Introduction

Lung cancer remains the prominent contributor of cancer-related mortality, with the worldwide 5-year survival rate of which is around 16.6% [1, 2]. Recently, with the wide application of low-dose computed tomography for early screening and the rapid development of target drugs for genetic mutations, the progress against lung cancer has achieved profound success. As of 2017, the mortality rate of lung cancer dropped from its peak by 51% among males and by 26% among females [2]. Nonetheless, the 5-year survival rate for lung cancer patients is still not well manifested. Additionally, the early-stage patients’ prognosis displays quite disparate from those of advanced-stage, with 5-year survival ranging from 85 to 6% [3], but its recurrence rate is still up to 90% after local resection or radical excision [4].

Currently, non-small cell lung cancer (NSCLC), accounting for approximately 85% of all lung cancer, has been endowed with several therapeutic options, including surgery, chemotherapy, radiation, target therapy and immunotherapy [5]. And surgery is the first choice of curative treatments for the medically operable. Owing to its readily entry into regional lymph nodes and apt to metastasize at an early stage, however, the recurrence rate accounts for approximately 27% to 38% for stage I NSCLC patients [6,7,8]. Therefore, intensive efforts have been directed to elucidate the molecular mechanism of premalignancy development and progression, and to identify potential molecular signatures for early diagnosis and interception.

Recently, long noncoding RNAs (lncRNAs) have attracted significant attentions in various cancers. LncRNA is a class of transcripts with length of more than 200 nucleotides that possesses limited or no protein-coding capacity, which is transcribed by RNA polymerase II, spliced, 5’capped, and polyadenylated [9]. It has been identified that lncRNAs were involved into diverse cellular, physiological and pathological process via a series of mechanism [10, 11], including serving as critical regulators of tumorigenesis and metastasis [12]. Furthermore, accumulating evidences revealed that lncRNA could disrupt miRNA-mediated degradation of target mRNAs by acting as “miRNA sponges” [13], indicating coding and noncoding RNAs could control one another through their ability to compete for miRNA binding locus, which termed as “ceRNA”. Under this hypothesis, a growing number of evidence revealed that ceRNA axis could contribute to tumorigenesis, progression and metastasis of cancer [14, 15]. For instance, researchers have demonstrated that LINC00336, an novel regulator of ferroptosis, could act as a ceRNA to affect tumor genesis and progression and mediate the expression of cystathionine-β-synthase (CBS) by sponging miR6852, which may serve as a potential therapeutic target of lung cancer [4B). Of these, the turquoise module was most highly correlated with DElncRNAs (Fig. 4B and C). Interestingly, we found that the correlation coefficient between royalblue module and progressive disease reached to 0.94, indicating that the royalblue module is a gene set specifically associated with progression of disease. The royalblue module is also the most relevant module to distant metastasis (cor = 0.72), indicating the correlation of lncRNAs with prognosis (Fig. 4D). By setting the module membership (MM) to > 0.8 and the gene significance (GS) to > 0.4 as threshold, we selected a total of 81 hub lncRNAs from the modules (Table S 5), most of which were involved in the recurrence or metastasis of disease. The results indicated that the DElncRNAs identified were involved in the prognosis of stage I LUAD patients.

Fig. 3
figure 3

Construction of ceRNA network. A-B The regulation relationship between miRNA and mRNA predicted by miRanda, and the relationship between miRNA and lncRNA predicted by miRanda, with the criteria of total score ≥ 150 and total energy ≤ -20. The green node represents lncRNAs; the red node represents miRNAs; the blue node represents mRNAs

Full size image

Subsequently, we found that 19 genes were associated with distant metastasis, 5 genes associated with locoregional recurrence and 11 genes associated with distant metastasis, and another 65 genes associated with new primary tumor. Of the hub genes identified, FENDRR, LINC00639 and RP4-676L2.1 were also predicted as ceRNAs in stage I LUAD constructed by our database. Furthermore, we constructed ceRNA network using the three lncRNAs by Cytoscape (Fig. 5A). Finally, ceRNA network containing 174 lncRNA-miRNA-mRNA regulatory relation for FENDRR, 937 lncRNA-miRNA-mRNA regulatory relation for LINC00639, and 54 regulatory relation for RP4-676L2.1. It showed that and LINC00639 were hub nodes that could target more miRNAs and mRNAs in the network. These findings indicated that DElncRNAs regulated mRNA expression via interaction with miRNAs.

Fig. 4
figure 4

WGCNA and identification of significant modules. A The soft-thresholding power in WGCNA. B The eigengene of each colored module were calculated to establish an adjacent matrix. C Cluster dendrogram obtained from lncRNAs data of stage I LUAD in TCGA dataset with average hierarchical linkage clustering. The color row underneath the dendrogram represents the module assigned by Dynamic Tree Cut. D Module-trait relationship heatmap. The row represents the modules, while the column represents the trait. The values in the box represents the correlation and P values

Full size image
Fig. 5
figure 5

The relationship and expression between lncRNAs and miRNAs. A The ceRNA network constructed by FENDRR, LINC00639 and RP4-676L2.1. B-C The FENDRR and miR-6815-5p expression levels of stage I LUAD tissues and paired normal tissues were tested by RT-PCR (n = 26). D Prognostic significance of FENDRR expression on PFS for stage I LUAD patients by the median value as cutoff

Full size image

Interestingly, one of subnetworks showed that lncRNA FENDRR only act as a sponge for hsa-miR-6815-5p to regulated mRNAs, including TNS1, PDLIM2, PPFIBP1, SCMH1, PLXDC1 and so on, which mainly involved in adherens junction, chromatin silence, blood vessel development (Figure S2A). Thus, we further explore the relationship between FENDRR and hsa-miR-6815-5p. In our analysis of RNA sequencing, the results showed that FENDRR and LINC00639 were significantly downregulated, while RP4-676L2.1 and hsa-miR-6815-5p was marly upregulated in stage I LUAD patients (P value < 0.05, Figure S 2B). Further, RT-PCR indicated that expression of FENDRR was significantly decreased and expression of hsa-miR-6815-5p was markedly increased in LUAD tissues compared with paired normal tissues (P value < 0.001) (Fig. 5B and C), while the expression of RP4-676L2.1 and LINC00639 were not significant (Figure S2C). Thus, the results revealed that down-regulation of FENDRR maybe involved in tumorigenesis via upregulating of hsa-miR-6815-5p.

To further explore the FENDRR identified, we evaluated its effects on recurrence by univariate Cox regression analysis in TCGA dataset. It showed that FENDRR was significantly associated with PFS (HRs = 1.69, 95%CI 1.07–2.68, P < 0.05), indicating that FENDRR could be regarded as a prognostic factor of tumor recurrence. Using median FPKM of FENDRR as cutoff value, we divided the patients into high-expression group and low-expression group. KM survival curve for the two groups indicated that PFS in the high FENDRR expression group displayed significantly longer than that in the low FENDRR expression group (Fig. 5D), while PFS for RP4-676L2.1 and LINC00639 between two groups were not significantly different (Figure S2D).

Discussion

Although it has achieved great improvement on clinical management for LUAD aided by the discovery of genetic mutations, there is still main obstacles on improving prognosis of patients resulting from its apt to migration and invasion. Therefore, the promise of utilization of lncRNAs in predicating clinical prognosis has attracted much attention in translational research. In this study, we identify three lncRNAs, FENDRR, LINC00639 and RP4-676L2.1, which might be related with prognosis of stage I LUAD patients. However, only FENDRR is predicted to be associated with PFS of patients, and validated to be down-regulated in LUAD tissues by RT-PCR assay. The primary results uncover the potential role of FENDRR in the progression of early-stage LUAD. Importantly, the study gives a novel hint of the mechanism by which FENDRR might involve in the progression of disease.

Previous studies have revealed that lncRNAs could exert their biological function via interaction with DNA, RNA and protein depending on their location within cells [25]. Although there is still a lot to learn about lncRNA, accumulating studies demonstrated its function on relieving target mRNAs degradation mediated by miRNA, playing a role as ceRNAs, which could act as vital regulators in various physiological and pathological process of tumor [26, 27]. To date, despite the plethora of reports, several lncRNAs have been identified to be involved in recurrence and metastasis of lung cancer [28,29,30]. In this study, we explored the most likely significant lncRNAs profile alternations of stage I LUAD, and revealed that FENDRR, LINC00639 and RP4-676L2.1 were predicted to be associated with progression of disease. However, only FENDRR were validated as predictive lncRNA of PFS by cox regression analysis, indicating its underlying function responsible for progression of LUAD.

LncRNA fetal-lethal non-coding developmental regulatory RNA (FENDRR), also named as FOXF1-AS1, is an intergenic lncRNA with consisting of seven exons. It is located at 3q13.31, 1,354 bp upstream of transcriptional start site, which is transcribed from a bidirectional promoter shared with the protein coding gene Foxf1 and Pitx2. Previous studies revealed that FENDRR could regulate cell migration, invasion, and lymphatic metastasis, demonstrating its inhibitory regulation in tumor progression [31]. It was reported that FENDRR was highly expressed in the lung, while was lowly expressed in the liver, colon and brain tissues, and the level of which was associated with prognosis of patients [32]. In breast cancer, the low level of FENDRR was associated with poor prognosis of patients, including a shorter survival and a shorter PFS [33]. Also, FENDRR was found to be related with survival of gastric cancer patients, the expression of which could be suppressed in gastric cancer associated fibroblasts by hypermethylation [34]. In agreement with previous studies [31], the level of FENDRR expression in tumor was also shown to be suppressed compared with paired normal tissue.

The mechanism of anti-malignant effects for FENDRR might involve in inhibiting cell migration, invasion and mediated stem-like properties by regulating epithelial mesenchymal transition (EMT) [35,36,37]. Previous studies reported that FENDRR could anchor PRC2 and/or TrxG/MLL complexes at its target promoters, increasing PRC2 occupancy and H2K27 trimethylation, which lead to the attenuation of target gene expression [38]. In gastric cancer, FENDRR was found to increase cell migration and invasion via up-regulation of FN1, MMP2 and MMP9 [34]. In vitro, FENDRR was revealed to decrease the IC50 for cisplatin in A549/DDP cells, and depressed chemotherapy resistance to cisplatin in NSCLC [37]. Hitherto, the contribution of FENDRR on progression of LUAD has not been well clarified. In our dataset, FENDRR could targeted miR-6815-5p in ceRNA network constructed, suggesting that it might exert function on regulation expression of target mRNAs by ‘sponging’ miR-6815-5p. To date, only one study was found that miR-6815-5p was significantly upregulated in exosomes of HPV16-infected cervical-vaginal fluid (CVF) based on microarray analysis, while its biological functions was not reported and unclear [39]. Hence, further studies on miR-6815-5p are warranted. To the best our knowledge, this is the first study where predicts the regulation of miR-6815-5p on targeted mRNAs in LUAD. Apart from the predication of lncRNAs on PFS, the main breakthrough of our study is centred around the regulation of miR-6815-5p on FENDRR, which needs to explore its molecular mechanisms to increase further confidence to this result.

Regarding RP4-676L2.1 and LINC00639, we speculated that they could be related with recurrence of disease by WGCNA using TCGA data set, suggesting their prognostic value in stage I LUAD. However, they were not significantly correlated with PFS of stage I LUAD patients by cox regression analysis. Surprisingly, RP4-676L2.1 and LINC00639 were also not successfully validated to be consistent difference of expression between tumor and normal lung tissue by qPCR. We speculated that it may lie with the bioinformatic analysis. Still, further studies are needed to validate using larger samples.

The discovery of lncRNAs contributes significantly to clinical prognosis in stage I LUAD patients. Its strength lies in not only testing using the sequencing dataset, but also validating using TCGA dataset by bioinformatics methods. Still, some limitations must be noted. Firstly, we speculated that FENDRR/miR-6815-5p axis may play an important role in the biological behavior of LUAD based on bioinformatics analysis and RT-PCR, but the specific mechanism still needs to be further verified such as the functional experiments of LUAD cells and RNA-RNA interaction verification experiments. Secondly, due to the low number of samples performed RNA-seq, the WGCNA was constructed using TCGA dataset, which might induce racial bias. This is not surprising, a substantial number of studies on lncRNAs have been validated using TCGA dataset. Thirdly, the lncRNAs identified from RNA-seq were inconsistent with the qPCR results, which may result from incorrect annotation in the bioinformatic analysis.

Conclusions

Taken together, this study contributed significantly to the wider knowledge of lncRNAs and ceRNAs involvement of progression of stage I LUAD. Out of the three lncRNAs validated, FENDRR is validated to involve in the recurrence of disease, which confirm previous findings. Based on the FENDRR-related ceRNA network constructed, it was revealed the regulation of miR-6815-5p on FENDRR for the first time, therefore open to further research to explore molecular mechanism. However, studies are still needed to establish the role of FENDRR/miR-6815-5p axis in the progression of early-stage LUAD.