Screening and verifying key genes with poor prognosis in colon cancer through bioinformatics analysis.

BACKGROUND: Colon cancer (CC) is one of the tumors with high morbidity and mortality in the world, and has a trend of younger generation. The molecular level of CC has not been fully elaborated. The purpose of this study is to screen and identify important genes with poor prognosis and their mechanisms at different levels.
METHODS: GSE74602 and GSE10972 gene expression profiles were downloaded from the Gene Expression Omnibus (GEO) database. There were 58 normal tissues and 58 CC tissues. Differentially expressed genes (DEGs) were screened out by using the GEO2R tool and Venn diagram. Then, the DAVID online database was used to perform the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Six hub genes with the highest correlation were screened out after the modular analysis of the protein-protein interaction (PPI) network by using Cytoscape's MCODE plug-in. Finally, the overall survival of key hub genes and potential pathways were verified in GEPIA and UALCAN database.
RESULTS: A total of 78 up-regulated DEGs were enriched in the mitotic nuclear division, cell division, cell proliferation, anaphase-promoting complex-dependent catabolic process and G2/M transition of the mitotic cell cycle. In total, 130 down-regulated DEGs were enriched in muscle contraction, bicarbonate transport, cellular response to zinc ion, negative regulation of growth, negative regulation of leukocyte apoptotic process and one-carbon metabolic process. CDK1, CCNB1, CDC20, AURKA, CCNA2 and TOP2A were the top six hub genes, mainly enriched in cell cycle pathways. Among them, CCNB1, CDK1, CDC20, CCNA2 were enriched in the G2/M phase. GEPIA and UALCAN database confirmed that CCNA2 and CCNB1 had a significant relationship with the poor prognosis of CC patients. Meanwhile, there was a positive correlation between the two.
CONCLUSIONS: Screening out genes with abnormal expression in CC help understand the initiation and progression of CC at the molecular level and explore candidate biomarkers for diagnosis, treatment and prognosis. 2020 Translational Cancer Research. All rights reserved.

Introduction

Colorectal cancer ranks third in incidence and second in mortality across the world (1). There are expected to be 147,950 newly diagnosed colorectal cancer patients in the United States by 2020, including 104,610 cases of colon cancer (CC) and 43,340 cases of rectal cancer. The median age of diagnosis for CC is 69 years, although CC has a trend towards younger age (2). Dietary patterns, body mass index, lifestyle factors will affect the morbidity. The adoption of best practices in cancer treatment and management, longer standing screening, and early detection programs have given rise to the colorectal survival rates in some developed countries. Current screening methods mainly include fecal occult blood test (FOBT), flexible sigmoidoscopy, and colonoscopy. The fecal occult blood test has low specificity. A common colonoscopy is invasive with poor comfort while painless colonoscopy costs more. The preliminary preparation process is more complex. Despite the rapid spread of screening programs since 2000, the proportion of cases diagnosed early in CC has only slightly increased, from 34% in the mid-1990s to 36% between 2012 and 2016 (3). Many researchers, therefore, expect to find CC-specific biomarkers for more convenient and accurate screening. These tumor markers can also be used for cancer risk stratification, treatment protocol formulation, prognostic analysis, and recurrence monitoring.

Understanding the regulation of gene expression is a good way to figure out the relationship between genotype and phenotype (4). Therefore, many scholars applied microarray technology and bioinformatics analysis to medical oncology, hoping to discover the differences between genetic and epigenetic phenotypes caused by tumors and to evaluate markers for disease diagnosis and treatment. However, analyzing independent microarrays from individual surveys often increase the false-positive rate, making it difficult to identify reliable key genes and signaling pathways. Based on this, we expected to screen reliable differentially expressed genes (DEGs) by overlapping related datasets. We present the following article in accordance with the MDAR checklist (available at http://dx.doi.org/10.21037/tcr-20-2309).

Methods

In this study, we downloaded GSE74602 and GSE10972 gene expression profiles from Gene Expression Omnibus (GEO) (Affymetrix GPL6104 platform, Illumina humanRef-8 v2.0 expression beadchip). There were 30 normal tissues and 30 CC tissues in GSE74602. There were 28 normal tissues and 28 CC tissues in GSE10972. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). Approval from the Institutional Review Board (IRB) was not needed since our data were extracted from the GEO database.

Identification of DEGs

The web tool GEO2R (Http://www.ncbi.nlm.nih.gov/geo/geo2r) was used to screen for DEGs in CC and non-cancer tissue samples. After the standardization of the micro matrix results, DEGs were identified with logFC (fold change) >1 and adj. P value <0.01. A total of 78 up-regulated and 130 down-regulated genes were detected on the ‘Draw Venn Diagram’ website.

Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses of DEGs

GO and KEGG pathway enrichment analyses were performed by DAVID online database, followed by the construction of the protein-protein interaction (PPI) network. Through the module analysis of the whole network by using the MCODE plug-in of Cytoscape, six hub genes with the highest degree of correlation were screened out.

Survival analysis to screen the candidate genes

Key hub genes and potential pathways were validated in the GEPIA and UALCAN database. This method was helpful to screen out genes with abnormal expression in CC, further understand the initiation and progression of CC at the molecular level, and explore candidate biomarkers for diagnosis, treatment and prognosis.

Statistical analysis

The web tool GEO2R is based on R software version 3.2.3 and the “limma”, “GEOquery” and “Biobase” packages. The GEPIA and UALCAN use the log-rank test, sometimes called the Mantel-Cox test, for the hypothesis evaluation. The GEPIA and UALCAN are based on R (version 3.3.2) and Perl (version 5.22.1) programs (5,6).

Results

Identification of DEGs in CCs

There were 30 normal tissues and 30 CC tissues in GSE74602. There were 28 normal tissues and 28 CC tissues in GSE10972. After standardization of the microarray results, DEGs (1,626 in GSE10972 and 249 in GSE74602) were identified. We used the ‘Draw Venn Diagram’ website to identify the common DEGs in GSE10972 and GSE74602. A total of 78 commonly up-regulated genes (log FC >0) and 130 commonly down-regulated genes were detected ( & ).

Figure 1

Venn diagram, protein-protein interaction (PPI) network and the most significant module of DEGs. 78 DEGs were up-regulated in the two datasets (logFC >0) among the mRNA expression profiling sets GSE10972, GSE74602 (A). 130 DEGs were down-regulated in the two datasets (logFC <0) among the mRNA expression profiling sets GSE10972, GSE74602 (B). A total of 208 DEGs were included in PPI network analysis (C). The most significant module was obtained from PPI network with 34 nodes and 536 edges (D). Upregulated genes are marked in light red; downregulated genes are marked in light blue. The PPI network of DEGs was constructed using Cytoscape. DEG, differentially expressed gene.

Table 1

All 208 commonly differentially expressed genes (DEGs) were detected from two profile datasets, including 130 downregulated genes and 78 up-regulated genes in the CRC tissues compared to normal CRC tissues

DEGs	Genes name
Upregulated	CDH3, SLCO4A1, IQGAP3, TPX2, MMP7, TRIB3, SERPINB5, CCNB1, FOXQ1, MMP3, ASPM, KRT80, LY6G6D, CDCA3, TCN1, ANLN, SLC7A5, CXCL2, CCNB2, PRC1, CDK1, PSAT1, PUS7, NFE2L3, CEP55, CEMIP, TOP2A, CLDN2, VSNL1, PHLDA1, CXCL1, SAPCD2, TMEM97, MCM4, CXCL8, AURKA, SLC12A2, REG1A, MAD2L1, DLGAP5, DPEP1, SOX9, HOXB8, PAICS, HMMR, REG1B, ASCL2, SCD, ETV4, PARPBP, CLDN1, STC2, CDCA7, COL11A1, CDC20, CENPN, CTHRC1UHRF1, ITGA2, KIAA0101, TESC, MYC, CCNA2, TTK, CDKN3, NME1, BUB1, NCAPG, TRIP13, MMP1, PTTG1, CDCA5, GINS2, UBE2T, CENPF, NUSAP1, CKS2, ESM1
Downregulated	PNCK, SPEG, LGALS2, HSPB6, FGL2, OTOP2, DPT, HSPB8, CCL23, MFAP4, ZG16, GUCA2B, WSCD1, CHP2, LRRN2, SCARA5, CLCA4, TCEAL2, ATP1A2, MYOM1, FAM129A, AKR1B10, MT1E, TUBAL3, PRKCB, CA4, CNN1, SEPP1, SDPR, CTSG, TRPM6, NXPE4, ADAMDEC1, ACKR1PY, CCL19, MYLK, HAND1, SLCO2A1, TMIGD1, DNASE1L3, CRYAB, SCIN, CKB, SLC26A3, CD36, CXCL12, PDE5A, MRGPRF, VIP, CA2, FLNC, CHGA, CLEC3B, MT1M, PKIB, ANPEP, ADH1A, CEACAM7, METTL7A, FHL1, ADTRP, PRPH, ST6GALNAC6, MT1G, CCL21, ABCA8, PDE9A, LMO3, TNFRSF13B, ACTG2, MT1F, KRT24, SLC16A9, BEST2, MYL9, INSL5, SLC51B, AQP8, ZBTB16, KCNMB1, DES, FABP4, TNFRSF17, NPTX1, SCGN, CLDN8, MS4A12, FAM107A, PLAC9, MYH11, PGM5, GPX3, SLC25A34, VSIG2, RNF150, NPY, ABI3BP, CALD1, GCNT2, SCNN1B, CASQ2, C2orf88, EPB41L3, CDKN2B, LAMA1, MAMDC2, RBPMS2, BEST4, CA1, SYNM, C7, TEX11, SLC4A4, C2orf40, SFRP1, PDK4, GFRA2, CFD, TMEM37, GPAT3, CPED1, RERGL, CHAIN, SLC51A, GUCA2A, SEMA6A, DHRS9, CA7, AFF3

KEGG and GO enrichment analyses of DEGs

We used DAVID software to analyze the biological classification of DEGs. GO-based biological process (BP) analysis indicated that the up-regulated DEGs were particularly enriched in the mitotic nuclear division, cell division, cell proliferation, anaphase-promoting complex-dependent catabolic process and G2/M transition of the mitotic cell cycle, while the down-regulated DEGs were enriched in muscle contraction, bicarbonate transport, cellular response to zinc ion, negative regulation of growth, negative regulation of leukocyte apoptotic process, and one-carbon metabolic process (). For cell component (CC), the up-regulated DEGs were mainly enriched in the spindle, spindle pole, midbody, nucleus, centrosome, nucleoplasm, and the down-regulated DEGs were significantly enriched in extracellular exosome, extracellular space, Z disc, sarcolemma, extracellular region, extracellular matrix (). In addition, molecular function (MF) analysis displayed that the up-regulated DEGs were mainly involved in protein kinase binding, chromatin binding, protein binding and cyclin-dependent protein serine/threonine kinase activity, and the down-regulated DEGs were particularly involved in carbonate dehydratase activity, chemokine receptor binding, transporter activity, structural constituent of muscle, chemokine activity and chloride channel activity (). Furthermore, KEGG analysis results revealed that the up-regulated DEGs were particularly enriched in the cell cycle, oocyte meiosis and progesterone-mediated oocyte maturation, while the down-regulated DEGs were mainly enriched in mineral absorption, bile secretion, nitrogen metabolism, proximal tubule bicarbonate reclamation and pancreatic secretion.

Table 2

Gene ontology analysis of differentially expressed genes in ovarian cancer

Expression	Category	Term	Count	P value	FDR
Up-regulated	GOTERM_BP_DIRECT	GO:0007067~mitotic nuclear division	15	4.10E−12	6.13E−09
	GOTERM_BP_DIRECT	GO:0051301~cell division	16	3.32E−11	4.96E−08
	GOTERM_BP_DIRECT	GO:0008283~cell proliferation	11	5.04E−06	0.007534
	GOTERM_BP_DIRECT	GO:0031145~anaphase-promoting complex-dependent catabolic process	6	2.71E−05	0.040467
	GOTERM_BP_DIRECT	GO:0000086~G2/M transition of mitotic cell cycle	7	3.35E−05	0.050149
	GOTERM_CC_DIRECT	GO:0005819~spindle	7	1.22E−05	0.014011
	GOTERM_CC_DIRECT	GO:0000922~spindle pole	6	9.79E−05	0.112351
	GOTERM_CC_DIRECT	GO:0030496~midbody	6	2.16E−04	0.247731
	GOTERM_CC_DIRECT	GO:0005634~nucleus	39	2.24E−04	0.256455
	GOTERM_CC_DIRECT	GO:0005813~centrosome	9	4.29E−04	0.491477
	GOTERM_CC_DIRECT	GO:0005654~nucleoplasm	24	8.62E−04	0.985099
	GOTERM_MF_DIRECT	GO:0019901~protein kinase binding	9	2.35E−04	0.285395
	GOTERM_MF_DIRECT	GO:0003682~chromatin binding	8	0.001607924	1.936332
	GOTERM_MF_DIRECT	GO:0005515~protein binding	51	0.004818525	5.700127
	GOTERM_MF_DIRECT	GO:0004693~cyclin-dependent protein serine/threonine kinase activity	3	0.009713212	11.1837
	KEGG_PATHWAY	hsa04110: cell cycle	11	5.80E−10	6.23E−07
	KEGG_PATHWAY	hsa04114: oocyte meiosis	8	1.63E−06	0.001746
	KEGG_PATHWAY	hsa04914: progesterone-mediated oocyte maturation	6	9.13E−05	0.09801
Down-regulated	GOTERM_BP_DIRECT	GO:0006936~muscle contraction	8	6.90E−06	0.010446
	GOTERM_BP_DIRECT	GO:0015701~bicarbonate transport	6	1.02E−05	0.015433
	GOTERM_BP_DIRECT	GO:0071294~cellular response to zinc ion	4	2.52E−04	0.381119
	GOTERM_BP_DIRECT	GO:0045926~negative regulation of growth	4	2.52E−04	0.381119
	GOTERM_BP_DIRECT	GO:2000107~negative regulation of leukocyte apoptotic process	3	6.38E−04	0.962218
	GOTERM_BP_DIRECT	GO:0006730~one-carbon metabolic process	4	0.001002	1.506587
	GOTERM_CC_DIRECT	GO:0070062~extracellular exosome	36	3.19E−04	0.385946
	GOTERM_CC_DIRECT	GO:0005615~extracellular space	22	3.98E−04	0.480914
	GOTERM_CC_DIRECT	GO:0030018~Z disc	6	0.001418	1.704013
	GOTERM_CC_DIRECT	GO:0042383~sarcolemma	5	0.002947	3.51202
	GOTERM_CC_DIRECT	GO:0005576~extracellular region	22	0.0037	4.390367
	GOTERM_CC_DIRECT	GO:0031012~extracellular matrix	8	0.004735	5.587538
	GOTERM_MF_DIRECT	GO:0004089~carbonate dehydratase activity	4	1.06E−04	0.133058
	GOTERM_MF_DIRECT	GO:0042379~chemokine receptor binding	3	0.001254	1.561892
	GOTERM_MF_DIRECT	GO:0005215~transporter activity	7	0.002604	3.218176
	GOTERM_MF_DIRECT	GO:0008307~structural constituent of muscle	4	0.002914	3.594352
	GOTERM_MF_DIRECT	GO:0008009~chemokine activity	4	0.004518	5.52181
	GOTERM_MF_DIRECT	GO:0005254~chloride channel activity	4	0.005935	7.194686
	KEGG_PATHWAY	hsa04978: Mineral absorption	7	2.26E−06	0.002651
	KEGG_PATHWAY	hsa04976: Bile secretion	6	3.64E−04	0.425875
	KEGG_PATHWAY	hsa00910: Nitrogen metabolism	4	4.33E−04	0.506185
	KEGG_PATHWAY	hsa04964: Proximal tubule bicarbonate reclamation	4	0.001086	1.264063
	KEGG_PATHWAY	hsa04972: Pancreatic secretion	6	0.001429	1.659887

GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; DEGs, differentially expressed genes; FDR, false discovery rate.

PPI network and modular analysis

A total of 208 DEGs were included in PPI network analysis. There were 171 nodes and 875 edges mapped in the PPI network for these common DEGs based on STRING database analysis (). The most significant module was obtained by using Cytoscape software. It had 34 nodes and 536 edges (). Genes with degree ≥10 in the module were identified as hub genes. These hub genes were all up-regulated. To further study the potential pathway of DEGs in the module, GO and KEGG enrichment analyses were performed again. The results showed that DEGs in module were mainly enriched in the cell cycle, oocyte meiosis, progesterone-mediated oocyte maturation, p53 signaling pathway (). We choose the top six genes, ranked by degree, including CDK1, CCNB1, CDC20, AURKA, CCNA2, TOP2A. CDK1 had the highest degree with 40, while AURKA, CCNA2 and TOP2A had the lowest degree with 37 (). Some EDGs (such as CCNB1, CDK1, CDC20, CCNA2, ) were mainly enriched in cell cycle pathway, especially in the G2/M phase ().

Table 3

Re-analysis of 34 selected genes via KEGG pathway enrichment

Category	Term	Count	P value	Genes
hsa04110	Cell cycle	10	2.78E−13	CCNB1*, CDK1*, MAD2L1, CCNB2, BUB1, TTK, CDC20*, PTTG1, CCNA2*, MCM4
hsa04114	Oocyte meiosis	8	7.36E−10	CCNB1*, CDK1*, MAD2L1, CCNB2, BUB1, AURKA*, CDC20*, PTTG1
hsa04914	Progesterone-mediated oocyte maturation	6	5.28E−07	CCNB1*, CDK1*, MAD2L1, CCNB2, BUB1, CCNA2*
hsa04115	p53 signaling pathway	3	0.007886	CCNB1*, CDK1*, CCNB2

*, genes were selected hub genes. KEGG, Kyoto Encyclopedia of Genes and Genomes.

Table 4

Top six hub genes with higher degree of connectivity

Gene	Degree of connectivity	Regulated
CDK1	40	Up
CCNB1	38	Up
CDC20	38	Up
AURKA	37	Up
CCNA2	37	Up
TOP2A	37	Up

Figure 2

Re-analysis of 6 selected genes by KEGG pathway enrichment. CCNB1, CDK1, CDC20, CCNA2 were enriched in the G2/M phase. KEGG, Kyoto Encyclopedia of Genes and Genomes.

Survival analysis of selected hub genes

GEPIA (http://gepia.cancer-pku.cn/) was utilized to analyze the survival data of the six hub genes (). We found that high expression of AURKA, CCNA2 and CCNB1 was significantly associated with worse survival (P<0.05), but CDC20, CDK1 and TOP2A had no significance (P>0.05). And then, UALCAN (http://ualcan.path.uab.edu/) was used to identify six hub genes survival data again (). The results showed that overexpression of CCNA2, CDC20 and CCNB1 would lead to poor prognosis (P<0.05), while AURKA, CDK1 and TOP2A had no significant correlation with prognosis (P>0.05). Summing up, the data from the two databases (GEPIA and UALCAN) both confirmed that the high expression of CCNA2 and CCNB1 had a significant correlation with poor prognosis, and the expression of these two genes was positively correlated ().

Figure 3

Significantly expressed 6 genes in CC patients compared to healthy people. GEPIA (http://gepia.cancer-pku.cn/) was utilized to identify 6 hub genes survival data (A-F). UALCAN (http://ualcan.path.uab.edu/) was used to identify 6 hub genes survival data again (G-L). CC, colon cancer.

Figure 4

The correlation between CCNA2 and CCNB1. CCNA2 and CCNB1 had a positive correlation.

Discussion

Despite the significant advances in diagnosis and treatment in the last decade, the prevalence of CC is still rising and five-year survival is still poor (7). Getting more information about the molecular mechanisms of CC is essential for the diagnosis and treatment of CC. DNA microarray technology has improved in sensitivity and selectivity in recent decades and has become a research tool widely used in molecular biology, genome research, cancer diagnosis and other fields (8).

In the present study, we obtained 78 commonly up-regulated genes and 130 commonly down-regulated genes by integrating two expression profiling datasets (GSE10972 and GSE74602) from the GEO dataset. To further understand the role of DEGs play in cancer, we performed the GO and KEGG pathways. The GO term analysis showed that up-regulated DEGs were mainly involved in the regulation of cell division and cell proliferation, while down-regulated DEGs mainly affected the regulation of inflammatory response, cell apoptosis, bicarbonate metabolism and muscle function. The regulation of cell proliferation is crucial to the growth, development and regeneration of eukaryotes, and is considered to be an important cause of cancer (9). Bicarbonate transporters may alter the rate of cancer cell proliferation by controlling the PH of the cytoplasm and extracellular space (10). Festi et al. found that cancer metastasis induced TGF-β release from bone, which in turn caused myasthenia by reducing ca2+ (11). The enriched KEGG pathways of up-regulated DEGs included cell cycle, oocyte meiosis and progesterone-mediated oocyte maturation. The down-regulated DEGs were significantly enriched in mineral absorption, bile and pancreatic secretion, carbon and nitrogen metabolism. Glutamine has multiple roles in cancer as it contributes its carbon backbone to the tricarboxylic acid (TCA) cycle and is an obligatory nitrogen source for purines and pyrimidines (12).

By analyzing the PPI network constructed with DEGs, the hub genes with the highest correlation were screened out, including CDK1, CCNB1, CDC20, AURKA, CCNA2 and TOP2A. Cyclin-dependent kinase 1 (CDK1) is a member of the serine/threonine-protein kinase family that is a vital determinant of mitotic progression. It affects the cell cycle from G2 to M by phosphorylating various substrates (13). CDK1 is overexpressed in a variety of cancers (14,15) and its accumulation in tumors of different sites is positively correlated with the degree of malignancy (16-23). Apart from the function in the cell cycle, CDK1 appears to be involved in protein translation, cell morphogenesis, protein secretion, and nuclear transport (24). CCNB1 works by switching cells from G2 to M but becomes dysfunctional in cancer cells (25). CCNB1 binds to cyclin-dependent kinase 1 (CDK1) to form a complex that phosphorylates the target substrate, and then alters cell cycle progression (26). FOXM1 promotes the development of human liver cancer cells by transcriptional activation of CCNB1, and the high level of CCNB1 is an independent indicator of poor prognosis in hepatocellular carcinoma patients (27). This occurs not only in HCC, but also in breast cancer, head and neck squamous cell carcinoma, and esophageal carcinoma (28). The expression of the CCNB1 gene in invasive pituitary adenoma was significantly higher than that in non-invasive pituitary adenoma, indicating that the CCNB1 gene influences the invasiveness of pituitary adenoma (29). Cell division cycle 20 (CDC20) is a key molecule in the cell cycle as it plays an important role in mitotic exit (30,31). Overexpression of CDC20 may lead to aneuploidy in the tumor cells (32). Abnormal levels of CDC20 expression was observed in varied human cancer, including cutaneous squamous cell carcinoma (33), oral squamous cell carcinoma (34), glioblastoma (35), and gastric cancer, bladder cancer (36), and cervical cancer (37). The overexpression of CDC20 may also indicate a poor prognosis in non-small cell lung cancer (38), pancreatic cancer (39), estrogen receptor-positive breast cancer (40), and hepatocellular carcinoma (41). AURKA encodes a serine-threonine kinase which controls G2/M conversion (42), centrosome maturation (43), and mitotic spindle formation (44). Abundant evidence suggests that inhibition of AURKA expression can effectively suppress cell proliferation, invasion and metastasis in some tumors (45-47). Wen et al. confirmed that the loss of AURKA was sufficient to restrict myelofibrosis and other Primary myelofibrosis phenotypes in vivo (48). Moreover, AURKA is considered to be a valid therapeutic target for neuroblastoma (49) and has been verified in phase I clinical trials (50). CCNA2 accumulates during the G1 phase and functions as a crucial regulatory factor of the G1/S and G2/M transition by activating CDK2 and CDK1 kinases, respectively (51,52). Fischer et al. suggested that p53 down-regulates CCNA2 and other genes through the p53-p21-dreamcde/ChR pathway to block G2/M cell cycle (53). Loukil et al. believed that the dysregulation of CCNA2 and RhoA would lead cells to escape from niche controls resulting in tumor metastasis (54). CCNA2 affects the activity of progesterone receptors through multiple mechanisms making it a key regulator of progesterone-induced breast cancer (55). TOP2A gene is responsible for encoding the ribozyme (type 2 topoisomerase, TOP2A) that plays a key role in the cell cycle. TOP2A can be used as a biomarker to predict the prognosis of a variety of cancers, such as bladder urothelial carcinoma (56), pancreatic cancer (57), prostate cancer (58) and osteosarcomas (59). TOP2A has become a major cellular target for many chemotherapeutic drugs and can be used early to identify patients who may be benefited from adjuvant or neo-adjuvant targeted therapy approaches. Some studies have shown that the change of TOP2A is an independent predictive marker of sensitivity to anthracyclines in adjuvant therapy for early breast cancer (60,61). TOP2A is also used as a marker for the efficacy of pegylated lyposomal doxorubicin (PLD) in epithelial ovarian cancer (62).

In this study, two GSE datasets were selected to screen DEGs and survival analysis was performed in GEPIA and UALCAN databases, which reduced the random errors that may be caused by a single data set and improved the reliability of the results. However, this study also has some limitations. First, the small sample size of this experiment limits the generalization of the conclusion. Second, this result has not been verified in clinical samples due to the accessibility of the data. In the following study, we can collect clinical samples and conduct molecular experiments to verify the six key hub genes we have discovered. In addition, more experiments can be conducted to clarify the upstream regulatory pathways and downstream mechanisms of key differential genes.

In conclusion, there were significant differences in expression levels of CDK1, CCNB1, CDC20, AURKA, CCNA2 and TOP2A between normal and CC tissues. The expression levels of CCNA2 and CCNB1 were negatively correlated with the survival time of patients. CCNA2 and CCNB1 may be potential biomarkers and therapeutic targets for CC.