Aurora Kinase A (AURKA) Regulates The Progression and Imatinib Resistance of Advanced Gastrointestinal Stromal Tumors

Background: Gastrointestinal stromal tumor (GIST) is a common tumor that originates from mesenchyme in the alimentary system. Compared to the typical gastrointestinal carcinomas, GISTs exhibit unique malignant behaviors. Bioinformatic tools and subsequent experiments were applied to investigate novel targets involved in GIST progression and imatinib resistance. Methods: Differences of gene expression pro�les between advanced and non-advanced GISTs were comprehensively analyzed based on Gene Expression Omnibus (GEO) dataset GSE136755. A protein-protein interaction (PPI) network was conducted to identify the potential target gene. Gene set enrichment analysis (GSEA) was used to elucidate relevant biological events of the target gene based on dataset GSE47911. Subsequently, immunohistochemistry and Kaplan-Meier analysis were performed to validate the prognostic value of the target gene in GISTs. Overexpression of the target gene was conducted to analyze its functions in proliferation, apoptosis, migration, and imatinib resistance of GIST/T1 cells. Results: In current study, a total of 606 differentailly expressed genes (DEGs) were screened based on dataset GSE136755 and the upregulated DEGs in advanced GISTs were mainly involved in cell division through functional annotations. The intersected hub gene, Aurora Kinase A (AURKA), was identi�ed by degree and bottleneck algorithms. GSEA revealed that AURKA was involved in cell cycle-related biological processes. Oncomine and GEPIA databases supported an elevated expression pattern of AURKA in most human malignances. Clinical assay demonstrated that AURKA could be an independent prognostic factor for GISTs. Additionally, overexpression of AURKA was experimentally demonstrated to promote cell proliferation and survival, and enhance imatinib resistance of GIST/T1 cells. Conclusions: These �ndings indicated that overexpression of AURKA promoted GIST progression and enhanced imatinib resistance, implying the potential of AURKA as a therapeutic target for GISTs.


Background
Gastrointestinal stromal tumor (GIST) is the most common mesenchymal tumor of the alimentary system and originates from the interstitial cells of Cajal (ICC) [1] .About 71% of GISTs present with KIT (71%) or PDGFRα (14%) mutations, leaving approximately 10% -15% of GISTs without any KIT or PDGFRα mutations which are reported as KIT/ PDGFRα wild type [2,3] .The malignant potential of GISTs is strati ed based on tumor size, mitotic index, and location according to the modi ed NIH criteria [4] .Based on these criteria, GISTs are classi ed as high-risk, intermediate-risk, low-risk, and very low-risk.The 5-year survival rate of advanced GIST patients is between 35% and 65% [5] .The main therapeutic option for primary localized GISTs is surgical resection.However, the recurrence rate for GISTs, even after complete surgical resection, is 40% -80% [5] .The median time to recurrence for most patients is approximately 12 -16 months [6] .Treatment with imatinib, a tyrosine kinase inhibitor (TKI) that targets KIT and PDGFRα, has improved the prognosis of GIST patients.However, when used to eliminate mature GIST cells, imatinib has a limited e cacy, and studies have suggested GIST persistence in prolonged TKI therapy [1] .Due to acquired resistance to imatinib, about 85% -90% of GIST patients usually experience disease progression within 20 -24 months [2,3,7] .Further researches are urgently needed to reveal the mechanism of GIST progression and explore novel therapeutic targets for imatinib resistance.
Currently, bioinformatic tools are being used to evaluate molecular signatures associated with progression and clinical outcomes in several types of malignancies [8][9][10] .In this study, the datasets GSE136755 and GSE47911 downloaded from the Gene Expression Omnibus (GEO) were used to evaluate the potential target genes involved in GIST progression.Among several hub genes in the dataset GSE136755, AURKA was considered as a key hub gene in GIST progression.Gene set enrichment analysis (GSEA) based on dataset GSE47911 recommended that AURKA appeared to promote GIST progression by regulating cell cycle processes.Subsequent clinical data analyses demonstrated the value of AURKA as a prognostic factor for GISTs.Furthermore, overexpression of AURKA was experimentally demonstrated to signi cantly promote cell proliferation and survival, and enhance imatinib resistance of GIST/T1 cells.

Data acquisition and differentially expressed genes (DEGs) identi cation
The dataset GSE136755 was downloaded from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) [11].The GSE136755 is based on the GPL17077 platform (Agilent-039494 SurePrint G3 Human GE v2 8x60K Microarray 039381), and includes clinicopathological information for 65 human GIST tumor samples without preoperative imatinib treatment.Primary GISTs with KIT mutations were selected to screen the differentially expressed genes (DEGs).High-risk GISTs were de ned as advanced samples (16 samples), while low-risk and very low-risk GISTs were considered as non-advanced samples (31 samples).Intermediate-risk GISTs were not included because they were not obviously special in terms of biological behaviors compared to the high-risk and low-risk GISTs.GEO2R, an R-associated web tool from the National Center for Biotechnology Information, was used to screen the DEGs between advanced and non-advanced GISTs [12] .The DEGs were identi ed using the cutoff values of: |log2FoldChange| > 1 and adju.p < 0.05.For hierarchical clustering analysis in Morpheus (https://software.broadinstitute.org/morpheus), the DEGs were reserved in text format.

Functional enrichment analysis of DEGs
The Database for Annotation, Visualization, and Integrated Discovery (DAVID) (http://david.ncifcrf.gov,version 6.8) is a web based bioinformatics resource that is used to extract functional annotation information of genes [13] .Gene ontology (GO) is a major bioinformatics tool for gene annotation and analysis [14,15] .The Kyoto Encyclopedia of Gene and Genome (KEGG) is a popular database for the analysis of advanced gene functions and potential signaling pathways in large-scale molecular data [16,17] .DAVID was used to perform GO and KEGG enrichment analyses of DEGs.The cut-off criterion was a false discovery rate (FDR) < 0.05.

Protein-protein interaction (PPI) network and module analysis
The Search Tool for the Retrieval Interacting Genes (STRING) (http://string-db.org) is an online database that is used to identify interactions among DEGs [18,19] .A con dence score ≥ 0.7 was set for conducting the PPI network.Cytoscape (version3.7.1), is an open-source bioinformatics software platform that is used for visualizing the PPI network and for further analyses [20,21] .The plugin-in Molecular Complex Detection (MCODE) of Cytoscape was performed to identify signi cant modules based on the PPI network topology.The criteria were: degree cut-off = 2, node score cut-off = 0.2, K-core = 2, and Max.depth = 100.Plug-in apps, ClueGO, were used to analyze and visualize the biological processes and pathways in signi cant modules.

Hub gene identi cation and analysis
Plugin-in cytoHubba of Cytoscape was used to identify hub genes based on the degree and bottleneck algorithms.The Oncomine database is an online platform that computes gene expression signatures, clusters, and gene-set modules [22] .Gene Expression Pro ling Interactive Analysis (GEPIA) database is a newly developed web server for cancer, normal gene expression pro ling and interactive analyses [23] .The cytoHubba was used to identify the hub genes, whose expression patterns were evaluated in common human malignancies using the Oncomine and GEPIA databases.
Clinicopathological features and KIT/PDGFRα mutation types were extracted from GSE136755 and from the raw data provided by Lagarde et al. [24] Correlations between key gene expression patterns and clinicpathological features as well as KIT/PDGFRα mutation types were statistically determined, with a P < 0.05 set as the threshold for statistical signi cance.

Gene set enrichment analysis (GSEA)
GSEA is a computational method that determines whether a priori de ned gene set shows statistically signi cant and concordant differences between two biological states [25,26] .The GSEA computes biological information from different perspectives and further elucidates on relevant biological events.

Immunohistochemistry and survival analysis
Between 2001 and 2015, a total of 49 patients admitted to the First A liated Hospital of Zhejiang University (Zhejiang Province, China), who were diagnosed with GISTs, were enrolled in this study.None of the patients had incomplete resection, neoadjuvant and adjuvant imatinib treatment, and a family history of GISTs.Clinical strati cation of GISTs was based on the modi ed NIH criteria [4] .Para nembedded GIST samples were obtained from the study participants and used to perform immunohistochemical (IHC) staining as well as survival analysis.All GIST tissue samples were provided by the Department of Pathology, the First A liated Hospital of Zhejiang University.The use of human tumor samples and clinical data in this study was approved by the Ethical Committee, the First A liated Hospital of Zhejiang University.Study participants were required to sign a written informed consent before enrollment.IHC staining was performed as previously described [28] .Brie y, tissue sections were incubated at 4℃ overnight with anti-human AURKA rabbit polyclonal antibody diluted 1:500 (NOVUS Biologicals, USA).A total of ve adjacent elds using ×400 magni cations in areas with the highest density of positive staining were scored according to the summation of the percentage of staining intensity.The immunostaining percentage was de ned as 0 (< 5%), 1 (< 20%), 2 (20-50%), and 3 (> 50%).Staining intensity was de ned from 0 (no staining) to 3 (strongest staining).The maximum score of IHC staining was 6, corresponding to more than 50% of the cells with the strongest staining intensity.The staining scores of less than the mean value were considered low expression, while scores more than the mean value were considered high expression.The follow-up time for all the patients was calculated from the surgical date to disease recurrence or last visit dates.

Cell culture
The GIST cell line (GIST/T1) donated by Prof. Wenbin Chen (Zhejiang University, Hangzhou, China), was obtained from the Cell Bank of the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China).Cells were maintained in RPMI-1640 medium (GIBCO) supplemented with 10% heatinactivated fetal bovine serum (GIBCO) in a humidi ed atmosphere with 5% CO 2 at 37℃.

Construction and transfection of lentiviral vectors for AURKA
The DNA fragment encoding the AURKA sequence was synthesized and inserted into the lentivirus expression vector, pLVX-IRES-tdtomato (TaKaRa, China).The resulting vector was identi ed as pLVX-AURKA-IRES-tdtomato.Lentiviral plasmids were transfected into HEK 293T cells with psPAX2 and pMD2.G at a ratio of 4:3:1 using Lipofectamine 2000 (Invitrogen, USA).Then, 48 hours after transfection, the virus was isolated.GIST/T1 cells were infected with lentivirus for 48 h, and the transfection e ciency was assessed by PCR and western blotting.

RNA extraction and real-time quantitative PCR (RT-qPCR)
Total RNA was extracted using the TRIzol reagent (Generay Biotech, China) according to the manufacturer's instructions.The extracted RNA was then treated with RQ1 RNase-Free DNase (Promega, USA).Then, reverse transcription was performed using PrimeScriptTM RT Master Mix (Takara, China) according to the manufacturer's instructions.RT-qPCR analysis was performed to con rm the expression levels of AURKA using CFX Connect Real-Time System (BIO-RAD, USA) with the SuperReal PreMix Color SYBR Green kit (Tiangen, China).The primer sets used for RT-qPCR are shown in Table 1.Gene expression levels were normalized against the internal control using the 2 −ΔΔCT method.

Western Blotting
Protein lysates from each sample were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and transferred to a polyvinylidene di uoride (PVDF) membrane (Merck Millipore, USA).The membrane was blocked with 5% non-fat milk or 5% BSA in TBST (tris-buffered saline with 0.1% Tween 20) for 1 h at room temperature.Then, the membrane was incubated with a primary antibody AURKA (1:1000 dilution, NOVUS Biologicals, USA) overnight at 4℃, washed thrice using 0.1% TBST buffer for 30 min, probed with Goat anti-Rabbit IgG-HRP secondary antibody (1:2000 dilution; Santa Cruz, USA), and washed thrice using 0.1% TBST buffer for 30 min.The HRP conjugated secondary antibody was detected and visualized using the enhanced chemiluminescence detection system (GE Healthcare, USA).Band intensity was quanti ed by densitometry using the ImageJ software (version 1.49; National Institutes of Health, USA) [29] .

Proliferation assay
Cell proliferation was assessed by the CCK8 assay (Tongren Chemical Society, Japan).Brie y, 1 × 10 5 cells per well were seeded in 96-well plates and incubated in a 5% CO 2 atmosphere at 37℃ for 24 h.Imatinib (3 μM) was then added to the culture medium for 48 h to evaluate the effect of AURKA on imatinib resistance.The medium was discarded, replaced with serum-free medium as well as CCK8 (10 μl) and incubated for 2 h.The Biokinetics Reader (MD corporate, USA) was used to detect absorbance at 450 nm.

Apoptosis assay
Cell apoptosis was assessed using the Annexin V-APC/7-AAD apoptosis kit (MultiSciences, China), according to the manufacturer's instructions.Brie y, 3 × 10 5 cells per well were seeded in 6-well plates and incubated in a 5% CO 2 atmosphere at 37°C for 24 h.Imatinib (3 μM) or solvent, were added to the culture medium for 48 h, after which the cells were collected.After washing twice in PBS at 4°C, cells were re-suspended in a binding buffer (500 μl).Annexin V-APC (5 μl) and 7-AAD (10 μl) were added into the suspension and incubated for 5 min at 4°C.The apoptosis index was examined by ow cytometry (ACCURI C6; BD, USA).

Migration assay
Cell migration was assessed in 24-well trans-well chambers with upper and bottom culture compartments separated by a polycarbonate membrane with 8 μm pore size.7.5 × 10 4 cells at logarithmic stage were seeded in the upper chamber.The culture medium with/without imatinib (3 μM), was added to the bottom chamber.Cells were incubated for 72 h and later xed in 4% paraformaldehyde.Non-migrated cells were washed and the inserts air-dried.Then, cells were incubated in 200 μl of 1% crystal violet for 10 min at room temperature.The number of cells that invaded the gel were counted in 3 elds (×100) using an inverted microscope (IX73; OLYMPUS, Japan).

Statistical analysis
All statistical analyses were performed using the SPSS 25.0 software (SPSS Inc., USA).Descriptive data were expressed as mean ± SD.Comparison of more than two mean values was performed by one-way analysis of variance (ANOVA) while the student's t-test was used to compare two mean values.Kaplan-Meier analysis was performed to establish disease-free survival (DFS) curves, while the log-rank test was used for survival curve comparison.The COX proportional hazards model was then used to perform multivariate analysis, while the Forward Likelihood Ratio method was used to identify independent variables.P < 0.05 was set as the threshold for statistical signi cance.

Identi cation of differentially expressed genes (DEGs)
A total of 65 human GIST samples without preoperative imatinib treatment were included in dataset GSE13675.A cohort of 47 samples with KIT mutations, consisting of 16 advanced (high-risk) GIST patients and 31 non-advanced (low-risk and very low-risk) GIST patients, was used to screen the DEGs using GEO2R.A total of 606 genes (244 upregulated and 362 downregulated) were identi ed using the cut-off criteria of adj.p < 0.05 and Log2FoldChange > 1.The top 50 up-and down-regulated genes were presented in a heatmap (Figure 1A, supplementary le 1).

Functional annotation of DEGs
To evaluate the biological clustering of DEGs, GO and KEGG analyses for the up-and downregulated DEGs were performed using DAVID.Based on GO analysis, upregulated DEGs were found to be signi cantly enriched in cell division, sister chromatid cohesion, mitotic nuclear division, microtubulebased movement, and mitotic metaphase plate congression.Cellular component (CC) of the upregulated DEGs was signi cantly enriched in midbody, kinesin complex, spindle, spindle microtubule, and condensed chromosome kinetochore.Molecular function (MF) of the upregulated DEGs was signi cantly enriched in microtubule motor activity and microtubule-binding.KEGG analysis showed that the upregulated DEGs were mainly involved in cell cycle and oocyte meiosis.GO analysis revealed that the down-regulated DEGs were signi cantly enriched in the interferon-gamma-mediated signaling pathway, type I interferon signaling pathway, and antigen processing and presentation of antigens.The CC of the down-regulated DEGs was signi cantly enriched in the integral component of the lumenal side of the endoplasmic reticulum membrane and ER to Golgi transport vesicle membrane while the MF of the downregulated DEGs was signi cantly enriched in peptide antigen binding and MHC class II receptor activity.KEGG analysis further revealed that the down-regulated DEGs were mainly involved in Graft-versus-host disease, Type I diabetes mellitus, Allograft rejection, Antigen processing and presentation, and Viral myocarditis.These results are presented in Figure 1B (supplementary le 2).

Module analysis through the PPI network of DEGs
To clarify the DEGs functionally, STRING was used to conduct a PPI network, which was composed of 603 nodes and 1768 edges.The PPI enrichment p-value was < 1.0E-16.The PPI network was visualized by Cytoscape and further analyzed by plug-in MCODE.Two most signi cant modules (module 1 and module 2) were identi ed and analyzed using GO, KEGG and REACTOME annotations to infer their biological functions.The module 1 (MCODE score = 36.667)was mainly involved in in cell cycle-related biological processes and signaling pathways, while the module 2 (MCODE score = 18.759) was mainly involved in immunological processes and signaling pathways.These results are presented in Figure 2A (supplementary le 2).

Identi cation of hub genes
To reveal the crucial genes underlying the regulation of GIST progression, we ltered hub genes among DEGs using the plugin-in cytoHubba of Cytoscape.Two algorithms, degree and bottleneck, were applied to weight the DEGs.The degree algorithm calculates the relevance and abundance of genes, while the bottleneck algorithm evaluates key genes positions in an entire regulatory network.According to the degree algorithm, the top 15 hub genes were CDK1, KIF11, KIF2C, CENPE, KIF20A, BUB1, CCNA2, CCNB1, AURKA, MAD2L1, CDCA8, KIF4A, CENPF, NDC80, and KIF23, with their scores ranging from 58 to 48.According to the bottleneck algorithm, the top 15 hub genes were AURKA, FN1, CD44, VEGFA, IL6, HLA-DQB1, HLA-DPA1, CXCL8, NT5E, ANK2, FOXM1, CHEK1, STAT1, CDC25A, and IFIH1, with their scores ranging from 64 to 8. A Venn diagram was used to identify the key hub genes by taking intersection of the two hub gene cohorts.The result showed that AURKA was the only overlapping hub gene (Figure 2B).

Gene Set Enrichment analysis (GSEA)
A total of 15 human gastric GIST samples in GSE47911 includes 6 high-risk cases, 1 intermediate-risk case, 3 low-risk cases, and 5 very low-risk cases.To further verify signi cant biological processes associated with AURKA expression, GSE47911 gene pro les were divided into two groups after which GSEA was performed based on AURKA expression level.Samples with the highest (25%, 4 samples) and lowest (25%, 4 samples) expression levels were selected for further analysis using GSEA.Cell cyclerelated gene sets were associated with elevated AURKA expression (Figure 2C, supplementary le 3).

Correlation between AURKA expression and clinicopathological features in GISTs
To evaluate the clinical signi cance of AURKA expression in GISTs, AURKA expression levels in 49 GIST tissues were assessed by IHC staining (Figure 3A).The correlations between AURKA expression and clinicopathological features (age, gender, location and risk strati cation) were determined (Table 2, supplementary le 4).AURKA expression was closely associated with tumor risk strati cation (Figure 3B; P < 0.001).The clinical signi cance of AURKA expression in GISTs was also evaluated using the data from GSE136755 and the raw data provided by Lagarde et al. [24] (Table 3, Table 4, and Figure 4, supplementary le 5).Findings from the GSE136755 revealed signi cant associations between AURKA expression and tumor risk strati cation (P < 0.001) as well as tumor stage (P < 0.001), while analyses of the raw data from Lagarde et al. [24] also showed a signi cant association between AURKA expression and tumor risk strati cation (P < 0.001) as well as tumor recurrence (P < 0.001) and metastasis (P < 0.001).However, apart from GSE136755 which revealed a signi cant association between AURKA expression and tumor location (P = 0.018), the data provided by Lagarde et al. did not establish a signi cant association between AURKA expression and tumor location (P = 0.156).
To determine the prognostic value of AURKA expression in GISTs, Kaplan-Meier survival analysis was performed.The range of observation time was 9 -79 months.As shown in Figure 3C, GIST patients with elevated AURKA expression exhibited poorer DFS than those with low AURKA expression level (43.25±6.94months vs 98.48±3.44 months P < 0.001).COX proportional hazard model showed that AURKA could be used as an independent prognostic marker for GISTs (P = 0.002).
Through analysis of the raw data provided by Lagarde et al. [24] , which contained 45 imatinib-sensitive and 15 imatinib-resistant samples, it was shown that there was a strong association between AURKA expression and imatinib-resistant gene mutations (Figure 4B, P = 0.018).

AURKA expression patterns in common human malignancies
To determine whether elevated AURKA expression is common in human digestive malignancies, mRNA expression levels of AURKA in stomach carcinoma, liver hepatocellular carcinoma, and colorectal carcinoma were evaluated using data from the GEPIA database.AURKA expression was found to be signi cantly up-regulated in all the above malignancies by comparison to the normal tissues.Findings from the Oncomine database also supported the upregulated expression of AURKA in most human malignancies (Figure 3D).

AURKA overexpression enhances GIST/T1 cell proliferation, survival, and migration
To assess the biological effects of AURKA expression in GISTs, AURKA was overexpressed in GIST/T1 cells transfected with the AURKA-expressing virus, which was de ned as the AURKA overexpression group (AURKA group).Normal GIST/T1 cells (Blank group) and GIST/T1 cells transfected with vacant plasmids (Vector group) were considered as the control groups.The transfection e ciency was determined by the red uorescence from tdtomato, and quanti ed by RT-qPCR and western blotting.Figure 5 showed that compared to the Blank and Vector groups, AURKA was overexpressed in GIST/T1 cells transfected with AURKA-expressing virus (the AURKA group).(supplementary le 6) The CCK8 assay was performed to assess the effect of AURKA overexpression on cell proliferation.Compared to the Blank and Vector groups, the overexpression of AURKA signi cantly enhanced GIST/T1 cell proliferation (P = 0.018) (Figure 6A).Imatinib treatment signi cantly inhibited cell proliferation in all the three groups.However, compared to the control groups, AURKA overexpressed cells still showed a relatively higher proliferation rate in the presence of imatinib (P < 0.001, Figure 6A).(supplementary le 6) We also established that AURKA overexpression markedly suppressed the apoptotic process in GIST/T1 cells (P < 0.001, Figure 6B).Similar result was observed in the circumstance of imatinib administration.
Compared to the Blank and Vector groups, AURKA overexpression obviously inhibited cell apoptosis after imatinib administration (P < 0.001, Figure 6B).The results suggested that the AURKA overexpression enhanced resistance of GIST cells to imatinib.(supplementary le 6) Apart from cell proliferation and apoptosis, the effect of AURKA overexpression on cell migration was also evaluated.The number of AURKA overexpressed cells that passed through the membrane was signi cantly higher than that of the other two control groups (for AURKA group, rate = 47.00 ± 5.06%; for Blank group, rate = 26.17± 3.66% for Vector group, rate = 30.00± 4.15%) in the absence of imatinib (P < 0.001, Figure 6C).However, imatinib signi cantly suppressed cell migration in all three groups (P < 0.001, Figure 6C) and effectively eliminated the contribution of AURKA overexpression to cell migration (P = 0.169, Figure 6C).(supplementary le 6)

Discussion
GIST is a common mesenchymal malignancy of the human alimentary system.Compared to gastrointestinal carcinomas, GIST is known to possess special biological features.For example, lymph node metastasis is not common in GIST and occurs preferably in young patients [31,32] .A 1-or 2-cm macroscopic margin may be su cient to achieve microscopically negative margins [33] .GIST responds poorly to conventional chemotherapy and radiotherapy [34] .As such, to better understand GIST biological behavior and inform the development of therapeutic strategies, it is conductive to establish the crucial genes that regulate the malignant behavior of GIST.Bioinformatics advances have been useful in exploring molecular targets for the progression and prognosis of GIST [11,35,36] .
In this study, gene pro les of 47 GIST samples from the dataset GSE136755 were selected for further analyses.Comparison of the gene pro les between advanced and non-advanced GISTs generated 244 upregulated DEGs and 362 downregulated DEGs.Functional annotation based on GO and KEGG analyses showed that upregulated DEGs were mainly enriched in cell cycle-related biological processes and signaling pathways, while the downregulated DEGs were mainly enriched in immune-related biological processes and signaling pathways.STRING database and Cytoscape software were used for further exploration of DEGs.Two important modules were extracted and visualized.Module 1 was consisted of upregulated DEGs and mainly involved in cell cycle-related biological processes and signaling pathways, while module 2 was consisted of downregulated DEGs and mainly involved in immunological processes and signaling pathways.This indicates that the difference in gene pro les between advanced and non-advanced GISTs is mainly re ected in cell cycle and tumor immunity.
Based on the degree and bottleneck algorithms, plug-in cytoHubba in Cytoscape software was used to screen for novel key genes associated with GIST progression.The degree algorithm calculates the relevance and abundance of genes, while the bottleneck algorithm evaluates key gene positions in an entire regulatory network.In this study, a signi cant key gene, AURKA, was identi ed using a Venn diagram.AURKA is a protein-coding gene that encodes a cell cycle-regulated kinase involved in microtubule formation and/or stabilization at the spindle pole during chromosomal segregation.It has been documented that AURKA promotes tumor progression by enhancing cell cycle progression, cell survival, genomic instability, epithelial-mesenchymal transition (EMT) and stem-like properties of cancer cells. [37]In most solid tumors, AURKA regulates cell cycle checkpoints and promotes the cell cycle process. [37]The GSEA based on GSE47911 gene pro les further validated the association between AURKA overexpression and cell cycle progression in GISTs.
To con rm the importance of AURKA expression in GISTs, we performed IHC staining to establish the associations between AURKA expression and clinicopathological characteristics of the enrolled 49 GIST patients.In advanced GISTs, the expression level of AURKA was found to be elevated.This result is consistent with the analyses of the data provided by GSE136755 and Lagarde et al. [24] Survival analysis further showed that AURKA overexpression was a potential independent prognostic factor for GIST patients.Furthermore, a series of in vitro experiments demonstrated that GIST cells with overexpressed AURKA had signi cant malignance advantages, such as cell proliferation, anti-apoptosis and cell migration.The ndings validated the results of bioinformatics analyses.Drug resistance is a major obstacle in cancer chemotherapy and greatly affects a patient's prognosis.
Adjuvant imatinib has been widely used as a rst-line therapeutic option for advanced GIST patients [38,39] .However, imatinib resistance has increased in recent years.By analyzing the raw data provided by Lagarde et al. [24] , a signi cant relationship was found between AURKA overexpression and gene mutations causing imatinib resistance.The GSE136755 data showed a similar but not signi cant result, which could be attributed to the small sample size.In addition, in vitro experiments showed that AURKA overexpression enhanced resistance of GIST cells to imatinib by promoting cell proliferation and inhibiting cell apoptosis.
The role of AURKA overexpression in tumor progression has been reported in a variety of human malignancies.AURKA phosphorylates RPS6KB1 and promotes cell proliferation as well as survival [40] .AURKA also stabilizes the transcription factor N-MYC, thereby promoting G1/S cell cycle transition and tumor cell proliferation [41] .Pharmacological inhibition of AURKA promotes the chemosensitivity of cervical cancer cells [42] .Compounds targeting AURKA, particularly alisertib, have been extensively studied in preclinical models, where they have shown synergistic effects with other targeted therapies, leading to tumor regression in a variety of cancer models [43] .Yeh et al. con rmed the contribution of the AURKA inhibitor, MLN8237, on the suppression of metastatic GISTs [44] .Findings from GEPIA and Oncomine databases also supported contribution of AURKA overexpression to tumorigenesis.This study still had the following limitations.First, the case number in each GIST cohort is not large.Therefore, we made up for this de ciency to some extent by incorporating different GIST cohorts for comprehensive analysis.Second, this study is the preliminary exploration and certi cation of AURKA as a therapeutic target.As for the potential of AURKA-targeting therapy, we have not carried out experiments for investigation.Further validation based on various in vitro and in vivo experiments needs to be performed.

Conclusions
In conclusion, our ndings demonstrated the signi cant overexpression of AURKA in advanced GISTs by bioinformatics analyses, which predicts a poor prognosis of GIST patients.Overexpression of AURKA was experimentally demonstrated to promote cell proliferation and inhibit cell apoptosis of GIST cells, which contributes to imatinib resistance, implying the potential of AURKA as a therapeutic target for GISTs.

Figures Figure 1
Figures

Figure 3 Expression
Figure 3

Figure 4 Expression
Figure 4

Figure 5 Construction
Figure 5

Table 1
Ethical approval and consent to participateThis study was by the Ethical Committee of the First A liated Hospital, Zhejiang University School of Medicine.XC and WW, study conception and design, writing, review, and revision of the manuscript; XC and JW, bioinformatics analyses; WF and SL, in vitro experiments; XC, data acquisition, analysis and interpretation.All authors read and approved the nal manuscript.43.OttoT, Sicinski P. Cell cycle proteins as promising targets in cancer therapy.Nat Rev Cancer.2017;17:93-115.44.Yeh CN, Yen CC, Chen YY, Cheng CT, Huang SC, Chang TW, et al.Identi cation of aurora kinase A as an unfavorable prognostic factor and potential treatment target for metastatic gastrointestinal stromal tumors.Oncotarget.2014;5:4071-86.Sequences of primers used for Quantitative real-time PCR analysis.

Table 2
Correlation between AURKA expression level and clinical features in 49 GIST patients *P < 0.05 was considered statistically signi cant Table3Correlation between AURKA intensity and clinical features based on GSE136755

Table 4
Correlation between AURKA intensity and clinical features based on the published raw data