Transforming Growth Factor β-Mediated Sox10 Suppression Controls Mesenchymal Progenitor Generation in Neural Crest Stem Cells§


  • Nessy John,

    1. Division of Cell and Developmental Biology, Institute of Anatomy, University of Zurich, Zurich, Switzerland
    2. Division of Biology, Institute of Cell Biology, ETH Zurich, Zurich, Switzerland
    Search for more papers by this author
  • Paolo Cinelli,

    1. Institute of Laboratory Animal Science, University of Zurich, Zurich, Switzerland
    Search for more papers by this author
  • Michael Wegner,

    1. Institute for Biochemistry, Emil Fischer Center, University of Erlangen-Nuernberg, Germany
    Search for more papers by this author
  • Lukas Sommer

    Corresponding author
    1. Division of Cell and Developmental Biology, Institute of Anatomy, University of Zurich, Zurich, Switzerland
    • Division of Cell and Developmental Biology, Institute of Anatomy, University of Zurich, Zurich CH-8057, Switzerland.
    Search for more papers by this author
    • Ph: 41-44-635-53-50; Fax: 41-44-635-68-95

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

  • Author contributions: N.J.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing and final approval of manuscript; P.C.: collection and/or assembly of data and final approval of manuscript; M.W.: provision of study material and final approval of manuscript; L.S.: conception and design, financial support, manuscript writing, and final approval of manuscript.

  • §

    First published online in STEM CELLSEXPRESS February 3, 2011.


During vertebrate development, neural crest stem cells (NCSCs) give rise to neural cells of the peripheral nervous system and to a variety of mesenchymal cell types, including smooth muscle, craniofacial chondrocytes, and osteocytes. Consistently, mesenchymal stem cells (MSCs) have recently been shown to derive in part from the neural crest (NC), although the mechanisms underlying MSC generation remains to be identified. Here, we show that transforming growth factor β (TGFβ)-mediated suppression of the NCSC transcription factor Sox10 induces a switch in neural to mesenchymal potential in NCSCs. In vitro and in vivo, TGFβ signal inactivation results in persistent Sox10 expression, decreased cell cycle exit, and perturbed generation of mesenchymal derivatives, which eventually leads to defective morphogenesis. In contrast, TGFβ-mediated downregulation of Sox10 or its genetic inactivation suppresses neural potential, confers mesenchymal potential to NC cells in vitro, and promotes cell cycle exit and precocious mesenchymal differentiation in vivo. Thus, negative regulation of Sox10 by TGFβ signaling promotes the generation of mesenchymal progenitors from NCSCs. Our study might lay the grounds for future applications demanding defined populations of MSCs for regenerative medicine. STEM CELLS 2011;29:689–699


Neural crest stem cells (NCSCs) are multipotent cells that during vertebrate development migrate from the dorsal neural tube to various embryonic locations and generate a versatile array of neural and non-neural cell types. The potential to generate chondrocytes and osteocytes has mainly been attributed to cranial neural crest (NC) cells [1]. In support of this, multipotent progenitors with neural as well as mesenchymal potentials were rarely observed in clonal cell culture assays of trunk NC cells [2, 3] and trunk NC was ineffective in forming cartilage after transplantation in vivo [4]. However, incubating trunk NC cells in tailor-made chondrogenesis-permissive media for a prolonged time period led to the emergence of mature chondrocytic markers [5, 6]. Furthermore, mesenchymal stem cells (MSCs) originating from the NCs were recently isolated from the trunk of mouse embryos [7]. How such MSCs arise from the NC and whether they actually generate mesenchymal derivatives in vivo remains to be addressed.

A structure that harbors NC-derived cells producing mesenchymal tissues in vivo is the pharyngeal apparatus (PA). Disturbances in the development of PA components result in severe congenital disorders, including DiGeorge syndrome (DGS) that is characterized by osteochondrogenic malformations and conotruncal heart defects [8]. Numerous mouse models have been exploited to understand the etiology and phenotype of DGS, which includes mice with conditional inactivation of transforming growth factor β (TGFβ) signaling components in NC cells [9, 10]. In particular, absence of the TGFβ receptor type 2 (Tgfbr2) in NCSCs leads to phenotypical defects in virtually all NC-derived PA structures [10]. In addition, mutations in TGFBR1 or TGFBR2 are correlated with craniofacial, cardiovascular, and skeletal disorders in human families [11].

Given the tight association of TGFβ signaling with the development of multiple NC-derived PA tissues, it is conceivable that TGFβ influences the development of a mesenchymal progenitor cell common to these tissues. Alternatively, TGFβ might elicit multiple and independent lineage-specific effects on committed progenitors to regulate the development of distinct PA structures. Here, we demonstrate that TGFβ signaling confers mesenchymal potential to NC cells, whereas suppressing their neurogenic potential. This transition from neural to mesenchymal potential is functionally mediated by downregulation of the NCSC transcription factor Sox10 and can be imposed on trunk NCSC, revealing a mechanism for the generation of mesenchymal progenitors from multipotent stem cells.


In Vivo Fate Mapping, Generation of Tgfbr2 Cko and Sox10−/− Mice

In vivo fate mapping of NC cells was performed by breeding Wnt1-Cre with R26R reporter transgenic animals [12, 13]. Tgfbr2 cko animals were obtained as described previously [10]. Littermates carrying only one allele of Tgfbr2 or lacking Cre transgene were used as control. Sox10 transgenic animals were bred and maintained in C3H Heston strain as described previously [14]. Genotyping was done by polymerase chain reaction (PCR) using genomic DNA. All animal experiments were conducted in accordance to the Veterinary Office of the Canton of Zurich, Switzerland.

Cell Culture Preparations and Staining

PA cell cultures were prepared by dissecting PA from E10.5 C57BL/6 embryos and digested in 0.025% trypsin (Gibco, Basel, Switzerland, and 0.35 mg/ml collagenase type1 (Worthington, Lakewood, NJ, in Hanks' balanced saline solution (Gibco) for 15 minutes, further digestion was stopped using 10% fetal calf serum (FCS; Gibco). Cells were collected by centrifugation at 1,200 rpm for 5 minutes and resuspended in media. NCSCs explants cultures were obtained as described previously [15]. For detailed media composition, culture conditions, specific differentiation conditions, antibody, and growth factor listings please refer to Supporting Information Experimental Methods.

Quantitative PCR

A total of 0.1–1 μg of total RNA from NCSC explants primed with or without TGFβ1 and NC explants in differentiation conditions were reverse transcribed with Oligo-dT primers (Invitrogen) and Superscript III (Invitrogen, Basel, Switzerland, Each experiment was repeated in quadruplicates, and quantitative PCR analysis was performed in triplicates and analyzed with Delta Ct-method. β-actin was used for normalization. Primers are listed in Supporting Information Table 1.


Generation of Mesenchymal Progenitors from NCSC Cultures by Short-Term TGFβ Signaling

The contribution of the NC to a MSC population from mouse embryonic trunk has recently been appreciated [7]. Additionally, ablation of TGFβ signaling resulted in differentiation and morphogenetic defects in mesenchymal derivatives of NC origin [10]. Therefore, we sought to assess whether generation of NC-derived mesenchymal progenitor cells can be promoted by TGFβ. To this end, wild-type trunk NCSCs were allowed to emigrate in culture from embryonic day 9 (E9) neural tube under low oxygen conditions and subsequently treated with different concentrations of TGFβ1 in defined medium for different time intervals (data not shown). On the basis of these preliminary experiments, we chose to prime NCSCs by addition of 1 ng/ml TGFβ1 for 4 hours and to examine the expression of the NCSC markers p75 neurotrophin receptor (p75NTR) and Sox10 transcription factor [16, 17] and of the bonafide marker of early osteochondrocytic lineage, Sox9 transcription factor [18, 19]. As previously reported for MSCs [20], Sox9 expression was induced by TGFβ1 in NCSCs after this priming period (Fig. 1C, 1D and see Table 1 for quantitative analysis). In contrast, although p75NTR expression remained unaltered in the vast majority of the NC cells, Sox10 protein expression was lost in NC cells treated with TGFβ1 for 4 hours (Fig. 1A, 1B and Table 1). This finding was confirmed by quantitative real-time PCR (qPCR) analysis, which revealed a highly significant downregulation of Sox10 mRNA expression in TGFβ1-primed NC cells relative to untreated control NCSCs (Fig. 1K).

Figure 1.

TGFβ1 priming confers mesenchymal potential on trunk NCSCs. (A–D): TGFβ1 priming of NCSCs. (A): Immunostaining for p75NTR (red) and Sox10 (green) expression in untreated NCSCs. Blue, 4′,6-diamidino-2-phenylindole staining of cell nuclei. (B): Expression of p75NTR and loss of Sox10 expression after priming with TGFβ1 for 4 hours. (C): Lack of Sox9 expression in control NCSCs. (D): Sox9 expression postpriming with TGFβ1 for 4 hours. (E–J): Mesenchymal potential of TGFβ1-primed neural crest (NC) cells. (E, F): SMA (red) and Calponin (green) expression in control and TGFβ1-primed NC explants, cultured in GM. (G, H): Collagen II expression in control and in TGFβ1-primed NC explants cultured in CM. (I, J): Alizarin red staining in control and TGFβ1-primed NC explants cultured in OM. (K): Sox10 quantitative real-time polymerase chain reaction (qPCR) analysis of NCSCs with or without TGFβ1 for 4 hours. (L–N): qPCR analysis of osteoblast lineage markers after 18 days of culture in OM. p value is compared between control and TGFβ1-primed cells cultured in OM. x-axis illustrates fold change where the value of control NCSCs is always 1. (O–R): Lack of neural potential in TGFβ1-primed NC cells. (O, P): Control and TGFβ1-primed cells cultured in FCS-forskolin and immunostained for NF160 (green) and FABP7 (red). (Q, R): NF160 (green) and SMA (red) expression in control and TGFβ1-primed cells treated with BMP2. (S): Summary of the experiment. Scale bars = 50 μm, n = 5. Data represent mean ± SEM. ***, p < .005; **, p < .01. Abbreviations: ALP, alkaline phosphatase; BMP2, bone morphogenetic protein 2; FABP7, fatty acid binding protein 7; FCS, fetal calf serum; FSC, fetal calf serum; NCSC, neural crest stem cell; OM, osteogenesis permissive medium; SMA, smooth muscle actin; TGFβ, transforming growth factor β.

Table 1. Marker expression in neural crest explants following growth factor treatment
inline image

Although TGFβ1 induces a smooth muscle cell fate in NCSCs on prolonged incubation [21], smooth muscle actin (SMA)-expressing smooth muscle cells were not present after 4 hours of TGFβ1 priming (data not shown). Furthermore, short-term TGFβ1 treatment did not affect cell survival (data not shown). To address whether the changes in Sox10 and Sox9 transcription factor expression correlated with altered mesenchymal potentials, sister plates with trunk NC explants were divided into two sets, with or without TGFβ1 priming. These explants were further cultured in different media, either used to maintain MSCs (basic undefined growth medium, GM; [7]) or to induce differentiation of mesenchymal lineages. To analyze the myogenic potential, TGFβ1-primed and unprimed NC explants were cultured in GM for 2 days. Although control cells only occasionally expressed the SMA and Calponin, priming of NC cells with TGFβ1 readily promoted smooth muscle generation even without further TGFβ1 treatment (Fig. 1E, 1F). Similarly, trunk NC explants without TGFβ1 priming did not form chondrocytes when cultured in chondrogenesis permissive medium (CM) for 10 days, as assessed by Collagen II immunostaining (Fig. 1G). In contrast, priming with TGFβ1 enabled trunk NC cells to respond to CM and to form chondrocytes after 10 days of culture (Fig. 1H). Chondrocytic differentiation was further confirmed by Alcian blue staining (data not shown). On culture in osteogenesis permissive medium (OM) for 18 days, control explants without TGFβ1 priming did not display alizarin red staining indicating the absence of osteoblast mineralization. In contrast, TGFβ1-primed NCSCs cultured in OM had the capacity for osteoblast differentiation, as visualized by alizarin red staining (Fig. 1I, 1J). To further validate the osteogenic potential of TGFβ1-primed NCSCs, we performed qPCR analysis using values normalized to control NCSCs, which had not been treated with TGFβ1. Although exposure of NCSCs to TGFβ1 for 4 hours barely induced the expression of the master transcriptional regulators Runx2 (an early marker in the lineage [22]) and Osterix (acting downstream to Runx2 [23]), culturing the cells for 18 days in OM led to robust upregulation of gene expression, with a significant twofold increase in Osterix expression in TGFβ1-primed cells as opposed to unprimed cells (Fig. 1L, 1M). In contrast, unprimed cells cultured for 18 days in OM did not express alkaline phosphatase (ALP), a marker indicative for mineralization during osteoblast differentiation (Fig. 1N). Intriguingly, however, a significant increase (fivefold) in ALP transcript expression was observed in TGFβ1-treated NCSCs cultured in OM compared with untreated NCSC in OM (Fig. 1N). These data confirm the results obtained by Alizarin red staining (Fig. 1I, 1J) and demonstrate that NCSCs are unable to give rise to fully differentiated osteoblasts unless primed with TGFβ1.

We next tested the effect of TGFβ1 priming on the neural potential of NC cells. On culturing control NC cells in complex medium containing FCS-forskolin [24], neurofilament 160 (NF160)-expressing neurons and fatty acid binding protein 7 (FABP7)-expressing glial cells were formed after 8 days (Fig. 1O). In contrast, in similar conditions, TGFβ1-primed NC cells neither differentiated into neurons nor into glia (Fig. 1P). To further ensure lack of neural potential, TGFβ1-primed cells were treated with bone morphogenetic protein 2 (BMP2), an instructive factor inducing autonomic neurogenesis in NCSCs [21]. TGFβ1-primed cells, instead of forming neurons, gave rise to smooth muscle cells, whereas the majority of the control cells formed neurons marked by the presence of NF160 and absence of SMA (Fig. 1Q, 1R). Quantification confirmed a significant loss in the neurogenic and gliogenic potential of TGFβ1-primed NC explants (Table 1). Thus, we conclude that TGFβ1-primed trunk NC cells display new potentials coherent with mesenchymal progenitors.

Fibroblast Growth Factor 8-Induced Sox9 Expression Is Not Sufficient to Elicit Competence for Mesenchymal Fates in NC Cells

As fibroblast growth factor 8 (FGF8) is also known to support mesenchymal fates in vivo [25, 26], we tested the effects of FGF8 (50 ng/ml) for 4 hours on trunk NC explants. Comparable with TGFβ1 treatment, FGF8-primed NC cells maintained p75NTR expression and upregulated Sox9 in virtually all cells (Fig. 2A, 2B and Table 1). In contrast, treatment with FGF8 did not downregulate Sox10 expression as efficiently as TGFβ1 (Fig. 2A and Table 1). Moreover, FGF8-primed NC cells did not readily form smooth muscle cells in GM, unlike TGFβ1-primed cells (Fig. 2C and Table 1). However, NC explants primed with FGF8 and subsequently treated with TGFβ1 were able to generate SMA and Calponin-expressing cells (Fig. 2D). Intriguingly, FGF8-primed NC cells also lacked the competence to respond to chondrogenic and osteogenic cues in appropriate culture conditions and did not form any chondrocytes (Fig. 2E) or osteoblasts (Fig. 2F).

Figure 2.

FGF8 priming is not sufficient to induce mesenchymal potentials in trunk NCSCs. (A): p75NTR (red) and Sox10 (green) expression in FGF8-primed NCSCs. (B): Sox9 expression in NC explants after priming with FGF8 for 4 hours. (C, D): SMA (red) and Calponin (green) immunostaining on FGF8-primed NC explants cultured with and without TGFβ1, respectively. (E): Collagen II expression in FGF8-primed NC explants cultured in CM. (F): Alizarin red staining in NC explants primed with FGF8 and cultured in OM. (G): NF160 (green) and FABP7 (red) immunostaining in FGF8-primed NC cells cultured in FCS-forskolin. (H): NF160 (green) and SMA (red) immunostaining on FGF8-primed NC cells cultured in BMP2. (I): Summary of the experiment illustrating lack of mesenchymal potential and persistence of Sox10 expression and neural potential in FGF8-primed NC cells. Scale bars = 50 μm, n = 4. Abbreviations: BMP2, bone morphogenetic protein 2; CM, chondrogenesis permissive medium; FABP7, fatty acid binding protein 7; FCS, fetal calf serum; FGF8, fibroblast growth factor 8; GM, growth medium; NCSC, neural crest stem cell; OM, osteogenesis permissive medium; SMA, smooth muscle actin.

In FCS-forskolin condition, FGF8-primed cells were capable of forming FABP7-expressing glial cells, although neurons were scarcely formed (Fig. 2G and Table 1). However, FGF8-primed cells possessed the potential for neurogenesis, as many NF160-positive neuronal cells emerged in explant cultured in the presence of BMP2 (Fig. 2H and Table 1). These data indicate that TGFβ cannot be substituted by FGF8 in promoting mesenchymal fates in NCSCs, although both growth factors upregulate Sox9.

Identification and Characterization of NC-Derived Mesenchymal Progenitor Cells in the PA

Our data provides a method to generate mesenchymal progenitors in cell culture from trunk NCSCs by short-term TGFβ signaling associated with loss of the NCSC factor Sox10. To address the in vivo relevance of these findings, we investigated whether NC-derived mesenchymal progenitors, similar to those generated in culture, could be isolated from the developing PA, a structure that orchestrates craniofacial skeletal development and morphogenesis in a TGFβ-dependent manner [10]. Using fine dissection methods, we manually isolated PA tissue from E10.5 mouse embryos and enzymatically dissociated these structures to obtain primary cultures (for details, see Materials and Methods and Supporting Information data). We found that the total population of dissociated PA cells at E10.5 was composed of 89% ± 1.3% recombined NC cells as observed by in vivo lineage tracing and X-Gal staining. Furthermore, 75% ± 0.5% of all cells expressed p75NTR, 6% ± 2.9% expressed Sox10, and 97% ± 0.7% expressed Sox9 (Fig. 3A–3D). Thus, NC-derived cells with a marker profile reminiscent of that of TGFβ1-primed trunk NC cells are present in the PA at E10.5.

Figure 3.

Characterization of NC-derived mesenchymal progenitor cells in the PA and their differentiation potential. (A–D): Characterization of PA cells. (A): Whole mount X-Gal staining of a Wnt1-Cre/R26R embryo at E10.5, arrow indicating the PA. (B): Adherent E10.5 PA culture displaying neural crest (NC)-derived X-Gal-stained cells. (C): Adherent PA cells stained for the neural crest stem cell markers p75NTR (red) and Sox10 (green). Note presence of p75NTR but absence of Sox10. (D): PA culture stained for Sox9. (E–J): Mesenchymal potential of PA cells. (E, F): SMA (red) and Calponin (green) expression in PA cells in GM and TGFβ1-supplemented GM. (G, H): Collagen II expression in PA cells cultured in GM and CM. (I, J): Alizarin red staining in PA cells cultured in GM and OM. (K, L): Lack of neural potential of PA cells. (K): PA cells cultured in FCS-forskolin cocktail stained negative for NF160 (green) and FABP7 (red). (L): Lack of NF160 and presence of SMA expression in BMP2-treated PA cells. Scale bars = 2 mm (A), 50 μm (B–L); n = 5. Quantification: SMA (E), 87% ± 1.5%; SMA (F), 91% ± 0.9%; FABP7 (K), 0% ± 0.2%; NF160 (L), 1% ± 0.5%. Abbreviations: BMP2, bone morphogenetic protein 2; CM, chondrogenesis permissive medium; FABP7, fatty acid binding protein 7; FCS, fetal calf serum; GM, growth medium; PA, pharyngeal apparatus; SMA, smooth muscle actin; TGFβ, transforming growth factor β.

We next analyzed the mesenchymal potential of PA cells and found that acquisition of smooth muscle fate is irrespective of TGFβ1 supplementation in GM (Fig. 3E, 3F). Cells cultured in CM for 10 days showed high cell density chondrocytic structures, whereas cells in control GM showed no signs of cartilage formation (Fig. 3G, 3H). PA cells cultured in OM for 18 days revealed osteogenic potential as revealed by alizarin red mineralization staining, whereas osteoblasts were not observed in control conditions (Fig. 3I, 3J).

To analyze whether NC-derived PA cells displayed neural potential, the cells were cultured in a complex differentiation medium supplemented with FCS-forskolin. PA cells, similar to TGFβ1-primed NC cells, failed to differentiate into neural cell types as shown by immunocytochemistry for NF160 and FABP7 (Fig. 3K). Similarly, in the presence of BMP2, PA cells formed smooth muscle cells, rather than adopting a neuronal fate, as shown by presence of SMA and absence of NF160 (Fig. 3L) Thus, based on marker expression and developmental potential, the PA at E10.5 harbors mesenchymal progenitors analogous to those obtained in culture on short-term treatment of NCSCs with TGFβ1.

Postmigratory NCSCs isolated from neurogenic structures, such as the enteric nervous system, exhibit extensive self-renewal capacity and can be readily passaged in sphere cultures [27]. To determine the self-renewal potential of PA-derived mesenchymal progenitor cells, single cells were cultured in sphere medium permissive for self-renewal of NCSCs [28]. PA cells possessed a limited sphere-forming potential and could only be passaged efficiently up to three passages (Supporting Information Fig. S1A). Immunostaining of dissociated secondary sphere cells for p75NTR, Sox10, and Sox 9 revealed the presence of cells with an expression pattern equivalent to that of primary PA-derived mesenchymal progenitors (Supporting Information Fig. S1B, S1C). Likewise, the differentiation potential of secondary sphere cells was similar to that of primary mesenchymal progenitors in the PA (Supporting Information Fig. S1D–S1K), suggesting maintenance of mesenchymal progenitor cells in terms of identity and plasticity. Thus, unlike neurogenic NCSCs, PA progenitors have a limited self-renewal capacity in sphere assays.

TGFβ Signaling Controls Mesenchymal Progenitor Cell Generation In Vivo

Given the role of TGFβ1 in the formation of mesenchymal progenitors in vitro, we performed in vivo characterization of the PA from Tgfbr2 conditional knockout (cko) embryos [10]. At E10.5, TGFβ signaling was active in the PA of control embryos, as assessed by the presence of phosphorylated Smad2 (p-Smad2), whereas TGFβ signal activity was abolished in the Tgfbr2 cko PA (Fig. 4A, 4B). Since, TGFβ1 treatment resulted in downregulation of Sox10, we assayed Sox10 expression in the PA of Tgfbr2 cko embryos in vivo. Although control PA was almost devoid of Sox10-expressing cells, PA of Tgfbr2 cko embryos displayed ninefold increase in the number of Sox10-expressing cells (Fig. 4C, 4D, for quantification refer to figure legend) along with an increase in mRNA transcript levels as assayed by in situ hybridization (data not shown). Although control PA was almost devoid of Sox10-expressing cells, PA of Tgfbr2 cko embryos displayed increase in the number of Sox10-expressing cells (Fig. 4C, 4D). An approximate ninefold increase in the percentage of Sox10 expression cells per PA was observed in the Tgfbr2 cko embryos (data not shown). Additionally, we also observed an increase in mRNA transcript levels in Tgfbr2 cko embryos as assayed by in situ hybridization (data not shown).

Figure 4.

Impaired mesenchymal differentiation and increased proliferation of Tgfbr2 cko PA cells in vivo. (A–J): In vivo characterization of the first pharyngeal arch at E10.5. (A, B): p-Smad2 expression in longitudinal cryosections of E10.5 control and Tgfbr2 cko embryos. (C, D): Sox10 expression in control and Tgfbr2 cko embryo. (E, F): Visualization of proliferation in vivo by means of p-HH3 immunostaining in control and Tgfbr2 cko PA. (G, H): Cell cycle regulator Cyclin D2 expression in control and Tgfbr2 cko PA. (I): Quantification of mitotic p-HH3-positive cells on longitudinal sections of E10.5 control and Tgfbr2 cko embryos. (J): Quantification of cell cycle exit on double labeling of BrdU (BrdU pulse performed at E10.5 for 24 hours) and Ki67 on control and Tgfbr2 cko sections in E11.5 embryos. Graph represents percentage of cells that exited cell cycle in 24 hours. Data represent mean ± SEM. ***, p < .005. (K–N): Meckel's cartilage at E13. (K, L): p-HH3 immunostaining in Meckel's cartilage in control and Tgfbr2 cko embryos. (M, N): Collagen II expression indicative for chondrocyte differentiation in Meckel's cartilage (white arrow) in control and Tgfbr2 cko embryos. White line marks the palatal shelf for orientation. (O, P): Immunostaining for Collagen II expression in thyroid cartilage (white arrow) in control and Tgfbr2 cko embryos. Blue, DAPI to visualize cell nuclei. Scale bar = 50 μm, n = 4. Quantifications: Sox10 (C), 5% ± 0.8%; Sox10 (D), 45% ± 1.3%; p-HH3 (E), 30% ± 0.7%; p-HH3 (F), 49% ± 3.6%. Ratio of cell cycle exit in (J), cont (21% ± 0.1%), cko (10% ± 0.3%). p-HH3 (K), 9% ± 0.5%, p-HH3 (L), 13% ± 0.6%. Abbreviations: BrdU, bromodeoxyuridine; p-HH3, phosphorylated histone H3.

In agreement with the known cytostatic functions of the TGFβ pathway, we also observed a significant increase in mitotic nuclei in Tgfbr2 cko PA at E10.5, as visualized by staining for phosphorylated histone H3 (p-HH3) (Fig. 4E, 4F, 4I). To assess possible mechanisms underlying TGFβ-mediated cell cycle regulation, we analyzed the expression of several cell cycle regulators (data not shown). Among them, only Cyclin D2 expression was altered in the Tgfbr2 cko PA, in that Cyclin D2 levels were drastically increased in the Tgfbr2 cko PA for a transient time period at E10.5 (Fig. 4G, 4H) but not at E11.5 (data not shown). Altered proliferation in the Tgbfr2 cko was confirmed by addressing the rate of cell cycle exit in the PA. In comparison with the control, only half of the proliferative cell population exited the cell cycle in the Tgfbr2 cko PA during the chosen time period (Fig. 4J). In sum, maintained expression of the NCSC factor Sox10 along with increased proliferation and decreased cell cycle exit in Tgfbr2 cko PA supports the hypothesis that TGFβ signaling in the PA is crucial for the transition of NCSCs to mesenchymal progenitor cells in vivo.

As interference with mesenchymal progenitor development in the PA of Tgfbr2 cko embryos likely leads to developmental defects in its derivatives, we examined Meckel's cartilage, a prominent NC-derived chondrocytic structure developing from the first PA. Similar to the increased proliferation in PA at E10.5, we observed an increase in proliferation in Meckel's cartilage at E13 in Tgfbr2 cko compared with the control tissue (Fig. 4K, 4L). In addition, we observed a delay in chondrocytic differentiation in Meckel's cartilage in Tgfbr2 cko at E13, as monitored by Collagen II expression (Fig. 4M, 4N; white arrow). Absence of chondrocytic differentiation marker in Tgfbr2 cko embryos at E13 was not only restricted to Meckel's cartilage but also observed in another NC-derived tissue, the thyroid cartilage (Fig. 4O, 4P; white arrow). In both structures, however, chondrogenesis was delayed rather than fully blocked, because the Meckel's and thyroid cartilage at E14 showed signs of differentiation, coupled with disturbed morphogenesis at the same time (Supporting Information Fig. S2A–S2D; white arrow). Additionally, smooth muscle cell differentiation was impaired in the aortopulmonary septum of Tgfbr2 cko embryos at E10.5 [10], affirming the requirement of TGFβ signaling for proper generation of mesenchymal lineages.

TGFβ Signal Inactivation Perturbs Plasticity and Proliferation of Mesenchymal Progenitor Cells Ex Vivo

Our results demonstrate that lack of TGFβ signaling alters the proliferation and differentiation rate of PA cells in vivo, affecting morphogenesis of NC-derived mesenchymal structures. Conceivably, this phenotype involves disturbed formation of mesenchymal progenitors in the PA. To address this hypothesis, we cultured control and Tgfbr2 cko PA cells from E10.5 embryos for 24 hours in GM, followed by immunostaining for differentiation (SMA) and proliferation (Ki67) markers (Fig. 5A, 5B). At this time point, a high fraction of postmitotic, fully differentiated SMA-expressing cells (Fig. 5A; arrow) was observed in the control. In contrast, Tgfbr2 cko cell cultures contained many undifferentiated proliferating cells (Fig. 5B; arrowhead) in addition to SMA-positive cells coexpressing the proliferation marker Ki67. After 2 days of culture, however, both control and Tgfbr2 cko PA cell cultures exhibited many SMA and Calponin-positive cells, indicating the presence of fully differentiated smooth muscle cells (Supporting Information Fig. S3A, S3B). Given that TGFβ signaling regulates the balance between cell proliferation versus differentiation in other systems [29], we quantified the percentage of proliferating and differentiating cells in control and Tgfbr2 cko PA cells at three different time points, 12 hours, 24 hours, and 36 hours. Interestingly, proliferation persisted and differentiation was delayed in Tgfbr2 cko cells as observed at all three time points (Supporting Information Fig. S3C). Likewise, chondrocytic differentiation was also delayed on TGFβ signal inactivation. After culturing cells for 6 days in CM, we observed Collagen II expression only in control cultures but not in the Tgfbr2 cko (Fig. 5C, 5D). Culturing PA cells in CM for 10 days, resulted in the presence of Collagen II-expressing cell clusters in both control and Tgfbr2 cko cells (Supporting Information Fig. S3D, S3E). Thus, smooth muscle formation and chondrogenesis was impaired but not completely abrogated in Tgfbr2 cko PA cells, pointing to the presence of cues partially substituting for TGFβ signaling [30]. In contrast, although control cell cultures readily produced osteoblasts when incubated in OM medium, mutant cells exhibited no signs of osteoblast differentiation even after 18 days of culture (Fig. 5E, 5F). Consistent with the known requirement of Sox10 for gliogenesis [14, 31] and with the maintained Sox10 expression in Tgfbr2 cko PA, mutant but not control cells occasionally formed glia (Fig. 5G, 5H).

Figure 5.

Inactivation of TGFβ signaling impairs formation of mesenchymal progenitors in the PA. (A, B): Control and Tgfbr2 cko PA cells, respectively, cultured in GM for 24 hours and immunostained for SMA (red) and Ki67 (green). Arrows indicate fully differentiated smooth muscle cells (A) and arrowheads indicate proliferating cells (B). (C, D): Collagen II immunostaining on control and Tgfbr2 cko cells cultured in CM for 6 days, revealing absence of chondrocytes in mutant cell cultures. (E, F): Alizarin red staining on control and Tgfbr2 cko cells cultured in OM for 18 days, illustrating absence of osteoblasts in mutant cell cultures. (G): Control PA cells stained for neural markers NF160 and FABP7, indicating absence of neurons and glia. (H): Tgfbr2 cko PA cells stained for neural markers NF160 and FABP7, indicating absence of neurons but not glia (6% ± 0.7% positive of all cells). (I): Graphic representation of sphere numbers per 10,000 cells from control and Tgfbr2 cko PA cultures in three consecutive passages. Data represent mean ± SEM. ***, p < .005. Scale bar = 50 μm, n = 3. Abbreviations: CM, chondrogenesis permissive medium; FABP7, fatty acid binding protein 7; FCS, fetal calf serum; GM, growth medium; OM, osteogenesis permissive medium; SMA, smooth muscle actin; TGFβ, transforming growth factor β.

To address the self-renewal capacity of Tgfbr2 cko PA cells, we performed sphere assays. In three consecutive passages, the number of spheres was significantly increased in floating cultures derived from Tgfbr2 cko PA at E10.5 (77 ± 4.9 in primary passage, 59 ± 6.5 in secondary passage, and 49 ± 1.8 in tertiary passage) as compared with control cultures (46 ± 6.6 in primary passage, 25 ± 4.1 in secondary passage, and 20 ± 2.5 in tertiary passage) (Fig. 5I). In sum, our data reveal an in vivo requirement of TGFβ signaling for the generation of mesenchymal progenitors from NCSCs. TGFβ signal inactivation perturbs the mesenchymal potential of PA cells that instead maintain certain features of NCSCs such as Sox10 expression and increased self-renewal capacity.

Acquisition of Mesenchymal Potential Through Suppression of Sox10

We have demonstrated that TGFβ signaling is involved in the generation of Sox10-negative mesenchymal progenitors from Sox10-positive NCSCs both ex vivo and in vivo. Hence, loss of Sox10 might represent the cardinal event in the generation of a NC-derived mesenchymal progenitor cells. To test this postulate, we prepared trunk NC explants from control embryos and from embryos carrying a targeted deletion of Sox10 (Sox10−/− embryos) [14, 32]. The cells were cultured in CM without prior TGFβ1 priming and immunostained for collagen II. Consistent with previous results, control NC explants were unable to adopt a chondrocytic fate, whereas Sox10−/− NC cells in CM expressed the chondrocytic marker (Fig. 6A, 6B). As Sox10+/− NC cells display haploinsufficiency [14, 32], we quantified the number of Collagen II-positive colonies in control, Sox10+/− and Sox10−/− explants cultured in CM without prior TGFβ1 priming. The graph (Fig. 6C) illustrates a striking relationship between Sox10 levels and chondrogenesis. Control explants contained negligible number of chondrocytic colonies per explant (1 ± 0.8), whereas Sox10+/− explants (5 ± 0.8) and Sox10−/− explants (11 ± 0.8) displayed an increasing mesenchymal potential with decreasing Sox10 levels.

Figure 6.

Lack of Sox10 confers mesenchymal potential on NCSCs and promotes cartilage differentiation in vivo. (A, B): Collagen II staining in control and Sox10−/− trunk NC explants cultured in CM without prior priming with TGFβ1. (C): Quantification of number of chondrocytic colonies per explant in control, Sox10+/−, and Sox10−/− cultures. (D, E): Collagen II expression representative for precocious differentiation of Meckel's cartilage (white arrow) in Sox10−/− as opposed to control E12 embryos. White line marks the palatal shelf for orientation. (F, G): Decreased number of proliferating cells as assessed by p-HH3 immunostaining in Meckel's cartilage of Sox10−/− E12 embryos (4% ± 0.25%) as compared with control (10% ± 0.75%). Data represents mean ± SEM. ***, p < .005. Scale bar = 50 μm, n = 3. (H): Model explaining mechanism of TGFβ-mediated development of mesenchymal progenitors from stem cells. TGFβ signaling confers multiple changes on NC cells, including decreased proliferation, loss of neural potentials, and gain in mesenchymal potentials. The nodal point in this process consists in TGFβ-mediated suppression of the transcription factor Sox10. In the pharyngeal apparatus (PA) in vivo, interfering with this sequence of events disrupts coordinated proliferation, mesenchymal fate acquisition, and timely differentiation, leading to disturbed morphogenesis of PA-derived mesenchymal structures in Tgfbr2 cko embryos. Abbreviations: CM, chondrogenesis permissive medium; NC, neural crest; NCSC, neural crest stem cell; p-HH3, phosphorylated histone H3; TGFβ, transforming growth factor β.

To address whether Sox10 negatively regulates cartilage formation in vivo, we examined the differentiation status of NC-derived Meckel's cartilage at E12 in control and Sox10−/− embryos. At this stage, chondrogenesis is normally not yet apparent, and control embryos therefore displayed only faint signs of differentiation (Fig. 6D). In contrast, in Sox10−/− embryos, Collagen II was precociously expressed in the primordium of Meckel's cartilage, indicating the predisposition of Sox10-negative cells to acquire a mesenchymal fate (Fig. 6E). Intriguingly, concomitant with precocious differentiation, Sox10−/− embryos also displayed reduced proliferation when compared with controls, as monitored by p-HH3 (Fig. 6F, 6G). Therefore, Sox10−/− embryos and Tgfbr2 cko embryos exhibit opposite phenotypes. Our combined data indicate that the TGFβ-mediated transition from NCSCs to mesenchymal progenitor cells, evident by a switch in potentials and proliferation rates, is mainly elicited by suppression of Sox10.


In this study, we demonstrate that short-term TGFβ1 signaling commands the transition from a NC stem to a progenitor cell state reminiscent of mesenchymal progenitors present in the PA. Intriguingly, however, this transition is characterized by TGFβ1-mediated suppression of neural potential accompanied by gain of mesenchymal potentials. A key player in this event is the NCSC transcription factor Sox10, the expression of which is negatively regulated by TGFβ1. Sox10-positive trunk NC cells are unable to respond to osteochondrocytic cues and to generate bone and cartilage cells. In contrast, loss of Sox10 on TGFβ1 priming or gene inactivation confers mesenchymal potential to NC cells. On the basis of our data, we propose a model in which the interplay between TGFβ1 signaling and Sox10 expression controls the switch from neural to mesenchymal potentials and also regulates proliferation in NC cells (Fig. 6H).

In contrast to cranial NC, it is assumed that trunk NC cells do not give rise to osteocytes and chondrocytes in vivo [33]. Nevertheless, NC-derived MSCs could be obtained from embryonic trunk cell populations [7], and trunk NC cells displayed osteochondrogenic potential after long-term culture in complex medium [6]. In this study, trunk NCSCs in high-density explant cultures were able to produce smooth muscle cells, osteoblasts, and chondrocytes when the cells were exposed to a brief pulse of TGFβ1 in defined minimal medium. Therefore, mesenchymal fates can efficiently be imposed on trunk NCSCs by short-term treatment with TGFβ1 to induce responsiveness toward myogenic and osteochondrogenic cues, whereas simultaneously suppressing the responsiveness to neurogenic signals.

Our study reveals that the SoxE transcription factors Sox9 and Sox10 are intimately associated with the transition from neurogenic NCSCs to mesenchymal progenitors. In particular, the TGFβ1-induced switch in NC cell plasticity is reflected by a change in the Sox code from a Sox10-positive/Sox9-negative to a Sox10-negative/Sox9-positive state. Although Sox9 and Sox10 are often coexpressed and act equivalently in NC induction and other processes of neural development [34, 35], they exhibit different expression patterns during NC lineage decisions and fulfill fundamentally distinct functions in the transition of NCSCs to mesenchymal progenitor cells. Sox9 expression is induced by TGFβ signaling during limb chondrogenesis [36] and in NCSCs, as shown here. However, Sox9 is neither sufficient as such to promote mesenchymal potential nor to prevent neural development of NCSCs. Indeed, FGF8 treatment induced Sox9 expression in NC cells as efficiently as TGFβ1 but did not confer responsiveness to osteochondrogenic factors. Furthermore, FGF8 but not TGFβ1-primed NC cells were able to produce neurons and glia in response to appropriate growth factors, despite the presence of Sox9.

On the basis of these findings, we focused on changes in Sox10 expression and function to determine the key step of TGFβ1-induced mesenchymal potential in NCSCs. Previously, we have shown that loss of Sox10 in cultured NCSCs prevents neural development and promotes smooth muscle cell differentiation [16]. Accordingly, multiple peripheral nervous system structures fail to develop properly in mice and Zebrafish lacking Sox10 [14, 37]. Overexpression of Sox10, on the other hand, counteracts TGFβ1-induced smooth muscle cell differentiation in cultured NCSCs [31]. Our data now reveal a novel function of Sox10 in controlling the emergence of mesenchymal progenitors by suppressing mesenchymal potential in NC cells. First, TGFβ1-induced Sox10 downregulation coincided with a gain in osteochondrocytic potential. Second, unlike TGFβ1, FGF8 treatment of NCSCs did not result in loss of Sox10 expression. Consistent with the role of Sox10 in conferring neural potential in NCSCs [31, 37] and in preventing mesenchymal progenitor formation, FGF8-treated NC cells were capable of generating neural but not osteochondrocytic cells. Third and most importantly, mutant NCSCs displaying reduced or no Sox10 expression had the capacity to generate mesenchymal cell types in vitro, even without prior exposure to TGFβ1. It is noteworthy, however, that deletion of Sox10 did not as effectively elicit non-neural potential in NCSCs as did short-term priming by TGFβ1, indicating additional TGFβ-dependent cues controlling mesenchymal fates.

In accordance with our model that Sox10 counteracts mesenchymal potential (Fig. 6H), we also observed precocious cartilage differentiation in Sox10−/− embryos in vivo. Evidence for premature osteocyte differentiation was not found in Sox10 mutants (data not shown), but the embryonic lethality of Sox10−/− embryos precluded an analysis of bone formation at later stages. This leaves open whether in vivo lack of Sox10 affects mesenchymal lineages other than cartilage. Nonetheless, we propose that TGFβ-mediated Sox10 downregulation not only regulates mesenchymal potential in NCSC cultures but also in NC-derived mesenchymal progenitors in the PA in vivo. Indeed, Tgfbr2 cko embryos exhibit a phenotype opposite to the one of Sox10−/− mutants, with impaired chondrogenic differentiation and persistent Sox10 expression in the PA. However, mesenchymal cell differentiation was delayed, but not fully blocked, on TGFβ signal inactivation. Moreover, molecular alterations including persistent Sox10 expression in the PA were transient in Tgfbr2 cko embryos. Therefore, our study points to a crucial role of TGFβ at early stages of PA development, whereas TGFβ-independent signals can apparently substitute in part for TGFβ during further PA development. These findings are in agreement with previous reports that, unlike in defined minimal medium, smooth muscle cells can form from NCSCs in complex medium, even if canonical signaling by TGFβ family factors is completely inactivated [30]. Furthermore, although TGFβ priming was required for NC cells to generate differentiated osteoblasts, early osteogenic transcription factors were also induced in untreated NC cells incubated in osteogenesis permissive conditions. Hence, TGFβ signaling appears to act in parallel with other pathways to promote mesenchymal potential and differentiation.

The suppression of NCSC features by TGFβ1 signaling is accompanied by changes in cell cycle control. Ex vivo, sphere formation and self-renewal by PA cells is increased on conditional Tgfbr2 inactivation. In vivo, cell cycle exit is decreased and mitosis increased in mutant PA and Meckel's cartilage. This is reminiscent of the developing midbrain where TGFβ counteracts neural stem cell maintenance by inhibiting cell cycle progression and self-renewal [29]. Despite the known antimitotic role of TGFβ1, we cannot exclude that changes in cell cycle progression of NC cells on TGFβ signal inactivation might occur indirectly, caused by persistence of Sox10. In fact, our study not only reveals an association of increased Sox10 expression with increased cell cycle progression and persistent Cyclin D2 expression in Tgfbr2 cko embryos but also strongly reduced mitosis in Meckel's cartilage of Sox10−/− embryos. Hence, Sox10 might play a hitherto unknown role in cell cycle control of NC cells. It remains to be addressed whether and how changes in proliferation on altered Sox10 activity are functionally coupled to changes in NCSC potentials.


In conclusion, we propose that the interaction between TGFβ1 signaling and the transcription factor Sox10 represents a nodal point in the spatiotemporal control and coordination of cell proliferation, mesenchymal cell differentiation, and morphogenesis of the PA. Interfering with this node at early stages prevents the timely appearance of multipotent mesenchymal progenitor cells, resulting in the broad developmental anomalies of mesenchymal structures obtained in vivo on TGFβ signal inactivation in NCSCs. Our model that mesenchymal progenitor cells are generated by NCSCs losing certain potentials while acquiring new ones might generally be applicable to other stem cells exhibiting neural and non-neural potential, such as embryonic and induced pluripotent stem cells. In particular, in view of the ill-defined nature and origin of MSCs and the heterogeneity of MSC preparations available to date, our findings might point to strategies for how defined MSC lines could be established for future applications in regenerative medicine.


We acknowledge Drs. A. McMahon, P. Soriano, and S. Karlsson for providing transgenic animals and Dr. C. Birchmeier for providing FABP7 antibody. We are thankful to Drs. N. Mantei and S. Falk for critical reading of the manuscript and to Dr. U. Suter for helpful discussions and scientific support. This work was supported by the Swiss National Science Foundation, by the National Center of Competence in Research “Neural Plasticity and Repair,” by the Vontobel Foundation, by the ETH Zurich, and the University of Zurich.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.