LncRNA NEAT1 promotes immunosuppression in gastric cancer under endoplasmic reticulum stress by maintaining the M⁶A methylation of SEMA3A in CAFs

Abstract

Background

Despite PD1 inhibitors offering new gastric cancer therapies, disease burden persists. Cancer-associated fibroblasts were found to be important in gastric cancer regarding tumor promotion and immunosuppression, but its role in gastric cancer under endoplasmic reticulum stress remained unknown.

Methods

We conducted mass spectrometry in cancer-associated fibroblasts and transcriptomic-seq in exosomes from gastric cancer, combined with single-cell RNA sequencing analysis and spatial transcriptome sequencing analysis in gastric cancer to uncover the cellular crosstalk. Endoplasmic reticulum stressed models were established, and the isolation of primary cells and exosomes were used to uncover the communication between gastric cancer cells and cancer-associated fibroblasts. RNA pull-down, RIP-qPCR, flow cytometry detection and ELISA were used to demonstrate the interaction and immunity function in gastric cancer under endoplasmic reticulum stress. Moreover, we developed αvβ3-targeted cationic liposomes delivering siRNA of lncRNA NEAT1 towards tumor cells in gastric cancer in vivo.

Results

We found m⁶A protein METTL3, SEMA3A and lncRNA NEAT1, which interacts with METTL3, were highly expressed in cancer-associated fibroblasts under endoplasmic reticulum stress. Endoplasmic reticulum stress stimulation enhanced exosome secretion from gastric cancer cells and significantly elevated the expression of lncRNA NEAT1 in cancer-associated fibroblasts. LncRNA NEAT1 upregulated METTL3 in CAFs, which resulted in the enhanced m⁶A methylation of SEMA3A mRNA and amplified Treg-mediating immunosuppression. The combination therapy of siNEAT1@Lip-cRGD and anti-PD1 boosted T-cell infiltration and suppressed tumor growth in vivo.

Conclusions

Our study unveiled the lncRNA NEAT1/METTL3/SEMA3A axis stimulated by endoplasmic reticulum stress that sustained gastric cancer immunosuppression, providing a rationale for targeting this pathway to improve immunotherapy efficacy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12943-026-02613-w.

Background

Most patients with gastric cancer (GC) have advanced tumor at the time of diagnosis and have missed the opportunity for radical surgery. For patients with advanced gastric cancer, traditional treatments such as chemotherapy are less effective and usually yield a survival benefit of no more than one year [1]. The clinical application of molecular targeted therapy has improved the prognosis of many tumors, but limited progress has been made in the treatment of advanced gastric cancer [2]. Although the emergence of immune checkpoint inhibitors targeting PD1 and PD-L1 provided new therapeutic strategies for gastric cancer patients, poor response rate of less than 20% remains unsolved [3]. At present, the greatest challenge of immunotherapy application in gastric cancer is the limited understanding of the immune escape mechanism between tumor cells and immune cells [4, 5]. In order to explore better immunotherapy combination strategy for gastric cancer, extensive exploration of the molecular mechanism of immune escape in gastric cancer should be wildly conducted.

The tumor microenvironment (TME) contributes to tumor development through multiple mechanisms [6]. Changes in the TME, such as hypoxia, oxidative stress, and calcium ion homeostasis imbalance, lead to the aggregation of unfolded or misfolded proteins in the endoplasmic reticulum [7]. Such disruption of homeostasis in the endoplasmic reticulum triggers endoplasmic reticulum stress (ERS) [8]. ERS plays a protective role in tumor development by activating multiple intracellular compensatory mechanisms [9]. So that, the consequences of ERS affecting different cells in the TME urgent to be explored [10]. Multiple studies have shown that ERS inhibits the occurrence of the antitumor immune response by regulating the function of immune cells in the TME, which revealing a new mechanism of tumor immune escape [11]. Regarding to gastric cancer, ERS can be observed throughout the processes of tumorigenesis and progression and closely related to the prognosis of gastric cancer and the formation of an immunosuppressive microenvironment [12]. However, the specific regulatory mechanism remains unclear.

As one of the most important components of the TME, cancer-associated fibroblasts (CAFs) secrete a variety of cytokines [13]. Thereby, CAFs promote tumor angiogenesis, disrupting homeostasis between tissue cells [14] and making the microenvironment more conducive to tumor growth [15]. Studies have revealed that CAFs promote the growth of breast cancer [16], liver cancer [17], gastric cancer [18], colorectal cancer [19], lung cancer [20], pancreatic cancer [21] and other common cancers [15]. CAFs promote cancer therapeutic resistance by producing soluble factors that can act on cancer cells through paracrine mechanisms [22]. Besides, CAFs contributes to metabolic reprogram and adaptation [23]. Intrastromal crosstalk between CAFs and other stromal cells plays an important role in cancer progression and treatment resistance [24]. CAFs regulate immune cell-mediated antitumor immunity by producing various cytokines, chemokines and other effector molecules [25]. CAFs promote the polarization or differentiation of immune cells into a specific subgroup of tumor-promoting cells and can also support the roles of immunosuppressive cells in activation, recruitment and immunosuppression and limit the cytokine production and cytotoxic effects of effector immune cells [26].

ERS is a prevalent phenomenon during gastric cancer development and progression, closely associated with patient prognosis and the formation of an immunosuppressive TME. However, the underlying causes and consequences of ERS in non-malignant stromal cells within the TME remain poorly understood. Our previous work has demonstrated the critical regulatory role of ERS in tumor drug resistance, though the precise molecular mechanisms remain elusive [27, 28]. Therefore, systematic investigation of ERS-mediated immunomodulatory pathways and their molecular basis may provide novel strategies to complement and enhance current immunotherapies for gastric cancer. Through multi-expression profiling of ERS-activated tumor cells and their secreted exosomes, along with CAFs in the ERS tumor microenvironment of gastric cancer, we revealed an immunosuppressive regulatory axis in this study.

Methods

Human tissue specimens

Between June 2020 and October 2021, ten patients diagnosed with gastric cancer were recruited from the Department of General Surgery at Nanfang Hospital. Tumor tissue specimens were collected within one hour postoperatively. A gastric cancer tissue sample, measuring approximately 1 cm³, was excised from the central region of the tumor. Additionally, a corresponding normal tissue sample of similar volume (1 cm³) was obtained from the adjacent non-neoplastic tissue. All procedures were conducted in compliance with ethical standards and with the informed consent of each participant.

Cell culture

Human gastric cancer cell lines AGS, HGC27 and MKN7 were obtained from ATCC, mouse forestomach carcinoma cell line MFC was obtained from iCell. Primary human gastric cancer-associated fibroblasts (h-CAFs) were obtained from Procell, and primary mouse gastric cancer-associated fibroblasts (m-CAFs) were utilized in this study. The AGS, HGC27 and MFC cell lines were maintained in RPMI 1640 medium (Solarbio) supplemented with 10% fetal bovine serum (FBS, Biological Industries). H-CAFs were cultured in human gastric cancer tissue fibroblast-specific complete medium (Procell), while m-CAFs were grown in high-glucose Dulbecco’s Modified Eagle Medium (DMEM, Solarbio) supplemented with 10% FBS. All media were supplemented with penicillin–streptomycin (Solarbio) as required. For subculturing, cells were detached using 0.25% trypsin–EDTA solution (Solarbio) upon reaching 80–90% confluency and were passaged at a 1:2 or 1:3 ratio. All cell lines were incubated at 37 °C in a humidified atmosphere containing 5% CO₂ using a cell culture incubator (ESCO, Singapore).

Isolation and validation of exosomes

Exosomes were isolated from the cell culture supernatants of MFC or HGC27 cells. Briefly, MFC and HGC27 cells were cultured in complete medium supplemented with 10% exosome-depleted fetal bovine serum. The supernatants were collected and subjected to sequential centrifugation steps: firstly at 300 × g for 10 min and at 4 °C to remove cells, followed by 3,000 × g for 10 min and at 4 °C to eliminate cell debris, finally at 10,000 × g for 10 min and at 4 °C to clear larger vesicles. The resulting supernatant was filtered through a 0.22 µm membrane. Then, the filtrate was ultracentrifuged at 100,000 × g for 70 min and at 4 °C to pellet the exosomes. The exosome pellet was resuspended in 200 µL of phosphate-buffered saline.

For downstream analyses, 50 µL of the resuspended exosomes was used for nanoparticle tracking analysis to determine particle size distribution, 50 µL was utilized for transmission electron microscopy to assess exosome morphology, and 100 µL was allocated for Western blot analysis. Exosomal RNA was extracted using the miRNeasy Mini Kit (QIAGEN, Germany) according to the manufacturer’s protocol. The expression of exosome-specific marker proteins (TSG101, CD9, and CD81) and the endoplasmic reticulum marker calnexin, as negative control, was evaluated by Western blotting. Nanoparticle tracking analysis was employed to quantify the size distribution and concentration of exosomal particles, while transmission electron microscopy was used to visualize the ultrastructure and morphology of the isolated exosomes.

Transcriptome sequence of exosome

Cellular supernatant was mixed with Ribo^TMExosome Isolation Reagent and exosome isolation was performed according to the manufacturer’s instructions (Ribobio, China). Exosomal RNA was extracted by Magzol Reagent (Magen, China). The quantity of RNA yield was assessed by using the Qubit (Thermo Fisher Scientific, USA) and the Agilent 2200 TapeStation (Agilent Technologies, USA). Briefly, rRNAs were removed from Exosomal RNA using QIAseqFastSelect RNA Removal Kit (QIAGEN, Germany) and fragmented to approximately 200 bp. Subsequently, the RNA fragments were subjected to first strand and second strand cDNA synthesis following by adaptor ligation and enrichment with a low-cycle according to instructions of NEBNext®Ultra™ RNA LibraryPrep Kit for Illumina (NEB, USA). The purified library products were evaluated using the Agilent 2200 TapeStation and Qubit. The libraries were sequenced by Illumina (Illumina, USA) with paired-end 150 bp at Ribobio Co. Ltd (Ribobio, China). The clean reads were obtained after removal of reads containing adapter, ploy-N and at low quality from raw data. HISAT2 was used to align the clean reads to the mouse reference genome mm10 with default parameters. HTSeq (RRID:SCR_005514) was subsequently employed to convert aligned short reads into read counts for each gene model. Differential expression was assessed by DEseq/DESeq2/edgeR/DEGseq using read counts as input. The Benjamini–Hochberg multiple test correction method was enabled. Differentially expressed genes were chosen according to the criteria of fold change > 2 and adjusted p-value < 0.05. All the differentially expressed genes were used for heat map analysis and KEGG ontology enrichment analyses. For KEGG enrichment analysis, a p-value < 0.05 was used as the threshold to determine significant enrichment of the gene sets.

Bioinformatics analysis approaches for single-cell sequencing (scRNA-seq) data

The scRNA-seq data of gastric cancers were retrieved from the Gene Expression Omnibus (GEO) database (GSE231539 [29], comprises 22 samples, including 12 gastric tumor tissue samples and 10 adjacent normal tissue samples). Processing and analysis of the scRNA-seq data were performed using Seurat (v5.2.1). Nearest neighbors were identified using the FindNeighbors function based on the top 30 dimensions. Cluster markers were identified using FindAllMarkers with test.use = "MAST", and cell types were annotated based on canonical marker genes. Differentially expressed genes (DEGs) between groups were identified using FindMarkers with test.use = "MAST"). Gene co localization analysis was performed using the FeaturePlot_scCustom function from scCustomize package. The ERS–CAFs score was calculated using AddModuleScore with the following genes: ('HSPA5', 'XBP1', 'ATF4', 'HSPA1A', 'HSPA1B', 'CCL2', 'DNAJB1', 'HSPA6', 'HIF1A', 'BAG3', 'VEGFA').

Bioinformatics analysis approaches for spatial transcriptome sequencing (ST-seq) data

The single-cell dataset GSE183904 comprises 40 samples, including 11 normal tissue samples and 29 tumor tissue and peritoneal tumor tissue samples [30]. A total of 26 primary tumor samples were selected for analysis. Genes expressed in more than 300 cells were retained, and cells with mitochondrial gene expression exceeding 10% were filtered out. Doublets were removed using the Scrublet package [31], and RNA contamination was eliminated using DecontX [32], resulting in a final dataset of 113,372 high-quality cells. After dimensionality reduction and clustering, batch effects were corrected using the Harmony algorithm. Cell types were annotated and visualized based on published studies [30, 33, 34]. In addition, malignant cells were identified using the InferCNV package.

Spatial transcriptomic data were obtained from GSE251950 [35] and processed using the Seurat v4.0 package in R. A SCTransform-based normalization and integration workflow was applied: each sample was independently normalized using SCTransform. Principal component analysis (PCA) was performed on the integrated data, followed by graph-based clustering based on the PCA results. To identify cell types at spatial transcriptomic spots, RCTD [36] was used to train cell type–specific signature profiles from the paired single-cell data, which were then applied to the spatial data for deconvolution to obtain the absolute abundance of each cell type. Using the MistyR package [37], a multiview spatial model was constructed. Specifically, the cell abundance matrix was input into the MistyR framework to build both intra-cellular and inter-cellular views, enabling quantification of the contributions of intrinsic cellular composition and local microenvironmental interactions to specific molecular phenotypes.

RNA extraction and real-time quantitative polymerase chain reaction (RT‒qPCR)

Total RNA was extracted from cell and tissue samples using TRIzol reagent (Novozymes, Nanjing, China). Subsequently, RNA was reverse transcribed into complementary DNA using HiScript II Q RT SuperMix for qPCR (Novozymes, Nanjing, China) in a 10 µL reaction volume, following the manufacturer’s protocol. Quantitative reverse transcription PCR (qRT‒PCR) was performed using a qRT‒PCR MIX (Yisheng, Shanghai, China) in a 10 µL reaction system containing cDNA as the template. Amplification was carried out on a fluorescence quantitative PCR analyzer (Quant Studio 6F, Thermo Fisher Scientific, Singapore) under the following cycling conditions: initial denaturation at 95 °C for 30 s, followed by 45 cycles of denaturation at 95 °C for 5 s and annealing/extension at 60 °C for 30 s. The expression levels of target genes were normalized to those of the housekeeping genes GAPDH or β-actin using the 2^−∆∆Ct method for relative quantification. For statistical analysis, data were processed using GraphPad Prism 9.0 (RRID:SCR_002798). Differences between groups were assessed by one-way analysis of variance (ANOVA), and a p-value < 0.05 was considered statistically significant. The qPCR primers used are listed in the following table.

Gene	Primer-F (5'–3')	Primer-R (5'–3')
GAPDH (human)	GGAGCGAGATCCCTCCAAAAT	GGCTGTTGTCATACTTCTCATGG
SEMA3A (human)	GTGCCAAGGCTGAAATTATCCT	CCCACTTGCATTCATCTCTTCT
NEAT1 (human)	GGCAGGTCTAGTTTGGGCAT	CCTCATCCCTCCCAGTACCA
Mut-SEMA3A (human)	TGTTCCCTGGGGTATGTCCT	CAGCTCATCAACCACCCCAA
Mut-NEAT1 (human)	TCTTCTAAATTGAGCCTCCGGT	GCGTTTAGCACAACACAATGAC
METTL3 (human)	TTGTCTCCAACCTTCCGTAGT	CCAGATCAGAGAGGTGGTGTAG
METTL14 (human)	GAACACAGAGCTTAAATCCCCA	TGTCAGCTAAACCTACATCCCTG
GRP78 (human)	ACATGGACCTGTTCCGCTCTA	TGGCTCCTTGCCATTGAAGA
GAPDH (mouse)	AGGCCGGTGCTGAGTATGTC	TGCCTGCTTCACCACCTTCT
SEMA3A (mouse)	TGTGCCAATTTCATCAAGGTCC	CTCTTCCCACGACCGTTTTCA
GRP78 (mouse)	CCAAAGATTCAGCAACTGGTG	ATCAAGCAGTACCAGATCACC
ITGAV (mouse)	TTGTTGCCGCCTTACGAGAA	TTGGCCTTCTGTTGCCCTAC

Western blotting (WB)

Total protein was extracted from cell or tissue lysates prepared using RIPA lysis buffer (Beyotime, China) supplemented with a phosphatase inhibitor cocktail (GBCBIO, China), a protease inhibitor cocktail (Leagene, China), and phenylmethylsulfonyl fluoride (PMSF, Solarbio, China). Protein concentrations were determined using a bicinchoninic acid protein assay kit (DINGGUO, China) according to the manufacturer’s instructions. For protein separation, SDS‒polyacrylamide gels were prepared using a PAGE Rapid Gel Preparation Kit (Epizyme, China). Proteins were resolved by electrophoresis and subsequently transferred onto a polyvinylidene difluoride membrane (Millipore, USA) using a wet transfer system. The membrane was blocked with 5% nonfat dry milk or bovine serum albumin in Tris-buffered saline with Tween-20 for 1 h at room temperature. Following blocking, the membrane was incubated overnight at 4 °C with primary antibodies specific to the target proteins. After washing, the membrane was incubated with horseradish peroxidase HRP-conjugated secondary antibodies (UELandy, China) for 1 h at room temperature. Protein bands were visualized using an enhanced chemiluminescence substrate, and images were captured using a CCD camera-based imaging system. The antibodies used in this assay are listed below.

Antibody	Company
GAPDH	UELandy, China
GRP78	Proteintech, USA
α-SMA	Bimake, USA
Vimentin	Bimake, USA
SEMA3A	Abcam, UK
METTL3	Abcam, UK
METTL14	Abcam, UK
YTHDF1	Abcam, UK
YTHDF2	Abcam, UK
YTHDF3	Abcam, UK
HA tag	Abcam, UK
Ubiquitin	Abcam, UK
Calnexin	Proteintech, USA
TSG101	Proteintech, USA
CD81	Proteintech, USA
CD9	Proteintech, USA

Isolation of primary cells

Tumors were excised from C57BL/6 mice (RRID: IMSR_JAX:000664) following subcutaneous inoculation with MFC cells. The tumor tissues were rinsed with phosphate-buffered saline containing 5% penicillin‒streptomycin‒gentamicin (Solarbio, China). Tissue samples approximately 1 mm³ in size were dissected and transferred into a pre-prepared enzymatic digestion solution. The digestion solution consisted of 20 mg/ml collagenase IV (Birfroxx, Germany), 20 μg/ml hyaluronidase (Solarbio, China), 20 μg/ml DNase I (Coolaber, China), and Dulbecco’s Modified Eagle Medium (DMEM, Solarbio, China). The tissue mixture was incubated at 37 °C for 1.5–2 h with gentle agitation to facilitate digestion. After digestion, the suspension was filtered through a 70 μm cell strainer to remove undigested tissue fragments and obtain a single-cell suspension. To eliminate erythrocytes, the cell suspension was treated with erythrocyte lysis buffer (Leagene, China) for 10 min at room temperature. The resulting cell suspension was then seeded into 6-well plates. Mouse-derived primary cancer-associated fibroblasts (m-CAFs) were isolated using the differential adherence method. For the isolation of mouse-derived primary regulatory T cells (m-Tregs), single-cell suspensions from the tumor tissues, as well as lymphocytes isolated from mouse spleens by mechanical dissociation, were stained with anti-CD4 and anti-CD25 flow cytometry antibodies (BioLegend, USA) for 30 min at 4 °C. Unbound antibodies were removed by washing the cells three times with PBS. CD4⁺CD25⁺ cells were subsequently sorted using a fluorescence-activated cell sorting (FACS) system (BD Aria III, USA).

Treg function assay

Regulatory T cells (Tregs) were fixed by incubating with 4% paraformaldehyde solution (Leagene, China) at room temperature for 15 min in the dark. Following fixation, the cells were stained with fluorescently labeled anti-IL-10 and anti-TGF-β antibodies (BioLegend, USA) and incubated at room temperature in the dark for 30 min. Unbound antibodies were removed by washing the cells three times with phosphate-buffered saline. The expression levels of IL-10 and TGF-β were analyzed using a flow cytometer (BD Aria III, USA).

RNA pull-down assay

Biotinylated SEMA3A mRNA probe and lncRNA NEAT1 probe were synthesized and purified. CAFs under ERS (or control) were lysed in RNA pull-down lysis buffer (20 mM Tris–HCl pH 7.5, 150 mM KCl, 1.5 mM MgCl₂, 0.5% NP-40, 1 mM DTT, protease inhibitor cocktail, RNase inhibitor). Lysates were centrifuged (12,000 × g, 15 min, 4 °C) and supernatants collected for binding reactions. Streptavidin magnetic beads (Pierce) were washed and blocked with yeast tRNA (0.1 mg/mL) and BSA (0.5 mg/mL) for 1 h at 4 °C. 2 µg of biotinylated RNA probe (NEAT1, SEMA3A mRNA, or antisense control) was incubated with pre-blocked beads for 30 min at 25 °C. Beads-bound probes were incubated with 500 µg of CAF lysate for 2 h at 4 °C with gentle rotation. Beads were washed 5 times with lysis buffer. Bound proteins were eluted by boiling in 1 × SDS loading buffer for 10 min. Western blotting with anti-METTL3, anti-SEMA3A, anti-YTHDF1, anti-YTHDF2, anti-YTHDF3 antibodies were used for detection.

M⁶A methylated RNA immunoprecipitation followed by quantitative PCR (MeRIP-qPCR)

Total RNA was extracted from ERS-stimulated h-CAFs using TRIzol. RNA (10 μg) was fragmented into 100–200 nt fragments with RNA fragmentation buffer (Thermo Fisher, Cat. No. AM8740) at 94 °C for 5 min, then terminated with 0.5 M EDTA. Protein A/G magnetic beads (50 μL) were washed 3 × with PBS, incubated with 5 μg anti-m⁶A antibody at 25 °C for 1 h, and washed 2 × with MeRIP lysis buffer (10 mM Tris–HCl pH 7.4, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1% NP-40, 1 × protease/RNAse inhibitor). Fragmented RNA was mixed with the antibody-bead complex and incubated at 4 °C overnight. Beads were washed 2 × with low-salt buffer (10 mM Tris–HCl pH 7.4, 150 mM NaCl, 0.1% NP-40, 0.5 mM EDTA) and 3 × with high-salt buffer (10 mM Tris–HCl pH 7.4, 500 mM NaCl, 0.1% NP-40, 0.5 mM EDTA). M⁶A-modified RNA was eluted with elution buffer (50 mM Tris–HCl pH 7.4, 10 mM EDTA, 0.5% SDS) containing 20 μg/mL proteinase K at 55 °C for 1 h. RNA was purified with an RNA purification column (Qiagen, Cat. No. 74104), reverse-transcribed into cDNA, and analyzed by qPCR with SEMA3A-specific primers.

RNA stability detection (actinomycin D treatment)

Cells in the logarithmic growth phase were seeded at an appropriate density in 6‑well plates and cultured until reaching 70–80% confluence. The medium was replaced with fresh complete medium containing actinomycin D at a final concentration of 5 μg/mL. Following drug addition, cells were harvested at predetermined time points. For RNA extraction at each time point, the medium was aspirated, and cells were washed twice with pre‑cooled, RNase‑free PBS to remove residual drug. TRIzol reagent was added to each well, and cells were thoroughly lysed by pipetting, followed by incubation at room temperature for 5 min. Total RNA was subsequently extracted, precipitated, washed, and resuspended according to standard protocols. RT‑qPCR was performed to quantify the remaining levels of target RNA.

RNA immunoprecipitation (RIP)-qPCR assay

H-CAFs (5 × 10⁶ cells) were harvested and washed twice with pre-cooled PBS. Add 1 mL of RIP Lysis Buffer (supplemented with 1 × protease inhibitor cocktail and 100 U/mL RNase inhibitor) to the cell pellet, resuspend thoroughly, and incubate on ice for 30 min. Centrifuge at 14,000 × g for 10 min at 4 °C, and collect the supernatant (cell lysate) for subsequent experiments. Take 50 μL of Protein A/G Magnetic Beads into a pre-cooled 1.5 mL RNase-free centrifuge tube, wash 3 times with 500 μL of RIP Wash Buffer (each wash: 3,000 × g, 30 s at 4 °C). Resuspend the beads in 100 μL of RIP Wash Buffer, then add 5 μg of anti-METTL3 antibody, anti-m⁶A antibody, or normal rabbit IgG (negative control) respectively. Incubate at 4 °C with gentle rotation for 2 h to form antibody-bead complexes. Immunoprecipitation of RNA–Protein Complexes: Add 100 μL of cell lysate to each antibody-bead complex tube, then supplement with Nuclease-free water to a final volume of 500 μL (ensure the final concentration of RIP Lysis Buffer is maintained). Add 100 U/mL RNase inhibitor to prevent RNA degradation, and incubate at 4 °C with gentle rotation overnight (12–16 h) to allow the formation of RNA–protein-antibody-bead complexes. After incubation, place the centrifuge tube on a magnetic stand to separate the beads. Discard the supernatant, and wash the beads 6 times with 500 μL of pre-cooled RIP Wash Buffer (each wash: 3,000 × g, 30 s at 4 °C) to remove non-specifically bound proteins and RNA. Add 150 μL of RIP Elution Buffer (containing 0.5% SDS and 10 mM EDTA) and 2 μL of Proteinase K (20 mg/mL) to the washed beads, mix thoroughly, and incubate at 55 °C for 30 min with occasional vortexing to elute the RNA–protein complexes and digest the proteins. After digestion, add 200 μL of Nuclease-free water to each tube, then extract total RNA using TRIzol Reagent according to the standard protocol. Determine the RNA concentration and purity using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). Use 1 μg of purified RNA for reverse transcription to synthesize cDNA using the PrimeScript™ RT Reagent Kit with gDNA Eraser, following the manufacturer’s instructions (reaction conditions: 42 °C for 5 min, 37 °C for 15 min, 85 °C for 5 s). Continue qPCR tests using the method mentioned above.

Immunofluorescence (IF)

When the cells reached 80–90% confluency (Biosharp, China), they were washed twice with phosphate-buffered saline and fixed with 4% methanol (Leagene, China) for 20–30 min at room temperature. Following fixation, the cells were permeabilized with 1% Triton X-100 solution (Leagene, China) for 10 min and then blocked with 5% bovine serum albumin (BSA, Sigma, USA) in PBS for 30 min at room temperature to prevent nonspecific binding. The cells were incubated with primary antibodies diluted in QuickBlock™ immunostaining primary antibody diluent (Beyotime, China) for 1 h at room temperature. After washing with PBS to remove unbound primary antibodies, the cells were incubated with fluorophore-conjugated secondary antibodies (Abcam, UK) for 1 h at room temperature in the dark. Subsequently, the nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI, Servicebio, China) for 5 min. The slides were mounted using an anti-fade fluorescence mounting medium (Servicebio, China) to prevent photobleaching. Finally, the slides were briefly rinsed in acetone to remove residual mounting medium and imaged using a laser scanning confocal microscope (Olympus, Japan).

Immunohistochemistry (IHC)

Gastric cancer tissue samples were fixed in 10% neutral buffered formalin (Leagene, China) for 24–48 h at room temperature. Following fixation, the tissues were processed and embedded in paraffin using an ABM-A paraffin embedding machine. Tissue sections of 6 μm thickness were cut using a microtome and mounted onto glass slides. The sections were deparaffinized by immersion in xylene and rehydrated through a graded series of ethanol solutions (100%, 95%, 80%, and 70%) followed by distilled water. Immunohistochemistry was performed using a commercially available IHC kit (Beijing Lanjieke Technology Co., Ltd., China). Tissue sections were subjected to antigen retrieval by heating in citrate buffer (pH 6.0) or EDTA buffer (pH 9.0) as appropriate. Endogenous peroxidase activity was quenched by incubating the sections with 3% hydrogen peroxide for 10 min. The sections were then blocked with 5% bovine serum albumin in PBS for 30 min at room temperature to reduce nonspecific binding. Primary antibodies against α-SMA (Bimake, USA), GRP78 (Cell Signaling Technology, USA), and SEMA3A (Abcam, USA) were diluted according to the manufacturer’s recommendations and applied to the tissue sections overnight at 4 °C. After washing with PBS, the sections were incubated with biotinylated secondary antibodies for 1 h at room temperature. Immunoreactivity was visualized using a 3,3′-diaminobenzidine chromogen substrate, and the sections were counterstained with hematoxylin to highlight nuclei. Finally, the slides were dehydrated, cleared in xylene, and mounted with a coverslip using a permanent mounting medium. Stained sections were scanned using a high-resolution slide scanner for further analysis.

Cell transfection

When the cells reached approximately 60% confluency, the culture medium was replaced with serum-free medium. For transfection, plasmids encoding HA-tagged ubiquitin (HA-Ub), lentiviral vectors expressing shRNA for SEMA3A was purchased from IGE Biotechnology; siRNA targeting METTL3, YTHDF1 and lncRNA NEAT1 were purchased form Ribobio; SEMA3A motif-mutated (hg19_dna range = chr7:83590695–83591062) vectors of pcDNA3.1 were diluted according to the manufacturer’s protocol. LncRNA NEAT1 was knocked out (sgRNA-1: GACCCCGGTGACGCGGCTGA, PAM: GGG; sgRNA-2: CCAAGTAATTAAGGCCATGT, PAM: AGG) and mutant (sgRNA-1: GTTAAAATGTTCTACAGCTT, PAM: AGG; sgRNA-2: TAGCCTGATGAAATAACTTG, PAM: GGG) utilizing Cas9 technology. The diluted nucleic acids were then mixed with Hieff Trans® Liposomal Transfection Reagent (Yeasen, China) to form liposome-nucleic acid complexes. The transfection complexes were added to the cells and incubated for 24–48 h under standard culture conditions. The silencing efficiency was validated by quantitative reverse transcription PCR using gene-specific primers and normalized to housekeeping genes.

Animal models

MFC cells were subcutaneously inoculated into C57BL/6 mice. After one week, the mice were randomized into control and experimental groups. For co-inoculation, MFC cells were mixed with either wild-type mouse cancer-associated fibroblasts (m-CAFs) or treated m-CAFs at a 2:1 ratio and injected into the dorsal subcutaneous tissue of the mice. Tumor growth was monitored by measuring tumor volume every three days using a caliper, and the volume was calculated using the formula: Volume = (Length × Width²)/2. To investigate the effects of endoplasmic reticulum stress, mice were divided into two groups: the ERS experimental group, which received intraperitoneal injections of tunicamycin at a dose of 0.35 mg/kg (100 µL per 20 g body weight), and the control group, which received an equivalent volume of dimethyl sulfoxide at the same dose and frequency. Injections were administered every three days.

Enzyme-linked immunosorbent assay (ELISA)

The expression levels of IL-10 and TGF-β in cell culture supernatants were quantified using a mouse-specific enzyme-linked immunosorbent assay kit (Cusabio, China) according to the manufacturer’s instructions. The supernatants were collected from m-Tregs isolated from mouse tumor tissues and spleens, which were co-cultured with different treatment groups. A standard curve was generated using serial dilutions of recombinant protein standards provided in the kit, and the concentrations of IL-10 and TGF-β were determined by fitting the absorbance values to a linear regression model.

Flow cytometry (FCM)

Fluorophore-conjugated specific antibodies were incubated with single-cell suspensions to bind surface or intracellular target antigens. The stained cell suspension was hydrodynamically focused into a single-cell stream by sheath fluid and passed through the detection zone. When each cell was interrogated by the laser, two types of signals were generated: light-scattering signals, in which forward scatter (FSC) reflected cell size and side scatter (SSC) indicated cellular granularity and internal complexity; and fluorescence signals, where fluorophores on the cell were excited and emitted light of specific wavelengths, which was collected through optical filters and converted into electronic signals by detectors. Signal analysis was performed using FlowJo software, enabling population gating, cell counting, and quantitative analysis of fluorescence intensity.

RNA in situ hybridization (RNA-ISH) and protein co-staining detection

The lncRNA NEAT1 probe was purchased from Pinpoease (283111‑A2). H‑CAFs were cultured on slide chambers and fixed with 4% formaldehyde in 1 × PBS for 30 min at room temperature. After two washes with 1 × PBS, cells were treated with Pre‑treatment Solution A from the PinpoRNA™ 2.0 Kit for 10 min at room temperature, followed by two additional PBS washes. Digestion was performed using Enzyme II for 15 min at room temperature. The probe hybridization solution was prepared in 1 × PBS, and hybridization was carried out at 40 °C for 120 min. The first fluorogenic substrate was applied and incubated for 30 min at room temperature. After completing the immunofluorescence staining procedure, nuclei were counterstained and slides were mounted.

Preparation and characterization of siNEAT1@Lip-cRGD

To fabricate siNEAT1@Lip-cRGD, the organic phase was obtained by dissolving SM-102, cholesterol, DOPE, DSPE-PEG₂₀₀₀ and DSPE-PEG₂₀₀₀-cRGD in anhydrous ethanol (molar ratios of 50:10:38.5:0.75:0.75), while aqueous phase was obtained by dissolving siNEAT1 in 25 mM citrate buffer (pH = 4). Then, siNEAT1@Lip-cRGD was prepare by rapid mixing of aqueous phase and organic phase at the volume ratio of 1:3. The weight ratio of ionizable lipid and siNEAT1 was 50:1. Finally, the formulated siNEAT1@Lip-cRGD was further ultrafiltered through a 3000-Da ultrafiltration tube to remove organic solution. The particle size distribution was assessed by dynamic light scattering (DLS). The morphology of siNEAT1@Lip-cRGD was observed by transmission electron microscope.

Cellular uptake of siNEAT1@Lip-Crgd

MFC cells were seeded in confocal dishes and allowed to stand overnight. Subsequently, the culture medium was replaced with fresh medium containing Cy5.5-siNEAT1@Lip-cRGD (2 ug/mL siNEAT1) or Cy5.5-siNEAT1@Lip (2 ug/mL siNEAT1). At the indicated time points, the cells were stained with LysoTracker Green and then observed using confocal laser scanning microscopy.

Antitumor study of siNEAT1@Lip-cRGD

To investigate antitumor ability of siNEAT1@Lip-cRGD, MFC tumor-bearing mice was constructed by subcutaneously injecting MFC cells (1 × 10⁶) into the right hind leg. When the tumor volume reach about 150 mm³, mice were randomly assigned into four groups and then intravenously treated with PBS, anti-PD1 (5 mg/kg), siNEAT1@Lip-cRGD (0.2 mg/kg) and siNEAT1@Lip-cRGD + anti-PD1 (equivalent dose of siNEAT1 and anti-PD1). The mice received siNEAT1@Lip-cRGD treatments every 2 days for a total of three administrations, followed by anti-PD1 treatments for a total of two administrations. At the end of experiment, the mice were euthanized and tumor tissues were collected for H&E, TUNEL and Ki67 staining.

Antitumor immunity assessment

To investigate the immune activation of combined treatment of Cy5.5-siNEAT1@Lip and anti-PD1, MFC tumor-bearing BALB/c mice (RRID: MGI: 2683685) were treated intravenously with various therapeutic agents. Subsequently, tumor tissues, lymph nodes and spleen were collected. After preparing a single-cell suspension, the cells were stained with various antibodies for flow cytometry analysis.

Statistical analysis

All experimental data were expressed as mean ± standard deviation (S.D.) and were analyzed using GraphPad Prism 10 (GraphPad Software, USA). Statistical comparisons between two groups were performed using an unpaired two-tailed Student’s t-test. For comparisons involving three or more groups, one-way analysis of variance (ANOVA) was applied, followed by appropriate post hoc tests to determine specific group differences. A p-value < 0.05 was considered statistically significant.

Results

SEMA3A is highly expressed in CAFs of ERS gastric cancer tissues and promotes the formation of an immunosuppressive TME

To investigate the role of ERS in CAFs within the tumor microenvironment of gastric cancer, we first analyzed the relationship between GRP78 protein expression and CAFs infiltration in GC tissues using data from The Cancer Genome Atlas (TCGA, https://www.cancer.gov/ccg/research/genome-sequencing/tcga). A positive correlation between ERS status in GC and the proportion of CAFs in the TME was demonstrated (Fig. 1A). To further model the GC TME, h-CAFs were co-cultured with the supernatant (SN) of human gastric cancer cells treated with tunicamycin (TM) to induce ERS (Supplementary Fig. 1A-C). Mass spectrometry analysis revealed that ERS regulated the expression of 114 proteins in h-CAFs, including 58 upregulated and 56 downregulated proteins (Supplementary Table 1). The corresponding proteomics data have been deposited to the PRIDE database (Project accession: PXD061686). Notably, these included m⁶A-modified proteins and SEMA3A, an immunosuppression-related protein (Fig. 1B). Recent studies have shown that high SEMA3A expression recruited M2 macrophages and excluded CD8⁺ T cells to form an immunosuppressive microenvironment [38]. Moreover, targeting SEMA3A/Nrp1 had enhanced chemo-immunotherapy sensitivity, and combining its blockade with PD-1 inhibitors had boosted T-cell function and anti-tumor efficacy in solid tumor models [39]. However, its regulatory role on other immune cells in the tumor immune microenvironment remains unclear. Analysis of TCGA data further revealed a positive correlation between CAF infiltration and SEMA3A expression in GC tissues (Fig. 1C). Kaplan–Meier survival analysis indicated that high SEMA3A expression was associated with poor prognosis in GC patients (Fig. 1D). To validate the effect of ERS on SEMA3A expression in h-CAFs, we measured SEMA3A mRNA and protein levels in h-CAFs treated with the supernatant of ERS-induced human gastric cancer cells. RT‒qPCR results showed that the supernatant from ERS-induced gastric cancer cells significantly upregulated SEMA3A expression in h-CAFs (Fig. 1E). WB analysis confirmed the overexpression of SEMA3A protein in h-CAFs when co-cultured with ERS-induced gastric cancer cells (Fig. 1F). IHC and IF staining of GC tissues further validated the high expression of SEMA3A in h-CAFs within ERS-GC (Fig. 1G-H). To establish animal models of ERS-GC, C57BL/6J mice were subcutaneously injected with gastric cancer cells treated with TM to induce ERS. IF and WB analysis of m-CAFs from these mice confirmed high SEMA3A expression in the ERS-GC TME (Fig. 1I-J). Given that SEMA3A is a secretory protein known to interact with neuropilin-1 (NRP1) on Tregs [40], we co-cultured m-Tregs with ERS-m-CAFs treated with the supernatant of ERS-GC cells to explore the immunomodulatory role of SEMA3A in the ERS-GC TME. Additionally, we manipulated SEMA3A expression in ERS-m-CAFs (Supplementary Fig. 1D) and analyzed the proportions of IL-10⁺m-Tregs and TGF-β⁺m-Tregs using flow cytometry. The results demonstrated that SEMA3A knockdown in m-CAFs reduced the proportions of IL-10⁺ m-Tregs and TGF-β⁺ m-Tregs, while SEMA3A overexpression increased these proportions (Fig. 1K and Supplementary Fig. 1E). ELISA further confirmed that SEMA3A downregulation in m-CAFs decreased IL-10 and TGF-β levels in the TME, whereas SEMA3A overexpression enhanced their expression (Fig. 1L).

Fig. 1.

SEMA3A is highly expressed in CAFs of ERS-GC and promotes immunosuppression. A TIMER analysis of gastric cancer sequencing data in TCGA database showed the correlation between GRP78 and h-CAFs infiltration. B The volcano map (left panel) and heat map (right panel) of mass spectrometry sequencing results of h-CAFs stimulated by ERS- gastric cancer cells supernatant compared with h-CAFs stimulated by normal gastric cancer cells supernatant. C TIMER analysis showed the correlation between SEMA3A and h-CAFs infiltration in gastric cancer. D Survival analysis showed the correlation between SEMA3A expression and prognosis in gastric cancer. E RT-qPCR results showed the mRNA expression of SEMA3A in h-CAFs stimulated by the supernatant from different gastric cancer cells with or without ERS, n = 3, **p < 0.01. F WB showed the protein level of SEMA3A in h-CAFs stimulated by the supernatant form with different gastric cancer cells with or without ERS. G Immunohistochemical results showed the expression of SEMA3A in gastric cancer tissues. H Immunofluorescence showed the expression of SEMA3A in human gastric cancer tissue under ERS. I Immunofluorescence showed the expression of SEMA3A in m-CAFs of ERS gastric cancer animal models. J WB showed the protein level of SEMA3A in m-CAFs. K Flow cytometry detected the level of IL-10⁺ m-Tregs (up panel) and TGF-ꞵ⁺ m-Tregs (down panel) when co-cultured, altered the expression of SEMA3A, with m-Tregs, n = 3, **p < 0.01. L Expression level of IL-10 (up panel) and TGF-ꞵ (down panel) were detected by ELISA after co-culturing the m-Tregs with m-CAFs altered the expression of SEMA3A, n = 3, **p < 0.01

M⁶A modification stabilized the high expression of SEMA3A in CAFs in the gastric cancer microenvironment under ERS

To investigate whether ERS influences the immunosuppressive effects of CAFs, we compared the impact of the expression of SEMA3A between normal CAFs and ERS-induced CAFs. Flow cytometry analysis revealed that the proportions of IL-10⁺Tregs and TGF-β⁺Tregs were significantly higher in the presence of ERS-CAFs compared to non-ERS CAFs (Fig. 2A and Supplementary Fig. 2 A). It is worth noticed that the expression of SEMA3A was positively correlated with the proportions of IL-10⁺Tregs and TGF-β⁺Tregs (Fig. 2A and Supplementary Fig. 2 A). ELISA results confirmed that SEMA3A promoted the secretion of IL-10 and TGF-β in the tumor microenvironment (Fig. 2B). Based on the mass spectrometry sequencing results shown in Fig. 1B and Supplementary Table 1, we assessed the protein expression of SEMA3A and m⁶A methyltransferases in h-CAFs treated with supernatants from gastric cancer cells under different conditions. WB analysis demonstrated that the supernatant from ERS-induced gastric cancer cells significantly increased the expression of SEMA3A and METTL3 in h-CAFs (Fig. 2C). To investigate the role of METTL3 in m⁶A modification of SEMA3A mRNA, we altered METTL3 expression in m-CAFs and h-CAFs (Supplementary Fig. 2B-D) and performed m⁶A methylated RNA immunoprecipitation followed by quantitative PCR (MeRIP-qPCR). The results indicated that METTL3 mediated the m⁶A modification of SEMA3A mRNA (Fig. 2D). By jointly analysis of the data from the UCSC Genome Browser (University of California Santa Cruz, Genomics Institute, genome.ucsc.edu) and m⁶A-TSHub database (http://180.208.58.19/tshub/), we mapped the binding sites of METTL3 on SEMA3A mRNA and identified prominent peaks at specific motifs (Fig. 2E). The highest-scoring site within the coding sequence (CDS) was selected for mutagenesis. RNA pull-down experiments revealed the binding between METTL3 and SEMA3A mRNA (Fig. 2F). Mutation of the METTL3 binding motif within SEMA3A mRNA significantly reduced the m⁶A modification level, as confirmed by RIP-qPCR (Fig. 2G). To examine the effect of METTL3 on SEMA3A mRNA stability, h-CAFs were treated with actinomycin D to inhibit transcription. The results demonstrated that METTL3 significantly enhanced the stability of SEMA3A mRNA (Fig. 2H). Additionally, RNA pull-down experiments identified YTHDF1 as an m⁶A reader protein bound to SEMA3A mRNA in h-CAFs (Fig. 2I). Recent studies have shown that YTHDF1, a member of the YTH domain family, stabilizes m⁶A-modified mRNAs [41]. Consistent with this, we modulated YTHDF1 expression in h-CAFs (Supplementary Fig. 2E-F) and found that YTHDF1 positively regulated the stability of SEMA3A mRNA (Fig. 2J).

Fig. 2.

The high expression of SEMA3A in ERS-CAFs is regulated by m⁶A modification in gastric cancer. A Co-cultured m-Tregs with m-CAFs after changing SEMA3A expression, the content of IL-10⁺ m-Tregs (up panel) and TGF-ꞵ⁺ m-Tregs (down panel) were detected by flow cytometry, n = 3, * p < 0.05, ** p < 0.01. B ELISA results shown the expression of IL-10 (up panel) and TGF-ꞵ (down panel) in TME after altering the expression of SEMA3A in m-CAFs, n = 3, ** p < 0.01. C The expression of SEMA3A and m⁶A proteins were detected by WB in h-CAFs stimulated by HGC27 supernatant. D M⁶A-MeRIP-qPCR results showed the m⁶A methylation level of SEMA3A after altering the expression of METTL3 in h-CAFs, n = 3, ** p < 0.01. E UCSC Genome Browser showed the binding sites of SEMA3A mRNA and METTL3. F RNA pull-down determined the binding of METTL3 and SEMA3A mRNA. G RIP-qPCR results showed the relative mRNA expression (left panel) and m⁶A methylation level (right panel) of SEMA3A in h-CAFs after mut the binding sites of SEMA3A mRNA, n = 3, ** p < 0.01. H The RNA stability of SEMA3A in h-CAFs was affected by the expression of METTL3. I The results of RNA pull-down revealed the combination of SEMA3A mRNA and different m⁶A reader proteins in h-CAFs. J The RNA stability of SEMA3A in h-CAFs was affected by the expression of YTHDF1

ERS induces the expression of the lncRNA NEAT1 in CAFs in the gastric cancer microenvironment

To elucidate the mechanisms by which METTL3 regulates the m⁶A modification of SEMA3A mRNA in h-CAFs under ERS, we performed transcriptome sequencing analysis. The corresponding data have been deposited in the GEO database under accession number GSE290384. This analysis identified 2,810 upregulated genes and 2,260 downregulated genes in ERS-h-CAFs compared to control h-CAFs (Table 1, Fig. 3A). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed that the differentially expressed genes were significantly enriched in pathways related to immune regulation, protein expression, and ERS (Fig. 3B). Using the ENCORI database (rnasysu.com/encori), we predicted potential interactions between the differentially expressed long non-coding RNA (lncRNA) NEAT1 (Fig. 3A) and METTL3 (Fig. 3C). RNAInter v4.0 interaction analysis further supported a strong interaction between lncRNA NEAT1 and METTL3 (Fig. 3D). RT-qPCR results confirmed the upregulation of lncRNA NEAT1 in ERS-h-CAFs (Fig. 3E). Knockout of lncRNA NEAT1 in h-CAFs significantly suppressed the expression of METTL3 mRNA, whereas lncRNA NEAT1 overexpression have no influence on METTL3 mRNA (Fig. 3F). WB analysis demonstrated that lncRNA NEAT1 positively regulated the protein expression of METTL3, METTL14, and SEMA3A in h-CAFs (Fig. 3G). Potential binding sites between lncRNA NEAT1 and METTL3 were mapped using the UCSC Genome Browser (Fig. 3H). Moreover, the latest study revealed four sites on lncRNA NEAT1 along with two METTL3 binding regions [42]. Among these, site 4 previously linked to etoposide treatment resistance, whereas the functions of sites 2 and 3 remained uncharacterized. We therefore synthesized a mutant lncRNA NEAT1 by introducing mutations from site 2 to site 3 (Supplementary Fig. 3 A), and synthesizing mutated lncRNA NEAT1 for RNA pulldown followed by WB experiment of METTL3. RNA pull-down experiments confirmed a direct interaction between lncRNA NEAT1 and METTL3, but not between lncRNA NEAT1 and SEMA3A (Fig. 3I). In addition, after mutating the binding sites of METTL3 in lncRNA NEAT1, the protein level of METTL3 significantly decreased (Fig. 3J). These findings suggested that lncRNA NEAT1 exerted its regulatory effects of SEMA3A by bounding and modulating the abundance of METTL3.

Table 1.

Differential expressed lncRNAs in the ERS h-CAFs compared with normal h-CAFs

Gene name	Description	Trans type	log2 (Fold_change)	p-value	q-value	Regulation	Significant	Mean FPKM
SNHG32	small nucleolar RNA host gene 32	lncRNA	1.79	0.00	0.00	up	yes	62.23265815
MIR503HG	MIR503 host gene	lncRNA	1.56	0.00	0.00	up	yes	11.04838889
ENSG00000261270	novel transcript, sense intronic to HERPUD1	lncRNA	3.24	0.00	0.00	up	yes	10.91760383
MIR22HG	MIR22 host gene	lncRNA	0.96	0.00	0.00	up	yes	27.01439457
NEAT1	nuclear paraspeckle assembly transcript 1	lncRNA	0.65	0.00	0.00	up	yes	49.75206354
MEG3	maternally expressed 3	lncRNA	0.64	0.00	0.00	up	yes	20.45368964
ENSG00000278730	novel transcript	lncRNA	0.86	0.00	0.00	up	yes	10.58569183
ENSG00000287853	novel transcript	lncRNA	0.78	0.00	0.00	up	yes	23.11446167
PKD1P6-NPIPP1	PKD1P6-NPIPP1 readthrough	lncRNA	0.68	0.00	0.00	up	yes	12.29579292

Top 10 differential expressed lncRNAs in the ERS h-CAFs compared with normal h-CAFs (mean FPKM > 10, q value < 0.01, p value < 0.01)

Fig. 3.

The expression of METTL3 in ERS-CAFs is regulated by lncRNA NEAT1 in gastric cancer. A Volcanic map of RNA sequencing showed that lncRNA NEAT1 was highly expressed in h-CAFs under ERS. B The results of GO analysis of RNA-seq showed that differential expressed RNAs were involved in immune response in h-CAFs under ERS. C ENCORI database predicted the binding effect between lncRNA NEAT1 and METTL3. D The RNAInter v4.0 interaction network analysis showed the relationship between lncRNA NEAT1 and METTL3. E RT-qPCR detected the expression of lncRNA NEAT1 in h-CAFs after stimulated by the supernatant of HGC27 with or without ERS, n = 3, *** p < 0.001. F RT-qPCR results showed the efficiency of knockout and overexpression of lncRNA NEAT1, along with the mRNA expression of METTL3 in h-CAFs, n = 3, *** p < 0.001. G The protein level of SEMA3A, METTL3 and METTL14 were detected after altered the expression of lncRNA NEAT1 in h-CAFs. H UCSC Genome Browser showed the binding sites of lncRNA NEAT1 and METTL3. I RNA pull-down determined the binding of lncRNA NEAT1 and METTL3. J WB detection of the protein level of METTL3 and SEMA3A in h-CAFs after mut the METTL3-binding sites in lncRNA NEAT1. K Co-IP assays were performed to detect the ubiquitination level and protein expression of METTL3 in h-CAFs across different experimental groups. L WB analysis of METTL3 protein levels in cells treated with cycloheximide (CHX, 100 μg/mL). M RIP-qPCR analysis of RNA precipitated by anti-METTL3 antibody or control IgG. The enrichment of SEMA3A mRNA and NEAT1 in the immunoprecipitates was quantified. N M⁶A-MeRIP-qPCR results showed the m⁶A methylation level of SEMA3A mRNA after altering the expression of lncRNA NEAT1 in h-CAFs, n = 3, ** p < 0.01. O RIP-qPCR results showed the relative mRNA expression of SEMA3A (left panel) and m⁶A methylation level of SEMA3A (right panel) in h-CAFs after mut the binding sites of METTL3, n = 3, ** p < 0.01. P The RNA stability of SEMA3A in h-CAFs was affected by the expression of lncRNA NEAT1 and METTL3

Co‑IP assays demonstrated that lncRNA NEAT1 knockdown significantly enhanced the ubiquitination of METTL3 in h-CAFs, while the proteasome inhibitor MG132 reversed its degradation (Fig. 3K). Consistent with its role in inhibiting METTL3 ubiquitination, lncRNA NEAT1 knockout shortened METTL3 protein half-life, providing functional evidence that lncRNA NEAT1 post-translationally stabilizes METTL3 (Fig. 3L). Supplementary Fig. 3B presented the potential E3 ubiquitin ligases predicted by Ubibrowser (http://ubibrowser.bio-it.cn/ubibrowser_v3/). These results indicated that lncRNA NEAT1 primarily stabilized METTL3 protein by inhibiting the ubiquitin‑proteasome pathway. RIP‑qPCR results showed that lncRNA NEAT1 depletion markedly reduced the binding of METTL3 to its downstream target SEMA3A mRNA, and the enrichment of lncRNA NEAT1 itself was abolished (Fig. 3M). These findings suggest that lncRNA NEAT1 may act as a molecular scaffold to facilitate the formation and stabilization of the METTL3–SEMA3A mRNA ternary complex. To further investigate the role of lncRNA NEAT1 in m⁶A modification of SEMA3A mRNA, we performed MeRIP-qPCR in h-CAFs with lncRNA NEAT1 knockout or overexpression. Additionally, METTL3 was overexpressed in lncRNA NEAT1-ko h-CAFs to assess its compensatory effects. The results showed that lncRNA NEAT1 knockout reduced the m⁶A modification level of SEMA3A mRNA, while lncRNA NEAT1 overexpression increased it (Fig. 3N). Meanwhile, the overexpression of METTL3 in lncRNA NEAT1-knockout h-CAFs partially restored the m⁶A modification level of SEMA3A mRNA (Fig. 3N). Moreover, mutation of the lncRNA NEAT1-METTL3 interaction sites decreased the binding of SEMA3A mRNA to METTL3, as demonstrated by RIP-qPCR (Fig. 3O). Such mutation also reduced the m⁶A modification level of SEMA3A mRNA (Fig. 3O). The above results suggested that lncRNA NEAT1 bound to and upregulated the protein level of METTL3, and then the METTL3 protein specifically increased the m⁶A modification level of its substrate, SEMA3A mRNA. RNA stability assays revealed that lncRNA NEAT1 knockout decreased the stability of SEMA3A mRNA, the overexpression of METTL3 in lncRNA NEAT1-knockout h-CAFs restored SEMA3A mRNA stability (Fig. 3P). In addition, in order to determine the subcellular site of the lncRNA NEAT1-METTL3 interaction, RNA-ISH/IF co-localization assays were performed. Results turned out that lncRNA NEAT1 and METTL3 showed clear nuclear co-localization in h-CAFs (Supplementary Fig. 3C).

ERS stimulates the production of gastric cancer exosomes and regulates the expression of the lncRNA NEAT1 in CAFs

To investigate the mechanisms underlying the high expression of lncRNA NEAT1 in CAFs under ERS as shown in Fig. 3A, we analyzed scRNA-seq data from the gastric cancer dataset GSE231539. Figure 4A illustrated the cellular composition of nontumor and tumor cell populations in the GSE231539 dataset. The expression of canonical cell type-specific markers was visualized using violin plots (Fig. 4B), while the t-distributed stochastic neighbor embedding (t-SNE) scatter plot highlighted the heterogeneity and feature similarities among epithelial cells and cancer cells (Fig. 4C, left panel). Violin plots further demonstrated higher expression of GRP78 (HSPA5) in tumor-derived epithelial/cancer cell populations compared to nontumor population (Fig. 4C, right panel). Further analysis revealed a positive correlation between GRP78 (HSPA5) expression and lncRNA NEAT1 expression in GRP78⁺ tumor cells (Fig. 4D). Moreover, the expression of SEMA3A in ERS-CAFs in the tumor microenvironment was found to be particularly prominent among the population of fibroblasts with positive SEMA3A expression (Fig. 4E). To further delineate the expression patterns of SEMA3A and lncRNA NEAT1 across cell types in the ERS‑enriched gastric cancer tumor microenvironment, we performed an integrative analysis of the single‑cell transcriptomic dataset GSE183904 and the spatial transcriptomic dataset GSE251950. Cell‑type signature profiles were trained from the paired single‑cell data and applied to deconvolute the spatial data, enabling estimation of the absolute abundance of each cell type. By constructing both intra‑cellular and inter‑cellular views, we quantitatively assessed the contributions of intrinsic cellular composition and local microenvironmental interactions to the observed molecular phenotypes. The analysis revealed significantly elevated expression of SEMA3A, lncRNA NEAT1, METTL3, and YTHDF1 in high-ERS state (HERS)‑CAFs and tumor cells (Fig. 4F-G and Supplementary Fig. 3D-F). Spatial mapping further demonstrated co‑localization of HERS‑CAFs with tumor cells, T cells, and Tregs (Fig. 4H and Supplementary Fig. 3G-I). Correlation analysis using data from the TCGA database also confirmed a positive relationship between GRP78 and lncRNA NEAT1 expression in human gastric cancer tissues (Fig. 4I). Additionally, survival analysis indicated that the high expression of GRP78 (HSPA5) and lncRNA NEAT1 were both associated with poor prognosis in gastric cancer patients (Supplementary Fig. 3J-K). Analysis of immune cell infiltration in the TME revealed that lncRNA NEAT1 expression was negatively correlated with the proportions of CD8⁺ T cells (Fig. 4J, left panel) and natural killer (NK) cells (Fig. 4J, middle panel), but positively correlated with Tregs infiltration (Fig. 4J, right panel). These findings suggest that lncRNA NEAT1 expression in h-CAFs may be influenced by lncRNA NEAT1 expression in gastric cancer tumor cells. Using transmission electron microscopy (TEM), we observed that ERS enhanced the secretion of exosomes from gastric cancer cells into the TME (Supplementary Fig. 3L). Detection of exosome- marker proteins and negative control proteins further confirmed that ERS-stimulated gastric cancer cells secreted a higher quantity of exosomes (Supplementary Fig. 3M). Based on these observations, we hypothesized that ERS promotes the secretion of lncRNA NEAT1-enriched exosomes from gastric cancer cells into the TME, which may subsequently modulate lncRNA NEAT1 expression in CAFs.

Fig. 4.

LncRNA NEAT1 is secreted by exosomes from ERS-GC to CAFs and promoted the immunosuppression. A Umap of the cell clusters segregated by non-tumor and tumor from human gastric cancer scRNA-seq data GSE231539. B Violin plot depicted the expression of canonical cell markers in each group of GSE231539. C TSNE plot highlighted the epithelial/cancer cells in GSE231539 (left panel), violin plot depicted the expression of GRP78 (HSPA5) in epithelial/cancer cell population segregated by non-tumor and tumor (right panel). D Positive correlation was found between GRP78 (HSPA5) and lncRNA NEAT1 in the GRP78⁺ cell group of GSE231539. E Comparison of ERS‑CAFs activity in SEMA3A‑positive fibroblasts between tumor and non‑tumor in GSE231539. F Spatial associations between the expression levels of lncRNA NEAT1, YTHDF1, METTL3, and SEMA3A with different cell types in gastric cancer microenvironment of the tumor tissue section in GSE251950 combined with GSE183904. G Spatial transcriptomics analyses of expression level of lncRNA NEAT1, YTHDF1, METTL3, and SEMA3A within the gastric cancer microenvironment in GSE251950 combined with GSE183904. H The importance of cell co-localization was quantified across different neighborhood scales (para: extended to two neighboring spots). Red boxes indicated the spatial co-localization levels between tregs and HERS-CAFs (left panel). Network graphs depicted the co-localization relationships between tumor cells and CAFs with T-cell subpopulations in gastric cancer spatial transcriptomics sections under para views (right panel), edges connecting cell types represented strong co-localization, defined by an importance score > 0.1. I Analysis of TCGA data revealed positive correlation between GRP78 (HSPA5) and lncRNA NEAT1 in gastric cancer tissue. J TCGA data analysis showed the expression of lncRNA NEAT1 correlated to the proportion of CD8⁺T cell (left panel), NK cell (middle panel) and Tregs (right panel) in gastric tumor tissues. K The volcano map of RNA-seq showed differentially expressed lncRNAs in ERS exosome of gastric cancer compared with normal exosome in gastric cancer. L LncRNA NEAT1 was found to be an up-regulated lncRNA in the exosomes of ERS-GC through RNA-seq. M RT-qPCR results revealed the expression of lncRNA NEAT1 in supernatant and exosome of gastric cancer cells, n = 3, *** p < 0.001. N RT-qPCR detected the expression of lncRNA NEAT1 in h-CAFs co-cultured with exosomes from gastric cancer cells (left panel) and m-CAFs co-cultured with exosomes from MFC cells (right panel), n = 3, *** p < 0.001. O Flow cytometry results showed the level of IL-10⁺ m-Tregs (left panel) and TGF-ꞵ⁺ m-Tregs (right panel) after co-culture m-Tregs with m-CAFs, which were stimulated by the supernatant of ERS-MFC, n = 3, ** p < 0.01. P The level of IL-10 (left panel) and TGF-ꞵ (right panel) were detected by ELISA in ERS-MFC models, n = 3, ** p < 0.01. Q WB detected the protein level of SEMA3A, METTL3 and METTL14 in h-CAFs after co-cultured with normal exosomes and ERS exosomes of GC. R The protein level of SEMA3A in each group was detected by WB after co-cultured h-CAFs with the supernatant of ordinary gastric cancer cells or the supernatant of ERS gastric cancer cells or the supernatant of gastric cancer cells added with exosome synthesis inhibitor GW4869

To validate this hypothesis, we performed transcriptome sequencing of exosomes derived from gastric cancer cells under ERS versus normal conditions. The corresponding data have been deposited in the GEO database under accession number GSE297015. A total of 192 lncRNAs were found to be significantly upregulated in exosomes from ERS-induced GC cells, including lncRNA NEAT1 (Fig. 4K-L, Table 2). RT-qPCR analysis confirmed that lncRNA NEAT1 was highly expressed in both the supernatant of ERS-induced GC cells (Fig. 4M, left panel) and the exosomes isolated from the supernatant of ERS-induced GC cells (Supplementary Fig. 3N and Fig. 4M, right panel). To assess the transfer of lncRNA NEAT1-enriched exosomes from gastric cancer cells to CAFs, we stimulated h-CAFs with exosomes from GC cells under ERS condition and m-CAFs with exosomes from different states of MFC models. LncRNA NEAT1 expression was significantly increased in h-CAFs co-cultured with exosomes from GC cells under ERS condition, as well as in m-CAFs co-cultured with exosomes from ERS-GC tumor (Fig. 4N). In contrast, no significant change in lncRNA NEAT1 expression was observed in m-Tregs under the same conditions (Supplementary Fig. 3 O). To evaluate the functional consequences of the lncRNA NEAT1-enriched exosomes in ERS-GC, we co-cultured m-Tregs with m-CAFs under exosomes of different states of the MFC models. Flow cytometry analysis showed that exosomes from ERS-MFC models promoted the generation of IL-10⁺ m-Tregs and TGF-β⁺ m-Tregs (Fig. 4O and Supplementary Fig. 3P). ELISA results further confirmed that exosomes from ERS-MFC models increased the levels of IL-10 and TGF-β in the TME (Fig. 4P). WB analysis revealed that exosomes from ERS-HGC27 cells upregulated the expression of METTL3 and SEMA3A in h-CAFs (Fig. 4Q). Moreover, the SN from ERS-HGC27 cells significantly upregulated METTL3 and SEMA3A expression in h-CAFs, whereas was largely abolished when exosome release from HGC27 cells was inhibited by GW4869 (Fig. 4R).

Table 2.

Differential expressed lncRNAs in the exosomes of ERS gastric cancer

ncRNA	RNA type	Gene	Gene type	log2(Fold_change)	p-value
NR_131224.1	ncRNA	H19	lncRNA	−1.612826224	0.025390628
NR_002728.3	ncRNA	KCNQ1OT1	lncRNA	−2.33471332	0.003539384
XR_001754939.2	ncRNA	LOC107987294	lncRNA	−1.959754081	0.015638697
NR_108049.1	ncRNA	CCAT1	lncRNA	−2.648448688	0.003989127
XR_001742582.2	ncRNA	LOC101929200	lncRNA	−8.43940817	2.69E-05
XR_001742612.1	ncRNA	MARCHF11-AS1	lncRNA	−2.859797655	0.004046966
NR_028272.1	ncRNA	NEAT1	lncRNA	1.479974908	0.043362154
XR_948625.2	ncRNA	LOC105379100	lncRNA	−8.297819016	4.69E-05
XR_946873.2	ncRNA	LOC105378604	lncRNA	−3.310583259	0.003419546

Top 10 differential expressed lncRNAs in the exosomes of ERS GC. According to the expression value of transcript, p value < 0.05

LncRNA NEAT1 in CAFs regulates METTL3 and promotes the SEMA3A-mediated tumor immunosuppressive effect

To elucidate the relationship between lncRNA NEAT1-mediated regulation of METTL3 expression and its role in promoting SEMA3A expression in ERS-CAFs, we overexpressed METTL3 in m-CAFs with lncRNA NEAT1 knockout. RT-qPCR results demonstrated that lncRNA NEAT1 knockout significantly reduced SEMA3A mRNA expression, while METTL3 overexpression in lncRNA NEAT1-knockout m-CAFs partially restored SEMA3A mRNA expression (Fig. 5A). Western blot analysis further confirmed that SEMA3A protein levels were lowest in the lncRNA NEAT1-knockout group, while had highest expression in the METTL3-overexpression group and moderately increased in the lncRNA NEAT1-knockout combined with METTL3-overexpression group (Fig. 5B). To investigate the impact of lncRNA NEAT1 on tumor immune regulation in gastric cancer, we compared the proportions of IL-10⁺ m-Tregs and TGF-β⁺ m-Tregs when co-cultured the m-Tregs with the lncRNA NEAT1-knockout m-CAFs, METTL3-overexpression m-CAFs, and lncRNA NEAT1-knockout combined with METTL3-overexpression m-CAFs. Flow cytometry analysis revealed that lncRNA NEAT1 knockout suppressed the generation of IL-10⁺ m-Tregs and TGF-β⁺ m-Tregs, whereas METTL3 overexpression promoted their production (Fig. 5C). Meanwhile, the combined intervention of lncRNA NEAT1 knockout and METTL3 overexpression attenuated these effects (Fig. 5C). ELISA results corroborated the finding of lncRNA NEAT1 knockout reduced IL-10 and TGF-β levels in co-cultured medium and METTL3 overexpression increased them, while the effects of METTL3 overexpression on IL-10 and TGF-β expression were partially restored in the lncRNA NEAT1-knockout background (Fig. 5D). Furthermore, we examined the effects of modulating METTL3 and SEMA3A expression in m-CAFs on IL-10⁺ m-Tregs, TGF-β⁺ m-Tregs. Flow cytometry analysis indicated that METTL3 knockdown in m-CAFs significantly decreased the proportions of IL-10⁺ m-Tregs and TGF-β⁺ m-Tregs, while SEMA3A overexpression increased these proportions (Fig. 5E). ELISA results demonstrated that reduced METTL3 expression in m-CAFs suppressed IL-10 and TGF-β levels in the co-culture medium, whereas SEMA3A overexpression enhanced their secretion (Fig. 5F).

Fig. 5.

LncRNA NEAT1 promotes the immunosuppressive function of CAFs by regulating the expression of SEMA3A through METTL3. A The expression of SEMA3A was detected by RT-qPCR in m-CAFs under ERS, n = 3, ** p < 0.01. B The protein level of METTL3 and SEMA3A was detected by WB in m-CAFs under ERS. C The level of Foxp3⁺ m-Tregs (up panel), IL-10⁺ m-Tregs (middle panel) and TGF-ꞵ⁺ m-Tregs (down panel) were detected by flow cytometry after the expression of METTL3 was rescued in m-CAFs and co-cultured with m-Tregs, n = 3, ** p < 0.01. D After the expression of METTL3 was restored in m-CAFs, the expression level of IL-10 (right panel) and TGF-ꞵ (left panel) were detected by ELISA in the TME, n = 3, ** p < 0.01. E The level of Foxp3⁺ m-Tregs (up panel), IL-10⁺ m-Tregs (middle panel) and TGF-ꞵ⁺ m-Tregs (down panel) were detected by flow cytometry after the expression of SEMA3A was rescued in m-CAFs and co-cultured with m-Tregs, n = 3, ** p < 0.01. F After the expression of SEMA3A was restored in m-CAFs, the expression level of IL-10 (up panel) and TGF-ꞵ (down panel) were detected by ELISA in the TME, n = 3, ** p < 0.01

The combination of siNEAT1@Lip-cRGD and anti-PD1 shows more effective than anti-PD1 regarding to the treatment of gastric cancer in vivo

In order to explore the anti-tumor effect of the suppression of lncRNA NEAT1, we developed a cyclic Arg-Gly-Asp (cRGD)-modified cationic liposome for the targeted delivery of siNEAT1 to gastric cancer cells (Fig. 6A). The cRGD peptide is known to actively target integrin-αvβ3 receptors, which are abundantly expressed on both gastric cancer cells and CAFs [43–46] (Supplementary Fig. 4A). As shown in Supplementary Fig. 4B-C, siNEAT1@Lip-cRGD exhibited a uniform spherical morphology with an average particle size of 155.5 nm. The size and polydispersity index (PDI) remained stable over 7 days (Supplementary Fig. 4D), indicating its potential suitability for tumor therapy. To evaluate the lysosomal escape capability of siNEAT1@Lip-cRGD, siNEAT1 was labeled with Cy5.5, and lysosomes were stained with LysoTracker Green. Fluorescence imaging revealed significant colocalization of LysoTracker Green and Cy5.5 (Fig. 6B), confirming that Cy5.5-siNEAT1@Lip-cRGD was internalized via the endocytic pathway. Prolonged incubation showed red fluorescence outside lysosomes, indicating effective lysosomal escape of Cy5.5-siNEAT1@Lip-cRGD. Notably, MFC cells treated with Cy5.5-siNEAT1@Lip-cRGD exhibited stronger fluorescence intensity compared to those treated with Cy5.5-siNEAT1@Lip (Supplementary Fig. 4E). This enhanced cellular uptake was blocked by pre-treating cells with free cRGD peptide (Fig. 6C-D), demonstrating the active targeting capability of Cy5.5-siNEAT1@Lip-cRGD and its improved intracellular delivery efficiency.

Fig. 6.

The combination therapy of siNEAT1@Lip-cRGD and anti-PD1 appears to be more efficient than PD1 monoclonal antibody therapy in gastric cancer. A Schematic drawing of the preparation of siNEAT1@Lip-cRGD. B CLSM images of MFC cells incubated with siNEAT1@Lip-cRGD for various hours. C CLSM images of MFC cells incubated with siNEAT1@Lip-cRGD or cRGD + siNEAT1@Lip-cRGD for 12 h. D The fluorescence quantitative analysis of MFC cells incubated with siNEAT1@Lip-cRGD or cRGD + siNEAT1@Lip-cRGD for 12 h, n = 3, ** p < 0.01. E Fluorescence images of major organs extracted from MFC tumor-bearing mice after treatment with Cy5.5, siNEAT1@Lip or siNEAT1@Lip-cRGD. F The average radiant efficiency of major organs extracted from MFC tumor-bearing mice after treatment with Cy5.5, siNEAT1@Lip or siNEAT1@Lip-cRGD, n = 3, ** p < 0.05. G Tumor volumes of MFC tumor-bearing mice after treatment with PBS, anti-PD1, siNEAT1@Lip-cRGD and siNEAT1@Lip-cRGD + anti-PD1, n = 5, ** p < 0.05, *** p < 0.001. H Tumor weights of MFC tumor-bearing mice after treatment with PBS, anti-PD1, siNEAT1@Lip-cRGD and siNEAT1@Lip-cRGD + anti-PD1, n = 5, *** p < 0.001

To validate the tumor-targeting ability of Cy5.5-siNEAT1@Lip-cRGD in vivo, MFC tumor-bearing mice were intravenously injected with Cy5.5-siNEAT1@Lip-cRGD. After 24 h, major organs were harvested for ex vivo fluorescence imaging. As shown in Fig. 6E, tumor tissues treated with Cy5.5-siNEAT1@Lip exhibited stronger fluorescence than those treated with free Cy5.5, likely due to the enhanced permeability and retention (EPR) effect. Notably, tumor tissues treated with Cy5.5-siNEAT1@Lip-cRGD displayed significantly higher fluorescence intensity, with an average radiant efficiency 1.67-fold greater than that of Cy5.5-siNEAT1@Lip-treated tumors (Fig. 6F), confirming the active targeting capability of Cy5.5-siNEAT1@Lip-cRGD in vivo. Next, the antitumor efficacy of siNEAT1@Lip-cRGD was evaluated in MFC tumor-bearing mice. Mice were randomly divided into four groups and treated via tail vein injection. Tumor volume increased rapidly in the PBS control group, while anti-PD1 monotherapy showed no significant tumor growth inhibition. In contrast, siNEAT1@Lip-cRGD alone moderately delayed tumor progression. Remarkably, the combination of siNEAT1@Lip-cRGD and anti-PD1 demonstrated the most potent tumor growth suppression, highlighting their synergistic antitumor effect (Fig. 6G). Tumor size and weight measurements further supported these findings, with the combination treatment group showing the most significant reduction in tumor burden (Fig. 6H). Additionally, body weight measurements (Supplementary Fig. 4 F) and histological analysis of major organs (Supplementary Fig. 4G) revealed no significant toxicity, indicating the favorable safety profile of siNEAT1@Lip-cRGD.

To assess the activation of antitumor immunity following the combined treatment of Cy5.5-siNEAT1@Lip-cRGD and anti-PD1 in MFC tumor-bearing mice, flow cytometry analysis was performed to quantify the proportion of CD8⁺ T cells in the spleen, lymph nodes, and tumor tissues (Fig. 7A). The results demonstrated that the combination of Cy5.5-siNEAT1@Lip-cRGD and anti-PD1 significantly increased the infiltration of activated CD8⁺ T cells while reducing the population of Tregs within the TME of MFC tumor-bearing mice (Fig. 7B). Furthermore, the combined treatment markedly suppressed the expression of immunosuppressive cytokines IL-10 and TGF-β (Fig. 7C-D). To further evaluate the therapeutic effects, tumor tissues were fixed and subjected to HE staining. As shown in Fig. 7E, the combined treatment of siNEAT1@Lip-cRGD and anti-PD1 induced the highest level of tumor cell apoptosis or necrosis, characterized by the absence of nuclei in tumor cells, compared to other treatment groups.

Fig. 7.

The combination therapy of siNEAT1@Lip-cRGD and anti-PD1 reverses the immunosuppressive microenvironment of gastric cancer. A The percentage of CD3⁺CD8⁺ cells in spleen (up panel), lymph node (middle panel) and tumor tissue (down panel) after treatment with PBS, anti-PD1, siNEAT1@Lip-cRGD and siNEAT1@Lip-cRGD + anti-PD1, n = 3, * p < 0.05, ** p < 0.01. B The percentage of CD3⁺CD4⁺ Foxp3⁺ cells (up panel), CD8⁺GZMB⁺ T cells (middle panel) and CD8⁺IFN-γ⁺ T cells (down panel) in tumor tissues after treatment with PBS, anti-PD1, siNEAT1@Lip-cRGD and siNEAT1@Lip-cRGD + anti-PD1, n = 3, ** p < 0.01. C The expressions of IL-10 and in tumor tissues after treatment with PBS, anti-PD1, siNEAT1@Lip-cRGD and siNEAT1@Lip-cRGD + anti-PD1, n = 3, * p < 0.05. D The expressions of TGF-β in tumor tissues after treatment with PBS, anti-PD1, siNEAT1@Lip-cRGD and siNEAT1@Lip-cRGD + anti-PD1, n = 3, * p < 0.05. E HE staining of tumor tissues extracted from MFC tumor-bearing mice after treatment with PBS, anti-PD1, siNEAT1@Lip-cRGD and siNEAT1@Lip-cRGD + anti-PD1

Discussion

Although immunotherapy has transformed the treatment landscape for solid tumors, durable and effective responses remain limited in patients with gastric cancer. The heterogeneity of the tumor microenvironment drives immune evasion, contributing to resistance to conventional therapies and immunotherapies [47]. Elucidating the molecular mechanisms underlying this resistance and targeting the immunosuppressive TME are critical for enhancing the efficacy of immunotherapy and improving patient outcomes. Stress conditions such as oncogene activation, nutrient deprivation, and hypoxia can disrupt endoplasmic reticulum homeostasis, triggering the unfolded protein response (UPR). This response regulates cancer progression through autonomous and nonautonomous cellular mechanisms, including stromal reprogramming, immune evasion, angiogenesis, and invasion [48]. Studies have demonstrated that hypoxia promotes the release of exosomes from breast cancer cells into the TME, enhancing tumor cell survival and invasive potential [49]. Components of the TME not only suppress antitumor immune responses but also facilitate tumor progression [50]. Tumor-associated macrophages (TAMs) play pivotal roles in tumor growth, angiogenesis, and immune regulation by participating in tumor initiation, progression, invasion, and metastasis. For instance, TAM-derived CXCL8 can inhibit CD8⁺ T cell function by upregulating PD-L1 expression on macrophages in GC [51]. Myeloid-derived suppressor cells (MDSCs) promote tumor progression through multiple mechanisms, including the upregulation of CXCL1 in GC cells and the induction of immunosuppression via S100A8/A9 [52]. Neutrophils have also been implicated in disease progression and are negatively correlated with patient survival [53]. CAFs, the most abundant stromal cells in the TME, contribute to tumor invasion and drug resistance in GC by fostering chronic inflammation and secreting immunomodulatory cytokines [54]. Therefore, investigating the interactions between cellular and molecular components within the TME of GC is essential for developing novel therapeutic strategies and improving patient survival.

The broad mechanism of action of immune checkpoint blockade therapy involves reversing T-cell inhibition by targeting inhibitory receptors (IRs) [55]. Blockade of IRs not only enhances T-cell activity but also modulates the function of immunosuppressive cells within the TME, particularly Tregs [56]. In recent years, neuropilin-1 (NRP1) has garnered significant attention in the field of immuno-oncology due to its dual functional roles [57]. Initially identified as a marker for Tregs, NRP1 has since been shown to play a critical role in maintaining the stability of Tregs within the TME and significantly suppressing the antitumor activity of CD8⁺ T cells [58]. NRP1 functions as a coreceptor for secreted semaphorins, including semaphorin 4A (Sema4A) [59] and semaphorin 3 A (Sema3A) [40]. Studies have demonstrated that NRP1 enhances immunosuppression by facilitating the recruitment of Tregs to tumor sites through its interaction with vascular endothelial growth factor (VEGF) and by maintaining the stability of tumor-specific Tregs via Sema3A signaling [60]. Consequently, Sema3A has emerged as a key secretory protein regulating Treg-mediated immune responses within the TME [61]. Through a series of experiments in this study, we found that lncRNA NEAT1 promoted m⁶A modification of SEMA3A through two stages: firstly, lncRNA NEAT1 bound to and upregulated the protein level of METTL3; secondly, the METTL3 enzyme enhanced its methylation activity, thereby specifically increasing the m⁶A modification level of its substrate SEMA3A mRNA.

Furthermore, through a series of experiments, our study systematically elucidates the dual molecular mechanisms by which lncRNA NEAT1 regulates METTL3. On one hand, immunoprecipitation assays showed that lncRNA NEAT1 knockout significantly enhanced METTL3 polyubiquitination, while proteasome inhibition rescued METTL3 degradation upon lncRNA NEAT1 loss. This suggested that lncRNA NEAT1 might interfere with E3 ligase interaction or recruit deubiquitinases to inhibit METTL3 degradation, providing direct evidence for lncRNA-mediated regulation of epigenetic enzymes via protein stability. On the other hand, RIP-qPCR results indicated a potential scaffolding role for lncRNA NEAT1. Loss of lncRNA NEAT1 reduced SEMA3A mRNA enrichment in METTL3 immunoprecipitates and abolished lncRNA NEAT1 self-enrichment, suggesting a lncRNA NEAT1/METTL3/SEMA3A mRNA ternary complex where lncRNA NEAT1 stabilized METTL3-substrate interaction. This dual mechanism explained how lncRNA NEAT1 deficiency concurrently reduced METTL3 levels and impaired its m⁶A modification activity on SEMA3A. However, this study has clear limitations that need to be addressed in future work. First, due to technical constraints, we were unable to directly assess the effect of lncRNA NEAT1 on the translation efficiency of METTL3 mRNA via polysome profiling. Although protein stability regulation has been identified as the primary mechanism, it cannot be completely ruled out that lncRNA NEAT1 may also fine-tune METTL3 translation rates by recruiting translation initiation factors or ribosomes. Second, the specific executor(s) responsible for lncRNA NEAT1-mediated stabilization of METTL3 protein remain unclear. Future studies should identify E3 ligases or deubiquitinases interacting with the lncRNA NEAT1/METTL3 complex through mass spectrometry screening and validate their functions. Third, the precise structural basis underlying the scaffold role of lncRNA NEAT1 in promoting the METTL3-SEMA3A mRNA interaction remains to be elucidated. Techniques such as truncation mutagenesis and CLIP-seq could be employed to map the key regions on lncRNA NEAT1 that bind to METTL3 and SEMA3A mRNA.

In recent years, the role of exosomes carrying lncRNAs in the TME has garnered significant attention [62]. Exosomes are key mediators of intercellular communication, playing critical roles in tumor immunity [63], angiogenesis [64], metastasis [65] and drug resistance [66] through diverse mechanisms. Accumulating evidence highlights the importance of exosomes in regulating the TME [67], with lncRNAs carried by exosomes serving as crucial epigenetic regulators in these processes [68, 69]. In this study, we elucidated a novel immunosuppressive mechanism involving tumor cells, CAFs and Tregs, mediated by exosome-transported lncRNAs in the TME of gastric cancer under ERS. We demonstrated that ERS enhances exosome secretion by gastric cancer cells. These exosomes, enriched with lncRNA NEAT1, were shown to upregulate METTL3 expression in CAFs. Elevated METTL3 levels in CAFs promoted m⁶A methylation of SEMA3A mRNA, enhancing its stability through the m⁶A reader protein YTHDF1. The resulting overexpression of SEMA3A in CAFs amplified the immunosuppressive function of Tregs, contributing to poor clinical outcomes in GC patients. While we have identified functionally critical sites on both lncRNA NEAT1 (sites 2–3) and SEMA3A (top-scoring CDS site), the remaining uncharacterized sites likely constitute a broader regulatory network. Our present work highlighted the necessity and importance of further investigating other modification sites and their potential crosstalk. Their individual or combinatorial roles in fine-tuning RNA–protein interactions, structural dynamics, and downstream signaling remain unexplored. Therefore, elucidating the interplay between different sites will not only reveal the full regulatory potential of m⁶A but also provide a more robust foundation for developing multi-target epitranscriptomic interventions in cancer.

While our study utilized TM to model ERS, we acknowledged its limitations in fully recapitulating the chronic, heterogeneous stress in clinical gastric cancer. Nevertheless, the clinical relevance of the identified lncRNA NEAT1/METTL3/SEMA3A axis was supported by multiple lines of evidence: TCGA analysis and IHC confirmed positive correlations between the ERS marker GRP78 and both CAF markers and SEMA3A expression. Data of scRNA-seq and ST-seq further linked GRP78⁺ CAFs with positive expression of lncRNA NEAT1, SEMA3A and METTL3 in TME. These findings suggested the mechanistic axis was operative in human tumors. Future studies employing physiological inducers like hypoxia or nutrient deprivation were proposed to further validate and generalize these conclusions. Building on these findings, we developed a targeted therapeutic strategy using the cationic liposome siNEAT1@Lip-cRGD to deliver siRNA against lncRNA NEAT1 in conjunction with anti-PD1 therapy for MFC tumors. Leveraging the high expression of integrin αvβ3 receptors on gastric cancer cells and CAFs, the formulation could achieve efficient cellular internalization and significant downregulation of lncRNA NEAT1 in both cell populations. Consequently, this combined approach demonstrated potent efficacy, underscoring lncRNA NEAT1 as a promising target for enhancing immunotherapy in GC patients. The combination therapy elicited a robust antitumor immune response and exhibited significant synergistic antitumor effects. The ERS–lncRNA NEAT1–METTL3–SEMA3A axis, supported by our data, may represent a promising biomarker for predicting an immunosuppressive tumor microenvironment and potential resistance to immune checkpoint inhibitors in gastric cancer.

Conclusion

Our study demonstrated that endoplasmic reticulum stress in gastric cancer enhances exosome secretion by gastric cancer cells, leading to the elevated expression of lncRNA NEAT1 in cancer-associated fibroblasts. This lncRNA NEAT1 upregulation stabilized SEMA3A mRNA expression through m⁶A methylation, thereby amplifying the immunosuppressive function of regulatory T cells. A graphical abstract summarizing these findings is provided in Fig. 8. Our study discovered a novel mechanism of lncRNA NEAT1 regulating the expression of METTL3 and stabling the expression of SEMA3A through m⁶A modification, by which CAFs suppressed the immune response in gastric cancer.

Fig. 8.

The graphic abstract of this work. LncRNA NEAT1 regulated the expression of METTL3 and promoted the m⁶A modification of SEMA3A in CAFs, by which suppressed the immune response in gastric cancer under endoplasmic reticulum stress

Abbreviations

GC

Gastric cancer

TME

Tumor microenvironment

ERS

Endoplasmic reticulum stress

CAFs

Cancer-associated fibroblasts

h-CAFs

Human gastric cancer-associated fibroblasts

DEGs

Differentially expressed genes

RT‒qPCR

Real-time quantitative polymerase chain reaction

m-CAFs

Mouse-derived primary cancer-associated fibroblasts

m-Tregs

Mouse-derived primary regulatory T cells

Tregs

Regulatory T cells

GRP78

78 KDa glucose-regulated protein

WB

Western-blotting

IHC

Immunohistochemistry

RIP-qPCR

RNA Immunoprecipitation (RIP)-qPCR Assay

IF

Immunofluorescence

TAMs

Tumor-associated macrophages

MDSCs

Myeloid-derived suppressor cells

TM

Tunicamycin

IRs

Inhibitory receptors

NRP1

Neuropilin-1

MeRIP-qPCR

M⁶A methylated RNA immunoprecipitation followed by quantitative PCR

ELISA

Enzyme-linked immunosorbent assay

RNA-ISH

RNA in situ hybridization

GEO

Gene Expression Omnibus

KEGG

Kyoto Encyclopedia of Genes and Genomes

lncRNA

Long non-coding RNA

scRNA-seq

Single-cell RNA-sequencing

TCGA

The Cancer Genome Atlas

ST-seq

Spatial transcriptome sequencing

PCA

Principal component analysis

t-SNE

T-distributed stochastic neighbor embedding

ENCORI

Encyclopedia of RNA Interactomes

GO

Gene ontology

lncRNA NEAT1

Long non-coding RNA nuclear enriched abundant transcript 1

CHX

Cycloheximide

NK

Natural killer

TEM

Transmission elmectron microscopy

RCTD

Robust cell type decomposition

HERS

High-ERS state

LERS

Low-ERS state

HSPA5

Heat shock protein family A member 5

SN

Supernatant

cRGD

Cyclic Arg-Gly-Asp

Lip

Liposome

PD1

Programmed cell death protein 1

PDI

Polydispersity index

CLSM

Confocal laser scanning microscopy

PBS

Phosphate buffer saline

GZMB

Granzyme B

HE

Hematoxylin-eosin

IFN-γ

Interferon gamma

UPR

Unfolded protein response

SEMA4A

Semaphorin 4A

SEMA3A

Semaphorin 3A

Foxp3

Forkhead box protein P3

VEGF

Vascular endothelial growth factor

METTL3

Methyltransferase-like protein 3

YTHDF

YTH domain-containing family protein

UCSC

University of California Santa Cruz

Authors’ contributions

Y. X. performed experiments in cells and a major contributor in writing the manuscript. R. X. analyzed the clinical data of TCGA and established the ERS cells. Z. C. conduced the experiments in animal models. H. Z. constructed the siNEAT1@Lip-cRGD. J. L. conducted the protein analysis. K. J. performed the sequencing data analysis. S. Y. conducted the mRNA analysis. X. Z. cultured cell lines in this study. Y. L. analyzed the m6A-qPCR in this study. D. X. uploaded all sequence data to GEO database. Q. Z. performed the flow cytometry analysis and immune cell analysis. L. Z. designed the siNEAT1@Lip-cRGD and tested its efficiency. Q. Z. confirmed the data in this study and was responsible for supervision and funding acquisition. All authors reviewed the manuscript.

Funding

This work was supported by grants from the National Natural Science Foundation of China (82373239) and the Natural Science Foundation of Guangdong Province (2024A1515012534), Director of Nanfang Hospital Foundation (2022A005) and Director of Nanfang Hospital Foundation (2023B055), Special Topics on Basic and Applied Basic Research in Guangzhou (SL2024A04J00648), the joint funding project of schools (institutes) and enterprises in Guangzhou (SL2024A03J00576), Guangzhou Major Medical Disciplines Project (2025–2027). These grants supported this work in the study design, collection, analysis and interpretation of the data and the decision to submit the paper for publication.

Data availability

Sequence data that support the findings of this study have been deposited in the PRIDE database with the accession code PXD061686 and in the GEO database with the accession code GSE290384 and GSE297015.

Declarations

Ethics approval and consent to participate

The animals were treated in accordance with the committee of Nanfang Hospital.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.