Integrated bioinformatics analysis to identify key genes related to the prognosis of esophageal squamous cell carcinoma.

BACKGROUND: Esophageal squamous cell carcinoma (ESCC) is a serious threat to human health and life. The National Center for Biotechnology Information Gene Expression Omnibus (NCBI-GEO) database provides valuable information on genes related to the pathogenesis and prognosis of ESCC, which helps us to make in-depth understanding about the disease and improve its prognosis.
METHODS: Four microarray profiles [GSE77861 (African Americans), GSE26886 (Germans), GSE17351 (Americans), and GSE45670 (Chinese)] from the NCBI-GEO including 49 ESCC tissues and 41 corresponding normal tissues were collected. Integrated bioinformatics methods, including protein-protein interaction (PPI) network analysis, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses, and Kaplan-Meier plotter were applied to determine the differentially expressed genes (DEGs) in ESCC together with their core functions and relationship with survival.
RESULTS: A total of 220 upregulated and 112 downregulated genes were identified as DEGs in ESCC, of which, 40 upregulated genes were core function genes. The DEGs were mostly involved in DNA replication and cell cycle pathways. Survival analysis and Bonferroni adjustment showed kinesin family member 18A (KIF18A) and TTK protein kinase (TTK) to be related to prognosis in ESCC.
CONCLUSIONS: The findings of the present study verified the previously proposed association between TTK and patient survival in ESCC, and identified KIF18A as ESCC prognosis-related gene markers for the first time. The underlying mechanism needs to be further investigated using larger sample size studies and biological experiments in future. 2021 Translational Cancer Research. All rights reserved.

Introduction

Esophageal cancer presents a serious threat to human health and life. In 2018, there were an estimated 572,034 new cases of esophageal cancer and 508,585 deaths, which saw this disease rank 7th for incidence and 6th for mortality out of all malignant tumors (1). In East Asia, esophageal squamous cell carcinoma (ESCC) is the major pathological type of esophageal cancer (2). Despite advancements in surgery, radiotherapy techniques, and systematic treatments, the prognosis of ESCC is still poor, with the 5-year survival rate standing at around 15% (3). Together with primary prevention, the promotion of early cancer screening programs, and the development of innovative therapy strategies, a comprehensive understanding of the genetic changes of ESCC would aid in improving the prognostic outlook for patients. Acquiring in-depth knowledge of the pathogenesis, progression, and prognosis of ESCC would help to deepen our perceptions of the disease, stratify patients by prognosis, and provide the foundation for precision medicine and targeted therapies. However, previous reports on ESCC genetic susceptibility genes have been inconclusive (4-8). More investigations on genetic alterations are needed to improve our understanding of ESCC and explore its underlying mechanisms; specifically, analysis of differentially expressed genes (DEGs) is an effective approach to identifying genes of interest in the disease.

Gene chip is a reliable and effective technique to quickly identify DEGs (9). DEGs identified using this technique can be deposited in public databanks for future reference. Access to such open data and integrated information can provide researchers with worthwhile hints for novel study ideas.

To identify genetic factors relating to the pathogenesis and prognosis of ESCC, and to explore the potential mechanisms, we conducted an integrated bioinformatics analysis based on the National Center for Biotechnology Information Gene Expression Omnibus (NCBI-GEO) microarray database. Unlike the previous studies (4,5), we selected microarray datasets based on the same gene chip to reduce heterogeneity generated from experimental tests. Also, we selected gene profiles with no restrictions on patients’ race, which might make our conclusion more generalizable. We present the following article in accordance with the MDAR checklist (available at http://dx.doi.org/10.21037/tcr-20-3220).

Methods

Microarray profiles

Six gene expression profiles were collected from that NCBI-GEO microarray database: GSE77861 (African Americans), GSE26886 (Germans), GSE17351 (Americans), and GSE100942, GSE45670, and GSE33810 (Chinese). Each of the 6 microarray profiles contained both ESCC and normal esophageal epitheliums tissues, and all expression data were built on GPL570 platform (HG-U133_Plus_2; Affymetrix Human Genome U133 Plus 2.0 Array).

Data processing of DEGs

GEO2R online tools were applied to identify DEGs between ESCC and normal esophageal epithelial tissues (10). The filter criteria included |logFC|>0 and adjusted P<0.05. Two Chinese datasets, GSE100942 and GSE33810, were excluded from further analysis, due to having an adjusted P value of ≥0.05. Finally, GSE77861, GSE26886, GSE17351, and GSE45670, including 49 cancer tissues and 41 normal tissues, were included in the genetic analysis. Common DEGs among the 4 microarray datasets were determined using a Venn diagram (http://bioinformatics.psb.ugent.be/webtools/Venn/). DEGs were considered to be upregulated if logFC >0, and downregulated if logFC <0.

PPI network analysis

The protein–protein interaction (PPI) network of the identified DEGs was constructed by estimating the probable interactions between the DEGs using the online instrument Search Tool for the Retrieval of Interacting Genes (STRING) in Cytoscape under the conditions of maximum number of interactions 0 and confidence score ≥0.4 (11,12). The Cytoscape Molecular Complex Detection app was applied to build modules of the PPI network, with the degree cutoff set at 2, maximum depth at 100, k-core at 2, and node score cutoff at 0.2.

Gene Ontology (GO) and pathway enrichment analysis

Based on high-throughput transcriptomic or genomic data, we used the bioinformatics tool Database for Annotation, Visualization and Integrated Discovery (DAVID) to conduct a GO analysis of hub DEGs to analyze their genetic features, including molecular functions (MFs), cellular components (CCs), and biological processes (BPs) (Bonferroni-corrected P<0.05/counts) (13,14). DAVID was also employed to visualize Kyoto Encyclopedia of Genes and Genomes (KEGG) biological pathways (P<0.05) (15,16).

Survival analysis

Associations between the identified hub genes and ESCC survival in public genome databases, including the European Genome-phenome Archive (EGA), The Cancer Genome Atlas (TCGA), and GEO (Affymetrix microarrays only), were evaluated using the online Kaplan-Meier plotter (17). The results were presented as hazard ratios (HRs) with 95% confidence intervals (CIs) and as the log-rank P value. Because of the large number of genes to be tested, we applied both Bonferroni-corrected P<0.05/the number of genes and false discovery rate (FDR) <0.2 to adjust for multiple comparisons.

This study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). All data used in our study are available in public databases, including the NCBI-GEO microarray database, EGA, and TCGA. Individual consent for this analysis was waived.

Results

A total of 2,720 upregulated genes and 2,725 downregulated genes were obtained from GSE77861 (African Americans); 15,133 upregulated genes and 6,263 downregulated genes were obtained from GSE26886 (Germans); 567 upregulated genes and 301 downregulated genes were obtained from GSE17351 (Americans); and 6,428 upregulated genes and 5,572 downregulated genes were obtained from GSE45670 (Chinese). The DEGs identified from 49 cancer tissues and 41 normal matching tissues in the above 4 microarray profiles were compared. After the exclusion of unknown genes, 332 common DEGs, including 220 upregulated genes and 112 downregulated genes, were identified ( and ).

Figure 1

Common DEGs in the 4 microarray datasets via Venn diagram. Blue represents GSE77861, pink represents GSE45670, green represents GSE26886, and yellow represents GSE17351. (A) A total of 221 upregulated DEGs (included 1 unknown gene) and (B) 112 downregulated DEGs. DEGs, differentially expressed genes.

Table 1

Common differentially expressed genes (DEGs) identified from 4 microarray datasets of esophageal squamous cell carcinoma tissues compared with normal esophageal tissues (n=332)

DEGs	Gene name
Upregulated genes (n=220)	KRT17 IPO9 CDC25B NUDT1 FOXM1 PLXNA1 DDX18 KIAA0930 KIF4A MEX3D MCM5 TMEM132A MIR1292 CD276 MIR1182 PAPD7 CDCA2 PTRH2 OTULIN RPN2 APOC1 GPNMB FZD2 KIF18B ENAH ATP2C1 TMPO MBD4 CSE1L CENPN RFC4 MIR664B TMEM9 UBE2S NIFK OCIAD2 NTMT1 TRAF4 WDR5 PHF20L1 DESI2 E2F6 FZD6 CUX1 CKAP5 HDGF ATP1B3 DHX36 NUDCD1 PAK2 ATP2B1 FAM64A PFDN2 ITPR3 U2SURP DDX11 IMP4 RIOK1 DSCC1 KNTC1 LRRC8D SAC3D1 AURKB GANAB GTSE1 NUF2 CDCA5 TEAD4 YWHAG IGF2BP2 WDR53 PSMB4 TMEM97 LAMC2 IMPDH1 ASXL1 IGF2R COLGALT1 BEND3 LPCAT1 TCOF1 DNAH14 HMGB3 HOMER3 MRGBP TSEN15 MYBL2 CENPO CKS1B BLM DIEXF CLCN7 PDCD2L NAPEPLD GABPB1 TFRC PSME4 PARP1 SERPINH1 CDCA3 SIX4 FLAD1 TCF3 DTL MEST GMNN CASC7 TTK SLC52A2 MAPKAPK2 FOXK2 HSPD1 RSRC1 COL7A1 KIF14 MCM10 ELF4 NELL2 CHPF2 SPAG5 AGRN CLPTM1L EML4 MCM2 TGS1 CBX3 MARCKSL1 MTERF3 CENPI MIR1178 ARFGAP1 BID NAXE ABL2 ITGB4 ABCC4 RIF1 HOXA10 RUVBL1 ADO ACVR1C SCARB1 TRIP13 FNDC3B HEATR1 DHX9 DDX28 UBE2C TOP2A HELLS USB1 KIF18A CACYBP MCM6 DERL1 CTHRC1 UHRF1 SGO2 ZAK CENPF NDUFB9 BYSL TPX2 ANP32E PUF60 DNMT3B CDC6 AURKA CHAF1A RAD1 DDX39A ASAP1 RUNX1 ITGA6 PRKDC NOL11 C1orf131 UBE2T PTDSS1 ECT2 CDH3 ARPC1B RAD51AP1 MIR7112 MMP10 EPHB4 HMCES PLAU UTP4 EIF3B FLVCR1 PCYT1A SLC25A32 YEATS2 SNRNP200 DHX33 APMAP C4orf48 ZNF281 ISG20L2 TNFAIP8L1 ADRM1 TMEM184B MED30 NCAPH ATAD2 DNMT1 CAD RPA1 UBAP2L TMEM138 REL MLEC RTKN KIAA1804 ORC6 GMPS DBR1 WDR66 CTSC
Downregulated genes (n=112)	C4orf3 CPEB3 ALAD PAQR8 UFL1 TFAP2B AHNAK RSU1 CHMP2B CUL4B KALRN CALCOCO2 FBXL5 RBP7 EPB41L3 PIAS1 FIG4 CCSER2 CITED2 ABHD5 SLC6A1 TRIOBP NCOA1 FBXL3 NOL12 BBIP1 PLA2G12A KANK1 PTN MDFIC EPB41L4A HSPB8 RAP1A SPAG16 LNPEP PGRMC2 FYCO1 ANKRD35 DDAH1 PRMT2 HACE1 GID4 SLC30A4 TLR3 STX17 NUCB2 UBL3 SECISBP2L KCTD6 SHROOM3 ANKHD1-EIF4EBP3 COBL RAB11A RRAGD CNPPD1 PHYHD1 AR PINK1 C15orf52 ALDH3A2 PDZRN3 BCAP29 VIT SH3BGRL2 KCNAB1 SAMD5 SSBP2 CTTNBP2 SASH1 PHC1 GNE ARL6IP5 MYZAP DCAF10 LOC100132167 FAM219B AZI2 USO1 HSDL2 PPP1R7 KAT2B SLC30A9 SMDT1 LMBRD1 SCIN EMP1 EDN3 AK9 MTERF4 HPGD CPPED1 GAB2 FAM13C DOCK9 MGLL RORA ATP5O TRIM13 ENSA GNAQ SNX3 EIF4E3 CAPN5 FAM189A2 PCCA DEPTOR DENND4C CAB39L ABLIM3 GGTA1P CAST GAB1

DEGs, differentially expressed genes.

STRING was used to identify the nodes with the most connections as core function genes (Figure S1). As shown in , there were 40 hub genes, all of which were upregulated.

Figure 2

The 40 most connected nodes in module analysis via Cytoscape. Each blue square represents a gene, and the connected lines represent interactions between genes.

The 40 core DEGs were subjected to GO and KEGG pathway enrichment analyses. In terms of BPs, the DEGs were significantly enriched in cell division, mitotic nuclear division, DNA replication and sister chromatid cohesion, chromosome segregation, DNA replication initiation, and G1/S transition of mitotic cell cycle. In regard to CCs, the DEGs were significantly enriched in the nucleoplasm, nucleus, chromosome, centromeric region, midbody, and spindle midzone. In terms of MFs, the DEGs were mostly involved in ATP binding, protein binding, ATP-dependent microtubule motor activity, plus-end-directed, and chromatin binding ().

Table 2

Gene Ontology analysis of 40 core function differentially expressed genes in esophageal squamous cell carcinoma

Category	Term	Count	%	Genes	Bonferroni P value	FDR
GOTERM_BP_DIRECT	GO:0051301~cell division	18	45	KIF14, CDC6, CKS1B, NUF2, KNTC1, TPX2, KIF18B, CENPF, AURKA, UBE2C, MCM5, NCAPH, FAM64A, SPAG5, CDCA2, CDCA5, HELLS, CDCA3	1.59E-16	7.75E-16
	GO:0007067~mitotic nuclear division	13	32.5	CENPN, CDC6, FAM64A, KNTC1, NUF2, TPX2, CDCA2, CENPF, AURKA, AURKB, CDCA5, HELLS, CDCA3	6.10E-11	2.98E-10
	GO:0006260~DNA replication	9	22.5	CDC6, RFC4, DTL, MCM2, CHAF1A, MCM10, MCM5, MCM6, DSCC1	5.71E-07	2.79E-06
	GO:0007062~sister chromatid cohesion	7	17.5	CENPN, KIF18A, KNTC1, NUF2, CENPF, AURKB, CDCA5	3.42E-05	1.67E-04
	GO:0007059~chromosome segregation	6	15	CENPN, SPAG5, NUF2, CDCA2, CENPF, TOP2A	1.31E-04	6.38E-04
	GO:0006270~DNA replication initiation	5	12.5	CDC6, MCM2, MCM10, MCM5, MCM6	2.29E-04	0.00112
	GO:0000082~G1/S transition of mitotic cell cycle	6	15	CDC6, MCM2, MCM10, CDCA5, MCM5, MCM6	9.84E-04	0.004811
GOTERM_CC_DIRECT	GO:0005654~nucleoplasm	26	65	CKS1B, KIF4A, FOXM1, AURKA, AURKB, MCM10, MYBL2, GTSE1, CDCA2, TOP2A, CDCA5, CENPN, CDC6, RAD51AP1, DTL, TPX2, CENPF, ATAD2, MCM2, UBE2C, MCM5, MCM6, RFC4, SPAG5, UBE2T, DSCC1	5.52E-10	6.61E-09
	GO:0005634~nucleus	29	72.5	FOXM1, KNTC1, AURKA, AURKB, MCM10, NCAPH, TOP2A, CDCA5, HELLS, TRIP13, KIF14, CDC6, RAD51AP1, DTL, KIF18A, TPX2, NUF2, KIF18B, ATAD2, CENPF, MCM2, ECT2, MCM5, MCM6, UHRF1, FAM64A, SPAG5, CHAF1A, UBE2T	6.32E-06	7.57E-05
	GO:0000775~chromosome, centromeric region	6	15	CENPN, NUF2, CENPF, CDCA5, HELLS, DSCC1	1.21E-05	1.45E-04
	GO:0030496~midbody	7	17.5	KIF14, KIF4A, SPAG5, CENPF, AURKA, AURKB, ECT2	2.74E-05	3.29E-04
	GO:0051233~spindle midzone	4	10	KIF14, CDC6, AURKA, AURKB	7.80E-04	0.009348
GOTERM_MF_DIRECT	GO:0005524~ATP binding	19	47.5	KIF14, CDC6, KIF4A, KIF18A, TPX2, KIF18B, TTK, ATAD2, AURKA, AURKB, MCM2, UBE2C, MCM5, MCM6, RFC4, TOP2A, UBE2T, HELLS, TRIP13	2.80E-08	3.72E-07
	GO:0005515~protein binding	36	90	CKS1B, KIF4A, FOXM1, KNTC1, TTK, AURKA, AURKB, MYBL2, MCM10, GTSE1, NCAPH, TOP2A, CDCA5, HELLS, CDCA3, TRIP13, KIF14, CDC6, RAD51AP1, DTL, TPX2, KIF18A, NUF2, CENPF, KIF18B, MCM2, UBE2C, ECT2, MCM5, MCM6, UHRF1, FAM64A, RFC4, SPAG5, CHAF1A, DSCC1	1.49E-06	1.98E-05
	GO:0008574~ATP-dependent microtubule motor activity, plus-end-directed	4	10	KIF14, KIF4A, KIF18A, KIF18B	5.16E-04	0.00686
	GO:0003682~chromatin binding	8	20	CENPF, ATAD2, CHAF1A, CDCA5, TOP2A, UBE2T, MCM5, HELLS	0.001531	0.020359

FDR, false discovery rate.

The KEGG pathway enrichment analysis revealed that the DEGs were mostly involved in the DNA replication and cell cycle pathways. The implicated genes included replication factor C subunit 4 (RFC4), minichromosome maintenance complex component 2 (MCM2), MCM5, MCM6, cell division cycle 6 (CDC6), and TTK protein kinase (TTK) (; ).

Figure 3

The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of the 40 core differentially expressed genes in esophageal squamous cell carcinoma. (A) KEGG pathway enrichment of hsa03030: DNA replication. Involved genes were replication factor C subunit 4 (RFC4), minichromosome maintenance complex component 2 (MCM2), MCM5, MCM6. (B) KEGG pathway enrichment of hsa04110: cell cycle. Involved genes were cell division cycle 6 (CDC6), TTK protein kinase (TTK), MCM2, MCM5, and MCM6.

Table 3

Kyoto Encyclopedia of Gene and Genome pathway analysis of 40 core differentially expressed genes in esophageal squamous cell carcinoma

Term	Count	%	P-value	Genes	FDR
hsa03030: DNA replication	4	10	2.11E-05	RFC4, MCM2 MCM5, MCM6	0.013148
hsa04110: Cell cycle	5	12.5	3.01E-05	CDC6, TTK, MCM2MCM5, MCM6	0.01875

FDR, false discovery rate.

Finally, we conducted a survival analysis of 39 identified core DEGs, as PICALM interacting mitotic regulator (FAM64A) was excluded due to a lack of survival data. With P<0.05 indicating statistical significance, 15 genes were found to be statistically significantly related to survival and are shown in . The survival curves of the other 24 genes are presented in Figure S2.

Figure 4

Kaplan-Meier plots of the 15 genes with statistical significance (P<0.05). Gene name is shown at the top of each figure part. Black line represents low expression of the identified gene, and red line represents high expression of the identified gene. Corresponding life table, hazard ratio, and log-rank P value are shown in each figure part.

The GO analysis of the 15 significant genes revealed that the most enriched BP, CC, and MF were cell division, nucleoplasm, and ATP binding, respectively (). KEGG pathway enrichment analysis again showed enrichment mainly in the cell cycle pathway ( and ).

Table 4

Gene Ontology analysis of 15 differentially expressed genes related to esophageal squamous cell carcinoma survival

Category	Term	Count	%	Genes	Bonferroni P value	FDR
GOTERM_BP_DIRECT	GO:0051301~cell division	8	53.33333	CKS1B, CDC6, NCAPH, SPAG5, TPX2, KIF18B, AURKA, CDCA3	5.17E-07	5.47E-06
GOTERM_CC_DIRECT	GO:0005654~nucleoplasm	10	66.66667	CKS1B, CDC6, RAD51AP1, SPAG5, TPX2, ATAD2, AURKA, MCM10, GTSE1, MCM6	0.002063	0.041061
GOTERM_MF_DIRECT	GO:0005524~ATP binding	8	53.33333	CDC6, KIF18A, TPX2, KIF18B, ATAD2, TTK, AURKA, MCM6	0.003221	0.075191

FDR, false discovery rate.

Table 5

Kyoto Encyclopedia of Genes and Genomes pathway analysis of 15 differentially expressed genes related to esophageal squamous cell carcinoma

Term	Count	%	P value	Genes	FDR
hsa04110: cell cycle	3	20	0.003111	CDC6, TTK, MCM6	1.480007

FDR, false discovery rate.

For multiple comparisons, we applied a Bonferroni correction of 0.05/39=0.001282 as Bonferroni-adjusted statistical significance. We also applied FDR <0.20 as a second reference. Eventually, 2 genes were identified as being significantly related to survival: kinesin family member 18A (KIF18A) (HR =0.24, 95% CI: 0.10–0.59, log-rank P=0.00072, FDR =0.10) and TTK (HR =0.20, 95% CI: 0.07–0.55, log-rank P=0.00057, FDR =0.10). Patients with a high expression of these genes had a better prognosis than those with a low expression.

Discussion

By performing a gene enrichment bioinformatics analysis of ESCC which did not factor in race, we identified 332 common DEGs from 4 microarray datasets (GSE77861, GSE26886, GSE17351, and GSE45670). Through module analysis, we identified 40 upregulated genes as core function genes, and the most involved pathways were the DNA replication and cell cycle pathways, which supported our previous knowledge of carcinogenesis. In the prognostic analysis, we identified 15 genes associated with survival. For these survival-related genes, the most enriched BP, CC, and MF were cell division, nucleoplasm, and ATP binding, respectively, and the most enriched pathway was the cell cycle pathway. Furthermore, for multiple comparisons, we applied the Bonferroni correction and FDR, and identified 2 genes, KIF18A (HR =0.24, 95% CI: 0.10–0.59, log-rank P=0.00072, FDR =0.10) and TTK (HR =0.20, 95% CI: 0.07–0.55, log-rank P=0.00057, FDR =0.10), as prognosis-related genes.

Although our findings present associations between KIF18A/TTK and ESCC development and prognosis, the exact mechanisms still are unclear, especially as research on these aspects of ESCC is lacking. A number of in vitro studies have explored the functional role of TTK in cancer, and have discovered the major pathways involved to be the TTK–AKT–mechanistic target of rapamycin kinase (mTOR) pathway in hepatocellular carcinoma and TTK/AKT interaction with B-Raf/mitogen-activated protein kinase 1 (MAPK1) signals in melanoma (18,19). KIF18A, KIF18B, and KIF19 are kinesin-8 family members. KIF18A together with dyneins, that are as the key component of canonical plus-end stepping in the mitotic spindle and the depolymerization of microtubules, is primarily involved in mitotic chromosome alignment, maintenance of chromosomal stability, and cell division. It also acts as the major driving force for cell migration (20-26). KIF18A dysregulation may give rise to genomic instability (27), eventually resulting in neoplasm development and metastatic progression (28).

Earlier studies have shown that the overexpression of human KIF18A is related to a poor prognosis of colorectal cancer (20), invasive breast cancer (29), and primary hepatocellular carcinoma (30). However, studies on KIF18A in ESCC are scarce. To the best of our knowledge, the present study is the first to reveal the relationship of KIF18A with the prognosis of ESCC. Taken together, previously published findings (20,29,30) and the results of the present study imply that KIF18A is a possible prognostic predictor; however, the molecular mechanisms need further elucidation.

In terms of treatment, the mechanism and molecules relevant to mitosis are well-established targets for microtubule-binding chemotherapeutic agents, such as taxol and vinca alkaloids (31), which is consistent with the fact that regimens containing paclitaxel are the preferred chemotherapeutic treatment for ESCC in Chinese clinical practice (32). Additionally, mitotic kinesins have also been considered as potentially valuable targets in the development of anticancer drugs (33).

TTK is a cancer/testis antigen (CTA) and an HLA-A2402-restricted epitope peptide (34). As well as in the testes, early-evolving embryos, the thymus, and placentas, the expression of CTA can be detected in various malignant tumors, and it is rarely detected in benign tissues (35,36). TTK participates in cell proliferation and migration through the phosphorylation of multiple amino acids, such as serine, threonine, and tyrosine hydroxyamino acids, and functions to maintain genomic integrity through regulating the spindle assembly checkpoint (35,37-39).

Previous studies have revealed the associations between TTK overexpression and the growth of pancreatic ductal adenocarcinoma (40), hepatocellular carcinoma (18), breast cancer (41), and melanoma (19). A previously published study found that TTK was a key component in ESCC progression and predicted a poor prognosis, although the molecular mechanisms were unclear (4).

TTK inhibitors suppress the activity of monopolar spindle 1 kinase, specifically deactivate the spindle assembly checkpoint, which results in chromosome separation dysfunction, heteroploidy, and ultimately, cell death. Thus, TTK inhibition has arisen as a promising therapeutic strategy for triple-negative breast cancer and malignant mesothelioma (37,42-46). Furthermore, the testis is an immunoprivileged organ, and CTAs are ideal candidates as tumor-related antigens (47). TTK has emerged as a valuable target for tumor immunotherapy, including cancer vaccines, chimeric antigen receptor T-cell immunotherapy, and immune checkpoint inhibitors, due to its cancer-specific expression and strong in vivo immunogenicity, together with its critical roles in cell proliferation and migration, and spindle assembly checkpoint regulation (42,47,48). In several pharmaceutical clinical trials on ESCC, TTK expression was reported as both a marker for the treatment response and a potential cancer vaccine target (49,50).

Our study has several limitations. First, a previous published study conducted using the laiurger microarray dataset (GSE38129 and GSE20347) recognized that the TPX2 microtubule nucleation factor, cyclin dependent kinase 1, and centrosomal protein 55 were associated with relapse-free survival (51). Instead, we selected another microarray dataset built on GPL570 platform, including diverse races, and chose overall survival as our study endpoint rather than relapse-free survival, which resulted in our conclusion being inconsistent with those of earlier studies. However, the discrepancy between our research and previous research revealed that the bioinformatics analysis results were unstable and emphasized the significance of molecular mechanism studies. Second, the survival data on ESCC were limited, and the small sample size reduced the reliability of the results; therefore, larger sample sizes are required in future research. Third, for both KIF18A or TTK, molecular mechanism studies were limited and none of them focused on ESCC. It should be recognized that gene enrichment analysis can only provide clues, and our findings should be further verified by biological experiments. Finally, bioinformatics analysis is a growing field. New analyses and gene expression microarray studies are in development, which might affect our conclusions. The challenge for researchers performing bioinformatics analyses is to integrate up-to-date analytical methods and datasets accordingly.

In conclusion, our analysis was the first to identify KIF18A related to ESCC prognosis and verified the previous association between TTK and ESCC survival. KIF18A is a key component of mitotic activity, and TTK is a good candidate for tumor immunotherapy. The results of the present study were in accordance with those of studies on the clinical use of paclitaxel and also shed light on directions for future anticancer immunotherapy research and development. However, our conclusions need to be further validated in studies with larger sample sizes and biological experiments.