Noncoding RNAs in tumor metastasis: molecular and clinical perspectives

Abstract

Metastasis is the main culprit of cancer-associated mortality and involves a complex and multistage process termed the metastatic cascade, which requires tumor cells to detach from the primary site, intravasate, disseminate in the circulation, extravasate, adapt to the foreign microenvironment, and form organ-specific colonization. Each of these processes has been already studied extensively for molecular mechanisms focused mainly on protein-coding genes. Recently, increasing evidence is pointing towards RNAs without coding potential for proteins, referred to as non-coding RNAs, as regulators in shaping cellular activity. Since those first reports, the detection and characterization of non-coding RNA have explosively thrived and greatly enriched the understanding of the molecular regulatory networks in metastasis. Moreover, a comprehensive description of ncRNA dysregulation will provide new insights into novel tools for the early detection and treatment of metastatic cancer. In this review, we focus on discussion of the emerging role of ncRNAs in governing cancer metastasis and describe step by step how ncRNAs impinge on cancer metastasis. In particular, we highlight the diagnostic and therapeutic applications of ncRNAs in metastatic cancer.

Introduction

Metastasis is defined as the spread of neoplastic cells from their original site to other regions of the body, thus resulting in the development of secondary tumors [1, 2]. It is now established that metastasis represents a complex and multi-step process termed metastatic cascade. This process can be parsed into distinct steps: local invasion at the primary site, entry into the vasculature (intravasation), transport and survival in the circulation, arrest, and exit from the distant capillary bed (extravasation), and colonization within the secondary organ parenchyma [1, 3]. Clinically, metastasis, accounting for more than 90% of all cancer mortality [1], causes not only direct damage to the affected organs, but also systemic syndromes including endocrine abnormalities, malignant effusions, hematologic disorders, and cachexia [4].

Previous studies have advanced to unveil the molecular basis of this life-threatening process. Several driving oncogenic mutations, including ERBB2 in breast cancer, BRAF in melanoma, ALK in lung cancers, and ZFP36L2 in colorectal cancer have been revealed to be crucial for metastatic disease [5, 6]. Epithelial–mesenchymal transition (EMT), an event during embryonic development, is exploited by cancer cells to acquire undifferentiated morphology and invasive traits, thereby facilitating metastasis [7]. Besides, extracellular matrix (ECM) degradation proteases, cell adhesion molecules, Rho/ROCK-mediated cytoskeleton motility, and angiogenetic process have also shown a substantial effect in programs enabling migration from primary sites [1]. What’s more, pro-survival signaling such as AKT pathways and signaling favoring immune evasion such as TGF-β pathways are key features of metastatic cancer that control the survival of disseminated tumor cells [5]. Aside from these general mediators of metastasis, context-dependent pathways mediating tumor-stroma interactions in the particular foreign microenvironments have essential roles for the adaptability and survival of metastatic tumor cells at ectopic sites. However, with the recent discovery of thousands of noncoding transcripts (noncoding RNAs, ncRNAs) which account for approximately 30% of the human genome, we must admit that there is still a long way to go to fully understand the mechanisms of cancer metastasis. In the human genome, only 1–2% of the DNA codes for proteins. In the past, the remaining 98% noncoding sequences were considered as “evolutionary junk”. Increasing evidence indicates that instead, these RNA transcripts have a tremendous impact on diverse physiological and pathological processes [8]. To date, high throughput transcriptome analyses have identified a myriad of ncRNAs that have surpassed the number of protein-coding genes in the human genome, and the number is still rapidly expanding [9, 10]. Indeed, many/most cancer-associated mutations reside within the non-coding regions rather than in the coding regions in the human genome [11]. Using the length of 200 nt as a criterion, those ncRNAs are divided into two subclasses: small ncRNAs (such as microRNAs (miRNAs), PIWI-Interacting RNAs (piRNAs)) and long noncoding RNAs (LncRNAs) [9].The characterization of such an expanded repertoire of ncRNAs is shedding new light on the molecular mechanisms of metastasis. For instance, the first metastasis-associated miRNA miR-10b, which promotes metastasis via regulating RhoC expression, was identified in breast cancer [12]. Moreover, the miR-200 family has been shown to play a key role in regulating EMT program [13]. Metastasis-associated lung adenocarcinoma transcript 1 (MALAT1), one of the first lncRNAs discovered, has been related to a high risk of metastatic progression in patients with lung cancer [14].

In this review, we mainly focus on recent findings of ncRNAs and their regulatory roles in distinct steps of the metastatic cascade to discuss how ncRNAs orchestrate a series of pathological events such as EMT, transendothelial migration, anoikis, immune evasion, dormancy, etc. We also provide a brief introduction to the biogenesis and function of ncRNAs and summarize the clinical value of ncRNAs in the diagnosis and treatment of cancer metastasis. Overall, we aim to provide a perspective for future research directions and hope that an understanding of the role of ncRNAs in cancer metastasis can help more bench discoveries be translated into clinical application.

Biogenesis and function of ncRNAs

Biogenesis and function of miRNAs

miRNA is a well-characterized class of small ncRNA, defined as single-stranded transcripts of ~ 20nt in length that regulate gene expression post-transcriptionally [15, 16].The biogenesis of miRNAs involves transcription by RNA polymerase II from corresponding miRNA-encoding genes [17], being cleaved by the Drosha and DGCR8 complex [18, 19], transported to the cytoplasm [20], processed by Dicer [21], and ultimately, formation of RNA-induced silencing complex (RISC) by association with Argonaute [22].

Mechanistically, the 5ʹ end (known as the seed sequence) of the miRNAs binds to the 3ʹ untranslated region (3ʹUTR) of target mRNAs by Watson–Crick pairing [22, 23]. This interaction can lead to inhibition of mRNAs of key metastasis genes by either blocking the translational machinery or promoting its degradation (Fig. 1). For instance, in prostate cancer, miR-34a binds to 3′ UTR of CD44 (an adhesion molecule) mRNA and result in its degradation, thereby inhibiting metastasis to the lung [24]. As another example, miR-506 targets the 3′-UTR of SNAI2 (a transcription factor) mRNA and inhibits its expression so as to upregulate metastasis-suppressing genes and attenuate cell migration and invasion in ovarian cancer [25]. miRNAs exert a far-reaching influence on gene regulation; it is estimated that as high as 60% of coding genes are affected by miRNAs [26].

Fig. 1.

Functional mechanisms of ncRNA. a miRNAs bind to the 3ʹUTR of target mRNAs by Watson–Crick pairing via their 5’ end, leading to inhibition of mRNAs by either blocking the translational machinery or promoting its degradation. b In the nucleus, lncRNAs can interact with nucleic acids, chromatin-modifying enzymes, and transcription factors to influence transcriptional programs of genes. In the cytoplasm, they can sponge miRNAs and/or proteins to block their functions, bind to proteins and guide them to specific subcellular localization, bind to mRNAs to regulate their translation, and act as scaffolds for different molecules to favor their intermolecular interactions. c circRNAs bind to transcripts or proteins to suppress their function or to act as scaffolds for enhanced intermolecular interaction. d piRNAs bind to transposon RNA to mediate transposon gene silencing. They can also associate with target transcripts to favor heterochromatin formation, leading to their epigenetic silencing of the target loci

Biogenesis and function of lncRNAs

LncRNA is a subclass of non-coding transcripts that is longer than 200nt [27]. LncRNAs share something in common with their protein-coding counterparts exemplified by the fact that they are both transcribed by RNA polymerase II, capped and generally polyadenylated, and spliced [28]. Similar to mRNAs, lncRNAs can arise from almost anywhere in the genome and in different directions [10]. Besides, lncRNAs form unique secondary structures which to some extent determine their distinct functions [29]. After biogenesis, lncRNAs either remain in the nucleus or are exported to the cytoplasm [30].

Compared to other ncRNAs, the mechanisms by which lncRNAs exert their functions are more complex and diverse. LncRNAs can cis- or trans-regulate genes at both transcriptional and post-transcriptional levels. Mechanistically, lncRNAs exhibit four distinct modes of action (Fig. 1), including function as decoys to block biomolecular functions [9], guidance the specific subcellular localization of proteins, function as scaffolds for intermolecular interactions, and interaction with nucleic acids, chromatin-modifying enzymes as well as transcription factors to influence transcriptional programs of genes. These impact various aspects of tumor metastasis. For instance, lncRNA Gas5 acts as an endogenous sponge to bind to and inactivate miR-222 in glioblastoma to attenuate cell invasion and migration [31], while its role as a protein decoy for glucocorticoid receptor has also been shown in various cancer cell lines[32]. In addition, lncRNA HOTAIR recruits polycomb repressive complex 2 (PRC2) to alter the H3K27me3 patterns and facilitate transcription of genes involved in metastasis in breast cancer [33].

Biogenesis and function of other ncRNAs

Other ncRNAs include circular RNA (circRNA), piRNA, small nucleolar RNA (snoRNA), pseudogenes, tRNA-derived small RNAs (tsRNAs), etc. Here we will briefly describe a few of the relatively well-studied types in cancer biology.

Circular RNA (circRNA) is a kind of single-stranded RNA in which the 3' and 5' ends join together to form a covalently closed loop, with the length of 100nt ~ over 4 kb [34, 35]. Comparing with lncRNAs, circRNAs are more stable than linear transcripts due to the lack of 3ʹ and 5ʹ ends, indicating a huge potential for cancer detection [36, 37]. So far, several action modes of circRNAs have been identified (Fig. 1). As one mechanism of action, circRNAs can act as sponges directly binding to biomolecules [38, 39], resulting in titrating away proteins or miRNAs which are related to metastasis. For example, circCCDC66 serves as a sponge for a group of miRNAs to protect several oncogenic mRNAs (e.g., MYC) from degradation, thus promoting colon cancer metastasis [40]. CircRNAs may also function as scaffolds to assemble various molecules involved in classical metastatic pathways, such as Wnt signaling pathway [41]. Of note, some studies even have proposed the protein coding potential of a subset of circRNAs [42, 43]. One of such cases reported in cancer metastasis is circPPP1R12A, which was shown to be translated into a functional protein and promote liver metastasis of colon cancer [44].

piRNA is a type of small ncRNA of 24 ~ 32nt in length whose canonical function is to silence transposons in germline cells thus maintain genomic integrity [45]. While miRNAs bind to AGO of Argonaute proteins to regulate gene expression, as the name suggests, piRNAs associate with the PIWI subfamily of Argonaute proteins [46]. Although thought as germline-specific in the first place, recent studies have begun to reveal functions of piRNAs in epigenetic modification and genome rearrangement in somatic cells, and emerging evidence suggests that this subclass of ncRNAs is also aberrantly expressed in multiple metastatic carcinoma types [47]. However, understanding of the biological mechanisms of piRNAs in cancer metastasis currently remains in its infancy (Fig. 1). Recent publications have brought new insight into the underlying mechanisms. In colorectal cancer, piR-1245 favors metastasis probably via a piRNA-dependent mRNA degradation of a panel of tumor suppressor genes [48]. Whereas in breast cancer, piR-36712 interacts with RNAs of SEPW1P (pseudogene of SEPW1) to promote targeting of SEPW1 RNA by miRNAs, thus augmenting P53-mediated metastasis-suppressing signaling [49].

The role of ncRNAs in metastatic cascade

The metastatic cascade is a simplified description in sequential order of how primary cancer cells disseminate and colonize ectopic sites [2]. At the outset of this journey, cancer cells break through the confinement of the primary tumor, attributing to the initiation of EMT [50]. Next, after detachment from the primary site, cancer cells enter into the blood vessels, which involves angiogenesis and proteolytic function [51]. Subsequently in the circulation, cancer cells are present as either single cells or as emboli with platelets and/or leukocytes to protect them from shear forces, immune attack, and anoikis [52]. In the capillary of distant organs, cancer cells arrest via physical trapping or selective adhesion and extravasate [51]. Afterward, the vast majority of extravasated cells will succumb, while the rest of them can survive in a dormant state [3]. Ultimately, cancer cells must adapt to and remodel the microenvironment to establish a macrometastatic lesion from micrometastases [1].

This whole cascade is orchestrated by a cohort of molecular pathways such as RhoGTP and EMT signaling, together with genes including cell adhesion molecules, matrix degradation protease, cytoskeleton, chemokines, and angiogenic factors [3]. There is mounting evidence that ncRNAs play a critical role in the regulation of these pathways and genes [8, 9]. In this section, we will classify the related studies by specific steps to discuss the regulatory landscape of pathways and genes by ncRNAs in distinct metastatic processes.

ncRNAs-mediated local invasion

Local invasion involves detachment from the primary site, disruption of the basement membrane, and penetration to the adjacent parenchyma [1]. Epithelial cell–cell adhesions are mediated mainly by cadherin, particularly E-cadherin, whose loss is robustly correlated with enhanced metastatic potential in multiple carcinoma types [1]. The dissolution of E-cadherin can be attributed to the EMT program, which also leads to dramatic alterations in cell shape from the polarized and organized epithelial morphology to the nonpolarized shape with heightened invasiveness [53]. Along with the loss of intercellular adhesions, cancer cells switch the pattern of cell–matrix adhesion molecules (e.g., integrins), favoring their passage to the stroma [50]. Integrin, a family of transmembrane linkers composed of α and β subunits, acts as a context-dependent function during metastasis through its inside-out and outside-in signaling [54]. In addition, active proteolysis (e.g., by matrix metalloproteinases (MMPs) and serine proteinases) and cytoskeleton-mediated motility (e.g., by Rho/ROCK regulated actin contractility) are also requisite [3] (Fig. 2).

Fig. 2.

Role of ncRNA in local invasion. Multiple ncRNAs that regulate local invasion are shown, acting on diverse hallmarks of cancer metastasis including EMT, cell adhesion, ECM degradation, and cell motility

The most canonical mechanism by which miRNAs regulate the EMT program would be the double-negative feedback loop formed by pairs of miRNAs and EMT-TFs (transcription factors that promote EMT-associated gene expression). For instance, miR-200 suppressed the expression of EMT-TFs ZEB at the post-transcriptional level, in turn, ZEB targeted miR-200 and repressed its transcription [55]. Similar feedback loops also existed between miR-34a and Snail [56], miR‑1/miR‑200b and SNAIL2 [57], and miR‑203 and SNAIL1 [58]. In addition to these loops, it has long been shown that a mass of miRNAs, such as miR-8084 [59], miR-484 [60], miR-708-3p [61], miR-506 [25], miR-4521[62] can modulate the level of EMT-TFs or EMT markers in a great variety of metastatic cancers.

A plethora of lncRNAs has been implicated in regulating EMT either by impinging on EMT-TFs and/ or EMT-associated markers. A previous study demonstrated that oncogenic lncRNA ZEB1-AS1 could induce EMT in hepatocellular carcinoma (HCC) cells by stimulating the transcription of EMT-TFs ZEB1 [63]. At the same time, lncRNA MEG3 was shown to promote EMT by inhibiting the expression of E-cadherin through epigenetic mechanisms in lung cancer cells. MEG3 recruited PRC2, an essential epigenetic transcriptional repressor, to induce inactive H3K27me3 markers on the promoter of E-cadherin gene [64]. In addition, lncRNA JPX also mediated the metastatic potential of lung cancer, where it sponged miR-33a-5p to abrogate the inhibitory effect on Twist [65].

Recently, novel circRNAs have also been proved to be the underlying mechanism of EMT program [66]. For instance, in triple-negative breast cancer, CircANKS1B functioned as a sponge for miR-148a-3p and miR-152-3p, thereby augmenting the translation of transcription factor USF1. Then, USF1 induced the transcription of TGF-β1, triggering TGF-β1/Smad signaling to activate EMT program [67]. At present, increasing numbers of laboratory studies are revealing the roles of circRNAs in EMT in a variety of cancer types, such as circNEIL3 in pancreatic cancer [68], circCORO1C in laryngeal squamous cell carcinoma [69], and CircNR3C2 in triple-negative breast cancer [70]. Likewise, piRNAs were introduced as new players in EMT, such as piR-1037 in human oral squamous cell carcinoma cells [71] and piR-932 in breast cancer stem cells [72], though the underlying molecular mechanisms have yet to be elucidated.

What’s more, miRNA has been involved in the regulation of integrin pattern. For example, in breast cancer, miR-31 downregulated the expression of α2β1 integrin (a metastasis repressor [73]) and facilitated metastasis to the lung [74]. In another model of breast cancer, miR-221/222 suppressed the level of α6β4 integrin (a metastasis promoter) and inhibited invasion [75]. Recently, MALAT1, one of the most well-studied metastasis-promoting lncRNA, has been shown to repress transcription factor TEAD to inhibit the expression of α6β4 integrin, leading to suppression of breast cancer metastasis in a transgenic mouse model [76]. Other than MALAT1, H19 [77], CASC9 [78], HOXD-AS1 [79] have also been shown to affect ECM-integrin interactions in multiple metastatic carcinoma types. Besides, evidence from pan-cancer bioinformatics analysis demonstrated that circRNA CDR1as exerted its effects on matrix organization, integrin binding, and collagen binding, indicative of a pro-metastatic role [80].

ncRNAs can also act as modulators for MMPs. One of such cases was that transcription-regulating kinase CDK8 stimulated colorectal cancer metastasis through effects on TGFβ-regulated miR-181b expression. Mechanistically, miR-181b targeted and inhibited the expression of TIMP3, the MMP inhibitor, thus resulting in the upregulation of MMPs [81]. LncRNAs are also regulators of MMPs in local invasion exemplified by recent evidence including ODRUL [82], MATN1-AS1 [83], PRECSIT[84], LSAMP-AS1 [85], CASC2 [86], and others. In gastric cancer, an in vivo research showed that cirRNA 0,000,096 upregulated MMP-2 and MMP-9 to confer increased tumor invasiveness [87]. As another example, circ0001361, newly identified in bladder cancer, was proved to enhance the local invasion of cancer cells by sponging miR-491-5p and upregulating MMP-9 [88]. Furthermore, piR-39980 increased the expression of MMP-2 and downregulates SERPINB1(an endogenous inhibitor of serine proteases) to promote migration and invasion in human osteosarcoma cells [89].

It is worth mentioning that miRNAs also regulate cellular motility, for instance, miR-124 [90, 91], miR-138 [91], miR-BART2-5p [92] and miR-200b [93], and so forth. have been demonstrated to target RHO/ROCK cytoskeletal remodeling pathway in cancer metastasis. Similarly, lncRNAs can affect the motility of cancer cells by regulating cytoskeleton reorganization. In HBV-related HCC, Dreh could suppress cancer cell migration by directly binding to intermediate filament vimentin, leading to its inactivation, and further altered the cytoskeleton structure [94]. As another example, lncMER52A, newly identified exclusively in HCC, was shown to enhance metastasis by stabilizing p120-catenin and triggering the p120-ctn/Rac1/Cdc42 axis [95]. Circular RNAs have also been implicated in the enhanced cellular motility during metastasis. A previous study showed that circSKA3, directly binding to integrin β1 and Tks5, contributed to the formation of actin-based invadopodia, thereby facilitating breast cancer invasion [96]. In short, ncRNAs have been implicated in every aspect of local invasion.

ncRNAs-mediated intravasation

Intravasation involves cancer cells that have undergone local invasion entering into nearby capillaries or lymphatic vessels [97]. This active process requires intercellular adhesions between cancer cells and endothelial cells (ECs), disruption of the endothelial integrity, and augmented motility to cross the barriers of microvessel walls [98]. Equally important is the ability to stimulate the formation of neovasculature within the primary malignant lesion, which is mediated by several angiogenic factors and growth factors such as VEGFA, ANGPTL4, PDGFB, FGF2, and so on [99]. In general, tumor-associated microvasculature is tortuous and leaky compared with normal blood vessels; therefore, neoangiogenesis is believed to facilitate intravasation [100] (Fig. 3).

Fig. 3.

Role of ncRNA in intravasation and survival in the circulation. Direct or indirect regulations of intravasation and CTCs survival by several ncRNAs are shown. These include diverse hallmarks of cancer metastasis such as a. angiogenesis b. enhanced vascular permeability, c. immune evasion, and d. anoikis resistance

Accumulating evidence has suggested multifaceted roles of ncRNAs in intravasation by the regulation of vascular permeability. In the local microenvironment of lung cancer, miR-23a was revealed to dissolve the tight junction and promote vascular permeability via targeting Zonula occludens protein-1 (ZO-1) in ECs, thus facilitating entry into the bloodstream [101]. A similar mechanism has been functionally described in HCC, where miR-103 disrupted endothelial integrity to promote transendothelial migration by post-transcriptional repression of VE-Cadherin, p120-catenin, and ZO-1 in ECs [102]. Recently, miR-629-5p was shown to break down the endothelial barrier through inhibition of Cadherin EGF LAG seven-pass G-type receptor 1(CELSR1) in ECs within tumor microenvironment of invasive lung cancer [103]. LncRNAs also regulate the tight junctions of endothelial cells to influence tumor cells entry into the circulation. For example, in gastric cancer, depletion of lncRNA MALAT1 in human umbilical vein endothelial cells (HUVECs) resulted in elevated vascular permeability in the local environment by inhibiting the expression of VE-cadherin and β-catenin [104].

Growing number of studies suggested that circRNAs can also act as regulators of vascular permeability. A recent study showed that exosomal circ-IARS derived from pancreatic cancer cells was captured by HUVECs, resulting in increased endothelial permeability. This was achieved through competitive inhibition of miR-122, thus upregulating the expression of downstream target RhoA, which led to increased expression of F-actin and decreased expression of ZO-1. Clinical data showed that Circ-IARS expression was positively correlated with liver metastasis of pancreatic cancer [105]. In another model of cholangiocarcinoma, circ-CCAC1 was demonstrated to dissolve the tight junction between ECs. Mechanistic investigations revealed that circ-CCAC1 interacted with EZH2 to suppress its nuclear translocation, thereby alleviating its inhibitory effect on SH3GL2, which downregulated ZO-1 and Occludin in ECs [106].

Several miRNAs, termed as angio-miRNAs, were shown to play a vital role in neovasculature formation and dissemination into the circulation. For instance, miR-205 enhanced the growth of blood vessels in ovarian cancer through the PTEN-AKT pathway [107]. A previous study reported that miR-135b induced angiogenesis in gastric cancer via downregulating the expression of FOXO1 [108]. Furthermore, miR-130a was revealed to function as a driver of angiogenesis in gastric cancer by suppressing the expression of c-MYB [109]. In contrast, miR-125b and miR-100 attenuated intravasation and metastasis in HCC. Encapsulated tumor cluster (VETC), a vascular pattern that facilitated the penetration of tumor cells into the circulation, was ablated by miR125b and miR-100. Mechanistic investigations revealed that these miRNAs directly inhibit the expression of Angiopoietin 2 (Angpt2), the key factor for VETC formation [110].

Notably, tumor cells usually suffer from hypoxia in the local microenvironment to give rise to new vessels, attributing to the induction of a vast array of lncRNAs. In non-small-cell lung cancer, lncRNA-p21 was significantly upregulated under hypoxic conditions, followed by a global induction of pro-angiogenic genes, including VEGFA, MMP2, PDGFB, and FGF2 [111]. In addition, lncRNA RAB11B-AS1 was induced by HIF2-mediated transcription in hypoxic breast cancer, thus enhancing angiogenesis and metastasis. Mechanistically, RAB11B-AS1 recruited RNA polymerase II to the promoter of VEGFA and ANGPTL4 genes, thereby upregulating the expression of these angiogenic factors [112]. Besides, lncRNAs can also negatively modulate intravasation. Adenomatous polyposis coli (APC) gene is a well-known tumor suppressor gene in colorectal cancer (CRC), and it has been reported recently that lncRNA-APC1 activated by APC exerted its anti-metastasis function by attenuating MAPK pathway-induced angiogenesis in endothelial cells in CRC [113].

In addition, a newly identified circRNA circPRRC2A was revealed as a player in metastatic renal cell carcinoma. CircPRRC2A sponged miR-514a-5p and miR-6776-5p (the negative regulators of TRPM3) to protect TRPM3 mRNA from degradation. Subsequently, the upregulated TRPM3 enhanced tumor angiogenesis [114]. In gastric cancer, a recent research showed that circSHKBP1 upregulated Vascular Endothelial Growth Factor A(VEGF) to confer enhanced tumor invasiveness. Regarding the molecular mechanism, circSHKBP1 acted as a sponge for miR-582-3p to attenuate the inhibitory effect on HUR, an RNA-binding protein that stabilized VEGF mRNA [115]. Indeed, a plethora of novel angio-circRNA in cancer metastasis has been identified recently, such as circRNA-MYLK [116], hsa-circ-001653 [117], circRNA-100338 [118], and so on.

Several in vivo studies have shown prominent roles that ncRNAs play in intravasation. One study using transgenic mice model for mammary tumor exhibited that fewer circulating tumor cells (CTCs) were collected from the peripheral blood of miR-10b-deficient mice compared with control mice, suggesting a positive impact of miR10b on intravasation [119]. In an orthotopic HCC model of mice, lncRNA-ATB knockdown led to an obvious reduction in the number of CTCs, indicating that lncRNA-ATB contributed to intravasation of HCC cells [120].

ncRNAs-mediated survival in the circulation

Once entering the bloodstream, CTCs are present as either single cells or emboli [52]. Emboli can be formed by tumor cells or by tumor cells with leukocytes, platelets, fibrins, etc. [52]. In fact, emboli can protect tumor cells from the threats of the innate immune system (mainly natural killer (NK) cells) and shear stress, two major challenges that CTCs have to overcome in the circulation [121]. For example, tumor cells interact with platelets through different cell surface receptors and ligands such as tissue factor and/or L- and P-selectins, promoting platelet coats formation as a physical shield against shear stress [122]. Other than that, platelets also favor immune evasion. Adhered platelets secrete TGF-β1 to counteract NK granule mobilization and interferon-γ secretion, and transfer platelet MHC-I to tumor cells to evade recognition, thus synergizing to confer resistance to NK-mediated cytolytic activity [123]. Furthermore, in the absence of adhesion to ECM that is necessary for cell survival, cancer cells are subject to anoikis, a class of apoptosis induced by a loss of integrin-dependent anchorage [124]. Anoikis induces cell death through the canonical intrinsic and extrinsic apoptotic pathways [125]. CTCs are enabled to anoikis resistance by induction of many mechanisms, including alterations in integrins expression, reactive oxygen species (ROS)-induced pro-survival signaling, constitutively activation of anti-apoptotic, pro-survival pathways such as PI3K/Akt and the Ras-Raf-Mek-Erk signaling, and so forth [98, 126] (Fig. 3).

In the circulation, platelet-derived microparticles (PMPs) are the main sources of miRNAs [127]. Although there is no direct evidence, previous research has demonstrated that miRNAs may serve as protective signals for CTCs against NK cells. A study reported that miR-296-3p enhanced the tolerance of prostate cancer cells to NK cells via directly repressing the expression of intercellular adhesion molecule 1 (ICAM-1). Subsequent in vivo experiment confirmed that miR-296-3p-ICAM-1 pathway had a positive role in guiding prostate cancer metastasis to the lung, suggesting that miR-296-3p likely increased survival of CTCs from the threats of NK cell [128]. Previous studies have illustrated the important role of breast cancer stem-like cell (BCSC) during metastasis, yet the underlying mechanisms are still being unraveled. miR20a was demonstrated to endow BCSC with resistance to NK cell cytotoxicity and facilitate its spreading to the lung. The processes were achieved via the inhibitory effect of miR20a on MICA and MICB, two stimulators for NK cell activation [129]. In addition, a previous study identified miRNA to be a downstream effector of TGF-β1-mediated immunosuppressive signaling against tumor-associated NK cells. TGF-β-induced miR-183 abrogated the effectiveness of NK cells by directly targeting DNAX activating protein 12 kDa (DAP12), which is a key transmembrane protein for cytolytic signal transduction in NK cells [130].

LncRNAs are also important innate immune system regulators. HOTAIR, one of the most well-studied pro-metastasis lncRNAs, was shown to inhibit NK cell activity via activating Wnt/β-catenin pathway [131]. While another novel lncRNA, lnc-CD56, upregulated CD56 in NK cells and contributed to their differentiation [132]. Aberrantly expressed circRNAs are also involved in the regulation of CTCs survival, though the direct evidence remains scarce. One study pointed to a critical role of circ_0000977 in immune evasion of pancreatic cancer cells from NK cells. Circ_0000977 acted as a decoy for miR-153, which directly targeted and repressed the expression of HIF1A, thus triggering attenuated cytotoxicity of NK cells and immune escape [133]. Conversely, CDR1as/ciRS-7 [80] and CircARSP91 [134] had been described as NK cell stimulators in multiple carcinoma types.

Aside from immune evasion, miRNAs are also involved in anoikis. Recently, miR-141 was identified to function as a crucial regulator for anoikis resistance and peritoneal metastasis in ovarian cancer. Notably, miR-141 directly targeted KLF12 mRNA and suppressed its expression, followed by Sp1-mediated survivin upregulation, which inhibited intrinsic apoptotic pathway [135]. This study revealed a novel pathway in anoikis resistance and implied a possible mechanism by which CTCs escape anoikis. Emerging evidence has suggested that various lnRNAs are also implicated in anoikis. In ovarian cancer cells, depletion of the key metastasis-associated lncRNA MALAT1 decreased metastasis and increased anoikis via affecting the alternative splicing program of pro-apoptotic gene KIF1B. Mechanistically, MALAT1 silencing suppressed the expression of splicing factor RNA binding protein fox-1 homolog 2 (RBFOX2), and led to preferential splicing of the pro-apoptotic full-length isoform of KIF1B [136]. Other metastasis-associated lncRNAs, such as HULC [137], HOTAIR [138], and LINC00958 [139], have been identified to drive anoikis resistance; however, the underlying molecular mechanisms warrant further investigation.

Notably, circRNA may also contribute to anoikis resistance. One study showed that lung cancer cells upregulated circUBAP2 to limit their susceptibility to anoikis, possibly owing to the elevated level of c-IAP1, Bcl-2, and surviving [140]. The role of piRNA in resisting attacks from the immune system in CTCs is still being unraveled, while they appear to be involved in regulating anoikis/ cell survival and apoptosis. For example, piR-004800 was reported to promote myeloma cell survival via the regulation of PI3K/Akt/mTOR pathway [141]. piR-8041, a tumor suppressor of glioblastoma, was also revealed to induce apoptosis probably via the inhibition of survival pathway MAPK/ERK [142].

Besides, the notion that lnRNAs are potential regulators in CTCs survival was further supported by profiling of CTCs gene expression in patients with metastatic castrate-resistance prostate cancer, which uncovered upregulated lncRNA SChLAP1 in CTCs, suggesting its pro-survival role for metastatic tumor cells in the bloodstream [143].

ncRNAs-mediated extravasation

In the context of extravasation, CTCs arrest in small capillaries of distant organs via physically trapping or specific adhesive interactions between CTCs and ECs, then they overcome the endothelial barrier to invade into the parenchyma [3]. Notably, a pattern, that organ tropism for different carcinoma types to form metastases, is ubiquitous. This metastatic preference can be explained by the anatomical layout of the vasculature, the organ-specific circulation pattern, and tumor cell–autonomous traits [51]. The structures of microvessels vary largely depending on organ types. For instance, the sinusoidal capillaries in the liver and bone are highly permeable and seem to be permissive to CTCs, whereas the continuous capillaries in brain and lung together with surrounding cells create an impermeable barrier known as the blood–brain barrier (BBB) and blood–gas barrier [144]. Accordingly, CTCs exploit further strategies to overcome these obstacles. First, adhesion molecules such as E-Selectin, N-cadherin, Integrins, CD44, and MUC1 are expressed to enhance the attachment of cancer cells to ECs [145]. Next, transendothelial migration of CTCs is further facilitated by EC junction opening through disruption of intercellular adhesions or induction of EC apoptosis [146]. Ultimately, CTCs squeeze between ECs, which requires dynamic changes in cell shape mediated by Rho/ROCK-regulated actin contractility [51] (Fig. 4).

Fig. 4.

Role of ncRNA in extravasation. ncRNAs are involved in the process of extravasation, as shown above, specifically including endothelium permeability, tumor cells adhesion to ECs, and tumor cells motility

Emerging evidence has suggested that ncRNAs are involved in the process of extravasation through modulation the adhesion of tumor cells to the ECs. As mentioned above, E-Selectin, an adhesion receptor that is expressed by vascular endothelial cells and mediates leukocytes and cancer cells rolling on the endothelium, has been widely investigated in cancer metastasis [145, 147]. In CRC, metastatic potential of cancer cells was alleviated by miR-146a and miR181b, which downregulated E-selectin in ECs by regulation of NF-κB pathway [148]. LncRNA HOTAIR has been reported to regulate CRC liver metastasis through an E-Selectin-associated mechanism. HOTAIR functioned as a decoy for miR-326 and positively regulated Fucosyltransferase 6 (FUT6) expression to promote fucosylation modification of CD44, which was required as an E-selectin ligand during CRC spreading [149]. Accordingly, it can be inferred that HOTAIR may facilitate metastasis by favoring the attachment of cancer cells to the ECs at the target organ.

miRNAs can also exert their effect on EC junction opening. A recent study revealed a mechanism regarding how CRC-secreted exosomal miR-25-3p disrupted the endothelium integrity at metastasis in the liver and lung in vivo. In the study, miR-25-3p directly targeted and inhibited the expression of Kruppel-Like Factor 4(KLF4), a transcription factor that maintains vascular integrity. Decreased KLF4 then reduced the expression of tight junction proteins ZO-1, occludin, and Caludin5 of lung and liver endothelial cells, thus favoring the extravasation of CTCs [150]. In addition, miR-181c was reported to be conducive to the breakdown of BBB. miR-181c was secreted by brain metastatic cancer cells encapsulated in extracellular vesicles and taken by brain endothelial cells where it directly targeted PDPK1, which resulted in abnormal actin dynamics and subsequent BBB disruption [151].

lncRNAs are also important determinants of vascular integrity. In lung metastasis from adenoid cystic carcinoma, lncRNA MRPL23-AS1 promoted metastasis via enhancing permeability of pulmonary endothelium. Mechanistically, exosomal MRPL23-AS1 derived from primary tumor induced transcriptional silencing of E-cadherin through binding of EZH2 to enhance the H3k27me3 of the E-cadherin promoter region. The novel RNA–protein complex augmented the expression of VEGF in lung ECs, resulting in the breaching of EC intercellular adhesions [152]. As another model of liver metastasis from gallbladder cancer, lncGALM released by cancer cells facilitated extravasation by promoting the leakiness of endothelial barrier. Regarding the molecular mechanism, lncGALM can directly bind to IL-1β mRNAs to increase their stability, thus favoring the apoptosis of liver sinusoidal endothelial cells [153]. Furthermore, a study revealed that the exosomes containing lncRNA GS1-600G8.5, which secreted by brain metastatic breast cancer cells, were internalized by brain microvascular endothelial cells, consequently leading to increased permeability of BBB and infiltration of cancer cells [154]. Several other lncRNAs have also emerged as key molecules to facilitate BBB disruption, such as LOC102640519 via the induction of LOC102640519/HOXC13/ZO-1 axis [155] and Snhg3 via the regulation of TWEAK/Fn14/STAT3 pathway [156].

The exact roles of circRNAs and piRNAs in extravasation have yet to be observed. However, based on current evidence, their involvement in this process can be proposed. Recent studies provided some evidence that circRNAs contribute to vascular permeability regulation in different disease models. For example, circ-USP1 could regulate vascular integrity in glioma endothelial cells. Mechanistically, circ-USP1 acted as a decoy for miR-194-5p and suppressed its activity to alleviate its inhibitory effect on transcription factor FLI1. As a result, FLI1 promoted the expression of tight junction proteins including claudin-5, occludin, and ZO-1 to enhance endothelium function [157]. Furthermore, circHIPK3 was reported to regulate retinal vascular leakage through sponging miR-30a-3p, thereby elevating the expression of vascular endothelial growth factor-C(VEGFC), frizzled class receptor 4(FZD4), and Wnt family member 2 (WNT2) [158]. In addition, circDLGAP4 [159] and circHECW2 [160] were reported to affect BBB integrity by regulating tight junction protein and mesenchymal marker expression, though their exact roles in extravasation to form brain metastasis have not been explored yet. Notably, no piRNA has been reported to influence extravasation directly; however, it can regulate the maintenance of blood–retinal barrier (BRB) in human retinal pigment epithelium through reorganization of tight junctions [161], indicating its potential role on the modulation of endothelium integrity and extravasation.

Intriguingly, miRNAs are also likely to modulate extravasation by dynamically altering cytoskeleton-mediated cellular motility. In melanoma, miR-146a augmented primary tumor and lung metastasis growth, whereas impaired CTCs extravasation owing to miR-146a-dependent inhibition of ITGAV and ROCK1. Consistently, miR-146a was downregulated in CTCs compared with primary tumor and distant lung colonies. Hence, it was appealing to speculate that miR-146a was implicated in the dynamic change of cell states during metastasis, where CTCs lowered the level of miR-146a to favor cell extravasation through ROCK1 activation and subsequently restored miR-146a expression to enhance subsequent colonization [162]. Furthermore, miRNAs have been investigated for their involvement in extravasation in vivo. For example, several models using cell tracking technique observed different miRNAs including miR-214, miR-143, and miR-145 were involved in the extravasation of CTCs [163–165].

ncRNAs-mediated colonization

Colonization, the process of extravasated carcinoma cells surviving and proliferating at foreign sites, represents one of the most inefficient steps in the metastatic cascade [3]. First, in order to survive at distant microenvironments, primary tumors release systemic signals to convert incompatible sites into more hospitable ones, known as the establishment of a premetastatic niche [166]. Next, upon reaching the distant sites, the majority of carcinoma cells suffer high rates of attrition (exceeds 99%), while a small minority persist as microcolonies for a long period of state called dormancy [97, 167]. Subsequently, to develop into macroscopic metastases, the founding cells depend on two attributes: adaptability to specific foreign microenvironments and capacity for self-renewal [1]. The first one can be explained by the “seed and soil” hypothesis proposed by Stephen Paget more than 100 years ago, stating that metastases develop only when “seeds (cancer cells) fall on congenial soil (organ microenvironment)” [168]. This dictates complex mechanisms to modify ectopic microenvironments, such as secretory signals or mobilization and recruitment of bone marrow-derived cells [123, 169]. The second one can be conferred by a program called mesenchymal–epithelial transition (MET) [170]. As discussed above, during the early course of the metastatic cascade, EMT is activated to enhance local invasion. It is now recognized that primary cancer cells that have undergone EMT are in a state of low proliferation, therefore, at distant sites disseminated cancer cells are proposed to go through a phenotypic reversion, that is, MET to reinitiate outgrowth [171] (Fig. 5).

Fig. 5.

Role of ncRNA in colonization. Dysregulated ncRNAs promote colonization in different organs through distinct mechanisms. The different stages of colonization include a establishment of a premetastatic niche, b dormancy, and c metastasis outgrowth

In the phase of premetastatic, miRNAs remodel the foreign microenvironment to strengthen the settlement of disseminated cancer cells in several manners. For instance, miR-122 reprogrammed the metabolism of premetastatic niche cells to achieve this end. More specifically, miR-122 derived from breast cancer cells directly targeted mRNAs of pyruvate kinase (PKM), citrate synthase (CS), and glucose transporter 1(GLUT1), thus interfering with glucose metabolism and result in glucose reallocation in niche cells in brain and lung [172]. In addition, miR-21 established a proinflammatory premetastatic niche to facilitate CRC cells seeding to the liver via binding to toll-like receptor 7(TLR7) and enhance IL-6 production [173]. Moreover, a recent study uncovered that increased vascular permeability in the lung premetastatic niche required the downregulation of miR-30 family [174].

So far there is no solid evidence regarding the exact mechanism of how lncRNAs function in pre-metastatic niche establishment. However, a number of dysregulated lncRNAs have been identified to be involved in pulmonary pre-metastatic niche formation of breast cancer by high-throughput sequencing. Functional enrichment analysis revealed that these lncRNAs were correlated with the TGF-β signaling pathway, pentose phosphate pathway, Hedgehog signaling pathway, and so on, while the molecular mechanisms underlying this process awaited further study [175]. Moreover, a recent study uncovered that ovarian cancer-derived lncRNA SPOCD1-AS induced the establishment of peritoneal pre-metastatic niche. The proposed mechanisms of this included mesothelial-to-mesenchymal transition of mesothelial cells and enhanced adhesion of cancer cells to mesothelial cells, and the above processes were achieved through interaction between lncRNA SPOCD1-AS and G3BP stress granule assembly factor 1(G3BP1) [176].

The contribution of circRNAs and piRNAs to colonization has been largely elusive. Using paired-end RNA sequencing, it was revealed that the primary and metastatic sites of ovarian cancer exhibited different circRNAs expression pattern [177], indicating possible roles of circRNAs on adaptability to the foreign microenvironment. Furthermore, in the tail vein injection mouse models of metastatic CRC, circPTK2 was shown to promote the formation of lung metastasis, suggesting its potential in colonization. The underlying molecular mechanisms involve the interaction of circPTK2 and vimentin [178].

For the subsequent dormant state, the roles of ncRNAs have been widely investigated. During the formation of bone marrow metastasis of breast cancer, a number of miRNAs, including miR-127, -197, -222, and -223 in stromal cells were transferred to breast cancer cells via gap junctions to inhibit the expression of C–X–C motif chemokine ligand 12(CXCL12), therefore, inducing cell cycle arrest in breast cancer cells [179]. Recently, a novel regulatory mechanism of dormancy in pulmonary metastasis of breast cancer was posited. It is widely recognized that breast cancer cells could spread from the primary site at very early stages of tumorigenesis and remain in quiescence for long periods [180]. lncRNA BORG was identified as a metastasis-promoting gene whose upregulation can reinitiate proliferation in dormant disseminated breast cancer cells. It was shown that in dormant metastatic tissues, the level of BORG expression was low; however, when faced with environmental stresses and signals (e.g., hypoxia, nutrient deprivation, TGF-β, BMP2, or BMP7), BORG was induced to reactivate these quiescent cells into proliferative states presumably through effectors including P21, P38/MAPK, ERK1/2, and others. Thus, this state plasticity of dormant disseminated tumor cells highlighted that targeting BORG would be a potential therapy for breast cancer lung metastasis [181]. A large number of cirRNAs and piRNAs had been reported to mediate cell cycle arrest [142, 182–185], hence they could be involved in the mechanisms of microcolonies and the state of long-term dormancy.

Ultimately, to foster macrometastatic growth, miRNAs orchestrate distinct adaptive programs, including bidirectional intercellular crosstalk between cancer cells and microenvironment. On the one hand, normal fibroblasts of distant organ sites are converted into cancer-associated fibroblasts (CAFs) upon the infiltration of metastatic cells. This process was demonstrated to be regulated by miR-1247-3p when HCC colonized the lung. Exosomal miR-1247-3p shed by HCC cells directly targeted and inhibited the expression of B4GALT3 in fibroblasts, resulting in fibroblasts activation via B4GALT3-β1-integrin-NF-κB axis. In turn, transformed CAFs further favored metastasis outgrowth by increased secretion of IL-6 and IL-8 [186]. On the other hand, miR-19a derived from astrocytes could reshape the metastatic cancer cells, leading to adaptive PTEN loss and resultant C–C motif chemokine ligand 2(CCL2)-dependent myeloid cell recruitment that promoted brain metastasis outgrowth [187]. Recently, in a model of mutant p53 breast cancer, miR-30d was also demonstrated to mediate CAFs recruitment in the lung metastasis [188].

Likewise, lncRNAs are implicated in the crosstalk between disseminated tumor cells and secondary microenvironment to establish overt metastasis. In breast cancer brain metastasis, lnc-BM augmented colonization by forming a positive-feedback loop between breast cancer cells and brain microenvironment. Mechanistic investigation revealed that increased lnc-BM associated with JAK2 and induced its kinase activity, thus leading to STAT3 phosphorylation. Active STAT3 promoted macrophages recruitment via upregulation of CCL2. In turn, attracted macrophages secreted oncostatin M (OSM) and IL-6 to further reinforce the JAK2/STAT3 pathway in metastatic breast cancer cells [189]. Conversely, in another model of breast cancer brain metastasis, lnc-XIST exhibited a negative role in metastasis formation by inhibition of miRNA-503-mediated M1–M2 polarization of microglia. This reprograming of microglia increased the production of immune-suppressive cytokines and inactivated T-cells, thereby facilitating the development of brain metastasis [190].

Besides, ncRNAs also serve as signals to initiate the MET program. A recent study indicated that the aforementioned negative feedback loops formed by pairs of miRNAs and EMT-TFs during EMT are also implicated in MET [191].In addition, several lncRNAs were reported to modulate MET program, such as CILA1 via modulation of the Wnt/β-catenin signaling pathway [192] and H19 via regulation of Twist expression [193]. Besides, circ-0005585 was suggested to promote MET program in ovarian cancer cells in a recent study [194]. Currently, the roles of piRNAs on MET remain largely unexplored, although as discussed above, they have been implicated in the EMT program in the context of local invasion.

Clinical applications of ncRNAs

ncRNAs as biomarkers for cancer metastasis

Given that the expression patterns of ncRNAs differ substantially across diverse cancers, and differential expression signatures are linked to distinct cancer states, it raises the possibility that ncRNAs can be appealing diagnostic and prognostic biomarkers [195]. Indeed, a plethora of published studies have provided evidence that ncRNAs can serve as novel tools for predicting cancer metastasis. In breast cancer, the primary lesion with low levels of miR-335 and miR-126 had a higher tendency to form metastasis [196], while patients with prostate cancer were more prone to develop metastasis when the expression of lncRNA PCAT-14 was low in the primary tumor [197]. In laryngeal cancer, higher hsa_circRNA_100855 level in the primary tumor was correlated with metastasis or advanced clinical stage [198]. Similarly, piR-1245, a novel oncogenic mediator, was significantly increased in patients with advanced and metastatic CRC [48].

Notably, ncRNAs including miRNAs, lncRNAs, circRNAs piwiRNAs [195] can all be detected in body fluids, including blood [199], urine [200], saliva [201] and sputum [202] with high abundance and stability. In general, in the extracellular space, ncRNAs are released from cells and encapsulated inside extracellular vesicles (mainly exosomes) to avoid digestion by the nuclease [195]. Alternatively, ncRNAs interact and form complexes with protein Argonaute 2 [203] or high-density lipoproteins [204]. Comprehensive analyses of exosomal ncRNAs have confirmed their abundancy in exosomes and uncovered distinct expression profiles of miRNAs, piwiRNAs and circRNAs between healthy individuals and cancer patients, making them potential tools to discriminate between healthy people and patients, and even define cancer states [205, 206]. Moreover, the new diagnostic tools are much less invasive than conventional biopsies. A growing body of evidence has demonstrated the utility of circulating ncRNAs to predict metastasis. CRC patients with high levels of serum miR-203 had a greater probability of liver or systemic metastasis [207]. The well-characterized pro-metastasis lncRNAs such as H19 [208, 209], MALAT1 [210, 211] and HOTAIR [33, 212] were present at high levels in the circulation across a variety of metastasis cancers. Plasmas level of circ-KIAA1244 was found to be downregulated in gastric cancer patients with lymphatic metastasis [213].

The piles of evidence have proved ncRNAs as potent biomarkers in preclinical studies and some ncRNAs indeed have entered into clinical trials. PCA3, a prostate-specific marker which can be detected in urine non-invasively, was the first ncRNA to be approved by the FDA for cancer diagnosis [214]. At present, there are hundreds of ongoing clinical trials evaluating ncRNA-associated biomarkers in multiple metastatic carcinoma types, including detection for dysregulated miRNAs as biomarkers for bone metastases (NCT03895216), analysis of circulating exosomal RNAs in the lung metastases of osteosarcoma (NCT03108677), and assessment of miRNAs of metastatic breast cancer patients (NCT02338167), and so forth.

ncRNAs as therapeutic targets for cancer metastasis

ncRNAs play crucial roles in cancer progression as either oncogenes or tumor suppressors; therefore, the development of therapeutic interventions by targeting ncRNAs has represented a burgeoning field of cancer research. However, approaches to modulate ncRNAs expression are far different from traditional agents. The development of RNA-targeting techniques has opened up new strategies to manipulate ncRNAs in vivo. For miRNA, antisense RNAs (antimiRs) are used to exert inhibitory functions, while double-stranded miRNA mimics are employed to replenish their expressions [215, 216]. AntimiRs are single-stranded RNAs that are complementary to the target miRNA. Mechanistically, antimiRs bind to the target miRNA strand and inactivate its function. miRNA mimics are a synthetic duplex form of oligoribonucleotides that aim to re-introduce the functions of the lost tumor suppressor miRNA counterparts [215]. For lncRNA, antisense oligonucleotides (ASOs) and small interfering RNAs are utilized to modulate its expression. In brief, small interfering RNAs are synthetic double-stranded RNAs that are incorporated into RISC and interact with target lncRNA to direct its cleavage. ASOs, conversely, are synthetic short DNAs that hybridize to the target lncRNA and create a DNA-RNA duplex to be digested by RNase H [216]. Currently, there are two major issues in the field of ncRNA-based therapeutics: the stability and affinity of drug molecules and the establishment of delivery vehicles to the target sites [215]. For the former, to optimize the pharmacokinetic properties of therapeutic RNAs, chemical modifications are required before their delivery, such as 2’ -O-methyl (2’-O-Me) and locked nucleic acid (LNA) modifications [217–219]. For example, in diffuse large B-cell lymphoma (DLBCL), cobomarsen, an LNA-based anti-miR-155 compound, has shown potential therapeutic efficacy in vitro and in vivo, as well as in a patient [220]. For the latter, oligonucleotides need delivery systems to improve cellular uptake and reduce systemic toxicity, among which lipid-based nanocarriers are widely applied [219]. Meanwhile, novel delivery approaches are being developed. For instance, in non-small cell lung cancer (NSCLC), hydrophobically modified let-7b miRNAs (hmiRNAs) which could enable efficient delivery without the need for lipid formulation has been demonstrated as an effective strategy in pre-clinical mouse models [221].

As most cases of metastasis cancer are unresectable at the time of diagnosis, the development of drugs that target ncRNAs provides promising therapeutic strategies. Pre-clinical studies manipulating a number of ncRNAs dysregulated during metastasis have been shown to exert prominent therapeutic effects (Fig. 6). Take miR-10b, the first recognized metastasis-associated miRNA for example. The addition of the nanodrug that contained miR-10b inhibitors significantly lowered the dose of doxorubicin, eliciting complete and persistent regression of metastatic lesions in mouse models of breast cancer [222]. Another well-studied pro-metastasis lncRNA MALAT1 has also been employed therapeutically. ASO-mediated silencing of lncRNA MALAT1 could greatly slow down tumor growth and impair metastasis formation in mouse models of lung and breast cancer [210, 223]. Of note, various miRNA-based therapeutics have entered clinical trials, with some of them have been under clinical development [224]. For instance, MRX34, a liposomal miR-34a mimic, had shown potent anticancer effectiveness in several preclinical studies, including mouse models of lung [225], liver [226], pancreatic [227], and prostate cancer [24]. Subsequently, MRX34 was approved for human trials in patients with advanced solid tumors, which was the first clinical trial for miRNA drug in history. However, whereas 4% overall response rate was achieved, recently the trial was terminated due to severe immune-related adverse events that led to four deaths. Although halted early, this study bought up the concern about toxic effects that come with miRNA drugs, which should catch the attention of drug development in the future [228].

Fig. 6.

ncRNAs as therapeutic targets for cancer metastasis. In mouse models of breast cancer, MALAT1 antisense oligonucleotides hybridize to MALAT1 and create a DNA–RNA duplex to be digested by RNase H, leading to impaired metastasis formation. Meanwhile, the addition of the nanodrug that contained single-stranded miR-10b antisense RNAs that are complementary to miR-10b counteracts the metastasis-promoting effect of miR-10b. Mechanistically, miR-10b antisense RNA directly binds to miR-10b and inactivate its function

Context-dependent roles of ncRNAs in metastasis

Various ncRNAs have been found to exhibit both tumor suppressive and oncogenic functions. A notable example is Malat1, which was originally identified as a predictor for metastasis in early stage non-small cell lung cancer [14]. Consistently, subsequent studies have reported the upregulation of Malat1 and its association with poor outcomes in a large number of metastatic cancer types [229]. Experimental studies also demonstrated metastasis-suppressive effect of Malat1 inactivation both in vitro and in vivo and the underlying mechanisms were proposed as the interaction of Malat1 with EZH2, SUZ12, and a number of miRNAs [230]. However, several recent studies have revealed a metastasis-suppressive role for Malat1 in CRC and breast cancer [76, 231, 232]. In contrast to previous studies, analysis of several large-scale data sets showed that Malat1 expression was significantly lower in human CRC and breast cancer, and particularly a reduction in metastatic than primary breast cancer tissues was observed [76, 231]. A positive correlation between MALAT1 expression and patient survival was also reported [76, 231]. Of note, using both transgenic and xenograft mouse models, knockout of Malat1 promoted lung metastasis of breast cancer and this phenotype could be reversed by re-expression of Malat1 [76]. Nevertheless, there were numerous methodological differences between these studies that may account for the divergent results [233]. The major difference appears to arise from the different strategies used for Malat1 gene deletion [76]. Indeed, manipulation of certain regions of Malat1-encoding gene could lead to significant upregulation of neighboring genes involving Neat1, Frmd8, Tigd3, and so on [234]. Therefore, it is unknown whether the phenotypes were specifically due to the loss of Malat1 lncRNA per se or the confounding effect of other gene expressions. In addition, it is also essential to validate the specific biological function of Malat1 through the performance of rescue experiments [76]. Furthermore, when it comes to clinical data, stratification of patient subgroups (e.g., TNM staging or molecular subtypes) and the source of the samples being analyzed (e.g., bulk sequencing that is affected by tumor purity and intratumor heterogeneity or single-cell sequencing) have emerged as potentially important considerations [233].

Similarly, multifaceted roles of circRNAs in distinct phases of cancer metastasis have been reported. A prime example of this is circHIPK3. While in CRC and cervical cancer, circHIPK3 displayed a pro-metastatic phenotype presumably through enhancing EMT [235], altering cell–ECM interactions [236], and generation of protruding actin structures [237] of cancer cells in the stage of local invasion, it suppressed bladder cancer metastasis via inhibition of angiogenesis [238]. Taken together, the context-depend effect of circHIPK3 may be largely determined by tumor types and/or downstream effector molecules. As ncRNAs are emerging as key therapeutic targets for many cancer types, it is of vital importance to consolidate the contradicting findings and comprehensively understand the underlying causes of the discrepancy in the role of ncRNAs in cancer metastasis.

Perspectives and conclusions

In the past decade, ncRNAs have been under vivid investigations in diverse physiological and pathological processes. Extensive studies support a prominent role of ncRNAs in orchestrating every step of the metastatic cascade (Table 1). Moreover, the expression profiling and identification of candidate ncRNAs in cancer metastasis have been dramatically promoted owing to the development of high-throughput sequencing technology. Although mounting evidence has revealed that ncRNAs are involved in the regulation of metastasis, their roles underlying specific steps of metastasis cascade warrant further investigation. Of note, it is important to realize that ncRNAs function pleiotropically in distinct steps. This notion is supported by many cases. miR-31 has been reported to exert its impact simultaneously on local invasion, intravasation, and colonization, and what’s more, even served as both positive and negative roles [74, 239]. Another prime example is the double negative feedback loops formed by pairs of miRNAs and EMT-TFs, where miRNAs hamper the EMT program at the early stage of metastasis cascade to inhibit local invasion, conversely favor the MET program at the late stage to promote colonization. This multifaceted nature of ncRNAs comes with many caveats. For example, this highlights the difficulty to target ncRNAs therapeutically because of the contradictory roles that ncRNAs play across distinct steps in metastatic cascade, as abrogate one process might lead to activation of another. Therefore, future work will be required to address the questions of the mechanisms by which ncRNAs regulate metastasis under more specific circumstances [240].

Table 1.

Some dysregulated ncRNAs and their putative roles in the metastasis cascade

ncRNA	Targets	Cancer type	Regulation	Ref
Local invasion
EMT
miR-200	ZEB	Multiple cancer types	Down	[55]
miR-34a	Snail	Multiple cancer types	Down	[56]
miR‑1	SNAIL2	Multiple cancer types	Down	[57]
miR‑203	SNAIL1	Multiple cancer types	Down	[58]
miR-4521	AKT/GSK3β/Snai1 pathway	Gastric cancer	Down	[62]
ZEB1 AS1	ZEB1	Hepatocellular carcinoma	Up	[63]
MEG3	E-cadherin and microRNA-200 family	Lung cancer	Up	[64]
JPX	miR-33a-5p /Twist1 axis	Lung cancer	Up	[65]
circANKS1B	TGF-β1/Smad signaling	Breast cancer	Up	[67]
circNEIL3	miR-432-5p	Pancreatic cancer	Up	[68]
circCORO1C	let-7c-5p/PBX3 axis	Laryngeal squamous cell carcinoma	Up	[69]
circNR3C2	miR-513a-3p	Breast cancer	Down	[70]
piR-1037	–	Oral squamous cell carcinoma	Up	[71]
piR-932	–	Breast cancer	Up	[72]
Cell adhesion
miR-31	Wnt/β-catenin signaling(integrins)	Breast cancer	Up	[74]
miR-221/222	β4 integrin	Breast cancer	Down	[75]
MALAT1	TEAD	Breast cancer	Down	[76]
H19	β3 and β4 integrins	Prostate cancer	Up	[77]
CASC9	LAMC2	Esophageal squamous cell carcinoma	Up	[78]
HOXD-AS1	HOXD3	Colorectal cancer	Down	[79]
circSKA3	β1 integrin	Breast cancer	Up	[96]
ECM degradation
miR-181b	TIMP3	Colorectal cancer	Up	[81]
ODRUL	miR-3182/MMP2 Axis	Osteosarcoma	Up	[82]
MATN1-AS1	RELA and MAPK signaling	Glioblastoma	Down	[83]
PRECSIT	STAT3 Signaling	Cutaneous Squamous Cell Carcinoma	Up	[84]
LSAMP-AS1	miR-183-5p and decorin	Prostate cancer	Down	[85]
CASC2	MMP-2	Gastric cancer	Down	[86]
Circ 0,000,096	MMP-2 and MMP-9	Gastric cancer	Up	[87]
Circ0001361	miR-491-5p/MMP-9 Axis	Bladder cancer	Up	[88]
CircCDR1as	Associated with matrix organization, integrin binding, and collagen binding	Pan-cancer	Up	[80]
piR-39980	MMP-2	Osteosarcoma	Up	[89]
Cell motility
miR-124	ROCK2	Hepatocellular carcinoma	Down	[90]
miR-138	RhoC	Cholangiocarcinoma	Down	[91]
miR- BART2-5p	RND3	Nasopharyngeal carcinoma	Up	[92]
miR-200b	ARHGAP18	Breast cancer	Down	[93]
Dreh	Vimentin	Hepatocellular carcinoma	Down	[94]
MER52A	p120-ctn/Rac1/Cdc42 axis	Hepatocellular carcinoma	Up	[95]
Intravasation
miR10b	–	Breast cancer	Down	[119]
lncRNA-ATB	–	Hepatocellular carcinoma	Up	[120]
Vascular permeability
miR-23a	ZO-1	Lung cancer	Up	[101]
miR-103	VE-Cadherin, p120-catenin and ZO-1	Hepatocellular carcinoma	Up	[102]
miR-629-5p	CELSR1	Lung cancer	Up	[103]
circ-IARS	miR-122/RhoA/ZO-1 axis	Pancreatic cancer	Up	[105]
circ-CCAC1	EZH2/ SH3GL2/ ZO-1 Occludin axis	Cholangiocarcinoma	Up	[106]
Angiogenesis
miR-205	PTEN-AKT pathway	Ovarian cancer	Up	[107]
miR-125b	angiopoietin 2	Hepatocellular carcinoma	Down	[110]
miR-100	MTOR-p70S6K signalling pathway	Hepatocellular carcinoma	Down	[110]
lncRNA-p21	VEGFA	Lung cancer	Up	[111]
RAB11B-AS1	VEGFA and ANGPTL4	Breast cancer	Up	[112]
lncRNA-APC1	MAPK pathway	Colorectal cancer	Down	[113]
CircPRRC2A	miR-514a-5p, miR-6776-5p and TRPM3	Renal cell carcinoma	Up	[114]
CircMYLK	miR-29a and VEGFA/VEGFR2	Bladder cancer	Up	[116]
circRNA-100338	VE-cadherin and ZO-1	Hepatocellular carcinoma	Up	[118]
circSHKBP1	miR-582-3p/HUR/VEGF axis	Gastric cancer	Up	[115]
Survival in the circulation
Immune evasion
miR-296-3p	ICAM-1	Prostate cancer	Up	[128]
miR-20a	MICA and MICB	Breast cancer	Up	[129]
Anoikis resista-nce
miR-141	KLF12/Sp1/survivin axis	Ovarian cancer	Up	[135]
MALAT1	RBFOX2/KIF1B axis	Ovarian cancer	Up	[136]
SChLAP1	–	Prostate cancer	Up	[143]
Extravasation
miR-214	–	Breast cancer, Pancreatic cancer, Melanoma	Up	[163, 164]
Endothelium permeability
miR-143	CREB1 and MEKK2	Breast cancer	Up	[165]
miR-25-3p	KLF4/ZO-1, occludin, and Caludin5 axis	Colorectal cancer	Up	[150]
miR-181c	PDPK1	Breast cancer	Up	[151]
MRPL23-AS1	E-cadherin VEGF	Adenoid cystic carcinoma	Up	[152]
GALM	IL-1β	Gallbladder cancer	Up	[153]
GS1-600G8.5	Associated with BBB permeability	Breast cancer	Up	[154]
Tumor cells adhesion to ECs
miR-146a and miR181b	NF-κB/E-selectin axis	Colorectal cancer	Down	[148]
Tumor cells motility
miR-146a	ITGAV and ROCK1	Melanoma	Down	[162]
Colonization
Establishment of a premetasta-tic niche
miR-122	glucose metabolism enzymes	Breast cancer	Up	[172]
miR-21	TLR7-IL-6 axis	Colorectal cancer	Up	[173]
miR-30	Skp2	Melanoma	Down	[174]
TUC339, UCA1, etc.	Associated with pre-metastatic niche formation	Breast cancer	Up or Down	[175]
SPOCD1-AS	G3BP1	Ovarian cancer	Up	[176]
Dormancy
miR-127, -197, -222, and -223	CXCL12	Breast cancer	Up	[179]
BORG	P21, P38/MAPK, ERK1/2, etc.	Breast cancer	Up	[181]
Metastasis outgrowth
miR-1247-3p	B4GALT3-β1-integrin-NF-κB axis	Hepatocellular carcinoma	Up	[186]
miR-19a	PTEN	Breast cancer	Up	[187]
miR-30d	Golgi apparatus	Breast cancer	Up	[188]
lnc-BM	JAK2/STAT3 pathway	Breast cancer	Up	[189]
lnc-XIST	miRNA-503	Breast cancer	Down	[190]
circPTK2	vimentin	Colorectal cancer	Up	[178]

The diagnostic and therapeutic applications of ncRNAs are evolving at a rapid pace, and it is expected that those ncRNAs-based biomarkers and drugs will be put into clinical use, with FDA-approved lncRNA PCA3 being a good precedent. One of the major issues for ncRNA biomarkers when they are translated from bench into clinical rests on the low specificity among different types of patients. To address this challenge, combining various candidate ncRNA biomarkers as an integrated tool has been shown to enhance the diagnostic value. In the future, prospective trials with larger patient cohorts are needed to confirm the utility of ncRNA biomarkers in terms of specificity and sensitivity [195]. Concerning ncRNA-based therapy, in addition to those noted above, novel targeting approaches to manipulate ncRNAs expression in cancer cells show potential for improved specificity and reduced off-target effects, such as splice-switching oligonucleotides, steric blocking oligonucleotides, and genome editing tool CRISPR/Cas9 [216]. At the same time, chemical modifications to maintain the in vivo stability and bioactivity should be optimized, cancer-cell-specific delivery systems are required to be developed, to further improve the safety and effectiveness of ncRNA-based therapy strategy.

Abbreviations

ncRNAs

Noncoding RNAs

EMT

Epithelial–mesenchymal transition

ECM

Extracellular matrix

miRNAs

MicroRNAs

piRNAs

PIWI-Interacting RNAs

lncRNAs

Long noncoding RNAs

MALAT1

Metastasis-associated lung adenocarcinoma transcript 1

RISC

RNA-induced silencing complex

3ʹUTR

3ʹ Untranslated region

Gas5

Growth Arrest Specific 5

HOTAIR

HOX Transcript Antisense RNA

PRC2

Polycomb repressive complex 2

circRNA

Circular RNA

snoRNA

Small nucleolar RNA

tsRNAs

TRNA-derived small RNAs

MMPs

Matrix metalloproteinases

EMT-TFs

EMT-related transcription factors

ZEB1-AS1

ZEB1 antisense 1

MEG3

Maternally Expressed 3

HCC

Hepatocellular carcinoma

JPX

JPX Transcript, XIST Activator

TEAD

TEA Domain Transcription Factor

CASC9

Cancer Susceptibility 9

MATN1-AS1

MATN1 Antisense RNA 1

PRECSIT

P53 Regulated Carcinoma-Associated Stat3 Activating Long Intergenic Non-Protein Coding Transcript

LSAMP-AS1

LSAMP Antisense RNA 1

CASC2

Cancer Susceptibility 2

SERPINB1

Serpin Family B Member 1

ECs

Endothelial cells

VEGFA

Vascular Endothelial Growth Factor A

VEGF

Vascular Endothelial Growth Factor A

ANGPTL4

Angiopoietin-Like 4

PDGFB

Platelet-Derived Growth Factor Subunit B

FGF2

Fibroblast Growth Factor 2

ZO-1

Zonula occludens protein-1

CELSR1

Cadherin EGF LAG seven-pass G-type receptor 1

HUVECs

Human umbilical vein endothelial cells

FOXO1

Forkhead Box O1

VETC

Encapsulated tumor cluster

Angpt2

Angiopoietin 2

RAB11B-AS1

RAB11B Antisense RNA 1

APC

Adenomatous polyposis coli

CRC

Colorectal cancer

TRPM3

Transient Receptor Potential Cation Channel Subfamily M Member 3

CTCs

Circulating tumor cells

ATB

lncRNA activated by TGF-β

NK cells

Natural killer cells

ROS

Reactive oxygen species

PMPs

Platelet-derived microparticles

ICAM-1

Intercellular adhesion molecule 1

BCSC

Breast cancer stem-like cell

MICA

MHC Class I Polypeptide-Related Sequence A

MICB

MHC Class I Polypeptide-Related Sequence B

DAP12

DNAX activating protein 12 kDa

KLF12

Kruppel-Like Factor 12

KIF1B

Kinesin Family Member 1B

RBFOX2

RNA binding protein fox-1 homolog 2

SChLAP1

SWI/SNF Complex Antagonist Associated With Prostate Cancer 1

BBB

Blood–brain barrier

FUT6

Fucosyltransferase 6

KLF4

Kruppel-Like Factor 4

MRPL23-AS1

MRPL23 Antisense RNA

EZH2

Enhancer of Zeste 2 Polycomb Repressive Complex 2 Subunit

VEGFC

Vascular endothelial growth factor-C

FZD4

Frizzled class receptor 4

WNT2

Wnt family member 2

BRB

Blood–retinal barrier

MET

Mesenchymal–epithelial transition

PKM

Pyruvate kinase

CS

Citrate synthase

GLUT1

Glucose transporter 1

TLR7

Toll-like receptor 7

G3BP1

G3BP Stress Granule Assembly Factor 1

CXCL12

C–X–C Motif Chemokine Ligand 12

CAFs

Cancer-associated fibroblasts

CCL2

C–C Motif Chemokine Ligand 2

OSM

Oncostatin M

XIST

X Inactive Specific Transcript

PCAT-14

Prostate Cancer-Associated Transcript 14

antimiRs

Antisense RNAs

ASOs

Antisense oligonucleotides

2’-O-Me

2’ -O-methyl

LNA

Locked nucleic acid

PCA3

Prostate Cancer Associated 3

Authors’ contributions

ZGZ and XWW: designed the article, QLL and ZZ: performed the literature search and drafted the work, ZGZ and XWW: revised the work.

Funding

This work was supported by National Natural Science Foundation of China (81821002 and 82003113) and 1.3.5 project for disciplines of excellence, West China Hospital, Sichuan University (2016105 and ZYGD20006).

Availability of data and materials

Not applicable.

Code availability

Not applicable.

Declarations

Conflict of interests

The authors declare that they have no competing interests.

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.