Genome-wide analysis of prognostic-related lncRNAs, miRNAs and mRNAs forming a competing endogenous RNA network in lung squamous cell carcinoma

Abstract

Purpose

As a type of cancer with the highest morbidity and mortality, lung squamous cell carcinoma (LUSC) has a very poor prognosis. Long-non-coding RNA (lncRNA) has recently attracted attentions because it can play the role of competing endogenous RNA (ceRNA) to inhibit microRNA (miRNA) functions. In this study, we aimed to find prognosis-related lncRNAs, miRNAs and mRNAs and construct a prognosis-related ceRNA network.

Methods

The original LUSC RNA-sequencing data and miRNA profiles data were downloaded from the cancer genome atlas (TCGA) database. Differentially expressed lncRNAs, miRNAs and mRNAs were then identified between patients with lymph node metastasis and no lymph node metastasis. Univariate Cox regression analysis was performed to find the survival-associated lncRNAs, miRNAs and mRNAs. Subsequently, prognostic-related ceRNA network was established. By multivariate Cox regression analysis, three lncRNA signatures and three mRNA signatures were developed and used for predicting LUSC patients’ survival.

Results

A total of 224 lncRNAs, 160 miRNAs, 913 mRNAs were identified between samples with lymph node metastasis and no lymph node metastasis. Univariate Cox regression analysis showed that, among them, 28 lncRNAs, 8 miRNAs, 105 mRNAs were significantly associated with patients’ overall survival time. Further pathway and enrichment analysis suggested that these mRNAs were associated with the regulation of transmembrane transport, regulation of blood circulation, plasma lipoprotein particle organization. Then we constructed a survival-related ceRNA network including 9 lncRNAs, 8 miRNAs and 23 mRNAs. Additionally, a multivariate Cox regression analysis demonstrated that three lncRNAs (AL161431.1, LINC02389, APCDD1L.DT) and three mRNAs (KLK6, SLITRK5, CCDC177) had a significant prognostic value. Risk score indicated that lncRNA signature and mRNA signature could independently predict overall survival in LUSC patients.

Conclusion

The current study provided a better understanding of the ceRNA network in the progression of LUSC and laid a theoretical foundation for LUSC prognosis.

Introduction

Lung cancer is the most prevalent malignancy and the leading cause of cancer-related death in China (Bray et al. 2018). According to the histological classification, it can be divided into two types: small cell lung cancer and non-small cell lung cancer. Non-small cell lung cancer can be divided into large cell lung cancer, lung adenocarcinomas and lung squamous cell carcinoma (LUSC). LUSC, often with poor cure rate, accounts for 30% of non-small cell lung cancers (Choi et al. 2017). Studies have shown that most patients with advanced LUSC often have lymphatic metastasis (Liu et al. 2019). Tumor metastasis is still the main cause of death in LUSC patients (Jiang et al. 2019). Thus, identification of biomarkers that are associated with the lymph node metastasis to predict the LUSC prognosis and clarification of the underlying molecular mechanisms become one of the main topics in lung research.

Long non-coding RNA (lncRNA) refers to a kind of RNA with a length of more than 200 nucleotides (Kopp and Mendell 2018). Initially, lncRNAs were considered as “dark matter” in the genome. But with the development of high-throughput RNA-sequencing, thousands of lncRNAs were identified and have been found to play important roles in the regulation of many diseases, including LUSC (Wong et al. 2018). Recent studies have shown that lncRNAs act as regulatory molecules impacting many cellular processes, such as cancer cell proliferation and metastasis (Kim et al. 2018; Wu et al. 2019b). Growing evidence suggests that lncRNA could regulate gene expression at the transcriptional levels (Wang et al. 2014), post-transcriptional levels (Tay et al. 2011) and epigenetic levels (Galupa and Heard 2015). MicroRNAs (miRNAs) are small non-coding RNAs with a length of 22 nucleotides. Classical theory has shown that miRNA could regulate target mRNA expression by binding to 3′ untranslated regions (3′UTRs) of mRNA (Galupa and Heard 2015). In cancer, the miRNA–mRNA network could modulate cellular proliferation and metastasis (Dong et al. 2019; Xue et al. 2019). Recently, a new theory was proposed that lncRNAs function as competing endogenous RNAs (ceRNA) by adsorbing miRNA to regulate target mRNA expression. They act as molecular sponges for miRNAs by binding to shared miRNA competitively, thereby inhibiting the mRNA degradation by miRNA (Tay et al. 2014; Thomson and Dinger 2016). Increasing evidences indicate that ceRNAs play vital roles in tumor growth and metastasis (Chen et al. 2019a; Yang et al. 2019). Therefore, it makes sense to construct a tumorigenesis and development-related ceRNA network of lncRNA–miRNA–mRNA interactions. To date, complex tumorigenesis-related ceRNA networks about lncRNA–miRNA–mRNA interactions have been explored in many tumors, such as LUSC (Ning et al. 2018). But the LUSC prognosis- related ceRNA network has not been studied.

To clarify these questions, in the present study, the original LUSC RNA-sequencing data and miRNA profile data were downloaded from the TCGA database. Combined with sample clinical parameters, we comprehensively analyzed and identified some prognostic RNAs, including lncRNAs, miRNAs and mRNAs. Based on these RNAs, we constructed a survival-related ceRNA network. Furthermore, we developed a prediction model based on lncRNA and mRNA signature. Through these efforts, we aimed to provide the basis for elucidating the regulatory mechanism of lncRNAs in the prognosis of LUSC and identify valuable biomarkers that can be used to predict prognosis of LUSC patients.

Materials and methods

Data resources and pretreatment

Public TCGA RNA-seq data, mature miRNAs data (the Illumina HiSeq platform) and the corresponding clinical information for LUSC were downloaded using the Genomics Data Commons data transfer tool (https://gdc.cancer.gov/access‑data/gdc‑data‑transfer‑tool.html). Annotation information for RNA-seq data was provided by the Ensemble database (https://asia.ensembl.org/index.html, Release 98). Mature miRNA annotation was provided by the miRBase database (https://www.mirbase.org/, Release 22.1). For RNA-seq data, mRNAs and lncRNAs which expressed more than 0 in more than 75% of samples were retained.

Differential expression analysis

Based on clinical information, there are 176 samples with lymph node metastasis (N1, N2, N3 and N4) and 319 samples without lymph node metastasis (N0) in the RNA-seq data. In mature miRNA data, there are 114 samples with lymph node metastasis and 216 samples without lymph node metastasis. The raw count value was analyzed by the edgeR package of R software to screen differentially expressed lncRNAs, miRNAs and mRNAs between lymph node metastasis and no lymph node metastasis. |log2(Fold change) |> 0.5 and P value < 0.05 were set as the thresholds. Differentially expressed lncRNAs, miRNAs and mRNAs were presented as a volcano plot. After normalization by edgeR, the raw count value was further converted by log2 (normalized value + 1) transformation and used for the next operation.

Univariate Cox regression analysis

According to clinical information, patients who were followed for 90–3650 days were included in the survival study. Univariate Cox regression analysis was used to find the associations between the expression of lncRNA, miRNA and mRNA, with overall survival. RNAs with a P value < 0.05 were considered as candidate prognosis-related biomarkers, and were screened to be used for subsequent ceRNA network construction and multivariate Cox regression analyses. Hazard ratios (HR) of statistically significant survival-associated lncRNAs (top 15), miRNAs, mRNAs (top 15) were described using forest plots.

ceRNA network construction

These candidate survival-related lncRNAs, miRNAs and mRNAs were used to build the ceRNA network. According to miRDB (https://www.mirdb.org/, Version 6.0), TargetScan (https://www.targetscan.org/, Release 7.2) and miRTarBase (https://mirtarbase.mbc.nctu.edu.tw/php/index.php, Release 7.0), miRNA–mRNA relationships were established. The correlations were recorded regardless of which of the three online databases. lncRNA–miRNA relationships were constructed using miRcode 11 (https://www.mircode.org/). Integrating miRNA–mRNA relationships and miRNA–lncRNA relationships, a ceRNA regulatory network consisting of lncRNA–miRNA–mRNA was developed. The ceRNA network was visualized using Cytoscape software (Shannon et al. 2003) (https://cytoscape.org/, Version 3.7.2).

Functional analysis

Pathway and process enrichment analysis of differentially expressed mRNAs and survival-related mRNAs were performed to explain the biological functions associated with lymph node metastasis and survival. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted with the clusterProfiler package of R software. The results were visualized using GOplot package of R software. To analyze the candidate survival-related mRNAs at the functional level, we employed Metascape (https://metascape.org, Release 3.5) online tool to perform pathway and process enrichment analyses. Protein–protein interaction (PPI) network analysis of candidate survival-related mRNAs was evaluated by the STRING database (https://stringdb.org, Version 11.0) as previously described (Ju et al. 2019).

Construction of lncRNA–mRNA expression-based risk score system

RNAs screened by univariate Cox analysis were further analyzed by multivariate cox analysis. RNAs with P < 0.05 were selected to construct the lncRNA and mRNA signature scores based on the following formula: the risk score = (β1*expression level of RNA1) + (β2*expression level of RNA2) + (β3*expression level of RNA3), β refers to the multivariate regression coefficient. The “SurvivalROC” package of R was performed to assess the prognostic performance by comparing the area under the curve (AUC) of time-dependent receiver–operator characteristic (ROC) curve within 3 years.

Statistical analysis

In this study, we used R software (3.6.2 version) and GraphPad Prism 7 to plot the figures. Kaplan–Meier (K–M) analysis was used to assess overall survival rate. P < 0.05 was considered statistically significant.

Results

Identification of differently expressed lncRNA, miRNAs, mRNAs in lymph node metastatic LUSC compared with lymph node nonmetastatic LUSC

We detected differently expressed RNAs in LUSC samples with lymph node metastasis and LUSC samples without lymph node metastasis. In total, 224 lncRNAs, 160 miRNAs, 913 mRNAs were associated with LUSC lymph node metastasis (Fig. 1a–c). Among them, 131 up-regulated lncRNAs and 93 down-regulated lncRNAs were obtained (Fig. 1a). 325 mRNAs were up-regulated and 588 mRNAs were down-regulated (Fig. 1c). For analysis of miRNA profiles, 160 miRNAs were screened, including 119 up-regulated miRNAs and 41 down-regulated miRNAs (Fig. 1b). Enrichment of these differently expressed genes represents the significance of a function. By GO analysis, we found that these genes are enriched in many biological processes, cellular components, molecular functions (Fig. 2). The most significantly enriched biological processes included “axoneme assembly”, “cilium movement”, “cornification” (Fig. 2). Other significant GO categories included “axoneme”, “blood microparticle”, “ciliary plasm”, “cation transmembrane transporter activity”, “inorganic cation transmembrane transporter activity” (Fig. 2). The significant pathways identified by KEGG pathway analysis included “neuroactive ligand-receptor interaction”, “metabolism of xenobiotics by the cytochrome P450”, “cAMP signaling pathway” (Fig. 3).

Fig. 1.

Volcano plot showing differently expressed RNAs in lymph node metastatic LUSC compared with lymph node nonmetastatic LUSC. a lncRNAs. b miRNAs. c mRNAs. The red points represent up-regulated RNAs and the baby blue represents down-regulated RNAs. Black points represent non-significant RNAs

Fig. 2.

GO analysis of differently expressed mRNAs through diverse GO categories. a Biological process. b Cell component. c Molecular function

Fig. 3.

KEGG analysis of differently expressed mRNAs

Survival-associated lncRNA, miRNAs, mRNAs in lymph node metastatic LUSC compared with lymph node nonmetastatic LUSC

The association between differently expressed RNAs (lncRNAs, miRNAs and mRNAs) with patients’ survival time were studied. Results showed that 28 lncRNAs, 8 miRNAs, 105 mRNAs were significantly associated with patients’ overall survival time. The top 15 survival-associated lncRNAs, mRNAs and all survival-associated miRNAs are present in Fig. 4. By pathway and process enrichment analysis, we found that these survival-associated genes were significantly enriched in “regulation of transmembrane transport”, “regulation of blood circulation”, “plasma lipoprotein particle organization” (Fig. 5a). Furthermore, we had input most significant survival-associated mRNAs into STRING to generate a protein–protein interaction network. Results showed significant hub genes, including APOB, HPR, ORM1, SERPIND1, AMBP and FGA (Fig. 5b).

Fig. 4.

Forest plots of hazard ratios of survival-associated RNAs. a Hazard ratios of top 15 survival-associated lncRNAs. b Hazard ratios of all significant survival-associated miRNAs. c Hazard ratios of top 15 survival-associated mRNAs. Red represents risk factors (hazard ratios > 1), blue represents protective factors (hazard ratios < 1)

Fig. 5.

Pathway enrichment and PPI analysis of survival-associated mRNAs. a Network of enriched terms across survival-associated mRNAs, colored by cluster ID. b Protein–protein interaction network of survival-associated mRNAs based on STRING database analysis and Cytoscape. Node size represents the degree of connection

Construction of survival-associated ceRNA network

To better understand the regulation relationship of survival-associated differently expressed RNAs (lncRNAs, miRNAs, mRNAs), a complex lncRNA–miRNA–mRNA network was established. As shown in Fig. 6a, the network was consisted with 57 interactions and 40 molecules, including 9 lncRNAs (Table 1), 8 miRNAs (Table 2) and 23 mRNAs (Table 3). Considering the key regulatory role of lncRNA, K–M plots of top three lncRNAs were used to reveal their values in predicting prognosis (Fig. 6b–d).

Fig. 6.

ceRNA network and Kaplan–Meier curves of lncRNAs. a Survival-associated ceRNA network in LUSC. The pink circle represents risk RNAs and blue circle represents protective RNAs. The diamond, triangle and ellipse represent survival-associated lncRNAs, miRNAs and mRNAs, respectively. b–j Kaplan–Meier curve estimates of survival time by different expression levels of lncRNAs in ceRNA network

Table 1.

lncRNAs in ceRNA network

lncRNA	logFC	HR (95% CI)
AC073365.1	0.68	0.95 (0.91–0.99)
AC007098.1	0.70	0.83 (0.75–0.93)
LINC00462	0.59	0.92 (0.87–0.98)
HOTTIP	− 0.77	0.92 (0.85–0.99)
AC025419.1	1.40	1.08 (1.00–1.17)
LINC01843	0.60	1.08 (1.01–1.15)
AC091078.1	0.69	0.90 (0.83–0.97)
LINC02137	0.54	0.91 (0.84–0.98)
LINC02389	0.53	0.91 (0.84–0.99)

FC fold change, HR hazard ratio, CI confidence interval

Table 2.

miRNAs in ceRNA network

miRNA	logFC	HR (95% CI)
miR-302a-5p	1.64	2.44 (1.42–4.20)
miR-302b-3p	1.59	2.58 (1.44–4.63)
miR-302d-3p	0.78	4.76 (1.69–13.39)
miR-3196	0.82	3.55 (1.37–9.20)
miR-4465	1.82	1.68 (1.10–2.57)
miR-3161	1.05	1.28 (1.04–1.57)
miR-2113	1.13	1.84 (1.09–3.11)
miR-483-3p	− 0.85	1.14 (1.00–1.31)

FC fold change, HR hazard ratio, CI confidence interval

Table 3.

mRNAs in ceRNA network

mRNA	logFC	HR (95% CI)
ELFN2	− 0.51	1.21 (1.06–1.39)
CCDC177	0.75	0.75 (0.62–0.91)
CRTAC1	− 0.97	1.12 (1.03–1.21)
EPHA7	0.97	0.89 (0.82–0.97)
HAPLN1	0.83	0.87 (0.76–0.99)
SLC28A1	− 0.51	1.89 (1.24–2.88)
SSTR1	− 0.66	1.28 (1.09–1.50)
CCDC141	− 0.63	1.27 (1.04–1.56)
ADRA2B	0.71	0.89 (0.80–0.99)
PLCXD3	− 0.72	1.19 (1.00–1.41)
SLITRK5	0.73	0.86 (0.75–0.99)
VGLL3	0.58	1.11 (1.00–1.24)
ZNF385B	− 0.56	1.14 (1.02–1.28)
EREG	0.69	1.11 (1.03–1.19)
NUDT11	0.52	0.90 (0.81–1.00)
PON1	− 1.09	1.34 (1.05–1.72)
RASGRF1	− 0.76	1.15 (1.03–1.28)
HPR	− 1.16	1.93 (1.37–2.72)
CLIC5	− 0.53	1.14 (1.03–1.27)
SPTA1	0.58	1.32 (1.05–1.64)
BHMT2	0.56	1.16 (1.02–1.33)
MT1A	− 0.68	1.25 (1.06–1.48)
AMH	0.60	0.83 (0.72–0.96)

FC fold change, HR hazard ratio, CI confidence interval

Establishment of prognostic prediction model

To identify the independent prognostic predictors of LUSC, multivariate Cox regression analysis was performed. With P value < 0.05, a group of lncRNA signatures including 3 lncRNAs (Table 4) and a group of mRNA signatures including 21 mRNAs were detected to have independent prediction value (Table 5). Furthermore, according to the P value, the top three significant lncRNAs (AL161431.1, LINC02389, APCDD1L.DT) and mRNAs (KLK6, SLITRK5, CCDC177) were used to construct prognostic prediction models, respectively. Using the three RNAs, we calculated a risk score for each sample. Based on the risk scores, LUSC patients were divided into two groups: high risk and low risk. K–M analysis showed that high risk groups were associated with worse prognosis of patients (Fig. 7a, b). ROC curves were also performed to assess the efficiency of these prediction models. The AUC of 3-lncRNAs and 3-mRNAs signature were 0.662 and 0.665, respectively (Fig. 7c, d). As shown in Fig. 8, the risk score could well assess patients’ prognosis.

Table 4.

Significant lncRNAs by multivariate Cox regression analysis

lncRNA	HR	P value
AL161431.1	1.06	0.009
LINC02389	0.90	0.031
APCDD1L-DT	1.09	0.045

HR hazard ratio

Table 5.

Significant mRNAs by multivariate Cox regression analysis

mRNA	HR	P value
KLK6	1.28	0.000
SLITRK5	0.67	0.001
CCDC177	0.59	0.001
AJAP1	1.42	0.001
S100A1	0.62	0.004
UTS2B	2.26	0.005
FUT6	0.74	0.011
ANKUB1	0.70	0.012
HPR	3.33	0.016
RP1	1.91	0.016
KCNK7	0.70	0.016
MYEOV	1.17	0.020
HAPLN1	0.76	0.020
KCNE5	2.04	0.020
DMRT2	1.21	0.020
AC010325.1	2.51	0.021
NAT8	0.11	0.022
R3HDML	2.73	0.025
C16orf96	1.57	0.028
PRR26	1.66	0.033
PLCXD3	0.67	0.047

HR hazard ratio

Fig. 7.

Prognostic prediction model based on three lncRNAs and three mRNAs. a, b Kaplan–Meier curve estimates of survival time by different risk grouped by three lncRNA-based risk score (a) and three mRNA-based risk score (b). c, d ROC curves analysis for prognostic prediction by three lncRNA-based risk score (c) and three mRNA-based risk score (d)

Fig. 8.

Prognostic signatures in distinguishing LUSC patients into high risk and low risk. a, b The distribution of lncRNA-based risk score (a) and mRNA-based risk score (b). c, d The distribution of patients’ survival status in different risk grouped by lncRNA-based risk score (c) and mRNA-based risk score (d)

The occurrence and development of tumor is a process involving multiple types of RNAs. Multiple types of RNA-based risk scores could increase prediction accuracy. Thus, we calculated risk scores using the above six RNAs. As shown in Fig. 9a, patients with high risk score had shorter survival time than patients with low risk score. The AUC of six RNA signatures was 0.701 (Fig. 9b).

Fig. 9.

Prognostic prediction model based on six RNAs. a Kaplan–Meier curve estimates of survival time by different risk groups by six RNA-based risk score. b ROC curve analysis for prognostic prediction by six RNA-based risk score

Discussion

Growing studies indicate that complex diseases, especially cancers, are caused by complex genomic variations and gene regulations (Nibbe et al. 2011). However, ceRNA hypothesis is considered as a novel way of lncRNAs to regulate mRNAs by competitive binding of miRNAs (Tay et al. 2014; Thomson and Dinger 2016). In the present study, to explore the possible molecular mechanism of LUSC prognosis, we took full advantage of huge genomic information provided by TCGA database to find a group of survival-associated lncRNAs, miRNAs and mRNAs. Furthermore, we constructed a prognostic ceRNA network. In addition, we developed a prediction model based on lncRNA and mRNA signature. These results indicate that ceRNA regulation play vital roles in LUSC prognosis.

In contrast to the previous studies, our present study was different distinctly. Lymph node metastasis is an indicator affecting the survival of patients. Compared with the previous studies of establishing a prognostic model based on differently expressed RNAs identified between tumor group and normal group, our research was to identify differently expressed RNAs between LUSC patients with lymph node metastasis and LUSC patients without lymph node metastasis. RNAs identified by this way are more likely to be prognostic biomarkers of patients. Furthermore, survival analysis of these RNAs was performed. Prognostic ceRNA network was constructed based on survival-associated RNAs. By survival analysis, we identified some survival-related markers. These markers have been reported in many cancers (Chen et al. 2019b; Das et al. 2019; Ji et al. 2018; Luo et al. 2019; Sun et al. 2019; Zhang et al. 2017; Zhou et al. 2018). However, little was known about their roles in LUSC progression and patients’ prognosis. Kamil et al. found that high expression of FLNC was associated with worse survival in glioblastoma multiforme (Kamil et al. 2019). In addition, the study reported that FLNC was not only associated with lymph node metastasis, but also with patients’ prognosis in esophageal squamous cell carcinoma (Tanabe et al. 2017). miR-483-3p was involved in the progression of many cancers. In the pancreatic adenocarcinoma, miR-483-3p inhibited the proliferation of tumor by in vitro and in vivo experiments. Furthermore, high expression of miR-483-3p could reduce patients’ survival time (Wang et al. 2015). In esophageal squamous cell carcinoma, miR-483-3p expression was also associated with patients’ prognosis (Pepe et al. 2018). lncRNA plays an important role in tumorigenesis and development through ceRNA mechanism, but its application in survival and prognosis is rare. Yue et al. found WARS2-IT1 was associated with prognosis of hepatocellular carcinoma patients (Yue et al. 2019). Ye et al. found that WARS2-IT1 was an independent prognostic marker by univariate and multivariate Cox analyses (Ye et al. 2019). In lung adenocarcinoma, LINC00261 was significantly down-regulated, leading to poor prognosis and recurrence (Shahabi et al. 2019). Shi et al. found that in small cell lung cancer, low expression of LINC00261 was related with TNM stage, lymph node status, distal metastasis, and poor survival of patients. Multivariate analysis showed that LINC00261 was an independent prognostic marker (Liu et al. 2017). Mechanism study showed that LINC00261 could regulate the Wnt signaling pathway and inhibit tumor metastasis by miR-522-3p/ SFRP and miR-105/FHL1 axis (Shi et al. 2019; Wang et al. 2019). In LUSC, LINC00261 was associated with survival and is an independent prognostic factor (Qi et al. 2019). However, LINC00261 was not found as an independent prognostic factor in our study. This may be attributed to different study purposes and data processing. There are still many novel lncRNAs not found in public reports according to pubmed search. These lncRNAs were associated with LUSC patients’ prognosis in our study. In brief, results demonstrated that these RNAs may help improve the accuracy of diagnosis and provide new targets for the treatment of LUSC.

Considering that proteins are the performers of molecular function. Enrichment and PPI analysis were performed to fully understand the biological functions of these abnormal survival-associated mRNAs and their interactions. In the present study, enrichment analysis showed that these abnormal mRNAs were mainly enriched in several pathways. Among these pathways, the most significant one was the “regulation of transmembrane transport” pathway. As is well-known that transmembrane transport is essential for molecular exchange for cell survival (Ming et al. 2019). Studies have shown that transmembrane protein was critical for cancer progression and metastasis (Qian et al. 2019). In addition, we also found pathways related to cancer, such as MAPK signaling pathway. Shain et al. reported that MAPK signaling pathway activation and ramp-up involved in melanoma evolution (Shain et al. 2018). Luk et al. found REX1 deficiency induced enhancement of MAPK signaling and drives hepatocarcinogenesis (Luk et al. 2019). All in all, these results provided clues for further mechanism studies.

Since Salmena put forward the hypothesis of ceRNA network (Salmena et al. 2011), more and more studies have shown that ceRNA network mediated by lncRNA is involved in tumor progression (Yang et al. 2019; Park et al. 2018). In the present study, we constructed ceRNA network using survival-associated RNAs. The purpose is to understand the correlation between lymph node metastasis and patient survival well. Through careful analysis of the ceRNA network, we found that miR-302d-3p, miR-302b-3p, miR-4465 could bind with a number of lncRNAs. There are several reports that these three miRNAs play vital roles in tumor progression. Li et al. reported that miR-302d-3p and miR-302b-3p could inhibit epithelial-mesenchymal transition and promote apoptosis in human endometrial carcinoma cells (Li et al. 2018). Wu et al. found lncRNA SNHG6 could bind with miR-4465 to promote cell proliferation and migration in ovarian clear cell carcinoma (Wu et al. 2019a). In addition, we also found three lncRNAs: HOTTIP, AC073365.1, AC091078.1 that bind with a number of miRNAs. HOTTIP is a popular lncRNA in recent years. It has been confirmed that it existed in tumor exosome (Zhao et al. 2018) and is related with tumor progression (Sun et al. 2017). Wang et al. found HOTTIP regulated the miR-615/IGF-2 pathway to promote renal cell carcinoma progression (Wang et al. 2018). Studies on the other two lncRNAs: AC073365.1, AC091078.1 were rare. Molecular mechanism of these newly discovered RNAs need further research.

However, several limitations in our study should be discussed. First, results need to be verified by other experimental methods. Second, novel lncRNAs that identified by our present study need to be further explored to understand the underlying molecular mechanism. Third, limited to the small number of survival-related miRNAs, we did not find an independent prognostic miRNA in multivariate Cox analysis. So prognostic prediction model had no miRNA. In summary, our present study delineates a ceRNA network that may help to improve the prognosis of LUSC patients. Our results provide clues for the molecular mechanism of LUSC progression and identified novel therapeutic targets for LUSC treatment.

Author contributions

QJ and YZ designed and supervised the study. SM, XL and HZ analyzed the data and wrote the original draft. SZ, YY, SY edited draft. All authors have read and approved the final manuscript.

Funding

This study was supported by The National Science Foundation for Young Scientists of China (Grant No. 81802415 to YZ), Shandong Provincial Natural Science Foundation (Grant No. ZR2018PH025 to YZ), The Doctoral Scientific Fund Project of the Affiliated Hospital of Qingdao University (Grant No. 2796 to QJ), Clinical Medicine + X Project, Medical College, Qingdao University.

Compliance with ethical standards

Conflict of interest

The authors disclose no conflicts.