Reduced DEFA5 Expression and STAT3 Activation Underlie the Submucosal Invasion of Early Gastric Cancers Open Access

Gastric cancer is the fourth leading cause of cancer-related deaths in the world, accounting for approximately 770,000 deaths annually [1]. The 5-year survival rate for gastric cancer with distant metastasis is low, at approximately 6.6%, whereas the 5-year survival rate for localized gastric cancer without lymph node metastasis is over 90% [2, 3]. Therefore, early diagnosis and treatment are imperative to improve the prognosis of gastric cancer. Endoscopic submucosal dissection (ESD) is recommended as a minimally invasive and curative treatment for patients with early gastric cancer (EGC) [4‒6], but some patients develop metastatic recurrence after ESD. Recently, a scoring system for predicting the cancer-specific survival of patients who underwent ESD was established [7]. This scoring system consisted of five risk factors: lymphatic invasion, tumor size >30 mm, positive vertical margin, venous invasion, and submucosal invasion ≥500 μm, and the presence of these risk factors were associated with poor cancer-specific survival after ESD. Another study reported that the depth of the T1–2, N0 gastric cancer was associated with the risk of recurrence after gastrectomy [8]. These studies have demonstrated that submucosal invasion is one of the core hallmarks of EGC with poor prognosis.

Gastric cancers develop from chronic atrophic gastritis and intestinal metaplasia [9]. However, how EGC progresses to advanced gastric cancer is not yet understood. Global gene expression analysis has been used for the molecular characterization of gastric cancer [10, 11]. We hypothesized that the transcriptomic change from intramucosal gastric cancer (IMGC) and submucosal invasive gastric cancer (SMGC) would unveil the molecular mechanism of submucosal invasion. Although transcriptome analysis has been widely used to characterize the molecular biology of gastric cancer, only a few studies have analyzed the gene expression profile of EGC [12, 13]. In particular, the difference between IMGC and SMGC, based on comprehensive gene expression patterns, has not been studied. In this study, a comprehensive analysis of gene expression profiles of IMGC and SMGC was conducted to elucidate the mechanism of submucosal invasion, and the genes and pathways involved in the submucosal invasion of EGC were identified.

From April 2017 to December 2018, patients with EGC, who underwent ESD or surgery at the University of Tokyo Hospital, were recruited for this study. Tumor location, macroscopic type, histologic type, and invasion depth were defined according to the Japanese classification of gastric carcinoma by the Japanese Gastric Cancer Association [14]. Of the 127 patients from whom written informed consent was obtained, 8 patients with differentiated IMGCs and 8 patients with differentiated SMGCs with submucosal invasion ≥500 μm were randomly selected for gene expression profiling. Moreover, from formalin-fixed, paraffin-embedded samples previously endoscopically resected at the University of Tokyo Hospital from 2015 to 2018, two sets of 50 cases each of differentiated IMGCs and SMGCs with submucosal invasion ≥500 μm were immunohistochemically analyzed, retrospectively. Cases of confirmed Helicobacter pylori (H. pylori) eradication at our hospital or other hospitals were defined as “eradication.” Those with endoscopically confirmed atrophic gastritis and with either a positive antibody test, urea breath test, stool test or histological examination were defined as “positive.” Those who neither had a history of H. pylori eradication nor atrophic gastritis and with a negative H. pylori test were defined as “negative.” Those who did not fall into any of the above categories were defined as “others.” This study was approved by the Research Ethics Committee of the Graduate School of Medicine and Faculty of Medicine, the University of Tokyo (Ethical Review Numbers: G10116 and 2020173NI).

One or two biopsy samples each from the tumor site and the normal mucosa, at least 2 cm away from the lesion, were obtained from each patient using endoscopic biopsy forceps. The collected specimens were immediately embedded in RNAlater (Thermo Fisher Scientific, Waltham, MA, USA) and stored at 25°C for 3–6 h. After that, the samples were stored at −30°C.

RNA extraction, gene expression profiling, and quantile normalization were performed as previously reported [15]. Thereafter, the total RNA was extracted using the miRNeasy Mini Kit (QIAGEN, Hilden, Germany). The quality of the extracted RNA was assured using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA), after which it was subjected to gene expression profiling using the SurePrint G3 Human Gene Expression 8 × 60K v3 microarray (Agilent Technologies). Measurements in the raw data were logged (base 2), and quantile normalization was performed using R v4.04. The expression value of gene corresponding to each probe was determined by subtracting the normalized signal intensity of the respective probe in the paired background from that in the tumor. Heatmaps were developed using ComplexHeatmap, version 2.7.7 [16]. The RNA consensus tissue gene data from the Human Protein Atlas (https://www.proteinatlas.org/) was used to determine tissue-specific gene expression [17]. Gene set enrichment analysis (GSEA; https://www.gsea-msigdb.org/gsea/index.jsp) [18, 19] was performed using curated gene sets from the Molecular Signatures database. The raw data have been deposited in the NCBI gene expression omnibus and are accessible under the GEO series accession number: GSE208099. The enrichment of various tissue-specific genes was calculated using TissueEnrich [20].

Formalin-fixed, paraffin-embedded samples (3 μm) were deparaffinized, and endogenous peroxidase activity was blocked using a 3% hydrogen peroxide-water solution for 5 min at 25°C. Antigen retrieval for defensin alpha 5 (DEFA5), MUC2, MUC5AC, MUC6, and CD10 was performed using citrate buffer (10 mm citrate buffer, pH 6.0), whereas antigen retrieval for phosphorylated STAT3 (pSTAT3) was performed using the antigen unmasking solution H-3301 (Vector Laboratories, Newark, CA, USA) with a 1:100 dilution at 120°C for 10 min in a pressure cooker (Panasonic, Osaka, Japan), followed by cooling at 25°C for 30 min. The primary antibody was applied at 37°C for 30 min. EnVision + Dual Link System-HRP (Agilent Technologies) was used as the secondary antibody, and the reaction was carried out at 37°C for 30 min. To visualize the antigen-antibody complex, the ImmPACT DAB substrate kit (Vector Laboratories) was used, and the sections were counterstained with hematoxylin. The DEFA5 expression level was defined as the mean of the number of positive cells in 10 fields of view at ×400 magnitude, randomly selected by the pathologist. The pSTAT3 expression level was defined as the mean of the number of positive tumor cells in 10 fields of view at ×400 magnitude, randomly counted using e-Count software (e-path Co., Ltd., Kanagawa, Japan). pSTAT3 expression levels <2.7 and ≥2.7 were considered negative and positive, respectively. For the evaluation of immunohistochemical (IHC) staining, MUC2, MUC5AC, MUC6, and CD10 levels <5% and ≥5% were considered negative and positive, respectively, based on a previously reported threshold [21]. The intestinal phenotype was defined as that which was positive for MUC2 or CD10 and negative for MUC5AC and MUC6, and the gastric phenotype as that which was positive for MUC5AC or MUC6 and negative for MUC2 and CD10. The mixed phenotype was defined as that which was positive for MUC2 or CD10 and MUC5AC or MUC6. The non-classified phenotype was defined as that which was negative for MUC2 and CD10 as well as for MUC5AC and MUC6.

The primary antibodies and dilution ratios used in this study were DEFA5 (sc-53997, Santa Cruz Biotechnology, Dallas, TX, USA), 1:50; pSTAT3 (sc-8059, Santa Cruz Biotechnology), 1:50; MUC2 (M7313, Agilent Technologies), 1:50; MUC5AC (NCL-MUC-5AC, Leica Microsystems GmbH, Wetzlar, Germany), 1:50; MUC6 (NCL-MUC-6, Leica), 1:50; and CD10 (NCL-CD10-270, Leica), 1:100.

The data were analyzed statistically using either JMP version 15 (SAS Institute Inc., Cary, NC, USA) or R version 4.0.4. All p values were two-sided, and p < 0.05 was considered statistically significant. For differential expression analysis of microarray data, the empirical Bayes method from the limma package was used [22]. Fisher’s exact test was performed to test the association between two categorical variables. The Wilcoxon rank sum test was performed to compare the expression levels of DEFA5 and pSTAT3 as determined by IHC staining. The Wilcoxon signed rank test was performed to compare the DEFA5 expression levels between normal matched mucosa and the tumor site.

To characterize the gene expression profiles of IMGC and SMGC, eight IMGC cases and eight early SMGC cases with submucosal invasion ≥500 μm were investigated in this study (Table 1). Biopsies of both the tumor site and the normal background mucosa were performed. Thereafter, microarray analysis was performed on 16 pairs of tumor and normal mucosa samples. The expression levels of individual genes were determined by subtracting the normalized signal intensity of each gene in the normal mucosa from that in the paired tumor site. To dissect the expression profiles of IMGC and SMGC, we extracted 1,000 differentially expressed genes (DEGs) (top 500 upregulated and top 500 downregulated genes) by subtracting the mean expression levels of each gene in IMGCs from the respective values in SMGCs. Hierarchical clustering using DEGs efficiently classified SMGCs and IMGCs. We detected 97 small intestine-specific genes among the DEGs, which were strongly biased toward downregulated genes (Fig. 1a), after annotating the DEGs based on tissue-specific genes identified by the Human Protein Atlas [17]; for example, small intestine-specific genes such as DEFA5, DEFA6, ITLN1, and CLCA1 were significantly downregulated in SMGCs (Fig. 1b). Next, we examined the expression of tissue-specific genes in DEGs using TissueEnrich, a tool that is used to calculate tissue-specific gene enrichment from an input gene set [20]. The analyses showed that small intestine-specific genes and duodenum-specific genes were highly enriched in the 500 downregulated genes (Fig. 1c), whereas these genes were not enriched in the 500 upregulated genes (Fig. 1d). These results suggest that downregulation of small intestine-specific genes may be involved in gastric cancer invasion. Similarly, when comparing the gene expression profiles of advanced and EGCs using the dataset reported by Vecchi et al. [12], small intestine-specific and duodenum-specific genes were found to be downregulated in advanced gastric cancer (Fig. 1e). Notably, a recent study that explored the characteristics of epithelial cells across different gastric conditions including normal gastric mucosa, chronic atrophic gastritis, intestinal metaplasia, and EGC, using single-cell RNA sequencing, revealed that the proportion of enterocytes increased along the cascade from gastritis to intestinal metaplasia but decreased in EGC [13]. These findings suggest that the downregulation of small intestine-specific genes may be involved, at every step, in the progression of EGC from intestinal metaplasia, to advanced gastric cancer.

Using microarray analysis, we found that small intestine-specific genes, particularly DEFA5, were downregulated in SMGCs than in IMGCs. To determine whether DEFA5 expression in normal mucosa differed between IMGCs and SMGCs, we analyzed the expression level of DEFA5 in the normal mucosae of IMGCs and SMGCs using our microarray data. We found no significant difference in DEFA5 expression in normal mucosae between IMGCs and SMGCs (Fig. 2a). We further compared the DEFA5 expression between tumor sites and normal mucosae. In SMGCs, DEFA5 expression at the tumor site was lower than that at the normal mucosal sites (Fig. 2b), whereas in IMGCs, DEFA5 expression did not differ significantly between the tumor site and normal mucosa (Fig. 2c).

To validate these findings at the protein level, IHC staining of DEFA5 was performed on 50 cases of IMGCs and SMGCs each; the clinical characteristics are summarized in Table 2. Significant differences between IMGCs and SMGCs in terms of H. pylori infectious status, extent of mucosal atrophy, location of tumor, and mucin phenotypic classification were not observed. However, there were significant differences in terms of the histological type, lymphatic invasion, venous invasion, and tumor size. IHC staining showed that DEFA5 was positive in 58% of the IMGC cases, whereas for SMGCs, it was positive in only 22% of the cases (Fig. 2d, e). In addition, the number of DEFA5-positive cells was significantly smaller in SMGCs than in IMGCs (Fig. 2f). To evaluate the DEFA5 expression based on the differentiation of tumor cells, we compared the DEFA5 expression between well-differentiated (tub1) (n = 73) and moderately differentiated (tub2) (n = 27) histological types. The number of DEFA5-positive cells was significantly greater in tub1 compared to that in tub2 (Fig. 2g). Next, we compared the DEFA5 expression between tub1 (n = 47) of IMGC and tub1 (n = 26) of SMGC. The number of DEFA5-positive cells was significantly fewer in SMGCs with tub1 compared to IMGCs with tub1 (Fig. 2h). These results showed that there was still a significant difference in DEFA5 expression between IMGCs and SMGCs after the differentiation level was adjusted.

To further characterize the transcriptomic differences between IMGCs and SMGCs, we performed GSEA. GSEA, using the HALLMARK gene set that represents well-defined biological states, revealed nine significantly enriched gene sets in SMGCs (q < 0.05). Among them, the epithelial-mesenchymal transition pathway, which is widely recognized to initiate the invasive and metastatic behavior of epithelial cancers [23], was the most upregulated in SMGCs (Fig. 3a). Subsequently, we performed GSEA using the C2 gene sets curated from various databases and literatures. Of the 6,366 gene sets, 326 were significantly upregulated and 128 were significantly downregulated in SMGCs (q < 0.05) (online suppl. Tables S1, S2; for all online suppl. material, see https://doi.org/10.1159/000531790). Notably, gene sets that contained DEGs between advanced gastric cancer and EGC as reported by Vecchi et al. [12] demonstrated the most prominent bidirectional correlation with SMGCs. Hence, the gene set that contained upregulated genes in advanced gastric cancer was upregulated in SMGCs, and the gene set that contained downregulated genes in advanced gastric cancer was downregulated in SMGCs (Fig. 3b, c). These results indicate that the transcriptomic signature of advanced gastric cancer may have been acquired during the progression from IMGCs to SMGCs. We focused on the STAT3 activation because both the HALLMARK_IL6_JAK_STAT3_SIGNALING in the HALLMARK gene set and the DAUER_STAT3_TARGETS_UP in the C2 gene set were upregulated in SMGCs. STAT3 is activated by various kinds of cytokines and is associated with gastric cancer progression. In our transcriptomic analysis, STAT3 activating cytokines, such as IL24, IL6, IFNG, IL26, IFNA2, and CSF2 [24‒26], were found to be significantly upregulated in SMGCs (Fig. 3d). These results suggest that aberrant activation of STAT3 may play a crucial role in submucosal invasion.

To confirm STAT3 activation in SMGCs, IHC staining of pSTAT3 was performed on the same samples used for IHC staining of DEFA5. The expression of pSTAT3 in the nucleus of tumor cells was marked in SMGCs (Fig. 4a, b). Of the 100 cases that were subjected to IHC staining, 40 were DEFA5-positive, 31 were pSTAT3-negative, and nine were positive for both DEFA5 and pSTAT3. Notably, by carefully observing the staining of DEFA5 and pSTAT3 in serial sections in these nine cases, we found that pSTAT3 was not expressed in the area where DEFA5 was expressed (Fig. 4c). These results suggest that STAT3 is activated in SMGCs and inversely correlated with DEFA5 expression.

In this study, we characterized the gene expression profiles of IMGCs and SMGCs via comprehensive transcriptomic analysis and IHC staining, for the first time. In addition, we have extracted genes which may potentially play a key role in submucosal invasion.

The expression of small intestine-specific genes was downregulated in SMGCs with reference to their gene in IMGCs. The downregulation of small intestine-specific genes was also observed at each step of the progression of intestinal metaplasia to early-stage gastric cancer as well as early-stage to advanced gastric cancer, suggesting that it may be related to the progression of gastric cancer. According to the Lauren classification, one of the criteria for classifying gastric cancer based on histopathological features, diffuse-type gastric cancer is known to have a poor prognosis, compared to intestinal-type gastric cancer. Recent single-cell RNA sequencing analysis has shown that diffuse-type and intestinal-type gastric cancers have distinct transcriptomic features, and that intestinal-type gastric cancers express small intestine-specific genes [27]. These results suggest that histopathological features and expression of small intestine-specific genes are associated with gastric cancer progression. In this study, only differentiated EGCs were included, and thus, all cases were classified as intestinal-type by Lauren classification. However, small intestine-specific genes were still significantly downregulated in SMGCs compared to IMGCs. This finding suggests that the loss of intestinal cell lineage at the gene expression level may contribute to gastric cancer progression, even when they are histopathologically classified as the intestinal type.

The role of the JAK-STAT signaling pathway in gastric cancer has been extensively studied. There is growing evidence that pSTAT3 is involved in tumorigenesis, proliferation, invasion, and metastasis of gastric cancer [28‒33]. In this study, GSEA has demonstrated that the JAK-STAT signaling pathway was activated in SMGCs. IHC analysis also showed that pSTAT3 markedly accumulated in SMGCs. Although further studies are needed to clarify the precise mechanism involved in the invasion of EGCs, these results have provided further evidence concerning the crucial role of the JAK-STAT signaling pathway in the submucosal invasion of gastric cancers.

We have demonstrated that pSTAT3 was upregulated in tumor cells in SMGCs. We also found that pSTAT3 was upregulated in stromal cells in SMGCs (data not shown). pSTAT3 in the stromal cells of the tumor microenvironment has been shown to play a crucial role in promoting tumor growth and metastasis by inducing the secretion of growth factors and cytokines that support tumor cell survival and proliferation [34]. The upregulation of various proinflammatory cytokines, found using microarray analysis, may be associated with pSTAT3 in stromal cells. Multifaceted approaches, such as spatial transcriptomic analysis and single-cell RNA-seq analysis, should clarify what cell type expresses proinflammatory cytokines in future studies.

In this study, DEFA5 was downregulated in SMGC. Wu et al. [35] reported that DEFA5 directly binds to BMI1 and decreases its recruitment to the promoter region of CDKN2a, which subsequently upregulates the expression of p16 and p19, leading to the cell cycle arrest of gastric cancer cells. However, in our analysis, the CDKN2a expression between IMGCs and SMGCs did not change significantly (data not shown). This suggests that DEFA5 may be involved in the submucosal invasion of gastric cancer via other molecular mechanisms, than directly binding to BMI1.

DEFA5 is an antimicrobial peptide usually secreted by the Paneth cells in the small intestine and plays a vital role in maintaining the homeostasis of the intestinal microbiota [36‒39]. Reduced DEFA5 expression causes intestinal dysbiosis, which leads to the pathogenesis of Crohn’s disease [40‒44], graft-versus-host disease [45, 46], and liver fibrogenesis [47, 48]. A recent study showed that the gastric mucosal microbiomes were altered in gastric cancer compared to superficial gastritis [49]. There is increasing evidence that tissue microbiomes can modulate cancer susceptibility and pathogenesis [50, 51]. Certain types of gut bacteria, such as Fusobacterium nucleatum, have been shown to be involved in the regulation of epithelial-mesenchymal transition, a process that is critical for tumor cell invasion [52]. This suggests that reduced expression of DEFA5 in SMGCs may cause dysbiosis and promote gastric cancer progression. However, it is important to note that the relationship between reduced DEFA5 expression and dysbiosis and tumor cell invasion is complex and still not fully understood. More research is needed to fully elucidate the mechanisms by which DEFA5 affects tumor cell invasion.

We have demonstrated that pSTAT3 was upregulated in the submucosal invasive lesions where DEFA5 was not expressed. This could be due to STAT3 possibly downregulating DEFA5. There are conflicting reports on whether STAT3 represses DEFA5 expression [53‒55], suggesting that the regulation of DEFA5 expression by STAT3 is case-dependent. To the best of our knowledge, no previous research has shown STAT3 binding to the promoter region of DEFA5. Herein, we evaluated the STAT3 binding to the promoter region of DEFA5 by searching the publicly available ChIP database (ChIP Atlas: https://chip-atlas.org/); the search, however, returned no hits. DEFA5 is a molecular marker characteristic of mature Paneth cells. Upon intestinal epithelial damage, Paneth cells are supposed to dedifferentiate and acquire proliferative capacity and pluripotency. This process of dedifferentiation requires the induction of Notch signaling via STAT3 phosphorylation [56]. Although the ChIP database did not provide evidence that STAT3 directly binds to the promoter of DEFA5, STAT3 phosphorylation may induce the dedifferentiation of Paneth-like cells expressing DEFA5 in gastric cancer and may indirectly suppress the expression of small intestine-specific genes including DEFA5. Therefore, further studies are needed to determine whether STAT3 can suppress DEFA5 expression in EGC. Another possibility is that the loss of DEFA5 induces STAT3 activation. Because DEFA5 is an antimicrobial peptide that maintains the microbiome, it is possible that reduced DEFA5 expression causes dysbiosis and induces aberrant expression of a diverse repertoire of chemokines and cytokines, which in turn activates the JAK-STAT signaling pathway. Restoring DEFA5 expression via the administration of recombinant DEFA5 or R-spondin is being investigated as a new therapeutic approach for inflammatory diseases associated with dysbiosis [46, 48, 57]. Further studies are required to clarify the mechanism of gastric cancer invasion caused by reduced DEFA5 expression, which may lead to a new therapeutic approach for gastric cancer via the restoration of DEFA5 expression.

In conclusion, we performed comprehensive gene expression analysis and IHC staining in IMGCs and SMGCs and identified that DEFA5 was downregulated in SMGCs. We also identified that the JAK-STAT signaling pathway was upregulated in SMGCs, and pSTAT3 was negatively correlated with DEFA5 expression. Although further studies are needed to clarify the detailed mechanism, reduced DEFA5 expression may play a crucial role in the submucosal invasion of EGC by upregulating pSTAT3.

We thank Natsuko Kageyama and Mitsue Yamamichi for their assistance in RNA extraction.

This study was approved by the Research Ethics Committee of the Graduate School of Medicine and Faculty of Medicine, the University of Tokyo (Ethical Review Numbers: G10116 and 2020173NI). Written informed consent was obtained for all the patients in that RNA extraction was performed. Informed consent was obtained in the form of opt-out on the web site for the patients in that IHC analysis was performed. The present study was performed in accordance with the Declaration of Helsinki.

The authors have no conflicts of interest to declare.

This work was supported by Grant-in-Aid for Early-Career Scientists (Grant Nos. 22K15959 and 21K15479), from the Japan Society for the Promotion of Science.

Conception and design: Sayaka Nagao, Yu Takahashi, and Nobutake Yamamichi. Acquisition of data: Sayaka Nagao, Yu Takahashi, Yuko Miura, Hiroya Mizutani, Daisuke Ohki, Yoshiki Sakaguchi, Seiichi Yakabi, Yosuke Tsuji, Keiko Niimi, and Naomi Kakushima. Material preparation: Sayaka Nagao, Yu Takahashi, Tamami Denda, and Yukihisa Tanaka. Analysis and interpretation of data: Sayaka Nagao, Yu Takahashi, and Yasunori Ota. Writing, review, and/or revision of the manuscript: Sayaka Nagao, Yu Takahashi, Tamami Denda, Yosuke Tsuji, Naomi Kakushima, Nobutake Yamamichi, Yasunori Ota, Kazuhiko Koike, and Mitsuhiro Fujishiro. All authors read and approved the final manuscript.

The raw data generated in this study have been deposited in the NCBI gene expression omnibus and are accessible under the GEO Series accession number: GSE208099.