Ethical statement
All animal experiments are in accordance with the "Regulations on the Administration of Laboratory Animals" (The National Science and Technology Commission of the People’s Republic of China, March 1, 2017, revised edition) and the National Institutes of Health Laboratory Animal Care and Use Guidelines (ISBN: 13:978-0- 309-15400-0, revised in 2011) to ensure the animal welfare of experimental animals. This study was approved by the Human Research Ethics Committees of Southwest Hospital, Army Medical University (AMU).
OC datasets selection from the GEO database
We selected and downloaded the raw data of four OC datasets from the GEO database. GSE40595 [25] contained 63 high grade serous ovarian cancer samples and 14 normal ovarian samples. GSE18520 [26] contained 53 advanced stages, high-grade primary tumor samples, and 10 normal ovarian samples. GSE38666 [27] contained 25 serous ovarian cancer samples and 20 para-cancer samples. GSE36668 [28] contained 4 serous ovarian borderline tumor samples, 4 well-differentiated serous ovarian carcinomas, and 4 normal ovarian samples. The four datasets were all based on the platform GPL570 (Affymetrix Human Genome U133 Plus 2.0 Array) to reduce the variability from the different experimental setup. Quality analysis was performed on raw data of the selected GEO datasets respectively by using the affyPLM package [29] in R software. Three tumor samples in GSE40595, one tumor sample in GSE18520, and one tumor sample in GSE38666 datasets were removed from the data processing because of the variance of sample quality. Thus, there were 155 tumor samples and 57 non-tumor samples included in our subsequent analysis in total.
Identificaction of proliferation-associated genes from the DEGs
To screen the DEGs in each GEO dataset, the limma package [30] was used with cutoff criteria of |log2 Fold Change (FC)|>1.5 and adjusted p-value<0.05. The heatmaps were drawn by the tools in the Omicshare platform (https://www.omicshare.com/). For visualization, an online Venn diagram tool (http://bioinformatics.psb.ugent.be/webtools/Venn/) was used to show the overlapped part of DEGs in the four GEO datasets. The GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis was conducted by DAVID (Resource 6.8) [31]. The top 10 GO categories with the smallest p-value and the genes in five proliferation-associated categories (“cell division”, “mitotic nuclear division”, “cell proliferation”, “mitotic sister chromatid segregation”, “chromo-some segregation”) were visualized by GOplot package in R [32]. The results of pathway enrichment analysis were visualized by using Omicshare tools.
Identification of candidate genes with early prognostic value
The overall survival analysis of the proliferation-associated genes was conducted by the Kal-plan-miere Plotter in all OC patients in the database. To obtain genes with more significant prognostic value, p<0.01 was set as the screening criteria. Survival analysis was also performed on the previously obtained prognostic genes only in Stage I-II patients with a cutoff of p<0.05. The different expression among stages of the selected genes with early-stage prognostic value was analyzed in the Gene Expression Profiling Interactive Analysis (GEPIA) database [33]. KIF15 was selected to be bioinformatically analyzed and functionally verified in a subsequent study.
Bioinformatical verification on the expression level of KIF15 in normal and OC tissues
The RNA-seq data of the OC samples in TCGA and normal ovarian samples in GTEx were downloaded from the UCSC Xena project (https://xena.ucsc.edu/). The OC samples (N=379 for cystic, mucinous, and serous neoplasms) was limited to RNA-seq data of FPKM with HTseq and the GTEx samples (N=88 for normal ovarian tissue) was limited to RNAseq data of FPKM. The downloaded RNA-Seq data of both datasets have been recomputed to minimize differences from distinct sources based on a standard pipeline. The corresponding clinical information of the OC dataset was downloaded from the TCGA database (https://portal.gdc.cancer.gov/). The data of KIF15 expression in multiple normal tissue samples of females was extracted and visualized by the ggpubr package in R. The differential expression of KIF15 was visualized by the beeswarm package in R. The analysis of KIF15 differential expression in five kinds of female-specific malignancies was conducted by the GEPIA online tool. The RNA-seq data of KIF15 expression in ovarian cancer cell lines were downloaded from the Cancer Cell Line Encyclopedia (CCLE,https://portals.broadinstitute.org/ccle) [34].
Immunohistochemistry on OC Tissue Microarray
The TMA (Alenabio, Xi’an, China) used in the study contained a total of 100 samples, including 80 ovarian cancer tissue samples of different histological types, 10 lymph node metastasis samples, and 10 non-tumor ovarian samples. The anti-KIF15 rabbit polyclonal antibodies (Sigma-Aldrich Cat# HPA035517) was used to conduct the immunohistochemical staining by a dilution rate of 1:100. The positive staining was quantified and classified into 5 levels: negative staining for 0 score; 1%-25% positive staining cells for 1 score; 26%-50% positive for 2 scores; 51%-75% positive for 3 scores and 76%-100% positive for 4 scores. Staining intensity was scored as negative (0), weak (1), moderate (2), and robust (3). All the pathologic sections were independently reviewed by two pathologists and the expression levels were graded by the product of positive staining percentage score and staining intensity score.
Cell Culture
The cell lines used in the study, including ovarian cancer cell lines SKOV3, OVCAR-3, A2780, and HO8910, cervical cancer cell lines Hela, Siha and C33A, lung adenocarcinoma cell line A549, pancreatic cancer cell line PANC-1 and glioblastoma cell line U87, were all purchased from the cell bank of Chinese Academy of Science (Shanghai, China). The cell lines were cultured according to the instructions online (http://www.cellbank.org.cn/).
Lentivirus transfection
Human KIF15 knocking down (KD) lentiviruses and negative control (NC) lentiviruses were constructed by Genechem (Shanghai, China). SKOV3 and HO8910 cells were seeded in 6-well plates the day before transduction to ensure the cells would grow to 30% to 40% confluence the next day, and then infected with lentivirus for 24 h at a Multiplicity Of Infection (MOI) of 20 and 10 respectively in the presence of polybrene (5 mg/mL, Genechem).
qRT-PCR analysis
The qRT-PCR was performed as previously reported [35]. The 2‒ΔΔct method was used to determine the expression of the KIF15 gene. All experiments were carried out in triplicate. The primers were purchased from Sangon Biotech (Shanghai, China).
KIF15:
Forward 5 ′‐ CTCTCACAGTTGAATGTCCTTG ‐ 3′
Reverse 5 ′‐ CTCCTTGTCAGCAGAATGAAG ‐ 3′
GAPDH:
Forward 5 ′‐ TGACTTCAACAGCGACACCCA ‐ 3 ′
Reverse 5 ′‐ CACCCTGTTGCTGTAGCCAAA ‐ 3 ′
Western Blot analysis
Western Blot was performed as previously described [35]. GAPDH was used as a loading control. The KIF15 rabbit polyclonal antibody (1:100) and GAPDH monoclonal antibody (1:1000, Santa Cruz Biotechnology Cat# sc-32233) was used.
Cell growth analysis by Celigo method
SKOV3 and HO8910 cells were transfected with KIF15-KD or NC lentivirus. The transfected cells were collected and then seeded into 48-well plates 2000 cells per well respectively. The number of cells with green fluorescence in each well was measured by a Cellomics ArrayScan System (Nexcelom, USA) once a day. The variable data of the green fluorescence signal were obtained for statistical analysis to construct 5-day cell proliferation curves. The green-fluorescence cells were also scanned to be counted by an image analysis software. The count of green-fluorescence cells at each time point was compared with that of day 1 to calculate the cell proliferation ratio for each time point and each experimental group. The fold change of cell proliferation was obtained to construct cell growth curves.
The cell proliferation ratio was computed as follows: fold change (NC vs experimental group)=proliferation ratio on day 5 for the NC group/proliferation ratio on day 5 for the experimental group. A fold change of proliferation ratio equal to or greater than 2 indicated that cell proliferation had been significantly slowed down.
Cell Counting Kit-8 (CCK8) assay
SKOV3 and HO8910 cells were plated into 96-well plates at some 2000 cells per well and transfected with the KIF15-KD or NC lentivirus. Cell proliferation was measured by using CCK8 Reagent (DOJINDO, JAPAN) respectively on days 1, 2, 3, 4, and 5 after transfection. The assays were performed in triplicate.
FACS assay by flow cytometry
Cells were seeded into 6-well plates and cultured in serum-free medium at 37°C for 24 h. Cells were transfected with the KIF15-KD or NC lentiviruses. Then the cells were harvested and analyzed by an AnnexinV-APC apoptosis kit (eBioscience, USA). Cell apoptosis was determined using the Guava InCyte software (Millipore, USA). All experiments were conducted in triplicate.
Caspase3/7 activity assay
To assess the activity of caspases 3 and 7, the Caspase-Glo 3/7 Assay (Promega, Germany) was conducted following the manufacturer’s instructions. The Caspase-Glo 3/7 Assay is based on the cleavage of the DEVD sequence of a luminogenic substrate by the caspases 3 and 7 and results in a luminescent signal. The fluorescence signal was measured at an excitation wavelength of 485 nm and an emission wavelength of 527 nm.
Subcutaneous transplantation of human OC cells in Balb/c nude mice
Female Balb/c nude mice of four weeks old were used in this experiment. A total of 2×107 transfected SKOV3 cells were subcutaneously injected into the right armpit of each mouse. The body weight and tumor diameter of each mouse were measured every week after cell transplantation. All mice were sacrificed on the 41st day after the initiation of cell injection. Before the mice were killed, the fluorescence images of xenograft tumors were photographed under a whole-body fluorescent imaging system (Lumina LT, Perkin Elmer, USA). Tumors were observed by both macroscopical and microscopical methods.
mRNA expression profiling
The SKOV3 cells of KIF15-KD and NC group were collected for mRNA expression profiling. Total RNA was isolated from cell samples by using an Agilent RNA 6000 Nano Kit (Agilent, USA), and the quality of total RNA was analyzed. Both the KIF15-KD and NC cell samples had three replicates. The mRNA expression profiling was conducted by using GeneChip prime view humans (901838, Affymetrix, USA). RNA labeling and hybridization were performed with a GeneChip Hybridization Wash and Stain Kit (Agilent, USA). The raw data obtained from mRNA expression profiling was quality-analyzed using R software as the aforementioned methods in Paragraph 2.2 before subsequent bioinformatic analysis.
Phospho-antibody arrays
To avoid batch difference, the cell samples applied in mRNA expression profiling were examined in this assay. The cell lysates of KIF15-KD and NC group were obtained and applied to a Cancer Signaling Phospho-Antibody Array ((PCS300, Full Moon Biosystems, USA). The phosphoantibody array detection was carried out in cooperation with Wayne Biotechnology (Shanghai, China) per the manufacturer’s protocols. The array contained 157 site-specific and phospho-specific antibodies and 147 non-phospho antibodies, each of which had 6 replicates. The slides were scanned by a GenePix 4000 scanner and the images were analyzed with GenePix Pro 6.0 (Molecular Devices, Sunnyvale, CA). The intensity of the fluorescence signal obtained from each antibody-stained region indicated the expression level of a certain protein. The extent of protein phosphorylation was measured by a ratio computation. The phosphorylation ratio was calculated as follows: phosphorylation ratio=phospho value/non-phosphor value [36]. The total proteome ratios were standardized by β-actin.
Pathway analysis by multiple bioinformatics methods
The DEGs in mRNA expression profiling were obtained by using R packages with a cutoff of |Fold Change|≥1.5 and p<0.05. The pathway analysis was performed with the plug-in ClueGO [37] in Cytoscape software (Version 3.7.1) with a cutoff of p<0.05. GSEA (Version 3.0), a pathway enrichment method was also used to analyze a level of gene sets. GSEA software uses the predefined gene sets from the Molecular Signatures Database (MSigDB v6.2) [38]. A gene set is a group of genes that share similar pathways, functions, chromosomal localization, or other features. In this study, we used all the C collection sets for GSEA analysis (i.e., H, C1-C7 collection in MsigDB). The list of ranked genes based on a score calculated as -log10 of p-value multiplied by the sign of fold change. The minimum and maximum criteria for the selection of gene sets from the collection were 10 and 500 genes, respectively. Pathway enrichment analysis on the results of the phos-pho-antibody arrays was also performed. According to the pathways obtained by using DAVID and GSEA, the phosphorylated proteins on the pathways were selected and the protein-protein interactive (PPI) networks were visualized by Cytoscape. The consistent genes and phosphorylated proteins on the previously obtained pathways both in mRNA expression microarrays and phospho-antibody arrays were visualized by Venn tools in the Omicshare platform.
Statistical analysis
SPSS 20.0 (IBM SPSS, Chicago, IL) software was used for statistical analyses. Values are presented as the mean±SD. Wilcox test was used to determine the significant expression difference of DEGs among ovarian cancer and non-tumor samples in GEO and TCGA datasets. The differences between NC and KD groups in proliferation and apoptosis assays were tested by the Student’s t-test. The different expression levels between the ovarian cancer tissue samples and adjacent non-tumor samples in TMA were tested with the Mann-Whitney test. p<0.05 was considered statistically significant.