Long non-coding RNAs in esophageal cancer: molecular mechanisms, functions, and potential applications

Esophageal carcinoma (EC), a serious malignant cancer, is the sixth leading cause of cancer-related death [1, 2]. Despite advances in multidisciplinary treatment, the 5-year relative survival rate remains less than 20% [3]. EC includes the following two primary pathological types: esophageal adenocarcinoma (EAC) and esophageal squamous cell carcinoma (ESCC) [2]. EAC is the leading histological type observed in patients from western countries, whereas ESCC has become the leading cause of EC in Asian countries and predominates over EAC worldwide [4, 5]. The pathogenesis of EC is complex and differs between EAC and ESCC. For EAC, the primary predisposing cause is metaplasia that is likely caused by chronic exposure to acid and bile reflux, such as in the case of Barrett’s esophagus and chronic gastroesophageal reflux disease [6]. However, the origin of ESCC carcinogenesis is not fully understood. Early-stage EC can be effectively treated with curative surgery, but for advanced cases, the therapeutic strategies are limited [7]. Unfortunately, EC patients are usually diagnosed at an advanced stage accompanied with lymphatic metastasis, and therefore they are not eligible for surgical resection [8]. The current standard treatment for these patients is concurrent definitive chemo- or radiotherapy, or a combination of both [9]. However, therapy resistance and tumor recurrence are major obstacles for EC therapy and are critical issues leading to poor prognoses [10]. Within the EC, a small number of cells termed cancer stem-like cells (CSCs) are considered to account for the initiation, recurrence, and therapeutic resistance of EC [10]. In recent years, compelling evidence has demonstrated the crucial roles of long non-coding RNAs (lncRNAs) in the pathogenesis and progression of EC.

LncRNAs, which are an emerging focus of current cancer research, are defined as endogenous cellular RNAs that are more than 200 nucleotides in length and are incapable of encoding protein [11, 12]. Initially, lncRNAs were considered as transcriptional “noise,” given their relatively low expression levels compared with mRNAs and their lack of protein-coding capacity [13]. However, in-depth studies in recent years revealed that lncRNAs possess certain characteristics of mRNAs; for instance, lncRNAs are transcribed by RNA polymerase II, equipped with a 3′ polyA tail and a 5′ cap, and contain a promoter and structure consisting of multiple exons [14, 15]. Accumulating evidence suggests that the aberrant lncRNA expression is associated with oncogenesis and the development of various cancers [16, 17]. LncRNAs have been shown to interact directly with DNA, RNA, and proteins to regulate several mechanisms, including the following: chromatin modification, RNA transcription, pre-mRNA splicing, mRNA translation, and other mechanisms that influence gene expression [18, 19]. Moreover, several lncRNAs have been functionally well-characterized in cancer pathogenesis and development and may be potential novel biomarkers for cancer diagnosis and prognosis, as well as therapeutic targets.

In this review, we focused our efforts on the recent findings regarding the molecular mechanisms and functional roles of lncRNA in EC oncogenesis and development. In addition, we discussed the potential implications of lncRNAs as biomarkers for the diagnosis and prognosis of EC.

LncRNAs may act as signals or guides for the recruitment of chromatin-modifying complexes to induce transcription, and they may even act as decoys that bind to transcription factors (TFs) to prevent the transcription factors from binding to target gene promoter regions, thereby suppressing transcription [20, 21]. In addition, lncRNAs can hybridize to pre-mRNAs, block the recognition of splice sites by spliceosomes, and regulate the alternative splicing of pre-mRNAs to produce alternate transcripts [17, 22]. An additional biological function of lncRNAs may include serving as “miRNA sponges” through interactions with miRNAs to inactivate these small regulatory RNAs and hence increase the expression of the miRNA target genes [23‐25]. Finally, lncRNAs may be involved in the modulation of protein localization, activity, and function [26]. In this section, we highlight the molecular mechanisms of lncRNAs in EC via their interactions with chromatin, DNA, RNA, and regulatory proteins (Fig. 1 and Table 1).

LncRNA-dependent chromatin regulation involves the recruitment and modulation of histone-modifying enzymes that induce chromatin modification at promoters and enhancers [27, 28]. In this manner, lncRNAs can regulate gene expression through histone modification, DNA methylation, and chromatin structure alteration [29, 30].

It has been demonstrated that many lncRNAs are associated with polycomb repressive complex 2 (PRC2), which is responsible for the trimethylation of lysine 27 on histone 3 (H3K27me3) and mediates the silencing of the target gene through local chromatin reorganization [31]. Enhancer of zeste homolog 2 (EZH2) and SUZ12 are subunits of the PRC2 complex. Wu et al. [32] demonstrated that cancer susceptibility candidate 9 (CASC9) downregulates the expression of PDCD4 via the recruitment of EZH2 to alter H3K27me3 levels at the promoter region of PDCD4. In addition, SET-binding factor 2 antisense RNA1 (SBF2-AS1) was demonstrated to bind to SUZ12 and guide PRC2 to the promoter of CDKN1A to decrease CDKN1A expression in ESCC [33].

The acetylation of histone H3 and H4 is another core mechanism through which chromatin structure and gene expression are altered [34]. In addition to recruiting EZH2, CASC9 also associates with the transcriptional coactivator CBP in the nucleus to increase the enrichment of CBP and H3K27 acetylation in the promoter region of LAMC2, thereby increasing LAMC2 expression [35].

DNA methylation is one of the most common and stable chromatin modification that is associated with gene inactivation [36, 37]. Lung cancer associated transcript 1 (LUCAT1) was originally identified in smoking-related lung cancer [38] and is also associated with colorectal cancer [39], clear cell renal cell carcinoma, and osteosarcoma [40]. A recent study demonstrated that LUCAT1 binds to DNMT1, the most abundant DNA methyltransferase in mammalian cells, and regulates its stability by inducing the ubiquitination of DNMT1 in ESCC [41]. The high levels of LUCAT1 in ESCC inhibit the expression of certain tumor suppressors through DNA methylation.

In addition, NMR, a novel lncRNA identified through microarray assays, was found to be upregulated in ESCC tissues and primarily located in the cell nucleus [42]. NMR interacts with the chromatin regulator BPTF [42], which was demonstrated to be involved in ATP-dependent chromatin remodeling and transcriptional regulation [43]. Hence, by recruiting BPTF to specific loci of chromatin, NMR upregulates the expression of MMP3 and MMP10 via ERK1/2 activation to promote ESCC tumorigenesis.

Following the transcription of RNA in the nucleus, a series of conserved processes are essential for the production of mature mRNAs that can be translated into proteins. LncRNAs modulate gene expression at the RNA level through the regulation of alternative splicing and the stability of mRNAs; additionally, lncRNAs act as miRNA sponges or competing endogenous RNAs (ceRNAs) [17, 26].

Several lncRNAs have been reported to interact with specific proteins to participate in global cellular processes in EC by regulating protein activity and function, modulating protein–protein interactions or directing the localization of proteins within cellular compartments to serve as structural components [26].

Through RNA pull-down assays and chromatin isolation by RNA purification (ChIRP), Xie and colleagues [78] demonstrated that LINC01503 could bind with both EBP-1 and ERK2 in the cytoplasm. Further analysis revealed that both basal and EGF- and IGF-induced phosphorylation of ERK1/2, Akt, p70S6K, and mTOR were significantly decreased following the knockdown of LINC01503. In addition, silencing LINC01503 expression increased the binding of EBP-1 to the PI3K subunit p85, suggesting that LINC01503 inhibits PI3K deubiquitination to activate the PI3K/Akt signaling pathway. Taken together, these findings suggest that LINC01503 contributes to ESCC cell proliferation, migration, and invasion through the activation of the ERK/MAPK and PI3K/Akt signaling pathways.

Snail, an important transcription factor influencing the EMT, binds to the E-box site in the promoter region of E-cad to suppress E-cad expression [79, 80]. This suppression triggers the EMT in a variety of cancer types, including ESCC [81, 82]. Thus, the nuclear localization of Snail is crucial for its role in the EMT progression [83]. A study by Zhang et al. [84] showed that Sprouty4-Intron 1 (SPRY4-IT1) directly increased the transcription and expression of Snail, as well as its nuclear localization, by directly binding with Snail in ESCC cells. SPRY4-IT1 is highly expressed in ESCC tissues, and overexpression of SPRY4-IT1 promotes the EMT in ESCC cells. This finding demonstrates that SPRY4-IT1 may act as an oncogene in ESCC progression via the regulation of Snail.

EZR-antisense 1 (EZR-AS1) interacts with and is part of the RNA polymerase II complex [85]. RIP assays have revealed that EZR-AS1 directly binds with SMYD3, a histone H3-lysine 4 (H3K4)-specific methyltransferase, causing SMYD3 redistribution and recruiting SMYD3 to the binding site in GC-rich regions downstream of the EZR promoter in ESCC cells. This recruitment results in the localized enrichment of SMYD3 and H3K4me3 in the EZR promoter. Lastly, EZR-AS1 was shown to promote ESCC cell migration via enhancing EZR transcription and expression.

Increasing evidence in the last decade indicates that lncRNAs function in a plethora of biological processes, including cell survival and apoptosis, cell cycle progression and proliferation, migration and invasion, stemness, and chemoradiotherapy (CRT) resistance (Table 2).

Although cancer is a complex and heterogeneous disease, one of the common features of cancer is that abnormal cells grow beyond control. In 2000, Hanahan and Weinberg proposed six properties that are the hallmarks of cancer [86]. These basic hallmarks include sustaining growth signaling, evading growth inhibitors, uncontrolled replicative immortality, tissue invasion and metastasis, promoting angiogenesis, and resisting cell death.

Acquired CRT resistance is one of the major obstacles in the treatment of EC [130]. Studies have shown that less than 50% of patients benefit from CRT treatment, and the remaining half patients present resistance to CRT [131]. Recently, several lines of evidence have suggested that lncRNAs are likely to play vital roles in CRT resistance in EC. Zhou et al. [132] examined 18 lncRNAs that were previously reported to be dysregulated in EC or involved in CRT resistance in cisplatin-resistant ESCC cell lines and samples from patients treated with dCRT. The authors detected that three lncRNAs (AFAP1-AS1, UCA1, and HOTAIR) were dysregulated in cisplatin-resistant cells compared with the parent cell line. Moreover, AFAP1-AS1 was significantly overexpressed in tumor tissues compared to the adjacent paired tissues. Furthermore, the overexpression of AFAP1-AS1 was strongly related to the response to dCRT and to the shorter progression-free survival (PFS) and OS of ESCC patients. High AFAP1-AS1 expression could predict resistance to CRT in patients with ESCC. Another lncRNA, LOC285194, also known as LSAMP antisense RNA 3, has been reported to be downregulated in several cancers, including EC and was found to be closely associated with a poor patient prognosis [133‐135]. Additionally, the low expression of LOC285194 could predict resistance to CRT [136]. As mentioned above, TUSC7 promotes cell apoptosis and inhibits chemotherapy resistance through the miR-224-dependent regulation of DESC1 [75].

Tumor radioresistance is very complex and heterogeneous. Although the mechanism underlying radioresistance is not well-understood, several signaling pathways have been demonstrated to be involved in radioresistance. The Wnt/β-catenin pathway is well-known to promote cell growth and survival and has been proven to modulate radioresistance in various cancers [137, 138]. WISP1, a Wnt- and β-catenin-responsive gene, mediates radioresistance primarily through suppressing irradiation-induced DNA damage and activating PI3K kinase [139]. Zhang and colleagues [47] reported that ESCC patients with high WISP1 expression had a significantly poorer prognosis compared with those with low WISP1 levels after radiotherapy. The authors further assayed the expression of 94 cancer-related lncRNAs in WISP1-overexpressed EC cells that received radiation, and they identified 14 upregulated lncRNAs and 5 downregulated lncRNAs. Among these lncRNAs, BOKAS was strongly associated with the irradiation-induced upregulation of WISP1. BOKAS is a natural antisense transcript of BOK, a member of the pro-apoptotic Bcl-2 family. Moreover, the downregulation of BOKAS decreased WISP1 expression and greatly enhanced irradiation-induced DNA damage in EC cells. Taken together, these findings indicate that BOKAS induces radioresistance via promoting the upregulation of WISP1.

CSCs only represent a small portion of cells within a given cancer, but they are believed to be responsible for self-renewal, metastatic ability, tumorigenicity, and therapeutic resistance [140‐143]. Although ECSCs play a critical role in EC, only a few lncRNAs have been discovered to be associated with the functions of these cells. As an example, MALAT1 has been demonstrated to be associated with tumor stem regulation in several cancer types [144‐146]. A recent study by Wang et al. [147] reported that the downregulation of MALAT1 repressed the cancer stem cell-like traits of ECSS through decreasing the expression of tumor stem genes OCT4 and Nanog.

It is recognized that the delayed diagnosis of EC results in metastasis and recurrence and is therefore a major obstacle for EC therapy. Recent studies have demonstrated that lncRNAs play a vital role in the pathological progression of EC. More importantly, lncRNAs have tissue and cell-type specificity. These patterns make lncRNAs attractive as potential biomarkers for the diagnosis and prognosis of EC (Table 3).

Emerging evidence has demonstrated that early diagnosis and effective intervention improves the survival of EC patients. LncRNAs are involved in EC oncogenesis and progression, and the presence of lncRNAs in the peripheral blood and body fluids of EC patients suggests that lncRNAs could serve as diagnostic biomarkers [89, 90]. Tong et al. [148] analyzed the levels of ten lncRNAs in 48 plasma samples and found that POU class 3 homeobox 3 (POU3F3), HNF1A-AS1, and SPRY4-IT1 were markedly higher in ESCC patients compared to healthy controls. In addition, in a cohort of 147 ESCC patients and 123 healthy volunteers, the receiver operating characteristics (ROC) curves demonstrated a strong separation between ESCC patients and healthy volunteers, with an area under the curve (AUC) of 0.842 (95% CI 0.794–0.890; p < 0.001) for POU3F3, with a 72.8% sensitivity and 89.4% specificity. In another study, Hu and colleagues [149] found that Linc00152, CASP8- and FADD-like apoptosis regulator-antisense 1 (CFLAR-AS1), and POU3F3 were significantly upregulated in a large cohort of 205 ESCC patients and 82 esophagus dysplasia patients compared to 210 healthy controls, with an AUC of 0.698, 0.651, and 0.584, respectively. The merged AUCs of the three lncRNAs were 0.765, while the AUC increased to 0.955 after merging the three factors with CEA. The circulating levels of the three lncRNAs were associated with poor postsurgery prognoses of ESCC patients in Kaplan–Meier curves. The authors also demonstrated the stability of the lncRNAs that were expressed in the human plasma, which is a crucial prerequisite for a biomarker. HOTAIR was shown to be significantly upregulated in ESCC tissues [150, 151]. A recent study demonstrated that the expression level of HOTAIR in the serum of ESCC patients (n = 50) was significantly higher compared to healthy controls (n = 20), with an AUC of 0.793 (95% CI 0.692 to 0.895, P < 0.01) and optimal cutoff values of 0.094 (sensitivity 56.0%, specificity 90.0%) [152]. In addition, the serum level of HOTAIR was positively correlated with the distant metastasis and TNM stage. MicroRNA-31 host gene (MIR31HG) is another EC-related lncRNA that is significantly upregulated in EC tissues compared to the adjacent normal tissues, as well as in ESCC plasma, compared to the healthy individuals [153]. In addition, plasma MIR31HG was found to differentiate between ESCC patients and healthy individuals by AUC analysis (95% CI 0.656 to 0.841, P < 0.01). These findings indicated that POU3F3, HOTAIR, and MIR31HG may be potential biomarkers for EC diagnosis. However, given that these lncRNAs have been shown to be dysregulated in cancers other than EC [154‐160], they may best serve as effective diagnostic biomarkers in EC in combination with other variables.

In recent years, great advances have been made in research into lncRNA-related prognostic biomarkers. The aberrant expression of several lncRNAs has been significantly associated with EC prognosis and may serve as potential prognostic predictors.

The expression of prostate cancer-associated ncRNA transcript 1 (PCAT-1) was markedly upregulated in 130 cancerous tissues compared to matched noncancerous tissues in ESCC [161]. High expression levels of PCAT-1 have been correlated with the depth of tumor invasion, lymph node metastasis, and TNM stage. Kaplan–Meier analysis has revealed that patients in the high PCAT-1 group (n = 65) had shorter survival times compared with those in the low PCAT-1 group (n = 39). The expression of lncRNA ZEB1-AS1 (ZEB1 antisense 1) was significantly upregulated in 87 ESCC tissues compared to the adjacent noncancerous tissues and was significantly associated with the depth of invasion and lymph node metastasis [145]. In addition, from the Kaplan–Meier survival curves, it was observed that the 5-year overall survival (OS) and disease-free survival (DFS) of ESCC patients with high levels of ZEB1-AS1 were shorter compared with those with low levels of ZEB1-AS1.

Additionally, the dysregulation of lncRNAs ATB [51], XIST [74], AK001796 [93], MALAT1 [147], nuclear transcription factor NF-κB interacting lncRNA (NKILA) [162], SPRY4-IT1 [163], and zinc finger antisense 1 (ZFAS1) [164] has also been demonstrated to be markedly associated with advanced lymph node metastasis, aggressive TNM stage, and shorter survival time. These lncRNAs may also serve as potential prognostic biomarkers for EC.

EC is the eighth most frequently diagnosed malignancy worldwide. Due to typically late diagnoses at the advanced stage, combined with lymphatic metastasis, the prognosis of EC patients is poor. Despite advancements in surgery, chemo- and radiotherapy treatment over the past decades, few encouraging improvements in the 5-year OS rate of EC patients have been achieved. Moreover, the molecular mechanisms underlying EC tumorigenesis and development are still elusive. Hence, a comprehensive understanding of the molecular pathogenesis and identification of potential biomarkers of this disease are urgently needed. It is now recognized that aberrant expression of lncRNAs is a crucial determinant for human cancer. In this review, we have summarized the molecular mechanisms of lncRNAs and how they function in EC by localizing to the chromatin and interacting with proteins and RNAs. Uncovering the underlying mechanisms of lncRNAs may help us to understand the pathogenesis and progress of EC, including cell apoptosis, proliferation, migration, stemness, and therapy resistance. Furthermore, lncRNAs have the potential to serve as promising biomarkers for diagnosing EC and predicting prognosis and relapse, and they may even be novel attractive targets for clinical therapy of EC. However, there remain significant gaps in our understanding of the functions of lncRNAs in EC; these gaps must be bridged before lncRNAs can be used in clinical practice.