E-cadherin is a calcium-dependent cell-cell adhesion molecule that plays a crucial role in establishing the epithelial architecture and maintaining cell polarity and differentiation[21,22]. GC cells can disseminate to distant organs, and there are dramatic alterations between cancer cells and extracellular-matrix components, indicating that alterations in cell-cell adhesion and cell-matrix adhesion can lead to tumor progression. In addition to its role in cell-cell adhesion, E-cadherin and the cadherin-catenin complex modulate various signaling pathways in epithelial cells, including Wnt signaling, Rho GTPases, NF-κB pathways and the epithelial-mesenchymal transition that accelerates cell motility and invasiveness[23-25]. Therefore, dysregulation of E-cadherin contributes to tumor invasion and progression. Promoter hypermethylation induced by Helicobacter pylori infection and mutation of CDH1 led to the inhibition of expression or activity of E-cadherin[21,26]. Dysfunction of E-cadherin is broadly involved in GC progression and predominantly contributes during invasion of the gastric wall and migration of cancer cells into the free abdominal space.

Annexin 1 (ANXA1) is a member of the calcium- and membrane-binding proteins and was initially characterized as a glucocorticoid-regulated anti-inflammatory protein[27]. Recent evidence suggests that ANXA1 has a wide range of cellular functions, such as membrane aggregation, phagocytosis, proliferation, apoptosis and tumorigenesis[28]. Overexpression of ANXA1 causes constitutive activation of the mitogen-activated protein kinase/extracellular signal-regulated kinase (ERK) pathway in various cells[28,29].

Cheng et al[30] found that high ANXA1 expression was significantly associated with increased serosal invasion and peritoneal metastasis, and poorer overall survival in GC patients. Furthermore, in vitro studies illustrated a novel regulatory mechanism involving formyl peptide receptors, ERK1/2 and ITGB1BP1, by which ANXA1 regulates GC cell invasiveness. An in vivo study revealed that shRNA-mediated inhibition of ANXA1 significantly suppressed formation of intraperitoneal nodules[30]. These results suggested that ANXA1 facilitates GC cell invasion in the gastric wall and spread to the abdominal cavity.

The melanoma-associated antigen (MAGE) genes encode multifunctional proteins that regulate the cell cycle, differentiation and survival[31,32]. A unique member of the type II MAGE family, neurotrophin receptor-interacting melanoma antigen-encoding gene homolog (NRAGE), also known as MAGE-D1, is located on the X chromosome and encodes an 86 kDa protein[33]. NRAGE is highly expressed in the nervous system and was originally reported to act as a pro-apoptotic factor required for the normal developmental apoptosis of sympathetic, sensory and motor neurons[34,35]. To date, there are conflicting reports regarding the expression and oncogenic significance of NRAGE in multiple cancers, including lung, breast and esophageal cancer, as well as melanoma[35,36].

We recently evaluated the expression and function of NRAGE in GC and found that NRAGE mRNA expression was positively correlated with that of the apoptosis-antagonizing transcription factor[37]. SiRNA-mediated knockdown of NRAGE significantly decreased proliferation, migration and invasion of GC cells. A stepwise elevation in NRAGE mRNA expression in GC tissues was observed with increasing disease stage, and high NRAGE expression was associated with serosal invasion of the tumor and positive lavage cytology and, subsequently, shorter survival[37]. Our results indicated that NRAGE may contribute to peritoneal dissemination of GC in terms of invasion of the gastric wall and detachment from the primary tumor to gain access to the peritoneal cavity.

Hypoxia is a hallmark of solid tumor formation and is associated with local invasion, metastatic spread, resistance to radiotherapy and chemotherapy, and poor prognosis in a number of human carcinomas[42]. The free intraperitoneal space is associated with a specific microenvironment and results in starved hypoxic GC cells that must survive under these environmental conditions until they can attach to the mesothelium[43]. Hypoxia-inducible factor-1α (HIF-1α) is a key transcription factor involved in the cellular response to hypoxic conditions and is also involved in angiogenesis and glycolysis[42,43].

Miao et al[20] showed a positive correlation between HIF-1α expression and GC peritoneal dissemination. Furthermore, GC stem/progenitor cells, which were identified using Hoechst 33342 staining, increased in primary GC cells under hypoxic conditions in vitro and showed an enhanced capacity for self-renewal but reduced differentiation mediated by HIF-1α. In mouse models, GC stem/progenitor cells preferentially resided in the hypoxic peritoneal zone of omentum-associated lymphoid tissues, also known as milky spots[20]. These findings supported the involvement of HIF-1α in the development of peritoneal dissemination, particularly in the adaptation to the hypoxic microenvironment in the abdominal cavity.

Phosphatase and tensin homolog (PTEN) is located on human chromosome 10q23, a locus that is highly susceptible to loss of heterozygosity, and it is one of the most frequently mutated tumor suppressors in multiple cancers[44]. PTEN negatively regulates the PI3K signaling pathway due to its lipid phosphatase activity, thereby inhibiting the activation of downstream components, such as AKT and NF-κB, stimulating cancer cell proliferation and growth[45,46]. PTEN could interact with and dephosphorylate focal adhesion kinase (FAK), leading to the inhibition of integrin-mediated cell spreading and cell migration[46,47]. Zhang et al[48] demonstrated that PTEN overexpression or knockdown in GC cells led to the down-regulation or up-regulation of FAK, respectively, and decreased or increased cell invasion, respectively. Using orthotropic GC nude mouse models, the researchers found that the metastatic peritoneum nodules were fewer and smaller in mice injected with GC cells overexpressing PTEN[48]. Accordingly, the PTEN/PI3K/NF-κB/FAK axis is believed to be one of the key pathways involved in the peritoneal dissemination of GC.

Chemokines are a superfamily of small, structurally related chemoattractant cytokines[49]. Chemokines bind to G protein-coupled receptors on leukocytes and stem cells and function through guanine nucleotide-binding proteins to initiate intracellular signaling cascades that promote migration toward the chemokine source[50]. Among all chemokine receptors, CXCR4 is of particular importance in solid cancer metastasis and migration[51,52]. CXCL12 is the only known ligand for CXCR4. It activates the CXCR4 receptor and attracts circulating CXCR4-expressing cells to peripheral tissues[51,53].

The CXCL12/CXCR4 axis regulates a wide variety of downstream signaling pathways related to chemotaxis, cell survival, and/or proliferation[54]. Yasumoto et al[55] evaluated the role of the CXCL12/CXCR4 axis in peritoneal dissemination of GC. The researchers first found that CXCR4 was abundant in the malignant ascites of mouse models using cells established from human malignant ascites of GC. CXCL12 was strongly expressed on peritoneal mesothelial cells, and higher levels of CXCL12 were detected in the malignant ascites fluid from patients with peritoneal dissemination of GC, compared to those in normal fluids in the peritoneal cavity.

CXCR4 expression in the primary tumors of patients with advanced GC was significantly associated with the occurrence of peritoneal dissemination. Chen et al[53] reported that the CXCL12/CXCR4 axis mediated cell migration via the mTOR pathway, and mTOR pathway inhibitors decreased CXCL12-stimulated cell migration in GC. Izumi et al[52] evaluated the tumor-promoting effects of CXCL12 derived from cancer-associated fibroblasts. CXCL12/CXCR4 activation by cancer-associated fibroblasts mediated integrin b1 clustering at the cell surface and promoted invasion of GC cells. Inhibition of CXCL12 decreased the invasive ability of GC cells via the suppression of integrin b1/FAK signaling. These results suggested that blocking the CXCL12/CXCR4 interaction or inhibiting downstream intracellular signaling pathways may be a useful strategy for cancer therapy[52].

To date, several preclinical and clinical studies have been conducted on anti-CXCL12 agents[56]. AMD3100 [recently renamed Mozobil (plerixafor injection)] was reported to inhibit CXCL12-induced tumor cell migration and downstream signaling and also to suppress CXCL12/CXCR4 signaling in murine tumor models. This drug is already approved for clinical use in patients with leukemia[57]. The therapeutic efficacy of several CXCR4 antagonists is also currently being tested in clinical trials[52].

Epidermal growth factor receptor (EGFR) is a member of a family of closely related growth factor receptor tyrosine kinases that includes HER2. Many family members have been identified as therapeutic targets for the treatment of various cancers[58]. In fact, treatment with trastuzumab, a human monoclonal antibody specific for HER2, has shown survival benefits in patients with advanced, HER2-positive GC[59]. To date, seven ligands for EGFR have been identified: epidermal growth factor (EGF), transforming growth factor-α; heparin-binding EGF-like growth factor (HB-EGF); amphiregulin (AREG); betacellulin; epiregulin; and epigen. Among them, AREG and HB-EGF have been reported to play a crucial role in tumor progression[60,61].

AREG and HB-EGF are synthesized as type I transmembrane protein precursors and are expressed on the cell surface as pro-amphiregulin and pro-HB-EGF, respectively[58,61,62]. Yasumoto et al[60] reported that AREG and HB-EGF were abundant in the ascites fluids from GC patients. AREG promoted the proliferation of CXCR4-expressing GC cells, and HB-EGF markedly induced migration of fibroblasts. HB-EGF and CXCL12 together enhanced TNF-α-converting enzyme-dependent AREG shedding from functional CXCR4-expressing GC cells[60]. These findings suggested that targeting tumor cells and their microenvironments via AREG and HB-EGF inhibition represents a promising treatment approach for peritoneal dissemination in GC.

Adhesion of GC cells to the peritoneum is a key step during development of peritoneal dissemination[66]. The integrin family of cell adhesion molecules serves as adhesion receptors for ECM proteins and cellular counterligands[66,67]. Nishimori et al[68] selected a GC cell line showing high peritoneal metastatic potential and found that these cells preferentially overexpressed a1-a6 integrins, in contrast to the parental cell line, which had a low peritoneal diffusion capability. Treatment with functional blocking antibodies to tumor integrins was found to decrease peritoneal dissemination.

Takatsuki et al[69] reported that integrin a3b1 played a critical role in cancer cell adhesion to the peritoneum. Monoclonal antibodies specific to integrin a3b1 inhibited GC cell adhesion to excised peritoneum and cell growth. In peritoneal mesothelial cells, mRNAs for laminin-5 and laminin-10/11, which have been identified as high-affinity ligands for integrin a3b1, were detected. Furthermore, pretreatment of excised peritoneum with an antibody to laminin-5 significantly inhibited the adhesion of GC cells[69]. Taken together, these findings have demonstrated that integrin strongly mediates the initial attachment of GC cells during peritoneal dissemination.

MMPs are a family of endogenous calcium- and zinc-dependent proteolytic enzymes that are capable of degrading most ECM components, as well as regulating other enzymes, chemokines and even cell receptors[70,71]. Matrix metalloproteinase 7 (MMP7) is a distinct family member with proteolytic activity against a wide range of biomolecules and is recognized as pivotal in the MMP family because it is the most potent member at activating other MMPs (i.e., MMP2 and MMP9) to degrade the ECM[72,73]. By degrading ECM proteins and regulating the activity of other biomolecules in the body, MMP7 may play a central role in stromal invasion of GC cells during formation of peritoneal dissemination[71,74]. Yonemura et al[75] reported that specific antisense oligonucleotides that inhibit MMP7 suppressed the invasive ability of GC cells without modifying cell proliferation in a mouse xenograft peritoneal dissemination model, leading to prolonged survival compared with control mice.

Connective tissue growth factor (CTGF) is a secretory protein and has been reported to be a multifunctional growth factor involved in wound healing, inflammation, cell adhesion, chemotaxis, apoptosis, tumor growth and fibrosis[76,77]. Additionally, CTGF promotes angiogenesis by regulating endothelial cell growth, migration, adhesion and survival[77,78]. Chen et al[79] evaluated the expression and oncological roles of CTGF in GC. CTGF overexpression or treatment with recombinant CTGF protein significantly inhibited GC cell adhesion. In vivo peritoneal dissemination models demonstrated that stable CTGF transfectants markedly decreased the number and size of the peritoneal nodules in the mesentery. Blocking integrin a3b1 inhibited GC cell adhesion to recombinant CTGF. Patients expressing low CTGF levels had a significantly higher prevalence of peritoneal dissemination and a lower probability of survival after surgery compared with those expressing high CTGF levels[79]. The authors concluded that CTGF was an anti-adhesion protein in peritoneal dissemination of GC and that recombinant CTGF may be a therapeutic option for patients with advanced GC.

Maternal embryonic leucine zipper kinase (MELK) is a cell cycle-dependent protein kinase that plays a key functional role in multiple cellular processes, such as proliferation, cell cycle progression, mitosis and spliceosome assembly[80]. MELK interacts with and phosphorylates CDC25B on Ser323 to regulate G2/M progression[81]. MELK has been reported to be frequently elevated in multiple cancers and is correlated with a poor prognosis[82,83]. In addition, MELK interacts with Bcl-GL through its amino-terminal region and suppresses apoptosis[81,83].

Du et al[84] found that MELK mRNA and protein expression were both elevated in GC tissues, and this was associated with chemoresistance to 5-fluorouracil. Knockdown of MELK significantly suppressed cell proliferation, migration and invasion of GC in vitro and in mouse xenograft peritoneal dissemination models in vivo; it decreased the percentage of cells in the G1/G0 phase and increased those in the G2/M and S phases. Moreover, knockdown of MELK decreased the amount of actin stress fibers, inhibited RhoA activity and the phosphorylation of FAK and paxillin, and prevented gastrin-stimulated FAK/paxillin phosphorylation[84]. Thus, MELK may be involved in the formation of peritoneal nodules in GC.

Angiogenesis in the subperitoneal space, which is predominantly mediated by VEGF, is an important step in peritoneal dissemination[85]. GC cells that reach the subperitoneal blood vessels proliferate by inducing angiogenesis and finally evolve into massive and firm peritoneal nodules[10,65]. VEGF secreted from cancer cells enhances tumor growth by inducing an angiogenic response in the peritoneal microenvironment, promotes vascular permeability in the peritoneum and contributes to the establishment of peritoneal nodules that generate abundant malignant ascites[87]. Accordingly, VEGF has a central role in the formation of peritoneal dissemination, and the development of antiangiogenic therapy targeting peritoneal mesothelial cells is a promising approach for regulating peritoneal dissemination of GC.

Iroquois homeobox 1 (IRX1) is a member of the Iroquois homeobox protein family and is involved in pattern formation in the embryo[88,89]. Jiang et al[90] first identified the tumor suppressive function of IRX1 in GC. Overexpression of the IRX1 gene is correlated with growth arrest and suppresses peritoneal spreading and long distance metastasis in GC. IRX1 transfection resulted in substantial suppression of peritoneal spreading with reduced angiogenesis (microvessel density) as well as vasculogenic mimicry formation in mouse xenograft models. In addition, the number of blood vessels was significantly reduced by treatment with supernatant from GC cells expressing recombinant IRX1[90]. These findings suggested that IRX1 acts as a growth facilitator of peritoneal nodules in GC via angiogenesis.

Peritoneal lavage cytology has been regarded as the most reliable method for detecting free GC cells; however, it lacks the sensitivity required for detection of residual cancer cells and prediction of peritoneal spread, and it is not uniformly performed in clinical practice[86,99,100]. To address this issue, molecular detection using reverse transcription-polymerase chain reaction (RT-PCR) analysis has been proposed as a detection method for micrometastasis in the abdominal cavity. Among the candidate molecular targets, carcinoembryonic antigen (CEA) has been suggested as a potent molecular marker[101]. We previously demonstrated that detection of CEA mRNA using RT-PCR of peritoneal washes had a high sensitivity and strong correlation with peritoneal recurrence and prognosis after curative surgery. In 242 patients without macroscopic peritoneal dissemination at the point of gastrectomy, positive CEA mRNA was the most important independent variable associated with peritoneal recurrence (hazard ratio = 1.57, P = 0.020)[102]. Moreover, we conducted a phase II clinical trial to evaluate the prognostic impact of postoperative S-1 monotherapy in GC patients with CEA mRNA positivity. As a result, the 3-year survival rate was similar between the study population and the historic control (67.3% vs 67.1%, respectively)[103]. To improve the diagnostic performance for detecting free peritoneal GC cells, a combination of molecular markers with high specificity might be necessary.

The Anosmin-1 (ANOS1) gene encodes a cell adhesion protein of the ECM[104]. ANOS1 contains a whey acidic protein domain and three fibronectin III domains, and it promotes the migration of gonadotropin-releasing hormone neurons from the olfactory placode to the hypothalamus during development[105]. ANOS1 induces neurite outgrowth and cell migration through fibroblast growth factor receptor 1 signal transduction pathways[106,107].

We recently reported that inhibiting ANOS1 expression decreased the proliferation, invasion and migration of GC cells. Notably, elevated ANOS1 levels in primary GC tissues were significantly associated with larger tumor size, serosal invasion, positive nodal status, and importantly, positive peritoneal lavage cytology[108]. Moreover, ANOS1 concentrations in the sera were lowest in healthy subjects and increased stepwise in patients with localized GC and those with disseminated GC. Thus, ANOS1 can be used to stratify patients according to risk of recurrence[108]. ANOS1 may, therefore, represent a biomarker for progression of GC phenotypes, including peritoneal spreading.

Phosphoglycerate kinase 1 (PGK1) is an ATP-generating enzyme of the glycolytic pathway and is regulated by HIF-1α[109]. To date, overexpression of PGK1 has been reported in breast and pancreatic cancers, and a close relationship between the regulation of the CXCR4/CXCL12 axis and PGK1 was shown in prostate cancer[109]. Zieker et al[14] found a significant overexpression of PGK1, the chemokine CXCR4 and its ligand CXCL12 in specimens from diffuse-type GC patients who had concomitant peritoneal dissemination. The expression levels of PGK1 were positively correlated with those of the CXCR4/CXCL12 axis and HIF-1α but were inversely correlated with VEGF expression[14]. PGK1 expression in the primary tumor may be a candidate biomarker for peritoneal dissemination of GC.

Dihydropyrimidinase-like 3 (DPYSL3) is a cell-adhesion molecule that is normally expressed in various organs[110,111]. However, recent studies showed that DPYSL3 was involved in metastasis of tumor cells in pancreatic cancer[112]. We recently evaluated the expression levels of DPYSL3 in GC cells and tissues and found that DPYSL3 mRNA expression levels were positively correlated with those of interacting genes (VEGF, FAK and ezrin)[113]. Elevated DPYSL3 expression levels were significantly associated with an aggressive phenotype, including serosal invasion, invasive growth type, and, importantly, positive peritoneal lavage cytology. Tissues from patients with stage IV GC showed increased expression of DPYSL3 mRNA. Consequently, high DPYSL3 mRNA expression in GC tissues was identified as an independent prognostic factor[113]. Although the significance of its expression in sera and ascites fluids has yet to be determined, our results indicated the potential of DPYSL3 as a biomarker of the progression of GC, including peritoneal dissemination.