Neural crest cells migrate throughout the embryo and differentiate into diverse derivatives: the peripheral neurons, cranial mesenchymal cells, and melanocytes. Because the neural crest cells have critical roles in organogenesis, detailed elucidation of neural crest cell differentiation is important in developmental biology. We recently reported that melanocytes could be induced from mouse ESCs. Here, we improved the culture system and showed the existence of neural crest-like precursors. The addition of retinoic acid to the culture medium reduced the hematopoiesis and promoted the expression of the neural crest marker genes. The colonies formed contained neural crest cell derivatives: neurons and glial cells, together with melanocytes. This suggested that neural crest-like cells assuming multiple cell fates had been generated in these present cultures. To isolate the neural crest-like cells, we analyzed the expression of c-Kit, a cell-surface protein expressed in the early stage of neural crest cells in vivo. The c-Kit-positive (c-Kit+) cells appeared as early as day 9 of the culture period and expressed the transcriptional factors Sox10 and Snail, which are expressed in neural crest cells. When the c-Kit+ cells were separated from the cultures and recultured, they frequently formed colonies containing neurons, glial cells, and melanocytes. Even a single c-Kit+ cell formed colonies that contained these three cell types, confirming their multipotential cell fate. The c-Kit+ cells were also capable of migrating along neural crest migratory pathways in vivo. These results indicate that the c-Kit+ cells isolated from melanocyte-differentiating cultures of ESCs are closely related to neural crest cells.
Neural crest cells are highly migratory multipotent cells that give rise to diverse derivatives, such as the peripheral nervous system, skin melanocytes, and the cranial mesenchymal cells. These cell types include peripheral neurons and glial cells, melanocytes, endocrine cells, smooth muscle, and bone . Neural crest cells emerge from the dorsal region of the fusing neural tube and migrate throughout the embryo. During their migration or once the target tissue has been reached, neural crest cells respond to various environmental factors, such as bone morphogenetic proteins (BMPs), epidermal growth factors , and Wnts , and differentiate into a variety of cells. In the trunk, neural crest cells migrate along three major routes, namely dorsolateral, ventrolateral, and ventromedial pathways, and differentiate into melanocytes, autonomic ganglia (sympathetic ganglia and parasympathetic ganglia), dorsal root ganglia (DRG), and glial cells . Because of this multidifferentiation potency of neural crest cells, disorder of or deficiency in neural crest cell development results in severe developmental anomalies, such as Waardenburg's syndrome (deafness with pigmentary abnormalities), Hirschsprung's disease (aganglionic megacolon) , and von Recklinghausen's disease .
The isolation of multipotent neural crest cells from various tissues has been reported. For example, Stemple and Anderson  described the isolation of multipotent neural crest stem cells (NCSCs) from explant cultures of embryonic day 10.5 (E10.5) rat neural tubes, Morrison et al.  isolated NCSCs from uncultured E14.5 fetal rat sciatic nerve, and Kruger et al.  discovered that NCSCs persist in the adult gut and isolated those cells. More recently, Fernandes et al.  isolated adult dermal precursors that exhibited properties similar to those of embryonic neural crest stem cells. Although these recent advances in the isolation of neural crest cells have enabled us to analyze their character, the generation and purification of neural crest cells from embryonic stem (ES) cells may be expected to lead to a better understanding of these cells. ESCs are an unlimited source of derivatives including neural crest cells, and their use should enable even biochemical analysis of neural crest cells. It is also a potential benefit that ESCs may be genetically manipulated to permit experiments to clarify the function of the specific genes involved in neural crest development.
Recent studies have demonstrated the differentiation of mouse ESCs into neuroectoderm in culture and the emergence of neural crest cells from this neuroectodermal population [10, 11]. These neural crest cells differentiated into sensory neurons, autonomic neurons , smooth muscle cells , and glial cells [10, 11] but not into melanocytes, another main neural crest derivative. More recently, sensory and sympathetic neurons were induced from human ESCs, and these neurons were developed from ESC-derived neural crest cells . Still, these studies did not identify neural crest cells themselves, which might possibly have emerged in the cultures.
We recently reported that melanocytes were induced from mouse ESCs when such cells were cocultured on monolayers of the bone marrow-derived stromal cell line ST2 . When ESCs were cocultured with ST2 cells, melanocytes differentiated from the ESCs after 21 days in culture. This melanocyte-inducing culture was considered to mimic the development of cutaneous melanocytes in vivo, since the ESC-derived melanocytes depended on c-Kit signals for their differentiation . As cutaneous melanocytes are derived only from neural crest cells, neural crest cells might have been generated in these melanocyte-inducing cultures. In this present study, we improved the ESC culture system to induce melanocytes as exclusively as possible, and using it, we identified and isolated neural crest-like cells that could differentiate into melanocytes, neurons, and glial cells.
Materials and Methods
Mouse D3 ESCs, green fluorescent protein (GFP)-D3 ESCs , SCL/tal-1 knockout J1 ESCs , and ST2 cells were maintained as previously described . For the induction of melanocytes and neural crest-like cells, 500 to 1,500 ESCs were inoculated into six-well plates (Greiner Bio-One, Tokyo, http://www.greiner-bio-one.co.jp) previously seeded with ST2 stromal cells and containing α-minimal essential medium (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) supplemented with 10% fetal calf serum, 10−7 M dexamethasone (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com), 20 pM fibroblast growth factor-2 (R&D Systems Inc., Minneapolis, http://www.rndsystems.com), 10 pM cholera toxin (Sigma-Aldrich), and 100 ng/ml human recombinant endothelin-3 (Peptide Institute, Inc., Osaka, Japan, http://www.peptide.co.jp) in 5% CO2 at 37°C. Usually, 100 nM all-trans retinoic acid (RA) (Sigma-Aldrich) was also present in the medium from day 0 to day 9 of the culture period. The medium was changed every 2 days, and the day when ESCs were seeded onto ST2 monolayers was defined as day 0.
Flow-Cytometric Analysis and Cell Sorting
Day 9–12 cultured ESCs were dissociated by incubating them for 6 minutes at 37°C in 1× Dispase II (diluted 1:10 in calcium- and magnesium-free phosphate-buffered saline [PBS]; Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com). The digestion was quenched with two volumes of staining medium (SM) (PBS containing 3% fetal calf serum). After centrifugation, the cells were triturated and resuspended in SM. The dissociated ESCs were then washed with SM, and blocked with rat anti-mouse Fc-γ receptor (2.4-G2; BD Biosciences, San Diego, http://www.bdbiosciences.com) on ice for 30–40 minutes. After another wash with SM, the cells stained with fluorescein isothiocyanate (FITC)-conjugated rat anti-mouse CD45 (30-F11; BD Biosciences) on ice, washed, and incubated with allophycocyanin-conjugated rat anti-mouse c-Kit (2B8; BD Biosciences). For some experiments, the cells were incubated with phycoerythrin-conjugated rat anti-mouse c-Kit (ACK2; a gift from Dr. Sin-ichi Nishikawa, Riken, Kobe, Japan). For in ovo transplantation experiments, the cells were incubated with phycoerythrin-conjugated rat anti-mouse CD45 (30-F11; eBiosciences, San Diego, CA, http://www.ebioscience.com) and allophycocyanin-conjugated rat anti-mouse c-Kit (2B8; BD Biosciences). The cells were washed and resuspended in SM containing 3 μg/ml propidium iodide (PI) (Calbiochem, La Jolla, CA, http://www.calbiochem.com). All cell sorting and analysis were performed with a FACSVantage dual-laser flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com). The sorted cells were directly inoculated into six- or 96-well plates previously seeded with ST2 stromal cells by using the CloneCyt Plus version 3.1 system (Becton Dickinson).
For analysis of neurons by flow cytometry, cultured ESCs were dissociated by incubating them for 5 minutes at 37°C in 0.25% trypsin (Invitrogen) and then fixed in 4% paraformaldehyde (PFA) in PBS for 15 minutes and made permeable by treatment with 0.1% Triton X-100 (Sigma-Aldrich) in 5% bovine serum albumin (BSA) PBS for 30 minutes. Thereafter, the cells were washed with SM and stained with mouse monoclonal antibody against mouse neuronal class III β-tubulin (TuJ-1) (BabCO, Berkeley, CA, http://store-crpinc.com) on ice, washed, and reacted with FITC-conjugated anti-mouse IgG (BioSource International, Camarillo, CA, http://www.biosource.com). Analyses were performed by using the FACSVantage.
Target cells were isolated from cultures at various times by using the FACSVantage. The total RNA was purified by using Isogen (Nippon Gene, Tokyo, http://www.nippongene.com), and first-strand cDNA synthesis was carried out with Superscript II reverse transcriptase (Invitrogen) primed with random hexamer in a 20-μl reaction mixture containing 1 μg of total RNA. A total of 0.5 μl of the first-strand cDNA mixture was used for polymerase chain reaction (PCR) conducted with recombinant Taq polymerase (Takara, Otsu, Japan, http://www.takara.co.jp) in a 50-μl volume. PCRs were performed as follows: 94°C for 2 minutes; 35 to 40 cycles of 94°C for 30 seconds; gene-specific annealing temperature for 30 seconds; and 72°C for 60 seconds. Five microliters of each PCR product was electrophoresed on an agarose gel and stained with ethidium bromide. The primers used for PCR were as follows: Snail, 5′-ATAGCGAGCTGCAGGACGCGTGTGT-3′ (forward) and 5′-AGGCCGAGGTGGACGAGAAGGACGA-3′ (reverse); Slug, 5′-AAGGCTTTCTCCAGACCCTGGCTG-3′ (forward) and 5′-CAGCCAGACTCCTCATGTTTATGC-3′ (reverse); Sox10, 5′-TTCAGGCTCACTACAAGAGTG-3′ (forward) and 5′-TCAGAGATGGCAGTGTAGAGG-3′ (reverse); Pax3, 5′-ATGGTTGCGTCTCTAAGATCCTG-3′ (forward) and 5′-GCGTCCTTGAGCAATTTGTC-3′ (reverse); Mitf-M, 5′-GCTGGAAATGCTAGAATACAG-3′ (forward) and 5′-TTCCAGGCTGATGATGTCATC-3′ (reverse); Oct4, 5′-GGCGTTCTCTTTGGAAAGGTGTTC-3′ (forward) and 5′-CTCGAACCACATCCTTCTCT-3′ (reverse); and glyceraldehyde-3-phosphate dehydrogenase, 5′-CTTCACCACCATGGAGAAGGC-3′ (forward) and 5′-GGCATGGACTGTGGTCATGAG-3′ (reverse).
The colonies were sequentially fixed in 4% PFA in PBS for 15 minutes, made permeable by immersion in 0.1% Triton X-100 in 5% BSA PBS for 30 minutes, washed in PBS, and blocked in a mixture of 3% goat serum and 5% BSA in PBS for 30 minutes. Primary antibodies, diluted in 5% BSA PBS, were then added and allowed to react for 30 minutes at room temperature. The primary antibodies used for immunohistochemistry were the following: anti-mouse neuronal class III β-tubulin (1:500; TuJ-1, BabCO), anti-mouse glial fibrillary acidic protein (GFAP) (1:500, Z0334; DakoCytomation, Glostrup, Denmark, http://www.dakocytomation.com), anti-mouse tyrosine hydroxylase (1:500; AB152; Chemicon, Temecula, CA, http://www.chemicon.com), and anti-mouse α smooth muscle actin (1:500, 1A4; Sigma-Aldrich). After having been washed in PBS, the cells were stained with the secondary antibodies in the same manner. The following second antibodies were used: FITC-conjugated anti-mouse IgG (1:500; Biosource), rhodamine-conjugated anti-mouse IgG (1:500; Jackson Immunoresearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com), FITC-conjugated anti-rabbit IgG (1:500; Medical Biological Laboratories Co., Ltd., Nagoya, Japan, http://www.mbl.co.jp), and rhodamine-conjugated anti-rabbit IgG (1:500; Chemicon). Colonies were examined by using an Olympus IX-71 fluorescein microscope (Tokyo, http://www.olympus-global.com).
In Ovo Transplantation
Fertile white leghorn chick eggs were incubated at 37°C in a humidified chamber until Hamburger and Hamilton stage 17–18. A small opening was made in the side of the shell, and Hanks' solution (Sigma-Aldrich) containing 10% india ink was injected under the blastoderm to visualize the embryo. c-Kit+/CD45− cells isolated by fluorescence-activated cell sorting from day 12 cultured GFP-ESCs were inoculated onto ST2 monolayers; 1 week later, the colonies that had formed were collected for transplantation. By use of capillary tubes with very gentle air pressure, the colonies were injected into the anterior, medial corner of one or two somites of each embryo, corresponding to the dorsal-most region of the neural-crest migratory pathway. The manipulated embryos were then incubated for an additional 3 days to stage 29–30 or 4 days to stage 34, fixed in 3.7% formaldehyde/PBS for 2 hours, and immersed in 30% sucrose/PBS overnight at 4°C. The embryos were embedded in OCT compound (Sakura Finetechnical, Tokyo, http://www.sakura-finetek.com/top.html), frozen at −70°C, sectioned at a 20-μm thickness, and placed on tissue-adhering slides. After having been washed in PBS, the slides were stained with TuJ-1 (1:500; Babco) and rhodamine-conjugated anti-mouse IgG (1:500; Jackson Immunoresearch Laboratories) as described above. The slides were examined with the Olympus IX-71 fluorescein microscope.
Melanocytes Induced from ESCs Were Generated from Neural Crest-Like Cells
Since melanocytes are generated from trunk neural crest cells , “caudalization” even in a culture system may cause effective induction of trunk neural crest cells and the consequent emergence of more melanocytes in the cultures. RA is often used as a caudalization factor in in vitro embryo cultures [15, 16], and it promotes the differentiation of neural crest-derived cells, such as melanocytes and adrenergic neurons, in quail neural crest cultures . Therefore, we evaluated the effect of RA on our cultures in which melanocytes differentiated from ESCs. The number of colonies generated in each well was slightly reduced in 10, 100, and 1,000 nM RA-containing cultures (Fig. 1A). However, the fraction of colonies containing melanocytes increased in an RA concentration-dependent manner up to 100 nM (Fig. 1B). At the highest concentration tested, 1,000 nM, the differentiation of melanocytes was totally inhibited. Also, at this concentration, the number of CD45-positive hematopoietic lineage cells, usually generated under the same conditions without RA, was reduced in number, as shown in Figure 1C. At each RA concentration, the number of melanocytes generated in each well was nearly the same, 4 × 105 per well (data not shown). We concluded from this experiment that we could more frequently and stably recapitulate the key event necessary for melanocyte induction by using 100 nM RA.
Since it was expected that this key event in these cultures was the induction of neural crest cells, we evaluated the expression of neural crest marker genes by reverse transcription (RT)-PCR. To exclude ST2 stromal cells used as the supporting cells in this culture, we started the cultures with GFP-ESCs  and purified GFP-ESC derivatives by flow cytometry. As shown in Figure 1D, Snail, Slug , and Pax3  expression continued from day 7 of the culture period; in the case of Snail, its expression was detected in undifferentiated ESCs. Considering that cell lineages other than melanocytes might also be generated in the present culture system , the relatively strong expression of these markers in them may reflect the expression pattern of these markers in the embryos. In addition to expression in the neural crest regions, Pax3 was detected in the somites after E8.5 . Snail and Slug were also detected in the primitive streak of E7.5 embryos or in the early mesoderm at E9.5 . Another neural crest marker, Sox10, was reported to be exclusively expressed in the neural crest regions at E8.5 [21, 22]. In accordance with this observation, we detected weak expression of Sox10 during days 8–9 of the culture period after 40 cycles of PCR; thereafter, its expression gradually increased. Slightly after Sox10 expression, the melanocyte lineage-specific marker Mitf-M became detectable and then increased gradually. These results suggest the possible induction of neural crest-like cells in light of the emergence of melanocyte lineage cells in the present culture system. Although the intensities of most of the individual markers were increased by the addition of RA, the overall expression patterns of these markers were similar, irrespective of RA (Fig. 1D).
To examine whether neural crest-derived cells other than melanocytes were generated in this culture, we immunostained the colonies that emerged with neuronal marker anti-β-tubulin (TuJ-1) or glial cell marker anti-GFAP after 21 days. TuJ-1-positive neurons expressing GFP were detected (Fig. 1E, i and ii); within the same colony, melanocytes were also found (Fig. 1E, iii). In addition, a number of colonies containing GFAP-positive glial cells were detected, as shown in Figure 1E, iv and v, and some of them also contained melanocytes (Fig. 1E, vi). These results imply the presence of neural crest-like cells capable of differentiating into neurons or glial cells or simultaneously into melanocytes in the present culture system.
Isolation of ESC-Derived Neural Crest-Like Cells by Flow Cytometry and Their Differentiation into Multiple Cell Lineages
To further characterize the neural crest-like cells induced from ESCs in this culture system, we took advantage of flow-cytometric purification of the target cells from the cultures. Among the various cell-surface proteins known to be expressed in the trunk neural crest, c-Kit receptor-type tyrosine kinase is expressed at the early stage of differentiation of melanocyte precursors from the trunk neural crest cells [23, 24]. Thus, we first evaluated c-Kit-positive cells emerged in this culture by flow cytometry. To distinguish them from ST2 cells, we used the forward and side scatter patterns of ST2, in which all of the ST2 cells were distributed in the population with higher side scatter values (Fig. 2A, left, indicated as gate I). In contrast, ESC-derived cells, in this case expressing GFP, were mostly distributed in the population with lower side scatter values (Fig. 2A, left, indicated as gate II). As shown in Figure 2A, right, ST2 cells were GFP-negative and mostly came into gate I. Instead, most of the cells of gate II were ESC-derived GFP-expressing cells. Using this simple discrimination method together with the CD45 marker specific for hematopoietic cells, which are also known to express c-Kit , we analyzed the expression of c-Kit during the culture period. In previous reports, phycoerythrin-conjugated anti-c-Kit antibody (PE-ACK2) was used for the isolation of skin melanoblasts [26, 27]. When we used this antibody for the analysis of the c-Kit expression, few c-Kit-positive cells were detected (Fig. 2B, left). To increase the sensitivity for detection of the c-Kit molecule, we used allophycocyanin-conjugated anti-c-Kit antibody (APC-2B8). Using APC-2B8, we could detect 20 times more c-Kit-positive cells than we could using PE-ACK2 (Fig. 2B, right). As shown in Figure 2C, c-Kit-positive cells appeared as early as day 9 of the culture; thereafter, their number increased during days 11–12 and continued to increase until day 16. Considering the c-Kit expression pattern and RT-PCR data (Fig. 1D), the neural crest cells would appear to have emerged at days 9–12 of the culture period, when c-Kit-positive cells first began to appear.
To test whether these c-Kit-positive cells corresponded to the neural crest-like cells described in the previous sections, we sorted c-Kit-positive and CD45-negative (c-Kit+/CD45−) cells by flow cytometry and evaluated the expression of trunk neural crest marker genes Sox10, Pax3, and Snail by RT-PCR. In the c-Kit+/CD45− fraction, Sox10 and Snail mRNAs were expressed apparently at a higher level than in the c-Kit-negative and CD45-negative (c-Kit−/CD45−) fraction (Fig. 2D). Pax3 was expressed in the c-Kit−/CD45− fraction, and an equal amount of the mRNA was expressed in the c-Kit+/CD45− fraction as well (Fig. 2D). Melanocyte lineage-specific Mitf-M was highly expressed in the c-Kit+/CD45− fraction; Oct4, which is known to be expressed in undifferentiated ESCs , was hardly detected in either fraction.
We next investigated the potential of sorted c-Kit+/CD45− cells as precursors of melanocytes. One thousand c-Kit+/CD45− cells sorted from the above cultures were recultured on ST2 cells and analyzed for the emergence of melanocytes. After 21 days in culture, both c-Kit+/CD45− and c-Kit−/CD45− cells formed colonies; that is, both fractions contained colony-forming precursor cells, but the c-Kit+/CD45− cultures clearly produced more melanocytes than the c-Kit−/CD45− cultures (Fig. 3A, 3B). Apparently, a greater number of colonies composed of melanocytes were formed in the wells inoculated with c-Kit+/CD45− cells (Fig. 3A). When endothelin-3, which is a critical signaling molecule in melanocyte differentiation both in vivo and in vitro, was antagonized by the addition of BQ788, an antagonist for endothelins, melanocytes did not emerge in these cultures (data not shown). This finding indicates that these melanocytes were the in vitro counterpart of cutaneous melanocytes.
Neurons and glial cells are other developmental fates of the neural crest , so we measured the number of these cells in cultures of sorted c-Kit+/CD45− cells. TuJ-1-positive neurons, detected by flow cytometry, were much more numerous in the cultures of the c-Kit+/CD45− fraction than in the cultures of the c-Kit−/CD45− fraction (Fig. 3C). The colonies containing GFAP+ cells were almost equal in number in both c-Kit+/CD45− and c-Kit−/CD45− cultures (Fig. 3D). We next immunostained neurons derived from the c-Kit+/CD45− cells with Brn3a, a sensory neuron marker, or with tyrosine hydroxylase (TH), an autonomic neuron marker, after day 21 of the culture period. TH and TuJ-1 double-positive neurons were detected in the colonies of the c-Kit+/CD45− cultures (Fig. 3E), whereas no Brn3a-positive neurons were detected in these cultures (data not shown). These observations demonstrate that the c-Kit+/CD45− cells derived from ESCs included a population of cells of the autonomic neuronal lineage, which usually differentiate from neural crest cells in vivo. Although less frequently, the c-Kit+/CD45− fraction, but not the c-Kit−/CD45− fraction, produced α-smooth muscle actin-positive cells (Fig. 3F). Interestingly, a number of colonies containing TuJ-1+ neurons, GFAP+ glial cells, and pigmented melanocytes were produced only in c-Kit+/CD45− cultures (Fig. 3G). To rule out the possible contamination of the c-Kit+/CD45− fraction by undifferentiated ESCs, we estimated the alkaline phosphatase activity characteristic of undifferentiated ESCs but found no alkaline phosphatase-positive cells among 50,000 c-Kit+/CD45− cells sorted (data not shown). These results suggest that these c-Kit+/CD45− cells likely have multiple fates, differentiating into melanocytes, neurons, glial cells, and sometimes smooth muscle cells.
Proliferative activity is characteristic of neural crest cells . In fact, even in vitro neural crest cells migrate out from the dissected neural tube and continue to proliferate to form large colonies. To assess the proliferative ability of c-Kit+/CD45− cells and their descendants, we measured the size of the individual colonies containing pigmented melanocytes, as shown in Figure 3I. The size of a colony is thought to reflect the synergetic effect of cell proliferation. Collectively, in most cases, the colonies were larger in those derived from c-Kit+/CD45− cells than in those formed by c-Kit−/CD45− cells. For example, c-Kit+/CD45− cells sorted from day-12 cultures contained more than 10 times as many huge (>5,000 μm in diameter) colonies as those formed by c-Kit−/CD45− cells. Thus, highly proliferative cells were concentrated in the c-Kit+/CD45− population.
Transplanted adult bone marrow cells were reported to differentiate into cells expressing neuron-specific markers in the adult brain [29, 30]. To exclude any possibility that ESC-derived hematopoietic cells differentiated into neural crest derivatives, we cultured SCL/tal-1-targeted J1 ESCs, which are completely defective in hematopoiesis [31, 32], on ST2 stromal cells; we then sorted 100 c-Kit+ cells and transferred them onto ST2 monolayers at day 12 of culture. The c-Kit+/CD45− cells appeared as early as day 9 of the culture without expression of CD45 and, like those arising from wild-type D3 ESCs, generated colonies comprising TuJ-1-reactive neurons, GFAP-positive glial cells, and pigmented melanocytes (Fig. 3H). Thus, c-Kit+/CD45− cells generated in this culture system did not arise from cells of the hematopoietic cell lineage.
Single c-Kit+/CD45− Cells Sorted from ESC Cultures Give Rise to Neurons, Glial Cells, and Melanocytes
To analyze more precisely the multipotency of the c-Kit+/CD45− cells, we examined whether a single c-Kit+/CD45− cell could differentiate into multilineage cells in vitro. From culture day 9 to culture day 12, c-Kit+/CD45− cells isolated by flow cytometry were inoculated at one cell per well onto ST2 monolayers in 96-well plates, as illustrated Figure 4A; the colonies that appeared were identified immunocytochemically after 21 days in culture. As shown in Table 1, 22.4%–56.1% of the sorted single c-Kit+/CD45− cells formed colonies in three separate experiments. Of these colonies, up to 9.2% contained multilineage cells, including melanocytes, neurons, and glial cells (a representative colony is shown in Fig. 4B), indicating that these colonies were derived from multipotential precursors. c-Kit+/CD45− cells also formed colonies composed of two cell types (i.e., up to 20.7% contained melanocyte and neurons, up to 6.6% contained melanocytes and glial cells, and up to 3.9% contained neurons and glial cells [a representative colony is shown in Fig. 4C]), indicating that these cells were derived from bipotential precursors. Up to 24.4% of the c-Kit+/CD45− cells formed colonies comprising a single cell type, indicating their derivation from unipotential precursors. In this single-cell culture, neurons that were both TH-positive and TuJ-1-positive were detected in the colonies arising from the c-Kit+/CD45− fraction (Fig. 4D). Although c-Kit−/CD45− cells formed colonies, colonies with multilineage cells never emerged, and there were less bipotential or single-potential colonies (Table 1). Next, we examined the relationship between the c-Kit expression level and the differentiation ability of the cells. c-Kit-low-expressing CD45− and c-Kit-high-expressing CD45− cells were inoculated at one cell per well onto ST2 monolayers in 96-well plates, and the colonies that appeared were identified immunocytochemically. Both cells showed equal colony-forming ability and differentiation ability (data not shown). It is thus likely that c-Kit expression level may not indicate the substantial differences of these two populations, at least in our in vitro assay system.
Table Table 1.. Frequencies of the production of different types of colonies from single c-Kit-positive or c-Kit−/CD45− cells
This single cell sorting experiment clearly demonstrates the multipotential nature of the c-Kit+/CD45− cells that differentiated from ESCs in the present culture system. In view of their similarity with neural crest cells in terms of multipotential cell fate, especially considering the fact that no other cell lineages than neural crest differentiate into melanocytes, the c-Kit+/CD45− population purified in this study contained a concentrated subpopulation of cells sharing the characteristic multipotential cell fate of neural crest cells.
c-Kit+/CD45− Neural Crest-Like Cells Showed Extended Motility upon Transplantation In Ovo
To confirm whether the c-Kit-positive cells exhibited migrational potency similar to that of neural crest cells in vivo, we isolated the c-Kit+/CD45− cells from melanocyte-differentiating cultures of GFP-ESCs and transplanted them into the chick neural crest migratory stream in ovo at Hamburger and Hamilton stage 17–18  (Fig. 5A). By 3 days later (stage 30), the transplanted cells had migrated to peripheral neural crest target sites. Many transplanted cells had migrated around the DRG (Fig. 5B). Some GFP+ cells were detected near the skin (Fig. 5C). Four days later (stage 34), GFP+ cells surrounded the dorsal aorta (Fig. 5D) and digestive tract (Fig. 5E). Some of these cells were reactive with TuJ-1 (Fig. 5D, lower photo). Thus, these transplanted c-Kit+/CD45− cells migrated along neural crest migratory pathways into target sites of the trunk neural crest in a manner similar to neural crest cells in vivo.
In our previous study, melanocytes were generated from mouse ESCs by culturing them on monolayers of ST2 stromal cells . Because melanocytes are derived from neural crest cell in vivo, we surmised that neural crest cells must have been generated prior to melanocyte differentiation in this culture system. In our present study, after the induction of melanocytes from ESCs, immunocytochemistry showed that some colonies contained TuJ-1-positive cells (neurons) and/or GFAP-positive cells (glial cells) in addition to melanocytes. This finding suggested that neural crest cells, which differentiate into neurons, glial cells, and melanocytes in vivo, were generated in our culture system. To isolate the neural crest cells from the culture, we first analyzed the expression of c-Kit, a potential neural crest marker protein [23, 24], and separated c-Kit+/CD45− cells from the culture. These cells frequently formed colonies comprising neurons, glial cells, and melanocytes. Even when a single c-Kit-positive cell was cultured, it produced a colony containing all three of these cell types. In the in ovo transplantation experiment, the c-Kit+/CD45− cells migrated along neural crest migratory pathways. Thus, our present results demonstrate that the c-Kit+/CD45− cells that differentiated from ESCs in this culture system are closely related to neural crest cells in vivo, both having multidifferentiation potential.
A recent study by Wilson et al.  showed that c-Kit begins to be expressed on the dorsal midline of the trunk neural tube from E9, precisely in the premigratory neural crest region. They found that the c-Kit+/CD45− cells migrated only into the ectoderm and developed into melanocytes. Even in vitro, c-Kit-positive neural crest cells developed only into melanocytes [23, 33]. In contrast to these studies, we found that our c-Kit+/CD45− cells developed into neurons and glial cells as well as into melanocytes. This discrepancy may be based on the difference in methodology between previous studies and our present study. The former studies detected c-Kit+/CD45− cells by histological immunostaining for c-Kit or staining for β-gal activity from a lacZ gene inserted within a c-kit homologous recombination construct, in which only cells expressing relatively more c-kit molecules might have the chance to be detected. Using flow cytometry and allophycocyanin-conjugated anti-c-Kit antibody (APC-2B8), we were able to detect and sort c-Kit+/CD45− cells with low expression, expressing only a small amount of c-kit molecules, as is clearly shown in Figure 2B. These “low” c-Kit+/CD45− cells, which were hardly detectable by immunostaining with phycoerythrin-conjugated antibodies, might only have the potential to differentiate into multilineage cells. Because there were no differences in colony forming or differentiation ability between c-Kit-low-expressing CD45− cells and c-Kit-high-expressing ones (data not shown), the c-Kit-expressing cell population in our cultures contained neural crest-like cells irrespective of the c-Kit expression level. In addition, there is evidence that a small population of neural crest cells that had already migrated into the epidermis had both melanogenic and neurogenic potential in vitro . These epidermal neural crest cells might be the in vivo counterparts of the low c-Kit-positive neural crest-like cells reported in this article.
Some studies have demonstrated that transplanted bone marrow-derived cells can migrate into the brain and differentiate into cells expressing neuron-specific marker antigens in central nervous system [29, 30] and vice versa . Hematopoietic cells are also known to express c-Kit . To exclude the possibility that hematopoietic cells had differentiated from ESCs and might have the potential to differentiate into neural crest derivatives, we collected c-Kit-positive cells in the CD45-negative fraction, which contained no hematopoietic cells. The c-Kit+/CD45−“nonhematopoietic” cells differentiated into neurons, glial cells, and melanocytes. To confirm this finding, we cultured SCL/tal-1-targeted J1 ESCs, which never differentiate into hematopoietic cells [31, 32], isolated the c-Kit-positive cells, and recultured them on ST2 monolayers. The c-Kit-positive cells appeared and generated colonies comprising neurons, glial cells, and melanocytes, just as did the wild-type D3 ESCs (Fig. 3H). This result indicates that the neurons and glial cells were the descendants of ESC-derived neural crest-like cells and had not differentiated from ESC-derived hematopoietic stem cells.
As we could successfully concentrate neural crest-like cells from the cultures by using c-kit molecules as a surface marker, up to 50% of c-Kit+ cells formed colonies containing melanocytes, indicating that c-Kit is not sufficient to exclusively discriminate neural crest precursors. To search for another neural crest-specific cell-surface marker molecule to more effectively enrich neural crest-like cells induced in the present culture system, we investigated c-Ret expression, because the expression of c-Ret was reported to occur in the early stage of sensory neuron differentiation from neural crest cells . Flow-cytometric analysis, however, showed no c-Ret-positive cells in our ESC cultures (data not shown). To achieve more effective enrichment of neural crest cells, further investigation for cell-surface marker molecules is necessary.
Because the growth factors BMP-2 and neuregulin (NRG)-1 were demonstrated to instruct NCSCs to differentiate into neurons and glial cells, respectively , we added BMP-2 or NRG-1 to cultures of purified c-Kit+/CD45− cells. In comparison with cultures without BMP-2 or NRG-1, the fraction of neurons or glial cells did not change significantly after the addition of BMP-2 or NRG-1. This means that ST2 stromal cells already had created an appropriate environment for the differentiation of neurons and glial cells in addition to that for melanocyte differentiation. In support of this view, mouse thymus neural crest cells were shown to differentiate into melanocytes, neurons, and glial cells on ST2 stromal cell monolayers . In a stromal cell-free NCSC culture system adopted from Morrison et al. , c-Kit+/CD45− cells purified by flow cytometry did not survive more than 10 days (data not shown). Thus, we conclude that ST2 stromal cells are critical for neural crest survival and differentiation in the present culture system.
Smooth muscle cells are another neural crest derivative also reported to be generated from NCSCs. Although α-smooth muscle actin-positive cells were generated on ST2 monolayers (Fig. 3F), their numbers were very small. This finding means that this culture system, in which differentiation into neurons, glial cells, and melanocytes could occur, is inadequate for the effective differentiation of smooth muscle. Factors such as transforming growth factor-β, which is known to promote NCSC differentiation into smooth muscle cells , might be necessary for such differentiation.
Characterization of neural crest differentiation has been hampered in part by the fact that only a few neural crest cells are generated during neural development in vivo. In our culture system, neural crest-like cells can be generated in unlimited numbers. Thus, purified populations of ESC-derived neural crest like-cells instead of neural tube-derived neural crest cells may be most useful for the study of certain aspects of neural crest biology, as well as for future cell therapy for various neural crest cell-related diseases.
The authors indicate no potential conflicts of interest.
We thank Dr. Shin-ichi Nishikawa (Riken, Kobe, Japan) for providing us with ACK2-PE antibody and other members of our laboratory for discussion and critical reading of the manuscript. This study was supported by grants from the Ministry of Education, Science, and Culture and the Agency for Science and Technology, Japan.