1. Introduction
Long non-coding RNAs (LncRNAs) are broadly defined as RNAs that usually do not encode for proteins and that are longer than 200 nucleotides. These are messenger RNA (mRNA)-like molecules that are transcribed by polymerase II, 5’ capped, and have a 3’ poly-A tail. Because they can form complex secondary structures, they often have functions. Many LncRNAs are preferentially found in the nucleus, where they participate in regulation of chromatin organization and transcription, as well as formation of nuclear speckles and regulation of splicing. In the cytoplasm, LncRNAs can regulate mRNA stability, bind to other non-coding RNAs, modulate protein post-translational modifications and protein function [
1,
2,
3,
4,
5].
LncRNAs have been studied in mammals since the early 90’s due to their involvement in developmental processes. For example, the
Xist (X-inactive specific transcript) lncRNA contributes to reshaping chromatin architecture to achieve X chromosome silencing in early embryonic development [
6]. The
H19 LncRNA is involved in genomic imprinting and regulation of the insulin growth factor 2 (
IGF2) and other genes involved in embryonic growth [
7,
8]. The study of
HOX genes, master regulators of embryonic development, led to the discovery of the LncRNA
HOTAIR (Homeobox transcript antisense intergenic RNA).
HOTAIR is transcribed from the antisense strand of the
HOXC gene.
HOTAIR has been reported to repress transcription of the
HOXD loci via interaction with PRC2 (polycomb repressive complex 2) [
9], although
in vivo models developed later report conflicting results regarding
HOXC or
HOXD genes regulation by
HOTAIR [
10,
11]. Because of the importance of physiological context to understand LncRNA molecular function, more recently other groups report exclusively
in vivo approaches using animal models of LncRNA genetic ablation. For example, a comprehensive study developed 18 knock-out (KO) mouse models for less well-known LncRNAs with human orthologs. Certain LncRNAs expression was highly tissue-specific (such as
Fendrr,
Manr and
linc–Cox2 expressed mainly in lung), supporting a unique physiological role, while other LncRNAs were more ubiquitously expressed. Three of the analyzed LncRNAs were required for embryonic development (
Fendrr, Mdgt, Peril), speaking to their fundamental functions [
12]. Thus, the involvement of LncRNAs in normal cell physiology and organism development suggests that these may also control disease-related processes such as cancer.
Lung cancer is the leading cause of cancer-related mortality in the U.S., and non-small cell lung cancer (NSCLC) is the most common subtype. Lung cancer is commonly diagnosed in late stages, where patients present distant metastasis with 9% of 5-year survival rate [
13]. Although the development of targeted therapies (e.g., tyrosine kinase inhibitors, TKIs) has improved patient outcomes, their clinical efficacy is often limited by both innate and acquired resistance, permitting tumor progression and recurrence leading to poor survival rates [
14]. It is imperative to better understand molecular drivers of tumorigenesis, metastasis, and therapy resistance in lung cancer to develop improved therapeutic strategies. LncRNAs are emerging as important molecules in cancer due to their oncogenic or tumor suppressive functions. Here, we summarize latest works providing
in vivo evidence available for the role of lncRNAs in lung cancer, therapy resistance, and their potential as biomarkers.
2. Role of LncRNAs in lung cancer
Because of the role of LncRNAs in regulating a diverse array of cellular functions, deregulation of their activities is involved in cancer. Some mechanisms reported for LncRNAs functions in cancer are: acting as miRNA sponges to modulate activity on their targets; interacting with histone-modifier enzymes to modulate known oncogene/tumor suppressor gene expression; interacting with transcription factors to repress/activate their transcriptional programs; acting as anti-sense molecules for tumor suppressor mRNAs, among other mechanisms [
15]. Importantly, regulation of LncRNA expression and function in cancer follows similar principles to that of known oncogenes and tumor suppressors, as it can be mediated by DNA methylation [
16], amplification or deletion [
17], and mutation or SNPs of DNA sequences [
18,
19]. In this section, we provide examples of well-studied LncRNAs that are reported to have oncogenic or tumor suppressive roles in lung cancer, as well as those with controversial functions.
MALAT1 (Metastasis-Associated Lung Adenocarcinoma Transcript 1) was one of the first lncRNAs described to be associated with cancer. In 2003, Ji
et al analyzed the gene expression profile in human primary lung cancer tumors that subsequently metastasized or those that did not metastasize and compared their transcriptional signatures. They identified a Metastasis-Associated Lung Adenocarcinoma Transcript 1, named
MALAT1, for its higher expression in primary tumors that metastasized. They also found a significant correlation between higher level of
MALAT1 expression in stage I lung cancer and worse survival outcomes [
20]. Several studies since then have described
MALAT1 function in normal physiology and cancer [
21,
22,
23]. Three independent groups developed an
in vivo approach to describing
MALAT1 physiological function. They found that
Malat1 is highly abundant in several mouse tissues and highly conserved across species. Genetic perturbation of the
Malat1 locus in mice (by genetic deletion [
24,
25] or genetic inactivation approaches [
26]) did not alter animal development, did not alter nuclear speckle formation, splicing, or mRNA stability. However, they described a role for
Malat1 in controlling neighboring genes expression in a tissue-specific manner that was not consistent between the three studies, especially that of
Neat1, another LncRNA [
24,
26]. In the context of lung cancer,
MALAT1 silencing did not show effects on lung cancer cell proliferation or viability
in vitro [
25]. To further understand the role of
MALAT1 in lung cancer metastasis, Gutschner
et al implanted EBC-1 lung cancer cells into nude mice and treated them with subcutaneous administration of an anti-sense oligonucleotide (ASO) targeting
MALAT1. After five weeks of treatment, all primary tumors were excised. Metastasis nodules were analyzed at 12 weeks, indicating fewer and smaller metastatic nodules in the treated group, suggesting a role for
MALAT1 in promoting metastasis. Through additional
in vitro studies, they report
MALAT1 inhibition results in aberrant expression of metastasis-associated genes in cell lines [
27]. Although these studies support a role for
MALAT1 in promoting metastasis and regulating certain genes expression, other evidence exists for different roles. Kim
et al describe an elegant study that challenges previously reported roles for
MALAT1 in metastasis. Using the
Malat1 knock-out (KO) mouse model developed by Nakagawa
et al (LacZ and Poly-A sequences were used as transcriptional terminators inserted 69 bp downstream of the transcription start site of
Malat1 without deletion of the DNA sequence), Kim
et al crossed these mice with a breast cancer model driven by MMTV-PyMT that mimics human disease. Surprisingly, they found a 7.2-fold increase in metastatic foci and 31-fold increase in percent of lung areas with metastatic lesions in
Malat1-KO mice as compared to
Malat1 WT mice, suggesting a role for
Malat1 in suppressing breast cancer metastasis to the lung. This phenotype was rescued by transgenic expression of
Malat1, suggesting that the RNA product itself diminished metastasis. Additionally, they identified
Malat1 interaction with TEAD proteins in mouse-derived tumors and cell lines, which suppressed TEAD-YAP interaction and, therefore, inhibited their pro-metastatic transcriptional program [
28]. This study and others highlight the importance of context (lung model vs. breast model) as well as experimental methodology to approach LncRNAs functional characterization (such as LncRNA genomic DNA loss vs. RNA loss reviewed in detail elsewhere [
21]). Of note, DNA elements themselves within a LncRNA locus may be responsible for regulatory functions that are independent from transcript function [
29,
30,
31,
32]. In summary,
MALAT1 promotes metastasis in lung cancer, but may show opposite functions in different types of cancer depending on cellular context.
The
GAS5 (growth arrest-specific 5) gene was first described as a G0-specific gene that is inhibited by serum and growth factors [
33].
In vivo,
Gas5 genetic deletion (
Gas5+/-) in mice decreased bone mass and impaired bone repair leading to osteoporosis. Mechanistically,
GAS5 positively influenced proper cell differentiation through interaction with UPF1 (a DNA/RNA helicase) to accelerate
SMAD7 mRNA decay [
34]. In lung cancer,
GAS5 was found downregulated in 72 NSCLC tumor samples as compared to their paired adjacent normal tissues, suggesting a tumor-suppressive role. Additionally, low
GAS5 expression was correlated with larger tumor size, lower differentiation levels, and higher staging of tumor-node metastasis [
35]. A xenograft model of
GAS5 overexpression (OE) showed that
GAS5 OE markedly decreased tumor size compared with control [
35]. Although this study did not explore a mechanism for
GAS5-mediated tumor suppression, other studies implicate a role for
GAS5 as miRNA sponge, to negatively influence cell cycle activator genes [
36] or positively influence
PTEN levels [
37]. Additionally, a recent study revealed
GAS5 is partially localized to the mitochondria where it modulates energy homeostasis by promoting de-acetylation of malate dehydrogenase, suppressing breast cancer [
38]. Taken together, a role for
GAS5 in halting the cell cycle as well as promoting cell differentiation in normal cells support its tumor suppressive role reported in lung cancer. Besides
GAS5, other LncRNAs have been studied for their tumor suppressive functions, such as
MEG3 and
TUG1, reviewed elsewhere [
39].
LUCAT1 (Lung Cancer Associated Transcript 1), first identified as smoke-induced and cancer associated LncRNA1 (
SCAL1) [
40], has higher expression in lung cancer as compared to normal controls, and is also found overexpressed in several cancer types [
41]. To our knowledge, a
LUCAT1-deficient mouse model has not been reported. Additionally, patients with tumors that express high levels of
LUCAT1 showed poorer overall survival as compared to those with lower
LUCAT1 expression. Moreover, high
LUCAT1 levels were associated with late staging in tumor-lymph node metastasis and higher tumor volume. In NSCLC cell lines A549 and SPC-A1,
LUCAT1 modulates
p21 and
p57 expression by promoting loci methylation through PRC2 [
42].
More recently, a role for
LUCAT1 in regulating immune responses has been described.
LUCAT1 genetic deletion in myeloid cells is found to enhance interferon-mediated gene transcription.
LUCAT1 acts as an immune suppressor by interacting with STAT1 and chromatin in the nucleus. It may also act by inhibiting NF-kB functions [
43]. These findings suggest a tumor-promoting role for
LUCAT1 that is tumor cell-intrinsic, in addition to a potential non cell-autonomous mechanism by inhibition of immune surveillance, although this mechanism remains to be explored in the lung cancer context.
HOTAIR has been vastly studied in cancer contexts [
44,
45]. In NSCLC, tumor samples and cell lines expressed higher
HOTAIR levels as compared to normal counterparts [
46]. Additionally, high
HOTAIR levels correlated with higher tumor grade and presence of lymph node metastases [
46,
47].
In vitro,
HOTAIR has been reported as a direct target of the Hypoxia-inducible-factor-1α (HIF-1α), therefore enhancing A549 NSCLC cells proliferation, migration, and invasion [
48].
In vivo, tail vein injections of SPC-A1 cells with or without siRNA targeting
HOTAIR showed that the knock-down condition reduced number of metastasis nodules found in the lungs of immunocompromised mice [
46].
HOTAIR silencing resulted in a decrease in matrix metalloproteinases (MMPs, which promote invasion and migration) expression and an increase in HOXA5 levels (a tumor suppressor) in cell lines, suggesting
HOTAIR acts through regulation of expression of cancer-related genes [
46]. Although this xenograft assay does not account for all steps required for a tumor cell to achieve metastatic colonization (extravasation, survival in blood, seeding of new site, proliferation in new site), and they measured colonization of lungs by a lung cancer cell line (same tissue), it raises the possibility that
HOTAIR may be involved in seeding and survival of cancer cells. Additionally, the absence of a competent immune system challenges interpretation of these results. Development of a
HOTAIR transgenic mouse model to understand
in vivo implications in lung cancer initiation and progression is necessary, similar to a
HOTAIR inducible system recently reported for breast cancer [
49]. This model showed that sustained
HOTAIR overexpression promotes breast cancer metastasis to lungs. Overall, with the data available,
HOTAIR seems to play an oncogenic role in lung cancer; however, robust mechanisms through which this lncRNA function remain to be uncovered.
Most studies focus on the contribution of a single lncRNA to cancer phenotypes. However, whether the lncRNAs described above are expressed simultaneously in tumors with unique or redundant functions remains to be explored in depth. For example, Esposito
et al showed that at least 80 oncogenic lncRNAs are active in NSCLC through a lncRNA-focused CRISPR screen. By further dissecting the role of two candidate lncRNAs,
CHiLL1 and
GCAWKR, they showed these have distinct cellular localization and non-overlapping targets. Importantly, ASOs targeting both these lncRNAs yielded additive effects, suggesting that they have cooperating functions in NSCLC progression [
50]. LncRNAs are generally expressed at lower levels than protein-coding genes [
51]. Because of this, we speculate that lncRNAs with redundant functions may be expressed simultaneously to compensate for a higher expression of their targets in disease conditions. By examining available TCGA lung adenocarcinoma datasets containing mRNA expression data, we did not find significant correlations (negative or positive) among the expression of lncRNAs described here. However, such an analysis in combination with functional studies could shed light on mutual exclusivity relationships between certain lncRNAs. Additionally, whether certain lncRNAs are predominantly expressed at different stages of tumor progression remains to be explored. A new online resource, lncRNAfunc
https://ccsm.uth.edu/lncRNAfunc, provides insights on differentially expressed lncRNAs across different cancer types and stages available in TCGA, as well as functional predictions [
52]. Although this analysis did not detect any correlations between the lncRNAs mentioned here and lung cancer stages, possibly due to lack of sufficient sample sizes, these lncRNAs did show correlation with stage in other cancers; for example,
LUCAT1 was correlated with cancer stage in kidney cancer.
In summary, LncRNAs have oncogenic and tumor suppressive roles in lung cancer, illustrated in
Figure 1. LncRNAs interact with protein-coding molecules resulting in activation or inactivation of specific signaling pathways in cancer cells. We speculate more lung cancer specific lncRNAs will be identified with genome-wide transcriptomic studies.
4. LncRNAs as biomarkers in NSCLC
LncRNAs are highly stable molecules that can be found in the systemic circulation. They are resistant to degradation due in part to their secondary structures, transport by exosomes, and stabilizing post-translational modifications [
79,
80]. Therefore, the study of LncRNAs in circulation is a plausible non-invasive method of detecting and following cancer progression. Currently, the most commonly used biomarkers for NSCLC diagnosis from circulation are carcinoembryonic antigen (CEA), cytokeratin-19 fragment (CYFRA21-1), squamous cell carcinoma antigen (SCCA), prolactin (PRL), and carbohydrate antigen 125 (CA125), which can be used individually or combined as a signature [
81,
82]. Additionally, LncRNAs can also be stably found in urine [
83] and even in nasal mucosa [
84], although the latter has not been explored in the context of cancer. In this section, we review evidence for the potential of LncRNAs as disease biomarkers in lung cancer.
Tang
et al analyzed LncRNAs expression in the blood from 232 patients diagnosed with NSCLC as compared to healthy controls. Expression levels of three LncRNAs (
RP11-397D12.4, AC007403.1, ERICH1-AS1) was higher in disease versus health states. Importantly, this expression pattern was stable even after freeze-thaw cycles [
85]. In a similar study, the LncRNAs
SPRY4-IT1, ANRIL, and
NEAT1 were found overexpressed in NSCLC versus healthy controls (N=50/group), and were detected stably after several freeze-thaw cycles and after samples exposure to room temperature for up to 24h [
86]. A more recent study analyzed exosomal LncRNAs from blood of NSCLC patients vs. Tuberculosis patients or healthy controls. They report higher levels of
RP5-977B1 in NSCLC compared to the two non-cancer groups. The diagnostic power of this LncRNA was greater than that of conventional biomarkers such as CEA and CYFRA21-1, and additionally worked for early-stage NSCLC, speaking to the promise of LncRNAs to detect early disease [
87].
HOTAIR has also been evaluated for its diagnostic value for pathological staging of NSCLC. It was determined to have similar power as that of biomarkers CEA and CYFRA21-1 [
88].
ESCCAL-1 is an oncogenic lncRNA, initially identified in esophageal cancer [
89]. Results from a large cohort lung cancer screen show that
ESCCAL-1 has increased expression in serum samples from lung cancer patients as compared to serum from patients with benign nodules or healthy individuals (unpublished data). These studies support the use of elevated lncRNAs expression individually or as signatures as biomarkers to predict disease or as staging tools in NSCLC. Clinical trials that evaluate the use of LncRNAs as biomarkers are currently ongoing for other cancers, according to the ClinicalTrials.gov database.
Comparable to tumor-promoting lncRNAs, tumor-suppressive lncRNAs such as
GAS5 also show potential as biomarkers in NSCLC. Liang
et al compared
GAS5 levels in the plasma of 90 NSCLC patients vs. 33 healthy controls. They detected significantly lower levels of
GAS5 in plasma derived from cancer patients. Additionally, they measured the dynamics of
GAS5 before and after surgery, detecting an increase in
GAS5 levels seven days after patients had surgery [
90]. As discussed above,
GAS5 has a clear tumor suppressive role in lung cancer, therefore, these studies are consistent and support the use of
GAS5 as biomarker for diagnosis and for responses to clinical intervention.
5. LncRNAs as therapeutic targets
The main approaches to targeting LncRNAs are similar to protein-coding genes: inhibiting oncogenic LncRNAs or restoring function of tumor-suppressive LncRNAs. Here, we briefly describe advances in LncRNA therapeutics as well as some challenges.
Therapeutic approaches to target LncRNAs with oncogenic functions mainly use ASOs (anti-sense oligonucleotides). The core mechanisms of action of ASOs involve promoting cleavage of their RNA targets, impeding their translation, or modulating splicing [
91]. There are several types of ASOs depending on their chemical modifications, such as locked nucleic acids and morpholinos, that may make them more resistant to degradation, provide better cellular availability, and lower off-target events [
91]. As mentioned in sections above, ASOs are a strategy widely used in cell line and animal-based assays to elucidate LncRNAs role in cancer [
27]. Although ASOs that target miRNAs have entered clinical trials [
4,
92], to our knowledge there are no therapies in clinical trials targeting LncRNAs directly. Small molecules are another strategy to interfere with oncogenic LncRNAs function [
93]. For example, quercetin, a recently developed small molecule, binds to a
MALAT1 triplex and modulates its transcript levels and functions
in vitro [
94]. Another small molecule recently identified, AC1NOD4Q, blocks the interaction between
HOTAIR and EZH2 (a PRC2 subunit), impeding methylation of downstream targets [
95]. Moreover, emerging strategies to silence LncRNAs in pre-clinical models use CRISPR-based approaches although these have not reached the clinical setting [
96].
Restoring function of tumor suppressive LncRNAs can be achieved by gene therapy or administration of synthetic RNA molecules approaches. Although gene therapy has not reached the clinic for LncRNAs as targets, there are pre-clinical models exploring it in non-cancer contexts. For example, the LncRNA
LeXis has been explored for exogenous administration using an adeno-associated virus vector in a murine model of Familial Hypercholesterolemia [
97].
There are several challenges for RNA-based therapies reported from the field of miRNAs. For example, in the phase-I clinical trial for MRX34, a double-stranded
miR-34a mimic encapsulated in liposomal nanoparticles, serious adverse events were immune-mediated toxicities, such as cytokine release syndrome, that resulted in 4 patient deaths and caused the trial to terminate early [
98]. The delivery of these therapies is another challenge, as they may accumulate in detoxifying organs such as kidneys and liver, causing associated toxicities [
99,
100]. Additionally, off-target events and on-target events in non-tumor tissues can account for toxicities. Therefore, there is a need to detect immune-related toxicities in pre-clinical studies and design strategies to reduce their prevalence, to engineer delivery strategies that target organs of interest with lower off-site effects, as well as to increase target molecule specificity.