Identification of CDT1 as a prognostic marker in human lung adenocarcinoma using bioinformatics approaches

Data acquisition

High-throughput RNA sequencing data and clinical information regarding lung cancer patients were obtained using the UCSC Xena browser (version: 2019-07-20, http://xenabrowser.net/datapages/). For external validation, transcriptome profiling data from patients with LUAD from the Gene Expression Omnibus datasets GSE118370, GSE140797 and GSE31210 were used.

R software was used to examine the levels of CDT1 expression in LUAD and other malignancies (version 4.0.4). R was used to perform data analyses in this study (version 4.0.4). All the raw codes used in the manuscript could be found at https://github.com/YeweiJia/CDT1-Jiang.

LinkedOmics analysis

The LinkedOmics database (http://www.linkedomics.org/login.php) contains multi-omics data and clinical information from the TCGA portal for 32 cancer types (Vasaikar et al., 2018). LinkedOmics is an online platform that integrates multi-omics data from TCGA to enable comprehensive analysis. We used the linkFinder module to evaluate DEGs related to CDT1.

Kaplan-Meier survival analysis

The Kaplan-Meier method is a widely used statistical technique for estimating the survival probability or survival function of a given population over time. In this study, we employed the Kaplan-Meier method to analyze the survival data of lung adenocarcinoma (LUAD) patients. The Kaplan-Meier estimator calculates the survival probability at each time point and generates a survival curve. The survival curve displays the proportion of patients surviving at different time intervals. The log-rank test or other appropriate statistical tests were performed to compare the survival curves between different groups. The p-value obtained from the statistical test provides information about the significance of the differences in survival between groups.

Identification of differentially expressed genes and building a nomogram

The R “Limma” package was applied to standardize the read counts and perform differential gene analysis with the threshold of LogFC >1.5 and <−1.5. The R‘rms’package and ‘survival’package were used to establish a prognostic nomogram to predict the survival risk.

Nomogram is a statistical tool that integrates multiple prognostic factors to estimate the probability of a specific outcome. We used a nomogram in our study to interpret and utilize the predictive model for assessing risk or prognosis based on specific patient characteristics or biomarkers.

Human tissue specimens

To investigate CDT1 levels in human LUAD, we obtained tumor tissues and paired adjacent non-tumorous tissues during radical resection of patients without prior chemotherapy or radiotherapy at the Department of Cardiothoracic Surgery, Hangzhou Traditional Chinese Medicine Hospital, affiliated with Zhejiang Chinese Medical University, from January 2022 to July 2022. Resected LUAD-adjacent non-tumor samples and matched tumor tissues were obtained and instantly stored in liquid nitrogen (Table 1). This study was approved by the ethics committee of the Hangzhou Traditional Chinese Medicine Hospital, affiliated with Zhejiang Chinese Medical University (Hangzhou, China; No. 2021LH005), and written consent was obtained from all patients. The grades and histological types of all tissue samples were independently verified by two professional pathologists.

RNA extraction and qPCR

Total RNA was extracted from the obtained tissues using TRIrizol reagent (Takara Bio, Shiga, Japan). cDNA was synthesized using the PrimeScript™ RT Master Mix (Takara Bio, Shiga, Japan). SYBR Premix Ex Taq™ II with a PCR detection system (Takara Bio, Shiga, Japan) was used to investigate the expression of the indicated genes. Transcriptional levels were normalized to those of the internal control gene, β-actin. The following primers were used: CDT1 forward, 5′-GGAGGTCAGATTACCAGCTCAC-3′ and reverse, 5′-TTGACGTGCTCCACCAGCTTCT-3′; β-actin forward, 5′-CATCCGCAAAGACCTGTACG-3′ and reverse, 5′-CCTGCTTGCTGATCCACATC-3′. The PCR was run as follows: initial denaturation at 95 °C for 35 s as an initial denaturation step, amplification for 40 cycles with denaturation at 95 °C for 5 s, and annealing at 60 °C for 30 s. The melting curve analysis was performed at the end of the PCR cycle. All reaction information was collected using the Applied Biosystems 7500 Real-Time PCR System (Thermo Scientific, Waltham, MA, USA), and normalized expression was calculated as the relative fold-change, using the formula 2^−∆∆CT.

Hematoxylin and eosin staining

The LUAD tissues and paired adjacent normal tissues were preserved in 4% paraformaldehyde, embedded in paraffin blocks, sliced at a thickness of 4 μm, dewaxed, dehydrated sequentially, and then stained with hematoxylin and eosin (H&E). Ultimately, all slides were independently examined and pathologically analyzed by two pathologists (Department of Pathology, Hangzhou Traditional Chinese Medicine Hospital, affiliated with Zhejiang Chinese Medical University), under an optical microscope (Olympus, Tokyo, Japan).

Immunohistochemistry and scoring analyses

Immunohistochemistry was conducted to determine CDT1 protein expression and prognostic value. LUAD tissues and paired adjacent normal tissues were stained by immunohistochemistry with anti-human CDT1 (1:300; Rabbit; 14382-1-AP; Proteintech, Rosemont, IL, USA), followed by IVision™ Poly-HRP goat anti-mouse/rabbit secondary antibody reagent (DD23-100; Xiamen Talent Biomedical Technology Co., Ltd., Xiamen, China) and DAB treatment. Then, the samples were observed under a microscope and photographed for further analysis. Each LUAD sample was evaluated based on staining intensity and positively stained cell percentage. The H-score was calculated as: (percentage of weak-intensity cells × 1) + (percentage of moderate-intensity cells × 2) + (percentage of strong-intensity cells × 3). Numbers 0, 1, 2, and 3 indicated positive cells. H-score values ranged between 0 and 3. A paired t-test was used to compare CDT1 expression between LUAD tissues and the paired non-tumor tissues.

GO and KEGG enrichment analyses

Gene ontology (GO) (http://www.geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (http://www.genome.jp/kegg/) pathways that are usually used to predict the biological processes as well as signaling pathways, in which some genes could participate, analyses were implemented using the hypergeometric test method, and an adjusted P-value < 0.05 was considered significant by Fisher’s exact test.

Tumor immune infiltration analysis

The R ‘ESTIMATE’, ‘CIBERSORT’ (Chen et al., 2018), and ‘GSVA’ (Hänzelmann, Castelo & Guinney, 2013) packages were used to analyze the immune infiltration of TME.

Statistical analysis

All statistical analyses were performed in R (v4.0.4) and Prism 8.0 (GraphPad Software Inc., San Diego, CA, USA). Kaplan–Meier curve and log-rank test were used to evaluate differences in survival rate between the groups. The student’s t-test or Wilcoxon test was performed for continuous data, and the Chi-square test or Fisher’s exact test was performed for categorical variables. Univariate and multivariate analyses using Cox proportional hazard modeling were performed to estimate the risk of death. Effects were considered significant if P < 0.05.

CDT1 expression was upregulated in LUAD, based on the TCGA and GEO databases

CDT1 was found to be highly elevated in a variety of malignancies when compared to normal tissues, according to a pan-cancer analysis utilizing Wilcoxon rank-sum and Wilcoxon signed-rank tests (Fig. 1A, P < 0.001). In order to show the expression of CDT1 in LUAD more clearly, the TCGA database’s paired tumor and normal nearby samples, as well as the unpaired samples, were also studied using the two different statistical analysis techniques. Results showed that the tumor group had a higher CDT1 expression than the normal group (Fig. 1B). Paired tumor tissues and normal tissues from two GEO datasets (GSE118370 and GSE140797) verified that CDT1 was upregulated in tumor tissues versus normal tissues (Fig. 1C), in addition, another cohort (GSE31210) indicated high expression level of CDT1 in LUAD sample compared to normal samples (Fig. 1C). Then, volcano plots of the three GEO datasets indicated the expression level of CDT1 among differentially expressed genes (DEGs) of the datasets (Fig. 1D).

Together, these results indicated CDT1 was significantly up-regulated in LUAD.

Correlation between CDT1 expression and poor prognosis

Based on the TCGA dataset from the UCSC Xena browser (version: 2019-07-20, http://xenabrowser.net/datapages/), the relationship between CDT1 expression and the prognosis of LUAD patients was assessed. The survival rate of patients with LUAD was calculated using R program software. The results showed that no difference in the survival rate between the male and female patients was observed (Fig. 2A). However, based on the cutoff (Fig. 2B), the high CDT1 expression group had considerably higher accumulated risk and worse OS than the low CDT1 expression group (log-rank P < 0.001; Figs. 2C and 2D). In addition, we also used GEO cohort to verify the survival rate, the data demonstrated that CDT1 high-expression group had a poorer survival rate compared to low-expression group (Fig. 2E).

In addition, in order to incorporate these prognostic factors, we constructed a nomogram that was used to predict or assess the risk or prognosis (Hartaigh et al., 2018) based on all the independent prognostic indicators for OS (Fig. 2F). Nomogram is a graphical tool used for predicting a specific outcome or event based on multiple variables. The nomogram visually represents a mathematical model that combines the effects of different predictor variables to estimate a particular outcome. Each predictor variable is assigned a numerical value, and by drawing a vertical line from each variable’s value to the points axis, the corresponding points for each variable are determined. Based on our results, we provided several predictors, including disease stages, age, and different expression levels of CDT1, and so on to establish a nomogram, a high total score predicted a high survival risk, whereas a low total score showed the opposite. Our data indicated in either in TCGA cohort or GSE31210, CDT1 high-expression group was deeply associated with poor survival rate. These results indicate that LUAD with increased CDT1 expression is associated with poor OS.

Taken together, these data revealed high expression level of CDT1 was positively associated with a poor survival rate.

Clinical samples to verify the expression level of CDT1 in LUAD and paired adjacent normal samples

Quantitative real-time PCR (qPCR) was first performed to evaluate the expression levels of CDT1 in LUAD and paired adjacent normal samples. Sixteen patient samples were used in the experiment. Patient information is shown in Table 1. The data showed that CDT1 was more highly expressed in the tumor group than in the normal group (Fig. 3A). H&E staining results are shown in Fig. 3B. Immunohistochemistry was then performed to explore the expression levels of CDT1 in LUAD and paired normal samples. Immunohistochemistry results indicated that the cytoplasmic staining level of CDT1 in tumor tissues is higher than that in adjacent normal tissues (Fig. 3C).

Our clinical data, including the qPCR and IHC, verified the high expression level of CDT1 in LUAD compared to paired adjacent normal tissue.

CDT1-related functional enrichment analysis

We first divided CDT1 into two groups according to the distribution, the R ‘limma’ package was used to analyze the Differentially Expressed Genes (DEGs). A total number of 466 DEGs were identified, including 417 upregulated genes and 49 down-regulated genes (Fig. 4A). GO and KEGG pathway enrichment analyses were then performed to analyze the functional enrichment. The GO analyses (Fig. 4B) indicated that the upregulated DEGs were mainly enriched in organelle fission, DNA replication, extracellular matrix, ATPase activity, and so on. The KEGG analysis (Fig. 4B) indicated the DEGs were primarily enriched in the Cell cycle, Carbon metabolism, DBA replication, p53 signaling pathway, and so on.

The LinkFinder module of the LinkedOmics database was used to explore the CDT1 co-expression network in LUAD. From this database, 3,182 genes were positively associated with CDT1. The Heatmap of the top 50 genes positively and negatively correlated with CDT1 are shown in Fig. 4C. The GO analysis indicated that CDT1 and the genes positively correlated with CDT1 were mainly involved in biological processes, including DNA replication, and cell cycle (Fig. 4D). The KEGG analysis indicated that CDT1 and the genes positively correlated with CDT1 were mainly involved in the cell cycle, DNA replication, and p53 signaling pathway (Fig. 4D).

It was obvious that the two analyses both aimed at DNA replication, and cell cycle, which were highly associated with LUAD progression.

Correlation analysis between CDT1 expression and immune infiltrates

In order to analyze the role of CDT1 expression in the tumor microenvironment (TME), the R ‘GSVA’ package and ‘ESTIMATE’ were used to analyze the relation among CDT1 expression, ESTIMATEScore, and immune infiltrates status. The landscape indicated that high expression of CDT1 was associated with higher tumor purity, lower ImmunScore, lower ESTIMATEScore as well as higher pro-tumor related immune cells infiltration (Fig. 5A).

In order to further confirm the relationship between CDT1 expression and immune cell infiltration, the R ‘CIBERSORT’ package was used to analyze the TME cell composition. The data (Fig. 6) demonstrated that high expression of CDT1 was significantly correlated with lower mast cells, and CD4 T cells expression. Earlier studies have shown that CD4 T cells promote pulmonary metastasis in mammary carcinomas by augmenting the protumor characteristics of macrophages (DeNardo et al., 2009). Mast cells can exert antitumor effects by stimulating immune responses, recruiting and activating other immune cells, and inducing tumor cell death (Oldford & Marshall, 2015). These findings suggest that the CDT1 high group is characterized by reduced mast cell expression and elevated CD4 T cell expression, indicative of a pro-tumor environment.

These data identified the positive relationship between a high expression level of CDT1 and a pro-tumor environment.

CDT1-related immune genes predicted poor OS in LUAD patients

We further investigated the differentially expressed genes (DEGs) related to CDT1 in the immune response of lung adenocarcinoma (LUAD). The volcano plot (Fig. 7A) initially revealed the DEGs between the high and low ESTIMATEScore groups. Subsequently, a Venn plot (Fig. 7B) was employed to analyze the interaction between CDT1 and the immune response. From this analysis, five genes (Table 2) were selected for subsequent survival analysis. The survival analysis results (Fig. 7C) demonstrated that elevated expression levels of EPYC, EREC, GPR87, and OR2I1P were associated with poor OS in LUAD patients.

These findings indicate a correlation between the expression of CDT1-related immune genes and poor survival outcomes in patients with LUAD.