Long noncoding RNAs (lncRNAs) were previously defined as RNA molecules longer than 200 nucleotides that are not translated into proteins [6
]. Recently, Spizzo et al.
amended this definition linking to the biological functions and described that lncRNAs are a class of RNA molecules that do not fit into any known class of small and structural RNAs, and possess regulatory roles in their primary or spliced form [109
]. The GENCODE Consortium is a part of the ENCODE (ENCyclopedia Of DNA Elements) project and aims to identify all gene features in the human genome. The GENCODE 7 release in 2012 indicates that the human genome contains at least 9,640 long noncoding RNA loci that can potentially encode 15,512 transcripts [110
]. LncRNA production is regulated by mechanisms similar to these of protein-coding genes, such as histone modifications and RNA splicing, and their expression shows tissue-specific patterns [111
]. Most lncRNAs are localized in nucleus and associated with chromatin, and some of them are preferentially processed into small RNAs [111
]. Banfai et al.
demonstrated that lncRNAs are rarely translated in two tested human cell lines, suggesting that ribosomes can differentiate the coding and noncoding transcripts for translation [112
Recent studies continue elucidating novel and unpredicted biological activities of the lncRNAs, which were previously ascribed as protein functions. Based on the regulatory mechanisms, the lncRNAs can be divided into four categories, including the lncRNAs that (1) regulate gene expression, (2) act as miRNA decoys to free target mRNAs, (3) regulate mRNA translation, and (4) regulate protein activities.
3.1. LncRNAs Regulating Gene Expression
The X-inactive-specific transcript (Xist) was one of the first lncRNAs discovered in mammals [113
]. Xist is encoded by the inactive X chromosome (Xi) and the genomic locus can transcribe a 17- to 20-kb RNA. This lncRNA binds the Xi in cis
to induce chromosome X silencing by recruiting the polycomb repressive complex 2 (PRC2) that induces histone H3-K27 methylation, a hallmark of gene inactivation [114
]. In this process, the transcription repressor Yin Yang 1 (YY1) confers allele-specific binding of Xist to the Xi with the involvement of two other noncoding RNAs, Jpx [116
] and Ftx [117
]. There is another lncRNA, Tsix, that is transcribed at the same locus of Xist but in the reverse direction and thus antisense to Xist [118
]. Tsix regulates imprinted and random X inactivation in development [119
]. An additional Xist-related RNA transcript is Xite that also plays a role in the X chromosome inactivation [120
The association of epigenetic silencing complexes with Xist to induce transcriptional silencing has been extended to several recently characterized lncRNAs, which also associate with the PRC2 and other chromatin repressive complexes [121
]. One of these lncRNAs is HOX Antisense Intergenic RNA (HOTAIR) that is a 2.2-kb transcript located at the HOXC gene cluster on chromosome 12 [122
]. HOTAIR also modulates gene expression through epigenetic regulation. Unlike Xist that acts in cis
, HOTAIR functions in trans
to recruit the PRC2 to the HOXD
locus on chromosome 2 to induce transcriptional silencing [122
]. In addition to associating with PRC2, HOTAIR also interacts with the LSD1/CoREST/REST histone modification complex, leading to both histone H3-K27 methylation and H3-K4 demethylation [123
]. Since HOTAIR plays an important role in the epigenetic regulation of its target genes, it is not surprising that its deregulation has been observed in different types of cancers. Recent studies suggest that HOTAIR overexpression is positively associated with increased tumor cell malignancy. Gupta et al.
reported that HOTAIR is overexpressed in both primary and metastatic breast cancer tissues, and its levels in the primary tumors could be used as a significant predictor of subsequent tumor metastasis and survival of the patients [124
]. Ectopically expressed HOTAIR could confer the breast epithelial cancer cells with invasive and metastatic potential while its depletion in breast cancer cells abrogated these activities [124
]. The role of HOTAIR in promoting oncogenesis has also been reported in other cancers. Yang et al.
compared HOTAIR levels between tumorigenic and adjacent non-tumorigenic tissues of hepatocellular carcinoma (HCC) samples and found that this lncRNA was expressed at higher levels in malignant tissues [125
]. In addition, the survival analysis of a cohort consisting of 60 HCC patients revealed that high HOTAIR expression could serve as an independent prognostic marker for disease recurrence and reduced patient survival [125
]. Another HCC-related study suggested that HOTAIR expression is a potential biomarker for lymph node metastasis from the primary tumors [126
]. Furthermore, HOTAIR upregulation was also observed in a cohort of patients diagnosed with the stage IV colorectal cancer (CRC) [127
]. In this study, Kogo et al.
indicated that high HOTAIR expression showed significantly positive correlation with the liver metastasis and poor patient outcome, further supporting that HOTAIR expression is a potential prognostic marker of multiple cancers. In pancreatic cancer, HOTAIR levels were also increased in tumorigenic tissues compared to the non-tumorigenic tissues, and associated with a more aggressive phenotype [128
]. The oncogenic role of HOTAIR in pancreatic cancer cell invasion was validated by its siRNA-mediated knockdown and overexpression studies [128
], consistent with the observations in the other aforementioned cancers. Interestingly, gene array studies showed only a small overlap of HOTAIR-regulated genes between pancreatic cancer and breast cancer [128
], suggesting that HOTAIR may regulate different sets of target genes in a cell type-specific manner.
EZH2 is a core component of the PRC2. EZH2 knockdown followed by chromatin immunoprecipitation demonstrated that HOTAIR-mediated gene repression could be either PRC2-depedent or -independent in pancreatic cancer cells, although the PRC2 is necessary for HOTAIR target gene repression in breast cancer cell [124
]. This discrepancy between the two cancer types suggests the presence of yet unidentified epigenetic mechanisms regulating HOTAIR-mediated transcriptional silencing.
Wang et al.
identified the lincRNA HOTTIP that is transcribed from the 5′ tip of the HOXA locus and modulates the activity of the WDR5-MLL complex, in which the WD40-repeat protein WDR5 binds the MLL complex to activate its histone H3-K4 methyltransferase activity [129
]. Chromosomal looping can make HOTTIP stay in the vicinity of its target genes and let it bind WDR5 to promote the WDR5/MLL complexes-mediated histone H3-K4 methylation, which leads to target gene activation. Thus, HOTTIP serves as a key intermediate to transmit information from higher order chromosomal looping into chromatin modifications [129
]. Another HOXA-related lncRNA is HOX antisense intergenic RNA myeloid 1 (HOTAIRM1) that is transcribed at a direction antisense to the HOXA gene [130
]. The knockdown of HOTAIRM1 reduced the expression of HOXA1 and HOXA4 during the myeloid differentiation in promyelocytic leukemia cells. Whether this lncRNA acts as a miRNA decoy to promote the expression of these HOX genes remains to be determined.
Metastasis-associated lung adenocarcinoma transcript 1 (MALAT1), also known as nuclear-enriched abundant transcript 2 (NEAT2), is one of the first identified cancer-associated lncRNAs [131
]. MALAT1 is a highly conserved noncoding transcript of over 8000 nts encoded by a locus on chromosome 11. It was initially recognized as a prognostic marker of increased metastatic risk for the patients of non-small cell lung carcinoma (NSCLC). Subsequent studies revealed that MALAT1 is localized in nuclear structures enriched with splicing and transcription factors, known as nuclear speckles, suggesting that this lncRNA may modulate alternative splicing of target genes [132
]. Using RNAi-mediated depletion for the components of the nuclear speckles, Miyagawa et al.
demonstrated that the pre-mRNA splicing activator RNPS1, the splicing coactivator SRm160 and the spliceosomal intron binding protein IBP160 promote MALAT1 localization to the nuclear speckles [134
]. Furthermore, MALAT1 depletion and delocalization from the nuclear speckles resulted in downregulation of two interferon-induced genes (OASL and IFI44) and a potential celiac disease susceptibility gene, SPINK4 [134
Despite these studies suggesting a role of MALAT1 in RNA splicing, a recent study by Gutschner et al.
showed that this lncRNA regulates gene expression but not alternative splicing in lung cancer cells [135
]. MALAT1 knockout was achieved by genomic integration of RNA destabilizing elements using zinc finger nucleases, leading to 1000-fold MALAT1 reduction. Using these MALAT1-knockout cells, the authors demonstrated that MALAT1 represses anti-metastatic genes and activates pro-metastatic genes in lung cancer cells, but does not affect genes regulating cell growth. In a xenograft mouse model, the MALAT1-depleted lung cancer cells showed reduced tumor formation compared to the cells with the intact MALAT1. This observation is in contrast to the findings reported by Yang et al.
showing that MALAT1 cooperates with Polycomb 2 protein (Pc2) in regulating the activation of the growth-control gene program in 293T cells [136
]. The different cell types employed in their experiments might contribute to the discrepancy between the two studies. Consistent with this prediction, MALAT1 is widely expressed in most normal human tissues, such as pancreas and lung, but absent in several other tissues including skin, stomach, bone marrow and uterus [131
]. This suggests that this lncRNA may possess tissue specific functions. In addition to its role in regulating lung cancer metastasis, MALAT1 is upregulated in uterine endometrial stromal sarcoma [137
], cervical cancer [4
] and hepatocellular carcinoma [138
], while its expression in the corresponding healthy tissues is undetectable or intermediate [131
]. Although the oncogenic role of MALAT1 in different cancers has been demonstrated by correlational and functional studies, the molecular mechanisms underlying its activities in regulating gene expression and RNA splicing remain undetermined. To date, several studies suggested the essential role of the 3′ end sequence and structure to its metastasis-promoting function, nuclear localization and stability [139
ANRIL is a large antisense ncRNA of the INK4b/ARF/INK4a locus [143
]. Yap et al.
demonstrated that ANRIL binds chromobox 7 (CBX7), a component of the polycomb repressive complex 1. This interaction contributes to the role of CBX in promoting EZH2-mediated H3-K27 methylation at the INK4b/ARF/INK4a locus and consequently represses the tumor suppresser INK4a gene. Consistently, both CBX7 and ANRIL are increasingly expressed in prostate cancer [144
GAS5 (growth arrest-specific transcript 5) is a lncRNA regulating growth arrest of T-cells and lymphocytes [145
]. Ectopic GAS5 increases apoptosis and reduces cell cycle progression. Consistently, its downregulation inhibits apoptosis and promotes cell cycle. Kino et al.
investigated the mechanism underlying the growth suppressive activities of GAS5 and discovered its role in blocking gene expression mediated by glucocorticoid receptor (GR) [146
]. GAS5 binds to the DNA-binding domain of GR and thus prevents its association with the glucocorticoid response element of the GR target genes with anti-apoptotic activities, such as inhibitor of apoptosis 2 (cIAP2). Thus, abundantly expressed GAS5 during starvation can sensitize the cells to apoptosis. The nonsense-mediated mRNA decay (NMD) is a system that controls the quality of gene transcripts and reduces errors in gene expression by eliminating RNAs with premature stop codons [147
]. Meanwhile, this mechanism also regulates the abundance of cellular transcripts, including ncRNAs. Recently, Zhang et al.
demonstrated a reciprocally negative regulation between GAS5 and miR-21 [148
]. While miR-21 represses GAS5 by targeting a sequence encoded by its exon 4, GAS5 inhibits miR-21 expression. Thus, GAS5 antagonizes the oncogenic activity of miR-21 [149
] through reducing its cellular levels. Consistently, miR-21 and GAS5 showed negative correlation in breast cancer specimens [148
]. The tumor suppressive role of GAS5 is supported by the identification of genetic susceptibility of its genomic locus, 1q25, to several cancers, including melanoma [150
], prostate cancer [151
], breast cancer [152
], colorectal cancer [154
] and B-cell lymphoma [155
]. Additionally, Tani et al.
indicated that GAS5 can be stabilized with the depletion of UPF1, an essential component of NMD or during starvation [156
] and the GAS5 introns encode multiple snoRNAs [157
Several other lncRNAs have also been demonstrated to regulate chromatin remodeling and gene transcription. PTENP1 is the PTEN pseudogene and encodes two antisense RNA (asRNA) transcripts, asRNA α and β [161
]. The α asRNA isoform can recruit DNMT3A, EZH2 and G9A to the PTEN promoter and repress its transcription. This will be further discussed below with other regulatory mechanisms of PTENP1. Evf2 is a polyadenylated lncRNA identified in embryonic brain cells [162
]. This lncRNA regulates the transcription of homeodomain transcription factors DLX5 and DLX6 through recruiting DLX and MECP2 to the DNA regulatory elements in the intergenic region of these two genes. As a p53 transactivated lncRNA, lincRNA-p21 is a key mediator of p53-dependent gene repression through a mechanism of recruiting heterogeneous nuclear ribonucleoprotein K (HNRNP) to these p53 target genes [163
]. Thus, inhibition of lincRNA-p21 affects the expression of a number of p53 repressed genes. Sheik et al.
identified an Oct4-activated lncRNA, AK028326, and discovered that this lncRNA regulates pluripotency in mouse embryonic stem cells [164
]. Interestingly, AK028326 activates Oct4 expression in a regulatory feedback loop.
3.2. LncRNAs Acting as miRNA Decoys to Free Target mRNAs
Since the regulation of gene expression by miRNAs was revealed, researchers have been using miRNA sponges, RNA molecules containing the target sequence or reverse complementary sequence of a miRNA to be sponged, as a tool to inhibit the function of miRNAs and release their target gene expression [165
]. Recent studies suggest that this approach naturally exists in cancers to modulate tumor suppressor and oncogene levels. In some literature, these decoy RNAs are also named as competitive endogenous RNAs (ceRNAs).
PTEN (phosphatase and tensin homolog) is a well characterized tumor suppressor with phosphatase activity. It is encoded at the 10q23.3 locus on chromosome 10 and frequently inactivated through diverse mechanisms in human cancers, highlighting its crucial role in oncogenesis. The 3′-UTR of the PTEN mRNA has 3,329 nts, markedly longer than the average 3′-UTR length (740 nts) of eukaryotic mRNAs [166
], implicating its vulnerability as a target of miRNAs. Thus, while PTEN inactivation can be achieved by gene deletion and epigenetic silencing in cancers, its expression is also regulated by multiple miRNAs. For example, the PTEN 3′-UTR contains potential binding sites for over 10 miRNAs overexpressed in glioblastoma multiforme, which is more than 2 times higher than any other tumor suppressor [167
]. PTEN expression can be repressed by miR-21, miR-221 and miR-222 [106
]. Recent studies revealed novel mechanisms regulating PTEN expression through its pseudogene PTENP1 (also called PTH2 or ψPTEN). Pseudogenes are dysfunctional relatives of their cognate genes but have lost the protein-coding ability due to premature stop codons, deletions/insertions or frameshift mutations, and thus cannot be translated into functional proteins [169
]. The PTEN pseudogene PTENP1 is highly transcribed in certain tissues and cells, suggesting that this lncRNA may have biological activities [170
]. The functional relationship between the PTEN and its pseudogene was first discovered by Poliseno et al.
]. In their study, the authors demonstrated that PTENP1 modulates endogenous PTEN transcript levels by acting as a molecular sponge for PTEN-targeting miRNAs; thus, the PTENP1 transcript serves as an effective decoy and exerts tumor suppressive functions (). This novel regulatory role of the PTENP1 can be extended to KRAS1P, the pseudogene of the oncogene KRAS. In a study by Poliseno et al.
, the overexpression of KRAS1P led to increased KRAS mRNA levels in prostate cancer DU145 cells through a mechanism of sequestering KRAS-targeting microRNAs, and consequently promoted cell proliferation [171
Figure 4 PTEN expression is regulated by multiple ncRNAs. (A) The PTEN pseudogene, PTENP1, can transcribe into the PTENP1 lncRNA that acts as a decoy to sponge the miRNAs targeting at the 3′-UTR of the PTEN mRNA; (B,C) The locus of the PTEN pseudogene (more ...)
As briefly discussed above, a recent study by Johnsson et al.
provided evidence of another regulatory mechanism of PTEN by its pseudogene PTENP1 (). The authors discovered that the PTENP1 locus can be transcribed from a reverse direction to create two isoforms of an antisense RNA (asRNA), α and β [161
]. The α isoform of the asRNA binds the PTEN promoter and recruits DNMT3A and two histone methyltransferases EZH2 and G9A, which mediate the methylation of histone H3-K27 and H3-K9, respectively, two well-characterized markers of gene repression [172
]; thus, the PTENP1 α asRNA negatively regulates PTEN gene expression through promoting the epigenetic silencing of its promoter. The β asRNA isoform interacts with the PTENP1 lncRNA through RNA-RNA pairing, which can maintain stable PTENP1 lncRNA levels in cytoplasm, increase its stability and facilitate its role as a microRNA sponge; thus, the PTENP1 β asRNA activates PTEN expression through facilitating the decoy activity of the PTENP1 lncRNA. Overall, the PTENP1 α and β asRNAs exhibit oncogenic and tumor suppressive roles, respectively, based on their effects on PTEN expression (). Whether the two asRNA isoforms are differentially expressed in human cancers remains to be determined.
Interestingly, PTEN expression can also be regulated at the translational level by the sponge effect of the transcript from another gene. The full length ZEB2 mRNA has a very long 3′-UTR (over 5,000 nts) and contains multiple potential binding sites of miRNAs that can also target the PTEN 3′-UTR, such as miR-181a and miR-200/miR-141 [173
]. Thus, the ZEB2 mRNA serves as a decoy or ceRNA of the PTEN mRNA () and reduced ZEB2 expression activates the PI3K/AKT pathway through downregulating PTEN.
There are several other examples showing that lncRNAs act as decoys to stabilize mRNAs. BACE-AS regulates the expression of β-secretase-1 (BACE1), a crucial enzyme in Alzheimer’s disease pathophysiology [174
]. Linc-MD1 is a muscle-specific lncRNA and activates the expression of MAML1 and MEF2C through its decoy role for miR-133 [175
3.3. LncRNAs Regulating mRNA Translation
As discussed above, MALAT1 was initially demonstrated to possess activities in regulating alternative splicing. MALAT1 associates with serine/arginine (SR) splicing factors at the nuclear speckle domains of nucleus and is involved in the process of alternative splicing [133
]. Thus, MALAT1 depletion or ectopic SR protein expression affects the alternative splicing of a similar set of pre-mRNAs. Additionally, MALAT1 alters the phosphorylation of the SR splicing factors, which is essential to their activities in regulating alternative splicing. However, this regulation may only be present in specific cell types or under particular conditions, because MALAT1-knockout mice were viable and fertile, and showed regularly localized nuclear speckle markers [176
]. Whether this regulatory mechanism contributes to the activity of MALAT1 in promoting tumor metastasis has not been determined.
A recent study from Zhang et al.
demonstrated the lncRNA, lincRNA-RoR, acts as a strong negative regulator of p53 [177
]. LincRNA-RoR reduces p53 expression in cells exposed to DNA damage stress, but not in unstressed cells, through directly binding to the heterogeneous nuclear ribonucleoprotein I (hnRNPI) and repressing p53 mRNA translation. As an autoregulatory feedback regulation, p53 transcriptionally induces lincRNA-RoR expression. The functional interplay between lincRNA-RoR and p53 may serve as additional surveillance to mediate cell response to genotoxic stresses.
As discussed above, another p53-transactivated lncRNA, lincRNA-p21, is involved in p53-dependent gene transrepression [163
]. A recent study from Yoon et al.
demonstrated that lincRNA-p21 also modulates translation [178
]. This lncRNA can associate with JunB and β-catenin mRNAs to reduce their translation rates. A RNA-binding protein, HuR, can recruit let-7/AGO2 to lincRNA-p21 to reduce its stability; thus, elevated HuR releases lincRNA-p21-mediated repression of JunB and β-catenin expression. Since lincRNA-p21 is transactivated by p53, the negative regulation of JunB and β-catenin by lincRNA-p21 is consistent to the tumor suppressive role of p53.
3.4. LncRNAs Regulating Protein Activities
Telomeres are the DNA-protein complexes at the end of eukaryotic chromosomes and essential to chromosome stability. As a frequently activated reverse transcriptase, telomerase adds DNA sequence repeats (“TTAGGG” in vertebrates) to the telomere regions of chromosome to maintain the telomere length. Azzalin et al.
discovered that telomeres can be transcribed into telomeric repeat-containing RNA (TERRA) [179
]. These molecules have different lengths and were predicted to play a role in the maintenance of telomere integrity through an unclear mechanism. A later study from Redon et al.
provided a possible mechanism underlying this activity of TERRA [180
]. The authors demonstrated that TERRA binds telomerase and thus acts as a potent competitive inhibitor for the telomeric DNA. Consistently, TERRA expression is significantly downregulated in multiple tumor cell lines, which plays a role in telomere maintenance and cell immortalization of cancer cells.
Wang et al.
demonstrated an indirect regulation of the histone acetyltransferase activities of CBP and p300 by ncRNAs [181
]. The 5′-regulatory region of the cyclin D1 gene encodes at least four ncRNAs. In response to ionizing radiation, these ncRNAs bind TLS (translocated in liposarcoma) protein at the chromatin of the cyclin D1 promoter region and cause an allosteric effect on this protein. This TLS conformational change promotes its activity of inhibiting CBP/p300-mediated histone acetylation and consequently silences cyclin D1 gene. These data indicate that ncRNAs encoded by a promoter can act as selective ligands to modulate the activities of transcription cofactors in response to genotoxic stresses.
LncRNAs can also regulate protein activity through altering their subcellular localization. Nuclear factor of activated T-cells (NFAT) represents a family of transcription factors regulating immune response. Some of NFAT proteins, such as NFAT1 and NFAT5, contribute to tumor metastasis and cell motility [182
]. The noncoding repressor of NFAT (NRON) is a ncRNA associated with multiple proteins [184
]. NRON binds phosphorylated NFAT1 to sequester NFAT in the cytoplasm. The depletion of NRON leads to NFAT dephosphorylation and nuclear import. Thus, ncRNAs can be a part of a scaffold to trap a latent transcription factor [185
] and regulate the expression of its target genes.