Directed Differentiation and Transplantation of Human Embryonic Stem Cell-Derived Motoneurons



Motoneurons represent a specialized class of neurons essential for the control of body movement. Motoneuron loss is the cause of a wide range of neurological disorders including amyotrophic lateral sclerosis and spinal muscular atrophy. Embryonic stem cells are a promising cell source for the study and potential treatment of motoneuron diseases. Here, we present a novel in vitro protocol of the directed differentiation of human embryonic stem cells (hESCs) into engraftable motoneurons. Neural induction of hESCs was induced on MS5 stromal feeders, resulting in the formation of neural rosettes. In response to sonic hedgehog and retinoic acid, neural rosettes were efficiently directed into spinal motoneurons with appropriate in vitro morphological, physiological, and biochemical properties. Global gene expression analysis was used as an unbiased measure to confirm motoneuron identity and type. Transplantation of motoneuron progeny into the developing chick embryo resulted in robust engraftment, maintenance of motoneuron phenotype, and long-distance axonal projections into peripheral host tissues. Transplantation into the adult rat spinal cord yielded neural grafts comprising a large number of human motoneurons with outgrowth of choline acetyltransferase positive fibers. These data provide evidence for in vivo survival of hESC-derived motoneurons, a key requirement in the development of hESC-based cell therapy in motoneuron disease.

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


Motoneurons are the key effector cell type for control of motor function, and loss of motoneurons is associated with a number of debilitating diseases such as amyotrophic lateral sclerosis (ALS) and spinal muscular atrophy [1,2]. Motoneurons are also regarded as a great model for probing mechanisms of vertebrate central nervous system (CNS) development, and the transcriptional pathways that guide motoneuronal specification are well characterized [3]. Recent studies have demonstrated the in vitro derivation of motoneurons from mouse [4,5] and human embryonic stem cells (hESCs) [6,7]. Current hESC-derived motoneuron differentiation protocols are based on embryoid body mediated neural induction followed by exposure to defined morphogens [6,7] such as sonic hedgehog (SHH) and retinoic acid (RA) acting as ventralizing and caudalizing factors, respectively. It has been suggested that the ability to undergo motoneuron specification under these conditions is temporally restricted to the earliest stages of neural induction [6,7]. Characterization of these cells in vitro and in vivo has been limited. Furthermore, there are currently no published data on the ability of hESC-derived motoneurons to secrete acetylcholine, the key neurotransmitter of spinal motoneurons, and to survive and maintain motoneuron characteristics in the developing or adult cord. In vivo survival and the ability for orthotopic integration are key requirements for the future applications in animal models of motoneuron disease. Here, we present a novel in vitro protocol for the efficient generation of hESC-derived motoneurons. Furthermore, we demonstrate in vivo survival and maintenance of motoneuron phenotype upon transplantation into the developing chick and adult rat spinal cord.

Materials and Methods

Culture of Undifferentiated hESC and Nonhuman Primate ESC Lines

H1 (WA-01, passages [P]36–52), H9 (WA-09, P44–60), RUES1-EGFP (P39–45), and the cynomolgus parthenogenetic line (Cyno1, P42–59) were cultured on mitotically inactivated mouse embryonic fibroblasts (Specialty Media, Phillipsburg, NJ, Undifferentiated hESCs and monkey ESCs were maintained under growth conditions and passaging techniques described previously for each given line: H1, H9, and RUES1-eGFP [8]; Cyno1 [9].

Neural Induction and Motoneuron Differentiation

MS5 stromal cells were maintained in α-minimal essential medium containing 10% fetal bovine serum and 2 mMl-glutamine as described previously [5]. Neural differentiation of hESCs was induced via coculture on MS5, and hESCs were plated at 5–20 × 103 cells on a confluent layer of irradiated (50 Gy) stromal cells in 6-cm cell culture plates in serum replacement medium containing Dulbecco's modified Eagle's medium, 15% knockout serum replacement (Invitrogen, Carlsbad, CA,, 2 mMl-glutamine, and 10 μM β-mercaptoethanol. At day 16, cultures were switched to N2 medium modified according to Johe et al. [10]. In some of the experiments, Noggin was added to the culture medium during neural induction on MS5. Neural rosettes were mechanically isolated around day 28 of differentiation. Isolated rosettes were gently replated on polyornithine (15 μg/ml)/laminin (1 μg/ml) coated culture dishes and maintained in N2 medium supplemented with ascorbic acid (AA) and brain-derived neurotrophic factor (BDNF) in the presence of RA/SHH (termed passage 1 [P1]). After 7–10 days (∼80% confluence), cells were mechanically passaged after exposure to Ca2/Mg2-free Hanks' balanced salt solution (HBSS) for 1 hour at room temperature and spun at 200g for 5 minutes. Cells were resuspended in N2 medium supplemented with BDNF and AA and replated again onto polyornithine/laminin coated culture dishes (50–100 × 103 cells per cm2) in the presence of RA/SHH (passage 2 [P2]). After an additional 10–15 days of culture, cells were differentiated in the absence of RA/SHH but in the presence of AA, BDNF, and glial cell line-derived neurotrophic factor (GDNF) (R&D Systems Inc., Minneapolis, Medium was changed every 2–3 days, and growth factors were added in various combinations and at various time points as described in the main text: SHH (500 ng/ml), BDNF (20 ng/ml), GDNF (20 ng/ml), Noggin (500 ng/ml) (all from R&D Systems), all-trans-Retinoic acid (1 μM), and AA (0.2 mM) (both from Sigma-Aldrich, St. Louis, Differentiation of hESCs into midbrain dopamine neurons for Affymetrix (Santa Clara, California, gene expression analysis followed our previously published protocol [11].


Cells were fixed in 4% paraformaldehyde/0.15% picric acid and stained with the following primary antibodies: rabbit polyclonal antibodies: Sox1 (United States Biological, Swampscott, Massachusetts,; 1:200), Pax6 and β-III Tubulin (Covance, Princeton, NJ,; 1:75, 1:2,000), Otx2 (F. Vaccarino; 1:1,000), Nkx6.1 (S. Heller; 1:5,000), BF1 (1:1,000); goat polyclonal antibodies: choline acetyl transferase (ChAT) (Chemicon, Temecula, CA,; 1:200); mouse monoclonal antibodies (IgG): Pax6, HB9 (MNR2), Isl1, Pax7, Lhx3 (Lim3), Nkx2.2 and HoxB4 (Developmental Studies Hybridoma Bank, Iowa City, IA,∼dshbwww; l:75), Tuj1 and HoxC8 (Covance; 1:500, 1:200), Olig2 (Qing Richard Lu; 1:1,000), human neural cell adhesion molecule (hNCAM, clone Eric1; Santa Cruz Biotechnology Inc., Santa Cruz, CA,; 1:200), human nuclear antigen (Chemicon; 1:200); guinea pig: HoxC6 (T. Jessell, 1:8,000). Appropriate Alexa488 and Alexa555 or Alexa568 labeled secondary antibodies (Molecular Probes, Eugene, OR,, biotinylated rabbit anti-goat followed by Alexa555 or Alexa568, and Alexa488-StreptAvidin conjugate and 4,6-diamidino-2-phenylindole (DAPI) counterstain were used for visualization. Image acquisition for cultured cells was performed using an inverted Olympus IX71 epifluorescence microscope with appropriate filter sets (Chroma, Rockingham, VT, using single channel acquisition on a Hamamatsu Orca CCD camera. Histological sections were imaged on an upright Olympus epifluorescence microscope (AX70) linked to a Photometrics CoolSNAP fx, acquired as single channel images, and merged using IPLab version 3.6. Confocal images such asFigure 5C were obtained via optical sectioning at 0.8 μm on a Leica (Heerbrugg, Switzerland, TCS SO2 AOBS set-up and reconstructed using a Leica Confocal Software Lite package.

Reverse Transcription-Polymerase Chain Reaction

Total RNA was extracted using the RNeasy kit and DNase I treatment (Qiagen, Hilden, Germany, to avoid genomic contamination. Undifferentiated hESCs and hESC progeny at day 28 of differentiation were manually dissected off the feeder layer to avoid contamination with stromal cells. Total RNA (1–2 μg each) was reverse transcribed (SuperScript; Invitrogen). Polymerase chain reaction (PCR) conditions were optimized, and linear amplification range was determined for each primer by varying annealing temperature and cycle number. PCR products were identified by size. All samples were treated with DNase to eliminate potential contaminating genomic DNA. Primer sequences, cycle numbers, and annealing temperatures are provided as supplemental data (supplemental online Table 1). All reverse transcription-PCR data shown are from H9; similar data were derived from H1 and Cyno1.

Affymetrix Analysis

Total RNA (5 μg) from day 15 of P2 motoneuron (MN) culture, day 14 of P2 dopamine (DA) culture, three samples of day 28 neural precursors cocultured on MS5 stroma, and three samples of undifferentiated hESCs (H1; passages 42–46) were processed by the Memorial Sloan-Kettering Cancer Center Genomics Core Facility and hybridized on Affymetrix U133A human oligonucleotide arrays. All analyses were done using the R statistical system and the BioConductor ( microarray analysis packages. The raw Affymetrix cell files were normalized and quantitated using the GC-RMA package, and the logarithm (base 2) of the expression levels was used in the subsequent analysis. To determine genes that were differentially expressed among each of the three differentiated samples (P2 MN, P2 DA neuron, and day 28 neural precursors) and the pooled data from the three undifferentiated samples of hESCs, a variant of thet test was used as implemented in the LIMMA package from BioConductor. In this method, the variance is corrected by adding a term that is computed from the global noise levels in the samples. This term improves robustness of the variance estimate, especially when dealing with small sample sizes. To account for multiple testing, we used the false discovery rate (FDR) method and applied a cutoff of 0.05 in the FDR to each list. This gave us three lists of sizes 1,426 (P2 MN), 651 (P2 DA neurons), and 44 (day 28 neural precursors). The larger number of transcripts in P2 MN and P2 DA versus day 28 neural precursor (group) indicates that these two groups are more distinct from undifferentiated hESCs than day 28 neural precursors.


For electrophysiological recordings, we mechanically transferred P2 cells at day 15 of differentiation onto astrocyte feeders that were derived in an autologous fashion from the respective hESC line or Cyno1 following protocols described previously [11]. Recordings were performed 10–15 days after replating in differentiation medium containing AA, BDNF, and GDNF as described above. Cells were placed on the stage of an inverted microscope (Axiovert 200; Carl Zeiss, Jena, Germany,, and the culture medium was replaced by a physiological saline solution containing 145 mM NaCl, 3 mM KCl, 1.5 mM CaCl2, 1 mM MgCl2, 5 mM glucose, and 10 mM HEPES (pH 7.4) and having 320–330 mOsm osmolarity. Recordings were performed at room temperature using the whole-cell patch-clamp technique. Patch pipettes were of resistance 4–5 MΩ and contained 140 mM potassium gluconate, 10 mM NaCl, 1 mM CaCl2, 10 mM HEPES, 0.2 mM EGTA, 3 mM ATP-Mg, and 0.4 mM GTP (pH 7.3), and osmolarity was adjusted to 295–305 mOsm with sucrose. Pipette-to-bath resistance was typically 5–10 MΩ. Data were acquired in current-clamp mode using the HEKA EPC 10 amplifier and acquisition system (HEKA, Lambrecht, Germany,

High-Performance Liquid Chromatography Analysis

High-performance liquid chromatography (HPLC) analysis was used to quantify acetylcholine (Ach) secreted from the hESC- or Cyno1-derived motoneurons. Ach was identified by electrochemical detection after enzymatic conversion of Ach into H2O2 by a solid-phase reactor (containing immobilized choline oxidase and acetylcholinesterase enzymes) [12]. Conditioned medium was collected at 50–65 days of differentiation for H9 (WA-09) cells and 40–55 days of differentiation for Cyno1 cells and stored at −80°C prior to analysis. Separation of the injected samples (25 μl per injection; ESA [Chelmsford, MA,] Autosampler 540) was achieved by isocratic elution in a mobile phase containing 100 mM Na2HPO4 (Sigma) (pH 8.0), 2 mM 1-Octanesulfonic acid sodium salt (Fluka, Buchs, Switzerland,, and 0.005% reagent MB (ESA) at 0.3 ml/minute. The oxidative potential of the analytical cell (ESA model 5040 with platinum target, Coulochem II) was set at 300 mV. Results were validated by coelution with acetylcholine standards under varying buffer conditions and detector settings. The limit of detection in the system was 50 fmol per injection.

Cell Counts and Statistical Analyses

Data for percentage of HB9/DAPI or HB9/β-III Tubulin cells were derived from a total of three independent experiments for each H9 (WA-09) and Cyno1 line. Cells were selected for quantification in uniform random fashion (fractionator). Each field was scored first for DAPI positive nuclei or β-III Tubulin positive cells and subsequently HB9/DAPI or HB9/β-III Tubulin colocalization. For each experiment, an average of greater than 8,000 cells were scored. The total number of cells analyzed in this study exceeded 25,000 cells. Percentages of HB9/DAPI or HB9/β-III Tubulin cells were compared with analysis of variance and Newman-Keuls post hoc analysis (Statistica 5.5; StatSoft, Tulsa, OK, Data are presented as mean ± SEM.

In Ovo Transplantation

Fertile eggs (CBT Farms, Chestertown, MD) were incubated at 37°C in a humidified incubator. hESC-derived motoneurons at days 15–18 in P2 stage were dissociated and transplanted into stage 16-17 chick spinal cord. Graft site extended over 1–2 somites. A small mechanical lesion was performed to generate space for graft. Eggs were placed back into the incubator for up to 9 days post-transplantation. Embryos were harvested and fixed with 4% paraformaldehyde for 1 day, incubated in 30% sucrose for 1–2 days, and embedded in optimum cutting temperature compound and cryosectioned for further immunohistochemical analysis.

In Vivo Transplantation into Adult Rat Spinal Cord

Three-month-old Sprague-Dawley rats were anesthetized with a mixture of ketamine/xylazine and placed in a stereotactic frame (Kopf, Tujunga, California, The skin over the cervical area was shaved and prepped in Betadine solution. An incision was performed through skin and fascia down to the C2 spinous process. The paraspinal muscles were dissected off the spinous process and laminae and the posterior bone elements were removed using a dental drill and small bone rongeurs. The median dorsal groove and vessel were used as the midline reference. A pulled glass microcapillary and pump were used to deliver the cells. Cells were harvested at day 15 of P2 differentiation (SHH/RA stage) and mechanically triturated after incubation in Ca2/Mg2-free HBSS for 1 hour at room temperature. Cells were resuspended at 100,000 cells per microliter for transplantation. Bilateral injections delivering 40,000 cells each were performed at the following coordinates: 0.5 mm lateral to the midline and 1.2 mm in depth from surface [13]. The dura was punctured with a 25-gauge needle to facilitate smooth entry of the microcapillary, which remained in place postinjection for a period of 10 minutes. The muscles were approximated with a 000 Vicryl suture and the skin with staples. The animals did not experience any obvious neurological deficit post operation. All animals received 20 mg/kg cyclosporin intraperitoneal 1 day before surgery and for the duration of the experiment. All animals received Sulfatrim medicated feeds as of the day of surgery. Crede maneuver (massaging of the bladder) was performed on all animals as a precautionary measure on day 1. Animals were sacrificed on day 1 (two animals) and 2 weeks (two animals) and 6 weeks (six animals) after transplantation.


We have previously described MS5 stromal feeder based coculture strategies for the directed differentiation of mouse and hESCs toward defined neural progeny [5,11,14]. The MS5 protocol belongs to the category of “stromal cell-derived inducing activity” based neural induction strategies [15,16] and can be adapted to the generation of a wide range of neuronal and glial cell types in mouse ESCs [5]. During hESC differentiation, MS5 based protocols yield large numbers of columnar neurepithelial structures termed “rosettes,” which can be further patterned and differentiated into various neuronal and glial cell fates such as midbrain dopamine neurons [11] or multipotent neural precursor cells [14]. Exposure of hESC-derived neuronal rosettes to fibroblast growth factor 8 (FGF8) and SHH enriches for ventral midbrain precursor cells that can be further differentiated into postmitotic dopamine neuron progeny. These data suggest that neural rosettes appropriately respond to developmental cues that specify anterior-posterior (AP) and dorsoventral (DV) identity during neural development. Here, we tested whether MS5 based neural induction of hESCs can be adapted for the efficient derivation of functional and engraftable motoneurons.

Neural induction of hESCs resulted in the formation of neural rosettes, as described previously [11], expressing Pax6 and Sox1 (Fig. 1A). These two transcription factors are known markers of the neural plate stage in mouse development [17, [18]–19]. At day 28 of differentiation on MS5, most cells in neural rosettes coexpressed Pax6 and Sox1 (Fig. 1A) and were strongly positive for the markers associated with anterior neural identity such as BF1 (FOXG1B) [20] and Otx2 (data not shown). In contrast, there was a lack of expression of posterior neural markers such as Gbx2 (anterior hindbrain) and HoxB4 (posterior hindbrain and spinal cord) (data not shown). These data suggest that, under default conditions, neural rosettes induced on MS5 progress toward an anterior neural patterning state.

Figure Figure 1..

Neural patterning and motoneuron derivation from human (h)ESCs. (A): Pax6 and Sox1 expression in neural rosettes at day 28 of differentiation on MS5 stroma. (B–C): Representative anterior-posterior (B) and dorsoventral (C) markers expressed in P2 rosette progeny (day 45 of differentiation) in the presence or absence (inset) of RA/SHH. (D–G): Immunocytochemical analysis of motoneuron markers at day 15 (D, E) and day 25 (F, G) of differentiation during P2 stage. (H): Percentage of HB9/DAPI cells for hESC and Cyno1 progeny at P2 after 15 days of RA/SHH treatment or under control conditions in the absence of RA/SHH. (I): Gene expression analysis of motoneuron markers during differentiation; day 0 (undifferentiated hESCs) and day 28 of differentiation on MS5; days 5, 10, and 15 during P2 differentiation. For schematic culture protocols, see Figure 2A. Scale bar in (A) corresponds to 200 μm for (A–C); 100 μm for (D, E); 25 μm for (F); 30 μm for (G). Abbreviations: ChAT, choline acetyl transferase; DAPI, 4,6-diamidino-2-phenylindole; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; hES, human embryonic stem; MN, motoneuron; P2, passage 2; RA, retinoic acid; SHH, sonic hedgehog; βTub, β-III Tubulin; VAChT, vesicular acetylcholine transporter.

We next tested whether progressive expression of anterior neural markers in rosettes can be modulated by appropriate extrinsic cues. Neural rosettes were mechanically isolated and maintained under feeder-free conditions on polyornithine/laminin-coated plates (P1 cultures). After 7 days of P1 culture, cells were replated to the P2 stage. Immunocytochemical analysis revealed maintenance of BF1 and Otx2 (Fig. 1B, inset) expression at day 10 of P2 cultures in the absence of extrinsically added patterning cues. However, exposure of P2 cultures to RA/SHH led to a dramatic decrease in BF1 with a corresponding increase in hindbrain and spinal cord markers such as HoxB4 and HoxC8 (Fig. 1B and supplemental online Fig. 1).

These data indicate that RA/SHH exposure is sufficient to suppress BF1 expression and to induce caudal identity markers. Progeny derived from human neural rosettes also showed appropriate modulation of DV fate markers in response to SHH exposure. Expression of ventral identity genes characteristic of the caudal CNS such as Nkx6.1 and Nkx2.2 required exposure of both RA and SHH (Fig. 1C and supplemental online Fig. 1).

We next tested whether RA/SHH-mediated induction of caudal and ventral patterning markers is associated with motoneuron specification and terminal differentiation into postmitotic motoneuron progeny. P1 cells were passaged and maintained for an additional 5–15 days in the presence of RA/SHH (P2 cultures). Gene expression analysis (Fig. 1I) confirmed the induction of transcripts related to motoneuron precursor cells (Nkx6.1, Olig2), early postmitotic motoneurons (neurogenin 2, Isl1), and more mature motoneurons (ChAT and vesicular acetylcholine transporter). Immunocytochemical studies at P2 confirmed the presence of large numbers of Olig2+ motoneuron precursors emerging from neuroepithelial structures. Cells expressing the postmitotic motoneuron marker Isl1 were arranged in the periphery of the Olig2+ precursor cell domains (Fig. 1D). After 10 days of P2 exposure to RA/SHH, large numbers of cells expressing the somatic motoneuron marker HB9+ were detected, and the percentage of HB9+ cells continued to increase in number until day 15 of RA/SHH treatment (Fig. 1E). Upon further differentiation in the absence of RA/SHH but in the presence of GDNF, BDNF, and AA, expression of more mature motoneuron markers was observed, including ChAT, the key enzyme required for the synthesis of Ach. An average of 26% of total cells expressed ChAT after 15 days of RA/SHH exposure followed by 5 days of differentiation in GDNF, BDNF, and AA. Less than 5% ChAT+ cells were present prior to day 10 of SHH/RA treatment. Cells treated identically but not subjected to SHH/RA exposure had less than 1% of cells positive for ChAT (supplemental online Fig. 2). Many ChAT+ cells coexpressed Lhx3 (Fig. 1F). Coexpression of ChAT and HB9 confirmed motoneuron identity of hESC progeny (Fig. 1G). Expression of Lhx3 in motoneuron progeny suggests that a subset of hESC-derived motoneurons expresses medial motor column markers, similar to the data reported recently for mouse embryonic stem cell-derived motoneurons [21]. The protocol was highly reproducible and yielded similar results for motoneuron progeny from a variety of hESC lines such as H1, H9, and RUES1-eGFP [22] (data not shown). The average number of HB9+ cells derived from hESC line (H9) was 20% out of total (DAPI+) cells (Fig. 1H). We next confirmed the robustness of the motoneuron induction protocol in Cyno1, a parthenogenetic monkey ESC-like cell line [9,23]. Motoneuron yield in Cyno1 was 43% HB9+ cells out of the total (DAPI+) population (Fig. 1H) and 65% of all (β-III Tubulin+) neurons.

Our study presents an MS5 based neural induction protocol for the efficient derivation of hESC-derived motoneurons. Motoneuron differentiation of hESCs in vitro has been characterized by the expression of a distinct set of markers such as HB9, Lhx3, and Isl1 [6,7]. However, there has been no previous report describing unbiased global expression profiles in specific hESC-derived neuronal populations. We performed genome-wide expression analysis using oligonucleotide arrays (Affymetrix U133A) to test whether known and novel candidate markers of motoneuron identity can be identified in hESC progeny. We compared expression profiles of hESC-derived cultures enriched in motoneuron (day 15 in P2 culture) versus midbrain dopamine neuron progeny (day 14 in P2 culture). A schematic representation of the experimental conditions is provided inFigure 2A. Transcripts related to hESC-derived motoneurons were defined as those with a signal threshold of greater than five times the levels detected in three independent samples of undifferentiated hESCs. The same criteria were used to define transcripts related to midbrain DA neuron differentiation and day 28 early neural precursor cells. A Venn diagram was drawn among all the transcripts related to at least one of the three categories of neural progeny. Based on this analysis, 976 transcripts were selectively enriched in motoneuron cultures, 204 transcripts in DA neuron cultures, and 19 transcripts in day 28 neural precursor cultures. A total of 429 transcripts were shared between P2 motoneuron and P2 DA neuron cultures (Fig. 2B). Motoneuron specific markers could be grouped into various categories including genes associated with ventral neural patterning and transcripts related to Hox and RA signaling. Key motoneuron markers such as HB9 (HLXB9) and Lhx3 were also grouped among motoneuron specific transcripts, validating the global expression data and independently confirming motoneuron identity of hESC progeny (Fig. 2C). Transcripts shared between motoneuron and midbrain DA neuron cultures but absent in early neural precursors and in undifferentiated hESCs included ion channel related markers defining pan-neuronal identity, a number of neurotrophic factor and neurotrophic factor receptor genes, and multiple neural and neuronal precursor genes (Fig. 2D). Key DA neuron specific transcripts comprised many of the known markers of midbrain DA neuron identity such as tyrosine-hydroxylase, Nurr1, and Engrailed-2 (Fig. 2E).

Figure Figure 2..

Genome-wide gene expression comparison among day 28 neural precursors and P2 motoneuron and P2 dopamine neuron culture using Affymetrix U133A expression arrays. (A): Schematic illustration of the protocol used for neural induction, MN and DA neuron differentiation of human (h)ESCs. Neural induction was achieved under serum-free conditions on MS5 stromal feeder cells (passage 0); hESC-derived neural rosettes were mechanically isolated at day 28 and replated onto polyornithine/laminin-coated plates. At day 35, P1 cells were replated in the presence of RA/SHH (P2) followed by terminal differentiation starting at day 50 of differentiation in the absence of RA/SHH but the presence of GDNF/BDNF/AA for MN differentiation. For DA neuron differentiation, SHH/FGF8 were treated for the first 7 days of P2 culture and then GDNF/BDNF/AA/TGFβ3/cAMP were added for further differentiation in the absence of SHH/FGF8 [11]. (B): Venn diagram presenting the overlap among transcripts selectively expressed in P2 MN enriched culture, P2 DA neuron enriched culture, and day 28 culture. (C–E): The top five genes most highly unregulated compared with hESCs are presented for any given group and supplemented with selected other highly expressed transcripts diagnostic for a given cell lineage. Fold changes of P2 MN versus hESCs are listed toward the right of the midline and fold changes of P2 DA versus hESCs are listed toward the left. (C): List of genes that are specifically enriched in the P2 MN culture. (D): List of genes that are commonly expressed in the P2 MN and DA neuron culture but not in the day 28 culture. (E): List of genes that are specifically expressed in the P2 DA neuron culture. Abbreviations: AA, ascorbic acid; BDNF, brain-derived neurotrophic factor; cult., culture; DA, dopamine; diff., differentiation; EN2, Engrailed-2; FGF8, fibroblast growth factor 8; GDNF, glial cell line-derived neurotrophic factor; hES, human embryonic stem; MN, motoneuron; NR4A2, Nurr1; P, passage; RA, retinoic acid; SHH, sonic hedgehog; TGFb3, transforming growth factor β3; TH, tyrosine-hydroxylase.

Work with mouse ESCs suggests that RA/SHH treatment induces motoneurons of rostral spinal cord identity expressing HoxC5 and HoxC6 [4]. In contrast, work with hESC-derived motoneurons suggested the derivation of motoneuron progeny with more caudal characteristics as assessed by the expression of HoxC8 [6]. However, the gene expression data presented indicate biased expression toward class 4–6 Hox genes (Fig. 2C) comparable with the profile reported for mouse ESC-derived motoneurons. Using immunocytochemistry, we observed that our hESC differentiation protocol yields motoneurons of both rostral (HoxC6) and more caudal (HoxC8) identity (Fig. 3A–3C). Overlap of HoxC6 and HoxC8 is indicative of posterior type brachial motoneuron identity [24,25]. We next tested in vitro function of motoneurons derived from H9 and Cyno1 using patch clamp studies (Fig. 3D). Single cell recordings demonstrated the induction of trains of action potentials in response to depolarizing currents. Mature electrophysiological signatures required coculture with astrocytes that were derived in an autologous fashion from the respective hESC line or Cyno1 following protocols described previously [11]. Biochemical evidence of motoneuron function was obtained by measuring Ach release in the conditioned medium of hESC- and Cyno1-derived motoneuron cultures after 15 days of RA/SHH exposure followed by 10 days of differentiation in the presence of AA, BDNF, and GDNF (Fig. 3E). HPLC analysis using electrochemical detection after enzymatic conversion of Ach into H2O2 by a solid-phase reactor [12] detected an average of 7.5–12 pmol Ach per milliliter medium (24 hours conditioning). Control P2 cultures treated identically but maintained in the absence of RA/SHH did not yield ChAT expressing neurons nor any detectable Ach signal (sensitivity <2 pmol/ml).

Figure Figure 3..

Human ESC-derived motoneuron characterization in vitro. (A–C): Hox gene expression at day 60 of motoneuron differentiation. (D): Electrophysiological analysis demonstrating action potentials in response to depolarizing currents in both human (h)ESC and Cyno1-derived motoneuron cultures. (E): High-performance liquid chromatography analysis of acetylcholine release in conditioned medium. Medium was collected from hESC or Cyno-derived passage 2 cultures after 10 days of differentiation with or without RA/SHH pretreatment. Scale bar in (A) corresponds to 100 μm for (A); 500 μm for (B, C). Abbreviations: Ach, acetylcholine; hES, human embryonic stem; ms, millisecond; mV, millivolt; RA, retinoic acid; SHH, sonic hedgehog.

The in vivo properties of hESC-derived motoneurons were tested after cross-species transplantation into the embryonic chick spinal cord at the time of motoneuron birth (HH16–17). The chick transplantation assay has been recently developed for reading out the in vivo potential of mouse ESC-derived motoneurons [4]. We used a hESC line expressing high levels of enhanced green fluorescent protein (eGFP) under control of a ubiquitous promoter (RUES1-eGFP [22]). Motoneuron progeny was isolated at days 15–18 of RA/SHH exposure and transplanted into the dorsal neural tube of stage HH16–17 chick embryos (Fig. 4A). A dorsal injection site was selected to test cell autonomous behavior of hESC-derived motoneurons in vivo in an ectopic location. This approach eliminates the possibility that local signals in the ventral cord exert instructive motoneuron inducing properties on uncommitted hESC progeny. Embryos were allowed to develop to stage HH26, HH34, and HH38 prior to analysis, corresponding to approximately 4.5, 8, and 12 days of chick development. At stage HH26, surviving hESC progeny located along the midline of the embryo could be readily detected by native eGFP signal (Fig. 4B). The location of human cells within the dorsal spinal cord was confirmed by histological analysis showing native eGFP expression and colocalization with human specific NCAM antibody (clone Eric1;Fig. 4C). The GFP+ cells also expressed markers specific to motor neuron fate including Nkx6.1 (Fig. 4D) and HB9 (Fig. 4E). These results suggest that hESC-derived motoneurons survive in the developing chick spinal cord and maintain expression of key motoneuron markers. The presence of human HB9+ and Nkx6.1+ cells at dorsal (ectopic) locations suggests that motoneuron specification occurred in vitro prior to implantation and independent of host-derived signals. However, at stage HH34, a significant proportion of cells was located ventrally outside the graft core on a migratory route toward the site of endogenous motoneurons in the ventral horn (Fig. 4F).

Figure Figure 4..

In vivo analysis of human (h)ESC-derived motoneurons in developing chick spinal cord. (A): Schematic illustration of transplantation procedure at stages HH16–17 in chick spinal cord. (B): Native eGFP fluorescence allowed detection of grafted hESC-derived cells in chick spinal cord (stage HH34). Area depicted corresponds to boxed area in inset. Inset provides an overview of the whole trunk region of the embryo. (C–E): Transverse sections at stage HH26 chick spinal cord. Human cells located in the dorsal spinal cord were identified by native eGFP signal and expression of human specific anti-neural cell adhesion molecule. Motoneuron identity was confirmed by Nkx6.1 (D) and HB9 (E) immunohistochemistry. (F): Transverse section of half spinal cord at stage HH34 showed hESC-derived motoneurons migrating toward ventral spinal cord. (G–H): Transverse sections at stage HH38 demonstrate human axonal projections out of the spinal cord. (G): Upper and lower-left panels: hNCAM+ fiber projecting out of the spinal cord via ventral root. Lower-right panel: human fibers reaching host skeletal muscle target. Inset: identity of target area revealed by expression of skeletal marker MF20 in adjacent section. (H): Long-distance axonal growth of hESC-derived motoneurons in vivo. Left panel: composite of transverse embryo section; region marked with (*) corresponds to the muscle target area depicted in (G), lower-right panel. Right panel: long-distance projection of human eGFP/hNCAM+ fiber. The location of the fiber corresponds to the boxed area in the left panel. Blue color in (C, F, G, H) represents 4,6-diamidino-2-phenylindole staining. Scale bar in (C) corresponds to 200 μm for (C); 125 μm for (F); 100 μm for (G) upper and lower-left panels, (H) right panel; 50 μm for (D, E) and (G) lower-right panel; 3 mm for (H) left panel. Abbreviations: eGFP, enhanced green fluorescent protein; hNCAM, human neural cell adhesion molecule.

A unique feature of motoneurons is their ability to extend fibers outside the CNS and to innervate peripheral muscle targets. Histological analysis at stage HH38 identified human axons exiting the spinal cord through the ventral root (Fig. 4G, upper and lower-left panels). Interestingly, many hESC-derived fibers outside the spinal cord were found in close proximity to axial and body wall musculature (Fig. 4G, lower right panel). Skeletal muscle identity was confirmed in adjacent sections using the skeletal muscle marker MF20 (inset). Interestingly, some human NCAM/eGFP+ processes were detected at distances of >3.5 mm from the spinal cord (Fig. 4H, human fibers [right panel], located in boxed area [left panel]). In summary, our data demonstrate that hESC-derived motoneurons are capable of exiting the developing chick spinal cord, exhibiting long-distance axon growth outside the CNS, and reaching peripheral targets in the trunk musculature.

Although our transplantation studies in the developing spinal cord demonstrate that hESC-derived motoneurons can survive and appropriately extend axonal fibers in response to local cues, applications in preclinical models of motoneuron disease will require survival and integration in the postnatal or adult CNS. To explore this question, we grafted hESC-derived motoneurons at the identical in vitro stage into the ventral spinal cord of 3-month-old Sprague-Dawley rats at the levels of C2 and C3 (Fig. 5A). Robust graft survival was observed 1 day, 2 weeks, and 6 weeks after transplantation. Survival was assessed by immunohistochemistry for eGFP and hNCAM (Fig. 5A, 5B). Whereas >20% of hESC-derived motoneuron progeny expressed HB9 (Fig. 5C) the first day after grafting, grafts at 2 weeks and 6 weeks after transplantation showed a decreasing number of HB9+ cells (<2% of all GFP+ cells expressed HB9 at 6 weeks after transplantation). The decrease in HB9 and Nkx6.1 expression was paralleled by an increase in the number of ChAT+ neurons from 4% of GFP+ cells within the first week after transplantation to 14% at 6 weeks. The temporary decrease in the percentage of ChAT+ cells compared with in vitro numbers suggest that ChAT+ cells may preferentially die or temporarily lose ChAT expression following grafting. Loss of HB9 expression in vivo followed a time course similar to that observed in hESC-derived motoneuron cultures in vitro. Furthermore, motoneurons in the developing chick spinal cord showed a similar loss in HB9 over time, and no HB9 expression was observed in endogenous motoneurons of the adult rat spinal cord. These data suggest that the switch from HB9 to ChAT expression corresponds to physiological motoneuron maturation rather than loss of in vivo phenotype. At 6 weeks after transplantation, the graft core was composed of a high percentage of eGFP/ChAT+ cells (Fig. 5D). We also observed single eGFP/ChAT double-labeled neurons located ventrally outside of the graft core with characteristic motoneuron morphologies (Fig. 5E). Current analysis was limited to the spinal cord and did not address fiber outgrowth outside the CNS. However, hNCAM staining revealed extensive fiber outgrowth and cell migration toward the ventral surface of the spinal cord (Fig. 5F). This suggests that hESC-derived motoneurons might be able to project toward the ventral root in the adult spinal cord, similar to our data presented in the embryonic chick spinal cord, albeit at a different time scale. We did not observe signs of tumorigenicity in any of the animals grafted. However, studies using larger groups of animals and assessing long-term survival beyond 6 weeks will be required to fully assess in vivo safety of the cells.

Figure Figure 5..

In vivo assessment of transplanted human (h)ESC-derived motoneurons in the adult rat spinal cord. (A–C): Coronal sections of spinal cord within the first 2 weeks after transplantation. Human graft was identified in orthotopic location in the ventral spinal cord by the expression of eGFP and human nuclear antigen. (B): Higher magnification image demonstrating colocalization of hNA and GFP+. (C): Expression of the motoneuron markers HB9 and Nkx6.1 in hESC-derived motoneuron grafts after transplantation. (D–F): Coronal sections of grafted adult rat spinal cord 6 weeks after transplantation. (D): hESC-derived eGFP+ cells expressed ChAT, a more mature motoneuron marker. (E): Reconstruction of confocalz-axis stack showing colocalization of eGFP and ChAT in a neuron located ventrally outside the graft core. (F): hNCAM+ fibers projecting toward ventral surface. In some instances, hNCAM+ cell bodies were aligned along human fibers (graft core is indicated by dashed line). Blue color in all figures represents 4,6-diamidino-2-phenylindole staining. Scale bar in (A) corresponds to 200 μm for (A), 150 μm for (F), 100 μm for (C) right panel; 60 μm for (C) left panel, 50 μm for (B, E), and 35 μm for (D). Abbreviations: ChAT, choline acetyl transferase; eGFP, enhanced green fluorescent protein; hNA, human nuclear antigen; hNCAM, human neural cell adhesion molecule.


This study demonstrates that hESC progeny is developmentally competent to respond to morphogens that determine AP and DV axis specification during early vertebrate development. Our protocols yield neural rosettes containing large numbers of neural precursors coexpressing Pax6 and Sox1 and maintained in vitro beyond the stage perviously defined critical for motoneuron specification. We previously showed that hESC-derived neural progeny at day 28 (expressing Pax6 and Sox1) are capable of midbrain dopaminergic differentiation [11]. The current study extends these findings to more caudal cell types, including motoneurons. Patterning of day 28 rosettes toward caudal fates is quite remarkable, given the majority of rosette cells expressing BF1, a marker traditionally associated with forebrain commitment. Our data also suggest that neural rosettes adopt anterior CNS markers by default. An anterior neural default model of CNS development has been postulated previously based on classic experiments in xenopus embryos. These studies showed that anterior CNS fates are first established and subsequently undergo a caudal transformation in response to secreted signals [26]. Morphogen-induced patterning of neural rosettes provides a powerful strategy to systematically generate neural cell types of distinct AP identity.

The protocols presented here offer a simple and efficient strategy to generate human motoneurons. Based on very conservative calculations, a single hESC plated at day 0 on MS5 for neural induction yields approximately 100 HB9+ motoneurons at day 50 of differentiation. These numbers suggest that therapeutically relevant numbers of motoneurons can be readily achieved. Although spinal motoneurons are derived from a single ventral pMN domain [27,28], they further acquire many different subtype identities based on positional identity, axonal projections, and gene expression. For translational applications of ESC-derived motoneurons, it will be essential to develop motoneuron subtype specific protocols that match the diseased population. There is evidence that most motoneurons derived from mouse ESCs using the RA/SHH protocol correspond to cervical or brachial level motoneurons based on Hoxc5 and Hoxc6 expression [4]. Similarly, many hESC-derived motoneuron in our protocol exhibit characteristics of brachial motoneurons. However, there is a slight caudal shift as compared with mouse ESC-derived motoneuron progeny toward HoxC6 and HoxC8 expression.

It is known that RA and FGF have antagonistic roles during development of the early chick spinal cord. FGF inhibits differentiation in the caudal stem zone, whereas RA provided rostrally by cells in the somitic mesoderm is required for neuronal differentiation and patterning of the ventral spinal cord. Mutual inhibition between the RA and FGF pathways not only controls neural differentiation but also underlies the progressive assignment of rostrocaudal identity by regulating Hox gene availability and activation [29,30]. In addition, recent work has shown that combinatorial actions of Wnt, RA, and FGF signals specify progenitor cell identities and motoneuron subtypes in the developing spinal cord [31]. Our current motoneuron differentiation protocols are based on exposure to high concentrations of RA, which may restrict AP identity. By decreasing the levels of RA and titrating FGF levels, hESC-derived rosette progeny may be directed toward more caudal motoneuron fates. Pretreatment with Wnt agonistic molecules may be another strategy to affect AP identity in motoneuron progeny.

We also successfully demonstrated that the transplantation of hESC-derived motoneurons in the developing chick embryo resulted in motoneuron survival and axonal outgrowth toward peripheral targets, despite the cross-species nature of the grafts, with considerable differences in developmental timing. Future studies will require detailed in vivo physiological analyses and retrograde tracing from specific muscle groups. However, such studies will be technically challenging due to the difficulties in following human axons over extended distances from the ventral horn to the peripheral muscle target within a single slice. Transplantation of specific motoneuron subtypes into the chick spinal cord will allow us to probe graft host interactions and to define the required subtypes necessary for restoration of motor function for a given spinal cord segment.

Finally, we show that hESC-derived motoneurons survive and maintain phenotype in the adult rat spinal cord for at least 6 weeks after transplantation. To our knowledge, this is the first study demonstrating in vivo survival of hESC-derived motoneurons, a prerequisite for future applications in preclinical models of motoneuron disease. Although our results provide an important first step in demonstrating functional engraftment, a number of key studies are required for future preclinical development. A key issue relates to the question of whether hESC-derived motoneurons can survive in animal models of motoneuron disease such as the SOD1G93A rat [32]. If robust survival can be achieved, it will be critical to test whether hESC-derived motoneurons are sensitive to endogenous inhibitors blocking axonal outgrowth in the adult CNS at levels similar to those observed with mouse ESC-derived motoneurons [33,34]. Additional steps will include questions related to the presence of guidance cues in the adult spinal cord, the capacity for long-distance axonal growth, and the ability to innervate appropriate peripheral targets. Finally, the relationship of target innervation and changes in behavioral parameters needs to be explored. Although the road to the clinical application of hESC-derived motoneurons remains extremely challenging, the ability to generate unlimited numbers of motoneuron progeny and the capacity for in vivo survival and integration in the developing and adult spinal cord are important first steps on this journey.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.


We thank S. Heller, Q.R. Lu, F. Vaccarino, and T. Jessell for Nkx6.1, Olig2, Otx2, and HoxC6 antibodies, H. Wichterle and J. Timmer for help in setting up chick embryo injections, and S. Noggle and A. Brivanlou for the RUES1-eGFP cell line. We also thank T. Bajwa and S. Clairmont for excellent technical assistance and M. Tomishima and S. Desbordes for critical review of the manuscript. This work was supported through grants from the ALS Association, Project ALS, the Michael W. McCarthy Foundation, and the Kinetics Foundation.