Author contributions: H.Z.: research conception and design, collection and/or assembly of data, data analysis, manuscript writing; J.W.P. and M.G.: collection and/or assembly of data, data analysis; G.L.: collection and/or assembly of data; L.C., T.C. and X.W.: provision of study material; X.-J.L.: research design; F.-P.C: research conception and design; R.-H.X.: research conception and design, data analysis, manuscript writing, final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLS EXPRESS August 7, 2009.
The multidrug transporter ABCG2 in cell membranes enables various stem cells and cancer cells to efflux chemicals, including the fluorescent dye Hoechst 33342. The Hoechst− cells can be sorted out as a side population with stem cell properties. Abcg2 expression in mouse embryonic stem cells (ESCs) reduces accumulation of DNA-damaging metabolites in the cells, which helps prevent cell differentiation. Surprisingly, we found that human ESCs do not express ABCG2 and cannot efflux Hoechst. In contrast, trophoblasts and neural epithelial cells derived from human ESCs are ABCG2+ and Hoechst−. Human ESCs ectopically expressing ABCG2 become Hoechst−, more tolerant of toxicity of mitoxantrone, a substrate of ABCG2, and more capable of self-renewal in basic fibroblast growth factor (bFGF)-free condition than control cells. However, Hoechstlow cells sorted as a small subpopulation from human ESCs express lower levels of pluripotency markers than the Hoechsthigh cells. Similar results were observed with human induced pluripotent stem cells. Conversely, mouse ESCs are Abcg2+ and mouse trophoblasts, Abcg2−. Thus, absence of ABCG2 is a novel feature of human pluripotent stem cells, which distinguishes them from many other stem cells including mouse ESCs, and may be a reason why they are sensitive to suboptimal culture conditions. STEM CELLS 2009;27:2435–2445
The superfamily of ATP-binding cassette (ABC) cell membrane transporters was initially characterized on the basis of their role in multidrug resistance, as a result of their ability to efflux various chemotherapeutic agents in an energy-dependent manner . ABCG2, also named MXR, BCRP, or ABCP, is a half-transporter of this family, expressed in a variety of normal and malignant cells to efflux chemotherapeutic agents, for example, mitoxantrone, anthracyclines, and campothecins , as well as the fluorescent dye Hoechst 33342 from the cells . Via fluorescence-activated cell sorting (FACS) of mouse bone marrow cells incubated with the dye, a subset of Hoechst− cells can be identified and sorted out as a side population (SP), which possesses hematopoietic stem cell properties . During differentiation of hematopoietic stem cells, Abcg2 expression sharply declines, suggesting Abcg2 as a stem cell marker . Abcg2-null mice almost completely lost lin−/c-Kit+/Sca-1+ SP cells, the residual SP cells lacked repopulating ability, and the Abcg2− hematopoietic cells were sensitive to mitoxantrone treatment in mice that underwent transplantation .
SP cells can also be isolated from many other tissues such as the liver, blood, lung, heart, gonad, intestine, and cornea , as well as from cancers such as acute myeloid leukemia, breast cancer, liver cancer, glioma, and lung cancer . In these cases, the SP cells all express ABCG2 and associate with normal or cancer stem cell properties. However, an increasing number of facts challenge the association between ABCG2, SP, and stem cells. For example, ABCG2 expression is correlated neither to SP nor to hematopoietic progenitor function in human umbilical cord blood . During mouse development, the embryo remains efflux inactive until the blastocyst stage when the inner cell mass becomes efflux active . Mouse embryonic stem cells (ESCs), which are usually derived from the inner cell mass, also express Abcg2 and efflux Hoechst like SP cells .
Surprisingly, when we counterstained live human ESCs for their nuclei with Hoechst 33342, we observed that only undifferentiated human ESCs were Hoechst+, whereas some spontaneously differentiated human ESCs were Hoechst−. We followed up with this observation to clarify whether human ESCs possess the SP property at all. Our results demonstrate that ABCG2 is not expressed in human ESCs, but is present in their early stage derivatives such as trophoblasts and neural epithelial cells. Human ESCs ectopically expressing ABCG2 become more tolerant of the toxicity of mitoxantrone, an ABCG2 substrate, and less dependent on bFGF to sustain self-renewal, than control cells. Similar results were observed with human induced pluripotent stem (iPS) cells. In contrast, Abcg2 is present in mouse ESCs [5, 10], and we found that it is absent in mouse trophoblast stem (TS) cells as well as differentiated trophoblasts. These results highlight the lack of ABCG2 as a novel feature that distinguishes human pluripotent stem cells from many other stem cells including mouse ESCs.
MATERIALS AND METHODS
Rat anti-ABCG2 (for Western blotting), rabbit anti-NANOG, and mouse anti-β-ACTIN antibodies were from Abcam (Cambridge, MA, http://www.abcam.com). Mouse anti-ABCG2 antibody (for immunostaining) and anti-ABCG2 antibody conjugated with fluorescein isothiocyanate (FITC) (for FACS) were from Millipore (Billerica, MA, http://www.millipore.com); mouse anti-PAX6, rat anti-TROMA-1 was from Developmental Studies Hybridoma Bank (Iowa City, IA, http://www.uiowa.edu/∼dshbwww); mouse antibodies against OCT3/4, TRA-1-60, and TRA-1-81 were from Santa Cruz Biotechnology (Santa Cruz, CA); and rabbit antibodies against AKT and phosphorylated AKT were from Cell Signaling Technology (Danvers, MA, http://www.cellsignal.com). Anti-OCT3/4 antibody conjugated with Alexa-694 (for FACS) was from BD Biosciences (San Jose, CA, http://www.bdbiosciences.com). Bone morphogenetic protein 4 (BMP4) was from R&D Systems (Minneapolis, http://www.rndsystems.com), and Hoechst 33342 was from Invitrogen (Carlsbad, CA, http://www.invitrogen.com). Mitoxantrone, verapamil, and leukemia inhibitory factor (LIF) were from Sigma-Aldrich (St. Louis, http://www.sigmaaldrich.com).
Human ES cell lines H9, HUES1, and CT2, and human iPS cell lines iPS(IMR90)-1 and iPS(foreskin)-1 were cultured on plates coated with Matrigel (BD Biosciences) in human ES medium, that is, Dulbecco's modified Eagle's medium (DMEM)/F12 containing 20% knockout serum replacer, 0.1 mM nonessential amino acids, 1 mM L-glutamine (all from Invitrogen), and 0.1 mM β-mercaptoethanol (Sigma-Aldrich), which was conditioned on mouse embryonic fibroblast (MEF) as feeders and then supplemented with 4 ng/ml bFGF (Millipore) . Some results were obtained from human ESCs cultured directly on the feeders in unconditioned human ES medium supplemented with 4 ng/ml bFGF [12, 13] or on Matrigel in defined TeSR1 medium .
Mouse ESCs derived from a 129 strain mouse blastocyst were cultured on MEF feeders in mouse ES medium, that is, DMEM (Invitrogen) containing 20% fetal bovine serum (FBS) (Hyclone, Logan, UT, http://www.hyclone.com), 0.1 mM nonessential amino acids, 1 mM L-glutamine, and 0.1 mM β-mercaptoethanol, in the presence of 1,000 unit/ml LIF. Mouse TS cells (courtesy of Dr. Janet Rossant) were cultured on gelatin-coated plates in 70% MEF-conditioned mouse ES medium, 30% TS medium, that is, RPMI 1640 (Invitrogen) supplemented with 20% FBS, 1 mM sodium pyruvate, 0.1 mM β-mercaptoethanol, 2 mM L-glutamine. Human recombinant fibroblast growth factor 4 (FGF-4, 25 ng/ml; Sigma-Aldrich) and heparin (1 mg/ml) were added to aliquots of TS cell medium and used immediately.
For trophoblast differentiation , human ES or iPS cells were treated with 100 ng/ml BMP4 for up to 7 days. The cells were harvested at various times of the treatment and subjected to analyses by immunostaining, reverse-transcription polymerase chain reaction (RT-PCR), or Western blotting. The differentiation of neural epithelial cells from human ESCs was carried out as described previously . Briefly, human ES cell colonies were detached from MEF feeder cells at day 0 and suspended in human ES cell medium (without bFGF) for 4 days. Then these human ES cell aggregates were cultured in a neural medium consisting of DMEM/F12, N2 supplement, and 2 μg/ml heparin in the absence of growth factors. After adherence to plastic surface on day 6, human ES cell aggregates flattened down and developed to columnar cells that were organized to rosette structure at days 8-10. Subsequently, these early rosette cells (primitive neural epithelial cells) formed neural tube-like rosettes with lumens (definitive neural epithelial cells) at days 14-17 of differentiation .
Differentiation of mouse ESCs was induced by culturing the cells on gelatin-coated plates in mouse ES medium, in the absence of LIF and presence of 100 ng/ml BMP4 for 5 days . Differentiation of mouse TS cells was induced by culturing the TS cells in the mouse ES medium, in the absence of FGF4 and heparin for 5 days .
Lentiviral Transduction of ABCG2 to Human ESCs
A 1.8-kb BamH1-EcoRI fragment coding for the full-length ABCG2 gene was subcloned into the BamHI and EcoRI sites by replacing EGFP in the pSIN4-EF2-EGFP-IRES-Neo plasmid (courtesy of Dr. James Thomson), resulting in pSIN4-EF2-ABCG2-IRES-Neo. Alternatively, EGFP cassette was deleted by BamHI to generate pSIN4-EF2-IRES-Neo as a negative control. Each of the two plasmids was used to prepare lentivirus in 293FT cell line (Invitrogen) with Superfect (Qiagen, Valencia, CA, http://www1.qiagen.com). The lentiviral supernatants were harvested at 48 hours and 72 hours after transfection, pooled, and supplemented with 6 μg/ml polybrene (Sigma-Aldrich) for infection of human ESCs. H9 cells at 2-3 days after split were transduced with the lentiviral supernatants derived from pSIN4-EF2-ABCG2-IRES-Neo to express ABCG2. Supernatants derived from pSIN4-EF2-IRES-Neo were used to transduce sibling H9 cells as a mock control. Forty-eight hours after transduction, 300 μg/ml neomycin was added daily for selection of cell clones expressing the transgenes. Stable clones expressing ABCG2 or the mock control were obtained 12-15 days after transduction and confirmed.
Hoechst Dye Efflux Assay
Human ES or iPS cells and mouse ES or TS cells cultured before reaching 75% confluence were live stained with 5 μg/ml Hoechst 33342 at 37°C under swirling motion for 10 minutes. For specific inhibition of ABCG2, some cells were simultaneously treated with 50 μM verapamil. Subsequently, the cells were washed three times with cold PBS, and examined under fluorescence microscope (Olympus, CKX41, Tokyo, Japan, http://www.olympus-global.com) with a digital camera (15.2, 64 Mp, Shifting Pixel; Diagnostic Instruments, Inc., Sterling Heights, MI, http://www.diaginc.com), which captures images with 360/40 and 457/20 filters for excitation and emission, respectively.
Cells were fixed with 4% paraformaldehyde for 10 minutes and incubated in PBS containing 5% rabbit serum and 0.4% Triton X-100 for blocking and permeablization, respectively. PBS containing 0.5% Tween 20 (PBS-T) was used to dilute antibodies and wash the cells in the following procedures. The cells were incubated with antibodies against ABCG2 (diluted 1:50), OCT4, TRA-1-60, TRA-1-81 (all diluted 1:1000), or TROMA-I (without dilution) at 37°C for 1 hour, followed by washing with PBS-T three times. Afterward, the cells were incubated with fluorochrome-conjugated, corresponding secondary antibodies at 37°C for 45 minutes and washed with PBS-T three times. Finally, the cells were examined under fluorescence microscope to capture both phase and fluorescent images.
Cells were lysed in RIPA lysis buffer supplemented with phenylmethylsulphonyl fluoride, sodium orthovanadate, and protease inhibitor cocktail solutions (all from Santa Cruz Biotechnology) at 10 μl/ml each. The cell lysates were stored at −80°C before use. Proteins in the lysates were separated in a 10% SDS-polyacrylamide gel and transferred electrophoretically to polyvinylidene fluoride membranes (Bio-Rad Laboratories, Irvine, CA, http://www.bio-rad.com). The membranes were blocked with 5% nonfat milk and incubated with antibodies against ABCG2 (diluted 1:50), NANOG (1:500), AKT (1:2,000), phosphorylated AKT (1:2,000) or β-ACTIN (1:2,000) at room temperature for 1 hour followed by PBS-T washing three times. Subsequently, the membranes were incubated with horseradish peroxidase-conjugated corresponding secondary antibodies (diluted 1:5,000) at room temperature for 30 minutes followed by washing with PBS-T three times. Finally, target protein bands on the membranes were visualized using Immobilon Western Chemiluminescent HRP Substrate (Millipore).
Flow Cytometry Analysis and Cell Sorting
For Hoechst staining analysis, human ESCs cultured at approximately 75% confluence were dissociated by incubation with Accutase (Innovative Cell Technologies, San Diego, http://www.innovativecelltech.com) at 37°C for 10 minutes, then centrifuged and resuspended in 400 μl cold PBS containing 2% FBS, 10 mM HEPES buffer, and 2 μg/ml propidium iodide (PI) (Sigma-Aldrich). The cells were then stained with 5 μg/ml Hoechst 33342 for 10 minutes at 37°C while swirling. Flow cytometric analysis and cell sorting were performed on BD FACSAria Cell-Sorting System (BD Biosciences). Live cells were gated in by excluding PI-positive cells in both detectors. The sorting of Hoechsthigh and Hoechstlow cells was then decided on the basis of Hoechst fluorescence emission of the live cells in both the blue and red wavelengths.
For bFGF withdrawal analysis, human ESCs were split to Matrigel-coated plates, cultured in TeSR1 medium without bFGF for 3 days, and dissociated with Accutase. The dissociated cells were washed with FACS buffer (PBS supplemented with 0.1% sodium azide and 2% FBS), fixed with 4% paraformaldehyde for 10 minutes, and permeabilized with 0.1% Triton X-100. Then cells were incubated with anti-OCT3/4 antibody conjugated with Alexa-694 and anti-ABCG2 antibody conjugated with FITC for 60 minutes, washed with the FACS buffer, and proceeded to flow cytometry analysis with CellQuest Pro (BD Biosciences).
Low-Density Array Analysis
Hoechsthigh and Hoechstlow cells sorted through flow cytometry were subjected to RNA isolation and reverse transcription using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com) according to the manufacturer's protocol. cDNA derived from approximately 100 ng RNA per sample was applied to TaqMan Human Stem Cell Pluripotency Low-Density Array card for real-time PCR on ABI 7900HT Fast System. The samples were tested in triplicate and the data analyzed with RQ2.1 software and displayed as change in cycle threshold (ΔCt) in a scatter plot. All the array cards, real-time PCR system, and software were from Applied Biosystems.
RNA was isolated from cells using TRIzol reagent (Invitrogen), and cDNA was synthesized from the RNA using ThermoScript (Invitrogen), according to the manufacturer's instructions. Gene expression was assessed through PCR with primers for specific genes (supporting information Table 1) under the following conditions: an initial 5-minute denaturation at 95°C; followed by 30 cycles of 45 seconds of denaturation at 95°C, 45 seconds of annealing at 55°C, and 45 seconds of extension at 72°C; and completed with a final extension at 72°C for 10 minutes.
Apoptosis Detection by TUNEL Assays
Forty-eight hours after mitoxantrone treatment of cells, nuclear fragmentation in the cells was analyzed using the terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick-end labeling (TUNEL) Assay Kit (Roche Applied Science, Indianapolis, IN, http://www.roche-applied-science.com) according to the manufacturer's instructions. The numbers of apoptotic cells (TUNEL positive) and total cells (4′,6-diamidino-2-phenylindole positive) were counted in five views per well of cells under fluorescent microscope at ×10 magnification. The percentages of the apoptotic cells over the total cells from the five views were averaged, and arcsine was transformed for statistical analysis using analysis of variance. Three independent experiments were repeated. The data were expressed as mean ± standard deviation.
ABCG2 is Absent in Human Pluripotent Stem Cells but Present in Their Derivative Trophoblasts and Neural Progenitor Cells
As introduced above, this study started with an incidental observation when we counterstained live human ESCs for their nuclei with Hoechst 33342. We observed that undifferentiated (OCT4+) human ESCs (H9) were Hoechst+, whereas some spontaneously differentiated H9 cells were Hoechst− (Fig. 1A). We have previously demonstrated that BMPs can induce human ES cell differentiation to trophoblasts , and BMP activities present in the Serum Replacer (Invitrogen) used for human ES cell culture contribute to the spontaneous differentiation to mixed cell lineages including trophoblasts . We then stained live trophoblasts differentiated from BMP4-treated H9 cells with Hoechst 33342 and found that they were indeed Hoechst−, whereas undifferentiated H9 cells were Hoechst+ and ABCG2− (Fig. 1B).
As ABCG2 mediates the efflux of the Hoechst dye , we tested whether ABCG2 is expressed in these cells. Both ABCG2 transcripts and protein were detected highly in the BMP4-treated H9 cells and slightly in spontaneously differentiated H9 cells, but absent in untreated, undifferentiated control cells (Figs. 1B and 2A–2D). Similar results (supporting information Fig. 1A–1C) were observed with other human pluripotent stem cell lines. Human ES cell lines CT2 (derived in our laboratory) and HUES1 (derived in Melton laboratory) , and human iPS cell lines iPS(IMR90)-1 and iPS(foreskin)-1 (derived in Thomson laboratory)  were ABCG2−/low and Hoechst+, whereas trophoblasts differentiated from these cell lines following BMP4 treatment were ABCG2+ and Hoechst−. Reanalysis of microarray database data from our previous studies  revealed that the transcript levels of ABCG2 and some other ABC family members such as ABCA4, ABCC1, ABCC2, ABCD1, and ABCD3 in BMP4-treated H1 human ESCs increase to 6.0-, 2.1-, 2.4-, 2.0-, 2.6-, and 3.4-fold, respectively, on day 7 of treatment, compared with paired, untreated H1 cells, where ABCG2 expression increased the most steadily and was the highest.
To exclude the possibility that different ABCG2 transcripts are expressed in human ESCs cultured in various conditions, we tested three ABCG2 transcripts termed a, b, and c, which are generated by alternative usage of exons 1a, 1b, and 1c, respectively . We detected only the transcripts a and b, but not c, in BMP4-treated H9 cells cultured on not only Matrigel in mouse embryonic fibroblast-conditioned medium  a system used throughout this study but also on the fibroblast feeders directly  or on Matrigel in the defined medium TeSR1  (Fig. 2E). However, ABCG2c could be detected in human KG-1 dendritic-like cells (data not shown).
Since neural stem and progenitor cells can be isolated as Abcg2+ cells from mouse forebrain , we tested whether human ES cell-derived neural epithelial cells also express ABCG2. Indeed, ABCG2 was detected among neural epithelial cells (PAX6+) differentiated from H9 cells (Fig. 3). This suggests that ABCG2 is not exclusively expressed in trophoblasts, but also in human ES cell-derived neural epithelial cells.
Abcg2 Is Present in Mouse ESCs, but Absent in Mouse Trophoblasts
The differential expression of ABCG2 in human ESCs and trophoblasts prompted us to test whether this difference is also present between mouse ES and trophoblasts cells. We confirmed that mouse ESCs were indeed Abcg2+ and Hoechst− as reported [5, 10], and also found that randomly differentiated mouse ESCs (which occurred upon withdrawal of leukemia inhibitory factor and addition of BMP4) remained largely Abcg2+ and Hoechst− (Fig. 4A). Next we tested Abcg2 expression in mouse trophoblasts. Mouse TS cells can be isolated from mouse blastocyst and maintained in mouse embryonic fibroblast-conditioned medium containing 25 ng/ml FGF4 and 1 μg/ml heparin, and they differentiate to trophoblasts when cultured in unconditioned medium in the absence of FGF4 and heparin . Interestingly, neither mouse TS cells nor their differentiated trophoblasts expressed Abcg2, thus both were Hoechst+ (Fig. 4B). The absence of Abcg2 in mouse TS cells was confirmed by Western blotting (Fig. 4C). These data suggest that human and mouse ESCs and trophoblasts have completely opposite patterns of ABCG2/Abcg2 expression.
Ectopic ABCG2 Expression Renders Human ESCs More Resistant to Mitoxantrone, and Less Dependent on bFGF
Recently Abcg2 has been reported to help sustain self-renewal of mouse ESCs by pumping out DNA-damaging metabolites such as protoporphyrin . We decided to test whether ectopic expression of ABCG2 would confer similar actions to human ESCs. We first transduced H9 cells with ABCG2 lentiviral particles or control viral particles. Stable cell clones were established following drug selection (see Materials and Methods). The ABCG2-expressing cells became ABCG2+ and Hoechst−, whereas H9 cells derived from the mock transduction (negative control) remained ABCG2− and Hoechst+ (Fig. 5A). Like control cells, the ABCG2-expressing cells remained positive for pluripotency markers (supporting information Fig. 2A). Following mitoxantrone treatment, the percentage of apoptotic (TUNEL+) cells was much lower in ABCG2-expressing H9 cells and H9-derived trophoblasts than in untreated or mock-transduced H9 cells, and the ABCG2 inhibitor verapamil diminished the differences (Fig. 5B). These data suggest that ABCG2 can protect human ESCs and trophoblasts from cytotoxic substrates. Nevertheless, their ability to differentiate to trophoblasts (supporting information Fig. 2B) and neural epithelial cells (supporting information Fig. 2C) was not compromised.
It has been known that cooperation of the two branches of FGF signaling, mitogen-activated protein kinase/extracellular signal-related kinase (ERK) and phosphoinositide-3 kinase (PI3K)/AKT, is required to sustain self-renewal of human ESCs [27, 28]. We also found that the ABCG2-expressing H9 cells had higher phosphorylation level of AKT (Fig. 5C), and the majority of them remained pluripotent (OCT4+ and NANOG+) in bFGF-free condition for up to 3 days, whereas the control cells had already mostly differentiated or died off (Fig. 5D, 5E). This suggests that ABCG2 may partially compensate for the inhibition of FGF signaling through activation of the downstream effecter AKT. However, it does not appear that inhibition of FGF signaling markedly affects expression of ABC transporters in human ESCs, based on our previous microarray database .
The “Side Population” Among Human ESCs Has Decreased Expression of Pluripotency Markers
Although human ESCs appeared mostly Hoechst+ in phenotype, we asked whether they also contain Hoechst− cells and, if yes, what would be their pluripotency status. To address these questions, we used FACS to sort H9 cells and obtained two subsets of cells: a major Hoechsthigh population and a minor Hoechstlow population (Fig. 6A). Their gene expression profiles were analyzed using the Low-Density Stem Cell Array card (Applied Biosystems). This card included real-time RT-PCR primers to detect expression of multiple well-defined genes validated as markers for pluripotency and differentiation as well as endogenous controls. As shown in Figure 6B, the Hoechsthigh cells had higher transcriptional levels (indicated by lower ΔCt values) of the pluripotency marker genes such as OCT4 , TERT , FGF4, GABR3 , and NR5A  than the Hoechstlow cells, nevertheless, almost no transcripts for differentiation marker genes were detected in either subset of the cells. These results were confirmed by testing some of the marker genes through regular RT-PCR, with trophoblasts used as a positive control for ABCG2 expression (Fig. 6C). Therefore, the SP-associated Hoechst efflux inversely correlates to the pluripotency of human ESCs, and the Hoechstlow cells may represent spontaneously differentiating (or poised-to-differentiate) ESCs in the given culture.
Human and mouse ESCs share great similarities in cell morphology, gene expression profile, differentiation ability, and so on . Many differentiation protocols for human ESCs have been adapted from those for mouse ESCs . However, dramatic differences in cell surface markers and growth factor requirements have been discovered between ESCs from the two species. For examples, human ESCs express SSEA3 and SSEA4 , whereas mouse ESCs express SSEA1 . Activation of FGF and transforming growth factor β (TGFβ) signaling and inhibition of BMP signaling are required to sustain human ES cell self-renewal . However, activation of BMP and LIF signaling and inhibition of FGF/Erk signaling are required to sustain mouse ES cell self-renewal . Interestingly, LIF/signal transducer and activator of transcription 3 signaling is dispensable for human ESCs , and TGFβ signaling is dispensable for mouse ESCs . In addition, human ESCs differentiate to trophoblasts in response to BMPs , whereas mouse ESCs do so only when their Oct4 expression is repressed . This study revealed another substantial difference between human and mouse ESCs. Human ESCs do not express ABCG2 and cannot efflux Hoechst, thus lacking SP property. In contrast, mouse ESCs are Abcg2+ and Hoechst−, thus possessing SP property. Moreover, human ES cell-derived trophoblasts are ABCG2+ and Hoechst−, whereas mouse trophoblasts are Abcg2− and Hoechst+.
Our observations are not cell line-dependent because similar results were obtained from multiple human ES and iPS cell lines (Fig. 1B and supporting information Figure 1A–1C). Neither were they caused by differences between various culture conditions or by differential expression of the three ABCG2 transcripts (Fig. 2E). However, spontaneous differentiation among human ESCs can indeed cause elevated ABCG2 expression in a mixed cell population (Fig. 2B). This might be accountable for the detection of ABCG2 in human ESCs (HUES1) in a previous report , in case the cells analyzed were from suboptimal culture conditions.
As mentioned above, Abcg2 promotes mouse ES cell self-renewal by maintaining homoeostasis of intracellular metabolites, and inhibition of Abcg2 causes mouse ES cell differentiation accompanied by decline of Nanog expression . Consistently, human ESCs ectopically expressing ABCG2 had increased tolerance to the cytotoxicity of mitoxantrone (Fig. 5B), and even could expand in a bFGF-free culture medium without obvious differentiation for up to 3 days, whereas the control cells had already mostly differentiated or died off by then (Fig. 5D, 5E). These data indicate that lack of ABCG2 may be a reason why human ES as well as iPS cells have higher demand for optimal culture conditions and delicate care to sustain self-renewal than mouse ESCs. Further study is necessary to elucidate how the human pluripotent stem cells maintain their homeostasis in the absence of ABCG2 and presence of bFGF, given the fact that none of the detected ABC transporters in human ESCs is remarkably regulated by bFGF .
It has been known that ABCG2 is expressed in late-stage trophoblasts in the placenta, protecting the fetus by expelling drugs, xenobiotics, and metabolites across the placental barrier at midgestational ages [41, 42]. However, no information is available about the role of ABCG2 at early stages of human embryos. Human ESCs provide us with a great opportunity to address this question, as they may behave similarly to the inner cell mass, whereas human ES cell-derived trophoblasts may somehow mirror the early stage trophoblasts present in the trophectoderm of the blastocyst. The absence of ABCG2 in human ESCs and presence in their derivative trophoblasts implicate that human trophoblasts might protect the inner cell mass from detrimental substances in the microenvironment during blastocyst implantation. On the other hand, Abcg2 is present in mouse ESCs as well as mouse inner cell mass , but absent in mouse trophoblasts (Fig. 4). Abcg2-null mice develop normally without identifiable defects during gastrulation and placental formation . These data indicate that, during mouse blastocyst implantation, the trophoblasts might protect the inner cell mass through other ABC transporters in the trophectoderm.
Our data suggest that the cell membrane transporter ABCG2 is absent in human pluripotent stem cells including ES and iPS cells, but present in their derivative trophoblast and neural epithelial cells. This is opposite to the expression pattern of mouse Abcg2, which is present in mouse ESCs but absent in mouse trophoblasts. The absence of ABCG2 in human pluripotent stem cells makes them a unique stem cell type that does not have the property of side population cells, thus they are less resistant to certain cytotoxic chemicals than side population cells. Interestingly, ectopic expression of ABCG2 in human ESCs renders them more tolerant of bFGF-free culture. The mechanism behind this observation awaits further investigations.
We thank Dr. James Thomson for human ES cell lines H1 and H9, human iPS cell lines iPS(IMR90)-1 and iPS(foreskin)-1, and pSIN4-EF2-EGFP-IRES-Neo lentiviral plasmid; Dr. Douglas Melton for human ES cell line HUES1; Dr. Janet Rossant for mouse TS cell line; and Dr. Zihai Li for human KG-1 dendritic-like cell line. We also thank Diane Gran for flow cytometric analysis and cell sorting, Dr. Zhi-bo Wang for technical assistance for Western blotting, Drs. Gordon Carmichael and Lixia Yue for critical reading of the manuscript, and all the members of Xu laboratory for kind help and support. This work was supported by Connecticut Stem Cell Research Grants 06SCB14 and 06SCD02 to R.-H.X., and the fund from China Scholarship Council to H.Z. The contents in this work are solely the responsibility of the authors and do not necessarily represent the official views of the State of Connecticut.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.