We utilized pan-cancer RNA-sequencing data and clinical information from TCGA (20) and acquired GTEx data from the UCSC Xena website (http://xena.ucsc.edu/) (21). Information regarding TMB and MSI was extracted from the TCGA dataset. Additionally, we gathered datasets from GEO databases (https://www.ncbi.nlm.nih.gov/geo/) for validation purposes in our pan-cancer analysis, specifically GSE19750, GSE13507, GSE42568, GSE39582, GSE13898, GSE6631, GSE66858, GSE15641, GSE167573, GSE26574, GSE106291, GSE50161, GSE43378, GSE76427, GSE30219, GSE51024, GSE183088, GSE26712, GSE28735, GSE62452, GSE32571, GSE116918, GSE87211, GSE21122, GSE30930, GSE15605, GSE65904, and GSE26899, GSE84437, GSE65144, and GSE17025.

We obtained cancer tissue data from the TCGA dataset and normal tissue data from both TCGA and GTEx datasets. Differences in AURKA expression between cancer and normal tissues were assessed using the “rma” function in R software, with significance defined as p < 0.05. A comprehensive list of cancer types and their corresponding abbreviations used in our study are provided in Supplementary Table 1 . Protein expression of AURKA in normal and pathological tumor tissues was analyzed using THPA (http://www.proteinatlas.org/), which provides immunohistochemistry results for AURKA protein expression in various tissues (22).

To evaluate the prognostic value of AURKA in patients with cancer from TCGA database, we conducted Kaplan–Meier (KM) analyses based on the median expression level of AURKA. Additionally, we performed a univariate Cox regression analysis to investigate the predictive significance of AURKA.

We employed the “cluster Profiler” package for Gene Set Enrichment Analysis (GSEA) to explore potential pathways associated with AURKA expression (23). Biological functions were examined based on KEGG and HALLMARK pathways.

The relative numbers of immune cells were obtained from the TIMER2.0 database (http://timer.cistrome.org/) using the CIBERSORT method (24). We analyzed the association between AURKA expression and the ESTIMATE score and immune score in the tumor microenvironment (TME) using SangerBox (http://sangerbox.com/Tool) (25). Spearman’s rho test was employed to detect associations between AURKA expression and the immune score and ESTIMATE score.

Copy number variations (CNVs) are linked to tumor progression, with higher CNV scores indicating increased aggressiveness (26). We acquired CNV data from the cBioPortal database (https://www.cbioportal.org/) (27). The TIDE score, which assesses tumor immune escape and predicts responses to immune checkpoint therapy (28), was utilized to explore the correlation between AURKA expression, T-cell lesions, and immunotherapeutic responses. We analyzed these relationships using data from the TIDE database (http://tide.dfci.harvard.edu/).

Our analysis of the correlation between AURKA and immunomodulators, including immune-activating genes, immunosuppressive genes, chemokines, and chemokine receptor genes, was conducted using the TISIDB database (http://cis.hku.hk/TISIDB/) (29).

The esophageal squamous mucosa cells EPC1 and EPC2 were generously provided by Dr. Hiroshi Nakagawa (30). Prostate cancer cells DU145, pancreatic cancer cells PaTu8988t, and EAC cells OE33 and OE19 were purchased from Sigma-Aldrich (Taufkirchen, Germany). These cells were cultured in RPMI-1640 medium (Thermo Fisher Scientific, Germany) supplemented with 10% fetal bovine serum (Sigma-Aldrich, Germany).

AURKA was silenced using antisense LNA GapmeRs (cat. no. 339511 LG00799137-DDA, Qiagen, Germany; sequence: 5′-T*C*T*A*G*C*T*G*T*A*A*T*A*A*G*T-3′). Following the manufacturer’s guidelines, cells were transfected with Lipofectamine™ 2000 transfection reagent and Plus™ reagent (Thermo Fisher Scientific, Germany). Initially, 25 × 10⁴ cells were seeded into six-well plates containing 2 ml of RPMI-1640 medium and cultured for 24 hours. The AURKA antisense LNA GapmeRs, as well as negative control and mock control, were designated as GapmeR, NC, and MC, respectively. Subsequently, 256 µl of optiMEM, 2.1 µl of Lipofectamine™ 2000, and 2.1 µl of Plus™ reagent were added to each tube, followed by the addition of 10.5 µl of the working solution (2 µM), with incubation at room temperature for 30 minutes. Finally, 1742 µl of complete medium and 258 µl of mixed transfection reagents were added to each well and cultured for 24 or 48 hours.

Total RNA was extracted using the RNeasy Plus Mini Kit (Qiagen, Germany). The reverse transcription reaction was carried out using the RevertAid RT Kit (Thermo Fisher Scientific, Germany). AURKA expression was quantified through qRT-PCR, with β-actin serving as the normalization reference (SYBR Green JumpStart Taq ReadyMix Kit, Sigma-Aldrich, Germany; AURKA: forward primer, GTCCCTGAGTGTCCTTGGC, reverse primer, GCAATGGAGTGAGACCCTCT; β-actin: forward primer, GTCTTCCCCTCCATCGTG, reverse primer, AGGGTGAGGATGCCTCTCTT).

For protein extraction, RIPA buffer was used, and the protein concentration was assessed following Bradford’s method. Total protein (20-µg sample) was separated using 12% sodium dodecyl sulfate polyacrylamide gels and transferred onto nitrocellulose membranes. Subsequently, the membranes were blocked with 5% milk powder and incubated with specific primary antibodies at 4°C overnight (AURKA, 1:1000 dilution, #14475; Cell Signaling Technology, USA; β-actin, 1:1000 dilution; A1978 Sigma-Aldrich, Germany), followed by incubation with a conjugated secondary antibody. Protein bands were visualized using an ECL chemiluminescence detection reagent (Millipore, Billerica, MA, USA) and analyzed and quantified using ImageJ software.

A total of 5,000 cells were seeded in 96-well plates and incubated for 24 hours. Subsequently, cells were treated with Aur-I (Aurora-A Inhibitor I), a specific inhibitor for AURKA (Sigma-Aldrich, Germany) or AURKA GapmeR, and incubated for 24 or 48 hours, respectively. Cells were stained with PrestoBlue™ cell viability reagent (Thermo Fisher Scientific, Germany) at 37°C for 10 minutes, and cell viability was assessed using SpectraMax M5 (Molecular Devices, Sunnyvale, CA, USA).

In a six-well plate, 500 cells were seeded and incubated for 24 hours. The medium was subsequently replaced with Ctr (0.1% dimethyl sulfoxide [DMSO]), 1 or 5 µM Aur-I, or AURKA GapmeR, and cells were incubated for 24 hours. The medium was then removed, and fresh medium was added for an additional 10-12 days. Cells were fixed with 3.7% paraformaldehyde, stained with 0.5% crystal violet, and the number of colonies was quantified using ImageJ software.

Cells were seeded and treated with Aur-I (1 µM and 5 µM) or AURKA GapmeR for 24 hours, followed by harvesting. Fixation was performed using 70% ethanol overnight. The cells were subsequently washed with PBS, stained with a propidium iodide staining solution, and analyzed through flow cytometry (BD LSRFortessa™ Cell Analyzer, BD Biosciences, Germany).

The Boyden chamber assay was carried out using 24-well trans-well inserts (ThinCert™, Greiner Bio-One, Germany). Cells were treated with AURKA GapmeR for 48 hours. The lower surface of the insert was coated with 10 µg/ml fibronectin and incubated at 37°C for 2 hours. Cells were seeded (approximately 5,000 cells) into the upper chamber in serum-free RPMI medium, while 500 µl of serum-containing medium was transferred to the lower chamber. Subsequently, cells were fixed with 3.7% paraformaldehyde, stained with 0.1 µg/ml DAPI (4′,6-diamidino-2-phenylindole) for 24 hours, and imaged using fluorescence microscopy (Axio-Observer, Zeiss, Germany). Cells from randomized regions of three independent experiments were counted. Cell counts were determined using ImageJ software.

Briefly, 3,500 cells (OE33 and OE19) were seeded into 96-well plates (X40 Mikrotiterplates (inertGrade), Brandt, Germany) and incubated for 24 hours. Cells were then treated with Aur-I (1 µM, 5 µM). Spheroids were monitored and measured every 4 hours using CELLCYTE X™ (CYTENA, Germany).

Bioinformatics analyses were conducted using R software (version 3.6.1). Group differences were assessed using unpaired Student’s t-test or one-way analysis of variance (ANOVA). Correlation between AURKA and variables, including immune score, ESTIMATE score, TMB, MSI, and TIDE, was evaluated using Spearman or Pearson correlation tests. Visualizations for in vitro experiments were generated using GraphPad Prism (version 9.0; La, Jolla, CA, USA). For differences obtained, the data are expressed as mean ± standard error of the mean. Differences were considered statistically significant at p-value < 0.05.

We conducted a comprehensive analysis by combining normal samples from the GTEx dataset with tumor samples from the TCGA dataset to investigate the differential expression of AURKA across 27 different human cancer types. Our analysis revealed that AURKA exhibited high expression in 26 cancer types, including ACC, BLCA, and others (all p<0.001), as depicted in Figure 1A . Notably, AURKA expression was notably low in LAML tumor tissues (p<0.001). Additionally, immunohistochemistry results sourced from the THPA database supported our findings, showing elevated protein levels of AURKA in esophageal, stomach, liver, and rectal cancer tissues ( Figure 1B ).

To establish the prognostic significance of AURKA, we employed KM and univariate Cox regression analyses using the TCGA pan-cancer data. Our analysis of OS revealed that higher AURKA expression was associated with worse prognoses in most cancers, including ACC, ESCA, LGG, PCPG, THCA, BLCA, LIHC, PRAD, THYM, BRCA, HNSC, LUAD, UCEC, KICH, LUSC, SARC, KIRC, MESO, SKCM, UVM, KIRP, and PAAD (all p<0.05), as shown in Supplementary Figure 1A . Intriguingly, AURKA appeared to function as a protective factor in READ (p=0.025), COAD (p=0.019), OV (p=0.048), STAD (p=0.027), and LAML (p=0.027). Furthermore, our analysis of disease-specific survival (DSS) demonstrated that AURKA was a risk factor in ACC, ESCA, LGG, PCPG, THCA, BLCA, LIHC, PRAD, BRCA, HNSC, LUAD, UCEC, KICH, LUSC, SARC, KIRC, MESO, SKCM, UVM, KIRP, and PAAD (all p<0.05), while high AURKA expression correlated negatively with prognoses in COAD (p=0.027), OV (p=0.02), and STAD (p=0.037; Supplementary Figure 1B ).

Moreover, our univariate Cox regression analysis for OS identified AURKA expression as an independent risk factor for OS in ACC, BLCA, BRCA, COAD, ESCA, HNSC, KICH, KIRC, KIRP, LAML, LGG, LIHC, LUAD, LUSC, MESO, PAAD, PCPG, PRAD, READ, SARC, SKCM, STAD, THCA, THYM, UCEC, and UVM ( Figure 2A ). In terms of DSS, AURKA expression was associated with 24 tumor types, with higher AURKA expression acting as a protective factor in COAD (p=0.003), OV (p=0.02), and STAD (p=0.04; Figure 2B ). We further explored the relationship between AURKA and the UICC tumor stage, revealing that AURKA was linked to advanced tumor stages in ACC (p=0.0023), KIRC (p<0.001), LIHC (p=0.02), UCEC (p=0.018), KIRP (p<0.001), LUSC (p=0.049), LUAD (p=0.021), and TGCT (p=0.043; Figure 3 ). These findings strongly suggest that AURKA is closely associated with patient outcomes and holds potential as a prognostic factor across a wide range of human cancer types.

To gain insights into the biological pathways involving AURKA, we conducted GSEA focusing on KEGG and HALLMARK pathways ( Figure 4 ). Our KEGG analysis unveiled a negative correlation between AURKA expression and pathways related to the cell cycle, oocyte meiosis, and homologous recombination ( Figure 4A ). Conversely, pathways such as arachidonic acid metabolism, linoleic acid metabolism, and asthma displayed a positive association with AURKA expression ( Figure 4B ). Similarly, the HALLMARK pathway analysis demonstrated that AURKA was negatively linked to pathways such as mTORC1 signaling, G2M checkpoint, and MYC target genes ( Figure 4C ), while it exhibited a positive correlation with pathways such as KARS signaling DN, myogenesis, coagulation, and bile acid metabolism ( Figure 4D ). These results underscore AURKA’s role in regulating the cell cycle and bile acid metabolism.

To gain a deeper understanding of the relationship between AURKA and tumor immunity, we investigated the association between AURKA expression and both ESTIMATE and immune scores in various human cancers. Our analysis revealed significant correlations between AURKA expression and the ESTIMATE score in a range of cancers, including ACC, BRCA, CESC, COAD, ESCA, GBM, HNSC, KIRC, KIRP, LGG, LIHC, LUAD, LUSC, PAAD, READ, SARC, SKCM, STAD, TGCT, THCA, and UCEC ( Supplementary Figure 2 ). Additionally, AURKA exhibited correlations with the immune score in ACC, CESC, COAD, ESCA, GBM, KICH, KIRC, KIRP, LUAD, LUSC, PAAD, READ, SARC, SKCM, STAD, TGCT, THCA, and UCEC ( Supplementary Figure 3 ).

To delve deeper into the association between AURKA and TIME), we explored the correlation between AURKA expression and immunomodulators ( Figure 5 ). In terms of immune-activated genes, AURKA exhibited positive correlations with MICB, PVR, CD276, and ULBP1 and negative correlations with C10orf54 in most cancers ( Figure 5A ). Concerning immune inhibitors, AURKA displayed a positive correlation with IL10RB in most cancers but a negative correlation in COAD ( Figure 5B ). We further investigated the relationship between AURKA and chemokines and their relevant receptors, revealing a significant negative correlation between AURKA and CCL14 in most cancers ( Figure 5C ). Notably, in THCA, we observed a positive correlation between AURKA and multiple chemokine receptor genes ( Figure 5D ). Additionally, our analysis using the TIMER2 database indicated a positive association between AURKA expression and the levels of activated CD4+ T-cells and eosinophil cells in tumors, while a negative correlation was observed with the level of activated mast cells in tumors ( Supplementary Figure 4 ).

TMB and MSI are pivotal factors in tumorigenesis and progression and serve as essential biomarkers for immunotherapy. Our analysis unveiled a positive correlation between AURKA expression and TMB in several human cancers, including UCS, BRCA, CHOL, KICH, LGG, LUAD, PAAD, PRAD, SARC, and STAD (all p<0.05). Conversely, AURKA expression exhibited a negative correlation with TMB in THYM ( Figure 6A ). In terms of MSI, we observed a significant correlation between AURKA expression and the MSI score in specific cancers such as UCEC, ESCA, and SARC, while a low AURKA expression inversely correlated with MSI in DLBC (p<0.001; Figure 6B ).

Furthermore, we explored the connection between AURKA expression and copy number variations (CNV). Our analysis demonstrated a robust correlation between AURKA expression and CNV in most cancers ( Figure 6C ), highlighting the widespread presence of AURKA and relevant CNV alterations in tumors. Finally, we employed the TIDE score to assess the potential efficacy of immune checkpoint inhibitors (ICIs) in cancer treatment. Our study revealed a strong correlation between AURKA expression and the TIDE score in cancers such as GBM, LGG, LUAD, KIRP, ESCA, LUSC, LAML, SKCM, UCEC, HNSC, CESC, COAD, LIHC, BLCA, and OVM ( Figure 6D ). These findings suggest that AURKA expression could serve as a reference marker for predicting the response to ICI therapy in cancer patients.

To validate the robustness of AURKA expression and its prognostic value, we analyzed pan-cancer data from GEO cohorts. As depicted in Figure 7 and Supplementary Figure 5 , AURKA exhibited higher expression in tumor tissues than in normal tissues across various tumors, including ACC, BLCA, BRCA, COAD, EAC, HNSC, KICH, KIRC, KIRP, LGG, LIHC, LUAD, MESO, OV, PAAD, PRAD, READ, SARC, SKCM, STAD, THCA, and UCEC, with the exception of LAML. Moreover, high AURKA expression was consistently associated with worse outcomes in ACC ( Figure 8A ), BLCA ( Figure 8B ), BRCA ( Figure 8C ), EAC ( Figure 8E ), HNSC ( Figure 8F ), KIRC ( Figure 8G ), LGG ( Figure 8I ), LUAD ( Supplementary Figure 6A ), MESO ( Supplementary Figure 6B ), PAAD ( Supplementary Figure 6D ), PRAD ( Supplementary Figure 6E ), SARC ( Supplementary Figure 6G ), and SKCM ( Supplementary Figure 6H ). In contrast, high AURKA expression correlated positively with the prognosis in COAD ( Figure 8D ), LAML ( Figure 8H ), OV ( Supplementary Figure 6C ), READ ( Supplementary Figure 6F ), and STAD ( Supplementary Figure 6I ). These findings corroborate our analysis of TCGA data.

We observed AURKA overexpression in various cancers, including ESCA ( Figure 1 ), and its close association with patients’ OS and disease-free survival (DFS; Supplementary Figures 1A, B ). AURKA also exhibited significant correlations with ESTIMATE and immune scores, TMB, MSI, CNV, and the TIDE score in most cancers, including ESCA ( Supplementary Figures 2 , 3 ; Figure 6 ). These findings suggest that AURKA may play a pivotal role in the immune response within tumor progression and in predicting the clinical outcomes of ESCA and other cancers. Moreover, higher AURKA expression was more frequently observed in patients under 65 years of age and in those at the M1 phase in EAC ( Supplementary Figure 7 ). We subsequently shifted our focus to validate and explore the function of AURKA in ESCA, particularly in EAC cells.

To begin, we compared AURKA expression between normal esophageal mucosa and EAC cell lines. Quantitative real-time polymerase chain reaction (qRT-PCR) results demonstrated that AURKA was highly expressed in OE33 and OE19 cells derived from EAC when compared to squamous epithelium cell lines ( Figure 9A ). Knockdown of AURKA using AURKA GapmeR in OE33 and OE19 cells led to decreased expression of both AURKA mRNA and protein levels ( Figures 9B, C ). Subsequently, a cell viability assay revealed that inhibiting AURKA using Aur-I or AURKA GapmeR resulted in suppressed cell viability ( Figures 9D, E ). We also investigated the role of AURKA in prostate cancer (DU145) and pancreatic cancer (PaTu8988t) cells. Treating these cells with AURKA-GapmeR or Aur-I resulted in decreased AURKA mRNA expression and reduced cell viability in both prostate cancer ( Supplementary Figures 8A–C ) and pancreatic cancer cells ( Supplementary Figures 8D–F ). Further molecular and functional analyses were conducted in EAC cell lines OE33 and OE19. The colony formation assay demonstrated a significant reduction in the number of cell colonies following AURKA inhibition ( Figures 9F, G , and Supplementary Figure 9A ). To explore the impact of AURKA inhibition on the cell cycle, OE33 and OE19 cells were treated with AURKA-GapmeR or Aur-I for 24 hours. These results indicated that AURKA inhibition arrested the growth of OE33 and OE19 cells within the G2/M phase ( Figures 9H–K , and Supplementary Figure 9B ). Furthermore, the Boyden chamber assay revealed a decrease in cell invasiveness upon inhibiting AURKA ( Figure 9L and Supplementary Figure 9C ). Additionally, 3D spheroid cultures were established to investigate the function of AURKA in the growth of EAC spheroids. OE33 and OE19 spheroids were treated with a vehicle control (0.1% DMSO), Aur-I at 1 µM, or 5 µM. The results demonstrated that inhibiting AURKA with Aur-I significantly (p < 0.05) reduced the spheroid area after 10 days ( Figure 9M ).