Journal of Cancer Research and Clinical Oncology (2020) 146:87–96

https://doi.org/10.1007/s00432-019-03090-z

ORIGINAL ARTICLE – CANCER RESEARCH

Candidate lncRNA–miRNA–mRNA network in predicting

hepatocarcinogenesis with cirrhosis: an integrated bioinformatics
analysis
Rui Zhang1,2,3 · Ying‑yi Jiang1,2,3 · Kun Xiao4 · Xiao‑quan Huang1,2,3 · Jian Wang1,3 · Shi‑yao Chen1,2,3,5

Received: 21 July 2019 / Accepted: 15 November 2019 / Published online: 22 November 2019
© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract
Purpose This study aimed to explore the potential competing endogenous RNA (ceRNA) network in forecasting HCC
development in patients with cirrhosis through a comprehensive bioinformatic analysis.
Methods Data mining from GEO and TCGA databases was employed to dig a spectrum of differentially expressed mRNA,
lncRNA and miRNA profiles. Their expression was confirmed by RT-PCR in matched HCC cohorts (n = 6/group). The
ceRNA network was constructed by co-expression analysis. Their reciprocal regulations and their roles in epithelial-to-
mesenchymal transition (EMT) process were validated by gain- and loss-of-function experiments at the cellular level.
Kaplan–Meier method was applied to reveal prognostic values.
Results By intersecting differentially expressed genes (DEGs) in GEO and TCGA data sets and Pearson correlation analy-
sis, 20 mRNAs, 24 miRNAs and 41 lncRNAs were identified. Of these, FOXD2-AS1, BLVRA and CYTH2 were markedly
upregulated in HCC tissues and HCC cells with high metastatic potential (MHCC97H) compared with their adjacent nor-
mal/cirrhotic tissues and L02 and MHCC97L cells. However, dysregulated miR-139-5p exhibited the opposite expression
pattern. Using miRanda algorithms, FOXD2-AS1, BLVRA and CYTH2 showed potential binding sites for miR-139-5p.
FOXD2-AS1 knockdown induced a marked increase in miR-139-5p and EMT inhibition. The loss of miR-139-5p led to an
increase in BLVRA and CYTH2 expression and EMT process. Conversely, miR-139-5p overexpression suppressed BLVRA
and CYTH expression and EMT process. FOXD2-AS1, miR-139-5p, BLVRA and CYTH2 highly correlated with prognosis
in patients with HCC.
Conclusion FOXD2-AS1/miR-139-5p/BLVRA or CYTH2 axis might be the underlying molecular mechanism that dissects
HCC development caused by cirrhosis.

Keywords Bioinformatics · Cirrhosis · Hepatocellular carcinoma · Differentially expressed genes · Long non-coding
RNAs · miRNAs

Rui Zhang, Ying-yi Jiang and Kun Xiao contributed equally to the
Abbreviations
work. ceRNA Competing endogenous RNA
EMT Epithelial-to-mesenchymal transition
Electronic supplementary material The online version of this DEG Differentially expressed genes
article (https​://doi.org/10.1007/s0043​2-019-03090​-z) contains
supplementary material, which is available to authorized users.

4
* Shi‑yao Chen Department of Liver Surgery, Key Laboratory
chen.shiyao@zs‑hospital.sh.cn of Carcinogenesis and Cancer Invasion of Ministry
of Education, Liver Cancer Institute, Zhongshan Hospital,
1
Department of Gastroenterology and Hepatology, Fudan University, 180 Fenglin Road, Shanghai 200032,
Zhongshan Hospital, Fudan University, 180 Fenglin Road, China
Shanghai 200032, China 5
Endoscopy Center and Endoscopy Research Institute,
2
Center of Evidence‑Based Medicine, Fudan University, 180 Zhongshan Hospital, Fudan University, 180 Fenglin Road,
Fenglin Road, Shanghai 200032, China Shanghai 200032, China
3
Shanghai Institute of Liver Disease, 180 Fenglin Road,
Shanghai 200032, China

13
Vol.:(0123456789)
88
HCC Hepatocellular carcinoma
ncRNA Non-coding RNAs
GEO Gene Expression Omnibus
TCGA​ The Cancer Genome Atlas
GO Gene Ontology
KEGG Kyoto Encyclopedia of Genes and Genomae
Background
Approximately 80% of patients with hepatocellular carci-
noma (HCC) patients have liver cirrhosis (European Asso-
ciation for the Study of the Liver, European Organisation for
Research and Treatment of Cancer 2012; Hucke et al. 2011;
Fattovich et al. 2004). Cirrhosis often contributes to hepatic
dysfunction and is considered as a premalignant state, which
increases the HCC risk and development. Recently, the cir-
rhotic microenvironment has been increasingly investigated
for its initiation and its relevance to HCC development (Fat-
tovich et al. 2004; El-Serag and Rudolph 2007; Schrader
et al. 2011). However, the underlying molecular mechanisms
remain to be fully deciphered.
In addition to translated functional RNAs, non-coding
RNAs (ncRNAs) fulfilling essential housekeeping functions
affect mRNA translation and are also widely recognized in
multiple biological processes. Based on the transcript size,
ncRNAs have two major classes: small ncRNAs (miRNAs)
and long ncRNAs (lncRNAs) (Sana et al. 2012). Micro-
RNA (miRNA) is an RNA molecule that negatively regu-
lates target genes via binding to the 3′ untranslated regions
(UTRs) of mRNA transcripts (Carleton et al. 2007). With
the length longer than 200 nucleotides, lncRNAs prefer-
entially expressed in specific tissues represent mRNA-like
transcripts (Zhang et al. 2014). It has been proposed that
dysregulated RNA transcripts and ncRNAs were well char-
acterized in cirrhosis and HCC samples (Sato et al. 2016;
Wu et al. 2017; Lendvai et al. 2019; Fu et al. 2016). Fur-
thermore, competitive endogenous RNA (ceRNA) network
[lncRNA sharing miRNA responsive elements (MREs) with
mRNA] was also involved in the both pathogenesis (Sal-
mena et al. 2011; Huang et al. 2018; Han et al. 2018). How-
ever, the potential involvement of the ceRNA network in
stimulating the transition from cirrhosis to HCC remains to
be established. Exploration of specific ceRNA elements that
predicts HCC transformation from cirrhosis might facilitate
more precise early detection, prevention or targeted thera-
peutic strategies.
Access to some public gene databases, such as Gene
Expression Omnibus (GEO) and The Cancer Genome Atlas
(TCGA), has become easier with the globalization of pop-
ulation-based cancer registries and high-throughput plat-
forms. This study aimed to implement a deep bioinformatics
analysis integrating GEO and TCGA databases, illustrate
13
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
aberrantly mRNAs and ncRNAs that composed of ceRNA
paradigm, and clarify their oncogenic roles in predictabili-
ties of the transformation from cirrhosis to the presence of
HCC. Understanding of the potential axis may allow the
delivery of disease-specific diagnostic and prognostic non-
invasive markers.
Materials and methods
Gene expression profile of cirrhosis and HCC
The expression profile data of cirrhosis and cirrhotic HCC
based on GPL571 [HG-U133A_2] Affymetrix Human
Genome U133A 2.0 Array platform were achieved from the
NCBI-GEO database (http://www.ncbi.nlm.nih.gov/geo/)
(GSE14323) (Barrett et al. 2005). A total of 19 normal liver
tissues, 41 cirrhotic tissues and 17 cirrhotic HCC tissues
were obtained.
The data of exon RNAseq [log2 (RPKM + 1)], mature
miRNA RNAseq [log2 (RPM + 1)] and the corresponding
clinical data of patients with HCC were publicly available
at TCGA (https​://genom​e-cance​r.ucsc.edu/). After filtering
out samples with log2 (RPKM + 1) = 0, the expression pro-
files of differential mRNAs, lncRNAs and miRNAs in paired
tumor and non-tumor tissues were determined.
Screening of differentially expressed genes
In the GEO database, the classical Bayesian method pro-
vided by limma package in the R/Bioconductor software
(Smyth GK. Limma: linear models for microarray data. In:
Bioinformatics & computational biology solutions using R
& Bioconductor 2011) (version 3.10.3, http://www.bioco​
nduct​or.org/packa​ges/2.9/bioc/html/limma​.html) was used
to analyze differentially expressed genes (DEGs) in the three
comparisons (cirrhosis vs normal, cirrhotic HCC vs nor-
mal and cirrhotic HCC vs cirrhosis). The adj. P value was
obtained through multiple checking and correcting by the
Benjamini and Hochberg method (Benjamini and Hochberg
1995). DEGs in the three comparisons with the filter crite-
ria of adj. P value < 0.05 and |logFoldChange| > 0.585 may
be candidates in the development of cirrhosis and cirrhotic
HCC. Likewise, differentially expressed mRNAs, lncRNAs
and miRNAs between peritumor and tumor in the TCGA
database were also obtained.
Mfuzz cluster and protein–protein interaction (PPI)
network analysis
In the GEO database, the R package Mfuzz (version 2.34.0,
http://bioco​nduct​or.org/packa​ges/relea​se/bioc/html/Mfuzz​
.html) was used for cluster analysis of DEGs according to
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
the development history of HCC (normal, cirrhosis and cir-
rhotic HCC) (Kumar and Futschik 2007). The optimal clus-
ter number was 14, and the membership threshold was 0.6
calculated using software default parameters. According to
the expression trend of 14 clusters, module genes were gen-
erated. The intersection of module DEGs in GEO and DEGs
in TCGA databases was acquired for subsequent analysis.
In accordance with the expressional trend (from normal
to cirrhosis and cirrhotic HCC), modular genes in the GEO
database were summarized as upregulated genes and down-
regulated genes. Protein–protein interaction (PPI) networks
were mapped to predict the interactions between gene-
encoded proteins using the Search Tool for the Retrieval of
Interacting Genes database (STRING, Version: 10.0, http://
www.strin​g-db.org/) database. A score of 0.4 (medium con-
fidence) was set as a threshold. Next, to access hub genes,
the visualized regulatory network was constructed using the
Cytoscape software (version 3.4.0, http://chian​ti.ucsd.edu/
cytos​cape-3.4.0/) (Shannon et al. 2003). CytoNCA plug-in
(version 2.1.6, http://apps.cytos​cape.org/apps/cyton​ca) was
used to analyze the topological properties of node networks
(Tang et al. 2015). By ranking the network topological prop-
erties of each node, hub proteins were obtained, which were
regarded as important nodes in protein interaction of the
PPI network.
Functional annotation of differentially expressed
genes and module genes
Gene Ontology (GO) annotation (Ashburner et al. 2000)
and Kyoto Encyclopedia of Genes and Genomae (KEGG)
pathway (Kanehisa and Goto 2000) were used for the data-
base for annotation, visualization and integrated discovery
(DAVID) (version 6.8, https​://david​-d.ncifc​r f.gov/). For
DEGs, P value < 0.05 and enrichment count of at least 5
were set as the cutoff criteria for significance. For module
genes, P value < 0.05 and enrichment count of at least 3 were
identified as the threshold to judge significant enrichment
during the development of cirrhotic HCC.
Cell culture and cell transfection
Human hepatocyte cell line (L02) and hepatoma cell lines
(MHCC97H and MHCC97L) that were authenticated by
STR analysis and tested for mycoplasma contamination
were obtained from the Liver Cancer Institute of Zhongshan
Hospital (Fudan University, Shanghai, China). They were
maintained in DMEM (Hyclone, USA) containing 10% FBS
(GIBICO, USA) in a 5% ­CO2 incubator at 37 °C.
Transfection assays with negative control (NC) mimic,
miR-139-5p mimics, miR-139-5p inhibitor, smart silencer
NC, and FOXD2-AS1 smart silencer in MHCC97H at a
final concentration of 100 nM were conducted using the
89
riboFECT CP transfection kits (RiboBio, Guangzhou,
China) according to the manufacturer’s protocol. The experi-
ment was performed in triplicate.
Patients and specimens
HCC specimens from two matched cohorts of six patients
with cirrhotic HCC and six with HCC who underwent surgi-
cal resection in 2019, were obtained from Zhongshan Hospi-
tal of Fudan University (Shanghai, China). Ethical approval
for this study was acquired from the Ethics Committee of
Zhongshan Hospital, and informed consent was obtained
from each patient. Fresh peri-tumors of HCC and cirrhotic
HCC were considered as normal and cirrhotic liver tissues,
respectively.
Quantitative real‑time polymerase chain reaction
(qRT‑PCR)
Total RNA of tissue specimens and cells were extracted
using TRIzol reagent (Ambion, CA, USA). For mRNA and
lncRNA analysis, they were reversely transcribed to cDNA
with a RevertAid First- Strand cDNA synthesis kit (Thermo
Fisher Scientific, USA) and subsequently amplified using
Maxima SYBR Green qPCR Master Mix kit (Thermo Fisher
Scientific, USA). The gene expression was normalized for
GAPDH levels calculated by the 2−ΔΔCt method. Primer
sequences of mRNA were shown in Table S1. The primers
of all lncRNAs were purchased from RiboBio (Guangzhou,
China).
For miRNA analysis, cDNA was synthesized using a
miRcute miRNA First-Strand cDNA Synthesis Kit and
qPCR analyses were performed with an miRcute miRNA
qPCR Detection Kit (Tiangen, Beijing, China). The gene
expression was determined using the equation 2−ΔΔCt where
the Ct value of U6 (RiboBio, Guangzhou, China) was used
for normalization. The primers of all miRNAs were pur-
chased from RiboBio (Guangzhou, China).
Western blotting analysis
Protein were extracted using the RIPA lysis buffer con-
taining PMSF (Beyotime, China) and protease inhibitors
(Roche, USA). After separation using 10% sodium dodecyl
sulfate-polyacrylamide gels, the samples were transferred
to polyvinylidene difluoride membranes (Millipore, MA,
USA). Blocking with 5% non-fat milk at room temperature
for 1 h, the membrane was incubated with primary antibod-
ies against E-cadherin (CST, 1:1000), N-cadherin (CST,
1:1000), BLVRA (Abcam, 1:5000) and CYTH2 (Affinity
Biosciences, 1:1000) at 4 °C overnight. Afterwards, the
membranes were incubated with corresponding secondary
13
90
antibodies at room temperature for 1 h and then detected
using an enhanced chemiluminescence detection kit.
Statistical analysis
The experimental results were presented as mean ± standard
deviation (SD) and statistically analyzed using SPSS (SPSS,
Chicago, USA). The samples from the TCGA database were
stratified into high- or low-expression groups according to
the median of gene expression. Survival curves with log-
rank test and Kaplan–Meier method were drawn for miR-
NAs, lncRNAs and mRNAs. A threshold of P < 0.05 was
considered statistically significant.
Fig. 1  Differentially expressed genes in the GEO database. a The
Venn diagram of three comparisons in liver samples from subjects
with normal liver, cirrhotic liver and cirrhotic HCC. b Heat-map
hierarchical clustering revealed differentially expressed genes in nor-
13
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
Results
DEGs in the GEO database
A total of 12,436 genes were annotated from GEO data-
base. As the first step of our analysis, DEGs were identi-
fied among three groups of liver samples from subjects with
normal liver, cirrhotic liver and cirrhotic HCC. As a conse-
quence, 1710 DEGs (cirrhotic liver vs normal liver), 1520
DEGs (cirrhotic HCC liver vs normal liver) and 69 DEGs
(cirrhotic HCC liver vs cirrhotic liver) were identified. By
combining DEGs in three comparisons, 2007 DEGs were
finally detected (Fig. 1a). A hierarchical-clustering heat map
of DEGs expression profiles was depicted in Fig. 1b. To bet-
ter understand the overall consequences of DEGs in cirrho-
sis and cirrhotic HCC and how these alterations functioned,
the biological process was explored with GO and KEGG
pathway analysis. As shown in Fig. 1c, the most enriched
mal liver, cirrhotic liver and cirrhotic HCC. c Enrichment analysis of
aberrantly expressed genes. The top 10 enriched GO terms and path-
ways analysis is shown
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
GO terms of immune response, inflammatory response and
cell adhesion, as well as significantly enriched pathways of
protein processing in the endoplasmic reticulum, focal adhe-
sion and phagosome were annotated, which were considered
to be closely linked with the progression of cirrhosis and
cirrhotic HCC.
Mfuzz clustering and PPI network construction
analysis for all DEGs in the GEO database
We started exploring the expression patterns of 2007
DEGs in GEO database. Combined, 14 clusters were
aggregated (Fig. 2a). Based on the course of disease
Fig. 2  Mfuzz clustering and PPI network construction analysis for all
differentially expressed genes in the GEO database. a Mfuzz cluster-
ing illustrating the patterns of dynamic changes in DEGs during the
progression of the normal liver to the cirrhotic liver and cirrhotic
91
development from normal to cirrhosis and cirrhotic HCC,
a total of six modules were revealed, consisting of up–up
module (cluster 1), up–smooth module (clusters 2 and 3),
up–down module (clusters 7, 9, 10, and 11), down–down
module (clusters 4, 5, and 12), down–smooth module
(clusters 6 and 8), and down–up module (clusters 13 and
14). To better understand the biological significance of
module DEGs, functional enrichment analysis was con-
ducted to illustrate the functional analysis of the six mod-
ule genes. According to the enriched results, GO biologi-
cal process (e.g., signal transduction, immune response,
and cell adhesion) and KEGG pathway (e.g., metabolic
pathways and focal adhesion) in the four modules genes
HCC. b, c Enrichment analysis showed the most enriched GO terms
and KEGG pathways. d The PPI network of upregulated and down-
regulated genes using the STRING database
13
92
that may be associated with progression of cirrhosis and
cirrhotic HCC were annotated (Fig. 2b, c).
With the purpose of predicting the interaction between
gene-encoded proteins, two PPI networks were constructed
from the six modules, of which one network was detected
from upregulated genes in the up–up, up–smooth and
up–down modules, and the other was constructed from
downregulated genes in the down–down, down–smooth
and down–up modules (Fig. 2d). According to the score of
degree centrality (DC), betweenness centrality (BC) and
closeness centrality (CC), key hub genes ranking in the top-
20 in the two networks were presented in Table S2 and S3.
DEGs in the TCGA database and the acquisition
of key DEGs
In the TCGA database, 404 HCC cohorts (355 HCC tissues
and 39 adjacent normal tissues) including the expression
profiles of 15841 mRNAs, 4317 lncRNAs and 752 miRNAs
were identified and the numbers of differential expressed
miRNAs, mRNAs and lncRNAs were presented in Table 1.
To comprehensively analyze the molecules responsible for
HCC caused by cirrhosis and their role in survival analy-
sis, DEGs in the TCGA database were intersected with six
module genes in the GEO database. As a result, 121 DEGs
were obtained, of which 42 genes were downregulated and
79 genes were upregulated. Of the 42 downregulated genes,
8 were from the up–smooth module and therefore were
excluded. Of the 79 upregulated genes, 14 were from the
down–smooth module and 15 from the down–down module.
Thus, 29 genes were eliminated. Finally, 34 downregulated
and 50 upregulated genes were obtained. The subsequent
analysis was based on these 84 genes (Table S4).
Construction of lncRNA–miRNA–mRNA regulatory
network
Based on our prior results, 84 differentially expressed
mRNAs, 306 differentially expressed lncRNAs and 262 dif-
ferential expressed miRNAs were uncovered, which might
play a role in regulating the tumor-formation effect of
cirrhosis. Next, their reciprocal co-expression relation-
ships and ceRNA network were examined. For this pur-
pose, the Pearson correlation coefficient between 84 key
Table 1  The number of differential expressed miRNA, mRNA and
lncRNA in TCGA database
miRNA mRNA lncRNA
Up 87 4234 255
Down 175 756 51
Total 262 4990 306
13
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
DEGs and differentially expressed miRNAs and lncR-
NAs was calculated. The relationship of miRNA–lncRNA
(r < − 0.5 and P < 0.05), miRNA–mRNA (r < − 0.5 and
P < 0.05) and mRNA–lncRNA (|r| > 0.5 and P < 0.05)
was used as the threshold value. Finally, a total of 312
lncRNA–miRNA–RNA regulatory relationships were
obtained. In this ceRNA network, 85 nodes including 20
mRNAs, 24 miRNAs and 41 lncRNAs were identified, sug-
gesting that they might be master regulators in the develop-
ment of cirrhotic HCC (Fig. 3).
FOXD2‑AS1/miR‑139‑5p/BLVRA or CYTH2 network
contribute to the development of cirrhotic HCC
Next, the potential lncRNA–miRNA–mRNA interaction
pairs were determined using miRanda algorithms. Addition-
ally, the results were further verified in six pairs of HCC
and cirrhotic HCC samples and at the cellular level. Con-
sequently, FOXD2-AS1, BLVRA and CYTH2 that signifi-
cantly increased and miR-139-5p that significantly decreased
in cirrhotic HCC samples and HCC cells with high‐meta-
static potentials were eventually identified as the underlying
elements of ceRNA-mediated regulatory network (Fig. 4a, b,
Table S5 and S6). The RT-PCR and western blotting results
showed that FOXD2-AS1 positively regulated the expression
of BLVRA and CYTH2, but negatively regulated the expres-
sion of miR-139-5p. In addition, miR-139-5p negatively
regulated the expression of BLVRA and CYTH2 (Fig. 4c,
d). The gain- and loss-of-function study was performed in
MHCC97H cells to characterize the role of FOXD2-AS1 and
miR-139-5p in the EMT process. As shown in Fig. 4d, the
depletion of FOXD2-AS1 or overexpression of miR-139-5p
remarkably decreased the level of N-cadherin but increased
the expression of E-cadherin. On the contrary, suppressed
miR-139-5p exerted the opposite effects. The survival anal-
ysis showed that FOXD2-AS1, miR-139-5p, BLVRA and
CYTH2 were significantly correlated with the overall sur-
vival of HCC (Fig. 4e). Collectively, it was proposed that
the FOXD2-AS1/miR-139-5p/BLVRA or CYTH2 network
promoted the cirrhotic HCC and EMT process.
Discussion
Increasing evidence has unveiled that cirrhotic livers char-
acterized by fibrous bands and surrounded by regenerative
nodules correlated with the risk of process of hepatocar-
cinogenesis (Borzio et al. 1995; Donato et al. 2001). Nev-
ertheless, how cirrhosis developed from normal liver and
progressed to cirrhosis and cirrhotic HCC remained still
undefined. Hence, a better understanding of the molecu-
lar alterations and mechanisms of HCC formation was
warranted for preventive therapies. The present study was
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
Fig. 3  Construction of the ceRNA network. Pearson correlation coefficient was calculated between differentially expressed genes and differential
miRNAs and lncRNAs
conducted to investigate potential lncRNA–miRNA–mRNA
regulatory network based on combined databases to eluci-
date the molecular signatures of cirrhosis progression to
HCC.
ceRNAs are transcripts that act as miRNA sponges,
modulating each other at the post-transcriptional level via
competitively binding to shared miRNAs (Qi et al. 2015).
In vitro and in vivo experiments, as well as integrated bio-
informatic analysis, showed that lncRNA-associated ceRNA
network separately modulated the biological processes of
liver fibrosis and HCC (Han et al. 2018; Hanson et al. 2018;
Mo et al. 2018; Fan et al. 2018; Zhang et al. 2018). Although
previous studies described how ceRNA network modulated
molecular diagnostic and prognostic targets in liver fibro-
sis and HCC, an important gap existed in the noninvasive
prediction of the transition from cirrhosis to cirrhotic HCC.
This study was novel in identifying aberrantly expressed
FOXD2-AS1/miR-139-5p/BLVRA or CYTH2 acting as
ceRNA network in initiating cirrhosis and the formation of
cirrhotic HCC. Moreover, this study further validated their
93
reciprocal regulatory relationship and role of FOXD2-AS1/
miR-139-5p/BLVRA or CYTH2 in the EMT process of
HCC. To our knowledge, our work is the first expression
profile that analyzes and validates the ceRNA-mediated
mechanism to signify how cirrhosis is converted into cir-
rhotic HCC.
In the GEO data sets, 2007 DEGs were found in three
comparisons of the cirrhotic HCC liver versus cirrhotic liver,
cirrhotic HCC liver versus normal liver, and cirrhotic liver
versus normal liver. In accordance with the disease devel-
opment from normal to cirrhosis and cirrhotic HCC, 2007
DEGs were fused into six modules. The PPI analysis was of
great importance for interpreting the molecular mechanisms
of the key cellular activities. The upregulated PPI and down-
regulated PPI networks in module genes were constructed to
elucidate the mechanism of modules genes. Subsequently,
top 20 of score in DC, BC and CC were identified as hub
genes in the progression of cirrhosis to cirrhotic HCC. For
example, CD163, FOS, IRF8, ISG15, LCK, NFKBIA,
PTPRC and SP110 were key proteins in the upregulated
13
94
Fig. 4  FOXD2-AS1/miR-139-5p/BLVRA or CYTH2 network con-
tribute to the development of cirrhotic HCC. a, b Expression of
FOXD2-AS1, miR-139-5p, BLVRA and CYTH2 in control, cirrho-
sis and cirrhotic HCC samples (n = 6/group) and at the cellular level
detected by RT-PCR. c Reciprocal regulatory relationship among
FOXD2-AS1, miR-139-5p, BLVRA and CYTH2 by the gain- and
PPI network. ALDH18A1, ALDH1L1, ALPL, HSPA5,
NUBP2, PDIA4, SEC63, VARS and YARS were key pro-
teins in the downregulated PPI network. Most of them (e.g.,
CD163, IRF8 and ISG15) were closely associated with the
13
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
loss-of function study in MHCC97H cells. d Protein expressions of
BLVRA, CYTH2, N-cadherin and E-cadherin in MHCC97H cells
with altered levels of FOXD2-AS1 and miR-139-5p. e FOXD2-AS1,
miR-139-5p, BLVRA and CYTH2 were significantly associated
with the overall survival in patients with HCC, as detected using the
Kaplan–Meier curve and a log-rank test. **P < 0.01; *P < 0.05
protumorigenic function and inflammatory response (Chen
et al. 2019; Jia et al. 2017; Yeung et al. 2018). Next, six mod-
ule DEGs in GEO data sets were intersected with DEGs in
TCGA. Using Pearson correlation analysis, ceRNA network
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
consisting of 20 mRNAs, 24 miRNAs and 41 lncRNAs were
found to be significantly associated with the occurrence of
cirrhotic HCC. After verification in six pairs of HCC and
cirrhotic HCC samples and at the cellular level, FOXD2-
AS1/miR-139-5p/BLVRA or CYTH2 axis was eventually
uncovered. Their mutual binding sites and reciprocal regula-
tory relationships were confirmed by miRanda algorithms
and cell experiments. Furthermore, FOXD2-AS1/miR-
139-5p/BLVRA or CYTH2 also significantly correlated with
the overall survival of HCC and EMT process. Our findings
were also consistent with previous investigations on either
precancerous or cancerous liver conditions (Sharma 2019;
Kubícková et al. 2016; Xu et al. 2019; Hua et al. 2018).
There were several limitations in our study. First, the find-
ings were mainly accessed from public data sets, lacked in-
depth in vitro and in vivo validations. Further elucidation
of molecular alterations in the transition from cirrhosis to
cirrhotic HCC may help in developing more rational, spe-
cific, and effective treatments. Second, whether FOXD2-
AS1 exerted its oncogenic effects through the specific regu-
lation of miR-139-5p and BLVRA/CYTH2 and the direct
regulatory mechanism of BLVRA, CYTH2, FOXD2-AS1
and miR-139-5p needed to be verified through additional
experiments.
In conclusion, the present study employed a comprehen-
sive data mining on multiple genomic platforms to dem-
onstrate FOXD2-AS1/miR-139-5p/BLVRA or CYTH2
by which cirrhosis facilitated HCC formation. Targeting
FOXD2-AS1/miR-139-5p/BLVRA or CYTH2 may be novel
therapeutic strategies against the transition from cirrhosis to
cirrhotic HCC.
Author contributions SYC conceived and designed the study. RZ, YYJ
and KX performed the experiments. RZ, YYJ, KX and JW made statis-
tical analysis. RZ and SYC analyzed the data and wrote the manuscript.
RZ, XQH and SYC revised the manuscript.
Funding This work was supported by the Innovation Fund of Sci-
ence and Technology Commission of Shanghai Municipality (No.
15411950501, 15411950507 and 17140902700).
Compliance with ethical standards
Conflict of interest The authors have declared that no competing inter-
est exists.
References
Ashburner M, Ball CA, Blake JA et al (2000) Gene Ontology: tool for
the unification of biology. Nat Genet 25(1):25–29
Barrett T, Suzek TO, Troup DB et al (2005) NCBI GEO: mining mil-
lions of expression profiles—database and tools. Nucleic Acids
Res 33(suppl 1):D562–D566
95
Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate:
a practical and powerful approach to multiple testing. J R Stat
Soc 57:289–300
Borzio M, Bruno S, Roncalli M et al (1995) Liver cell dysplasia is
a major risk factor for hepatocellular carcinoma in cirrhosis: a
prospective study. Gastroenterology 108(3):812–817
Carleton M, Cleary MA, Linsley PS (2007) MicroRNAs and cell cycle
regulation. Cell Cycle 6:2127–2132
Chen T, Chen J, Zhu Y et al (2019) CD163, a novel therapeutic target,
regulates the proliferation and stemness of glioma cells via casein
kinase 2. Oncogene 38(8):1183–1199
Donato MF, Arosio E, Del Ninno E et al (2001) High rates of hepato-
cellular carcinoma in cirrhotic patients with high liver cell prolif-
erative activity. Hepatology 34(3):523–528
El-Serag HB, Rudolph KL (2007) Hepatocellular carcinoma: epi-
demiology and molecular carcinogenesis. Gastroenterology
132:2557–2576
European Association for the Study of the Liver, European Organisa-
tion for Research and Treatment of Cancer (2012) EASL-EORTC
clinical practice guidelines: management of hepatocellular carci-
noma. J Hepatol 56:908–943
Fan H, Lv P, Mu T et al (2018) LncRNA n335586/miR-924/CKMT1A
axis contributes to cell migration and invasion in hepatocellular
carcinoma cells. Cancer Lett 429:89–99
Fattovich G, Stroffolini T, Zagni I et al (2004) Hepatocellular carci-
noma in cirrhosis: incidence and risk factors. Gastroenterology
127(5, Suppl 1):S35–S50
Fu N, Niu X, Wang Y et al (2016) Role of LncRNA-activated by trans-
forming growth factor beta in the progression of hepatitis C virus-
related liver fibrosis. Discov Med 22(119):29–42
Han X, Hong Y, Zhang K (2018) TUG1 is involved in liver fibrosis and
activation of HSCs by regulating miR-29b. Biochem Biophys Res
Commun 503(3):1394–1400
Hanson A, Wilhelmsen D, DiStefano JK (2018) The role of long non-
coding RNAs (lncRNAs) in the development and progression of
fibrosis associated with nonalcoholic fatty liver disease (NAFLD).
Noncoding RNA 4(3):18
Hua S, Lei L, Deng L et al (2018) miR-139-5p inhibits aerobic gly-
colysis, cell proliferation, migration, and invasion in hepatocel-
lular carcinoma via a reciprocal regulatory interaction with ETS1.
Oncogene 37(12):1624–1636
Huang Y, Xiang B, Liu Y et al (2018) LncRNA CDKN2B-AS1 pro-
motes tumor growth and metastasis of human hepatocellular carci-
noma by targeting let-7c-5p/NAP1L1 axis. Cancer Lett 437:56–66
Hucke F, Sieghart W, Schoniger-Hekele M et al (2011) Clinical char-
acteristics of patients with hepatocellular carcinoma in Austria—
is there a need for a structured screening program? Wien Klin
Wochenschr 123:542–551
Jia Y, Han S, Li J et al (2017) IRF8 is the target of SIRT1 for the inflam-
mation response in macrophages. Innate Immun 23(2):188–195
Kanehisa M, Goto S (2000) KEGG: kyoto encyclopedia of genes and
genomes. Nucleic Acids Res 28(1):27–30
Kubícková KN, Subhanová I, Konícková R et al (2016) Predictive role
BLVRA mRNA expression in hepatocellular cancer. Ann Hepatol
15(6):881–887
Kumar L, Futschik ME (2007) Mfuzz: a software package for soft
clustering of microarray data. Bioinformation 2(1):5–7
Lendvai G, Szekerczés T, Gyöngyösi B et al (2019) MicroRNA expres-
sion in focal nodular hyperplasia in comparison with cirrhosis
and hepatocellular carcinoma. Pathol Oncol Res 25(3):1103–1109
Mo Y, He L, Lai Z et al (2018) LINC01287/miR-298/STAT3 feedback
loop regulates growth and the epithelial-to-mesenchymal tran-
sition phenotype in hepatocellular carcinoma cells. J Exp Clin
Cancer Res 37(1):149
Qi X, Zhang DH, Wu N et al (2015) ceRNA in cancer: possible func-
tions and clinical implications. J Med Genet 52(10):710–718
13
96
Salmena L, Poliseno L, Tay Y et al (2011) A ceRNA hypothesis: the
Rosetta Stone of a hidden RNA language? Cell 146:353–358
Sana J, Faltejskova P, Svoboda M et al (2012) Novel classes of non-
coding RNAs and cancer. J Transl Med 10:103
Sato M, Ikeda H, Uranbileg B et al (2016) Sphingosine kinase-1, S1P
transporter spinster homolog 2 and S1P2 mRNA expressions
are increased in liver with advanced fibrosis in human. Sci Rep
6:32119
Schrader J, Gordon-Walker TT, Aucott RL et al (2011) Matrix stiff-
ness modulates proliferation, chemotherapeutic response,
and dormancy in hepatocellular carcinoma cells. Hepatology
53(4):1192–1205
Shannon P, Markiel A, Ozier O et al (2003) Cytoscape: a software
environment for integrated models of biomolecular interaction
networks. Genome Res 13(11):2498–2504
Sharma MC (2019) Annexin A2 (ANXA2): an emerging biomarker and
potential therapeutic target for aggressive cancers. Int J Cancer
144(9):2074–2081
Smyth GK. Limma: linear models for microarray data. In: Bioinformat-
ics & computational biology solutions using R & Bioconductor
2011:397–420
Tang Y, Li M, Wang J et al (2015) CytoNCA: a cytoscape plugin for
centrality analysis and evaluation of protein interaction networks.
BioSystems 127:67–72
13
Journal of Cancer Research and Clinical Oncology (2020) 146:87–96
Wu MS, Wang CH, Tseng FC et al (2017) Interleukin-17F expression
is elevated in hepatitis C patients with fibrosis and hepatocellular
carcinoma. Infect Agent Cancer 12:42
Xu K, Zhang Z, Qian J et al (2019) LncRNA FOXD2-AS1 plays
an oncogenic role in hepatocellular carcinoma through epi-
genetically silencing CDKN1B (p27) via EZH2. Exp Cell Res
380(2):198–204
Yeung TL, Tsai CC, Leung CS et al (2018) ISG15 promotes ERK1
ISGylation, CD8+ T cell activation and suppresses ovarian cancer
progression. Cancers (Basel) 10(12):464
Zhang K, Shi ZM, Chang YN et al (2014) The ways of action of long
non-coding RNAs in cytoplasm and nucleus. Gene 547:1–9
Zhang Z, Wang S, Liu W (2018) EMT-related long non-coding RNA in
hepatocellular carcinoma: a study with TCGA database. Biochem
Biophys Res Commun 503(3):1530–1536
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.