The Role of Non-coding RNAs in Oncology

Abstract

For decades, research into cancer biology focused on the involvement of protein-coding genes. Only recently was it discovered that an entire class of molecules, termed non-coding RNA (ncRNA), plays key regulatory roles in shaping cellular activity. An explosion of studies into ncRNA biology has since shown that they represent a diverse and prevalent group of RNAs including both oncogenic molecules and those that work in a tumor suppressive manner. As a result, hundreds of cancer-focused clinical trials involving ncRNAs as novel biomarkers or therapies have begun, and these are likely just the beginning.

eTOC – Chinnaiyan & Slack

Slack and Chinnaiyan explore the diverse and context-dependent roles of ncRNAs, including circRNAs, lncRNAs, miRNAs, piRNAs, and tsRNAs, in cancer. They provide insight into the prospect of therapeutic targeting and use of ncRNAs as biomarkers with an up-to-date summary of clinical and preclinical studies.

Although now one of the hottest topics in biomedical science, the importance of non-coding RNA (ncRNA) was largely unrecognized until recently. RNA was once thought to mostly serve as a messenger that carried instructions encoded in DNA so that other molecules, like the ribosome, could use the code to make proteins. However, in the last 30 years, researchers have discovered that multiple types of RNA exist, and that among the most important is non-coding RNA- the type that is not involved in producing proteins. The discovery of tens of thousands of ncRNA species has revolutionized the field, altering the way that researchers think about physiology and the development of disease (Adams et al., 2017; Bartel, 2018; Evans et al., 2016; Rupaimoole and Slack, 2017). Non-coding RNAs constitute more than 90% of the RNAs made from the human genome, but most of the >50,000 known ncRNAs have been discovered only in the past 10 years and remain largely unstudied (Deveson et al., 2017; Esposito et al., 2019; Kopp and Mendell, 2018; Ransohoff et al., 2018).

Still, there are many ncRNAs that have since been shown to play key roles in both normal cellular function and disease, including cancer, and this information is being actively translated into the clinic. Some small ncRNAs are so stable that they survive in the bloodstream and could be the basis for accurate and sensitive screens for major human cancers in a few drops of blood (Agaoglu et al., 2011; Imaoka et al., 2016; Toiyama et al., 2013). Additionally, ncRNAs can be therapeutically targeted, and the delivery of ncRNAs can be based on an existing foundation of what has been learned regarding delivery of RNAi and oligonucleotides targeting protein-coding mRNAs (Levin, 2019; Pecot et al., 2011; Wu et al., 2014). In fact, the field of RNA medicine has seen a renaissance (Levin, 2019) with the recent approval of the first RNAi drug Onpattro™ (patisiran; reduces levels of TTR for treatment of the neurodegenerative disease hereditary transthyretin amyloidosis) (Adams et al., 2018) and the clinical success of the RNA-targeting oligonucleotide drug Spinraza® (nusinersen; increases levels of full-length SMN2 for treatment of the neuromuscular disease spinal muscular atrophy) (Wurster et al., 2019). In addition, clinical trials with drugs based on a class of ncRNAs called microRNA (miRNA), either therapies that increase or decrease the target miRNA, have begun for cancer (Beg et al., 2017; Seto et al., 2018; van Zandwijk et al., 2017).

In this review, we will discuss ncRNAs in relation to cancer cell biology and their relevance to current clinical practice. We first examine the intricacies of the different classes of ncRNAs (microRNAs (miRNAs), transfer RNA (tRNA)-derived small RNAs (tsRNAs), PIWI-interacting RNAs (piRNAs), long ncRNAs (lncRNAs), pseudogenes, and circular RNAs (circRNAs), Supplemental Table 1) and provide fundamental examples of the far-reaching roles that these molecules have in affecting cancer processes. We then discuss how these basic science insights in ncRNA biology are being used to develop next-generation diagnostics and therapies in cancer. As the so-called “dark matter” of the genome continues to be brought into the light, it is evident that targeting ncRNA signaling has great potential to impact cancer patient care.

Overview of classes of ncRNAs and their association with cancer

For decades, the miniscule protein-coding portion of the genome was the primary focus of medical research. The sequencing of the human genome showed that only about 2% of our genes ultimately code for proteins, and many in the scientific community believed that the remaining 98% was simply non-functonal “junk” (Mattick and Makunin, 2006; Slack, 2006). However, the ENCODE project revealed that the non-protein coding portion of the genome is copied into thousands of RNA molecules (Djebali et al., 2012; Gerstein et al., 2012) that not only regulate fundamental biological processes such as growth, development, and organ function, but also appear to play a critical role in the whole spectrum of human disease, notably cancer (for recent reviews, see (Adams et al., 2017; Deveson et al., 2017; Rupaimoole and Slack, 2017)). Trailblazing research led to our understanding of how ncRNA molecules perform multiple vital roles in the coding, decoding, regulation, and expression of genes as well as how they communicate with each other (Anastasiadou et al., 2018b; Esquela-Kerscher and Slack, 2006; Gregory and Shiekhattar, 2005; Huarte and Rinn, 2010; Kasinski and Slack, 2011; Krichevsky et al., 2003; Rinn and Huarte, 2011; Tay et al., 2014). This knowledge of ncRNA regulatory roles revealed specific ncRNA networks based on complementary base pairing at work in different cancer types (Anastasiadou et al., 2018b), which in turn, opened up possibilities for scientists in this field to develop specific cancer therapeutic and preventive strategies focused on the ncRNAs made from the human genome (for recent reviews, see (Cieslik and Chinnaiyan, 2018; Rupaimoole and Slack, 2017)).

Cancer is characterized by cells that grow (proliferate) out of control, are able to spread to other tissues (metastasize), and lose the ability to die through the orderly process of cell death (apoptosis). The discovery of ncRNA has added a new dimension to the understanding of how cancer develops, and how it may be treated, by providing a window into the impact of the rest of the genome. Deregulated ncRNA expression, and subsequent downstream signaling processes that we detail in later sections, have been directly implicated in cancer development and progression. Genetic alterations in genes encoding ncRNAs have been found to be associated with cancer; however, compared to protein-coding genes, the list of genetic examples from studies thus far is considerably shorter. The most notable example is likely deletion of 13q14.3 in chronic lymphocytic leukemia (CLL) that deletes the miR-15/16 tumor suppressors (Calin et al., 2002). Conversely, amplification of chromosomal regions encoding oncogenic ncRNAs are also found in cancer, including amplification of lncRNAs FAL1 (Hu et al., 2014) and PVT1 (Tseng et al., 2014). Single nucleotide polymorphisms (SNPs) in the genes of lncRNAs H19 (Hua et al., 2016), ANRIL (Pasmant et al., 2011), and CCAT2 (Ling et al., 2013) have also been associated with varying risks of cancer development. In addition to genetic alterations within transcribed regions, mutations in the promoters of ncRNAs can lead to altered gene expression levels, such as recurrent driver mutations in the promoters of lncRNAs NEAT1 and RMRP in breast cancer (Rheinbay et al., 2017). Besides aberrations in sequences encoding the ncRNA itself, mutations or dysregulation of enzymes involved in the biogenesis of ncRNAs are implicated in cancer, such as Drosha and Dicer involved in miRNA processing (Rupaimoole and Slack, 2017). In addition to these genetic mechanisms, up- or downregulation of ncRNA expression associated with cancer can occur through epigenetic, transcriptional, or post-transcriptional processes (see recent reviews (Adams et al., 2014; Anastasiadou et al., 2018a; Rupaimoole and Slack, 2017)).

Non-coding RNAs can be divided into different classes, broadly based upon their size. The small ncRNAs important in cancer include miRNAs, tsRNAs, and piRNAs. At the opposite end of the size spectrum are the lncRNAs, which are characterized as untranslated RNAs greater than 200 nucleotides in length, and include subclasses such as pseudogenes and circRNAs.

MicroRNAs

Near the turn of the millennium, the first miRNAs, lin-4 and let-7, were identified through developmental studies in C. elegans (Lee et al., 1993; Reinhart et al., 2000). MicroRNAs are short ncRNAs of approximately 22 nucleotides in length that regulate the expression of other RNAs, notably mRNAs through binding between the 5’ end (known as the “seed”) of the miRNA with complementary sequences in target RNAs. Genes encoding miRNAs are transcribed by RNA Pol II and processed through an evolutionarily conserved pathway. In the canonical processing pathway, this longer primary transcript, called the pri-miRNA, forms a characteristic hairpin structure that is recognized by the Microprocessor complex (consisting of Drosha and DGCR8), cleaved into a pre-miRNA about 60 nucleotides in length, and exported to the cytoplasm via an Exportin 5 and Ran-GTP complex. The ends of the pre-miRNA are then cleaved by Dicer to form a miRNA duplex, which consists of 5’ phosphates and 2 nucleotide overhangs on each 3’ end. One strand of the miRNA duplex, the guide strand, is loaded onto an Argonaute protein and selected to form the RNA-induced silencing complex (RISC) containing the mature 22 nucleotide miRNA (see recent reviews by (Anastasiadou et al., 2018a; Bartel, 2018)). The mature miRNA functions by binding to the 3’ untranslated regions (3’UTR) of mRNAs and inhibiting their use by either degradation or translational repression (Bartel, 2018; Esquela-Kerscher and Slack, 2006). Multiple studies have attempted to annotate the number of miRNAs in different species. For humans, higher estimates from miRBase v22 have placed the number of mature miRNAs at 2654, but algorithms from other databases, such as MirGeneDB2.0, decrease this number to 588 high confidence miRNAs (Fromm et al., 2015; Kozomara et al., 2018). Despite discrepancies in their absolute numbers, it is evident that miRNAs have far-reaching effects on downstream processes, as more than 60% of coding genes are potential targets of miRNAs (Friedman et al., 2009). In addition, hundreds of miRNAs are conserved at their seed region across phylogeny (Bartel, 2018), suggesting key roles in developmental or physiological processes in animals.

Regarding cancer, miRNAs provide a powerful new avenue for the discovery of novel genetic risk factors (Ryan et al., 2010). Among the small ncRNA species, miRNAs are by far the most extensively studied in cancer compared to tsRNAs and piRNAs. MicroRNAs have been found to be altered in all cancer types studied (Volinia et al., 2006), and alterations in miRNAs have been demonstrated to play a crucial role in affecting molecular and cellular processes of the cancer state (Esquela-Kerscher and Slack, 2006; Nicoloso et al., 2009). Although researchers are still learning the extent of the contribution of miRNAs to cancer, these small ncRNAs seem to function in one of two ways—as tumor suppressors or oncogenes (commonly referred to as oncomiRs) that promote cancer growth or metastasis (Rupaimoole and Slack, 2017). Although small, miRNAs are powerful, with each molecule often able to regulate more than one target, and, vice versa, mRNAs are frequently targeted by several miRNAs (Bartel, 2018). As such, miRNAs function as master regulators that control the expression of thousands of coding and non-coding genes, including most of the insidious oncogenes, such as RAS, MYC, and EGFR, and the critical tumor suppressors, including TP53, PTEN, and BRCA1. Table 1 provides a list of selected miRNAs (and examples from the other classes of ncRNAs) that have in vivo experimental evidence to either support an oncogenic, tumor suppressive, or context-dependent role. Many of these studies involved creation of transgenic mouse models and/or delivery of miRNA mimetics or antimiRs, synthetic oligoribonucleotides that either replenish or decrease levels of the target miRNA. Table 1 further provides instances of the cancer-related molecular mechanisms that each miRNA has been shown to regulate. Here, we highlight a few examples from each category, with Figure 1 and Figure 2 also providing visual representations of one oncogenic and one tumor suppressive pathway, respectively.

Table 1.

Oncogenic or tumor suppressive non-coding RNAs with in vivo experimental evidence

Name	ncRNA	Cancer types examined	In vivo experimental techniques used	Cancer-related mechanisms and/or functions of ncRNA	References
Oncogenic ncRNAs
miR-10b	miRNA	breast, glioblastoma	antimiRs, CRISPR-Cas9 knockdown in mouse xenografts and allografts, transgenic knockout mouse models	targets several transcripts that encode regulators of cell cycle progression, migration, invasion, and metastasis	Ma et al., 2010, Kim et al., 2016, El Fatimy et al., 2017
miR-21	miRNA	lung, B cell lymphoma	transgenic knockout, overexpression mouse models	targets transcripts that encode negative regulators of RAS signaling, leading to increased proliferation and decreased apoptosis	Hatley et al., 2010, Medina et al., 2010
miR-31	miRNA	lung, breast	overexpression in mouse xenografts, transgenic knockout, overexpression mouse models	targets transcripts that encode regulators of RAS, WNT, and TGFβ signaling to increase proliferation, stem cell renewal, and metastasis	Edmonds et al., 2016, Lv et al., 2017
miR-155	miRNA	lymphoma	transgenic overexpression mouse model, treatment with antimiRs	targets SHIP1 transcript, a negative regulator of AKT, to increase proliferation and survival	O’Connell et al., 2009, Babar et al., 2012, Cheng et al., 2015
miR-221	miRNA	liver	overexpression, treatment with antimiRs in mouse xenografts	targets transcripts of tumor suppressors and cell cycle inhibitors (e.g., p27, PTEN) to increase proliferation and decrease apoptosis	Pineau et al., 2010, Park et al., 2011
LeuCAG3’tsRNA	tsRNA	liver	LNA knockdown in PDX	enhances translation of transcripts encoding ribosomal proteins, leading to increased ribosome biogenesis and proliferation	Kim et al., 2017
ARLNC1	lncRNA	prostate	shRNA or ASO knockdown in mouse xenografts	interacts with the mRNA encoding AR, a nuclear receptor, to promote oncogenic AR signaling, proliferation, and suraval	Zhang et al., 2018
CamK-A	lncRNA	breast	shRNA, siRNA knockdown in mouse xenografts/PDX	interacts with and controls activity of kinases that modulate calcium-induced NF-KB signaling, leading to remodeling of the tumor microenvironment	Sang et al., 2018
CCAT1	lncRNA	colorectal, esophageal	shRNA, siRNA knockdown in mouse xenografts	interacts with transcription factors (e.g., SOX2, p63) to activate expression of genes involved in increasing proliferation and decreasing apoptosis	Kim et al., 2014, Jiang et al., 2018
CTBP1-AS	lncRNA	prostate	siRNA knockdown, overexpression in mouse xenografts	recruits chromatin modifying, splicing factors to promoter of a nuclear receptor corepressor (CTBP) to decrease its expression, leading to increased oncogenic AR activity	Takayama et al., 2013
DSCAM-AS1	lncRNA	breast	shRNA knockdown in mouse xenografts	interacts with proteins of the hnRNP complex involved in RNA processing and mediates proliferation, invasion, and metastasis	Niknafs et al., 2016
EPIC1	lncRNA	breast	shRNA knockdown in mouse xenografts	interacts with MYC transcription factor and increases its activation of target genes, leading to enhanced cell cycle progression	Wang et al., 2018
FAL1	lncRNA	ovarian breast	shRNA, siRNA knockdown in mouse xenografts	stabilizes components of PRC1 chromatin modifying complex to mediate expression of genes involved in proliferation and suraval	Hu et al., 2014
HOTAIR	lncRNA	breast	siRNA knockdown, overexpression in mouse xenografts	recruits PRC2, LSD1/CoREST/REST chromatin modifying complexes, scaffolds transcription factors at target promoters of genes involved in invasion, metastasis, and proliferation	Gupta et al., 2010, Li et al., 2016b
LINK-A	lncRNA	breast	shRNA knockdown in mouse xenografts, transgenic overexpression in mammary gland, LNA knockdown	interacts with kinases that control HIF1α activity, glycolysis, enhances degradation of tumor suppressors (RB, p53) and antigen peptide-loading complexes to promote immune evasion	Lin et al., 2016, Hu et al., 2019
lncARSR	lncRNA	RCC	shRNA knockdown, overexpression in mouse xenografts	interacts with transcriptional coactivator YAP and acts as a ceRNA for miRNAs that target RTK transcripts, leading to enhanced suraval and propagation of tumor-initiating cells	Qu et al., 2016a, Qu et al., 2016b
PCAT-1	lncRNA	prostate	overexpression in mouse xenografts	represses expression of BRCA2 tumor suppressor to impact DNA damage repair	Prensner et al., 2014a
PVT1	lncRNA	colorectal, gastric	CRISPR-Cas9 or shRNA knockdown, overexpression in mouse xenografts	activates oncogenic signaling (MYC, STAT3) and represses expression of tumor suppressors (p15, p16), resulting in increased proliferation, angiogenesis and decreased apoptosis	Tseng et al., 2014, Kong et al., 2015, Zhao et al., 2018
SAMMSON	lncRNA	melanoma	GapmeR knockdown in PDX	interacts with and controls subcellular localization of proteins that regulate mitochondrial homeostasis and metabolism	Leucci et al., 2016
SChLAP1	lncRNA	prostate	shRNA knockdown in mouse xenografts	interacts with and antagonizes activity of the SWI/SNF chromatin modifying complex to promote invasion and metastasis	Prensner et al., 2013
THOR	lncRNA	lung, melanoma	CRISPR-Cas9 knockdown, overexpression in mouse xenografts, transgenic knockout, overexpression in zebrafish	binds IGF2BP1 to stabilize interactions with oncogenic target mRNAs, in turn stabilizing those transcripts and promoting proliferation	Hosono et al., 2017
BRAFP1	pseudogene	B cell lymphoma	transgenic overexpression mouse model	acts as a ceRNA for miRNAs that target the BRAF transcript, leading to increased BRAF expression, MAPK signaling, and proliferation	Karreth et al., 2015
circCCDC66	circRNA	colorectal	siRNA knockdown in mouse xenografts	sponges several miRNAs that target oncogenic transcripts (e.g., MYC), promoting proliferation, migration, and invasion	Hsiao et al., 2017
circCTNNB1	circRNA	gastric	shRNA knockdown, overexpression in mouse xenografts	binds a DEAD-box RNA helicase (DDX3) to increase transcription factor (YY1) activation of WNT/p-catenin target genes	Yang et al., 2019
Tumor suppressive ncRNAs
let-7	miRNA	lung	antimiRs, mimetics in mouse xenografts or transgenic lung cancer mouse models	targets several transcripts that encode oncogenes, including RAS, leading to decreased cell cycle progression and proliferation	Trang et al., 2010, Trang et al., 2011
miR-15a/16-1	miRNA	prostate, leukemia	knockdown, overexpression in mouse xenografts, transgenic knockout mouse	targets several transcripts that encode cyclins, CDKs, and anti-apoptotic proteins, thereby increasing apoptosis and inhibiting proliferation	Bonci et al., 2008, Klein et al., 2010
miR-34a	miRNA	lung, prostate, breast	overexpression, mimetics, antimiRs in mouse xenografts, mimetics in transgenic lung cancer mouse models	targets several oncogenic transcripts encoding cyclins, CDKs, cell adhesion molecules, RTKs, and other non-RTKs, resulting in decreased proliferation, invasion, and suraval	Liu et al., 2011, Trang et al., 2011, Kasinski et al., 2012, Adams et al., 2016b
miR-122	miRNA	liver	transgenic knockout mouse models, overexpression in transgenic liver cancer mouse models	targets expression of several genes involved in lipid metabolism, proliferation, and inflammation	Hsu et al., 2012, Tsai et al., 2012
miR-506	miRNA	ovanan	mimetics in mouse xenografts	targets SNAI2 transcript to decrease its expression and inhibit migration, invasion, and EMT	Yang et al., 2013a
GAS5	lncRNA	glioblastoma	overexpression in mouse xenografts	acts as a nuclear receptor response element mimic for GR, decreases oncogenic miRNA expression, leading to enhanced apoptosis, decreased proliferation, invasion, and migration	Zhang et al., 2013, Zhao et al., 2015
LET	lncRNA	liver, colorectal	shRNA knockdown, overexpression in mouse xenografts	interacts with and destabilizes a dsRNA binding protein (NF90), a key factor involved in regulation of HIF-1a levels and cell invasion	Yang et al., 2013b
MEG3	lncRNA	lung	overexpression in mouse xenografts	increases p53 tumor suppressor levels to activate apoptosis and decrease proliferation	Lu et al., 2013
PTENP1	pseudogene	RCC, bladder	overexpression in mouse xenografts	acts as a ceRNA for miRNAs that target the PTEN tumor suppressor transcript, leading to increased apoptosis, decreased proliferation, migration, and invasion	Yu et al., 2014, Zheng et al., 2018
circHIPK3	circRNA	bladder	overexpression in mouse xenografts	sponges miRNAs to suppress expression of an endoglycosidase (HPSE), thereby decreasing levels of pro-migratory and angiogenic factors	Li et al., 2017
Context-dependent ncRNAs
miR-26a	miRNA	glioma, leukemia, liver, colorectal	overexpression in transgenic glioma, liver cancer mouse models, transgenic overexpression mouse models	targets both tumor suppressor (PTEN) and cyclin (CCND2/E2) transcripts to either increase or decrease proliferation, depending on the context	Huse et al., 2009, Kota et al., 2009, Mavrakis et al., 2011, Zeitels et al., 2014
miR-29	miRNA	leukemia, glioblastoma	mimetics in mouse xenografts, transgenic overexpression mouse models	promotes proliferation in B cells but can also target and decrease expression of transcipts that enhance lipid synthesis (SCAP, SREBP-1), proliferation (CDK6), and apoptosis (MCL-1)	Garzon et al., 2009, Santanam et al., 2010, Ru et al., 2016
MALAT1	lncRNA	lung, breast	genetic knockout, overexpression in mouse xenografts and allografts, transgenic knockout, overexpression mouse models, ASOs in mouse xenografts or transgenic models	promotes expression of genes involved in metastasis, conversely, shown to also inhibit metastasis through interaction with/inhibition of a pro-metastatic transcription factor (TEAD)	Gutschner et al., 2013, Arun et al., 2016, Kim et al., 2018
NEAT1	lncRNA	prostate, skin, pancreatic	shRNA knockdown, overexpression in mouse xenografts, transgenic knockout mouse models	mediates oncogenic nuclear receptor (ER) signaling, prevents DNA damage and activation of p53 tumor suppressor, leading to proliferation, invasion, and decreased apoptosis, conversely, shown to also prevent transformation and proliferation in other settings	Chakravarty et al., 2014, Adriaens et al., 2016, Mello et al., 2017
NKILA	lncRNA	breast	shRNA knockdown, overexpression in mouse xenografts/PDX	binds/inhibits NF-KB and downstream inflammation, which increases apoptosis, reduces invasion, promotes activation-induced cell death in CTLs, T>1 cells, leading to immune evasion	Liu et al., 2015, Huang et al., 2018

Abbreviations: AR, androgen receptor, ASO, antisense oligonucleotide, CDK, cyclin-dependent kinase, ceRNA, competitive endogenous RNA, circRNA, circular RNA, CTL, cytotoxic T lymphocyte, EMT, epithelial-to-mesenchymal transition, ER, estrogen receptor, GR, glucocorticoid receptor, hnRNP, heterogeneous nuclear ribonucleoprotein, LNA, locked nucleic acid, lncRNA, long non-coding RNA, miRNA, microRNA, ncRNA, non-coding RNA, PDX, patient-derived xenograft, RCC, renal cell carcinoma, RTK, receptor tyrosine kinase, tsRNA, tRNA-derived small RNA

Figure 1. Oncogenic ncRNAs and cancer-promoting mechanisms.

Simplified examples of pathways towards oncogenesis for the different classes of ncRNAs discussed in the review. (A) miRNA: miR-155 directly targets and decreases expression of SHIP1, a hematopoietic cell-specific phosphatase that hydrolyzes phosphatidylinositol-3,4,5-triphosphate [PI(3,4,5)P₃] to phosphatidylinositol-3,4-bisphosphate [PI(3,4)P₂]. Decreased SHIP1 levels can drive AKT signaling, proliferation, and survival, leading to lymphoma. As represented in the figure, clinical trials with anti-miR-155 (NCT02580522/NCT03713320) are underway. (Therapeutic clinical trials targeting the other classes of ncRNAs have yet to be initiated, and are, thus, not depicted.) (B) tsRNA: LeuCAG3’tsRNA (abbreviated Leu3’ in the figure) directly binds and enhances translation of RPS15 and RPS28 transcripts by unfolding of their secondary structures. LeuCAG3’tsRNA can, therefore, increase the levels of these small ribosomal proteins and biogenesis of ribosomes, promoting proliferation of hepatocellular carcinoma cells. (C) lncRNA: IGF2BP1 is an RNA binding protein that forms a messenger ribonucleoprotein (mRNP) complex with other proteins. Interaction of the lncRNA THOR with IGF2BP1 within the mRNP complex leads to stabilization and increased translation of IGF2BP1 target mRNAs (represented by gray curved lines). These proteins (represented as different color ovals) include known oncogenes with wide-reaching effects that can lead to cancers such as melanoma. (D) pseudogene: BRAFP1 acts as a competitive endogenous RNA for BRAF by binding shared miRNAs. This leads to increased BRAF expression and downstream proliferative MAPK signaling that can drive cancers like lymphoma. (E) circRNA: circCTNNBI binds to DDX3 and increases its interaction with YY1 transcription factors, resulting in enhanced transactivation of YY1 target promoters. This includes promoters for many genes involved in Wnt/β-catenin signaling.

Figure 2. Tumor suppressive ncRNAs and pathways inhibiting tumor progression.

Examples of ncRNAs that can function as tumor suppressors and representative functional mechanisms of inhibition. Examples of tumor suppressive tsRNAs have not yet been mechanistically detailed and are, thus, not included in this figure. (A) miRNA: Among the targets of miR-16-1 are several transcripts that encode cyclins and cyclin-dependent kinases important for the G₀/G₁-S phase transition of the cell cycle. Decreased expression of these cell cycle regulators can prevent cancer development, and the locus encoding miR-16-1 is often deleted in chronic lymphocytic leukemia. A recent clinical trial (NCT02369198) examined the effects of replenishing this tumor suppressive miRNA in malignant pleural mesothelioma and non-small cell lung cancer using a TargomiR strategy. (B) lncRNA: MEG3 enhances the stability of p53 by directly binding to the protein. MEG3 can also inhibit expression of MDM2, which results in stabilization of p53. Both of these pathways lead to increased p53 activity, inhibition of proliferation, and activation of apoptosis. (C) pseudogene: PTENP1 functions as a competitive endogenous RNA for PTEN by sponging shared miRNA, thereby increasing levels of PTEN. Elevated PTEN levels can cause inhibition of AKT signaling, activation of apoptosis, inhibition of proliferation, and prevent malignancies such as renal cell carcinoma. (D) circRNA: circHIPK3 is an example of a tumor suppressive circRNA that acts as a sponge for miRNA (miR-558 in particular). miR-558 has been suggested to associate with the HPSE (heparanase) promoter and increase expression of heparanase. Therefore, sponging of miR-558 by circHIPK3 decreases heparanase levels which in turn leads to decreased MMP9 and VEGF. These changes result in decreased invasion, migration, and angiogenesis and can inhibit bladder cancer progression.

Oncogenes:

Research has shown that miRNAs can function as oncogenes, promoting abnormal cell growth and contributing to tumor formation. These miRNAs may directly inhibit the activity of tumor suppressors or work indirectly by removing the genetic brakes on oncogene activity. For example, miR-155 can promote abnormal B-cell proliferation, setting in process a series of changes that eventually lead to the development of leukemia and lymphoma (Babar et al., 2012; O’Connell et al., 2009). Importantly, delivery of antimiRs targeting miR-155 can inhibit tumor growth (Babar et al., 2012; Cheng et al., 2015) (Figure 1A). Another miRNA, miR-21, is overexpressed in pre B-cell lymphoma and has been implicated in other cancers, such as lung cancer, through targeting negative regulators of Ras signaling (Hatley et al., 2010; Medina et al., 2010). In glioblastomas, an oncogenic miRNA, miR-10b, is expressed at higher levels than in normal brain tissue and is required for tumor growth (El Fatimy et al., 2017). These oncogenic miRNAs exhibit the phenomenon of oncogene addiction (oncomiR addiction), since the tumors are dependent on the continued expression of the miRNAs for survival (Cheng and Slack, 2012), and are, thus, important potential targets in anti-cancer therapy. Recently, the potential for targeting oncomiRs uniquely overexpressed in an individual patient’s tumors was demonstrated, opening the way to personalized miRNA therapy as we discuss at the end of the review (Gilles et al., 2018).

Tumor suppressors:

Research suggests that miRNAs can also act as tumor suppressors, and when their function is lost, so is their protective power. For instance, the conserved miRNA let-7 represses RAS, a family of oncogenes implicated in about one-third of all human cancers (Johnson et al., 2007; Johnson et al., 2005). Consequently, let-7 expression can reduce levels of RAS, suggesting that it acts as a tumor suppressor and can be used as a new and promising therapeutic agent (Trang et al., 2010; Trang et al., 2011). As another example, miR-15a and miR-16-1 typically act as tumor suppressors but, when mutated or deleted, are associated with the development of chronic lymphocytic leukemia (CLL), the most common type of leukemia (Calin et al., 2002; Calin et al., 2005; Klein et al., 2010) (Figure 2A). Furthermore, miR-34a, a member of the conserved, redundant miR-34 family of miRNAs (Concepcion et al., 2012), regulates the expression of several oncogenes and is a direct downstream target of p53 (Adams et al., 2016a; Adams et al., 2016b; Kasinski and Slack, 2012; Liu et al., 2011). These miRNAs are also disabled in other types of cancer, such as multiple myeloma, mangle cell lymphoma, and prostate cancer (Volinia et al., 2006).

Context-dependent:

Some miRNAs function as either tumor suppressors or oncogenes, depending on the context. A notable example is miR-29, which helps prevent disease progression in the indolent form of B-CLL, but is also elevated in acute myeloid leukemia and the more aggressive form of B-CLL, implying that this miRNA can also function as an oncogene (Pekarsky and Croce, 2010).

Transfer RNA-derived small RNAs (also called tRNA fragments)

Just in the last decade, a new group of clinically-relevant small ncRNAs was discovered that are derived from tRNAs, which we will collectively refer to as tRNA-derived small RNAs (tsRNAs) (Cole et al., 2009; Lee et al., 2009); these molecules have also been referred to as tRNA fragments (tRFs) or tRNA-derived stress-induced RNAs (tiRNAs). Many aspects of their biogenesis and downstream mechanisms of action are still being unraveled, along with implementation of a universal nomenclature system. Transfer RNAs are transcribed by RNA Pol III as pre-tRNAs and then undergo modification and processing to produce mature tRNAs. A tRNA can generate different types of tsRNAs, depending on where cleavage occurs and whether it is in the pre-tRNA transcript or mature tRNA. The enzymes responsible for generation of tsRNAs are still being fully elucidated, but roles for ribonucleases such as Angiogenin, Dicer, and RNase Z have been noted in some instances (see recent reviews by (Balatti et al., 2017; Kumar et al., 2016; Saikia and Hatzoglou, 2015)). Several tsRNA databases have already been constructed for this new class of small ncRNAs to assist in their characterization. MINTbase v2.0 analyzed all datasets from The Cancer Genome Atlas (TCGA) and identified 26,531 unique tsRNA sequences from cancerous tissues (only fragments resulting from mature tRNAs were included in the algorithm) (Pliatsika et al., 2018). Despite an impressive number of distinct tsRNA molecules in cancer cells and some unique mechanisms of action, tsRNAs overlap with miRNAs in some aspects. Like miRNAs, tsRNAs have been found associated with Argonaute proteins, can mediate translational repression of mRNAs via binding to target 3’UTRs, and can be either oncogenic or tumor suppressive (Goodarzi et al., 2015; Kuscu et al., 2018; Maute et al., 2013). In a recent, exciting study, it was discovered that a tsRNA derived from a Leucine tRNA plays a role in global protein translation by regulating the expression of genes coding for ribosomal components (Table 1, Figure 1B). Importantly, the researchers show that this tsRNA is upregulated in liver tumors and that targeting the tsRNA with a LNA oligonucleotide in vitro and in vivo causes liver tumor cells to undergo cell death (Kim et al., 2017). These results demonstrate that tsRNAs are another potential new target for anti-cancer therapy.

PIWI-interacting RNAs

PIWI-interacting RNAs are a class of small RNAs approximately 21-35 nucleotides in length normally associated with the PIWI subfamily of Argonaute proteins and transposon silencing in germline development (Aravin et al., 2006; Girard et al., 2006). Species-specific pathways exist for production of piRNA precursors, and piRNA sequences are not very conserved. In mammals, piRNAs are processed from long, single-stranded transcripts whose genomic loci are clustered throughout the genome and transcribed by RNA Pol II, and approximately 20,000 piRNAs exist in the human genome (Ng et al., 2016). As noted above, to mediate their effects, miRNAs associate with the ubiquitously expressed AGO family of Argonaute proteins. In contrast, as their name implies, piRNAs are loaded onto proteins of the PIWI family of Argonaute proteins, which are often restricted to gonadal cells. The classic function of piRNAs is to silence transposons, and they do this in two ways. In the first mechanism, piRNAs guide PIWI proteins to nascent transposon transcripts and generate repressive chromatin states at target transposon promoters to silence their transcription. In the second scenario, piRNAs guide the PIWI complex to the transposon mRNA, where it cleaves the transcript (see (Ng et al., 2016; Ozata et al., 2019) for recent reviews). Although classically thought of as functioning only in gonadal cells, recent work shows that some of the tens of thousands of piRNAs are expressed in somatic tissues, albeit it at very low levels, and misexpressed in cancers, suggesting that these small ncRNAs might also be useful biomarkers (Martinez et al., 2015; Mei et al., 2015; Ng et al., 2016). However, the functional roles of piRNAs in somatic tissue and cancer are still being elucidated.

Long ncRNAs

Some of the earliest lncRNAs to be discovered were XIST and H19 (Bartolomei et al., 1991; Brown et al., 1991). Although initially identified through studies characterizing X-chromosome inactivation and embryonic development, respectively, these two genes are now among a long list of lncRNAs to be mechanistically linked to several types of cancers. Long ncRNAs are characterized as non-coding transcripts greater than 200 base pairs in length transcribed by RNA Pol II from independent promoters. Similar to protein-coding genes, their genomic locations are marked by enrichment of H3K4 trimethylation at the transcriptional start site and H3K36 trimethylation throughout the gene body. Long ncRNA transcripts consist of multiple exons that are spliced through canonical mechanisms into a mature transcript and usually include 5’caps and 3’poly(A) tails. However, lncRNAs have fewer exons and are expressed at lower levels overall compared to protein-coding transcripts (Cabili et al., 2011; Derrien et al., 2012; Iyer et al., 2015). Interestingly, lncRNAs are also not highly evolutionarily conserved, with only 5-6% of lncRNAs harboring conserved sequences (Iyer et al., 2015). In regards to this, there is a line of thinking that highly conserved IncRNAs may be more likely to be functional. However, there are primate-specific lncRNAs which likely could be involved in disease processes. RNA-sequencing (RNA-seq) has shown that there are much higher absolute numbers of unique lncRNA transcripts compared to protein-coding genes. A recent extensive cancer-centric study, MiTranscriptome, analyzed 7256 RNA-seq libraries primarily from human tissues (5298 primary tumors, 281 metastases, and 701 normal/benign) and identified 58,648 lncRNAs. Furthermore, these datasets allowed for an in-depth exploration of the landscape of lncRNAs in cancer and led to the identification of 7941 lncRNAs that were cancer- and/or lineage-specific (Iyer et al., 2015).

Compared to small ncRNAs, lncRNAs exhibit extensive mechanistic diversity to carry out their functional roles and, therefore, require a lengthier discussion in this area. Long ncRNAs can function either in cis or trans, meaning they mediate local effects near their own sites of transcription (cis), or they operate at distant genomic or cellular locations (trans). Different lncRNAs have also been shown to be able to influence gene expression at all levels-epigenetic, transcriptional, and post-transcriptional. Multiple lncRNAs bring other regulatory molecules (e.g., mRNAs, miRNAs, DNA) into proximity with one another and with proteins (e.g., chromatin modifying complexes, transcription factors, E3 ligases, RNA-binding proteins (RBPs)), essentially creating a flexible molecular scaffold that fosters the chemical interactions necessary to sustain cellular activity (see recent reviews, (Anastasiadou et al., 2018b; Kopp and Mendell, 2018)).

In terms of lncRNAs functioning by directly binding to protein complexes, one of the most well-characterized mechanisms is guiding of chromatin modifying complexes to target gene promoters to influence transcriptional repression/activation. Noted examples include the following: HOTAIR binds PRC2 and LSD1/CoREST/REST on its 5’ and 3’ ends, respectively, to target the complex to promoters and modulate histone methylation levels (Tsai et al., 2010); SChLAP1, an aggressive prostate cancer-specific lncRNA, interacts directly with the SWI/SNF nucleosome remodeling complex (Prensner et al., 2013); ANRIL interacts with components of PRC1 and PRC2 to silence genes in the CDKN2B/2A tumor suppressor gene cluster (Yap et al., 2010). Long ncRNAs also commonly bind transcription factors, which can have broad downstream effects on cellular transcriptional programs. Intriguingly, the GAS5 lncRNA acts as a mimic of a glucocorticoid response element (GRE), directly binding the DNA-binding domain of the glucocorticoid receptor (GR) and preventing it from activating its target genes, including those that prevent apoptosis (Hudson et al., 2014; Kino et al., 2010). Other direct interaction partners include PANDA and NF-YA transcription factors that are critical activators of p53-targeted cell death genes (Hung et al., 2011), PCGEM1 enhancement of c-Myc activity as a master regulator of metabolism (Hung et al., 2014), NKILA directly binding to the NF-KB/IKB complex to block phosphorylation of IKB by IKK (Liu et al., 2015), and DINO interacting with and stabilizing p53 following DNA damage (Schmitt et al., 2016). Long ncRNAs are often also found in direct contact with RBPs that regulate mRNA processing and stability. The effects of HOTAIR are mediated by binding to hnRNPA2/B1, which function as “matchmakers” to target HOTAIR/PRC2 to mRNA transcripts (Meredith et al., 2016). Recently, the novel, ultraconserved lncRNA THOR was discovered and shown to function as an oncogene by stabilizing the binding of IGF2BP1 to target mRNAs (Hosono et al., 2017) (Figure 1C). Finally, lncRNAs can function as scaffolds for regulatory molecules found in nuclear speckles and paraspeckles, exemplified by lncRNAs NEAT1 and MALAT1 (Clemson et al., 2009; Tripathi et al., 2010).

In addition to proteins, lncRNAs can directly bind to nucleic acids to mediate their molecular mechanisms of action. Two interesting examples of this include the recently identified lncRNA ARLNC1 that interacts with AR mRNA to regulate its cytoplasmic levels in prostate cancer (Zhang et al., 2018), and LincRNA-p21, which directly binds JUNB and CTNNB1 transcripts to repress their translation (Yoon et al., 2012). Another common mechanism involves lncRNA acting as a competitive endogenous RNA (ceRNA) or “sponge” for miRNAs. One of the first lncRNAs to be identified, H19, acts as a decoy for several tumor suppressor miRNAs, with let-7 as one notable example (Kallen et al., 2013). TUG1 has been shown to act as a sponge for miRNAs that target PTEN in prostate cancer (Du et al., 2016), as well as those that can target SOX2 and MYC in glioma cells (Katsushima et al., 2016). Long ncRNAs have furthermore been found to interact directly with DNA. GA-rich motifs in target promoters of genes involved in the TGFβ pathway attract GA-rich sequences of the MEG3 lncRNA via RNA-DNA triplex formation, and PRC2 is pulled along via interaction with MEG3 (Mondal et al., 2015).

Given the wide range of regulatory mechanisms and the diversity of downstream pathways affected, it is not surprising that many lncRNAs, including several of those highlighted above, have been found through in vivo experiments to be important contributors to cancer progression and represent possible therapeutic targets (Table 1). Due to their poor species conservation, most of these studies have relied upon modulating expression of lncRNAs in human cancer cell lines xenografted in mice. Like miRNAs, lncRNAs have been found to function either as tumor suppressors, oncogenes, or exhibit context-dependent roles (Table 1).

Oncogenes:

Many lncRNAs whose cellular roles have been elucidated function as oncogenes that promote tumor growth and are often overexpressed in cancer. HOTAIR is one of the most well-studied oncogenic lncRNAs that was initially characterized as a regulator of the HOX family of genes, which help control cellular identity (Rinn et al., 2007). However, a more global role for HOTAIR in controlling gene repression through targeting of PRC2 and LSD1/CoREST/REST was quickly noted (Gupta et al., 2010; Tsai et al., 2010). HOTAIR overexpression has been associated with poor outcomes in breast and several other cancers, possibly by increasing metastasis and tumor invasiveness (Balas and Johnson, 2018; Bhan et al., 2017; Gupta et al., 2010; Li et al., 2016b). Several novel oncogenic lncRNAs have been functionally described only in the past few years, including THOR (Hosono et al., 2017) (Figure 1C) and ARLNC1 (Zhang et al., 2018) mentioned previously, as well as SAMMSON (Leucci et al., 2016), DSCAM-AS1 (Niknafs et al., 2016), lncARSR (Qu et al., 2016a), CamK-A (Sang et al., 2018), and EPIC1 (Wang et al., 2018) (Table 1), and this list will undoubtedly continue to grow. SAMMSON has garnered much attention in particular lately as a melanoma-specific IncRNA that is required for growth and survival of cells, regardless of TP53, BRAF, or NRAS mutational status (Leucci et al., 2016). Mechanistically, SAMMSON promotes cell growth through interactions with p32, CARF, and XRN2 proteins that help to balance ribosomal RNA (rRNA) maturation and protein synthesis in the cytosol and mitochondria (Leucci et al., 2016; Vendramin et al., 2018). Importantly, GapmeR silencing of SAMMSON in patient-derived xenograft (PDX) models inhibits tumor growth, demonstrating the therapeutic potential of this lncRNA (Leucci et al., 2016).

Tumor suppressors:

Some lncRNAs act as safeguards against cancer development by preventing proliferation, activating apoptosis, maintaining genomic stability, or promoting tumor suppressor expression. For instance, MEG3 is one of the most well-characterized tumor suppressive lncRNAs. In addition to regulating the TGFβ pathway noted above, MEG3 downregulates MDM2 expression and increases p53 protein levels (Figure 2B); regulation of these and other pathways leads to decreased cell proliferation (Lu et al., 2013; Mondal et al., 2015; Zhang et al., 2010; Zhou et al., 2007). Consistent with its status as a tumor suppressor, copy loss of MEG3 and increased CpG methylation of its promoter is often noted in cancers (Balas and Johnson, 2018; Zhang et al., 2010). GAS5 is another lncRNA with known tumor suppressive roles often downregulated in cancers and with direct in vivo experimental evidence in breast and glioblastoma models (Bhan et al., 2017; Zhang et al., 2013; Zhao et al., 2015). In addition to affecting GR signaling, GAS5 is also involved in interactions with miRNAs that ultimately can lead to decreased cell proliferation and increased apoptosis, as well as decreased migratory potential (Hudson et al., 2014; Kino et al., 2010; Zhang et al., 2013; Zhao et al., 2015).

Context-dependent:

Many lncRNAs display both tumor suppressive and oncogenic functions. For example, in breast cancer, the lncRNA NKILA negatively regulates NF-KB signaling and downstream inflammation. In mouse xenograft models with human breast cancer cell lines overexpressing NKILA, this translates into reduced metastasis and increased survival, suggesting a tumor suppressive role (Liu et al., 2015). However, it was also recently shown that increased NKILA can promote tumor immune evasion, an oncogenic property, through activation-induced cell death of cytotoxic T lymphocytes (CTLs) and T_H1 cells (Huang et al., 2018).

Pseudogenes

Pseudogenes are a subclass of lncRNA transcripts that are similar in sequence to coding genes but have become inactivated and no longer result in functional proteins. RNA-seq studies have found the expression of thousands of pseudogenes from across the genome, with many of these shown to be specific to particular cancer types (Kalyana-Sundaram et al., 2012). Although pseudogenes are sometimes dismissed as useless evolutionary relics, research suggests that they perform vital functions and can both protect against or contribute to the development of cancer. One way that pseudogenes function is to act as decoys that divert miRNAs and other molecules away from their coding gene counterparts. The PTENP1 pseudogene is almost identical in size and structure to a coding gene, PTEN, which is a powerful tumor suppressor. When functioning normally, the pseudogene PTENP1 can act as a decoy to sponge miRNAs that would otherwise reduce the production of PTEN (Poliseno et al., 2010; Yu et al., 2014; Zheng et al., 2018) (Figure 2C). By acting in this manner as a ceRNA, the PTENP1 pseudogene functions as a tumor suppressor (Table 1). Other pseudogene ceRNAs can promote cancer growth. Recently, abnormal levels of the pseudogene BRAFP1 were shown to cause aggressive B-cell lymphoma in a mouse model. When overexpressed, the pseudogene operated as a miRNA decoy that increased the amount of BRAF protein (Figure 1D). In fact, the mice with the abnormal pseudogene developed cancer as rapidly as those with the protein-coding counterpart BRAF oncogene (Karreth et al., 2015). This suggests that far from being “dead” genes, pseudogenes may contribute directly to cancer development. Although the research is just beginning, it is becoming clearer that pseudogenes are more active in physiology and cancer than previously appreciated.

Circular RNAs

Within the IncRNA group are also circRNAs, which, as their name implies, are single-stranded covalently closed RNA molecules. These molecules are often generated through a back-splicing process, but the exact mechanisms of circRNA biogenesis remain to be fully elucidated. However, it has been shown that the parent gene expression of circRNAs is not a predictor of circRNA expression. Factors that do seem to be associated with circRNAs are longer flanking intron lengths and increased presence of intronic repetitive and reverse complement elements (Chen et al., 2019; Ivanov et al., 2015; Jeck et al., 2013; Vo et al., 2019). Several databases of circRNAs found in cell lines have emerged over the past few years, but studies centered on the direct identification of circRNA species in clinical samples were only recently conducted. In one study, a global analysis of circRNAs using clinical tumor samples (2000+) was performed across more than 40 cancer sites and led to the generation of the MiOncoCirc database. Notably, these efforts identified over 160,000 significantly expressed circRNAs (Vo et al., 2019). Another recent study, specifically in localized prostate cancer, identified 76,311 circRNAs through RNA-seq of prostate tumor specimens (Chen et al., 2019). Combined, these studies demonstrate the high prevalence of circRNAs in cancer. Interestingly, overall global circRNA abundance has been demonstrated to be negatively correlated with cell proliferation (Bachmayr-Heyda et al., 2015; Vo et al., 2019).

Despite widespread expression of circRNAs, most mechanisms of action and functional roles are still being elucidated. For the handful of circRNAs that have been functionally studied, mechanisms similar to other lncRNAs have been described, as well as the ability to function as oncogenes or tumor suppressors (Table 1). As one mechanism of action, like lncRNAs, circRNAs have been noted to sponge miRNAs, as is the case for circCCDC66, which has oncogenic functions in colorectal cancer (Hsiao et al., 2017) and circHIPK3, which has tumor suppressive roles in bladder cancer (Li et al., 2017) (Figure 2D). Interactions of circRNAs with miRNAs may also function to stabilize the miRNA binding partner, which was recently suggested to occur with circCSNK1G3 and miR-181 in prostate cancer, leading to enhanced cell proliferation (Chen et al., 2019). Circular RNAs have also been proposed to bind directly to proteins to either act as a scaffold for complex assembly, such as for circCTNNBI and DDX3 (Yang et al., 2019) (Figure 1E), or to sponge the protein of interest to prevent it from mediating its actions (Kristensen et al., 2018). It is also important to note that while we have discussed circRNAs in the context of ncRNAs, there have also been studies showing that some circRNAs may be translated (Legnini et al., 2017; Pamudurti et al., 2017). While functional studies and assessments of the therapeutic potential of circRNAs are only in their infancy, it is clear that circRNAs possess enhanced stability (Vo et al., 2019), making them prime candidates for novel cancer biomarkers, discussed further below.

Clinical relevance of ncRNAs

A search through the Clinicaltrials.gov database with keywords associated with ncRNAs shows hundreds of clinical trials involving ncRNAs (Supplemental Table 1), with many of these being cancer trials and the vast majority, over 800 trials, involving miRNAs. Given the high specificity of detection methods for ncRNAs and their tissue-specificity, as well as a plethora of supporting data in discovery studies, it is not surprising that the vast majority of these trials are diagnostic in nature, where the ncRNAs are used as biomarkers for a disease state or a prognostic marker for disease outcome. However, given the recent success of RNAi-based and oligo-based drugs (Levin, 2019; Pecot et al., 2011), a number of therapeutic clinical trials with ncRNAs have begun (Rupaimoole and Slack, 2017), albeit with mixed success. Below, we highlight notable ncRNA diagnostic and therapeutic cancer trials that are currently active, or just recently completed, and have the potential to impact patient care.

Non-coding RNAs as cancer biomarkers

Non-coding RNAs are useful molecular biomarkers since deregulated ncRNA expression is observed across diverse cancers. Numerous large-scale discovery studies in the pre-clinical setting have shown the promise of ncRNAs as diagnostic and prognostic biomarkers, with prominent examples outlined in Table 2. Notable examples can be found across both small (miRNAs) and long ncRNA (lncRNAs, circRNAs) classes. Although they are only beginning to be functionally described, circRNAs are a particularly intriguing class of biomarkers due to their enhanced stability, resulting from covalently closed ends resistant to exoribonucleases (Schwanhausser et al., 2013; Vo et al., 2019). Importantly, as shown in Table 2, several biomarkers can be found in blood or urine, where acquisition is much less invasive than that required for obtaining tumor biopsies. Expression of a given biomarker can detect the presence of cancer compared to normal samples, or predict survival, metastasis, or therapy response once cancer has been found.

Table 2.

Non-coding RNA cancer biomarkers

Name	ncRNA class	Cancer	Example of biomarker utility	Source material	References
miR-10b	miRNA	pancreatic	detection; prognostic for MFS and OS	tissue	Preis et al., 2011
miR-16	miRNA	lung	prognostic for OS	blood	Wang et al., 2013
miR-21	miRNA	breast	prognostic for OS	tissue	Yan et al., 2008
		colorectal	detection; prognostic for OS	blood; tissue	Toiyama et al., 2013
		leukemia	prognostic for OS	blood	Rossi et al., 2010
		lung	prognostic for CSS and RFS	tissue	Saito et al., 2011
		prostate	detection	blood	Agaoglu et al., 2011
miR-34a	miRNA	lung	prognostic for RFS	tissue	Gallardo et al., 2009
miR-155	miRNA	leukemia	detection; prognostic for OS	blood	Ferrajoli et al., 2013
miR-221	miRNA	prostate	detection	blood	Agaoglu et al., 2011
miR-375	miRNA	prostate	prognostic for OS	blood	Huang et al., 2015
miR-506	miRNA	gastric	prognostic for OS	tissue	Sakimura et al., 2015
		ovarian	prognostic for OS and PFS	tissue	Yang et al., 2013a
		pancreatic	prognostic for OS	tissue	Li et al., 2016a
miR-1290	miRNA	colorectal	detection; prognostic for OS and RFS	blood; tissue	Imaoka et al., 2016
		prostate	prognostic for OS	blood	Huang et al., 2015
CamK-A	lncRNA	breast	prognostic for OS and RFS	tissue	Sang et al., 2018
CCAT1	lncRNA	colorectal	prognostic for OS, CSS, and RFS	tissue	McCleland et al., 2016; Ozawa et al., 2017
CCAT2	lncRNA	colorectal	prognostic for OS and RFS	tissue	Ozawa et al., 2017
EPIC1	lncRNA	breast	prognostic for OS	tissue	Wang et al., 2018
FAL1	lncRNA	ovarian	prognostic for OS	tissue	Hu et al., 2014
H19	lncRNA	gastric	detection	blood	Zhou et al., 2015
HOTAIR	lncRNA	breast	prognostic for OS and MFS	tissue	Gupta et al., 2010
		colorectal	detection; prognostic for OS	blood; tissue	Kogo et al., 2011; Svoboda et al., 2014
		ESCC	prognostic for OS	tissue	Li et al., 2013
		ovarian	prognostic for OS in carboplatin-treated patients	tissue	Teschendorff et al., 2015
		pancreatic	prognostic for OS	tissue	Kim et al., 2013
HOTTIP	lncRNA	liver	prognostic for OS	tissue	Quagliata et al., 2014
HULC	lncRNA	liver	detection	blood; tissue	Panzitt et al., 2007; Xie et al., 2013
LINK-A	lncRNA	breast	prognostic for RFS	tissue	Lin et al., 2016
lncARSR	lncRNA	RCC	prognostic for OS and RFS; prognostic for PFS in sunitinib-treated patients	blood; tissue	Qu et al., 2016a; Qu et al., 2016b
MALAT1	lncRNA	lung	prognostic for OS in early-stage disease	tissue	Ji et al., 2003
		prostate	detection	blood; urine	Ren et al., 2013; Wang et al., 2014
NEAT1	lncRNA	prostate	detection; prognostic for MFS, BCR, and CSS	tissue	Chakravarty et al., 2014
PCA3	lncRNA	prostate	detection	urine	Hessels et al., 2003
PCAT-1	lncRNA	prostate	detection	tissue	Prensner et al., 2011
PCAT-14	lncRNA	prostate	detection; prognostic for OS, BCR, MFS, and CSS	tissue	Shukla et al., 2016; White et al., 2017
SChLAP1	lncRNA	prostate	prognostic for BCR, MFS, and CSS	tissue	Prensner et al., 2014b; Mehra et al., 2016
UCA1	lncRNA	bladder	detection	urine	Wang et al., 2006
circAR	circRNA	prostate	detection of mCRPC versus primary prostate cancer	tissue	Vo et al., 2019
circCCDC66	circRNA	colorectal	detection; prognostic for OS	tissue	Hsiao et al., 2017
circCTNNB1	circRNA	gastric	prognostic for OS	tissue	Yang et al., 2019
ciRS-7	circRNA	colorectal	detection; prognostic for OS	tissue	Weng et al., 2017

Abbreviations: BCR, biochemical recurrence; circRNA, circular RNA; CSS, cancer-specific survival; ESCC, esophageal squamous cell carcinoma; IncRNA, long non-coding RNA; mCRPC, metastatic castration-resistant prostate cancer; MFS, metastasis-free survival; OS, overall survival; PFS, progression-free survival; RCC, renal cell carcinoma; RFS, recurrence-free survival

Of the biomarkers listed in Table 2, PCA3 is the first and only ncRNA to receive FDA-approval as a cancer biomarker test to date. PCA3 is a prostate-specific marker often overexpressed in prostate cancer, and importantly, can be detected easily through non-invasive urine collection (Hessels et al., 2003). When combined with urine TMPRSS2:ERG levels, another transcript often overexpressed in prostate cancer, the diagnostic power of PCA3 is even stronger (Tomlins et al., 2011; Tomlins et al., 2016). With expression data from thousands of ncRNAs now available across disease sites, the potential to identify new single biomarkers (as in Table 2) or panels of biomarkers is endless. The four highlighted below involve compelling science and are illustrative examples of the hundreds of ongoing clinical trials centered on ncRNA-related biomarkers in cancer.

“Addition of microRNA Blood Test to Lung Cancer Screening Low Dose CT” (NCT03452514)

A trial sponsored by Hummingbird Diagnostics and conducted at Harvard Medical School and Lahey Clinic is currently enrolling to determine whether a miRNA profile (HMBDx, proprietary to Hummingbird) is superior to the specificity of low dose computed tomography (LDCT) for the diagnosis of lung cancer. Target enrollment is 400 participants who meet the criteria for lung cancer screening. Here, a blood draw will be used to screen for a serum miRNA signature (Keller et al., 2011a; Keller et al., 2011b; Leidinger et al., 2011) prior to LDCT and correlated with future outcome after a minimum 12 months of follow-up. Although the ability of CT to detect lung nodules is high, there are the caveats of false positives and repeated exposure to radiation. Being able to use a simple, non-invasive blood-based miRNA test would be a significant step forward for lung cancer diagnostics.

“Evaluating the Expression Levels of MicroRNA-10b in Patients with Gliomas” (NCT01849952)

Glioblastoma is one of the most aggressive types of cancer, and the development of novel early biomarkers and therapeutic targets is critically needed. As mentioned above, the oncogenic miR-10b is highly expressed in glioblastomas and required for tumor growth (El Fatimy et al., 2017). This multi-site clinical trial is being sponsored by and conducted at Dartmouth-Hitchcock Medical Center, in addition to Tufts Medical Center, Massachusetts General Hospital, and University of Vermont. The trial will test whether miR-10b expression levels in primary tumor, blood, and cerebrospinal fluid samples from glioblastoma patients can serve as a prognostic and diagnostic marker of glioblastoma. Patient survival, as well as tumor grade and genotypic variations, will be correlated to miR-10b expression levels in an anticipated 200 participants with brain tumors of glial origin.

“DICER1-related Pleuropulmonary Blastoma Cancer Predisposition Syndrome: A Natural History Study” (NCT01247597)

This large observational study, aiming to enroll 1500 participants, is sponsored by the National Cancer Institute (NCI). As mentioned above, Dicer1 is a ribonuclease involved in the miRNA biogenesis pathway. (Hill et al., 2009). This trial seeks to identify the specific types of cancer and benign neoplasms associated with an inherited mutation in DICER1, in particular in pleuropulmonary blastoma (PPB), a rare lung tumor that appears in young children. Researchers are collecting further medical history and genetic information on individuals and close relatives of individuals who have PPB or other rare tumors associated with PPB (e.g., cystic nephroma, nasal chondromesenchymal hamartoma, ovarian Sertoli-Leydig cell tumors, ocular medulloepithelioma) to better understand this disease.

“Non-coding RNA in the Exosome of the Epithelia Ovarian Cancer” (NCT03738319)

This case-control observational study sponsored by Peking Union Medical College Hospital will use next-generation sequencing to analyze the expression of miRNAs and lncRNAs in blood samples from 160 patients with high grade serous ovarian cancer (HGSOC) and benign gynecologic diseases. Differential expression of miRNAs and lncRNAs will be compared between cancer and control groups, and candidate ncRNAs will be validated as biomarkers for the detection and prognosis of ovarian cancer. Completion of this trial will hopefully lead to promising clinical ncRNA biomarkers that can be used to aide in early detection of this disease, which is often not diagnosed until an advanced stage (Giannopoulou et al., 2018).

Therapeutic targeting of ncRNAs in cancer clinical trials

As discussed throughout this review, ncRNAs are important regulatory molecules that can either drive or prevent oncogenic processes (Table 1, Figures 1 and 2). Accordingly, development of effective therapies to silence (for oncogenes) or overexpress (for tumor suppressors) ncRNAs has been an active area of investigation in recent years (Adams et al., 2017; Arun et al., 2018; Rupaimoole and Slack, 2017). In vivo inhibition of ncRNAs often requires oligonucleotide targeting of the ncRNA as well as a delivery method. For targeting of miRNA in which overexpression is desired, synthetic oligoribonucleotides consisting of a miRNA duplex are used (miRNA mimetics). Conversely, for achieving inhibition of oncogenic miRNA, single-stranded antisense RNAs are employed (antimiRs or antagomiRs) (Rupaimoole and Slack, 2017). Chemical modifications are often made to RNA drugs in order to increase stability and affinity in vivo. Common examples include locked nucleic acid (LNA) modification to lock the confirmation of the sugar phosphate backbone of the RNA and substitutions to the 2’-OH of the ribose sugar, such as 2’-O-methyl (2’-O-Me) or 2’-O-methoxyethyl (2’-O-MOE) modifications (Adams et al., 2017). In addition to optimization of targeting design, RNA drugs often require an in vivo delivery method, with lipid-based nanocarriers being the most well-established, and exciting new advances in peptide and polymer delivery systems also being developed (Adams et al., 2017; Rupaimoole and Slack, 2017). Long ncRNAs can be inhibited in vivo through approaches like antisense oligonucleotide (ASO) technology, which results in RNase H-dependent degradation of the target lncRNA (Arun et al., 2018). Specific chemical modifications can also be made to lncRNA-targeting ASOs to increase their stability, such as LNA, 2’-O-Me, or 2’-O-MOE mentioned above, S-constrained ethyl (cEt) modifications, or “GapmeR” RNA-DNA-RNA hybrid ASOs. This optimization has even allowed for development of the next-generation of ASOs with increased potency that are freely uptaken in vivo, without a delivery method (Hong et al., 2015; Yamamoto et al., 2015).

To date, most studies employing ncRNA-targeting drugs in the cancer setting have been in the pre-clinical stage, but these studies have nonetheless highlighted the promise of this strategy. For example, ASO-based depletion of the lncRNA MALAT1 can impact growth and metastasis of lung and breast cancer cells in murine models (Arun et al., 2016; Gutschner et al., 2013). Below, we expand upon the two therapeutic clinical trials that have been completed, and one which has reported preliminary results, all of which targeted miRNAs. Since no therapeutic trials directly targeting the other classes of ncRNAs have been completed, an alternative strategy harnessing the H19 lncRNA promoter in a clinical trial in bladder cancer is discussed. As research and technology progresses, we are sure to see more ncRNA-based therapies advance to the clinic.

“A Multicenter Phase 1 Study of MRX34, MicroRNA miR-RX34 Liposomal Injection” (NCT01829971/NCT02862145)

This international study was sponsored by Mirna Therapeutics (now Synlogic Therapeutics) and conducted at multiple sites in the USA and Korea. It evaluated the safety of MRX34 in patients with unresectable primary liver cancer, select advanced or metastatic cancer with or without liver involvement, or hematologic malignancies. The drug was a liposomal miR-34 mimetic given intravenously. As mentioned above, miR-34 is a tumor suppressive miRNA that is a direct target of p53 and regulates the expression of several oncogenes (Adams et al., 2016a; Adams et al., 2016b; Kasinski and Slack, 2012; Liu et al., 2011). Forty-seven patients received drug and a maximum tolerated dose (MTD) was determined, while a handful of patients experienced partial responses (Beg et al., 2017). From this study, the authors concluded that MRX34 showed some anti-tumor activity and was safe. Unfortunately, in a Phase 1b study, multiple patients experienced immune-related toxicities, some patients experienced severe (Grade 4) cytokine release syndrome, and the trial was terminated. Since the final results of the trial have not been published, it is unclear why these patients experienced severe adverse events (SAEs), but it should be noted that both lipid-based nanoparticles and small double-stranded RNAs have been reported to cause immune toxicity (discussed in (Levin, 2019)). A clinical trial (NCT02716012) using the same lipid nanoparticle, known as a Smarticle, but with a different dsRNA (targeting the protein-coding gene CEBPA (Reebye et al., 2018; Voutila et al., 2017)) from MiNA Therapeutics recently reported no SAEs at the 2019 European Society for Medical Oncology Congress.

“MesomiR 1: A Phase 1 Study of TargomiRs as 2nd or 3rd Line Treatment for Patients with Recurrent MPM and NSCLC” (NCT02369198)

This study, sponsored by the Asbestos Diseases Research Foundation and EnGeneIC Limited and conducted at the University of Sydney, evaluated the MTD of TargomiRs in patients with recurrent malignant pleural mesothelioma (MPM) and non-small cell lung cancer (NSCLC). The intravenously-delivered TargomiR drug was a minicell containing a double-stranded, 23 base pair, synthetic miRNA mimic based on tumor suppressive miR-16, discussed previously (Reid et al., 2013) (Figure 2A). The drug delivery vehicle was constructed from nonliving bacterial minicells coated with a targeting moiety, an anti-EGFR antibody. This specifically targeted the particle and its miRNA payload to EGFR-expressing cancer cells. 26 patients were treated with TargomiRs, and a MTD was determined (van Zandwijk et al., 2017). One patient showed a partial response to therapy, leading the investigators to conclude that their studies supported further trials, perhaps in combination with other current therapies.

“Safety, Tolerability, and Pharmacokinetics of MRG-106 in Patients with Mycosis Fungoides (MF), CLL, DLBCL (Diffuse Large B-Cell Lymphoma), or ATLL (Adult T-Cell Leukemia/Lymphoma)” (NCT02580552/NCT03713320)

The objectives of these studies, sponsored by miRagen Therapeutics and conducted at multiple sites in the USA and Canada, are to evaluate the safety and potential efficacy of cobomarsen, also known as MRG-106, in patients diagnosed with a subset of lymphomas and leukemias. Cobomarsen is a LNA designed to inhibit the activity of miR-155 (Figure 1A), an oncogenic miRNA found at high levels in these types of cancers and important for the growth and survival of MF cancer cells (Seto et al., 2018). In preliminary results presented at the American Society for Hematology in 2018, the authors showed that cobomarsen was well tolerated in 43 patients treated either intratumorally or systemically. Furthermore, some patients had durable partial responses.

“Codex: Study of Inodiftagene Vixteplasmid (BC-819) in Unresponsive NMIBC” (NCT03719300)

H19 is a lncRNA that possesses several oncogenic mechanisms of action and is often upregulated in different types of cancer (Bhan et al., 2017). Due to its upregulation in cancer cells, researchers have used its promoter to deliver cancer-specific expression of downstream sequences. One compound, BC-819 (also known as DTA-H19), is a DNA plasmid expressing diphtheria toxin A (DTA) under the regulation of the H19 promoter, delivered as a complex with polyethyleneimine (PEI). BC-819 has completed Phase 1/2a trials in pancreatic, ovarian, and bladder cancers, with current clinical development focused on bladder cancer, where H19 is often suppressed in normal cells but upregulated in tumor cells (Gofrit et al., 2014). This Phase 2, multi-site trial sponsored by Anchiano Therapeutics will examine 140 patients with NMIBC (non-muscle-invasive bladder cancer) who are unresponsive to Bacillus Calmette-Guerin (BCG) therapy, a standard treatment. BC-819 will be directly instilled into the bladder, and several measures of clinical efficacy will be assessed throughout the study. While not directly targeting a ncRNA, this is an interesting example of how properties of ncRNAs, like H19 cancer-specific expression, can be harnessed for novel cancer therapeutic interventions.

Perspectives and future opportunities

Looking forward, ncRNAs are likely to continue to better enable researchers and clinicians to discover molecular signatures that help fine-tune diagnoses and differentiate malignant from benign tissue, as well as one type of cancer from another. These tests will also likely be prescribed to determine prognosis, such as determining risk of metastasis or probability of responding to treatments like chemotherapy. Although the therapeutic applications are still at an early stage, it is expected that drugs based on ncRNAs involved in cancer pathways will one day be prescribed for cancer patients. For identifying candidate biomarkers or therapeutic targets, focusing on lineage and disease-specific ncRNAs seems to be the ideal approach, as is the case with PCA3 mentioned above in prostate cancer (Hessels et al., 2003). Level of expression is also a key factor for developing ncRNA-based diagnostics and therapeutics. As noted previously, lncRNAs are often expressed at lower levels compared to protein-coding genes (Derrien et al., 2012; Iyer et al., 2015), so it is important to identify those candidates that can be readily detected. Furthermore, development of diagnostic biomarkers found in urine or blood is ideal to spare patients the non-invasive procedures often associated with tissue collection.

Compared to biologics or small-molecules, there are important benefits and challenges associated with RNA-based therapeutics that should be taken into consideration. RNA drugs are based on nucleotide hybridization, so targets can be easily approached; one just needs to identify the sequence and test candidate therapeutics for activity. With small-molecules or biologics, lengthier and laborious screens or structure-based design approaches must be employed. There is also the potential to design RNA-based approaches that can easily target multiple combinations of RNAs. Despite these advantages, the delivery method for RNA-based drugs has to be taken into account and optimized, and this can often be a challenging process. Oral formulations are also more difficult for RNA-based therapeutics compared to small-molecules and biologics. As noted above, there have further been issues with immune-related toxicity and other adverse events from RNA-based drugs. Improvement of oligonucleotide chemistry and delivery methods are continuously being pursued to mitigate these issues in future studies. Finally, the cost of gram quantity small-molecules can be much less than RNA-based drugs. We point the reader to excellent recent reviews (and references within) on the topics of the preceding two paragraphs for more discussion (Adams et al., 2017; Levin, 2019; Rupaimoole and Slack, 2017; Setten et al., 2019).

Members of our group recently showed the potential of targeting oncomiRs that were upregulated in an individual cancer patient’s tumor using patient-derived organoids (PDO) and xenografts (PDX) (Gilles et al., 2018), opening up the exciting possibility of personalized RNA medicine. Currently, each cancer patient is assigned to a dedicated care team consisting of a specialized oncology surgeon, medical oncologist, radiation oncologist, and pathologist. It is not too far-fetched to envision that tumor boards will add an integrated ncRNA workflow to the standard of care for every cancer patient. This could include the following components: a complete RNA-seq (including ncRNAs) performed on the cancer patient’s tumor and serum; a complete whole-genome sequence on the tumor and germline genome; an integrative bioinformatics and pathology analysis to quantify malignant pathologic phenotypes in the patient’s cancer and to identify significantly altered genomic and ncRNA alterations; a pathology report that includes an interpretation of the tumor pathology, DNA and RNA data, and potential therapeutic vulnerabilities, which would be presented at tumor boards to aid in therapeutic decisions; PDO and PDX models grown and therapeutic approaches tested on the PDO and PDX (e.g., traditional drug testing and ncRNA-based drug screening centered on the RNA profiles); recommendations for possible ncRNA-based therapy; a searchable file of the patient’s DNA/RNA and prognostic data placed in the electronic medical record of the patient for later reanalysis if necessary and if additional ncRNA information becomes known. The discovery of ncRNAs has opened up a new chapter in the history of medicine, one that promises to revolutionize the way that cancer and other diseases (discussed in (Adams et al., 2017; Levin, 2019; Rupaimoole and Slack, 2017)) are diagnosed and treated, and we have just begun to scratch the surface.