Regulation and Expression of the ATP-binding Cassette Transporter ABCG2 in Human Embryonic Stem Cells

Abstract

The expression and function of several multidrug transporters (including ABCB1 and ABCG2) have been studied in human cancer cells and in mouse and human adult stem cells. However, the expression of ABCG2 in human embryonic stem cells (hESCs) remains unclear. Limited and contradictory results in the literature from two research groups have raised questions regarding its expression and function. In this study, we used quantitative real-time PCR, Northern blots, whole genome RNA sequencing, Western blots, and immunofluorescence microscopy to study ABCG2 expression in hESCs. We found that full-length ABCG2 mRNA transcripts are expressed in undifferentiated hESC lines. However, ABCG2 protein was undetectable even under embryoid body differentiation or cytotoxic drug induction. Moreover, surface ABCG2 protein was coexpressed with the differentiation marker SSEA-1 of hESCs, following constant BMP-4 signaling at days 4 and 6. This expression was tightly correlated with the down-regulation of two microRNAs (i.e., hsa-miR-519c and hsa-miR-520h). Transfection of microRNA mimics and inhibitors of these two microRNAs confirmed their direct involvement in the regulation ABCG2 translation. Our findings clarify the controversy regarding the expression of the ABCG2 gene and also provide new insights into translational control of the expression of membrane transporter mRNAs by microRNAs in hESCs.

INTRODUCTION

Comprehensive analyses of several classical ATP-binding cassette (ABC) transporters (e.g., ABCB1, ABCC1, and ABCG2) have shed light on their roles in cytotoxic drug efflux, glutathione (GSH)-conjugated substrate export, and drug resistance in cancer cells (for reviews, see refs (1-3)). However, the exact physiological functions of these multidrug transporters remain largely unknown. Further, knockout of the Abcg2 gene in mice has not yielded embryonic lethal phenotypes, but leads to hypersensitivity to dietary phototoxins and to protoporphyria (4). These data suggest that Abcg2 is not required for viability or fertility in mouse embryos, but might be critical in the cellular fate commitment, differentiation, and homeostasis of progenitor cells.

Indeed, the multidrug efflux pump ABCG2 has been implicated as the cause of the “side population” which helps define adult stem cells of various tissues and tumors, including placental trophoblasts, neural stem cells or progenitors, and hematopoietic progenitors (5-9). The capacities of human embryonic stem cells (hESCs) for self-renewal, propagation, and maintenance of the pluripotent state in vitro make regenerative medicine a feasible goal (10-13). However, the role of ABCG2 in hESC self-renewal and in maintaining pluripotency is not clear. The limited information in the literature reveals contradictory results on expression of ABCG2 (14-16). Sarkadi and coworkers reported expression of high levels of ABCG2 mRNA transcripts and ABCG2 protein in two hESC lines, HUES1 and HUES9(14, 15), whereas Zeng et al. showed HUES1, WA09 (H9), and CT2 hESCs lack ABCG2 mRNA as well as ABCG2 protein (16).

Thus, to determine the role of ABCG2 in the regulation of cellular homeostasis, pluripotency, and self-renewal of hESCs, its expression in hESCs must first be clarified. In this study, we examined the expression of ABCG2 in hESCs by quantitative real-time PCR, Western blot, and immunofluorescence microscopy. We found that ABCG2 mRNAs were expressed in all examined hESC lines. Despite an intermediate level of ABCG2 mRNAs, the level of the ABCG2 protein was very low or undetectable. The lack of protein expression was further linked to two microRNAs (miRNAs) that regulate ABCG2 mRNA translation. Our studies also suggest new roles of ABCG2 in the regulation of cellular differentiation, indicating that it could be a surface marker to define a specific differentiation state and to rapidly identify altered signaling that impairs hESC growth.

MATERIALS AND METHODS

Cell Culture Medium, Growth Factors, Antibodies, and Chemical Reagents

The reagents used in this study are as follows: hESC culture medium DMEM/F12, Knockout Serum Replacer (KSR), DMEM medium, and Dulbecco’s Phosphate-Buffered Saline (D-PBS) free of Ca^2+/Mg²⁺ from Invitrogen Inc. (Calsbad, CA); Heat-inactivated fetal bovine serum (FBS) from both Gemini (Gemini Bio-Products, West Sacramento, CA) and Hyclone (Logan, UT); Oct-4 (mouse monoclonal antibody or mouse mAb, IgG2b, sc-5279), SSEA-1 (mouse IgM, sc-21702) and SSEA-4 (mouse mAb IgG3, sc-21704) purchased from Santa Cruz Biotechnology (Santa Cruz, CA); anti-NANOG polyclonal antibody from COSMO Bio Co. Ltd. (Japan), anti-BCRP (clone BXP21) from Kamiya Biomedical Company (Seattle, WA), phycoerythrin (PE) labeled (clone 5D3, Catalog number or Cat No: 12-8888) or unlabeled anti-human ABCG2 (clone 5D3, Cat No: 14-8888-82) from eBioscience Inc. (San Diego, CA); BMP-4 (Cat No: 314-BP-010) and FGF-2 (Cat No: 233-FB) from R&D Systems (Minneapolis, MN); doxorubicin and mitoxantrone from Sigma-Aldrich (St. Louis, MO).

Cancer Cell Lines

The H460 parental cell line and its mitoxantrone-resistant derivative (H460/MX) were cultured as previously described (17). H460/MX cells express high levels of ABCG2 protein. The parental and resistant lines were used as negative and positive controls for Western blotting and immunofluorescence staining of ABCG2 protein expression, respectively. In addition, MCF-7/MX, a mitoxantrone-resistant subline derived from human MCF-7 breast cancer cells and approximately 4000-fold more resistant to mitoxantrone, was also used as a positive control for ABCG2 expression (18).

Human ES Cell Lines

The hESC lines used in this study were: hESBGN-01, hESBGN-02, and hESBGN-03 (NIH codes: BG01, BG02, and BG03) from BresaGen, Inc. (Athens, GA); HES-2 (NIH codes: ES02) from ES International (Singapore); I 3 and I 6 (NIH codes: TE03 and TE06) from Technion-Israel Institute of Technology (Haifa, Israel); HSF-1 and HSF-6 (NIH Codes: UC01 and UC06) from University of California at San Francisco (San Francisco, CA); H1 and H9 (NIH Codes: WA01 and WA09) from Wisconsin Alumni Research Foundation (WiCell Research Institute, Madison, WI); and Sahlgrenska 1 (NIH Code: SA01) from Cellartis AB (Göteborg, Sweden). Eight out of the above 10 hESC lines were used for screening ABCG2 mRNA expression (Figure 1). We focused on the 4 hESC lines (i.e., WA01, WA09, BG01, and BG03) for the cellular and molecular analyses, particularly on the WA01 and WA09 cell lines due to their continuity on the NIH Human Embryonic Stem Cell Registry. More detailed information regarding these hESC lines is available on the website of the NIH Stem Cell Unit (http://stemcells.nih.gov/research/nihresearch/scunit/).

Figure 1.

Human ES cell characterization and ABCG2 mRNA expression. (A): Phase images show morphology of WA01 (H1) hESC colonies grown on MEF feeder layers. Indirect immunofluorescence microscopic analysis of NANOG expression was carried out in undifferentiated (Undif) and spontaneously differentiated colonies using a polyclonal antibody against NANOG. Representative images and immunostaining of multiple hESC lines are shown. (B): Flow cytometric analysis of hESC marker expression in WA01 cells grown as colonies on MEFs. One representative experiment is shown. (C): Quantitative real-time PCR (qPCR) analysis of ABCG2 mRNA and 18S ribosomal RNA (18S Ribo) in 8 hESC lines grown on MEFs. Mean cycle threshold (Ct) values with standard deviation (bars) from triplicate PCR reactions are shown. (D-G): Northern blotting analysis of ABCG2 mRNA expression in hESCs. Approximately 20 μg of total RNAs from MCF-7/MX (positive control with exceptionally high ABCG2 expression), H460 (parental control), H460/MX (positive control), and 5 undifferentiated hESC lines (i.e., TE06, WA01, UC06, BG03, and BG02) were used for Northern blotting with a 102-base oligonucleotide probe as described in Material and Methods. (D): ABCG2 gene structures and the locations of the probes designed for real-time PCR and Northern blot. (E): The autoradiograph was obtained by exposure of the membrane to the film for 30 min. The RNA gel (presented in the lower panel) shows the amount and integrity of the total RNAs prior to transferring to the nylon membrane. (F): Autoradiograph with 20-hour exposure. (G): An 8-day prolonged exposure of lanes 3 to 8 to enhance the detection of ABCG2 mRNA expression.

Human ES Cell Culture on Feeder Layer

All hESC lines were initially cultured according to the suppliers’ protocols and adapted to a simple protocol as outlined below. MULTIWELL^TM 6 WELL polystyrene plates (Cat No 353046, Becton Dickinson Labware, Franklin Lakes, NJ) were coated with 0.1% gelatin for 1 hour at 37°C. Mouse embryonic fibroblasts (MEFs) derived from the CF1 strain, at passage numbers 5 and 6 (designated as p5 and p6, respectively), were cultured in DMEM medium (supplemented with 10% FBS, 2 mM L-glutamine and 0.1 mM non-essential amino acids) and irradiated at a dose of 8,110 rads with an X-ray irradiator (Faxitron X-ray Corporation, Wheeling, IL). The irradiated cells were plated at a density of 1.88 × 10⁵ cells per well (i.e., 1.96 × 10⁴ cells/cm²) and incubated at 37°C for 24 hours. Human ES cells were plated on top of an MEF feeder layer as small clumps (~ 50 μm in diameter) in the hESC medium containing 80% DMEM/F12 medium, 20% KSR, 2 mM L-glutamine, 0.1 mM non-essential amino acids, 0.1 mM β-mercaptoethanol, and 4 ng/mL of FGF-2. Colonies that exhibited morphologically differentiated cells (~5%) were manually removed. For cell passage, the hESC colonies were washed twice with D-PBS, and incubated with 1 mg/mL collagenase IV for 10 to 30 min. Detached colonies were transferred to a 15 ml tube and allowed to sediment for 5 to 10 min at room temperature. The supernatants containing residual MEFs were removed and the colonies resuspended with 5 ml hESC medium to repeat the sediment step. Finally, the colonies were triturated into small clumps and passaged at a ratio of 1 to 3 every 3 to 5 days.

Embryoid Body Differentiation

Human ES colonies grown on MEFs were dissociated with 1.5 mg/mL Collagenase IV in DMEM/F12 medium and processed as described for cell passage. The colonies were then lightly triturated in 1 ml differentiation medium, i.e., normal hESC growth medium without FGF-2. These hESC clumps were further plated on a 60-mm suspension culture dish with additional 6 ml of differentiation medium. The medium was changed every 2 days by sedimentation as described above. The embryoid bodies (EBs) were collected at different time points for analysis.

Feeder-Free Culture of hESCs

Human ES cells were cultured on hESC-qualified BD Matrigel (BD Biosciences, Bedford, MA) in mTeSR™1 medium from StemCell Technologies (Vancouver, BC, Canada). The semi-confluent colonies were rinsed with D-PBS and incubated at 37°C with 2 mg/mL dispase (Invitrogen Inc.) for 10 to 15 min. The colonies were collected, washed, triturated as small hESC clumps as described above, and plated on 2.5% BD Matrigel for passage or for desired experiments such as cytotoxic assays (with doxorubicin and mitoxantrone) and BMP-4 induced differentiation.

Flow Cytometry

Isolated human ES colonies, grown on MEFs, were dissociated with Accutase™ cell detachment solution (Innovative Cell Technologies, San Diego, CA). The cell pellets were rinsed once with D-PBS, resuspended in FACS buffer (DMEM/F12 medium containing 10% FBS) and incubated on ice for hESC surface marker staining. The cells used for localization of intracellular markers (e.g., Oct-4) were fixed in 4% paraformaldehyde and incubated at room temperature for 10 min and then resuspended in 0.1% Triton X-100 (also diluted in D-PBS) followed by incubation on ice for 15 min and centrifugation. The pellets were then resuspended in FACS buffer and incubated with desired antibodies in a round-bottom 96-well plate (Fisher Scientific). Approximately 1 to 2.5 × 10⁵ cells (in 50 μl of volume) were used for each reaction. Antibodies used were SSEA-1, SSEA-4, Tra-1-81 (kind gift from Dr. P. Andrews, Sheffield University, UK), and Oct-4. Cells were incubated at 4°C for 1 hour, washed with FACS buffer, and incubated with Alexa Fluor® 488-conjugated secondary antibodies (1:100 dilution) at 4°C for 30 min. Cells were washed again with FACS buffer and resuspended in 125 μl FACS buffer. FACSCalibur analyzed these cells with a 96-well plate HTS attachment (BD Biosciences, San Jose, CA). Data were collected and analyzed using PlateManager and CellQuest Pro software (BD Biosciences).

RNA Preparations

RNA was extracted using the basic Trizol protocol (Invitrogen Inc.) with slight modifications. Briefly, hESC cells were dissolved in 1 ml of the Trizol solution and stored in a -80°C freezer prior to use. Human ES cell lysates were further subjected to chloroform extraction and isopropanol precipitations. A DNA-Free kit (Applied Biosystems/Ambion, Austin, TX) was used to remove genomic DNA contamination. The quality of RNAs was agarose gel-verified and concentration was determined using an ND-1000 UV-Vis Spectrophotometer (NanoDrop Technologies).

Quantitative Real-Time PCR

Complementary DNA (cDNA) was synthesized by reverse transcription of 1 μg RNA in a 20-μL-reaction using a High Capacity cDNA kit with RNase inhibitor (Applied Biosystems). The reaction was followed by 10 min at 25°C, 120 min at 37°C, and 5 sec at 85°C. We used the TaqMan quantitative real-time PCR Microfluidic Platform (TaqMan Low Density Array or TLDA, Applied Biosystems) for the confirmation of ABCG2 gene expressions. The ABCG2 probe used for real-time PCR is located at the exon 5 and 6 boundary of the ABCG2 gene (TaqMan® Gene Expression Assay ID: Hs00184979, Applied Biosystems). cDNA was mixed with 2X TaqMan Universal PCR Master Mix (Applied Biosystems), loaded on the TLDA card, and run on an ABI Prism 7900 HT Sequence Detection System (Applied Biosystems) based on the manufacturer’s instructions. TLDA values were median-centered as previously described (19).

Whole Genome RNA Sequencing

RNA preparations and 36-bp paired mRNA sequencing using an Illumina Genome Analyzer IIx were performed according to the manufacturer’s instructions (SAIC-Frederick, Inc., Frederick, MD). Amplified and finally sequenced reads were mapped to human transcript sequences (version hg19) using "TopHat" (20). We used RealTimeAnalysisVersion (RTA) 1.6, SeqControlSoftwareVersion (SCS) 2.6, and GERALD Pipeline Version 1.6. Quality data was performed for total yield, clusters passing filters, and percent alignment. All our samples met the following requirements: a minimum of 500 Mega bases per sample, a minimum of 75% of clusters passing filter, a minimum of 70% of sequences aligning to the references, and an error rate less than 2%. Gene-specific and transcript-specific expression denoted as FPKM (Fragments Per Kilobase per Million reads) were calculated using "Cufflinks" (21). These transcripts can be considered as “absolute” data for each sample. Thus, for any given gene(s), if the sequencing reads are greater than zero, we can assume that the gene has some expression.

Real-Time PCR Analysis of miRNA Expression

Total RNAs enriched with miRNAs were extracted using the miRNeasy Mini Kit (Cat No: 217004, QIAGEN). Reverse transcription steps were performed using the TaqMan® MicroRNA Reverse Transcription Kit and stem-loop reverse primers for desired assays (Applied Biosystems). Twenty or fifty nanograms of RNAs were used for cDNA synthesis by a thermal cycler program (16°C, 30 min; 42°C, 30 min; 85°C, 5 min, and 4°C for termination). TaqMan® Small RNA Assay reagents, also purchased from Applied Biosystems, for miRNAs were: hsa-miR-519c (Assay ID, 001163; miRBase accession number, MIMAT0002832; mature miRNA Sequence: 5′-AAA GUG CAU CUU UUU AGA GGA U-3′), hsa-miR-520h (Assay ID, 001170; miRBase accession number: MIMAT0002867; mature miRNA sequence: 5′-ACA AAG UGC UUC CCU UUA GAG U-3′), and U6 snRNA control (Assay ID, 001973). Five microliters of cDNAs were used for real-time PCR using TaqMan® Universal PCR Master Mix II. The reaction was followed by 2 min at 50°C, 10 min at 95°C, and then 40 cycles PCR (15 sec at 95°C and 60 sec 60°C) run on the ABI Prism 7900 HT Sequence Detection System as described above. The cycle threshold (Ct) values were obtained by using the RQ Manager 1.2 software. Thereafter, we calculated ΔCt values (i.e., Normalized ΔCt = Ct^{miR-519c or miR-520h} - Ct^{U6 snRNA}), ΔΔCt (i.e., ΔΔCt = Normalized ΔCt^treated – Normalized ΔCt^controls), and fold change of the miRNAs (i.e., 2^−ΔΔCt).

Oliogonucleotide Probes and Radiolabeling

We designed a 102-base oligonucleotide probe specific for ABCG2 coding sequences based on the sequence information from the National Center for Biotechnology Information (Accession number: NM_004827, version 2). The probe template sequences are: 5′-CGG ATT AAC AGG GTC ATT CAA GAG TTA GGT CTG GAT AAA GTG GCA GAC TCC AAG GTT GGA ACT CAG TTT ATC CGT GGT GTG TCT GGA GGA GAA AGA AAA AGG −3′. The sequences of this probe template overlap with the probes designed for real-time PCR as described as above and designated as G2-971S. The probe template was synthesized and SDS-PAGE purified by Lofstrand Labs Limited (Gaithersburg, MD). The Northern blot probe was generated by a random priming and [³²P]-labeling method using G2-971S as the template DNA. High specific activity (i.e., > 1 × 10⁹ dpm/μg) was achieved by this radiolabeling method,

Northern Blotting

Eight samples of total RNAs, approximately 20 μg each, were denatured in a formaldehyde loading buffer, loaded into a 1% agarose gel that contained formaldehyde and 3-(N-morpholino) propanesulfonic acid (MOPS) along with 3 μg of RNA ladder (Invitrogen), and electrophoresed at 90 volts for 3 hours. The gel was transferred to a Nytran SuPerCharge nylon membrane (Whatman Inc. Piscataway, NJ) with a ‘TurboBlotter′ and 20X saline sodium citrate (SSC) buffer overnight. The membrane was then UV-linked, air-dried, and prehybridized with 6X SSC, 5X Denhardt′s solution, and 0.5% sodium dodecyl sulfate (SDS) at 68°C for 3 hours. The membrane was further hybridized with the radiolabeled probe (2.5 × 10⁶ dpm/mL) in HYBE buffer at 68°C for 24 hours. The membrane was finally washed in 2X SSC and 0.1% SDS at 68°C with three buffer changes over a period of 90 min and autoradiographed at −80°C using an intensifier screen.

Western Blotting

Cells were lysed in RIPA buffer [20 mM Tris, pH 7.2, 1% (v/v) Triton X-100, 1% (w/v) sodium deoxycholate, 0.1% (v/v) sodium dodecyl sulfate (SDS), 1 mM EDTA, 150 μM NaCl, 3% (v/v) protease inhibitor cocktail (Sigma, product number P8340), 50 mM sodium fluoride, 1 mM sodium vanadate]. Postnuclear supernatants from lysates centrifuged for 15 min at 16,000 × g at 4°C in a Hermle MR-2 centrifuge were solubilized in Protein Solubilizing Mixture [25% (w/v) sucrose, 2.5% (w/v) SDS, 0.05% bromophenol blue, 25 mM Tris, and 2.5 mM EDTA], resolved by SDS-PAGE, transferred to nitrocellulose, and evaluated by Western blot analysis using enhanced chemiluminescence (ECL). Briefly, after transferring proteins to membranes using the iBlot Dry Blotting system (Invitrogen), the membranes were treated with blocking buffer (0.1% Triton X-100, 15mM Tris, pH 7.4, 150 mM NaCl, 5% goat serum) for 2 hours at room temperature and incubated overnight at room temperature in a primary antibody (1:1000, BXP-21) diluted in blocking buffer. The membranes were then washed with Tris buffered saline (TBS; 15 mM Tris, 150 mM NaCl), twice with TBS supplemented with 0.1% Triton X-100, and once with high salt TBS (15 mM Tris, pH 7.4, 500 mM NaCl). Subsequently, membranes were incubated with a secondary antibody diluted in blocking buffer for 1 hour at room temperature and washed as above. The immunoblots were covered with ECL reagents and exposed to film. The images were analyzed by the ImageJ program (National Institutes of Health, Bethesda, Maryland).

Immunohistochemistry

Human ES colonies were grown in 6-well plates and fixed in 4% paraformaldehyde solution at room temperature for 20 min. The cells were blocked with 10% normal goat serum in the presence or absence of 0.1% Triton X-100 (diluted in D-PBS) at room temperature for 1 hour, then reacted with primary antibodies at 4°C overnight, and subsequently incubated with Alexa Fluor®-conjugated secondary antibodies (Invitrogen) in 5% blocking solutions for 1 hour at room temperature. Finally, the cells were stained with Hoeschst 33342 solution or DAPI. The samples were examined under an Axiovert 200 fluorescence microscope (Zeiss, Jena, Germany) equipped with ApoTOME AxioVision Rel 4.6 (Zeiss, Germany) acquisition systems. All images used for quantitative analysis were obtained under unsaturated exposure conditions. The images were analyzed by the ImageJ program.

Micro-RNA Inhibitor and Mimic Transfection

Human ES cells (6.0 x 10⁵ cells per well in 12-well plate) were transfected with miRNA inhibitors and mimics for both miR-519c and miR-520h. MicroRNA inhibitors and mimics were purchased from Dharmacon (Thermo Scientific, Lafayette, CO). These reagents include: miRIDIAN Hairpin Inhibitor hsa-miR-519c-3p (Cat No: IH-300786-05-0005) (abbreviated as miR519i), miRIDIAN Mimic hsa-miR-519c-3p (Cat No: C-300786-03-005) (i.e., miR519m), miRIDIAN Hairpin Inhibitor hsa-miR-520h (Cat No: IH-300833-05-0005) (i.e., miR520i), and miRIDIAN Mimic hsa-miR-520h (Cat No: C-300833-03-0005) (i.e., miR520m). Both miRIDIAN miRNA Hairpin Inhibitor Negative Control #1 (Cat No: IN-001005-01-05, sequence: UCA CAA CCU CCU AGA AAG AGU AGA) and miRIDIAN miRNA Mimic Negative Control #2 (Cat No: CN-002000-01-05, sequence: UUG UAC UAC ACA AAA GUA CUG) were used as controls in these experiments. A non-targeting miRIDIAN miRNA Transfection Control labeled with Dy547 (Cat No: IP-004500-01-05) was included in each of the transactions to monitor the delivery of miRNAs into hESCs. Human ES cells were transfected with the above inhibitors and mimics ranging from 0 to 160 nM by using the DharmaFECT Duo Transfection Reagent (Cat No: T-2010-02, Thermo Scientific) and following the manufacturers′ procedures. Optimal transfection efficiency for each concentration in both WA01 and WA09 cells was determined by imaging the Dy547 fluorescence of the live cells under a microscope at 24 hours after the transfection. The transfection efficiencies were calculated based on the percentage of Dy547 positive cells over all imaged cells. The average transfection efficiency was expressed as the mean values of Dy547 fluorescence intensity in all imaged cells and used for normalization of ABCG2 expression. Only cells or wells that have similar transfection efficiencies were used for BMP-4 induction. The cells were washed with fresh mTeSR1™ medium and then treated with 100 ng/mL of BMP-4 for an additional 48 hours to enhance the detection of ABCG2 expression. At 48 hours after BMP-4 induction, the transfected cells were fixed by 4% paraformaldehyde and co-stained with the anti-SSEA-1 and anti-ABCG2 monoclonal antibodies.

Quantitative analysis of ABCG2 expression was performed using the ImageJ program. To avoid analytical bias, we randomly assigned cell numbers from the regions of interest (ROI) in phase images. Due to the redundant expression of SSEA-1 on the plasma membrane after BMP-4 induction, we used the SSEA-1 immunofluorescence signal as the plasma membrane marker to determine ABCG2 (5D3) expression. The background of fluorescence images was subtracted using the Rolling Ball Radius method of the ImageJ program. ABCG2/5D3 immunofluorescence intensity (FI) was then measured at the single pixel level at the plasma membrane region of each randomly assigned cell. ABCG2 expression (i.e., FI) was normalized to the average transfection efficiency of each well. Subsequently, ABCG2 FI values were sorted according to increasing intensity and presented as line graphs (raw data) to visualize the heterogeneous expression of ABCG2 (i.e., cumulative ABCG2 expression). Statistical analysis of the mean (columns) and standard deviation (bars) of ABCG2 FI values from individual cells (~190 to 450 cells) are presented as histograms. P values were derived from an unpaired and two-sided student′s t-test. As indicated above, it is essential to determine the average transfection efficiency (for normalization) and perform desired experiments in the same set of cells/wells. In contrast to flow cytometric methods that require dissociation of cells from the culture dishes, the immunofluorescence-based methods described in our study provide a powerful way to achieve this aim. Moreover, the immunofluorescence method and data analysis provide insight into the regulation of ABCG2 at single-cell resolution rather than in a bulk population.

Statistical Analysis

Statistical analyses were performed as described previously (22).

RESULTS

Human ES Cell Culture Systems and mRNA Gene Expression

Rigorous quality control assays were used to monitor hESC culture, including virus and mycoplasma tests, cytogenetic karyotyping, comparative genomic hybridization (CGH) analysis of genomic integrity, immunostaining of hESCs with anti-NANOG, and flow cytometric analysis of a panel of hESC markers that include SSEA-1, SSEA-4, Tra-1-81, and Oct-4 (Figures 1A, 1B). Human ES cell samples with fewer than 5% of spontaneously differentiated cells (SDCs) were used for the analyses. For detailed characterizations, see the website of the NIH Stem Cell Unit. Thus, samples prepared from our hESC culture are suitable for genetic and molecular analysis of ABC transporter expression.

Enrichment of ABCG2 mRNA Transcripts in hESCs

We focused on the expression of ABCG2, a half-ABC transporter that plays a major role in mitoxantrone and anthracycline drug resistance. ABCG2 is of particular interest in hESCs due to its importance in adult stem cells and also its disputed expression in hESCs. In this study, we examined ABCG2 mRNA expression in 9 hESCs lines (BG01, BG02, BG03, ES02, SA01, TE03, UC01, UC06, and WA01) by real-time PCR and Northern blotting. All hESC lines showed similar mRNA levels, as determined by their cycle threshold (Ct) values (Figure 1C). When the Ct values of real-time PCR are converted to fold-differences, ABCG2 has a 110-fold higher mRNA signal than that of ABCB1 (also known as MDR1) (data not shown). Besides the sensitive real-time PCR method, we also employed Northern blots to examine the size and expression levels of ABCG2 mRNAs (Figure 1D-G). Figure 1E shows that mitoxantrone-selected MCF-7 (MCF-7/MX) cells had exceptionally high mRNA expression, displaying approximately a 3.5-kilobase (kb) fragment after 30-min exposure of the membrane to the film (Figure 1E, lane 1). However, H460/MX cells had much weaker expression of ABCG2 compared with MCF-7/MX cells (Figure 1F, lanes 1 and 2). Non-drug selected H460 cells had an ABCG2 mRNA level slightly lower than that of hESCs (Figure 1G), which is basically consistent with the real-time PCR result showing that H460 cells had an ABCG2 mRNA level (Ct = 26.6) comparable to that of hESCs (Figure 1C). All 5 hESC lines examined by Northern blotting had ABCG2 mRNAs around 2.8 to 3.8 kb, consistent with the sizes of the full-length ABCG2 mRNAs (Figure 1G). No alternative splicing events were found in these hESCs (Figure 1G).

Intermediate ABCG2 mRNA Expression Revealed by Whole Genome mRNA Sequencing

To further confirm the ABCG2 mRNA expression profile, we used an Illumina Genome Analyzer IIx to quantitatively verify gene-specific and transcript-specific expression of ABCG2 as defined by the average fragments per kilobase per million reads (FPKM) (Figure 2). We mapped gene-specific transcripts to approximately 21, 510 individual genes in three sequencing experiments of the representative line BG01. These transcripts are classified into three categories based on the cutoff values of the average FPKM, which include high (i.e., FPKM ≥ 5), intermediate (i.e., 5 > FPKM ≥ 1), and low to no (i.e., FPKM < 1) mRNA expression (Figure 2). Under this assay condition, ABCG2 and several other genes (e.g., CD9, REST, and TERT that are highly expressed in hESCs) had a stable FRKM expression profile (Figure 2). Our data indicated that ABCG2 mRNA expression in the three samples from two different BG01 cultures had similar FPKM values (i.e., 1.6, 1.5, and 1.8), within the range of intermediate mRNA expression. Moreover, within this range, we also detected pluripotent gene transcripts from POU5F1, KLF4, and TERT (Figure 2). Taken together, these data suggest that undifferentiated hESCs express an intermediate level of ABCG2 mRNAs without the presence of alternative splicing forms.

Figure 2.

Whole genome mRNA sequencing and ABCG2 mRNA expression. Analysis of ABCG2 mRNA expression by an Illumina Genome Analyzer IIx in two BG01 samples (sample 1 and sample 2). (A and B): Sample 1 was sequenced in duplicate for verifying the reproducibility of the assay. (C): mRNA sequencing of a different passage of BG01 cells for analysis of the stability of gene expression within the same line. Gene-specific and transcript-specific expression of ABCG2 are defined by the average fragments per kilobase per million reads (FPKM). The levels of transcripts are classified into three categories (as indicated) based on the cutoff values of the average FPKM.

Lack of ABCG2 Protein Expression in hESCs

We further examined ABCG2 protein expression in BG01, WA01, and control cells by immunostaining using an anti-ABCG2 monoclonal antibody (clone 5D3) that specifically recognizes an extracellular epitope of ABCG2. In the control experiments, parental H460 cells did not express surface ABCG2 (Figure 3A) compared with their drug-resistant H460/MX cells, which have significant amounts of surface ABCG2 expression (Figure 3B). No positive ABCG2 (5D3) colonies were found in BG01 (Figure 3C) or WA01 cells (Figure 3D). Similar results were found in both BG03 and WA09 cells. To explore the linkage of ABCG2 with the spontaneous differentiation of hESCs, we examined two additional hESC lines (i.e., TE03 and TE06) because these two cell lines displayed spontaneously differentiated cells with higher frequencies than both BG01 and WA01 cell lines. However, only a few isolated ABCG2⁺ colonies were found in TE03 cells (Supplemental Figure 1A). The estimated frequency of ABCG2⁺ colonies was approximately 3 out of 100 in the TE03 line. In addition, we found that morphologically undifferentiated TE06 cells usually do not express ABCG2 (5D3) (Supplemental Figure 1B), which could serve as a negative control in this experiment. The TE06 line seems to have a higher frequency of ABCG2⁺ colonies (~ 5 to 10%) and often generates spontaneously differentiated cells (SDCs) (~5 to 8%) under the hESC growth conditions (Supplemental Figure 1C). However, the majority of SDCs did not express ABCG2 (5D3). ABCG2⁺ cells were also found in morphologically undifferentiated cells (Supplemental Figure 1D).

Figure 3.

ABCG2 protein expression in hESC and embryoid bodies. Surface plasma membrane ABCG2 expression was determined by immunostaining directly with PE-labeled anti-human CD338 (ABCG2) (clone 5D3) in (A) H460 (negative control) and (B) H460/MX (positive control). (C and D): Surface coexpression of ABCG2 and SSEA-4 was determined by immunostaining undifferentiated WA01 (H1) and BG01 cells with anti-ABCG2 (clone 5D3) and anti-SSEA-4. (E-G): Total cellular lysates from H460 (parental control), H460/MX (high ABCG2 expression control), undifferentiated (undif) hESCs (i.e., BG01 and WA01), embryoid bodies (EBs) of BG01 and WA01 (BG01, EB at day 8 and WA01, EB at days 5 and 10) were used for Western blotting with the anti-ABCG2 (BXP21) monoclonal antibody as described. Three independent experiments (E, F, and G) are shown. A long exposure (Long Expo) blot is shown in the right panel of Figure 2F to enhance the detection of trace amounts or low levels of ABCG2 expression. Abbreviations are: Kd, kilodalton, d5, d8, and d10, EBs obtained at days 5, 8, and 10 respectively. Scale bars, 50 μm.

ABCG2 protein expression was also verified by Western blot using the BXP21 monoclonal antibody in BG01 and WA01 cells. Both parental H460 and its mitoxantrone-resistant derivative line H460/MX were used as negative and positive controls for ABCG2 protein expression. ABCG2 protein in H460 cells was only weakly expressed and constituted approximately 3.4% of total ABCG2 in H460/MX cells (Figures 3E, 3F; lanes 1 and 2). However, despite similar mRNA levels to H460 cells, undifferentiated BG01 and WA01 cells had undetectable ABCG2 protein (Figures 3E, lanes 3 and 4; 3F, lane 3; 3G, lane 2). Thus, our data suggest that hESC colonies barely express surface ABCG2 or intracellular ABCG2 under the routine growth conditions. Interestingly, differentiation of hESCs as embryoid bodies (EBs) for BG01 and WA01 cells at days 8 and 10 also did not induce the synthesis of ABCG2 protein (Figures 3F, lane 4; 3G, lanes 3 and 4). Thus, our data suggest that ABCG2, not likely to be useful as a marker for spontaneous or EB differentiation, might be associated with specific differentiation signals.

Regulation of ABCG2 by BMP-4 Signaling

To verify whether ABCG2 could be induced by specific differentiation signals, we treated cells with 50 or 100 ng/mL of recombinant BMP-4, a member of the bone morphogenetic protein family originally identified as a potent inducer of osteogenesis (23). BMP-4 efficiently converted undifferentiated WA01 colonies into a differentiated state at day 6 (Figure 4, A to C). These differentiated cells manifested co-expression of SSEA-1 and down-regulation of the pluripotency marker SSEA-4 (Figure 4, A to C). When WA01 colonies were treated with BMP-4 (100 ng/mL), the mean membrane fluorescence intensity (FI) of SSEA-1 increased by 3.2-fold (P value = 0.0015) compared with the untreated control at the 24-hour time point, further elevated by 19-fold (P = 1.2 × 10⁻¹⁴) at day 4, and stabilized at day 6 (Figure 4D1). Correspondingly, only a 1.3-fold increase (P = 0.09) in the membrane fluorescence intensity of ABCG2 was observed in WA01 colonies after 24-hour BMP-4 treatment compared to the untreated control (Figure 4D2). ABCG2 signal was significantly elevated by 3.4-fold (P = 4.7 × 10⁻¹²) at day 4 and by 7.0-fold (P = 3.4 × 10⁻¹²) at day 6 (Figure 4D2) compared with the untreated control. These data indicate that expression of SSEA-1 precedes the emergence of ABCG2⁺ cells under these conditions. There is a step-wise increase in the SSEA-1/ABCG2 correlation coefficient (r) (ranging from – 0.13 to 0.46) in WA01 cells (Figure 4D3). The SSEA-1/ABCG2 expression patterns mediated by BMP-4 were also found in WA09 cells (Figure 4E). However, no ABCG2 protein was observed at day 4 in BG03 cells despite up-regulation of SSEA-1 (Supplemental Figure 2A). These data suggest that latent expression of ABCG2 protein might be tightly regulated by constant BMP-4 signaling, regardless of the presence of intermediate levels of ABCG2 mRNAs.

Figure 4.

Indirect immunofluorescence analysis of surface ABCG2, SSEA-1, and SSEA-4 coexpression during differentiation. (A): WA01 cells (control); (B and C): WA01 cells treated with 50 ng/mL or 100 ng/mL of recombinant BMP-4 for 6 days. C1 indicates BMP-4 concentration 1 (i.e., 50 ng/mL) and C2 denotes BMP-4 concentration 2 (i.e., 100 ng/mL). The images were acquired with an Axiovert 200 fluorescence microscope (Zeiss). Bar, 50 μm. (D and E): Quantitative analysis of the coexpression of surface ABCG2 and SSEA-1 in WA01 cells (D) and WA09 cells (E) under the treatment of the cells with 100 ng/mL of BMP-4 for 0 (control), 24, 96, and 144 hours. Mean plasma membrane fluorescence intensity (FI or IF) from a representative pixel per cell was measured by the ImageJ program. The average of the mean fluorescence intensity (FI) and its standard deviation (SD) from 50 individual cells are shown. The Pearson correlations between ABCG2 and SSEA-1 are shown in the right panels. Abbreviations: CE, coefficient; hr, hour(s).

Induction and Selection of ABCG2 by Cytotoxic Drugs

A well-known function of ABCG2 is to confer multidrug resistance in human cancer cells. ABCG2 is frequently induced or selected by cytotoxic drugs such as mitoxantrone and anthracyclines. To determine whether this pharmacological function is present in hESCs, we treated BG03 hESCs with mitoxantrone and doxorubicin. In general, the hESCs were hypersensitive to these cytotoxic drugs. Short periods (~ 6 hours) of cytotoxic drug treatment failed to induce cell surface expression of ABCG2 (data not shown). Supplemental Figure 2B showed that the BG03 control did not express SSEA-1 or ABCG2. However, rare mitoxantrone-resistant BG03 colonies (SSEA-4⁺/ABCG2⁺⁺) could be selected under severely cytotoxic conditions such as 1 μM mitoxantrone treatment (Supplemental Figure 2C). In addition, low levels of SSEA-1 and ABCG2 could be induced by 10 nM doxorubicin at 24 hours (Supplemental Figure 2D).

Regulation of ABCG2 by miRNAs

Our above experimental results indicate that the ABCG2 gene in hESCs has an mRNA⁺/protein⁻ expression pattern, which is subject to regulation by differentiation signaling and cytotoxic induction. These regulatory processes enhance ABCG2 protein expression (Figure 4, Supplemental Figure 2). The mechanisms that regulate the mRNA⁺/protein⁻ state of the ABCG2 gene are unknown in hESCs. Previous study revealed that two miRNAs (i.e., hsa-miR-519c and hsa-miR-520h) play an important role in regulating ABCG2 protein translation in human cancer cells (24, 25). To verify whether these miRNAs are also associated with the regulation of ABCG2 expression in hESCs, we performed quantitative real-time PCR to examine the levels of the two miRNAs in WA01, WA09, BG01, and BG03 cells. Our results showed these miRNAs were relatively low in WA09 cells compared with WA01, BG01, and BG03 cells (Figures 5A, 5B). After BMP-4 induction, WA01 cells exhibited 2.0-, 2.8-, and 9.2-fold decreases in miR-519c expression at 24-, 96-, and 144-hours, respectively (Figure 5A, columns 1-4). However, no obvious decrease in miR-519c was found in WA09 cells at the 24-and 96-hour time points (Figure 5A, columns 5-7). Nonetheless, a 2-fold latent reduction of miR-519c was still observed in WA09 cells at day 6 after BMP-4 treatment (Figure 5A, column 8). The down-regulation patterns of both miRNA-519c and miRNA-520h in WA09 cells were similar (Figures 5A, 5B). Further, down-regulation of miR-519c was apparently correlated with extracellular membrane ABCG2 expression (i.e., correlation coefficient r = 0.90) (Figure 5C). Consistently, reduction of miRNA-520h was also correlated well with ABCG2 expression (r = 0.89) (Figure 5D). Taken together with the known function of the two miRNAs in the regulation of the ABCG2 gene in cancer cells, our data indicate that down-regulation of the two miRNAs is tightly correlated with translation control of ABCG2 mRNA in hESCs.

Figure 5.

MicroRNAs and ABCG2 expression. Relative expression of hsa-miR-519c (A) and hsa-miR-520h (B) in WA01, WA09, and their BMP-4 (100 ng/mL) treated cells (as described in Figures 3D and 3E). Fold of down-regulation of the two miRNAs was based on the cycling threshold (Ct) values from real-time PCR experiments. Both BG01 and BG03 cells were included as untreated controls for comparison. (C and D): The correlations between fold-expression of the two miRNAs and surface ABCG2 expression as determined by quantitatively analyzing immunofluorescence intensity of the anti-ABCG2 (5D3) staining.

Regulation of ABCG2 Expression by both miRNA Inhibitor and Mimics

MicroRNA mimics for miR-519c and miR-520h (designated as miR-519m and miR-520m respectively) are chemically modified double-stranded oligonucleotides designed to mimic the function of the two endogenous mature miRNAs, whereas miRNA hairpin inhibitors for miR-519c and miR-520h are RNA oligonucleotides with novel secondary structure synthesized to restrain the functions of the two endogenous mature miRNAs. To establish the role of miR-519c and miR-520h in the regulation of ABCG2 expression, we transfected these miRNA mimics and miRNA hairpin inhibitors into WA01 and WA09 cells and then induced the cells with 100 ng/mL of BMP-4. In general, we could achieve transfection efficiencies up to 91% under our experimental conditions (Figures 6A, 7A). The average transfection efficiencies used for normalization in all imaged cells were similar to each other under different treatments in both WA01 and WA09 cells (Figures 6B, 7B).

Figure 6.

Regulation of ABCG2 expression by miRNA inhibitors. MicroRNA hairpin inhibitors (miR-519i and miR-520i) are synthesized to restrain the functions of the two endogenous mature miRNAs miR-519c and miR-520h, respectively. Both WA01 and WA09 cells were transfected with miR-519i and miR-520i. At 24 hours, the average transfection efficiency was determined by measuring the mean Dy547 fluorescence intensity values of all imaged cells in 5 to 7 regions of interest (ROI). The cells were then induced with 100 ng/mL of BMP-4 for an additional 48 hours. ABCG2 expression was determined by immunostaining the cells with the mouse mAb (clone 5D3). The membrane ABCG2 immunofluorescence intensity was measured by the ImageJ program and normalized to the transfection efficiency. Each transfection condition in each independent experiment was done at least in duplicate. One of two independent experiments with similar results is shown. (A): Representative images of WA01 cells transfected with Dy547-labeled control miRNAs for monitoring transfection efficiency (TE) at 24 hours, with black arrowheads indicating typically transfected cells (TC) and white arrowheads pointing to free fluorescence particles (FP) of the Dy547-labeled miRNA controls. (B): The average transfection efficiencies in WA01 cells treated with different miRNA inhibitors are as indicated. (C and D): Representative SSEA-1 and ABCG2 (5D3) immunofluorescence images of WA01 cells treated with different miRNA inhibitors, with arrowheads indicating plasma membrane (PM) regions with increasing fluorescence intensity. (E and F): Cumulative expression of ABCG2 (5D3) was based on the average immunofluorescence intensity measured randomly from 191 to 234 individual cells. (G and H): Histogram presentation and statistical analysis of ABCG2 (5D3) expression of the left panel. Columns mean immunofluorescence intensity; Bars, standard deviations; Scale bars, 50 μm. Abbreviations: ABCG2 (FI), normalized ABCG2 (5D3) expression based on immunofluorescence intensity; Neg Cont, the miRNA inhibitor negative control; PM, the plasma membrane.

Figure 7.

Regulation of ABCG2 expression by miRNA mimics. Two miRNA mimics designated as miR-519m and miR-520m are designed to mimic the function of the two endogenous mature miRNAs miR-519c and miR-520h, respectively. Both WA01 and WA09 cells were transfected with miR-519m and miR-520m. The procedures and methods used for BMP-4 induction, determination of transfection efficiency, ABCG2 (5D3) expression, and normalization are the same as described in Figure 6. Each transfection condition in each independent experiment was done at least in duplicate. One of two independent experiments with similar results is shown. (A): Representative images of WA09 cells transfected with Dy547-labeled control miRNAs for monitoring transfection efficiency at 24 hours. (B): The average transfection efficiencies in WA09 cells treated with different miRNA mimics. (C and D): Representative SSEA-1 and ABCG2 (5D3) immunofluorescence images of WA09 cells treated with different mimics. (E and F): Cumulative expression of ABCG2 (5D3) was based on the average immunofluorescence intensity measured randomly from 191 to 400 individual cells. (G and H): Histogram presentation and statistical analysis of ABCG2 (5D3) expression. Scale bars, 50 μm. Abbreviations: ABCG2 (FI), normalized ABCG2 (5D3) expression based on immunofluorescence intensity; Neg Cont, the miRNA mimic negative control. (I): Hypothetical regulation of ABCG2 mRNA⁺/protein⁻ expression in undifferentiated hESCs. Undifferentiated hESCs under healthy hESC culture conditions with FGF-2 generate homogeneous cell types and a uniform pluripotent state that lacks surface ABCG2 protein expression. The potential mechanisms that control the ABCG2 protein levels might be associated with consistent differential signaling that enhances translation of ABCG2 mRNA, proteasome degradation and pluripotent factor-related inhibition of ABCG2 expression. ABCG2 mRNA⁺/protein⁻ cells or colonies under hESC culture conditions may not have the ability to survive under cytotoxic drug selection with ABCG2-related substrates (e.g., mitoxantrone, abbreviated as MX). (J): Regulation of ABCG2 mRNA⁺/protein⁻ expression in undifferentiated hESCs by miRNAs that target the 3′ untranslated region (3′-UTR) of the ABCG2 gene. The model shows that the inhibition of ABCG2 mRNA translation by miRNAs in undifferentiated cells could be derepressed by constant differentiation signals (such as BMP-4) and miRNA inhibitors (such as miR-519i and miR-520i). MicroRNA mimics may serve as inhibitors of ABCG2 protein expression. Abbreviations: ABCG2 (FI), normalized ABCG2 (5D3) expression based on immunofluorescence intensity; FI, fluorescence intensity; hESCs, human embryonic stem cell; miRNA, microRNA; Neg Cont, the miRNA mimic negative control.

Our results indicated that both miR-519i and miR-520i (at 160 nM) significantly increased ABCG2 protein levels in WA01 cells by 2.6-fold (P = 2.8 × 10⁻²⁰) and 1.9-fold (P = 1.3 × 10⁻¹⁴) respectively, compared with the negative inhibitor control (Figures 6D-F). However, under the equivalent concentration (160 nM) of both inhibitors (80 nM each), the two inhibitors antagonized their own modulatory effects on ABCG2 expression, comparable to that effect of the negative control (P = 0.36) (Figures 6D-F). A moderate increase (i.e., 1.3-fold) in ABCG2 expression related to the negative control was also evident in WA09 cells after treatment with the optimal 80 nM miR-519i (P = 0.008) (Figures 6G, 6H). Neither miR-520i (P = 0.85) nor miR-519i combined with miR-520i had a modulatory effect on ABCG2 expression in WA09 cells (Figures 6G, 6H).

Moreover, our results also indicated that both miR-519m and miR-520m at 80 nM could significantly decrease ABCG2 protein levels in WA09 cells by 32% (P = 1.4 × 10⁻⁵) and 40% (P = 8.1 × 10⁻⁹) respectively, compared with the negative control (Figures 7D-F). Consistently, the inhibitory effect of the mimics on ABCG2 expression was significantly attenuated when WA09 cells were treated with both miR-519m and miR-520m (Figure 7F, columns 2-4). Nonetheless, no obvious down-regulation of surface ABCG2 was found after the expression of miR-519m or miR-520m or both in WA01 cells (Figures 7G, 7H). These results suggested that the miRNA inhibitors and mimics for miR-520c and miR-520h are directly involved in the regulation of ABCG2 protein expression in hESCs in a cellular context-dependent manner.

DISCUSSION

We believe our studies have provided sufficient evidence to resolve the controversy about ABCG2 expression in hESCs by showing that undifferentiated hESCs have intermediate levels of ABCG2 mRNA (mRNA⁺) but no protein (protein⁻) (Figure 7I). This ABCG2 mRNA⁺/protein⁻ expression pattern is different from previous studies reporting mRNA⁺/protein⁺ (14, 15) and mRNA⁻/protein⁻ in hESCs (16). The failure to detect ABCG2 mRNA in the latter study might have simply resulted from the insensitivity of the detection method (16), whereas the mRNA⁺/protein⁺ pattern could reflect a more differentiated hESC population. Despite intermediate levels of ABCG2 mRNAs, hESCs maintain tight post-translational control that limits protein synthesis even with cytotoxic and differentiation stress signals. Surprisingly, the majority of spontaneously differentiated hESC cells did not express surface ABCG2 (Supplemental Figures 1C, 3B). Nevertheless, the mechanisms that result in the mRNA⁺/protein⁻ pattern we observed have not previously been elucidated. The potential mechanisms that control the ABCG2 protein levels might be associated with translational inhibition of ABCG2 mRNA, proteasome degradation of nascent synthesis of the protein, or pluripotent-factor related inhibition of ABCG2 expression (Figure 7I).

To determine the potential mechanisms that regulate ABCG2 protein levels, we focused on the BMP-4-induced models of ABCG2 expression. In these models, we have shown that expression of the stage-specific embryonic antigen-1 (SSEA-1), which has been observed to increase upon hESC differentiation, precedes the emergence of surface ABCG2 expression at day 4 or day 6 after constant BMP-4 signal induction (Figure 4). We were able to show that at least two previously described miRNAs (miR-519c and miR-520h) are involved in the regulation of the mRNA⁺/protein⁻ state of the ABCG2 gene. The functionality of the two miRNAs, targeting the 3′ untranslated region (3′-UTR) of the ABCG2 gene, was demonstrated in human cancer drug-resistant cells (24, 25). Transfection of miRNA inhibitors and mimics for miR-519c and miR-520h has proved that these two endogenous mature miRNAs are directly involved in the regulation of surface ABCG2 expression (Figures 6, 7). Interestingly, these miRNA inhibitors or mimics acting on the two endogenous miRNAs are cell-type dependent. No cooperativity was observed between the miRNA inhibitors or between the miRNA mimics, suggesting that miR-519c and miR-520h exert their inhibitory effects on ABCG2 expression via different pathways. Thus, an emerging model for the regulation of ABCG2 in hESCs depicts a translational repression state (i.e., mRNA⁺/protein⁻) in undifferentiated hESCs or a miRNA-dependent derepression state (i.e., mRNA⁺/protein⁺) mediated by differentiation signaling or by cytotoxic stress (Figure 7J).

It is not clear why hESCs under culture conditions supplemented with FGF-2 have been reported to show relatively high ABCG2 protein expression (14, 15). One simple explanation is that these hESCs might represent partially differentiated cells under even slight modification of the conditions (e.g., 15% KSR in hESC medium), although these cells still retain significant amounts of hESC marker proteins (14, 15). This question should be answered by examining the presence of SSEA-1. An alternative possibility might be that these hESCs have mouse ES cell (mESC)-like properties such as expression of SSEA-1 and ABCG2 in undifferentiated mESCs. As we know, the mESC culture system, distinguished from hESCs, relies on both BMP-4 and Stat3 for self-renewal (26). In this system, both SSEA-1 and ABCG2 are pluripotent markers in mESCs, which are down-regulated in differentiating mESCs. We have also revealed considerable plasticity of the surface ABCG2 expression during hESC culture (Supplemental Figures 1D, 3C-D). The frequency of ABCG2⁺ cells is very low (~1 to 3%) in most hESC lines and could be higher in some hESC lines such as TE06 (Supplemental Figure 1D). It is also possible that hESCs could adapt to an mESC-like state under certain growth conditions. Recent studies have indicated the hESC and mESC states are interconvertible (27).

It has been also shown that curcumin, a co-inhibitor of several ABC drug transporters such as ABCB1 and ABCG2 and fatty acid 5Δ and 6Δ desaturase, promotes ES pluripotent states and delays neuronal differentiation (28), indirectly suggesting that ABCG2 might play a role in differentiation. It has been demonstrated that Abcg2 mediates cell protection in cardiac side population (SP) cells when the cells were exposed to oxidative stress such as hydrogen peroxide, which appears to be linked with increased γ-glutathione reductase protein expression (29). Recent work using mESCs demonstrates that Abcg2 protects ES cells from the accumulation of protoporphyrin IX and elevated levels of reactive oxygen species (ROS) during ES colony expansion (30). Pharmacological inhibition of ABCG2 by its inhibitor fumitremorgin C (FTC) resulted in a pronounced decrease in Nanog staining in these mESCs (30). These data suggest that ABCG2 may be part of the defense machinery that sustains the self-renewal of mESCs through reducing ROS-mediated cellular damage or apoptosis.

In principal, the cell protection functions of ABCG2/Abcg2 in both hESCs and mESCs should be fundamentally similar. However, this protection could be achieved in different ways. Mouse ES cells have a high plating efficiency when they are dissociated to single cell suspension. Thus, Abcg2 may function in mESCs in a pluripotent state at the single-cell level. In contrast, ABCG2 confers a survival advantage to the topoisomerase inhibitor mitoxantrone only in some ABCG2⁺ differentiated cells and in very rare ABCG2⁺ colonies (Supplemental Figure 2C). Hence, these data suggest that a selection, rather than induction mechanism is involved in this protection, reminiscent of doxorubicin and etoposide drug-resistant clones derived from the Luria-Delbrück fluctuation analysis (31, 32) and of doxorubicin single-step selected variant cells in human breast cancer cells (33).

In summary, here we shed light on the controversy concerning ABCG2 expression in hESCs, and show that ABCG2 is a late-stage differentiation marker of BMP-4-related pathways in in vitro hESC culture. Cell surface expression of this transporter could be used as a marker to define cell type, cellular state, or to rapidly identify signals that control hESC growth. Our data provide molecular clues to how ABCG2 is regulated by miRNAs during differentiation.

Supplementary Material

Supp Fig S1-S3

Supplemental Figure 1. Direct immunofluorescence analysis of surface ABCG2 expression. Surface membrane ABCG2 expression was determined by directly immunostaining with PE-labeled anti-human CD338 (ABCG2) (clone 5D3) in (A) a differentiated TE03 colony that overexpressed ABCG2, (B) undifferentiated TE06 cells (TE06 undif), (C) spontaneously differentiated TE06 cells (TE06 Dif), and (D) mixed populations of TE06 cells that contain morphologically differentiated cells with a negative ABCG2 status as indicated by arrowheads. The ABCG2⁺ cells were found in morphologically undifferentiated cells as indicated by arrows. Both TE03 and TE06 cells were grown on 2.5% Matrigel in mTeSR™1 medium. Under this growth condition, there were approximately 5 to 8% of spontaneously differentiated cells observed in TE06 cells (C), whereas the TE03 line had approximately 2% of spontaneously differentiated cells (A). The differentiated colonies, based on morphologies and other immunostaining criteria, were counted under microscope. The images were acquired under an Axiovert 200 fluorescence microscope (Zeiss). Bar, 50 μm

Supplemental Figure 2. (A): Analysis of surface ABCG2, SSEA-1, and SSEA-4 coexpression during BMP-4 directed differentiation. Surface ABCG2, SSEA-1, and SSEA-1 were co-stained with anti-SSEA-1, anti-SSEA-4, and anti-ABCG2 (clone 5D3) in BG03 control cells and BG03 cells treated with 50 ng/ml or 100 ng/ml of recombinant BMP-4 for 4 days. (B-D): Immunofluorescence analysis of surface ABCG2, SSEA-1, and SSEA-4 coexpression under cytotoxic stresses. Surface membrane coexpression of ABCG2, SSEA-1, and SSEA-1 was determined by immunostaining the cells with anti-ABCG2 (clone 5D3) in (B) BG03 control, (C) BG03 cells treated with 1 μM mitoxantrone for 24 hours, and (D) BG01 cells treated with 10 nM doxorubicin for 24 hours. Both BG01 and BG03 cell lines were cultured on 2.5% BD Matrigel in mTeSR™1 medium. Less than 1% of the colonies were observed to have spontaneously differentiated cells under these culture conditions. Homogeneously staining patterns were observed in all experimental panels except in Supplemental Figure 2C, in which only one colony survived and was photographed from one well of a 6-well plate. All other previously seeded colonies (~ 100 colonies) were killed by the mitoxantrone (MX) treatment. The images were acquired with an Axiovert 200 fluorescence microscope (Zeiss). Bar, 50 μm

Supplemental Figure 3. Surface ABCG2 expression in BG03 cells. Surface membrane ABCG2 expression was determined by immunostaining of the cells with PE-labeled anti-human CD338 (ABCG2) (clone 5D3): (A) undifferentiated BG03 cells; (B) spontaneously differentiated BG03 cells, (C) morphologically undifferentiated BG03 colonies (2 out of ~ 100 colonies) stained with the 5D3 monoclonal antibody; (D) Enlarged view of C; (E) Two 5D3 positive colonies shown in D were isolated by a colony pick-up method under a Leica microscope, triturated into small clumps, and plated on a new well coated 2.5% BD Matrigel. All clumps formed nice colonies after the immediate passage, without recurrent 5D3 staining and spontaneous differentiation. Bar, 50 μm