Cancer-related histologic data were collected through the TCGA (The Cancer Genome Atlas) data portal (https://cancergenome.nih.gov). Relevant clinical information and RNA-seq data from the TCGA database were obtained for analysis in this study.

Single-cell RNA sequencing data were obtained from the GEO database (Gene Expression Omnibus, https://www.ncbi.nlm.nih.gov/geo/) for the GSE183904 cohort information (containing 10 normals and 23 tumors). Downstream analysis of single-cell RNA sequencing (scRNA-seq) was performed according to the standard process of Seurat (https://github.com/satijalab/seurat). Each sample was evaluated for genes/features that were present in at least three cells, with consideration given to cells exhibiting a range of 500 or more features but not exceeding 6000 features. Cells exhibiting mitochondrial RNA percentages exceeding 20 were excluded from the analysis. The integration of 33 single-cell sequencing samples was performed using Harmony. Run Principal Component Analysis (RunPCA) is employed to conduct linear data dimensionality reduction and identify the primary 30 principal components. Employ FindNeibors (ndims = 1:30) to compute the distance among samples and utilize Run t-distributed Stochastic Neighbor Embedding (RunTSNE) (ndims = 1:30) to execute nonlinear dimensionality reduction on the samples. Cellular markers were obtained from the article attachment (https://aacrjournals.org/cancerdiscovery/article/12/3/670/681898/Single-Cell-Atlas-of-Lineage-States-Tumor).

The survival curves of S level versus tumor patients and RCN3 level versus GC patients were plotted according to the clinical data of GC patients. The analysis of survival was conducted utilizing the Kaplan–Meier method. p 0.001 was taken to obtain the differentially expressed genes, and the selection of differentially expressed genes was conducted using criteria of $|$log${}_2$FC(Fold Change)$|$ $>$1 and p 0.001 for screening purposes. Limma R was used to obtain differentially expressed gene (DEGs) with p-value 0.05 and threshold log${}_2$FC $>$1. The single-cell data had a cell proportion threshold of 0 (thresh.pc = 0), a differential fold of 0 (thresh.fc = 0, only.pos = TRUE), and a p-value of 0.05 (thresh.p = 0.05).

The CellChat R package (https://github.com/sqjin/CellChat) was employed to visualize networks of interactions between cells. The Seurat object with annotations was transformed into a CellChat entity. Receptors or ligands that exhibited increased expression in a specific cell type were detected using the identify Over ExpressedGenes function, and subsequently, gene expression data was integrated into the protein interaction network using the projectData function. The identification of ligand-receptor interactions was conducted by screening for cases where either the ligand or the receptor exhibited overexpression, utilizing the OverExpressedInteractions function. The probability value of each interaction was computed by CellChat, and the significance of the interaction was determined. The probability value calculation not only factored in the expression matrix but also incorporated prior knowledge of the interaction.

RPMI 1640 medium (11875119), fetal bovine serum (10099-141) were purchased from Gibco (Grand Island, CA, USA). CD68 (ab303565) antibody, CD206 (ab64693) antibody, iNOS (ab178945) antibody, $\alpha$SMA (ab5831), $\beta$-tubulin (ab7291) antibody were purchased from Abcam (Cambridge, UK). RCN3 (IC219670) antibody was purchased from BD Biosciences (Franklin Lakes, NJ, USA). p-STAT1 ELISA kit (JL19643-48T) was purchased from Jianglai Biological Company (Shanghai, China). MycoAlert mycoplasma test kit was purchased from LONZA, Basel, Switzerland. Fludarabine (NSC 118218) was purchased from MedChemExpress (MCE, Newark, NJ, USA).

Human gastric mucosal epithelial cells GES-1 and human monocytic leukemia THP-1 were purchased from ATCC. GES-1 was cultured in RPMI 1640 (10% fetal bovine serum (FBS), 1% double antibody) complete medium at 37 °C, 5% CO${}_2$. THP-1 was cultured in (10% FBS, 1% Dual Antibody, 0.05 mM $\beta$-mercaptoethanol) complete medium at 37 °C, 5% CO${}_2$. All cell lines underwent biweekly testing for mycoplasma using the MycoAlert Mycoplasma Detection Kit (Stratagene, San Diego, CA, USA) by the authors. All cell lines were validated by STR profiling and tested negative for mycoplasma. Type II collagenase (BS164) and type IV collagenase (BS165) were purchased from Biosharp (Shanghai, China).

In this study, we collected 40 peripheral blood samples from patients with locally advanced or metastatic gastric cancer undergoing immunotherapy at the Second Affiliated Hospital of Qiqihar Medical University, as well as formalin-fixed and paraffin-embedded sections of the primary foci of tissues from 40 patients with gastric cancer. The general information, clinical characteristics, imaging and follow-up data were obtained from the patients. This study was approved by the Hospital Ethics Committee of the Second Affiliated Hospital of Qiqihar Medical University ([2021]1215-4), and all patients signed an informed consent form. The study was performed with reference to the outline of the Declaration of Helsinki.

Reagents were employed as per the kit instructions to extract total RNA from cell/tissue samples. Next, complementary DNA (cDNA) was synthesized from RNA using the BioScript All-in-One cDNA Synthesis SuperMix (Bimake, Houston, TX, USA). RT-PCR analysis was conducted by employing FastStart Universal SYBR Green Master Mix (manufactured by Roche, located in Mannheim, Germany) and CFX96 TM Real-time PCR System (developed by Bio-Rad, headquartered in Hercules, CA, USA) for the purpose of RT-PCR analysis. The relative levels of gene expression were determined using the 2–(${}^{\mathrm{\Delta }\mathrm{\Delta }}$Ct) method. To normalize the data, the average threshold cycle (Ct) value for each transcript was compared to that of GAPDH. The primer sequences used in this study can be found in Table 1.

Tissue/cell contents were obtained through RIPA cell lysate, and total protein was collected by centrifugation. Protein concentration was quantified according to the bicinchoninic acid (BCA) assay. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) electrophoresis was performed. The proteins that were separated underwent transfer onto a PVDF membrane (ISEQ00010, Millipore, Bedford, MA, USA) utilizing a transfer device. The PVDF membrane was incubated for 1 hour with 5% skimmed milk, followed by washing with tris buffered saline (TBS) . The primary antibody (CD68 antibody (1:2000), CD206 antibody (1:2000), iNOS antibody (1:2000), $\beta$-tubulin antibody (1:2000), RCN3 antibody (1:2000), (Cell Signaling Technology, Danvers, MA, USA)) of ferroptosis marker protein was incubated at 4 °C overnight. Subsequently, the secondary antibody was added and incubated for 1h, and the corresponding secondary antibody was diluted 1:2000 according to the instructions. Finally, the immunoblot was developed in an exposure unit for about 3 minutes, ensuring the bands faced upwards. Protein bands were assessed and quantified using the Odyssey infrared imaging system (LI-COR Biosciences, Jersey, New Jersey, USA) and grayscale values of the protein bands were determined through Image-Pro Plus.

A diluted standard solution of 50 µL was introduced into the reaction wells, followed by the addition of 50 µL each of the sample and biotin-labeled antibody for testing purposes. The template was covered and incubated at 37 °C for 1 hour. After the liquid was extracted from the wells, a washing solution was introduced into each well and agitated for a duration of 30 seconds, with this process being repeated three times. The wells were then supplemented with 80 µL of affinity streptavidin-HRP and incubated for 30 min at 37 °C. Subsequently, a total volume of 50 µL of substrate A and B was added to each well, followed by an additional incubation at 37 °C for 10 min while avoiding exposure to light. After removing the plate, 50 µL of termination solution was swiftly added, and the optical density (OD) value of each well was measured subsequent to the addition of the termination solution.

RCN3-specific short hairpin RNA (shRNA) (Cat#TRCN0000029495; target sequence: GACAGAAACAAAGATGGCTAT) and non-target control shRNA (Cat# SHC016-1EA) lentiviral plasmids were purchased from Millipore igma (Burlington, MA, USA). The transfer of plasmids into HEK293T cells was facilitated using TurboFect Transfection Reagent (Thermo Scientific, Waltham, MA, USA). The supernatants were harvested 48 hours and 72 hours after transfection, followed by lentivirus concentration using the Lenti-X Concentrator (Takara Bio, San Jose, CA, USA). The GSCs were transfected with lentiviral particles containing RCN3-shRNA to establish stable expression of shRNAs. Subsequently, puromycin (1 µg/mL) or inoculin (5 µg/mL) was used for screening and selecting stable clones. Recombinant lentiviral particles were employed to produce control cells by expressing shRNAs that do not target any specific genes. To prevent the potential ascription of artifacts to a particular clone, we utilized a combination of diverse clones. Lentiviral particles were produced in 293T cells by utilizing the lentiviral shuttle vector FUGW, which contained either human RCN3 (NM_020650.3) or an empty vector. After 24 h of infection, the cells were subjected to further experiments. Human RCN3 siRNA sequence was: RNA oligo sequences 21nt guide UUCAGCAAUCACGAUGUCCCG; 21nt passenger GGACAUCGUGAUUGCUGAAAC.

The cells were subjected to formaldehyde treatment, followed by cell collection for sonication and fragmentation. Antibodies targeting specific proteins were then added, which not only bound to the target protein-DNA complexes but also interacted with each other. Subsequently, Protein A was introduced to form complexes with the antibody-target protein-DNA interactions, resulting in precipitation. The precipitated complexes underwent washing to eliminate any non-specific binding, followed by elution and acquisition of enriched target protein-DNA complexes. Subsequently, de-crosslinking and purification processes were conducted on the enriched DNA fragments for PCR analysis.

The cells or tissue sections were fixed with polymethanol, followed by three washes with PBS for 5 minutes each time. Subsequently, the tissue or cell sections were incubated in 2% BSA for 30 minutes and then washed twice with PBS. The sample was incubated with primary antibodies, including RCN3 antibody (1:2000) and $\alpha$SMA antibody (1:2000), at room temperature for 1 hour. Subsequently, the sample underwent three rounds of PBS washing, with each round lasting for 5 minutes. Next, the secondary antibody (1:2000) was introduced and allowed to incubate at ambient temperature for a duration of 30 minutes. Subsequently, PBS was used to wash the sample three times. The nuclei were stained with DAPI for 10 minutes, followed by three washes with PBS and blocking the plate. Finally, the samples were visualized under a microscope (E0226, Beyotime Biological Co., Ltd., Shanghai, China).

The collagenase digestion method was used to culture gastric CAFs and gastric NFs. The tissue blocks were digested with DMEM medium containing 0.2% type IV collagenase and 0.1% type II collagenase at 37 °C for 3 hours. Subsequently, the mixture was filtered through sterile filters with a mesh size of 200, followed by centrifugation at a speed of 1000 rpm/min for 5 minutes. The supernatant was discarded, and the sediment was resuspended in PBS before undergoing two additional rounds of centrifugation. The cells were subsequently resuspended in culture medium supplemented with 10% FBS and transferred to culture dishes after a 24-hour incubation period, the medium was changed once every 1 day.

The cumulative survival rates were determined using Kaplan–Meier survival curves. Statistical analysis of the data was performed using GraphPad Prism9 software (Dotmatics, Boston, MA, USA). The data were presented as mean $\pm$ standard deviation (mean $\pm$ SD). Statistical differences between the means of two groups were compared by t-test, and statistical differences between multiple groups were compared by one-way ANOVA. p 0.05 was statistically different.

First, we downloaded the TCGA dataset for EPIC analysis, which showed that the proportion of CAFs was significantly upregulated in GC (p 0.001) (Fig. 1A), and subsequent survival analysis showed that a higher number of CAFs predicted a poor prognosis (Fig. 1B). The above results suggested that the number of CAFs cells increased in GC patients and were gastric cancer factors. This study also screened CAFs for molecular markers. The single-cell (GSE183904) data was also downloaded for differential analysis of CAFs. Seven differentially expressed genes were screened after combined analysis: COL6A3, CTHRC1, FN1, RCN3, P4HA3, ADAM12, COL11A1 (Fig. 1C). Previous studies revealed that COL6A3, CTHRC1, FN1, RCN3, P4HA3, ADAM12, and COL11A1 were differentially expressed in CAFs of various cancers [17, 18, 19, 20, 21] and could be used as a prognostic biomarker. However, the role of RCN3 in GC is not known yet; therefore, this study chose RCN3 to explore. TCGA differential analysis showed that RCN3 was highly expressed in GC (t = –7.8347, p-value = 2.16 $\times$ 10${}^{-9}$) (Fig. 1D), and survival analysis showed that RCN3 could be used as a marker of GC prognosis (Fig. 1E).

Fig. 1.

Gastric cancer (GC) dataset analysis. (A) EPIC analysis of cancer-associated fibroblast (CAF) level. (B) Survival probability curves of CAF levels versus tumor patients. (C) Venn plots of Differential gene coanalysis. (D) Relative expression of RCN3. (E) Survival probability curves of RCN3 levels versus GC patients. (F) Uniform Manifold Approximation and Projection (UMAP) plot. (G) Distribution of RCN3 in various types of cells. (H) RCN3 expression in tumor cells and normal cells (Containing 10 normals and 23 tumors), ***p 0.001. DEG, differentially expressed gene.

The results of single-cell clustering analysis and Uniform Manifold Approximation and Projection (UMAP) were categorized into nine clusters (Fig. 1F), in which RCN3 was significantly expressed in CAFs of GC (Fig. 1G). RCN3 was demonstrated in Fig. 1H, where RCN3 was significantly highly expressed in tumor cells compared with normal cells (t = 26.568, p-value 2.2 $\times$ 10${}^{-16}$).

Analyzing the association of RCN3 with clinicopathological features, we found that high expression of RCN3 was associated with T-staging of tumor size (p 0.001; Table 2). The univariate Cox regression analysis (Fig. 2A) reported six significant risk factors for GC patients OS, including tumor age (HR, 1.51), TNM stage (HR of T, N, M: 1.26, 1.99, and 1.33, respectively), stage (HR, 1.52), and RCN3 (HR, 2.26), whereas only age and RCN3 were identified as independent risk factors for GC patients OS by multivariate Cox regression analysis (HR of age, 1.65; HR of RCN3, 1.98; Fig. 2B).

Fig. 2.

Univariate and multivariate Cox analysis of GC patients. (A) Univariate Cox analysis. (B) Multivariable Cox analysis.

To validate the results of the analysis described above, clinical GC tissue samples were collected for analysis by immunofluorescence, and the findings showed RCN3 was expressed in CAFs and was highly expressed in patients (Fig. 3A). The isolated CAFs and NFs were subjected to RCN3 expression analysis, revealing significantly elevated protein and mRNA levels of RCN3 in CAFs (Fig. 3B,C).

Fig. 3.

RCN 3 is highly expressed in CAFs and regulates macrophage polarization. (A) Immunofluorescence plots of the GC clinical samples (Scale bar = 50µm). (B) Western blot detected the protein level of RCN3 in each group. (C) Reverse Transcription-Polymerase Chain Reaction (RT-PCR) detected the expression of RCN3 in each group. (D) CellChat interaction analysis. (E) RT-PCR detected the expression of RCN3 in GES-1 cells. (F) Western blot detected the protein levels of CD68, iNOS, and CD206 in macrophage THP-1 cells. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001. n = 3.

Studies have shown that CAFs interact with other cells and play an important role in colorectal cancer [22]. CellChat analysis found that CAFs interacted with macrophages in an enhanced manner compared to normal tissue (Fig. 3D). We therefore co-cultured THP-1 cells, a human monocytic leukemia cell line, induced to differentiate into macrophages with phorbol ester (PMA), following overexpression of RCN3 in the GES-1 cell line. Consequently, we observed significant alterations in macrophage polarization and an enhanced M1 phenotype (Fig. 3E,F).

Previous literature revealed that RCN3 interacted with PCSK6 and regulated the polarization of macrophages [23]. In this study, single-cell analysis showed that PCSK6 is highly expressed in macrophages, while genecards results suggested that PCSK6 might be on the macrophage membrane (Fig. 4A). Then, an immunofluorescence assay was performed to test the expression of PCSK6. The results showed that PCSK6 was increased on the cell membranes of macrophages of GC tissues compared with normal tissues (Fig. 4B). To investigate whether RCN3 regulates macrophage polarisation by acting on PCSK6, PCSK6-KD THP-1 cells were constructed (Fig. 4C). The co-culture of GES-1 cells overexpressing RCN3 with PCSK6 knockdown THP-1 cells, the results found that knockdown of PCSK6 reversed the effect of overexpression of RCN3 on macrophage polarisation (Fig. 4D). To further explore the role of RCN3 with PCSK6, CHIP experiment was performed and found that the RCN3 protein can regulate the expression of PCSK6 (Fig. 4E).

Fig. 4.

RCN 3 acts with PCSK 6 to regulate macrophage polarization. (A) single cell analysis of PCSK6 expression. (B) Immunofluorescence graph of the GC clinical samples (Scale bar = 50µm). (C) RT-PCR showed the expression of PCSK6 in THP-1 cells. (D) Western blot showed the protein levels of CD68, INOS, and CD206 in macrophage THP-1 cells. (E) CHIP assay showed that RCN3 protein can regulate the expression of PCSK6. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001, n = 3.

To investigate how PCSK6 involved in macrophage polarisation, we collected supernatants from GES-1 cells overexpressing RCN3 and co-cultured them with THP-1 cells, which resulted in an observed increase in the phosphorylation of STAT1 (Fig. 5A). The inhibitor of STAT1 (Fludarabine), in the meantime, can reverse the macrophage polarization induced by RCN3 (Fig. 5B).

Fig. 5.

PCSK 6 regulates macrophage polarization by STAT1. (A) ELISA assay to detect the phosphorylation level of STAT1. (B) Western blot detected the expression of CD68, INOS, and CD206. *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001. n = 3.