We collected three gene datasets, GSE81277, GSE21034 and GSE35988, from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). GSE81277 included the RNA sequence data of LNCaP PCa cells resistant to vinblastine. Three vinblastine sensitive and three resistant samples were obtained from GSE81277. Furthermore, GSE21034 and GSE35988 included the RNA and clinical data of patients with PCa. Additionally, the clinical data of patients with PCa were collected from both The Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov/) and Chinese Prostate Cancer Genome and Epigenome Atlas (CPGEA) (http://www.cpgea.com) databases.

The primary RNA-sequence data obtained from different databases were normalized using R software (version 4.0.3). Based on the document comments, the expression matrix including probe ID was substituted by the corresponding gene ID. The genes with |log2FC > 1| and P < 0.05 were considered as critical genes. The genes were reflected in volcano map made by R software “Enhancedvolcano” package. Additionally, the clinical data from different databases were downloaded for further study.

The pathways of the enriched hub genes were analyzed using the online tool Metascape (http://metascape.org/), and the bubble map was constructed using R software “ggplot2” package. Furthermore, the PPI network was constructed using STRING (https://cn.string-db.org/).

Online tools, such as UALCAN (http://ualcan.path.uab.edu/), gene expression profiling interactive analysis (GEPIA) (http://www.gepia.cancer-pku.cn/) and Tumour Immune Estimation Resource (TIMER) (https://cistrome.shinyapps.io/timer/) were used for analysis.

Based on the clinical data from different databases, the expression of hub genes in different tumour stages was analyzed. Additionally, using GEPIA the hub genes that influence overall survival (OS) and disease-free survival (DFS) in patients with PCa were also analyzed.

The correlation and mutation type of the identified hub genes and immune cells in PCa were analyzed using TIMER.

Clinical PCa and CRPC specimens were collected from Tongji Hospital, School of Medicine, Tongji University. The collection method was approved by the Ethics Committee of Tongji Hospital, School of Medicine, Tongji University (SBKT-2021-220). Patients who provided the samples were informed of the experiment and gave informed consent.

PCa cell lines were purchased from the Chinese Academy of Science Cell Bank (Shanghai, China). The human CRPC cell lines C4-2 and 22Rv1 were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Catalog No. R8758, Sigma, Darmstadt, Germany) containing 10% fetal bovine serum (FBS) (Catalog No. 10091, Gibco, Thermo Fisher Scientific, Waltham, MA, USA). The cells were cultured in a humid environment with 5% CO2 and 95% air at 37°C. Vinblastine (Catalog No. S4505) was purchased from SelleckChem (Houston, TX, USA). The CRPC cells were treated with vinblastine (3.25 nmol/L) for 24 h.

Cell transfection assays were performed with Lipofectamine 2000 (Catalog No. 11668019, Thermo Fisher Scientific). shRNA lentivirus was constructed for specific gene knockdowns. The shRNAs were purchased from the Youze Biotechnology Company. Additionally, blank control lentivirus (shControl) without knockdown specific genes was also constructed by the Youze Biotechnology Company. The shRNA sequence was as follows: shCCNB1: GCAGCACCTGGCTAAGAATGCAGCACCTGGCTAAGAAT and shAURKA: CCGGCCTGTCTTACTGTCATTCGAACTCGAGTTCGAATGACAGTAAGCAGGTTTTTG.

The total RNA was extracted from CRPC cell lines utilizing TRIzol Reagent (Sigma–Aldrich, St. Louis, MO, USA, Catalog No. T9424). cDNA was transcribed using the reverse transcription kit (Advantage^® RT-for-PCR Kit, Takara Bio Inc., Kusatsu, Japan, Catalog No. 639505). Finally, we measured the volume of cDNA using a real-time PCR kit (TB Green^® Premix Ex Taq™ II, Takara Bio Inc., Catalog No. RR420A) according to the manufacturer’s instructions. The primers of CCNB1, AURKA and GAPDH are shown in Table 1 . The 2^−ΔΔCt method was used to quantify mRNA expression levels.

Rabbit monoclonal anti-CCNB1 antibody (Catalog No. ab156447), anti-AURKA antibody (Catalog No. ab108353) and anti-GAPDH antibody (Catalog No. ab9485) were purchased from Abcam (Abcam UK, Cambridge, UK).

Tissue samples and cell line proteins were extracted with RIPA lysis buffer. Protein samples were treated with Dual Colour Protein Loading Buffer (Thermo Fisher Scientific, Waltham, MA, USA). Sodium dodecyl-sulfate polyacrylamide gel electrophoresis (10%) was used to separate proteins, which were then transferred to nitrocellulose membranes (Merck KGaA, Darmstadt, Germany). Protein-Free Rapid Blocking Buffer (Thermo Fisher Scientific) was utilized to block the membranes. Then, the membranes were incubated at 4°C overnight with primary antibodies against CCNB1 (1:1000), AURKA (1:1000) and GAPDH (1:1000) (Abcam UK, Cambridge, UK). On the second day, the membranes were washed thrice using 1×TBST (10 min/cycle). Then, the membranes were incubated at normal temperature for 1.5 h with a matched secondary antibody (Catalog No. A0208, HRP-labeled Goat Anti-Human IgG (H+L), Beyotime Biotechnology, Shanghai, China). Finally, the membranes were exposed to X-ray film.

Cell proliferation ability was detected using Cell counting kit-8 (CCK-8) (Dojindo, Japan). Briefly, cells were placed in 96-well plates (3000 cells/well) and cultured with 200 µL RPMI 1640 + 10% FBS for 0 h, 24 h, 48 h or 72 h. After culturing, cells were detected using CCK-8, following the manufacturer’s instructions. Absorbance at 450 nm was measured using a spectrophotometer (LD942, Beijing, China).

The matrix data were analyzed using R version 4.0.3 (Institute for Statistics and Mathematics, Vienna, Austria; https://www.r-project.org). Comparisons between two groups were performed using the Wilcoxon test, and the Kruskal–Wallis test was employed for comparisons between more than two groups. Hazard ratios (HRs), 95% confidence interval (95% CI) and P values were used as statistical metrics. Twotailed,P < 0.05 was deemed as statistically significant.

The potential hub genes that correlated with PCa resistance to vinblastine were identified. In the GEO database, we found a gene dataset, GSE81277, which included the RNA-sequence data of both vinblastine-sensitive and -resistant LNCaP cells. Subsequently, the volcano map revealed nine hub genes (CDC20, LRRFIP1, CCNB1, GPSM2, AURKA, EBLN2, CCDC150, CENPA and TROAP) from LNCaP PCa cell samples that correlated with vinblastine resistance ( Figure 1A ). Further, we constructed a heat map to reflect the specific expression of each hub gene in the GSE81277 dataset ( Figure 1B ).

The potential pathways and the correlation of each hub gene were analyzed using Metascape. The pathways enriched by the hub genes are illustrated in a bubble map. We found these genes mainly enriched in “microtubule cytoskeleton organization involved in mitosis” ( Figure 2A ). Then, using STRING, we constructed the PPI network to find the correlation between each hub gene. The network revealed that apart from EBLN2 and CCDC150, each hub gene correlated with other genes ( Figure 2B ).

After identifying the nine hub genes that were associated with vinblastine resistance in PCa, we examined the expression of these genes in PCa. We collected the sequence data of patients with PCa from different databases, such as TCGA, CPGEA and GEO. In TCGA, apart from LRRFIP1, all other gene expressions changed in PCa. Moreover, only GPSM2 was downregulated whereas other genes were upregulated in PCa ( Figure 3A ). As the TCGA data included samples from Western populations, we further tested the expression of these hub genes in different populations, including the Asian population. Using the CPGEA database, we collected the RNA-sequence data of the Chinese population. The data revealed a change in the expressions of these nine genes in Chinese patients with PCa. Furthermore, GPSM2 and EBLN2 were downregulated whereas other hub genes were upregulated ( Figure 3B ). Finally, the GSE21034 dataset was used to verify the results. In GSE21034, we found that the expression of LRRFIP1, CCNB1, GPSM2 and, AURKA changed in PCa samples, with GPSM2 being downregulated ( Figure 3C ). As methylation level can affect the function of genes, we further examined the methylation level of these nine genes in the TCGA database. However, data on EBLN2 could not be found and thus, only the methylation level of eight genes was tested. The methylation level of GPSM2, EBLN2, CCDC150 and TROAP showed a statistically significant change compared to the other genes ( Figure S1 ).

As vinblastine resistance-related hub genes have been reported to influence PCa occurrence, we further evaluated these genes’ role in PCa progression. As the Tumour-Node-Metastasis (TNM) staging has been widely used to determine the severity of PCa (10), we evaluated the expression of these nine hub genes in different TNM tumour stages. Depending on the clinical data from different databases, the function of these vinblastine resistance-related hub genes in PCa progression was analyzed. In the TCGA database, we found that GPSM2 expression was downregulated when PCa progressed from the T2 stage to the T3 stage ( Figure 4A ). As LIRRFIP1 was not differentially expressed between normal and tumour tissues, LIRRFIP1 was not included in the study. However, the other genes did not significantly affect the primary tumour (T) stage ( Figure 4A ). In Chinese patients with PCa, the nine genes also did not show a significant effect on the T stage ( Figure S2A ). Furthermore, the nine genes also did not significantly influence node metastasis ( Figures S2B, C ). Then, using the clinical data from GSE21034, we classified the patients with PCa into a primary tumour group and metastasis group based on the occurrence of distant metastasis. CDC20, CCNB1, AURKA, CDCC150 and TROAP were observed to influence distant metastasis ( Figure 4B ).

The GSE35988 dataset was used to verify the above results. In GSE35988, apart from LRRFIP1, all other genes influenced PCa occurrence ( Figure 5A ). Further, using the GSE35988 dataset, we classified patients with PCa into a primary tumour group and a CRPC group. Apart from GPSM2, other genes were upregulated in the CRPC samples ( Figure 5B ).

The influence of the eight hub genes on the survival status of patients with PCa was evaluated using GEPIA. LRRFIP1 did not significantly affect PCa occurrence; therefore, it was excluded from the study process. Apart from GPSM2 and EBLN2, all other genes were observed to influence the DFS of patients with PCa ( Figure 6 ). Furthermore, CDC20 and CDCC150 influenced the OS of patients with PCa ( Figure S3 ).

Immune infiltration has been reported to play an important role in PCa (11). Hence, the correlation between these vinblastine resistance-related hub genes and immune cells in PCa was analyzed using TIMER. However, EBLN2 could not be identified using TIMER, thereby excluding it from the study. Apart from LRRFIP1 and TROAP, all other gene mutation types were associated with immune cell infiltrations in PCa ( Figure S4 ).

Next, the specific correlation between these gene expressions and immune cells in PCa was evaluated. LRRFIP1 was not observed to be associated with CD4+T cells in PCa ( Figure 7B ). Moreover, AURKA was not correlated with Purity cell, and CD8+T cell in PCa ( Figure 7E ). Furthermore, CCDC150 was not correlated with CD4+T cell, CD8+T cell and Macrophage cell in PCa ( Figure 7F ). Additionally, TROAP was not correlated with CD4+T cells and Macrophage cells in PCa ( Figure 7G ). However, other genes showed a correlation with all immune cells in PCa ( Figure 7 ).

Finally, the function of these hub genes was verified using clinical samples and PCa cells. Depending on that the above results, CCNB1 and AURKA were upregulated in vinblastine-resistant PCa cells, PCa samples and CRPC samples. Additionally, they also affected the patient’s survival status, and even PCa progression. hence, they were selected as the study objectives and thought they may critical for CRPC resistant to vinblastine. C4-2 PCa cells, which were cultured from LNCaP cells and grown in an environment without androgen, were considered as a cell line model of CRPC and another PCa cell line: 22Rv1 which can also growth without androgen were used in the study (12). Further, these two types of cells are sensitive to vinblastine.

The protein levels of CCNB1 and AURKA were analyzed in clinical normal prostate tissues, primary PCa samples and CRPC samples. These genes were found to be upregulated in tumour tissues and highly expressed in CRPC samples ( Figures 8A, B ), indicating their role in the occurrence of both PCa and CRPC. Next, we constructed CCNB1 and AURKA knockdown CRPC cell lines. The knockdown lentivirus decreased the expression of CCNB1 and AURKA at both mRNA and protein levels in C4-2 cells ( Figures 8C–E ) and can also inhibit the expression of CCNB1 and AURKA protein in 22Rv1 cells ( Figure S5A ). Then, the C4-2 and 22Rv1 cell lines were treated with vinblastine for 24 h to determine their proliferation ability. We observed that vinblastine decreased cell proliferation of C4-2 ( Figure 8F ). Further same results were observed in 22Rv1 cells ( Figure S5B ). Next, we transfected the specific knockdown lentivirus into C4-2 and 22Rv1 cells to determine the influence of the two genes on CRPC cells proliferation ability with vinblastine treatment. CCNB1 and AURKA knockdown decreased the proliferation ability of the C4-2 cells ( Figures 8G, H ) and 22Rv1 cells ( Figure S5C, D ). Thus, CCNB1 and AURKA have a potential role in the occurrence of PCa and CRPC and can influence CRPC resistant to vinblastine.