Identification of biomarkers associated with partial epithelial to mesenchymal transition in the secretome of slug over-expressing hepatocellular carcinoma cells

Abstract

Background

Hepatocellular carcinoma (HCC) is the second leading cause of cancer-related deaths worldwide. Complete epithelial to mesenchymal transition (EMT) has long been considered as a crucial step for metastasis initiation. It has, however, become apparent that many carcinoma cells can metastasize without complete loss of epithelial traits or with incomplete gain of mesenchymal traits, i.e., partial EMT. Here, we aimed to determine the similarities and differences between complete and partial EMT through over-expression of the EMT-associated transcription factor Slug in different HCC-derived cell lines.

Methods

Slug over-expressing HCC-derived HepG2 and Huh7 cells were assessed for their EMT, chemo-resistance and stemness features using Western blotting, qRT-PCR, neutral red uptake, doxorubicin accumulation and scratch wound healing assays. We also collected conditioned media from Slug over-expressing HCC cells and analyzed its exosomal protein content for the presence of chemo-resistance and partial EMT markers using MALDI-TOF/TOF and ELISA assays, respectively.

Results

We found that Slug over-expression resulted in the induction of both complete and partial EMT in the different HCC-derived cell lines tested. Complete EMT was characterized by downregulation of E-cadherin and upregulation of ZEB2. Partial EMT was characterized by upregulation of E-cadherin and downregulation of vimentin and ZEB2. Interestingly, we found that Slug induced chemo-resistance through downregulation of the ATP binding cassette (ABC) transporter ABCB1 and upregulation of the ABC transporter ABCG2, as well as through expression of CD133, a stemness marker that exhibited a similar expression pattern in cells with either a complete or a partial EMT phenotype. In addition, we found that Slug-mediated partial EMT was associated with enhanced exosomal secretion of post-translationally modified fibronectin 1 (FN1), collagen type II alpha 1 (COL2A1) and native fibrinogen gamma chain (FGG).

Conclusions

From our data we conclude that the exosomal proteins identified may be considered as potential non-invasive biomarkers for chemo-resistance and partial EMT in HCC.

Introduction

Hepatocellular carcinoma (HCC) is the fifth-most common cancer and the second leading cause of cancer-related deaths worldwide [1]. For epithelial tumors to metastasize, a fraction of the transformed cells loses its epithelial traits and acquires a mesenchymal phenotype, which is characterized by an elongated spindle-like morphology, a motile phenotype, a loss of adherens and tight junctions and attachment to the extracellular matrix via focal adhesions [2]. This process is known as epithelial to mesenchymal transition (EMT). The process of EMT is characterized by a down-regulation of epithelial markers such as E-cadherin, and an up-regulation of mesenchymal markers such as vimentin and EMT transcription factors (EMT-TF) such as Snail, Slug, Twist1, ZEB1 and ZEB2 [3, 4]. Snail and Slug can bind to the promotor of the CDH1 gene (coding for E-cadherin) and, by doing so, repress its expression. In addition, these transcription factors can up-regulate the expression of proteins that facilitate cell migration such as vimentin, fibronectin 1 and matrix metalloproteases [5]. It has been found that in vitro over-expression of Slug or Snail is sufficient to induce EMT not only in normal epithelial cells [6, 7], but also in cancer cells [8, 9]. There is, however, accumulating evidence indicating that Slug-mediated EMT can be either complete [10] and partial [11]. In case of partial EMT, cells simultaneously express epithelial and mesenchymal traits. Hence, they can initiate metastasis with incomplete loss of epithelial traits and/or incomplete gain of mesenchymal traits [12]. Circulating tumor cells (CTCs) usually exhibit a partial EMT phenotype reminiscent of epithelial cells for their cell-cell adhesion traits and mesenchymal cells for their migratory traits [13]. Taking advantage of both collective migration [12, 14] and clustering in the bloodstream [15, 16], it has been found that CTCs with a partial EMT phenotype may be apoptosis resistant and can be up to 50 times more metastatic than individually migrating CTCs [17]. Therefore, CTCs exhibiting a partial EMT phenotype may pose a higher metastatic risk for patients [18].

Extracellular vesicles (EVs) can be classified in at least three subgroups including apoptotic bodies, micro-vesicles and exosomes [19]. Exosomes that are 40–150 nm in diameter are derived from endosomal pathways [20]. It has been suggested that exosomes, as carriers of proteins, miRNAs and part of the cellular secretome, may contribute to the EMT process [21]. As exosomes are found in body fluids that can be collected non-invasively, they represent a great potential for diagnostic and therapeutic purposes [22, 23].

Here, we hypothesized that over-expression of the EMT inducer Slug may affect the chemo-resistance and stemness of HCC cells. To test this hypothesis, we investigated chemo-resistance and EMT-associated biomarkers from exosomes secreted by these cells. We found that fibronectin 1 (FN1) [24–26], collagen type II alpha 1 (Col2A1) and fibrinogen gamma chain (FGG) may serve as useful non-invasive biomarkers for chemo-resistance and partial EMT in HCC.

Materials and methods

Cell lines and experimental design

HCC-derived HepG2 cells were obtained from the ATCC (Middlesex, UK) and HCC-derived Huh7 cells were kindly provided by Dr. Mehmet Öztürk (Dokuz Eylül University, Turkey). Both cell lines were stably transfected with a Slug plasmid (see below). From each cell line two different Slug over-expressing clones were selected, called HepG2-Sl-Clon1, HepG2-Sl-Clon2 and Huh7-Sl-Clon1 and Huh7-Sl-Clon2. As controls HepG2 and Huh7 cells that were stably transfected with a pcDNA3.1/Neo empty vector were used. In addition, transient transfections were carried out and the resulting cells were called HepG2-Sl (Tr) and Huh7-Sl (Tr). Control cells transiently transfected with a pcDNA3.1/Neo empty vector were called HepG2-EV (Tr) and Huh7-EV (Tr), respectively.

Cell culture conditions, morphological observations and reagents

HepG2 and Huh7 cells were cultured in DMEM medium (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% fetal bovine serum (Sigma-Aldrich), penicillin (100 U/ml) and streptomycin (100 μg/ml) (Sigma-Aldrich) at 37 °C in a humidified atmosphere containing 5% CO₂. Changes in cellular morphologies were assessed using an inverted microscope (Olympus, BX 50, Center Valley, PA, USA). Cisplatin (Sigma-Aldrich) was dissolved in water and doxorubicin (Sigma-Aldrich) was dissolved in DMSO. Both were stored as stock solutions at −20 °C.

Stable and transient transfection assays

SlugMyc_pcDNA3 (Addgene ID 31698) was a gift from Paul Wade [27]. Stable transfections with SlugMyc_pcDNA3 and the empty pcDNA3.1/Neo (Invitrogen) plasmids were carried out using a calcium phosphate-mediated chemical transfection method [28]. Briefly, 150,000 cells per well were seeded into a 6-well plate and allowed to adhere overnight. Next, the culture medium was replaced by fresh medium 2 h prior transfection. To this end, 2× HeBS (Hepes buffered saline) and DNA/CaCl₂ were mixed and immediately vortexed. After 20 min the mixture was evenly distributed to the culture wells after which the cells were incubated for 48 h under standard growth conditions. Subsequent selections were carried out using 500 μg/ml G418 (Sigma-Aldrich, St. Louis, MO, USA) for 3–4 weeks. Single colonies were picked using cloning cylinders (Milipore) and transferred to individual wells of a 96-well plate. Cells growing under selection pressure were subsequently transferred to 6-well plates. More than 40 colonies from both Huh7 and HepG2 were generated in order to select Slug over-expressing clonal cell lines. For each assay empty vector transfected cells were used as controls.

Transient transfections were carried out using Turbofect Reagent (Thermo Scientific, Germany) according to the manufacturer’s protocol. Briefly, 350,000 HepG2 and 140,000 Huh7 cells per well were seeded in 2 ml DMEM in a 6-well plate. After 24 h incubation 8 μg plasmid DNA was diluted in 400 μl serum-free DMEM after which 8 μl Turbofect transfection reagent was added for the transfection of HepG2 cells. For the transfection of Huh7 cells 4 μg plasmid DNA was diluted in 400 μl serum-free DMEM after which 8 μl Turbofect transfection reagent was added.

In-cell Western assay

For the selection of Slug overexpressing cells, individual cells were examined for Slug expression using an in-cell Western method. For this, cells were seeded into each well of a 96-well plate at a density of 50,000 cells/ml. After 48 h, the cells were fixed in ice-cold methanol for 10 min at −20 °C. Next, the wells were washed with 200 μl PBS three times (3 min each) and subsequently blocked in Odyssey blocking buffer (Licor, Lincoln, NE, USA) for 1 h at room temperature. The wells were then incubated with an anti-Slug antibody (Cell Signaling Technology CST, Danvers, MD, USA, Cat. No: C19G7, 1:100 dilution) in 0.1% Tween 20 containing Odyssey blocking buffer overnight at 4 °C with shaking. Next, the wells were washed 3 times with PBS/0.1% Tween 20 (PBS-T) and incubated with an Infrared IRDye800CW secondary antibody (1:1000) and a cell tag stain (1: 2000) in Odyssey blocking buffer/0.2% Tween-200 for 1 h at room temperature. After this incubation, the wells were washed with PBS-T and imaged on an infrared scanner (Licor, Odyssey Clx) using both 700 and 800 nm wavelength channels. The expression of Slug was determined through absorbance at 800 nm. These signals were background-subtracted from those of wells treated with only secondary antibody, and then normalized to cell numbers through division by the 700 nm signal. Data were acquired using the ImageStudio software tool, exported and analyzed in Excel.

Western blotting

To determine protein expression, 750,000 cells were seeded in 75 cm² flasks. When the cells reached 80% confluency, they were washed with ice-cold PBS and scraped off in RIPA buffer containing a protease inhibitor cocktail (cOmplete™ Protease Inhibitor Cocktail, Roche Mannheim, Germany). The extracted protein samples were subsequently mixed with β-mercaptoethanol and 4× protein sample loading buffer and incubated at 90 °C for 5 min. Next, the samples were subjected to SDS-PAGE in stain-free gels (TGX Stain-free, Bio-Rad, Richmond, CA, USA) after which the proteins were visualized using a gel documentation system (Gel Doc EZ, Bio-Rad). These proteins were transferred to nitrocellulose membranes (Licor, Odyssey Clx, Lincoln, NE) using a semi-dry electro-transfer device (Transblot turbo, Bio-Rad) after which the membranes were blocked in skim milk (Sigma-Aldrich) and incubated with antibodies directed against Slug (CST, Cat. No: C19G7, 1:500 dilution), CD133 (Invitrogen, Cat. No: PA5–38014, 1:1000 dilution), E-cadherin (CST, Cat. No: 24E10, 1:1000 dilution) and vimentin (CST, Cat. No: D21H3-XP, 1:100 dilution) at 4 °C overnight. Next, the membranes were washed 3 times in TBS-T buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.1% Tween 20) and incubated with a secondary antibody for 1 h at room temperature. Finally, the membranes were washed 3 times in TBS-T buffer after which the proteins were visualized using a chemiluminescence detection system (ChemiDoc MP, Bio-Rad). As a loading control β-actin was used. The data were quantified using Image Lab software and normalized to the expression β-actin. Three independent biological replicates were carried out for all experiments, except for those with HepG2-Sl-Clon1 and HepG2-Sl-Clon2, as these cells lost their Slug over-expression after a few passages.

qRT-PCR

In order to determine the mRNA expression levels of ABCB1, ABCC1, ABCC2, ABCC3, ABCC5, ABCG2, CD133, COL2A1, FGG, FN1, SLUG and ZEB2 in stably transfected cells and the COL2A1, FGG, FN1 and SLUG mRNA expression levels in transiently transfected cells, total RNA was isolated using a Nucleospin RNA kit (Macherey-Nagel, Düren, Germany) according to the directions of the manufacturer. Next, cDNA synthesis was carried out using a Protoscript II kit (NEB, Ipswich, USA) according to the manufacturer’s protocol after which quantitative real-time PCR (qRT-PCR) amplification was carried out using a Brilliant SYBR Green qPCR master mix (Agilent Technologies, Palo Alto, CA, USA) in conjunction with a real-time PCR platform (Mx3005P, Agilent Technologies, Palo Alto, CA, USA). The mRNA expression data were normalized to ACTB and analyzed using an efficiency corrected relative quantification method as reported before [29]. The primers were obtained from Sentegen (Ankara, Turkey) and their sequences are listed in Table 1.

Table 1.

Primer sequences used for qRT-PCR

Gene Name (Accession Number)	Forward Primer	Reverse Primer
ABCB1 (NM_000927.4)	5’-TGACATTTATTCAAAGTTAAAAGCA-3′	5’-TAGACACTTTATGCAAACATTTCAA-3′
ABCC1 (NM_004996.3)	5’-AGTGGAACCCCTCTCTGTTTAAG-3′	5’-AACAGCAGCACGGTGTAGAA-3′
ABCC2 (NM_000392.4)	5’-TCCTTGCGCAGCTGGATTACAT-3′	5’-TCGCTGAAGTGAGAGTAGATTG-3′
ABCC3 (NM_003786.3)	5’-CAGAGAAGGTGCAGGTGACA-3’	5’-CTAAAGCAGCATAGACGCCC-3’
ABCC5 (NM_001320032.1)	5’-AGAGGTGACCTTTGAGAACGCA-3’	5’-CTCCAGATAACTCCACCAGACGG-3’
ABCG2 (NM_001348989.1)	5’-CCGCGACAGCTTCCAATGACCT-3’	5’-GCCGAAGAGCTGCTGAGAACTGTA-3’
ACTB (NM_001101.3)	5’-GAGCGCGGCTACAGCTT-3’	5’-TCCTTAATGTCACGCACGATTT-3’
CD133 (NM_001145847.1)	5’-AGTCGGAAACTGGCAGATAGC-3’	5’-GGTAGTGTTGTACTGGGCCAAT-3’
COL2A1 (NM_001844.4)	5’-AACCAGATTGAGAGCATCCG-3’	5’-AACGTTTGCTGGATTGGGGT-3’
FGG (NM_000509.5)	5’-CAGGATGGATCTGGTTGGTGG-3’	5’-GGAGTAGATGCTTTTGAGTAAGTGC-3’
FN1 (NM_212474.2)	5’-TGGGCAACTCTGTCAACGAA-3’	5’-CCACTCATCTCCAACGGCAT-3’
SLUG (NM_003068.4)	5’-CGAACTGGACACACATACAGTG-3’	5’-CTGAGGATCTCTGGTTGTGGT-3’
ZEB2 (NM_001171653.1)	5’-ATCTGAAAGAACACCTGCGAAT-3’	5’-TGTAGGAACCAGAATGGGAGAA-3’

Neutral red uptake assay

EMT is generally considered to be associated with drug resistance [30]. In in order to determine the concentration of cisplatin required to kill 50% of the cells (IC₅₀ values) undergoing complete or partial EMT, a neutral red uptake assay was carried out as reported before [31]. Briefly, HepG2-Sl-Clon1, HepG2-Sl-Clon2, Huh7-Sl-Clon1, Huh7-Sl-Clon2, HepG2-EV and Huh7-EV cells were seeded at densities of 10,000 cells/well in 96-well culture plates (Thermo Scientific, Germany) in a total volume of 100 μl. The cells were allowed to adhere overnight and subsequently treated with different concentrations (1.5–100 μM) of cisplatin for 72 h. Next, the media were removed after which 200 μl fresh medium containing 50 μg/ml neutral red was added to each well and incubated for an additional 3 h. The dye was then removed and each well was washed rapidly with 200 μl phosphate buffer saline (PBS) followed by the addition of 100 μl dye extraction solution (acetic acid-water-ethanol 1:49:50). Next, the plates were incubated for 15 min at room temperature with shaking on a GFL 3012 shaker (Gesellschaft für Labortechnik mbH, Burgwedel, Germany) to extract the dye. Finally, the absorbance was measured using an ELx 808 Ultra Microplate Reader (BioTek Instruments, Winooski, VT, USA) equipped with a 540 nm filter. Each assay was repeated at least three times. Viability was evaluated through comparison with untreated cells. Drug resistance was evaluated by comparing the IC₅₀ values of control and Slug over-expressing cells.

Doxorubicin accumulation assay

To determine the accumulation of doxorubicin in Slug over-expressing cells, HepG2-Sl-Clon1, HepG2-Sl-Clon2, Huh7-Sl-Clon1 and Huh7-Sl-Clon2 cells, along with control empty vector transfected cells were seeded in each well of a 6-well cell culture plate (3 × 10⁵ cells per well) onto 35 mm sterile coverslips (Thermo Scientific) and allowed to adhere overnight. Subsequently, the cells were incubated with 0.18 μM doxorubicin (Sigma-Aldrich) for 24 h, after which they were fixed with 4% paraformaldehyde and washed with PBS. Next, the cells were treated with 1% Triton X-100 solution to permeabilize, washed with PBS, incubated with 0.05 μg/ml DAPI solution and washed again with PBS. Subsequently, the coverslips were removed from the wells and mounted on glass slides with 50% glycerol, after which images were captured using a fluorescence microscope (Olympus, BX 50, Center Valley, PA, USA). Data were quantified using ImageJ software (NIH, Bethesda, MD, USA) and doxorubicin signals were normalized to DAPI signals.

Scratch wound healing assay

To determine cell motility, a scratch wound healing assay was carried out. For this, Huh7-EV and Huh7-Sl-Clon1 cells were seeded in 6-well cell culture plates (Thermo Scientific) and cultured until 100% confluence. Next, scratches were made with sterile 200 μl pipette tips after which the wounded monolayers were washed immediately twice with PBS and incubated in serum free medium. Wound healing rates were determined by comparing images captured after 48 and 96 h with the those captured at 0 h using an inverted microscope (Olympus, BX 50, Center Valley, PA, USA). The images were quantified using ImageJ software (NIH, Bethesda, MD, USA).

Exosomal extracellular protein isolation

For secretome analyses of Slug over-expressing cells, exosomal extracellular proteins were isolated. Exosome isolations were carried out according to a differential ultracentrifugation protocol as reported before [32]. Briefly, Huh7-EV and Huh7-Sl-Clon1 cells were seeded at a density of 750,000 cells in 75 cm² flasks. When the cells reached 70% confluence they were serum starved for 24 h, after which 90 ml conditioned medium (CM) was collected. The CM was serially centrifuged at 300×g for 10 min, at 2000×g for 10 min, at 10, 000×g for 30 min and, finally, at 100,000×g for 70 min in a Beckman Type 45 Ti rotor using a Beckman Optima LE 80 K ultracentrifuge. The supernatants were discarded and the exosome pellets were dissolved in a Laemmli sample buffer (0.02% Bromophenol blue, % 4 SDS, 20% Glycerol, 120 mM Tris-HCl, pH 6.8) yielding extracellular proteins. The temperature was maintained at 4 °C during all steps.

MALDI-TOF/TOF assay

The isolated exosomal extracellular proteins were subjected to SDS-PAGE using standard protocols (see above). The gels were stained with Coomassie Brilliant Blue G250 and documented using a Gel Doc EZ system (Bio-Rad Richmond, CA, USA). Further analyses were carried out by the selection of differentially staining bands. These bands were excised from the gels and subjected to zip-tip and MALDI TOF/TOF analyses as previously reported [33].

ELISA assay

The upregulation of proteins (as shown by MALDI-TOF/TOF) in the secretome of Slug over-expressing cells was confirmed by ELISA. In the stably transfected group, the culture media of Huh7-Sl-Clon1 and control cells grown to confluence were replaced with serum-free media after which the respective cells were incubated for an additional 24 h and collected. In the transiently transfected group, the culture media of HepG2-Sl (Tr), Huh7-Sl (Tr) and control cells were collected after 72 h of transfection. The resulting conditioned media (CM) were centrifuged at 1000×g for 20 min at 4 °C after which the amounts of the proteins of interest were determined using ELISA kits according to the manufacturer’s instructions. Respective ELISA kits were used to determine fibronectin 1 (FN1, Abcam, Cambridge, UK), collagen type II alpha 1 (COL2A1, Elabscience, China) and fibrinogen gamma chain (FGG, Aviva Systems Biology, San Diego, CA, USA). For quantification, the plates were read in a plate reader (Epoch2, BioTek Instruments, Winooski, VT, USA).

Bioinformatics and statistical analyses

Normalized expression data of E-GEOD-69667 were downloaded from the Array Express Database (https://www.ebi.ac.uk/arrayexpress/experiments/E-GEOD-69667/) as TPM values. COL2A1, FN1 and FGG expression data were extracted using R software version 4.4.3, and presented as fold changes. Three independent biological replicates were carried out, unless indicated otherwise. Statistical analyses and graphing were performed using Graphpad (GraphPad Prism 6, San Diego, CA, USA). One-way ANOVA with Dunnet post hoc test or Student t-test were used for differential analyses, and p-values < 0.05 were considered to be statistically significant.

Results

Slug over-expression results in complete EMT in HepG2 cells and partial EMT in Huh7 cells

Based on the analysis of Slug over-expression in HepG2 and Huh7 transfected cells using an in-cell Western assay (see materials and methods), two clones from each cell line were selected, i.e., HepG2-Sl-Clon1, HepG2-Sl-Clon2 and Huh7-Sl-Clon1 and Huh7-Sl-Clon2, respectively (Fig. 1A). Slug over-expression in the selected cell lines was confirmed by Western blotting and qRT-PCR. The HepG2-S1-Clon1 cells showed a 7 and 12-fold upregulation of Slug protein and mRNA, respectively, whereas the Huh7-Sl-Clon1 cells showed a 9 and 17-fold upregulation of Slug protein and mRNA, respectively (Fig. 1B). Next, the expressions of the canonical EMT markers E-cadherin, vimentin and ZEB2 were examined. We found that the E-cadherin protein expression level was 3-fold downregulated and that the ZEB2 mRNA expression level was 132-fold upregulated in Slug over-expressing HepG2 cells, indicating a successful induction of complete EMT in these cells. Interestingly, we could not detect any vimentin expression in the HepG2 cells (Fig. 1C). In Slug over-expressing Huh7 cells we observed a 3-fold upregulation of E-cadherin protein, a 33-fold downregulation of ZEB2 mRNA and 20-fold downregulation vimentin protein expression. Together, these data indicate that Slug over-expression induces a complete EMT in HepG2 cells and a partial EMT in Huh7 cells.

Fig. 1.

Expression of EMT markers in Slug over-expressing cells. A Selection of Slug over-expressing clonal cell lines by in-cell Western analysis. The signals from the 700 nm channel indicate cell numbers and the signals from the 800 nm channel indicate Slug primary antibody binding to secondary antibody. The first lane was used for background subtraction and the second lane was used for data analysis in both channels. As the Slug expression increases the color in the merged lane changes from orange to green. The Slug protein expression level was significantly higher in SlugMyc_pcDNA3 transfected Huh7 and HepG2 cells (ICW: in-cell Western). The experiment was carried out using three independent biological samples. B Relative protein expression of Slug, vimentin and E-cadherin in HepG2-EV, HepG2-Sl-Clon1, Huh7-EV and Huh7-Sl-Clon1 cells determined by Western blotting. Slug was upregulated in HepG2-Sl-Clon1 cells and Huh7-Sl-Clon1 cells. E-cadherin was downregulated in HepG2-Sl-Clon1 cells and upregulated in Huh7-Sl-Clon1 cells. Vimentin expression could not detected in HepG2 cells and was downregulated in Huh7-Sl-Clon1 cells. C Relative mRNA expression of SLUG and ZEB2 in HepG2-EV, HepG2-Sl-Clon1, Huh7-EV and Huh7-Sl-Clon1 cells determined by qRT-PCR. Slug was upregulated in HepG2-Sl-Clon1 cells and Huh7-Sl-Clon1 cells. ZEB2 was upregulated in HepG2-Sl-Clon1 cells and downregulated in Huh7-Sl-Clon1 cells. Since the HepG2-Sl-Clon1 cells lost the overexpression of Slug after passaging, the experiments were carried out only once in this group. All experiments with Huh7-Sl-Clon1 cells were performed as three independent biological replicates. Statistical comparisons were carried out using Student t-test or ANOVA followed by Dunnet’s post hoc (*p < 0.05, **p < 0.01, ***p < 0.001)

Since the Slug over-expressing Huh7 cells did not show the classical signs of EMT induction, we decided to examine the morphologies of the respective cells under an inverted microscope. While Slug over-expressing HepG2 cells showed a clear fibroblast-like morphology, none of the Slug over-expressing Huh7 cells examined showed a classical spindle shaped morphology characteristic of cells undergoing EMT (Fig. 2). EMT is known to be associated with a loss of cell-cell adhesion molecules, along with the acquisition of a spindle-like morphology, which implies loss of cellular polarity and eventually enhanced migration [2]. Since at present limited data are available regarding the motility of cells undergoing partial EMT, we set out to investigate the motility of the HCC cells undergoing partial EMT using a scratch wound healing assay. The data obtained indicate that the wound closure rate was relatively high in Slug over-expressing Huh7 cells, suggesting that also partial EMT results in an enhanced motility of these cells (Fig. 3). As this experiment was carried out in serum-reduced medium, the contribution of cell proliferation to the wound closure rate was likely to be low.

Fig. 2.

Morphology of transiently transfected HepG2 and Huh7 cells. Arrows indicate a fibroblast-like morphology of Slug over-expressing HepG2 cells. Similar spindle shaped fibroblast-like morphologies were not seen in any of the Slug over-expressing Huh7 cells

Fig. 3.

Scratch wound healing of Huh7-Sl-Clon1 and Huh7-EV cells. Huh7-Sl-Clon1 cells were significantly more motile than Huh7-EV cells. The experiments were carried out using three independent biological replicates. Statistical comparisons were carried out using Student t-test (***p < 0.001)

Slug over-expression results in chemo-resistance of cells undergoing complete or partial EMT

As the process of EMT is often accompanied by enhanced drug resistance [34], we next set out to analyze the effect of Slug over-expression on cisplatin resistance using a neutral red uptake assay. IC₅₀ values were calculated by comparing the viabilities of cells that were exposed to 2-fold serially diluted cisplatin concentrations. While the IC₅₀ of cisplatin for HepG2-EV cells was found to be 11.2 μM, we found that it was increased to 27.16 μM and 19.4 μM for HepG2-Sl-Clon1 and HepG2-Sl-Clon2 cells, respectively. In case of Huh7 cells, we found that the IC₅₀ of cisplatin for Huh7-EV cells was 11.8 μM, and that it was increased to 23.4 μM and 15.65 μM for Huh7-Sl-Clon1 and Huh7-Sl-Clon2 cells, respectively (Fig. 4). These increases in IC₅₀ values suggest that chemo-resistance was developed by HCC cells undergoing both partial and complete EMT.

Fig. 4.

Chemo-resistance of cells undergoing complete and partial EMT. Neutral-red uptake indicates an increase in IC50 value for cisplatin in all Slug over-expressing cells lines. The experiments were conducted using three independent biological replicates. Statistical comparisons were carried out using ANOVA followed by Dunnet’s post hoc (***p < 0.001)

The impact of Slug over-expression on chemo-resistance was also determined using a doxorubicin accumulation assay. This assay is based on the premise that drug resistant cells pump out doxorubicin, leading to less accumulation of doxorubicin within the cells which can be monitored using fluorescence microscopy. We found that Slug over-expression resulted in a significantly decreased doxorubicin accumulation in both Huh7 and HepG2 cells (Fig. 5).

Fig. 5.

Doxorubucin accumulation in cells undergoing complete and partial EMT. Intracellular doxorubucin levels were decreased in both cell types indicating enhanced chemo-resistance in the different Slug over-expressing cells. The experiments were done using three independent biological replicates. Statistical comparisons were carried out using ANOVA followed by Dunnet’s post hoc (**p < 0.01, ***p < 0.001)

Additionally, we investigated a putative mechanism of chemo-resistance development by determining the expression of a series of drug transporters, i.e., ABCB1, ABCC1, ABCC2, ABCC3, ABCC5 and ABCG2 using qRT-PCR. We found that the expression of 4 of these transporters (ABCC1, ABCC2, ABCC3 and ABCC5) was not significantly affected (data not shown). In contrast, we found that the expression of ABCG2 was 10 and 7-fold upregulated in HepG2-Sl-Clon1 and Huh7-Sl-Clon1 cells, respectively. Unexpectedly, we found that the expression of ABCB1 was 7 and 3-fold downregulated in HepG2-Sl- Clon1 and Huh7-Sl- Clon1 cells, respectively (Fig. 6).

Fig. 6.

qRT-PCR analysis of ABCB1 and ABCG2 mRNA levels in cells undergoing complete and partial EMT. Since the HepG2-Sl-Clon1 cells lost Slug over-expression after passaging, the experiments were carried out only once in this group. All experiments with Huh7-Sl-Clon1 cells were performed as three independent biological replicates. Statistical comparisons were carried out using Student t-test (*p < 0.05)

Slug over-expression results in stemness in cells undergoing complete or partial EMT

It has been reported that Slug over-expression may result in stemness phenotypes in ovarian [35] and hepatocellular carcinoma cells [9]. In addition, chemo-resistance has also been reported to be associated with stemness phenotypes [35–37]. Based on these findings, we set out to investigate the expression of CD133, a stemness marker, in HCC cells undergoing complete and partial EMT. We found that CD133 protein was upregulated 4 and 5-fold and that CD133 mRNA was upregulated 1.5 and 6-fold in HepG2-Sl-Clon1 and Huh7-Sl-Clon1 cells (Fig. 7A, B), respectively, indicating that stemness may be induced by Slug regardless the acquisition of complete or partial EMT.

Fig. 7.

Expression of CD133, a stemness marker, in Slug over-expressing cells. A Relative CD133 protein expression in HepG2-EV, HepG2-Sl-Clon1, Huh7-EV and Huh7-Sl-Clon1 cells determined by Western blotting. CD133 protein was upregulated in HepG2-Sl-Clon1 cells and Huh7-Sl-Clon1 cells. B Relative CD133 mRNA expression in HepG2-EV, HepG2-Sl-Clon1, Huh7-EV and Huh7-Sl-Clon1 cells determined by qRT-PCR. CD133 mRNA was upregulated in HepG2-Sl-Clon1 cells and Huh7-Sl-Clon1 cells. Since the HepG2-Sl-Clon1 cells lost Slug over-expression after passaging, the experiments were carried out only once in this group. All experiments with Huh7-Sl-Clon1 cells were performed as three independent biological replicates. Statistical comparisons were carried out using Student t-test (*p < 0.05, **p < 0.01)

FN1, COL2A1 and FGG as partial EMT biomarkers in HCC

Non-invasive biomarkers that can act as signatures of chemo-resistance may be of great importance for clinical cancer management and/or the prediction of recurrence [38]. Since so far partial EMT has been less studied than complete EMT, we set out to identify possible partial EMT biomarkers in HCC cell culture supernatants. For this, we isolated exosome proteins from conditioned media of Huh7-Sl-Clon1 and Huh7-EV cells. Subsequent MALDI-TOF/TOF analysis of differentially expressed exosome proteins revealed the presence of three proteins that were highly enriched in Slug over-expressing Huh7 cells (Fig. 8A). These proteins were identified as fibronectin1 (FN1) with a molecular weight > 262 kDa, collagen type II alpha 1 (COL2A1) with a molecular weight of 210 kDa and fibrinogen gamma chain (FGG) with a molecular weight of 50 kDa (Fig. 8B). Since the predicted molecular weights of FN1, COL2A1 and FGG are 262 and 141 and 50 kDa, respectively, we assume that FN1 and COL2A1 are post-translationally modified and that the respective modified protein levels are elevated in the HCC cells undergoing partial EMT. FGG did not appear to undergo any modification in these cells. Subsequent qRT-PCR analyses revealed a 2.8 and 5.8-fold upregulation of respectively COL2A1 and FGG mRNA in Huh7-Sl-Clon1 cells compared to its corresponding empty vector transfected cells. In contrast, we observed a 2-fold downregulation of FN1 mRNA (Fig. 8C). We also assessed the proteins levels of these markers in the conditioned media using ELISA. By doing so, we found that the total (modified and unmodified) COL2A1 and FN1 protein levels were significantly upregulated in Slug over-expressing Huh7-S1-Clon1 cells compared to its corresponding empty vector transfected cells. The FGG levels could not be determined in the conditioned media since they were below the detection limit of the kit used (Fig. 8C).

Fig. 8.

Biomarker screening for chemoresistance and partial EMT from the secretome of Slug over-expressing cells. A SDS-PAGE analysis of molecular weight marker (M), exosomal and extracellular proteins isolated from Huh7-EV (EV) and Huh7-Sl-Clon1 (Sl) cells. A representative image of differentially expressed proteins is shown. B Empty rectangles indicate the bands that were excised and subjected to zip-tip and MALDI-TOF/TOF analyses. C ELISA of COL2A1, FGG and FN1 from conditioned media. COL2A1 and FN1 protein expression were increased in Huh7-Sl-Clon1 cells compared to control Huh7-EV cells. The FGG level could not be measured due to its low abundance. D qRT-PCR analysis of COL2A1, FGG and FN1 mRNA. COL2A1 and FGG mRNA expression were increased in Huh7-Sl-Clon1 cells compared to control Huh7-EV cells. A decrease in FN1 mRNA was observed in Huh7-Sl-Clon1 cells compared to control Huh7-EV cells. All experiments were carried out using three independent biological replicates. Statistical comparisons were carried out using Student t-test (*p < 0.05, **p < 0.01, ***p < 0.001)

In order to additionally compare the expression levels of COL2A1, FGG and FN1 in HCC cells undergoing both complete and partial EMT, we transiently transfected HepG2 and Huh7 cells with the Slug over-expression plasmid. Subsequent ELISA analyses of the conditioned media from these cells revealed a statistically significant upregulation of the total (modified and unmodified) COL2A1 and FN1 protein levels in both Slug over-expressing HepG2 and Huh7 cells compared to their corresponding empty vector transfected cells (Fig. 9A). Similar to the stably Slug transfected Huh7 cells, the FGG levels in both transiently Slug over-expressing HepG2 and Huh7 cells could not be determined in the conditioned media as its level was below the detection limit of the kit used. qRT-PCR analyses did not reveal any changes in the expression of the COL2A1 and FGG mRNAs, whereas a 10-fold downregulation of FN1 mRNA was observed in Slug transfected HepG2-Sl (Tr) cells. In contrast, we observed a 15 and 14-fold upregulation of the COL2A1 and FGG mRNAs, respectively, and a 10-fold downregulation of the FN1 mRNA in Slug transfected Huh7-Sl (Tr) cells compared to its corresponding empty vector transfected cells (Fig. 9B).

Fig. 9.

A ELISA analysis of FN1 and COL2A1 proteins in conditioned media. COL2A1 and FN1 protein expression were increased in HepG2-Sl (Tr) and Huh7-Sl (Tr) cells compared to the respective control cells. B qRT-PCR analysis of COL2A1, FGG and FN1 mRNAs. Upregulated COL2A1 and FGG and down regulated FN1 levels were observed in Huh7-Sl (Tr) derived media. HepG2-Sl (Tr) cells showed a stable COL2A1 and FGG expression and a downregulated FN1 expression. C Analysis of a publicly available RNA sequencing dataset (GSE69667) revealed a higher COL2A1 expression in cells exhibiting a partial EMT phenotype (36 h) compared to cells with either epithelial (0 h) or mesenchymal (72 h) phenotypes. All experiments were performed using three independent biological replicates. Statistical comparisons were carried out using Student t-test (*p < 0.05, **p < 0.01, ***p < 0.001)

To further substantiate the roles of FN1, COL2A1 and/or FGG in the induction of either partial or total EMT, we examined a publicly available RNA sequencing dataset (GSE69667) [39]. In this study, the authors induced EMT in A549 lung adenocarcinoma cells with TGFβ in a time course of 0–96 h, and they reported that within a timeframe of 12–36 h the cells exhibited a partial epithelial/mesenchymal phenotype, whereas more epithelial and more mesenchymal characteristics were apparent at the earlier and later time points, respectively. Data for the 0, 36 and 72 h time points, representing epithelial, partial epithelial/mesenchymal and mesenchymal phenotypes, were extracted for analyses of our genes of interest. By doing so, we found that the COL2A1 mRNA expression level was increased 3-fold in the cells exhibiting a partial epithelial/mesenchymal phenotype compared to those exhibiting epithelial and mesenchymal phenotypes (Fig. 9C). Analysis of the same dataset revealed that FN1 was highly expressed at all time points, with a slight increase in expression at the later time points (Fig. 9C). We also found a decrease in the expression of FGG at all time points (data not shown) indicating that, unlike in HCC, this gene may not serve as an EMT biomarker in lung adenocarcinoma.

Discussion

Chemo-resistance has frequently been associated with the over-expression and activation of EMT markers such as the transcription factor Slug in a variety of cancer types [35, 40–42]. Here, we exogenously over-expressed Slug in two different HCC-derived cell lines and report three primary events: (1) induction of complete and partial EMT in HepG2 and Huh7 cells, respectively, (2) induction of chemo-resistance and stemness in HCC cells undergoing both partial and complete EMT and (3) upregulation of post translationally modified fibronectin 1, collagen type II alpha 1 and native fibrinogen gamma chain, proteins that may be considered as potential biomarkers of chemo-resistance and partial EMT in HCC.

In HepG2 cells undergoing complete EMT the mRNA expression level of the mesenchymal marker ZEB2 was markedly upregulated whereas that of the epithelial marker E-cadherin was downregulated, in conformity with a canonical EMT program. These findings are consistent with a previous report [9]. Surprisingly, however, we found that ZEB2 and vimentin were downregulated while E-cadherin was upregulated in Slug expressing Huh7 cells undergoing partial EMT. Some literature reports have suggested an involvement of upregulated E-cadherin in EMT, and a transient upregulation of E-cadherin during partial EMT in HGF-induced (non-malignant) canine kidney MDCK cells has been noted [11]. In addition, it has been found that some advanced carcinomas can adopt mesenchymal features while retaining characteristics of well-differentiated epithelial cells [12]. It has also been suggested that murine EpH4 mammary tumor cells may be more aggressive and/or invasive without the involvement of classical epithelial or mesenchymal markers such as vimentin and E-cadherin [43]. In addition, it has been found that over-expression of a matrix metalloprotease can promote the invasion of MDCK cells in vivo, indicating that these non-malignant epithelial cells may become invasive through a partial EMT induction [44].

The repression of E-cadherin has been reported to be initiated by members of the Snail family and completed by members of the ZEB family of proteins [45]. Most other genes that are repressed during EMT have, however, been found to be inhibited by ZEB family members, irrespective of the EMT-inducing signal [46]. In the current work, one of the main differences between Slug-mediated complete and partial EMT may result from differences in ZEB2 expression. We found that Slug-mediated downregulation of ZEB2 in Huh7 cells results in partial EMT, suggesting that ZEB2 may be indispensable for the completion of EMT. Very recently, it has been reported that the transcription factor FOXO1 may reverse EMT in HCC cells by regulating the expression of EMT markers and transcriptional activators [47]. Over-expression of FOXO1 was found to repress the expression of ZEB2 without affecting the expression of Slug in Huh7 cells, again indicating a pivotal role of ZEB2 in partial EMT induction. Moreover, ZEB1 and ZEB2 have been found to be regulated by MYC-associated zinc finger protein (MAZ) in HCC cells. Silencing of MAZ resulted in a significant inhibition of HCC cell proliferation, tumorigenesis, invasion and migration [48], which further supports a role for ZEB2 in the decision to complete the EMT program. It has been reported that ZEB2 over-expression may significantly enhance cell motility, invasiveness and vasculogenic mimicry in hepatocellular carcinoma HepG2 cells through upregulation of the expression of VE-cadherin, Flt-1 and Flk-1 and activation of MMPs [49]. Based on these results, as well as those from our current study, we suggest that ZEB2 may act as a key regulator in the decision-making process for partial or complete EMT after Slug-induced initiation.

Chemo-resistance is a complex and dynamic process that is regulated through various mechanisms, including the activation of anti-apoptotic pathways, the expression of ABC transporters that can pump drugs out of the cell and DNA repair systems that can block the effects of DNA damage-inducing therapeutic agents [50]. It has also been found that chemo-resistance may be associated with EMT in pancreatic cancer [34] and prostate cancer [51]. As yet, however, chemo-resistance in the context of complete and partial EMT has been poorly studied. Here, we show that chemo-resistance can occur in HCC cells undergoing both complete and partial EMT. Chemo-resistance was evaluated by comparing the IC₅₀ values of cisplatin, the accumulation of doxorubicin and the expression of ABC transporters in cells that showed either partial or complete EMT. Slug over-expression was found to lead to the development of resistance to cisplatin in cells undergoing both complete and partial EMT. The accumulation of doxorubicin was found to be decreased, implying the presence of more effective efflux pumps, which was corroborated by the observation of a significant increase in the expression of the ABC transporter ABCG2. These results are consistent with previous findings indicating that EMT and chemo-resistance are interrelated in HCC cells [52, 53]. Our data indicate that both complete and partial EMT may lead to the development of chemo-resistance, at least in part through upregulation of the ABCG2 transporter in HCC cells. The expression of ABCB1 was found to be downregulated in cells undergoing both complete and partial EMT, whereas the overall resistance to cisplatin was increased more than 2-fold in cells undergoing both types of EMT. Therefore, it may be suggested that Slug fine-tunes chemo-resistance by regulating the expression of the ABCB1 and ABCG2 transporters in HCC. Several other studies have reported similar data suggesting that Slug over-expression may enhance chemo-resistance in different cancer types [54–56]. Conversely, it has been suggested that Slug inhibition may enhance multidrug resistance through the upregulation of ABCB1 and ABCG2 in HCC-derived MHCCLM3 and SMMC-7721 cells [57]. Together, these results imply an EMT-mediated plasticity and heterogeneity in HCC. Two cancer cell lines originating from the same primary tumor site, but propagated in different cellular contexts, may exhibit different responses to the same effectors. Together with this notion, we can conclude that drug resistance in HCC may be affected by altering the expression of Slug. Up- or downregulation of Slug expression may result in enhanced or inhibited chemo-resistance depending on the cellular context. Hence, a well-defined chemo-resistance profile of a given tumor seems to be a prerequisite for considering a possible Slug targeted therapy.

CD133 is a biomarker that is used to isolate HCC stem-like cells. CD133 positive cancer cells have been reported to exhibit higher proliferative, self-renewal and differentiation capacities compared to its negative counterparts [58]. CD133 positive cells may also exhibit an increased colony-forming efficiency in soft agar in vitro and an increased ability to form tumors in vivo [59]. It has been shown that liver cancer patients with high CD133 expression levels may exhibit shorter overall survival and higher relapse rates than those with low CD133 expression levels [60]. It has recently also been reported that in inflammatory liver microenvironments normal liver stem cells may transform to EMT-positive metastatic cancer stem cells [61]. Here, we found that CD133 upregulation was similar in cells undergoing complete and partial EMT, increasing both the motility and chemo-resistance of HCC cells.

We next aimed to identify potential chemo-resistance and partial EMT biomarkers in conditioned media from Slug over-expressing HCC cells using MALDI-TOF/TOF. The biomarkers that we identified were: post-translationally modified fibronectin1 (FN1), collagen type II alpha 1 (COL2A1) and native fibrinogen gamma chain (FGG). We conclude that these three proteins may serve as potential biomarkers for chemo-resistance and partial EMT in HCC. Interestingly, we found that FN1 mRNA was downregulated in HepG2 cells transiently over-expressing Slug and Huh7 cells stably and transiently over-expressing Slug, whereas the total (native and post-translationally modified) FN1 protein level was increased in the conditioned media. We conclude that upregulation of the FN1 protein may serve as a biomarker for both partial and complete EMT in HCC, although the fold increase in FN1 secretion in Huh7 cells undergoing partial EMT was 2.4 compared to a 1.1-fold increase in HepG2 cells undergoing complete EMT. These observations were substantiated using a publicly available RNA-sequencing dataset (GSE69667) showing that the expression of FN1 was consistently high in A549 lung adenocarcinoma cells, irrespective their epithelial, partially epithelial/mesenchymal or mesenchymal status. Here, we show that chemo-resistance in HCC cells may be associated with partial EMT, and that FN1 may serve as a chemo-resistance biomarker regardless of EMT status. Fibronectin is a large glycoprotein whose primary role is to attach cells to the extracellular matrix [62]. It may form dimers through disulfide bridge formation at the C-terminus. Each chain is approximately 2350 amino acids long and contains six domains harboring specific binding sites for integrins, sulfated proteoglycans, fibrin, denatured forms of collagen and DNA [63]. Human FN harbors seven potential N-glycosylation sites [64]. Alterations in glycan structure and aberrant expression of FN are known to be involved in the proliferation, migration and invasion of cancer cells [65–68]. FN1 may be implicated in increased invasiveness of cells showing partial EMT since it is known that the oligosaccharide part of FN1 markedly enhances its affinity to gelatin, and promotes the adhesion and spread of fibroblasts [64, 69]. The post-translational addition of oligosaccharides to FN1 may also protect the molecule from proteolytic degradation, thereby enhancing its half-life [64]. These FN1 modifications may explain our contradictory findings of decreases in FN1 mRNA levels at one hand, whereas on the other hand post-translational modifications of the protein (such as glycosylation) may enhance its half-life as well as enhance its secretion by cells undergoing both partial or complete EMT.

In the past, oncogenic alterations of COL2A1 have been reported [70–72]. It has also been reported that the initiator methionine tRNA may upregulate the secretion of type II collagen from stromal fibroblasts to promote tumor growth and angiogenesis [70]. Also, type II collagen has been defined as a specific biomarker for mesenchymal chondrosarcomas [72]. Here we found that, while total and post-translationally modified COL2A1 protein and its mRNA levels were increased by 20 and 15-fold, respectively, in transiently Slug over-expressing cells exhibiting partial EMT, only a 1.5-fold increase in secreted Col2A1 protein and no change in its mRNA level were noted in HepG2 cells undergoing complete EMT. Based on these findings, we conclude that a dramatic upregulation of COL2A1 mRNA and protein may serve as a biomarker for partial EMT. This conclusion is substantiated by our analysis of a publicly available RNA sequencing dataset in A594 cells where a substantial increase in COL2A1 expression was seen in cells undergoing partial EMT compared to cells with epithelial or mesenchymal phenotypes.

Fibrinogen is a plasma glycoprotein synthesized in the liver that is composed of three structurally different subunits, i.e., alpha (FGA), beta (FGB) and gamma (FGG). Fibrinogen gamma (FGG) and the plasma level of fibrinogen have been reported to be abnormal in HCC patients and it has been claimed that elevated plasma fibrinogen levels may serve as useful predictors of disease progression [73]. Here, we found that the FGG mRNA levels were significantly increased in HCC cells undergoing partial EMT. Future studies are needed to establish whether post-translationally modified FN1 and COL2A1 protein levels and native FGG protein levels are also elevated in sera of patients exhibiting partial EMT, thereby firmly establishing the potential of these proteins to serve as biomarkers of partial EMT and chemo-resistance in HCC.

Author contributions

Oğuzhan Karaosmanoğlu carried out the experiments and wrote the manuscript, Sreeparna Banerjee supervised the research and wrote the manuscript, and Hülya Sivas was responsible for the overall supervision of the research.

Compliance with ethical standards

Human and animal studies

The authors did not perform experiments involving human participants or animal models in this study.