Identification of a novel CpG methylation signature to predict prognosis in lung squamous cell carcinoma

Abstract

BACKGROUND:

DNA methylation plays a vital role in modulating genomic function and warrants evaluation as a biomarker for the diagnosis and treatment of lung squamous cell carcinoma (LUSC).

OBJECTIVE:

In this study, we aimed to identify effective potential biomarkers for predicting prognosis and drug sensitivity in LUSC.

METHODS:

A univariate Cox proportional hazards regression analysis, a random survival forests-variable hunting (RSFVH) algorithm, and a multivariate Cox regression analysis were adopted to analyze the methylation profile of patients with LUSC included in public databases: The Cancer Genome Atlas (TCGA), and the Gene Expression Omnibus (GEO).

RESULTS:

A methylated region consisting of 3 sites (cg06675147, cg07064331, cg20429172) was selected. Patients were divided into a high-risk group and a low-risk group in the training dataset. High-risk patients had shorter overall survival (OS) (hazard ratio [HR]: 2.72, 95% confidence interval [CI]: 1.82–4.07, $P<$ 0.001) compared with low-risk patients. The accuracy of the prognostic signature was validated in the test and validation cohorts (TCGA, $n=$ 94; GSE56044, $n=$ 23). Gene set variation analysis (GSVA) showed that activity in the cell cycle/mitotic, ERBB, and ERK/MAPK pathways was higher in the high-risk compared with the low-risk group, which may lead to differences in OS.Interestingly, we observed that patients in the high-risk group were more sensitive to gemcitabine and docetaxel than the low-risk group, which is consistent with results of the GSVA.

CONCLUSION:

We report novel methylation sites that could be used as powerful tools for predicting risk factors for poorer survival in patients with LUSC.

Keywords

Lung squamous cell carcinoma methylated sites overall survival prognosis signature

1. Introduction

Lung cancer is one of the most frequently diagnosed malignant tumors and a leading cause of cancer related mortality worldwide [1]. In most cases, lung cancer is diagnosed at an advanced stage, and the 5-year survival rate remains below 20% [2]. According to histological types, lung cancers are classified into small cell lung cancer (SCLC), or non-small cell lung cancer (NSCLC). The latter includes lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC). NSCLC represents 87% of all lung cancers, and approximately one-third of NSCLC are reported as LUSC [3]. Data from several recent studies confirm that the current treatment strategies (e.g. epidermal growth factor receptor tyrosine kinase inhibitors and immune checkpoint inhibitors) are effective in improving overall survival (OS) rates in patients with LUAD [4, 5, 6, 7]. In contrast, the lack of early diagnostic biomarkers and of effective treatment protocols has resulted in sustained poor OS rates in patients with LUSC [8]. Therefore, it is essential to identify prognostic biomarkers and therapeutic targets to improve OS rates in patients with LUSC.

Aberrant DNA methylation may result in silencing of tumor-suppressor genes and can potentially lead to the development of cancer. DNA methylation has been reported as a promising target for the development of robust predictive, prognostic, and diagnostic biomarkers [9]. In addition, a previous study uncovered the differences between DNA methylation and expression patterns in LUAD and LUSC, and evaluated their biological characteristics [10]. The methylation profiles of GALR1, ZMYND10, NTSR1, AGTR1, and SLC5A8 may be useful for NSCLC diagnosis [11]. Moreover, the DNA methylation pattern for TP63, TBX5, TCF21, HNF1B, CLDN1, TBX5, and ADHFE1 was reported as a potential diagnostic biomarker for LUSC [12]. VAX1, IRX1, AdCyAP1, and CH25H were also suggested as prognostic biomarkers for LUSC [13]. However, more biomarkers are urgently needed to improve the treatment of LUSC according to patient’s prognosis.

In this study, we identified and validated potential methylation sites as a biomarker to predict LUSC prognosis by utilizing the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases. Identification of the methylation signature of patients’ tumor DNA will allow stratification of patients into a high-risk and a low-risk group. we observed that patients in the high-risk group were more sensitive to gemcitabine and docetaxel than the low-risk group. These results may have important implications for target therapy, which may result in improved OS in patients with LUSC.

2. Materials and methods

2.1 Clinical cohorts, RNA-Sequencing, and DNA methylation

RNA-sequencing and DNA methylation data as well as clinical features were obtained from TCGA database (https://cancergenome.nih.gov/). A total of 521 samples (480 patients vs. 41 adjacent normal tissues) were used to analyze mRNA expression, and 359 samples (324 patients vs. 35 adjacent normal tissues) were used to analyze DNA methylation. An additional 23 patient samples were added from the GEO database (GSE56044, https://www.ncbi.nlm.nih.gov/geo/). Natural log half maximal inhibitory concentration (IC50) for various compounds and all LUSC cell lines were downloaded from the genomics of drug sensitivity in cancer database (GDSC, https://www.cancerrxgene.org/). DNA methylation data were evaluated using the Illumina Human Methylation450 platform (Illumina Inc., California, USA). The Illumina methylated arrays had 482,421 CpG sites in the genome [14], and probe-level data represented a beta value for every gene. mRNA expression data were validated using the Affymetrix Human Genome U219 array. Finally, we obtained 302 samples from the TCGA and GEO databases specific to DNA methylation, and demographic information such as gender and age at all stages were collected. All TCGA methylation datasets were randomly divided into the training cohorts ( $n=$ 187), and the test cohorts ( $n=$ 92). The GEO methylation dataset for 23 patients was validated. The technical details used to select the DNA methylated sites for LUSC prognostic prediction are shown in Fig. 1.

Figure 1.

Schematic showing the selection of a DNA methylation signature to predict prognostic markers. 1. Different screen, fold change $=$ 2. FDR $<$ 0.05; 2. In total, 22 methylated genes include 620 methylated sites; 3. Univariable Cox analysis, $P$ -value $<$ 0.05; 4. Random survival forest variable hunting algorithm; 5. Multivariable Cox analysis, $P$ -value $<$ 0.05.

2.2 Identification of differentially expressed and methylated genes

edgeR software is a bioconductor package for differential expression analysis of gene expression data [15]. edgeR software was used to screen for differentially expressed genes (DEGs) and differentially methylated genes (DMGs) between patients and adjacent normal tissue. $|$ log FC $|$ $\geqslant$ 1, $P<$ 0.05, and false discovery rate (FDR) $<$ 0.05 indicated statistically significant differences.

2.3 Statistical analysis

A machine learning approach was used to build the predictive model as described previously [16]. Firstly, a univariate Cox proportional hazard regression analysis was adopted based on a previously published report to assess the correlation between survival time and the methylation of each gene site in the training cohorts [17]. We then used the random survival forests-variable hunting (RSFVH) algorithm to screen for methylated genes, of which 10 were screened out [18, 19]. Finally, in order to identify the reliable and powerful predictive prognostic methylated sites, multivariate Cox regression analysis was adopted to build a prognostic model. Based on the model, the prognosis risk was accessed according to the following methylation equation, where N is the number of methylated sites of signature, $\textit{Meth}_{i}$ is the methylation of the gene sites, and $\textit{Coef}_{i}$ is the multivariate Cox regression coefficient:

$\displaystyle\text{Risk Score (RS)}=\sum_{i=1}^{N}\textit{Meth}_{i}*\textit{% Coef}_{i}$

Figure 2.

Univariate Cox proportional hazard regression analysis of differentially methylated genes in the training group and multivariate Cox regression analysis of the 65 methylated sites in the training group that were considered statistically significant in both groups ( $P<$ 0.05).

Risk Score (RS) was calculated based on the multi-node weighted sum of risk scores and the coefficient of the signature for each patient as described previously [20]. The median RS was calculated as the cut-off value enabling division of the training cohorts, the test cohorts, and the validation cohorts into a high-risk and low-risk group, respectively. Kaplan-Meier survival analysis was used to compare the prognoses of the two groups using the log-rank test. Receiver operating characteristic (ROC) curve analysis was plotted to confirm the ability of the model to predict the 5-year survival. Multivariable Cox regression analysis was used to analyze the independent survival prediction of the methylation signatures. Student’s t-test was used to screen for the differential expression and methylation genes between tumor and adjacent tissues. All statistical analyses were conducted using R 3.5.1 (www.r-project.org). pROC, survival, and random ForestSRC were all downloaded from Bioconductor (https://bioconductor.org).

2.4 Generation of the nomogram

To develop a more sensitive prediction tool for clinical practice, we generated a nomogram integrating the methylation signature and two different clinical risk factors (age and stage) to predict 1-year, 3-year, and 5-year survival. The nomogram score for each variable based on the point scale was calculated. After calculating the total nomogram score, we obtained the predicted probability of 1-year, 3-year, and 5-year survival for each patient, as described previously [21].

2.5 Gene set variation analysis (GSVA) of different pathways between the high and low-risk groups

GSVA was designed to assess the concerted behavior group of functionally related genes, between the high-risk group and the low-risk group. The GSVA package implemented a non-parametric unsupervised method, and a $p$ value of $<$ 0.05 was considered statistically significant (Tables S1 and S2).

2.6 Analysis of drug sensitivity in the high-and low-risk group, using GDSC

The GDCS database contains pharmacogenomics data including drug screening results for thousands of cancer cell lines and over 200 anti-cancer drugs, genomic variants and methylation data. Inhibitory concentration 50% (IC50) is the concentration of the drug that kills half of tumor cells in vitro [22, 23]. Data are available for more than 1000 cell lines including 15 LUSC cell lines. Chemotherapy is recommended as the first-line treatment for patients with advanced or metastatic NSCLC [24, 25]. Using the training model, methylation data from LUSC cells were processed, and drug sensitivity was calculated for both the high-risk and low-risk groups. A scatter plot was developed to show the IC50 of commonly used chemotherapy drugs for both the high-risk and low-risk groups.

3. Results

Table 1
Summary of patient demographics and characteristics

Characteristic	Training ( $N=$ 187)	Test ( $N=$ 92)	Validation ( $N=$ 23)
Gender
Female	117 (62.5%)	26 (41.9%)	9 (39.1%)
Male	70 (37.5%)	66 (58.1%)	14 (60.9%)
Age
$<$ 67 years	80 (42.8%)	41 (44.6%)	4 (17.3%)
$\geqslant$ 67 years	107 (57.2%)	52 (55.4%)	19 (82.7%)
Stage
I	87 (46.5%)	41 (44.6%)	22 (95.7%)
II	66 (35.3%)	34 (36.9%)	1 (4.3%)
III	33 (17.6%)	14 (18.5%)	0 (0.0%)
IV	1 (0.6%)	0 (0.0%)	0 (0.0%)
Vital status
Living	117 (62.6%)	64 (69.6%)	13 (56.5%)
Dead	72 (37.4%)	28 (30.4%)	10 (43.5%)

Figure 3.

Methylated site signature predicts survival rate of patients with LUSC. (A–B) Kaplan-Meier survival curves divided LUSC patients into high-risk and low-risk groups using the methylated gene site signature in the training and test groups. Log-rank test was used to calculated $P$ -values. (C–D) Area under the curve (AUC) of receiver operating characteristic (ROC) curves used to assess the sensitivity and specificity of the prognostic model.

3.1 Identification of DEGs and DMGs

To evaluate gene expression and methylation patterns of LUSC, significant differential analysis was performed in LUSC and compared to adjacent normal samples. We identified 7,332 DEGs (4583 downregulated, and 2749 upregulated, Table S3) and 389 DMGs (20 downregulated, and 369 upregulated, Table S4) ( $|$ Fold change $|$ $\geqslant$ 2, $P<$ 0.05, FDR $<$ 0.05, Fig. S1). We also identified 134 DEGs both differently expressed and methylated between LUSC and adjacent normal tissues (Table S5). Among these 134 genes, we found 22 differentially methylated genes that showed an opposite expression pattern between the two groups (Table S6). Overall, we screened 620 methylated gene sites from these 22 genes (Table S7), which were then used for further prognosis analysis to explore the methylation signature in patients with LUSC.

Figure 4.

Methylated site signature was validated in the GSE56044 dataset. (A) Methylated site signature as a prognostic biomarker was validated in the GSE56044 dataset, which was also used to divide LUSC patients into a high-risk and a low-risk group. (B) Area under the curve (AUC) of receiver operating characteristic (ROC) curve was used to determine the sensitivity and specificity of the prognostic model.

Figure 5.

The signature of independent prognostic factors and nomogram development. (A) Combined clinical information was used for multivariate Cox regression analysis to study OS in LUSC patients in the training group (B) The nomogram combined the methylated gene site signature and clinical-related variables to predict patients’ OS.

3.2 Identification of methylated gene signature in the training cohorts

All 279 patients were randomly divided into two groups (training group, $n=$ 187; validation group, $n=$ 94). The training group including the complete clinical data was used to further explore the association of survival with 475 methylated sites. Survival time and survival status were included as dependent variables in the univariate Cox proportional hazard regression analysis of the 475 methylated sites. Sixty-four methylation sites were found significantly correlated with the OS ( $P<$ 0.05; Table S8). Next, the random forest algorithm was used to identify the methylated gene sites signature from the 64 sites. Ten sites were picked up based on the strong relation to patient survival based on the permutation feature importance (PFI) score defined by the RSFVH algorithm (Fig. S2). Next, in order to obtain the most powerful predictive prognostic methylated sites, a multivariate Cox regression analysis was developed for the three identified methylated gene sites (cg06675147, cg07064331, and cg20429172) (Fig. 2 and Table S9). The risk score (Table S10), combining the three methylation sites, cg06675147, cg07064331, and cg20429172 was determined as follows:

$\displaystyle\text{RS}=\left(5.13\times\textit{meth}_{\text{cg06675147}}\right% )+\left(-89.62\times\textit{meth}_{\textit{cg07064331}}\right)+\left(-2.24% \times\textit{meth}_{\textit{cg20429172}}\right)$

3.3 Confirmation of the survival power of the methylation signature in the training and test datasets

The methylation signature was used to calculate each patient’s RS. The median RS was considered as the cut-off point to classify the training cohorts into the low-risk group ( $n=$ 94) and high-risk group ( $n=$ 93). Using Kaplan-Meier survival analysis, the survival rates were calculated. We found that survival rates in patients with low-risk group were more than 60% at 5 years compared with less than 30% in patients with high-risk (hazard ratio [HR]: 2.72, 95% confidence interval [CI]: 1.82–4.07, $P<$ 0.001, Fig. 3A). Using the established prognosis model, the test cohorts were divided into a low-risk group (51 patients, 54.3%) and a high-risk group (43 patients, 45.7%). The 5-year OS was less than 40% in high-risk patients and 70% in low-risk patients of the test group (HR: 1.77, 95% CI: 1.03–3.46; $P=$ 0.036, Fig. 3B). We used ROC curve analysis to assess the accuracy of our prognostic model. The larger the area under the ROC curve, the better the ability of the model is to predict the prognosis of patients with LUSC. In the training dataset, the predictive precision of the prognostic signature was reliable (AUC ${}_{\text{Signature}}=$ 0.734, Fig. 3C). We obtained similar outcomes in the test cohorts (AUC ${}_{\text{Signature}}=$ 0.704, Fig. 3D). Our results suggest that the methylated site signature may be a powerful prognostic biomarker.

3.4 Validation of the robust signature in the independent validation cohorts

To confirm that the identified three-methylated gene site-based classifier had equal prognostic value in different patients, the signature of the GSE56044 data series was evaluated in relation to OS in patients with LUSC (Table S11). Using the established coefficient of the module, the training cohort was divided into a low-risk group (11 patients, 47.8%) and a high-risk group (12 patients, 52.2%). The corresponding 5-year OS was less than 40% in the high-risk group and 80% in the low-risk group (Fig. 4A). In the validation dataset, the predictive ability of the signature was similarly reliable (area under the curve [AUC] ${}_{\text{Signature}}=$ 0.742, Fig. 4B), which demonstrated that the methylation signature in our study was a robust prognosis biomarker.

Figure 6.

GSVA analysis of differentially-activated pathways between the high- and low-risk groups.

3.5 Independent prognostic predictors and development of a nomogram to predict OS

The relationship between the signature RS and the clinicopathological characteristics (e.g., age and stage) was analyzed by multivariate Cox regression analysis. The relationship demonstrated that the signature was an independent prognostic predictor of OS in the training dataset (high-risk group vs. low-risk group, HR $=$ 2.88, 95% CI: 1.88–4.20, $P<$ 0.001, Fig. 5A). To develop a more sensitive predictive tool for clinical practice, we generated a nomogram integrating the methylation signature and two clinical risk factors (age and stage) to predict the rate of 1-year, 3-year, and 5-year survival. The model can be used to obtain a nomogram score for each variable based on the point scale. After calculating the total nomogram score, we obtained the predicted probability of survival at 1-year, 3-year, and 5-year for each patient. The nomogram showed that the methylation signature is the largest contributor to 1-year, 3-year, and 5-year OS, followed by age and stage (Fig. 5B).

3.6 Evaluation of differential pathway activity between the high-risk and low-risk group

GSVA was used to screen for DEGs in different pathways between the high- and low-risk groups in the training dataset. We identified three pathways of interest using the KEGG, BIOCARTA, and REACTOME datasets (non-parametric, $p$ value $<$ 0.05), which were cell cycle/mitosis, ERBB, and ERK/MAPK (Fig. 6).

3.7 Identification of signature methylation sites in LUSC cell lines that are resistant or sensitive to drugs

High-risk and low-risk groups were identified using the methylation data of gemcitabine and docetaxel-treated LUSC cells based on the training model in the GDSC database. A scatter plot presents the IC50 of common chemotherapy drugs in the high-risk and low-risk groups (Fig. 7). Results show that patients in the high-risk group were more sensitive to gemcitabine and docetaxel than the low-risk group.

Figure 7.

Signature methylation sites in LUSC cell lines resistant and sensitive to chemotherapeutic drugs.

4. Discussion

Lung cancer is one of the most aggressive tumors and the leading cause of cancer death worldwide. Until the last decade, metastatic NSCLC was associated with a $<$ 5% 5-year OS rate. The presence of specific biomarkers dictates response to different treatments in NSCLC [26]. Since the rate of molecular alterations (e.g., EGFR mutations, ROS1 rearrangements, BRAF V600E mutations, and ALK rearrangements) that predict response to therapy are low (approximately 30%) [27], there is an urgent need for more biomarker-directed therapies, that could improve patient OS. A number of studies on DNA methylated gene signatures reported prognostic biomarkers for patients with NSCLC [28, 29, 30, 31]. However, more precise methylated gene signatures and stable modules to predict prognosis are needed to implement personalized treatment strategies for patients with LUSC.

We evaluated 475 methylated gene sites from 22 differentially methylated genes which had opposite differential expression patterns. By applying different statistical approaches and a machine-learning method, we identified a three-gene site methylation signature. Furthermore, we validated the methylated gene signature in the TCGA, which was independently associated with OS in patients with LUSC. Moreover, we established a methylated gene site nomogram to predict OS. We found that the cell cycle/mitosis, ERBB, and ERK/MAPK pathways were different between the high-risk and low-risk groups in the training dataset, which lead to variations in OS. Finally, high-risk and low-risk groups were defined from the methylation data for gemcitabine and docetaxel-treated LUSC cells based on the training module in the GDSC database. It is well established that gemcitabine and docetaxel disrupt DNA synthesis and as such are commonly used as chemotherapy drugs [32, 33, 34]. We found that patients in the high-risk group were more sensitive to gemcitabine and docetaxel than the low-risk group. This observation is consistent with previous results of the GSVA, which showed that the cell cycle was different between the high-risk and low-risk group in the training dataset. These results may have important implications for precision therapy, which may result in improved OS in patients with LUSC.

We determined a set of three methylated gene sites consisting of TRIM58, FOXF1, and C16orf86 that predicts OS in three patient cohorts. The TRIM58 promoter was previously shown to be methylated and frequently silenced in early-stage lung adenocarcinoma [35]. As a novel epigenetic signature, TRIM58 has an important role in the early diagnosis of lung cancer [36]. In addition, it was found that TRIM58 methylation was associated with eight prognostic genes in LUSC [37]. It was also previously reported that FOXF1 is hypermethylated and downregulated in chronic obstructive pulmonary disease [38]. It was shown to be an important modulating factor in the lung, influencing important biological events such as maintenance of endothelial barrier function and prevention of edema after lung injury [39], promotion of lung regeneration after partial pneumonectomy [40], and mediation of mesenchymal stem cell fusion-induced reprogramming [41]. C16orf86 is an epigenetically-regulated factor which is conferred by genetic variation located at independent loci [42]. Little is known about the role of C16orf86 in tumor progression.

Our work has several limitations. We obtained a low number of normal adjacent tumor tissue compared to available pathological tissue. We did not investigate for the specific mechanism accounting for methylation of the sites involved in LUSC. We lack prospective data to evaluate the methylated sites signature. Despite these limitations, the significant and consistent correlation of the identified methylated sites signature with the OS found in three independent groups highlights the potentially powerful and significant prognostic signature of these sites in LUSC.

Taken together, our results show that we successfully identified a prognostic signature which may help predict the OS in patients with LUSC. Importantly, we developed a 3-methylated gene site nomogram to predict patients’ prognosis. In keeping with previous results, we found that high-risk patients were more sensitive to gemcitabine and docetaxel than low-risk patients. Our observations suggest that activity in relation to the cell cycle/mitosis may be different between the high-risk and low-risk groups. Our data also suggest that this signature could serve as a precise and robust prognostic biomarker and as a potential therapeutic target in patients with LUSC.

Funding

The work is supported by the National Natural Science Foundation of China (81672267, 81871859) and the Natural Science Foundation of Guangdong Province (2016A030313632).

Supplementary data

The supplementary files are available to download from https://dx-doi-org.web.bisu.edu.cn/10.3233/CBM-201564.

sj-docx-1-cbm-10.3233_CBM-201564.docx - Supplemental material

Supplemental material, sj-docx-1-cbm-10.3233_CBM-201564.docx

sj-xls-1-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-1-cbm-10.3233_CBM-201564.xls

sj-xls-2-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-2-cbm-10.3233_CBM-201564.xls

sj-xls-3-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-3-cbm-10.3233_CBM-201564.xls

sj-xls-4-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-4-cbm-10.3233_CBM-201564.xls

sj-xls-5-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-5-cbm-10.3233_CBM-201564.xls

sj-xls-6-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-6-cbm-10.3233_CBM-201564.xls

sj-xls-7-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-7-cbm-10.3233_CBM-201564.xls

sj-xls-8-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-8-cbm-10.3233_CBM-201564.xls

sj-xls-9-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-9-cbm-10.3233_CBM-201564.xls

sj-xls-10-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-10-cbm-10.3233_CBM-201564.xls

sj-xls-11-cbm-10.3233_CBM-201564.xls - Supplemental material

Supplemental material, sj-xls-11-cbm-10.3233_CBM-201564.xls

Footnotes

Conflict of interest

The authors declare that they have no conflict of interest.

References

Jemal

Bray

Center

M.M.

et al., Global cancer statistics, CA Cancer J Clin 61(2) (2011), 69–90.

Kemp-Jacobsen

Johansen

J.S.

Mellemgaard

et al., AHRR (cg05575921) methylation extent of leukocyte DNA and lung cancer survival, PLoS One 14(2) (2019), e0211745.

Martinez

Behrens

Agorreta

et al., 5 protein-based signature for resectable lung squamous cell carcinoma improves the prognostic performance of the TNM staging, Thorax 74(4) (2019), 371–379.

Uchida

Seki

Mizuno

et al., Involvement of dual-strand of the miR-144 duplex and their targets in the pathogenesis of lung squamous cell carcinoma, Cancer Sci 110(1) (2019), 420–432.

Soria

J.C.

Ohe

Vansteenkiste

et al., Osimertinib in Untreated EGFR-Mutated Advanced Non-Small-Cell Lung Cancer, N Engl J Med 378(2) (2018), 113–125.

Peters

Camidge

D.R.

Shaw

A.T.

et al., Investigators AT: Alectinib versus Crizotinib in Untreated ALK-Positive Non-Small-Cell Lung Cancer, N Engl J Med 377(9) (2017), 829–838.

Mok

T.S.K.

W.L.

Kudaba

et al., Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial, Lancet 393(10183) (2019), 1819–1830.

Paz-Ares

Tan

E.H.

O’Byrne

et al., Afatinib versus gefitinib in patients with EGFR mutation-positive advanced non-small-cell lung cancer: overall survival data from the phase IIb LUX-Lung 7 trial, Ann Oncol 28(2) (2017), 270–277.

Koch

Joosten

S.C.

Feng

et al., Analysis of DNA methylation in cancer: location revisited, Nat Rev Clin Oncol 15(7) (2018), 459–466.

10.

Yang

Wang

and Liu

, Exploring and comparing of the gene expression and methylation differences between lung adenocarcinoma and squamous cell carcinoma, J Cell Physiol 234(4) (2019), 4454–4459.

11.

Guo

Yan

et al., Identification and validation of the methylation biomarkers of non-small cell lung cancer (NSCLC), Clin Epigenetics 7; (2015), 3.

12.

Shi

Y.X.

Wang

et al., Genome-wide DNA methylation profiling reveals novel epigenetic signatures in squamous cell lung cancer, BMC Genomics 18(1) (2017), 901.

13.

Gao

Zhuang

Zhou

et al., Prognostic value of aberrantly expressed methylation gene profiles in lung squamous cell carcinoma: A study based on The Cancer Genome Atlas, J Cell Physiol 234(5) (2019), 6519–6528.

14.

Bibikova

Barnes

Tsan

et al., High density DNA methylation array with single CpG site resolution, Genomics 98(4) (2011), 288–295.

15.

Robinson

M.D.

McCarthy

D.J.

and Smyth

G.K.

, edgeR: a Bioconductor package for differential expression analysis of digital gene expression data, Bioinformatics 26(1) (2010), 139–140.

16.

Yin

Zhang

et al., Identification of DNA methylation signature to predict prognosis in gastric adenocarcinoma, J Cell Biochem (2019), 1–8.

17.

Guo

J.C.

Chen

et al., Protein-coding genes combined with long noncoding RNA as a novel transcriptome molecular staging model to predict the survival of patients with esophageal squamous cell carcinoma, Cancer Commun (Lond) 38; (2018), 4.

18.

Meng

Jin-Cheng

Jue

et al., Protein-coding genes, long non-coding RNAs combined with microRNAs as a novel clinical multi-dimension transcriptome signature to predict prognosis in ovarian cancer, Oncotarget 8(42) (2017), 72847–72859.

19.

Guo

J.C.

C.Q.

Wang

Q.Y.

et al., Protein-coding genes combined with long non-coding RNAs predict prognosis in esophageal squamous cell carcinoma patients as a novel clinical multi-dimensional signature, Mol Biosyst 12(11) (2016), 3467–3477.

20.

Fan

and Liu

, Discovery of a novel six-long non-coding RNA signature predicting survival of colorectal cancer patients, J Cell Biochem 119(4) (2018), 3574–3585.

21.

Tian

Zhu

Yan

et al., Differentially expressed lncrnas in gastric cancer patients: a potential biomarker for gastric cancer prognosis, J Cancer 8(13) (2017), 2575–2586.

22.

Qiu

Lin

et al., A novel mutation panel for predicting etoposide resistance in small-cell lung cancer, Drug Des Devel Ther 13; (2019), 2021–2041.

23.

Szalai

Subramanian

Holland

C.H.

et al., Signatures of cell death and proliferation in perturbation transcriptomics data-from confounding factor to effective prediction, Nucleic Acids Res 47(19) (2019), 10010–10026.

24.

Passiglia

Bironzo

and Scagliotti

G.V.

, First-line immune-chemotherapy combination: the right strategy to fight squamous non-small cell lung cancer? Transl Lung Cancer Res 8(4) (2019), 546–549.

25.

Ramalingam

and Belani

, Systemic chemotherapy for advanced non-small cell lung cancer: recent advances and future directions, Oncologist 13(Suppl) (2008), 5–13.

26.

Arbour

K.C.

and Riely

G.J.

, Systemic Therapy for Locally Advanced and Metastatic Non-Small Cell Lung Cancer: A Review, JAMA 322(8) (2019), 764–774.

27.

Skoulidis

and Heymach

J.V.

, Co-occurring genomic alterations in non-small-cell lung cancer biology and therapy, Nat Rev Cancer 19(9) (2019), 495–509.

28.

Sandoval

Mendez-Gonzalez

Nadal

et al., A prognostic DNA methylation signature for stage I non-small-cell lung cancer, J Clin Oncol 31(32) (2013), 4140–4147.

29.

Hao

Luo

Krawczyk

et al., DNA methylation markers for diagnosis and prognosis of common cancers, Proc Natl Acad Sci U S A 114(28) (2017), 7414–7419.

30.

Wang

Duan

Meng

Q.H.

et al., Integrated analysis of DNA methylation profiling and gene expression profiling identifies novel markers in lung cancer in Xuanwei, China, PLoS One 13(10) (2018), e0203155.

31.

Wang

Deng

Xin

et al., Prognostic and predictive value of three DNA methylation signatures in lung adenocarcinoma, Front Genet 10(2019), 349.

32.

Lai

C.J.

Lin

C.Y.

Liao

W.Y.

et al., Promotes Migration and Invasion of Docetaxel-Resistant Prostate Cancer Cells Likely via Induction of Hippo-Yap Signaling, Cells 8(4) (2019).

33.

Yardley

D.A.

Brufsky

Coleman

R.E.

et al., Phase II/III weekly nab-paclitaxel plus gemcitabine or carboplatin versus gemcitabine/carboplatin as first-line treatment of patients with metastatic triple-negative breast cancer (the tnAcity study): study protocol for a randomized controlled trial, Trials 16; (2015), 575.

34.

Kurimot

Kimura

Usami

et al., Comparing the cost-effectiveness of FOLFIRINOX, nab-paclitaxel plus gemcitabine, gemcitabine and S-1 for the treatment of metastatic pancreatic cancer, Mol Clin Oncol 7(1) (2017), 125–130.

35.

Kajiura

Masuda

Naruto

et al., Frequent silencing of the candidate tumor suppressor TRIM58 by promoter methylation in early-stage lung adenocarcinoma, Oncotarget 8(2) (2017), 2890–2905.

36.

Diaz-Lagares

Mendez-Gonzalez

Hervas

et al., A Novel Epigenetic Signature for Early Diagnosis in Lung Cancer, Clin Cancer Res 22(13) (2016), 3361–3371.

37.

Zhang

Cui

et al., TRIM58/cg26157385 methylation is associated with eight prognostic genes in lung squamous cell carcinoma, Oncol Rep 40(1) (2018), 206–216.

38.

Yoo

Takikawa

Geraghty

et al., Integrative analysis of DNA methylation and gene expression data identifies EPAS1 as a key regulator of COPD, PLoS Genet 11(1) (2015), e1004898.

39.

Cai

Bolte

et al., FOXF1 maintains endothelial barrier function and prevents edema after lung injury, Sci Signal 9(424) (2016), ra40.

40.

Bolte

Flood

H.M.

Ren

et al., FOXF1 transcription factor promotes lung regeneration after partial pneumonectomy, Sci Rep 7(1) (2017), 10690.

41.

Wei

H.J.

Nickoloff

J.A.

Chen

W.H.

et al., FOXF1 mediates mesenchymal stem cell fusion-induced reprogramming of lung cancer cells, Oncotarget 5(19) (2014), 9514–9529.

42.

Lemire

Zaidi

SH..

Ban

et al., Long-range epigenetic regulation is conferred by genetic variation located at thousands of independent loci, Nat Commun 6; (2015), 6326.

Identification of a novel CpG methylation signature to predict prognosis in lung squamous cell carcinoma

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSION:

Keywords

1. Introduction

2. Materials and methods

2.1 Clinical cohorts, RNA-Sequencing, and DNA methylation

2.3 Statistical analysis

2.5 Gene set variation analysis (GSVA) of different pathways between the high and low-risk groups

2.6 Analysis of drug sensitivity in the high-and low-risk group, using GDSC

3. Results

Table 1 Summary of patient demographics and characteristics

3.3 Confirmation of the survival power of the methylation signature in the training and test datasets

3.4 Validation of the robust signature in the independent validation cohorts

3.6 Evaluation of differential pathway activity between the high-risk and low-risk group

3.7 Identification of signature methylation sites in LUSC cell lines that are resistant or sensitive to drugs

Funding

Supplementary data

sj-docx-1-cbm-10.3233_CBM-201564.docx - Supplemental material

sj-xls-1-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-2-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-3-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-4-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-5-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-6-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-7-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-8-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-9-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-10-cbm-10.3233_CBM-201564.xls - Supplemental material

sj-xls-11-cbm-10.3233_CBM-201564.xls - Supplemental material

Footnotes

Conflict of interest

References

Table 1
Summary of patient demographics and characteristics