Molecular interplay of pro-inflammatory transcription factors and non-coding RNAs in esophageal squamous cell carcinoma

Introduction

Esophageal squamous cell carcinoma (ESCC) originates from epithelial tissues lining the inner esophageal cavity, a tube like muscular tissue connecting pharynx to stomach. Histologically, the innermost layers of the esophagus, which are exposed to food intake, are protected by a non-keratinizing stratified squamous epithelium that undergoes continuous shedding and replacement. This epithelium rests on a basement membrane (the lamina propria) and consists of a proliferating layer of basal cells which differentiate and move upward to replace shed cells. Esophageal metaplasia and dysplasia, characterized by anomalous growth and improper differentiation of squamous epithelium, serve as the pre-cancerous indicators for ESCC. Patients with Barrett’s esophagus, a form of gastroesophageal reflex disease (GERD) characterized by the replacement of esophageal squamous epithelium with glandular epithelium, are at a higher risk of developing ESCC ¹

ESCC is the sixth most common cancer in the developing world and typically progresses aggressively with high metastatic rates. Smoking combined with alcohol abuse alone constitutes more than 90% risk factor for ESCC. ² However, other factors such as diet, aging, obesity, human papilloma virus (HPV) infection, and genetic susceptibility have also been associated with ESCC. ³ Multimodal treatment options for ESCC include surgery, chemotherapy, and radiotherapy; however, tumor recurrence is common. Disease diagnosis typically occurs at late stages of cancer development, at which point the 5-year survival probability may be as low as 4.5% for patients with distant metastases. In 2016 alone, there were 16,910 estimated new cases of ESCC and 15,960 deaths in the United States (http://seer.cancer.gov/statfacts/html/esoph.html). However, ESCC is more common in developing countries, particularly in the “Asian esophageal cancer belt” which extends from northeast China to the middle east, with an annual frequency as high as 100 cases per 100,000. ⁴ In China alone, estimated new incidences and deaths related to esophageal cancer were projected to reach 478,000 and 375,000, respectively, in 2015. ⁵ Hence, this is still one of the deadliest forms of cancer due to its aggressive behavior and high metastatic rates.

A fundamental understanding of molecular pathways driving tumor formation, invasion, and metastasis is essential to the development of new therapeutic strategies for ESCC. Mutations in tumor suppressor genes and pathways commonly mutated in other cancers such as TP53, p16INK4a, p15INK4b, and Notch have also been found in ESCC. ⁶ In addition to intrinsic factors such as genomic lesions, mutations, and chromosomal abnormalities, cancer cells also depend on other factors such as adjacent cell types or non-canonical pathways for their growth and metastasis. One of the classic examples of tumor cell dependency on other cells types is the existence of cancer-associated fibroblasts (CAFs) which secrete growth factors, remodel tumor microenvironment, enhance angiogenesis, and assist in tumor invasion.^7–9 However, inflammation, a physiological response initiated by non-immune and immune cells against injury, stress, infection, and irritants has been tightly linked to cancer progression.^10,11 Furthermore, the development of ESCC is more frequent in patients with Barrett’s esophagus which is caused by chronic inflammation called as reflux esophagitis, underscoring the importance of inflammation in ESCC development. ¹² Numerous transcription factors have been implicated in cancer-associated inflammatory pathways.^13–15 Although an in-depth discussion of all these factors is beyond the scope of this review, we will be focusing on two key mediators of inflammatory pathways in ESCC: nuclear factor kappa B (NF-κB) and signal transducer and activator of transcription 3 (STAT3). We will also examine the role of non-coding RNAs involved in the regulation of these transcription factors.

Non-coding RNAs

Non-coding RNAs are endogenously made RNA transcripts with no protein-coding potential. Based on size, they can be broadly classified into small (<200 nt in length) and long non-coding RNAs (lncRNAs; <200 nt in length). Several small and lncRNAs are primarily involved in house-keeping functions of the cells. Small nuclear RNAs (snRNAs, also known as U-RNAs), with an average size of 150 nt in length, are essential for pre–messenger RNA (mRNA) processing/splicing events while transfer RNAs (tRNAs, 76–90 nt in size) serve as an adaptor molecule linking amino acid sequence of the protein with the triplet codons in the mRNA. Small nucleolar RNAs (snoRNAs), named due to their localization in the nucleolar compartment, guide chemical modifications of snRNAs, tRNAs, and ribosomal RNAs (rRNAs). ¹⁶ However, there are other small RNA populations possessing regulatory functions on host transcriptome. Piwi interacting RNAs (piRNAs) are derived from mobile repetitive elements and mainly guard the genomic integrity in germline cells by targeting transposons and repetitive elements by both epigenetic and post-transcriptional mechanisms. ¹⁷ Endogenous small interfering RNAs (endosiRNAs) are derived from naturally occurring double-stranded RNA (dsRNA) structures within the cells. These dsRNA structures are likely the byproducts of sense and anti-sense transcripts originating from bidirectional transcription or complementary mRNA sequences transcribed from different genomic loci or during viral infections. ¹⁸ In mammals, their key function appears to be in mediating anti-viral response and in silencing transposable elements.^19–21 Ribosomal RNAs (rRNAs), with size ranging from 121 nt to 5 kb, are the most abundant RNA species present in the cells. In ribosomes, rRNAs constitute about 60% of its molecular weight and are crucial for protein synthesis. However, the primary focus of this review will be on the other two non-coding RNA species, microRNAs (miRNAs) and regulatory lncRNAs due to their extensive functional characterization and emerging interest on their role in cancer progression.

MicroRNAs

MicroRNAs, the classic example of small non-coding RNAs, are ~22 nt long single-stranded RNA molecules with a widespread regulatory role in both the development and disease progression. Almost two decades ago, the first miRNA, lin-4 (human homologue: miR-125) which regulates developmental timing in Caenorhabditis elegans by targeting lin-14, was identified. ²² As of June 2014, there are more than 28,000 mature miRNAs documented in a total of 193 species (miRBASE release 21). The annotated numbers of mature miRNAs are 2588 and 1915 in human and mouse, respectively (http://www.mirbase.org/cgi-bin/browse.pl), and it has been predicted that roughly 40% of mRNAs can be targeted by miRNAs. ²³ A vast majority of miRNAs are transcribed by RNA polymerase II as primary miRNA transcripts (pri-miRNA). Pri-miRNAs are several kilobases long and incorporate post-transcriptional modifications such as capping and poly-adenylation identical to those of cellular mRNAs.^24,25 However, Borchert et al. ²⁶ identified miR-517a/b/c as a RNA polymerase III derived miRNA due to their interspersed presence within Alu repeats which also uses RNA polymerase III for their own transcription. The pri-miRNA with the unique stem-loop precursor structure is then recognized by a class II RNAse III type enzyme, Drosha, and dsRNA binding domain protein, DGCR8 complex, to release the hair-pin shaped precursor miRNA (pre-miRNA) of 60 to 110 nt in length with a 2 nt 3′ overhang. ²⁷ Pre-miRNAs are then exported to cytoplasm where they are processed by another RNAse III type enzyme, DICER, to generate duplex double-stranded miRNAs of ~22 bp in length. ²⁸ From the duplex miRNA, through a helicase mediated RNA unwinding, the “guide strand” (mature miRNA) is retained in the RNA-induced silencing complex (RISC) whereas the complementary strand is likely degraded. ²⁹ Deregulation of miRNA biogenesis by alteration in the expression of factors involved in the biogenesis pathway has been noted in many cancers.^30,31 In general, miRNA processing machinery is lower in cancer compared to normal tissues, indicating a decrease in global miRNA levels. ³⁰ However, Sugito et al. ³² reported the overexpression of DROSHA complex in ESCC which leads to poor prognosis. In another study, both DICER and DROSHA have shown decreased expression in ESCC cell line compared to normal cells. ³³ Moreover, DICER expression is positively correlated with radiation sensitization and longer patient survival in ESCC. ³⁴

LncRNAs

In addition to miRNAs, lncRNAs are also considered to be the major part of transcriptional output in cells, in addition to protein-coding genes. As the name suggests, lncRNAs are currently defined as any non-coding RNA greater than 200 nt in size. LncRNAs show post-transcriptional modifications identical to cellular mRNAs such as capping, methylation, and poly-A addition vastly due to the RNA polymerase II mediated transcription. Unlike protein-coding genes, lncRNAs often show very low conservation across the species but a few lncRNAs such as those located within Hox gene clusters do show syntenic conservation in mammals. ³⁵ In recent past, contrary to the definition of lncRNAs, some of them do encode short peptides with a defined function. Given the recent discoveries of micro peptides with functional significance, it will not be surprising if the number of lncRNAs encoding short micro peptides rise in substantial proportions.^36–38 However, compared to protein-coding genes, the number of annotated lncRNAs has increased proportional to the genome size, and based on the recent estimates, there are more than 100,000 annotated lncRNA transcripts and approximately 15,000 experimentally verified lncRNAs across 24 different tissue types (source: www.lincipedia.org).^39,40 Moreover, unlike miRNAs which have more simplistic classification methods, the classification of lncRNAs remains challenging due to the complexity of different types of lncRNAs. For instance, lncRNAs can be classified based on transcript length (long vs very long), association with protein-coding genes (intronic, intergenic, anti-sense), association with other genomic regulatory elements (enhancer, promoter, or upstream), association with subcellular structures (chromatin, nuclear bodies, or polycomb repressive complex 2 (PRC2) associated), and based on structure and function (enhancer lncRNAs, competing endogenous RNAs, or circular RNAs; for detailed reviews, see St Laurent et al. ⁴¹ and Hombach and Kretz ⁴² ). Mechanistic understanding of lncRNAs is currently the focus of many laboratories across the globe, and the studies so far suggest that the function of lncRNAs is to regulate coding genes by several diverse mechanisms. These may include transcriptional interference in cis or trans, impacting the stability of the mRNAs, regulating epigenetic landscape of target genes, and acting as a competing endogenous RNAs for microRNA target genes. ⁴³ In addition, several studies have highlighted their potential regulatory functions in physiological scenarios and involvement in pathogenesis.^44–47 This is not surprising as it is evident from recent literature that many lncRNAs are detected in secreted body fluids serving as potential biomarkers for cancers.^48,49

Pro-inflammatory transcription factors, mechanism, and relevance in esophageal cancer

Inflammation is a natural host response activated in both immune and non-immune cells by a multitude of factors such as stress, injury, and infection. Inflammation is classified as acute or chronic and is mediated by distinct cytokines/chemokines in different cells. Many parallels can be drawn between tumor microenvironment and inflammation in terms of the cytokine secretion profile, immune cells within the tumor milieu, and the extra cellular matrix remodeling. Indeed, chronic wounds which are characterized by pro-inflammatory environments, pose a potential risk for tumor development, and hence, cancer is often referred to as an overhealing wound. ⁵⁰ It is now clear that, in addition to carcinogenic lesions intrinsic to the tumor, inflammation provides a conducive environment for the proliferation, invasion, metastasis, and angiogenesis of cancer cells. ⁵¹ Although diverse transcription factors mediate inflammatory pathways, NF-κB and STAT3 play a pivotal and well-documented role in linking inflammation and cancer.

NF-κB/Rel

About 30 years ago, NF-κB was discovered on a quest to identify protein factor(s) that bind to the octomeric repeats found in the enhancers of immunoglobulins. ⁵² Though initially identified in immune cells, NF-κB is ubiquitously expressed in all tissues. ⁵³ NF-κB is by far, one of the most studied transcription factors with more than 57,000 publications (source: PubMed) starting from its original discovery. NF-κB is a structurally related family of proteins, characterized by the presence of a Rel homology domain at the N-terminus of the protein sequence with DNA binding/dimerization activities.

NF-κB proteins can be broadly subdivided into class I and class II based on the motifs present in the C-terminal half of the protein. Class I comprises NF-κB1 and NF-κB2, otherwise known as p50/p105 and p52/100, respectively. Class I proteins possess a long C-terminal domain consisting of multiple ankyrin repeats which exert an inhibitory effect on their DNA binding activity. Limited proteolysis by 26S proteasome machinery or translational arrest leads to truncated versions of these proteins (p50 from p105 and p52 from p100) which cannot transactivate gene expression on its own. Heterodimerization with class II factors is necessary for their ability to activate gene transcription due to the presence of transactivation domains present in class II NF-κB factors. Class II factors include c-Rel, RelB, and RelA (p65). RelA-p50 is the most predominant form of NF-κB heterodimer found in most cell types though other homodimer/heterodimer combinations have also been reported.

Canonically, the NF-κB signaling pathway is regulated by inhibitory kappa B (IκB) proteins. There are several IκB proteins (IκB α, IκB β, IκB γ, and IκB ε) which interact with NF-κB with different affinities and show tissue specific expression. This suggests that the regulation of NF-κB is under stringent spatial and temporal controls. ⁵⁴ In most cell types, NF-κB is retained in the cytoplasm as an inactive form by its interaction with IκB. Activation of NF-κB signaling is initiated by phosphorylation of IκB subunit by a specific serine-specific kinase known as IκB kinase (IKK). The IKK complex consists of two catalytic subunits, IKKα and IKKβ, and the sensing scaffold protein IKKγ. IKKγ is also known as NF-κB essential modulator (NEMO) and relays upstream signaling events to the catalytic subunits. ⁵⁵ In the canonical NF-κB activation pathway, pro-inflammatory ligands such as tumor necrosis factor alpha (TNFα) or interleukin 1 beta (IL-1β) or lipopolysaccharide (LPS; via Toll like receptors) lead to the recruitment and activation of IKKs which inactivate IκB by phosphorylation. IκB phosphorylation triggers the dissociation of IκB from the NF-κB subunit followed by the proteasomal degradation. This leads to the translocation of active NF-κB dimer into the nucleus resulting in transcriptional activation of target genes (Figure 1).

OPEN IN VIEWER

Figure 1.

Schema showing simplistic view of NF-κB and STAT3 signaling pathways. Non-canonical NF-κB pathway is not shown here due to its limited relevance to ESCC. In NF-κB canonical pathway, ligand activation either by viral/bacterial infection or pro-inflammatory stimulus leads to activation of IκB kinase. IκB kinase in turn phosphorylates the IκB subunit of NF-κB complex causing the dissociation of IκB subunit from NF-κB complex, followed by poly-ubiquitination and proteasomal degradation. In the absence of IκB subunit, the p50-RelA complex translocates to the nucleus to turn on transcriptional activation of target genes, which includes IκB and miR-21 which forms a negative and positive feedback loop for NF-κB pathway, respectively. In addition, the transcriptional targets of NF-κB also include IL-6, which is a major upstream activator of JAK–STAT3 pathway. Mechanistically, STAT3 activation is simpler than NF-κB pathway. Ligand (IL-6 and EGF) binding to their target receptors leads to the activation of receptor-associated tyrosine kinase, JAK1, which phosphorylates latent monomeric STAT3 present in cytoplasm at residue Tyr 705. Phosphorylation at Tyr 705 leads to dimerization of STAT3 monomers and their subsequent translocation to the nucleus and transcriptional activation of target genes. Notably, miR-21 is a bona fide transcriptional target of STAT3 in epithelial cells suggesting key molecular link between NF-κB and STAT3 pathways.

Non-canonical NF-κB activation, a process seen in B and T cell organ development, is independent of IκB. In this pathway, an entirely different set of ligands (such as CD40 or B cell–activating factor (BAFF)) activate NF-κB inducing kinase (NIK) which in turn phosphorylates IKKα dimer without the NEMO subunits. The phosphorylated IKKα phosphorylates the two serine residues C-terminus to the ankyrin repeats in p100 leading to the liberation of active p52/RelA dimer which then shuttles to the nucleus for target gene activation. One notable fact about NF-κB pathway is that IκB is itself one of the transcriptional targets of NF-κB. Hence, the activation of NF-κB pathway also activates IκB transcription, a negative regulator of the same, leading to a feedback loop. Hence, after a transient NF-κB activation, nascent IκB synthesized by this activation, enters the nucleus, binds to NF-κB subunits, and transports the complex back to the cytoplasm to its latent stage, ultimately shutting this pathway.

Constitutive activation of NF-κB is often seen in many epithelial cancers including ESCC. ⁵⁶ This is most likely due to the fact that NF-κB has been implicated in all the essential steps/processes that drive efficient tumorigenesis. ⁵⁷ NF-κB has more than hundreds of validated transcriptional targets (http://www.bu.edu/nf-kb/gene-resources/target-genes/), out of which some of them are key players in determining the ability of cancer cells to indefinitely proliferate (cMyc and cyclin D1), inhibit apoptotic signals (ciAPS, c-FLP, and members of the bcl-2 family), migrate (IL-8), undergo angiogenesis (vascular endothelial growth factor (VEGF)) and undergo invasion/metastasis (matrix metalloproteinase (MMP)-2/9)).^56,58 Tumor cells achieve constitutive activation of NF-κB by mutations in genes encoding the NF-κB transcription factors themselves or in genes that control NF-κB activity (such as IκB genes). Alternatively, the tumor cells can overexpress growth factors/inflammatory ligands which can induce NF-κB activation. However, the involvement of NF-κB in cancer is by no means resolved. Contrasting evidence indicates a tumor-suppressive role of NF-κB in certain cancers through transcriptional activation of Fas ligand. ⁵⁹ Currently, there is a considerable focus on targeting inflammatory signaling pathways involved in cancer invasion and metastasis. NF-κB is one of the prime target molecules in this scenario.

Non-coding RNAs regulating/regulated by NF-κB pathway in ESCC

MicroRNAs may have a profound effect on the NF-κB signaling pathway and therefore on its effects in cancer (Figure 2). The list of microRNAs that are regulated by NF-κB is on a steady rise though very few miRNAs are implicated in regulating or regulated by NF-κB pathway in ESCC.^60,61 NF-κB has been shown to transcriptionally activate oncogenic miRNAs while some microRNAs are known to repress NF-κB either directly or through targeting of the regulatory factors involved in the NF-κB pathway. Gong et al. ⁶² reported the role of miR-138, a tumor suppressor miRNA downregulated in ESCC, which inhibits NF-κB signaling via suppressing poly-ubiquitination of TNF receptor–associated factor 2 (TRAF2) and receptor-interacting protein 1 (RIP1) which are negative regulators of NF-κB signaling. Besides, an inverse correlation of miR-138 levels and NF-κB expression in ESCC patient samples provides further support for this regulation. Let-7, a tumor suppressor miRNA downregulated in multiple forms of cancer, including ESCC, ⁶³ downregulates Ras and IL-6, well-known oncogenes which are upstream activators of the NF-κB pathway. ⁶⁴ Other miRNAs may have pro-oncogenic effects through stimulation of NF-κB signaling. For example, miR-21 belongs to a class of miRNAs known as oncogenic miRNAs (oncomiRs) which are widely known to be upregulated in multiple cancers including ESCC.^65,66 Notably, miR-21 promotes NF-κB signaling by targeting phosphatase and tensin homolog (PTEN), thereby enhancing the Akt signaling pathway, an upstream activator of NF-κB signaling. In a retrospective study with ESCC patients, Bahmanyar et al. ⁶⁷ showed an 80% increased incidence of ESCC in patients with gastric ulcer caused by Helicobacter pylori infection. H. pylori infection also leads to the upregulation of miR-155, which shows positive correlation with NF-κB activation in mouse models of hepatocarcinogenesis. ⁶⁸ In support of this, elevated plasma miR-155 has been shown as a biomarker for early diagnosis of ESCC ⁶⁹ highlighting H. pylori infection as another upstream event in NF-κB activation through miR-155.

OPEN IN VIEWER

Figure 2.

Schematic shows the molecular interplay between NF-κB and STAT3 which is regulated by/regulating the non-coding RNAs pertaining to ESCC. Upregulated non-coding RNAs are depicted in black against a gray background, while the ones which are downregulated are depicted in white against dark gray background in ESCC. Solid lines indicate that the known regulation of this pathway in ESCC while dashed lines suggest that the indicated relationship may/may not exist in ESCC. Long non-coding RNAs are shown in rounded rectangle shapes.

Likewise, the NF-κB pathway has the capability to influence transcription of miRNAs relevant to carcinogenesis. The role of NF-κB in miR-146a transcription is a classic example of a negative feedback loop between a miRNA and transcription factor. In monocytic cells, LPS-induced activation of NF-κB pathway leads to recruitment of active NF-κB dimer to the miR-146 promoter leading to rapid induction of both miR-146a and miR-146b. ⁷⁰ However, two key modulators of NF-κB signaling pathway, TRAF6 and IRAK1 (interleukin-1 receptor-associated kinase 1) were found to be the direct target of miR-146, indicating a negative regulatory loop. This observation is reflected by the global downregulation of miR-146 in ESCC tissues compared to tumor adjacent tissues, likely due to the effect of this negative feedback loop. ⁷¹

Even though, the oncogenic role of NF-κB is well known, NF-κB has also shown to activate miRNAs that possess both tumor-suppressive and tumor-promoting roles in a context-dependent manner. For instance, miR-31 has been shown to be a direct target of NF-κB in skin keratinocytes leading to hyperproliferation in psoriatic epidermis. ⁷² In agreement with this, miR-31 was found to be one of the significantly upregulated miRNAs in esophageal hyperplasia induced by zinc deficiency. ⁷³ Furthermore, serine threonine kinase 40 (STK40), a key negative regulator of NF-κB signaling, ⁷⁴ was shown to be the direct target of miR-31 suggesting a feed forward mechanism by which NF-κB and miR-31 cooperatively promote ESCC proliferation. However, miR-31 also exerts a p21-dependent tumor-suppressive role in ESCC cell lines independent of p53, ⁷⁵ and overexpression of miR-31 in ESCC cell lines leads to decreased migration, proliferation, and invasion highlighting a dual role of NF-κB → miR-31 axis in regulating ESCC progression. ⁷⁶ In another finding, miR-34a, a well-known tumor suppressor activated by p53 in many cancers was shown to have NF-κB binding at its promoter. Moreover, in ESCC cell lines, overexpression of NF-κB was shown to induce its transcription. ⁷⁷ In a screen to identify whether miRNAs are involved in radio resistance of ESCC, Su et al. ⁷⁸ reported the downregulation of miR-301a in radioresistant compared to radiosensitive cell line derived from same parental host. Notably, miR-301a potentially activates NF-κB pathway, by targeting NF-κB repressing factor (NKRF) in pancreatic adenocarcinoma. ⁷⁹ In an independent study, upregulation of miR-98 was shown to confer radiosensitivity to ESCC cell line and NF-κB was predicted as one of the targets of miR-98. ⁸⁰ The reason for discrepancy between the above studies in addressing whether NF-κB is truly a tumor promoter or suppressor is unclear, but it might highly be due to context-dependent effect of NF-κB signaling in tumor progression, though this requires further investigation. ⁸¹ Taken together, the link between miRNAs and NF-κB pathway appears to reveal a complex relationship in ESCC tumorigenesis (Table 1).

OPEN IN VIEWER

Table 1.

List of small and long non-coding RNAs regulating/regulated by NF-κB pathway.

MiRNA/lncRNA	Target of NF-κB	Effect on NF-κB	Target	Up/down in ESCC	References
MiR-138	No	Inhibition	Unknown	Decrease	82
Let-7	No	Inhibition	Ras/IL-6	Decrease	63, 64
MiR-21	Yes	Activation	PTEN	Increase	64–66, 83
MiR-155	No	Activation	C/EBPbeta	Increase a	68, 69
MiR-146	Yes	Activation	TRAF6/IRAK	Decrease	70, 71
MiR-181b	Yes	Activation	CYLD	Increase	84, 85
MiR-31	Yes	Activation	STK40	Unknown	73, 75
MiR-34a	Yes	Unknown	Unknown	Unknown	77
MiR-301	No	Activation	NKRF	Decrease	79
MiR-98	No	Inhibition	NF-κB	Unknown	80
HOTAIR	No	Activation	NA	Increase	46, 86, 87
NKILA	Yes	Inhibition	NA	Unknown	88

NF-κB: nuclear factor kappa B; MiRNA: microRNA; lncRNA: long non-coding RNA; ESCC: esophageal squamous cell carcinoma; PTEN: phosphatase and tensin homolog; C/EBPbeta: CCAAT-enhancer-binding proteins; TRAF6: tumor necrosis factor receptor–associated factor 2; IRAK1: interleukin-1 receptor-associated kinase 1; STK40: serine/threonine kinase 40; NKRF: NF-κB repressing factor; NKILA: NF-κB interacting lncRNA; NA: Not available.

a

In patients’ plasma samples.

The advent of next-generation sequencing methods at affordable costs led to a paradigm shift in research focus from coding genes to lncRNAs in human physiology and pathology. In this line, many lncRNAs have been shown as excellent biomarkers for ESCC, though the mechanistic details on how these lncRNAs are regulated or how they regulate the coding genome is an active area of investigation. ⁸⁹ HOTAIR, named for Hox intergenic anti-sense RNA and originally identified in the context of limb development, was later shown to be involved in both breast cancer as well as ESCC.^45,46,86 In ESCC, HOTAIR promotes metastasis by epigenetic modulation of target genes in vivo, and its expression is correlated with poor survival. Given the recent study demonstrating the activation of the NF-κB pathway by HOTAIR in ovarian cancer chemo resistance, it is tempting to speculate a similar mechanistic link between NF-κB and HOTAIR in ESCC. ⁸⁷ In addition to the upstream regulation of NF-κB by lncRNAs, the study by Liu et al. ⁸⁸ profiled the non-coding interacting partners of this complex in cancer. This led to the identification of NF-κB interacting lncRNA (NKILA) which is upregulated by NF-κB, binds to NF-κB/IκB complex directly inhibiting IKK induced phosphorylation and activation of NF-κB pathway. Low NKILA expression was also correlated with breast cancer metastasis and poor patient prognosis. However, the role of NKILA in ESCC and its relevance in NF-κB activation in ESCC are yet to be determined.

STAT3

The STAT family consists of seven protein members including STAT1, STAT2, STAT3, STAT4, STAT5 a/b, and STAT6. STAT3 was originally identified during an attempt to find the intermediate proteins which transduced/relayed growth factor (epidermal growth factor (EGF)) and cytokine (IL-6) signaling for the activation of gene transcription in the nucleus.^90,91 Originally, named as acute-phase response factor, STAT3 consists of three different isoforms differing in molecular weight from 722 to 770 amino acids in length. Like other STAT proteins, STAT3 exists in a latent inactive form in the cytoplasm. Binding of IL-6 and EGF ligands to their target receptors, leads to the activation of receptor–associated tyrosine kinases widely known as Janus kinases (JAKs), most notably JAK1. JAK1 phosphorylates STAT3 at residue Tyr 705. STAT3 phosphorylation is seen within minutes after ligand binding and reaches its maximum phosphorylation between 15 and 60 min after binding, following which it gradually decreases. Phosphorylation at Tyr 705 leads to dimerization of STAT3 monomers and their subsequent translocation into the nucleus and transcriptional activation of target genes.

Multiple studies have highlighted STAT3 overexpression in ESCC. Timme et al. ⁹² reported abundant expression of cytoplasmic STAT3 and nuclear phospho-STAT3 in ESCC tissue sections, and siRNA knockdown of STAT3 leads to reduced proliferation of ESCC cells. The increased expression of STAT3 may also result from activation of the Wnt/β-catenin signaling pathway as STAT3 harbors transcription factor (TCF)-binding element within its promoter. ⁹³ IL-6, one of the key upstream activator of STAT3 phosphorylation was also found to be overexpressed in ESCC specimens compared to adjacent non-malignant epithelium. In addition, serum IL-6 levels also correlated with the development of distant metastases and lower treatment response due to persistent activation of STAT3.^94,95 STAT3 can also be activated by EGF, a growth factor signaling often amplified or overexpressed in ESCC. ⁹² The importance of STAT3 signaling in ESCC is exemplified by its known targets which affects multiple facets of cancer progression, such as, cell proliferation and anti-apoptosis (Oct-1, c-myc, Bcl-2, and Bcl-XL), angiogenesis (VEGF), immune evasion (CXCL10 inhibition), and invasion (MMP-2).^96,97 STAT3 inhibitors, either in the form of JAK1/2 inhibitors or STAT3 dimerization blockers, are currently undergoing clinical trials for cancer. ⁹⁸

NF-κB and STAT3 interplay in ESCC

STAT3 and NF-κB are accomplices with respect to tumor progression and metastasis in ESCC. Both proteins have many targets in common and are known to interact with each other. The physical interaction of STAT3 with the active NF-κB complex in the nucleus precludes the binding of IκB, leading to perpetual maintenance of NF-κB signaling. However, IL-6, a major inducer of STAT3 phosphorylation, is directly bound and activated by NF-κB suggesting a feed forward positive loop between STAT3 and NF-κB in cancer. ⁹⁹ In addition, miR-21 which promotes tumorigenesis through NF-κB activation, is a direct target of IL-6 through STAT3^83,84 Recently, in MCF10A cells, NF-κB was shown to enhance miR-181b levels by increasing IL-6 transcription, which in turn activates STAT3 phosphorylation leading to its recruitment in miR-181b promoter and transcriptional activation. ⁸⁴ MiR-181b directly binds to 3′-untranslated region (3′-UTR) and represses the translation of CYLD, a tumor suppressor and deubiquitinating enzyme known to inhibit NF-κB signaling forming a positive feedback loop (NF-κB → IL-6 → STAT3 → miR-181b → CYLD → NF-κB) leading to persistent activation of this pathway. In ESCC, this reciprocal activation was recently shown to regulate the proliferation of esophageal cancer stem–like cells ⁸⁵ (Figure 2).

Non-coding RNAs regulating/regulated by STAT3 pathway in ESCC

As with NF-κB, miRNAs are key regulators of the STAT3 pathway in ESCC (Figure 2). The let-7 family of microRNAs are known tumor suppressors in several cancers. Lower levels of let-7 are associated with poor prognosis and let-7b and let-7c have been shown to be suppressed in ESCC. ¹⁰⁰ The downregulation of let-7 has a direct impact on STAT3 activation. Let-7c overexpression prevents cisplatin-induced upregulation of STAT3 phosphorylation due to the direct targeting of IL-6 mRNA by let-7c in ESCC cells. This indicates that let-7 modulates ESCC chemosensitivity by suppressing IL-6/STAT3 activation in ESCC. ⁶³ It is worth noting that STAT3 exerts a negative impact on let-7c by activating the transcription of Lin-28, a protein which precludes the processing of pre-let-7, thereby preventing mature let-7 biogenesis. ¹⁰¹ MiR-143 and miR-124 also fall under the same category of miRNAs which negatively regulate STAT3 levels in ESCC. Liu et al. ⁸² showed lower expression of miR-143 in ESCC patient sections and miR-143 expression is inversely correlated with lymph node metastasis. STAT3 was found to be a direct target of miR-143, and miR-143 overexpression significantly affected cell proliferation as well as epithelial–mesenchymal transition (EMT) pathways in ESCC cell lines. Similar results were reported by another group for miR-124, which binds to STAT3 and suppresses its translation. Ectopic expression of miR-124 led to the inhibition of cell proliferation through a block in the G1/S phase transition, eventually leading to apoptosis.^102,103 Given the pivotal role of STAT3 in tumorigenesis and cancer progression, it is evident that the transition from normal cells to cancer cells involves the suppression of miRNAs that negatively regulate STAT3 signaling.

However, being an oncogenic activator, STAT3 is known to induce the transcription of various oncomiRs such as miR-21 and miR-181b in order to potentiate a positive feedback loop between NF-κB and STAT3, as discussed earlier. ⁸⁴ MiR-17-92 cluster, alternatively known as oncomiR-1, is one of the well-studied miRNA clusters. Six oncogenic miRNAs (miR-17, miR-18a, miR-19a, miR-20a, miR-19b-1, and miR-92a-1) are derived from this cluster coding for single primary transcript of about 800 nt in size. ¹⁰⁴ All six miRNA members of this cluster were recently shown to be upregulated in ESCC. Notably, each member of this cluster had distinct positive correlation with different aspects of ESCC (miR-18a with tumor stage, miR-92a with clinical stage, miR-19b and miR-17a with lymph node metastasis). ¹⁰⁵ Given that the miR-17-92 cluster is regulated by IL-6/STAT3 signaling in endothelial cells through a conserved STAT3 binding site in its promoter, it is tempting to speculate a similar activation of these oncomiRs by STAT3 in ESCC. ¹⁰⁶ Even though the list of miRNAs targeted or targeting IL-6/STAT3 signaling is expanding currently, very few miRNAs have been linked with STAT3 in ESCC,108 implying that there is more to come in future.

Several lncRNAs are currently known to regulate/regulated by IL-6/STAT3. OLAIP2 blocks phosphorylated STAT3 homodimer formation, ¹⁰⁷ while lnc-DC promotes STAT3 phosphorylation. ¹⁰⁹ Lnc-TCF7 is induced by IL-6/STAT3 and promotes EMT in hepatocellular carcinoma. ¹¹⁰ However, none of these has been implicated in ESCC. Significant attention is currently paid to the identification of novel lncRNAs in ESCC, and the link between these lncRNAs and IL-6/STAT3 pathways, if any, has yet to be elucidated (Table 2). ¹¹¹

OPEN IN VIEWER

Table 2.

List of small and long non-coding RNAs regulating/regulated by STAT3 pathway.

MiRNA/lncRNA	Target of STAT3	Effect on STAT3	Target	Up/down in ESCC	References
MiR-143	No	Inhibition	IL-6	Decrease	82
Let-7	No	Inhibition	STAT3	Decrease	63, 64, 101
MiR-124	No	Inhibition	STAT3/CDK4	Decrease	102, 103
MiR-21	Yes	Activation	PTEN	Increase	64–66, 83
MiR-181b	Yes	Activation	CYLD	Increase	84, 85
MiR-17-92	Yes	Activation	BMPR2	Increase	104–106
OLAIP2	No	Inhibition	NA	Unknown	107
Lnc-DC	No	Activation	NA	Unknown	109
Lnc-TCF7	Yes	Unknown	NA	Unknown	110

STAT3: signal transducer and activator of transcription 3; MiRNA: microRNA; lncRNA: long non-coding RNA; ESCC: esophageal squamous cell carcinoma; IL-6: interleukin 6; CDK4: cyclin-dependent kinase 4; PTEN: phosphatase and tensin homolog; BMPR2: bone morphogenetic protein receptor type II; NA: Not available.

Conclusion

This review summarizes the current state of knowledge on the molecular interaction between the pro-inflammatory transcription factors, NF-κB and STAT3, and non-coding RNAs in ESCC. In summary, both NF-κB and STAT3 are regulated by a diverse array of oncomiRs/tumor suppressor microRNAs. MicroRNAs achieve this feat by directly targeting the key components of the pathway or by altering the regulators of this pathway. However, NF-κB and STAT3 exploit non-coding RNAs as a means to execute their function, especially in ESCC proliferation, invasion, and metastasis. Even though discovered only a couple of decades ago, non-coding RNAs are already in the pharmaceutical race as candidate therapeutic targets for treating cancer. MiR-34a, a tumor suppressor miRNA, marketed by miRNA therapeutics as MRX34 is currently under clinical trials for treating breast cancer patients. ¹¹² LncRNAs have shown significant association with cancer prognosis and are currently proposed to be potential biomarkers for predicting the clinical outcome for multiple cancers. ⁴⁸ However, the available literature that is summarized in this review represents only a fraction of non-coding RNAs that are currently investigated for their potential in clinical therapeutics. ¹¹³ More extensive work is required to unveil upstream regulators and downstream effector non-coding RNAs of NF-κB/STAT3 pathway, in order to develop alternative treatment options for ESCC.

Discussion and future perspectives

Despite the wide scientific interest in elucidating the function of non-coding RNAs in ESCC, we still do not have a holistic picture of the entire non-coding RNAs spectrum (both short and long) regulated by pro-inflammatory pathways. Mechanistically, many more questions remain unanswered with reference to the link between NF-κB/STAT3 and non-coding RNAs. In addition to transcriptional regulation, many miRNAs are regulated post-transcriptionally either at the interface between pri-miRNA and pre-miRNA or between pre-miRNA and mature miRNA. For instance, SMAD4, upon activation of transforming growth factor beta (TGFβ)/bone morphogenetic protein (BMP) signaling, binds to the DROSHA–associated RNA helicase p68 and is recruited to pri-miR-21 and controls the processing of pre-miR-21. ¹¹⁴ Available evidence thus far indicates that both STAT3 and NF-κB can regulate miRNA biogenesis by controlling the transcription of the regulators of miRNA processing. STAT3 promotes the transcription of lin-28, a post-transcriptional inhibitor of Let-7 miRNA biogenesis, while NF-κB was recently shown to activate the transcription of DICER.^87,101 However, it is still unclear whether STAT3/NF-κB physically interacts with the players of the miRNA biogenesis pathway. The second key question is what are the long non-coding RNAs that are transcriptionally controlled by NF-κB/STAT3? Currently, there are no reports which have investigated the NF-κB/STAT3 binding sites in the non-coding portion of genome either computationally or by chromatin immunoprecipitation (ChIP)-seq analysis. Nevertheless, in a recent report examining the lncRNA expression profiles in B cell lymphoma, the authors reported a considerable overlap between NF-κB/STAT3 ChIP-seq peaks with some of the lncRNAs expressed in their dataset indicating an unknown territory of lncRNAs controlled by NF-κB/STAT3 in cancer. ¹¹⁵ Conceivably, the future research will shed light on to these queries.

From a therapeutic viewpoint, as the role of NF-κB and STAT3 pathways in promoting ESCC pathogenesis is evident, targeting these transcription factors or the upstream activators of these pathways is becoming an attractive option in treating ESCC. Despite this fact, targeting these pathways is fraught with potential side effects due to the broad-spectrum role of these transcription factors in both development and human physiology. ¹¹⁶ Particularly, STAT3 is also involved in protection against myocardial ischemia/reperfusion injury, liver injury, and obesity, ⁹⁸ while administration of NF-κB inhibitors can lead to long-term immunosuppression or excessive inflammasome activation induced by IL-1β.^116,117 Due to the spatio-temporal complexity of NF-κB and STAT3 signaling, targeting the upstream activators of these pathways also poses some challenges. TNFα antagonists used for treating rheumatoid arthritis and inflammatory bowel disease to suppress inflammation have been shown to result in malignancy in many cases. ¹¹⁸ TNFα can activate both NF-κB and STAT3, directly or indirectly, and the above observation might initially look counterintuitive. However, TNFα can either induce apoptosis or NF-κB activation depending on the cell type, environment, and dose; ¹¹⁸ and hence, using NF-κB/STAT3 pathway as therapeutic targets may result in unpredictable outcomes. Significant progress has been made in the field of targeting non-coding RNAs in vivo. MicroRNAs mimics or inhibitors can be efficiently modified (locked nucleic acids, 2’-O-methoxy, phosphorothioate) to improve intracellular delivery and stability in vivo. ¹¹⁹ Similar approaches are exploited to target nuclear localized long non-coding RNAs, which are difficult to target with conventional siRNAs. ¹²⁰ Under these circumstances, considering non-coding RNAs that are linked to these pathways as therapeutic targets represents a potential and novel approach which requires future mechanistic studies for validation.