Gene expression profiling of MYC-driven tumor signatures in porcine liver stem cells by transcriptome sequencing

Abstract

AIM: To identify the genes induced and regulated by the MYC protein in generating tumors from liver stem cells.

METHODS: In this study, we have used an immortal porcine liver stem cell line, PICM-19, to study the role of c-MYC in hepatocarcinogenesis. PICM-19 cells were converted into cancer cells (PICM-19-CSCs) by overexpressing human MYC. To identify MYC-driven differential gene expression, transcriptome sequencing was carried out by RNA sequencing, and genes identified by this method were validated using real-time PCR. In vivo tumorigenicity studies were then conducted by injecting PICM-19-CSCs into the flanks of immunodeficient mice.

RESULTS: Our results showed that MYC-overexpressing PICM-19 stem cells formed tumors in immunodeficient mice demonstrating that a single oncogene was sufficient to convert them into cancer cells (PICM-19-CSCs). By using comparative bioinformatics analyses, we have determined that > 1000 genes were differentially expressed between PICM-19 and PICM-19-CSCs. Gene ontology analysis further showed that the MYC-induced, altered gene expression was primarily associated with various cellular processes, such as metabolism, cell adhesion, growth and proliferation, cell cycle, inflammation and tumorigenesis. Interestingly, six genes expressed by PICM-19 cells (CDO1, C22orf39, DKK2, ENPEP, GPX6, SRPX2) were completely silenced after MYC-induction in PICM-19-CSCs, suggesting that the absence of these genes may be critical for inducing tumorigenesis.

CONCLUSION: MYC-driven genes may serve as promising candidates for the development of hepatocellular carcinoma therapeutics that would not have deleterious effects on other cell types in the liver.

Key Words: Hepatocellular carcinoma; MYC; Stem cells; Gene expression; RNA sequencing

Core tip: It is well-established that cancer stem cells not only drive tumor growth in the liver, in general, but also that the proto-oncogene c-MYC plays a critical role in that process. However, little is known about genes induced and regulated by MYC to generate tumors, and, in particular, those involved in liver stem cells. In this study, we examined the role of MYC protein in hepatocarcinogenesis using an immortal porcine liver stem cell line, PICM-19. Interestingly, MYC-overexpression silenced the expression of six genes in PICM-19 cells (CDO1, C22orf39, DKK2, ENPEP, GPX6, SRPX2), suggesting that they may be critical in inducing tumorigenesis.

INTRODUCTION

Hepatocellular carcinoma (HCC) is the most common form of liver cancer, and is the third leading cause of cancer-related deaths worldwide[1]. It occurs as a consequence of exposure to a variety of risk factors, and appears primarily in liver disease patients with underlying cirrhosis[2]. While surgical resection and whole liver transplantation are the most effective treatment options currently available for the treatment of HCC, recurrence is quite common in patients who have had a resection, and the 5-year survival rate post-surgery is < 50%[3].

Hepatocarcinogenesis is a multistep process involving a number of cellular and molecular factors. Among these, the proto-oncogene c-MYC (herein, referred to as MYC) has been shown to be a critical regulator of malignant transformation during the early stages of liver cancer[4]. Dysregulation of MYC expression correlates with poor prognosis in human cancers, including HCC[5]. Its overexpression, and subsequent induction of its target genes, causes the malignant conversion of preneoplastic liver lesions[4]. Conversely, silencing of MYC results in the inhibition of migration, invasion and proliferation of human liver cancer cells[6]. Therefore, the study of MYC-expressing liver stem cells could be highly relevant to our understanding of the pathogenesis in human HCC. Due to the involvement of the MYC oncoprotein in angiogenesis and various cellular processes such as cell cycle regulation, differentiation, growth, metabolism, adhesion and motility, MYC, and its downstream target genes have become attractive therapeutic aims of cancer research[7].

It is now apparent that the capacity to sustain tumor growth resides in a small population of stem cells [termed “cancer stem cells” (CSCs)] present in most cancers, including HCC[8-10]. CSCs possess greater colony-forming efficiency, higher proliferation potential, and the necessary phenotype of “stemness” to form tumors in animals[5]. To support this, CD133-expressing progenitor cells were shown to drive the recurrence of liver cancer following chemotherapy, suggesting that CSCs are resistant to treatment[11]. A possible mechanism for the enhanced survival of CSCs in tumors is the activation of the Akt (also known as protein kinase B) and B-cell lymphoma 2 (Bcl-2) signaling pathways[11]. Therefore, a critical step towards developing effective cancer treatments is to understand cellular, molecular and biochemical differences between normal stem cells and CSCs.

To decipher the role of MYC and MYC-targeted genes in HCC, we have developed a liver cell model of MYC oncogene transformation based on the overexpression of MYC in a porcine liver stem cell line, PICM-19[12]. The PICM-19 cell line originated from the spontaneous differentiation of cultured pig epiblast tissue and was, therefore, derived from pig embryonic stem cells[13]. The cell line is unique in its ability to differentiate into either of the two cell types that comprise the parenchyma of the developing liver, i.e., fetal hepatocytes or bile duct epithelium, and grows indefinitely without change in stem cell character. In this study, we have MYC-transformed the PICM-19 cell line into CSCs, and studied their transcriptome profiles by RNA sequencing (RNA-seq). We determined that MYC induced the expression of a large number of genes with diverse cellular processes in PICM-19 cells, most of which were not previously implicated in HCC. Our data identified novel therapeutic targets while underscoring the importance of MYC in tumor generation in primary liver stem cells.

MATERIALS AND METHODS

Cell culture

The PICM-19 porcine liver stem cell line was established from a pig embryo’s inner cell mass, i.e., embryonic stem cells, and was cultured as described previously[12,13]. Briefly, cells were grown in 10% DMEM/199 medium and plated as feeder-free cultures using Matrigel (BD Biosciences, San Jose, CA). This growth medium is a 50:50 mixture of DMEM low glucose (Invitrogen, Carlsbad, CA) and Medium 199 (Invitrogen) supplemented with 10% fetal bovine serum, 2-mercaptoethanol and nucleosides. In subsequent steps, cells were fed daily with conditioned medium obtained from γ-irradiated STO feeder cells, a mouse embryonic fibroblast cell line (Cat. #CRL-1503, ATCC, Rockville, MD). PICM-19-CSCs, generated by transfecting the PICM-19 cell line (see below), were also maintained under these growth conditions. All cells were maintained at 37 °C in a 5% CO₂ humidified incubator.

Plasmids

The expression vector pUNO1-hMYC was generated by cloning the human c-MYC open reading frame (ORF) into the multiple cloning site of the plasmid pUNO1-mcs (InvivoGen, San Diego, CA) downstream of a strong elongation factor (EF)-1α/human T-lymphotropic virus (HTLV) hybrid promoter active in most cell types. pUNO1-mcs contains the blasticidin resistance gene driven by a CMV promoter and enhancer in tandem with the bacterial EM7 promoter. This allows the amplification of the plasmid, and, after transfection into mammalian cells, the blasticidin selection of stable transfectants. Another plasmid, pUNO1-MYC-IRES-Luc was also constructed by cloning the firefly luciferase ORF downstream of MYC ORF separated by an internal ribosome entry site (IRES) sequence to maintain expression of both MYC and luciferase (luc) genes under the control of the hybrid promoter in this plasmid. Large scale preparations of both vectors were prepared by transforming them into competent cells of Escherichia coli strain DH5α, and by extracting them using the Qiagen Plasmid Maxi kit (Qiagen, Valencia, CA).

Luciferase assay

PICM-19 cells were successfully transfected with the pUNO1-MYC-IRES-Luc plasmid using the mouse macrophage nucleofection kit (Amaxa Biosystems, Gaithersburg, MD) and the program A-13 on the nucleofector I device (Amaxa). Following nucleofection, cells were plated in 12-well plates and incubated overnight at 37 °C. Growth medium was then replaced with the fresh medium containing 5 μg/mL blasticidin (InvivoGen) to select for positive transfectants. Individual colonies that formed were further grown, and assessed for MYC expression using reverse transcription polymerase chain reaction (RT-PCR) (data not shown). The clone that showed the highest MYC expression was then used in further experiments mentioned below. Next, cells of this clone were plated in 6-well plates, and 24 h post-plating they were resuspended in fresh medium and were treated with the Bright-Glo luciferase assay substrate (Promega, Madison, WI) to measure luciferase activity using the IVIS^® Imaging System (Xenogen Corporation, Alameda, CA).

Western blotting

Western blot analysis of cellular proteins extracted from PICM-19 and PICM-19-CSCs was performed using mouse anti-human c-MYC antibody (Cat. #sc-40, Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Ten μg of total protein from these cells was loaded onto a 10% denaturing SDS-PAGE gel. Following separation, proteins were transferred to a nitrocellulose filter and treated with blocking solution containing 5% milk powder. Primary antibody at a concentration of 1:500 was then added to the filter and incubated for 2 h in a rotating chamber at 4 °C. After several washes in buffer, the secondary anti-mouse IgG antibody (Santa Cruz Biotechnology) was added at a concentration of 1:1000 and the filter incubated for 2 h at RT. The blot was washed and probed with ECL solution to visualize protein bands (Thermo Fischer Scientific, Rockford, IL).

Tumorigenicity assay

Five-week old NOD/SCID mice were purchased from Charles River Breeding Laboratories (Wilmington, MA). Freshly cultured PICM-19 and PICM-19-CSCs were treated with trypsin to detach and harvest them in PBS. After washing twice in PBS, 1 × 10⁶ cells were resuspended in 100 μL PBS, and were injected into the flanks of the immunodeficient mice (n = 3). This study was approved by the University of Minnesota Institutional Animal Care and Use Committee. Animals were observed on a daily basis for signs of declining health and lethargy. Tumor formation was monitored twice weekly by measuring the width and length of tumors. Animals with tumors that grew to a diameter of 1.5 cm, as measured by calipers, were sacrificed. To alleviate pain and stress in animals, carbon dioxide was administered in the fume hood prior to euthanasia. Tumors were then harvested and immediately fixed in 10% formalin for 24 h. These tissue samples were cut in 5 μm-thick sections on a microtome. Sections were deparaffinized with xylene, rehydrated in three five-minute steps with 100% ethanol, 90% ethanol, and 70% ethanol, and stained with hematoxylin and eosin (HE) for 1 h. Slides were rinsed with RNase-free water and mounted under glass coverslips with Clarion mounting medium (Santa Cruz Biotechnology). They were observed using a Nikon DXM1200C microscope.

RNA isolation and library preparation

Total cellular RNAs were isolated from PICM-19 and PICM-19-CSCs using the RNeasy Mini kit (Qiagen). They were quantified using a fluorimetric RiboGreen assay. Total RNA integrity was assessed by capillary electrophoresis using BioAnalyzer 2100 (Agilent Technologies, Santa Clara, CA, United States). For samples to pass the initial quality control (QC) step, it was necessary to quantify > 1 μg and to have a RNA Integrity Number of ≥ 8. They were then converted to Illumina sequencing libraries using the Truseq RNA Sample Preparation Kit (Cat. # RS-122-2001, Illumina, San Diego, CA). For amplification and library preparation, 1 μg of total RNA was purified using oligo-dT coated magnetic beads, fragmented and reverse transcribed into cDNA. The cDNA was fragmented, blunt-ended, and ligated to indexed (barcoded) adaptors and amplified using 15 cycles of PCR. Final library size distribution was validated using capillary electrophoresis and quantified using fluorimetry (PicoGreen) and by Q-PCR. Indexed libraries were normalized, pooled and size selected to 320bp +/- 5% using Caliper’s XT instrument.

Cluster generation and sequencing

Truseq libraries were hybridized to a paired-end flow cell and individual fragments were clonally amplified by bridge amplification on the Illumina cBot. Once clustering was complete, the flow cell was loaded onto the Illumina HiSeq 2000 sequencer and sequenced using Illumina’s SBS chemistry. Upon completion of read 1, a 7 base pair index read was performed. Finally, the library fragments were resynthesized in the reverse direction and sequenced from the opposite end of the read 1 fragment thus producing the template for paired end read 2.

Primary analysis and de-multiplexing

Base call (.bcl) files for each cycle of sequencing were generated by Illumina Real Time Analysis software. The base call files and run folders were then exported to servers maintained at the Minnesota Supercomputing Institute, where primary analysis and de-multiplexing were performed using Illumina’s CASAVA software 1.8.2. De-multiplexed FASTQ files generated by the CASAVA workflow were analyzed by the mapping and aligning software as described below.

Read mapping and RNA sequencing analysis

Raw paired-end RNA-seq reads in FASTQ format were assessed for base call quality, cycle uniformity and contamination using FastQC (http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc/). The QC-filtered paired-end reads were mapped to pig genome (UCSC susScr3 assembly) via Tophat 2[14] using the UCSC pig annotation in GTF format. Resulting Binary Alignment and Mapping (BAM) files were then visualized using the Integrative Genomics Viewer (IGV)[15] to ensure reasonable behavior of expected control invariant genes [e.g., glyceraldehyde 3-phosphate dehydrogenase (GAPDH)], and to confirm expected behavior of positive control genes (e.g., MYC). BAM files were subsequently analyzed via Cuffdiff (default parameters using the iGenomes GTF file as shown previously[16]) to quantify the expression level(s) of each known gene in units of FPKM (Fragment mapped per kilobase of genes per million mapped reads), and tested for differential expression. A matrix of FPKM transcript abundance levels for each gene was extracted and downloaded into CummeRbund[17] to assess sample clustering using Multi-Dimensional Scaling and Principle Component Analysis, which revealed sample outliers. Differential gene sets were filtered to remove those with fold changes < 1.5 (up- or downregulated) and with a BH FDR-corrected P-value > 0.05.

Real time PCR

The cDNA from PICM-19 and PICM-19-CSCs was synthesized with 1 μg of total RNA using SuperScript II reverse transcriptase (Invitrogen Life Technologies, Carlsbad, CA) and oligo dT6-12 primers (Sigma-Genosys, The Woodlands, TX). Quantitative real-time PCR was performed using the FullVelocity SYBR Green QPCR master mix (Stratagene, La Jolla, CA) following the manufacturer’s specifications. The 25 μL final reaction volume consisted of pre-made reaction mix (SYBR Green I dye, reaction buffer, Taq DNA polymerase, and dNTPs), 0.3 mmol/L of each primer, and 0.5 ng cDNA in water. Reaction conditions for PCR for the Mx3000P QPCR System (Stratagene) were as follows: polymerase activation at 95 °C for 5 min, 40 denaturation cycles of 95 °C for 10 s, annealing at 60 °C for 10 s and elongation at 72 °C for 10 s. A melting curve was performed to assess primer specificity and product quality using the same conditions. The relative product levels were quantified using the 2(-∆∆C_T) method[18]. Data are presented as relative induction of each porcine gene, normalized to β-actin, and are representative of three independent experiments.

RESULTS

Expression of MYC protein in PICM-19 stem cells

We initially determined that PICM-19 cells can be transfected with an expression vector driven by the EF-1α/HTLV hybrid promoter. pUNO1-MYC-IRES-Luc was constructed by cloning the open reading frame (ORF) of firefly luciferase (luc) gene downstream of the MYC ORF separated by an IRES sequence. This vector also contains the blasticidin resistance gene as a selectable marker. The function of the genetic construct was tested by assaying the expression of luciferase in transfected cells. After transfection, feeder-independent PICM-19 cells were cultured (Figure 1A) and transfectants were selected using blasticidin. Individual clones were then screened for MYC expression, and the clone that had the highest MYC expression as detected by RT-PCR (data not shown) was assayed for luciferase activity in 24-well plates using D-luciferin as the substrate. Transfected PICM-19 cells expressed luciferase demonstrating that they took up and incorporated the plasmid DNA (Figure 1B). This finding also indicated that the stem cells were expressing the MYC ORF located upstream of luc ORF in the expression vector. This was confirmed by Western blotting assay using anti-human MYC antibody (Figure 1C).

Figure 1 Generation of PICM-19-CSCs. A: Feeder-independent PICM-19 cells were grown in 10% DMEM/Medium 199 on Matrigel in 24-well plates (see Materials and Methods). They were transfected with pUNO1-MYC-IRES-Luc plasmid and tested for luciferase expression using luciferin as substrate; B: Untransfected PICM-19 cells did not show bioluminescence (left panel), whereas transfected cells expressed luciferase (right panel); C: Detection of MYC protein in PICM-19-CSCs by Western blotting using the anti-human mouse c-MYC monoclonal antibody. As expected, PICM-19-CSCs, but not the untransfected PICM-19 stem cells (control) expressed MYC.

Conversion of PICM-19 liver stem cells into cancer cells by MYC overexpression

We subcutaneously injected 1 × 10⁶ cells into the back flanks of NOD-SCID mice to establish that MYC-overexpressing PICM-19 cells (PICM-19-CSCs) can generate tumors in mice. PICM-19-CSCs, but not the control PICM-19 stem cells, generated tumors in vivo (Figure 2). When the tumors reached 1.5 cm in diameter, animals were sacrificed and the tumors were sectioned to be analyzed by HE staining. The tumors were composed of a uniform mass of cells that were actively undergoing cell division. This experiment demonstrated the proof-of-principle that PICM-19-CSCs are capable of producing tumors in vivo.

Figure 2 PICM-19-CSCs generate tumors after implantation in mice. A: PICM-19 cells were transfected with 3 μg of pUNO1-hMYC by nucleofection. 1 x 10⁶ PICM-19 cells expressing MYC were implanted into the back flanks of NOD-SCID mice (n = 3) by subcutaneous injections. Tumors that grew to 1.5 cm in diameter were harvested; B and C: Tumors were sectioned and HE staining was performed, and cells were visualized by confocal microscopy (B: × 40; C: × 400). These images show that the PICM-19-CSCs induced tumors were uniform, and the cells were actively dividing in the PICM-19-CSC tumors, as seen by mitotic cell division [indicated by arrows in (C)].

Molecular characterization of PICM-19 and PICM-19-CSCs

To understand the differences between PICM-19 and PICM-19-CSCs at the molecular level, we examined the RNA-seq profiles of both these cell types, and found sets of genes that were expressed in each cell line (Tables 1 and 2). These genes were normalized to the expression of housekeeping gene GAPDH, and were analyzed. Only a two-fold change or greater was considered significant. In this process, we identified a set of six genes that were expressed only in PICM-19 stem cells (Table 1). The expression of these six genes (CDO1, C22orf39, DKK2, ENPEP, GPX6 and SRPX2) was undetectable in PICM-19-CSCs, as a result of MYC protein overexpression. Their absence in PICM-19-CSCs suggested their potential use as therapeutic targets for liver cancers associated with MYC overexpression. Among these, CDO1 is a tumor suppressor[19], while Dickkopf-related protein 2 (DKK2) is an inhibitor of the Wnt signaling pathway[20]. Both these genes are epigenetically silenced in human cancers. While the overexpression of DKK2 led to an inhibition in renal cancer progression through apoptotic and cell cycle pathways[20], forced expression of cysteine dioxygenase type 1 (CDO1) decreased tumor cell growth of different tumor cell lines and in mice[19]. Lack of DDK2 expression in PICM-19-CSCs suggested that a putative epigenetic control mechanism to silence DKK2 also exists in MYC-induced liver cancer.

Table 1 Genes expressed by PICM-19 liver stem cells.

Gene ID	Gene abbreviation	Function	Fold Δ1
C22ORF39	Human chromosome 22 orf39 homolog	Unknown	11.33
CDO1	Cysteine dioxygenase type 1	Tumor suppressor	6.28
DKK2	Dickkopf Wnt signaling pathway inhibitor 2	Cell signaling	12.20
ENPEP	Glutamyl aminopeptidase	Proteolysis	4.82
GPX6	Glutathione peroxidase 6	Detoxification	2.18
SRPX2	Sushi-repeat containing protein, X-linked 2	Angiogenesis	17.22

¹Normalized to GAPDH expression.

Table 2 Genes expressed in PICM-19-CSCs but not in PICM-19 cells.

Gene ID	Gene abbreviation	Fold Δ1
VEGF	Vascular endothelial growth factor	41718.7
snoZ40	Small nucleolar RNA Z40	16658.2
ssc-mir-195a-1	MicroRNA mir-195a-1	6514.83
ssc-mir-125b-1	MicroRNA mir-125b-1	3627.42
ssc-mir-423	MicroRNA mir-423	3627.42
SNORD53	Small nucleolar RNA, C/D box53	2811.10
ssc-mir-421	MicroRNA mir-421	2811.10
ssc-mir-133a-2	MicroRNA mir-133a-2	2558.12
ssc-mir-365-1	MicroRNA mir-365-1	2558.12
SNORD16	Small nucleolar RNA, C/D box 16	1497.27
ssc-mir-31	MicroRNA mir-31	1144.09
SERP2	Stress-associated ER protein family member 2	1137.16
SNORD94	Small nucleolar RNA, C/D box 94	1126.85
SNORD67	Small nucleolar RNA, C/D box 67	816.01
Y_RNA	Non-coding RNA	727.41
GRIA	FAD-dependent oxygenase	370.13
CH242-74M17.6	GAS locus protein, porcine-specific	322.43
CH242-74M17.6	GNAS complex locus, porcine-specific	322.43
SNORA36	Small nucleolar RNA, C/D box 36	199.10
RBP1	Retinol-binding protein 1	182.72
TIMP-3	TIMP metallopeptidase inhibitor 3	173.55
RNF130	Ring finger protein 130	147.15
MT-ND6	Mitochondrial NADH dehydrogenase 6	141.82
PCDH7	Protocadherin 7	90.60
ND4L	NADH dehydrogenase subunit 4L	89.11
U12	Non-coding RNA	85.94
MT1A	Metallothionein 1A	84.09
ZFX	Zinc finger protein X-linked	79.94
CH242-17J7.1	GNAS complex locus, porcine-specific	65.88
RNPC1	RNA binding motif protein 38	65.10
GEM1	Gremlin 1	54.41
CADM4	Cell adhesion molecule 4	39.92
CYP1A1	Cytochrome P450, family 1, subfamily A1	39.89
TFAP2A	Transcription factor AP-2 alpha	39.27
AFP	Alpha-fetoprotein	33.23
CFD	Complement factor D	26.62
GNG	Two-component system response regulator GnG/NtrC	24.20
RNF128	Ring finger protein 128	23.79
MIF	Macrophage migration inhibitory factor	23.69
IFI27l2	Interferon, alpha-inducible protein 27-like 2	22.72
PDE6H	Phosphodiesterase 6H, cGMP-specific	22.54
KLF5	Krüppel-like factor 5	22.53
PRKD1	Protein kinase D1	21.27
SPINT2	Serpin peptidase inhibitor, Kunitz type, 2	18.54
MYC	Myelocytomatosis oncogene	17.64
CD44	Cell surface glycoprotein 44	17.14
GPC3	Glypican 3	16.54
ECHS1	Enoyl-coA hydratase, short-chain, 1	15.42
S100A14	S100 calcium binding protein A14	15.41
LONRF3	LON peptidase N-terminal domain and ring finger 3	15.25
CU463271.1	Protein encoded by CU463271.1 locus	14.86
MMP16	Matrix metalloproteinase 16	14.28
S100A16	S100 calcium binding protein A16	14.14
SLC22A17	Solute carrier family 22 member A17	14.07
ID1	Inhibitor of DNA binding 1	13.34
ATP5E	ATP synthase H+ transporting epsilon	13.14
ISG12(a)	ISG12 protein-like	13.03
UQCRH	Uniquinol-cytochrome c reductase hinge protein	12.44
DKK2	Dickkopf homolog 2	12.20
IL17RA	Interleukin 17 receptor A	12.13
STXBP6	Syntaxin binding protein 6	11.63
F2	Coagulation factor II	11.57
COX17	Cytochrome c oxidase copper chaperone	10.05
MGMT	O-6-methylguanine-DNA methyltransferase	9.96
SDF4	Stromal cell derived factor 4	9.50
PDPN	Podoplanin	9.46
DLK2	Delta-like 2 homolog	9.24
WNT1	Wingless-related MMTV integration site family member 1	9.03
CISD1	CDGSH iron sulfur domain 1	9.00
MYD88	Myeloid differentiation primary response 88	8.71
SDC4	Syndecan 4	8.68
EPCAM	Epithelial cell adhesion molecule	8.59
RPS29	Ribosomal protein S29	8.32
P14ARF	p14 ARF protein (porcine-specific)	8.29
ELAVL2	ELAV-like neuron-specific RNA binding protein 2	8.23
FZD3	Frizzled homolog 3	8.17
XBP1	X-box binding protein 1	8.09
TRIAP1	TP53 regulated inhibitor of apoptosis 1	8.08
ATP5O	ATP synthase H+ transporting O subunit	8.02
VKORC1	Vitamin K epoxide reductase complex, subunit 1	7.80
FAM84B	Family with sequence similarity 84, member B	7.70
ECI2	Enoyl-coA delta isomerase 2	7.69
ARHGDIB	Rho GDP dissociation inhibitor beta	7.41
C20ORF27	Human chromosome 22 orf39 homolog	7.31
RAB6B	RAS oncogene family member	7.28
DDT	D-dopachrome tautomerase	7.14
AGMAT	Agmatine ureohydrolase (Agmatinase)	7.12
CHCHD2	Coiled-coil-helix-coil-helix domain 2	6.92
EPHA7	Eph receptor A7	6.72
OSR2	Odd-skipped related 2	6.71
CFI	Complement factor I	6.55
NDUFB11	NADH dehydrogenase 1 alpha subcomplex 11	6.37
SLC35F1	Solute carrier family 35 member F1	6.37
NUDT14	Nucleoside diphosphate linked moiety X-type motif 14	6.35
SH3GL3	SH3 domain GRB2-like 3	6.32
NDUFA12	NADH dehydrogenase 1 alpha subcomplex 12	6.30
CDO1	Cysteine dioxygenase type 1	6.28
DYNLT1	Dynein, light-chain, Tctex-type	6.27
HSBP1	Heat shock factor binding protein 1	6.20
CCK	Chloecystokinin	5.98
ANXA1	Annexin A2	5.85
FCGRT	Fc fragment of IgG, receptor, transporter, alpha	5.84
AK5	Adenylate kinase 5	5.83
RDH11	Retinol dehydrogenase 11	5.58
RPS18	Ribosomal protein S18	5.56
TGOLN2	Trans-golgi network protein 2	5.55
RNF150	Ring finger protein 150	5.54
HIFX	Hypoxia-inducible factor X	5.47
CFH	Complement factor H	5.43
NPC2	Niemann-Pick disease, type C2	5.34
DCTPP1	dCTP pyrophosphatase 1	5.25
CLVS1	Clavesin 1	5.24
HSF4	Heat shock transcription factor 4	5.23
SERINC2	Serine incorporator 2	5.08
MSMB	Microseminoprotein, beta	5.03
CHID1	Chitinase domain containing 1	4.98
LYSMD2	LysM, putative peptidoglycan-binding, domain 2	4.95
SH3BGRL2	SH3 domain binding glutamic acid-rich protein-like 2	4.86
POU3F1	POU domain, class 3, transcription factor 1	4.77
TCN2	Transcobalamin II	4.76
SORT1	Sortilin 1	4.75
PTGES	Prostaglandin E synthase	4.65
CCND2	Cyclin D2	4.63
GXYLT2	Glucoside xylosyltransferase 2	4.50
CRTAC1	Cartilage acidic protein 1	4.33
MFGE8	Milk fat globule-EGF factor 8	4.29
TNC	Tenascin C	4.28
PCBD1	Pterin-4 alpha-carbinolamine	4.27
PDE10A	Phosphodiesterase 10A	4.22
RAB27B	RAS oncogene family member	4.18
INCENP	Inner centromere protein antigens 135/155kDa	4.14
GJB1	Gap junction protein, beta 1	4.09
CEBPG	CCAAT/enhancer binding protein, gamma	4.04
DUROC-SLA1	Src-like adaptor 1 (pig-specific)	4.04
ITIH4	Inter-alpha-trypsin inhibitor heavy chain family, member 4	4.03
PEBP1	Phosphatidylethanolamine binding protein 1	4.02
ABTB2	Ankyrin repeat and BTB (POZ) domain containing 2	4.00
GJB5	Gap junction protein, beta 5	3.95
CREB3	cAMP responsive element binding protein 3	3.86
DPM2	Dolichyl-phasphate mannosyltranferase polypeptide 2	3.85
GEM	GTP binding protein overexpressed in skeletal muscle	3.85
DOCK8	Dedicator of cytokinesis 8	3.84
TMP-SLA-2	Transmembrane protein, pig-specific	3.83
CYB5B	Cytochrome b5 type B	3.80
TSAN6	Tsan6, Pig-specific protein	3.75
RNASEH2C	Ribonuclease H2, subunit C	3.71
NARS	Asparaginyl-tRNA synthetase	3.70
LSM6	U6 snRNP-associated protein	3.69
MTFP1	Mitochondrial fission process 1	3.66
SLC16A1	Solute carrier family 16 member A1	3.65
RARB	Retinoic acid receptor, beta	3.64
SDSL	Serine dehydrarase-like	3.64
MPC2	Mitochondrial pyruvate carrier 2	3.63
PARVB	Parvin, beta	3.60
RENBP	Renin binding protein	3.55
POC1A	POC1 centrolar protein A	3.54
GUSB	Glucuronidase, beta	3.53
CH242-9E10.1	GNAS complex locus, porcine-specific	3.52
REEP3	Receptor accessory protein 3	3.51
CPT1A	Carnitine palmitoyltransferase 1A	3.40
BHLHB9	Basis helix-loop-helix domain containing B9	3.37
FAAH	Fatty acid amide hydrolase	3.37
MRPL20	Mitochondrial ribosomal protein L20	3.37
HTR2A	5-hydroxytryptamine (serotonin) receptor 2A, G protein-coupled	3.36
TRERF1	Transcriptional regulating factor 1	3.35
XYLB	Xylulokinase	3.32
RBPMS2	RNA binding protein with multiple splicing 2	3.28
FGFR3	Fibroblast growth factor receptor 3	3.27
SLC22A23	Solute carrier family 22 member A23	3.21
NDUFB1	NADH dehydrogenase 1 alpha subcomplex 1	3.18
ZFAND2A	Zinc finger, AN1-type domain 2A	3.16
EPB41	Erythrocyte membrane protein band 4.1	3.11
IGSF23	Immunoglobulin superfamily, member 23	3.11
FAM122C	Family with sequence similarity 122, member C	3.10
AMIGO2	Adhesion molecule with Ig-like domain	3.03
SHROOM4	Shroom family member 4	3.02
MAT1A	Methionine adenosyltransferase 1A	3.01
PARN	Poly(A)-specific ribonuclease	3.00
PFKFB1	6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 1	3.00
TCF25	Transcription factor 25	2.98
PRODH2	Proline dehydrogenase (oxidase) 2	2.95
CHCHD10	Coiled-coil-helix-coil-helix domain 10	2.91
CH242-217H9.3	GNAS complex locus, porcine-specific	2.85
CASC5	Cancer susceptibility candidate 5	2.84
IGLON5	IgLON family member 5	2.84
CDH17	Cadherin 17	2.82
SLC44A4	Solute carrier family 44 member A4	2.82
LRCH2	Leucine-rich repeats and calponin homology domain 2	2.76
DUROC-NEU1	Sialidase 1 (pig-specific)	2.75
CDC45	Cell division cycle 45	2.71
BCKDHB	Branched chain keto acid dehydrogenase E1	2.69
HNF4A	Hepatocyte nuclear factor 4, alpha	2.67
SCG5	Secretogranin V	2.66
BOLA1	BolA family member 1	2.65
PYGL	Phosphorylase, glycogen, liver-specific	2.64
SCARB1	Scavenger receptor class B, member 1	2.64
SERPIND1	Serpin peptidase inhibitor, clade D, member 1	2.63
GAS2L3	Growth arrest-specific 2 like 3	2.61
EFNA2	Ephrin A2	2.60
PDZK1IP1	PDZK1 interacting protein 1	2.59
VIMP	VCP-interacting membrane protein	2.59
PRR5L	Proline-rich 5-like	2.57
ACVR2B	Activin A receptor, type IIB	2.56
ATP9A	ATPase, class II, type 9A	2.55
PLAUR	Plasminogen activator, urokinase receptor	2.53
LDLR	Low density lipoprotein receptor	2.51
FLRT3	Fibronectin leucine-rich transmembrane protein 3	2.50
ADAMTSL3	ADAMTS-like 3	2.49
CINP	Cyclin-dependent kinase 2 interacting protein	2.48
WBSCR22	Williams Beuren syndrome chromosome region 22	2.48
ALDH9A1	Aldehyde dehydrogenase 9 family member 1	2.46
SSC.82438	Sus scrofa Amphiregulin	2.44
NME3	NME/NME23 nucleoside diphosphate kinase 3	2.43
CCS	Copper chaperone for superoxide dismutase	2.40
PITX2	Paired-like homeodomain transcription factor 2	2.38
MRPL3	Mitochondrial ribosomal protein L3	2.33
BNC1	Basonuclin 1	2.32
RPS16	Ribosomal protein S16	2.32
TM7SF2	Transmembrane 7 superfamily member 2	2.30
CRELD2	Cystein-rich with EGF-like domains 2	2.22
HHEX	Hematopoietically expressed homeobox	2.20
RPL31	Ribosomal protein L31	2.20
TGFA	Transforming growth factor, alpha	2.20
IDS	Iduronate 2-sulfatase	2.18
NRG4	Neuregulin 4	2.17
DGAT1	Diacylglycerol O-acyltransferase 1	2.16
GNAZ	Guanine nucleotide binding protein	2.14
ARL3	ADP-ribosylation factor-like 3	2.12
FABP5	Fatty acid binding protein 5	2.12
KIF12	Kinesin family member 12	2.12
IKBKG	Inhibitor of kappa B	2.09
PDZRN3	PDZ domain containing ring finger 3	2.09
PHF19	PHD finger protein 19	2.09
MRPL24	Mitochondrial ribosomal protein L24	2.07
ARAP2	ArfGAP with RhoGAP domain, ankyrin repeat and PH domain 2	2.04
ITGF3	ItgF3-like protein	2.02
POLA2	Polymerase (DNA directed), alpha 2	2.02
ADAMTS7	ADAM metallopeptidase with thrombo- Spondin type 1 motif 7	1.45

¹Normalized to GAPDH expression.

In contrast, ectopic expression of MYC protein upregulated the expression of 237 genes in PICM-19-CSCs not found in the PICM-19 stem cell profile (Table 3). New genes upregulated by MYC included those involved in cell growth and proliferation, adhesion, migration, protein transport, cell cycle, metabolism, inflammation, and tumorigenesis underscoring the broad reach of MYC protein in regulating diverse cellular processes. However, genes coding RNAs involved in transcriptional and post-transcriptional regulatory mechanisms dominated the top 25 MYC-upregulated genes (Table 2), consistent with the active cell growth and proliferative nature of PICM-19-CSCs.

Table 3 Genes upregulated by MYC in PICM-19-CSCs.

Gene ID	Gene abbreviation	Function	log2 ratio
CH242-278B18.2	GNAS Complex, porcine-specific	Unknown	10.25
NFKB1	Nuclear factor κB	Inflammation	8.94
SEMA3F	Semaphorin 3F	Cell signaling	8.27
NME1-NME2	NME1-NME2 read-through	Inflammation	8.22
SERPINA1	Serpin protease inhibitor, clade A1	Inflammation	8.17
VLDLR	Very low density lipoprotein receptor	Metabolism	8.01
APOE	Apolipoprotein E	Metabolism	7.57
ITGA3	Integrin, alpha 3	Migration, Cell adhesion, Invasion	7.41
IGF2BP2	Insulin-like growth factor 2 mRNA binding protein 2	Gene regulation	7.33
MT-ATP6	Mitochondrial ATP synthase 6	Metabolism	7.21
HIF1A	Hypoxia-inducible factor 1α	Angiogenesis	7.12
APOA1	Apolipoprotein A1	Metabolism	7.06
ITGB7	Integrin, beta 7	Cell adhesion	7.03
NKAIN1	Na+/K+ transporting ATPase	Ion transport	7.00
FGG	Fibrinogen gamma chain	Cell adhesion	6.79
CTSB	Cathepsin B	Carcinogenesis	6.74
NTN1	Netrin 1	Inflammation	6.65
JUP	Junction plakoglobin	Cell adhesion	6.61
PLG	Plasminogen	Cell proliferation	6.60
MT-ND2	Mitochondrial NADH dehydrogenase 2	Metabolism	6.37
SERPINA3-2	Serpin protease inhibitor, clade A3-2	Inflammation	6.31
RBP4	Retinol binding protein 4	Metabolism	6.18
FTL	Ferritin, light polypeptide	Iron metabolism	6.08
DUSP4	Dual specificity phosphatase 4	Cell proliferation	6.02
MT-CO1	Mitochondrial cytochrome c oxidase I	Inflammation	6.00

The farnesoid X receptor (FXR), a member of the nuclear receptor superfamily, plays an important role in regulating the progression of several cancers including HCC; and miR-421, which was highly upregulated in PICM-19-CSCs, suppresses the expression of FXR to induce tumorigenesis and migration in HCC cells[21]. MiR-421 also regulates deleted in pancreatic cancer 4 (DPC4; also known as Smad4), a critical tumor suppressor shown to be involved in the progression of pancreatic cancer[22].

In addition to these regulatory RNAs, protocadherin-7, tissue inhibitor of metalloproteinase 3 and vascular endothelial growth factor (VEGF) were highly expressed in PICM-19-CSCs compared to PICM-19 parental stem cells. The ratio between the matrix metalloproteinsases (MMPs) and their inhibitors (TIMPs) is an important factor in balancing tissue homeostasis. Recent studies have shown that the TIMP family of proteins not only inhibits several disintegrin-metalloproteinases (ADAMs and ADAMTSs) but also has contrasting functions including anti- and pro-apoptotic[23]. The sequestration of TIMP-3 by binding to the extracellular matrix (ECM) or through endocytosis may help to protect cells from apoptosis while still allowing it to control pericellular proteolysis via its potent inhibition of MMPs, ADAMs and ADAMTSs[23]. In PICM-19-CSCs, only TIMP-3 was upregulated (Table 2), and ADAM23 and ADAMTS2 were suppressed (Table 4).

Table 4 Genes downregulated by MYC in PICM-19-CSCs.

Gene ID	Gene abbreviation	Function	log2 ratio
ASPN	Asporin	Extracellular matrix	8.20
IGF1	Insulin-like growth factor 1	Cell growth	7.44
OGN	Osteoglycin	Metabolism	6.52
GPM6B	Glycoprotein M6B	Membrane trafficking, cell-to-cell communication	6.23
ISLR	Immunoglobulin superfamily containing leucine-rich repeat	Unknown	6.19
ANGPT1	Angiopoietin 1	Angiogenesis	5.53
WDR66	WD repeat domain 66	EMT modulator	5.49
RSPO2	R-spondin 2 homolog	Wnt modulator	5.46
COL6A2	Collagen, type VI, alpha 2	Fibrosis	5.14
TNR	Tenascin R	Extracellular matrix	5.03
FAM5C	Family with sequence similarity 5, member C	Unknown	4.97
FIGF	c-Fos iduced growth factor	Angiogenesis	4.92
ADAM23	ADAM metallopeptidase domain 23	Metabolism	4.84
MAB21L2	Mab-21-like 2	Cell survival	4.83
PDE1A	Phosphodiesterase 1A	Metabolism, Apoptosis	4.83
CLMP	CXADR-like membrane protein	Cell adhesion	4.74
COL28A1	Collagen, type XXVIII, alpha 1	Fibrosis	4.74
GJA1	Gap junction protein, alpha 1	Oxidative stress	4.67
SELENBP1	Selenium binding protein 1	Tumor suppressor	4.58
CPED1	Cadherin-like and PC-esterase domain containing 1	Unknown	4.52
COL5A3	Collagen, type V, alpha 3	Fibrosis	4.51
ADAMTS2	ADAM metallopeptidase with thrombospondin typ1 motif 2	Metabolism	4.50
PDE8B	Phosphodiesterase 8B	Metabolism	4.47
SGCE	Sarcoglycan, epsilon	Cytoskeleton	4.42
GDPD2	Glycerophosphodiester phosphodiesterase domain containing 2	Cell growth	4.00

It has been well established that VEGF is highly produced in HCC, and the inhibition of VEGF results in the suppression of angiogenesis[24]. Moreover, miR-26a targets VEGF and hepatocyte growth factor (HGF) to cause this effect via the HGF-cMet pathway[24] as well as the phosphatidylinositol 3-kinase (PIK3C2α)/Akt/hypoxia-inducible factor (HIF)-1α pathway in HCC cells[25]. In our study, downregulation of miR-26a and upregulation of HIF-1α in PICM-19-CSCs demonstrates that MYC is also a player in this tumor-inducing molecular mechanism of HCC.

Genes not regulated by the MYC protein in PICM-19-CSCs

Next, we studied the profiles of PICM-19 and PICM-19-CSCs to identify the genes whose expression was not altered in stem cells due to MYC overexpression. For this, we selected genes that showed more than a two-fold increase when normalized with the GAPDH. We found that the expression of 28 genes was not affected by the MYC protein (Table 5). These genes included a number of microRNAs (miRNAs) and small nucleolar RNAs (SnoRNAs) that regulate post-transcriptional gene expression, and are expressed at very high levels. As MYC has no influence on their expression, it appears that the genes of this 28-gene set may be primarily involved in maintaining the cellular integrity and stemness.

Table 5 Genes expressed by both PICM-19 and PICM-19-CSCs that were not affected by MYC overexpression.

Gene ID	Gene abbreviation	Function	Fold Δ1
SSC-MIR-425	MicroRNA mir-425	Gene regulation	7925.28
SNORD53	Small nucleolar RNA, C/D box 53	Gene regulation	4969.50
SSC-MIR-378-1	MicroRNA mir-378-1	Gene regulation	3962.64
SSC-MIR-99A	MicroRNA mir-99a	Gene regulation	3761.27
SSC-MIR-101-2	MicroRNA mir-101-2	Gene regulation	3400.56
SSC-LET-7F-2	MicroRNA let-7f-2	Gene regulation	2814.68
5s rRNA	5S ribosomal RNA	Protein synthesis	2255.63
SSC-MIR-210	MicroRNA mir-210	Gene regulation	1730.53
U6	Small nucleolar RNA U6	RNA interference	1187.43
SSC-MIR-196A-2	MicroRNA mir-196a-2	Gene regulation	1127.81
SCARNA15	Small Cajal body-specific RNA 15	Gene regulation	287.24
SNORA72	Small nucleolar RNA, H/ACA box 72	Gene regulation	255.25
SCARNA11	Small Cajal body-specific RNA 11	Gene regulation	163.50
RAMP2	Receptor (G protein-coupled) activity modifying protein 2	Protein transport	56.66
CCNI2	Cyclin I family member 2	Cell cycle	10.67
P4HTM	Prolyl 4-hydroxylase, transmembrane	Hypoxia	6.25
SPDYA	Speedy/RINGO cell cycle regulator family member A	Cell cycle	5.56
PARK2	Parkin RBR E3 ubiquitin ligase	Unknown	5.09
GRPR	Gastrin-releasing peptide receptor	Cell proliferation	4.52
ATG10	Autophagy related 10	Autophagy	3.76
PHYHIP	Phytanoyl-CoA hydroxylase interacting protein	Cell proliferation	3.41
DPEP1	Dipeptidase 1	Metabolism	2.72
DUROC-C2	Nuclear receptor corepressor 2	DNA binding	2.64
HIST1H2BH	Histone cluster 1, H2bh	Histone modification	2.58
PHKG1	Phosphorylase kinase, gamma 1	Glycogenolysis	2.52
TMEM79	Transmembrane protein 79	Protein secretion	2.04
APH1B	APH1B gamma secretase subunit	Protein transport	2.01
IRAK4	Interleukin-1 receptor-associated kinase 4	Cell signaling	2.01

¹Normalized to GAPDH expression.

Differential expression of MYC-regulated genes in PICM-19-CSCs

In order to identify genes that are involved in MYC-induced oncogenesis, we studied global transciptome profiles of PICM-19-CSCs and compared them with those obtained with PICM-19 stem cells. While these genes are expressed by both PICM-19 and PICM-19-CSCs, a significant number of genes were differentially regulated by MYC. While MYC upregulated 4173 genes in PICM-19-CSCs, it downregulated 4248 genes. Statistical analysis of the data has revealed 616 upregulated and 391 downregulated genes that had a log₂ ratio of at least 2. Analysis of gene expression of these two cell types showed that MYC overexpression resulted in changes in gene expression among genes involved in diverse cellular processes. This is not surprising as it is known that MYC regulates about 15% of the genes in mammalian genomes[26].

Among the genes upregulated by MYC are ribosomal genes demonstrating active protein synthesis in CSCs which is associated with cell growth, proliferation, and metabolism. While our data show only 8 of these ribosomal genes, others such as the initiation factors (elF-2a, eIF4A, eIF4E, eIF4G, eIF5A), MrDb, nucleolin, and ribosomal proteins (S11, S16) were also induced > 1.5 fold (data not shown). These genes were previously shown to be regulated by the MYC protein[27].

MYC is also considered an important regulator of the cell cycle. We found that the expression of several cell cycle genes (cdc45, CCND2, CIPN) was enhanced in PICM-19-CSCs. Adhesion molecule with Ig-like domain 2 (AMIGO2), a leucine-rich repeat family member implicated in decreased cell adhesion and migration, and an increase in chromosomal instability and tumorigenicity in human gastric adenocarcinoma[28], was also upregulated in PICM-19-CSCs.

The majority of the top 25 genes upregulated by MYC are those involved in metabolism, adhesion and inflammation (Table 3). Integrin genes ITGA3 and ITGB4, and plakoglobin JUP were upregulated in PICM-19-CSCs. These three genes were recently found to be responsible for metastasis seen in oral squamous cell carcinoma[29]. Their expression in CSCs is significant because integrin adhesion receptors link the ECM to actin cytoskeleton and transmit biochemical signals across the plasma membrane[30]. Binding of integrins to their cognitive receptors trigger signaling cascades to induce diverse cellular processes such as cell survival, apoptosis, proliferation, and motility[29]. JUP expression was also associated with colorectal tumor metastasis[31].

Interestingly, markers of inflammation in the liver, apolipoproteins (APOA1 and APOE) and the SERPINA1 gene encoding alpha1-antitypsin (AAT), and another member of the AAT family, SERPINA3-2 were also induced in PICM-19-CSCs. While APOE was shown to confer protection against liver damage caused by hepatitis C virus (HCV) infection[32], its expression level was found to be significantly higher in HBV associated liver cirrhosis[33]. In a recent HCC study on familial aggregation, APOA1 and APOE were identified by mass spectroscopy of the proteins as being among nine highly overexpressed in individuals with HCC[34]. Previous studies have shown that the protein level of APOE was increased in human HCC[35], and that human hepatoma-derived cell lines express high levels of APOE and very low density lipoprotein (VLDL). Consistent with this, PICM-19-CSCs displayed an upregulated expression of VLDLR (Table 3). AAT is the most abundant liver-derived glycoprotein in the plasma, and a diagnostic and prognostic marker of HCC[36,37]. It is associated with the progression from liver cirrhosis to cancer in HCV-infected individuals[38]. Collectively, upregulation of APOA1, APOE, VLDLR and AAT in PICM-19-CSCs suggested their involvement in MYC-induced liver cancer.

We found that MYC overexpression resulted in downregulation of a number of genes (Table 4) that were previously implicated in tumor reduction and/or inhibition of human cancers. These genes differ from those that were silenced; and the reduced expression could be a result of either direct or indirect MYC targeting (Table 1). Among these, c-Fos induced growth factor, FIGF (also known as vascular endothelial growth factor D), is active in angiogenesis and endothelial cell growth. Its putative role in human cancers is unclear. While it was shown to enhance lymphatic node metastasis in tissue sarcomas[39], its expression was significantly lower in the serum of metastatic patients with differentiated thyroid sarcoma suggesting that FIGF may prevent metastasis. Interestingly, MYC caused the downregulation of FIGF in PICM-19-CSCs (Table 4) indicating that, in liver stem cells, FIGF has an anti-tumor function. Osteoglycin, a protein that decreases gelatinase activity and suppresses lymphatic metastasis[40], was also downregulated in PICM-19-CSCs.

Collagens are structural components of the ECM and are primarily synthesized by hepatic stellate cells in the liver to induce fibrosis[41]. A recent proteomic study with platelet-derived growth factor C transgenic and phosphatase and tensin homolog null mouse models of liver fibrosis and its progression to HCC demonstrated that the protein levels of collagens COL6A2 and COL28A1 were downregulated in fibrosis, and COL5A3 was upregulated in fibrosis; whereas only COL6A2, but not COL5A3 and COL28A1, was increased in tumors generated in both mouse models[42]. We found that COL5A3, COL6A2, COL28A1 were significantly downregulated in PICM-19-CSCs. Thus, in liver cells, MYC may be involved in the regulation of COL5A3 and COL28A1.

Among regulatory RNAs upregulated in PICM-19-CSCs, miR-125b was recently shown as a negative regulator of p53[43]. While the knockdown of miR-125b elevated p53 levels and induced apoptosis in zebra fish and human cells[43], its overexpression promoted human glioma cell proliferation and inhibited apoptosis by targeting the pro-apoptotic Bcl-2 modifying factor[44].

MiR-31 is an oncomir that is upregulated in PICM-19-CSCs. In colorectal cancer cells, miR-31 activated the rat sarcoma (RAS) pathway and inhibited the function of the RAS p21 GTPase activating protein 1 (RASA1) in promoting tumorigenesis and cancer cell growth[45]. It was also shown to promote oncogenesis in the cholangiocarcinoma HCCC-9810 cell line by inhibiting RASA1, thereby downregulating the mitogen-activated protein kinase signaling pathway[46]. Based on these findings, it appears that miR-31 may also be inhibiting the RASA1 in liver cancers. Paradoxically, miR-195, which was shown to suppress tumorigenicity and cause cell cycle arrest by blocking the G1/S transition via the repression of retinoblastoma-E2 transcription factor signaling through cyclin D1, CDK6, and E2F3 in HCC cell lines[47], was upregulated in PICM-19-CSCs.

High levels of miR-365 were found in PICM-19-CSCs. The expression levels of miR-365 vary in different malignancies. It is was overexpressed in both cells and clinical specimens of cutaneous squamous cell carcinoma[48], where it was found to suppress to Mybl2 activity[49]. siRNA silencing of Mybl2 in proliferating Caco-2 cells increased the expression of the cell-cycle regulators cdk2, cyclin D2, and c-MYC; and decreased the expression of cdc25B and cyclin B2[50]. Thus, it is possible that MYC protein suppresses the expression of Mybl2 through the upregulation of miR-365, as seen in the PICM-19-CSCs.

Another key and common feature of CSCs was activation of NF-κB signaling, which is known to increase stress resistance and survival[10]. PICM-19-CSCs had a high level of NF-κB expression when compared to PICM-19 cells. Similarly, the expression of anti-apoptotic genes Bcl-xl and Mcl-1 were elevated in PICM-19-CSCs. Thus, the blocking of anti-apoptotic effects of these proteins may be a viable strategy to eliminate tumor cells. Sabutoclax, a pan-inhibitor of Bcl-2 family of protein that was shown recently to target Mcl-1[51], may be a suitable drug for this purpose.

Selenium-binding protein (SELENBP1) is a tumor suppressor and lower SELENBP1 levels are associated with poor prognosis of liver, lung, colorectal and stomach cancers[52,53]. Decreased SELENBP1 expression has been associated with increased cell motility, proliferation and inhibition of apoptosis, and vascular invasion in HCC[53]. Its expression was suppressed in PICM-19-CSCs suggesting that MYC regulates this gene in the liver to promote cell proliferation and invasion. In a recent proteomic study conducted with serum and tumor tissue specimens of c-MYC transgenic mice, it was shown that APOE, SELENBP1 and retinol binding protein 4 are differentially regulated[54]. Our data support these findings and confirms that MYC differentially regulates the expression of these proteins in liver cells.

Previous studies have also identified MYC as a downstream target of the Wnt signaling, a pathway that is highly active in HCC[55,56]. Dysregulated Wnt signaling occurs in 30% of HCCs. Also, deletion of adenomatous polyposis coli (APC), a negative regulator of Wnt, in the adult mouse liver, leads to an increase in the nuclear β-catenin and MYC[57]. Gene expression array analysis of colorectal cancer cells has shown that MYC is required for the majority of Wnt targeted gene activation after the APC loss[58]. In our study, we found a high level of Wnt expression (Table 2) demonstrating that the Wnt pathway was activated in PICM-19-CSCs.

Validation of MYC-regulated genes in PICM-19-CSCs

To further validate these RNA-seq data, four genes that were upregulated, four genes downregulated, and two genes that were silenced in PICM-19-CSCs, when compared with PICM-19 cells (Tables 1, 3-4) were selected for quantitative real-time PCR studies on the basis of their roles in diverse cellular processes. PCR was performed using porcine-specific primers for CTSB, HIF-1A α-subunit, SERPINA3-2, ITGA3, SELENBP1, RSPO2, FIGF, ANGPT1, CDO1 and DKK2 (Table 6). While two of them (HIF-1α and SELENBP1) were suggested as biomarkers for HCC, others were identified in studies with multiple human cancers. As shown in Figure 3, induced expression of MYC in PICM-19-CSCs resulted in the alteration of gene expression for all ten genes studies. These findings clearly illustrate differences in the expression of these genes between PICM-19 stem cells and PICM-19-CSCs.

Figure 3 RNA sequencing of PICM-19 liver stem cells and PICM-19-CSCs. A: Quality of the RNA extracted from PICM-19 and PICM-19-CSCs was assessed by electrophoresis (left panel) and by Nanodrop measurement to determine (RIN) for each cell type. RIN for PICM-19 and PICM-19-CSCs was 9.50 and 9.30, respectively. After the completion of sequencing, data was analyzed with TopHat and Cufflinks (see Materials and Methods), differential gene expression was calculated as fragments per kilobase of transcripts per million mapped reads (FPKM). Data thus obtained was plotted as a B: Density plot; C: Scatter Matrix plot; D: Box plot. Differentially expressed genes for each cell type were shown as log₁₀ fold change. log₁₀ fold change of 2 or greater was considered significant. In total, 25096 genes were differentially expressed in both cell types, whereas PICM-19 and PICM-19-CSCs displayed unique expression of 6 and 237 genes, respectively.

Table 6 Procine-specific primers used in real-time polymerase chain reaction.

Gene	Gene ID	GenBank#	Primer sequences	Product size (bp)
CTSB	100370961	NM_001097458.1	F: AGCCTGGAACTTCTGGACAA	190
			R: TTTGTAGGACGGGGTGTAGC
HIF-1A α-subunit	396696	NM_001123124.1	F: GCCAGAACCTCCTGTAACCA	204
			R: CCTTTTCCTGCTCTGTTTGG
SERPINA3-2	396686	NM_213787.1	F: TGAAAGTTTCCCAGGTGGTC	200
			R: CTAGGGCTTGGTGACTTTGC
ITGA3	100517053	XM_005668879.1	F: CTTCAAACGGAACCAGAGGA	203
			R: AGAAGCTTTGTAGCCGGTGA
SELENBP1	100152724	XM_001929643.3	F: GGTACCAGCCTCGACACAAT	203
			R: GTTGTGCAGGAATCGGATCT
RSPO2	100154008	NM_001293141.1	F: GTCCAGATGGTTTTGCACCT	202
			R: CTCCTGGATTCAGCAATGGT
FIGF	100155670	XM_001928382.3	F: TCCCCAACCCTGTCATTTTA	202
			R: TCCTTCTCCATTCCTTGGTG
ANGPT1	397009	NM_213959.1	F: GGGGGAGGTTGGACTGTAAT	203
			R: TGTGAATAGGCTCGGTTTCC
CDO1	100312964	NM_001167643.1	F: TCTGAAGATGCTGCAAGGAA	208
			R: CGTATCGAAGGGTGGACTGT
DKK2	100519672	XM_003129269.2	F: CTGGTACCCGCTGCAATAAT	198
			R: GACAGGGATCTCCTTCATGC
β-actin	45269029	AY550069.1	F: GGACCTGACCGACTACCTCA	180
			R: GGCAGCTCGTAGCTCTTCTC

Collectively, the vast repertoire of genes that are either directly or indirectly induced by MYC protein expression in PICM-19-CSCs underscores the importance of its involvement in the induction of diverse cellular pathways that ultimately may lead to liver cancer.

DISCUSSION

CSCs have been identified in a number of cancers, including HCC, as a unique population of cells that possess self-renewing and tumor-initiating properties[5,8]. Mounting evidence strongly suggests that CSCs are resistant to anti-cancer therapies[11,59]. Therefore, targeting CSCs may be an effective strategy to eliminate the drug resistant and metastatic tumor cells[60]. Here, we have developed CSCs that are capable of forming tumors in mice by overexpressing human MYC in the liver stem cell line, PICM-19. The involvement of MYC in murine and human HCC malignant transformation has been well documented[2,4,7]. Cellular targets of MYC include genes from almost every biochemical and regulatory pathway, including growth, metabolism, cell cycle progression, differentiation and apoptosis[26,61]. Our results provide a snapshot of differential gene expression induced by the MYC protein in an in vitro model of liver stem cells. PICM-19-CSCs were able to form tumors in immunodeficient mice within three weeks, demonstrating their elevated proliferation and tumor-producing potential in vivo. Analysis of gene expression profiles in PICM-19-CSCs has revealed novel “molecular signatures” that were either up- or downregulated as a consequence of MYC overexpression. A number of genes were both up- and downregulated by MYC overexpression in PICM-19 cells demonstrating their potential involvement in MYC-induced tumorigenesis. We also found that MYC expression silenced six genes completely (Table 1), suggesting that the inhibition of these genes is potentially crucial for tumor induction by MYC. To our knowledge, this is the first report to investigate transcriptome profiles driven by a single oncogene in liver stem cells, and cells of porcine origin, an important animal for modeling human physiology and disease.

The use of MYC to generate CSCs is highly relevant to study HCC because MYC is a key component of hepatocarcinogenesis. High-level expression of MYC was also shown to be essential for the proliferation and survival of other CSCs, such as in glioma[62]. In a recent report, chemoresistant CSCs were isolated from MYC-driven mouse liver tumors[63]. Side-population (SP) cells in these tumors expressed CD44, a marker of hepatic progenitor cells, and displayed an upregulation in the expression of multidrug resistance gene 1 (MDR1), which conferred resistance to doxorubicin, paclitaxel and SN38[63]. We found that PICM-19-CSCs had a significant upregulated expression of MDR1 when compared to PICM-19 cells, suggesting that MDR1 can be an effective therapeutic target for MYC-driven liver cancers. Interestingly, CD44 expression was found only in PICM-19-CSCs but not in PICM-19 cells, demonstrating that MYC regulates CD44.

We discovered several notable similarities in comparing our RNA-seq data with those recently published from CD90⁺ liver CSCs isolated from HCC patients[64]. Among these, upregulation of glypican 3 (GPC3), apolipoprotein E (APOE), Polo-like kinase 2 and peroxisome proliferator-activated receptor γ, and downregulation of interleukins 8 and 12, suppressor of cytokine signaling 3, and chemokine ligands CXCL12 and CCL2 also occurred in PICM-19-CSCs. Interestingly, neither PICM-19 nor PICM-19-CSCs expressed the progenitor cell markers CD90 and CD133. This observation suggests species-specific differences between cells of human and porcine origin.

MYC also induced the expression of hepatic stem cell marker EpCAM, a target of the Wnt/β-catenin signaling pathway, in PICM-19-CSCs. EpCAM-expressing cells in HCC tumor specimens were previously shown to possess stem-like features and were able to initiate tumors within the liver[65]. PICM-19 cells expressed MYC, Oct4, Sox2, Klf4, and Nanog, which are factors that are commonly used to generate induced pluripotent stem cells. When MYC protein was overexpressed in PICM-19 cells the expression of these proteins was elevated, suggesting that such reprogramming factors are critical for the proliferation of CSCs, and may play a role in tumor-initiation, as shown recently in Ewing sarcoma[66] and breast cancer[67].

By comparing global transcriptome sequencing data between PICM-19 and PICM-19-CSCs, we have confirmed the involvement of a number of MYC-regulated genes involved in development (Sox2, Klf4, Nanog, Oct4, Amd1, Med12), stemness and self-renewal (Wnt/β-catenin), cell cycle (Cyclin D1, GPC3), metabolism (APOA1, APOE), and markers of hepatic stem/progenitor cells (EpCAM, KRT19). In fact, some of these were found to be expressed by side-population (SP) cells isolated from liver cancer cell lines[68]. Their upregulation in CSCs makes them potential candidates as therapeutic targets in liver cancers where MYC is overexpressed. In addition to these genes, PICM-19-CSCs also displayed high levels of a novel protein Sal-like protein 2 (SALL2), a member of the zinc finger transcription factor family. Recently, a family member SALL4 was found to be expressed in liver cancers[69], where it regulates the stemness of EpCAM⁺ CSCs[70]. While we did not detect SALL4 in PICM-19-CSCs, the induction of SALL2 suggests a certain degree of redundancy in the function of these proteins.

Hypoxia plays an important role in the recurrence of HCC, and HIF-1α is critical part of this process. After insufficient radiofrequency ablation of HCC or sustained treatment with sorafenib, tumor recurrence occurs via signaling pathways involving HIF-1α[71,72], and the expression level of HIF-1α was recently shown to correlate with unfavorable outcomes in patients with HCC[73]. We found that the expression of HIF-1α was significantly upregulated in CSCs. Another protein overexpressed in CSCs is cathepsin B (CTSB). Even though the role of CTSB in HCC is unclear, it has been shown to interact directly with the hepatitis B spliced protein-encoded by the HBV in hepatoma cell lines promoting tumor cell migration and invasion[74]. The alpha-1-antichymotrypsin 2 (SERPINA3-2) is a porcine-specific gene that is also significantly upregulated in CSCs (Figure 4). It shares 71% identity with human SERPINA3, a serine protease inhibitor[75]. While its involvement in HCC is unknown, it has been reported recently that the expression of SEPRINA3 is elevated in cholangiocarcinoma[76].

Figure 4 Validation of RNA-seq data using real-time polymerase chain reaction. Expression of ten genes was tested by real-time polymerase chain reaction using total RNA isolated from PICM-19 and PICM-19-CSCs (see Materials and Methods). Data are presented as mean values from three independent experiments with a variance of < 10%.

Epithelial-mesenchymal transition is a crucial event in the tumor progression of HCC. It has been reported recently that the expression of integrin α3 (ITGA3) is a molecular marker of cells undergoing EMT and of cancer cells with aggressive phenotypes[77]. ITGA3 expression is upregulated in breast cancer[77], and colorectal cancer[78]; and ITGA3-knockout mice were found to be considerably less susceptible to skin carcinogenesis[79], suggesting that ITGA3 is a critical gene in multiple cancers. While its involvement has not been demonstrated in HCC, we observed that its expression is enhanced in CSCs. This result indicated that ITGA3 may play a role in MYC-induced liver cancer.

The R-spondin proteins are secreted agonists of the canonical Wnt/β-catenin signaling pathway. Although they are unable to initiate Wnt signaling, they can enhance responses to low-dose Wnt proteins[80]. One of these proteins, R-spondin 2 (RSPO2) possesses tumor-suppressive activity and was recently shown to inhibit tumor growth in colorectal cancer cells by inhibiting Wnt signaling[81]. Induction of Wnt signaling and downregulation of RSPO2 was also observed in human neurofibromas and malignant peripheral nerve sheath tumors[82]. We found that RSPO2 was downregulated in CSCs, suggesting that it might act as a tumor suppressor in liver cells as well.

The expression of SELENBP1 is markedly reduced in multiple tumor types. Downregulation of SELENBP1 correlated with increased DNA damage and oxidative stress in tissue samples of patients with HCC[83]. Recent studies have shown that the reduced expression of SELENBP1 correlated with poor survival in breast cancer patients[84], and with the degree of colorectal cancer differentiation[85], suggesting that its expression could serve as a potential biomarker for these cancers. We found down-regulation of SELENBP1 in CSCs, demonstrating that that MYC-overexpression can result in this effect.

Tumor metastasis is associated with accelerated cell proliferation, dedifferentiation and loss of cell-cell contacts. FIGF and ANGPT1, two genes involved in angiogenesis, were found to be down-regulated in canine mammary carcinoma[86], breast cancer[87], and prostate cancer[88]. In our studies, the expression of both these genes was significantly lowered in CSCs, suggesting that their reduced expression is a factor in MYC-induced tumorigenesis in liver cells.

Aberrant DNA methylation was shown to play a role in carcinogenesis and disease progression in many cancers[89], and DNA methylation analysis of cancer genes might reveal novel prognostic biomarkers[90]. Cysteine dioxygenase 1 (CDO1), a tumor suppressor, is one such gene that is silenced by promoter methylation in multiple human cancers[19]. For instance, in lymph node-positive breast cancer patients, promoter DNA methylation of CDO1 causes its downregulation, making it an ideal biomarker for disease outcome prediction[91]. A recent DNA methylation study on human lung squamous cell carcinoma has also shown that CDO1 is hypermethylated and downregulated in lung cancer[92]. Even though forced expression of CDO1 resulted in markedly decreased tumor cells growth in human cancer cell lines and in mice[19], little is known about its role in HCC. In our studies, we found that CDO1 is silenced in CSCs, indicating that CDO1 may play a role in MYC-induced carcinogenesis.

Among Wnt antagonists, the Dickkopf (DKK) family of proteins act as tumor suppressors and are frequently silenced in various cancers[93,94]. It has been previously shown that DKK2 is epigenetically silenced in renal cancer, where it inhibits tumor progression through apoptotic and cell cycle pathways involving Bcl2 and cyclin D1[20]. In our study, we found that the expression of DKK2 is silenced in CSCs after MYC-overexpression (Figure 4), suggesting that the inhibition of Wnt antagonist is a mechanism of carcinogenesis mediated by the MYC protein.

Approximately, 70000 cancer deaths in the United States each year are associated with changes in the MYC gene or its expression[61]. Despite enormous progress achieved during the past several decades, long-term patient survival remains very low. Advances in stem cell biology provide opportunities to develop novel approaches to reduce morbidity and mortality associated with liver cancer. Our study is an important contribution to understanding the potential of CSC-targeted therapy for HCC. Novel genes identified in our investigation may serve as promising candidates to develop effective therapeutic options for HCC.

ACKNOWLEDGMENTS

We thank Dr. Aaron Becker (BioMedical Genomics Center, University of Minnesota) for RNA library preparation and sequencing, and Drs. John Garbe and Ying Zhang (Minnesota Supercomputing Institute) for help with Galaxy Suite software analysis.

COMMENTS

Background

It is now well-established that cancer stem cells (CSCs) not only drive tumor growth in the liver, in general, but also that the proto-oncogene c-Myc plays a critical role in that process. However, little is known about the genes that are induced and regulated by the MYC protein to generate tumors, and, in particular, those involved in liver stem cells.

Research frontiers

Hepatocellular carcinoma is a multifactorial disease in which the MYC protein plays a central role. However, a number of different proteins and signaling pathways are activated in liver cancer. Cancer stem cells were isolated from hepatocellular carcinoma (HCC) specimens and by culturing human HCC cell lines. But it is unclear as to which cellular pathways are activated by the MYC protein.

Innovations and breakthroughs

This is the first study to demonstrate the regulation of global gene expression by a single oncoprotein (MYC) in porcine liver stem cells using a novel stem cell line, PICM-19, that is capable of propagating in cell culture for a prolonged time period without losing its phenotypic characteristics. By overexpressing the MYC protein, these stem cells were converted into cancer cells. By using a cutting-edge deep sequencing technique (RNA-seq), the authors have identified six genes (CDO1, C22orf39, DKK2, ENPEP, GPX6, SRPX2) that might be critical in carcinogenesis induced by the MYC protein in PICM-19 cells.

Applications

MYC-driven genes identified in this study may serve as promising candidates for the development of HCC therapeutics without the deleterious effects on non-cancerous stem cells and on other cells present in the liver.

Terminology

CSC - cancer stem cells; RNA-seq - RNA sequencing.

Peer-review

In this study, the authors identify the overall genes expression in a liver stem cell after induction by c-MYC plasmid transfection in generating hepatic tumors. The authors used an immortal porcine liver stem cell line, PICM-19, to identify MYC-driven differential gene expression. Transcriptome sequencing was carried out by RNA sequencing and genes identified by this method were validated using real-time PCR. MYC-driven genes identified in this study may serve as promising candidates for the development of HCC therapeutics that would not have deleterious effects on other cell types in the liver.