Long non-coding RNA (lncRNA) and epithelial-mesenchymal transition (EMT) in colorectal cancer: a systematic review

ABSTRACT

Background

Colorectal cancer (CRC) is a leading cause of cancer-related death. Epithelial-mesenchymal transition (EMT) is a major process in tumor metastasis development. This systematic review aims to describe the role of long non-coding RNA (lncRNA) in EMT in CRC.

Methods

The electronic databases, PubMed, Cochrane, and EMBASE, were searched from January1990 to June 2019 to identify studies examining lncRNA and their role in mediating EMT in CRC. Studies examining clinical specimens and/or in vitro experiments were included.

Results

In 61 identified studies, 54 lncRNAs were increased in CRC compared to normal colorectal epithelium. Increased lncRNA expression was frequently associated with worse survival. Many lncRNAs mediate their effect through competitive endogenous RNA or transcription factor regulation. The ZEB1, 2/E-cadherin, Wnt/β-catenin signaling, and chromatin remodeling pathways are discussed in particular.

Conclusions

lncRNAs are major regulators of EMT and predictor adverse outcome in CRC patients. Future research must focus on delineating lncRNA function prior to potential clinical use.

Introduction

Colorectal cancer (CRC) is the fourth most commonly diagnosed cancer and the third leading cause of cancer-related death in the United States.¹ While the prognosis for patients with localized disease is good, with 5-year overall survival rates of approximately 90%, there is a significant decrease in the survival of patients with regional lymph node disease (70%) and distant metastasis (14%). Although screening for colon and rectal cancer has been shown to improve survival, many patients are still diagnosed with advanced stage disease at initial presentation.²

Over the past 20 years, there has been significant research into the process of tumor metastasis. One of the critical signaling mechanisms in this process is epithelial-to-mesenchymal transition (EMT). EMT is a process whereby cells transition from an epithelial to a mesenchymal phenotype, through a series of changes in gene expression. Since the initial description by May et al, EMT has become a major focus of cancer research.³ Additionally, it is a critical component of normal embryological development, involved in gastrulation, neural crest formation, and heart morphogenesis.⁴

At a macroscopic level, cells shift from a locally proliferative and relatively static cell phenotype to a more motile and invasive cell phenotype, which is essential for cancer metastasis.^5,6 At the cellular level, epithelial-like cells lose cell polarity, cell-to-cell adhesion, and gain a migratory and invasive phenotype leading to tumor metastasis.⁷ Epithelial cells express a number of different cell surface markers, most notably E-cadherin, but also cytokeratins, occludins, and claudins among others.^8,9 These markers are significant factors in cell-cell adhesion and apical-basal cellular polarity. In contrast, mesenchymal cells are characterized by the expression of markers including vimentin, N-cadherin, and fibronectin.⁹ These markers are associated with front-rear polarity and stem cell characteristics.

During cancer development, mesenchymal cells are associated with resistance to apoptosis and possess an ability to invade through tissue.^6,10 The mesenchymal characteristics that result from EMT activation enable tumor cells to undergo a number of the steps in tumor metastasis including; local invasion at the primary tumor site, blood vessel intravastation, extravasation into distant tissues, and enable survival at distant tissues.⁷ Activation of EMT is also associated with chemotherapy drug resistance in in vitro and clinical patient studies.^11,12

EMT is a complex process and is regulated through different signaling mechanisms. These factors include transcription factors, epigenetic modifiers, and non-coding RNAs. Overall, the EMT program is managed by a number of master regulatory transcription factors, namely; snail family transcriptional repressor 1 (SNAI1), TWIST-related protein 1 (TWIST), snail family transcriptional repressor 2 (SLUG), zinc finger E-box-binding homeobox -(ZEB)1, and ZEB2.^13,14 However, these master transcription factors can be regulated by a number of different upstream regulators, which adds a significant layer of complexity.

Epigenetic regulation of EMT can occur through different processes in methylation and acetylation.^11,15 Methylation at CpG islands of the promoter region of genes involved in EMT repression has previously been demonstrated.¹⁶ Chromatin modifiers such as histone deacteylase 1/2 (HDAC1/2) and lysine-specific histone demethylase 1A (LSD1) have been implicated in EMT and are hypothesized to control the conversion between different intermediate EMT states.^17,18 Treatment with a demethylating agent, in vivo, induces an epithelial-like phenotype in both naïve and chemoresistant cell lines.^19,20

The field of microRNA, and more recently long non-coding RNA, represent a stimulating avenue of research for EMT. microRNAs (miR) have been clearly defined as regulators of EMT, the miR-200 family being the most notable.¹¹ The miR-200 family mediates EMT by decreasing the expression of ZEB1 and ZEB2 via post-transcriptional repression. In addition, ZEB1 and ZEB2 act on the promoter region of the miR-200 family to transcriptionally repress miR-200 family expression. Decreased expression of the miR-200 family leads to a mesenchymal phenotype, thereby promoting EMT.²¹ TGF-β is a major regulator of EMT, inducing ZEB1 and SNAIL1 expression, which is also targeted by the miR-200 family to repress EMT.^21,22 The miR-200 family has also been shown to induce proliferation through different signaling pathways such as the miR-200/RECK/SKP2, CDKN1B axis, and the miR-200/RASSF2/KRAS/ERK1,2 axis.^23,24

lncRNAs are a recently described class of molecules that have roles in normal cellular function and tumorigenesis. They do not code for protein and are defined as being longer than 200 base pairs.²⁵ Similar to messenger RNA, lncRNA are transcribed by RNA polymerase II, but can mediate their action through a number of different mechanisms; competitive endogenous RNA (ceRNA), enhancer, scaffold, signal, and guide function (Figure 1).^26,27 Many lncRNAs are differentially expressed during EMT²⁸ and are involved in EMT activation^29-31 and repression.^25,32,33

Figure 1.

Illustration of the different mechanisms through which lncRNAs can signal. (a) Competitive endogenous RNA function (ceRNA): the lncRNA acts as a functional decoy or “sponge” for other molecules, which prevents the other molecules from executing their function, such as in the case of lncRNA-miRNA sponging; (b) Enhancer function: the lncRNA can act as an enhancer or transcription factor-like molecule in cells to promote gene expression; (c) Guide function: the lncRNA can recruit proteins to a nuclear site,; for instance, to assist in chromatin remodeling; (d) Scaffold function: the lncRNA can act to bring proteins spatially close to each other to aid in ribonuclear protein formation; (e) Signal function: lncRNA acts as a signal for a molecule such as a transcription factor to promote or repress gene expression.

lncRNAs are critical mediators of cell signaling and play a major role in tumor signaling and metastasis. However, the expression of lncRNAs can vary widely across tumor types, suggesting differing roles of lncRNAs in different cancers.³⁴ While a number of reviews have demonstrated the role of lncRNAs in colorectal cancer and in other cancer types, the specific role of lncRNAs in EMT has not been described. Therefore, the aim of this systematic review is to describe in vitro and clinical studies that specifically examine the roles of lncRNAs in different EMT signaling pathways in colorectal cancer. This study also aims to identify current gaps in the literature and to identify future avenues of research and potential therapeutic targets.

Methods

This study was conducted using the “Preferred Reporting for Systematic Reviews and Meta-analysis” guidelines.^35,36 The study methodology was registered with the PROSPERO database (Reference: 159776).

Study selection

The electronic PubMed (NCBI, Bethesda, MD), Cochrane Central Register of Controlled Trials (CENTRAL) (John Wiley & Sons, Hoboken, NJ), and EMBASE (Elsevier, Amsterdam, The Netherlands) databases were systematically searched for all studies relating to lncRNA and their role in EMT in colorectal adenocarcinoma. No restrictions were placed upon the language. The search was restricted from January 1st 1990 to June 31st 2019. The following search terms were used: “colorectal cancer,” “colorectal adenocarcinoma,” “colon cancer,” “rectal cancer,” “colon adenocarcinoma,” “rectal adenocarcinoma,” “lncRNA,” “long non-coding RNA,” “Epithelial-mesenchymal transition,” and “EMT.” The Boolean search terms “AND” and “OR” were used to define the search. Searches from all databases were collated and duplicate references removed.

Inclusion & exclusion criteria

Studies were included if they investigated the molecular signaling role of lncRNAs in EMT in colorectal cancer and/or investigated tumoral expression of lncRNAs in human clinical samples along with including related survival outcomes of patients with colorectal cancer. Studies were excluded if they did not specifically pertain to colorectal cancer, were focused on microRNA signaling, or were review papers.

Data extraction & outcomes

Two reviewers (SO’B and CB) extracted data from the selected studies. Titles and abstracts were screened for relevance by these two reviewers, and in the case of uncertain relevance to the review, the senior author reviewed the paper. Other significant articles from the references of the identified papers were included. The following data were extracted from each of the selected studies: authors, year of publication, journal of publication, main study hypothesis, use of human tumor tissue, colorectal cancer cell line selection and use, experimental techniques used, main study findings, and study limitations.

Results

One hundred and forty-nine studies were identified from the queried databases. Additional studies were manually identified from the references. The search process is illustrated in Figure S1. Ninety-five records remained after removal of duplicates. Thirty-four articles were excluded [reports on cancers other than CRC (n = 14), articles focusing on microRNA (n = 7), and review articles (n = 13)], leaving 61 studies for inclusion in the final analysis.^37–97

In vitro studies

A composite figure of the lncRNAs and their associated individual signaling pathways is shown in Figure S2 and lncRNAs in which the signaling pathway is not fully explored are described in Figure S3.

Thirty-one distinct lncRNAs were found to regulate EMT through a number of different signaling cascades, some of the most common including the ZEB1, ZEB2/E-cadherin, vimentin pathway, Wnt/β-catenin pathway, JAK/STAT pathway, the mTOR pathway, and the MAPK/ERK pathway (Table 1).^37–91

Table 1.

In vitro data from selected studies.

			Expression and Functional changes
Author	lncRNA	Mechanism of regulation	Increased Proliferation	Increased Invasion, Migration	Increased tumor growth In vivo
ZEB1, 2/E-cadherin, vimentin
Chen37	XIST	XIST/miR-200b/ZEB1	+	+	N/A
Chen38	H19	H19/?/E-cadherin, Vimentin, Snail	+	+	N/A
Liang39	H19	H19/miR-200a, miR-138/ZEB1, Vimentin	N/A	+	+
Rokavec92	H19	H19/miR-138, miR-200a/Vimentin, ZEB1/2	N/A	+	N/A
Rigoutsos41	N-BLR	N-BLR/miR-141, miR-200 c/ZEB1	+	+	N/A
Chen40	UICLM	UICLM/miR-215/ZEB2	+	+	N/A
Shen42	SPRY4-IT1	SPRY4-IT1/?/E-cadherin	+	+	N/A
Jin43	SPRY4-IT1	SPRY4-IT1/miR-101/?	+	+	N/A
Wang44	SNHG6	SNHG6/miR101/ZEB1	+	+	N/A
Yang45	HCP5	HCP5/miR139/ZEB1	+	+	N/A
Wang46	B3GALT5-AS1	B3GALT5-AS1/miR203/ZEB1, SNAIL	↓B3GALT5-AS1 → ↑ proliferation	↓B3GALT5-AS1 → ↑migration, invasion	↓B3GALT5-AS1 → ↓liver metastasis
Tao48	CHRF	CHRF/miR-489/TWIST-1	N/A	+	N/A
Jiang47	SNHG15	SNHG15/Slug	N/A	+	+
Wnt/β-catenin Signaling
Jia49	PlncRNA-1	PlncRNA-1/miR-204/β-catenin	↑PlncRNA-1 → ↑proliferation, ↓apoptosis	↑PlncRNA-1 → ↑migration, invasion	N/A
Ding50	H19	H19/miR-29b-3p/PGRN	+	+	N/A
Sun93	TUG1	TUG1/miR-600/KIAA1199	N/A	+	N/A
Yue51	CYTOR	CYTOR/CK1/β-catenin	N/A	+	N/A
Yuan52	CTD903	CTD903/?/Wnt/β-catenin	N/A	+	N/A
Yu54	SLC04A1-AS1	SLC04A1-AS1/β-catenin/GSKβ	+	+	N/A
Li53	lncTCF7	lncTCF7/?/Wnt	N/A	+	N/A
Chromatin Remodeling and epigenetic modulation
Li55	MALAT1	MALAT1/miR-218/EZH2	N/A	+	N/A
Xiong56	MALAT1	Oxymatrine/MALAT1/?	N/A	+	N/A
Liu57	XIST	XIST/miR-137/EZH2	N/A	+	N/A
Li58	HOXD-AS1	HOXD-AS1/miR-217/AEG-1, EZH2	+	+	N/A
Lin59	HIF1A-AS2	HIF1A-AS2/miR-129-5p/DNMT3A	+	+	N/A
Sun60	TUG1	TUG1/?/HDAC 1/2/3	N/A	+	N/A
Wang61	SATB2-AS1	SATB2-AS1/p300/H3K27, H3k9/SATB/HDAC1/SNAIl	+	+	N/A
JAK-STAT3 signaling
Wu62	BC200	BC200/STAT3/cyclins, MMP	+	+	N/A
Xue94	AB073614	AB073614/STAT3/?	+	+	N/A
mTOR signaling
Jahangiri95	UCA1	UCA1/mir-143/mTOR	+	+	N/A
Chen63	ZFAS1	Sp1/ZFAS1/miR-150/VEGFA/VEGFR/akt/mTOR	+	+	N/A
MAPK/ERK signaling
Li64	H19	H19/miR-194/FoxM1	N/A	+	N/A
Shan65	SNHG7	SNHG7/miR-216b/GALNT1	+	+	N/A
Guo66	BANCR	BANCR/?/MEK/ERK	+	+	+
Wang67	NNT-AS1	NNT-AS1/?/MAPK/ERK	+	+	+
Li68	SLC25A25-AS1	SLC25A25-AS1/?/ERK, p38	↓SLC25A25-AS1 → ↑proliferation, chemoresistance	N/A	N/A
TGF-β signaling
Takahashi69	PVT-1	PVT-1/?/TGF-β	+	+	N/A
Kong70	LINC01133	TGF-β/LINC01133/SRSF6	N/A	↓LINC01133 → ↑invasion, migration	N/A
Other signaling pathways
Neuropilin 2 signaling
Liu72	XIST	XIST/miR486/NRP2	+	+	N/A
HIF1-α signaling
Zhang73	CPS1-IT1	CPS1-IT1/HIF1-α/LC-1,LC-2	+	+	N/A
Zhang74	CPS1-IT1	CPS1-IT1/HIF1-α/LC-1,LC-2	N/A	N/A	↑CPS1-IT1 → ↓tumor size
OCT4 signaling
Han75	CRCMSL	CRCMSL/HMGB2/OCT4	↑CRCMSL→ ↓proliferation	↑CRCMSL → ↓migration, invasion	↑CRCMSL → ↓tumor size
Caspase signaling
Yue76	lncRNA-ATB	lncRNA-ATB/?/E-cadherin, Caspase 3, 9	N/A	+	N/A
Sun77	LINC00959	Linc00959/?/Caspase 3, 9	+	+	N/A
NF-κB signaling
Wang78	CYTOR	CYTOR/NCL-Sam86/NfKB	N/A	+	N/A
Notch signaling
Yang79	FOXD2-AS1	FOXD2-AS1/?/Notch	+	+	N/A
Mechanisms not as yet explored
Han80	AFAP1-AS1	AFAP1-AS1/?	+	+	N/A
Bo81	AFAP1-AS1	AFAP-AS1/?/E-cadherin, MMP9, vimentin, ZEB1	N/A	+	N/A
Wu83	HOTAIR	HOTAIR/?/E-cadherin, MMP9, vimentin	+	+	N/A
Wu96	HOTAIR	HOTAIR/IGF2BP2/?	+	+	N/A
Dou97	HOTAIR	HOTAIR/?/	+	+	N/A
Tong84	HOXA-AS2	HOXA-AS2/?/E-cadherin, vimentin, N-cadherin	+	+	N/A
Rezanejad85	VIM-AS1	VIM-AS1/?/E-cadherin, Snail, vimentin	↓VIM-AS1 → ↑proliferation, ↓apoptosis	↓VIM-AS1 → ↑migration	N/A
Gu86	URHC	URHC/?/E-cadherin, N-cadherin, vimentin, snail	+	+	N/A
Zhou87	GHET1	GHET1/?/E-cadherin, vimentin	+	+	N/A
Li88	FOXP4-AS1	FOXP4-AS1/?/p15, p21, p27, KLF2	+	N/A	+
Lu89	PANDAR	PANDAR/?/E-cadherin	+	+	N/A
Chen90	FEZF1-AS1	FEZF1-AS1/?/FEZF1	+	+	+
Li91	XLOC_010588	XLOC_010588/?/E-cadherin, vimentin, Slug	N/A	+	N/A

↑- Increased, ↓- Decreased, N/A- not applicable, *- co-cultured with cancer-associated fibroblast contained media.

The majority (14) of the lncRNAs identified in this study mediate their effect through a ceRNA mechanism, by sponging a target miRNA (Figure S2 and Table 1). In contrast, seven lncRNAs mediate their effect via a transcription-factor like mechanism, and one lncRNA acts through a scaffold mechanism to mediate an effect on EMT. Ten of the lncRNAs do not have the mechanism fully delineated and represent a significant area for further research (Figure S3).

Pathway affected

ZEB1, 2/E-cadherin, vimentin signaling

The ZEB1, 2/E-cadherin, vimentin axis is a well characterized signaling pathway critical to the process of EMT. Ten of the lncRNAs were regulators of this pathway (H19, SNHG6, SPRY4-IT1, CHRF, N-BLR, XIST, UICLM, B3GALT5-AS1, SNHG15) (Table 1),^{37–46,50,92} the majority of which (9/10) mediated their effect through a ceRNA mechanism. The other lncRNA mediated its effect through a transcription factor regulation mechanism (SNHG15).⁴⁷ The miR-200 family was the most common miRNA target of the lncRNAs examined (XIST, N-BLR, and H19).^{37–39,41,50,92} As expected, the major transcription factors that regulate EMT were targeted by lncRNAs to regulate the process (ZEB1, ZEB2, TWIST, SNAIL, SLUG). Few studies have examined the interaction between lncRNAs that target the same molecule in a signaling cascade. For example, SNHG6 and SPRY1-IT1 both directly target miR-101, which acts on the transcription factor ZEB1.^43,44 However, it is not clear what effect SNHG6 knockdown would have on the expression and other downstream functions of SPRY1-IT1. These examples indicate that a systems biology approach is needed for exploring the interaction between lncRNAs,

Wnt/β-catenin signaling

The Wnt/β-catenin signaling pathway is another critical mediator of EMT that is regulated by lncRNAs. Seven lncRNAs targeted the Wnt/β-catenin signaling cascade (H19, PlncRNA-1, TUG1, SLCO4A1-AS1, CYTOR, CTD903, lncTCF7).^{49,50,52,53,60,93,98} Five of the lncRNAs converged on β-catenin to mediate their effect, and three of the lncRNAs mediated their effect through a ceRNA mechanism (H19, PlncRNA-1, TUG1).^{49,50,60,93,98} SLCO4A1-AS1 and CYTOR regulated upstream molecules of β-catenin through a transcriptional factor mechanism.^51,54 Both lncTCF7 and CTD903 have been shown to mediate EMT through Wnt/β-catenin signaling, but the exact mechanism of action was not fully examined.^52,53

Interestingly, some identified lncRNAs mediate EMT through a number of different signaling cascades by different mechanisms. H19 is a well-characterized lncRNA that regulates Wnt/β-catenin signaling, ZEB1, 2/E-cadherin, vimentin signaling, and MAPK/ERK signaling through different ceRNA mechanisms.^38,39,50,51 This demonstrates that lncRNAs can have complex roles in regulating a specific cellular characteristic, such as EMT. Although examining single signaling pathways is critical to understanding a lncRNA’s mechanism of action, it is important to examine lncRNAs in a network signaling approach, as it may help identify potential therapeutic targets.

Epigenetic regulation- chromatin remodeling

Epigenetic modification of gene expression through chromatin remodeling, such as histone methylation and acetylation, has been shown to mediate EMT. Four lncRNAs identified in this study regulate EMT by targeting different core molecules associated with methylation, such as EZH2 (enhancer of zeste homolog 2) and DNMT3A (DNA methyltransferase 3 alpha).^55,58,59,61 EZH2 is the enzymatic component of the polycomb repressive complex 2, which is critical in histone methylation. DNMT3A is an enzyme that transfers methyl groups to CpG islands, leading to gene repression. Two lncRNAs regulate EMT by targeting core molecules associated with acetylation, p300, and HDAC1 (histone deactylase 1).^59,60 Four lncRNAs directly target miRNAs as their mechanism of action but, in contrast, SATB2-AS1 has a more complex signaling mechanism. It acts through a scaffold mechanism to recruit p300 and to acetylate both H3K27 and H3K9 at the promoter region of SATB2. In turn, SATB2 leads to the subsequent recruitment of HDAC1 to the promoter of SNAIL, thereby silencing SNAIL and inhibiting EMT.

Other signaling cascades

JAK-STAT3 signaling

Both BC200 and lncRNA AB073614 act through a transcriptional factor mechanism to reduce STAT3 phosphorylation in the JAK/STAT signaling cascade.^62,94 This leads to the modulation of matrix metalloproteinases and induces EMT.

mTOR signaling

The mTOR pathway is a complex signaling family involved in EMT. ZFAS1 and UCA1 induce EMT through a ceRNA mechanism but target different molecules in the pathway.^63,95 Through binding with miR-150, ZFAS1 targets VEGFA and subsequently Akt. Few studies examine upstream regulators of lncRNA, and interestingly, the authors demonstrate an upstream regulator of ZFAS1 through the SP1 transcription factor.⁶³ UCA1 directly binds miR-143, which targets mTOR to induce EMT. However, the authors demonstrate the complexity of the tumor microenvironment and its effect on lncRNA expression by co-culturing the cells with cancer-associated fibroblast-conditioned media.⁹⁵ This suggests that lncRNA expression can be changed through exogenous agents, indicating a potential avenue for research into their therapeutic modulation.

MAPK/ERK signaling

As previously mentioned, H19 acts through a ceRNA mechanism in both Wnt/β-catenin and in ZEB1, 2/E-cadherin, and vimentin signaling. H19 also directly binds miR-194, which targets FoxM1, a downstream target of the MAPK/ERK signaling pathway.⁶⁴ SNHG7 also affects MAPK/ERK signaling by targeting GALNT1, a downstream molecule in the MAPK/ERK signaling pathway.^65,99 Three other lncRNAs (BANCR, NNT-AS1, SLCA25-AS1) mediate their effect on EMT through MAPK/ERK signaling, but further work is needed to identify their direct targets.^66–68

TGF-β signaling

The TGF-β signaling superfamily is a complex mediator of EMT. Both PVT1 and LINC001133 mediate EMT through TGF-β signaling but further work is needed to identify their direct targets.^69,70

Neuropilin 2 signaling

XIST targets neuropilin 2 signaling through a ceRNA mechanism with miR-486.⁷² This study is limited as it does not explore downstream targets of NRP2, marking an area of future research.¹⁰⁰

OCT4 signaling

CRCMSL also functions through a ceRNA-like mechanism, but binds HMGB2 in the cellular cytoplasm and prevents its shuttling to the nucleus. This prevents the interaction between HMGB2 and OCT4 in the nucleus, promoting a mesenchymal phenotype.⁷⁵ This is a different mechanism compared other included studies that demonstrate lncRNAs bind miRNAs, and shows the complexity in studying lncRNAs.

Caspase signaling

Both lncRNA-ATB and LINC00959 mediate an effect on EMT through caspase signaling,^76,77 but the exact signaling mechanism is unclear. However, it suggests an important avenue of investigation as caspase signaling is typically associated with apoptosis.

NOTCH signaling

FOXD2-AS1 mediates EMT through NOTCH signaling, but the exact mechanism requires further investigation.⁷⁹

NF-κB signaling

CYTOR acts through a transcriptional factor mechanism to regulate Wnt/β-catenin signaling; it can also act through a scaffold mechanism by mediating the interaction between NCL and Sam86 in the NF-κB signaling pathway.⁷⁸ In a hypoxia induction model, CPS1-IT1 induces HIF1-α signaling to mediate EMT.^73,74

Function affected

Chemoresistance

Previous studies have demonstrated that a mesenchymal phenotype is associated with chemoresistance in colorectal cancer.^101,102 Only three of the included studies examine lncRNAs regulating EMT in the context of chemoresistance.^41,55,68 N-BLR regulates EMT through miR-200 family signaling, thereby affecting the downstream expression of ZEB1 and ZEB2.⁴¹ N-BLR also regulates chemoresistance through this mechanism. The miR-200 family is an extensively studied regulator of chemoresistance.¹¹ Similarly, MALAT1 is involved in CRC chemoresistance. MALAT1 expression is increased in a chemoresistant cell line model compared to a parental cell line, but this effect on chemoresistance is mediated through epigenetic regulation of EZH2, by directly binding miR-218.⁵⁵ In contrast, SLC25A25-AS1 is decreased in colorectal cancer compared to normal tissue, and knockdown leads to chemoresistance and increased proliferation, through MAPK/ERK signaling.⁶⁸ Many studies do not examine the role of lncRNAs in chemoresistance, and there is significant scope to examine their role as a potential modulator of this process to improve patient outcomes.

Cell proliferation

Thirty-one studies demonstrated that increased lncRNA expression was associated with increased cellular proliferation in vitro.^{37,38,40–45,49,50,54,58,59,61–63,65-67,69,72,73,77,79,80,83,84,86–90,94-97}

In contrast, decreased expression of four lncRNAs was associated with increased cellular proliferation.^46,68,75,85

Migration and invasion

Thirty-nine studies demonstrated increased lncRNA expression was associated with increased cellular migration and invasion in vitro.^{37–45,47-66,69,72,73,76–81,83,84,86,87,89–97} In contrast, decreased expression of four lncRNAs was associated with increased cellular migration and invasion.^46,70,75,85

Tumor growth in vivo

Seven studies demonstrated increased lncRNA expression was associated with increased tumor growth in vivo.^{39,47,65,66,74,88,90} In contrast, decreased expression of two lncRNAs was associated with increased tumor growth in vivo.^46,75

Human studies

Fifty-three lncRNAs were increased in expression in cancer tissue, while 8 lncRNAs were decreased in expression when compared to normal adjacent colorectal epithelium (Table 2). Interestingly, a majority of studies examine lncRNAs that are increased in expression compared to normal colon epithelium, suggesting their role as biomarkers, but also underscoring their carcinogenic function. As is evident in Table 2, for the majority of presented lncRNA, increased expression is associated with an increase in tumor stage (higher incidence of metastatic disease), supporting their role in EMT.

Table 2.

Clinical data from the selected studies.

Author	lncRNA	Number of Patients (controls)	Tumor stage		lncRNA expression in cancer
			I/II	III/IV	vs. normal colon epithelium	Association with clinico-pathologic variables
ZEB1, 2/E-cadherin, vimentin
Chen37	XIST	115 (115)	47	68	↑	↑Stage
Chen38	H19	96 (96)	54	42	↑	↑Stage
Liang39	H19	30 (30)	N/A	N/A	↑	-
Chen40	UICLM	137 (122)	42	80	↑	↑Stage
Rigoutsos41	N-BLR	462 (28)	226	236	↑	-
Shen42	SPRY4-IT1	96 (96)	N/A	N/A	↑	-
Jin43	SPRY4-IT1	88 (88)	N/A	N/A	↑	-
Wang44	SNHG6	77(77)	N/A	N/A	↑	-
Yang45	HCP5	135(135)	64	71	↑	↑Stage
Wang46	B3GALT5-AS1	64(64)	25	39	↓	↓Stage
Jiang47	SNHG15	329 (41)	N/A	N/A	↑	-
Tao48	CHRF	80 (80)	N/A	N/A	↑	-
Wnt/β-catenin Signaling
Jia49	PlncRNA-1	50 (50)	18	32	↑	↑Stage
Ding50	H19	185 (185)	87	98	↑	↑ poor differentiation ↑ T3 + T4 stage ↑ distant metastasis
Yue51	CYTOR	100 (100)	32	68	-	↑Stage
Yuan52	CTD903	115 (115)	62	53	↑	↑ colon cancer vs rectal cancer
Li53	TCF7	58 (58)	27	31	↑	-
Yu54	SLC04A1-AS1	50(15)	20	30	↑	↑Stage
Chromatin Remodeling and epigenetic modulation
Li55	MALAT1	68 (68) **48 (0) ‡ 46 (0)	21	47	↑	↑Stage
Xiong56	MALAT1	58 (58)	25	33	↑	-
Liu57	XIST	20 (20)	N/A	N/A	↑	-
Li58	HOXD-AS1	136 (136)	42	94	↑	↑Stage
Lin59	HIF1A-AS2	92 (92)	52	40	-	↑Stage
Sun60	TUG1	120 (120)	69	51	↑	↑ tumor grade ↑ tumor depth ↑ liver metastasis
Wang61	SATB2-AS1	216(40)	123	53	↓	↑ Stage
JAK-STAT3 signaling
Wu62	BC200	82 (82)	N/A	N/A	↑	↑Stage
mTOR signaling
Chen63	ZFAS1	112 (112)	65	47	↑	↑Stage
MAPK/ERK signaling
Li64	H19	214 (214)	45	169	↑	↑Stage
Shan65	SNHG7	48 (48)	24	24	↑	↑Stage
Guo66	BANCR	60 (60)	28	32	↑	↑Stage
Wang67	NNT-AS1	70 (70)	36	34	↑	↑ tumor stage ↑ vessel invasion ↑ lymph node-metastasis ↑ tumor grade
Li68	SLC25A25-AS1	585 (19), 30 (30)	16	14	↓	-
TGF-β signaling
Takahashi69	PVT1	312 CRC samples -Set 1: 148 (0)† -Set 2: 164 (164)	81	83	↑	↑ Stage ↑ lymph node metastasis ↑ venous invasion ↑ expression
Kong70	LINC01133	219 (219)	N/A	N/A	↓	-
Eide71	MIR31HG	1097(41)	615	482	↑	-
Other signaling pathways
Neuropilin 2 signaling
Liu72	XIST	317 (317)	168	149	↑	-
HIF1-α signaling
Zhang73	CPS1-IT1	24 (0)	N/A	N/A	↓	-
Zhang74	CPS1-IT1	24 (24)	N/A	N/A	↓	-
OCT4 signaling
Han75	CRCMSL	108 (20)	64	44	↓	-
Caspase signaling
Yue76	LncRNA-ATB	102 (60)	25	35	↑	↑ Stage
Sun77	LINC00959	87 (87)	38	49	↓	↑ Stage
NF-κB signaling
Wang78	CYTOR	138 (138)	N/A	N/A	↑	-
Notch signaling
Yang79	FOXD2-AS1	45 (45)	NR	NR	↑	-
Mechanisms not as yet explored
Han80	AFAP1-AS1	15 (15)	N/A	N/A	↑	-
Bo81	AFAP1-AS1	1061 (97)	N/A	N/A	↑	↑ Stage
Ye82	CCAT1	37 (37)	22	15	↑	↑ Stage
Wu83	HOTAIR	156 (120)	57	63	↑	↑ Stage
Tong84	HOXA-AS2	60 (60)	N/A	N/A	↑	-
Rezanejad85	VIM-AS1	35 (35)	22	13	↑	↑ Stage
Gu86	URHC	77 (77)	31	46	↑	↑ Stage ↑ tumor size ↑ lymph node metastasis
Zhou87	GHET1	20 (20)	N/A	N/A	↑	-
Li88	FOXP4-AS1	48 (48), 174 (56)**	24	24	↑	↑ Stage
Lu89	PANDAR	124 (124)	46	78	↑	↑ Stage
Chen90	FEZF1-AS1	34 (34) 153 (56) FFPE	92	61	↑	↑ Stage
Li91	XLOC_010588	111 (70)	71	33	↑	↑ Stage ↑ lymph node metastasis

↑- Increased, ↓- Decreased, ** Oxaliplatin- based chemotherapy, ‡ FOLFOX chemotherapy, ** microarray data set, † laser capture microdissection. N/A- not applicable.

Association with overall survival and recurrence-free survival

Differential expression of 35 lncRNAs was associated with worse overall survival in clinical patient samples (Figure 2). This is in agreement with the process of EMT being associated with tumor metastasis. For 29 of these lncRNAs, increased expression was associated with worse overall survival, while decreased expression of six lncRNAs was associated with worse overall survival. Only 14 studies examined for differences in recurrence-free survival (Figure 3). Thirteen studies demonstrated that increased lncRNA expression was associated with worse recurrence-free survival, with a single study demonstrating that decreased lncRNA expression was associated with worse recurrence-free survival. The clinical association of differential lncRNA expression supports the hypothesis that these lncRNAs have a major role in tumor signaling and in patient prognosis. However, the absence of this association does not rule out that a lncRNA has a role in tumor signaling. In fact, many of these studies demonstrate an association between lncRNA expression and clinical tumor stage or other prognostic indicators (Table 2). This suggests that a given lncRNA may have a role in a specific facet of EMT, but when examined alone, may not have an association with adverse clinical outcomes.

Figure 2.

Differential expression of lncRNAs associated with worse overall survival.

The lncRNAs are organized by signaling pathway. Increased expression of the lncRNAs in the green box is associated with worse overall survival. Decreased expression of the lncRNAs in the red box is associated with worse overall survival. N/A- Not available.

Figure 3.

Differential expression of lncRNAs associated with worse recurrence-free survival.

The lncRNAs are organized by signaling pathway. Increased expression of the lncRNAs in the green box is associated with worse overall survival. Decreased expression of the lncRNAs in the red box is associated with worse overall survival. N/A- Not available.

Discussion

The aim of this systematic review was to describe the role of lncRNAs in the process of EMT in CRC and to discuss their mechanism of action as well as their interaction in particular signaling pathways. We also describe a number of areas of potential research regarding lncRNAs.

The majority of the studies identified lncRNAs, which are upregulated in CRC. These lncRNA were identified from screening sequencing data sets or had previously been associated with tumorigenesis and EMT in different types of cancer. Cancer cells, both in vivo and in vitro, were shown to phenotypically change their behavior following knock-down of these lncRNAs. Several investigators identified a mechanism partly mediated by a lncRNA-miRNA-mRNA mechanism or a competitive endogenous RNA-type mechanism. Additionally, lncRNAs were able to directly affect tumor-signaling pathways by interacting with key-signaling molecules. Finally, while knock down of various lncRNAs in an in vitro CRC model will lead to phenotypic changes, the responsible molecular mechanism has not been explored in a number of studies and provides a large area for further research.

Differential expression of lncRNAs within the tumor microenvironment is another interesting area of lncRNA investigation. The majority of studies identified in this review utilize whole tumor RNA isolation, which lends itself to contamination with stromal and white blood cell RNA. For this reason, investigators run the risk of biasing results.¹⁰³ Previous studies have demonstrated the heterogeneity of gene expression in colon cancer. This is particularly important in the study of EMT, as there is progressive change in gene expression, such as with the miR-200 family.¹⁰³ Laser capture microdissection allows for the selection of cancer cells without stromal or other “contamination.” However, the microenvironment represents an avenue of further investigation with recent studies examining the role of exosome-derived lncRNAs in tumor signaling.¹⁰⁴ For example, Jahangiri et al. demonstrated that co-culturing cancer-associated fibroblasts leads to increased UCA-1 expression, with a subsequent shift to a mesenchymal phenotype.¹⁰⁵

A number of studies did not explore the lncRNA mechanism of action in detail. As lncRNAs have multiple potential mechanisms of action and often act through multiple mechanisms at once, these can be difficult to delineate. Some authors have suggested mechanism discovery pipelines to attempt to refine this process.²⁷ In particular, the development of RNA-RNA binding and RNA-protein binding bioinformatics algorithms may assist in describing potential mechanisms of action.¹⁰⁶ Others have created bioinformatics algorithms to predict lncRNA-miRNA binding, one of the most commonly described mechanisms of action.¹⁰⁷ Combining these bioinformatics resources with exploratory studies may provide novel areas of research. Such a systems biology approach lends itself to the discovery of new lncRNAs and their mechanisms of action.

While the majority of studies included in this review group colon and rectal adenocarcinoma together, there are significant molecular and clinical differences between these two types of large bowel cancers. Right-sided colon adenocarcinoma tumors are more likely to be highly microsatellite instable tumors, whereas left-sided colon adenocarcinomas are more likely to be chromosomal instable tumors.¹⁰⁸ BRAF mutations more commonly occur in colon adenocarcinomas, and rectal adenocarcinomas typically have APC and TP53 mutations.¹⁰⁹ Clinically, rectal cancer has a tendency toward pulmonary metastases, whereas colon cancer has a tendency to develop liver metastases.^110,111 Typical adjuvant chemotherapy strategies for colon cancer include fluorouracil-based therapies, but for high microsatellite instable tumors, there does not appear to be a survival benefit for such treatment.¹¹² As there are such clear differences between colon and rectal cancer, it is important for investigators to report on differences in lncRNA expression between left-sided, right-sided, and rectal cancers, as well as to report on specific details of the cancer as above.

There are some limitations to this study. Although a large number of lncRNAs are associated with adverse clinical outcomes, and are shown to mediate an effect on EMT-related genes, their mechanism is not fully explored. Therefore, the effect of these lncRNAs on EMT is difficult to fully summarize. While EMT is a critical component of tumor metastasis, this study focuses only on EMT and does not go into detail on the effect of lncRNAs on other hallmarks of carcinogenesis. Few studies delineate between lncRNA expression in the tumor microenvironment compared to expression directly from tumor cells themselves. This limits our ability to confirm whether the expression is increased in colon cancer cells themselves. Some of the included studies examining lncRNA expression in clinical samples are small and should be confirmed in larger data sets.

lncRNAs play a major role in tumor signaling and in EMT in CRC. This review describes clinical studies and in vitro mechanisms of the lncRNAs where available. Many lncRNAs, however, have multiple mechanisms of action, and further research is needed. Tumoral lncRNA expression may have a prognostic role in the clinical setting, which, therefore, warrants further work and validation in independent clinical samples.

Funding Statement

This work was supported by the John W. Price and Barbara Thruston Atwood Price Trust and a grant from the Mary K. Oxley Foundation. This work was supported by the University of Louisville Cancer Education Program [National Institutes of Health Grant: NIH 2R25CA134283-06A1].

Author contributions

SOB and CB performed the literature search, SOB, CB, JH, CF, KS, MP, and JB extracted and verified the study data, SOB, CB, JH, CF, KS, MP, and JB synthesized study data and created figures and tables. SOB, CF, and SG wrote the paper and all authors were involved in final draft changes. SOB and SG developed the study and were in charge of overall direction and planning.

Disclosure of Potential Conflicts of Interest

Susan Galandiuk receives a stipend as Editor-in-Chief of the journal “Diseases of the Colon and Rectum.” The other authors do not have conflicts of interest.

Supplementary data:

Supplemental material for this article can be accessed here.