The STAD cell lines AGS and MKN45 were cultured in F12K and DMEM containing 10-15% fetal bovine serum at 37°C in a humidified atmosphere containing 5% CO₂.

The AGS cells were cross-linked with formaldehyde and sonicated using a Bioruptor (Diagenode, Belgium), and the precipitate was removed via centrifugation. Protein A/G magnetic beads (Pierce, USA) were incubated overnight with a histone-3-lysine-27 acetylation (H3K27ac) antibody (Abcam, USA) at 4°C. The DNA fragments were eluted and recycled for chromatin immunoprecipitation sequencing (ChIP-seq) on a BGISEQ-500 platform (BGI-Shenzhen, China).

Raw ChIP-seq data of MKN45 cells and 11 STAD tumors were downloaded from the Gene Expression Omnibus database. The downloaded raw data and our sequencing data were ordered via trimming, Bowtie2 read mapping, filtering, and MACS2 peak calling. Bam files were converted into bigwig files using bamCoverage. Hockey stick plots and positional information of super-enhancers were analyzed using the rank ordering of super enhancers algorithm (5). The Integrative Genomics Viewer was used to visualize the bigwig files of the ChIP-seq data.

The STAD dataset was downloaded from the Cancer Genome Atlas (TCGA; https://cancergenome.nih.gov/), which includes 375 tumor samples and 32 non-malignant stomach tissue samples. Patients with STAD who were included in the TCGA were divided into high and low expression groups according to the median. GraphPad software was used to draw all survival-related graphs. Statistical significance was set at p<0.05.

Immune cell abundance was evaluated using six algorithms, namely ImmuCellAI (11), CIBERSORT (12), EPIC (13), quantiSeq (14), TIMER (15), and xCELL (16). The correlation of the lncRNA expression matrix with immune cell abundance was analyzed.

Immune markers including immunomodulators (immunoinhibitors and immunostimulators), MHC molecules, chemokines and chemokine receptor were downloaded from TISIDB (http://cis.hku.hk/TISIDB/download.php). The expression of these markers was from TCGA database. The correlation of the lncRNA expression matrix with immune markers level was also analyzed in STAD.

The TM4SF1-AS1 small interference RNA (siRNA) sequences were designed and synthesized by GenePharma (Suzhou, China). The siRNA was diluted using sterile DEPC water and transfected into AGS and MKN45 cells using Lipofectamine RNAiMAX Transfection Reagent (Invitrogen, USA). The siRNA sequences of TM4SF1-AS1 are listed in Table 1 .

RNA was isolated from cells using the TRIzol method. Cells in a 6-well plate were lysed with 1 mL Trizol in each well, and 200 μL chloroform was added, shaken, mixed, and centrifuged at 12,000 rpm at 4°C for 10 min after being incubated for 5 min. Next, 500 μL isopropanol was added, mixed, and centrifuged. After being washed twice with 75% alcohol, the RNA precipitate was dissolved in diethyl pyrocarbonate water. RNA was reverse transcribed to first-strand cDNA using Hifair III 1st Strand cDNA Synthesis SuperMix for qPCR (gDNA digester plus) (Yeasen, China).

ChamQ Universal SYBR qPCR Master Mix (Vazyme, China) was used to perform qRT-PCR on a CFX96 Real-Time System (Bio-Rad, USA). Relative mRNA expression was calculated using the 2^−ΔΔCT method and normalized to that of GAPDH. The primers are shown in Table 2 .

CD8+ T cells were separated from human PBMCs using the EasySep™ Human CD8+ T Cell Isolation Kit (STEMCELL Technologies, Canada). Samples were prepared at the indicated cell concentrations within the volume range. The isolation cocktail was added to the required tube, mixed, and incubated. Vortex RapidSpheres™, add it to the sample and mix them. The recommended medium was added to top up the sample to the indicated volume, then mixed by gently pipetting the sample up and down 2–3 times. The tube was placed into the magnet without a lid and incubated. The magnet was then picked up and the magnet and tube were inverted in one continuous motion, pouring the enriched cell suspension into a new tube.

First, stomach cancer cells were stained using the CellTrace CFSE Cell Proliferation Kit (Invitrogen, USA) for 20 min at 37°C. Activated CD8+ T cells (2 × 10⁴) were generated by incubation with ImmunoCult™ Human CD3/CD28 T Cell Activator (STEMCELL Technologies, Canada) at 37°C and 5% CO₂ for up to 3 days, and were then added to cancer cells. Both activated CD8+ T cells and cancer cells were collected to be stained with propidium iodide (3.75 mM solution, 1:500 final dilution). These cells were harvested and analyzed using flow cytometry after 12 h (Beckman CytoFLEX, USA).

Hi-C data of two STAD samples, T2000877 (GSM3333325) and T990275 (GSM3356360), were downloaded from the Gene Expression Omnibus database (GSE118391; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE118391). The results of this analysis were visualized and graphed using WashU tool (http://epigenomegateway.wustl.edu/browser/).

Each of the five enhancer regions (E1–E5) and their negative control (NC) were amplified using PCR or gene synthesis, then cloned into the pGL3-promoter vector. The primers used for PCR amplification of each enhancer are listed in Table 2 .

The vectors were transfected into STAD AGS and MKN45 cells, and pRL-TK plasmids were co-transfected as a normalization control. Luciferase assays were conducted using the Dual-Luciferase Reporter Assay System (Promega, USA).

Total RNA was extracted from cells using the TRIzol method as previously described, then qualified and quantified using a Nano Drop and Agilent 2100 bioanalyzer (ThermoFisher Scientific, USA). The mRNA library was constructed and sequenced on a BGIseq500 system (BGI-Shenzhen, China).

GSEA 4.1.0 software (http://www.broadinstitute.org/gsea/index.jsp) was downloaded and employed to analyze critical gene enrichment pathways between different groups based on FPKM values, which reflect relative gene expression.

All data are presented as mean ± SD. The Student’s t-test or analysis of variance was performed using either SPSS or GraphPad 8.0.1 software. Values were considered statistically significant at p<0.05 (*, p<0.05; **, p<0.01; ***, p<0.001).

LncRNA plays a crucial role in biological processes and are regulated by SEs in several types of cancer. However, no studies on the relationship between lncRNAs and super-enhancers in STAD have been conducted to date. We first performed ChIP-seq with the H3K27ac antibody in AGS cell lines to identify SE-associated lncRNA in STAD, and downloaded raw data of 11 tissue samples and MKN45 cell lines from GEO database (GSE117953). We identified 1722, 1273, 1514, 2150, 2017, 1434, 1536, 1935, 983, 1978, and 2393 SE-associated lncRNAs in 11 STAD samples, respectively (T980401, T990275, and other 9 samples), and 1049 and 719 in two STAD cells, respectively (AGS and MKN45). Hockey stick plots of two representative tumor samples and both cells showed some SE-associated lncRNA ( Figures 1A–D ). Unsurprisingly, some common and vital lncRNAs that promote tumor progression are at the forefront of the H3K27ac signal, such as MALAT1, H19, and CCAT1 ( Figures 1A–D ). We also identified some novel SE-associated lncRNA in STAD ( Figures 1A–D ). According to the results of ChIP-seq, SE-associated lncRNAs in AGS and MKN45 cell lines intersected, and 386 common SE-associated lncRNAs were obtained ( Figure 1E ). The 386 SE-associated lncRNAs of the above two cell lines and the SE-associated lncRNAs obtained in the same way, which were present in over six STAD tissue samples, were intersected and 308 common SE-associated lncRNAs were obtained ( Figure 1F ). We analyzed 308 SE-associated lncRNAs in STAD samples and cells using ChIP-seq data against the H3K27ac antibody.

To further determine the expression profile and clinical value of SE-associated lncRNA, we first analyzed the expression changes of all lncRNAs in patients with STAD from TCGA database. The expression of 2961 lncRNAs in STAD patients were significantly upregulated or downregulated compared with that of normal samples ( Figure 2A ). The 2961 lncRNAs with altered expression in patients with intersected with the 308 SE-associated lncRNAs from the previous results, and 74 SE-associated and differentially expressed lncRNAs were obtained in total ( Figure 2B ). The heatmap revealed overall differential expression of these 74 SE-associated lncRNAs in normal stomach and STAD samples ( Figure 2C ). We next utilized the least absolute shrinkage and selection operator (LASSO) algorithm to evaluate the prognosis role of 74 SE-associated lncRNAs in STAD patients. So that four SE-associated lncRNAs were included in this prognosis model according to the minimum criteria. We first observed the risk scores in alive and dead STAD patients, and found that the risk scores in dead STAD patients was indeed higher than those in alive patients ( Figure 2D ). We divided these SE-associated lncRNAs into two groups, high risk and low risk on basis of the RiskScore signature. High risk group was showed to be correlated with worse prognosis in STAD patients compared with low risk group ( Figure 2E ). The distribution of risk scores in STAD patients were showed ( Figure 2F ). Taken together, we elicited 74 differentially expressed SE-associated lncRNAs and assessed their correlation with the clinical importance of STAD patients.

Immune cells always affect the occurrence and development of tumors in the tumor microenvironment, and the relationship between immune cells and tumors can be used to develop specific drugs to inhibit tumor progression, such as PD1 and PDL1 inhibitors. We then examined whether SE-associated lncRNAs in STAD are related to the infiltration of immune cells. We used six algorithms to analyze the correlation between SE-associated lncRNAs and immune cells infiltration, namely ImmuCellAI, CIBERSORT, EPIC, quantiSeq, TIMER, and Xcell, respectively. It was found that all 74 differentially expressed SE-associated lncRNAs in STAD were significantly correlated with the immune cell infiltration using the ImmuCellAI ( Figure 3A ) and Xcell algorithms ( Figure 3B ). These immune cells included T and B cells, DC, neutrophils, and macrophage cells. In addition, most differentially expressed SE-associated lncRNAs in STAD were found to be significantly correlated with the immune cell infiltration by the other four algorithms, namely CIBERSORT ( Supplementary Figure S1A ), EPIC ( Supplementary Figure S1B ), quantiSeq ( Supplementary Figure S1C ), and TIMER ( Supplementary Figure S1D ). These data indicated that a correlation exists between differentially expressed SE-associated lncRNAs and immune cell infiltration in STAD.

In the tumor immune microenvironment, immune-related genes play a vital role in the regulation of tumor killing, including immunomodulators and chemokines. Therefore, we investigated whether SE-associated lncRNAs in STAD are related to immune makers. The list of immunomodulators and chemokines was downloaded from TISIDB database (http://cis.hku.hk/TISIDB/download.php). We first analyzed the correlation between SE-associated lncRNAs and immunomodulators in STAD and found that both immunoinhibitors and immunostimulators exhibited a significant correlation with SE-associated lncRNAs ( Figures 4A, B ). In addition, SE-associated lncRNAs were significantly correlated with MHC molecules ( Figure 5A ).

We then analyzed the correlation between SE-associated lncRNAs and chemokines in STAD and found that chemokines and chemokine receptors were significantly related to SE-associated lncRNAs ( Figures 5B, C ). In summary, these data show that SE-associated lncRNAs in STAD are related to immunomodulators and chemokines, and they participate in tumor immune regulation of tumors.

We selected a CD8+ T cell-related lncRNA, TM4SF1-AS1, as a representation to investigate the role of SE-associated lncRNAs in tumor immune regulation of STAD. First, the expression of TM4SF1-AS1 in STAD samples was analyzed, and the expression of TM4SF1-AS1 was found to be significantly upregulated in STAD samples, which differed from normal stomach samples ( Figure 6A ). We then observed the abundant differences in TM4SF1-AS1 expression in CD8+ T cells and classified subtypes as either high or low TM4SF1-AS1 expression. The abundance of CD8 naïve cells in the group with high TM4SF1-AS1 expression was found to be significantly lower than that of samples with low TM4SF1-AS1 expression. The abundance of CD8⁺ T, Tc, and Tgd cells was the opposite ( Figure 6B ).

We also performed an in vitro T cell killing assay to observe the killing effect of activated T cells on tumor cells with TM4SF1-AS1 interference. First, we found that the #3 and #4 siRNA against TM4SF1-AS1 had the best knockdown efficiency ( Supplementary Figures S2A, B ), and we selected siRNA#4 for further experimentation. Subsequently, the T cell killing assay showed that TM4SF1-AS1 interference significantly weakened the killing effect of activated CD8+ T cells on tumor cells in a T cell concentration-dependent manner ( Figures 6C, D ).

Furthermore, we analyzed immunophenoscore in CTLA- PD1-, CTLA- PD1+, CTLA+ PD1-and CTLA+ PD1+ patients between the two groups with high and low TM4SF1-AS1 expression, respectively. Immunophenoscore information of CTLA- PD1-, CTLA- PD1+, CTLA+ PD1- and CTLA+ PD1+ patients were downloaded from TICA database (https://tcia.at/home). The immunophenoscore of CTLA^- PD1+ and CTLA+ PD1+ samples in the group with high TM4SF1-AS1 expression was significantly lower than that of samples with low TM4SF1-AS1 expression, while no significant difference was observed in CTLA- PD1- and CTLA+ PD1- patients between the two groups ( Figure 6E ). We analyzed the expression of PD1, PD-L1, PD-L2, and CTLA4 in both groups. The expression of PD1, PD-L1, and PD-L2 was significantly higher in the group with high TM4SF1-AS1 expression ( Figure 6F ). and the expression of CTLA4 did not differ significantly between the high and low TM4SF1-AS1 expression groups ( Figure 6G ). In summary, these results indicate that TM4SF1-AS1 is involved in the T cell-mediated killing effect of tumor cells and monitors the immune response to anti-PD1 therapy.

Since TM4SF1-AS1 was chosen as a representation of SE-associated lncRNAs in STAD, we further verified the regulation of the SE on TM4SF1-AS1. ChIP-seq data revealed one to two SE peaks in TM4SF1-AS1 in 11 STAD tumor samples and both AGS and MKN45 cell lines ( Figure 7A ). Subsequently, Hi-C data revealed that a spatial interaction exists between the SE regions and the TM4SF1-AS1 promoter ( Figure 7B ). We then cloned the five enhancer sequences and their negative control sequences into the PGL3-promotor vector for dual luciferase experiments. The fluorescence intensity of E2 and E3 was significantly higher than that of the NC group ( Figures 7C, D ). We confirmed that TM4SF1-AS1, a SE-associated lncRNA, is indeed regulated by its super-enhancer in STAD based on the above results.

Finally, we performed RNA-seq on AGS and MKN45 cells after treating them with TM4SF1-AS1 interference to further understand the tumor immune-related processes of TM4SF1-AS1 involved in STAD. First, we selected two siRNA targets, siTM4SF1-AS1#3 and siTM4SF1-AS1#4, for RNA-seq in two STAD cells. Next, we analyzed differentially expressed genes by comparing siTM4SF1-AS1 with siNC. We set the cut-off as log2|FC| ≥1 and Q<0.05, and identified both downregulated and upregulated genes following TM4SF1-AS1 knockdown in both AGS ( Figures 8A, B ) and MKN45 cells ( Figures 8C, D ). The shared differentially expressed genes in groups between siTM4SF1-AS#3 and siTM4SF1-AS#4 of AGS and MKN45 cell lines were 304 and 337, respectively ( Figures 8E, F ). Heatmap showed the expression change of shared differentially expressed genes in groups of siTM4SF1-AS#3, siTM4SF1-AS#4 in comparison with siNC group ( Figures 8G, H ). GSEA analysis revealed that TM4SF1-AS1 can regulate T cell-related biological processes in AGS ( Figures 8E, F ) and MKN45 cells ( Figures 8G, H ). KEGG pathway analysis revealed that TM4SF1-AS1 modulates several biological processes and pathways, including protein processing in endoplasmic reticulum (ER), cell cycle, microRNAs in cancer, cellular senescence, and virus infection in AGS ( Figures 8I, J ) and MKN45 cells ( Figures 8K, L ). These data further indicate that TM4SF1-AS1 is involved in tumor processes and immune regulation in STAD.