enriched top 20 ranking KEGG pathways of the DEPs are shown in Figure 4A. Coincidentally, all the identified F I G U R E 4 The pathway analysis of the differentially expressed proteins between the gastric cancer (T) and adjacent tissues (P). A, The significantly enriched top 20 ranking KEGG pathway of the DEPs. B, The oxidative phosphorylation signalling pathway (P-value = 6.2e−49) played a crucial role in pathogenesis of GC. The green indicates the down-regulated protein expression in the gastric cancer group. C, PPI network analysis of DEPs involved in oxidative phosphorylation by STRING database. D, The hub network of oxidative phosphorylation DEPs involved in OXPHOS were down-regulated. As shown in the pathway network, OXPHOS was significantly down-regulated in GC tissues compared to that in adjacent tissues ( Figure 4B). The interaction and correlation networks of the 76 DEPs involved in OXPHOS are shown in Figure 4C. For an in-depth investigation into the DEPs in GC tissues, a hub subnetwork was constructed from the main network through the Cytohubba plugin ( Figure 4D), which the TCGA database would use to verify the proteins for candidate markers.

| Gene set enrichment analysis (GSEA) confirmed down-regulation of oxidative phosphorylation in GC
Despite the crucial role of specific DEPs identification in an in-  Figure 4B,D, respectively.

| UQCRQ, NDUFB7 and UQCRC2 may be diagnostic biomarkers in GC
Based on the hub subnetwork of OXPHOS ( Figure 4D), we successfully retrieved the mRNA expression data of 10 DEPs from TCGA data sets using starBase, which contained 375 GC and 32 normal samples. The mRNA expression levels of three genes were significantly down-regulated in GC, including UQCRQ (P-value = .0012, Figure 6A), NDUFB7(P-value = 1.0e−6, Figure 6B) and UQCRC2 (P-value = .0064, Figure 6C). We also analysed the expression level of UQCRQ, NDUFB7 and UQCRC2 protein in the above 10 pairs of GC tissues and corresponding adjacent tissues by Western blot.
Compared with adjacent tissues, the expression level of UQCRQ, NDUFB7 and UQCRC2 protein in GC tissues were significantly reduced ( Figure 7).
Next, we assessed the prognostic value of these three proteins.

| D ISCUSS I ON
GC ranks the fifth and third in morbidity and cancer-associated mortality in all types of malignant tumours globally. 13 Therefore, developing novel diagnostic and prognostic biomarkers of GC is essential. In recent years, quantitative proteomics analyses have been increasingly applied to provide new insights for the treatment of GC.
Recently, the mitochondria have received considerable attention in cancer research. Mitochondria have been extensively investigated to address carcinogenesis due to their diverse and complicated functions, including biosynthetic metabolic modulation, cellular signalling and cell death, in addition to their role in energy production. 8 Emerging evidence has revealed that altered protein and number of mitochondria, 8  tumorigenesis. In addition, we found that Ndufb7 is significantly down-regulated in GC.
UQCRQ, a subunit of ubiquinol-cytochrome c reductase complex III, part of the mitochondrial respiratory chain, encodes a ubiquinone-binding protein of low molecular mass. In a previous study, mitochondrial dysfunction caused by UQCRQ defection has been reported to be involved in the pathogenesis of ulcerative colitis. 17 UQCRC2 is a part of the ubiquinol-cytochrome c reductase complex, also called complex III. UQCRC2, a key subunit of mitochondrial respiratory complex III, plays an important role in maintaining the structural and functional integrity of the mitochondria. 8 Warburg first demonstrated that aerobic glycolysis was employed by tumour cells with rapid proliferation, which exerted an irreversible injury on OXPHOS. 18 Additionally, excessive mitochondrial ROS generation could promote carcinogenesis and tumour progression. 8 Mitochondrial dysfunction has been prevalently reported in tumour cells. Nevertheless, the notion that both mitochondrial metabolism and glycolysis are involved in tumour cells is controversial and widely challenged, 8 and mitochondrial function plays a decisive role in the maintenance of cancer. 19 Therefore, there have been various assessments and discussions over the changes in mitochondrial function and protein expression in different human malignancies. 8  to be significantly down-regulated in GC. However, the increased expression was positively related to a better prognosis, indicating that these proteins might be promising diagnostic and prognostic biomarkers for GC. Nonetheless, these results were obtained by bioinformatics analysis and require further validation.

ACK N OWLED G EM ENTS
We are grateful to the laboratory staff in the Key Laboratory

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

CO N S E NT
All the patients signed the informed consent.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used in this study are available from the corresponding authors upon request.