Modulation of FOXD3 Activity in Human Embryonic Stem Cells Directs Pluripotency and Paraxial Mesoderm Fates§


  • Brigitte L. Arduini,

    Corresponding author
    1. Laboratory of Molecular Vertebrate Embryology, The Rockefeller University, New York, New York, USA
    • Brigitte L. Arduini, Rensselaer Stem Cell Center, Center for Biotechnology and Integrative Studies, Rensselaer Polytechnic Institute, Troy, NY, USA

      Ali H. Brivanlou, The Rockefeller University, 1230 York Ave, New York, New York 10065, USA

    Search for more papers by this author
  • Ali H. Brivanlou

    Corresponding author
    1. Laboratory of Molecular Vertebrate Embryology, The Rockefeller University, New York, New York, USA
    • Brigitte L. Arduini, Rensselaer Stem Cell Center, Center for Biotechnology and Integrative Studies, Rensselaer Polytechnic Institute, Troy, NY, USA

      Ali H. Brivanlou, The Rockefeller University, 1230 York Ave, New York, New York 10065, USA

    Search for more papers by this author
    • Telephone: 212-327-8656; Fax: 212-327-8685

  • Author contributions: B.L.A.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; A.H.B.: conception and design, data analysis and interpretation, and manuscript writing.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS August 9, 2012.


Transcription factor Foxd3 has been described in model systems as a key member of the pluripotency network in mice as well as being involved in the formation of many critical vertebrate cell types in vivo. Yet virtually nothing is known about roles of FOXD3 in human development and conflicting reports exist regarding its expression in human embryonic stem cells (hESCs). We find that FOXD3 is expressed at both the RNA and protein levels in undifferentiated hESCs and report a Foxd3 expression domain in paraxial mesoderm derivatives of wild-type mouse embryos. Furthermore, increasing FOXD3 activity in hESCs is sufficient for rapid and specific generation of mesenchymal cell types of the paraxial mesoderm, even under pluripotency maintenance conditions. Gene expression diagnostic of chondroblasts, skeletal myoblasts, osteoblasts, and adipoblast is observed within 48 hours of FOXD3 induction, as are morphological and genetic hallmarks of epithelial-to-mesenchymal transition. FOXD3-overexpressing cells can be maintained for several passages, while downregulation of the transgene leads to further differentiation. Loss-of-function also leads to differentiation, toward endoderm and mesoderm. Taken together, these data indicate that a balance of FOXD3 activity is required to maintain pluripotency. STEM Cells2012;30:2188–2198


Foxd3 is a multifunctional protein with context-dependent roles in a wide variety of early embryonic and extraembryonic cell populations. This winged helix transcription factor is expressed iteratively during vertebrate development, in the dorsal and paraxial mesoderm and neural crest, as well as the inner cell mass and trophectoderm of mammalian embryos [1–11]. Although Foxd3 function is crucial for formation of each of these essential populations in model organisms, almost nothing is known about the role of FOXD3 in the human milieu.

Loss-of-function analyses in mice indicate that Foxd3 function is required in the epiblast, and that the primitive streak fails to form in Foxd3−/− embryos [12]. These embryos die at approximately embryonic day 6.5 (E6.5), due to loss of epiblast via cell death. Furthermore, it was not possible to generate mouse embryonic stem cells (mESCs) from Foxd3−/− blastocysts [12]. Conditional knockout studies showed that while mESCs with floxed alleles of Foxd3 grow normally, Cre-mediated excision of Foxd3 results in failure of self-renewal. Additionally, many of these Foxd3−/− mESCs precociously upregulate differentiation genes and are eliminated by apoptosis, consistent with the in vivo results [13]. Together, these studies indicate that Foxd3 is required prior to the establishment of germ layers in mouse embryos, both to maintain self-renewal and to block differentiation. No data are available on these requirements in humans. Furthermore, discrepancies exist in the literature as to whether FOXD3 is expressed in human blastocysts or ESCs at all [2, 14-17].

In anamniotes, foxd3 is expressed in the organizer of Xenopus embryos and the tailbud of zebrafish, and has been implicated in the formation of mesoderm in these organisms [6, 18-20]. Expression of Xenopus foxd3 nonautonomously induces mesoderm-specific genes brachyury and goosecoid, in a Nodal-dependent fashion [18]. Subsequently, foxd3 is expressed in paraxial or somitic mesoderm [6, 18]. Foxd3 loss-of-function studies in teleosts reported morphological somite defects and abrogation of muscle-specific myf5 expression [19, 21], and these effects appear to be downstream of or in parallel with pax3 and myod, respectively. Thus, foxd3 appears to play roles in both the broad formation of the mesodermal germ layer and later in development of specific mesodermal domains. However, no studies have described a similar expression domain in paraxial mesoderm of a mammalian system, and little is known about the extent of Foxd3 participation in the developing somites in any vertebrate.

Studies involving overexpression of Foxd3 indicate that in certain cell populations it may be sufficient to cause morphological changes, such as epithelial-to-mesenchymal transition (EMT) within the chick neural tube [22, 23] or convergent extension in Xenopus dorsal mesoderm [18]. In contrast, constitutive overexpression in human embryonal carcinoma and myeloid cell lines did not result in morphological changes or upregulation of differentiation genes [4, 24]. To date, no Foxd3 overexpression studies have been undertaken in mouse or human ESCs (hESCs). We report that Foxd3 is expressed in paraxial mesoderm derivatives in wild-type mouse embryos, and that overexpression of human FOXD3 in hESCs maintained under pluripotency conditions results in dramatic loss of pluripotency gene expression, widespread EMT, and rapid paraxial mesoderm differentiation. Abrogation of FOXD3 function also leads to reduction of pluripotency gene expression, with concomitant differentiation toward definitive endoderm and mesoderm. These data suggest that, rather than acting in a binary fashion to promote self-renewal, a specific level of FOXD3 activity is required to maintain hESCs. Furthermore, our findings indicate that FOXD3 plays an evolutionarily conserved role in mammalian paraxial mesoderm development and point to the possibility of a broader role than previously described. As paraxial mesoderm cell fates are particularly elusive in differentiation of hESCs, this system provides a quick and efficient way to study somitic development downstream of specification by FOXD3.


hESC Culture and Transgenesis

hESC line RUES2 was maintained as previously described [25] in plates coated with Matrigel (BD Biosciences, Franklin Lakes, NJ, and mouse embryonic fibroblast-conditioned HUES medium (CM) with 20 ng/ml basic fibroblast growth factor (bFGF; Invitrogen, Grand Island, NY, For experiments, RUES2(CAG:TetOn; TRE:Flag-FOXD3) cells were induced with 2 μg/ml doxycycline (dox; Sigma, St. Louis, MO, added to the culture medium. RUES2(CAG:TetOn; TRE:Flag-FOXD3) cells were induced and passaged in CM+bFGF plus dox or differentiated in CM+bFGF minus dox. RUES2(TT:shFOXD3) cells were induced with 100 ng/ml dox. Media were refreshed daily.

For differentiation of hESCs toward neuroectoderm, RUES2 cells were seeded to Matrigel-coated plates in CM plus bFGF and expanded for 2 days. Cells were then switched to N2B27 medium for 4 days, followed by 6 days of differentiation in N2 medium. No exogenous growth factors were added.

Stable transgenic lines were nucleofected essentially as described [26]. Stable transfectants were selected with blasticidin and/or puromycin, followed by clonal passaging. Briefly, hESCs were pretreated for 1 hour with 10 μM ROCK inhibitor Y-27632 (Calbiochem, Darmstadt, Germany, and incubated in 0.05% trypsin-EDTA (Gibco, Grand Island, NY, Single colonies were lifted and transferred to a new Matrigel-coated dish containing CM+bFGF with ROCK inhibitor. Clonal passaging was repeated from two to four times per line, followed by expansion and analysis. Embryoid bodies (EBs) were generated from each cell line as previously described [27]. Cell clumps were harvested with dispase (Invitrogen) and plated in nonconditioned HUESM suspension on agar-coated plates. EBs were harvested at 14 days and 28 days of growth, processed for RT-PCR, and assayed for expression of genes from each of the primary germ layers. All cell lines described were submitted to Cell Line Genetics LLC (Madison, WI,, per company instructions, for karyotype analysis and were found to be normal.

Plasmid Construction

The coding sequence of human FOXD3 was amplified from undifferentiated RUES2 cDNA and subcloned into an enhanced PiggyBAC (epB) transposable element vector [26] downstream of the TRE-Tight promoter (Clontech, Mountain View, CA, and in-frame with a synthetic flag tag. Several short hairpin sequences were designed against the FOXD3 coding sequence, synthesized (IDT, Coralville, IA, and subcloned individually into the synthetic intron of the epB-bsd-TT-intRFP vector (A. Rosa and A.H. Brivanlou, unpublished data). Two short hairpin RNAs (shRNAs) achieved knockdown of FOXD3 in hESCs, shFOXD3 #1 (sense sequence 5′-GCCCAAGAACAGCCTAGTG-3′) and shFOXD3 #3 (sense sequence 5′-GTCCGAGGACATGTTCGACAA-3′). Both shRNAs gave similar results, although knockdown by shFOXD3 #3 was more efficient than shFOXD3 #1. Data presented are for shFOXD3 #3 unless otherwise stated. Human FOXA2 coding sequence was amplified from clone 3940139 (Open Biosystems, Lafayette, CO,; Accession Number BC011780) and subcloned into the epB-puro-TT vector in-frame with a synthetic Flag tag.

RNA Analysis

RNA was isolated from adherent and EB culture using Trizol reagent (Invitrogen, Grand Island, NY,, treated with a DNA-free kit (Ambion, Grand Island, NY,, and reverse transcribed with Transcriptor First Strand cDNA Synthesis kit (Roche, Indianapolis, IN, Real time RT-PCR was performed with LightCycler 480 SYBR Green I (Roche) and analyzed with Matlab. Results were normalized to ATP5O expression and displayed as fold change relative to uninduced cells. Semiquantitative RT-PCR was performed with GoTaq Green reagents (Promega, Madison, WI, Primer sequences and number of cycles are provided as supporting information.

Immunostaining and Western Blot

Cells grown in adherent culture on Matrigel-coated optical plastic dishes (ibidi, Planegg/Martinsried, Germany, were fixed and immunostained essentially as previously described [28]. Secondary antibodies were AlexaFluor488-, AlexaFluor555-, and AlexaFluor647-conjugated donkey anti-rabbit/anti-mouse/anti-goat antibodies (Molecular Probes). Cells were counterstained with either Sytox or 4′,6-diamidino-2-Phenylindole (DAPI) (Molecular Probes, Grand Island, NY, Images were taken on either a Zeiss LSM 510 confocal microscope (AHB lab) or an Olympus IX71 inverted microscope (Rockefeller University Bio-Imaging Resource Center). Images used for direct comparison of conditions were taken together and processed identically in LSM Viewer or ImageJ. Western blot was performed as previously described [28]. Secondary antibodies were horse radish peroxidase (HRP) anti-rabbit/anti-mouse, blots were developed with enhanced chemiluminescence (ECL) detection reagent (Amersham Biosciences, Pittsburgh, PA, Primary antibodies and dilutions are provided as supporting information.

Cell Counts

Cells were fixed and processed for immunofluorescence of PHOSPHO-HISTONEH3 or activated CASPASE3 as described above. Plates were imaged with an ImageXpress Microtiter Plate Reader available through the Rockefeller University High-Throughput Screening Resource Center. Cells were counted using Matlab and results were manually checked for accuracy.


Wild-type (B6CBA) embryos were collected from timed matings by dissection in ice-cold phosphate-buffered saline (PBS). Embryos were fixed in 4% paraformaldehyde, washed in PBS, and stored at 4°C. Whole-mount immunofluorescence was performed essentially as described [29], except that embryos were permeabilized with distilled water rather than methanol. Whole-mount embryos were cleared with benzoic acid-benzyl benzoate and imaged on a Zeiss LSM 510 confocal microscope. For sections, embryos were kept in 30% sucrose at 4°C overnight, mounted and frozen in optimal cutting temperature (OCT) compound, and sectioned at 12 μm.


Endogenous FOXD3 Expression in hESCs

Although some studies have reported FOXD3 RNA expression in hESCs and human blastocysts [2, 16, 17], others indicate that FOXD3 is not expressed in pluripotent hESCs [14, 15]. This is likely due to the GC-rich (72%) and repetitive nature of the transcript. We found that while the transcript is not detected by typical PCR protocols when targeting the 5′ end (data not shown), the transcript is readily identified in undifferentiated hESCs by semiquantitative RT-PCR with primers directed at the 3′ end of the transcript (Fig. 1, supporting information Table S1). As at least one reference indicates that Foxd3 protein may be present in either the nucleus or the cytoplasm in mice [7], we next examined protein localization. Consistent with the case in mESCs, FOXD3 is expressed and localized to the nucleus in undifferentiated hESCs (Fig. 1). Upon differentiation toward neuroectoderm in adherent, serum-free conditions (B.L. Arduini and A.H. Brivanlou, unpublished data), FOXD3 is downregulated at the RNA level and the protein level and is not coexpressed with PAX6, a gene diagnostic of neuroepithelium (Fig. 1).

Figure 1.

Endogenous expression of human FOXD3. (A): FOXD3 protein is coexpressed with OCT4 and localized to the nucleus of undifferentiated human embryonic stem cells (hESCs) (undiff). After 10 days of differentiation toward mixed ectodermal cell fates (diff 10d), FOXD3 is downregulated and is not coexpressed with PAX6 in presumptive neural epithelia. Scale bars = 50 μm. (B): Semiquantitative RT-PCR confirms that FOXD3 RNA is present in undifferentiated hESCs but below the level of detection in differentiated ectodermal cultures. (C): Immunofluoresence for Foxd3 (red) and Pax3 (green) in whole mount embryonic day 10 (E10) wild-type mouse embryos. Scale bars = 100 μm. (D): Immunofluorescence on transverse sections at midtrunk levels in E10 wild-type mouse embryos. Scale bars = 100 μm. Arrowhead points to epidermis. Abbreviations: D, dermomyotome; M, mesenchyme; NT, neural tube.

We also examined expression of Foxd3 in mouse embryos using immunofluorescence. In addition to known expression domains in the inner cell mass and neural crest ([4, 5, 9, 11] and data not shown), we find that Foxd3 is expressed adjacent to Pax3+ dermomyotome at midtrunk levels in E10.0 wild-type mice (Fig. 1), consistent with somite expression observed in nonmammalian systems [6, 8].

Overexpression of FOXD3 in hESCs Leads to EMT

In order to investigate the role of FOXD3 in human development, we generated stable, conditional transgenic hESC lines using the epB transposable element system [26]. Either untagged or N-terminally flag-tagged human FOXD3 was subcloned downstream of the TRE-Tight promoter (Fig. 2). Stable RUES2 cell lines were generated in which TetOn activator was constitutively expressed and FOXD3 was induced by TetOn with addition of dox to the culture medium, hereafter RUES2(Flag-FOXD3). Both tagged and untagged FOXD3 gave the same results in multiple clonal lines, data presented are for the flag-tagged version.

Figure 2.

FOXD3 overexpression generates epithelial-to-mesenchymal transition. (A): Schematic of gain-of-function transposable element construct. (B): Progressive cell morphology change in RUES2 cells overexpressing FOXD3. Scale bars = 100 μm. (C): Semiquantitative RT-PCR for SNAIL2 and E-CADHERIN (CDH1) expression in RUES2(Flag-FOXD3), undifferentiated (−), or 96 hours of FOXD3 overexpression (+). Embryoid body cDNA was used as positive control. (D): Western blot for CDH1 protein in RUES2 (Flag-FOXD3), undifferentiated (−), or 96 hours of FOXD3 overexpression (+). TUBA was used as a loading control. (E): Immunofluorescence for CDH1 (red) at 0, 24, 48, and 96 hours of induction. Scale bars = 50 μm. Abbreviations: bsd, blasticidin resistance; CAG, chick beta-globin constitutive promoter; Dox, doxycycline; pA, SV40 polyadenylation sequence; puro, puromycin resistance; TetOn, rtTA activator; TRE, TRE-Tight inducible promoter; UBC, human UBIQUITINASE C constitutive promoter; TUBA, ALPHA-TUBULIN.

Quantitative RT-PCR at 0–144 hours of induction indicates tight regulation of the transgene and ∼ 100-fold activation over uninduced cells (supporting information Fig. S1). Immunofluorescence with an anti-Flag antibody shows relatively uniform expression across the cell population and Western blot analysis indicates that a maximal level of protein is reached within 12 hours of dox induction (supporting information Fig. S1). Within 36 hours of dox induction, some cells become mesenchymal in character, separating from the edges of colonies, and SNAIL2, a key gene associated with EMT, is upregulated at the RNA level at 96 hours of induction (Fig. 2). With continued transgene expression, progressively more cells undergo this morphological change. By 144 hours of induction, RUES2(Flag-FOXD3) cultures no longer contain the tightly packed colonies that are characteristic of undifferentiated hESCs. However, RNA levels of epithelial adhesion molecule E-CADHERIN (CDH1) remain unchanged at 96 hours of induction (Fig. 2). Noting that CDH1 function is also modulated by post-transcriptional means, we investigated expression of CDH1 protein by Western blot and immunofluorescence. CDH1 levels are comparable in uninduced and induced cells at 96 hours (Fig. 2). Conversely, CDH1 strikingly demarcates the cell membrane in uninduced cells but is no longer strongly localized to the cell membrane in FOXD3-overexpressing cells at 96 hours (Fig. 2). Together, these data indicate that FOXD3 overexpression is sufficient to induce morphological and molecular hallmarks of EMT in hESCs.

FOXD3 Overexpression Leads to Paraxial Mesoderm Differentiation

Many cell types undergo EMT during embryogenesis. In order to characterize the nature of FOXD3-overexpressing cells, we analyzed expression of pluripotency- and differentiation-associated genes. Mesoderm-specific genes BRACHYURY and FOXC2 are upregulated upon FOXD3 overexpression, based on real time RT-PCR analysis (Fig. 3). While BRACHYURY is expressed in a pan-mesodermal fashion in vivo, FOXC2 expression is restricted to the presomitic domain of the mesoderm [30]. FOXF1, diagnostic of more ventral mesoderm, was not expressed at any of the time points examined, nor were the dorsal/organizer genes FOXA2 and GOOSECOID (GSC) upregulated (Fig. 3). While BRACHYURY expression is downregulated after 48 hours of induction, FOXC2 expression persists (Fig. 3). These data suggested specification of paraxial or somitic mesoderm. Indeed, genes diagnostic of somite derivatives, including MYF5 (skeletal muscle), SOX9 and COLLAGEN2 (chondroblasts), RUNX2 (osteoblasts), and PPARG (adipoblasts), were upregulated upon FOXD3 overexpression (Fig. 3 and data not shown). Notably, while KDR/FLK1 and PDGFRA are both expressed in undifferentiated cells, both genes are downregulated, while PDGFRB expression increases slightly (Fig. 3). In vivo, these genes are coexpressed in mesendoderm, but later diverge, with KDR expressed in endothelial derivatives and PDGFRB expressed by chondroblasts. These data are consistent with specification of paraxial, but not dorsal, lateral, or ventral, mesoderm.

Figure 3.

Pan-mesodermal and paraxial mesodermal genes are upregulated by FOXD3 overexpression. (A): Real time RT-PCR analyses of gene expression from 0 to 144 hours of induction by doxycycline (dox). Red dashed lines indicate a value of 1 (control condition). Transcripts for FOXF1, PAX3, and PAX7 were below detectable limits at all time points examined (cycles >40.00). (B, C): Immunofluorescence for differentiation-associated proteins at 144 hours of dox induction. Scale bars = 100 μm.

Some of these genes, as well as mesenchymal morphology, are also expressed by neural crest cells in vivo. To investigate the possibility that FOXD3-overexpressing cells were neural crest cells, we examined additional markers at the RNA and protein levels. Upregulation of TFAP2A was observed consistently at 96 and 144 hours of induction, while significant expression of SOX10 was not observed at any time point examined (Fig. 3). Additionally, HNK1 and P75 are expressed by distinct populations in undifferentiated and FOXD3-overexpressing hESCs, while doubly positive cells are rare (Fig. 3). TFAP2A is known to be expressed in limb mesenchyme [31], while P75 and HNK1 immunoreactivities have been described in somite-derived chondroblasts and myoblasts, respectively [32–34]. Thus, the temporal and spatial profiles of gene expression remain consistent with a paraxial mesoderm fate for RUES2(Flag-FOXD3) cells.

Genes diagnostic of ectoderm (PAX6) and endoderm (SOX17 and GATA6) were not upregulated (supporting information Fig. S2 and data not shown). Human chorionic gonadotropin (HCG) subunits alpha and beta5 are upregulated before 36 hours of induction. Conversely, significant expression of trophectoderm-associated CDX2, GCM, and MESP2 was not observed at any of the time points examined (supporting information Fig. S2), suggesting that HCG expression reflects mesoderm rather than trophectoderm differentiation. Pluripotency-associated genes OCT4, NANOG, and LEFTYB are strongly downregulated at 96 hours of FOXD3 overexpression (supporting information Fig. S2). Immunofluorescence for OCT4 and NANOG at 96 hours of induction reveals that the reduction of protein is relatively uniform across the population, rather than variable between cells (supporting information Fig. S2). Taken together, these data suggest that continuous FOXD3 overexpression leads to rapid and selective specification of paraxial mesoderm.

Next, we asked whether these effects were a general feature of winged helix transcription factors. Using a FOXA2 transgene flag-tagged identically to the FOXD3 construct described above, we generated stable, conditional transgenic lines. Continuous overexpression of FOXA2 did alter the morphology and transcriptional profile of cells within 96 hours (supporting information Fig. S3). However, these changes were significantly different from those seen upon FOXD3 overexpression. Pluripotency genes OCT4 and NANOG were only slightly downregulated, while genes diagnostic of endoderm and mesoderm were upregulated even at 96 hours of induction. Furthermore, cells took on a slightly flattened morphology resembling that seen upon FOXD3 loss-of-function discussed below, rather than the mesenchymal phenotype of FOXD3 overexpression. These data are consistent with previously published reports of wild-type and Flag-tagged FOXA2 activity and the well-established role of this transcription factor in endoderm development [35–39].

FOXD3-Overexpressing Cells Maintain a Stable Phenotype but Undergo Further Differentiation upon Transgene Downregulation

In order to determine the growth and maintenance properties of RUES(Flag-FOXD3) cells during transgene induction, we performed immunofluorescence and cell counting for proliferating (PHOSPHO-HISTONEH3+) and dying (activated CASPASE3+) cells (Fig. 4). Cell counts of at least 1,000 cells at each time point indicate that the rate of proliferation decreases with continued FOXD3 overexpression. Cell death is increased relative to uninduced cells at multiple time points.

Figure 4.

FOXD3 overexpression affects cell proliferation and cell death. (A): Percentage of P-HISH3-positive cells detected by immunofluorescence, compared to total nuclei identified by sytox counterstain. Error bars indicate SEM (*, p <.1; **, p <.01; ***, p <.0001). (B): Percentage of activated CSP3-positive cells detected by immunofluorescence, compared to total nuclei identified by sytox counterstain. Error bars indicate SEM (*, p <.1; **, p <.01). Abbreviations: CSP3, CASPASE3; P-HISH3, PHOSPHO-HISTONE H3.

However, cell counts for total number of cells indicate that FOXD3-overexpressing cells continue to expand over 8 days of transgene activation (Fig. 5). After 144 hours of induction, RUES2(Flag-FOXD3) cells maintained in dox for up to three passages (approximately 21 days) retained a mesenchymal morphology (Fig. 5). Furthermore, these cells can be passaged by trypsinization without ROCK inhibitor treatment, a property distinct from undifferentiated wild-type or transgenic RUES2 cells. Passaged RUES2(Flag-FOXD3) cells maintained in dox continued to express proteins that were present at 144 hours, including SOX9, TFAP2A, HNK1, and P75 (Fig. 5) and could be cryogenically preserved, thawed, and re-expanded (data not shown). In contrast, retraction of dox after 144 hours leads to formation of clustered foci of cells (Fig. 5). These foci were formed within 14 days of dox retraction (20 total days of differentiation) and expressed COLLAGEN2A1, suggestive of chondroblast differentiation.

Figure 5.

FOXD3 gain-of-function human embryonic stem cells can be maintained in dox or further differentiated with retraction of dox. (A): Growth curve of RUES2(Flag-FOXD3) cells for 8 days in the presence of dox. (B): Comparison of RUES2(Flag-FOXD3) cell morphology at 8 days and after three passages (∼ 3 weeks total induction). Scale bars = 50 μm. (C): Immunofluorescence for TFAP2A and SOX9 at passage 3. Scale bars = 50 μm. (D): Immunofluorescence for P75 and HNK1 at passage 3. Scale bars = 50 μm. (E): Bright-field image of RUES2(Flag-FOXD3) cells at 20 days of differentiation (6 days in the presence of dox followed by 14 days without dox). Scale bars = 50 μm. (F): Immunofluorescence for COL2A1 in RUES2(Flag-FOXD3) cells at 20 days of differentiation. Scale bars = 50 μm. Abbreviation: COL2A1, COLLAGEN 2A1.

FOXD3 Loss-of-Function Leads to Mixed Differentiation Toward Definitive Mesoderm and Endoderm Fates

In order to achieve FOXD3 loss-of-function, we subcloned shRNAs against the FOXD3 coding sequence into a synthetic intron within the tagRFP gene (intRFP). TagRFP expression is driven by the TRE-Tight promoter in an ePB transposable element that also contains a constitutive TetOn cassette and blasticidin resistance (ePB-bsd-TT-intRFP, Fig. 6). Upon dox induction, shFOXD3 is released via intron splicing and TagRFP is expressed as a lineage tracer. Multiple shRNAs gave similar results (Fig. 6, supporting information Fig. S4). A slight reduction of FOXD3 RNA was observed with real time RT-PCR after 96 hours of shRNA induction (Fig. 6). Loss of FOXD3 protein was confirmed by both Western blot and immunofluorescence at 96 hours of induction (Fig. 6).

Figure 6.

FOXD3 loss-of-function leads to precocious differentiation. (A): Schematic of loss-of-function transposable element construct. (B): Western blot for FOXD3 indicates reduction at the protein level. TUBA was used as a loading control. (C): Immunofluorescence for FOXD3 at 96 hours −/+ dox (RUES2(shFOXD3 #3)). Scale bars = 20 μm. (D): Bright-field imaging of live cells at 96 hours −/+ dox. Scale bars = 100 μm. (E): Real time RT-PCR for pluripotency and differentiation genes at 96 hours −/+ dox. PAX6, FOXC2, FOXF1, and CHORDIN transcripts were not detected (>40 cycles). Abbreviations: bsd, blasticidin resistance; pA, SV40 polyadenylation signal; RFP, intRFP; TetOn, rtTA activator; TRE, TRE-Tight inducible promoter; TUBA, ALPHA-TUBULIN; UBC, human UBIQUITINASE C constitutive promoter.

Real time RT-PCR after 96 hours of dox induction revealed upregulation of differentiation-associated genes BRACHYURY, FOXA2, GSC, MIXL1, and CDX2, while pluripotency gene expression was not significantly reduced (Fig. 6). Using immunofluorescence to examine these effects at single-cell resolution, we found that differentiation genes are upregulated in patches, while OCT4 was only slightly reduced across the population (Fig. 7, supporting information Fig. S4). CDH1 protein, while still present, appeared to be less restricted to the membrane in a subset of FOXD3 loss-of-function cells (Fig. 6, supporting information Fig. S4).

Figure 7.

Upregulation of differentiation markers in RUES2(shFOXD3 #3) cells. (A): Immunofluorescence for BRACHYURY (T), FOXA2, E-CADHERIN (CDH1), and OCT4 at 96 hours −/+ dox. Although nonspecific adherence to matrigel is evident in FOXA2 immunofluorescence experiments, nuclear staining is evident only in the plus dox condition. Scale bars = 20 μm. (B): Percentage of P-HISH3-positive cells detected by immunofluorescence, compared to total nuclei identified by sytox counterstain. Error bars indicate SEM. (C): Percentage of activated CSP3-positive cells detected by immunofluorescence, compared to total nuclei identified by sytox counterstain. Error bars indicate SEM (*, p <.2). (D): Model of FOXD3 activity in cell fate decisions from human embryonic stem cells. Abbreviations: CSP3, CASPASE3; P-HISH3, PHOSPHO-HISTONEH3; RFP, red fluorescent protein.

We further analyzed expression of FOXQ1 and CXCR4 (definitive endoderm) and LAMININ BETA1 (extraembryonic endoderm). Definitive endoderm genes are upregulated upon FOXD3 knockdown, while extraembryonic genes are not (supporting information Fig. S4). Furthermore, upregulation of FOXA2 but not GSC or CHORDIN suggests that these cells do not represent the dorsal-most mesodermal fates (Fig. 6, supporting information Fig. S4). Together, these data indicate that FOXD3 function is necessary to repress differentiation, consistent with observations in mESCs [13]. Upon reduction of FOXD3 activity, hESCs appear to differentiate spontaneously toward embryonic fates supported by culture conditions.

Additionally, we examined cell proliferation and apoptosis in FOXD3 loss-of-function cells using PHOSPHO-HISTONEH3 and activated CASPASE3 immunoreactivity, respectively (Fig. 7). We find that while proliferative capacity remains relatively unchanged, a small but consistent increase in cell death is observed in loss-of-function cells at 2 and 3 days of induction relative to uninduced cells in multiple trials. Taken together, these data indicate that FOXD3 activity is critical for maintenance or self-renewal, as well as blocking differentiation, consistent with results in mESCs [13].


While it has been demonstrated that FOXD3 is necessary to derive and maintain mESCs, overexpression studies have not been undertaken in either mouse or human ESCs. We find that upregulation of FOXD3 in hESCs leads to increased expression of mesodermal genes BRACHYURY and FOXC2 as well as TBX6, SOX9, COLL2, PDGFRB, PPARG, and RUNX2, genes associated with paraxial mesoderm derivatives. HNK1 and P75, antigens associated with both somitic and neural crest development, are expressed by distinct populations of RUES(Flag-FOXD3) cells in dox. These data are consistent with specification of distinct chondrogenic and myogenic precursors, rather than neural crest cells that would likely coexpress these proteins. The relatively late expression of TFAP2A is also consistent with the formation of somite-derived limb mesenchyme, as opposed to neural crest cells that would be expected to upregulate TFAP2A early in the differentiation program. Although COLL2 is sometimes used as a marker of notochord, SOX9 and COLL2 are also expressed by chondroblasts arising in the somites [40]. The mesenchymal nature of the FOXD3-overexpressing cells and the lack of vacuolated cells even upon differentiation suggest that the COLL2+ cells are chondroblasts. However, it remains possible that the conditions presented here do not support formation of large vacuoles normally seen in notochord cells in vivo. Thus, our data indicate that increasing the level of FOXD3 activity in undifferentiated hESCs is sufficient to selectively specify paraxial mesoderm derivatives within 48 hours.

Paraxial mesoderm gives rise to a number of developmentally and clinically important cell populations, including chondroblasts that contribute to vertebra, intervertebral discs and articular cartilage, and myoblasts that generate the skeletal muscle essential for coordinated movement (see [40] for review). Our results are particularly striking as specification of chondroblasts and skeletal myoblasts from hESCs has been elusive, and current protocols require several weeks of differentiation before enrichment of these cell types [41, 42]. Loss-of-function analyses in zebrafish and flounder found that foxd3 is necessary for expression of myf5 in muscle progenitors [19, 21] and ectopic formation of paraxial muscle tissue has been observed in frog embryos injected with foxd3 RNA [18, 43]. However, to our knowledge, this is the first study to implicate FOXD3 directly in the formation of somitic cell types other than myoblasts, and attributes to FOXD3 a more general role within paraxial mesoderm development. This is reminiscent of neural crest development, in which Foxd3 is absolutely required for only a subset of neural crest derivatives but appears to have broader roles in cell fate diversification and maintenance of multipotency [44, 45]. Taken together, these data indicate that a strict level of FOXD3 activity is required for self-renewal and maintenance of pluripotency. Overall, our results refine the role of FOXD3 in regulation of pluripotency and also suggest that stoichiometry of FOXD3 with other pluripotency factors may be a critical aspect of stemness.

Conserved Roles for FOXD3

Previous studies show that Foxd3 misexpression is sufficient to generate EMT in the neural tube of chick embryos and convergent extension movements in Xenopus explants [18, 22, 23]. We report for the first time that this is also the case in hESCs. Both chondrogenic and myogenic subpopulations undergo EMT within the somites in vivo. In FOXD3-overexpressing hESCs, SNAIL2 expression is upregulated and E-CADHERIN protein is relocalized concomitant with morphological changes toward mesenchymal cell types. Although it remains unclear which, if any, of these effects are direct, the system presented here provides an excellent platform to dissect the role of FOXD3 in EMT as well as to study the process of EMT itself at single-cell resolution.

At the transcriptional level, we show that FOXD3 upregulates MYF5 and SOX9, known direct targets, but does not affect expression of PAX3 or MYOD. Notably, although Myf5 is a direct, positively regulated target of Foxd3 in other systems, there is a significant delay between FOXD3 overexpression and MYF5 upregulation in RUES2(Flag-FOXD3) cells. This suggests that additional changes are necessary before FOXD3 can activate MYF5 transcription, possibly expression of mesoderm-specific cofactors of FOXD3. These data are consistent with observations in teleosts and chick [19, 21, 22], placing Foxd3 near, but not at the top, of genetic cascades governing formation of multiple tissue types.

We found that both gain- and loss-of-function of FOXD3 in hESCs lead to mesodermal development. While gain-of-function generates paraxial mesoderm, reduction of FOXD3 activity yields mixed endodermal and mesodermal populations. The latter result is also observed in conditional knockout mESC lines, which upregulate mesoderm and endoderm genes upon Cre-mediated excision of Foxd3 [13]. However, previous studies in Xenopus and zebrafish suggest that even significantly reduced levels of foxd3 may support mesoderm development [18, 20]. Thus, it may be that the loss-of-function reported here is hypomorphic in nature, and that although it is not enough to maintain self-renewal, sufficient FOXD3 activity is retained to sustain mesodermal differentiation. In addition, FOXD3 loss-of-function in hESCs results in a cell death phenotype similar to that observed in conditional Foxd3 knockout mESCs, identifying another evolutionarily conserved function of this gene.


In conclusion, we have shown that human FOXD3 is a critical regulator of pluripotency, and that a restricted window of FOXD3 activity is necessary for the maintenance of hESCs. Conditional modulation of FOXD3 activity will provide a foundation for study of somitic derivatives of hESCs as well as investigation of the roles of this transcription factor in other essential cell populations. This work suggests that both conserved and previously unappreciated roles for FOXD3 remain to be discovered.


We thank Joanna Krzyspiak, Dena Sweeney and Sacha Jordan for technical contributions and Blaine Cooper for administrative assistance. We are grateful to Alessandro Rosa for providing the epB-bsd-TT:intRFP vector and advice on shRNA design, and to Aryeh Warmflash and Benoit Sorre for help with Matlab. This work was supported by NYSTEM award C024160 to A.H.B.


The authors indicate no potential conflicts of interest.