Oncomine database (https://software.oncomine.com/resource/login.html) is useful tool for gene-wide expression analyses (Rhodes et al., 2004). Oncomine was used to summarize DTYMK expression in tumor tissues and their normal counterparts across human cancers. Parameters for disease screens included: p Value less than 0.001, fold change over 1.5, Gene rank: Top 10%, data type mRNA.

GEPIA (http://gepia.cancer-pku.cn/index.html) contains RNA sequencing expression data of tumors and normal samples from the TCGA and the GTEx projects (Tang et al., 2017). The comparison of DTYMK expression between tumors and normal samples and in different tumor stages was performed using this platform. The effects of DTYMK expression on prognosis (overall survival and disease-free survival) were also studied.

Genetic alterations of DTYMK in cancers were analyzed using cBioPortal database (http://www.cbioportal.org/), which comprises molecular profiles and clinical attributes from The Cancer Genome Atlas (Cerami et al., 2012; Gao et al., 2013).

Clinical Proteomic Tumor Analysis Consortium (CPTAC) Confirmatory/Discovery dataset in UALCAN database (http://ualcan.path.uab.edu/) was useful for analysis DTYMK protein expression (Chandrashekar et al., 2017; Chen et al., 2019).

TIMER web server (https://cistrome.shinyapps.io/timer/) is a comprehensive resource for systematical analysis of immune infiltrates across diverse cancer types (Li et al., 2016; Li et al., 2017). The abundances of six immune infiltrates (B cells, CD4⁺ T cells, CD8⁺ T cells, Neutrophils, Macrophages and Dendritic cells) were estimated using the TIMER algorithm. The analysis of DTYMK expression in tumors and control samples as well as its relationship with immune infiltration was carried out using TIMER.

The Kaplan-Meier plotter (http://kmplot.com/analysis/) can be harnessed for assessing the impact of genes on patient survival in 21 cancer types (Nagy et al., 2021). The effect of DTYMK on survival (overall survival and disease-free survival) was evaluated in all cancer types.

Based on TCGA data, the univariate Cox regression analysis and multivariate Cox regression analysis to further clarify the relationship between DTYMK expression and the prognosis in LIHC and LUAD. The time-dependent receiver operating characteristic (ROC) analysis was used to compare the predictive accuracy of DTYMK expression in LIHC and LUAD.

PrognoScan (http://www.prognoscan.org/) was used to evaluate the relationship between gene expression and patient prognosis across a large collection of publicly available cancer microarray datasets (Mizuno et al., 2009). We searched the item “DTYMK” and screened datasets which showed the significant relationship between DTYMK expression and patient survival. Based on the analysis results, a web tool, Sangerbox (http://sangerbox.com/Tool) was then employed to draw a forest plot.

TISIDB (http://cis.hku.hk/TISIDB/index.php) integrates genomics, transcriptomics and clinical data of 30 cancer types from The Cancer Genome Atlas (TCGA) as well as other data for tumor and immune system interaction (Ru et al., 2019). TISIB was used to evaluate DTYMK expression in different immune subtypes and molecular subtypes of tumors. We also used TISIB to explore the relationship between DTYMK and immune infiltration, MHC expression, immune inhibitors, immunostimulators, chemokines and chemokine receptors.

GeneMANIA (http://genemania.org/) helps to determine related genes of input genes with a large set of functional association data (Warde-Farley et al., 2010). We used GeneMANIA to define the protein-protein interaction network of DTYMK.

The STRING database (https://www.string-db.org/) comprises and predicts physical and functional associations between proteins (Szklarczyk et al., 2021). Fifty interactive proteins of DTYMK were predicted using STRING. Key parameters included: minimum required interaction score (medium confidence 0.400), max number of interactors to show (no more than 50 interactors). Kyoto encyclopedia of genes and genomes (KEGG) enrichment and gene ontology (GO) analysis of these fifty interactive genes were performed using The Database for Annotation, Visualization and Integrated Discovery (DAVID, https://david.ncifcrf.gov/home.jsp) (Huang da et al., 2009a; Huang da et al., 2009b).

The LinkedOmics database (http://www.linkedomics.org/) encompasses multi-omics data and clinical information from TCGA (Vasaikar et al., 2018). This platform allows researchers to analyze cancer multi-omics data within and across tumor types. Here, the LinkFinder module was employed to generate a volcano plot and heatmap plot for gene correlations with DTYMK in LIHC and LUAD. The LinkInterpreter module was used for Gene Set Enrichment Analysis (GSEA) for correlated genes in LIHC and LUAD.

Drug bank (https://go.drugbank.com/) combines detailed drug data with comprehensive drug target information (Wishart et al., 2006; Wishart et al., 2008; Knox et al., 2011; Law et al., 2014; Wishart et al., 2018). The pharmaco-transcriptomics module was used to search those pharmaceutical compounds that significantly influenced DTYMK expression.

Lung cancer cell lines including 95C and 95D were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). A549, H1299, MIA Paca-2 and HepG2 cells were provided by American Type Culture Collection (ATCC, Manassas, VA, United States). 95C, 95D and H1299 were cultured with Roswell Park Memorial Institute (RPMI)-1640 medium (Thermo Fisher Scientific, Waltham, MA, United States) supplemented with 10% fetal bovine serum (FBS, PAN Biotech, Aidenbach, Germany). A549, MIA Paca-2 and HepG2 were cultured with Dulbeccos Modified Eagles Media (DMEM) medium (high glucose, Thermo Fisher Scientific) supplemented with 10% FBS and penicillin/streptomycin (100 U/mL, North China Pharmaceutical Company, Shijiazhuang, China). All cell lines were cultured under 37°C and 5% CO₂ conditions.

Two shRNAs targeting DTYMK (NM_012145) were designed using BLOCK-iT™ RNAi Designer (https://rnaidesigner.thermofisher.com/rnaiexpress/sort.do) and synthesized by Sangon Biotech (Shanghai, China). Paired primers of DTYMK shRNAs or scrambled shRNAs were annealed (heating at 90°C for 15 min and cooling at room temperature for 2 h) and then ligated into the EcoR I/Age I linearized pshRNA-EGFP-P2A-Puro plasmid (designated as pshDTYMK-EGFP-P2A-Puro and pshNC-EGFP-P2A-Puro). Sequences for shDTYMK01: forward primer: CCG​GGG​GAA​CAA​GTG​CCG​TTA​ATT​ATT​CAA​GAG​ATA​ATT​AAC​GGC​ACT​TGT​TCC​CTT​TTT​G; Reverse primer: AATTCAAAAA GGG​AAC​AAG​TGC​CGT​TAA​TTA​TCT​CTT​GAA​TAA​TTA​ACG​GCA​CTT​GTT​CCC; Sequences for shDTYMK02: Forward primer: CCG​GGC​AAA​TCG​CTG​GGA​ACA​AGT​G TTC​AAG​AGA​CAC​TTG​TTC​CCA​GCG​ATT​TGC​TTT​TTG; Reverse primer: AAT​TCA​AAA​AGC​AAA​TCG​CTG​GGA​ACA​AGT​GTC​TCT​TGA​ACA​CTT​GTT​CCC​AGC​GAT​TTG​C.

After cloning and validation by sequencing, pshDTYMK-EGFP-P2A-Puro and pshNC-EGFP-P2A-Puro were used to generate lentiviruses with packaging vectors including psPAX2 and pMD2.G and the transfection reagent Polyethylenimine (PEI, Polysciences, Warrington, PA, United States) as detailed in a previous study (Dey et al., 2017).

To establish stable knockdown cell lines, HepG2 and H1299 cells were transfected with Lenti-shDTYMK and ctrl viruses (10 MOI) and selected using puromycin (2-4 μg/ml) for 2 weeks. Western blot was employed to validate knockdown efficiencies as detailed below.

After DTYMK knockdown, HepG2 and H1299 cells were seeded on 6-well plates at 400 cells/well and allowed to grow for 10 days. When more than 50 cells were observed in each colony, culture medium was removed and washed twice with PBS. Tumor cells were fixed using absolute methanol for 15 min and stained with 0.1% crystal violet. After washing with distilled water three times, photos were obtained using microscopy (Olympus, Tokyo, Japan).

HepG2 and H1299 cells were seeded on 6-well plates at 5×10⁵ cells/well and allowed to grow for 24 h. When cell confluence reached 70%, culture medium was removed and cells were washed twice with PBS. Cells were scratched using a sterilized pipette tip and then cultured with DMEM (HepG2) or 1640 (H1299) without FBS. Wound healing was examined and photos were captured under a light microscope after 48 h. Wound healing areas were calculated using ImageJ (version 1.53, NIH, Bethesda, MA, United States). Wound closure was calculated using the algorithm: wound closure % = [(A_0h-A_48h)/A_0h] × 100%.

HepG2 and H1299 cells were seeded on the upper chambers of Transwells (Corning, Corning, NY, United States) at 5 × 10⁴ cells/well in medium without FBS. Chambers were then immersed into medium containing 10% FBS and tumor cells were allowed to migrate for 24 h. Migratory cells were fixed with 4% formaldehyde, penetrated by absolute methanol and stained using 0.1% crystal violet. Migratory cells were observed under the microscope and counted in four different fields.

Western blot analysis of DTYMK expression was performed according to a previous study (Miao et al., 2020). Preparation of protein from 95C, 95D, A549, MIA Paca-2, HepG2 and H1299 cells was performed using the Mammalian Protein Extraction Kit (CWBIO company, Beijing, China) supplemented with proteinase inhibitor (dilution 1:99). Protein concentrations were determined using the Bicinchoninic assay (BCA) method (CWBIO company). Proteins were separated by 10% SDS-PAGE gels and then transblotted into polyvinylidene fluoride (PVDF) membrane. After washing with TBS-Tween and blocking using 5% skim milk powder, the PVDF membrane was incubated with the DTYMK polyclonal antibody (1:1000, #15360-1-AP, Proteintech, Rosemont, IL, United States). β-actin (1:1000, Proteintech) was used as an internal loading control. After washing, the membrane was incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (1:5000, Proteintech) for 1 hour. Electrochemiluminescence (ECL) reagent was used to visualize protein bands and ImageJ software (version 1.53, NIH) used to measure the mean gray-value of protein bands.

Data were presented as Mean ± SD. DTYMK expression comparison among different cancers in TIMER was calculated using the Wilcoxon rank-sum test. Patient survival comparisons between individuals with high and low expression of DTYMK was carried out using GEPIA, Kaplan-Meier plotter and PrognoScan. DTYMK co-expressing genes were obtained from LinkedOmics and screened using Spearman’s correlation. For in vitro experiments, the difference among groups was calculated by One-way ANOVA and multiple comparisons was performed using the Tukey post-test. p values of less than 0.05 were considered as statistically significant.

To characterize DTYMK expression in various of cancer types, DTYMK transcription was analysed using the Oncomine platform (Figure 1A). Colorectal cancer, lymphoma, and breast cancer were the top three malignancies showing increased DTYMK expression in this platform (Supplementary Table S1). DTYMK expression was then compared between tumor tissues and normal controls by TIMER using data derived from the TCGA database. Significant augmentation of DTYMK expression was observed in 19 cancer types, including bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), and cervical squamous cell carcinoma (CESC) etc. (Figure 1B). Genetic alteration of DTYMK in cancers included gene mutations, structural variants, amplification and deep deletion (homozygous deletion). Although deep deletion of DTYMK was common to most cancers, the frequency was low. For instance, there was 7.84% (20/255) in sarcoma (Figure 1C). Gene amplification of DTYMK was more common in cancers such as ovarian serous cystadenocarcinoma (OV), uterine carcinosarcoma (UCS), pancreatic adenocarcinoma (PAAD), LUAD, thymoma (THYM), and LIHC (Figure 1C). GEPIA was then employed to validate DTYMK mRNA expression. GEPIA combined the data from TCGA and GTEx databases, in which more tumor and normal samples were enrolled. DTYMK expression was increased in 19 cancer types including BLCA, CESC and colon adenocarcinoma (COAD) (Figure 1D). Using the CPTAC dataset from the UALCAN platform, DTYMK protein expression in several cancers was further validated. DTYMK protein expression was reduced in breast cancer and clear cell renal cell carcinoma when compared with normal tissues (Figure 2). In contrast, significantly enhanced DTYMK expression was found in COAD, LUAD, OV, and UCEC tissues (Figure 2). These results demonstrated that DTYMK expression was upregulated in most cancer types and might play an essential role in human cancer progression.

To further explore DTYMK expression in human cancers obtained from different stages, we analyzed the relationship between DTYMK and cancer stages using GEPIA. In adrenocortical carcinoma (ACC), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC) and LUAD, stage IV tumor tissue was associated with higher expression of DTYMK compared to other stages (Figures 3A–C,E). However, DTYMK expression was decreased in stage IV LIHC tissues (Figure 3D).

Immune subtypes of solid tumors have been divided into six categories according to the transcriptomic profiles of over 10,000 patients from The Cancer Genome Atlas (TCGA) cancer types (Thorsson et al., 2018). The six categories include C1 (wound healing), C2 (IFN-gamma dominant), C3 (inflammatory), C4 (lymphocyte depleted), C5 (immunologically quiet), C6 (TGF-β dominant) (Thorsson et al., 2018). C1 tumors are characterized by high induction of angiogenesis and proliferation; C2 are characterized by the highest M1/M2 macrophage polarization (towards M1); C3 are characterized by elevated Th17 and Th1 genes; C4 are characterized by macrophage infiltration and suppressed Th1 lymphocytes; C5 are characterized by the lowest lymphocyte and highest macrophage response and C6 are characterized by highest TGF-β signature and high lymphocyte infiltration (Thorsson et al., 2018). The variation of DTYMK expression in different subtypes was found in 19 cancers (Figures 3F–X). The top five (ranked by p value) cancers with high DTYMK expression included LUAD (p = 1.65e-35), BRCA (p = 3.79e-30), lower grade glioma (LGG) (p = 5.98e-21), LIHC (p = 8.04e-18), and STAD (p = 3.36e-15). We noticed that DTYMK expression was the lowest in C3 subtypes (inflammatory, high Th1 and Th17 infiltration) of human cancers, especially in BLCA, BRCA, ESCA, LUAD, LUSC, OV, PAAD, prostate adenocarcinoma (PRAD), READ, SARC, and STAD. DTYMK expression also differed in different molecular subtypes of cancers. Significant alteration of DTYMK expression was shown in 11 cancer types (Figures 4A–K). The top five cancers (ranked by p value) included BRCA (p = 1.18e-25), LGG (p = 4.02e-21), UCEC (p = 1.41e-10), PRAD (p = 8.61e-06), and OV (p = 3.78e-04). In BRCA, DTYMK expression was higher in basal-like, human epidermal growth factor receptor 2 (Her2)-enriched and luminal B subtypes compared with luminal A and normal-like subtypes. In LGG, DTYMK expression was the highest in glioma cytosine-phosphate-guanine island methylator phenotype (G-CIMP)-low subtype and the lowest in chromosome 1p/19q codeletion (Codel) subtype. In UCEC, DTYMK expression was downregulated in copy number low (CN_low) subtype. In PRAD, DTYMK expression was increased in speckle type BTB/POZ protein (SPOP) subtype and decreased in friend leukemia integration 1 (FLI1) subtype. In OV, DTYMK expression was downregulated in differentiated and mesenchymal subtype but upregulated in immunoreactive and proliferative subtype. Our results suggested that DTYMK expression varied among different immune and molecular subtypes.

The TIMER database was employed to elucidate whether DTYMK impacted on immune infiltration in human cancers. The results showed that DTYMK expression was significantly correlated with the infiltration of six types of immune cells, which included B cell, CD8⁺ T cell, CD4⁺ T cell, macrophage, neutrophil and dendritic cell (Supplementary Figures S2A,B). Among all the tumors, LGG, LIHC, LUAD, and STAD were the most closely correlated cancer types with DTYMK expression (Figures 5A–AB). Notably, DTYMK expression was positively correlated with the infiltration of B cell (LGG: p = 3.53e-07; LIHC: p = 1.35e-13), CD4⁺ T cell (LGG: p = 3.02e-14; LIHC: p = 1.42e-04), macrophage (LGG: p = 1.25e-10; LIHC: p = 8.57e-12), neutrophil (LGG: p = 1.28e-06; LIHC: p = 1.49e-05) and dendritic cell (LGG: p = 4.32e-13; LIHC: p = 5.44e-13) in LGG and LIHC (Figures 5A–N). In contrast, in LUAD and STAD, DTYMK expression was negatively correlated with the infiltration of B cell (LUAD: p = 5.76e-09; STAD: p = 2.05e-08), CD8⁺ T cell (LUAD: p = 2.35e-02; STAD: p = 1.89e-02), CD4⁺ T cell (LUAD: p = 1.82e-07; STAD: p = 3.00e-13), macrophage (LUAD: p = 3.38e-04; STAD: p = 1.56e-18), and dendritic cell (LUAD: p = 3.11e-06; STAD: p = 1.85e-07) (Figures 5O–AB). These results suggested DTYMK might play distinct roles in immune infiltration in different cancer types.

The relationship between DTYMK and immune infiltration was further confirmed using the TISIDB database, which explored 28 types of infiltrated lymphocytes (Supplementary File S1). Although DTYMK was positively correlated with the infiltration of activated CD4⁺ T cells and CD8⁺ T cells in LGG, LIHC, LUAD, and STAD, effector memory CD4⁺ T and CD8⁺ T infiltration was inversely correlated with DTYMK expression (Figure 6A). The infiltration of CD4⁺ T cellsubsets, Th1, Th2 and Th17, was also negatively correlated with DTYMK expression in LGG, LIHC, LUAD, and STAD (Figure 6A). Moreover, follicular helper T cell infiltration was positively correlated with DTYMK in LGG, but was negatively correlated in LUAD and STAD (Figure 6A). Lastly, the infiltration of innate immune cells including eosinophils and natural killer (NK) cells was also negatively correlated with DTYMK in LIHC, LUAD and STAD (Figure 6A). These results revealed the contrasting effects of DTYMK on adaptive and innate immune cells infiltration in different cancers.

The correlation between DTYMK and the major histocompatibility complex (MHC), immune inhibitors, immune stimulators, chemokines, and receptors expressed in the tumor microenvironment (TME) were also inspected across 30 cancer types (Figures 6B–F; Supplementary File S1). Interestingly, DTYMK positively impacted MHC class II molecules (HLA-DP, HLA-DM, HLA-DO, HLA-DQ, HLA-DR) expression in LGG but negatively impacted MHCII expression in LUAD and STAD (Figure 6B). Poliovirus receptor-related 2 (PVRL2), an immunoinhibitor, binds with the coinhibitory receptor T cell immunoreceptor with Ig and ITIM domains (TIGIT) and negatively regulates lymphocyte activity (Stamm et al., 2018). It has been shown that interleukin-6 (IL-6)/IL-6R signaling is involved in the stimulation of Th17 cell differentiation from naive T cells (Yamashita et al., 2011). Our further analysis uncovered that high expression of DTYMK positively influenced PVRL2 expression but negatively influenced IL-6R expression in LGG, LIHC, LUAD and STAD (Figures 6C,D; Supplementary File S1). These results suggested that DTYMK expression not only affected immune cell infiltration, but also influenced major histocompatibility complex (MHC) expression, immune cell activities through regulating immune inhibitors, chemokines/cytokines and their receptor expression.

To study the molecular role of DTYMK in tumorigenesis we used the GeneMANIA platform to plot a protein-protein interaction (PPI) network for DTYMK. DTYMK strongly interacted with thymidine kinase 1 (TK1) (Figure 7A). To clarify DTYMK functions in tumor cells, the top 50 proteins interacting with DTYMK were obtained from the STRING database (Supplementary File S2) and processed for KEGG and GO enrichment analyses using the DAVID platform (Figures 7B–E; Supplementary File S3–S6). As shown in the figure, “pyrimidine metabolism”, “purine metabolism” and “metabolic pathways” are critical pathways in which DTYMK functions in human cancers (Figure 7B).

Considering that DTYMK expression was upregulated and also correlated with poor prognosis and immune infiltration in LIHC and LUAD, we then focused on the functions of DTYMK in LIHC and LUAD. DTYMK co-expressing genes in LIHC and LUAD were analyzed using the Linkedomics database (Supplementary Files S7, S8). All correlated genes were shown by Volcano plots (Figures 8A, 9A). The top 50 genes that positively and negatively correlated with DTYMK were indicated in the heatmap plot (Figures 8B,C, Figures 9B,C). GSEA analysis of DTYMK-correlating genes was then carried out. Coinciding with the enrichment results we determined using DAVID, the results of KEGG enrichment underlined that positively correlated genes of DTYMK were involved in “cell cycle,” “pyrimidine metabolism,” and “purine metabolism” pathways in LIHC and LUAD (Figures 8D, 9D). What is noteworthy is that negatively correlated genes were enriched in “Th1 and Th2 cell differentiation” or/and “Th17 cell differentiation” in LIHC and LUAD (Figures 8D, 9D). The results emphasized that DTYMK might be implicated in regulating helper T cell differentiation. Additionally, the intersection between DTYMK interactive genes by STRING and DTYMK co-expressing genes showed that TK1 was the only common gene for LIHC (Figure 8E), while TK1, proliferating cell nuclear antigen clamp associated factor (KIAA0101), NME/NM23 nucleoside diphosphate kinase 1 (NME1), mitotic arrest deficient 2 like 1 (MAD2L1) were common genes for LUAD (Figure 9E). The positive correlations between DTYMK and TK1, KIAA0101, NME1, and MAD2L1 were further validated in GEPIA platform (Figures 8F, 9F–I). Using the drug bank database we found that the metabolism of four drugs led to the upregulation of DTYMK while nine drugs led to the downregulation (Table 1). Their effects on DTYMK expression should be noted considering its potential role in human cancers progression.

Given that DTYMK expression varied across human cancers and correlated to cancer stages and immune subtypes as well as immune cell infiltrations, it is conceivable that DTYMK might be a potential prognostic marker. This was first evaluated in human cancers using GEPIA. Higher DTYMK expression represented worse OS in cancers including ACC (p = 0.0088), LGG (p = 6.6e-07), LIHC (p = 1.4e-05), LUAD (p = 0.00057), mesothelioma (MESO) (p = 0.036), PAAD (p = 0.011), skin cutaneous melanoma (SKCM) (p = 0.0012) and uveal melanoma (UVM) (p = 1.7e-05) (Figures 10A,C–I). Intriguingly, DTYMK acted as a protective factor for OS of diffuse large B-cell lymphoma (DLBC) patients (p = 0.0092) (Figure 10B). Furthermore, the relationship between DTYMK expression and patients’ relapse-free survival (RFS) was also explored. Higher expression of DTYMK was correlated with shorter RFS in cancers including ACC (p = 0.00068), BLCA (p = 0.045), KIRP (p = 0.0069), LGG (p = 1.6e-05), LIHC (p = 0.011), LUAD (p = 0.0045), PAAD (p = 0.0086), PRAD (p = 0.0066), SKCM (p = 0.02), and UVM (p = 0.02) (Figures 10J–S).

The relationship between DTYMK expression and patient prognosis was then determined using Kaplan-Meier analysis. Data used in the Kaplan-Meier plotter, including gene expression, RFS and OS information were sourced from GEO, EGA and TCGA. It was found that higher DTYMK expression was correlated with shorter OS in cancers like ESCA (p = 0.0084), KIRC (p = 2.9e-06), KIRP (p = 0.014), LIHC (p = 1.2e-05), LUAD (p = 2.5e-05), and PAAD (p = 0.015) (Figures 11A–E,G). However, increased expression of DTYMK in LUSC indicated a prolonged OS (p = 0.028) (Figure 11F). Upregulated DTYMK expression also reflected worse RFS in LIHC (p = 0.011), LUAD (p = 0.0042), PAAD (p = 0.026), SARC (p = 0.01), and UCEC (p = 0.025) (Figures 11H–L).

The results from the GEPIA and Kaplan-Meier analysis showed that DTYMK expression significantly correlated with survival (OS and RFS) of patients with LIHC and LUAD. To further clarify the prognostic role of DTYMK in LIHC and LUAD, the Cox proportional hazard regression model was established to evaluate prognostic factors. LIHC and LUAD patients were divided into groups with high or low DTYMK expression based on median DTYMK expression. Both the univariate Cox analysis and multivariate Cox analysis results demonstrated that higher DTYMK expression correlated with worse prognosis in LIHC and LUAD (Figures 12A,B,D,E). We then employed receiver operating characteristic (ROC) analysis to demonstrate the diagnostic efficacy of DTYMK in LIHC and LUAD (Figures 12C,F). The area under curve (AUC) value was 0.71, 0.661 and 0.699 for 1, 3, and 5 years OS of LIHC, while the AUC value was 0.627, 0.597, and 0.588 for 1, 3, and 5 years OS of LUAD. The results indicated that DTYMK expression might powerfully predict prognosis in LIHC and LUAD.

PrognoScan database facilitates the investigation of gene association with clinical outcome by encompassing a collection of publicly available cancer microarray datasets. Our results showed that DTYMK expression significantly influenced the prognosis of some cancers (Figure 13; Supplementary File S9). In some cancers like bladder, blood, brain, breast, eye, lung, ovarian, skin and soft tissue cancers, elevated expression of DTYMK resulted in poorer prognosis (Figure 13; Supplementary File S9). On the contrary, DTYMK played a protective role in colorectal cancer (Figure 13; Supplementary File S9).

To validate DTYMK function in tumor cells, we examined DTYMK expression in tumor cell lines of various origins (lung cancer including 95C, 95D, A549, and H1299; pancreatic cancer MIA PaCa-2; liver cancer HepG2) (Figure 14A). Two short hairpin RNAs were designed to downregulate DTYMK expression in HepG2 cells and H1299 cells. Compared with ctrl shRNA, shDTYMK02 but not shDTYMK01 markedly reduced DTYMK expression in both tumor cell lines (Figures 14B,C). shDTYMK02 was used to stably disrupt DTYMK expression in cancer cell lines. Subsequently, knockdown of DTYMK impeded cell migration and clonal formation in HepG2 and H1299 cells (Figures 14D–L). Our results confirmed the role of DTYMK in regulating tumor cell migration and clonal formation.