Abstract Background The dysregulation of SOXs is related to tumor invasion, metastasis, proliferation, apoptosis, and epithelial-mesenchymal transition. This research sought to investigate the function and mechanisms of SOX21 in gastric cancer (GC). Methods Multiple databases were included to determine the hub transcription factors in GC. In addition, RT-qPCR and Western blot were used to validate gene expression in tissues from GC patients. CCK-8, EdU, colony formation, wound healing, Transwell assays, and a xenograft tumor model were used to determine the function of SOX21 in GC. The targets of SOX21 were predicted and verified using ChIP, dual-luciferase reporter, and functional assays. SOX21 DNA methylation in GC cells was determined by qMSP. Rescue experiments were carried out in GC cells with DNMT1 silencing alone or in combination with SOX21 silencing. Results SOX21 was downregulated in GC tissues and cells. Ectopic expression of SOX21 inhibited cell growth, invasion, and migration, and induced apoptosis of GC cells. CKS2 was a target of SOX21, and overexpression of CKS2 promoted cell viability and mobility in GC cells overexpressing SOX21. The downregulation of SOX21 was related to the DNA hypermethylation catalyzed by DNMT1. The silencing of SOX21, by contrast, overturned the anti-tumor effects of sh-DNMT1 in vitro and in vivo. Conclusion Our data showed that DNMT1 overexpression upregulated CKS2 expression via hypermethylation of SOX21, thus promoting GC cell proliferation and growth, indicating that the DNMT1/SOX21/CKS2 axis could be a target for GC treatment. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-025-14577-z. Keywords: Gastric cancer, DNMT1, SOX21, CKS2, Hypermethylation Background Gastric cancer (GC) ranks in the five place regarding both incidence and cancer-associated mortality in the world in 2022 [[34]1]. Whereas perioperative chemotherapy is the conventional management in Western populations, D2 gastrectomy followed by adjuvant chemotherapy remains the gold standard in East Asia [[35]2]. Unfortunately, the 5-year survival of patients with advanced GC is still lower than 30%, and more effective predictive biomarkers are needed [[36]3]. In light of the imperative for efficacious and well-tolerated treatments for patients with advanced GC, targeted therapy and immunotherapy have emerged as adjuncts to traditional cytotoxic chemotherapy [[37]4]. Therefore, identifying genetic alterations involved in oncogenesis using bioinformatics algorithms such as Gene Expression Omnibus (GEO) might enable a better understanding of the mechanisms underlying the GC progression and recognize new targets for the prediction of therapy and prognosis [[38]5]. The dysregulation of sex-determining region Y-related box (SOX) members is tightly linked to tumor invasion, metastasis, proliferation, apoptosis, and epithelial-mesenchymal transition (EMT) [[39]6, [40]7]. SOXs are a conserved group of transcription factors (TF) that govern DNA binding through a highly conserved high-mobility group domain, and both the upregulation and downregulation of SOX TFs may initiate cancer progression [[41]8]. SOX21, a member of the SOXB2 group, has been identified to be a risk gene in glioblastoma multiforme along with SOX6 [[42]9]. In this study, we identified SOX21, a hub TF downregulated in GC, and cyclin-dependent kinases regulatory subunit 2 (CKS2) as a target of SOX21. CKS2 has been reported to promote the progression of certain malignancies by positively regulating proliferation, invasion, and migration, and the downregulation of CKS2 can induce cancer cell apoptosis [[43]10]. For instance, CKS2 is overexpressed in hepatocellular carcinoma tissues and is tightly linked to dismal outcomes in these patients [[44]11]. Relative to normal brain samples, the expression of CKS2 was enhanced in glioma tissues, and overexpression of CKS2 was a risk factor for the prognosis of overall survival in patients with glioma [[45]12]. Even though CKS2 has been identified as one of the GC-related genes and correlated with biological aggressiveness and poor prognosis of GC [[46]13], its upstream modifiers have not been revealed in GC. The TF E2F1 enhanced CKS2 expression by binding to its promoter in pediatric retinoblastoma [[47]14], indicating that a TF in GC might control its expression as well. In the current study, the functional role of SOX21 on GC cell proliferation and tumor growth was assessed in vitro and in vivo by carrying out gain-of‐function assays. Moreover, the upstream and downstream molecular mechanisms of SOX21 in GC were also explored. Materials and methods Patients and tissues In this study, we analyzed the tumor and paracancerous tissues of 13 patients with GC who underwent surgery in the Affiliated Chuzhou Hospital of Anhui Medical University between June 2020 and December 2023. None of the patients with GC had received chemotherapy or radiation therapy before surgery. The clinical information of all patients is listed in Table [48]1. This study was approved by the Ethics Committee of the Affiliated Chuzhou Hospital of Anhui Medical University. All patients offered signed written informed consent for participation. Table 1. Clinical information of 13 patients with gastric cancer Variables Patient number (n = 13) Age(years)  ≥ 60 8  < 60 5 Gender  Male 7  Female 6 Depth of tumor (pT)  T1 2  T2 3  T3 5  T4 3 Nodal stage (pN)  N0 1  N1 6  N2 4  N3 2 Distant metastasis (pM)  M0 12  M1 1 TNM stage  I 2  II 5  III 5  IV 1 [49]Open in a new tab Cell culture Human GC cell lines SNU-16 (CRL-5974), AGS (CRL-1739), Hs 746 T (HTB-135), and NCI-N87 (CRL-5822) were all purchased from ATCC (Manassas, VA, USA). AGS cells were cultured in F-12 K Medium (30-2004), Hs 746 T in DMEM (30-2002), and SNU-16 and NCI-N87 in RPMI-1640 medium (30-2001), with 10% FBS (30-2020, ATCC) added in all media (all acquired from ATC). Human gastric mucosal epithelial cells GES-1 (SNL-304, Sunncell, Wuhan, Hubei, China) were used in a specialized medium for GES-1 cells (SNLM-304, Sunncell). All cells were maintained in a cell incubator with 5% CO[2] at 37 °C. Generation of stable cells Lentiviral vectors containing shRNAs targeting DNMT1 or SOX21 (Table [50]2), overexpression of SOX21, and overexpression of CKS2 (all from VectorBuilder, Guangzhou, Guangdong, China) were used to infect AGS and NCI-N87 cells, and Polybrene was added (1 µL/mL, C0351-1 ml, Beyotime, Shanghai, China). After 24 h, 2 µg/mL puromycin (HY-K1057, MedChemExpress, Monmouth Junction, NJ, USA) was added to establish stable GC cell populations. Table 2. ShRNA information shRNA Target sequence hDNMT1[shRNA#1] GACGACCCTGACCTCAAATAT hDNMT1[shRNA#2] CCCGAGTATGCGCCCATATTT hDNMT1[shRNA#3] GAGGTTCGCTTATCAACTAAT hSOX21[shRNA#1] GCCTCCCTGTTTGTACTATTT hSOX21[shRNA#2] ACGGAGGTGGAGGAGTAACTT hSOX21[shRNA#3] GCTACATGATCCCGTGCAACT [51]Open in a new tab DNMT1 DNA-methyltransferase 1, SOX21 sex determining region Y-box 21, shRNA short hairpin RNA RNA extraction and RT-qPCR Total RNA was extracted from GC cells, GC tissues, or adjacent tissues using TRIzol reagent (15596018CN, Thermo Fisher Scientific Inc., Waltham, MA, USA). The cDNAs were generated using the Evo M-MLV Reverse Transcription Kit (AG11707, Accurate Biology, Changsha, Hunan, China), and TB Green Premix Ex Taq II (RR820Q, Takara Holdings Inc., Kyoto, Japan) was used for quantitative PCR. The relative gene expression was examined using the 2^−ΔΔCt method. The primer sequences were: DNMT1: forward, 5’-AGGTGGAGAGTTATGACGAGGC-3’ and reverse, 5’-GGTAGAATGCCTGATGGTCTGC-3’; SOX21: forward, 5’-CCGAGTGGAAACTGCTCACAGA-3’ and reverse, 5’-GGCGAACTTGTCCTTCTTGAGC-3’; CKS2: forward, 5’- GAGGAGACTTGGTGTCCAACAG-3’ and reverse, 5’-GATTTGACGATCCCCAGATAAACT-3’; GAPDH: forward, 5’-GTCTCCTCTGACTTCAACAGCG-3’ and reverse, 5’- ACCACCCTGTTGCTGTAGCCAA-3’. Western blot analyses Proteins were extracted from GES-1 cells, GC cells, GC tissues, or adjacent tissues using RIPA buffer (R0010, Solarbio, Beijing, China) and quantified using the BCA Protein Assay Kit (PC0020, Solarbio). Proteins were then transferred to a PVDF membrane (IPFL07810, Sigma-Aldrich Chemical Company, St Louis, MO, USA) following electrophoresis using the SDS-polyacrylamide gel electrophoresis sample preparation kit (89888, Thermo Fisher Scientific). The 5% skimmed milk was used to seal the membranes before primary and secondary antibodies (ab6721 from Abcam or CSB-[52]PA573747 from Cusabio, Wuhan, Hubei, China) were added. The primary antibodies to SOX21 (1:1000, ab220024), CKS2 (1:500, ab155078), and GAPDH (1:2500, ab9485) were purchased from Abcam (Cambridge, UK), and the primary antibodies against DNMT1 (1:1000, 5032), E-Cadherin (1:1000, 3195), N-Cadherin (1:1000, 4061), and Vimentin (1:1000, 5741) were purchased from Cell Signaling Technologies (Beverly, MA, USA). Protein bands were identified using BeyoECL Plus (P0018S, Beyotime). The relative expression of the relevant proteins was examined using ImageJ software. CCK-8 CCK-8 (40203ES60, Yeasen, Shanghai, China) was used to examine GC cell proliferation. GC cells (1500 cells/well) were incubated at 37 °C with 5% CO[2] for 2 h in 96-well cell culture plates (FCP962, Beyotime). After the supplementation of 10 µL of CCK-8 at 0, 24, 48, and 72 h, the cells were incubated for 4 h. The optical density (OD) value at 450 nm was read using a microplate reader [[53]15, [54]16]. Colony formation assay GC cells were seeded into 6-well cell culture plates (FCP060, Beyotime) at 1000/well and cultured in a serum-free medium at 37 °C, 5% CO[2] with medium renewal at 3-d intervals. After 14 days of incubation, the cells were fixed in 10% formalin (60535ES60, Yeasen), stained with 0.1% crystal violet stain solution (60506ES60, Yeasen), photographed, and observed under a light microscope. Wound healing assay The GC cells were cultured in six-well plates with 8 × 10^5 cells per well. When the GC cells adhered, a straight line was drawn in each well using a sterilized 200 µL tip. The cells were washed and incubated in a serum-free medium under the conditions of 37 °C, 5% CO[2]. The 0 h, 12 h, and 24 h images were taken and recorded. Transwell assays Transwell cell culture plates (CLS3464, Sigma-Aldrich) were used, with Matrix-Gel (C0371-50 ml, Beyotime) applied in the invasion assay and without Matrix-Gel in the migration assay. GC cells (4 × 10^4 cells) suspended in a medium without FBS were loaded into the apical chamber, and a medium containing 10% FBS was placed in the basolateral chamber. After a 24-h incubation at 37 °C with 5% CO[2], migrated and invaded cells were fixed, stained, and observed under a light microscope. TUNEL assay CG cells (5 × 10^5 cells/per well) were incubated in 24-well plates at 37 °C for 14 h. The cells were fixed with 4% paraformaldehyde (PFA) for 1 h, permeabilized with 0.1% Triton X-100 (HFH10, Thermo Fisher Scientific), and stained using TUNEL Apoptosis Detection Kit (C1086, Beyotime) in the dark for 60 min, and counterstained with DAPI (D9542, Sigma-Aldrich) for 10 min. Images were acquired and photographed using fluorescence microscopy. EdU assay The EdU staining kit (ab219801, Abcam) was used to detect cell proliferation. After 60% of the GC cells had adhered to the wall, the cells were labeled using a 10 µM EdU stock solution for 2 h. GC cells were fixed with 4% PFA for 30 min and washed, followed by the addition of 100 µL of 1X permeabilization buffer. The GC cells were incubated with EdU staining solution for 2 h, counter-stained with 5 µg/mL DAPI for 30 min at room temperature (both in the dark), and photographed under a microscope. Methylation-specific qPCR (qMSP) Genomic DNA was extracted from GC cells using a DNA extraction kit (9765, Takara). The DNA was denatured and sulfated using the MethylCode Bisulfite Conversion Kit (MECOV50, Thermo Fisher Scientific). Subsequently, the DNA was desulfated and purified, and real-time qPCR was performed using a Rapid MS-qPCR kit (A-P-1028, IVDSHOW, Zhangjiakou, Hebei, China) according to the manufacturer’s instructions. The percentage of methylation for each sample was calculated by the following formula: %Methylation = 100/[1 + 2^ΔCt(meth−unmeth)]% [[55]17]. MSP primers were as follows: Left M primer 5’-TTTTGGGATATTTTAATTTTTCGG-3’, Right M primer 5’-AAAAAACTCACCAACAAACGC-3’; Left U primer 5’-GTTTTTGGGATATTTTAATTTTTTGG-3’, Right U primer 5’-AAAAAACTCACCAACAAACACC-3’. Chromatin Immunoprecipitation (ChIP) According to the protocol of the ChIP kits (RK20258, ABclonal, Wuhan, Hubei, China), 1 × 10^7 GC cells were ice-bathed with cell fixative to cross-link for 15 min, and the reaction was terminated by glycine. The cells were resuspended by adding a cell lysis buffer, and the precipitate was collected by centrifugation. After a 10-min incubation with ChIP sonication buffer, the cells were sonicated and centrifuged to collect the supernatant. The supernatant was incubated with nuclease-free water, NaCl, and RNase A for 0.5 h and with Proteinase K at 65 °C for 2 h. DNA was purified using a universal DNA purification kit (DP214, TIANGEN Biotech Co., Ltd., Beijing, China) and analyzed for chromatin fragmentation. The supernatant was incubated with anti-DNMT1 (1:50, 5032, Cell Signaling Technologies) or anti-SOX21(1:50, sc-293461, Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) at 4 °C for 3 h. Normal rabbit IgG (1:100, ab171870, Abcam) or mouse IgG (1:20, HY-[56]P80757, MedChemExpress, Monmouth Junction, NJ, USA) was used as a control. The supernatant was transferred to a tube containing A/G magnetic beads (26162, Thermo Fisher Scientific) and incubated for 2 h at 4 °C. After adsorption of the magnetic beads, the DNA was purified by centrifugation and amplified by qPCR using either the SOX21 promoter-specific primer or the CKS2 promoter-specific primer. Luciferase reporter assay The promoters of SOX21 and CKS2 were introduced into the pGL3-basic vector (VT1554, Youbio, Changsha, Hunan, China) to obtain the pGL3-SOX21 vector and the pGL3-CKS2 vector, respectively. GC cells with low expression of DNMT1 were transfected with pGL3-SOX21 vector and pRL-TK vector (VT1568, Youbio), and GC cells overexpressing SOX21 were transfected with pGL3-CKS2 vector and pRL-TK. The GC cells were lysed after 36 h, and Renilla luciferase (R-luc) was used to normalize firefly luciferase (F-luc) activity to evaluate the reporter transcription using the Renilla-Firefly Luciferase Test Kit (16185, Thermo Fisher). Tumor xenograft model The animal experiments involved in this study were approved by the Animal Ethics Committee of the Affiliated Chuzhou Hospital of Anhui Medical University. Xenograft experiments were performed using male BALB/c nude mice (5 weeks, 14–16 g) (Vital River, Beijing, China). AGS cells (2 × 10^7) were administered into the axilla of mice. Tumor volume was recorded every 5 days using a vernier caliper, and the volume was calculated using the formula volume = (length × width × width)/2. After 25 days, euthanasia was performed by intraperitoneal injection of 200 mg/kg sodium pentobarbital, and the tumors were weighed. Immunohistochemistry (IHC) Mouse tumor tissues were paraffin-embedded and sectioned at 5-µm. After dewaxing, hydration, and antigen retrieval using citrate buffer (c9999, Sigma-Aldrich) for 15 min at 99 °C, treatment with 3% H[2]O[2] at room temperature for 10 min, the sections were sealed using the serum for 45 min and probed with the primary antibodies to Ki67 (1:50, PA5-114437, Abcam) and cleaved-caspase-3 (1:400, 9661, Cell Signaling Technologies) overnight at 4 °C, followed by incubation with HRP-conjugated secondary antibody (1:1000, ab6721, Abcam). The reaction product was developed using a DAB kit (PK10005, ProteinTech Group, Chicago, IL, USA). The slides were counter-stained with hematoxylin (C0107-100 ml, Beyotime), viewed under a microscope, and photographed. The rate of positive cells was measured. Data analysis Data are expressed as mean ± SEM. Depending on whether the samples were paired or not, a paired t-test or a t-test was used. We used one-way or two-way ANOVA, followed by Tukey’s post hoc test, when comparing more than two groups of data. All statistical calculations were carried out using GraphPad Prism (Version 8.0, GraphPad Software Inc., San Diego, CA, USA). If the p-value was < 0.05, the difference was defined as statistically significant. Results SOX21 is lowly expressed in GC tissues and cells [57]GSE13911 (expression data from 38 primary gastric tumors and 31 adjacent normal samples), [58]GSE118916 (gene expression profile in 15 pairs of GC tumor and adjacent non-tumor tissues), and [59]GSE208099 (16 biopsies from both tumor site and background normal mucosa) datasets in the GEO database were analyzed for differentially expressed genes (Fig. [60]1A) using Benjamini & Hochberg as the p-value correction condition, and adjusted p < 0.05 and|Log[2]FoldChange| >1 as the threshold. A total of 13,898 genes related to GC were downloaded from the Genecards database ([61]https://www.genecards.org/) by searching with the keyword GC. The human TF list was downloaded from HumanTFDB ([62]http://bioinfo.life.hust.edu.cn/HumanTFDB). A total of 16 differentially expressed TFs were obtained by taking intersections via Jvenn ([63]https://jvenn.toulouse.inrae.fr/app/example.html) (Fig. [64]1B). Subsequently, a PPI network was formed using these 16 TFs on the String database ([65]https://cn.string-db.org/) and imported into Cytoscape to select the top 10 as hub genes (Fig. [66]1C) through the clusteringCoefficient algorithm via the CytoHubba plugin. Among them, the role or expression of NEUROD1 [[67]18], FOSL1 [[68]19], SOX2 [[69]20], LEF1 [[70]21], MYC [[71]22], ESRRG [[72]23], HOXC6 [[73]24], and HOXBs [[74]25] have been revealed in GC, which makes SOX21 as our TF of interest. We found that SOX21 was lowly expressed in all three GEO datasets ([75]GSE13911 (Log[2]FoldChange = −1.078), [76]GSE118916 (Log[2]FoldChange = −3.822), and [77]GSE208099 (Log[2]FoldChange = −3.433)). SOX21 was predicted to be lowly expressed in stomach adenocarcinoma (STAD) in the UALCAN database ([78]https://ualcan.path.uab.edu/index.html) as well (Fig. [79]1D). Fig. 1. [80]Fig. 1 [81]Open in a new tab The hub TF SOX21 is downregulated in GC tissues and cell lines. A Differentially expressed genes in the [82]GSE13911, [83]GSE118916, and [84]GSE208099 datasets. B The intersection of differentially expressed genes in the three datasets, GC-related genes in the Genecards database, and the human TF list. C The PPI network of 16 differentially expressed TF and the hub 10 genes. D The expression of SOX21 in the STAD-UALCAN dataset. E The mRNA expression of SOX21 in adjacent and tumor tissues in GC patients was examined using RT-qPCR (n = 13). F The protein expression of SOX21 in adjacent and tumor tissues in GC patients was analyzed using western blot analysis (n = 13). Paired t-tests were used to compare the data between the two groups. Data are expressed as means ± standard errors of the means In our enrolled patients (n = 13), the mRNA and protein expression of SOX21 was reduced in tumor tissues relative to paired paracancerous tissues (Fig. [85]1E, F). Overexpression of SOX21 represses GC cell viability and mobility We examined the expression of SOX21 in GC cells (SNU-16, AGS, Hs 746 T, and NCI-N87) and GES-1 cells by western blot analyses. SOX21 was poorly expressed in GC cells compared to GES-1 cells (Fig. [86]2A). Since SOX21 expression was relatively low in AGS and NCI-N87 cells, we used lentivirus overexpressing SOX21 to infect these cell lines. Overexpression efficiency was detected by RT-qPCR and western blot analysis (Fig. [87]2B, C). The CCK-8 and colony formation assays were conducted to evaluate cell growth. It was found that the upregulation of SOX21 led to hampered cell viability (Fig. [88]2D) and a reduced number of colonies formed (Fig. [89]2E). The migration of GC cells was detected using a wound healing assay, and GC cells overexpressing SOX21 healed slower at both 12 h and 24 h (Fig. [90]2F). Transwell assay was conducted for migration and invasion evaluation. There were a few migrated and invaded GC cells when SOX21 was overexpressed (Fig. [91]2G). Lastly, we found that the apoptotic rate was enhanced upon overexpression of SOX21 using TUNEL assays (Fig. [92]2H). Fig. 2. [93]Fig. 2 [94]Open in a new tab Overexpression of SOX21 inhibits proliferation, migration, and invasion and promotes apoptosis in GC cells. A The protein expression of SOX21 in GC cells and GES-1 cells was examined using western blot analysis. B The mRNA expression of SOX21 in GC cells after infection with oe-NC or oe-SOX21 was examined using RT-qPCR. C The protein expression of SOX21 in GC cells after infection with oe-NC or oe-SOX21 was examined using western blot analysis. D The OD value of GC cells was read at 0, 24, 48, and 72 h using the CCK-8 assay. E The proliferation of GC cells was examined using colony formation assays. F The wound healing rate of GC cells at 12 h and 24 h after scratches. G The migration and invasion of GC cells were examined using the Transwell assay. H Detection of apoptosis in GC cells by TUNEL assay. Unpaired t-tests were used to compare the data between two groups, and ANOVA and Tukey’s post hoc test were used to compare the data between multiple groups. Data are expressed as means ± standard errors of the means of three independent experiments SOX21 slows the GC cell growth in vivo Only one cell line, AGS, which showed lower SOX21 expression, was administered subcutaneously into the mice for the consideration of animal suffering reduction. The volume was measured every five days (Fig. [95]3A). The mice were euthanized after 25 days, and the tumor weights were recorded (Fig. [96]3B), which showed that the subcutaneous xenograft tumors in nude mice overexpressing SOX21 were reduced in size and weight. The high expression of SOX21 was observed in the tumor tissues of mice administered with AGS cells overexpressing SOX21 (Fig. [97]3C). The expression of proliferation-associated protein Ki67 and apoptosis-associated protein cleaved-caspase-3 was detected by IHC (Fig. [98]3D). The expression of Ki67 was decreased, and the expression of cleaved-caspase-3 was increased in the tumor tissues formed by AGS cells overexpressing SOX21. Fig. 3. [99]Fig. 3 [100]Open in a new tab Overexpression of SOX21 inhibits GC growth in vivo. A The representative images of subcutaneous xenograft tumors in nude mice and the tumor growth curve within 25 d. B The weight of tumors formed by AGS cells with oe-NC or oe-SOX21 at day 25. C The expression of SOX21 in the tumor tissues was examined using western blot analysis. D The representative immunohistochemical images and positive cells of Ki67 and cleaved-caspase-3 in the tumor tissues formed by AGS cells with oe-NC or oe-SOX21. Unpaired t-tests were used to compare the data between two groups, and ANOVA and Tukey’s post hoc test were used to compare the data between multiple groups. Data are expressed as means ± standard errors of the means (n = 6) High expression of SOX21 transcriptionally suppresses CKS2 expression in GC cells The GEO datasets ([101]GSE13911, [102]GSE118916, and [103]GSE208099) were intersected with the GC-related genes in GeneCards, and 240 differential genes were obtained (Fig. [104]4A). KEGG pathway enrichment analysis was performed by the KOBAS database ([105]http://bioinfo.org/kobas), ranked according to p-value, and the three pathways with the smallest p-value were selected for the next analysis. Among them, hsa05200: Pathways in cancer pathway is closely related to GC (Fig. [106]4B). Interaction analysis was performed through the STRING database. The clustering Coefficient algorithm in Cytoscape revealed that CKS2 was at the most central position in the pathway (Fig. [107]4C). We further found that CKS2 was highly expressed in the [108]GSE13911 (Log[2]FoldChange = 1.468), [109]GSE118916 (Log[2]FoldChange = 1.02), and [110]GSE208099 (Log[2]FoldChange = 1.863) datasets, and were also highly expressed in the STAD-UALCAN database (Fig. [111]4D). SOX21 was found to have binding sites on the promoter of CKS2 by JASPAR ([112]https://jaspar.elixir.no/) website (Fig. [113]4E). STAD (from TCGA [[114]26], and stomach samples were selected on the GEPIA database ([115]http://gepia.cancer-pku.cn/). Spearman’s correlation analysis revealed that SOX21 was inversely correlated with CKS2 expression (Fig. [116]4F). Fig. 4. [117]Fig. 4 [118]Open in a new tab High expression of SOX21 transcriptionally suppresses CKS2 expression in GC cells. A The intersection of differentially expressed genes in the three datasets and GC-related genes in the Genecards database. B The KEGG enrichment analysis of 240 differentially expressed genes in the three datasets and GC-related genes in the Genecards database. C The PPI network of genes is enriched in the hsa05200: Pathways in cancer. D The expression of CKS2 in the STAD-UALCAN database. E The binding relation between SOX21 and CKS2 was predicted on the JASPAR website. F The negative correlation between SOX21 and CKS2 was predicted using the GEPIA database. G The mRNA expression of CKS2 in adjacent and tumor tissues in GC patients was examined using RT-qPCR (n = 13). H The protein expression of CKS2 in GC cells and GES-1 cells was examined using western blot analysis. I The binding relation between SOX21 and the CKS2 promoter was verified using a ChIP assay. J The binding relation between SOX21 and the CKS2 promoter was examined using luciferase assays. K The mRNA expression of CKS2 in GC cells after infection with oe-NC or oe-SOX21 was examined using RT-qPCR. L The protein expression of CKS2 in GC cells after infection with oe-NC or oe-SOX21 was examined using western blot analysis. Paired or unpaired t-tests were used to compare the data between two groups, and ANOVA and Tukey’s post hoc test were used to compare the data between multiple groups. Data are expressed as means ± standard errors of the means of three independent experiments To validate our prediction results, RT-qPCR was performed to detect the expression of CKS2 in tumor tissues and paired paracancerous tissue samples from GC patients. The expression of CKS2 in tumor tissues was higher than that in paracancerous tissues (Fig. [119]4G). The expression of CKS2 was assessed in GC cells (SNU-16, AGS, Hs 746 T, and NCI-N87) and GES-1 cells. As expected, CKS2 was highly expressed in GC cells (Fig. [120]4H). ChIP assay showed that SOX21 bound to the CKS2 promoter (Fig. [121]4I). Dual-luciferase reporter gene assay showed that GC cells overexpressing SOX21 had lower luciferase activity, i.e., overexpression of SOX21 repressed the transcriptional activity of the CKS2 promoter (Fig. [122]4J). Finally, we also analyzed the effect of oe-SOX21 treatment on the transcriptional and translational levels of CKS2 in GC cells by RT-qPCR and western blot analysis. The overexpression of SOX21 significantly downregulated the levels of CKS2 (Fig. [123]4K, L). Overexpression of CKS2 overturns the anti-proliferative and pro-apoptotic properties of SOX21 overexpression on GC cells GC cells stably overexpressing SOX21 were infected with CKS2 overexpression lentiviral vector to obtain the oe-SOX21 + oe-NC and oe-SOX21 + oe-CKS2 groups, and the overexpression efficiency was detected by RT-qPCR and western blot analysis (Fig. [124]5A, B). The infected GC cells were subjected to EdU and colony formation assays. The ability of overexpression of SOX21 to inhibit the proliferation of GC cells was compromised by ectopic expression of CKS2 (Fig. [125]5C, D). The protein expression of E-Cadherin, N-Cadherin, and Vimentin was detected. The downregulation of N-Cadherin and Vimentin and upregulation of E-Cadherin induced by SOX21 were reversed by overexpression of CKS2, indicating GC cells reverted to a more mesenchymal phenotype (Fig. [126]5E). The migration and invasion of GC cells were detected using Transwell assays, and overexpression of CKS2 reversed the suppressing effects of overexpression of SOX21 on GC cell migration and invasion (Fig. [127]5F). Apoptosis was detected using TUNEL staining (Fig. [128]5G), which showed that overexpression of CKS2 attenuated the promotion of apoptosis by overexpression of SOX21. Fig. 5. [129]Fig. 5 [130]Open in a new tab Overexpression of CKS2 impairs the ability of SOX21 to inhibit the proliferation of GC cells. A The mRNA expression of CKS2 in GC cells infected with oe-SOX21 + oe-NC or oe-SOX21 + oe-CKS2 was examined using RT-qPCR. B The protein expression of CKS2 in GC cells infected with oe-SOX21 + oe-NC or oe-SOX21 + oe-CKS2 was examined using western blot analysis. C The proliferation of GC cells was examined using colony formation assays. D The EdU-positive cells in GC cells overexpressing both SOX21 and CKS2. E The protein expression of E-Cadherin, N-Cadherin, and Vimentin in GC was examined using western blot analysis. F The migration and invasion of GC cells were examined using the Transwell assay. G Detection of apoptosis in GC cells by TUNEL assay. Unpaired t-tests were used to compare the data between two groups, and ANOVA and Tukey’s post hoc test were used to compare the data between multiple groups. Data are expressed as means ± standard errors of the means of three independent experiments High expression of DNMT1 promotes SOX21 promoter methylation The reason for SOX21 downregulation in GC may be affected by methylation modification by DNA methylases. We predicted the CpG islands in the SOX21 promoter (chr13: 94707622–94709622) from the MethPrimer database ([131]http://www.urogene.org/cgi-bin/methprimer/methprimer.cgi) and found that the presence of CpG islands (Fig. [132]6A). Lowly expressed SOX21 was found to have hypermethylated modifications in GC through the Mexpress database ([133]https://mexpress.ugent.be/) (Fig. [134]6B). To further investigate the mechanism by which SOX21 is methylated in GC, we explored the GC data from the Meth450 platform in the LinkOmics database ([135]http://www.urogene.org/cgi-bin/methprimer/methprimer.cgi) using Spearman’s analysis to investigate the correlation between DNA methylation transferases DNMT3A, DNMT1, DNMT3B, DNMT3L and the methylation level of SOX21 (Fig. [136]6C). Only the DNMT1 expression was positively correlated with the SOX21 DNA methylation (R = 0.2041). DNMT1 was found to be overexpressed in STAD through the UALCAN database (Fig. [137]6D). Fig. 6. [138]Fig. 6 [139]Open in a new tab High expression of DNMT1 promotes DNA methylation of the SOX21 promoter. A Prediction of CpG islands in the SOX21 promoter region using the MethPrimer database. B The methylation of the SOX21 promoter region was predicted on the Mexpress database. C The correlation between DNMT3A, DNMT1, DNMT3B, DNMT3L, and SOX21 methylation levels in GC from the Meth450 platform in the LinkOmics database using Spearman analysis. D DNMT1 expression was predicted in the STAD-UALCAN database. E The protein expression of DNMT1 in adjacent and tumor tissues in GC patients was examined using western blot analysis (n = 13). F The protein expression of DNMT1 in GC cells and GES-1 cells was examined using western blot analysis. G The mRNA expression of DNMT1 in GC cells infected with shRNAs targeting DNMT1 was examined using RT-qPCR. H The protein expression of DNMT1 in GC cells infected with shRNAs targeting DNMT1 was examined using western blot analysis. I The methylation level of the SOX21 promoter in GC cells infected with sh-NC or sh-DNMT1 #2 was examined using the MSP assay. J The binding relation between DNMT1 and the SOX21 promoter was verified using a ChIP assay. K The binding relation between DNMT1 and the SOX21 promoter was examined using luciferase assays. L The mRNA expression of SOX21 in GC cells infected with sh-DNMT1 #2 was examined using RT-qPCR. M The protein expression of SOX21 in GC cells infected with sh-DNMT1 #2 was examined using western blot analysis. Paired or unpaired t-tests were used to compare the data between two groups, and ANOVA and Tukey’s post hoc test were used to compare the data between multiple groups. Data are expressed as means ± standard errors of the means of three independent experiments Western blot analysis demonstrated that DNMT1 was overexpressed in GC tissues relative to adjacent tissues (Fig. [140]6E). The expression of DNMT1 in GC cells and GES-1 cells was detected using western blot analysis. DNMT1 was highly expressed in all GC cells (Fig. [141]6F). To validate the regulatory effects of DNMT1 on SOX21, we silenced its expression in AGS and NCI-N87 cells using three shRNAs and verified the mRNA and protein expression of DNMT1 and SOX21 (Fig. [142]6G, H). The expression of DNMT1 was reduced, and the expression of SOX21 was elevated following DNMT1 knockdown. We chose the one with the best knockdown efficiency (sh-DNMT1 #2) for the subsequent experiments. The MSP assay detected a reduced level of SOX21 promoter methylation in GC cells infected with sh-DNMT1 #2 (Fig. [143]6I). ChIP assay revealed that DNMT1 bound to the SOX21 promoter (Fig. [144]6J). Moreover, GC cells with DNMT1 knockdown had higher luciferase activity, i.e., low DNMT1 expression promoted the transcriptional activity of the SOX21 promoter (Fig. [145]6K). Lastly, DNMT1 knockdown led to increased SOX21 expression (Fig. [146]6L, M). Silencing of SOX21 abates the anti-proliferative and pro-apoptotic properties of DNMT1 knockdown on GC cells Three shRNAs targeting SOX21 were used to infect GC cells with DNMT1 knockdown. It was found that the silencing of SOX21 repressed SOX21 expression without alteration in DNMT1 expression (Fig. [147]7A, B). sh-SOX21 #3, with the best efficacy, was used for the following assays. EdU and CCK-8 assays were conducted to assess cell proliferation and viability. The repression of sh-DNMT1 #2 on cell proliferation and viability was reversed by SOX21 knockdown (Fig. [148]7C, D). Transwell assays demonstrated that the repressed migration and invasion by sh-DNMT1 #2 were rescued by SOX21 knockdown (Fig. [149]7E). Moreover, the apoptotic cells induced by sh-DNMT1 #2 were reduced by sh-SOX21 #3 (Fig. [150]7F). Fig. 7. [151]Fig. 7 [152]Open in a new tab Silencing of SOX21 abates the anti-proliferative and pro-apoptotic properties of sh-DNMT1 on GC cells. A The mRNA expression of SOX21 in GC cells infected with sh-DNMT1 + shRNAs targeting SOX21 was examined using RT-qPCR. B The protein expression of SOX21 in GC cells infected with sh-DNMT1 + shRNAs targeting SOX21 was examined using western blot analysis. C The EdU-positive GC cells after infection. D The OD value of GC cells was read at 0, 24, 48, and 72 h using the CCK-8 assay. E The migration and invasion of GC cells were examined using the Transwell assay. F Detection of apoptosis in GC cells by TUNEL assay. ANOVA and Tukey’s post hoc test were used to compare the data between multiple groups. Data are expressed as means ± standard errors of the means of three independent experiments The DNMT1/SOX21/CKS2 axis is involved in GC progression in vivo AGS cells infected with sh-NC, sh-DNMT1 #2, sh-DNMT1 #2 + sh-NC, sh-DNMT1 #2 + sh-SOX21 #3, and oe-SOX21 + oe-NC or oe-SOX21 + oe-CKS2 were administrated into nude mice. The tumor volume and weight were repressed in mice administered with sh-DNMT1 #2, which were reversed by sh-SOX21 #3 (Fig. [153]8A, B). As expected, the expression of DNMT1 was also influenced by sh-DNMT1 #2, but not sh-SOX21 #3, nor oe-CKS2. The restoration of SOX21 by sh-DNMT1 #2 was again repressed by sh-SOX21 #3. The expression of CKS2, as the downstream molecule, was governed by both DNMT1 and SOX21 (Fig. [154]8C). The downregulation of Ki67 and upregulation of cleaved-caspase-3 by sh-DNMT1 were reversed by sh-SOX21 as well (Fig. [155]8D). Likewise, the anti-tumor effects of oe-SOX21 were also abated by overexpression of CKS2, as evidenced by larger tumor burden and higher Ki-67 staining and lower cleaved-caspase-3 staining (Fig. [156]8A-C). Fig. 8. [157]Fig. 8 [158]Open in a new tab DNMT1/SOX21/CKS2 axis is involved in GC progression in vivo. A The representative images of subcutaneous xenograft tumors in nude mice and the tumor growth curve within 25 d. B The weight of tumors formed by AGS cells with sh-NC, sh-DNMT1 #2 + sh-NC, sh-DNMT1 #2 + sh-SOX21 #3, oe-SOX21 + oe-NC, or oe-SOX21 + oe-CKS2 at day 25. C DNMT1, SOX21, and CKS2 expression in the tumor tissues was examined using western blot analysis. D The representative immunohistochemical images and positive cells of Ki67 and cleaved-caspase-3 in the tumor tissues formed by AGS cells. ANOVA and Tukey’s post hoc test were used to compare the data between multiple groups. Data are expressed as means ± standard errors of the means (n = 6) Discussion The key findings of our study include the discovery that SOX21 overexpression limits GC growth by targeting CKS2. Our detailed investigation shows that DNMT1 represses SOX21 in GC cells by elevating the DNA methylation of the SOX21 promoter region, which results in CKS2 overexpression in GC cells. In this study, we identified that SOX2 and SOX21 were the most differentially expressed TFs between GC and adjacent tissues in three GEO datasets and that there was an interaction between them. A significant decline in the number of cells expressing SOX2 and SOX21 TFs within the subgranular zone of 5xFAD mice was observed in comparison to their non-transgenic counterparts [[159]27], indicating their possible role in Alzheimer’s Disease. Eenjes et al. reported that SOX21 inhibited the differentiation of airway progenitors by antagonizing SOX2-induced expression of specific genes in airway differentiation [[160]28]. However, different from the well-known effects of SOX2 on various cancers, including GC [[161]29], there is scarce evidence showing the function of SOX21 in cancer, in particular GC, which highlights the novelty of our research. SOX2, in conjunction with the long noncoding RNA SOX21-AS1, regulates the characteristics of breast cancer stem cells [[162]30, [163]31]. Caglayan et al. showed that the induction of SOX21 in the tumor cells resulted in significantly smaller tumor size, and mice injected with glioma cells orthotopically into the brain survived longer when SOX21 was elevated [[164]32]. In this study, we found the reduced expression of SOX21 in both GC tissues and cells and constrained GC cell proliferation in vitro and tumor growth in vivo in the presence of overexpression of SOX21, indicating the possible anti-tumor properties of SOX21 upregulation. SOX21, a member of the SOXB2 subfamily, has repressor activity and can act as a tumor suppressor during gliomagenesis since SOX21 mediates tumor suppressor response by modulating p53 levels [[165]33, [166]34]. Therefore, we linked its tumor-suppressing effects to the downstream oncogenes. CKS2 expression was significantly elevated in breast cancer samples relative to normal breast tissues, and its expression was higher in breast cancer patients with unsatisfactory prognoses [[167]35]. Moreover, CKS2 knockdown suppressed cell proliferation, induced S and G2/M phase arrest and apoptosis in vitro, and slowed tumor growth in vivo in pancreatic cancer [[168]36]. As for its role in GC, CKS2 was identified to have high-importance values in multiple GC predictive models [[169]37]. In more detail, the positivity rates for CKS2 in benign gastric tumor tissue, low-grade intraepithelial neoplasia tissue, high-grade intraepithelial neoplasia tissue, and GC tissue gradually increased, and upregulation of CKS2 was linked to tumor size, Lauren classification, number of lymph node metastases, and TNM stage in patients with GC [[170]38]. In this study, the anti-tumor effects of SOX21 overexpression were overturned by CKS2 overexpression, further substantiating the conclusion derived from prediction and regulatory assays that CKS2 is a downstream target of SOX21. In addition, both SOX21 and CKS2 have been reported to regulate the process of EMT (the expression of E-cadherin, N-cadherin, and Snail) [[171]39, [172]40], which was further validated through our in vitro assays. Considering the importance of EMT to cancer metastasis [[173]41], we provided preliminary findings regarding the role of the SOX21/CKS2 axis on metastasis. Aberrant DNA methylation occurs in the early stage of GC, leading to cancer-specific genetic and epigenetic changes [[174]42]. Mitchell et al. showed that SOX21 was one of the six genes showing very low methylation in non-neoplastic colorectal tissue, suggesting their role as candidate biomarkers for stool-based assays in colorectal cancer [[175]43]. Here, DNMT1 was found to be responsible for the downregulation of SOX21 in GC. DNMT1 has been recently reported to be implicated in many malignancies, including pancreatic, breast, bladder cancers, and GC [[176]44]. For instance, DNMT1-mediated promoter hypermethylation was found to decrease the expression of protocadherin 10 in GC tissues and cells, while this study failed to validate the function of DNMT1 in GC [[177]45]. Lanka et al. described the discovery of novel DNMT1 inhibitors using fragment-based drug design, pharmacophore modeling, and molecular dynamics simulations, which showed promising binding affinities and drug-like properties, highlighting their potential as anticancer agents [[178]46]. Here, the downregulation of DNMT1 repressed the malignant aggressiveness of GC cells and tumor growth in mice, which was reversed by SOX21 as well. There are some limitations in this study. Firstly, although the utilization of extensive datasets, such as TCGA, furnishes pivotal insights into gene expression patterns and disease mechanisms, the analysis is susceptible to technical and biological biases, including tumor heterogeneity and sample purity [[179]47]. Moreover, liquid biopsy has emerged as a promising non-invasive method for early cancer detection and management, and DNMT1-mediated methylation changes could theoretically be captured through such non-invasive diagnostic platforms [[180]48]. Lastly, immunotherapy has achieved great progress recently, including circulating biomarkers in melanoma [[181]49], PD-1/PD-L1 inhibitors in GC [[182]50], and small extracellular vesicles in influencing immune responses [[183]51]. Considering that high CKS2 expression was significantly correlated with high PD-L1 expression and low CD8 positive rate in lung adenocarcinoma [[184]52], further validation of the function of CKS2 and its relationship with the immune responses in GC is necessary. Conclusion In summary, this present study illustrates that SOX21 was lower-expressed in GC tissues and can serve as a tumor-suppressor gene in GC. We find that SOX21, mediated by DNMT1-catalyzed DNA methylation, regulates the CKS2 transcription in GC. Therefore, developing DNMT1 or CKS2 inhibitors might be a new direction for GC patients. Supplementary Information [185]Supplementary Material 1.^ (80.5MB, pdf) Acknowledgements