Unraveling Molecular Differences of Gastric Cancer by Label-Free Quantitative Proteomics Analysis

Gastric cancer (GC) has significant morbidity and mortality worldwide and especially in China. Its molecular pathogenesis has not been thoroughly elaborated. The acknowledged biomarkers for diagnosis, prognosis, recurrence monitoring and treatment are lacking. Proteins from matched pairs of human GC and adjacent tissues were analyzed by a coupled label-free Mass Spectrometry (MS) approach, followed by functional annotation with software analysis. Nano-LC-MS/MS, quantitative real-time polymerase chain reaction (qRT-PCR), western blot and immunohistochemistry were used to validate dysregulated proteins. One hundred forty-six dysregulated proteins with more than twofold expressions were quantified, 22 of which were first reported to be relevant with GC. Most of them were involved in cancers and gastrointestinal disease. The expression of a panel of four upregulated nucleic acid binding proteins, heterogeneous nuclear ribonucleoprotein hnRNPA2B1, hnRNPD, hnRNPL and Y-box binding protein 1 (YBX-1) were validated by Nano-LC-MS/MS, qRT-PCR, western blot and immunohistochemistry assays in ten GC patients’ tissues. They were located in the keynotes of a predicted interaction network and might play important roles in abnormal cell growth. The label-free quantitative proteomic approach provides a deeper understanding and novel insight into GC-related molecular changes and possible mechanisms. It also provides some potential biomarkers for clinical diagnosis.


Introduction
Gastric cancer (GC) has long been recognized as among the world's major malignancies, ranked fifth in incidence and third in mortality since 2012 [1][2][3]. In China, approximately 405,000 new cases and 325,000 deaths from GC have been reported, making it the second most prevalent disease and the third in cancer-related deaths [2].
GC is a complex and multi-factorial process disease, resulting from the interaction between genetic and environmental factors that potentially deregulate cell oncogenic signaling pathways to promote

Overall Protein Changes Identified by Label-Free Quantitative Strategy
To quantitatively compare GC tissue proteome, proteins were extracted from surgically resected fresh cancer and adjacent tissues of three patients, randomly selected from ten patient samples. After in-solution tryptic digestion, each of three samples were run in triplicate by LC-MS/MS, and a total of 3639 and 3543 proteins were identified by Proteome Discoverer 1.4 in three pairs of GC and adjacent tissues, respectively ( Figure 1A and Table S1, Supplementary Material). After evaluating MS data quality, all three samples were analyzed with Progenesis LC-MS software, using an algorithm based on the pair-wise features detection at LC-MS level. 726, 662 and 662 dysregulated proteins (ě2-fold) were quantified in three GC patients ( Figure 1B and Table S1, Supplementary Material), while 146 dysregulated proteins were reliably quantified after limited selection in these proteins. Of 146 proteins, 81 proteins were downregulated, 65 proteins were upregulated (adjacent/tumor ratio ě2-fold or ď0.5-fold, p-value < 0.05) and 22 proteins were first found to be related with GC (

Overall Protein Changes Identified by Label-Free Quantitative Strategy
To quantitatively compare GC tissue proteome, proteins were extracted from surgically resected fresh cancer and adjacent tissues of three patients, randomly selected from ten patient samples. After in-solution tryptic digestion, each of three samples were run in triplicate by LC-MS/MS, and a total of 3639 and 3543 proteins were identified by Proteome Discoverer 1.4 in three pairs of GC and adjacent tissues, respectively ( Figure 1A and Table S1, Supplementary Material). After evaluating MS data quality, all three samples were analyzed with Progenesis LC-MS software, using an algorithm based on the pair-wise features detection at LC-MS level. 726, 662 and 662 dysregulated proteins (≥2-fold) were quantified in three GC patients ( Figure 1B and Table S1, Supplementary Material), while 146 dysregulated proteins were reliably quantified after limited selection in these proteins. Of 146 proteins, 81 proteins were downregulated, 65 proteins were upregulated (adjacent/tumor ratio ≥2-fold or ≤0.5-fold, p-value < 0.05) and 22 proteins were first found to be related with GC (Table S1, Supplementary Material). identified from three cases in tumor or adjacent tissues respectively. One thousand seven hundred forty-four proteins identified appear in both tumor and adjacent tissues; (B) The number of proteins with more than twofold differential expression in three cases, respectively, and the number of proteins shared in two or three cases.

Functional Annotation of Proteins between GC and Adjacent Tissues
To reveal any possible function of the 146 dysregulated proteins, in silico analysis was performed. Ingenuity Pathway Analysis (IPA) aids in the integration of complex omics data and provides insight into regulatory mechanism and biological functions based on published studies [21]. The heatmap of "disease and function" of 146 dysregulated proteins by IPA was shown in Figure 1. Venn diagrams of total identified proteins and dysregulated proteins. (A) Total proteins identified from three cases in tumor or adjacent tissues respectively. One thousand seven hundred forty-four proteins identified appear in both tumor and adjacent tissues; (B) The number of proteins with more than twofold differential expression in three cases, respectively, and the number of proteins shared in two or three cases.

Overall Protein Changes Identified by Label-Free Quantitative Strategy
To quantitatively compare GC tissue proteome, proteins were extracted from surgically resected fresh cancer and adjacent tissues of three patients, randomly selected from ten patient samples. After in-solution tryptic digestion, each of three samples were run in triplicate by LC-MS/MS, and a total of 3639 and 3543 proteins were identified by Proteome Discoverer 1.4 in three pairs of GC and adjacent tissues, respectively ( Figure 1A and Table S1, Supplementary Material). After evaluating MS data quality, all three samples were analyzed with Progenesis LC-MS software, using an algorithm based on the pair-wise features detection at LC-MS level. 726, 662 and 662 dysregulated proteins (≥2-fold) were quantified in three GC patients ( Figure 1B and Table S1, Supplementary Material), while 146 dysregulated proteins were reliably quantified after limited selection in these proteins. Of 146 proteins, 81 proteins were downregulated, 65 proteins were upregulated (adjacent/tumor ratio ≥2-fold or ≤0.5-fold, p-value < 0.05) and 22 proteins were first found to be related with GC (Table S1, Supplementary Material). identified from three cases in tumor or adjacent tissues respectively. One thousand seven hundred forty-four proteins identified appear in both tumor and adjacent tissues; (B) The number of proteins with more than twofold differential expression in three cases, respectively, and the number of proteins shared in two or three cases.

Functional Annotation of Proteins between GC and Adjacent Tissues
To reveal any possible function of the 146 dysregulated proteins, in silico analysis was performed. Ingenuity Pathway Analysis (IPA) aids in the integration of complex omics data and provides insight into regulatory mechanism and biological functions based on published studies [21]. The heatmap of "disease and function" of 146 dysregulated proteins by IPA was shown in

Functional Annotation of Proteins between GC and Adjacent Tissues
To reveal any possible function of the 146 dysregulated proteins, in silico analysis was performed. Ingenuity Pathway Analysis (IPA) aids in the integration of complex omics data and provides insight into regulatory mechanism and biological functions based on published studies [21]. The heatmap of "disease and function" of 146 dysregulated proteins by IPA was shown in Figure 2, most of these proteins were involved in cancers (117/146, 80.14%) and gastrointestinal disease (99/146, 67.80%) ( Table 1 and Table S2 (Supplementary Material)). Their main functions concern cellular growth and proliferation, nucleic acid metabolism, small molecule biochemistry, cell death and survival, cellular movement ( Table 2 and Table S2 (Supplementary Material)). Protein Analysis Through Evolutionary Relationships (PANTHER) is a comprehensive database used to analyze protein family, gene ontology and pathways for proteins with different abundances between adjacent and tumor tissues [22]. PANTHER analysis showed that 146 dysregulated proteins could be categorized into 25 protein classes in which nucleic acid binding proteins comprised the largest group (8.9%) ( Figure 3A). According to the Meta-analysis, these proteins were most associated with metabolic (25.5%), and cellular (17.0%) processes, among others ( Figure 3B). This study also revealed that molecular functions of these proteins were mostly concerned with catalytic (34.5%) and binding (22.6%) activity ( Figure 3C). We performed additional analysis using the Database for Annotation, Visualization and Integrated Discovery (DAVID) in order to further shed light on functional annotation of these dysregulated proteins. DAVID contains an integrated biological knowledgebase and extracts biological meaning from large gene/protein lists at systematically [23]. The results in Table S3 show that 146 differentially express proteins possessed various molecular functions and biological process. Further examination of the group of 65 upregulated proteins, DAVID revealed that their main functions is RNA and protein binding, pointed to biological processes of RNA splicing and processing, as well as metabolic processes ( Figure 3D,E and Table S3, Supplementary Material).
To obtain credible signaling pathways where dysregulated proteins may participate, STRING and Reactome were selected to find enriched pathways together with PANTHER and DAVID. STRING is a global scale database that annotates protein interactions and associations at various levels [24]. Reactome is an expert-authored, peer-reviewed knowledgebase of human reactions and pathways that functions as a data mining resource and electronic textbook [25]. Comprehensive analysis by four publicly available pathway tools revealed many enriched pathways take part in GC, for example metabolic pathways, gene expression, the citric acid (TCA) cycle and respiratory electron transport, mitochondrial dysfunction, oxidative phosphorylation, mRNA splicing ( Figure 3F and Table S4, pathways that functions as a data mining resource and electronic textbook [25]. Comprehensive analysis by four publicly available pathway tools revealed many enriched pathways take part in GC, for example metabolic pathways, gene expression, the citric acid (TCA) cycle and respiratory electron transport, mitochondrial dysfunction, oxidative phosphorylation, mRNA splicing ( Figure  3F and Table S4, Supplementary Material).  STRING predicted protein-protein interaction of 146 dysregulated proteins showed that most of them could interact with each other and formed strong networks with three dynamic clusters.

Validation of Dysregulated hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 by Nano-LC-MS/MS, qRT-PCR and Western Blot
Taken together, a number of evidence suggests that activated nucleic acid binding proteins might play a vital role in cell growth, proliferation and metastasis in GC pathogenesis. To verify the activation of these kinds of proteins, Nano-LC-MS/MS, qRT-PCR and western blot were employed to validate the differential expression of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 in ten pooled or individual GC patients' tissue samples. By Nano-LC-MS/MS analysis based on relative abundances of Peptide-Spectrum Match (PSMs), significant up-regulations of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 were confirmed (Tables S5 and S6, Supplementary Materila) [27]. QRT-PCR and western blot both demonstrated a significant up-regulation of three hnRNPs and YBX-1 transcription and translation in the most of the ten GC tissues compared with adjacent tissues, with only 1-2 cases being the exception ( Figure 5A-C). These results also demonstrated that qRT-PCR, western blot and a coupled label-free MS approach have consistency in mRNA and protein abundance quantification.  Taken together, a number of evidence suggests that activated nucleic acid binding proteins might play a vital role in cell growth, proliferation and metastasis in GC pathogenesis. To verify the activation of these kinds of proteins, Nano-LC-MS/MS, qRT-PCR and western blot were employed to validate the differential expression of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 in ten pooled or individual GC patients' tissue samples. By Nano-LC-MS/MS analysis based on relative abundances of Peptide-Spectrum Match (PSMs), significant up-regulations of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 were confirmed (Tables S5 and S6, Supplementary Materila) [27]. QRT-PCR and western blot both demonstrated a significant up-regulation of three hnRNPs and YBX-1 transcription and translation in the most of the ten GC tissues compared with adjacent tissues, with only 1-2 cases being the exception ( Figure 5A-C). These results also demonstrated that qRT-PCR, western blot and a coupled label-free MS approach have consistency in mRNA and protein abundance quantification. (B) Western blots (n = 10) of hnRNPs and YBX-1. N represent adjacent tissue and T represent tumor tissue; (C) Grayscale scanning of western blots bands. The ratio was compared to β-actin and statistically analyzed. Significance of differences between GC and adjacent tissues are displayed by ** p-value < 0.01 or * p-value < 0.05.

Expression and Distribution Detection of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 by Immunohistochemistry
Proteins highly abundant in tumors might be detectable with better reproducibility by immunohistochemistry (IHC), which can also reveal the distribution status of particular proteins in cells and tissues. With specific antibodies hybridization, IHC results showed a stronger and higher density positive nuclei staining for all three hnRNPs and YBX-1 in GC tissues than in adjacent normal tissues ( Figure 6). Cancer cells undergoing cell proliferation have larger and multiple nuclei (arrows shown) compared to normal cells. Furthermore, by observing protein expression and distribution, the slides stained with hnRNPA2B1, hnRNPD and YBX-1 antibodies displayed a weak (1+) to moderate (2+) glandular epithelium cell cytoplasm and nucleus staining in adjacent tissues, whereas there were only strong (3+) nucleus staining in tumor tissues were found. HnRNPL did not appear in cytoplasms. Immunohistochemistry not only confirmed the expression of four upregulated proteins in accordance with qRT-PCR and western blot results, but also revealed nucleo-cytoplasmic shuttling of hnRNPA2B1, hnRNPD and YBX-1.

Expression and Distribution Detection of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 by Immunohistochemistry
Proteins highly abundant in tumors might be detectable with better reproducibility by immunohistochemistry (IHC), which can also reveal the distribution status of particular proteins in cells and tissues. With specific antibodies hybridization, IHC results showed a stronger and higher density positive nuclei staining for all three hnRNPs and YBX-1 in GC tissues than in adjacent normal tissues ( Figure 6). Cancer cells undergoing cell proliferation have larger and multiple nuclei (arrows shown) compared to normal cells. Furthermore, by observing protein expression and distribution, the slides stained with hnRNPA2B1, hnRNPD and YBX-1 antibodies displayed a weak (1+) to moderate (2+) glandular epithelium cell cytoplasm and nucleus staining in adjacent tissues, whereas there were only strong (3+) nucleus staining in tumor tissues were found. HnRNPL did not appear in cytoplasms. Immunohistochemistry not only confirmed the expression of four upregulated proteins in accordance with qRT-PCR and western blot results, but also revealed nucleo-cytoplasmic shuttling of hnRNPA2B1, hnRNPD and YBX-1.

Figure 6.
Representative immunohistochemical staining for sectioned formalin fixed GC and adjacent tissues. Specific antibodies of Anti-hnRNPA2B1 (Santa Cruz, TX, USA), Anti-hnRNPD (Proteintech, Chicago, IL, USA), Anti-hnRNPL (Santa Cruz) and Anti-YBX-1 (Santa Cruz) were hybridized respectively. IHC results showed that the morphology of tubular glands disappeared in cancer sections when compared to adjacent tissues. Cancer sections have stronger and higher density nuclei staining, high ratios of nucleus/cytoplasmic area, different shaped nuclei including megakaryocytes and polykaryocytes (arrows). Weak cytoplasmic staining were only seen in hnRNPA2B1, hnRNPD and YBX-1 hybridized normal sections. The magnification is 400×; scale bar: 20 μm.

Discussion
Since the introduction of label-free proteomics, a broad variety of studies aiming for discovery of biomarkers or drug targets for diseases have been performed [16]. However, this method has been seldom used in GC studies. In 2012, by using shotgun proteomic approach and label-free quantitation analysis, Aquino et al. revealed tissue-type proteins were very distinct from each other in control, cancer and resection margin biopsies, only 11, 22, and 29 proteins (p-value ≤ 0.05) were quantified in cancer vs. control, resection margin vs. cancer and resection margin vs. control, respectively [28]. Moreover, resection margin biopsies proteins may be related to tumor nourishment and metastasis [28]. In 2013, by using a combinatorial approach of Con-A affinity chromatography, SDS-PAGE, LC/MS/MS and label-free comparative glycoproteomic quantification strategy, Uen et al. found 17 differentially expressed glycoproteins with 10 upregulated and 7 downregulated in plasma from GC patients versus healthy volunteers [29]. In 2015, by using SDS-PAGE and a coupled label-free MS approach, Qiao et al. identified 297, 419, and 265 dysregulated proteins with ≥2 folds in SGC-7901, MGC-803 and HGC-27 cells respectively when compared with GES-1 cells, and provided evidence showing that filamin C is a tumor suppressor, inhibiting cancer cells metastasis [30]. In our study, by using filter-aided sample preparation (FASP) method followed by a coupled label-free MS approach on whole protein extract from surgically resected GC patients' fresh tumor and matched adjacent tissues, we identified and quantified a higher number of dysregulated proteins. In three independent cases with matched samples, a total of 3639 and 3543 proteins in cancer and adjacent tissues were identified. For better quantification, each of three case samples was performed in triplicates on LC-MS/MS and statistical analysis was carried out. A total of 146 dysregulated proteins with more than twofold differential expression were quantified between tumor and adjacent tissues, 81 of which were downregulated, while the other 65 proteins were upregulated in tumor tissues.

Discussion
Since the introduction of label-free proteomics, a broad variety of studies aiming for discovery of biomarkers or drug targets for diseases have been performed [16]. However, this method has been seldom used in GC studies. In 2012, by using shotgun proteomic approach and label-free quantitation analysis, Aquino et al. revealed tissue-type proteins were very distinct from each other in control, cancer and resection margin biopsies, only 11, 22, and 29 proteins (p-value ď 0.05) were quantified in cancer vs. control, resection margin vs. cancer and resection margin vs. control, respectively [28]. Moreover, resection margin biopsies proteins may be related to tumor nourishment and metastasis [28]. In 2013, by using a combinatorial approach of Con-A affinity chromatography, SDS-PAGE, LC/MS/MS and label-free comparative glycoproteomic quantification strategy, Uen et al. found 17 differentially expressed glycoproteins with 10 upregulated and 7 downregulated in plasma from GC patients versus healthy volunteers [29]. In 2015, by using SDS-PAGE and a coupled label-free MS approach, Qiao et al. identified 297, 419, and 265 dysregulated proteins with ě2 folds in SGC-7901, MGC-803 and HGC-27 cells respectively when compared with GES-1 cells, and provided evidence showing that filamin C is a tumor suppressor, inhibiting cancer cells metastasis [30]. In our study, by using filter-aided sample preparation (FASP) method followed by a coupled label-free MS approach on whole protein extract from surgically resected GC patients' fresh tumor and matched adjacent tissues, we identified and quantified a higher number of dysregulated proteins. In three independent cases with matched samples, a total of 3639 and 3543 proteins in cancer and adjacent tissues were identified. For better quantification, each of three case samples was performed in triplicates on LC-MS/MS and statistical analysis was carried out. A total of 146 dysregulated proteins with more than twofold differential expression were quantified between tumor and adjacent tissues, 81 of which were downregulated, while the other 65 proteins were upregulated in tumor tissues.
Although proteomics approaches focusing on the differences between tumor and adjacent tissues can reveal a number of proteins relevant to tumor, functional annotations of carcinogenesis require bioinformatics and biostatistical tools for analysis, which have become indispensable to handle and to interpret the vast amount of data. In our study, IPA, PANTHER, STRING, DAVID and Reactome were comprehensively used to exploit and explain any possible information related to GC carcinogenesis. By elucidating, 117 and 99 molecules of 146 dysregulated proteins were classified to be related with cancer and gastrointestinal disease development, respectively. Their functions concerning cellular growth and proliferation, nucleic acid metabolism, small molecule biochemistry, cell death and survival, cellular movement were all relative to cancer pathogenesis. Moreover, nucleic acid binding proteins (8.9%) were the most abundant group identified to be predominantly catalytic in nature (34.5%) and greatly involved in metabolic processes (25.5%). Kocevar et al. has partially confirmed our hypothesis by identifying 30 different proteins involved in metabolism, development, death, response to stress, cell cycle, cell communication, transport, and cell motility processes in GC [34]. Signal pathway prediction revealed that metabolic pathways, gene expression, oxidative phosphorylation, mitochondrial dysfunction, mRNA splicing were involved in GC carcinogenesis. Protein-protein interaction prediction uncovered an integrated network with three protein clusters. The clusters showed a recruitment of proteins with functional synergy, such as nucleic acid binding associated proteins, metabolizing associated proteins and extracellular matrix associated proteins. It was clear that all these proteins could somehow interact with each other so as to control the cells fate together. In the network consisting of 65 upregulated proteins, a series of hnRNPA2B1, hnRNPD, hnRNPL, hnRNPR together with other nucleic acid binding proteins, such as YBX-1, NPM1, RAN, SNRPF, were gathered in the center of the network, and functionally reacted with some confirmed GC-related proteins such as SFN with carcinogenesis and poor prognosis of GC [32,33], PCNA with abnormal cell proliferation [33], VIM with epithelial-mesenchymal transition in tumor [36] These data remind us that activation of the group of nucleic acid binding proteins might push stomach cells into entering an abnormal proliferation cycle.
To confirm our findings and hypothesis that nucleic acid binding proteins may form a network center and may play a vital role in abnormal cell proliferation during GC development, four upregulated molecules of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 from the center of the network in Figure 4 were selected to perform further experiments by Nano-LC-MS/MS, qRT-PCR, western blot and immunohistochemistry. The results are consistent with the findings by the coupled label-free MS approach, strongly supporting differential expression of four molecules between GC and adjacent tissues in ten matched case samples.
HnRNPs are a set of primarily nuclear proteins that bind to nascent transcripts with important roles in multiple aspects of nucleic acid metabolism, including the packaging of nascent transcripts, alternative splicing and translational regulation. Their potential roles in tumor development and progression including the inhibition of apoptosis, angiogenesis and cell invasion have been reported [37]. Aberrant expression of hnRNPs in cancers have been detected, for example hnRNPA2B1 in breast, pancreatic and gastric cancer, hnRNPB1 in lung cancer, esophagus cancer, leukemia and lymphoma, hnRNPC1/C2 in lung cancer, hnRNPA2B2 in thyroid cancer and hnRNPM4 in lung cancer [38]. In our study, hnRNPA2B1, hnRNPD, hnRNPL and hnRNPR expression were found to be increased in GC tissues, but hnRNPC expression was decreased. Among these proteins, hnRNPD and hnRNPR were first reported to be relevant with GC. In immunohistochemistry, we observed that hnRNPA2B1 and hnRNPD expression have nucleo-cytoplasmic shuttling phenomenon. Elevated expression of hnRNPA2B1 in GC has been reported [39], which is keeping in line with our results. Translocation of hnRNPA2B1 with c-myc, c-fos, p53, and Rb from nucleolus to cytoplasm during tumor cells differentiation [40]. HnRNPA2B1 could act as a novel regulator of oncogenic K-ras, modulating PI3K/AKT/mTOR signal pathway in K-ras-dependent pancreatic ductal adenocarcinoma cells [41,42]. HnRNPD could enhance the expression of c-myc, c-fos and c-jun, and regulated liver cancer cell proliferation in transgenic mice [43]. Thyroid carcinoma may recruit cytoplasmic hnRNPD to disturb the stability of mRNAs encoding cyclin-dependent kinase inhibitors, leading to uncontrolled growth and progression of tumor cells [44]. As illustrated above, many reports have shown that hnRNPs may be involved in proliferation and metastasis in GC development.
YBX-1 is a multi-functional protein that participates in a wide variety of DNA/RNA-dependent events, and acts as a versatile oncoprotein with an important role in carcinogenesis [45]. In our study, the expression of YBX-1 was upregulated and the distribution of its nucleo-cytoplasmic shuttling was indicated in GC tissues. Previous research reported that YBX-1 was predominantly localized in cytoplasm, particularly in the perinuclear region in benign cells [45]. Environmental stresses, such as adenovirus infection, hyperthermia, oxidative stress, UV irradiation or DNA-damaging drugs, could induce YBX-1 relocation from cytoplasm to nucleus [45]. YBX-1 can activate E2F, PI3K/Akt/mTOR and Ras/Raf/MEK/ERK pathways to promote cancer cell proliferation [46]. YBX-1 expression correlated significantly with lymph node status and perforation and as a potential prognostic biomarker in intestinal-type GC [47]. YBX-1 can promote GC development in both cancer cells and cancer vascular cells [48]. Nuclear YBX-1 expression was significantly associated with Her2 expression, poor prognosis and metastasis in GC patients [49]. In brief, YBX-1 may become a potential biomarker and target molecule in GC therapy.
There are some limitations in our study that need to be addressed. Firstly, the 146 dysregulated proteins were screened from only three GC patients between cancer and adjacent tissues, therefore it is needed to confirm these dysregulated proteins among multiple specimens to avoid heterogeneity and remove artificial differences owning to differential proteins losses during protein or peptide preparation. Second, the sample size still needed to be expanded, particularly for qRT-PCR, western blot and immunohistochemistry analysis. Tissue microarray should be employed to validate the results and candidate molecules.

Clinical Tissue Samples
Ten cases of GC and adjacent normal tissues were collected from GC patients who underwent gastric resection at the Department of Digestive Diseases, Xijing Hospital Affiliated to Fourth Military Medical University between November 2012 and January 2013. GC patients were male, ages from 40 to 70 years old, with primary and low to moderate differentiated gastric adenocarcinoma. Radiotherapy, chemotherapy, and immunotherapy were not performed before surgery. Cancer tissues were taken from the core area of tumor, avoiding inclusion of necrotic and adjacent non-cancerous tissues. Adjacent tissues were prepared from non-cancerous regions at least 5 cm apart from the core area of tumor. All samples were verified by two pathologists after surgery. Clinical pathologic characteristics are described in Table S7

Protein Preparation
Whole proteins were extracted from tissues as previously reported [17]. Matched GC and adjacent fresh tissues were excised immediately following gastrectomy, cut into small blocks, and rinsed with ice-cold PBS. About 100 mg GC or adjacent tissues were homogenized in RIPA lysis buffer (Pierce, Thermo Scientific, Waltham, MA, USA) containing protease inhibitor cocktail (Roche, Basel, Switzerland). Tissue lysates were centrifuged at 15,000ˆg for 20 min at 4˝C, and supernatants were collected. After protein concentration was determined by BCA (bicinchoninic acid) protein assay kit (Pierce, Thermo Scientific, Germany), protein aliquots were stored at´80˝C.

In-Solution Tryptic Digestion
Proteins from three pairs of GC and adjacent tissues were processed in-solution digestion respectively as previously described [17,26]. The samples were dialyzed with ammonium bicarbonate, reduced with DL-Dithiothreitol (DTT, Sigma-Aldrich, St. Louis, MO, USA), and alkylated with iodoacetamide (IAA, Sigma-Aldrich). Trypsin (Sigma-Aldrich) digestion was performed at 37˝C for 24 h, and C18 spin columns (Millipore, Waltham, MA, USA) were used to purify the peptides. Peptides were dried in SpeedVac and stored at´80˝C until analysis.

SDS-PAGE and in-Gel Tryptic Digestion
About 100 µg protein were run on 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Proteins from 10 samples of GC or adjacent tissue lysates were pooled to avoid heterogeneity. After coomassie blue staining, the protein bands were cut into 10 equal shares from the gel, processed in-gel digestion with trypsin using the standard protocol [50]. After trypsin digestion, Ziptip C 18 micropipette tips (Millipore) were used to purify the peptides prior to adding 0.1% formic acid for Nano-LC-MS/MS analysis.

Nano-LC-MS/MS Analysis
The method was performed as previously described [26,50]. In brief, peptides of three individually or ten pooled tryptic samples were fractionated by high pressure liquid chromatography (HPLC, Thermo EASY-nLC System, Waltham, MA, USA) respectively: buffer A (0.1% (v/v) formic acid in Milli-Q water) and buffer B (0.1% formic acid in 100% acetonitrile). Peptides were eluted from the column at a constant flow rate of 300 nL¨min´1 with a linear gradient of buffer B from 5% to 35% over 120 min. Then following full scan with LTQ Velos Pro tandem mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA): LTQ MS/MS depended scans with collision-induced dissociation (CID) mode. Each of the three tryptic samples must be loaded three times with the same methods. MS/MS data were analyzed with Proteome Discoverer 1.4 (Mascot and SEQUEST, Waltham, MA, USA) according to manufacturer's instruction and searched against human uniprot protein database (UniProtKB [51] 3 November 2014, 140,992 sequences) for protein identification. Peptide mass tolerance was set to 0.8 Da, fragment mass tolerance was set to 10 ppm and a maximum of two missed cleavages was followed. Variable modification was oxidation of methionine, static modification was carbamidomethylation of cysteine. Protein identification was considered valid if at least one peptide and the p-value <0.05, the proteins not satisfying these defined criteria were rejected, the threshold for accepting MS/MS spectra was false discovery rate (FDR) 0.05.

Label-Free Quantification
Label-free quantification of three matched samples was acquired and analyzed in Progenesis LC-MS software (version 4.1, Milford, MA, USA) as previously described [50,52]. Mass spectrometer raw-files were transformed to .mzxml with ReAdw-program and imported into Progenesis LC-MS software. The ion intensity maps of all six runs were examined for defects. One sample was set as the reference, data processing was aligned, and peptide ions with charge state of +1 or >4 and difference ratio of proteins (adjacent/cancer) are <2.0-fold or >0.5-fold were excluded. For quantification, the unique peptides validated by MS (p-value < 0.05) were chosen and calculated by summing the abundances of all peptides allocated to a specific protein. The software calculates ANOVA and q-values, which were used to deduce differentiating peptides.

Bioinformatics Analysis
Limited selections were used to screen label-free quantitative data before our results were necessary for differential analysis: proteins with the same peptides found in two or three patients were considered; retrieve the credibility is ě95%, false positive is <5% in the database; difference ratio of proteins is ě2.0-fold and p-value is ď0.05; identified proteins must be redundancy by all artificial. Dysregulated proteins were subjected to bioinformatics analysis tools for enrichment categories of functional annotation, networks and diseases-related proteins, for example PANTHER 9.0 [53], IPA [54], STRING (version 9.1) [55], Reactome [56] and DAVID (Bioinformatics Resources 6.7) [57].

Validation of Dysregulated Proteins by qRT-PCR, Western Blot and Immunohistochemistry
Tissues were ground with mortar and total RNA was extracted with RNAfast 1000 (Pioneer biotechnology, Xi'an, China), and reverse transcribed with PrimeScript™ RT Reagent kit with gDNA Eraser (TaKaRa, Tokyo, Japan) according to the respective manufacturers' instructions. Diluted aliquots of reverse transcribed cDNAs were used as templates in qRT-PCR containing SYBR Green PCR Master Mix (TaKaRa) with Applied Biosystems 7500 Fast Real-Time PCR system (Life technologies, Waltham, MA, USA). Triplicate reactions were carried out for each sample to ensure reproducibility. Gene expression was quantified using the comparative cycle threshold (C t ) method. Primers sets are listed in Table S7 (Supplementary Material).
Proteins from adjacent and tumor tissues were separated on SDS-polyacrylamide gels were transferred to PVDF membranes by electro blotting. Membranes were washed and blocked, then were incubated with the specific primary antibodies (Table S7, Supplementary Material) at 4˝C overnight. Membranes were washed with PBST and incubated with the respective HRP-conjugated secondary antibody. Signals were visualized with the WesternBright™ Sirius Highest sensitivity chemiluminescent HRP Substrate (Advansta, Menlo Park, CA, USA) and intensities recorded with a ChemiDoc-It 510 Imager (UVP, Upland, CA, USA). Band intensities were quantified using the Image J and normalized to total protein on the membrane.
Immunohistochemistry analysis was performed with GC and adjacent tissues fixed in 10% formalin, embedded in paraffin, and sectioned at 3-4 microns. Tissue sections were deparaffinized, hydrated, subjected to thermal treatment in Tris-EDTA in a pressure cooker at boiling for 20 min for antigen retrieval, exposed to endogenous peroxidase blocking and incubated with primary antibodies (Table S7, Supplementary Material) for 2 h at room temperature. The reaction was visualized with 3, 3 1 -diaminobenzidine tetra hydrochloride (DAB, Zhongshan, China) as chromogen. Finally, sections were counterstained with hematoxylin, dehydrated, mounted, and tissue slides were evaluated under a microscope (Olympus IX71, Olympus Corporation, Tokyo, Japan). For evaluating protein expression, the intensity of staining was scored as negative, weak, moderate, or strong (score 0, 1, 2, or 3, respectively) [26].

Data Analysis
SPSS 19.0 software (SPSS Inc., Armonk, NY, USA) was used for statistical analysis. p-value < 0.05 was considered statistically significant. Statistical analyses between two groups were performed using a two-tailed Student's t-test.

Conclusions
In summary, we confidently identified 146 dysregulated proteins including 22 proteins first reported on paired GC and adjacent tissues by a coupled label-free MS approach. Additionally, we noted the majority of dysregulated proteins were involved in cancers and gastrointestinal disease. Moreover, four possible key carcinogenesis molecules of hnRNPA2B1, hnRNPD, hnRNPL and YBX-1 were validated by Nano-LC-MS/MS, qRT-PCR, western blot and immunohistochemistry. They were located in a predicted interaction network keynotes and their nucleo-cytoplasmic shuttling may play an important role in gastric carcinogenesis. Although further studies are needed to validate their functional roles and molecular mechanism, these findings provide an overview and deeper understanding about GC-related molecular changes, and provide a group of potential diagnostic/prognostic biomarkers and therapeutic molecular targets for clinical intervention of GC.