WNT/β-catenin signaling mediates human neural crest induction via a pre-neural border intermediate

Neural crest (NC) is a transient multi-potent cell population that emerges between the forming neural plate and the non-neural/surface ectoderm. The ‘new head’ model (Gans and Northcutt, 1983) proposed that the NC was an evolutionary advancement that distinguished vertebrates from other chordates, and linked the derivatives of NC and cranial placodes to the predatory life style of vertebrates. Interestingly, NC and cranial placodes were further proposed to derive from single precursors (Baker and Bronner-Fraser, 1997; Gans and Northcutt, 1983), and phylogenetic studies suggested that their lineages diverted after the neural-epidermal segregation in the early ectoderm (Groves and LaBonne, 2014; Patthey and Gunhaga, 2014; Schlosser, 2008). In avian embryos, neural induction is thought to initiate slightly before or during gastrulation (Streit et al., 2000; Wilson et al., 2000,, 2001), and specification and lineage assays identified the lateral epiblast at early gastrula stages as poised to form NC independently of neural fate (Basch et al., 2006; Patthey et al., 2009). In pre-gastrula amphibian embryos, evidence for establishment of a neural boundary via planar signaling was reported (Savage and Phillips, 1989; Zhang and Jacobson, 1993). A recent study in amphibians proposes a shared genetic regulatory network of blastula pluripotent cells with developing NC cells at the neural border (Buitrago-Delgado et al., 2015). These studies in avian and amphibian embryos hint at an early fate specification for NC. Recent human embryo studies have validated several markers of pre-migratory NC cells (Betters et al., 2010) but, similar to rodents and other mammals, no information has been published regarding earlier stages of specification or induction of NC. Interspecific differences in NC development have been postulated (Barriga et al., 2015); it remains unclear when mammalian NC cells are specified during development and if their induction involves mesoderm and neural tissues. There is an urgent need to develop and optimize models for mammalian NC development to study these aspects.

Embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs), derived from mouse, primate and human blastocysts, have been extensively employed for studies of developmental processes that otherwise are difficult to carry out in vivo (Evans, 2011; Pera and Trounson, 2004; Rodda et al., 2002; Sasai, 2013; Spagnoli and Hemmati-Brivanlou, 2006). The current methods for generating human NC tissue, however, often involve dense self-organizing cell conglomerates (embryoid bodies, neural rosettes, or confluent cultures) that produce autocrine signaling. They also include steps and components that have precluded the elucidation of the specific signaling pathways and tissue contributions required for the early stages of human NC formation.

Here, we set out to develop a suitable model in which to study the contributions of specific signaling pathways and early embryonic tissues during human NC formation. We establish a culture system that generates human NC cells validated with markers demonstrated in human embryos (Betters et al., 2010). Furthermore, our protocol demonstrates cost effectiveness, high efficiency and unprecedented speed, and dispenses with the use of serum replacement, heregulin, activin antagonists, and the pre-differentiation (ESC expansion) period employed by previous protocols (Chambers et al., 2013; Fukuta et al., 2014; Menendez et al., 2011; Mica et al., 2013; Umeda et al., 2015). Our study demonstrates a key role for canonical WNT signaling in human NC formation, confirming contributions from bone morphogenetic protein (BMP) and fibroblast growth factor (FGF) pathways and supports an independent origin of NC from PAX6⁺ neural progenitors. Furthermore, our study suggests a distinct cellular origin of human NC cells through a novel early intermediate that precedes and is differentially regulated from the neural border cells.

To enable a clearer analysis of the signaling pathways and possible tissue contributions involved in early human NC development, we employed low-density cultures of dissociated human ESCs (hESCs) in defined serum free-media under WNT activation (Fig. 1A). We chose to study the effect of WNT signaling because it has been shown to be instrumental in NC induction in both human and animal models (Garcia-Castro et al., 2002; Menendez et al., 2011; Mica et al., 2013). We plated 20,000 hESCs/cm² in DMEM/F12 medium, plus B27 supplement and bovine serum albumin, and employed CHIR 99021 (CHIR, a GSK3 inhibitor) to activate WNT signaling. CHIR-treated WA01 hESCs, Y6 iPSCs and RIV9 iPSCs displayed robust expression of the characteristic NC markers SOX10, PAX7 and TFAP2A by day 5 (Fig. 1B-D; Fig. S1; data not shown). In the absence of CHIR, human ESCs did not generate SOX10- or PAX7-expressing cells (Fig. 1B). The differentiating cells expanded by 9.3±1.5 times under CHIR treatment and resulted in 63.1±9.6% SOX10⁺, 78.6±7.9% PAX7⁺ and 84.4±7.6% TFAP2A⁺ cells (Fig. 1C). The vast majority (∼95%) of the SOX10⁺ cells co-expressed PAX7 and TFAP2A (Fig. 1D). The remaining SOX10⁺ cells co-expressed either TFAP2A (∼5%) or PAX7 (<0.5%). Transcriptional analysis of a panel of NC markers (PAX3, PAX7, TFAP2A, SNAI2, ZIC1, MSX1, MSX2, NR2F1) revealed that expression levels of these markers were increased by 30- to 2000-fold (Fig. 1E). CHIR substitution by Wnt3a in our culture conditions also led to robust NC induction (as revealed by SOX10-PAX7 immunostaining, and PCR of NC markers; data not shown).

NC-like cells have been reported to arise spontaneously from cultures plated at high densities (Leung et al., 2013). Indeed, CHIR^– control cultures seeded at 50,000-100,000 cells/cm² generated PAX7⁺ and SOX10⁺ cells (Fig. S2A,B). Importantly, inhibition of WNT signaling through DKK1 or the tankyrase inhibitor XAV 939 (Fig. S2C,D) repressed the induction of NC markers, suggesting a role of paracrine WNT signaling in NC induction.

Upon CHIR treatment, expression of the hESC markers OCT4 (POU5F1 – Human Gene Nomenclature Database), NANOG and FOXD3 was greatly reduced by day 3 of differentiation (Fig. 1F). FOXD3, which is also expressed by NC cells, increased dramatically from day 4 onwards. In accordance with this, expression of the early NC genes (or so-called neural border genes) PAX7, PAX3, MSX1 and TFAP2A (Basch et al., 2006; de Croze et al., 2011; Sato et al., 2005; Tribulo et al., 2003) was increased at day 3 by 7- to 50-fold (normalized to undifferentiated ESCs) and more robustly by day 4 (Fig. 1G; Fig. 3D). Their expression continued to rise or was maintained at day 5, when the majority of the SOX10⁺ cells arose (Fig. 1B; Fig. 6A), indicating the acquisition of a definitive NC fate.

Previous NC induction protocols frequently employed activin/TGFβ antagonism using the type I Activin/TGFβ receptor inhibitor SB 431542 (Chambers et al., 2013; Menendez et al., 2011; Mica et al., 2013). Transient (3 days) and continuous (5 days) treatment with SB 431542 in our differentiating cultures had either no promoting effect on early NC gene induction (PAX3 and PAX7) or severely repressed the expression of mature markers (SOX10 and FOXD3) (Fig. 1H), suggesting that short-term or prolonged activin/TGFβ antagonism had a negative effect on human NC induction.

To assess further the human NC-like cells generated by WNT induction, we tested their capacity to contribute to terminal derivatives associated with NC. CHIR-treated cells subjected to terminal differentiation (see Materials and Methods) form ectomenchymal and non-ectomenchymal derivatives (Fig. 1I), including chondrocytes, osteoblasts, peripheral neurons [HuC/D⁺ (ELAVL3/4), TUJ1⁺ (TUBB3) PRPH⁺, ISL1⁺], glial precursors (S100β⁺ GFAP^+/−) and melanoblasts (MITF⁺ SOX10⁺). Similar results were obtained with WNT3A-treated cells (Fig. S3), demonstrating the multipotency of our WNT-induced NC cells.

To determine the mechanism of action of CHIR in our NC induction model, we measured transcription of AXIN2, a canonical WNT target gene (Lustig et al., 2002). We found that it progressively increased day by day under CHIR treatment, whereas its transcription remained unchanged in non-treated cultures (Fig. 2A). We therefore performed knockdown of CTNNB1 (β-catenin), a key mediator of the transcriptional response of canonical WNT signaling. Transduction of CTNNB1 short hairpin (sh)RNA dramatically repressed protein expression and transcription (>80%) of the CTNNB1 gene, compared with control (luciferase) shRNA transduction (Fig. 2B,C; Table S2). Although repression of CTNNB1 did not repress expression of OCT4 at the ESC stage (Fig. 2D), induction of SOX10⁺ cells (Fig. 2E) and transcription of key NC markers including PAX7, PAX3, MSX1, SNAI2 and FOXD3 was dramatically reduced (Fig. 2F). Reducing CTNNB1 expression partially rescued expression of selected non-neural ectoderm genes [DLX5 and ECAD (CDH1), but not, for example, GATA3] (Fig. 2G) suggesting a requirement for restricted WNT signaling for non-neural ectoderm induction (Dincer et al., 2013; Leung et al., 2013).

Given the proposed fate relationships between placodes and NC, we decided to address the possible contribution of these relationships in our system. We found that the gene encoding the epithelial cell-cell adhesion molecule E-cadherin (CDH1, which is also highly expressed in hESCs), and the nuclear factors FOXC1, EYA2 and ISL1 were highly induced in control day 5 cultures (Fig. 3A). Upon CHIR administration, however, these surface ectoderm and preplacodal ectoderm-associated transcripts were repressed to basal levels (Fig. 3A). In a time course assay, we observed that the early non-neural ectoderm markers DLX5 and GATA3 were progressively induced in control cultures, and were induced 1 to 2 days earlier than the preplacodal ectoderm marker SIX1 and the epidermal ectoderm marker ΔNP63 (an isoform of TP63), respectively, confirming the sequential induction of competence and maturation markers during human placodal ectoderm cell induction (Fig. 3B,C) (Dincer et al., 2013; Leung et al., 2013). CHIR treatment curtailed the progressive induction of early (DLX5 and GATA3) and mature (ΔNP63 and SIX1) non-neural ectoderm markers, suggesting that canonical WNT activation restricts non-neural ectoderm induction (Fig. 3B,C) and simultaneously promotes NC induction (Fig. 3F).

Competence factors common for both NC and cranial placodes have also been identified, including Tfap2a and Zic1 (de Croze et al., 2011; Hong and Saint-Jeannet, 2007; Kwon et al., 2010; Phillips et al., 2006; Streit and Stern, 1999; Tribulo et al., 2003). Both TFAP2A and ZIC1 were induced at high levels at the first few days of cranial placode (−CHIR) induction and displayed significantly higher expression levels than the NC (+CHIR)-inducing condition at differentiation day 3 (Fig. 3D,E). Under the NC (+CHIR) condition, both genes were only induced by day 4 or 5, suggesting that they were mature NC markers. Induction of these factors, therefore, was differentially regulated during human NC and cranial placode development, and supports an early segregation of NC and cranial placode progenitors prior to the establishment of the neural border stage (Fig. 3F).

The BMP signaling pathway has also been strongly implicated in NC induction in model organisms (Stuhlmiller and Garcia-Castro, 2012a). We found that blockage of BMP signaling with the BMP inhibitor noggin at 300 ng/ml from day 0, day 1 or day 2, for 24 h, led to a dramatic inhibition of human NC induction to levels below 5% of the total cell population (Fig. 4A,B). BMP blockade from day 3 to day 4 instead only partially curtailed NC induction (28.7±6% SOX10⁺ cells). Noggin treatment from day 0 to day 1 also eliminated PAX7⁺ and TFAP2A⁺ cells, as indicated by the absence of staining of day 5 cultures with SOX10, PAX7 and TFAP2A antibodies (Fig. 4C). In agreement with the antibody staining results, noggin treatments of 1 day, 3 days or 5 days duration all repressed transcription of NC markers, and 24 h of noggin treatment from day 0 to day 1 was sufficient to significantly repress NC gene transcription, indicative of an early essential role of BMP activity in NC induction (Fig. 4D). Noggin treatments correspondingly promoted transcription of neural-related genes, including SOX1, HES5 and SOX2, in agreement with the role of BMP repression in neural induction (Fig. 4E). Neural induction appeared to require continuous BMP repression as 5 days of noggin treatment led to the highest expression level of HES5, a mature neuroepithelial marker (Fig. 4E).

We found that noggin treatment from day 0 to day 1 did not affect expression of the WNT target AXIN2 in a time course experiment (measured at day 1, day 3 and day 5) suggesting that BMP signaling might have a direct role during the early phase of NC induction (data not shown). WNT has been suggested to cooperate with BMP signaling to induce neural plate border formation in avian embryos (Patthey et al., 2009). Continuous or transient treatment with BMP4 ligand (20 ng/ml) with CHIR, however, dramatically repressed induction of the early NC genes PAX3 and PAX7 (Fig. 5A). BMP4 ligand also repressed expression of neural-related genes (PAX6 and SOX2; Fig. 5B) but promoted the transcription of genes indicative of non-neural ectoderm (e.g. GATA3, DLX5 and ECAD; Fig. 5C). We further performed a BMP-titration assay to better define the role of BMP in NC induction. Expression of SOX10, PAX7 and TFAP2A displayed no obvious gains or losses upon BMP4 addition from day 0 to day 1 at a dosage range of 0.005-0.01 ng/ml. However, at 0.05 ng/ml BMP4 or above, expression of SOX10 and PAX7 was severely reduced, whereas TFAP2A expression was detected ubiquitously, suggestive of non-neural ectoderm identity (Fig. 5D).

NC cells are believed to derive from neural plate cells, but avian experiments have suggested an independent origin (Basch et al., 2006), and the early requirement for BMP signaling in human NC induction also deviates from this notion. The neural induction program in human embryos and ESCs has been studied (Greber et al., 2011; Hou et al., 2013; Zhang et al., 2010). Although there was transient and variable upregulation of the pluripotency/neural markers SOX2, SOX1 and OTX2 during our NC cell induction, other pre-neural and mature neural progenitor markers, such as LHX2 and HES5, were never induced (data not shown; Fig. S4A,B). Our results do not support a robust execution of the pre-neural and neural program during induction of human NC cells.

PAX6 has been proposed as the earliest definitive neuroectoderm marker in human embryos and ESCs (Chambers et al., 2009; Gerrard et al., 2005; Surmacz et al., 2012; Zhang et al., 2010). We performed triple labeling of OCT4, PAX6 and SOX10 in day 1, day 3 and day 5 control and CHIR-treated cultures, to denote ESC, neuroectoderm and NC fates, respectively (Fig. 6A). Under control (−CHIR) and NC-inducing (+CHIR) treatments, we observed progressive reduction of OCT4⁺ cells from day 1 to day 5. A small amount of PAX6⁺ cells were induced in the control cultures starting at day 3 but there was no further increase of PAX6⁺ cells at day 5, consistent with our findings that control cultures mainly formed non-neural cell types. Similar to control cultures, in the presence of CHIR, minimal PAX6⁺ cells (≤5%) were induced at all time points. Some SOX10⁺ cells (∼10% of the total population), however, started to arise at day 3. By day 5, the number of SOX10⁺ cells climbed to ∼60% or higher but the number of PAX6⁺ cells barely reached 5% of the total population (Fig. 6A,B). No co-expression or spatial relationship of the SOX10⁺ and the small population of PAX6⁺ cells were noted at day 5. To address possible contributions of PAX6 in the CHIR-mediated NC induction, we performed PAX6-knockdown experiments. Application of PAX6 shRNA to CHIR-treated cells led to robust reduction of PAX6 protein and transcript (Fig. 6C,D). Notably, PAX6 knockdown cultures did not exhibit a reduction in NC markers and, instead, selected markers such as SOX10 and FOXD3 displayed higher expression (Fig. 6E; data not shown). We conclude that human NC cells can be generated independently of PAX6⁺ neuroectoderm cells.

FGF signaling has been implicated in NC induction in animal models, either acting directly on NC precursors or via mesoderm induction (Mayor et al., 1997; Stuhlmiller and Garcia-Castro, 2012b; Yardley and Garcia-Castro, 2012). Treatment of differentiating hESCs with PD 173074, an FGFR1- and FGFR3-specific inhibitor, repressed the expression of the FGF targets SPRY1 and SPRY2 by 40-50% (Fig. 7A). Examination of NC markers at differentiation day 5 revealed decreased expression of all examined NC genes when PD 173074 was administered continuously for 5 days (Fig. 7B), suggesting that FGF signaling is required for human NC induction in accordance with previous findings from animal models (Stuhlmiller and Garcia-Castro, 2012b).

WNT signaling is known to promote mesoderm induction in differentiating hESCs (Funa et al., 2015; Sumi et al., 2008). We found that T transcription (brachyury, a marker of mesoderm) was transiently upregulated in our cultures, peaking at day 3 (∼10-fold compared with undifferentiated ESCs), before returning to undifferentiated levels at day 5 (Fig. S4C). Transcription of T, as well as another mesoderm marker, TBX6, could be promoted by the presence of FGF2, consistent with the role of FGF signaling in mesoderm induction (Fig. 7C; data not shown) (Vallier et al., 2009). Despite detectable T transcription, T⁺ cells were not detected in our differentiating cultures, whereas CHIR plus FGF2 treatment readily induced T⁺ cell clusters (∼5% of the cell clusters; Fig. 7D). There was also no evidence of robust transcriptional upregulation of other mesoderm precursor or progenitor markers during CHIR-induced differentiation (e.g. MIXL1, MESP1, cardiac-specific NKX2-5) (Fig. S4D; data not shown). The induction of T⁺ cells upon FGF treatment supports the idea that mesoderm induction in our system requires exogenous FGF signals.

In chick embryos, the earliest NC marker identified with a restricted expression at the border of the forming neural plate is PAX7 (Basch et al., 2006). Our unpublished results in rabbit and mouse embryos support Pax7 association with early NC development in mammals. We performed co-labeling of PAX7 and OCT4 to assess the transition of hESCs to PAX7⁺ NC progenitors and found PAX7⁺OCT4⁺ and PAX7⁺OCT4⁻ cell populations in the day 3 cultures (Fig. 8A, yellow arrows and white arrowheads, respectively). Interestingly, the majority of cells expressing a high level of PAX7 displayed no or very low OCT4. We proposed that during the WNT-induction of human NC cells, differentiating hESC might transit from OCT4⁺ ESCs to OCT4^lowPAX7^low cells, before acquiring an OCT4⁻PAX7⁺ status associated with a NC progenitor stage.

As a first step aiming to identify earlier events during NC induction that might precede the OCT4-PAX7 transition, we focused on genes functionally involved in early NC induction in animal models. Gain- and loss-of-function experiments in frog embryos reveal early functions of Gbx2 and Sp5 in the NC induction cascade and in the regulation of neural plate border gene expression (Li et al., 2009; Park et al., 2013). Gbx2^−/− mice display a reduction of post-otic NC cells and a NC migration patterning defect (Byrd and Meyers, 2005). We found that both were specifically induced by CHIR (Fig. 8B,C). Their expression peaked at differentiation day 2 and day 3, respectively. Overexpression of Zic3 induces ectopic NC tissues in amphibian embryos (Nakata et al., 1997) and is suggested to be an early NC specifier in teleost embryos (Garnett et al., 2012). We found that ZIC3 was transiently induced during differentiation either in the presence or absence of CHIR (Fig. 8D). CHIR, however, promoted a further increase/maintenance of ZIC3, which peaked at differentiation day 2. ZEB2, a causative gene in human Hirschsprung's disease associated with failure to form NC-derived enteric ganglia (Wakamatsu et al., 2001) that is expressed very early in the developing mouse neural fold and NC (Miyoshi et al., 2006; Van de Putte et al., 2003), was specifically induced by CHIR (Fig. 8E). Similar to GBX2, SP5 and ZIC3, ZEB2 expression peaked between differentiation day 2 and day 3. We consequently dubbed these genes as pre-neural plate border or simply ‘pre-border’ (pB) genes because their induction and peak of expression preceded the classical neural plate border genes such as PAX7, PAX3, MSX1 and TFAP2A (Fig. 1G; Fig. 3D).

Because pB genes and ‘neural border’ (NB) genes were induced with different temporal kinetics, we wished to test whether these two gene groups were differentially regulated. We therefore examined the response of pB (GBX2, SP5, ZIC3 and ZEB2) and NB (PAX3, PAX7, MSX1 and TFAP2A) genes to alterations of WNT, BMP and FGF signaling. Knocking down CTNNB1 by shRNA consistently inhibited expression of all pB and NB genes, except TFAP2A (which was also expressed highly in control non-neural ectoderm cultures) (Fig. 8F), suggesting that most, if not all, pB and NB genes were regulated by WNT/CTNNB1 signaling. Intriguingly, blocking BMP signaling with noggin did not affect expression of pB genes, but significantly repressed all NB genes (Fig. 8G). Addition of FGF2 ligand also inhibited expression of all NB genes, but had little effect on pB genes except GBX2 (Fig. 8H). In conclusion, we propose that human NC induction undergoes a stage-wise progression following a sequence of pB, NB and mature NC stages, with pB genes mainly regulated by canonical WNT signaling and partially by FGF signaling but not by BMP signaling (Fig. 8I).

hESCs resemble mouse epiblast stem cells in terms of signaling requirements and gene regulatory network for self-renewal and maintenance. The exact embryonic equivalent of ESCs in humans, however, is still under debate. hESCs appear to display low but heterogeneous endogenous WNT activity (Blauwkamp et al., 2012; Davidson et al., 2012). The absolute requirement of WNT activity in hESC renewal is contested but it appears that nuclear β-catenin signaling is dispensable for ESC maintenance (Kim et al., 2013) and WNT activation leads to differentiation of hESCs (Davidson et al., 2012; Dravid et al., 2005). Combining various chemically defined conditions, cell plating strategies, and activation of signaling pathways, such as activin, Nodal and FGF, WNT activation via GSK3 inhibition or constitutively active β-catenin can direct the differentiation of hESCs into a number of embryonic cell types, including primitive streak/mesendoderm (Funa et al., 2015; Gouti et al., 2014; Nakanishi et al., 2009; Sumi et al., 2008; Tsakiridis et al., 2015), paraxial mesoderm (Shelton et al., 2014; Tan et al., 2013; Umeda et al., 2012), lateral plate mesoderm (Tan et al., 2013), caudal neural plate, and axial stem cell progenitors (Denham et al., 2015; Tsakiridis et al., 2015).

Here, we report a robust and efficient method of inducing NC cells from hESCs. Our approach is based on (1) a precise seeding density of dissociated hESCs, (2) a defined medium composition, and (3) the small molecule GSK inhibitor CHIR 99021 to activate WNT signaling. Through this protocol, hESCs sequentially lose stemness/pluripotency markers and acquired characteristics of NC cells. We found that markers associated with non-neural ectoderm, placodal and epidermal tissues arise in the absence of WNT induction, as suggested by a previous study (Leung et al., 2013). WNT activation instead represses the non-neural ectoderm program and promotes NC induction. Using knockdown assays, we demonstrated that β-catenin is responsible for the activation of a set of early pB genes largely independently of BMP or FGF signaling. Interestingly, during in vitro hESC differentiation into primitive streak and NC precursors, β-catenin binding was observed in the genomic regions close to, or within the loci of, all four pre-border genes GBX2, SP5, ZEB2 and ZIC3 (Funa et al., 2015). Our findings are also in line with our previous work in the amniote avian embryo, which develops from a flat blastodisc in a similar manner to humans and most mammals. Through fate-mapping and specification assays, we proposed that the earliest NC populations arise independently of neural and mesodermal tissues, from precursors that precede the standard neural plate border characteristics (Basch et al., 2006).

The ‘binary competence’ model of neural crest/cranial placode induction proposed a segregation of NC and cranial placode competence at the neural plate border stage (Schlosser, 2008). Contrary to this model, we found that WNT directs a segregation of early NC and cranial placode cell fates prior to the induction of the neural plate border markers; pB genes such as GBX2, ZEB2, SP5 and ZIC3 are induced by WNT/CTNNB1 signaling 1 or 2 days prior to induction of the NB genes PAX3, PAX7 and MSX1. Moreover, expression of neural border markers common to both NC and cranial placodes, such as TFAP2A and ZIC1, displayed distinct kinetics during the course of human NC and cranial placode induction (Fig. 3D,E; Leung et al., 2013) further supporting the idea that segregation of NC and cranial placode fates occurs very early in our cultures. It has been suggested that NC competence segregates with the neuralized domain of the ectoderm (Pieper et al., 2012; Schlosser, 2008). We have examined the expression dynamics of a number of neural-related transcripts, including LHX2, PAX6, SOX1, SOX2, OTX2 and HES5 (Greber et al., 2011; Hou et al., 2013; Zhang et al., 2010). Variable and mild (∼10-fold or less) increases in expression were observed in only a subset of these genes (PAX6, SOX1, SOX2, OTX2) and others were not induced at all (LHX2 and HES5). Time course antibody staining for PAX6 also revealed minimal induction of PAX6⁺ cells and, most importantly, we did not observe any correlation in time and space for the induction of PAX6⁺ cells and SOX10⁺ NC cells. Furthermore, knockdown assays also suggested that reduced PAX6 function did not negatively interfere with NC induction. An early origin of human NC cells independent of neural progenitors has been suspected (Chambers et al., 2013), and our embryological evidence has also suggested the capacity of non-neural ectoderm to form NC without acquiring a neural character (Yardley and Garcia-Castro, 2012). Our data substantiate this notion by showing that human NC cells form independently of PAX6⁺ neuroectoderm.

Neural induction in hESCs requires inhibition of either BMP or activin, or both (Chambers et al., 2009; Pera et al., 2004; Smith et al., 2008; Surmacz et al., 2012). The fact that administration of noggin blocked expression of NC and/or neural border markers suggests that human NC induction requires BMP activation, which is distinct from neural induction. The effect of noggin on NC induction appears to be delayed as treatment of noggin from day 0 to day 1 did not translate into repression of pB genes as only NB genes were blocked. It is also worth noting that transcripts for BMP ligands including BMP2, BMP4 and BMP7, can be readily detected under CHIR treatment (data not shown), supporting the suggested contribution of this pathway during later events in NC development (Faure et al., 2002; Liem et al., 1997,, 1995; Selleck and Bronner-Fraser, 1995; Stuhlmiller and Garcia-Castro, 2012b).

Alternative neural induction pathways have been proposed. For instance, induction of human neural tube cells posterior to the forebrain has been shown to require endogenous WNT ligand activity (Lupo et al., 2013). Also, retinoic acid/FGF activity can posteriorize human neural progenitors, consistent with the roles proposed for these signaling pathways in animal models. We observed robust induction of engrailed 1 (EN1), but not caudal HOX genes, suggesting that our cultures might have induced cells equivalent to those that reside at the mid/hindbrain level (data not shown). Pre-migratory NC and dorsal neural tube cells share expression of a number of genes, including En1, Bmp4, Bmp7, Wnt1, Wnt3, Mafb and Lmx1a (Krispin et al., 2010). However, as discussed above, our cultures did not display definitive features of neural precursors. Furthermore, we obtained high yields of SOX10⁺PAX7⁺TFAP2A⁺ triple-labeled cells. These observations argue against a co-induction of a substantial percentage of dorsal neural tube progenitors in our cultures.

Our study established a rapid (5 days) and efficient protocol for generating a high yield of NC cells. Other studies have relied on surface markers such as HNK1 (human natural killer-1) and p75NTR (NGFR) (Fukuta et al., 2014; Jiang et al., 2009; Lee et al., 2010,, 2007; Menendez et al., 2011; Umeda et al., 2012), that are respectively known to label only few human NC cells or to have robust expression in non-NC cell populations (Betters et al., 2010). We instead relied on PAX7, TFAP2A and SOX10, which have been identified as robust markers co-expressed in a specific fashion in embryonic human NC (Betters et al., 2010). Our protocol simplifies previous approaches by eliminating the use of the BMP inhibitor noggin and/or the activin receptor inhibitor SB 431542. By modulating the initial seeding density thereby allowing efficient induction in the absence of a pre-differentiation period (Chambers et al., 2013; Fukuta et al., 2014; Menendez et al., 2011; Mica et al., 2013; Umeda et al., 2012), we attempt to limit the possible effects of paracrine signals of unknown nature that are produced by neighboring cells, and provide a sensitive model responsive to engineered cues for development. Simultaneously, we significantly shorten the total length of time required for the induction protocol [from ∼9 days with the current shortest protocol (Fukuta et al., 2014) to 5 days with our protocol]. Our newly devised protocol appears to be a useful supplement to the current induction protocols in generating human NC cells, and will certainly facilitate future studies on human NC development.

WA01 (WiCell), WA09 (WiCell), Y6 iPSC (Yale Stem Cell Center) and RIV9 iPSCs (University of California Riverside) of passages 30 to 70 were maintained on plastic surfaces coated with Matrigel (08-774-552, Fisher Scientific) in serum-free medium (mTeSR1, STEMCELL Technologies). Cultures were passaged every 6 to 7 days using Dispase (STEMCELL Technologies) according to the manufacturer's instructions.

Human ESC/iPSC colonies digested with Accutase cell detachment solution (Innovative Cell Technologies) were re-suspended in differentiation medium plus CHIR 99021 (3 µM; #4423 Tocris Bioscience) plated at the indicated densities on Matrigel-coated surfaces. Differentiation medium contained DMEM/F12 medium supplemented with 2% B27 supplement (Life Technologies), 1× Glutamax (Life Technologies) and 0.5% bovine serum albumin (wt/vol) (A9647, Sigma-Aldrich). The ROCK inhibitor Y-27632 (10 µM; Tocris Bioscience) was added for the first two days. Noggin (300-500 ng/ml; R&D systems), BMP4 (0.001-20 ng/ml; R&D Systems), SB 431542 (20 µM; Tocris Bioscience), FGF2 (5-20 ng/ml; R&D Systems), PD 173074 (200 nM; Tocris) and PD 184352 (5 µM; Tocris) were added as indicated. Small molecules and medium were replenished daily.

Day 5 CHIR-treated cells were plated at 20,000-80,000 cells/cm² on a Matrigel-coated surface with NC induction medium, supplemented with 10 µM Y-27632 overnight. Medium was changed by 24 h post-plating to terminal differentiation media. For peripheral neurons, melanoblasts and glial progenitors, differentiation medium contained DMEM/F12 with 2% B27 supplement, 1× Glutamax, 3 µM CHIR 99021, 1 µM SU-5402 (Tocris) and 2.5 µM N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester (DAPT; Calbiochem, 565784; 10 μM). For osteoblast differentiation, medium contained 50 mg/ml ascorbic acid, 10 mM β-glycerophosphate and 50 nM 1α,25-OH vitamin D3 (zur Nieden et al., 2003). Media were changed every 3-4 days. Cultures were harvested at the indicated time and processed for Alizarin Red staining. For chondrocyte differentiation, dissociated day 5 cells were pelleted at 1200 rpm (280 g) at a concentration of 1×10⁶ cells/ml and incubated in 1 ml of MesenCult-ACF Chondrogenic Differentiation Medium (05455; Stem Cell Technologies). Cells were fed every 3-5 days. At day 21, pellets were fixed with 10% formalin or 4% paraformaldehyde, paraffin sectioned and processed for Alcian Blue staining.

Total RNAs were extracted using TRIzol reagent (Life Technologies). Total RNA (0.5-1 µg) was reverse-transcribed using SMART MMLV reverse transcriptase (Clontech Laboratories). Real-time polymerase chain reaction (PCR) was carried out using the iTaq universal SYBR Green supermix (Bio-Rad) and the IQ5 Real-Time PCR Detection System (Bio-Rad). The primer concentration was 160 nM. Forty cycles of reactions were performed followed by melt curves to confirm specificity of primer sets. Three to six biological replicates were measured. Primer sequences are listed in Table S1.

Human shRNA sequences for CTNNB1 and PAX6 were designed using a set of rules that maximize knockdown efficiency and minimize off-targets (http://www.broadinstitute.org/rnai/public/resources/rules). These sequences were retrieved from the RNAi consortium database (http://www.broadinstitute.org/rnai/public/). Target sequences tested are listed in Table S2. Annealed shRNA oligos with the loop sequence TTCAAGAGA are cloned between the AgeI and EcoRI sites of pLKO.1 puro (a gift from Bob Weinberg, plasmid #8453, Addgene) (Stewart et al., 2003).

For lentivirus generation, the helper constructs psPAX2 and pMD2.G (both a gift from Didier Trono, plasmids #12260 and #12259, respectively, Addgene) together with the transfer vector were transfected into the 293T/17 cell line (CRL-11268, ATCC) using the calcium phosphate precipitation method. Viruses were partially purified using Lenti-X concentrator (Clontech Laboratories). For lentivirus infection, WA01 ESCs, plated at 20,000 cells/cm² were incubated in mTeSR1 medium (STEMCELL Technologies) containing titrated amount of concentrated lentivirus and polybrene (6 µg/ml) for 4-8 h. Antibiotic selection began at 48 h post infection with puromycin (Tocris, 4089) at 1 µg/ml.

Immunofluorescence was performed on cells cultured on Nunc Lab-Tek Chamber Slide System (Thermo Scientific, # 177399). Cultures were fixed (4% paraformaldehyde, 12 min), permeabilized (0.4% Tween-20, 12 min) and blocked (10% fetal bovine serum, 1 h) at room temperature. Primary antibody and secondary antibody incubation were performed in 4% fetal bovine serum overnight at 4°C and 1 h at room temperature, respectively. Primary antibodies are listed in Table S3. Secondary antibodies were species- or subtype-specific Alexa secondary antibodies (Life Technologies). Images were taken using a Nikon Eclipse 80i microscope and a SPOT cooled camera or a Zeiss M1 imager. Images were processed in Adobe Photoshop CC version 14.2.1. and were adjusted simultaneously in exactly the same manner for experimental and control sets.

Cells were lysed using RIPA lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM EDTA, 0.5% sodium deoxycholate, 1% Igepal, 0.1% SDS), plus protease inhibitors (Roche Diagnostics), followed by sonification and centrifugation to remove cell debris. Protein quantification was performed using the Quick Start Bradford Protein Assay (Bio-Rad). Ten-µg protein samples in Lammeli buffer with DTT were separated on a 12% SDS-PAGE gel and transferred to PVDF membrane (Bio-Rad). Membrane was blocked with 5% milk and incubated with the primary antibodies mouse monoclonal anti-β-catenin (610153, BD Biosciences; 1:2000), mouse IgG1 anti-PAX6 (Developmental Studies Hybridoma Bank; 1:500) or mouse monoclonal anti-β-actin (A5316, Sigma-Aldrich; 1:10,000) for 1 h at room temperature. Membrane was then incubated with a horseradish peroxidase-conjugated secondary antibody goat anti-mouse IgG (H+L) (Thermo Fisher Scientific, 31432). Signal was developed using the Clarity Western ECL Kit (Bio-Rad) and detected using x-ray films.

Chart drawing and statistical analyses were performed using the GraphPad software Prism (version 6.0f). Asterisks indicate statistical significance: *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001. Unpaired t-tests and one-way ANOVA tests were respectively performed for comparisons of two or more means.

We thank Dr zur Nieden and Nicole Sparks (UCR) for osteogenic culture assistance, Yale undergraduate helpers Laura Miyares, Cayla Broton and Kevin Kwang for assistance in cell counting and real-time PCR analyses, and the Valerie Horsley laboratory for the maintenance of the stem cell facility. WA01 and WA09 cell lines were obtained from the WiCell Institute through the Yale Stem Cell Center. The monoclonal antibodies against TFAP2A (3B5), PAX6 and PAX7 were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development and maintained by the University of Iowa, Department of Biology, Iowa City, IA 52242, USA.