Thirty ESD specimens from patients with early gastric cancer were selected from all ESD specimens available at the Department of Pathology (Peking University First Hospital) from 2016 to 2018. Clinical information for these specimens was abstracted from patient medical records. In addition, five fresh ESD specimens were collected. Pathological evaluation was conducted based on the 5th World Health Organization classification system [18]. The distribution of neoplastic lesions included 23 cases of low-grade dysplasia (LGD), 24 cases of high-grade dysplasia (HGD), 30 cases of intramucosal carcinoma, and 11 cases of submucosally invasive carcinoma (pT1b1, 7/30; pT1b2, 4/30). A total of 19 cases of well-differentiated gastric adenocarcinoma and 11 cases of poorly differentiated gastric adenocarcinoma were evaluated (Table 1).

The 30 included patients comprised 24 male and 6 female patients, with ages ranging from 43 to 79 years (median, 65 years). A total of 12 cases involved the pyloric antrum, whereas the other 18 cases involved the body or fundus. Clinical information and follow-up data were obtained for all cases. The follow-up period was defined as starting from the date of initial diagnosis and ending at the date of the patient’s death, progression, relapse, or last follow-up visit. The follow-up duration ranged from 10 to 54 months (median, 36.5 months). Frozen sections from five additional ESD specimens derived from patients with early gastric carcinomas were subjected to SA-β-gal staining.

Frozen sections from fresh ESD specimens were fixed in 4% formalin and evaluated using an SA-β-gal staining kit (GenMed Scientifics, Inc., Wilmington, DE, United States) according to the manufacturer’s instructions. Briefly, tissues were stained with the SA-β-gal staining solution overnight at 37°C; we counted 100 cells in random fields, and calculated the percentage of SA-β-gal-positive cells (i.e., blue cells).

Formalin-fixed, paraffin-embedded (FFPE) sections were deparaffinized with serial xylene treatments and hydrated in graded alcohols. Endogenous peroxidase activity was quenched using 0.3% hydrogen peroxide for 60 min. Antigen retrieval was carried out by heating the specimen in citrate buffer (20-mM citrate buffer, pH 6.0) at 95°C for 20 min. After blocking with horse serum (1:100 in phosphate buffered saline, PBS), the sections were incubated with primary antibodies in various dilutions overnight at 4°C. The specimens were incubated using a Dako Envision Flex amplification kit (Dako, Glostrup, Denmark) for 60 min, and colour development was completed using a freshly prepared diaminobenzidine solution (Dako). The sections were then counterstained using Mayer’s haematoxylin. For the negative control, the primary antibody was replaced with phosphate buffered saline (PBS) or normal rabbit serum. The relevant antibodies and their associated information are summarized in Supplementary Table S1.

We found that cGAS, STING, and IRF3 demonstrated cytoplasmic staining, while STAT6, Ki67, minichromosome maintenance complex component 7 (MCM7), and p53 showed nuclear staining. The evaluation was performed in whole slides. Since the intensive staining for each of these factors was similar, the positive stained glandular epithelial cells were semiquantitatively estimated and the resultant percentages represent the positive glandular epithelial cells against total glandular cells for each of groups. Stromal cells were not counted. For associations with clinicopathological parameters and Kaplan–Meier single-factor analysis, positive cases were defined as those with either >10% or >20% positive tumour cells (i.e., different thresholds were evaluated), while negative cases were defined as those with ≤10% or ≤20% positive tumour cells.

DNA was extracted from FFPE blocks derived from 30 cases using a QIAamp DNA FFPE tissue kit (Qiagen, Hilden, Germany). The DNA library was constructed using the capture method, and paired-end sequencing was performed using a NextSeq 500 Sequencer in combination with the NextSeq™ 500 High Output Kit (Illumina, Inc., San Diego, CA, United States) according to the manufacturer-recommended protocols for gastrointestinal tumour-related genes (Burning Rock, Guangzhou, China). The average sequencing depth was 1,000× for tissue samples. Single nucleotide variants, copy number variants, and fusion were called in the pipeline. Target regions were captured using designed probes spanning 41 genes (Supplementary Table S2).

Sequence data were mapped onto the reference human genome (hg19) using Burrows–Wheeler Aligner version 0.7.10. Local alignment optimization, duplication marking, and variant calling were performed using the Genome Analysis Tool Kit version 3.2 and VarScan version 2.4.3.

All statistical analyses were performed using SPSS statistical software (version 17.0, Chicago, IL, United States), Microsoft Excel 2007 (Seattle, WA, United States), and GraphPad Prism software (San Diego, CA, United States). The data obtained from immunohistochemistry (IHC) and association between expressed factors and other parameters were analysed using chi-square tests, Fisher exact tests, Mann–Whitney U tests, and Spearman correlation analysis. The distributions of IHC results between different lesion types were analysed using nonparametric Friedman tests. Progression-free survival curves were plotted using the Kaplan-Meier method and were compared using log-rank tests. Differences were considered statistically significant given a two-sided p value of <0.05.

Fresh frozen sections from five ESD specimens of early gastric carcinomas were subjected to SA-β-gal staining. Results showed positive staining in the cytoplasm of precancerous and cancerous cells in all evaluated cases (>10% SA-β-gal-positive cells) but not in the inflammatory gastric mucosa (Figure 1A).

Expression of cGAS, STING, IRF3, and STAT6 factors involved in the mediation of SASP activity was analysed in the 30 ESD specimens. cGAS staining was mainly detected in the cytoplasm of precancerous and cancerous cells (Figure 1B), while STING staining was detected in the cytoplasm of cancer cells and stromal cells such as lymphocytes (Figure 1B). In addition, the positive expression of IRF3 presented as cytoplasmic staining in precancerous and cancerous lesions, while the positive expression of STAT6 mainly presented as nuclear staining in precancerous and cancerous lesions (Figure 1B). The distributions of cGAS, STING, IRF3, and STAT6-positive expression was similar in high-grade dysplastic and cancerous lesions, with diffuse positive staining. Focal or scattered weak staining was detected in low-grade dysplastic cells (Figure 1B). The distribution of positive expression between different lesions was analysed using non-parametric Friedman tests. Positive staining for cGAS, STING, IRF3, and STAT6 in the evaluated carcinomas was statistically significantly different from that detected in normal or inflammatory (N/I) cells or in LGD; there were no differences in expression between HGD and the evaluated carcinomas (Figure 1C).

On performing Spearman correlation, we found that cGAS expression positively correlated with IRF3 and STAT6 expression, but not with STING expression (Figure 1D; Table 2). We also found significantly positive correlation between IRF3 and STAT6 expression, but not between STING and IRF3 or STAT6 expression (Table 2). These results suggest that SASP may be prevalently activated in early gastric carcinogenesis.

SASP activation is induced by the binding of cGAS to free cytoplasmic DNA generated following DNA damage [3, 4]. To determine DNA damage in early gastric cancer, factors involved in DDR were estimated using IHC. The expression of the Fanconi anemia group D2 (FANCD2) protein, tumour suppressor p53 binding protein 1 (TP53BP1), and replication protein A (RPA2) indicated nuclear staining to varying degrees in precancerous and cancerous lesions (Figure 2A). Expression levels of FANCD2, TP53BP1, and RPA2 in each lesion (i.e., from every case) were calculated; their distributions are shown in Figure 2B. The expression levels of FANCD2, TP53BP1, and RPA2 were statistically significantly elevated in HGD and in carcinomas, according to the results of non-parametric Friedman tests (Figure 2B).

Mutual correlations were also statistically evaluated using Spearman correlation tests, the results of which indicated statistically significantly positive correlations between FANCD2, TP53BP1, and RPA2 expression, with a closer correlation detected between FANCD2 and TP53BP1 expression (Figure 2C; Table 2).

All cases were subjected to targeted sequencing using a panel of 41 genes expressed in gastric cancers. The detected genomic mutations are summarized in Supplementary Table S3. In addition to the scattered distribution of breast cancer genes (BRCA), the ataxia telangiectasia mutated gene (ATM), adenomatous polyposis coli (APC), phosphatase and tensin homolog (PTEN), and Kirsten rat sarcoma viral oncogene (KRAS) mutations, our results showed that 17 cases (56.67%) harboured TP53 gene mutations in the coding sequence hot spots from exon 4 to exon 9 (Table 3). Moreover, 20 cases were also stained for immunohistochemical analysis. In the evaluation of p53 expression, we found that cases with genomic mutations always presented with strongly positive staining (i.e., in ≥90% of cells) for HGD and for carcinomas, while cases without mutations exhibited weak and scattered positive cells (similar to the findings described in a recent report) [19]. All nine cases with missense mutation were positive on IHC. In addition, one of four cases with frameshift mutations was positive, while one case with a splice-site mutation was negative on IHC. Seven cases with wild-type sequencing were negative. Positive staining for p53 was only present in HGD and in cancer cells; these specimens showed approximately 90% positive cells (Figure 3A; Table 3).

Proliferation zone expansion with an increasing Ki67 index is generally characteristic of gastric carcinogenesis [20]. In this study, Ki67 and MCM7 expressions were demonstrated using IHC, wherein these factors showed nuclear staining in gastric precancerous and cancerous lesions (Figure 4A). Positive staining for Ki67 and MCM7 was mainly located in the proliferation zone of inflammatory gastric mucosa, whereas scattered nuclear staining was found in LGD. Positive cells were markedly increased in HGD and in cancer cells (Figure 4A). The distributions of Ki67 and MCM7 in each lesion demonstrated that high expression levels were present in HGD and in carcinomas (Figure 4B). Friedman test analysis showed that both N/I and LGD differed from HGD and from carcinomas, while there was no difference between N/I and LGD or between HGD and carcinoma (Figure 4B).

Finally, we found a moderately strong correlation between Ki67 and MCM7 immunostaining (p < 0.001, r = 0.697, Table 2).

We performed statistical analyses to clarify correlations between expression levels for SASP, DDR, and proliferation factors in early gastric cancers. The results are summarized in Table 2. cGAS, STING, and IRF3 expression was positively correlated with TP53BP1 expression (p = 0.014, r = 0.443; p = 0.039, r = 0.380; and p < 0.001, r = 0.625), while IRF3 and STAT6 expression was each positively correlated with FANCD2 expression (p = 0.008, r = 0.477, and p = 0.025, r = 0.410, respectively). FANCD2 and RPA2 expressions were positively associated with TP53 mutation status (p = 0.005 and p = 0.025, respectively) (Figure 3B). Moreover, TP53BP1 and RPA2 expressions was closely correlated with Ki67 (p = 0.010, r = 0.462; and p = 0.041, r = 0.375) and MCM7 expression (p = 0.011, r = 0.460; and p = 0.011, r = 0.458), while FANCD2 expression was positively correlated with Ki67 expression (p = 0.024, r = 0.412). Finally, IRF3 expression was positively correlated with the expression DDR (TP53BP1, FANCD2, and RPA2) and cell proliferation factors (Ki67 and MCM7) (Table 2).

Next, the expression profile in early gastric cancer was generated based on the expression of SASP, DDR, and proliferation factors in reference to TP53 mutation status (Figure 5). Roughly, we detected three groups that could be classified as follows: a low expression group with SASP, DDR, and proliferation factors detected at a prevalence of <20%; a moderate expression group with SASP, DDR, and proliferation factors detected at a level ranging from 20% to 60%; and a high expression group with SASP, DDR, and proliferation factors detected in more than 60% of cells. These expression profile groups correlated with TP53 mutation status, such that low and moderate expression groups were clustered in those with TP53 wild-type status, while moderate and high expression groups were mainly detected in those positive for TP53 mutations (Figure 5).

To identify the possible biological role of the expression of SASP factors in early gastric cancer, associations between the expression of SASP factors and clinical features were estimated in the present study. In following up on the clinical course of each of the 30 cases, we found that four cases underwent subtotal gastrectomy (three cases) or total gastrectomy (one case) following the ESD procedure. Three patients with early gastric cancer relapsed; these patients were determined to have intramucosal carcinoma (pT1a) and well-differentiated adenocarcinoma (without radical operation).

Associations between the expression of SASP factors and other clinicopathological parameters were analysed using either a >10% or >20% cut-off. Cases with lesions located in the gastric pyloric antrum showed a statistically significantly elevated expression of IRF3, as compared with cases with lesions located in the body/fundus (>10%, p = 0.024; >20%, p = 0.015). Moreover, cases without relapse exhibited a markedly elevated expression of STAT6 (>10%, p = 0.014; >20%, p = 0.009). Poor differentiated tumour showed a statistically significantly elevated expression of cGAS, as the cut-off value was >10% (p = 0.009, detailed data not shown), but not >20% (p = 0.142). No statistically significant associations were found between the expression of SASP factors and other clinicopathological parameters, including sex, age, and tumour stage. Results for the >10% threshold are shown in Table 4.

Kaplan–Meier single-factor analysis using either a >10% or >20% cut-off and associated log-rank tests revealed that cases showing STAT6-positive expression demonstrated a longer progression-free survival time than those with STAT6-negative expression (>10%, p = 0.012; >20%, p = 0.025) (results for the >10% threshold are shown in Figure 6), suggesting that STAT6 expression may be an important predictive factor in early gastric carcinoma. cGAS, STING and IRF3 expression showed no statistically significant effect on progression-free survival (>10%, Figure 6). In addition, FANCD2, TP53BP1, RPA2, MCM7, and Ki67 expression as well as TP53 genomic mutations exhibited no statistically significant effects on progression-free survival (results for the >10% threshold are shown in Supplementary Figure S1).