Abstract
Background: Peritoneal dissemination represents a poor prognostic indicator in gastric cancer. Despite a comprehensive molecular characterization of this disease, no peritoneal dissemination-specific signature has been identified, limiting the tailoring of the surgical and oncological treatments. In this review, we outline the available literature focusing on the role of the different molecular pathways involved in the acquisition of peritoneal metastatic dissemination. Summary: According to our results, several molecular determinants are associated with peritoneal carcinomatosis and are involved in several cellular and molecular carcinogenetic processes. However, a comprehensive understanding of the complex molecular landscape of gastric carcinosis is still lacking. Key Messages: More efforts should be made toward the integration of molecular and histologic data to perform a risk prediction assessment of peritoneal dissemination based on molecular profiling and histological evaluation.
Introduction
Gastric cancer (GC) is the fifth most common malignancy in both sexes, with more than 1 million new diagnoses in 2020 [1]. With almost 800,000 deaths, GC also represents the third cause of cancer-related death worldwide [1, 2].
Peritoneal dissemination (PD) is a typical metastatic pattern of GC and is characterized by a dismal prognosis. Nearly 20% of patients present synchronous peritoneal metastases, which can be found at the time of initial GC diagnosis or during radical surgery [3]. Metachronous PD occurs as peritoneal recurrence after curative resection and accounts for up to 50% of all recurrences [3]. Accurate staging of disease is mandatory to define patients’ treatment strategy [2, 4]. However, in some cases, radiological staging fails to detect PD [5, 6]. Laparoscopic exploration with peritoneal lavage might be suggested as a diagnostic option [7‒9]; however, it remains an invasive procedure that requires general anesthesia.
GC is a highly heterogenous group of neoplasms. The phenotypical and morphologic heterogeneity might be responsible for the different metastatic patterns observed in patients with advanced GC [1‒13]. A better comprehension of the mechanism underlying different patterns of clinical behavior of the disease could allow identification of predictors of PD and might guide GC treatment to a specific target, achieving encouraging outcomes. Previous clinical and multi-omic analysis propose a subtype of GC patients at high risk of PD. Further insights are needed to better understand the complexity of the molecular mechanisms underlying peritoneal carcinomatosis. Multiple predictors demonstrated a potential as PD markers in GC [14]; however, most of them need further confirmation by clinical trials and additional validation in order to achieve an accurate patients’ risk stratification.
The present review summarizes the molecular processes and the main molecular and genomic actors associated with PD risk in GC. Although these molecules may have an action at multiple steps, we have illustrated their role according to their primary contribution to carcinomatosis. The classification of molecules was based on the principal pathway and the main process in which they have a role in determining PD (Table 1).
Epithelial-Mesenchymal Transition
Epithelial-mesenchymal transition (EMT) is a process characterized by the progression from an epithelial cell phenotype to a mesenchymal phenotype, distinguished by fibroblast-like features [15]. EMT seems to be involved in the major steps of tumorigenesis [16]. Though it is considered a reversible process, it is only one of the various steps needed for tumor invasion and metastatic spread [17, 18]. Many molecules seem to be implicated in determining PD thanks to this process in GC.
ARL4C expression positively correlated with the EMT gene set in GSEA. Hu et al. [19] identified ADP-ribosylation factor-like 4c (ARL4C) as a PD-associated gene via EMT. Immunohistochemical analysis, conducted on 146 patients, showed that ARL4C was overexpressed in GC cells, and its expression was associated with the depth of invasion and PD and was a poor prognostic factor.
In recent molecular biological studies, ribophorine II (RPN2) also induced EMT and metastatic activity [20]. Fujimoto et al. [21] performed immunohistochemical staining to detect RPN2 and p53 in the primary lesion and adjacent normal gastric mucosa of 242 gastric adenocarcinoma patients. RPN2 expression was associated with a higher incidence of depth of wall invasion, lymph node metastasis, lymphatic invasion, venous invasion, PD, high histopathological stage, and p53 expression [21].
Moreover, zinc finger E-box-binding homeobox 1 (ZEB1) is one of the EMT-inducible genes that play a crucial role in tumor progression in different cancers [22‒24]. Immunohistochemical staining on 134 GC samples showed that ZEB1 expression was significantly associated with PD and that it was an independent prognostic factor [25].
Smad-interacting protein 1 (SIP1, also known as ZEB2) is another zinc finger E-box-binding homeobox family protein involved in EMT which seems to have a critical role in tumor progression in some cancers [26, 27]. Analyzing 134 samples of patients affected by GC, elevated SIP1 level was correlated with PD and poor prognosis [28].
KAP1 (KRAB-associated protein 1) is a transcriptional corepressor of Kruppel-associated box zinc finger proteins. Patients with high KAP1 expression in gastric adenocarcinoma samples demonstrated a higher incidence of peritoneal carcinomatosis and significantly poorer overall survival [29].
Yonemura et al. [30] studied the expression of E-cadherin (ECD) and S100A4 in GC. ECD is the strongest molecule in homophilic adhesion of epithelial cells and has a significant role in forming epithelial architecture. In carcinogenesis, irreversible inactivation of ECD is frequently observed [31, 32]. As a result, the adhesive ability of cancer cells deteriorates, and cell dispersion is facilitated. At the same time, S100A4 protein modifies cell motility by its ability to change cytoskeletal dynamics [33]. The power of S100A4 to interact with nonmuscle myosin supports the view that it takes part in cellular motility [34]. Yonemura et al. [30], by analyzing 92 GC patient samples, demonstrated that S1004A+/ECD expression was associated with PD [30]. Interestingly, the study of Cho et al. confirmed these results, showing that S100A4 expression was more frequently observed in patients with advanced GC, positive lymph node metastasis, and PD [35].
Chen et al. [36] in a previous study demonstrated that FNDC1 (fibronectin type III domain containing 1) has a higher frequency of mutations in patients with peritoneal metastasis (PM) of GC by whole-exome sequencing of postoperative tumors samples and adjacent normal tissues. Afterward, they found that FNDC1 promotes the invasiveness of GC and correlates with PM and poor prognosis [37].
The mesothelial-to-mesenchymal transition contributes to PD in neoplastic disease [38]. Signal transducer and activator of transcription 3 (STAT3) has been frequently implicated in the mesenchymal transition of human peritoneal mesothelial cells [39]. Based on that statement, Yang and Xu [40] showed that an increased expression of activated STAT3 was associated with PM in GC.
Recent evidence indicated a direct and indirect association of some microRNAs (miRNAs) with the EMT process [41, 42]. Kurashige et al. [43] reported that miR-200b was an EMT regulator through migration and invasion inhibition via targeting the mRNAs of ZEB1 and ZEB2 in GC cells. Subsequently, they demonstrated that the expression of miR-200b was significantly lower in GC samples of patients with PM [44].
Inflammation
It is well established that chronic inflammation can create an environment favorable for carcinogenesis and promote tumor initiation and progression [45]. The inflammatory cytokine tumor necrosis factor α (TNF-α) is an acute-phase reaction protein that promotes the progression of cancer by activating invasive growth factors [46] and also induces cancer invasion and metastasis [47]. Still, it can also increase immune response, angiogenesis, cell proliferation, and invasion [48]. TNF-α-related mechanisms showed that TNF-α plays a significant role in GC progression [49]. Especially, expression of TNF-α was demonstrated to be correlated with the risk of PM [50]. Many other inflammatory response factors seem to be involved in determining PD in GC.
CXCR7 (chemokine C-X-C motif receptor 7) mRNA, CXCL1 (chemokine C-X-C ligand 1), and CXCL1R (CXCL1 receptor) expression appeared to be associated with PD [51, 52]. CXCL12 and CXCR4 showed similar behavior. CLCX12 and CXCR4 positivity in primary gastric carcinomas correlated with peritoneal carcinomatosis [53].
AnxA1 (annexin A1) was initially characterized as a glucocorticoid-regulated anti-inflammatory protein [54]. However, N-terminally truncated AnxA1 has a proinflammatory role by promoting neutrophil transendothelial migration [55]. AnxA1 was shown to be a potential marker in gastric carcinogenesis by genome-wide complementary DNA microarray studies [17, 56]. Additionally, high AnxA1 expression was associated with PM and an independent risk factor for poor prognosis [57]. Furthermore, positive AnxA1 nuclear staining had an independent association with PD [58].
Colony-stimulating factor-1 (CSF-1) is a hematopoietic growth factor involved in cell differentiation, proliferation, and activation via binding to its receptor, CSF-1R, expressed on microglia and macrophages [59]. Several studies have shown that the overexpression of CSF-1 and CSF-1R is associated with progression in various types of cancer [60, 61]. Based on this evidence, Okuda et al. [62] demonstrated that the expression of CSF-1 and CSF-1R was significantly associated with the presence of PM in GC. CKLF, like MARVEL transmembrane domain containing 6 (CMTM6), is involved in the activation and chemotaxis of immune cells and affects the proliferation and invasion of tumor cells [63, 64].
Programmed death ligand-1 (PD-L1) is a negative immunoregulator that inhibits the activation of T cells and induces the apoptosis of antitumor T cells, which increases the occurrence and development of various tumors [65‒68]. The high expression of CMTM6 and PD-L1 correlated with Bormann type, N stage, PM, and TNM stage [69]. Despite this evidence, a previous study by Son et al. [70] demonstrated how tumor immune microenvironment is different between primary gastric tumor and metastatic gastric tumor. CD8+ T-cell, PD-L1+ cell, and PD-L1+ CK+ cell densities and deficient DNA mismatch repair were significantly lower in metastatic compared to primary GC. These results suggested that the microenvironment of PM was less immunologically active than that of primary tumors in GC patients. This evidence highlighted the controversial role of inflammation and immunity mechanisms in GC development and spread. Further evaluations of immune-related biomarkers in metastatic GC should be considered.
PD-1/PD-L1 blockade therapy is widely used in patients with metastatic GC. However, the effect on PM remains unclear due to the lower expression of PD-L1 in PM. Real-world data from the DELIVER trial demonstrated that the presence of PM was a prognostic factor for survival outcomes in patients treated with the anti-PD-1 nivolumab. A controlled phase II trial has been planned to assess the effectiveness and safety of hyperthermic intraperitoneal chemotherapy combined with the anti-PD-1 antibody camrelizumab in patients with advanced GB with PM [71].
Adhesion and Migration
Peritoneal-free cancer cells may have the ability to attach to the peritoneal surface. Some adhesion molecules have been reported to promote the attachment of GC cells to mesothelium and have a role in PD [72]. On the other hand, anti-adhesion and deconstruction of extracellular matrix is fundamental for cancer cells to migrate and disseminate to peritoneal-free space [73]. Claudins are major tight-junction components and display four transmembrane domains [74]. Ohtani et al. [75] showed that decreased claudin-4 (CLDN4) expression was associated with PM and lower OS.
HABP1 (hyaluronic acid-binding protein 1) is an adhesion molecule that was observed to be overexpressed in several types of cancers, including breast cancer [76], ovarian cancer [77], and endometrial cancer [78]. Elevated HABP1 expression was shown to have a role in the progression of GC; it was correlated with the depth of invasion, lymph node metastasis, liver metastasis, PM, and poor prognosis [79].
The adhesion of cancer cells to extracellular matrix proteins (ECMPs) is central in the multistep metastatic process [80]. It has been confirmed that cell-ECMP interactions are mediated by heterodimeric transmembrane glycoproteins, known as integrins [81]. The expression of alfa3 beta1 integrin was demonstrated to be associated with the presence of liver and PMs in GC [82]. Li et al. [83] demonstrated that beta integrin-h3 was significantly higher in cases with PM and positive peritoneal lavage cytology.
Recent research has revealed that tensin is not only an adhesion molecule, but it plays a role in several biological processes such as cell motility, migration, and apoptosis, as well as intercellular adhesion by mediating signal transductions through tyrosine phosphorylation [84‒86]. GC with higher tensin 4 (TNS4) mRNA expression exhibited histologically poorer grade, deeper invasion into the serosa, higher rate of positive lymph node metastasis, or PD [87].
Previous studies noticed that collagen alterations in the tumor microenvironment are associated with cancer dissemination and prognosis [88‒90]. Furthermore, the radial alignment of collagen at the tumor-stroma boundary increases the invasiveness of cancer cells [89, 90]. Another study demonstrated that in GS, there was a significantly higher 3-year cumulative PM rate in patients with high collagen signature [91].
Mesothelin is an adhesion molecule overexpressed in many tumors, such as mesothelioma, pancreatic cancer, ovarian cancer, GC, nonsmall cell lung cancer, triple-negative breast cancer, and cholangiocarcinoma [92‒95]. However, the correlation between mesothelin expression and GC progression is still unclear [95, 96]. Multivariate survival analysis showed that high mesothelin expression was independently associated with poor recurrence-free survival, overall survival, and peritoneal recurrence [97].
CD44 is a cell surface adhesion molecule that recognizes hyaluronate and mediates various functions, such as cell-cell and cell-matrix adhesion, lymphocyte homing, and T-cell adhesion and activation [98]. Yamamichi et al. [99] observed that expression of CD44s mRNA in patients with PD of GC was significantly higher than in patients without peritoneal involvement.
MUC1 (anti-adhesion mechanism) is a membrane-associated glycoprotein involved in several functions, such as defense against mechanical and infectious insults, lubrication, and acid resistance [100]. Wang et al. [101], by analyzing 76 GC samples, found that overexpression of MUC1 correlates with PD and poor prognosis.
Matrix metalloproteinase is a broad family of proteolytic enzymes which can degrade several substrates. Previous studies revealed that they are also expressed by cancer cells [102]. For example, matrix metalloproteinase-1 (MMP-1) is a proteolytic enzyme that degrades type I and III collagen, which are the main components of gastric stroma and peritoneum [103]. Inoue et al. [104] demonstrated that MMP-1 expression was associated with PD and poor prognosis in GC.
Another study showed that the expression of PAR-1 (protease-activated receptor-1) and MMP-1 correlated with PD. PAR-1 and combined PAR-1 and MMP-1 expression were independent prognostic factors [105]. Furthermore, MMP-7 expression was revealed to be associated with PD. Survival rate was lower in patients with GC with positive MMP-7 in GC samples [106]. DJ-1 appeared to be significantly upregulated in GC with PM, acting with a similar process. DJ-1 promotes in vitro invasion, migration, and in vivo peritoneal metastatic abilities of GC cells via MMP-2 and MMP-9 activity [107]. Also, Yang et al. [108] found that the expression rates of MMP-9 mRNA were higher in patients with PM and in patients with vascular endothelial growth factor (VEGF) expression.
Urokinase-type plasminogen activator (uPA) is another proteolytic enzyme; it has the ability to activate plasminogen to plasmin, which degrades many ECMPs [109, 110]. High uPA expression correlated with PD [111]. Ding et al. [112] observed that uPA, uPAR (uPA receptor), and PAI-1 expressions in metastatic peritoneal lesions were significantly higher than in normal peritoneal tissues of non-PM patients.
Proliferation, Apoptosis, Cell Cycle
Within the process of migration and adhesion to the peritoneum, cancer cells have to face a hostile environment. For this reason, cancer cells require intrinsic mechanisms of tumor maintenance and proliferation [72, 113]. Below, we review the main molecular actors that may contribute to the ability of GC cells to survive and proliferate in a free peritoneal microenvironment.
Overexpression of REGIV was shown in colorectal adenomas with severe dysplasia and adenocarcinoma, indicating the involvement of REGIV in colorectal carcinogenesis [114]. REGIV protein expression was also described in goblet cells of intestinal metaplasia and goblet-like cell vesicles of GC [115]. REGIV was revealed to be associated with peritoneal recurrence after surgery in GC and was an independent prognostic factor [116].
Epidermal growth factor receptor (EGFR) is a member of a family of closely related growth factor receptor tyrosine kinases. Anomalous activation of EGFR signaling can result in the dysregulation of cell growth ultimately leading to cancer [117]. In particular, EGFR and ErbB2 have been recognized as targets in the treatment of various cancers [118]. Saito et al. [119] observed that EGFR positivity rate of metastatic tumors (70.1%) in patients with PM was significantly higher than that of (37.5%) patients with liver metastasis.
In the carcinogenetic process, the increase in GnT-V (N-acetylglucosaminyltransferase V) activity and its cell surface products result from increased transcription driven by activation of the Ras-Ets and protein kinase B signaling pathway [120, 121]. Some studies have demonstrated that increased expression of GnT-V in breast and colorectal cancer correlates with distant metastases and a poor prognosis [122‒124]. In GC, GnT-V expression was shown to be significantly correlated with PD [125].
The hepatocyte growth factor/mesenchymal-epithelial transition factor (c-Met) pathway plays a key role in carcinogenesis, promoting migration, invasion, survival, and suppressing apoptosis via MAPK and PI3K-AKT pathways [126]. Metastasis-associated colon cancer 1 (MACC1) was observed to be elevated in numerous cancer tissues [127‒130]. Recently, it was demonstrated that MACC1 may be involved in the growth of blood vessels, lymphangiogenesis, and metastasis of GC, but little is known regarding its role in GC development [131, 132]. Guo et al. [133] analyzed the expressions of MACC1 and c-Met proteins. In a sample of 98 gastric carcinoma and nontumorous tissues, their expressions were detected by immunohistochemistry (IHC). Expression of the MACC1 protein in GC tissue was correlated with lymph node metastasis, PM, and hepatic metastasis [133]. Previously, Tsugawa et al. [134] have already observed that amplification of oncogene c-Met was associated with PD in GC.
Carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) belonging to the immunoglobulin superfamily and the carcinoembryonic antigen (CEA) family is a transmembrane protein. It has been reported that CEACAM1 is downregulated in breast, prostate, and endometrial cancer tissues [135‒137]. However, CEACAM1 overexpression is correlated to malignant grade of lung adenocarcinoma, advanced colon cancer, and hepatocellular carcinoma [138‒140]. Whether CEACAM1 is a suppressor or promoter for malignant phenotype of cancer cells still remains controversial. Conducting immunohistochemical assay on GC samples, Takeuchi et al. [141] have shown that CEACAM1 expression is the independent risk factor for PD.
Endoglin is upregulated in numerous tumors and correlated with proliferation and invasion and metastasis as well [142]. A recent study demonstrated that expression of endoglin in the peritoneal tissue of GC patients was correlated with PD and shorter survival [143]. VEGF expression also revealed to be a significant parameter in predicting PD in GC [144]. In patients with PM, the expression rates of integrin β3 mRNA and VEGF protein were higher than in patients without peritoneal involvement [145]. Furthermore, high expression of VEGF and decreased p53 expression were correlated with the presence of peritoneal disease [146].
Previous studies suggested that protein tyrosine phosphatase superfamily plays a key role in tumorigenesis and metastasis development [147‒150]. In particular, phosphatase of regenerating liver-3 (PRL3) was already observed to be related to lymph node metastasis [151]. Two studies showed that PRL-3 was overexpressed in GC tissues with PM compared with that in normal gastric tissues [152, 153].
ECD is one of the subclasses of the cadherin family which plays a key role in maintaining tight junctions between cells in epithelial tissues [154]. It has been suggested that downregulation of cadherin-mediated intercellular adhesion is involved in the initial steps of tumor invasion and metastasis development [32, 155]. Several studies revealed that ECD inactivation is significantly higher in GC with PD than in GC without peritoneal involvement [156‒158].
Connective tissue growth factor (CTGF) is involved in numerous biologic and pathologic processes, such as angiogenesis, osteogenesis, fibrosis in kidneys and skin, and tumor development [159, 160]. GC CTGF expression was revealed to be significantly correlated with the development of PD and a shorter survival rate [161, 162].
Epiregulin (EREG) is a member of the family of EGF-like ligands and is a wide receptor-binding ligand, that directly activates EGFR and ErbB4 homodimers, but also stimulates ErbB2 and ErbB3 heterodimers, resulting in the activation of the downstream signaling pathways. By means of reverse transcription polymerase chain reaction (RT-PCR) and IHC, high EREG protein expression in GC was significantly associated with TNM stage distant metastases as well as poor overall survival [163].
Sphingosine kinase 1 (SPHK1), which catalyzes the phosphorylation of sphingosine to S1P, promotes cell growth, proliferation, and anti-apoptosis. In particular, SPHK1 plays an oncogenic role in promoting survival and invasion in some tumors [164, 165]. In GC, SPHK1 expression was associated with peritoneal recurrence and poor prognosis [166].
Furthermore, long noncoding RNAs seem to play a role in tumor development and disease spreading. Metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) is one of the most studied long noncoding RNAs. It regulates alternative splicing by modulating the phosphorylation and distribution of pre-messenger RNA splicing factors [167], which leads to the activation of a growth control program [168]. Overexpression of MALAT1 was observed in various types of solid human carcinomas [169‒171].
HOTAIR (HOX-antisense intergenic RNA), another lncRNA, was also observed to promote cancer metastasis development. Increased expression of HOTAIR was found in several tumors [172] and seemed to correlate with presence of metastasis in breast and colon cancer patients [173, 174]. In GC, levels of MALAT1 and HOTAIR were revealed as independent risk factors for PD [175].
Metabolism
Several cellular metabolic changes are necessary for the survival of neoplastic cells in free abdominal space [72, 113]. The hypoxic microenvironment has a recognized role in regulating cell stemness, which can determinate tumor metastasis and stimulate a more aggressive neoplastic phenotype [176, 177]. These cellular responses are primarily controlled by the hypoxia-inducible factors [178‒180]. It was demonstrated that HIF-1a protein in GC correlated with the development of PD [181]. Furthermore, procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2 (PLOD2), is regulated by hypoxia-inducible factor-1 (HIF-1) and mediates extracellular matrix remodeling, alignment, and mechanical properties [182, 183]. PLOD2 expression was a significant and independent predictive factor for peritoneal recurrence in GC [184].
Phosphatidylinositol transfer protein cytoplasmic 1 (PITPNC1) was observed to favor malignant secretion of pro-metastatic factors, and it was also upregulated in metastatic breast cancer, melanoma, and colon cancer [185]. Elevated expression of PITPNC1 in GC (phosphatidylinositol transfer protein, cytoplasmic 1) demonstrated to be a predictor of omental metastasis [186].
Fatty acid synthase (FASN) is an enzyme of lipogenesis; it is controlled by hormones, growth factors, and diet. In tumor tissues, the fatty acid supply is highly determined on de novo synthesis thanks to FASN. Indeed, some studies have demonstrated that FASN is overexpressed in several cancers [187‒190]. In GC, FASN expression was significantly associated with PD and with shorter OS [191].
Methionine sulfoxide reductases (MSRs) are repair enzymes that reduce methionine sulfoxide residues [192]. They are important in protecting oxidative damage by eliminating cellular reactive oxygen species [193]. Moreover, MSR-B3 plays a role in tumorigenesis in breast cancer and is a predictor of poor survival in bladder urothelial carcinoma patients [194, 195]. In GC, MSR-B3 was shown to be a significant risk factor for the presence of PM and for unfavorable prognosis [196].
Miscellaneous
In literature, there are many molecules that are demonstrated to be correlated to peritoneal spread of gastric adenocarcinoma, but in some cases their mechanism of action is still unknown. Zang et al. [197] observed that LPPR4 (lipid phosphate phosphatase-related protein type 4) was significantly overexpressed in GC patients samples with PD. In vivo, LPPR4 promotes migration, invasion, and adhesion of GC cells through the Sp1/integrin α/FAK signaling pathway [197]. In GC, synaptotagmin XIII (SYT13) expression seemed to be significiantly higher in patients with PM [198].
Recently, tartrate-resistant acid phosphatase-5 (ACP5), which is considered to be a classic marker for bone resorption and osteoclast differentiation [199], was revealed to be a potent proinvasive gene [200, 201]. To date, no studies have reported the significance of ACP5 in human GC. Kawamura et al. [202] observed that ACP5 (tartrate-resistant acid phosphatase type-5) expression was significantly associated with PD and a poor prognosis in GC. Sawaki et al. [203], analyzing GC specimen, showed that TNNI2 expression was associated with PM and that it was an independent marker of peritoneal recurrence after surgery.
Telomerase activity may play a role in cancer evolution by favoring the selection and survival of neoplastic clone subpopulations. Telomerase is a ribonucleoprotein that synthesizes TTAGGG tandem repeats at each telomeric region to prolong it back to its original length [204]. The enzyme is effective in embryonic cells and stem cells, but its activity is undetectable in normal differentiated somatic cells [205]. Cells that can overcome this restriction have the potential for extended survival and unlimited proliferation [206]. In GC, high telomerase activity was revealed to correlate with the presence of PM [207].
It was found that also miRNAs may participate in regulation of cell proliferation, differentiation, metabolism, and apoptosis [208‒210]. Moreover, an increasing number of studies suggested that miRNAs are involved in the metastatic spread and progression of GC [211‒214]. However, evidence from a previous study has shown that miR-125a-5p was downregulated in GC [215]; the specific mechanisms of miR-125a-5p in metastasis development and invasion in GC have not been completely understood. Cao et al. [208] observed that in GC the expression of miR-125a-5p was markedly decreased in patients with PM compared with patients without PM. Also, downregulation of miR-125a-3p was correlated with PD and poor prognosis. In vitro assays showed that miR-125a-3p suppressed the proliferation of GC cells [209]. Shimura et al. [216] observed that miR-30a-5p, -659-3p, and -3917 were significantly overexpressed in GC of patients with PD compared to patients without peritoneal involvement.
Increasing evidence demonstrated that also lncRNAs are involved in the tumorigenesis and play a role in several cellular process [217]. LINC00958, originally reported by Seitz et al. [218] in bladder cancer, was observed to be metastasis associated and plays an oncogenic role. The investigation of LINC00958 was subsequently extended to glioma [219] and endometrial cancer [220], suggesting that LINC00958 was associated with metastases of cancers. In GC, LINC00958 expression pronouncedly correlated with PD [221].
Molecular markers of PD, when validated in clinical practice, might also play a role in defining prognosis of GC patients. Lee et al. [221] developed a transcriptomic signature based on a six-gene panel for risk stratification and characterization of high-risk patients with peritoneal carcinomatosis which might help in clinical decision-making.
Molecular Markers in Peritoneal Lavage
Several molecules, such as CA 125 [222], IL-17 [223], MMP-7 [224], surviving [225] MUC2, FABP1, TFF1, and MASPIN [226], were identified in literature as marker of peritoneal spreading of GC in peritoneal lavage (Table 2). CEA was the most recurrently studied and in literature was observed that positive CEA alone [224, 227‒231] or in combination with CK20 [230, 231] is a marker of peritoneal recurrence of disease and of a shorter survival. Moreover, SYT123 mRNA in peritoneal lavages was correlated with peritoneal recurrence [228], while Ber-EP4 was a predictor of positive peritoneal cytology and poor prognosis [232, 233].
Many different molecules were revealed to be associated with positive peritoneal cytology. Dopa decarboxylase levels were demonstrated to be a good predictor of diagnosis of synchronous PD in GC patients [234]. Levels of lysophosphatidic acid were also observed to be elevated in a patient with peritoneal disease and could be considered a biomarker of PD [235]. Telomerase activity in peritoneal lavage was observed to be associated with PD [207, 236, 237] and CD01 gene promoter DNA methylation was demonstrated to be a marker of peritoneal micrometastasis [238].
Furthermore, GC-derived peritoneal exomoses were considered a possible marker of positive cytology. Ohzawa et al. [239] demonstrated that exosomal miR-29s were correlated with PD and peritoneal recurrence.
Defining molecular markers of PD in peritoneal lavage could complement to cytology. Moreover, if these indicators are present also in metastatic peritoneal tissue, they might be considered a target for tailored therapy in patients with PD.
Molecular Profiling of Malignant Ascites and PD Cells
Most of the molecular profiling of GC was developed based on the analysis of primary tumor samples. However, GC primary lesion and peritoneal disease seem to be characterized by different molecular patterns [13, 14]. Molecular profiling of PM and malignant ascites could contribute to understand PD mechanisms better.
Wang et al. [240] conducted the first whole-exome and transcriptome sequencing of peritoneal carcinomatosis cells from 43 patients affected by GC. They revealed two molecular subtypes, “mesenchymal-like” (M) and “epithelial-like” (E), with different response rates to chemotherapy (31% vs. 71% in M vs. E, respectively). Compared to the E subtype, the M subtype demonstrated a higher expression of mesenchymal signature genes, lower expression of epithelial markers, and it showed to be more genomically and chromosomally stable. Furthermore, the M subtype had a higher expression of transforming growth factor-β pathway genes, submitting an immune-suppressive microenvironment. In addition, immune checkpoint TIM-3, its ligand galectin-9, and VISTA were highly expressed in this subtype [240].
Afterward, Tanaka et al. [241] conducted a comprehensive multi-omic analysis of malignant ascitic fluid samples and their corresponding tumor cell lines from 98 GC patients. They showed higher rates of receptor tyrosine kinase and mitogen-activated protein kinase pathway alterations compared to primary GC. Moreover, the analyses identified two distinct molecular subtypes: one presenting active super enhancers at the ELF3, KLF5, and EHF loci and a second characterized by transforming growth factor-β pathway activation through SMAD3 super-enhancer activation and high expression of transcriptional enhancer factor TEF-1 (TEAD1) [241].
Lim et al. [242], by analyzing whole-exome sequences of normal gastric tissues, primary tumors, and malignant ascites from 8 GC patients, revealed a distinctive mutational signature. The comparative analysis demonstrated some recurrent mutations for GC peritoneal carcinomatosis: mutations in COL4A6, INTS2, and PTPN13; mutations in druggable genes including TEP1, PRKCD, BRAF, ERBB4, PIK3CA, HDAC9, FYN, FASN, BIRC2, FLT3, ROCK1, CD22, and PIK3C2B; and mutations in metastasis-associated genes including TNFSF12, L1CAM, DIAPH3, ROCK1, TGFBR1, MYO9B, NR4A1, and RHOA. Notably, gene ontology analysis revealed the significant enrichment of mutations in the Rho-ROCK signaling pathway-associated biological processes in malignant ascites.
Diffuse GC
Focusing on the classic histological classification of GC, one of the oldest and most used is Lauren’s classification, which divides gastric adenocarcinoma into intestinal (54%), diffuse (32%), and indeterminate type (15%). The intestinal type has a tubular or glandular differentiation and is often sporadic and usually associated with H. pylori infection. The diffuse type, which corresponds to poorly cohesive carcinoma according to the World Health Classification System (2019), is histologically composed of discohesive cells and molecularly characterized by the loss of ECD expression [243].
Besides having different microscopic appearances, the two subgroups have also distinct clinicopathological and molecular features. Several studies demonstrated that intestinal GC tends to metastasize to the liver, while diffuse GC is frequently associated with PD [244‒247]. However, the molecular mechanisms underlying the correlation between diffuse type and carcinomatosis are still obscure.
Lorelei-like GPI-anchored protein 1 (LLGL1) has been associated with loss of cellular adhesion and dissemination of cancer cells in many cancer types. Desuki et al. [248] reported that loss of LLGL1 expression significantly correlated with loss of ECD expression, with the diffuse type of GC, and with peritoneal carcinomatosis.
Kim and colleagues [249] analyzed the expression of CLDN18.2 and ECD by means of immunofluorescence in a cohort of 77 diffuse metastatic GC patients. The expression of these two markers was significantly lower in patients with PM than those without PM at the time of diagnosis. Furthermore, it was lower in patients who developed PM from diagnosis to death than in those who never did [249].
Conclusion
PD is a major cause of mortality in GC patients, and current treatment options are limited. For this reason, existing and future research should aim at tailoring the surgical and oncological treatments. To this date, several molecular determinants of PM have been identified. However, most of the molecular processes and molecular factors that we summarized in our review play a role in the general carcinogenetic cascade. A specific molecular signature of peritoneal carcinomatosis in GC and a comprehensive understanding of its complex molecular landscape are still lacking. More in vitro and in vivo studies and molecular profiling of metastatic peritoneal lesions from GC patients are needed. Of note, it is essential to integrate this growing amount of molecular data with the histopathologic characterization to find a morpho-molecular signature of gastric carcinomatosis.
Conflict of Interest Statement
Matteo Fassan reports personal fees (as speaker bureau or advisor) from Roche, MSD, GSK, Astellas Pharma, and Diaceutics and received research grants from Astellas Pharma, QED therapeutics, and macrophage pharma, unrelated to the current work.
Funding Sources
Matteo Fassan is supported by a grant from the Italian Health Ministry/Veneto region research program NET-2016-02363853 and AIRC 5 per mille 2019 (ID 22759 program).
Author Contributions
Valentina Mari, Valentina Angerilli, Giada Munari, Marco Scarpa, and Quoc Riccardo Bao performed literature research; Salvatore Pucciarelli, Matteo Fassan, and Gaya Spolverato: concept of the study and review of the manuscript; and Valentina Mari, Valentina Angerilli, and Giada Munari: draft of the manuscript. All the authors approved the final version of the manuscript.
References
Additional information
Valentina Mari and Valentina Angerilli: co-first authorship.Matteo Fassan and Gaya Spolverato: co-last authorship.