Introduction

Gastric cancer (GC) remains a significant global health issue, accounting for 1,089,103 new cases in 2020 and an estimated 768,793 deaths, making it the fifth most common and fourth deadliest cancer worldwide [1]. Incidence rates are twice as high in men as in women. Most patients are diagnosed at an advanced stage where outcomes are poor, with a 5-year survival rate of less than 40%, even with multimodal treatment approaches [2]. The primary reasons for these disappointing results are tumor aggressiveness and resistance to therapy. Objective response rates to conventional chemotherapeutic regimens range from 20—40% [3]. Targeted therapies for gastric cancer include trastuzumab for HER2-positive cases [4], ramucirumab as a second-line therapy [5], and pembrolizumab for tumors expressing PD-L1, as a third-line treatment. In recent decades, the personalized treatment of tumors using targeted drugs at various stages of carcinogenesis has become widespread in oncological practice [6].

Promising directions in the development of personalized approaches for the treatment of GC include studying its genomics and molecular phenotype. Since 2013, several research groups have attempted to create molecular genetic classifications of GC, the largest of which are TCGA [7] and ACRG [8].

To adapt molecular genetic classifications to routine practice, several groups have proposed their immunophenotypic analogues, which do not require complex and costly research methods. R. Gonzales and colleagues demonstrated the possibility of using IHC markers to classify tumors into CIN (p53) and MSI (MMR proteins) subtypes and using in situ hybridization to identify the EBV-associated subtype [9]. Two other groups also proposed IHC detection of E-cadherin to recognize genomically stable tumors [10, 11], since the lack of E-cadherin expression is characteristic of diffuse, the most common histological type among GS-subtype of gastric adenocarcinomas according to TCGA observations [7]. A comparison of the immunohistochemical classifications of gastric cancer available in the literature is presented in table 1.

Таблица 1. Сравнение существующих иммуногистохимических классификаций рака желудка

The analysis of attempts to classify GC by immunohistochemistry [9, 11—13] shows that in all cases after identifying MMR-deficient, EBER-positive, and E-cadherin aberrant subtypes, there remains a large (more than 50% of cases) and quite heterogeneous group of tumors, often referred to as having aberrant p53 expression. N. Setia and co-authors proposed further subdivision of this group based on mucin expression, although no differences in clinical and prognostic parameters were found between the identified groups.

Thus, significant progress has been made in the development of molecular classifications of GC in recent years, but there is a need for further refinement of existing approaches.

Material and Methods

Tissue Samples

The study used surgical samples (full-sized surgical material obtained during gastrectomy) from 310 patients with a confirmed diagnosis of GC, operated on from 2010 to 2018. All patients had not received chemotherapy, radiotherapy or any immunotherapy prior to surgery. Exclusion criteria included: carcinoma in situ, insufficient material in paraffin blocks, verified neuroendocrine cancer, gastric lymphoma, or gastrointestinal stromal tumor of the stomach. The age of the patients ranged from 22 to 85 years (median age — 63 years). The median follow-up period for patients was 83 months. Collected clinical and pathological features include gender, age (median age 63 years), maximum size (median 8 cm), macroscopic type by R. Bormann, histological type (WHO 2019), Lauren type (diffuse, intestinal, intermediate), TNM stage, tumor grade (WHO 2010: G1, G2, G3 and WHO 2019: LG and HG), tumor site (proximal: cardia, body/fundus; distal — antrum; subtotal, total lesion), lymphovascular and/or venous invasion, and overall survival (months). At least three gastrointestinal pathologists reviewed all H&E slides (NVD, DVK, PGM).

Biomarker Panel

Expression for 23 different biomarkers was assessed (MSH2, MSH6, MLH1, PMS2, MUC2, CDX2, MUC5AC, CD10, E-cadherin, β-catenin, Claudin-1, Claudin-3, Claudin-4, CD44, p53, PD-L1 (SP142), PD-L1 (SP236), HER2, CD4, CD8, CD68, CD1a, LMP-1) along with EBER in situ hybridization. The specifications and dilution, location (membranous, nuclear, and cytoplasmic) for all biomarkers are detailed in table 2. At least two gastrointestinal pathologists reviewed all immunohistochemical stains and in situ hybridization (NVD and PGM).

Таблица 2. Характеристика биомаркеров

Immunohistochemistry

Immunohistochemical reactions were conducted using the UltraVision Quanto detection system (Thermo Fisher Scientific, USA), following the manufacturer’s protocol in an Autostainer 480S immunostainer (Thermo Fisher Scientific, USA). Deparaffinization, rehydration, and antigen retrieval were performed using Thermo Dewax and HIER Buffer L (pH 6.0), Thermo Dewax and HIER Buffer M (pH 8.0), and Thermo Dewax and HIER Buffer H (pH 9.0) at 95—98°C for 20 minutes in a pre-treatment module (PT-Module, Thermo Fisher Scientific, USA). The buffer selection was tailored individually for each specific antibody according to the manufacturer’s instructions. Only for the LMP-1 marker, the manufacturer recommended a buffer with pH 9.0 for deparaffinization and antigen retrieval; however, its use resulted in significant background staining, leading to the selection of the buffer with pH 6.0. The antibody incubation time was 20—30 minutes following the manufacturer’s guidelines.

Reactions with the HER2/neu marker were performed using the HercepTest kit (Dako/Agilent Technologies, USA) according to the manufacturer’s instructions. PD-L1 expression was determined using Roche Ventana, USA SP142 and SP263 clones and the OptiView DAB IHC Detection Kit. PD-L1 reactions were carried out using the Ventana BenchMark Ultra device (Roche Ventana, USA) following the standard protocol no later than 24 hours after the preparation of paraffin sections.

In situ hybridization

In situ hybridization was conducted using primers to small Epstein-Barr virus RNAs (INFORM EBER, Roche Ventana, USA) and the ISH iVIEW Blue Detection Kit visualization system (Roche Ventana, USA). In situ hybridization reactions were carried out using the Ventana BenchMark Ultra device (Roche Ventana, USA). The results were evaluated according to established standards [21, 22]. The reaction was considered positive if the EBER signal was localized in the nuclei of tumor cells, but there was no cytoplasmic staining [21]. A positive control using a tissue sample from EBV-associated nasopharyngeal carcinoma was included in each set of reactions.

The obtained results were compared with the main clinical-morphological characteristics of GC.

Scoring Criteria

The results were evaluated using a DM2500 light microscope (Leica Microsystems, Germany) by two independent researchers. The SCN400 slide scanner and DM4000B/DFC495 microscope (Leica Microsystems, Germany) were used to obtain microphotographs.

PD-L1 expression was assessed using the Combined Positive Score — CPS [14]. A positive PD-L1-status was established at CPS>1. Tonsil and placenta tissues were used as controls.

The assessment of reactions with microsatellite instability markers was conducted according to previously published methodologies [11, 12]. Tumors were classified as MMR-deficient if there was a lack of nuclear expression of at least one of the markers (MLH1, MSH2, MSH6, and PMS2), with a positive reaction in surrounding normal epithelial and smooth muscle cells, as well as lymphocytes. If there was positive expression of the markers, tumors were considered MMR-positive, i.e., without defects in the DNA repair system genes. Cases with no expression in both tumor cells and surrounding normal cells were excluded from analysis.

Mucin expression was evaluated following previously published methodologies [15—18]. CDX2 was assessed according to the methodology of K. Roessler and colleagues [19]. Depending on the expression of MUC2, CD10 and MUC5AC, all tumors were divided into 4 groups (MUC status): tumors with expression of only intestinal differentiation markers (MUC2 and/or CD10, MUC-I status), tumors with expression of the gastric differentiation marker (MUC5AC, MUC-G status), tumors with expression of MUC2, CD10 and MUC5AC (mixed, MUC-mix status), tumors in which there was no expression of all markers (MUC-null status).

The evaluation of claudin expression was conducted by two independent researchers following the methodology of T.L. Hwang et al. [4]. The expression of E-cadherin, β-catenin, and p53 was assessed using the methodology proposed by N. Setia [11]. CD44 expression was evaluated as the proportion of positively stained cells among all tumor cells. HER2 status was determined using the commonly accepted assessment system [20].

The assessment of immune microenvironment markers (CD4, CD8, CD68, CD1a) involved quantitative counting of cells within the tumor infiltrate in three fields of view, calculating the average. Three fields of view were selected based on the areas of greatest positivity (hot spots). The methods for assessing biomarker expression are summarized in table 2.

Statistics

Statistical analysis was performed using PASW Statistics 22 software. For comparing quantitative data between two unrelated samples, Student’s T-test was used for normally distributed parameters, and the Mann-Whitney U test was applied for parameters not normally distributed or those on an ordinal scale. The Kruskal-Wallis test was utilized for multiple comparisons of quantitative data. If significant differences were identified after multiple comparisons, pairwise group comparisons were subsequently performed using the Mann-Whitney U test. For multiple comparisons of qualitative data, the Chi-square test was used, followed Fisher’s tests as necessary. Correlation analysis was performed using Spearman’s rank correlation coefficient.

Survival analysis was conducted using the Kaplan-Meier method. Survival characteristics were presented as the mean survival time and standard error of the mean, 95% CI for the mean survival time, and the median survival. Survival comparisons were made using the log-rank test. The predictors of survival were evaluated using a multivariable Cox regression model with a stepwise Wald inclusion algorithm.

Results

A Five-Tier Classification of Gastric Adenocarcinoma

Out of 310 cases, mismatch repair (MMR) deficiency (absence of reaction with one or several MMR markers) was identified in 25 cases (8.1%). Among these 25 MMR-deficient tumors, only in three cases (12%) was there an absence of MSH2/MSH6 expression, whereas in the remaining cases, MLH-1/PMS2 expression was absent (88%). Therefore, relying solely on MLH-1/PMS2 to assess microsatellite instability in GC is impractical, as it may lead to missing some patients with microsatellite instability.

Tumors with positive Epstein-Barr encoding region (EBER) expression occurred in 9% of our sample, which aligns with literature data [23—25]. In situ hybridization results did not correlate with the expression of the Epstein-Barr virus immunohistochemical marker LMP-1 (r=-0.069, p=0.229). The frequency of aberrant E-cadherin expression was 19.4%.

As a result of multistage statistical analysis, we formulated a step-by-step algorithm to determine whether a tumor belongs to a molecular subtype (Fig. S1). During the algorithm development process 17 variants of subdivision into molecular subtypes were analyzed. The following subtypes were statistically shown to be unreasonable: EBER/PD-L1, E-cadherin/β-catenin-aberrant, PD-L1-positive, as the selection of these subtypes does not reflect differences in patient survival. The detailed steps in the development of an algorithm to separate GC into molecular subtypes are summarized in fig. 1.

Рис. 1. Алгоритм разбивки случаев рака желудка на молекулярные подтипы с помощью иммуногистохимического метода.

Based on the analysis of survival curves and other statistical data, the most preferable classification option is the one that provides for the isolation of the MMR-negative subtype at the first stage (25 cases, 8.1%), at the second stage EBER-positive subtype (28 cases, 9%), then E-cadherin-aberrant subtype (53 cases, 17.1%) and finally p53-null-subtype (204 cases, 65.8%), because in this variant the differences between subtypes have the highest significance (p=0.001).

P53-null-subtype is the most numerous and heterogeneous group of lesions, which requires further subdivision. To identify criteria for separation, we reanalyzed the expression of immunohistochemical markers (β-catenin, p53, PD-L1 SP163, PD-L1 SP142, MUC2, CD10, MUC5AC, MUC-status, HER2-status, claudin-1, claudin-3, claudin-4) within this subtype. It was found that patients with positive CDX-2 expression have significant survival benefits compared to those with negative CDX-2 expression; thus, the remaining group was divided into CDX2-positive (49.7%) and CDX2-negative subtypes (16.1%). These subtypes were delineated and characterized for the first time. Heat map of the clinical and morphologic characteristics of the identified subtypes of GC is presented in fig. 2 and table 3.

Рис. 2. Тепловая карта клинико-морфологических характеристик выделенных подтипов рака желудка.

Цветовые обозначения различных признаков приведены в соответствии с табл. 3. Переменные, показанные белым цветом, не были доступны для интерпретации.

Таблица 3. Цветовые обозначения признаков для тепловой карты, изображенной на рис. 2

Comparative survival analysis between the identified subtypes is presented in table 4 and fig. 3.

Таблица 4. Попарные сравнения выживаемости пациентов с различными молекулярными подтипами РЖ

Рис. 3. Кривые выживаемости Каплана-Мейера при РЖ в соответствии с 5-уровневой классификацией.

Clinicopathological Features of the Five Subtypes of Gastric Cancers

The MMR-deficient subtype occurs in 8.1% of GC cases, more often including older patients (average age 69 years, median age — 70 years), with a predominance of females. This subtype mainly consists of tubular adenocarcinomas of intestinal type according to P. Lauren (88% vs 54% in non-microsatellite instable GC, p=0.010), without signet ring cells (84% vs 52.6%, p=0.002), located distally in the stomach (60% vs 33%, p=0.022), and having a saucer/cup-like macroscopic form (68% vs 36%, p=0.010). Thus, by most clinicopathological characteristics, our identified MMR-deficient subtype of GC corresponds to the data in the literature [2, 26]. However, unlike other authors, we found no evidence of less frequent metastasis to regional lymph nodes (p=0.280) or differences in clinical stages (p=0.081). The MMR-deficient subtype is characterized by positive expression of MUC2 (p=0.017), higher expression of CD44 (p=0.01, fig. 4), and positive expression of PD-L1 (both clone SP263 and clone SP142) (p<0.05, fig. 4). This subtype also has an increased density of CD8+ cell infiltration compared to the CDX2-positive subtype (p=0.005) and an increased density of CD1a+ cells compared to the E-cadherin-aberrant subtype (p=0.038). The overall five-year survival rate of patients in the MMR-deficient tumor group was 76% (p=0.05, CI: 0.25—1.01, HR 0.59, median survival 91 vs 35 months for non-microsatellite instable GC; fig. 3), thus the tumors are characterized by a favorable prognosis.

Рис. 4. Иммуногистохимическая характеристика молекулярных подтипов РЖ.

MMR-дефицитный подтип характеризуется отсутствием реакции с одним или несколькими MMR маркерами, более высокой экспрессией CD44, положительной экспрессией PD-L1, повышенной плотностью инфильтрации CD8+ клеток по сравнению с CDX2-позитивным подтипом; EBER-позитивный подтип — положительной реакцией с EBER, MUC- null фенотипом, положительной экспрессией PD-L1, повышенной плотностью инфильтрации CD8+ клеток по сравнению со всеми другими подтипами; E-cadherin-aberrant — аберрантной экспрессией E-cadherin, более частой экспрессией MUC2 и claudin-3; CDX2-положительный подтип — положительной экспрессией CDX2, MUC2, CD10 и фенотипом MUC-I; CDX2-отрицательный подтип — отрицательной экспрессией CDX2, положительной экспрессией MUC5AC и фенотипом MUC-G.

The EBER-positive subtype is found in 9% of cases, consistent with the literature [23—25]. Literature suggests EBER-positive tumors more commonly occur in the proximal sections of the stomach and predominantly affect males [23, 24, 27], which our study fully confirms: unlike the EBER-negative expression group, the EBER-positive group exhibits a significant predominance of male patients (82.1% males in EBER-positive GC vs 17.9% in EBER-negative GC, p=0.003) and more frequent proximal localization (78.6% vs 52.3%, p=0.026). It is newly revealed that the EBER-positive group of adenocarcinomas has fewer diffuse-infiltrative forms of cancer (3.6% vs 20.6%, p=0.166), a predominance of tubular histological type (89.3% vs 53.9%, p=0.040), lower differentiation grades (77% vs 44.4%, p=0.008), and a predominance of the intermediate subtype by P. Lauren (71.4% vs 33%, p=0.000). For the EBER-positive subtype, normal p53 expression is characteristic (p=0.014), with no increased expression of CLAU-3 (p<0.05) and a MUC-null phenotype (p<0.05, fig. 4), as well as positive PD-L1 expression (both clone SP263 and clone SP142) (p<0.05, fig. 4) and increased infiltration density of CD4+, CD8+ (fig. 4), CD68+, and CD1a+ cells compared to all other subtypes (p=0.000). The EBER-positive subtype has a favorable prognosis, with an overall five-year survival rate of 45.5% (p=0.99, CI: 0.12—8.86, HR 1.02, median survival 64 months for EBER-positive GC vs 37 months for EBER-negative GC; fig. 3).

The E-cadherin-aberrant subtype occurs in 17.1% of GC cases, more often includes female patients (56.7% females in E-cadherin aberrant GC vs 43.3% p=0.035); it includes dyscohesive carcinomas (51.7% vs 17.6%, p=0.000) and tubular adenocarcinomas with low differentiation (100% vs 42.2%, p=0.000), with subtotal/total localization (21.7% vs 7.7%, p=0.005), type 3 and 4 by R. Bormann (30% vs 16.4%, p=0.023), larger tumor sizes (more than 8 cm, 29.3% vs 17.2%, p=0.001), diffuse type by P. Lauren (55% vs 17.7%, p=0.000). The E-cadherin-aberrant subtype is characterized by aberrant β-catenin expression (p=0.000, fig. 4), more frequent expression of MUC2 (p=0.013, fig. 4), and claudin-3 (p=0.000, fig. 4). The HER2/neu expression was found in 3.8% of cases. The E-cadherin-aberrant subtype has an extremely poor prognosis, with an overall five-year survival rate of 42.4%, (p=0.019, CI: 0.55—2.39, HR 1.14, median survival 16 months for E-cadherin aberrant GC vs 38 months for E-cadherin non-aberrant GC; fig. 3).

The CDX2-positive subtype occurs in 49.7% of GC cases; it includes tumors of intestinal and intermediate type by P. Lauren, located proximally in the stomach with significantly smaller sizes. This subtype often includes papillary and mucinous GC with less depth of invasion, no lymph node involvement, and lower clinical stage. The CDX2-positive subtype is characterized by positive expression of MUC2 (p<0.050), CD10 (p=0.019), and MUC-I phenotype (p=0.030). The HER2/neu expression was detected in 10.4% of cases. The CDX2-positive subtype has a relatively favorable prognosis (p=0.032, CI: 0.43—1.45, HR 0.95, median survival for CDX2-positive GC 47 months vs 23 months for CDX2-negative GC; fig. 3).

The CDX2-negative subtype occurs in 16.1% of GC cases; it includes tumors of diffuse and intermediate type by P. Lauren, primarily located distally in the stomach and having significantly larger sizes. This subtype often includes papillary carcinomas of the stomach, as well as carcinomas with a greater depth of invasion and a higher clinical stage. The CDX2-negative subtype is characterized by positive expression of MUC5AC (p=0.02) and the MUC-G phenotype (p=0.045). The HER2/neu expression was identified in 12% of cases. The CDX2-negative subtype is characterized by an unfavorable prognosis compared to CDX2-positive subtype (p=0.029, CI: 0.50—2.45, HR 1.12, median survival for CDX2-negative GC 23 vs 47 months for CDX2-positive GC), MMR-deficient subtype (p=0.001, CI: 0.25—1.01, HR, median survival 91 months for MMR-deficient GC), EBER subtype (p=0.021, CI: 0.12—8.86, HR, median survival 64 months for EBER subtype) (fig. 3).

The validity of identifying CDX2-positive and CDX2-negative molecular subtypes was confirmed by regression analysis (fig. 5). Fig. 5 shows that CDX2-positive subtype is associated with a 2-fold decrease in the odds of a fatal outcome (p=0.003). On the other side CDX2-negative subtype is associated with a 2.072-fold increase in the odds of fatal outcome (p=0,035).

Discussion

The identified subtypes have promising implications for the targeted therapy of gastric malignancies. High expression of PD-L1 in MMR-deficient and EBER-positive subtypes creates a rationale for the use of immune checkpoint inhibitors in treatment, particularly atezolizumab and durvalumab. According to ACRG, MSI subtype tumors were characterized by a high number of mutations, with 44% of samples from this group showing mutations in ARID1A, and 42% in phosphoinositide-3-kinase and tensin homolog (PTEN), which are targets for rapamycin (mTOR signaling pathway). Thus, rapamycin may also be a promising marker for therapy in patients with MMR-deficient subtype.

Рис. 5. График отношения шансов для клинико-морфологических характеристик рака желудка, а также новых предложенных молекулярных подтипов, составленный с помощью регрессионной модели пропорциональных рисков Кокса.

In the E-cadherin aberrant subtype, increased expression of claudin-3 was found. It is known that claudin-3 and claudin-4 are receptors for Clostridium perfringens enterotoxin (CPE), which causes the disruption of epithelial cells [28]. Therefore, it may be possible to use CPE for targeted therapy of gastric adenocarcinomas characterized by increased expression of claudins 3 and 4, particularly in tumors of the E-cadherin aberrant subtype [29—33].

In the MMR-deficient subtype, high expression of CD44 was detected, which serves as both a marker of cancer stem cells and a receptor for hyaluronic acid. Vectors using hyaluronic acid, currently under development, may be used for targeted therapy of tumors expressing CD44 [34], representing an additional possible therapeutic approach to treating MMR-deficient tumors.

In gastric adenocarcinoma cells and areas of intestinal metaplasia, a positive correlation of claudins 3 and 4 with the expression of CDX2, which controls the differentiation of intestinal epithelium, was found. An increase in the expression of CDX2, as well as claudins 3 and 4, is most characteristic of intestinal-type gastric cancer. In an in vitro experiment, recombinant CDX2-expressing vectors were transferred into CDX2-negative gastric cancer cells, after which the level of claudin 3 and 4 expression increased, while the expression of other claudins did not change [29, 37, 38]. Thus, it can be suggested that CDX2 plays a role in regulating the expression of claudins 3 and 4 in gastric cancer [36] and possibly the use of CPE for targeted therapy not only of the E-cadherin aberrant subtype but also of the CDX2-positive.

In CDX2-positive and CDX2-negative subtypes, HER2/neu expression is determined in some cases, widely used in combined therapy for gastric cancer. However, since the CDX2-positive and CDX2-negative subtypes are identified for the first time in this study, further detailed study of the molecular profile of these subtypes is necessary to find the most suitable therapeutic targets. It should be noted that CDX2 expression has been previously studied in GC, but this is the first time that a molecular subtype based on this marker has been identified (table 1). Moreover, the literature data on the prognostic value of CDX2 in GC are contradictory: a number of studies have confirmed the association with a favorable prognosis [39—45], while some researchers have not revealed the prognostic importance of this marker [46, 47]. Thus, in the present research, for the first time on a large sample of patients (more than 300) in the examination of GC surgical material by means of immunohistochemical method, it is shown that positive expression of CDX2 is an important factor of favorable prognosis in patients with GC in p53-null-subtype and can serve as a basis for dividing patients into two groups: CDX2+ and CDX2-.

Conclusion

A new molecular classification of gastric cancer has been developed, including 5 molecular subtypes — MMR-deficient subtype (median survival 91 months), EBER-positive subtype (median 64 months), E-cadherin-aberrant subtype (median 16 months), CDX2-positive subtype (median 47 months), CDX2-negative (median 23 months). The CDX2-positive and CDX2-negative subtypes are identified for the first time. CDX2-positive subtype is associated with a 2-fold decrease in the odds of a fatal outcome, in contrast CDX2-negative subtype is associated with a 2.072-fold increase in the odds of fatal outcome. The relevance of distinguishing a subtype associated with aberrant expression of p53, as described by other authors, is not confirmed.

Author contribution:

Conceptualization and design of the study — Danilova N.V., Andreeva Yu.Yu., Malkov P.G.

Collection and processing of data –Danilova N.V., Khomyakov V.M., Chaika A.V.

Statistical analysis –Danilova N.V.

Writing –Danilova N.V., Kalinin D.V., Porubayeva E.E.

Editing –Danilova N.V., Andreeva Yu.Yu., Malkov P.G.

Funding. No funding was received to assist with the preparation of this manuscript.

Ethics approval. Retrospective analyses of anonymized diagnostic left over material have been approved by the local ethics committee (Lomonosov Moscow State University, Russia).