Center for Stem Cell Biology, Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, Tennessee, USA
Patricia A. Labosky, Ph.D., Center for Stem Cell Biology, Department of Cell and Developmental Biology, 2213 Garland Avenue, 9465 MRB IV, Vanderbilt University School of Medicine, Nashville, Tennessee 37232-0494, USA. Telephone: 615-322-2540; Fax: 615-322-6645===
The Foxd3 forkhead transcription factor is required for maintaining pluripotent cells in the early mouse embryo and for the establishment of murine embryonic stem cell (ESC) lines. To begin to understand the role of Foxd3 in ESC maintenance, we derived ESC lines from blastocysts that carried two conditional Foxd3 alleles and a tamoxifen-inducible Cre transgene. Tamoxifen treatment produced a rapid and near complete loss of Foxd3 mRNA and protein. Foxd3-deficient ESCs maintained a normal proliferation rate but displayed increased apoptosis, and clonally dispersed ESCs showed a decreased ability to self-renew. Under either self-renewal or differentiation-promoting culture conditions we observed a strong, precocious differentiation of Foxd3 mutant ESCs along multiple lineages, including trophectoderm, endoderm, and mesendoderm. This profound alteration in biological behavior occurred in the face of continued expression of factors known to induce pluripotency, including Oct4, Sox2, and Nanog. We present a model for the role of Foxd3 in repressing differentiation, promoting self-renewal, and maintaining survival of mouse ESCs.
Disclosure of potential conflicts of interest is found at the end of this article.
Author contributions: Y.L.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing; P.A.L.: conception and design, financial support, collection and assembly of data, data analysis and interpretation, manuscript writing.
Murine embryonic stem cells (ESCs) are derived from the inner cell mass (ICM) of blastocyst-stage embryos and can be maintained indefinitely in vitro while retaining the ability to subsequently differentiate into all the cells of the adult animal. Understanding properties of ESCs and how self-renewal and pluripotency are regulated will have a large impact on developmental biology studies and regenerative medicine. Several transcription factors are required for ESC self-renewal and pluripotency, including Oct4, Sox2, and Nanog, and inactivation of these genes leads to loss of pluripotent stem cells and aberrant differentiation into extraembryonic trophoblast in the case of Oct4 and Sox2, or primitive endoderm in the case of Nanog [1, , , –5]. Recently, overexpression of a cocktail of transcription factors (Oct4, Sox2, c-Myc, and Klf4 or Oct4, Sox2, Lin-28, and Nanog) has resulted in the induction of pluripotency in somatic cells [6, , , –10]. These induced pluripotent stem cells (iPSCs) have all the properties of ESCs, but the mechanism of this induction is still unclear. Identification of factors immediately downstream of these transcription factors will be crucial.
Foxd3 is a forkhead transcription factor required for maintenance of progenitor cells in the ICM, trophoblast, and neural crest lineages [11, –13]. Foxd3−/− embryos die shortly after implantation, and cells in the mutant ICM and epiblast undergo extensive programmed cell death . ESCs express Foxd3, and expression is dramatically downregulated when cells are induced to differentiate , suggesting that Foxd3 expression in pluripotent stem cells is functionally significant. Together, this work illustrates the important role that Foxd3 plays in maintaining multipotent progenitor cells from divergent embryonic lineages. However, the early lethality of Foxd3−/− embryos and the consequent inability to establish Foxd3−/− ESC lines hampered efforts to study the role that Foxd3 plays in ESC maintenance.
To circumvent this problem, we derived ESC lines in which Cre-mediated inactivation of Foxd3 function can be temporally regulated. These Foxd3fl/fl;CAGG Cre-ERTM (Foxd3fl/fl;Cre-ER) ESCs were indistinguishable from normal ESCs in culture, and the Foxd3 coding region was deleted when cells were cultured in the presence of 4-hydroxytamoxifen (TM). Using this inducible system, we demonstrated that Foxd3 was not required for cell proliferation but that mutant ESCs underwent increased apoptosis, indicating that Foxd3 was required for ESC survival. Mutant ESCs were defective in their ability to form colonies from single cells, illustrating a requirement for Foxd3 in stem cell self-renewal. At the same time, although they were maintained under differentiation-inhibiting conditions, Foxd3 mutant ESCs do not respond to these cues and undergo extensive differentiation despite the maintenance of expression of multiple stem cell genes. Together, our results shape a deeper understanding of the biological roles of this transcription factor in murine ESCs and allow us to propose a model that will further our comprehension of mechanisms regulating maintenance of self-renewal and multipotency, the defining characteristics of all stem cells.
Materials and Methods
Generation of Inducible Foxd3 Mutant Murine ESC Lines
Foxd3fl mice were maintained on a 129S6/SvEvTac (Taconic Farms, Germantown, NY, http://www.taconic.com) genetic background . Mice carrying a tamoxifen-inducible variant of Cre recombinase (CAGG Cre-ERTM, also called Tg(cre/Esr1)5Amc)  (Jackson Laboratory, Bar Harbor, ME, http://www.jax.org) were backcrossed four times to 129S6/SvEvTac mice. The resulting animals were crossed to Foxd3fl/fl mice and a line of Foxd3fl/fl;CAGG Cre-ERTM established. These were interbred, and blastocysts were harvested at 3.5 days post coitum (dpc) using standard methods [16, 17]. Blastocysts were cultured on irradiated STO fibroblasts in ESC medium supplemented with 50 μM MEK1 inhibitor PD98059 (Cell Signaling Technology, Beverly, MA, http://www.cellsignal.com). After 3–4 days, ICM outgrowths were isolated and trypsinized in microdrops, and cell suspensions were transferred to fresh feeder layers. After 4–5 days, ESC lines were obvious in the cultures. Lines were cryopreserved at passages 3–4, and samples were lysed for DNA extraction. Individual cell lines were genotyped for the Foxd3fl allele and the presence of Cre using polymerase chain reaction (PCR) as described . Animal care was in accordance with Vanderbilt University Institutional Animal Care and Use Committee guidelines.
ESCs were cultured on irradiated mouse embryonic fibroblast (MEF) feeder cells using standard protocols . 4-Hydroxytamoxifen (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) was dissolved in ethanol at a 1 mM stock concentration and added to the medium on a daily basis at a concentration of 2 μM unless otherwise specified. Differentiation of ESCs was carried out by generating embryoid bodies (EBs) in suspension culture on ultralow-attachment dishes (Corning Life Sciences, Acton, MA, http://www.corning.com/lifesciences) in ES medium without addition of leukemia inhibitory factor (LIF).
Cell Number Analysis and Self-Renewal Assay
For cell number analysis, ESCs were plated at a density of 5 × 104 cells per well in 12-well dishes. Two to 4 days after plating, cells were dissociated with trypsin and trituration to a single-cell suspension and manually counted with a hemocytometer. Cell numbers were measured for multiple cell lines from at least three independent experiments.
To assay for self-renewal, ESCs treated with or without 2 μM TM for 2 days were dissociated and plated at clonal density (10 cells per cm2) on MEF feeders in 12-well dishes. At day 4, cells were fixed and stained for alkaline phosphatase activity according to the manufacturer's instructions (Millipore, Billerica, MA, http://www.millipore.com) and counterstained with 4,6-diamidino-2-phenylindole (DAPI). Colony formation efficiency was calculated by the counting the number of alkaline phosphatase (AP)-positive colonies divided by the total number of cells plated. Number of cells per colony was determined by manually counting DAPI-stained cells. Positively stained cells were manually counted by two individuals who were blinded to the genotypes and TM treatment. Colony formation efficiency and number of cells per colony were determined from three independent experiments.
ESCs were washed with phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde (PFA) in PBS for 30 minutes at room temperature. EBs were washed with PBS and fixed with 4% PFA at 4°C for 1 hour. After sinking in 30% sucrose at 4°C overnight, EBs were embedded in Tissue-Tek OCT Compound (Sakura Finetek, Torrance, CA, http://www.sakura.com), and cryosectioned at 10 μm. For immunofluorescence, cells or sections were permeabilized and blocked using 0.1% Triton X-100 (Sigma-Aldrich) and 1% goat serum (Jackson Immunoresearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com) in PBS. Primary antibodies were diluted in blocking buffer and incubated overnight at 4°C. Secondary antibodies diluted in blocking buffer were incubated at room temperature for 1 hour. Cells were counterstained with DAPI. Images were captured using a Zeiss Observer A.1 fluorescent microscope and AxioCam MRc5 camera (Carl Zeiss, Jena, Germany, http://www.zeiss.com). Cells were counted manually by two individuals.
Primary antibodies used include rabbit anti-Foxd3  (1:1,000), rabbit anti-phospho-histone 3 (pH3; 1:500; Millipore), mouse anti-Oct4 (1:500; Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com), rabbit anti-Sox2 (1:3,000; kind gift from Larysa Pevny), rabbit anti-Nanog (1:500; Invitrogen, Carlsbad, CA, http://www.invitrogen.com), mouse anti-Foxa2 (1:50; Developmental Studies Hybridoma Bank, Iowa City, IA, http://dshb.biology.uiowa.edu), and rabbit anti-Foxa2 (1:500; kind gift from Chin Chiang). Secondary antibodies used were Cy3 (Jackson Immunoresearch Laboratories) and Alexa 488 (Invitrogen) goat anti-rabbit or goat anti-mouse (1:500).
Proliferation and Terminal Deoxynucleotidyl Transferase dUTP Nick-End Labeling Assays
5-Bromo-2′-deoxyuridine (BrdU; 500 μg/ml; Sigma-Aldrich) was added to the culture medium for 1 hour. Cells were then fixed with 4% PFA in PBS, washed with PBS, and treated with 2 N HCl for 20 minutes. Immunocytochemistry was performed as described above using a mouse anti-BrdU (1:50; Becton Dickinson) antibody. Terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assays were performed using the In Situ Cell Death Detection Kit (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com) according to the manufacturer's instructions. BrdU-positive, pH3-positive, and TUNEL-positive cells were counted, and the percentages were calculated against total number of DAPI-stained cells. Data were collected from more than three independent experiments, with more than 7,000 DAPI-positive cells counted for each treatment.
RNA Isolation and Reverse Transcription PCR
Total RNA was isolated using the RNeasy Mini Kit or RNeasy Micro Kit (Qiagen, Hilden, Germany, http://www1.qiagen.com) followed by TURBO DNase treatment (Ambion, Austin, TX, http://www.ambion.com) according to the manufacturer's instructions. cDNA synthesis was carried out with 500 ng of total RNA using the SuperScript III quantitative reverse transcription (qRT)-PCR kit (Invitrogen). Real-time qRT-PCR was performed with the ABI Prism 7900 Sequence Detection System using SYBR Green Master Mix (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com). Samples were run in duplicate, and relative levels of each mRNA were examined by comparing cycle threshold values for each reaction among samples using the Hypoxanthine phosphoribosyl transferase (Hprt) gene as reference. Relative expression levels of each mRNA were measured from at least three independent experiments and compared by Student's t test. Primer sequences are listed in supplemental online Table 1.
Reduction of Foxd3 Expression by 4-Hydroxytamoxifen Treatment
Foxd3 is required for maintenance of the ICM and the establishment of ESC lines . Therefore, to generate a genetic lesion of Foxd3 in ESCs, we needed to develop a regulatable system. The Foxd3 conditional allele (Foxd3fl)  was designed so that Cre-mediated recombination deletes the entire coding region of Foxd3 gene. The CAGG Cre-ERTM transgenic line expresses a tamoxifen-regulated fusion protein of Cre fused to a modified ligand-binding domain of the estrogen receptor (ER) . Foxd3fl/fl;CAGG Cre-ERTM (Foxd3fl/fl;Cre-ER) mice appeared normal and healthy and were intercrossed to obtain blastocysts for ESC derivation. We obtained experimental lines (Foxd3fl/fl;Cre-ER) and control lines without the Cre-ER transgene (Foxd3fl/fl) in equal numbers. The genotyping scheme is shown in Figure 1A. PCR of genomic DNA isolated from wild-type, Foxd3fl/fl;Cre-ER, and Foxd3fl/fl ESCs confirmed the lack of wild-type and the presence of Foxd3fl alleles in untreated Foxd3fl/fl;Cre-ER and Foxd3fl/fl cells and the presence of the Cre transgene in Foxd3fl/fl;Cre-ER cells (Fig. 1B). When cultured in complete ESC medium without TM, ESCs of both genotypes appeared similar to wild-type ESCs and displayed comparable growth rates. When TM was added to the culture medium, the presence and intensity of the PCR product from the Foxd3fl allele was unchanged in Foxd3fl/fl control ESCs. However, recombination of the Foxd3fl allele was detected in TM-treated Foxd3fl/fl;Cre-ER experimental cells (Fig. 1B). Cre-mediated recombination was not complete, because some Foxd3fl PCR product was detected in the TM-treated cultures (Fig. 1B). Real-time quantitative PCR revealed that after 3 days of TM treatment, the PCR signal from the Foxd3fl allele in TM-treated Foxd3fl/fl;Cre-ER cultures ranged from 13% to 24% of levels in untreated ESCs (n = 3 experiments; data not shown).
We analyzed four independent cell lines of each genotype for Foxd3 expression at the mRNA and protein levels; E1, E2, E3, and E4 were experimental lines, and C1, C2, C3, and C4 were control lines. Immunocytochemistry with a Foxd3 antibody revealed that 70%–90% of the ESCs in untreated cultures expressed Foxd3 (Fig. 1E), consistent with the expression of Foxd3 in the ICM . However, when cultured in the presence of TM, Foxd3 mRNA and protein levels were significantly reduced in experimental cells but not control cells (Fig. 1C, 1D; also shown in Fig. 3A, 3C). Real-time qRT-PCR analysis revealed a greater than 90% reduction of Foxd3 mRNA in Foxd3fl/fl;Cre-ER ESCs 1 day after TM treatment, and after 2 days mRNA levels were nearly undetectable (Fig. 1D). Quantification of the number of Foxd3-expressing cells by immunocytochemistry showed that after 1 day of TM treatment, approximately 15% of the cells still contained Foxd3 protein (Fig. 1E, arrowheads). Two and 3 days of TM treatment resulted in further reduction of Foxd3-positive cells in the cultures (7.4% and 5.8%, respectively) and a decrease in total Foxd3 protein as measured by Western blot analysis (data not shown), suggesting that Cre-mediated recombination occurred in the vast majority of cells under these conditions.
Foxd3 Is Required for ESC Self-Renewal and Survival
To examine whether Foxd3 plays a role in growth and maintenance of ESCs, we quantified the effect of TM treatment on cell numbers in experimental and control ESCs. When Foxd3fl/fl;Cre-ER ESCs were cultured in the presence of 2 μM TM, the number of surviving cells after 2–4 days was significantly reduced compared with control cultures (Fig. 2A). Control Foxd3fl/fl ESCs cultured in the absence or presence of 2 μM TM had a similar number of surviving cells after 2–4 days of TM treatment (Fig. 2A). These results suggest that loss of Foxd3, but not treatment with TM, was responsible for the decreased cell numbers and colony size (supplemental online Fig. 1A) in ESCs. A higher concentration of TM (4 μM) showed a nonspecific toxic effect: control cell lines cultured in the presence of 4 μM TM had a reduced number of cells (Fig. 2A). Therefore, we used 2 μM TM for the rest of our experiments. Serial passaging of Foxd3fl/fl;Cre-ER ESCs with TM treatment (10 passages) resulted in fewer surviving cells at each passage; however, cells did not die off, and continued culture was possible in most cases (supplemental online Fig. 1B). However, many of these cells were Foxd3-positive, suggesting that they were selectively able to proliferate despite continued TM in the medium (data not shown).
Self-renewal of stem cells is defined by the ability of a single cell to give rise to a new colony of stem cells [18, –20]. Cells were cultured with or without TM for 2 days and plated at clonal density (10 cells per cm2). AP staining was performed after 4 days, and immunocytochemistry confirmed that the mutant AP-positive colonies no longer expressed Foxd3 (Fig. 2B). The efficiency of colony formation of untreated experimental cells was 44.4%, whereas TM-treated cells formed colonies at only 10.6% (Fig. 2C), suggesting a defect in self-renewal of Foxd3-deficient ESCs. In addition, although both untreated and 2 μM TM-treated cultures of Foxd3fl/fl;Cre-ER ESC lines contained AP-positive colonies (Fig. 2B), treated colonies contained three- to fourfold fewer cells than control colonies (Fig. 2D). When the same assays were performed with Foxd3fl/fl control ESCs, we saw no difference in either colony size or efficiency of colony formation with (42.7%) or without (45.6%) TM treatment (supplemental online Fig. 1C, 1D). These experiments demonstrate that Foxd3 is required for self-renewal of ESCs.
The reduced cell numbers and colony size that we observed in ESC cultures could be caused by increased cell death and/or decreased cell proliferation. We measured programmed cell death by TUNEL labeling, and although control cultures without TM treatment exhibited few apoptotic cells, we observed more Foxd3fl/fl;Cre-ER ESCs treated with TM undergoing apoptosis at 2 and 3 days after TM treatment compared with controls (Fig. 3A, arrows). Quantification of the percentage of TUNEL-positive cells showed a twofold increase in apoptosis after 2 or 3 days of TM treatment (Fig. 3B). There was no increase of apoptosis in Foxd3fl/fl control ESCs treated with TM (Fig. 3A). These data indicate that Foxd3 plays an important role in maintaining ESC survival. In mutant colonies, Foxd3-positive ESCs (Fig. 3A, arrowheads) and TUNEL-positive cells (Fig. 3A, arrows) were observed, but Foxd3/TUNEL double-positive cells were rarely seen (2 of 972 cells counted at day 2 and 3 of 1,034 for day 3), suggesting a cell-autonomous function of Foxd3 in preventing ESCs from undergoing programmed cell death. To examine proliferation of Foxd3 mutant cells, we used BrdU incorporation to identify cells in S-phase of the cell cycle and pH3 as a marker for cells in G2-M phase. In mutant cultures, many Foxd3-deficient ESCs were BrdU-positive (Fig. 3C, arrows), suggesting that Foxd3 is not required for ESCs to enter S-phase of the cell cycle. Quantification of BrdU-positive cells showed no significant difference between mutant and control cultures (Fig. 3D). Similarly, the percentage of pH3-positive cells in Foxd3 mutant ESC cultures was similar to that of untreated cultures (Fig. 3E, 3F). Together, these data show that the deletion of Foxd3 has no significant effect on ESC proliferation but results in increased cell death.
Maintenance of Stem Cell Gene Expression in Foxd3 Mutant ESCs
Pluripotent ESCs are maintained in an undifferentiated state by the function of a set of key transcription factors, including Oct4, Sox2, and Nanog [21, 22], and overexpression of these transcription factors in combination with Klf4, c-Myc, and/or Lin28 can reprogram somatic cells to adopt a stem cell phenotype [7, , –10, 23]. To examine whether the loss of Foxd3 affects the expression of these stem cell genes, we analyzed mRNA levels and protein expression of multiple stem cell transcription factors in Foxd3 mutant ESCs. Immunocytochemistry for Oct4, Sox2, and Nanog showed that all three proteins were expressed in Foxd3 mutant ESCs after 3 days of TM treatment (Fig. 4A). Double immunocytochemistry for Oct4 and Foxd3 confirmed that most Oct4-positive cells in mutant cultures were Foxd3-deficient (Fig. 4A), suggesting that Foxd3 is not required for maintaining the expression of these stem cell genes. A time-course analysis of mRNA levels for these genes along with Zfx, Esrrb, c-Myc, Tbx3, and Klf4 by qRT-PCR revealed a small decrease of transcript levels of some of the genes after days 1 and 2 of TM treatment, and by day 3, expression levels were either unchanged or slightly increased (Fig. 4B). Oct4 protein levels in the treated experimental cultures were also maintained as measured by Western blotting (data not shown). These findings indicate that Foxd3 is not required to maintain the expression of the stem cell genes analyzed and suggest that Foxd3 may function downstream of these genes, or in a parallel pathway, in mouse ESCs.
Aberrant Differentiation of Foxd3 Mutant ESCs
When ESCs are cultured in the presence of mitotically inactivated embryonic fibroblast feeders, serum, and LIF, the overwhelming majority of the cells maintain an undifferentiated phenotype. To examine the differentiation status of Foxd3 mutant ESCs cultured under conditions that normally inhibit differentiation, we used qRT-PCR to monitor expression of differentiation markers for several lineages. A two- to fourfold increase in mRNA levels of primitive endoderm markers Foxa2, AFP, and Sox17 in TM treated cultures was observed (Fig. 5A), indicating increased differentiation to primitive endoderm of Foxd3-deficient ESCs. Trophectoderm is an extraembryonic lineage that contributes to the placenta, and normally ESCs do not contribute to trophectoderm in chimeras or in vitro . Despite this lineage restriction, Foxd3 mutant ESCs showed 12-, 3.4-, and 4.4-fold increases in mRNAs for trophectoderm markers Cdx2, Fgfr2, and Csh1/PL1, respectively (Fig. 5B). After implantation, the ICM gives rise to primitive ectoderm/epiblast cells, which then contribute to cells in the three primary germ layers. One of the first differentiation markers expressed in epiblast cells in vivo, Fgf5 , was increased twofold in Foxd3 mutant ESCs (Fig. 5C). Goosecoid (Gsc) and Brachyury (T), molecular markers for mesendoderm , increased 5.4- and 12-fold, respectively (Fig. 5C). Together, these results suggest that Foxd3 normally functions to maintain ESCs in an undifferentiated state by repressing differentiation toward extraembryonic and embryonic lineages. Interestingly, expression levels of neuroectoderm markers Nestin, Pax6, and Sox1 either decreased or remained the same in mutant cells (Fig. 5D), suggesting that Foxd3 selectively represses commitment of epiblast cells into mesoderm and endoderm lineages, but not ectoderm.
During lineage diversification ESCs upregulate lineage-specific genes while downregulating self-renewal genes. One possible mechanism for precocious and aberrant differentiation is the disruption of this coordinated regulation, with the induction of lineage-specific gene networks at the same time that self-renewal gene expression is maintained. To test this possibility in Foxd3 mutant ESCs, we performed double immunocytochemistry for the endoderm marker Foxa2 and stem cell markers Oct4 and Sox2 (Fig. 5E). Both TM-treated Foxd3fl/fl;Cre-ER cultures (Fig. 5E) and untreated cultures (data not shown) contained very few cells that were double-positive for Foxa2 and Oct4 or Sox2 (Fig. 5E, arrowheads; 4%–5% cells in both mutant and control cultures), suggesting the coordination between differentiation and self-renewal was not disrupted in Foxd3 mutant ESCs. Surprisingly, the percentages of Oct4-positive and Foxa2-positive cells in control and mutant cultures were similar (Fig. 5F). Because we observed increased apoptosis in Foxd3 mutant cultures, we performed TUNEL analysis and immunolocalization for either Oct4 or Foxa2 to examine which population of cells preferentially undergoes apoptosis. Oct4 cells are mostly in the internal portion of the colonies, whereas Foxa2 cells are at the periphery (Fig. 5E, 5G). In control cultures, TUNEL/Oct4 double-positive cells were rarely observed (Fig. 5H; 1.5% of total Oct4-positive cells); in mutant cultures, the percentage of TUNEL/Oct4 double-positive cells was similar (Fig. 5G, 5H; 2%). In contrast, we measured 6.6% TUNEL/Foxa2 double-positive cells in control cultures versus a significantly higher 28.7% TUNEL/Foxa2 double-positive cells in mutant cultures (Fig. 5G, 5H). Furthermore, in mutant cultures, not only did we observe TUNEL/Foxa2 double-positive cells that were DAPI-positive (Fig. 5G, arrowheads), we also observed many cells at the edge of colonies that were TUNEL/Foxa2 double-positive but DAPI-negative, suggesting that these cells had lost their nuclear integrity (Fig. 5G, arrows). Together, these data indicate that Foxd3 mutant ESCs undergo aberrant differentiation, and these differentiated cells have a higher tendency for apoptosis.
Precocious Differentiation of Foxd3 Mutant Embryoid Bodies
ESCs not only have the capacity to self-renew in vitro, they are also capable of differentiating into multiple cell types. When ESCs are grown in suspension culture without LIF, the cells aggregate to form EBs. EBs differentiate in a manner similar to the early ICM [27, –29]. First, cells in the outer layer of the EBs differentiate to primitive endoderm that forms a thick basement membrane separating the primitive endoderm layer from the inner core of the EBs . Next, EBs undergo cavitation similar to formation of the proamniotic cavity in the embryo . Cells in the inner core resemble epiblast/primitive ectoderm and undergo differentiation into cells from all three germ layers.
Foxd3fl/fl;Cre-ER ESCs formed EBs when cultured in the presence of 2 μM TM. However, the EBs were much smaller than those from cells without TM treatment (supplemental online Fig. 2A), consistent with reduced cell survival in the absence of Foxd3. qRT-PCR confirmed that Foxd3 mRNA was decreased in TM-treated cells in day 2–8 EBs (supplemental online Fig. 2B, 2C). We examined primitive endoderm differentiation by qRT-PCR analysis of Foxa2 mRNA levels in control and mutant EBs and observed increased Foxa2 expression in Foxd3 mutant EBs at all stages examined (Fig. 6A). In control day 4 (D4) EBs, Foxa2 protein was detected primarily in the outer layer of primitive endoderm cells (Fig. 6B). In contrast, mutant EBs contained more Foxa2-positive cells, with many located within the core of the EBs. By D8, in addition to the outer layer Foxa2-positive primitive endoderm cells, a few Foxa2-positive cells were observed in the inner core of control EBs. The Foxa2-positive cells in Foxd3 mutant EBs were more numerous and widespread in the inside of the EBs (Fig. 6B). Comparing the percentage of Foxa2-positive cells in control and mutant EBs, we found a fivefold increase and a threefold increase in mutant EBs at D4 and D8, respectively (Fig. 6C). qRT-PCR analysis showed increased expression of another early endoderm marker, Sox17, in mutant EBs (Fig. 6D). mRNA levels of Foxa2 and Sox17 in mutant D2–D8 EBs were 3-fold to 18-fold higher than those in control EBs (Fig. 6A, 6D; note the logarithmic scale of these graphs). Two late endoderm markers, AFP and Alb1, showed a similar increase of expression in mutant EBs (Fig. 6E, 6F). Expression levels of the trophoblast marker Cdx2, and the mesendoderm marker Gsc, were also upregulated in mutant EBs, whereas the neuroectoderm marker Sox1 was slightly decreased (Fig. 6G–6I). Together, these data show that Foxd3 mutant EBs undergo spontaneous differentiation into the same repertoire of lineages that was demonstrated with mutant ESCs in the presence of LIF: mesendoderm and trophectoderm, but not neuroectoderm.
Foxd3 Maintains ESC Survival and Self-Renewal While Repressing Differentiation
ESCs are unique in their ability to remain self-renewing and pluripotent in vitro. These processes are tightly regulated by a network of transcription factors, only some of which are well understood . Our results establish two intertwined roles for Foxd3 in ESCs: this protein is required to maintain ESC survival and self-renewal while simultaneously repressing differentiation into either normal differentiation pathways (endoderm and mesoderm) or abnormal ones (trophectoderm) (Fig. 7, model).
The role of Foxd3 in regulating ESC survival is in agreement with our in vivo studies demonstrating similar roles for Foxd3 in both early embryos  and neural crest progenitors . Foxd3−/− embryos exhibit increased programmed cell death at 6.5 dpc, and in vitro cultures of mutant blastocysts also contain more TUNEL-positive cells than wild-type blastocyst cultures . Deletion of Foxd3 in the neural crest, another multipotent cell lineage, resulted in considerable apoptosis of neural crest progenitor cells . Our experiments here monitoring the expression of residual Foxd3 in treated cells together with TUNEL revealed very few, if any, double-positive cells, strongly suggesting a cell-autonomous function of Foxd3 in maintaining ESC survival. Furthermore, our results show that Foxd3 mutant ESCs maintaining Oct4 expression are protected from programmed cell death, whereas those mutant cells that have differentiated inappropriately (marked by Foxa2) have a much higher tendency to undergo apoptosis. These results suggest that the increased apoptosis in mutant ESCs could be a secondary consequence of aberrant differentiation. However, our results do not rule out the possibility that Foxd3 may directly regulate genes in apoptotic pathways. Interestingly, the proliferation capacity of mutant ESCs is comparable to that of control cells, even as their ability to self-renew is greatly decreased. Therefore, our results point to a specific function of Foxd3 in maintaining ESC survival and self-renewal, without affecting proliferation.
The early lethality of Foxd3 mutant embryos makes it difficult to study the role of Foxd3 in regulating differentiation . Our ESC data presented here demonstrate upregulation of multiple differentiation markers for several cell lineages in Foxd3 mutant ESCs and EBs, suggesting that Foxd3 may act as a gatekeeper in pluripotent ESCs to prevent cells from inappropriate differentiation, thereby maintaining the pool of stem cells. Similarly, in the trophoblast lineage, Foxd3 controls differentiation of trophoblast stem cells to maintain a multipotent progenitor cell population .
Positioning Foxd3 in the Transcription Factor Network Regulating ESC Self-Renewal and Pluripotency
Recent genome-wide studies suggest that Oct4, Sox2, and Nanog form a regulatory feedback circuit to maintain self-renewal and pluripotency in ESCs [32, 33]. These are not the only crucial factors; a parallel pathway involving transcription factors Esrrb and Tbx3 also represses differentiation to maintain ESC self-renewal and pluripotency . Expression of Oct4, Sox2, and/or Nanog in concert with Klf4, c-Myc, and/or Lin-28 reprograms human or murine somatic cells from multiple sources to adopt stem cell behavior and generate iPSCs that share all the properties of ESCs [6, , , –10]. In this study, we showed that mRNA levels of Pou5f1, Sox2, Nanog, Esrrb, Tbx3, Klf4, and c-Myc are maintained or slightly upregulated in Foxd3 mutant ESCs, suggesting that Foxd3 acts downstream of, or in parallel with, these factors to regulate self-renewal and inhibit spontaneous differentiation of ESCs. In fact, iPSCs induced by the expression of OCT4, SOX2, NANOG, and LIN28 from human somatic cells  express FOXD3 (James Thomson, personal communication), suggesting that FOXD3 could be one of the downstream executors to induce stem cell-like properties in iPSCs.
Context-Dependent Function of Foxd3 During Development
Our results add to previous findings suggesting that Foxd3 is a multifunctional protein playing a role in many divergent processes in the embryo and in progenitor cells. As a transcription factor, Foxd3 has been shown to act either as a repressor or an activator, depending upon the cellular context. During Xenopus mesoderm induction, FoxD3 functions as a repressor to induce mesoderm, and elegant structure-function analyses identified a strong transcriptional repression domain in the C terminus of Xenopus FoxD3 [34, 35]. Within this domain is a Groucho corepressor interaction motif conserved in Xenopus, zebrafish, chick, and mammals. In Xenopus, Foxd3 binds DNA and recruits Grg4 (Groucho-related gene 4) to repress transcription , likely by the recruitment of histone deacetylases that alter chromatin conformation and result in decreased transcription. It is unclear whether mammalian Foxd3 functions in the same manner as Xenopus FoxD3; mammalian Foxd3 has several sequence domains that are divergent from both Xenopus and zebrafish FoxD3, which might suggest that these orthologs function by different mechanisms. Expression of the murine homologs of Grg1–4, called Tle1–4, has been implicated in a variety of embryonic functions, including segmentation and neurogenesis (reviewed in [36, 37]), but thus far a role for these proteins in ESC pluripotency has not been examined. Embryonic expression of the murine Groucho homologs is widespread, making it likely that other cofactors might also be relevant to the regulation of Foxd3 protein function in a context-specific manner. Polycomb group (PcG) transcriptional repressors repress many developmental regulatory genes in murine ESCs . Because neither Groucho nor PcG proteins bind DNA directly, it is likely that they are recruited to target genes by sequence-specific DNA-binding transcription factors such as Foxd3.
In many circumstances Foxd3 acts as a repressor, but in other situations Foxd3 has transcriptional activator activity. In the human embryonic kidney 293 cell line, ectopically expressed Foxd3 activates expression of the endoderm genes Foxa1 and Foxa2 . Other work demonstrated that Foxd3 activates expression of a reporter driven by either the Nanog or the Pou5f1 promoter in ESCs, and ESCs transfected with short interfering RNA (siRNA) for Foxd3 have reduced expression of Nanog and Pou5f1 . These results support an activator role for Foxd3 maintaining Pou5f1 and Nanog expression. However, cotransfection of Foxd3 and a reporter preceded by multimers of a 30-base pair Foxd3-binding site from the Nanog promoter, Foxd3 overexpression causes decreased expression of this reporter in a dose-dependent manner, suggesting that Foxd3 can act as a repressor in ESCs . ESCs transfected with Foxd3 siRNA show reduced expression of Pou5f1 and Nanog, but the efficiency of the siRNA is unknown. Both Oct4 and Sox2 levels must be maintained in a very narrow window to maintain pluripotency of ESCs, and Oct4 and Nanog have a complex coregulation with biphasic regulation of Nanog by Oct4, such that a steady-state concentration of Oct4 maintains Nanog expression, whereas an elevated level of Oct4 represses Nanog mRNA levels . Therefore, it is clear that the cross-regulation of these self-renewal factors is complex, and individual transcription factors may function as repressors or activators in the same cell type.
These proposed divergent mechanisms of action for Foxd3 as either a repressor or an activator could indicate a role for Foxd3 as a convergent point for multiple pathways. Other Fox proteins have been proposed to function as so-called “pioneer factors” to modify chromatin; FoxA1 binds the Albumin1 enhancer prior to Alb1 transcription, and this binding of FoxA1 appears to open the chromatin and allow access to other transcription factors [41, 42]. Similar function has been ascribed to Foxe1 binding to the thyroperoxidase promoter and thereby opening chromatin . It is possible that Foxd3 is acting at promoters of differentiation-specific genes to allow access to other transcription factors that modulate transcription. In fact, this has been suggested by chromatin immunoprecipitation assays demonstrating that Foxd3 binds the inactive Alb1 promoter in ESCs and may modulate methylation of a CpG to control chromatin conformation . This is supported by our data in Foxd3 mutant ESCs indicating that along with other markers of endoderm development, Alb1 transcription is precociously increased in EBs, and it raises the possibility that Foxd3 may be acting at a key step in regulating multiple early lineage loci. Future work will be focused on identifying the molecular partners and targets of Foxd3 that function to carry out these essential roles in both ESCs and other multipotent precursors.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank Alison Dell for preliminary data on this project and Audrey Frist and Alison LeGrone for technical help. Thanks also go to Chin Chiang for the Foxa2 antibody, Larysa Pevny for the Sox2 antibody, Mark Magnuson's laboratory for use of the real-time PCR machine, and James Thomson and Ron Stewart for sharing unpublished data. Thanks also go to Dan Kessler, Mark Magnuson, Nathan Mundell, Brian Nelms, and Chris Wright for thoughtful comments on the manuscript. This work was supported by NIH Grant HD36720 (to P.A.L.). Y.L. is currently affiliated with Cell and Developmental Biology, College of Life Sciences, University of Dundee, Dundee, U.K.