1. IntroductionDespite its decreasing incidence, gastric cancer (GC) remains one of the most common causes of death for neoplasms worldwide. Recent epidemiological trends have indicated a relative increase in the rate of the diffuse histotype at the expense of the intestinal type in the West. This could explain the lack of improvements in survival and the increasing rates of peritoneal carcinomatosis in recent periods [1,2]. The intestinal and diffuse types of GC, aside from their morphological features, show evident pathological and clinical differences. The former is more common in males and older patients, whereas the diffuse type usually affects younger patients. The risk of lymph node metastasis is significantly higher in the diffuse type, as it is for extra-regional and para-aortic nodes, and the propensity for peritoneal recurrence is relevant, especially when the serosa is involved [3,4]. The aggressive features and peculiar pattern of spread have caused the consideration of different and more specific treatment approaches for this subgroup of GC, such as super-extended lymphadenectomy and locoregional treatments for the prevention of peritoneal recurrence [4]. However, in several reports, the Lauren histotype did not show an independent prognostic value, and this indicated that relevant heterogeneity may exist in the context of diffuse-type tumors [5,6,7].In the last years, two molecular classifications of GC (The Cancer Genome Atlas Program, TCGA, and Asian Cancer Research Group, ACRG) have been introduced, and extensive research is now ongoing to explore potential clinical applications of different molecular subgroups [8,9,10]. Both classifications showed a simple division of GC into four subgroups; the ACRG divided GC into microsatellite instability (MSI) and microsatellite-stable tumors (MSS) group, with subsets of MSS tumors for epithelial-to-mesenchymal transition (EMT), TP53-positive, and TP53-negative groups [9]. The EMT is a process in which epithelial cells transform into mesenchymal-like cells that are capable of migration, invasion, and metastasis [11]. This process is associated with the downregulation of epithelial markers, such as E-cadherin, upregulation of mesenchymal markers, and involvement of the zinc finger E-box binding homeobox 1 (ZEB-1), a transcriptional repressor of E-cadherin expression and EMT promoter [12,13,14]. The stem-cell marker CD44, a non-kinase cell-surface transmembrane glycoprotein, is also a recognized immunohistochemical marker of the EMT phenotype [12].The clinical significance of the EMT has been investigated in several neoplasms, and it has been found to be associated with advanced stages and poor prognosis, including in GC [9,12,15]. However, a strong association between EMT status and Lauren histology has been found, as more than 80% of EMT-positive tumors are of the diffuse histotype [9]. Genomic analyses showed that the signaling pathways of the EMT identify the so-called mesenchymal type of GC, and most of such cases are of diffuse histology [16]. Interestingly, poorly cohesive GC also showed a worse prognosis when associated with EMT signature [17]. To date, it is unclear if the association with diffuse histology acts as a confounding factor, or, alternatively, if EMT status could identify a more aggressive phenotype in the context of the diffuse-type group.

The aim of the present retrospective study was to evaluate the potential impact of EMT status on the clinical–pathological characteristics and survival of patients with MSS diffuse-type tumors treated with gastrectomy and lymphadenectomy, with a long follow-up period.

2. Materials and Methods 2.1. PatientsPatients with primary GC who submitted to surgical resection at the Department of General Surgery and Surgical Oncology, University of Siena, Italy, between 1997 and 2012 were considered. For the present study, the following selection criteria were adopted: (1) availability of a tumor specimen in the tissue bank; (2) the patients submitted to up-front resective surgery, and neo-adjuvant or non-resected cases were excluded; (3) MSS tumors, with the exclusion of MSI tumors; (4) diffuse histotype according to Lauren classification, with the exclusion of intestinal or mixed types. All clinical and pathological data were retrospectively retrieved from a prospectively collected database, as previously reported [1,18]. Informed consent for the use of clinical data and for molecular analysis was obtained from all patients at the time of hospital recovery. All clinical, pathological, and molecular procedures or analyses were performed in accordance with the institutional review committee of University of Siena. 2.2. Surgical Approach, Histopathology, and Staging

After routine clinical staging with a CT scan, patients with potentially resectable GC were addressed to surgery. The aim of the surgical procedure was complete resection of the tumor with negative resection margins and extended (D2, D2plus) lymphadenectomy. For tumors located in the lower and middle third, a subtotal gastrectomy was performed, maintaining a proximal margin of at least 5 cm from the tumor; in all other cases, a total gastrectomy was performed. After surgery, cases were classified as R1/R2 when microscopic (peritoneal cytology, resection margins) or macroscopic residual tumor was present.

Immediately after surgical resection, lymph node mapping was performed in the fresh specimen according to the Japanese classification, as previously reported. The number of removed and involved lymph nodes for each station was recorded. Lymph node stations were also divided into perigastric (stations 1 to 7), second level (stations 8 to 12) and third level stations (stations 13 to 16) according to the Japanese guidelines [19]. For TNM staging, all cases were re-classified according to the 8th edition of the AJCC/UICC TNM classification. All other pathological factors were classified as previously reported in other studies by our group and according to the last WHO classification of digestive system tumors [18]. 2.3. Microsatellite Analysis and Immunohistochemistry (IHC)Frozen samples of tumor tissue were retrieved for each patient in our biological bank. Microsatellite analysis was evaluated by using 5 quasi monomorphic mononucleotide repeats, namely, BAT-26, BAT-25, NR -24, NR-21, and NR-27, as previously reported in detail [18]. We considered a tumor as having MSI whenever 2 or more markers showed instability on 5 loci (MSI-H).

For IHC, all procedures were automatically carried out on representative 2 µm thick paraffin sections from each case by Benchmark Ultra (Ventana, Monza, Italy) by using extended antigen retrieval and with DAB as chromogen. The following antibodies and dilutions were used: E-cadherin (NCH-38, #M3612 Dako Denmark A/S, 1:100), CD44 (ab6124, #F10-44-2, Abcam, 1:100), and ZEB-1 (2C1a, #ab50887 Abcam, 1:10). Counterstain with Mayer’s Hematoxylin was applied for 2 min to each section. For each case, the IHC expression in the malignant cells was compared to the immunoreaction in normal epithelial and/or stromal cells (positive control). All markers were scored independently by two observers who were blinded to clinical data and to each other.

For E-cadherin, a three-tiered score of positivity was applied by evaluating the percentage of stained tumor cells (0–10% = none; 11–89% = 1; >90% = 2), whereas for CD44 and ZEB-1, a two-tiered score was used (presence: immunostaining in 10% or more of the epithelial tumor cells; absence: immunostaining in less than 10%).

Different patterns of expression (membranous, cytoplasmic, nuclear) were considered for different markers. The expression of E-cadherin was classified as follows: (1) “normal/positive”: membranous immunostaining in more than 10% of the epithelial tumor cells; (2) “lost/negative”: absence of immunostaining or positivity in less than 10% of the epithelial tumor cells; (3) “abnormal”: cytoplasmic or heterogeneous expression (between 10 and 90% of the tumoral cells were immunopositive at the membrane and cytoplasmatic level). Altered expression of E-cadherin was defined as either complete loss of membrane staining or an aberrant cytoplasmic staining pattern [20,21]. Membranous expression for CD44 (>10%) and nuclear expression for ZEB-1 (>10%) were considered for the assessment of marker positivity [22,23]. Samples with differences in assessments between the two investigators were re-evaluated, and a consensus decision was made. Because the staining patterns sometimes varied within the same tumor, particularly when the differentiation status varied, the final score was based on the dominant pattern, and the highest value was used for the statistical analysis.We considered all samples in which E-cadherin was aberrant (negative/abnormal) and at least one of the EMT-related markers (CD44 or ZEB-1) was positive as showing an EMT phenotype [12,24] (Figure 1 and Figure 2). 2.4. Additional Treatments and Follow-UpAdjuvant chemotherapy was proposed after surgery according to the tumor stage, R-status, and the patient’s general conditions and age. Then, patients were subjected to regular outpatient follow-up examinations according to a standard protocol [5]. The follow-up was closed in December 2021. Follow-up data were available for all patients under study. The median follow-up period was 30 months (range: 1–240) for the entire series and 96 months (range: 16–240) for patients who were still alive at the last check-up or had non-cancer-related deaths. 2.5. Statistical Analysis

Statistical analysis was performed by using the SPSS statistical software (21.0). Numerical variables were expressed as the mean and standard deviation (SD) when normally distributed and as the median and interquartile range (IQR) if not normally distributed. Statistical associations between clinicopathological characteristics and EMT status were assessed with a χ2 test for categorical variables and with a t-test or non-parametric tests (Mann–Whitney) for continuous variables. The linear regression (R2) was used to explore the correlations between two linear variables. Survival curves were estimated by using the Kaplan–Meier method. Death due to cancer (cancer-specific survival) was considered as the endpoint; as such, deaths due to causes other than gastric cancer were considered censored cases at the time of death. Survival curves were compared by using the log-rank test. A p-value < 0.05 was considered statistically significant.

4. DiscussionThe new molecular classifications of GC are recent and important steps in the management of this complex disease, and the flourishing of the literature is ongoing in order to clarify their potential clinical implications in clinical practice [4]. However, few clinical-oriented studies with detailed surgical data have been reported in the literature, and short follow-up periods are generally available for survival analyses.The MSI group is the most studied in the literature, and it has been demonstrated to be associated with older age, less advanced stage, limited lymph node involvement, and better prognosis [8,9,15,18,25]. The EMT group of the ACRG classification is also very interesting from a clinical point of view [9]. The EMT is a process in which epithelial cells are transformed into cells with mesenchymal phenotypes, loss of cellular adhesion, enhanced invasive properties, and abundant stroma. These alterations, together with microenvironment remodeling, facilitate the aggressiveness, metastasis, and chemo-resistance of GC [11,14].Some reports have suggested that the EMT phenotype is correlated with younger age, diffuse/poorly differentiated histology, advanced TNM stage, and worse prognosis of GC [9,11,12,15,16,26,27,28,29]. The strong association with the diffuse histotype or undifferentiated histology may act as a confounding factor, and it is unclear if this association could explain the clinical behavior of EMT-positive tumors [9,12,28,29]. For this reason, we focused our analysis on MSS diffuse-type tumors. A separate analysis of molecular features according to the Lauren histotype was also suggested by other authors [20]. To our knowledge, this is the first report to identify EMT status as a strong marker of tumor progression and poor prognosis in MSS diffuse-histotype GC patients with a long follow-up time.In the ACRG study, more than 80% of the EMT cases were diffuse, but less than 50% of the diffuse-type cases belonged to the EMT subgroup [9]. The percentage of diffuse-type cases showing an EMT phenotype in our series (54%) was like the rate found in the ACRG study, as well as other reports [9,20,29]. In several studies, the Lauren classification was not used, and this did not allow a comparison with the present series [12,15].Most clinical and pathological variables analyzed in the present study were similar according to the EMT status; however, the EMT-positive group showed a strong association with extensive lymph nodal metastases. The extended lymphadenectomy (median number of 41 removed lymph nodes) and the routinary mapping of lymph node stations allowed an accurate topographic analysis of nodal spread according to the molecular type. The median numbers of positive lymph nodes and involved nodal stations were notably higher in the EMT-positive group; only three patients (6.7%) had negative lymph nodes, and 71% of cases were pN3. Although it is well known that the Lauren histotype affects the incidence, number, and distant location of lymph node metastases in GC, the EMT status could characterize a phenotype with more extensive nodal spread. The strong linear correlation between the numbers of removed and positive nodes is indicative of the biological aggressiveness of EMT-positive tumors and their propensity to invade multiple nodal chains (Figure 3). However, according to the present data, this seems to happen, above all, for perigastric nodes, whereas the overall involvement rates of second- and third-level nodal stations were similar in the two groups under study. As such, from a surgical point of view, the extent of lymphadenectomy should not change according to EMT status in the diffuse histotype of GC.Another important finding of the present study is the prognostic impact of EMT status in diffuse-type tumors. Most EMT-positive patients died due to GC during follow-up, with a median survival of 26 months vs. 51 months for negative cases, and the prognosis was particularly poor in advanced nodal stages and neoplasms involving the serosa. A greater propensity of EMT-positive tumors for peritoneal dissemination was also observed; on the contrary, the long-term survival and peritoneal spread of EMT-negative cases resemble the characteristics of the intestinal type. As such, the EMT phenotype may play a crucial role in the clinical correlation between diffuse-type and peritoneal carcinomatosis. Although the diffuse type has been proposed to be included in a unique pathological/molecular category, the present data indicate that a phenotype with more aggressive clinical behavior can be identified [30]. Interestingly, EMT status was equally predictive of worse survival in both signet- and non-signet poorly cohesive carcinomas of the WHO classification, which are emerging prognostic factors in GC [31,32]. These aspects should be better elucidated in future specific studies.

The present study included patients who were treated in the pre-neoadjuvant era and then submitted to up-front surgery, even for locally advanced forms. Although this may be considered a limitation, the analysis of the correlations among EMT status, tumor advancement, and lymph node spread is reliable in neoplasms that are not submitted to neo-adjuvant treatments. These results indicate that patients with EMT-positive diffuse-type tumors have a poor probability of surviving in the long term when submitted to up-front surgery. The prognostic value of EMT status in patients submitted to neo-adjuvant chemotherapy remains unclear and should be analyzed in future studies.

In our series, a favorable prognostic impact of negative EMT status was observed in patients submitted to adjuvant chemotherapy. Interestingly, this phenotype was associated with better prognosis at younger ages, whereas the long-term survival of aged patients was generally poor in both groups, independently from EMT status. It is important to note that almost all patients (93%) in the younger group were treated with adjuvant chemotherapy, in comparison with 48% of the aged patients. These data may fit with the hypothesis that EMT status could affect the response to chemotherapy in GC [11,33], but specific studies should be designed in the future.The limitations of the present study include its retrospective design, its relatively small sample size, and the methods used for EMT evaluation. Although IHC may represent a limitation when compared with genomic analysis, this method is strongly encouraged for the molecular characterization of GC due to its limited costs. When confirmed in further studies, it could be an important gain for potential applications in a clinical setting [20,27,28].The present findings need confirmation in external validation studies, and correlation analyses that are performed on endoscopic biopsies are necessary before drawing clinical and therapeutic implications. Potential changes in EMT markers induced by chemotherapy should also be considered. However, the present results indicate that the analysis of EMT status could expand the molecular characterization of GC. Preoperative biopsy of GC should focus on the assessment of MSI, EBV, and EMT phenotypes. The evaluation of this panel of molecular markers could entail indications for neo-adjuvant therapy, up-front surgery, and, in EMT-positive cases, a multimodal approach that includes hyperthermic intraperitoneal chemotherapy (HIPEC) or pressurized intraperitoneal aerosol chemotherapy (PIPAC), in consideration of the greater propensity for peritoneal spread [34]. Potential therapies with EMT inhibitors are under evaluation in clinical studies, and this could be another important step in a specific therapeutic approach to diffuse GC [35,36].