NSUN2/m5C/TPI1 axis promotes glycolysis and immune evasion in breast cancer Authors: Xinming Song, Qihang Li, Ziteng Xiao, Haiqing Luo, Xiangyong Li Published: 2026 ABSTRACT BackgroundAerobic glycolysis and immune evasion drive tumor development. 5-methylcytosine (m5C) methyltransferase NSUN2 acts as an oncogene in breast cancer (BC), but the molecular mechanisms remain incompletely elucidated. This study aimed to investigate the role of NSUN2 in BC and the underlying mechanisms.MethodsIn vitro experiments were conducted to determine cell proliferation, glycolysis, and immune evasion. Tumor growth was evaluated using a tumor-bearing mouse model. The interplay among NSUN2, YBX1, and TPI1 was evaluated by RNA immunoprecipitation, methylated RNA immunoprecipitation, and dual-luciferase reporter assay.ResultsThe results showed that NSUN2 and TPI1 were highly expressed in BC cells. NSUN2 knockdown inhibited BC cell proliferation, glycolysis, and immune evasion, which was reversed by TPI1 overexpression. NSUN2 enhanced TPI1 stability by facilitating the m5C modification, which was recognized by YBX1. Moreover, overexpression of TPI1 reversed the inhibition of tumor growth, glycolysis, and immune cell infiltration induced by NSUN2 silencing in vivo.ConclusionsIn conclusion, NSUN2 accelerates the progression of BC by promoting glycolysis and immune evasion. Mechanistically, NSUN2 promotes m5C methylation of TPI1 in a YBX1-dependent manner, which is associated with increased PD-L1 expression and immune-evasive phenotypes. These findings suggest that NSUN2 may be an effective therapeutic target for BC.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12967-026-08390-w. Introduction Breast cancer (BC) is a significant global public health burden, with 2.26 million cases in 2020, and is a leading cause of cancer-related mortality among women [1]. Although BC mortality rate is declining in developed countries, it continues to increase in developing countries, with substantial disparities [2], which may be attributable to the lack of timely diagnosis and treatment [3]. A variety of therapeutic approaches have been maturely applied in the treatment of BC; however, advanced BC remains incurable [4]. Therefore, elucidating the pathogenesis of BC remains necessary to identify effective therapeutic strategies and targets. One of the hallmarks of cancer is metabolic reprogramming, especially glycolysis, which occurs under aerobic conditions, a phenomenon known as the “Warburg” effect. This metabolic alteration promotes tumor growth and chemoradiotherapy resistance in cancers, including BC [5, 6]. Additionally, blockade of immune checkpoints to promote T cell activation is a promising cancer treatment option [7]. Thus, immunotherapy is a potential strategy for BC treatment. Nevertheless, tumor cells often develop strategies to evade the immune system, a process termed immune escape, thereby promoting cancer progression [8]. Immune escape has limited the expected efficacy of immunotherapy in BC [9]. However, the mechanisms underlying glycolysis and immune evasion remain largely unknown. Epitranscriptomic modifications play crucial roles in regulating gene expression and are often dysregulated in cancer [10]. 5-methylcytosine (m5C) is an emerging RNA modification that has gained more attention. It has been indicated that m5C modification is commonly increased in cancers and promotes oncogene expression, accelerating tumor cell growth, invasion, and metastasis [11]. NSUN2 is a methyltransferase enzyme responsible for catalyzing m5C modification of tRNA, mRNA, miRNA, and long non-coding RNA [12]. NSUN2 is highly expressed in many malignancies, including BC, and is closely associated with BC tumorigenesis and development [13, 14]. Further studies are needed to fully elucidate the mechanism of NSUN2 action in BC. The present study aimed to elucidate the role of NSUN2 in BC, focusing on its impacts on glycolysis and immune evasion. Moreover, the epitranscriptomic regulation mediated by NSUN2 through m5C methylation was also investigated. Our study may identify a potential therapeutic target for BC intervention and treatment. Materials and methods Bioinformatic analysis The expression of NSUN2 in pan-cancer and BC, genes associated with NSUN2, and the expression of TPI1 in BC were predicted using the UALCAN online database (https://ualcan.path.uab.edu/). NSUN2-related genes were also obtained from the LinkedOmics (http://www.linkedomics.org) and UALCAN databases according to the following criteria: Pearson correlation coefficient |r|>0.3 and nominal p < 0.05. Multiple testing correction was performed using the Benjamini-Hochberg method when using the LinkedOmics database, and genes with a false discovery rate (FDR)<0.05 were considered statistically significant for downstream analysis. The overall survival of BC patients with high or low NSUN2 expression was predicted using the GEPIA database (http://gepia.cancer-pku.cn/) with a cutoff ranging from 25 to 75%. The prognostic significance of TPI1 in BC was forecasted using the Kaplan-Meier plotter analysis tool (https://kmplot.com/analysis/). KEGG enrichment analysis of NSUN2-related mRNAs was performed by the FUNRICH tool (http://www.funrich.org/). Gene Set Enrichment Analysis (GSEA; http://www.broad.mit.edu/gsea/) was performed to assess the relation between NSUN2 and the glycolysis/gluconeogenesis pathway. Potential m5C sites in TPI1 were acquired from the RNAm5Cfinder database (http://www.rnanut.net/rnam5cfinder/). To assess whether NSUN2 was related to immune infiltration, BC RNA-seq data were downloaded from TCGA database. Immune landscape analysis was performed by the CIBERSORT package (version 0.1.0) with 1000 permutations. Only samples with CIBERSORT output p < 0.05 were considered reliable for subsequent correlation analysis. The “corrplot” packages (version 0.92) were used to analyze the correlation between immune cells and NSUN2 expression, and the results were visualized using the “ggplot2” package (version 3.4.1). Tissue collection This study was approved by the Ethics Committee of Affiliated Hospital of Guangdong Medical University. Fifty patients with BC were diagnosed and subjected to surgery in our hospital. Tumor tissues and adjacent normal tissues were collected from these patients during surgery. After washing using the normal saline, tumors were frozen in liquid nitrogen and stored at −80 °C. Written informed consent was signed by each patient. According to the expression of NSUN2 or TPI1 in the tumor tissues, the patients were divided into NSUN2/TPI1 high-expression group (n = 25) and NSUN2/TPI1 low-expression group (n = 25). The clinical features are shown in Tables 1 and 2.Table 1Correlation analysis of NSUN2 expression level and clinical featuresSubjectsnLow (n = 25)High (n = 25)χ2p-valueAge (years)1.2820.258 <50241410 ≥50261115Tumor size5.1280.024* <50 mm26179 ≥50 mm24816Differentiation11.69<0.001** I/II28208 III/IV22517Tumor number2.9220.087 1281711 ≥222814Distant Metastasis8.3330.004** YES20515 NO302010Table 2Correlation analysis of TPI1 expression level and clinical featuresSubjectsnLow (n = 25)High (n = 25)χ2p-valueAge (years)0.3210.571 <50241311 ≥50261214Tumor size2.8850.089 <50 mm261610 ≥50 mm24915Differentiation15.91<0.001** I/II28217 III/IV22418Tumor number1.2990.255 1281612 ≥222913Distant Metastasis12.000<0.001** YES20416 NO30219 Table 1Correlation analysis of NSUN2 expression level and clinical featuresSubjectsnLow (n = 25)High (n = 25)χ2p-valueAge (years)1.2820.258 <50241410 ≥50261115Tumor size5.1280.024* <50 mm26179 ≥50 mm24816Differentiation11.69<0.001** I/II28208 III/IV22517Tumor number2.9220.087 1281711 ≥222814Distant Metastasis8.3330.004** YES20515 NO302010 Correlation analysis of NSUN2 expression level and clinical features Table 2Correlation analysis of TPI1 expression level and clinical featuresSubjectsnLow (n = 25)High (n = 25)χ2p-valueAge (years)0.3210.571 <50241311 ≥50261214Tumor size2.8850.089 <50 mm261610 ≥50 mm24915Differentiation15.91<0.001** I/II28217 III/IV22418Tumor number1.2990.255 1281612 ≥222913Distant Metastasis12.000<0.001** YES20416 NO30219 Correlation analysis of TPI1 expression level and clinical features Cell culture Human BC cell lines (MCF-7, MDA-MB-231, MDA-MB-468, BT-549, SK-BR-3, and T-47D) were purchased from American Type Culture Collection (ATCC; Manassas, VA, USA) and maintained in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Human normal mammary epithelial cell line (MCF-10A, ATCC) was maintained in DMEM/F12 supplemented with 10% FBS, 0.3 g/L L-glutamine, 20 ng/mL EGF, 10 μg/ml insulin, 500 μg/mL hydrocortisone, and 40 mg/L gentamicin as previously described [15]. Mouse BC cells 4T1 (ATCC) were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium supplemented with 10% FBS and 1% penicillin/streptomycin. All these cells were cultured at 37 °C with 5% CO2. Cell transfection Short hairpin RNA (sh)-NSUN2, sh-negative control (NC), sh-YBX1, overexpressing plasmids (oe-NSUN2, oe-TPI1, oe-ALYREF, oe-YBX1, oe-YBX2, oe-YTHDF2, oe-RAD52, and oe-FMRP) were synthesized by Genomeditech (Shanghai, China). MCF-7 and MDA-MB-231 cells in the logarithmic growth phase were seeded into six-well plates and cultured until 80% confluency. These cells were transfected with the above plasmids using Lipofectamine 2000 (Invitrogen, San Diego, CA, USA) for 48 h. In vivo study Sh-NSUN2 lentivirus (Lv-sh-NSUN2), sh-NC lentivirus (Lv-sh-NC), TPI1 overexpressing lentivirus (Lv-TPI1), and negative control lentivirus (Lv-NC) were acquired from Genomeditech. 4T1 cells were stably infected with these lentiviruses. These cells were suspended in PBS for subsequent use. The animal study was approved by the Ethics Committee of Affiliated Hospital of Guangdong Medical University (YJYS2025272). Female BALB/c mice (6 weeks old) were purchased from Shanghai SLAC Laboratory Animal Co., Ltd (Shanghai, China). All mice were randomly assigned into seven experimental groups: Lv-sh-NC, Lv-sh-NSUN2, Lv-sh-NC+Lv-NC, Lv-sh-NSUN2+Lv-NC, Lv-sh-NSUN2+Lv-TPI1, Lv-sh-NSUN2+Lv-TPI1+isotype control, and Lv-sh-NSUN2+Lv-TPI1+anti-CD8α, with 5 mice per group. The mice were subcutaneously injected with 200 µL cell suspension (5 × 106 cells) in the right flank. Tumor volume was measured once a week, lasting for 4 weeks. The formula: volume = 0.52 × length × width2 was used to calculate the volume. After the last test of tumor volume, all mice were sacrificed by inhaling isoflurane. Tumors were removed from mice, imaged using a digital camera, and weighed. Volume detection during tumor growth and analysis of experimental endpoints were performed by an investigator who was blinded to the group allocation. To assess whether tumor growth was related to CD8+ T cell-mediated immunological surveillance, after 4 days of tumor cell inoculation, 100 μg of rat anti-mouse CD8α mAb (ab22378, Abcam) or its isotype control (rat IgG2b; ab18541, Abcam) was injected intraperitoneally into the mice twice a week. Tumor volume and weight were measured as mentioned above. Immunohistochemistry (IHC) Tumor tissues of patients with BC and mice were fixed in 4% paraformaldehyde and embedded in paraffin. Paraffin sections (5 μm) were deparaffinized and dehydrated. For mouse tissue IHC, the sections were boiled in 0.1 M citrate buffer for antigen retrieval. For the block, the sections were incubated with 10% normal goat serum for 1 h. Primary antibodies were added to incubate with the sections overnight at 4 °C, and the secondary antibody was incubated with the sections for 30 min, followed by 3,3′-diaminobenzidine staining. Images were captured using a light microscope. Antibodies were listed below: anti-NSUN2 (ab259941, 1:200, Abcam, Cambridge, MA, USA), anti-Ki67 (ab15580, 1:500, Abcam), anti-CD8A (ab217344, 1:2000, Abcam), anti-TPI1 (ab196618, 1:200, Abcam), anti-GLUT1 (ab115730, 1:250, Abcam), anti-LDHB (ab52488, 1:2000, Abcam), anti-HIF1α (ab114977, 1:500, Abcam), anti-PD-L1 (PA5-20343, 1:100, Invitrogen), and horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (ab6721, 1:1000, Abcam). For human tissue IHC, the experiment was performed using the PD-L1 clone 22C3 pharmDx kit (Dako/Agilent, Santa Clara, CA, USA) on the Autostainer Link 48 platform (Dako/Agilent) according to the manufacturer’s instructions. PD-L1 expression in human tumor tissues was quantified as follows: Proportion scores were noted as 0, 1, 2, 3, 4, representing < 5%, 5–25%, 26–50%, 51–75%, and >75%. Intensity scores were noted as 0, 1, 2, 3, indicating negative, weak, moderate, and strong staining. IHC scores was calculated as follows: proportion score × intensity score. Quantitative real-time polymerase chain reaction (qPCR) Total RNA from BC cells and tumor tissues was extracted using the TRIzol reagent (Invitrogen). RNA concentration was detected using the NanoDrop 2000 system. cDNA was prepared using the HiScript IV RT SuperMix for qPCR (+gDNA wiper) (Vazyme, Nanjing, China). qPCR was performed using SupRealQ Purple Universal SYBR qPCR Master Mix (U+) on the CFX96 Real-time PCR system. Thermal cycling conditions were 95 °C for 30 s, 40 cycles consisting of 95 °C for 10 s and 60 °C for 30 s. Relative mRNA expression levels were calculated using the 2-ΔΔCt method. Data were normalized to internal control β-actin expression. Cell counting kit-8 (CCK-8) MCF-7 and MDA-MB-231 cells in 96-well plates were incubated with 10 μL CCK-8 solution (Dojindo, Kumamoto, Japan) at 37 °C for 4 h. Afterwards, the absorbance was detected at 450 nm using a microplate reader. EdU assay The BeyoClick™ EdU cell proliferation kit with Alexa Fluor 594 (Beyotime, Shanghai, China) was used to assess cell proliferation. BC cells in six-well plates and incubated with 10 μM EdU working solution for 3 h. After removing the culture solution, 4% paraformaldehyde was used to fix the cells, and then 0.3% Triton X-100 was used for permeabilization. Click reaction solution was prepared and incubated with the cells in the dark for 30 min. The cell nucleus was stained with 4’, 6-diaminyl-2-phenylindole (DAPI) for 10 min in the dark. The fluorescence signals were captured under a laser-scanning confocal microscope. Determination of glucose consumption and lactate concentration glucose consumption was measured using the glucose uptake-Glo assay kit (Promega), and lactate concentration was measured using the lactate-Glo assay kit (Promega). The procedures were performed according to the manufacturer’s instructions. Determination of extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) BC cells were seeded in 24-well XF Cell culture plates (Seahorse Bioscience, Billerica, MA, USA). After adding corresponding reagents, ECAR and OCR were automatically determined using the Seahorse XFe24 analyzer (Seahorse Bioscience). Cell co-culture BC cells were seeded into 24-well plates. The activated human peripheral blood CD8+ T cells (Delf, Hefei, China) were seeded into the same plates at an effector-to-target ratio of 5:1. The cells were co-cultured for 16 h as previously described [16]. Measurement of the cytotoxic rate The lactate dehydrogenase (LDH) cytotoxicity assay kit (Beyotime) was used to evaluate the toxicity of CD8+ T cells on BC cells. The co-culture medium was removed, and BC cells were incubated with the LDH release reagent for 1 h. After centrifuging at 400 g for 5 min, the supernatant (120 μL) was collected and incubated with LDH detection solution for 30 min in the dark. The absorbance was measured at 490 nm using a microplate reader. Immunofluorescence (IF) To detect PD-L1 expression, BC cells were fixed with 4% paraformaldehyde for 0.5 h and penetrated by 0.5% Triton X-100 for 15 min. After blocking with normal goat serum, the cells were incubated with primary antibody against PD-L1 (ab279292, 1:100, Abcam) overnight at 4 °C, and incubated with Alexa Fluor 594-conjugated goat anti-mouse IgG (ab150116, 1:200, Abcam) for 1 h. After 4’, 6-diaminyl-2-phenylindole (DAPI) staining of the nucleus, fluorescence signals were captured under a laser scanning confocal microscope. IF and fluorescence in situ hybridization (FISH) were performed to observe the localization of NSUN2 and TPI1 in BC cells, respectively. IF was conducted as mentioned above. Antibodies were used as follows: anti-NSUN2 (ab259941, 1:1000, Abcam) and Alexa Fluor 594-conjugated goat anti-rabbit IgG (ab150080, 1:500, Abcam). After IF, the cells were washed with PBS and fixed with 2% paraformaldehyde. The cells were hybridized with a specific fluorescently labeled TPI1 mRNA probe at 80 °C for 10 min and immediately transferred to 37 °C overnight. After washing and sealing, the cells were observed under the laser scanning confocal microscope. Western blot Proteins were extracted from BC cells using radio immunoprecipitation assay (RIPA) reagent. After detecting protein concentration using a BCA kit (Beyotime), Proteins were separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The following antibodies were used: anti-NSUN2 (ab259941, 1:1000, Abcam), anti-PD-L1 (ab205921, 1:1000, Abcam), anti-β-actin (ab8227, 1:1000, Abcam), and HRP-conjugated goat anti-rabbit IgG (ab6721, 1:5000, Abcam). Protein bands were visualized using SuperFemto ECL Master Mix (Vazyme). RNA immunoprecipitation (RIP) RIP kit was purchased from BersinBio (Guangzhou, China). BC cell samples were lysed using RIP lysis buffer. After removing DNA using DNase, the lysate was collected. The lysate was incubated with antibodies targeting NSUN2, YBX1, and IgG overnight at 4 °C, and protein A/G magnetic beads were added for further incubation for 1 h. After washing, the magnetic beads were collected, and RNA was isolated for mRNA expression detection. M5C methylated RIP (MeRIP) The M5C MeRIP kit was purchased from BersinBio. BC cells were lysed by TRIzol, and total RNA was isolated. Next, RNA was broken into fragments of about 300 nt in length. The RNA samples were incubated with 4 μg m5C or IgG antibody at 4 °C for 4 h, followed by incubation with protein A/G magnetic beads at 4 °C for 1 h. The magnetic beads were collected, and RNA was isolated for TPI1 expression detection. Dual-luciferase reporter assay Wild-type (wt) TPI1 sequences and the corresponding mutated sequences at the m5C sites were cloned into the pGL3 basic vector (Promega) to construct reporter plasmids. BC cells were transfected with the reporter plasmids, pRL-TK vector (Promega), and sh-NSUN2, sh-YBX1, or YBX1 overexpression plasmids using Lipofectamine 2000. The dual luciferase reporter gene assay kit was used to measure firefly and Renilla luciferase activities. Actinomycin D (ACD) treatment BC cells after transfection were treated with 2 μg/mL ACD for 0, 4, 8, and 12 h. Total RNA was isolated for TPI1 expression detection. Flow cytometry Tumors from mice were digested using collagenase IV (1 mg/mL) and DNase I (20 μg/mL) at 37 °C for 40 min. The single-cell suspension was collected by centrifugation at 300 g for 5 min after filtration through a 70 μm cell sieve. The cells (1 × 106) were incubated with 1 μg of anti-CD16/CD32 at 4 °C for blocking. Then, the cells were incubated with APC-Cy7-conjugated anti-CD45 (103115, Biolegend), FITC-conjugated anti-CD3 (100203, Biolegend), APC-conjugated CD4 (100412, Biolegend), and BV421-conjugated CD8 (301036, Biolegend) at 4 °C for 30 min. To assess activated CD8+ T cells, after CD8 antibody incubation, the cells were incubated with the fixation/permeabilization buffer and then incubated with PE-conjugated anti-IFN-γ (505807, Biolegend) or PE-conjugated granzyme B (372208, Biolegend). To evaluate macrophages, following CD45 antibody incubation, the cells were incubated with PerCP-Cy5.5-conjugated anti-CD11b (101227, Biolegend) and APC-conjugated anti-F4/80 (123116, Biolegend). The stained cells were detected using a flow cytometer (Beckman Coulter, Miami, FL, USA). Singlet cells were gated using FSC-A vs FSC-H, and live cells were gated according to the viability dye-negative population. CD45+ cell gate analysis was used to identify leukocyte populations. CD3+ T cells were identified within CD45+ cells. CD4+ and CD8+ T cells were gated from the CD3+ population. IFN-γ+ and granzyme B+ populations were gated from CD8+ T cells, while macrophages were identified as CD45+CD11b+F4/80+ cells. Isotype-matched immunoglobulins were used as controls. Statistical analysis The GraphPad Prism 7.0 software was used for statistical analysis, and the results were expressed as means ± standard deviation (SD). Differences were compared using Student’s t-test and one-way ANOVA followed by Tukey’s post hoc test. Correlation between PD-L1 and NSUN2 or TPI1 expression was assessed using Spearman’s rank correlation test. p < 0.05 indicates statistical significance. All in vitro experiments were performed with three independent biological replicates, while in vivo experiments were performed with five biological replicates. Results NSUN2 expression is increased in BC To investigate whether NSUN2 is involved in BC progression, its expression in pan-cancer. As shown in Fig. 1A, NSUN2 expression was upregulated in various cancers, including BC. Figure 1B also illustrated that NSUN2 was highly expressed in BC. Additionally, NSUN2 expression was predicted at different stages of BC. The results showed that the expression of NSUN2 was highest in stage 2 (Fig. 1C). Next, the overall survival of patients with BC was predicted, and the results showed that compared with the low NSUN2 expression group, patients with high NSUN2 expression had worse survival (Fig. 1D), indicating that high NSUN2 expression is associated with poor survival.Fig. 1NSUN2 expression is increased in BC. (A) expression of NSUN2 in pan-cancer. (B) expression of NSUN2 in normal and tumor samples in BC. (C) expression of NSUN2 in different stages of BC. (D) overall survival of patients with BC with high or low NSUN2 expression. (E) the relationship between NSUN2 expression and immune infiltration. (F) qPCR and (G, H) western blotting were conducted to determine NSUN2 levels in tumor and adjacent normal tissues from patients with BC. Data represent n = 50 individual patient samples. (I) NSUN2 mRNA expression and (J, K) protein levels were detected using qPCR and western blotting in MCF-10A cells and BC cell lines (MCF-7, MDA-MB-231, MDA-MB-468, BT-549, SK-BR-3, and T-47D), respectively. Data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001. **p < 0.01. *p < 0.05 NSUN2 expression is increased in BC. (A) expression of NSUN2 in pan-cancer. (B) expression of NSUN2 in normal and tumor samples in BC. (C) expression of NSUN2 in different stages of BC. (D) overall survival of patients with BC with high or low NSUN2 expression. (E) the relationship between NSUN2 expression and immune infiltration. (F) qPCR and (G, H) western blotting were conducted to determine NSUN2 levels in tumor and adjacent normal tissues from patients with BC. Data represent n = 50 individual patient samples. (I) NSUN2 mRNA expression and (J, K) protein levels were detected using qPCR and western blotting in MCF-10A cells and BC cell lines (MCF-7, MDA-MB-231, MDA-MB-468, BT-549, SK-BR-3, and T-47D), respectively. Data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001. **p < 0.01. *p < 0.05 Next, we analyzed the relationship between NSUN2 expression and immune infiltration. As shown in Fig. 1E, NSUN2 expression had a moderate positive correlation with M0 macrophages and a moderate negative correlation with CD8+ T cells, suggesting that NSUN2 may regulate immune infiltration. Then, we collected tumor and adjacent normal tissues from 50 patients with BC. NSUN2 expression was found to be higher in the tumor tissues than that in the paired normal tissues (Fig. 1F–H). Patients with high NSUN2 were related to large tumor size, late differentiation, and distant metastasis (Table 1). We also cultured BC cell lines and detected NSUN2 expression. We found that NSUN2 mRNA expression was higher in MCF-7, MDA-MB-231, BT-549, SK-BR-3, and T-47D cell lines, compared with that in MCF-10A cells, while there was no significant difference between MDA-MB-468 and MCF-10A cells (Fig. 1 and J). The protein levels of NSUN2 in BC cell lines were consistent with its mRNA expression (Fig. 1K). The results demonstrated that NSUN2 is highly expressed in BC. Knockdown of NSUN2 suppresses proliferation, glycolysis, and immune evasion of BC cells We used in vitro experiments to confirm the function of NSUN2. MCF-7 and MDA-MB-231 cells were chosen and transfected with sh-NC and sh-NSUN2. The expression of NSUN2 was decreased following sh-NSUN2 transfection (Fig. 2A). Cellular proliferation was evaluated using CCK-8 and EdU assay. We found that NSUN2 knockdown inhibited BC cell proliferation (Fig. 2B and C). Additionally, knockdown of NSUN2 reduced glucose consumption (Fig. 2D), lactate concentration (Fig. 2E), and ECAR (Fig. 2F and G), as well as increased OCR (Fig. 2H and I). qPCR results showed that NSUN2 knockdown downregulated GLUT1, LDHB, and HK2 expression (Fig. 2J–L). Moreover, BC cells were co-cultured with CD8+ T cells, and immune escape was evaluated. The cytotoxic rate of BC cells induced by CD8+ T cells was increased after NSUN2 knockdown (Fig. 2M). IF, qPCR, and western blotting results showed that NSUN2 knockdown reduced the levels of PD-L1 in BC cells (Fig. 2N–R). These findings suggest that silencing of NSUN2 inhibited BC cell proliferation, glycolysis, and immune evasion.Fig. 2Knockdown of NSUN2 suppresses proliferation, glycolysis, and immune evasion of BC cells. (A) BC cells were transfected with sh-NC and sh-NSUN2, and NSUN2 expression was detected by qPCR. (B) BC cell viability was assessed using CCK-8. (C) EdU assay was performed to assess the proliferation capability of BC cells. (D) glucose consumption and (E) lactate concentration were tested using commercial kits. (F, G) ECAR and (H, I) OCR were measured using Seahorse analysis. (J) GLUT1, (K) LDHB, and (L) HK2 expression was detected using qPCR. (M) the cytotoxic rate of BC cells was evaluated using the LDH assay kit. (N) qPCR, (O, P) western blotting, and (Q, R) IF assay were performed to determine the levels of PD-L1 in BC cells. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001 Knockdown of NSUN2 suppresses proliferation, glycolysis, and immune evasion of BC cells. (A) BC cells were transfected with sh-NC and sh-NSUN2, and NSUN2 expression was detected by qPCR. (B) BC cell viability was assessed using CCK-8. (C) EdU assay was performed to assess the proliferation capability of BC cells. (D) glucose consumption and (E) lactate concentration were tested using commercial kits. (F, G) ECAR and (H, I) OCR were measured using Seahorse analysis. (J) GLUT1, (K) LDHB, and (L) HK2 expression was detected using qPCR. (M) the cytotoxic rate of BC cells was evaluated using the LDH assay kit. (N) qPCR, (O, P) western blotting, and (Q, R) IF assay were performed to determine the levels of PD-L1 in BC cells. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001 Silencing of NSUN2 inhibits tumor growth in vivo Genes that are dysregulated in tumors are often involved in the progression of cancer. Here, we analyzed whether NSUN2 affected BC progression using a tumor-bearing mouse model. We found that knockdown of NSUN2 reduced tumor weight and volume (Fig. 3A–C). Glycolysis provides nutrients and energy for tumor growth. Thus, we measured several glycolysis-related factors and found that silencing of NSUN2 downregulated the expression of GLUT1, LDHB, and HK2 (Fig. 3D). Additionally, IHC was carried out to measure the levels of NSUN2, Ki67 (proliferation-related factor), and CD8A (T-cell surface glycoprotein) in tumors. The results showed that NSUN2 knockdown decreased NSUN2 and Ki67 levels; however, CD8A levels were elevated by NSUN2 knockdown (Fig. 3E–H). The results showed that silencing of NSUN2 inhibited tumor growth, which is associated with inhibition of glycolysis, proliferation, and immune response.Fig. 3Silencing of NSUN2 inhibits tumor growth in vivo. (A) images of tumors from the tumor-bearing mouse model in each group. (B) tumors were weighted. (C) tumor volume was detected every week during the tumor growth process. (D) the mRNA levels of glycolysis-related factors GLUT1, LDHB, and HK2 were measured using qPCR. (E-H) the levels of NSUN2, Ki67, and CD8A in tumor tissues were measured using IHC. Scale bar = 100 μm. All data are presented as mean ± SD from n = 5 mice per group. ***p < 0.001 Silencing of NSUN2 inhibits tumor growth in vivo. (A) images of tumors from the tumor-bearing mouse model in each group. (B) tumors were weighted. (C) tumor volume was detected every week during the tumor growth process. (D) the mRNA levels of glycolysis-related factors GLUT1, LDHB, and HK2 were measured using qPCR. (E-H) the levels of NSUN2, Ki67, and CD8A in tumor tissues were measured using IHC. Scale bar = 100 μm. All data are presented as mean ± SD from n = 5 mice per group. ***p < 0.001 The expression of NSUN2-related TPI1 is upregulated in BC To explore the underlying molecular mechanism, we screened the mRNAs associated with NSUN2 using the LinkedOmics database. Numerous mRNAs were found to be positively or negatively correlated with NSUN2 (Fig. 4A–C). We also used the UALCAN database to predict genes positively associated with NSUN2. The Venn diagram showed a total of 394 mRNAs were positively associated with NSUN2, which was predicted using both the UALCAN and LinkedOmics databases (Fig. 4D). KEGG enrichment analysis indicated that these mRNAs are related to several pathways, such as cellular senescence, glycolysis/gluconeogenesis, DNA replication, and p53 signaling pathway (Fig. 4E). We focused on the glycolysis/gluconeogenesis since we explored the function of NSUN2 from the perspective of glycolysis, and GSEA analysis showed that the glycolysis/gluconeogenesis gene set was significantly enriched in samples with high expression of NSUN2 (Fig. 4F). Only TPI1 was enriched in the glycolysis pathway, which was chosen for the follow-up studies. Pearson correlation analysis results showed that NSUN2 and TPI1 had a positive correlation (Fig. 4G). TPI1 expression was increased in BC tissue samples (Fig. 4H) and increased with disease stage progression (Fig. 4I), compared with the normal samples. Kaplan-Meier plotter analysis showed that patients with high TPI1 expression were associated with a worse prognosis (Fig. 4J). Next, TPI1 expression was measured in BC cell lines, and the results showed that TPI1 was highly expressed in these BC cells, including MCF-7, MDA-MB-231, MDA-MB-468, BT-549, SK-BR-3, and T-47D, compared with normal MCF-10A cells (Fig. 4K). Additionally, high TPI1 expression was observed in BC tumor samples (Fig. 4L). High TPI1 expression was related to tumor differentiation and distant metastasis (Table 2). The results demonstrate that TPI1 expression was elevated in BC, and suggest that NSUN2 may regulate TPI1 to regulate glycolysis in BC cells.Fig. 4The expression of NSUN2-related TPI1 is upregulated in BC. (A) NSUN2-associated gene results from the LinkedOmics database. (B) genes positively correlated to NSUN2 from the LinkedOmics database. (C) genes negatively correlated to NSUN2 from the LinkedOmics database. (D) mRNAs positively associated with NSUN2 were predicted using both the UALCAN and LinkedOmics tools and shown using the venn diagram. (E) KEGG enrichment analysis revealed the pathways related to NSUN2-associated mRNAs. (F) GSEA analysis was used to evaluate the relation between NSUN2-correlated mRNAs and glycolysis/gluconeogenesis. (G) the correlation between NSUN2 and TPI1 was predicted using the Pearson analysis through the LinkedOmics database. (H) expression of TPI1 in normal and tumor samples in BC. (I) expression of TPI1 in different stages of BC. (J) the prognosis of patients with BC with high or low TPI1 expression was predicted using the Kaplan-Meier plotter online tool. (K) TPI1 mRNA expression was detected using qPCR in MCF-10A cells and BC cell lines (MCF-7, MDA-MB-231, MDA-MB-468, BT-549, SK-BR-3, and T-47D). Data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). (L) TPI1 expression in BC tumors and adjacent normal tissues. Data represent n = 50 individual patient samples. ***p < 0.001. **p < 0.01. *p < 0.05 The expression of NSUN2-related TPI1 is upregulated in BC. (A) NSUN2-associated gene results from the LinkedOmics database. (B) genes positively correlated to NSUN2 from the LinkedOmics database. (C) genes negatively correlated to NSUN2 from the LinkedOmics database. (D) mRNAs positively associated with NSUN2 were predicted using both the UALCAN and LinkedOmics tools and shown using the venn diagram. (E) KEGG enrichment analysis revealed the pathways related to NSUN2-associated mRNAs. (F) GSEA analysis was used to evaluate the relation between NSUN2-correlated mRNAs and glycolysis/gluconeogenesis. (G) the correlation between NSUN2 and TPI1 was predicted using the Pearson analysis through the LinkedOmics database. (H) expression of TPI1 in normal and tumor samples in BC. (I) expression of TPI1 in different stages of BC. (J) the prognosis of patients with BC with high or low TPI1 expression was predicted using the Kaplan-Meier plotter online tool. (K) TPI1 mRNA expression was detected using qPCR in MCF-10A cells and BC cell lines (MCF-7, MDA-MB-231, MDA-MB-468, BT-549, SK-BR-3, and T-47D). Data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). (L) TPI1 expression in BC tumors and adjacent normal tissues. Data represent n = 50 individual patient samples. ***p < 0.001. **p < 0.01. *p < 0.05 PD-L1 expression positively correlates with NSUN2 or TPI1 in BC To assess the clinical relationship between PD-L1 and either NSUN2 or TPI1, we evaluated PD-L1 protein expression using IHC in 50 BC tumor tissues. PD-L1 negative or positive expression results are shown in Fig. S1A. After quantifying the IHC results, Spearman’s rank correlation analysis was performed to assess the correlation. The results showed that PD-L1 IHC scores were positively correlated with NSUN2 mRNA expression (r = 0.567, p < 0.001) and with TPI1 mRNA expression (R = 0.679, p < 0.001) (Fig. S1B and C). These data support a clinical association between the NSUN2/TPI1 axis and PD-L1 expression in BC tissues. Overexpression of TPI1 promotes BC cell proliferation, glycolysis, and immune evasion We then evaluated the impact of TPI1 on cellular functions. TPI1 overexpression plasmids were transfected into BC cells, and its expression was upregulated (Fig. 5A). Overexpression of TPI1 enhanced cell viability (Fig. 5B) and increased EdU positive cells (Fig. 5C). In terms of glycolysis, TPI1 overexpression elevated glucose consumption (Fig. 5D), LA concentration (Fig. 5E), and ECAR (Fig. 5Fand G), but reduced OCR (Fig. 5Hand I). The expression of GLUT1, LDHB, and HK2 was upregulated by TPI1 overexpression (Fig. 5J–L). In terms of immune evasion, TPI1 reduced the cytotoxic effect of CD8+ T cells on BC cells (Fig. 5M). Furthermore, the results of qPCR, western blot, and IF showed that overexpression of TPI1 elevated PD-L1 expression in MCF-7 and MDA-MB-231 cells (Fig. 5N–R). The results showed that TPI1 functions as an oncogene in BC.Fig. 5Overexpression of TPI1 promotes BC cell proliferation, glycolysis, and immune evasion. (A) BC cells were transfected with oe-NC and oe-TPI1, and TPI1 expression was detected by qPCR. (B) BC cell viability was assessed using CCK-8. (C) EdU assay was performed to assess the proliferation capability of BC cells. (D) glucose consumption and (E) lactate concentration were tested using commercial kits. (F, G) ECAR and (H, I) OCR were measured using Seahorse analysis. (J) GLUT1, (K) LDHB, and (L) HK2 expression was detected using qPCR. (M) the cytotoxic rate of BC cells was evaluated using the LDH assay kit. PD-L1 levels were detected using (N) qPCR, (O, P) western blotting, and (Q, R) IF. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001 Overexpression of TPI1 promotes BC cell proliferation, glycolysis, and immune evasion. (A) BC cells were transfected with oe-NC and oe-TPI1, and TPI1 expression was detected by qPCR. (B) BC cell viability was assessed using CCK-8. (C) EdU assay was performed to assess the proliferation capability of BC cells. (D) glucose consumption and (E) lactate concentration were tested using commercial kits. (F, G) ECAR and (H, I) OCR were measured using Seahorse analysis. (J) GLUT1, (K) LDHB, and (L) HK2 expression was detected using qPCR. (M) the cytotoxic rate of BC cells was evaluated using the LDH assay kit. PD-L1 levels were detected using (N) qPCR, (O, P) western blotting, and (Q, R) IF. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001 NSUN2 stabilizes TPI1 through YBX1-recognized m5C methylation NSUN2 is a methyltransferase that catalyzes m5C methylation. Hence, we investigated whether NSUN2 affected m5C modification of TPI1. RIP results showed that NSUN2 interacted with TPI1 (Fig. 6A). In addition, knockdown of NSUN2 reduced TPI1 mRNA expressed in BC cells (Fig. 6B). IF-FISH results showed that NSUN2 was mainly located in the nucleus but also distributed in the cytoplasm, while TPI1 mainly located in the cytoplasm (Fig. 6C), further demonstrating that NSUN2 could interact with TPI1. Moreover, knockdown of NSUN2 decreased m5C levels of TPI1 (Fig. 6D). m5C methylation sites in TPI1 were predicted using the RNAm5Cfinder database, and many sites were found (Fig. 6E). Three sites, 193, 273, and 962 sites with the highest possibility scores, were chosen for further study. Dual-luciferase reporter assay showed that compared with the sh-NC, NSUN2 knockdown only reduced the luciferase activity when co-transfected with site 962 wt reporter plasmids in MCF-7 cells, but did not change those in sites 193 and 273 wt group (Fig. 6F–H). The findings suggested that site 962 is a key functional m5C-responsive site candidate in TPI1, which was similarly observed in the MDA-MB-231 cells (Fig. 6I). Furthermore, NSUN2 knockdown reduced the mRNA stability of TPI1 (Fig. 6Jand K). In short, knockdown of NSUN2 suppresses m5C modification of TPI1 and reduces its mRNA stability.Fig. 6NSUN2 stabilizes TPI1 through YBX1-recognized m5C methylation. (A) RIP assay was performed to assess the interaction between NSUN2 and TPI1. (B) qPCR was conducted to detect TPI1 expression after NSUN2 knockdown. (C) the localization of NSUN2 and TPI1 in BC cells was observed using IF-FISH. (D) the effect of NSUN2 on m5C methylation of TPI1 was assessed using MeRIP. (E) m5C sites in TPI1 were predicted using the RNAm5Cfinder database. (F-H) the candidate m5C sites in TPI1 were evaluated using the luciferase reporter assay in MCF-7 cells. (I) the candidate m5C site 962 in TPI1 was evaluated using the luciferase reporter assay in MDA-MB-231 cells. (J, K) the effect of NSUN2 knockdown on TPI1 mRNA stability was measured after ACD treatment. (L, M) the successful transfection of overexpressing plasmids was detected using qPCR. (N, O) the regulation of overexpressed readers on TPI1 expression was measured using qPCR. (P) the successful transfection of sh-YBX1 was detected using qPCR. (Q) the regulation of YBX1 on TPI1 expression was measured using qPCR. (R) the interaction between NSUN2 and TPI1 was evaluated using RIP. (S-V) the binding relationship between YBX1 and TPI1 was evaluated using dual-luciferase reporter assay. (W) the effect of NSUN2 overexpression and YBX1 knockdown on YBX1 expression was measured by qPCR. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001. **p < 0.01. *p < 0.05 NSUN2 stabilizes TPI1 through YBX1-recognized m5C methylation. (A) RIP assay was performed to assess the interaction between NSUN2 and TPI1. (B) qPCR was conducted to detect TPI1 expression after NSUN2 knockdown. (C) the localization of NSUN2 and TPI1 in BC cells was observed using IF-FISH. (D) the effect of NSUN2 on m5C methylation of TPI1 was assessed using MeRIP. (E) m5C sites in TPI1 were predicted using the RNAm5Cfinder database. (F-H) the candidate m5C sites in TPI1 were evaluated using the luciferase reporter assay in MCF-7 cells. (I) the candidate m5C site 962 in TPI1 was evaluated using the luciferase reporter assay in MDA-MB-231 cells. (J, K) the effect of NSUN2 knockdown on TPI1 mRNA stability was measured after ACD treatment. (L, M) the successful transfection of overexpressing plasmids was detected using qPCR. (N, O) the regulation of overexpressed readers on TPI1 expression was measured using qPCR. (P) the successful transfection of sh-YBX1 was detected using qPCR. (Q) the regulation of YBX1 on TPI1 expression was measured using qPCR. (R) the interaction between NSUN2 and TPI1 was evaluated using RIP. (S-V) the binding relationship between YBX1 and TPI1 was evaluated using dual-luciferase reporter assay. (W) the effect of NSUN2 overexpression and YBX1 knockdown on YBX1 expression was measured by qPCR. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001. **p < 0.01. *p < 0.05 M5C methylation is recognized by “readers”. We then explored “readers” who might identify the TPI1 m5C site. We chose several readers and transfected their overexpressing plasmids into BC cells, and the results of qPCR confirmed the successful transfection (Fig. 6Land M). Next, we found that the expression of TPI1 was upregulated after overexpressing ALYREF and YBX1, and was downregulated by YTHDF2 overexpression (Fig. 6N), while overexpression of YBX1 and YBX2 increased TPI1 expression in MDA-MB-231 cells (Fig. 6O). Therefore, YBX1 was selected for further study. Silencing of YBX1 decreased TPI1 expression in BC cells (Fig. 6Pand Q). In addition, YBX1 interacted with TPI1 (Fig. 6R). The luciferase activity was increased after overexpression of YBX1 and was reduced after YBX1 knockdown in the TPI1 wt group (Fig. 6S–V), verifying their binding relationship. Overexpression of NSUN2 elevated TPI1 expression, which was decreased by YBX1 knockdown (Fig. 6W). In summary, YBX1 is a “reader” that recognizes m5C modification of TPI1 mediated by NSUN2. The promotion of glycolysis and immune evasion induced by TPI1 depends on YBX1 The functions of YBX1 were evaluated using rescue experiments. We overexpressed TPI1 and knocked down YBX1 in BC cells. Cell viability and EdU positive cells were increased after TPI1 overexpression, which was reversed by YBX1 silencing (Fig. S2A-C). In addition, knockdown of YBX1 abrogated the elevation of decreased glucose consumption, LA concentration, and ECAR, and the reduction of OCR induced by TPI1 (Fig. S2D-I). The upregulation of GLUT1, LDHB, and HK2 expression caused by TPI1 overexpression was counteracted after YBX1 knockdown (Fig. S2J-L). Furthermore, overexpression of TPI1 reduced the cytotoxic effect of CD8+ T cells on BC cells and upregulated PD-L1 in BC cells, whereas knockdown of YBX1 rescued these effects induced by TPI1 (Fig. S2M-R). In short, the promoting effect of TPI1 in BC depends on YBX1. NSUN2 knockdown inhibits tumor growth by downregulating TPI1 expression The role of TPI1 in vivo was investigated in the absence of NSUN2. We found that overexpression of TPI1 abrogated the inhibition of tumor weight and volume induced by NSUN2 knockdown (Fig. 7A–C). Additionally, the expression of GLUT1, LDHB, and HK2 downregulated by NSUN2 knockdown was counteracted by TPI1 overexpression (Fig. 7D), suggesting that TPI1 promotes glycolysis in tumors. Then, we conducted flow cytometry to evaluate the tumor-infiltrating immune cells. As shown in Fig. 7E, we successively screened out immune cells, CD45 cells, and CD3 cells through a gating strategy. CD4+ T cells and macrophages were not obviously affected by NSUN2 or TPI1. Knockdown of NSUN2 increased CD8+ T cells, which was counteracted by TPI1 overexpression. Activation of CD8+ T cells was monitored by IFN-γ and granzyme B levels. We found that silencing of NSUN2 increased the percentage of IFN-γ+CD8+ T or granzyme B+CD8+ T cells, whereas TPI1 reversed the effects on these cells induced by NSUN2 silence. Besides, IHC results showed that NSUN2 knockdown decreased the levels of TPI1, Ki67, GLUT1, LDHB, HIF1α, and PD-L1, as well as increased CD8A expression, while TPI1 overexpression rescued their levels modulated by NSUN2 knockdown (Fig. 7F), Together, the results suggested that silencing of NSUN2 inhibited tumor proliferation, glycolysis, and immune evasion in vivo by decreasing TPI1 expression, thereby suppressing tumor growth.Fig. 7NSUN2 knockdown inhibits tumor growth by downregulating TPI1 expression. (A) images of tumors from the tumor-bearing mouse model in each group. (B) tumors were weighed. (C) tumor volume was detected every week during the tumor growth process. (D) the mRNA levels of glycolysis-related factors GLUT1, LDHB, and HK2 were measured using qPCR. (E) the effects of NSUN2 and TPI1 on the levels of CD4+ T cells (CD4+CD3+ T cells), CD8+ T cells (CD8+CD3+ T cells), activated CD8+ T cells (IFNγ+CD8+ T cells and granzyme B+CD8+ T cells), and macrophages (F4/80+CD11b+CD45+ cells) in tumors were analyzed by flow cytometry. (F) the levels of TPI1, Ki67, CD8A, GLUT1, LDHB, HIF1α, and PD-L1 in tumor tissues were measured using IHC. Scale bar = 100 μm. All data are presented as mean ± SD from n = 5 mice per group. ***p < 0.001. ###p < 0.001 NSUN2 knockdown inhibits tumor growth by downregulating TPI1 expression. (A) images of tumors from the tumor-bearing mouse model in each group. (B) tumors were weighed. (C) tumor volume was detected every week during the tumor growth process. (D) the mRNA levels of glycolysis-related factors GLUT1, LDHB, and HK2 were measured using qPCR. (E) the effects of NSUN2 and TPI1 on the levels of CD4+ T cells (CD4+CD3+ T cells), CD8+ T cells (CD8+CD3+ T cells), activated CD8+ T cells (IFNγ+CD8+ T cells and granzyme B+CD8+ T cells), and macrophages (F4/80+CD11b+CD45+ cells) in tumors were analyzed by flow cytometry. (F) the levels of TPI1, Ki67, CD8A, GLUT1, LDHB, HIF1α, and PD-L1 in tumor tissues were measured using IHC. Scale bar = 100 μm. All data are presented as mean ± SD from n = 5 mice per group. ***p < 0.001. ###p < 0.001 The effects of NSUN2/TPI1 on tumor growth depend on CD8+ T cells To assess whether immune evasion is involved in tumor growth, we injected anti-CD8α into the tumor-bearing mice to deplete CD8+ T cells. The results showed that as compared with the isotype control group, anti-CD8α decreased tumor weight and volume (Fig. S3A-C). Collectively, inhibition of CD8+ T-cell infiltration promoted tumor growth caused by NSUN2 silencing and TPI1 overexpression. NSUN2 knockdown inhibits BC cell proliferation, glycolysis, and immune evasion by decreasing TPI1 expression Rescue experiments were carried out to explore the role of TPI1 in vitro in the absence of NSUN2. Cell proliferation suppressed by NSUN2 knockdown was counteracted by TPI1 overexpression (Fig. 8Aand B). Glucose consumption, lactate concentration, and ECAR were reduced, and OCR was increased by NSUN2 knockdown, whereas TPI1 overexpression abolished the effects induced by NSUN2 silence (Fig. 8C–H). Besides, TPI1 increased GLUT1, LDHB, and HK2 expression in sh-NSUN2 transfected cells (Fig. 8I–K). Immune evasion was further evaluated. We found that silencing of NSUN2 increased the cytotoxic rate, whereas TPI1 reversed this inhibition (Fig. 8L). The levels of PD-L1 downregulated by NSUN2 silence were abrogated by overexpression of TPI1 (Fig. 8M–Q). The results demonstrated that silencing of NSUN2 inhibits the proliferation, glycolysis, and immune evasion of BC cells in vitro via decreasing TPI1 expression.Fig. 8NSUN2 knockdown inhibits BC cell proliferation, glycolysis, and immune evasion by decreasing TPI1 expression. (A) BC cell viability was assessed using CCK-8. (B) EdU assay was performed to assess the proliferation capability of BC cells. (C) glucose consumption and (D) lactate concentration were tested using commercial kits. (E, F) ECAR and (G, H) OCR were measured using Seahorse analysis. (I) GLUT1, (J) LDHB, and (K) HK2 expression was detected using qPCR. (L) the cytotoxic rate of BC cells was evaluated using the LDH assay kit. (M) the mRNA expression of PD-L1 in BC cells was measured using qPCR. (N, O) western blotting and (P, Q) IF assay were conducted to determine PD-L1 levels. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001. ###p < 0.001. ##p < 0.01. #p < 0.05 NSUN2 knockdown inhibits BC cell proliferation, glycolysis, and immune evasion by decreasing TPI1 expression. (A) BC cell viability was assessed using CCK-8. (B) EdU assay was performed to assess the proliferation capability of BC cells. (C) glucose consumption and (D) lactate concentration were tested using commercial kits. (E, F) ECAR and (G, H) OCR were measured using Seahorse analysis. (I) GLUT1, (J) LDHB, and (K) HK2 expression was detected using qPCR. (L) the cytotoxic rate of BC cells was evaluated using the LDH assay kit. (M) the mRNA expression of PD-L1 in BC cells was measured using qPCR. (N, O) western blotting and (P, Q) IF assay were conducted to determine PD-L1 levels. All data are presented as mean ± SD from n = 3 independent biological replicates (each performed with technical triplicates). ***p < 0.001. ###p < 0.001. ##p < 0.01. #p < 0.05 Discussion Although the role of m5C in tumor progression has received extensive attention, its precise modification mechanism remains unclear. Our study provides compelling evidence that NSUN2 plays a critical role in BC progression through m5C methylation of TPI1. Specifically, we investigated the effect of NSUN2 on glycolysis and immune evasion, offering novel mechanistic insights into its oncogenic role. NSUN2 is a key enzyme regulating m5C methylation and is closely related to the activation of the oncogene MYC [17], suggesting it may be a central regulator in cancer. NSUN2 expression is frequently elevated in multiple malignancies, including colorectal, lung, thyroid, and gastric cancers [18–21], and similarly, in BC. Previous studies have shown that NSUN2 knockdown inhibits BC cell proliferation, migration, and invasion while inducing apoptosis through stabilization of HGH1 mRNA, thereby attenuating tumor progression [22]. Wang et al. [23] have found that high expression of NSUN2 is associated with poor survival, and silencing of NSUN2 inhibits proliferation and migration in vitro and in vivo. Additionally, Yi et al. [24] have reported that overexpression of NSUN2 facilitates BC cell migration, invasion, and proliferation. Consistent with these observations, our results confirmed that NSUN2 knockdown inhibited tumor growth in vivo. Importantly, our work revealed for the first time that NSUN2 regulates glycolysis and immune evasion—both critical processes in BC biology. BC cells are highly dependent on glycolysis for energy and biosynthetic substrates, and epigenetic changes can promote aerobic glycolysis [6, 25]. Glycolysis is intimately linked to immune escape; lactate accumulation can impair anti-tumor immune cell function by altering intracellular pH [26]. Notably, the glycolytic enzyme HK2 activates NF-κB and increases PD-L1 expression in BC [27]. The interplay between tumor glycolysis and immune evasion has emerged as a critical research focus. Our study demonstrated that NSUN2 knockdown suppressed BC cell proliferation, glycolysis, and immune evasion both in vitro and in vivo, suggesting that NSUN2 facilitates BC progression by driving metabolic reprogramming and immune suppression. The promotion of glycolysis depends on overexpression or increased activity of glycolytic-related enzymes [6]. Our study also reveals a mechanistic link between NSUN2 and TPI1. TPI1 is a glycolysis-related enzyme that catalyzes the interconversion of dihydroxyacetone phosphate and d-glyceraldehyde-3-phosphate, maintaining triosephosphate equilibrium [28]. TPI1 deficiency is associated with hemolytic anemia and neurological disorders, potentially leading to fatal outcomes [29]. Beyond metabolic diseases, TPI1 is implicated in lung fibrosis, osteoarthritis, and various cancers [30–32]. In BC, TPI1 facilitates tumor cell glycolysis, proliferation, and metastasis by stabilizing CDCA5 [33], whereas atractylenolide-1 inhibits colony formation and induces apoptosis through downregulation of TPI1 [34]. In the present study, we found that NSUN2 knockdown reduces TPI1 m5C methylation, with site 962 as a key m5C-responsive site candidate, consequently decreasing TPI1 stability. This modification site will be validated by mass spectrometry or single-base resolution bisulfite sequencing in future studies. TPI1 overexpression promotes BC glycolysis and immune evasion, confirming its role as an oncogene in BC, consistent with this prior report [33]. Rescue experiments demonstrated that TPI1 overexpression counteracts the anti-tumor effects of NSUN2 silencing, underscoring that NSUN2 drives BC progression by upregulating TPI1. These discoveries provide direct mechanistic evidence linking NSUN2-mediated m5C modification to glycolysis and immune escape. The recognition of NSUN2-mediated m5C-modified TPI1 by YBX1, which is an m5C reader protein [35], was confirmed in our study, illustrating the complexity of post-transcriptional regulation in BC. YBX1 is involved in BC cell growth, cell cycle, migration, invasion, and apoptosis [36, 37]. Here, we found that knockdown of YBX1 reversed the promotion of malignant cellular behaviors induced by TPI1, suggesting that YBX1 participates in BC progression via recognizing m5C modification of TPI1. While this previous study has revealed that NSUN2/YBX1-mediated m5C methylation of HGH1 is involved in BC progression [22], our study expands this axis to glycolysis and immune escape, identifying TPI1 as a novel m5C target. The NSUN2/YBX1/m5C/TPI1 axis also regulates PD-L1 expression, thereby linking metabolic reprogramming to immune evasion. This mechanism may represent a promising therapeutic target for disrupting immune evasion in BC. Additionally, a recent study by He et al. [38] also implicated NSUN2-mediated m5 C modification in reprogramming glucose metabolism to favor tumor immune evasion in hepatocellular carcinoma. Together, these findings position the NSUN2–m5C axis as a critical bridge connecting tumor metabolism and the immune microenvironment, with multiple downstream targets and context-specific functions. Tumor cells enhance glycolysis to alter the tumor immune microenvironment, stabilizing immune checkpoints and thus driving immune escape [26]. PD-L1 is a well-known immune checkpoint that is highly expressed in BC cells and is associated with T cell function inhibition [39]. PD-L1-targeted immunotherapies are transforming the management of various solid tumors. Although BC was long considered non-immunogenic, recent studies have shown that specific subtypes, such as triple-negative BC, can respond to PD-L1 blockade [40, 41]. However, the expression of PD-L1 varies greatly across subtypes of BC, ranging from 0 to 83% [42], limiting response rates to PD-L1 inhibitors. Therefore, elucidating PD-L1 regulatory mechanisms is essential for optimizing immunotherapy strategies. In the present study, we found that silencing of NSUN2 reduced PD-L1 expression by decreasing TPI1 expression. The promotion of immune evasion by NSUN2 and TPI1 was a novel finding of our study. These results suggest a role for PD-L1 as a potential downstream effector molecule and provide insight into the interaction between metabolism reprogramming and immunity in BC. However, the precise molecular link warrants further investigation. Given the role of TPI1 in glycolysis, its effect on PD-L1 might be indirect, mediated via metabolic intermediates and associated signaling pathways impacting PD-L1 transcription or translation. Future studies will identify the precise signaling cascade connecting TPI1 to PD-L1 regulation, and evaluate whether NSUN2 inhibition synergizes with PD-L1 blockade in BC immunotherapy. In conclusion, our study sheds light on the multifaceted role of NSUN2 in BC. The results demonstrated that silencing of NSUN2 suppresses the progression of BC by inhibiting cell proliferation, glycolysis, and immune evasion. Mechanistically, NSUN2 stabilizes TPI1 mRNA by promoting its m5C methylation in a YBX1-dependent manner, which is associated with increased PD-L1 expression and immune-evasive phenotypes. This study reveals an important regulatory pathway in BC and highlights the potential of NSUN2 as a therapeutic target. Our study delineates a crucial role of m5C modification in BC. Future work is necessary to integrate the roles of other RNA modifications, such as m6A and m1A methylation, in coordinating the metabolism and immune evasion, comprehensive understanding of the role of post-transcriptional regulation in the progression of BC. Our functional studies were conducted in models representing both TNBC (MDA-MB-231, 4T1) and luminal A (MCF-7) subtypes, indicating the potential broad relevance of the NSUN2/YBX1/m5C/TPI1 axis. Future work will evaluate the activity and clinical association of this axis across all major BC subtypes to fully assess its translational potential and identify patient populations most likely to benefit from its therapeutic targeting. Moreover, we will use new techniques such as spatially resolved CRISPR screening (Perturb-DBiT) [43], co-profiling of transcriptomes and high-plex proteins (spatial CITE-seq) [44], and simultaneous mapping of chromatin, RNA, and protein (DBiT-seq) [45] to explore the spatial dynamics, cellular heterogeneity, and multilayer regulatory circuits mediated by the NSUN2/TPI1 axis in the tumor microenvironment, ultimately guiding more precise therapeutic strategies. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1: Figure S1. Correlation of PD-L1 expression and NSUN2 or TPI1 expression in tumor tissues from patients with BC. (A) Representative IHC images of PD-L1 expression in BC tumor tissues. Upper panel: PD-L1 positive; lower panel: PD-L1 negative. (B) Correlation between PD-L1 IHC score and NSUN2 mRNA expression was assessed using Spearman’s rank correlation test. (C) Correlation between PD-L1 IHC score and TPI1 mRNA expression was assessed using Spearman’s rank correlation test. Supplementary material 2: Figure S2. The promotion of glycolysis and immune evasion induced by TPI1 depends on YBX1. (A) BC cell viability was determined via the CCK-8 assay. (B, C) EdU assay was employed to evaluate the proliferative capacity. (D) Glucose uptake by BC cells and (E) lactate production were quantified using commercially detection kits. (F, G) ECAR and (H, I) OCR were analyzed using the Seahorse analyzer. qPCR was used to quantify the expression of (J) GLUT1, (K) LDHB, and (L) HK2 in BC cells. (M) Cytotoxicity in BC cells was determined by measuring LDH. (N) qPCR, (O, P) western blotting, and (Q, R) IF staining were performed to assess PD-L1 levels in BC cells. All data are presented as mean ± SD from n=3 independent biological replicates (each performed with technical triplicates). ***P<0.001. ###P<0.001. ##P<0.01. #P<0.05. Supplementary material 3: Figure S3. The effects of NSUN2/TPI1 on tumor growth depends on CD8+ T cells. Tumor-bearing mice were injected with anti-CD8α or an isotype control. (A) Images of tumors from the tumor-bearing mouse model in each group. (B) Tumors were weighed. (C) Tumor volume was detected every week. All data are presented as mean ± SD from n=5 mice per group. ***P<0.001. Supplementary material 1: Figure S1. Correlation of PD-L1 expression and NSUN2 or TPI1 expression in tumor tissues from patients with BC. (A) Representative IHC images of PD-L1 expression in BC tumor tissues. Upper panel: PD-L1 positive; lower panel: PD-L1 negative. (B) Correlation between PD-L1 IHC score and NSUN2 mRNA expression was assessed using Spearman’s rank correlation test. (C) Correlation between PD-L1 IHC score and TPI1 mRNA expression was assessed using Spearman’s rank correlation test. Supplementary material 2: Figure S2. The promotion of glycolysis and immune evasion induced by TPI1 depends on YBX1. (A) BC cell viability was determined via the CCK-8 assay. (B, C) EdU assay was employed to evaluate the proliferative capacity. (D) Glucose uptake by BC cells and (E) lactate production were quantified using commercially detection kits. (F, G) ECAR and (H, I) OCR were analyzed using the Seahorse analyzer. qPCR was used to quantify the expression of (J) GLUT1, (K) LDHB, and (L) HK2 in BC cells. (M) Cytotoxicity in BC cells was determined by measuring LDH. (N) qPCR, (O, P) western blotting, and (Q, R) IF staining were performed to assess PD-L1 levels in BC cells. All data are presented as mean ± SD from n=3 independent biological replicates (each performed with technical triplicates). ***P<0.001. ###P<0.001. ##P<0.01. #P<0.05. Supplementary material 3: Figure S3. The effects of NSUN2/TPI1 on tumor growth depends on CD8+ T cells. Tumor-bearing mice were injected with anti-CD8α or an isotype control. (A) Images of tumors from the tumor-bearing mouse model in each group. (B) Tumors were weighed. (C) Tumor volume was detected every week. All data are presented as mean ± SD from n=5 mice per group. ***P<0.001.