Diagnostic and Therapeutic Implications of microRNAs in Non-Small Cell Lung Cancer

microRNAs (miRNAs), endogenous suppressors of target mRNAs, are deeply involved in every step of non-small cell lung cancer (NSCLC) development, from tumor initiation to progression and metastasis. They play roles in cell proliferation, apoptosis, angiogenesis, epithelial-to-mesenchymal transition, migration, invasion, and metastatic colonization, as well as immunosuppression. Due to their versatility, numerous attempts have been made to use miRNAs for clinical applications. miRNAs can be used as cancer subtype classifiers, diagnostic markers, drug-response predictors, prognostic markers, and therapeutic targets in NSCLC. Many challenges remain ahead of their actual clinical application; however, when achieved, the use of miRNAs in the clinic is expected to enable great progress in the diagnosis and treatment of patients with NSCLC.


Introduction
microRNAs (miRNAs) are small, non-coding RNAs that are 20-25 nucleotides in length [1,2], and are endogenous suppressors of target genes [3]. They have complementary sequences to 3 -untranslated regions (3 -UTRs) of target mRNAs and bind to these regions through Watson-Crick base pairing. Perfectly matched binding between miRNAs and 3 -UTRs leads to mRNA cleavage and degradation, while imperfectly matched binding leads to translational repression. This binding involves 7-8 nucleotides of the miRNAs, which is called the seed sequence; therefore, one miRNA can target multiple mRNAs, and one mRNA can be targeted by multiple miRNAs. Through suppressing target genes, miRNAs regulate diverse physiological and pathological conditions, including cancer. miRNAs can either promote or repress cancer development and progression according to their target genes. Numerous miRNAs can act as oncogenes by negatively regulating tumor suppressors. For example, miR-21 expression is up-regulated in colon cancer, and promotes cell growth and invasion by repressing the tumor suppressor PTEN [4]. Conversely, let-7 inhibits cellular proliferation by negatively regulating the KRAS oncogene in lung cancer [5].
Lung cancer manifests as a malignant tumor caused by uncontrolled cell growth in the bronchiolar and alveolar epithelium of the lungs. Lung cancer is the most common cancer worldwide in both incidence (11.6% of the total cases) and mortality (18.4% of the total cancer deaths) [6]. Long-term smoking is well-known to be the main cause of lung cancer, while environmental effects (e.g., air pollution and particulate matters) and genetic variations (e.g., KRAS and EGFR mutations) can and synergistically suppress cancer cell migration by regulating actin cytoskeleton, adherens junctions, and focal adhesion [23]. Furthermore, "arm switching" between two strands of pre-miRNA duplexes can occur during cancer development, which can be used as a cancer biomarker [20].

Functional Mechanisms of miRNAs
In most cases, miRNAs recognize mRNA targets via binding to specific sequences at their 3 -UTRs; however, some miRNAs can bind to 5 -UTRs or coding sequences of mRNA targets [24]. For example, miR-1254 interacts with the 5 -UTR of CCAR1 and enhances its stability [25], while miR-20a represses DAPK3 by binding to its protein coding sequences (exon 2) [26]. Even promoter regions can be targeted by miRNAs; let-7i binds to the core promoter region of IL2 and up-regulates IL2 promoter activity [27]. Recently, numerous studies have shown that long non-coding RNAs are also binding partners of miRNAs [28].
Mature miRNAs loaded onto the RISC mediate gene silencing via two mechanisms: mRNA decay and translational repression [29,30]. Ago proteins have an RNase H-like domain; thus, RISC can cleave mRNA targets when they are perfectly complementary to miRNAs, which is common in plants [31]. In contrast, animal miRNAs bind to their target mRNAs at seed sequences of 7-8 nucleotides in length thorough partial complementarity, which instead induces translational repression and mRNA decay. miRNA-loaded Ago proteins bind to GW182, a scaffolding protein [32] that recruits poly(A)-binding protein (PABP) and deadenylase complexes (CCR4-NOT and PAN2-PAN3). Deadenylation then causes degradation of target mRNAs. In addition, RISC recruits decapping factors, such as DCP1, DCP2, and DDX6, which promotes decapping and degradation of target mRNAs [29,30].
GW182 binding to RISC represses initiation of translation by breaking the interaction between PABP and eIF4G, which stimulates ribosome recruitment [29]. GW182 also recruits translation repressors working at the inhibition step, such as DDX6 [33] and eIF4E-binding protein 4E-T [34]. In addition, miRNAs have been shown to dissociate the eIF4A complex, thus inhibiting ribosome binding and scanning [35]. Translational repression and mRNA decay are both crucial mechanisms of miRNA-mediated gene silencing; however, these two mechanisms do not always work synergistically [36], and the preference between the two mechanisms may depend on cellular context or miRNA:mRNA binding characteristics [29,37]. The biogenesis and functional mechanisms of select miRNAs are summarized in Figure 1.

Regulation of miRNA Expression
miRNA expression can be regulated through transcriptional and post-transcriptional mechanisms [38]. Like mRNAs transcribed by RNA polymerase II complex, miRNAs are also under the transcription control of various signaling pathway components and transcription factors [39]. Promoter methylation is also one of major mechanisms for regulating miRNA expression. miR-34a is downregulated in colon cancers with liver metastases, which is due to hypermethylation on the promoter region [40]. Upregulation of oncogenic miR-21 in various types of cancer is associated with hypomethylation on the promoter region [41]. In addition, histone modification on the promoter of miR-200b/200a/429 cluster causes silencing of miR-200 family members and promotes stemness of breast cancer cells [42].
After transcription and processing/maturation, miRNA levels are further regulated by diverse endogenous factors. Ago2 protein increases miRNA abundance by promoting miRNA processing and enhancing miRNA stability, which is independent of its RNase function [43]. RNA binding protein FXR1 stabilizes miR-301a-3p and facilitates p21 targeting in oral cancer [44]. QKI-5 directly interacts with miR-196b-5p and reduces its stability [45]. Numerous competing endogenous RNAs (ceRNAs) are also involved in the regulation of miRNA levels. ITGA1 mRNA functions as a sponge against miR-181b and relieves ADCY9 targeting [46]. FN1 mRNA acts as a ceRNA for miR-200c and modulates epithelial-to-mesenchymal transition (EMT) in breast cancer cells [47]. Moreover, one of the main features of long non-coding RNAs [48] and circular RNAs [49] is their functioning as ceRNAs, thus preventing the interaction between miRNAs and their target genes.

Roles of miRNAs in Progression and Metastasis of NSCLC
From the moment of tumor initiation to distant metastasis, epithelial cancer cells, such as NSCLC cells, must undergo a series of steps that are necessary for the acquisition of their pathological properties [50]. First, certain epithelial cells gain a selective growth advantage via genetic or epigenetic events. Through the EMT, highly proliferative cancer cells lose cell-to-cell and cell-to-matrix adhesion, and separate from the primary tumor. These cells acquire additional invasive abilities, and invade into basement membranes and stromal extracellular matrix (ECM). The invasive cancer cells further invade into surrounding blood or lymphatic vessels (intravasation). Then, they are transported via the circulatory system to distant sites and exit from blood vessels (extravasation). Once at the new site, the cancer cells reinvade, adapt to the new environment, and, finally, proliferate to form secondary metastatic tumors. miRNAs are closely involved in every step of NSCLC progression and metastasis, and precisely control the expression and activity of key factors during these processes. Select miRNAs that are known to regulate NSCLC development are shown in Table 1.

Angiogenesis and Hypoxia
The formation of a network of new blood vessels that can supply enough oxygen and nutrients to the tumor tissues is essential for optimal tumor outgrowth. Cancer cells facilitate growth of vascular endothelial cells and promote angiogenesis by secreting vascular endothelial growth factors (VEGFs) [50]. miR-128 and miR-195 directly target VEGFs thereby suppressing tumorigenesis and angiogenesis [65,66]. miR-200b targets the VEGF receptors Flt-1 and KDR, and suppresses cancer cell invasion and metastasis [67,68]. On the opposite side, miR-130b promotes tumorigenesis by targeting TIMP-2, an inhibitor of metalloproteinase-2 and angiogenesis [69]. An inadequate supply of oxygen around the tumor microenvironment causes hypoxia. HIF1α, which promotes angiogenesis, is induced under hypoxic conditions. Several studies have reported that HIF1α is also targeted by several miRNAs, such as miR-130a, miR-199a, and miR-200c [70][71][72].

Survival and Immune Escape
Lack of attachment to the ECM triggers cell death (anoikis) in epithelial cells. Once they invade the vascular or lymphatic system, cancer cells need to acquire anoikis resistance [83]. miRNAs promote or inhibit anoikis by targeting key molecules involved in anoikis signaling pathways [84]. miR-34a and miR-451 enhance the susceptibility of lung cancer cells to anoikis [85,86]. miR-148a inhibits anchorage-independent growth by targeting MMP15 and ROCK1 in NSCLC [87]. In contrast, exosomal miR-222 promotes cell survival under anchorage-independent conditions by directly targeting SOCS3 [88]. Cancer cells must survive the attacks of the immune system to progress, and miRNAs are involved in both immune attacks on tumors and immune escape [89,90]. miR-451 suppresses cell proliferation and metastasis in lung cancer cells by directly targeting PSMB8, which is one subunit of an immunoproteasome and modulates inflammatory responses [91]. miR-200 family members target PD-L1 and control immunosuppression in NSCLC cells [92]. miR-138, miR-140, and miR-142 have also been reported to target PD-L1 [93][94][95]. On the contrary, miR-197 enhances PD-L1 expression through the regulation of the cyclin-dependent kinase CKS1B and STAT3 pathway [96].

MET and Metastatic Colonization
A subset of cancer cells that survive within the systemic circulation will ultimately colonize at a distant metastatic site. In contrast to the initial stages of cancer development, these cancer cells lose their mobility and invasiveness and regain cell-to-cell and cell-to-matrix adhesiveness, which is a reverse process of EMT known as the mesenchymal-to-epithelial transition (MET) [97]. Although metastasis occurs independently of MET in some cases [98], various miRNAs associated with EMT also regulate MET. Ectopic expression of miR-200 family members induces MET in highly metastatic lung cancer cells [99]. Additionally, miR-147 induces MET and reverses drug resistance [100]. miR-29b promotes MET and prevents lung fibrosis [101]. Selection of the final metastatic destination and colonization at various organ sites, such as the bones, brain, and lymph nodes, is also affected by a variety of miRNAs [102].

Clinical Implications of miRNAs in NSCLC
Considering that multiple miRNAs have functions similar to their target genes and that a single miRNA can regulate several mRNAs, a panel of miRNAs is considered a better biomarker than individual miRNAs for clinical applications [74]. Since cancer-associated miRNA biomarkers can be easily detected in tissue, blood, or other bodily fluids, circulating miRNAs grant several potential advantages for clinical application, including high stability in serum, ease of non-invasive detection in circulation, and a convenient screening method [103]. Circulating miRNAs also afford a chance to overcome the problem of tumor heterogeneity by allowing for the collection of all pathological signals from many disparate portions of primary tumors and metastatic sites.

NSCLC Subtype Classifiers
LUAD and LUSC are the major histological subtypes of NSCLC. LUSC is most commonly associated with tobacco use, while LUAD is commonly associated with non-smokers and women [104], suggesting the existence of several underlying major differences not only in biological patterns, but also in molecular characteristics, between histological subtypes. For example, activating mutations in epidermal growth factor receptor (EGFR) and mutations in anaplastic lymphoma kinase (ALK) fusion proteins usually occur in LUAD, but not in LUSC, rendering therapy targeted at these genes ineffective for LUSC [105]. MiRNAs can be used to distinguish subtypes among NSCLC. In a previous report, four miRNAs (miR-205, miR-93, miR-221, and miR-30e) were shown to be highly expressed in LUSC, and five miRNAs (miR-29b, miR-29c, let-7e, miR-100, and miR-125a-5p) were highly expressed in LUAD [106]. Through an analysis of three miRNome profiling datasets, Hu et al. identified that miR-375, miR-203, and miR-205 were differentially expressed miRNAs that could be used to distinguish LUSC from other NSCLC subtypes [107]. A recent study using machine learning approaches identified miR-944 and miR-205 as useful for classifying tumors into the LUAD and LUSC subtypes [108].

Diagnostic Markers
To date, serum tumor markers have not been employed for early lung cancer screening due to limitations in their effectiveness, sensitivity, and specificity. However, circulating miRNAs have demonstrated potential advantages for use in in clinical screening methods [103]. Currently, several articles have shown that many kinds of circulating miRNAs can be used to detect lung cancer [109][110][111][112]. Recently, in an analysis of serum samples from 1566 lung cancer and 2178 non-cancer participants, the diagnostic accuracy, sensitivity, and specificity of the combined expression levels of two miRNAs (miR-1268b and miR-6075) were all 99%, regardless of the histological type and pathological Tumor, Node, Metastasis (TNM) stage of the NSCLC [110]. In another study of 2856 participants, a 14-miRNA signature distinguished patients with lung cancer from patients with non-tumor lung diseases with an accuracy of 92.5%, sensitivity of 96.4%, and specificity of 88.6%. In addition, the expression level of miR-17-3p distinguished patients with lung cancer from those with non-tumor lung diseases with the highest significance and an Area Under the Receiver Operating Characteristics (AUROC) value of 0.899 [112].

Drug-Response Predictors
Furthermore, miRNAs could be useful for predicting tumor responsiveness to chemotherapy or different therapeutic approaches. In a study with drug-resistant NSCLC cell lines, several markers were identified as being predictive of the degree of responsiveness to therapy, with miR-192, miR-194, miR-205, miR-30a, and miR-30c demonstrated to be predictive factors for a positive response to chemotherapy [113]. In an analysis of 148 LUAD patients who were negative for EGFR mutations or ALK translocations and who received maintenance treatment with pemetrexed, progression-free survival duration for patients expressing different levels of circulating miR-25, miR-145, and miR-210 were significantly different in the pemetrexed-treated group, suggesting these three miRNAs are predictors for the efficacy of maintenance treatment [114]. Recently, with the advent of immune checkpoint blockade therapy, immunotherapy has shown promising results in various types of cancer including lung cancer [115]. In a comparative analysis between responders and non-responders to PD-1/PD-L1 inhibitors, miR-320 family members, such as miR-320d, miR-320c, and miR-320b, were identified as potential biomarkers for predicting the efficacy of immunotherapy in advanced NSCLC. In addition, the level of exosomal miR-125b-5p was dramatically downregulated in the partial response-post samples, indicating that miR-125b-5p levels might be useful for monitoring the efficacy of anti-PD-1/PD-L1 treatment [116]. When the T-cell suppressor miR-125b-5p is downregulated during immunotherapy, patients may achieve increased T-cell function and respond well to immunotherapy. Of note, plasma-derived exosomes detected in patients are mainly released from tumor cells, which can more accurately and dynamically reflect the state and function of tumor cells.

Prognostic Markers
A prognostic biomarker should ideally provide information on the overall disease outcome in patients, such as disease recurrence or disease progression, independent of the treatment regimen. The discovery of prognostic factors could contribute to classifying patients by prognosis and identifying high-risk cases requiring aggressive approaches [117]. Similar to oncogenes and tumor suppressor genes, oncogenic miRNAs and tumor suppressive miRNAs are playing major roles in accurately identifying lung cancer prognoses. Numerous studies have shown that individual miRNAs play prognostic roles in NSCLC patients [52,54,58,61,74,[76][77][78]80,91,96,99]. In a recent analysis using the TCGA database, two different prognostic miRNA signatures (a four-miRNA signature for LUAD: miR-375, miR-148a, miR-29b-1, and miR-584; and a four-miRNA signature for LUSC: miR-4746, miR-326, miR-93, and miR-671) were found to be independent prognostic factors in LUAD and LUSC patients [118]. Machine learning algorithms are indeed useful methods for analyzing large volumes of data, such as genetic information produced by next-generation sequencing technologies. We previously applied a neural network-based algorithm called Cascaded Wx framework to extract miRNA markers most highly associated with LUAD patient survival [119]. Through subsequent profiling of miRNA expression levels in LUAD patient samples, miR-374a and miR-374b, both EMT-related miRNAs, were identified as potential prognostic markers associated with poor survival in LUAD patients [120].

Therapeutic Targets
MiRNAs have various functions within cancer cells, with some having a high specificity for cancer-associated pathways, and they are one of the most promising therapeutic targets with well-characterized expressions. Biologically, the attractiveness of using miRNAs for cancer treatment comes from their ability to target multiple genes involved in multiple cancer-related pathways. Adding tumor suppressive miRNAs or reducing oncogenic miRNAs in cancer cells could be effective as a therapeutic strategy. In in vivo lung cancer models for therapeutic applications, miR-15/16 [121], miR-29b [122], miR-7 [123], miR-34a [124], let-7 [125], miR-200c [126], and miR-145 [127] were tested using different delivery systems. Two phase I clinical trials have been conducted for advanced solid cancers, including NSCLC (Table 2) [128][129][130]. The first-in-human, phase I study of a microRNA-based cancer therapy using MRX34, a liposomal mimic of miR-34a, was conducted but closed early due to unexpectedly severe immune-mediated toxicities with a modest overall response rate of 4% [129]. In contrast, in a phase I MesomiR-1 trial, TargomiRs, comprised of a miR-16-based microRNA mimic packaged in EDV TM nanocells that are targeted with an anti-EGFR-specific antibody, was tested in patients with advanced NSCLC or malignant pleural mesothelioma. Overall, TargomiR treatment has been well tolerated and shown to be safe in patients. Interim data indicated that disease control was achieved in five of six patients after 8 weeks of protocol treatment [128,130].

Pitfalls and Challenges for Clinical Application of miRNAs
Two phase I clinical trials have provided early data elucidating the efficacy of novel miRNA-based treatment approaches [128,129]. Considering the fact that phase I studies are predominantly comprised of previously heavily treated patients who have partially or fully failed other treatment modalities, the potential of partial treatment response can be high (Table 2). However, in a MRX34 trial, treatmentattributed serious adverse events tended to occur late after the completion of daily MRX34 infusions. The serious adverse events included sepsis, hypoxia, cytokine release syndrome, and hepatic failure, which is a pattern suggestive of immune-mediated toxicity. These adverse events were not observed in the pre-clinical tests with MRX34 in animal models, including non-human primates [129]. These findings suggest that the development of appropriate drug delivery systems for these miRNA-based therapies is essential. The potentially wide-ranging impacts of miRNAs on the regulation of gene expression may result in unexpected side effects in normal cells through nontargeted delivery or unintentional targeting due to the expression of cancer-related antigens. Thus, to avoid "off target" effects, such as systemic immune activation, the effective and specific targeting of miRNA therapeutics to cancer tissues which spares normal tissues is an essential, but as yet unresolved, challenge to overcome.

Conclusions
To date, miRNAs have been widely known for their key roles in both tumorigenesis and tumor suppression, and have been extensively studied in the field of NSCLC. miRNAs represent a powerful tool to help in the diagnosis, prognosis, and prediction of response to various treatments of NSCLC to improve patient survival rates. miRNA mimics of tumor suppressive miRNAs or miRNA inhibitors (such as Antagomirs) against oncogenic miRNAs may have therapeutic potential. However, despite the potential clinical benefits of miRNA mimics, recent phase I trials have shown unexpected adverse events. Thus, the future development of innovative methods of miRNA delivery will be required to avoid evoking an undesirable immune response.

Conflicts of Interest:
The authors declare no conflict of interest.