Human embryonic stem cells (hESCs) have been directed to differentiate into neuronal cells using many cell-culture techniques. Central nervous system cells with clinical importance have been produced from hESCs. To date, however, there have been no definitive reports of generation of peripheral neurons from hESCs. We used a modification of the method of Sasai and colleagues for mouse and primate embryonic stem cells to elicit neuronal differentiation from hESCs. When hESCs are cocultured with the mouse stromal line PA6 for 3 weeks, neurons are induced that coexpress (a) peripherin and Brn3a, and (b) peripherin and tyrosine hydroxylase, combinations characteristic of peripheral sensory and sympathetic neurons, respectively. In vivo, peripheral sensory and sympathetic neurons develop from the neural crest (NC). Analysis of expression of mRNAs identified in other species as NC markers reveals that the PA6 cells induce NC-like cells before neuronal differentiation takes place. Several NC markers, including SNAIL, dHAND, and Sox9, are increased at 1 week of coculture relative to naive cells. Furthermore, the expression of several NC marker genes known to be downregulated upon in vivo differentiation of NC derivatives, was observed to be present at lower levels at 3 weeks of PA6-hESC coculture than at 1 week. Our report is the first on the expression of molecular markers of NC-like cells in primates, in general, and in humans, specifically. Our results suggest that this system can be used for studying molecular and cellular events in the almost inaccessible human NC, as well as for producing normal human peripheral neurons for developing therapies for diseases such as familial dysautonomia.
Pluripotent human embryonic stem cells (hESCs) have been directed to differentiate into neuronal cells using many cell-culture techniques. Several central nervous system (CNS) cell types with potential clinical importance have been produced from hESCs, including dopaminergic and cortical pyramidal-like neurons (reviewed in ). However, for understanding of pathogenesis and drug discovery/treatment of peripheral neuropathies such as Familial Dysautonomia (FD), it will be necessary to produce peripheral neurons from hESCs. Availability of human peripheral sensory neurons (PSNs), for example, would permit the study of how the truncated protein product of the IKBKAP gene, the gene mutated in the vast majority of FD patients (reviewed in ), leads to specific degeneration of this type of neuron in this disease. In addition, normal human sensory neurons could be useful for screening of new and improved drugs for treating FD and other peripheral neuropathies [3, 4]. It is even possible that such neurons could eventually be used for cell-replacement therapy in patients who have significant sensory neuron loss.
Sasai and his colleagues have shown that neurons are efficiently induced from ESCs of mice  and primates  when cocultured with the mouse stromal line PA6. This induction was named “stromal-derived inducing activity” (SDIA). Others have used this technique to enrich hESC cultures for neuronal precursors . Adding BMP4 to the culture medium after several days of coculture of SDIA-mouse ESC leads to the generation of a population of neurons coexpressing peripherin (a molecule characteristic of neurons with peripheral axons: sensory, autonomic, and motor ) and Brn3a, (a transcription factor characteristic of PSNs and a small population of CNS neurons ). Expression of this combination of markers is thought to be characteristic almost exclusively of PSNs . Rhesus monkey ESCs generated peripherin+/Brn3a+ sensory neurons spontaneously at low levels when cultured with PA6, and there was enrichment for this phenotype when low concentrations of BMP4 were added to the medium after several days of coculture. These cultures also included cells that were tyrosine hydroxylase (TH)+/peripherin+, a characteristic of peripheral sympathetic neurons.
The peripheral nervous system (PNS) develops from a unique progenitor/stem cell population: the neural crest (NC) (reviewed in ). PA6 induction of peripheral sensory and sympathetic neurons from murine ESC was apparently preceded by the appearance of markers consistent with an NC intermediate because at early stages of the induction process, several molecules known to be expressed in the NC were detected by immunocytochemistry or polymerase chain reaction (PCR) . These include snail, Sox9, FoxD3, Msx-1, and p75NGFR . The development of the NC has been the topic of intense study in amphibians, avians, and rodents, but little is known about the NC of primates, including that of humans. To date, there have been no reports of the expression of these or other putative NC markers in the NC of primates, including that of humans.
In this study, we applied a modification of the SDIA induction method to two lines of human ESCs to produce cells with the characteristics of human peripheral neurons, both sensory and sympathetic. As was observed for rhesus ESCs, peripheral neurons are generated using SDIA without added BMP4. Similar to the mouse, but demonstrated here for the first time for primates, the peripheral ganglion-like neurons appear to pass through an NC-like stage. SDIA induction of peripheral neurons has great promise not only for study and drug discovery/testing for peripheral neuropathies, but for investigation of human NC development, which occurs at stages of embryonic development that are difficult to access.
Materials and Methods
hESCs (HES-1 [XX]  and HUES 7 XY and HUES1 [XX] cell lines) were cultured on mitotically inactivated mouse embryonic or human neonatal fibroblast feeder layers in gelatin-coated tissue culture dishes and passaged every 6–7 days in 80% knockout Dulbecco's modified Eagle medium (DMEM) supplemented with 20% knockout serum replacement, 1 mM glutamine, 1% nonessential amino acids, 75 U/ml penicillin, 50 μg/ml streptomycin, 0.1 mM β-mercaptoethanol, and 4 ng/ml basic fibroblast growth factor. The mouse PA6 cell line, obtained from the Riken Cell Bank (Riken, Japan, http://www.brc.riken.jp/lab/cell/english), was cultured on gelatin-coated dishes in 90% DMEM, 10% fetal calf serum, 4.5 gm/l D-glucose, 1 mM L-glutamine, 75 U/ml penicillin, and 75 μg/ml streptomycin.
Approximately 104 PA6 cells were seeded on gelatin-coated 13-mm coverslips in 24-well dishes. Confluent cultures were treated with trypsin/EDTA for 4 minutes, and the fibroblast feeder layer was removed manually, leaving the hESC colonies. The colonies were then triturated, and approximately 1,000 hESCs were placed in each PA6-containing well. The medium was then changed to 90% BHK-21 medium/Glasgow modified Eagle medium (MEM), 10% knockout serum replacement, 2 mM glutamine, 1 mM pyruvate, 0.1 mM nonessential amino acid solution, and 0.1 mM β-mercaptoethanol. Medium was changed 4 and 6 days after hESC plating. On day 8, the medium was changed to 90% BHK-21 medium/Glasgow MEM, 100 μM tetrahydrobiopterin, 2 mM glutamine, 1 mM pyruvate, 0.1 mM nonessential amino acid solution, N2 supplement X1, and 0.1 mM β-mercaptoethanol. Subsequently, medium was replaced every 2 days.
Coverslips were rinsed in phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde for 30 minutes. After a rinsing in PBS, the coverslips were incubated for one hour in blocking solution containing 1% bovine serum albumin, 5% horse serum and 0.5% Triton in PBS. The following antibodies were used at the indicated dilutions: β-III-tubulin/Tuj-I (1:600; Promega, Madison, WI, http://www.promega.com), neurofilament (1:600; Sigma, St. Louis, http://www.sigmaaldrich.com), peripherin (1:1100; Chemicon, Temecula, CA, http://www.chemicon.com), Brn3a (1:100; Chemicon), TH (1:100; Chemicon), NCAM (1:200; Chemicon), p75 (undiluted; Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com), E-cadherin (1:60, NeoMarker; Lab Vision, Fremont, CA, http://www.labvision.com), and AP2 (Developmental Studies Hybridoma Bank, Iowa City, IA, http://www.uiowa.edu/∼dshbwww).
Sections were rinsed repeatedly in PBS and then incubated with the primary antibodies for 1 hour at room temperature or overnight at 4°C. Secondary antibodies (Alexa 488, 594, and bio-tinylated antimouse and antirabbit followed by Extravidin-conjugates) were then applied, and nuclei were stained with Hoechst (0.1 μg/ml). Coverslips were then rinsed in PBS, mounted on microscope slides in 90% glycerol, 10% PBS, 1% n-propyl-gallate, and sealed with nail polish.
RNA was extracted using TriReagent (Sigma #T9424) and the Aurum Total RNA kit (#732–6820; Bio-Rad Laboratories, Hercules, CA, http://www.bio-rad.com), according to the manufacturers' protocols. Reverse transcription (RT)–PCR was performed using Ready-To-Go RT-PCR beads (#27–9266–01; GE Healthcare, Piscataway, NJ, http://www5.amershambiosciences.com/aptrix/upp01077.nsf/content/homepage_country_select) or Ready-Mix reverse-iT one-step kit (#AB-0844/LD; ABgene). Primer sequences (forward, reverse) and length of the amplification products were as follows:
Four-week cocultures were immunostained for peripherin (peri+) and Brn3a (brn+). Colonies were classified as containing double-stained cells (peri+/brn+), containing single-stained cells, or negative for both. A double-stained colony was one in which cells that express both Brn3a in the nucleus and peripherin in the cytoplasm were clearly identified. A few colonies were so thick that it was not possible to determine accurately whether Brn3a and peripherin were in the same cells, so these were not included in the quantification. Three separate experiments were performed for quantification purposes, and more than 400 colonies were counted. One-week cultures were quantified for the presence of AP2/p75 cells; a total of 163 colonies were scored.
SDIA Induction of Neuron-Like Cells from hESCs
Cells with the molecular characteristics of neurons are induced from murine and primate ESCs when cocultured with the mouse stromal line PA6 [5, 6]. In this study, we repeated this experiment with modified conditions and cultured two lines of hESCs with PA6cells(SDIA). After 7 days of culturing hESCs with PA6, a distinct morphological change in the hESC colonies was observed. Undifferentiated hESCs cultured with mouse or human foreskin fibroblasts appeared as dense, round, or oval monolayer colonies. hESC colonies, by contrast, were irregular in outline after 7 days of SDIA treatment. At that time, the large majority of colonies differentiated into primarily NCAM+ neural precursor cells (Figs. 1A–1C). Many colonies, primarily the larger ones, contained cells expressing the non-neural ectodermal marker E-cadherin+ in their centers (Figs. 1B, 1C). Several colonies expressed a few β-3-tubulin (Tuj-1)–staining cells (a universal and early marker for neurons), but most of these Tuj-1+ cells did not extend axons (not shown). The results are consistent with other observations that PA6 induces neural differentiation from hESCs (i.e., ).
Three weeks after seeding hESCs on PA6 cells, massive neuronal differentiation was observed using immunohistochemisty. More than 50% of the colonies were Tuj-1+ with extensive networks of stained axons (Figs. 1E, 1F). When hESCs were grown with the same media but on gelatin or laminin without a PA6 cell feeder layer, very few cells survived after 2 weeks of coculture, and none expressed neural markers (not shown).
Generation of Peripheral Neuron-Like Cells from hESCs
Previous studies have shown that SDIA could be used to generate both central and peripheral neurons from murine and primate ESCs. A study using SDIA on hESCs showed that it is neurogenic for hESCs as well, but the presence of peripheral neurons in the cultures was not investigated . To ascertain whether peripheral neurons are present in SDIA-induced cultures, we immunostained for the protein peripherin, which is present in neurons with axons outside the CNS . After 4 weeks of SDIA treatment, 51% of the colonies contained Tuj-1+/peri− CNS-like neurons, and 34.5% contained Tuj-1+/peripherin+ PSN-like neurons (402 colonies, 3 independent experiments) (Figs. 2A, 2B). The number of peripherin+/Tuj-1+ cells was widely variable between the colonies. RT-PCR analysis confirmed that peripherin mRNA was expressed in 3-week cultures, but not 1-week cultures (see below).
Most of the Tuj-1+ colonies also contained cells and processes expressing TH (Fig. 2C). Several neuronal types express TH, including catecholaminergic CNS neurons and peripheral sympathetic ganglion neurons. To explore whether the TH+ neurons in the SDIA cultures were CNS- or PNS-like, we double-stained colonies induced for 3 weeks with SDIA for peripherin and TH, a combination characteristic of sympathetic neurons. Of the 30% of the colonies that contained TH+ cells, approximately half contained TH+/peri+ cells (Figs. 2D–2I). Less than 1% of the cells in these colonies displayed the TH+/peri+ PNS-like catecholaminergic phenotype.
Next, we tested the cultures for the presence of another peripheral neuron subpopulation, sensory ganglion neurons (PSNs). There is no single marker thought to be specific for PSNs. Therefore, in previous studies of PSN differentiation, double-staining with antibodies to Brn3a (a transcription factor characteristic of PSNs and a small population of CNS neurons ) and peripherin has been used as a criterion for PSN identity . When we used this combination of antibodies to stain cultures induced by SDIA for 4 weeks, cells positive for both Brn3a and peripherin were observed in some of the colonies (Fig. 3). Some of these putative sensory neurons migrated away from the colonies, and, as might be expected from neurons derived from the migratory NC, these migrated clusters sometimes comprised half the number of the cells in the neighboring dense colony. The percentage of colonies that contained cells expressing Brn3a, peripherin, and double-stained cells was determined in three separate experiments. Quantitative analysis revealed that double-stained cells were present in 16.5% ± 3.1% of the more than 400 colonies observed, approximately the same proportion of colonies that contained peripherin+-only cells (18% ± 3.1%). Approximately half of the colonies contained Brn3a+ cells (51.7% ± 4.2%), and some colonies did not contain cells staining for either marker (13.7% ± 2.2%, Fig. 4). Double-stained neurons were also observed at 3 weeks, but not at 2 weeks of coculture (not shown).
Immature PSNs (like many other neurons) are bipolar, and the majority of the peri+/brn3a+ neurons we observed had this morphology (Figs. 3E, 3G). However, mature PSNs have a unique pseudounipolar structure, a morphology that arises from the fusion of the proximal segments of the two initial processes, as described by His around a century ago (Fig. 3H). We observed a few pseudounipolar peri+/brn+ cells and a number of double-stained cells with morphologies intermediate between immature and mature sensory neurons (Figs. 3E–3G). Peri+/brn+ cells with more than two processes exiting from the soma were never observed, in contrast to the frequent Tuj-1+ and TH+/peri+ multipolar neurons.
PSNs express tyrosine-kinase receptors (Trks) that bind trophic factors of the neurotrophin family (reviewed in ). In the chick embryo, TrkC is initially expressed in migrating NC and in virtually all cells in the nascent dorsal root ganglia (DRG), and then is downregulated in most DRG cells, and only remains in large, proprioceptive neurons . By contrast, the low-affinity neurotrophin receptor p75 is expressed both in migrating neural crest cells  and many postmitotic neurons in DRG. RT-PCR analysis of SDIA-treated hESCs revealed that TrkC was induced in 1-week cocultures when compared with naive hESCs, but subsequently its expression was lower at 3 weeks of cocultures (Fig. 5). p75, by contrast, was expressed only at low levels in the 1-week cocultures, and was highly induced in the 3-week cultures. Although other cell types express these receptors, these results are consistent with the known pattern of expression of these receptors in developing PSNs in the chick embryo. Peripherin mRNA was first observed at 3 weeks of coculture, as was the case for the protein (Fig. 5).
SDIA Induces Differentiation of NC-Like Cells
Most PSNs develop from the NC in vertebrate embryos (with the exception of those derived from ectodermal placodes in the head). To determine whether the differentiation of the PSN-like neurons we observed might be preceded by NC-like cells, we examined 7-day SDIA-induced hESCs for the presence of molecules expressed in murine NC. Unfortunately, there is no one specific marker indicative of NC identity, so we examined a number of different molecules used as NC markers in several species by immunostaining and RT-PCR.
RT-PCR analysis was performed for a series of NC markers on undifferentiated hESCs and on hESCs after 1 and 3 weeks of SDIA treatment. High expression of the early mammalian  NC cell marker  SNAIL was observed after 1 week of SDIA induction and was dramatically reduced after 3 weeks of treatment (Fig. 6). Other transcripts associated with NC development in the mouse also induced by 1 week of SDIA treatment included Sox9, dHAND, and MSX1. The upregulation of these genes in the 1-week cultures, and the subsequent decrease in their expression by 3 weeks of culture, are consistent with the presence of NC-like cells in the 1-week cultures, and their subsequent differentiation into sensory-like and sympathetic-like neurons by the third week.
A few genes considered to be NC markers did not show this pattern of early up- and later downregulation. FoxD3, expressed in the premigratory NC, was present in the 1-week cocultures. However, it was also expressed in naive hESCs (Fig. 6, ) as well as at 3 weeks of culture, so its detection was less informative than the aforementioned transcripts as to the presence of NC-like cells in the cultures. Immunocytochemistry showed that AP2 was expressed after 1 week of SDIA treatment, and this was confirmed by the presence of its mRNA. However, AP2 is expressed by epidermal cells as well as NC, and it is therefore not surprising that its mRNA expression continued to rise until 3 weeks, similarly to the increase in E-cadherin expression. It is therefore likely that our culture conditions are permissive for epidermal cells to differentiate and/or multiply. Pigmented cells were not observed in the cultures, and we did not stain for smooth muscle actin, because the SDIA method, coupled with the differentiation medium we used, has already been shown to inhibit the production of melanocytes and mesenchymal NC derivatives .
Immunocytochemical evidence also supported the induction of NC from hESCs using SDIA. Cells expressing AP2+/ NCAM, a combination thought to be characteristic of NC cells , were present in 42.3% ± 5.5% of the colonies after 1 week of coculture (163 colonies observed in two experiments) (Fig. 1D). Approximately 50% of the colonies also contained immunostained cells positive for the low-affinity neurotrophin receptor p75, which is expressed in migrating murine crest cells  (not shown).
We designed the PCR primers in this study to be specific for human mRNAs. Control experiments showed that the (murine) PA6 cells grown alone did not express any of the mRNAs for human NC markers. The PA6 cells expressed murine, and not human, actin transcripts, as expected (not shown).
In this study, we show that cells with the molecular and morphological characteristics of PNS and CNS neurons are generated from hESCs by coculturing them with PA6 stromal cells. Within the peripheral lineage, both sensory and sympathetic ganglion-like neurons were generated. This is the first report showing that cells with the specific molecular characteristics of peripheral sensory and sympathetic neurons can be generated from hESCs. Transplantation of hESCs into somites of chick embryos at the time of NC migration resulted in some human cells with primitive neuronal morphology colonizing the DRG and in the formation of ganglion-like clusters consisting of neurofilament-positive cells ; however, precise phenotypic identity of the putative peripheral neurons was not determined in that study.
Very recently, a report was published claiming to have produced sympathetic-like neurons from hESCs, using a different procedure . However, only single immunostaining for either peripherin or TH was shown. This left the identity of both populations of single-stained cells ambiguous. It has been shown in rats, although not in primates, that there is a small population of peripheral sensory neurons that transiently expresses TH . It is therefore theoretically possible that some of the TH+/peripherin+ cells we observed were not sympathetic-like, but rather sensory-like neurons.
As was found for primate ESCs, and in contrast to the situation for mouse ESCs, we found that the induction of peripheral neurons by PA6 did not require the addition of exogenous BMP4 . Our preliminary experiments adding BMP4 to the cultures have not shown a dramatic increase in PSNs, but we have not yet examined its effect on the induction of sympathetic-like neurons. Brn3a+/peripherin+ sensory-like neurons and TH+/peripherin+ autonomic-like neurons were detected only after at least 3 weeks of coculture. DRGs are present in human embryos from Carnegie stage 12 (approximately week 4 ). If “naive” hESCs are considered to be from the equivalent of 5-day blastocysts, the combined 21 + 5 days are approximately consistent with the timing of the generation of PSNs in the human embryo. There is no immunocytochemical or molecular information about the expression of peripherin or Brn3a in human embryos, so it is difficult to compare the hESC exactly with the in vivo situation.
Although molecular data are important for identifying derivatives of hESCs, additional types of evidence are necessary. The presence of Brn3a+/peripherin neuron-like cells with the morphologies expected of PSNs, bipolar, pseudounipolar, and intermediate in the SDIA cultures, was an important confirmation of this identity. Embryonic PSNs are characterized by expression of high-affinity receptors to the neurotrophins of the Trk family and the low-affinity neurotrophin receptor p75. We found that mRNA for TrkC was expressed after 1 week of coculture and was downregulated at 3 weeks. This is consistent with observations in the chick that TrkC is expressed on most migrating NC cells and all early DRG cells, but with differentiation, is limited to only the population of large DRG neurons . It is also possible that there was a selective death of TrkC-expressing cells in our cultures, but our experimental protocol does not allow us to distinguish between these possibilities. The low-affinity receptor for neurotrophins, p75, by contrast, is expressed initially by migrating neural crest cells and later by substantial numbers of neurons in the DRG in chicks and rodents. We again observed a parallel to this in the SDIA cultures: p75 message was virtually nonexistent in naive hESCs and in 1-week cocultures, but appeared simultaneously with the peripherin in the 3-week cultures. Again, this is the first demonstration of expression of these receptor molecules in hESC-derived cells. The earliest stages at which parallel human in vivo data have been published are for DRG from a Carnegie stage 15–embryo, which is approximately one developmental month later. In that study, it was found that approximately the same number of human fetal DRG cells expressed p75 (37%), and a much smaller percentage expressed TrkC (34%) .
Trunk NC induction and migration occur at embryonic stages in humans that are particularly inaccessible (Carnegie stages 10–12, approximately 21–24 days postconception ), before pregnancy is usually noticed and before elective abortions are usually performed. There is, therefore, a great paucity of human embryos from these stages, and it was of particular interest to examine the expression of genes that are considered markers of NC in lower species in our hESC cultures. A high level of one of the earliest and most definitive markers for NC in mammals, SNAIL, was observed after 1 week of coculture. After 3 weeks of coculture, there was striking downregulation of this molecule. This pattern of high expression after 1 week and reduced expression after 3 weeks was observed for several other NC markers: Sox9, dHAND, and MSX1. This pattern of expression is consistent with the possibility that SDIA generates peripheral neuron-like cells via an NC intermediate, just as is the case in vivo. In the study of Sasai et al. , PSN- and SG-like neurons were shown to be induced from rhesus ESCs. In their study, however, no evidence was presented for the expression of NC cell markers in the primate ESC cultures. Our data are, therefore, the first molecular information on NC development in humans in particular, and as far as we could tell, in primates altogether. Our analysis of the molecular expression in hESC-derived putative human NC demonstrates the tremendous potential of hESCs for the study of molecular events of human embryonic development at early organogenetic stages that are normally difficult or impossible to study in normal human embryos.
The PSNs were a small proportion of the cells within the colonies. Only approximately 10% of the colonies with double-stained cells contained 10%–20% Brn3a+/peri+ PSNs, and half contained less than 5% double-stained cells. Preliminary experiments extending the cocultures to 6 weeks did not improve the yield of PSNs. This suggests that cells that were Brn3a+/peri- at 3 weeks could not further differentiate to express peripherin afterward. We are now attempting to improve the efficiency and purity of SDIA induction of NC-like cells, using sorting with cell-surface markers such as p75 and addition of growth factors. These techniques may allow us to isolate and expand a human NC population for further molecular and cellular characterization. In addition, generation of expandable progenitor population should allow us to produce sufficient numbers of PSNs to develop a system for drug testing, gene therapy, and, potentially, cell transplant therapy for diseases in which peripheral neurons are damaged or degenerate such as FD.
This work was supported by a very generous grant from the Familial Dysautonomia Foundation, Inc., New York. We thank Chaya Kalcheim (Hebrew University, Jerusalem) and Benny Motro (Bar-Ilan University, Ramat-Gan, Israel) for helpful discussions of these results. Chaya Morgenstern provided excellent logistical and technical support, and Natalie Nagorsky and Orel Pasder helped maintain the hESCs. Human foreskin fibroblasts were a gift of InterPharm Laboratories, Ness Ziona, Israel. The hESC line HUES-7 was provided by Doug Melton (Harvard University, Cambridge, MA). The AP2 antibody was developed by Trevor Williams and obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the NICHD and maintained by the Dept. of Biol. Sciences, University of Iowa, Iowa City, IA.