PLK1 overexpression as a dual-role biomarker and therapeutic vulnerability in pulmonary adenocarcinoma

Data acquisition and preprocessing

This research employed Xiantao Academic (https://www.xiantao.love/) to conduct bioinformatics analysis and visualize data. The platform enabled extensive analyses such as pan-cancer analysis, differential expression assessments, clinical relevance evaluation, functional clustering, and related statistical analyses. Transcriptomic profiles and corresponding clinicopathological metadata for LUAD cohorts (TCGA-LUAD) were acquired through GDC portal (http://portal.gdc.cancer.gov) and UCSC XENA browser (http://xenabrowser.net/datapages/). STAR-aligned and Toil-pipeline normalized transcript-level TPM matrix has been used for downstream analysis (Vivian et al., 2017). Independent validation was performed using dataset GSE115002 from NCBI’s GEO repository. Representative immunohistochemical (IHC) images of PLK1 protein expression in normal lung tissue and lung adenocarcinoma have been obtained from Human Protein Atlas (HPA) (http://www.proteinatlas.org/) (Uhlén et al., 2015). Gene mutation and copy number alteration (CNA) data have been obtained from cBioPortal (http://www.cBioportal.org/) (De Bruijn et al., 2023). Transcriptomic data are log2(TPM + 1) transformed before analysis. Normal tissue samples and incomplete clinical annotations are excluded.

Differential expression analysis

To investigate the expression pattern of PLK1 (ENSG00000166851.15), we performed the following comparisons: Pan-cancer analysis (Fig. 1A): expression of PLK1 on TCGA and Genotype-Tissue Expression (GTEx) cohorts. Unpaired analysis (Fig. 1B): Tumor vs. normal LUAD samples using Wilcoxon rank sum test. Paired sample analysis (Fig. 1C): Matched tumor-normal LUAD samples using paired t-test. External validation (Fig. 1D): Analysis of GSE115002 using Wilcoxon test. Statistical computations were executed in R (v4.2.1) using stats, car, and ggplot2 packages.

Diagnostic value assessment

The diagnostic accuracy of PLK1 was evaluated through time-dependent AUC assessment (Fig. 1E) using package pROC (v1.18.0). Curves and AUC values were visualized using ggplot2 (v3.4.4).

Association with clinicopathological features

Association of PLK1 expression with Clinicopathological Features (e.g., tumor stage, TNM classification, gender) was tested by Kruskal-Wallis test (Figs. 2A–2I). Relevant data were extracted from TCGA; incomplete cases were excluded from subgroup analysis.

Survival analysis

PLK1’s survival prediction capacity was quantified through Kaplan–Meier survival curves and univariate Cox proportional hazards regression (Figs. 3A–3I). Patients were categorized according to median expression level. Survival (v3.3.1) and survminer (v0.4.9) packages were used. Logrank test was used to evaluate statistical significance. Clinical follow-up data were supplemented by survival metadata from Liu et al. (2018).

Nomogram construction

To construct a prognostic model of overall survival, optimal expression cutoff values were determined using surv_cutpoint function with bootstrap resampling (200 iterations, 80 samples each). Multivariable survival modeling was performed through proportional hazards regression to identify clinicopathologically independent predictors (Table 1). The prognostic prediction system was developed with the rms package (v6.3-0), and calibration plots were used for 1-, 3- and 5-year survival intervals (Fig. 4).

Differential gene expression and functional enrichment

Differentially expressed genes (DEGs) of high and low PLK1 expression groups (top 50% vs. bottom 50%) were identified using DESeq2 (v1.36.0) with default parameters (Love, Huber & Anders, 2014). Genes with an adjusted p-value (FDR) < 0.05 and an absolute log2 fold change > 1 were defined as statistically significant. The analysis using edgeR (v3.38.2) on the same raw count data yielded consistent results (Fig. 5A). The top 20 genes strongly co-expressed with PLK1 were selected according to Spearman correlation and visualized using heatmap (Fig. 5B). GO and KEGG pathway enrichment analyses were performed using clusterProfiler (v4.4.4) (Yu et al., 2012), considering terms with an FDR <0.05 as significantly enriched. Gene ID annotation was performed using org.Hs.eg.db and visualisation with custom ggplot2 scripts (Fig. 5C).

Immune cell infiltration study

Immune cell infiltration was estimated using ssGSEA with the GSVA package (v1.46.0) (Fig. 6A) (Hänzelmann, Castelo & Guinney, 2013). The immune cell gene sets (n = 24) were selected from Bindea et al. (2013). The correlation between PLK1 expression and immune cell abundance (e.g., Th2 cells and immature dendritic cells) was assessed using Spearman rank correlation (Figs. 6B, 6C). Immune infiltration scores were compared between PLK1-high and PLK1-low using Wilcoxon rank sum test (Figs. 6D–6G). All plots were generated using ggplot2.

External validation analysis

The robustness of our findings was assessed through external validation using independent data and tools. The association between PLK1 expression and overall survival in LUAD was examined using the Kaplan–Meier plotter online tool (https://kmplot.com/analysis/index.php?p=home), which aggregates data from multiple GSE datasets within the GEO repository (Posta & Gyorffy, 2025). Furthermore, the GSE115002 dataset was utilized to reconfirm key associations, including the link between PLK1 expression and clinical stage, the functional enrichment profiles of PLK1-co-expressed genes, and the correlation between PLK1 expression and immune cell infiltration levels via ssGSEA. Analytical methods were consistent with those applied in the primary analysis. The corresponding data have been provided in the Supplemental Information.

Cell culture and PLK1 specific inhibitor treatment

The A549 cell line, derived from human lung adenocarcinoma was generously provided by Prof. Lijun Shi from the First Affiliated Hospital of Guangxi Medical University. The cell line was authenticated by short tandem repeat (STR) profiling within the last six months and routinely tested to confirm the absence of mycoplasma contamination. A549 cells were cultured in RPMI 1640 medium (cat.no. 61870036; Gibco) with 10% fetal bovine serum (FBS, cat.no. A5256701; Gibco) at 37 °C with 5% CO₂. Cells were maintained at 37 °C in a humidified atmosphere with 5% CO₂ and routinely passaged at 80–90% confluence using 0.25% Trypsin-EDTA (cat. no. 25300054; Gibco). All cell culture procedures were performed under standard sterile conditions. In order to inhibit PLK1 activity we used selective small-molecule inhibitor GSK461364 (Cat. no. G873280, CAS: 929095-18-1; Macklin Biochemical Co., Shanghai, China). This ATP-competitive inhibitor exhibits high selectivity for PLK1, with a half-maximal inhibitory concentration (IC50) of approximately 2.2 nM and more than 390-fold selectivity over PLK2 and PLK3 (Olmos et al., 2011). Consequently, based on preliminary dose–response experiments and published pharmacokinetic data, a working concentration of 40 nM GSK461364 was selected for all subsequent functional assays to ensure efficacy within the selective window for PLK1 inhibition (Gilmartin et al., 2009). Dimethyl sulfoxide (DMSO, cat.no. A68908; Innochem) at a final concentration of 0.1% was used in an equivalent volume as the vehicle control. All downstream assays were performed immediately after treatment. This pharmacological inhibition approach (rather than siRNA-mediated gene silencing) was chosen to recapture therapeutic drug effects and to enhance its translational relevance for future studies.

Cell proliferation assessment (CCK-8)

The proliferative potential of A549 cells was quantitatively analyzed using the CCK-8 colorimetric assay (cat.no. C0037; Beyotime). After 12-hour pre-culture in 96-well plates (1.5 × 10³ cells/well), the medium was replaced with fresh solutions containing: Vehicle control (0.1% DMSO) and 40 nM GSK461364. At designated time points (0–4 days post-treatment), 10 μL CCK-8 reagent was added to each well and incubated for 1 h at 37 °C. Optical density measurements at 450 nm were acquired using a GloMax Discover multimode microplate reader (Promega Corporation).

Clonogenic survival analysis

A549 cells (1 × 10³ cells/well) seeded in 6-well plates received 24-hours GSK461364 treatment (40 nM), followed by PBS washing and 6-day maintenance in drug-free medium for clonogenic assessment. Post-treatment procedures included: (1) PBS washing to remove non-adherent cells, (2) fixation with 4% paraformaldehyde (15 min, cat. no. P00999; Beyotime), and (3) staining with 0.5% crystal violet solution (15 min, cat.no. C0121). ImageJ software was used for quantifying Colony formation assay.

Apoptosis analysis

To induce apoptosis, A549 cells (5 × 10⁵ cells/well) were treated with 40 nM GSK461364 or vehicle control (0.1% DMSO) for 24 h. Quantitative apoptosis assessment was then conducted using the FITC Annexin V/PI dual-staining kit (C1062M; Beyotime Biotechnology). Following treatment, cells were harvested via trypsinization (0.25% trypsin-EDTA, cat.no. 25200056; Gibco). Cell pellets were resuspended in binding buffer and dual-stained with Annexin V-FITC/PI (15 min, dark). Accuri C6 flow cytometer (BD Biosciences) was used for detection and flowJo software 10.9.0 was used for analysis.

Gating strategy: The intact cell population was first identified on an FSC-A vs. SSC-A dot plot to exclude debris. Single cells were then selected using an FSC-H vs. FSC-A gate to exclude doublets. Apoptotic populations were finally quantified on an Annexin V-FITC vs. PI dot plot.

Cell cycle assay

A549 cells (5 × 10⁵/well) in 6-well plates were treated with 40 nM GSK461364 or vehicle control for 24 h. Floating and adherent cells (harvested via 0.25% trypsin digestion) were pooled, fixed in 75% ethanol (−20 °C, 16 h), then subjected to RNase A treatment (100 µg/mL, 37 °C, 30 min) prior to PI staining (50 µg/mL, 15 min dark incubation). Cell cycle distribution was quantified using a BD Accuri C6 flow cytometer analyzed with FlowJo v10.9.0.

Gating strategy: After gating on the intact single-cell population (FSC-A vs. SSC-A) and excluding doublets (FSC-H vs. FSC-A), the DNA content of single cells was analyzed on a PI-A histogram. The cell cycle phase distribution (G0/G1, S, G2/M) was modeled and quantified using the built-in “Cell Cycle” analysis module in FlowJo.

Western blotting

To assess the effects of PLK1 inhibition on cell cycle regulatory proteins, A549 cells were treated with 40 nM GSK461364 or vehicle control (0.1% DMSO) for 24 h prior to protein extraction. Cellular proteins were lysed in RIPA buffer supplemented with protease inhibitor cocktail. After centrifugation (12,000× g, 15 min), protein concentrations were quantified spectrophotometrically (NanoDrop). 25 µg of protein per lane was separated on 10% SDS-polyacrylamide gels and electrophoretically transferred to PVDF membranes (beyotime, cat.no. FFP24). The transfer was conducted at 100 V for 60 min on ice. Membranes were blocked with QuickBlock buffer for 15 min at room temperature prior to overnight incubation (4 ° C) with primary antibodies: Cyclin B1 (cat. no. ab181593, clone EPR17060, 1:2000; Abcam), CDK1 (cat.no. ab32094, clone YE324, 1:5000; Abcam), and β-actin (cat.no. 66009-1-lg, 2D4H5, 1:10,000; Proteintech). HRP-conjugated goat anti-mouse (cat.no. P0946, 1:1000; Beyotime) or goat anti-rabbit (#P0948, 1:1000; Beyotime) secondary antibodies were applied for 1 h at RT, and proteins were detected using an ECL substrate.

Statistical analysis

Differences in PLK1 expression were assessed through Wilcoxon tests or paired t-tests. Survival was evaluated using Kaplan–Meier/log-rank tests, while associations were analyzed using Spearman’s correlation. Experimental comparisons utilized t-tests/ANOVA after confirming normality. Benjamini–Hochberg FDR correction was applied as needed (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). Data from triplicate experiments are presented as mean ± SEM and were analyzed with GraphPad Prism v10 (experimental data) and R v4.2.1 (bioinformatics data) to ensure methodological consistency.

Upregulation of PLK1 expression in LUAD tissues

Analysis of TCGA and GTEx databases revealed differential PLK1 expression across various cancer types (Fig. 1A). In LUAD patients, tumor tissues demonstrated significantly elevated PLK1 expression compared to adjacent normal tissues (Figs. 1B, 1C). Validation using the GSE115002 dataset (52 paired samples) confirmed higher PLK1 levels in tumor tissues (p < 0.001, Fig. 1D). ROC curve analysis indicated high diagnostic accuracy of PLK1 expression with an AUC of 0.981 (0.964–0.998) (Fig. 1E). IHC images from the HPA database corroborated enhanced PLK1 protein expression in LUAD specimens (Fig. 1F). Genomic alteration analysis via cBioPortal showed PLK1 amplifications, deep deletions, and missense mutations occurring in <2% of cases (Fig. 1G). These multi-platform findings substantiate PLK1 overexpression in LUAD.

Association between PLK1 expression and clinicopathological features in LUAD

Analysis of TCGA data revealed significant correlations between PLK1 expression levels and nine clinicopathological parameters (Fig. 2). Elevated PLK1 expression showed strong associations with advanced T stage (p < 0.01, Fig. 2A), N stage (p < 0.01, Fig. 2B), M stage (p < 0.05, Fig. 2C), and higher pathological stage (p < 0.01, Fig. 2D). Significant relationships were also observed with gender (p < 0.001, Fig. 2E), age (p < 0.05, Fig. 2F), primary therapy outcome (p < 0.01, Fig. 2G), overall survival events (p < 0.001, Fig. 2H), and disease-specific survival events (p < 0.001, Fig. 2I). These findings indicate that PLK1 is not only upregulated in LUAD but also significantly associated with key clinicopathological features.

Elevated PLK1 levels are linked to unfavorable outcomes in LUAD patients

Kaplan–Meier survival analysis of TCGA data stratified by median PLK1 expression revealed significantly worse overall survival in high-expression patients (HR = 1.91, p < 0.001; Fig. 3I). Subgroup analyses demonstrated consistent prognostic associations across multiple clinical contexts: younger patients (<65 years) showed heightened risk (HR = 2.03, p = 0.002; Fig. 3A), as did male subjects (HR = 1.76, p = 0.008; Fig. 3B) and those achieving R0 resection (HR = 1.76, p = 0.001; Fig. 3C). Elevated PLK1 expression maintained predictive value across tumor stages—T2 (HR = 2.18, p < 0.001; Fig. 3D), N0 (HR = 2.29, p < 0.001; Fig. 3E), and M0 (HR = 2.05, p < 0.001; Fig. 3F)—as well as pathological stages I–III (HR = 1.93, p < 0.001; Fig. 3G). Notably, patients with complete/partial treatment response (CR/PR) still exhibited increased mortality risk (HR = 2.03, p = 0.002, Fig. 3H). These stratified findings establish PLK1 overexpression as a multi-contextual prognostic biomarker in LUAD.

PLK1 as independent prognostic determinant in LUAD survival

Cox regression analyses incorporating age, gender, and TNM stage parameters identified elevated PLK1 expression as an independent risk factor for reduced overall survival (HR = 1.692, 95% CI [1.186–2.413], p = 0.004). Multivariate assessment confirmed significant prognostic contributions from advanced T stage (HR = 2.134, 95% CI [1.407–3.235], p < 0.001) and nodal involvement (N stage: HR = 2.028, 95% CI [1.435–2.867], p < 0.001), as detailed in Table 1. These findings establish PLK1 overexpression as a distinct biological predictor of clinical outcomes independent of conventional staging parameters.

PLK1-incorporated nomogram for LUAD prognostication

A predictive nomogram integrating three multivariate-validated prognostic determinants—T stage, N stage, and PLK1 expression—was developed to estimate 1-, 3-, and 5-year survival probabilities (Fig. 4A). Calibration analysis demonstrated strong concordance between model predictions and observed outcomes across all timepoints (Fig. 4B). The model achieved a concordance index of 0.692 (95% CI [0.668–0.716]), indicating moderate discriminative capacity. Comparative assessment revealed superior predictive performance of this composite model over individual prognostic parameters (Fig. 4B), establishing its clinical utility for stratified survival prediction in LUAD management.

Functional characterization of PLK1-associated genes in LUAD

Comparative analysis of high vs low PLK1 expression groups using DESeq2/edgeR identified 1,786 protein-coding DEGs (—log₂FC— > 1, adj.p < 0.05), comprising 912 upregulated and 874 downregulated genes (Fig. 5A). Pearson correlation analysis revealed 275 strongly PLK1-associated genes (r > 0.6, adj.p < 0.05), with top 20 positive correlations visualized (Fig. 5B). Functional enrichment of these co-expressed genes demonstrated significant involvement in nuclear division processes (GO: BP), chromosomal organization (GO: CC), and DNA catalytic activities (GO: MF), along with cell cycle regulation pathways (KEGG) (Fig. 5C). Quantitative enrichment metrics identified 340 BP terms, 72 CC terms, 40 MF terms, and 16 KEGG pathways (adj.p/q < 0.05), with detailed annotations cataloged in Table S1.

PLK1 expression is correlated with altered immune cell infiltration in LUAD

TCGA-based analysis revealed PLK1 expression positively correlated with Th2 cells (r = 0.733, p <0.001, Fig. 6B), Tγδ cells, NK CD56dim cells, regulatory T cells (Tregs), and activated dendritic cells (aDCs). Conversely, inverse associations were observed with mast cells, immature dendritic cells (iDCs: r = −0.316, p < 0.001, Fig. 6C), follicular helper T cells (TFH), eosinophils, Th17 cells, conventional dendritic cells (cDCs), CD8+ T cells, B cells, central memory T cells (Tcm), total T cells, macrophages, and plasmacytoid dendritic cells (pDCs) (Fig. 6A). Comparative analysis between PLK1-high/low groups demonstrated significant differential infiltration of Th2 cells, iDCs, and CD8+ T cells (p < 0.05; Figs. 6D–6G). These correlative findings are consistent with a model in which high PLK1 expression may contribute to an immunosuppressive TME, potentially through coordinated associations with Th2 polarization, impaired dendritic cell maturation, and reduced cytotoxic T cell infiltration.

External validation of key findings

To reinforce the robustness of our findings, we validated the core results in independent external datasets. Initial analysis of aggregated GEO cohorts confirmed that high PLK1 expression was significantly associated with poorer overall survival (pooled HR = 1.95, 95% CI [1.64–2.33], p < 3.3e-14). Subsequent analysis in the GSE115002 validation cohort consistently recapitulated our primary findings: high PLK1 expression was significantly correlated with advanced clinical stage (p = 0.033, Cohen’s d = −0.771, indicating a medium effect size toward higher stage), and PLK1-co-expressed genes were again predominantly enriched in cell cycle-related pathways. Furthermore, the significant association between high PLK1 expression and an immunosuppressive tumor microenvironment was confirmed by ssGSEA. These consistent results across independent cohorts substantially strengthen the generalizability of our conclusions. Detailed results are presented in Fig. S1.

PLK1 inhibition suppresses oncogenic activity in A549 cells

Pharmacological PLK1 inhibition significantly attenuated A549 proliferative capacity, evidenced by reduced CCK-8 viability (Fig. 7A) and diminished colony formation (Figs. 7B, 7F). Cell cycle profiling revealed G2/M phase arrest (Figs. 7D, 7H) accompanied by elevated apoptosis rates (Fig. 7C,7G). Western blot analysis demonstrated concomitant accumulation of Cyclin B1 and CDK1 (Fig. 7E), aligning with prior pathway enrichment findings implicating PLK1 in cell cycle regulation (Fig. 5C). These multimodal findings establish PLK1 as a critical regulator of oncogenic phenotypes in LUAD through cell cycle modulation and inhibiting cell survival.