Long non-coding RNAs in Colorectal Cancer: Progression and Future Directions.

Identification of the colorectal adenoma-carcinoma sequence with its corresponding genetic and epigenetic alterations has significantly increased our knowledge of the etiopathogenesis of colorectal cancer (CRC). However, the molecular mechanisms of colorectal carcinogenesis and metastasis haven't been clearly elucidated. Long non-coding ribonucleic acids (lncRNAs) are key participants of gene regulations rather than "noises". Accumulative studies have implicated that the aberrant expressions of lncRNAs are tightly corelated to CRC screening, diagnosis, prognosis and therapeutic outcomes. Our review focuses on recent findings on the involvement of lncRNAs in CRC oncogenesis and the lncRNA-based clinical implications in patients with CRC.

Introduction

Genomic studies over the past several decades have confirmed that only approximately 2% of human gene transcripts are stably translated into proteins, generating numerous non-protein-coding ribonucleic acids (ncRNAs)1. Mounting evidence suggests that these non-coding transcriptional “noises” encompass a wide variety of subclasses, including not only “housekeeping” ncRNAs [(such as ribosomal RNAs (rRNAs), transfer RNAs (tRNAs), small nuclear RNAs (snRNAs, also known as u-RNAs), small nucleolar RNAs (snoRNAs), and Ribonuclease P (RNase P)] but also “regulatory” ncRNAs2. The growing numbers of “regulatory” ncRNAs contain two main groups: the short ncRNAs [such as microRNAs (miRNAs), short interfering RNAs (siRNAs), and PIWI-interacting RNAs (piRNAs)] and long noncoding RNAs (lncRNAs)3. All of these ncRNAs form a network-like RNA infrastructure4.

Over recent years, the different subclasses of short regulatory ncRNAs, such as miRNAs, have been relatively well recognized and studied based on their structural features and biologic functions. For example, we previously reported that the elevation of oncofetal miR-17-5p targets the nucleolar and coiled-body phosphoprotein 1 (NOLC1, also known as P130) gene, and subsequently activates the Wnt/β-catenin pathway during oncogenesis and CRC progression5. Moreover, the weak miR-150 expression status of CRC patients links with a poorer response to adjuvant chemotherapy and is considered a potential prognostic biomarker in CRC6.

Most recently, our knowledge has been advanced by the emergence of next-generation sequencing (NGS) technology and the development of bioinformatics methods to investigate lncRNAs7. lncRNAs are typically defined as endogenous noncoding RNA molecules (> 200 nucleotides) that govern many genes expressions at the epigenetic, transcriptional, or post-transcriptional levels8. Building upon their genomic proximity to protein-coding genes, lncRNAs may be crudely divided into different groups: intergenic transcripts (lincRNAs), intronic transcripts, overlapping transcripts either in sense or in antisense, and processing-type transcripts that cannot be placed in any of the other categories9, 10. The past decade has witnessed a steep rise in interest in lncRNA research. Referring to the latest database, NONCODE v4.0 (http://www.bioinfo.org/noncode/), more than 54,000 and 46,400 lncRNA genes have been identified in humans and in mice, respectively11. However, a majority of lncRNAs remain to be characterized. Advances in understanding the basic mechanisms of the involvement of lncRNAs in cancer have raised the field of cancer research well beyond original expectations. Changes of lncRNAs expression profiles have been observed in many human CRC, and such discovery may open a new avenue for the lncRNA-based clinical applications in CRC12.

As the fourth leading reason of cancer-related mortality, CRC is associated with tremendous social and economic burden worldwide13. The pathogenetic mechanisms underlying CRC development are complex. The “adenoma-carcinoma sequence” multi-step model proposed by Fearon and Vogelstein suggests that accumulating genetic and epigenetic mutations in the colorectum drive epithelial dysplasia and proliferation, ultimately resulting in CRC14.

Both intrinsic (such as age, diabetes, obesity, adenomatous polyposis, and inflammatory bowel disease) and extrinsic (such as smoking, inadequate fiber, fish and vitamin D intake, hard-drinking , and high-fat diet) factors have been shown to be associated with CRC risk. Regarding intrinsic factors, We previously conducted a systematic review in approximately 9,000,000 participants from 41 prospective studies15. We confirmed that both general obesity and central obesity are positively associated with CRC risk15. Regarding extrinsic factors, we previously summarized 9 prospective studies with approximately one million participants for meta-analysis, and found inverse correlation between vitamin D intake/ blood 25-hydroxyvitamin D levels and CRC risk16. Moreover, although previous studies have documented the dysregulation of many oncogenes [including the Kirsten rat sarcoma viral oncogene homolog (KRAS) and v-myc avian myelocytomatosis viral oncogene homolog (MYC, also known as c-Myc)] and oncosuppressive genes (including the adenomatous polyposis coli (APC), β‑catenin (CTNNB1), deleted in colorectal cancer (DCC) and P53) in CRC, the molecular and genetic basis of colorectal carcinogenesis and metastasis remain largely unknown17. Therefore, there is still an urgent need to improve our understanding of the CRC etiopathogenesis and to develop additional tumor markers and therapeutic targets for this disease.

In this review, we will first provide an overview of the biology and function of lncRNAs. We will then focus on the involvement of lncRNA dysregulation in CRC etiopathogenesis. Finally, we will discuss the latest advance of the clinical implications of lncRNAs and their association with the diagnosis, prognosis and potential treatment of CRC.

Origins and functions of lncRNAs

Understanding the origins and molecular mechanisms of lncRNAs is critical for characterizing lncRNA function. The emergence of functional noncoding RNAs can be divided into the following non-mutually exclusive hypotheses18: (1) frame disruptions of a protein-coding gene generates a lncRNA; (2) two untranscribed and divided sequences are juxtaposed following chromosomal rearrangement and are transformed into a multi-exon lncRNA; (3) duplication of a noncoding gene following retrotransposition gives rise to lncRNAs; (4) tandem duplication events of neighboring repeats give rise to a lncRNA; and (5) a transposable element (TE) inserted into a gene emerges as a new lncRNA. However, the origins of many lncRNAs remain elusive except for several typical transcripts, such as X-inactive specific transcript (Xist). Recent studies reported by Derrien at al10 and Ulitsky at al19 suggest that a novel lncRNA gene can occur de novo either from untranscribed sequence or from TE, which is in contrast to protein coding genes.

While the origin and evolution of lncRNAs are still obscure, even less is known regarding their precise molecular functions. Although previously regarded as “junk DNA” because of their overwhelming burden of transposons, pseudogenes, and simple repeats, there is mounting evidence that lncRNAs exert a diverse spectrum of regulatory functions, including X chromosome inactivation (XCI), genomic imprinting, tissue regeneration, cell cycle control, apoptosis, aging and carcinogenesis1, 20.

To better understand their functional significance, lncRNAs can be further reclassified based on their molecular mechanisms into following categories (Figure ): (1) signaling molecules (e.g., Xist, lincRNA-p21, COLDAIR, or HOTAIR), which are transcribed in a time- or space-specific manner in response to diverse stimuli, and reflect the integrated transcriptional activity; (2) molecular decoys (e.g., MALAT1, PTENP1, IPS1 or PANDA), which bind and competitively inhibit transcription factors, miRNAs or other targets from chromatin, but do not exert any additional functions; (3) molecular guides (e.g., Xist, COLDAIR, lincRNA-p21 or HOTAIR), which can recruit chromatin-modifying enzymes to specific target sequences, either in cis or in trans; and (4) molecular scaffolds (e.g., TERC, HOTAIR, ANRIL), which can work as central platforms and assemble proteins together to form architecture of ribonucleoprotein complexes21. In fact, several lncRNAs have multiple modes of action that are critical to their biological function21. For example, the Hox transcript antisense intergenic RNA (HOTAIR), located at chromosome 12q13.13, is a lincRNA that may function not only as an anatomic signal that guides chromatin-modifying complexes to their targets but also as a molecular scaffold for histone modification complexes9. Thus, it is apparent that the molecular mechanisms of genes regulation and cellular functions mediated by lncRNAs may be far more sophisticated than by miRNAs.

Methods of lncRNA detection or quantification

Several techniques are currently available that can be used for lncRNA detection and quantification, including RNA sequencing (RNA-seq), RNA microarray, reverse transcription polymerase chain reaction (RT-PCR), fluorescence in situ hybridization (FISH), and RNA blot analysis12. Most lncRNAs operate as RNA-protein complexes, and advanced methods for studying lncRNA-protein interactions include the following: RNA interference (RNAi), RNA Pull-down, RNA-binding protein immunoprecipitation (RIP), chromatin isolation by RNA purification (ChIRP), ChIRP-sequencing (ChIRP-seq), crosslinking-immunopurification (CLIP) and CLIP-sequencing (CLIP-seq)22.

Meanwhile, the available computational approaches and bioinformatics resources allow researchers to better characterize the functional lncRNAs and to shed additional light on the roles of lncRNAs in CRC23. The related database, annotation tools and other bioinformatics resources available are listed in Table .

lncRNA expression profiles in CRC

Research focused on the molecular mechanisms of CRC has highlighted the stepwise progression following the adenoma-carcinoma sequence. Accumulated evidence indicates that CRC-associated lncRNAs acts as master gene regulators through different mechanisms. With the aid of tools ranging from global lncRNA expression profiles with RNA-seq or RNA microarray to selected lncRNAs expression examining with qRT-PCR or FISH, several studies have detected a number of consistently and reproducibly altered CRC-associated lncRNAs both in vivo and in vitro38-42.

The first CRC-related lncRNA was identified in 1996 when Hibi and colleagues43 reported that the endogenous H19 gene is frequently abundant in CRC specimens, and the overexpression of H19 plays an important role in CRC development. Since that discovery, accumulating reports of dysregulated expression of lncRNAs in CRC have suggested that aberrant lncRNAs are involved in all stages of CRC (Table ). Recent studies are aimed at determining whether lncRNAs can be used as novel predictors for CRC screening, diagnosis, prognosis and therapeutic outcomes.

Involvement of lncRNAs in CRC pathogenesis

It is widely accepted that CRC tumorigenesis is a multistage process76. CRC develops typically from normal mucosal epithelium to an benign adenoma and ultimately progresses to malignant tumor77. Understanding the tumorigenic signaling pathways that involve lncRNAs would provide further insights into CRC pathogenesis.

lncRNAs participate in many known signaling pathways involved in CRC pathogenesis, including the Wnt/β-catenin signaling pathway, epidermal growth factor receptor (EGFR)/insulin-like growth factor type I receptor (IGF-IR) signaling pathway (KRAS and phosphatidylinositol-3-kinase (PI3K) pathways), transforming growth factor-beta (TGF-β) signaling pathway, p53 signaling pathway, and the epithelial-mesenchymal transition (EMT) program (Figure ).

The Wnt/β-catenin pathway

The disruption of the Wnt/β-catenin pathway contributes to CRC initiation78. The β-catenin destruction complex, comprising of multiprotein like Axin, Glycogen Synthase Kinase 3β (GSK-3β), APC, casein kinase 1 (CK1), protein phosphatase 2A (PP2A), activator protein 2α (AP-2α), are switch controls for Wnt signaling79. Malfunction of this destruction complex leads to stimulation of the Wnt signaling via the translocation of accumulated free β-catenin into the nucleus and subsequent activation of T-cell factor(TCF)/lymphoid enhancing factor (LEF) targets, such as cyclin D1 (CCND1), E2F transcription factor-4 (E), and proto-oncogene c-Myc80.

Inactivation of the oncosuppressive gene APC is the main rate-limiting step of Wnt signaling, occurring in more than 80% of sporadic CRCs78, 81. The APC gene has been reported to be modulated by a lncRNA called H1972. H19 is a paternally imprinted and maternally expressed transcript located at 11p15.582. Oncofetal H19 is elevated during embryogenesis, but is inactivated postnatally and re-expressed during tumorigenesis83. Based on their findings that H19 knockout mice never spontaneously develop tumors, Yoshimizu et al. demonstrated that H19 acts as a oncosuppressive gene rather than as an oncogene72. Meanwhile, in the absence of H19 expression in the APC mutant CRC murine model, the number of adenomas, especially smaller polyps, were significantly higher than in its presence, suggesting that H19 could control the early event of CRC tumorigenesis72.

AP-2α belongs to β-catenin destruction complex84. In vivo and in vitro studies showed that AP-2α served as a oncosuppressive protein by interacting with APC and β-catenin84. One of our recent study has shown that new-found colorectal cancer-associated lncRNA (CCAL) negatively regulated AP-2α and then opened the way to Wnt signaling in CRC tumorigenesis44.

The E2F family of transcription factors are in charge of the cell cycle (G1/S transition) and the action of oncosuppressive genes85. Yochum and colleagues49 revealed a β-catenin-regulated lncRNA derived from the E locus termed E antisense transcript. E antisense transcript is expressed in HCT116 β-catenin signaling-stimulated CRC cells49. Both β-catenin and TCF4 could bind to the 3'-untranslated region (3'-UTR) sequences of E, driving expression of this lncRNA that reduces E levels and decreases binding of E to promoters of targets such as proliferating cell nuclear antigen (PCNA) and cyclin A2, ultimately activating the expression of genes repressed by E49.

The MYC region upstream contains an large tumor-type-specific super enhancer86. Xiang et al.41 showed a super-enhancer located at locus 8q24.21, 515 kb MYC upstream region, that transcribes a human CRC-specific lncRNA CCAT1-L. CCAT1-L is known as an isoform of the recently described lncRNA colon cancer-associated transcript-1 (CCAT1)87, which is exclusively localized to the nucleus, and its expression is upregulated in CRC cell lines and CRC tissue samples41. Furthermore, CCAT1-L participates in a positive control of MYC transcription by interacting with a zinc finger protein, CCCTC-binding factor (CTCF), and facilitating CTCF binding to MYC locus41. In addition, Ling et al.46 concluded that a novel gene transcript named colon cancer-associated transcript-2 (CCAT2), which maps to the 8q24.21 genomic region, is a downstream target of Wnt. CCAT2 is overexpressed in microsatellite-stable (MSS) CRC samples46. Higher MYC expression was found in CRC cells overexpressing CCAT2, and transient knockdown of CCAT2 reduced MYC expression46. Additionally, there is a complex feedback loop between CCAT2 and Wnt signaling. On the one hand, CCAT2 could broadly enhance Wnt activities by binding to TCF7L2, modulating its transcriptional activity46. On the other hand, TCF7L2 knockdown consistently reduced CCAT2 expression, indicating that TCF7L2 is necessary for the maintenance of high CCAT2 expression levels46.

The observations described above suggest that alteration of lncRNA expression participates in the early event of the etiopathogenesis of CRC.

EGFR/IGF-IR signaling (KRAS and PI3K pathways)

The epidermal growth factor receptor (EGFR) is one of the members of human EGFR family of receptor tyrosine kinases88. Functional stimulation of their receptors occurs in a broad spectrum of solid tumors including CRC88. Like EGFR, insulin-like growth factor-I receptor (IGF-IR) is also a ubiquitous transmembrane tyrosine kinase that plays crucial roles in colorectal tumorigenesis and progression89. Binding of their respective ligands to the EGFR and IGF-IR lead to their autophosphorylation and subsequent activation of the KRAS/Raf/mitogen-activated protein kinase (MAPK) and PI3K/AKT/mammalian target of rapamycin (mToR) signaling cascades, contributing to uncontrolled cellular proliferation, adhesion, angiogenesis, migration, metastasis, and survival90.

The newly identified IGF driven gene termed colorectal neoplasia differentially expressed (CRNDE), located at locus hCG_1815491 on chromosome 16, is activated early in CRC and potentially interacts with chromatin-modifying complexes to engage in gene expression47. Ellis et al.48 reported that CRNDE is one of the downstream targets of both the PI3K and KRAS signal cascades. Upregulation of CRNDE expression promotes the metabolic shift of neoplastic colon cells from oxidative metabolism to aerobic glycolysis, contributing to the “Warburg effect”, a hallmark of CRC cell metabolism48.

Another lncRNA that is involved in the pathogenesis of CRC is a pseudogene called phosphatase and tensin homolog pseudogene 1 (PTENP1) located at 9p13.3 91. PTENP1 is highly homologous to the crucial tumor-suppressor gene PTEN, an inhibitor of PI3K pathway signaling70. Poliseno et al.70 showed that the bona fide tumor suppressor gene PTENP1, which contains a relatively shorter 3'UTR than that of PTEN, functions as a decoy to bind to the PTEN-targeting miRNAs, allowing for the expression of the PTEN protein. Moreover, deletion of PTENP1 in CRC directly correlates with reduced PTEN expression70.

The TGF-β signaling pathway

TGF-β is a 25-kDa cytokine that signals via transmembrane serine-threonine kinase receptors(STKR) and intracellular mediators of Smad family92. The TGF-β signal pathway plays dichotomous roles in the CRC progression93. In the early phase of tumorigenesis, TGF-β signaling acts as a cancer suppressor. In contrast, in the late cancer stages, it accelerates CRC invasiveness. Improved understanding of how the TGF-β signaling cascade mediates interactions with lncRNA will provide important insights into colorectal carcinogenesis, development and progression94.

Some researchers have demonstrated the involvement of lncRNAs in TGF-β signaling pathways that are associated with various cancer types, including CRC50. TGF-β signaling is induced by the aberrant expression of lncRNAs. The human PVT-1 oncogene, which encodes the lncRNA PVT-1, is involved in cancer pathophysiology95. Takahashi and colleagues58 conducted gene expression microarray analyses using PVT-1 knockdown of CRC cell lines RKO and HCT116 and found that the proteins relating to the canonical TGF-β signaling (such as tumor suppressor gene Smad4) and apoptosis were significantly activated by PVT-1 knockdown.

On the other hand, lncRNAs can also be activated by TGF-β. For example, Yuan and colleagues54 discovered a novel TGF-β-induced lncRNA termed lncRNA-activated by TGF-β (lncRNA-ATB) that promotes the invasion-metastasis cascade in hepatocellular carcinoma through Smad-independent signaling. lncRNA-ATB competitively binds to and sponges miR-200s away from their mRNA targets, zinc-finger E-box binding homeobox family proteins (including ZEB1 and ZEB2), which encode the key EMT-promoting transcription factors, thus promoting EMT54. Notably, lncRNA-ATB is also significantly upregulated by TGF-β in the Smad4-deficient CRC cell line SW480, indicating that lncRNA-ATB also promotes CRC progression through the noncanonical TGF-β pathway50, 54.

Taken together, these findings suggest an existence of a complex feedback loop between the lncRNAs and the TGF-β signaling pathway.

The p53 pathway

The tumor suppressor p53, known as “a cellular gatekeeper”, provides an intrinsically powerful defense against CRC. As a master transcription factor, p53 not only upregulates anti-proliferative and proapoptotic genes such as cyclin-dependent kinase inhibitor 1 (p21) (G1 arrest), cyclin B1 (G2 arrest), p53-upregulated modulator of apoptosis (PUMA; apoptosis) and the plasminogen activator inhibitor 1 (PAI-1; cellular senescence), but it also suppresses the genes involved in control of cell proliferation such as polo-like kinase-1(PLK1) and apoptosis (Bcl-2)96, 97. Loss of p53 function, through either mutations in p53 itself or perturbations in pathways signaling to p53, is a frequent alteration in the majority of human cancers, including CRC61.

Strong evidences have unraveled the link between p53 and the lncRNA network98. Several lncRNAs, such as MEG3, MALAT1, p53-eRNAs, and Wrap53, serve as p53 regulators, while other lncRNAs, such as Loc285184, lncRNA-p21, H19,and PANDA, serve as p53 effectors99. For example, the human maternally expressed gene-3 (MEG3), with a length of 1595 nucleotides, is an imprinted ncRNA located at human 14q32.3100. As a p53 regulator, MEG3 downregulates expression of the well-known negative p53 regulator, mouse double minute 2 homolog (MDM2), selectively regulates p53 targets expression, ultimately inhibiting CRC pathogenesis and progression67. Furthermore, the expression of MEG3 is under epigenetic control, and because of hypermethylation in the MEG3 promoter, MEG3 loss is common in many human tumors (including CRC) and tumor cell lines (including colon cell lines HCT116 and H29)67, 68, 101, 102. Thus, MEG3 could represent a tumor suppressor gene.

Unlike the p53 regulator MEG3, the lncRNA Loc285194 (also referred to as LSAMP antisense RNA 3) serves as a p53 effector. The Loc285194 gene, located at 3q13.31 and consisting of 4 exons for a total of 2105 nt in length, was originally identified to be downregulated in osteosarcoma samples and subsequently reported to be downregulated in CRC66. Through in vitro and in vivo analyses, Liu and colleagues65 further reported that the Loc285194 upstream region is a putative transcription target of p53. As an endogenous sponge, Loc285194 inhibits CRC cell growth by binding to and repressing oncogenic miR-21165. These results imply Loc285194 as a potential CRC suppressor.

In addition to Loc285194 expression, which is triggered by p53, Huarte and colleagues103 identified a p53-activated lincRNA called lincRNA-p21 that played an important role in triggering apoptosis. lincRNA-p21 is also a direct p53 transcription target near the cyclin-dependent kinase (CDK) inhibitor p21 (CDKN1A) locus, a canonical transcriptional target of p53104. Loss of lincRNA-p21 affects many genes that are usually suppressed by p53, whereas expression of this lincRNA functions as a transcriptional repressor by modulating the localization of heterogeneous nuclear ribonucleoprotein K (hnRNP-K), a well-known inhibitor complex in the p53 pathway103.

In light of these findings, we conclude that lncRNAs also belong to the p53-regulatory network and are involved in CRC tumorigenesis.

The EMT program

Distant metastasis is the primary cause of CRC-related mortality105. Recent preclinical or clinical research has implicated the activation of the EMT program and its reverse program in the metastatic process in CRC106. The hallmarks of EMT include the downregulation of E-cadherin (CDH1) and upregulation of non-epithelial cadherins, like N-cadherin (CDH2)1. Several reasons contribute to the disturbance of controlled epithelial balance, including the altered expression levels of lncRNAs107.

Several researches have highlighted the roles of lncRNAs in regulating the epithelial phenotype by controlling EMT inducers such as ZEB1, ZEB2 and Snail zinc finger family proteins [including Snail1 and Snail2 (also known as Slug)]108. For example, metastasis-associated lung adenocarcinoma transcript-1 (MALAT1) has been confirmed to be an important mediator of TGF-β-induced EMT109. Upregulated MALAT1 promotes EMT by regulating expression levels of ZEB1, ZEB2 and Slug and activating Wnt signal during the metastasis of bladder cancer110. Because the overexpression of MALAT1 was previously confirmed to be involved in CRC progression, we hypothesized that MALAT1 may also play critical roles in CRC cell EMT programs; this hypothesis remains to be validated.

A second lncRNA functioning as a positive moderator of EMT progression in CRC is H19. H19 was found to be upregulated in two EMT model, including Methotrexate (MTX) resistant HT-29 CRC cell, TGF-β1-treated HT-29 CRC cell and TGF-β1-treated SW620 CRC cell73. H19 modulated ZEB1 and ZEB2 expression levels by functioning as decoys for miR-138 and miR-200a, promoting EMT transition in CRC73. Moreover, lncRNA H19 enhanced the invasive potential of cancer cells by inhibiting E-cadherin via induction of the expression of miR-675-dependent Slug107.

A third lncRNA participating in EMT-controlled CRC metastasis is called HOTAIR. Wu and colleagues111 revealed that HOTAIR overexpression promotes CRC cell migration and invasiveness, while depletion of HOTAIR suppresses the EMT program, including upexpression of E-cadherin and loss of vimentin and matrix metalloproteinase-9 (MMP9). These examples demonstrate the functional importance of lncRNAs in invasion and metastasis.

Clinical implications of lncRNAs in CRC

Effect of above-mentioned lncRNAs in CRC etiopathogenesis clearly revealed that dysregulated lncRNAs in CRC may be of significant clinical value (Figure ). Several lncRNAs have emerged as promising biomarkers for susceptibility screening, diagnosis, prognosis, and lncRNA-based therapies for CRC12. Specifically, polymorphisms within lncRNAs encompass regions that may contribute to susceptibility to CRC, and circulating lncRNAs may be promising markers for the early detection of CRC. Meanwhile, the fact that abnormal lncRNA expression in CRC correlates with patient survival or treatment outcome provides the theoretical rationale for utilizing lncRNAs to be promising therapeutical targets.

Polymorphisms within lncRNA-encompassing regions as susceptibility screening markers

Genome-wide association studies (GWAS) combined with single nucleotide polymorphism (SNP) arrays and NGS technology have identified powerful associations of multiple variants, such as SNPs and other genetic abnormalities located on the chromosomal region 8q24 with the onset of CRC112. Several highly conserved lncRNAs were discovered to be transcribed from the "gene-desert" region of 8q2446, 68, 113. Therefore, oncogenic SNPs located within these lncRNAs transcripts may involve in CRC carcinogenesis.

For instance, the lncRNA CCAT2, which maps to the 8q24.21 region, encompasses a cancer-related rs6983267 SNP that is faithfully linked to CRC oncogenesis, progression, and prognosis through regulating MYC transcription46. Using pyrosequencing data on heterogeneous cell lines (DLD1 and KM12SM cells) and CRC samples, Ling et al.46 found that the G allele of rs6983267 produces more CCAT2 transcripts and induces higher MYC expression than the T allele. The variant genotype of rs6983267 affects the expression levels and functions of CCAT2, providing an alternative explanation of the risk of SNP-conferred CRC.

A second lncRNA involved in cancer-related gene polymorphisms is known as prostate cancer non-coding RNA-1 (PRNCR1). Li et al.68 identified 4 new functional CRC susceptibility variants (rs13252298, rs7007694, rs16901946, and rs1456315) in the lncRNA PRNCR1 at the 8q24 locus. The authors further demonstrated that the rs13252298 and rs1456315 alleles may contribute to a decreased CRC risk, whereas polymorphisms of the others may promote CRC growth and differentiation 68. This evidence suggests that SNPs in the lncRNA locus likely increase CRC susceptibility.

Considering the results of these two studies, it is anticipated that with further information from GWAS, more polymorphisms in the lncRNA locus will be discovered, allowing clinicians to better assess the genetic susceptibility to CRC screening.

Potential of lncRNA expression in CRC early diagnosis

Due to the screening of fiber-optic colonoscopies, fecal occult blood test (FOBT), and stool-based DNA assays, 5-year CRC-related mortality rates have slightly declined to date. Nonetheless, colonoscopies are invasive and cost intensive, whereas FOBT and stool-based DNA assays have hyposensitivity and low specificity. Until now, only a few functionally characterized serum candidates, such as carcino-embryonic antigen (CEA) and carbohydrate antigen 19-9 (CA-19-9), have been clinically utilized. However, neither CEA nor CA19-9 meets expectations for early detection or prognosis of CRC. Consequently, there is an urgent need for seeking novel biomarkers which are sensitive or specific enough.

Compared with the protein-coding RNAs or their proteins, the use of a lncRNA as a biomarker is of great advantage because its expression is a better indicator of tumor status114. The lncRNA expression patterns, either in tissues or in circulation, have the potential to be diagnostic markers for CRC in clinical routine and could lessen the number of unnecessary colonoscopies follow-up procedures.

For example, Tanaka and coworkers52 found a lncRNA termed long QT intronic transcript-1 (LIT1/KCNQ1OT1) which was frequently (40%) detected to be loss of imprinting (LOI) in the CRC tissues, whereas negative in peritumoral samples. Coincidentally, Nakano et al.53 supported the fact that LOI of LIT1 was observed only in CRC tissue specimens at a high frequency (53%), but negative(0%) in adjacent normal samples, thereby making LIT1 a potential for the diagnosis of CRC.

Furthermore, Pedersen et al.62 reported that the lncRNA called colorectal adenocarcinoma hypermethylated (CAHM) was hypermethylated in 81% of adenomas and 71% of CRC but not in normal tissue (8%); meanwhile, methylated CAHM was observed in the plasma of 55% of CRC compared with 4% of adenomas and 7% of subjects without neoplasia, showing the promise of CAHM as a blood candidate for early detection of CRC.

Another potentially diagnostic lncRNA is CRNDE47. Graham and colleagues47 reported that CRNDE upregulation is an early event in CRC. In tissue samples, the expression levels of the CRNDE transcript were found to be elevated in more than 90% of tissues from colorectal adenomas and adenocarcinomas, whereas levels were very low in normal tissues or inflammatory bowel disease (IBD) samples. In plasma samples, CRNDE transcript was positive in 13 of 15 CRC patients but in only 1 of 15 controls, affording a sensitivity and specificity of 87% and 93%47.

Moreover, CCAT1 is another promising RNA biomarker candidate42. The 2628 nucleotide-long lncRNA CCAT1, located in the vicinity of c-Myc on chromosome 8q24.21, was first identified from blood and fecal samples from CRC patients in 2005115. Studies from the same group subsequently examined the CCAT1 levels in human biospecimens and found out that CCAT1 was markedly upregulated across the colon adenoma-carcinoma sequence45; meanwhile, peripheral blood samples of CCAT1 was shown to be overexpressed (up to 1700 fold) in approximately 40% of CRC, but not in that of healthy volunteers42. The researchers first described the use of a CCAT1-specific peptide nucleic acid (PNA)-based molecular beacons (TO-PNA-MB) to be a powerfulpredictor for specific identification of CRC in vitro, ex vivo, and in vivo38.

Taken together, the ectopic presences of these lncRNAs in tissues, stools, and blood samples raise hope for using combining optimized lncRNAs for detection of CRC.

Potential of lncRNA expression for CRC prognosis and prediction

Currently, the only comprehensive criteria employed in clinical practice to provide an indication of CRC prognosis or to guide adjuvant chemotherapy choices is the American Joint Committee on Cancer (AJCC) tumor node metastasis (TNM) staging system116. Regrettably, the TNM system hasn't provided the clinicians with the optimal staging allocation, but has resulted in substantial under- or over-treatment of certain CRC patients over the past decades117, 118.

Recent studies suggest that both oncosuppressive and oncogenic lncRNAs may have potential to be used as efficient biomarkers of CRC in clinical routine.

For example, Zhai et al.64 revealed a contradiction regarding the function of tumor suppressor lincRNA-p21 in CRC progression. On the one hand, the expression levels of lincRNA-p21 in CRC tissue samples were significantly lower compared with the tumor-adjacent samples. On the other hand, as the CRC stage progresses, the expression levels of this lincRNA in tumor tissues were significantly elevated64. The finding indicates that the disruption of the p53/lincRNA-p21 network may contribute to CRC development and malignant transformation64. Another notable candidate is non-coding RNA expressed in aggressive neuroblastoma (ncRAN). This lncRNA located at 17q25.1 is dramatically downregulated in CRC tissues and CRC cell lines, such as DLD1, LoVo and HCT11669. Moreover, decreased expression of ncRAN leads to CRC metastasis and is correlated to poorer overall survival (OS) in CRC patients69. A third example of a tumor suppressive lncRNA is Loc285194. Qi et al66 showed that the prevalent loss of Loc285194 in CRC tumor tissues was positively related to a larger tumor size, a poorer TNM stage, an increased number of distant metastases and a poorer disease-free survival (DFS)66. Hence, Loc285194 might function as a novel prognostic indicator in CRC.

Apart from tumor suppressive lncRNA, the group of oncogenic lncRNAs whose expression is increased may also serve as potential prognostic factors for CRC.

For example, Kogo et al.39 found that the upregulation of HOTAIR was frequent in stage IV CRC tissues with liver metastases. Using gene set enrichment analysis, the authors further revealed that HOTAIR expression was significantly correlated to the polycomb-repressive complex-2 (PRC2) gene function, thus contributing globally to cancer progression39, 119. Ge et al57 reported that the lncRNA prostate cancer-associated ncRNA transcript-1 (PCAT1) gene, located on the chromosome 8q24 725 kb upstream of the c-Myc, was more frequently upregulated in CRC tumor tissues than in normal adjacent tissues. Moreover, upregulation of PCAT1 implied a poorer OS time in patients with CRC57. Therefore, this lncRNA may be used to complement the TNM staging system to predict prognosis57. Our recent study shown that overexpression of lncRNA CCAL in tumor tissues of patients with CRC was connected with a poorer OS and a worse therapeutic outcome, implying CCAL as good indicator for guideness of clinical chemotherapy.

Compared with single utilization mode, the ensemble features of lncRNAs may have an obvious advantage for CRC prognosis. For example, Hu and colleagues120 performed a lncRNA expression profiling in large CRC cohorts, identifying a set of six lncRNAs (AK024680, AK123657, CR622106, BX649059, BX648207, and AK026784) that correlated with the DFS of CRC. Based on this finding, the authors successfully classified CRC patients into high- or low-risk subgroup for survival prediction120. Furthermore, the six-lncRNA signature combined with the AJCC TNM staging system showed a stronger power for DFS prediction in the ROC analysis, implying that the stage-independent lncRNA signatures may have potential to refine the current prognostic model120.

Although these results above are promising, further studies remain necessary to make sure that the lncRNAs may provide more information than current standard diagnostic procedures.

lncRNAs as a future therapeutic approach in CRC

In spite of the fact that long-term rate of survival of patients with early-phase CRC has increased dramatically over the past decades, successful therapeutic strategies for patients with advanced CRC remain limited, requiring novel aggressive therapy in addition to surgical intervention.

At present, anticancer strategies that focus on cancer-specific expression of oncogenic lncRNAs as target molecules are under development. For instance, the paternally-imprinted, oncofetal lncRNA H19 has been found to be a promising target for CRC therapy121. BC-819 (also termed DTA-H19) is a novel double-stranded DNA plasmid carrying the A fragment of diphtheria toxin (DTA) gene under the control of H19 promoter122-124. Intra-tumoral or intra-arterial treatment with DTA-H19 leads to overexpression of DTA in the colorectal tumor, causing a significant suppression of subcutaneous or metastatic colorectal tumor growth in vivo121, 125.

Alternatively, genetic therapies may be designed to treat CRC by manipulating the expression of lncRNAs. For instance, an RNA duplex of siRNAs exhibits high knockdown efficacy of many oncogenic lncRNAs in CRC cells. Silencing of lncRNA expression via administration of specific siRNA oligonucleotides targeting the oncogenic lncRNA CRNDE48, HOTAIR39, 111, CCAT246, Pvt1 oncogene (PVT1)58, and uc.73A59 sequences have been confirmed to exhibit significant anticancer effects, such as decreasing invasion or proliferation either in vivo or in vitro. In addition to knockdown of the oncogenic lncRNAs, restoring endogenous tumor suppressive lncRNAs function in patients with CRC may also be beneficial. One example is in the case of the p53-induced lncRNA Loc285194. Delivery of Loc285194 by plasmid expression vector into CRC cells in xenograft models or in the p53 wild-type CRC cell line HCT-116 significantly suppresses tumor cell growth65 Although the nucleic acid-based approach for CRC treatment is in the development phase, the lncRNA-based therapeutics for CRC still has several difficulties to overcome. For example, the challenge of achieving a better delivery system with efficacy, safety and specificity remains to be solved.

To implement these experiment-based research advances into medical practice, future robust and rigorous investigations into the therapeutic potential of lncRNAs are required.

Conclusions

To date, we and others have taken important initial steps on the road to figure out the biological functions of lncRNAs in CRC. Given the exponential growth of newly discovered cancer-associated lncRNAs, future studies must specifically address the super-secondary structure (motifs), secondary structure, or tertiary structure of lncRNAs. As we move forward, we will refine our global view of the gene regulatory network involving lncRNAs and ultimately establish more effective measures or strategies for the field of lncRNA as a potential biomarker for clinical diagnosis, prognosis and therapeutics in patients with CRC. However, further study employing a greater number of clinical samples will be required to confirm the sensitivity and specificity for these lncRNA as clinical biomarkers in CRC patients.

Table 1

The lncRNA bioinformatics resources.

Bioinformatics Resources	Web Link	Reference
NONCODE v4	http://www.noncode.org	24
H-InvDB rel 8.3	http://www.h-invitational.jp	25
LncRNADisease	http://cmbi.bjmu.edu.cn/lncrnadisease	26
LincSNP	http://bioinfo.hrbmu.edu.cn/LincSNP.	27
Rfam 11.0	http://rfam.janelia.org	28
Human Body Map lincRNAs	http://www.broadinstitute.org/genome_bio/human_lincrnas/	29
lncRNAdb	http://www.lncrnadb.org/	30
ncFANs	http://www.ebiomed.org/ncFANs/	31
Noncoder	http://noncoder.mpi-bn.mpg.de	32
NRED	http://nred.matticklab.com/cgi-bin/ncrnadb.pl/	33
ChIPBase	http://deepbase.sysu.edu.cn/chipbase/	34
LNCipedia 2.0	http://www.lncipedia.org/)	35
DIANA-LncBase	http://www.microrna.gr/LncBase/	36
iSeeRNA	http://www.myogenesisdb.org/iSeeRNA/	37

Table 2

LncRNAs associated with CRC.

LncRNAs and references	Cytoband	Expression level	Types	Mode of action	Potential clinical significance
CCAL44	3q29	Elevated	Oncogenic	Activates Wnt signaling by suppressing AP-2α	Prognosis and therapeutics
CCAT138, 42, 45	8q24.21	Elevated	Oncogenic	N/A	Screening, diagnosis, and prognosis
CCAT1-L41	8q24.21	Elevated	Oncogenic	Regulates chromatin interactions at the MYC locus	N/A
CCAT246	8q24.21	Elevated	Oncogenic	Up-regulates and Responds to Wnt signaling; underlies metastatic progression and chromosomal instability	Screening, diagnosis, and prognosis
CRNDE47, 48	16q12.2	Elevated	Oncogenic	Responds to EGFR signaling; regulates genes involved in central metabolism	Diagnosis
E2F4 antisense49	16q21-22	Elevated	Oncogenic	Activated by Wnt signaling; contributes to carcinogenesis by reducing the level of the E2F4 cell cycle repressor	NA
HOTAIR39, 50	12q13.13	Elevated	Oncogenic	Correlates with PRC2 function; participates in EMT	Prognosis
HULC51	6p24.3	Elevated	Oncogenic	N/A	Diagnosis
KCNQ1OT1/LIT152, 53	11p15.5	LOI	Oncogenic	Correlates with epigenetic status at the KvDMR1	Diagnosis
lncRNA-ATB54	14q11.2	Elevated	Oncogenic	Activated by TGF-β; promotes EMT	Prognosis and therapeutics
MALAT155	11q13.1	Elevated	Oncogenic	Promotes EMT by regulating ZEB1, ZEB2 and Slug expression; activates Wnt signaling	Diagnosis and prognosis
ncNRFR56	1p13.2	Elevated	Oncogenic	Inhibits the oncosuppressive function of let-7 by binding to target mRNAs	Diagnosis
PCAT157	8q24.21	Elevated	Oncogenic	N/A	Prognosis
PVT158	8q24	Elevated	Oncogenic	Generates antiapoptotic activity via TGF-β signaling	Prognosis
uc.73A59	2q22.3	Elevated	Oncogenic	Reduces apoptosis	Diagnosis, prognosis, and therapeutics
CUDR60	19p13.12	Elevated	Oncogenic	N/A	N/A
BANCR61	9q21	Decreased	Oncosuppressive	Suppresses CRC proliferation via targeting p21	N/A
CAHM62	6q26	Decreased	Oncosuppressive	N/A	Screening and diagnosis
lncRNA-LET63	15q24.1	Decreased	Oncosuppressive	Repressed by histone deacetylase 3; contributes to hypoxia-mediated metastasis	NA
lincRNA-p2164	6p21.2	Decreased	Oncosuppressive	Acts with hnRNP-K as a transcriptional coactivator of p53	Prognosis
Loc28519465, 66	3q13.31	Decreased	Oncosuppressive	Induced by p53; inhibits tumor cell growth in part through reducing miR-211 level	Prognosis
MEG367, 68	14q32.2	Decreased	Oncosuppressive	Induces accumulation of TP53 by down-regulating MDM2	NA
ncRAN69	17q25.1	Decreased	Oncosuppressive	N/A	Prognosis and therapeutics
PTENP170	9p13.3	Decreased	Oncosuppressive	Regulates cellular levels of PTEN by binding to the PTEN-targeting miRNAs	therapeutics
BA318C17.171	20p12.1	Decreased	N/A	N/A	N/A
H1940, 72, 73	11p15.5	Elevated or LOI	Oncogenic/oncosuppressive	Decreases RB expression by acting as the precursor of miR-675; controls the number of polyps in the APC murine model of CRC; promotes EMT via functioning as miR-138 and miR-200a sponges	Prognosis and therapeutics
PRNCR168	8q24	N/A	N/A	Contains CRC- related SNPs within its locus	Screening
XIST74, 75	Xq13.2	Elevated	N/A	N/A	N/A