We employed the Biomarker Exploration of Solid Tumors (BEST) network tool to juxtapose RRM2 expression levels between lung cancer and standard tissues within the GSE68571 and TCGA-LUAD datasets. The gene expression data was standardized by converting them into Z-scores to facilitate comparative analysis (29, 30).

We downloaded STAR-counts data and corresponding clinical information for TCGA-LUAD tumors from the TCGA database (https://portal.gdc.cancer.gov). We then extracted data in TPM format and performed normalization using the log2(TPM + 1) transformation. After retaining samples that included both RNA seq data and clinical information, we ultimately selected 516 samples for further analysis. The GTEx data we used is from the V8 version, detailed information can be found on the official GTEx website (https://gtexportal.org/home/datasets). Statistical analysis was conducted using R software, version v4.0.3. Results were considered statistically significant when the p-value was less than 0.05 (30).

To evaluate the prognostic significance of RRM2 mRNA expression in LUAD, we first assessed its association with overall survival (OS) in patients with lung cancer. Survival analysis was performed using the KM plotter online platform (http://www.kmplot.com/analysis/) (31), The Biomarker Exploration of Solid Tumors (BEST) network tool was used to compare RRM2 expression levels between tumor and normal tissues in the GSE68571 dataset. Gene expression data were standardized by conversion to Z-scores to facilitate cross-sample comparison. In addition, hazard ratios (HRs), log-rank p-values, and 95% confidence intervals (CIs) were calculated.

We then downloaded STAR-counts RNA-seq data and corresponding clinical information for LUAD samples from the TCGA database (https://portal.gdc.cancer.gov). The data were transformed into TPM format and normalized using a log2(TPM + 1) transformation. After filtering to include only samples with both RNA-seq and clinical data, a total of 516 samples were retained for downstream analyses. Consistency validation was further performed using the GEO dataset (GSE68571). Univariate and multivariate Cox proportional hazards regression analyses were conducted to identify independent prognostic factors. Forest plots were generated using the “forestplot” R package to visualize p-values, HRs, and 95% CIs for each variable. Based on the multivariate Cox regression results, a prognostic nomogram was constructed using the “rms” package to predict the 5-year overall recurrence probability. All statistical analyses were performed using R software (version 4.0.3), and results with a p-value < 0.05 were considered statistically significant.

Upstream binding miRNAs of RRM2 were predicted by several target gene prediction programs, consisting of PITA, RNA22, miRmap, microT, miRanda, PicTar, and TargetScan. Only the predicted miRNAs that commonly appeared in more than two programs as mentioned above were included for subsequent analyses (32). These predicted miRNAs were regarded as candidate miRNAs of RRM2.

We downloaded STAR-counts data and corresponding clinical information for LUAD tumors from the TCGA database (https://portal.gdc.cancer.gov). Gene expression was converted to TPM and log-transformed (log2[TPM + 1]). Samples with both RNA-seq and clinical data were retained (n = 516) ssGSEA scoring. Immune-cell signatures (published marker gene sets) were scored per sample using GSVA (R) with method = “ssgsea”. Scores represent relative abundance of immune populations.

Group comparison and correlations. Samples were stratified into RRM2-high vs. RRM2-low (median split). Differences in ssGSEA scores between groups were tested by two-sided Wilcoxon rank-sum. Associations between RRM2 expression and immune-cell scores were quantified by Spearman’s ρ. Significance was set at P < 0.05 (two-sided). Immune-infiltration patterns were cross-checked using TIMER and TISIDB. TIMER’s SCNA module was used to compare infiltration across RRM2 copy-number states (deep deletion, arm-level deletion, diploid/normal, arm-level gain, high amplification), using TIMER’s default statistics (33).

The data used in the above research all come from the files in the Supplementary File of the GEO database (https://www.ncbi.nlm.nih.gov/geo/). The Seurat package was utilized to generate objects and filter out cells of poor quality, while also carrying out a standard data preprocessing procedure. The number of genes, the number of cells, and the percentage of mitochondrial content were calculated. The filtering criteria were genes detected in fewer than 5 cells and cells with fewer than 200 detected genes.

We retain genes detected in at least 12346 cells and filter out cells with fewer than 200 detected genes, as well as cells with a high mitochondrial content (>5%). After discarding low-quality cells, we retain 2000 cells for downstream analysis. To normalize each cell, we scale the UMI counts using scale factor = 10,000. After logarithmically transforming the data, we use the ScaleData function (v4.1.0) in Seurat.

We apply the corrected normalized data to standard analysis, extracting the top 2000 variable genes for Principal Component Analysis (PCA). For the visualization and clustering of UMAP (or TSNE), we retain the top 20 principal components. Cell clustering is performed using the FindClusters function (resolution = 0.8) implemented in the Seurat R package (34).

The A549 and H460 cells were incubated at 37°C in a 5% CO₂ atmosphere. The primary antibodies for RRM2 polyclonal antibody were purchased from Cell Signaling Technology (Danvers, MA, United States). The β-actin and secondary antibodies were purchased from Protein Technology Group, Inc. (Wuhan, China). The siRNA of RRM2 were purchased from jtsbio Biotech (jtsbio Biotech Co.,Ltd, Wuhan, China). jtsbio Biotech designed and established the let-7c-5p overexpression plasmid (jtsbio Biotech Co.,Ltd, Wuhan, China) (Supplementary Table 1). Plasmid and siRNA transfection was performed with Lipofectamine^® 3000 following the manufacturer’s instructions.

Cells were seeded at a density of 5 × 10³ cells/well. Add 10 µl of CCK8 solution at 24, 48, and 72 hours, and measure the absorbance at 450 nm after 2 hours.

Cells were seeded at a density of 4×10⁴ cells in 24-well transwell inserts. After 24 hours, they were fixed with 4% paraformaldehyde for 10 minutes, stained with crystal violet solution for 12 minutes, and images were collected.

Cells were collected, and total RNA was extracted using Trizol reagent (Invitro-gen Inc., Carlsbad, CA) according to the manufacturer’s protocol. The expression levels of let-7c-5p and RRM2 mRNA were normalized to the GAPDH (Supplementary Table 2). The relative fold changes in target gene expression between the control group and the experimental group were calculated using the 2^−ΔΔCT method.

Homogenize cells in RIPA lysis buffer. Determine protein concentration using a BCA assay kit, boil the mixture at 98 °C in a Dual-Color protein loading buffer, separate equal amounts of protein by SDS-PAGE, and transfer to a PVDF membrane. Subsequently, incubate the membrane with primary and secondary antibodies. Perform enhanced chemiluminescence to visualize the protein bands. Finally, apply ImageJ for quantitative analysis.

All results are expressed as mean ± SD. Comparisons among groups were conducted using one-way ANOVA and Tukey’s multiple comparison tests. A p-value of less than 0.05 was considered statistically significant. Data analysis was performed using the BEST, TIMER, and TISIDB platforms, which automatically applied the Benjamini-Hochberg method for FDR correction in multiple hypothesis testing and reported effect sizes (such as correlation coefficients and normalized enrichment scores) in correlation and enrichment analyses. Analyses were performed in R 4.0.3 using GSVA and base/ggplot2 functions for statistics and plotting.

Transcriptomic profiling of LUAD tissues from the GSE68571 and TCGA-LUAD cohorts demonstrates significant up-regulation of RRM2 in tumors compared to normal tissues (Wilcoxon test, p=0.00053 and p<2.2×10^-16, respectively; Figures 1A, B). RRM2 expression levels were found to escalate with advancing tumor stage (Kruskal-Wallis test, p=2.273×10^-33), and correlated strongly with TNM-T classification (p=1.08×10^-31), suggesting its role in metastatic progression (Figures 1C, D). Collectively, these findings indicate that RRM2 may contribute to the malignant progression of LUAD and could represent a potential therapeutic target. Its stage-dependent overexpression and significant association with patient survival highlight the potential clinical utility of RRM2 for risk stratification and precision oncology applications.

Kaplan–Meier survival analyses consistently demonstrated that patients with high RRM2 expression had significantly worse OS compared with those with low expression across multiple datasets. In the TCGA-LUAD cohort, elevated RRM2 levels were strongly associated with reduced survival probability (HR = 1.82, 95% CI: 1.55–2.12, p = 2.4 × 10^-14; Figure 2A), which was further confirmed in two additional LUAD datasets (HR = 1.39, 95% CI: 1.13–1.71, p = 0.0019; HR = 1.63, 95% CI: 1.38–1.93, p = 9.6 × 10^-9; Figures 2B, C). The external validation cohort (GSE68571) also supported these findings, showing significantly shorter OS in the high RRM2 group (log-rank p < 0.01; Figure 2D).Time-dependent ROC curve analysis yielded AUC values of 0.621 (1-year), 0.597 (3-year), and 0.586 (5-year), indicating limited but consistent prognostic discrimination across time points (Figures 2E, F).Univariate Cox regression identified RRM2, TNM stage, and smoking status as significant predictors of OS (Figure 2G). Importantly, multivariate Cox analysis confirmed RRM2 as an independent prognostic factor for LUAD (HR = 1.30, 95% CI: 1.14–1.53, p < 0.01), after adjusting for age, sex, race, tumor stage, and smoking status (Figure 2H). Collectively, these findings demonstrate that RRM2 overexpression is significantly associated with unfavorable clinical outcomes and may serve as an independent prognostic biomarker in LUAD, providing potential value for risk stratification and precision oncology applications.

The pivotal role of ncRNAs, notably miRNAs in modulating gene expression is well-established. In LUAD, RRM2 has been associated with tumorigenesis, and its expression is hypothesized to be under miRNA-mediated control. To probe this hypothesis, an exhaustive analysis was undertaken to uncover miRNAs capable of binding and modulating RRM2 expression. Utilizing Cytoscape software, a network was delineated to graphically represent the interactions between RRM2 and a spectrum of miRNAs. This network schematizes the putative regulatory relationships between RRM2 and 20 distinct miRNAs, encompassing let-7f-5p, miR-6845-5p, let-7b-3p, among others. The connections between RRM2 and these miRNAs imply potential regulatory influences they may exert on RRM2 expression (Figure 3A). This visualization serves as a foundation for further investigation into the intricate interplay between RRM2 and miRNAs in the context of LUAD.

To delve deeper into the interplay between RRM2 and the aforementioned miRNAs, an expression correlation analysis was meticulously conducted. The findings, tabulated and presented with correlation coefficients (R-values) and their corresponding statistical significance (P-values), delineate the relationship between each miRNA-RRM2 pair. Notably, RRM2 exhibited significant inverse correlations with let-7f-5p (R = -0.14525, P = 0.001), let-7b-3p (R = -0.23631, P = 7.29E-08), and let-7b-5p (R = -0.32575, P = 6.1E-14), suggesting that these miRNAs may function as tumor suppressors by negatively regulating RRM2 expression. In contrast, a positive correlation was between RRM2 and miR-4788 (R = 0.28105, P = 0.000000303), implying a potential oncogenic role in enhancing RRM2 expression (Figure 3B). Focusing on let-7c-5p, a miRNA of particular interest due to its significant negative correlation with RRM2, we compared its expression levels between 512 LUAD cancer samples and 20 standard samples using box plots. The analysis revealed a substantial down-regulation of let-7c-5p in cancer samples (P = 1.9e-18), hinting that its diminished expression might contribute to the up-regulation of RRM2 and potentially foster tumorigenesis (Figure 3C). The prediction significance of let-7c-5p in LUAD was further evaluated using KM survival plots. The analysis indicated that patients with elevated let-7c-5p expression levels exhibited improved OS rates (Log-Rank p = 0.043, Hazard Ratio = 0.74), suggesting that let-7c-5p could serve as a potential prediction biomarker and therapeutic target for LUAD (Figure 3D). These results underscore the intricate regulatory network involving RRM2 and miRNAs in LUAD and highlight the potential of miRNAs as therapeutic agents.

To verify the role of let-7c-5p in the development of NSCLC, we measured the expression of let-7c-5p in NSCLC cell lines H273, H23, A549, and H460, as well as in normal lung epithelial cells 16HBE, using RT-qPCR. The results showed that the expression of let-7c-5p was lower in all NSCLC cell lines compared to 16HBE cells (Supplementary Figure S1). Among them, the A549 and H460 cells showing the lowest expression of let-7c-5p were transfected with let-7c-5p mimic or mimic control for subsequent use (Figure 4A). RT-qPCR assay presented the transfection effectiveness of RRM2 siRNA in A549 and H460 cells (Figure 4B). The results of the CCK8 analysis indicate that the number of proliferating A549 and H460 cells decreases following the overexpression of let-7c-5p (Figures 4C, D). From a molecular perspective, it was found that the let-7c-5p mimic reduces the level of RRM2 (Figure 4E). The transwell experiment demonstrated that the absence of RRM2 significantly inhibited the cell migration and metastasis capabilities (Figure 4F). The above results indicate that let-7c-5p regulates RRM2, which inhibits the proliferation and migratory capacity of LUAD cells in vitro.

Utilizing the ssGSEA methodology, we quantified the associations between RRM2 expression and the levels of immune cell infiltration in LUAD (Figures 5A, B). Our findings revealed a positive correlation between RRM2 expression and the infiltration of Natural Killer (NK) CD56bright cells, TH2 cells, TFH, T effector memory cells (Tem), and Macrophages, while a negative correlation was with Th17 cells, dendritic Cells (DCs), neutrophils, cytotoxic cells, regulatory T cells (Tregs), T memory cells (Tcm), pDCs, CD8⁺ T cells, and B cells. To further elucidate these relationships, scatter plots were employed to illustrate the correlation between RRM2 expression and various immune cell populations in LUAD (Figures 5C–O). These plots include: B cells (Figure 5C), CD4⁺ T cells (Figure 5D), CD8⁺ T cells (Figure 5E), Macrophages (M0) (Figure 5F), Macrophages (M1) (Figure 5G), DCs (Figure 5H), Neutrophils (Figure 5I), Cytotoxic Cells (Figure 5J), Tregs (Figure 5K), Tcm (Figure 5L), Tem (Figure 5M), NK cells (Figure 5N), Monocytes (Figure 5O). Box plots (Figure 5P) were used to display the infiltration levels of different immune cells in LUAD, categorized by copy number alterations. These analyses indicate that the expression level of RRM2 is closely associated with the immune cell infiltration patterns in LUAD. The correlation coefficients and P-values provide a quantitative measure of the strength and statistical significance of these associations. Our data imply that elevated RRM2 expression may be associated with a subdued anti-tumor immune response. The use of TIMER and TISIDB databases further corroborates the connection between RRM2 expression and immune cell infiltration, potentially influencing LUAD prognosis and treatment responses. These findings highlight the complex interplay between RRM2 expression and immune cell dynamics in LUAD and may have implications for developing immunotherapeutic strategies.

To delineate the immune landscape of LUAD, single-cell transcriptomic data comprising 12,346 cells were analyzed using t-distributed stochastic neighbor embedding (t-SNE). This dimensionality reduction approach revealed seven transcriptionally distinct T-cell clusters, including effector memory CD8 T cells, MAIT cells, Non-Vd2 γδ T cells, CD4⁺ T cells, T regulatory cells, Terminal effector CD8 T cells, and Th1/Th17 cells (Figure 6A). The clear separation of these clusters in the two-dimensional embedding underscores the intrinsic transcriptional heterogeneity and functional specialization of the T-cell compartment in LUAD.

To identify genes contributing to cellular heterogeneity, the relationship between standardized variance and average expression was examined (Figure 6B). Among analyzed, 2,000 were classified as highly variable (highlighted in red). Notably, RRM2 emerged among the top variable genes, alongside immune activation associated genes such as FOS, CCL3, CD177, and HSPA6, indicating its potential involvement in immune cell proliferation or activation processes.

The spatial distribution map of RRM2 expression across the t-SNE projection (Figure 6C) demonstrated a strikingly localized pattern, with the highest density observed within a small subset of cells positioned at the lower region of the embedding. This subpopulation corresponded to the effector memory CD8 T-cell cluster, while most other immune subtypes exhibited minimal RRM2 expression. The density gradient further confirmed a concentrated enrichment of RRM2-positive cells, suggesting that RRM2 expression is functionally confined to a proliferative or metabolically active subset.

Violin plot analysis (Figure 6D) corroborated these findings, showing that RRM2 expression was markedly elevated in effector memory CD8 T cells, whereas it remained low or undetectable in MAIT cells, T regulatory cells, Th1/Th17 cells, and other subsets. Given the established role of RRM2 in deoxyribonucleotide synthesis and DNA repair, its preferential expression in effector memory CD8 T cells likely supports enhanced proliferative potential and metabolic demand during immune activation.

Collectively, these data identify RRM2 as a proliferation-associated marker preferentially expressed in activated cytotoxic T cells, implicating its involvement in sustaining immune effector responses within the LUAD tumor microenvironment (TME).