Neural crest stem cell multipotency requires Foxd3 to maintain neural potential and repress mesenchymal fates

At the onset of their migration from the neural tube, neural crest (NC) cells are a heterogeneous pool of multipotent neural crest stem cells (NCSCs) and fate-restricted progenitors that give rise to a wide variety of cell types, including neurons, glia, melanocytes, vascular smooth muscle cells (VSMCs), chondrocytes and osteoblasts (Le Douarin and Kalcheim, 1999). In vitro clonal analyses and in vivo cell transplantation and labeling experiments established that NC is both multipotent and self-renewing (Baroffio et al., 1991; Bronner-Fraser and Fraser, 1988; Bronner-Fraser et al., 1980; Ito and Sieber-Blum, 1991; Sieber-Blum and Cohen, 1980; Trentin et al., 2004). NCSCs can be isolated from neural tube explants or embryonic and postnatal NC derivatives, including fetal peripheral nerves, heart, skin and the adult enteric nervous system (ENS) (reviewed in Crane and Trainor, 2006; Delfino-Machin et al., 2007; Teng and Labosky, 2006). Additionally, NCSCs can be isolated by flow cytometry based on their expression of the low-affinity nerve growth factor receptor, p75 (Stemple and Anderson, 1992). Combinatorial Wnt and bone morphogenic protein (BMP) signaling synergistically maintains NCSC multipotency in vitro by suppressing differentiation (Kleber et al., 2005), and overexpressing the HMG-box transcription factor Sox10 in NC maintains multipotency by preserving gliogenic potential and inhibiting neural differentiation (Kim et al., 2003). Despite these significant advances in characterization of NCSCs, relatively little is known about the molecules required for establishing and controlling their self-renewal and/or multipotency.

NCSCs may share common transcriptional mechanisms controlling self-renewal and multipotency with other embryonic progenitor populations. Our earlier work established that Foxd3 functions in embryonic stem cells (ESCs) and trophoblast stem cells (TSCs) to maintain self-renewal and repress differentiation, supporting the notion of a conserved role for Foxd3 in stem cell self-renewal and multipotency (Hanna et al., 2002; Liu and Labosky, 2008; Tompers et al., 2005). In the NC, Foxd3 is expressed in pre-migratory and early migrating cells, and its expression decreases as cells differentiate into most derivatives, suggesting a link between Foxd3 expression and multipotency (Labosky and Kaestner, 1998). Ectopic expression of Foxd3 in avian NC inhibited neural differentiation (Dottori et al., 2001), supporting the hypothesis that Foxd3 largely maintains an early, uncommitted NCSC state. Loss-of-function studies in Xenopus laevis, zebrafish and mice demonstrate dramatic losses of distinct NC derivatives and suggest a central role for Foxd3 in early NC maintenance (Lister et al., 2006; Montero-Balaguer et al., 2006; Sasai et al., 2001; Stewart et al., 2006; Teng et al., 2008). An NC-specific deletion of murine Foxd3 caused severe defects in most NC derivatives; however, this requirement was not identical throughout the NC: although Foxd3 is required for establishment of the ENS, it is dispensable for cardiac outflow tract septation (Teng et al., 2008). This led us to hypothesize that intrinsic differences between these overlapping NC populations might reflect divergent molecular requirements controlling multipotency and lineage commitment. However, direct roles for Foxd3 in NCSC multipotency and self-renewal, or the extent to which Foxd3 controls lineage allocation of NC has not been directly examined.

To address these unanswered questions, we combined an NC-specific deletion of Foxd3 with in vivo lineage mapping and in vitro clonal analysis of isolated NC to demonstrate a cell-autonomous requirement for Foxd3 in NC self-renewal and multipotency. In vivo, the NC-specific deletion of Foxd3 resulted in loss of neural derivatives of the cranial, vagal and cardiac NC, with aberrant differentiation of mesenchymal NC derivatives and ectopic expansion of the NC-derived VSMC domain. In single-cell analyses of multipotency, Foxd3 played a crucial role in negatively regulating myofibroblast differentiation, thereby maintaining newly generated NCSCs in an uncommitted multipotent state. We therefore provide the first genetic identification of a specific gene requirement for both self-renewal and multipotency of NC. Together with published findings, we describe a model in which Foxd3 maintains NCSCs by inhibiting non-neural differentiation, drawing important gene-regulatory parallels between disparate stem and progenitor cell populations.

The Foxd3 null alleles Foxd3^tm1.Lby and Foxd3^tm2.Lby (called Foxd3^– throughout) were used interchangeably in combination with Foxd3^tm3.Lby, the Foxd3 conditional allele (Foxd3^flox). These alleles were described previously (Hanna et al., 2002; Teng et al., 2008). The Wnt1-Cre transgenic line (Danielian et al., 1998) was used to delete Foxd3^flox and to lineage-map NC-using reporter strains Gt(ROSA)26Sor^tm1Sor (called R26R^lacZ) and Gt(ROSA)26Sor^tm1(EYFP)Cos (called R26R^YFP) (Soriano, 1999; Srinivas et al., 2001). The tamoxifen-inducible transgenic line (CAGG Cre-ER™) (Hayashi and McMahon, 2002) was used for inducible deletion of Foxd3. All lines are on a mixed genetic background and maintained in accordance with protocols approved by the Vanderbilt University Institutional Animal Care and Use Committee (IACUC).

Embryos were fixed in 4% paraformaldehyde (PFA) in PBS for 4 hours or overnight, and histology was performed using standard procedures (Presnell and Schreibman, 1997). Whole-mount immunostaining (Wall et al., 1992) and 5-bromo-4-chloro-3-indolyl–d-galactoside (X-gal) staining (Nagy et al., 2003) were performed as described. The following antibodies were used: chicken-anti-green fluorescent protein to detect yellow fluorescent protein (YFP) expression from the activated R26R^YFP allele (1:500, Abcam), goat-anti-Sox10 (1:20, Santa Cruz), goat-anti-SM22α (1:200, Abcam), mouse-anti-neurofilament (2H3, 1:1000, Developmental Studies Hybridoma Bank), mouse-anti-β-III tubulin (1:500, Covance, TUJ1), mouse-anti-α-smooth muscle actin (1:200, Sigma), mouse-anti-glial fibrillary acidic protein (1:200, Sigma), rabbit-anti-Foxd3 [1:500 (Tompers et al., 2005)], rabbit anti-smooth muscle myosin heavy chain (1:200, Biomedical Technologies), rabbit anti-Sox9 (1:500, Chemicon), rabbit anti-Runx2 [1:500 (Yang et al., 2004)], rabbit-anti-cleaved caspase3 (1:200, Cell Signaling), rabbit-anti-p75 (1:200, Promega), rabbit anti-peripherin (1:1000, Chemicon). Secondary antisera were purchased from Jackson ImmunoResearch and 4′,6-diamidino-2-phenylindole (DAPI) (1:5000, Molecular Probes) was used to detect nuclei. For sequential detection of Foxd3 and p75, anti-Foxd3 and Cy3-conjugated secondary antibody labeled sections were incubated in unconjugated anti-rabbit IgG (1:15) before immunodetection of p75. The Vectastain ABC Kit (Vector Laboratories) was used for colorimetric immunohistochemical experiments. Terminal deoxynucleotidyl transferase (TUNEL) analysis was performed using the In Situ Cell Death Detection Kit (Roche).

Cardiac/vagal neural tube (from mid-otic placode to somite four) and trunk neural tube segments (from somites 16 to 22) of embryos 9.0-9.5 days post coitum (dpc) were isolated by microdissection and dissociated from surrounding tissues with collagenase/dispase (Roche) using procedures modified from Stemple and Anderson (Stemple and Anderson, 1992). Explants were cultured in hypoxic conditions (3-6% oxygen and 5% CO₂) (Morrison et al., 2000) in wells coated with 30 μg/ml Fibronectin (Gibco) in self-renewal medium containing: Dulbecco's modified Eagle's medium low glucose (Invitrogen), 30% neurobasal medium (Invitrogen), 15% chick embryo extract (CEE), 2% B27 (Invitrogen), 1% N2 (Invitrogen), 117 nM retinoic acid (Sigma), 50 μM β-mercaptoethanol (Sigma), 20 ng/ml insulin-like growth factor (IGF) 1 and 20 ng/ml basic fibroblast growth factor (bFGF) (R&D Systems) using procedures modified from Morrison et al. (Morrison et al., 1999) and Stemple and Anderson (Stemple and Anderson, 1992). After 48 hours, neural tubes were physically removed. For adherent clonal cultures, NC cells were plated at low density (25 cells/cm²) ensuring that individual cells formed spatially distinct colonies. After 6 days, medium was changed to differentiation-promoting medium (10 ng/ml bFGF and 1% CEE) for 8 days before analysis of colony composition (Morrison et al., 1999; Mosher et al., 2007). Sphere cultures were maintained 8-10 days in low-adherence dishes (Costar) undisturbed and at low density (<1 cell per 10 μl of medium) to minimize fusion, and dissociated according to Fasano et al. (Fasano et al., 2007). All statistics were mean ± s.e.m. and P-values calculated with Student's t-test.

Vagal NC (from somites one to six) and cardiac NC (from the mid-otic placode to somite four) are initially overlapping cell populations that innervate the digestive tract and diaphragm, and function in cardiac development by mediating outflow tract septation, contributing VSMCs to the great vessels and generating parasympathetic cardiac ganglia (Allan and Greer, 1997; Hildreth et al., 2008; Hutson and Kirby, 2007; Young and Newgreen, 2001). Foxd3 is differentially required in these NC populations; NC-specific Foxd3-null embryos have no ENS, whereas cardiac outflow tract septation is overtly normal (Teng et al., 2008). This striking difference led us to examine temporal expression of Foxd3 in Wnt1-Cre; R26R^YFP lineage-traced NC from distinct axial origins: cranial, cardiac, vagal and trunk NC. Foxd3 expression overlapped with YFP-positive cells in pre-migratory NC at the dorsal-lateralmost region of the headfolds at 8.0 dpc (four-somite stage), and expression was maintained in migratory cranial NC at 8.5 dpc (see Fig. S1A,B in the supplementary material), in agreement with our previous work (Labosky and Kaestner, 1998). At vagal/cardiac somite levels, Foxd3 was co-expressed with YFP in ventral migrating NC at 8.5 dpc (Fig. 1A). Beginning at 9.0 dpc (14 somites), Foxd3 expression was extinguished in ventralmost NC cells (Fig. 1B, arrowhead) and by 9.5 dpc, Foxd3 expression was restricted to presumptive cranial and dorsal root ganglia (DRG), with no expression detected in cranial or cardiac NC mesenchyme (Fig. 1C and see Fig. S1C,D in the supplementary material). At 10.5 dpc, Foxd3 was co-expressed with Sox10, an NCSC and glial marker in NC at the periphery of vagal and trunk DRG (Fig. 1D,E). This expression pattern was similar to previously reported co-expression of p75, an NCSC and neuronal progenitor marker, and Sox10 (Sonnenberg-Riethmacher et al., 2001), that led us to examine expression of Foxd3 and p75. At 10.5-11.5 dpc, Foxd3 and p75 were co-expressed in NC at the level of the developing esophagus and trachea, and in cranial ganglia (Fig. 1F,G). Foxd3 and p75 were downregulated in cardiac NC before entry into the outflow tract (Fig. 1F) and in cranial NC-derived mesenchyme (Fig. 1G). These expression patterns indicate that Foxd3 is present in NCSCs and support a role for Foxd3 in segregation of neural, glial and ectomesenchymal lineages of the NC.

Foxd3 NC-conditional null mice die perinatally, presumably as a result of respiratory failure (Teng et al., 2008). To determine if this lethality is associated with defects in NC-mediated diaphragm innervation, we evaluated NC with Wnt1-Cre activated R26R^lacZ and X-gal staining. In Foxd3^flox/+; Wnt1-Cre control embryos at 16.5 dpc the phrenic nerves projected to the antral right and left leaflets of the diaphragm and elongated dorsally and ventrally through the central region of the muscle (Fig. 2A, top). By contrast, only a few NC-derived cells were present in a small cluster in the left leaflet of diaphragms from NC-conditional nulls (Fig. 2A, bottom). Examination of R26R^YFP with neurofilament expression (2H3) confirmed a paucity of neurons projecting into the diaphragm in mutant embryos (Fig. 2B). Additionally, mutant vagal NC were not present in the esophagus (Fig. 2C). These data indicate a vital role for Foxd3 in development of neural derivatives of the vagal NC.

To determine if Foxd3 is required for normal differentiation of cardiac NC-derived neuronal lineages, we examined contribution of lineage-traced control (Foxd3 heterozygous) and Foxd3 mutant NC to the parasympathetic cardiac neurons using β-III tubulin (TUJ1) immunodetection. From 12.5-17.5 dpc, NC contributed to the dorsal cardiac ganglia in heterozygous control embryos, but in mutants, NC cells were not present in this location (n=12 mutant embryos; Fig. 2D,E). In Foxd3 mutant embryos, NC did not contribute to parasympathetic innervation, and although non-NC neurons [presumably of placodal origin (Kirby, 1988)] were present, only rudimentary ganglia formed (Fig. 2E).

Our data suggest a model in which Foxd3 is initially expressed in all NC but then extinguished in cardiac NC before outflow tract septation, as a requisite step in the differentiation and generation of VSMCs in the great vessels. Although Foxd3-deficient NC mediate outflow tract septation, it was unknown whether the normal contribution of VSMC progenitors to pharyngeal arch arteries required Foxd3 expression in NC progenitors. Previous lineage-tracing studies demonstrated that cardiac NC gives rise to VSMCs in the aortic arch, but NC-derived cells do not populate the descending aorta, forming a distinct boundary with mesoderm-derived VSMCs at the junction of the left fourth and sixth arch arteries with the dorsal aorta (Jiang et al., 2000; Le Lievre and Le Douarin, 1975). To investigate if Foxd3 is required for cardiac NC to contribute appropriately to the great vessels, we compared distributions of lineage-traced control and conditional Foxd3-null NC in the aorta. At 16.5 dpc, NC cells populated the aortic arch of both Foxd3 heterozygous control and mutant embryos (Fig. 3A,B). The sharp boundary between NC- and mesoderm-derived VSMC was apparent at the level of the ductus arteriosus in control embryos (Fig. 3A, arrow). However, this border was no longer respected in Foxd3 mutant embryos and NC-derived cells were detected caudally, throughout the descending aorta, to at least the level of the diaphragm. Confocal analysis of the aorta subjected to YFP immunodetection co-labeled with smooth muscle 22α (SM22α) or smooth muscle myosin heavy chain (SM-MHC) showed Foxd3 mutant NC in the VSMC layer of the descending aorta (Fig. 3C-H). Further analysis of VSMC markers in transverse sections from lineage-traced 14.5 and 16.5 dpc embryos confirmed contribution of mutant NC to VSMCs within the tunica media of the descending aorta (see Fig. S2F-H in the supplementary material; data not shown). The relative contribution of NC cells in the descending aorta at 14.5-16.5 dpc was greater at regions proximal to the normal NC boundary (average YFP-positive vascular area=31% just caudal to the ductus arteriosus, n=4) compared with more distal regions (8.2% YFP-positive, at the diaphragm), demonstrating a rostral-to-caudal gradient of ectopic mutant NC. These data may indicate that Foxd3 mutant cardiac NC-derived VSMCs are unable to maintain their position rostral to this border, or alternatively, other initially non-VSMC NC progenitors, presumably from vagal and/or trunk NC, may inappropriately acquire VSMC fates and aberrantly contribute to the descending aorta.

The ectopic location of Foxd3 mutant NC-derived VSMCs in the aorta raised the question of whether other mesenchymal NC populations were similarly affected by loss of Foxd3. Cranial NC generate both neural (cranial ganglia) and mesenchymal (facial bone and cartilage) derivatives, therefore we examined differentiation and patterning of this NC population. Analysis of TUJ1 expression (neurons) in lineage-traced embryos revealed that, in contrast to control embryos (Fig. 4A), mutant NC showed limited contribution to cranial nerves, with co-expression of TUJ1 and YFP primarily in DRG (Fig. 4B). During early stages of endochondrial bone formation (11.5-12.5 dpc), chondrogenesis was not apparent in control embryos: Runx2, marking osteoblast progenitors, and collagen-II, a chondrocyte marker, were not detected (data not shown). Sox9 expression (marking osteo-chondroprogenitors) was only faintly detected in the mandibular component of the first pharyngeal arch at 11.5 dpc (Fig. 4C). In 12.5 dpc control embryos, Sox9 was uniformly expressed in the distal mandibular precartilage primordium (Fig. 4E-E′), and this expression domain largely overlapped with that of Runx2 (Fig. 4F-F′, arrowheads). In stark contrast, mutant NC showed precocious induction of Sox9 at 11.5 dpc (Fig. 4D). By 12.5 dpc, Sox9 was strongly expressed in the medial region of the developing mandible with lower expression at the lateral edges (Fig. 4G-G′, arrow) whereas Runx2 was highly expressed at the lateral edges (Fig. 4H-H′, arrowheads). Our results indicate that Foxd3-null cranial NC cells preferentially formed non-neural lineages and that differentiation of mesenchymal NC lineages was accelerated.

The observation that Foxd3 is expressed in early, presumably multipotent, NC, and downregulated in later migratory NC, raised the hypothesis that Foxd3 is required to maintain the multipotent nature of NC. To address this in a tightly controlled manner, we turned to a culture system in which neural tube-derived NCSCs can be maintained as undifferentiated cells (Stemple and Anderson, 1992). Approximately 96% of cells actively migrating away from neural tube explants demonstrated Wnt1-Cre-based reporter activation (see Fig. S3 in the supplementary material). Nuclear Foxd3 expression was detected in control NC cultures but undetectable in mutant NC (Fig. 5A,B), demonstrating efficient deletion of the coding region. Heterozygous control NC cultured for 48 hours in self-renewal medium maintained expression of NCSC markers p75 and Sox10, and few cells expressed smooth muscle α-actin (SMA), a marker of myofibroblast and VSMC differentiation (Fig. 5C,E). By contrast, almost all mutant NC cells from both vagal and trunk levels were devoid of p75 [decreased by 99% in vagal and trunk NC cultures (P<0.05); Fig 5D,F and see Figs S4 and S5 in the supplementary material]. Similarly, p75 expression was reduced in vivo in fate-mapped vagal and trunk NC in mutant embryos at 9.5 dpc (see Fig. S4D,E,I,J in the supplementary material), suggesting a role for Foxd3 in establishing or maintaining p75-expressing NCSCs. Additionally, expression of Sox10, normally detected in NCSCs and differentiated glia, was reduced in mutant cultures. The few Sox10-positive cells present did not co-express p75 (Fig. 5F), suggesting differentiation towards the glial lineage. Mutant explant cultures contained increased numbers of SMA-expressing cells (Fig. 5D), demonstrating that mutant NC differentiate towards the myofibroblast lineage under conditions that normally maintain undifferentiated NCSC.

To further evaluate if Foxd3 is required to repress differentiation of individual NC cells towards multiple lineages, thereby ensuring continued multipotency, we performed in vitro culture of NC cells at clonal density (Stemple and Anderson, 1992). Individual control NC cells typically proliferated to form round, tightly packed colonies whereas the majority of mutant colonies consisted of loosely-packed fibroblast-like cells, consistent with the data above that showed a tendency for mutant NC to adopt a myofibroblast fate (Fig. 5G,H). To examine differentiation status at a molecular level, NC colonies were cultured for six days under self-renewing conditions and evaluated for expression of p75 or markers of neural (Peripherin), glial [glial fibrillary acidic protein (GFAP)], and myofibroblast (SMA) differentiation. Control NC colonies expressed p75 (Fig. 5I), and significant neural, glial and myofibroblast differentiation was not detected (Fig. 5K). In stark contrast to controls, clonally derived mutant NC colonies rarely expressed p75 (Fig. 3J), and although many NC colonies (60-70%) contained SMA-positive cells, the remaining non-myofibroblast colonies did not exhibit aberrant neural and glial differentiation (Fig. 5L). These data indicate that Foxd3 is required to repress myofibroblast differentiation and maintain NCSCs in an undifferentiated state.

Consistent with increased apoptosis in the neural tube of Foxd3 NC-mutant embryos (Teng et al., 2008), the loss of p75 and Sox10 in mutant NC could reflect selective apoptosis of NCSCs. However, apoptosis was only modestly increased in Foxd3 mutant NC in vitro, measured by cleaved-caspase3 detection or TUNEL assay (see Fig. S5 in the supplementary material). Comparison of control and mutant outgrowths showed no significant differences in initial cell numbers or migration index despite loss of p75 expression (see Fig. S4C,H in the supplementary material). In a complementary approach, we failed to detect restoration of NCSC numbers when neural tube explants were exposed to an apoptosis inhibitor [the caspase inhibitor, Z-VAD-FMK, (Maynard et al., 2000; Uesaka et al., 2007)]. Using TUNEL assays to detect caspase-dependent and independent apoptosis, Z-VAD-FMK reduced cell death in both control and mutant NC, but p75 expression was not restored (see Fig. S5 in the supplementary material). This suggests a direct cell-autonomous role for Foxd3 in maintaining NCSC multipotency, rather than simply increasing cell survival.

Self-renewal is a defining characteristic of stem cells, and Foxd3 is crucial for this property in ESCs (Hanna et al., 2002; Liu and Labosky, 2008). Therefore, to determine if Foxd3 regulates self-renewal in NCSCs we used neurosphere culture assays to measure this crucial stem cell property. Compared with wild-type NC, Foxd3 mutant NC generated smaller neurospheres at much lower frequency (Fig. 6A-D). Analysis of secondary neurosphere formation from dissociated individual primary neurospheres showed that Foxd3 mutant NC was significantly impaired in self-renewal capacity (vagal NC reduced by 64%; trunk NC by 47%; Fig. 6E). To control for the smaller size of primary Foxd3 mutant neurospheres, we also dissociated pools of primary neurospheres, plated equivalent cell numbers, and again observed similar reductions in self-renewal from mutant neurospheres (reduced 63% for vagal NC, 55% for trunk NC; Fig. 6F-H). Taken together, our findings reveal a conserved role for Foxd3 in both NC and ESCs in the maintenance of stem cell self-renewal.

If Foxd3 plays a role in maintaining multipotency of NCSCs, it is possible that individual mutant NC may have a restricted ability to generate progeny towards one or more NC lineages. In vivo, Foxd3 mutant NC gave rise to mesenchymal lineages, but showed limited capacity to generate neurons (Figs 2 and 3). Similarly, NC contributed some neurons and glia to smaller than normal DRG, but derivatives more distal to the neural tube were absent (Teng et al., 2008). To directly address whether individual NC cells have reduced lineage potential, clonally-derived colonies from vagal and trunk NC were cultured under differentiation-promoting conditions, and lineage potency measured by immunocytochemistry for differentiation markers as above and scored as described (Morrison et al., 1999) (Fig. 7A). Compared with wild-type (or heterozygous; see Fig. S6C in the supplementary material) control cultures, Foxd3 mutant NC had significantly reduced numbers of neuron (N)-containing colonies (mutant vagal and trunk N-containing colonies reduced by 61 and 69%, P<0.001; see Fig. S6A,B in the supplementary material), supporting a role for Foxd3 in maintaining neural potential, consistent with in vivo results (Fig. 2). Further quantification indicated that individual mutant cells had additional defects in potency. For example, a reduced proportion of bi-potent cells produced both neurons and myofibroblasts (N+M; Fig. 7B,C). Most significantly, mutant NC rarely (1% of vagal NC, 2% of trunk NC, P<0.001) formed multipotent (neural/glial/myofibroblast, N+G+M) colonies compared with control NC (10% of vagal NC, 16% of trunk NC). The rare mutant multipotent colonies were significantly smaller than controls, and almost always showed a strong differentiation bias towards a single fate (N, G or M), with few cells adopting other fates. Coincident with loss of multipotency, mutant clonal cultures displayed an increased incidence of myofibroblast (M-only) colonies, but the percentage of neural or glial (N-only or G-only) colonies was unaffected. This altered spectrum of potential in control versus mutant NC was similar in vagal and trunk NC, demonstrating that loss of multipotency, and an increased bias toward the myofibroblast lineage, is an equivalently penetrant phenotype along the anteroposterior axis (Fig. 7B,C). In a complimentary approach, inducible deletion of Foxd3 after NC migrated from the neural tube resulted in reduction of Sox10 expression, and recapitulated these changes in NC potency (loss of N+G+M and bias towards M-only; see Fig. S7 in the supplementary material). In all of the above assays, Foxd3-heterozygous NC were similar to wild type in p75 expression, self-renewal and lineage composition of clonal cultures (see Fig. S5C in the supplementary material; data not shown), consistent with absence of NC defects in heterozygous embryos or mice. Together, these data clearly demonstrate that Foxd3 is required to maintain multipotency in NCSCs, but is not required for differentiation to any specific lineage.

Our findings demonstrate a cell-autonomous requirement for the transcription factor Foxd3 in maintaining both self-renewal and multipotency of NC progenitors. Foxd3 plays a pervasive role in preserving these definitive stem cell properties throughout the NC; the conditional inactivation of Foxd3 affected cranial, cardiac, vagal and trunk domains similarly. Clonal analyses of Foxd3 mutant NC indicated a severe reduction in self-renewal and an increased bias towards myofibroblast-directed NC, which was associated with an almost complete loss of multipotent NCSCs. Overall, neural differentiation was reduced in mutant cultures, suggesting that multipotent (N+G+M) cells and neural-containing bipotent cells (N+M or N+G) undergo a selective loss of neural potential, moving towards myofibroblast or glial fates. However, the frequency of neural-only or glial-only restricted progenitors was not significantly altered, indicating that lineage entry is not impaired by loss of Foxd3. Our current model is that Foxd3 permissively maintains multipotent NCSCs in an uncommitted state, rather than functioning as part of an instructive cue to instill neural or glial competency.

Our results support a model in which Foxd3 functions as a gatekeeper in multipotent NCSCs to repress mesenchymal lineages and preserve neural fates and multipotency (Fig. 7D). NC-specific inactivation of Foxd3 caused profound deficits in innervation of several target organs, and NC cells aberrantly adopted mesenchymal cell fates. This is consistent with the notion that Foxd3 functions in opposing programs controlling neural versus mesenchymal potential of NC. Without Foxd3, individual NC cells committed to one of the three major lineages (neural, glial or myofibroblast) may be unaffected, whereas multipotent and bi-potent progenitors destined to more distally located neural and/or glial fates are unable to repress differentiation toward mesenchymal lineages. An alternative, although not mutually exclusive, explanation is that the progressive reduction in mutant NC self-renewal we describe here may selectively exhaust the neural-containing progenitor pool, so that by the time NC reach distant locations, their overall neural potential is dramatically reduced or absent. In either of these possibilities, we suggest that Foxd3 promotes self-renewal and multipotency independent of NCSC survival mechanisms because blocking caspase activity in Foxd3 mutant NC cultures did not restore expression of NCSC-associated proteins.

Distinct NC cell fates are thought to arise via progressive restrictions in lineage potential before and during NC migration (Krispin et al., 2010; Young et al., 2003). Foxd3 is expressed in pre-migratory NC and is maintained in early migrating NC progenitors, but is downregulated in cranial and cardiac NC mesenchyme well before chondrocyte, osteoblast or VSMC differentiation. Our findings demonstrate an early requirement for Foxd3 in repression of ectomesenchymal fates shortly after cells leave the neural tube, before arrival at the pharyngeal arches or cardiac outflow tract. During craniofacial development, Sox9 is required for specification of bi-potent osteochondroprogenitors (Akiyama et al., 2002; Lefebvre et al., 1997; Mori-Akiyama et al., 2003) and subsequently, Runx2 acts as a determining factor for differentiation of osteoblasts (Ducy et al., 1997; Komori et al., 1997). Loss of Foxd3 in the NC resulted in ectopic activation of both Sox9 and Runx2 in cranial NC, therefore premature endochondrial bone and cartilage formation may underlie the severe craniofacial defects observed in Foxd3 NC mutant embryos (Teng et al., 2008). In addition to controlling the timing of cranial NC differentiation, we also provide evidence that Foxd3 is required to restrict NC progenitors from VSMC and myofibroblast fates. Normally, cardiac NC generates VSMCs in precise regions of the aortic arch and great vessels (Jiang et al., 2000; Le Lievre and Le Douarin, 1975). The contribution of NC-derived VSMCs extends from the ascending aorta caudally to the precise location where the ductus arteriosus connects to the aortic arch. Further caudally, VSMCs in the descending aorta are derived from mesoderm (Esner et al., 2006; Wasteson et al., 2008; Wiegreffe et al., 2007). Our lineage-tracing experiments showed that Foxd3 mutant NC extended caudally well past the aortic boundary, but other NC-mesoderm boundaries in the pulmonary arteries were still respected (data not shown). In multiple locations in vivo and in vitro the neural potential of mutant NC was decreased, whereas mesenchymal potential was increased. Therefore, it is tempting to speculate that mutant NC located in the distal aorta are derived from cells that would normally give rise to neural and/or glial cells in the nearby gastrointestinal tract or sympathetic ganglia that underwent an inappropriate lineage decision to a mesenchymal fate. Lineage analysis of NC populations captured on the basis of emigration and Foxd3 expression status could begin to define these transitional cell states in a prospective manner, and live cell imaging in the whole embryo could resolve some of these unanswered questions.

One surprising discovery we describe here is that Foxd3 functions to preserve multi (or bi-) potency by maintaining the neural lineage choice open in multipotent NCSCs. Therefore, regulation of Foxd3 expression is important with respect to the ability of NC to interpret environmental signals effecting lineage allocation. Combinatorial Wnt and BMP signaling synergistically maintain NCSC multipotency and suppress differentiation (Kleber et al., 2005; Lee et al., 2004; Wurdak et al., 2005), and transforming growth factor (TGF) β signaling promotes non-neural lineage differentiation from NCSCs (Shah et al., 1996; Wurdak et al., 2005). Our findings here are consistent with the possibility that Foxd3 functions at a crucial early step in NCSC multipotency and self-renewal by permissively maintaining responsiveness to Wnt/BMP signaling and/or functions in opposition to TGFβ signaling to promote neural fates. In contrast to factors such as Bmi1 and Hmga2 that specifically regulate self-renewal independent of initial NC multipotency (Molofsky et al., 2003; Nishino et al., 2008), Foxd3 is required in newly generated NC for both multilineage potential and self-renewal. It is likely that these properties are intimately coupled by Foxd3; reduced self-renewal of mutant NC may be because of increased differentiation and not as a result of direct cell cycle regulation. Proliferation of Foxd3 mutant NC is not reduced in vivo, or in Foxd3 mutant ESCs despite diminished self-renewal and multipotency (Liu and Labosky, 2008; Teng et al., 2008).

In addition to Foxd3, the only other transcription factor definitively linked to multipotency of NC is Sox10, which maintains NCSCs by inhibiting differentiation into neural cell fates to allow for gliogenesis (Kim et al., 2003). Heterozygous loss of Sox10 results in defects in glial and melanocyte lineages and, similarly to Foxd3 mutant embryos, Sox10 homozygous mutants fail to form the ENS (Southard-Smith et al., 1998). Sox10 expression is markedly reduced in both Foxd3 mutant embryos (Teng et al., 2008) and NC cultures. Although it is possible that Foxd3 may regulate Sox10 expression and vice versa, it is also likely that they are coordinately regulated (Drerup et al., 2009). Sox10 preservation of glial fates may act in synchrony with Foxd3-mediated maintenance of neural competency within NCSCs to ensure multipotency. Future studies will address whether these two transcription factors act in a single common pathway or in parallel to regulate NCSC multipotency.

The precise molecular mechanisms by which Foxd3 controls self-renewal and multipotency are incompletely understood, but presumably involve direct regulation of gene transcription and/or regulation of chromatin modifications (Liber et al., 2010; Xu et al., 2009). Although primarily characterized as a transcriptional repressor, Foxd3 has both activator or repressor activity, depending upon the cellular context (Guo et al., 2002; Lee et al., 2006; Steiner et al., 2006; Sutton et al., 1996; Yaklichkin et al., 2007). A distinct possibility is that interactions with differentially expressed co-factors modulate Foxd3 function. Consistent with this idea, Foxd3 mediates repression of Mitf (thereby inhibiting melanocyte cell fate) via a protein-protein interaction with Pax3, and not by direct effects on the Mitf cis-regulatory sequences (Thomas and Erickson, 2009). Together with our previous findings, the model emerging is of a complex, context-dependent role for Foxd3 as a multifunctional protein regulating gene expression in distinct cellular programs. A further possibility is that Foxd3 controls multipotency and self-renewal of NC by its central participation in the overall repression of gene regulatory networks promoting NC differentiation to non-neural lineages.

Our work now adds NCSCs to ESCs and TSCs as multipotent progenitor pools with a strict requirement for Foxd3 in maintaining multipotency and self-renewal (Hanna et al., 2002; Liu and Labosky, 2008; Tompers et al., 2005). This molecular conservation of function suggests that Foxd3 may be a fundamental integrator and/or effector of the upstream pathways modulating `stemness'. Determining if Foxd3 regulates common target genes or signaling pathways in these divergent progenitor populations could lead to a clearer fundamental understanding of the levels of regulation involved in multipotency, and assist with development of methods for accessing control of populations for use in stem cell-based therapies; these will be important future directions.

We thank A. Frist, A. LeGrone and J. Kappa for technical assistance; E. Armour for preliminary apoptosis data; X. Yang for advice and the Runx2 antibody; and D. Bader, S. Huppert, B. Nelms, J. Saint-Jeannet, M. Southard-Smith and C. Wright for helpful discussions and crucial reading of the manuscript. The 2H3 monoclonal antibody was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD. Olympus FV1000 confocal microscopy was performed in part through use of the VUMC Cell Imaging Shared Resource. This work was supported by an R01 grant from the NIH (HD36720) to P.A.L., Vanderbilt University Medical Center Academic Program Support, and predoctoral fellowships from the A.H.A. (0615209B) and NIH (NS065604) to N.A.M. Deposited in PMC for release after 12 months.