All-Trans Retinoic Acid-Induced Deficiency of the Wnt/β-Catenin Pathway Enhances Hepatic Carcinoma Stem Cell Differentiation

Retinoic acid (RA) is an important biological signal that directly differentiates cells during embryonic development and tumorigenesis. However, the molecular mechanism of RA-mediated differentiation in hepatic cancer stem cells (hCSCs) is not well understood. In this study, we found that mRNA expressions of RA-biosynthesis-related dehydrogenases were highly expressed in hepatocellular carcinoma. All-trans retinoic acid (ATRA) differentiated hCSCs through inhibiting the function of β-catenin in vitro. ATRA also inhibited the function of PI3K-AKT and enhanced GSK-3β-dependent degradation of phosphorylated β-catenin. Furthermore, ATRA and β-catenin silencing both increased hCSC sensitivity to docetaxel treatment. Our results suggest that targeting β-catenin will provide extra benefits for ATRA-mediated treatment of hepatic cancer patients.


Introduction
Cancer stem cells (CSCs) represent a small but unique population of cancer cells that can sustain their self-renewal and multipotentiality. These cells are also considered cancer-initiating cells (CICs) because their tumorigenic potency allows CSCs/CICs to form de novo tumors [1][2]. Clinical studies suggest that CSCs/CICs are resistant to traditional chemo-or radio-therapies due to their specialized microenvironments [3][4]. It has been proposed that both microenvironment-providing signals and their intracellular molecule transductions are essential for protecting and maintaining CSCs/CICs. Several important pathways such as Notch, Hedgehog and Wnt have been identified in the maintenance or differentiation of CSCs/CICs (5).
Liver cancer is one of the most common and lethal cancers worldwide [5][6]. Unfortunately, clinical therapies for liver cancer are still not available due to the heterogeneous nature of the cellular components within the tumors [5]. In hepatocellular carcinoma (HCC), cell membrane proteins such as CD133, CD13, CD90, EpCAM and CD44 have been widely applied to isolate analysis, decision to publish, and preparation of the manuscript. National Natural Science Foundation of China (No. 31160237), http://www.nsfc.gov.cn/, study design, data collection and analysis, decision to publish, and preparation of the manuscript. National Natural Science Foundation of China No. 31260276), http://www.nsfc.gov.cn/, study design, data collection and analysis, decision to publish, and preparation of the manuscript. Yunnan Province Science and Technology Innovation Team (No. 2011CI123), http:// www.ynstc.gov.cn/, study design, data collection and analysis, decision to publish, and preparation of the manuscript. Talent Program of Yunnan Province (No. W8110305), http://www.ynstc.gov.cn/ study design, data collection and analysis, decision to publish, and preparation of the manuscript. Foundation of School of Life Sciences of Yunnan University (No. 2012S301), http://www.ynusky.ynu.edu.cn/, study design, data collection and analysis, decision to publish, and preparation of the manuscript. Natural Science Foundation of Yunnan Province (No. 2011BC003), http://www.ynstc.gov.cn/, study design, data collection and analysis, decision to publish, and preparation of the manuscript. antibodies for 10 minutes at 4°C. The stained cells were further analyzed by an Accuri C6 flow cytometer (Becton Dickinson, USA). The APC-, PE-and FITC-conjugated mouse IgG1s were used synchronously as the isotype controls.

Sorting CD133 + hCSCs
The CD133 + hCSCs were isolated using magnetic beads according to the standard procedure. In brief, cells were trypsinized to obtain an individual cell suspension and incubated with magnetic beads with anti-human CD133 antibodies (Miltenyi Biotec, Germany) for 30 minutes at 4°C. Cells were loaded onto LS columns and CD133 + cells were separated from other cells using a QuadroMACS™ Separator (Miltenyi Biotec, Germany). CD133 + cells were collected and suspended in DMEM/F12 media supplemented with 20 ng/ml EGF, 10 ng/ml FGF2 and 2% B27 (Miltenyi Biotec, Germany).

Drug treatments
Stock concentrations of ATRA (Sigma, USA), 9-cis retinoic acid (Sigma, USA) and docetaxel (Sigma, USA) were prepared in dimethyl sulfoxide (DMSO). The three drugs were further diluted into DMSO to obtain different working concentrations.

Immunohistochemistry
Human HCC specimens were obtained from the Department of Pathology of the Affiliated Calmette Hospital of Kunming Medical University. Immunohistochemical staining was conducted according to the standard procedures. Briefly, paraffin-embedded sections were incubated with antibodies of rabbit anti-human RARβ (Anbo Biotechnology Company, China), RALDH1 (Abgent, USA), RDH10 (Abgent, USA) and ADH1 (Beijing Biosynthesis Biotechnology Co., Ltd, China) overnight at 4°C. Sections were further stained with goat anti-human secondary antibodies conjugated with horseradish peroxidase (Santa Cruz Biotechnology, USA). Signals were developed with DAB reagent (Boster Biological Technology Ltd., China). Sections were examined under a light microscope (Olympus, Japan). Histological results were finally confirmed by a pathologist and with surgery conducted at the Affiliated Calmette Hospital of Kunming Medical University. qPCR Total RNAs were purified from cells using RNAiso Reagent (TaKaRa, China), and reversetranscription (RT) reactions were conducted to obtain cDNAs using a PrimeScriptTM RT Reagent Kit (TaKaRa, China). SYBR Primix Ex TaqTM (TaKaRa, China) was used for realtime fluorescence PCR (qPCR). Experiments were carried out using a real-time PCR machine (Applied Biosystems, USA).

Cell proliferation assay
Cells were seeded into 96-well plates at a density of 500 cells per well. Next day, various doses of DOC and ATRA were used to treat the cells. Cell proliferation was analyzed by a CellTiter 96 1 Aqueous One Solution Cell Proliferation Assay kit (Promega, USA). Cells were incubated for 2 hours, with cell proliferation then determined based on color change. We used a Spectra-Max M2 (Molecular Devices, USA) to test OD values at 490 nm.
Knockdown of β-catenin mRNA β-catenin shRNA/PLKO.1 and PLKO.1 empty plasmids were purchased from Sigma (USA). To knockdown the mRNA expression level of β-catenin, two stably-interfered cell lines were established according to the standard experimental procedure. Briefly, β-catenin shRNA/ PLKO.1 (or PLKO.1) plasmids together with pMD2.G and psPAX2 plasmids were mixed with FuGENE HD transfection reagent (Roche, Germany) and incubated at RT for 20 minutes, with the mixture then applied to the transfect 293T cells. Lentiviral particles were harvested 48 hours after transfection and added to the hCSCs for an extra 48 hours. Stable cancer stem cells were further screened with 10 μg/ml puromycin (Tocris, UK) for 15 days.

Bioinformatics
The Oncomine (Compendia Bioscience; http://www.oncomine.org) database was used for bioinformatic analysis. Relevant analytical parameters are provided in the figures.

Ethics committee approval
The Ethics Committee of the Affiliated Calmette Hospital of Kunming Medical University approved this study. Under this supervision, written informed consent from the donors or the next of kin was obtained for use of the sample in this research.

Statistical analysis
All experiments were repeated three times. Statistical analyses were performed using SPSS 13.0 software. Cell proliferation and qPCR were assessed by One Way ANOVA.

Retinoic acid metabolism was abnormally increased in liver cancers
The human liver is an actively RA-synthesizing organ. To address RA metabolism in HCC, we investigated three kinds of RA synthesis-related enzymes in HCC, including three alcohol dehydrogenases (ADH1, ADH2, ADH 3), two retinol dehydrogenases (RDH1, RDH10) and three retinaldehyde dehydrogenases (RALDH1, RALDH 2, RALDH 3) (Fig 1A). Patient-specific gene profiling was analyzed according to the online clinic liver cancer database (Oncomine: https://www.oncomine.org). Results from a total of 1,911 individual cancer patients demonstrated that the mRNA expressions of RALDH1, ADH1 and RDH10 were highest in HCC patients (Fig 1B-1D, dark blue histogram bars). Among a total of 16 different cancer types, the mRNA expression of RALDH1 showed a significant (p < 0.000001) 20-fold increase in HCC compared with that in other cancer types ( Fig 1B). Similar results were obtained for mRNA expressions of ADH1 and RDH10 in HCC (Fig 1C and 1D). We compared the mRNA expressions of three rate-limiting enzymes, RALDH1, RALDH2 and RALDH3, between liver cancers and normal liver tissues, and found that the mRNA expressions of RALDH1 and RALDH2 were slightly but significantly increased in HCC compared with that in normal liver (S1A-S1D Fig). Our findings demonstrated that RA-related metabolizing activities were abnormally increased in HCC, suggesting that RA was an enriched signal provided by the local microenvironment of liver cancers. Binding of all-trans retinoic acid activated retinoic acid receptors in CD133 + hepatic cancer stem cells To understand the role of RA signaling in CD133 + hCSCs, we first used flow cytometry to examine the percentages of CD133 + cells in different hepatic carcinoma cell lines (HepG2, Huh-7 and PLC-PRF-5). Interestingly, we found that about 11.7% of HepG2 cells were CD133-expressing (S2A Fig). These CD133 + cells were further isolated by magnetic beads and cultured in DMEM with highly concentrated growth factors EGF/bFGF. After 4 days culture, we observed that CD133 + cells generated typical spheres (S2D left Fig). Typical spheres were also generated from CD133 + PLC-PRF-5 cells (S10E Fig). In addition, CD133 + cells coexpressed higher levels of stem cell markers, such as SOX2, NANOG and OCT4, compared with that of CD133 -HepG2 cells (S2B Fig). These findings indicated that the HepG2 CD133 + cells were self-renewing hCSCs. Next, we examined and found the expression of RARs in CD133 + hCSCs was down-regulated (S3A- S3C Fig). For instance, the RARβ protein was lower expressed in CD133 + hCSCs than it was in CD133non-hCSCs (S3D Fig). Notably, we also detected that the expression level of the RARβ protein was much higher in HCC specimens than that in normal liver tissues (S3E Fig). However, no significant differences in the expression levels of RALDH1, ADH1 and RDH10 proteins were observed in HCC specimens compared with normal liver tissues (S10B Fig All-trans retinoic acid-induced differentiation of CD133 + hepatic cancer stem cells was dependent on β-catenin Since RA is an endogenous differentiation signal, the effects of ATRA on the differentiation of CD133 + cells was further determined. We found that ATRA (10 −9~1 0 −5 M) decreased the number of CD133 + hCSCs in a concentration-dependent manner (Fig 2A). Consistently, we observed a significant down-regulation in stem cell marker proteins after ATRA treatment ( Fig  2B and 2C). Similar results were investigated when 9-cis retinoic acid treated CD133 + hCSCs (S10A and S10C Fig). Furthermore, 9-cis retinoic acid treatment decreased the mRNA level in β-catenin (S10D Fig). These findings suggest that decreased CD133 + hCSCs were caused by the direct differentiation effect of ATRA. Interestingly, ATRA-induced differentiation of CD133 + hCSCs was associated with the activation of RARs (S4 and S5 Figs).
Previous studies suggested that ATRA inhibited cancer cell growth through down-regulating β-catenin protein expression (39)(40). We hypothesized that ATRA-induced differentiation of CD133 + hCSCs was likely related to the function of β-catenin. We found that ATRA treatment (10 −9 to 10 −5 M) indeed decreased the protein level of β-catenin in CD133 + hCSCs ( Fig  2D). At the same time, increased protein phosphorylation of β-catenin was observed after ATRA treatment (Fig 2D). It has been suggested that the Wnt/β-catenin signaling pathway is required to maintain the quiescent status of stem cells. Consistent with these findings, our results demonstrated that knockdown of β-catenin mRNA decreased the protein expression of stem cell markers in CD133 + hCSCs (Fig 2F and 2G) and impaired their stemness (S6A and S6B Fig). The regulation of β-catenin might be dependent on the PI3K-AKT pathway because ATRA decreased the protein level of PI3K and the protein phosphorylation of AKT (Fig 2E). The above findings support that ATRA-induced differentiation of CD133 + hCSCs was directly dependent on the function of β-catenin (Fig 3K).

Differentiated CD133 + hepatic cancer stem cells were sensitive to docetaxel
We investigated whether ATRA-induced differentiation could provide a possible strategy to target CD133 + hCSCs. ATRA treatment (10 −7~1 0 −5 M) alone did not impact the survival or proliferation of CD133 + hCSCs (Fig 3A and 3B). Similarly, knockdown of β-catenin mRNA did not change the survival or proliferation of CD133 + hCSCs compared with that of the PLKO.1 vector control (Fig 3G and 3H, black: vector control, green: β-catenin shRNA). However, treatment with 10 −8 M DOC decreased the survival and proliferation of CD133 + hCSCs  (Fig 3E and 3F). Consistently, we observed many more apoptotic CD133 + hCSCs in the combined treatments (S7A and S7B Fig).
To confirm these results, we combined ATRA (10 −6 M), β-catenin knockdown and DOC (10 −8 M) and obtained an even more efficient targeting effect (Fig 3I and 3J). These results demonstrated that differentiation of CD133 + hCSCs increased their sensitivity to DOC treatment.

Discussion
Although the liver is the only organ for retinoid metabolism, measurement of retinoid concentrations in liver cancer patients has not been reported. RA biosynthesis requires retinol dehydrogenases (RDH1, RDH10), alcohol dehydrogenases (ADH1, ADH2, ADH3) and retinaldehyde dehydrogenases (RALDH1, RALDH2, RALDH3) [35]. We analyzed the mRNA expression of the above dehydrogenases in HCC patients, and found that their mRNA expressions were the highest among 16 different cancers (Fig 1). The mRNA expressions of RALDH1 and RALDH2 were also slightly higher in HCC than in normal livers (S1 Fig), and protein expressions of RALDH1, ADH1 and especially RDH10 were also detected in hepatic carcinoma tissues (S10B Fig). To our best knowledge, this is the first time the activities of RA biosynthesis have been demonstrated in HCC patients, suggesting that RA could be enriched in the HCC microenvironment.
To determine whether RA is involved in the differentiation of hCSCs, we isolated selfrenewing CD133 + hCSCs from the hepatocellular carcinoma cell line HepG2. We observed relatively low expressions of RARs and β-catenin in CD133 + hCSCs compared with those in CD133non-hCSCs (S2C, S3A and S3D Figs). Interestingly, binding of ATRA to RARs activated the phosphorylation of the receptors (S5 Fig). In vitro, ATRA directly differentiated CD133 + hCSCs, as indicated by the decreased protein expression of stem cell markers (Fig 2B  and 2C). Further studies demonstrated that β-catenin was involved in ATRA-induced differentiation since its knockdown also decreased the protein expression of stem cell markers (Fig 2F  and 2G). Another form of vitamin A, 9-cis retinoic acid, that activates both retinoid X and retinoic acid receptors, attenuated the mRNA expression of β-catenin and stem cell markers in hepatic CSCs (S10A, S10C and S10D Fig). Consistent with previous findings [39][40], we concluded that RA in HCC functions mainly as a differentiation signal for hCSCs. We found that ATRA induced the down-regulation of β-catenin protein expression, but increased the phosphorylation level of β-catenin (Fig 2D). Phosphorylated β-catenin can be further degraded though GSK3β [63] and the PI3K-AKT pathway may influence β-catenin phosphorylation by indirectly regulating the protein functions of GSK3β [57,59]. To verify this, we examined the effects of ATRA on the function of the PI3K-AKT pathway. Our results demonstrated that ATRA down-regulated the expression and phosphorylation of PI3K and AKT (Fig 2E). Therefore, we identified a potential RAR-mediated cytoplasmic signaling pathway that enhanced the GSK-3β-dependent degradation of β-catenin through the inactivation of the PI3K-AKT pathway (Fig 3K). In addition, our results also suggested that Wnt/β-catenin signaling was required to maintain the undifferentiated status of hCSCs.
ATRA has been clinically applied to treat cancer patients diagnosed with acute promyelocytic leukemia (APL) [63][64][65]. New clinical phase trials have also been conducted to develop ATRA as a potential anti-drug for other solid cancers [41,[54][55][56]. However, we found that ATRA treatment alone did not obviously change cell survival or proliferation of CD133 + hCSCs (Fig 3A and 3B), with similar results also obtained when treating CD133non-hCSCs (S9 Fig). These findings suggest that solid hepatic cancer cells, in particular CD133 + hCSCs, may be resistant to ATRA treatment. Our results were consistent with previous studies in which lower drug levels of ATRA were clinically associated with relapses and ATRA-resistance in patients with APL [66]. This drug-related resistance might be because ATRA-induced differentiation needs molecular collaboration from other downstream pathways such as β-catenin. For instance, we found that ATRA-induced down-regulation of β-catenin increased hCSC sensitivity to docetaxel treatment (Fig 3E and 3F). Furthermore, combined inhibition of ATRA and β-catenin silencing also increased hCSC sensitivity to docetaxel treatment (Fig 3I and 3J).
Our research demonstrated that β-catenin potentially plays an essential role during ATRAinduced differentiation of hepatic cancer stem cells. Targeting β-catenin may provide extra benefits for hepatic cancer patients as the current treatment strategy mainly focuses on ATRAmediated differentiation.

Conclusions
Our results showed that ATRA differentiated hCSCs through a cytoplasmic signal pathway potentially composed of PI3K/Akt, GSK3β and β-catenin. β-catenin facilitated the undifferentiated status of hCSCs, suggesting that targeting β-catenin will provide extra benefits for ATRAmediated treatment of hepatic cancer patients.