Patients and specimens

Thirty-seven pairs of GC tumor and non-tumor tissues were collected at Sun Yat-Sen University Cancer Center (SYSUCC) (Guangzhou, China). These tissues were used for qPCR and western blotting analyses. A total of 94 pairs of GC tumor and non-tumor tissues were collected at The First Affiliated Hospital, Sun Yat-sen University (SYSUFAH) (Guangzhou, China). These patients did not receive any treatment except for surgery. The study was approved by the Committees for Ethical Review of Research Involving Human Subjects at SYSUCC and SYSUFAH. This study was conducted in accordance with ethical guidelines of the Declaration of Helsinki.

A total of 40 tumor tissues were collected from GC patients receiving anti-PD1 immunotherapy combined with chemotherapy (SYSUCC). Thirty-two patients received post-surgery treatment and 8 received both before- and post-surgery treatment. The main chemotherapy regimen was FOLFOX (Fluorouracil + Oxaliplatin) or XELOX (Capecitabine + Oxaliplatin). Patients were divided into two groups according to the therapy effect: responders (R) and non-responders (NR). Responders were identified as the patients experiencing a confirmed complete response (CR) or partial response (PR). Non-responders were identified as the patients experiencing a confirmed progressive disease (PD) or stable disease (SD). This study was approved by the Committees for Ethical Review of Research Involving Human Subjects at SYSUCC. This study was conducted in accordance with ethical guidelines of the Declaration of Helsinki.

Cell lines

Human GC cell lines AGS, MKN45 and HGC27 were obtained from American type culture collection (ATCC, MD) and cultured in RPMI1640 (Gibco BRL, NY) supplemented with 10% fetal bovine serum (ExCell Bio, Shanghai, China) and 1% penicillin/streptomycin (Invitrogen, NY, USA). All cells were incubated with 5% CO2 at 37 °C in a humidified incubator. The cell lines were validated by STR profiling.

Animal studies

All animal experiments were approved by the Animal Ethics Committee at Sun Yat-sen University Cancer Center (SYSUCC). BALB/c nude mice (female, 4-weeks old) were purchased from the Beijing Vital River Laboratory Animal Technology (Beijing, China). Animals were housed under specific pathogen-free conditions. To evaluate the xenograft growth, 1 × 106 HSPA4-overexpressing cells (MKN45-HSPA4) or control cells (MKN45-Vec) suspended in 100 µl phosphate buffered saline (PBS) were injected into the flanks of nude mice (n = 5). The length (L) and width (W) of the tumor were measured every 5 days. Tumor volumes were calculated as volume (mm3) = L×W2 × 0.5. Tumor tissues (50 mg per tumor) were frozen rapidly in liquid nitrogen, and ground using tissue grinder. RIPA lysis buffer was used to lyse the tissues on ice for 30 min. The supernatant was collected after centrifuging at 12,000 g for 20 min at 4 °C. Western blotting analysis was performed to analyze the expression of HSPA4 in xenografts.

Plasmids and siRNAs

pEZ-Lv105-HSPA4, psi-LVRU6GP-shRNAs targeting HSPA4 and the corresponding control plasmids were purchased from GeneCopoeia (Guangzhou, China). Lentivirus was packaged using the Lenti-Pac™ HIV expression Packaging Kit (GeneCopoeia, Guangzhou, China) in 293FT cells (Invitrogen, NY, USA). siRNAs targeting ALKBH5 were obtained from RiboBio (Guangzhou, China). siRNAs were transfected into cells using Lipofectamine 3000 (Invitrogen, NY, USA) according to the manufacturer’s instructions.

Cell proliferation assay

Cells were plated at a density of 500 (AGS) or 1000 (HGC27 and MKN45) cells per well in 96-well plates. After 24 h, cell viability was detected by Cell Counting Kit-8 (CCK-8; Dojindo, Kumamoto, Japan) according to the manufacturer’s instructions. OD450 was examined once a day for five days. Three independent assays were repeated.

Foci formation assay

A total of 500 cells (AGS) per well were seeded into 6-well plates and cultured for 12 days. Cell colonies (> 50 cells/colony) were fixed with 75% ethanol, stained with 1% crystal violet and counted. Three independent assays were repeated.

5-ethynyl-2’-deoxyuridine (EdU) assay

EdU assay were performed using Cell-Light 5-ethynyl-2’-deoxyuridine (EdU) Apollo488 in vitro Kit (Ribobio, Guangzhou, China). Briefly, GC cells were plated in 6-well plates, and incubated with 1× EdU assay buffer for 2 h. After washing, the cells were incubated with Apollo®488 staining for 30 min. The EdU(+) cells were analyzed by flow cytometry.

Quantitative real time PCR (qRT-PCR)

Total RNA was extracted from cells using TRIzol Reagent (Invitrogen, NY) and subjected to RT-PCR using gene-specific primers. SYBR Green Master Rox (Vazyme, Nanjing, China) was used for qRT-PCR analysis. The primers’ sequences were listed in Supplementary materials.

Cycloheximide chase assay and western blotting

Cells were treated with 10 µM cycloheximide (CHX) (Santa Cruz Biotechnology, Santa Cruz, CA) and protein lysates were harvested at different time points. Western blotting analysis was performed according to the standard protocol. Chemiluminescence signals were captured and analyzed by Quantity One system (Bio-Rad, Hercules, CA, USA). Antibodies are listed in the Supplementary materials.

Immunofluorescence (IF) staining

Paraffin-embedded formalin fixed sections were de-paraffinized with xylene, rehydrated with graded ethanol, then incubated with 3% hydrogen peroxide for 10 min at 37 °C. Antigen retrieval was performed in EDTA buffer (pH 8.0) for 8 min. The slides were blocked with 4% bovine serum albumin for 30 min at room temperature, then slides were incubated with anti-CD58 (1:50), anti-ALKBH5 (1:2000) at 4 °C overnight in a moist chamber. After thorough washing, the slides were incubated with Alexa Fluor 633– or 488–conjugated secondary antibodies (Invitrogen, NY). Subsequently, all slides were mounted with anti-fade reagent 4′,6 diamidino-2-phenylindole (DAPI, Thermo Fisher Scientific, MA). Images were captured with an OLYMPUS FV2000 fluorescence microscope.

Dot blot assay

Total RNA was spotted onto a Hybond-N + membrane (Byotime, Shanghai, China) and cross-linked using UV CROSSLINKERS (Giangarlo Scientific, NY). The membrane was incubated in blocking buffer (5% skim milk) for 1 h at room temperature and then with an anti-m6A antibody overnight at 4 °C. After washing, the membrane was incubated with an anti-mouse antibody for 1 h at room temperature. Finally, the membrane was incubated with ECL luminescent solution, and chemiluminescence signals were captured and analyzed by Quantity One system (Bio-Rad, Hercules, CA, USA). Methylene blue (MB) was used to interact with RNA and as a loading control.

Coimmunoprecipitation (co-IP) assay

Co-IP assays were performed using a Pierce Direct Magnetic IP/Co-IP kit (Thermo Scientific, Rockford, IL) according to the manual. Briefly, cells were lysed with IP lysis buffer. 500 µl lysis solution at a concentration of 1.5 mg/ml was incubated with antibody-conjugated magnetic beads overnight at 4 °C. The pull-down samples were subjected to immunoblotting assay.

Methylated RNA immunoprecipitation (Me-RIP)

Me-RIP assays were performed using a riboMeRIP m6A Transcriptome Profiling Kit (RiboBio, Guangzhou, China). Briefly, 18 µg total RNA was interrupted by RNA fragmentation buffer. Then, N6-methyladenosine (m6A) methylated RNAs were immunoprecipitated with 5 µg anti-m6A antibody. After immunoprecipitation, washing and elution, the eluted m6A RNA fragments were purified and recycled with Monarch® RNA Cleanup Kit (New England Biolabs, Ipswich, MA). Reverse transcription was performed using the Evo M-MLV RT kit with gDNA Clean for qPCR (Accurate Biology, Changsha, China). Primers were designed according to the m6A modification sites predicted by SRAMP (http://www.cuilab.cn/sramp). QRT-PCR was performed using SYBR Green SuperMix (Vazyme, Nanjing, China) and Roche 480 Real-Time PCR system (Roche, Basel, Switzerland).

CD8+ T cells culture

Peripheral blood mononuclear cells (PBMCs) from healthy donors were collected from buffy coats by Ficoll gradient centrifugation. CD8+ T cells were isolated from PBMCs using CD8+ microbeads (Miltenyi Biotec, North Rhine-Westphalia, Germany) and cultured in complete RPMI medium (RPMI1640, 10% FBS, 10 mmol/L HEPES, 1 mmol/L sodium pyruvate, 2 mmol/L L-glutamine, 0.05 mmol/L 2-mercaptoethanol, and penicillin/streptomycin) (Invitrogen, NY, USA).

In vitro T cells activation and FACS analysis

CD8+ T cells were seeded in CD3 antibody (2 µg/ml) coated plates, and stimulated with CD28 antibody (2 µg/ml) and IL-2 (10 ng/ml) for 24 h. For co-culture experiments, tumor cells (2 × 104 per well for HGC27 and MKN45; 1 × 104 per well for AGS) were plated in 96-well plates. Twelve hours later, 1 × 104 stimulated CD8+ T cells were added to the tumor cells in complete RPMI1640 medium with 10 ng/ml IFN-γ. The Plates were centrifuged at 400 g for 5 min to ensure cell-to-cell contact and the co-cultures were kept for 72 h at 37 °C and 5% CO2 in a humidified incubator. Then CD8+ T cells and tumor cells were analyzed with flow cytometry respectively.

To evaluate intracellular expression of IFN-ɤ, GZMB, and TNF-α of CD8+ T cells, CD8+ T cells were co-cultured with tumor cells for 16 h, brefeldin A (BFA, BioLegend, San Diego, CA) was added 6 h before cells were analyzed.

Bioinformatic analysis of RNA-seq and single cell RNA-seq data of GC samples

The gene expression data of GC tumor tissues and normal gastric tissues were obtained from GEPIA (http://gepia.cancer-pku.cn/) that based on samples from TCGA and GTEx databases [27] or Gene Expression Omnibus (GEO) datasets (GSE54129) [28], which have been normalized using the locally weighted scatter plot smoothing (Lowess). The LinkedOmics [29] was analyzed for genes that correlated negatively with ALKBH5 based on the TCGA program in GC tissues. The correlation between ALKBH5 and CD58 was obtained from LinkedOmics based on GC datasets of TCGA and GEO (GSE13861) [29, 30]. The correlation between HSPA4 and tumor proliferation signature was analyzed by R GSVA package, the parameter method=’ssgsea’ was selected, and finally the correlation was analyzed by the Spearman’s correlation. To identify the correlation between CD58 and immune score, RNA-seq profiles were downloaded from TCGA GC dataset, and the immune score was download from TIMER2.0 [31]. The correlation between CD58 and CD8+ T cells immune score was drawn by R ggstatsplot package.

The single-cell RNA-seq (scRNA-seq) data of 3 GC tumor tissues were downloaded from GEO under accession number GSE163558 [32]. The data were analyzed in R v.4.2.1 using Seurat v.4.2.0 package for data preprocessing, UMAP nonlinear dimensionality reduction, and cell cluster recognition and clustering. According to the Cellmarker database (http://xteam.xbio.top/CellMarker/), the cell population expressing CD8A and GZMK (granzyme K) (cluster 1) was annotated as CD8+ T cells, and the cell population expressing CD24, KRT19 (keratin 19), KRT18 (keratin 18) and EPCAM (epithelial cell adhesion molecule) (cluster 2, 3, 7, 11 and 15) was annotated as tumor cells, from which tumor cell clusters with high expression (cluster 3, 7, 11 and 15) and low expression (cluster 2) of ALKBH5 were extracted. The remaining cell clusters not covered by these markers were labeled as clusters of other cells. CellChat v.1.5.0 package was used to analyze the cell–cell contact interactions between CD8+ T cells and ALKBH5 high-expression or low-expression tumor cells.

The bulk RNA-seq data of stomach adenocarcinoma with anti-PD1 immunotherapy information were collected from Sequence Read Archive (SRA) data (https://ncbi.nlm.nih.gov/sra) [33]. We used TIGER (Tumor Immunotherapy Gene Expression Resource) (http://tiger.canceromics.org/) to analyze the gene signatures for discerning therapy responses [34].

Multiplex fluorescent immunohistochemistry (mfIHC) assay

Multiplex fluorescent immunohistochemistry assay was performed to spatially visualize and quantify the following markers: HSPA4, ALKBH5, CD58 and CD8. The Formalin-fixed paraffin-embedded GC tumor tissues were analyzed by using the Polaris® system (PerkinElmer, Waltham, MA, USA) with the customized Opal 5-color Manual IHC Kit (Panovue, Beijing, China) according to the manufacturer’s instructions. Firstly, slides were deparaffinized with xylene, rehydrated with graded ethanol. Antigen retrieval was performed in EDTA buffer (pH 8.0) for 10 min before each antibody incubation. The following antibody dilutions paired with Opal tyramide signal amplification reagent were applied in the following order for each panel: HSPA4 (1:500) with Opal 520, ALKBH5 (1: 6000) with Opal 690, CD58 (1:400) with Opal 570; CD8 (Operating fluid) with Opal 480. Slides were incubated with first primary antibody at 4 °C overnight in a moist chamber, and then incubated with paired Opal tyramide signal amplification reagent at room temperature for 1 h. After thorough washing, the slides were incubated with HRP-labeled secondary antibody at room temperature for 10 min. Repeat the above steps until the last set of antibody was stained. Finally, slides were nuclear counterstained with DAPI (Abcam, Cambridge, UK). Slides were scanned at high resolution (×20, 0.5 µM pixel) using Polaris system. Positive stain quantification and spatial cell analysis was performed using HALO 3.3 image analysis software (Indica Labs, Albuquerque, NM).

Statistical analysis

GraphPad Prism 7.0 and R v.4.2.1 were used for data analyses. Image J was used to quantify the grayscale values of western blotting images. Normalization was performed by: values of target protein bands/values of internal control bands. For mfIHC results, we calculated the best cut-off value for the average fluorescence intensity of HSPA4 using the R package maxstat (Maximally selected rank statistics with several p-value approximations version: 0.7–25), set the minimum sample size of the group to be greater than 25%, the maximum sample number of the group to be less than 75%, and the best cut-off value is 36.8. Based on this, the patients were divided into HSPA4 upregulation and normal expression groups. Kaplan-Meier analysis with the log-rank test was performed to analyze the probability of differences in the overall survival (OS). P < 0.05 was considered statistically significant. One-way analysis of variance tests was performed to compare the significant differences among more than two groups. Other significant differences were analyzed by unpaired t-test. All results are presented as the mean ± SD.