Differentiation of Human Embryonic Stem Cells to Dopaminergic Neurons in Serum-Free Suspension Culture



The use of human embryonic stem cells (hESCs) as a source of dopaminergic neurons for Parkinson's disease cell therapy will require the development of simple and reliable cell differentiation protocols. The use of cell cocultures, added extracellular signaling factors, or transgenic approaches to drive hESC differentiation could lead to additional regulatory as well as cell production delays for these therapies. Because the neuronal cell lineage seems to require limited or no signaling for its formation, we tested the ability of hESCs to differentiate to form dopamine-producing neurons in a simple serum-free suspension culture system. BG01 and BG03 hESCs were differentiated as suspension aggregates, and neural progenitors and neurons were detecz after 2–4 weeks. Plated neurons responded appropriately to electrophysiological cues. This differentiation was inhibited by early exposure to bone morphogenic protein (BMP)-4, but a pulse of BMP-4 from days 5 to 9 caused induction of peripheral neuronal differentiation. Real-time polymerase chain reaction and whole-mount immunocytochemistry demonstrated the expression of multiple markers of the midbrain dopaminergic phenotype in serum-free differentiations. Neurons expressing tyrosine hydroxylase (TH) were killed by 6-hydroxydopamine (6-OHDA), a neurotoxic catecholamine. Upon plating, these cells released dopamine and other catecholamines in response to K+ depolarization. Surviving TH+ neurons, derived from the cells differentiated in serum-free suspension cultures, were detected 8 weeks after transplantation into 6-OHDA–lesioned rat brains. This work suggests that hESCs can differentiate in simple serum-free suspension cultures to produce the large number of cells required for transplantation studies.


The use of human embryonic stem cells (hESCs) as the source of neural cells for transplantation therapies has several advantages [1]. hESCs are a source of many well-characterized human stem and progenitor cell types that correspond to the cells found in a developing embryo. Recent studies have documented important advances in the culture of hESCs for cell therapy. This work indicates that it will be possible to propagate normal and undifferentiated hESCs without feeder cells in highly defined and controlled culture conditions, allowing the generation of master cell banks to support cell transplants [24]. In addition to the ability to propagate and expand hESCs in well-defined conditions, it will be important to develop methods to differentiate the cells using simple and well-defined culture conditions. Ideally, differentiation should be carried out in a serum-free environment using approaches that can be easily scaled for the production of large numbers of differentiated cell types. The cells cultured in these conditions should respond to known developmental modulators in a way predicted from their normal function in vivo or shown in other differentiation studies of hESCs or nonhuman embryonic stem (ES) cells.

Previous studies have shown that grafts of fetal midbrain dopaminergic neurons could survive, reinnervate, and function in patients with Parkinson's disease (PD) and provide a proof of principal for this approach [5]. However, two double-blind controlled trials revealed significant issues that need to be addressed before larger trials are warranted [6, 7]. One concern is the difficulty in standardizing the fetal midbrain tissue used for implantation, which will be critical for consistent clinical outcomes. The generation of midbrain dopamine neurons from hESCs would provide cell populations that could be expanded, characterized, and standardized in vitro, providing optimal populations for studies in animal models of PD [8]. This approach would also provide a useful model for many aspects of human neurogenesis, including examination of the molecular and developmental controls of the midbrain lineage and functional analyses of their cellular and physiological characteristics. Several methods to generate midbrain dopaminergic neurons from mouse and primate ES cells currently exist, some of which have led to recovery of symptoms in rat models of PD after cell implantation [911]. Many of these rely on complex multistep protocols requiring the exposure of the cells to stromal cell lines [12, 13], multiple growth factors [9, 14], or expression of transgenes such as Nurr1 [11]. The utility of these methods to direct differentiation from hESCs has not yet been reported, and use of cocultures, added signaling factors, or transgenic approaches could lead to additional regulatory as well as cell production barriers that will complicate their use in eventual therapies and increase their cost.

We have previously reported effective neural differentiation of hESCs in serum-free conditions under the influence of the HepG2-conditioned medium MedII [15]. In this study we report the differentiation of hESCs to midbrain dopaminergic neurons in a simple serum-free suspension system. This occurred in the absence of added growth factors or neural-inducing agents, demonstrating that it was driven by signaling within suspension aggregates. We showed that this differentiation was initially inhibited by bone morphogenic protein (BMP)-4, but later BMP signaling induced peripheral neuronal differentiation. These effects of BMP-4 were the same as those previously observed in cultures of mouse or nonhuman primate ES cells, demonstrating that cell fates can be easily manipulated by the addition of exogenous factors in our culture system. The differentiated tyrosine hydroxylase–positive (TH+) neurons were susceptible to 6-hydroxy-dopamine (6-OHDA), plated cultures released dopamine and other catecholamines upon depolarization, and surviving TH+ neurons were detected 8 weeks after transplantation to the 6-OHDA–lesioned rat brain. Our approach represents a simple and potentially scalable platform for the large-scale derivation of dopaminergic neurons for studies in animal models of PD and the molecular, cellular, and physiological examination of this differentiation pathway.

Materials and Methods

Human Embryonic Stem Cell Culture

The NIH-registered BG01 and BG03 cell lines were used in this work (http://stemcells.nih.gov/index.asp). Microdissection-passaged hESCs were cultured and passaged as described [15, 16], whereas collagenase/trypsin-passaged hESCs were grown in 20% knockout serum replacer (KSR) human ES medium. This medium consisted of 50/50 Dulbecco's modified Eagle's medium (DMEM)/F12 (Invitrogen, Grand Island, NY, http://www.invitrogen.com) supplemented with 20% KSR (Invitrogen), ×1 nonessential amino acids (Invitrogen), 20 mM L-glutamine (Invitrogen), 0.5 U/ml penicillin/0.5 U/ml streptomycin (Invitrogen), 4 ng/ml fibroblast growth factor (FGF)-2 (Sigma, St. Louis, http://www.sigma-aldrich.com), and 0.1 mM β-mecaptoethanol (Sigma) with or without 10 ng/ml human leukemia inhibitory factor (LIF; Chemicon, Temecula, CA, http://www.chemicon.com), which did not noticeably affect the maintenance or differentiation of hESCs. Collagenase/trypsin hESCs were passaged by treatment with 1 mg/ml collagenase (Invitrogen) for 4 minutes, followed by 0.05% trypsin/EDTA (Invitrogen) for 30 seconds and trituration to single cells or small clumps. Fetal calf serum (FCS), 10% (HyClone, Logan, UT, http://www.hyclone.com) in DMEM/F12, was added to the hESC suspension, followed by centrifugation, aspiration, and resuspension in culture medium. hESCs were replated at 1.5 × 105 cells, on 1.2 × 106 mouse embryonic fibroblasts (MEFs), per 35-mm dish and proliferated to 0.85 to 1.0 × 106 hESCs after 3–4 days.

BG01 hESCs were also passaged as clumps using EDTA-free trypsin (Invitrogen). Cells were grown on 1.2 × 106 MEFs per 35-mm dish in 20% KSR human ES medium (without LIF) that had been conditioned on MEFs for 24 hours [4]. These cells were passaged by treating with 0.05% EDTA-free trypsin (Invitrogen) for 30 seconds, removing the feeder layer by pulling it off the plate with watchmakers forceps, scraping the adherent hESC colonies off the dish, and gently triturating with a P1000 pipette until the colonies were disaggregated to clumps of ∼10 to 100 cells. The trypsin was neutralized with 10% FCS in DMEM/F12, and the colony clumps were centrifuged and replated at a density to maintain more than 200 colonies per 35-mm dish.

Derivation of Collagenase/Trypsin-Passaged and SSEA-4–Enriched BG01 Cells

Undifferentiated BG01 hESCs were adapted to collagenase/trypsin passaging and enriched by magnetic sorting using an anti-stage specific embryonic antigen-4 (SSEA-4) antibody (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, http://www.uiowa.edu/∼dshbwww/) and the MACS separation system (Miltenyi Biotec, Auburn, CA, http://www.miltenyibiotec.com/) according to the manufacturer's instructions. Differentiated colonies were excised from a culture of microdissection-passaged BG01 hESCs. The culture was maintained for 5 to 10 passages using collagenase/trypsin disaggregation as described above, before SSEA-4 immunomagnetic enrichment. For enrichment, the cells were harvested enzymatically as described, and the enzymes were inactivated by adding 10% fetal bovine serum (FBS)/10% KSR human ES medium and passed through a cell strainer (Becton, Dickinson, Franklin Lakes, NJ, http://www.bd.com). For blocking, cells were pelleted and resuspended in staining buffer (SB) (5% FBS, 1 mM EDTA, 0.5 U/ml penicillin and 0.5 U/ml streptomycin, in Ca2+/Mg2+-free phosphate-buffered saline [PBS]). The cells were pelleted and resuspended in 1 ml primary anti-SSEA-4 antibody diluted 1:10 in SB and incubated at 4°C for 15 minutes. SB, 9 ml, was then added, and the cells were pelleted and washed (10 ml SB was added and repelleted). A total of 1 × 107 cells were resuspended in 80 μl SB and incubated with 20 μl magnetic goat anti-mouse immunoglobulin G (IgG) MicroBeads at 4°C for 10 minutes. SB, 1.9 ml, and then fluorescent-conjugated secondary antibody, 2 μl (Alexa-488 conjugated goat anti-mouse IgG [Molecular Probes, Eugene, OR, http://www.probes.com]), were added to enable fluorescent analysis of the separation. The sample was incubated for 5 minutes at 4°C and then brought to 10 ml with SB, pelleted, washed, resuspended in 500 μl SB, and applied to a separation column that had been equilibrated with 3 × 500 μl SB and prepositioned on the selection magnet. The flow-through and three washes with 500 μl SB were collected, which presumably contained a SSEA-4 population. The column was removed from the magnet, 500 μl SB was forced through with a plunger, and the presumed SSEA-4+ cell population was collected in a 15-ml tube. A total of 9.5 ml 20% KSR human ES medium was added, and the cells were pelleted and resuspended in 1 ml of the same medium. A total of 105 SSEA-4–enriched hESCs were plated on MEFs on 35-mm dishes, and the cells were maintained in 20% KSR ES medium and passaged withcollagenase/trypsin as described above.

To examine the effectiveness of the enrichment, aliquots of the starting population, the flow/wash sample, and SSEA-4–enriched sample were analyzed by flow cytometry. Typically, 85% of the cells in the starting hESC populations were SSEA-4+, which was enriched to >99% SSEA-4+ cells after immunomagnetic selection. The nonretained flow through exhibited ∼60% SSEA-4+ cells. A secondary antibody alone as negative control exhibited background staining on only 0.5% of cells.

Neural Differentiation of hESCs

Trypsin/collagenase-passaged cultures were treated with collagenase, and whole hESC colonies were removed from the feeder layer using a fire-drawn Pasteur pipette needle, washed with DMEM, and placed in suspension culture in differentiation medium. Microdissection-passaged BG01 and BG03 cultures were harvested for differentiation by excising whole colonies using glass needles. The differentiation media used were either a MedII/FGF2 medium (DMEM/F12, 1 × N2 [Invitrogen], 20 mM L-glutamine, 0.5 U/ml penicillin, 0.5 U/ml streptomycin, 4 ng/ml FGF-2, and 50% serum-free MedII) or a DMEM/N2 medium (DMEM, 1 × N2, 20 mM L-glutamine, 0.5 U/ml penicillin, 0.5 U/ml streptomycin). MedII was made as described previously [17], except the base medium used for conditioning was DMEM/N2 medium (above). Cultures were differentiated for 2–6 weeks in suspension, and the media changed every 5–7 days. For adherent culture, differentiated aggregates were cut into pieces with glass needles or razor blades and were plated on dishes or Permanox slides coated with 20 μg/ml polyornithine (Sigma) and 1 μg/ml laminin (Sigma) in MedII/FGF2 medium or Neurobasal medium (Invitrogen) containing 1 × B27 (Invitrogen), 5% FCS (Hyclone), 2 ng/ml glial-derived neurotrophic factor (GDNF) (R&D Systems, Minneapolis, http://www.rndsystems.com/), 10 ng/ml brain-derived neurotrophic factor (BDNF) (R&D Systems), 20 mM L-glutamine, 0.5 U/ml penicillin, and 0.5 U/ml streptomycin.

Immunostaining and Histochemistry

Whole-mount immunostaining of cell aggregates was performed in 15-ml tubes using 200- to 500-μl volumes for antibody binding and 2- to 5-ml volumes for washes with 1× PBS. Immunostaining of adherent cells used the same solutions. Cultures were fixed in 4% paraformaldehyde (PFA) (Fisher Scientific, Hampton, NH, http://www.fisherscientific.com) and 4% sucrose (Sigma) in 1× PBS. Samples were blocked with 3% goat serum (Invitrogen), 1% polyvinyl pyrolidone (Sigma), and 0.3% Triton X-100 (Sigma) in 1× PBS (block buffer; Triton X-100 was omitted for cell-surface immunostaining) and then incubated with primary antibody diluted in block buffer for 1–2 hours at room temperature. Samples were then washed and incubated for 1–2 hours in secondary antibodies diluted 1:1,000 in block buffer, followed by washing. The secondary antibodies were goat anti-rabbit, anti-sheep, anti-rat, or anti-mouse antibodies (Molecular Probes) conjugated with Alexa-350 (blue), 488 (green), 568 (red), or 647 (far red). Nuclei were stained with 5 ng/ml 4′,6′-diamidino-2-phenylindole (DAPI; Sigma). Whole-mount suspension immunostainings were mounted on glass slides and gently flattened with a coverslip to enable visualization. Individual color channels were captured separately with a Q Imaging digital camera on a NIKON E1000 or TE 2000E microscope and merged in Adobe Photoshop. Confocal and 2-photon confocal imaging was performed using a Leica TCS SP2 Spectral Confocal Microscope. Negative controls using secondary antibody alone did not exhibit staining. The primary antibodies (supplier, catalog number, and dilution) used were microtubule-associated protein 2 (MAP2) (Sigma, M4403, 1:500), Nestin (Chemicon, AB5922, 1/200), Nestin (Chemicon, MAB5326, 1:200), Vimentin (Chemicon, CBL202, 1:200), OCT-4 (Santa Cruz Biotech, Santa Cruz, CA, http://www.scbt.com; sc-5279, 1:100), βIII tubulin (Sigma, T8660, 1:500), Neurofilament H (Sternberger Monoclonals, Lutherville, MD, http://home.att.net/∼sternbmonoc/home.htm; SMI32, 1:500), HuC/D (Molecular Probes, A-21271, 1:500), TH (Pel-Freez Biologicals, Rogers, AR, http://www.pelfreez-bio.com; P60101-0, 1:100), TH (Pel-Freez, P40101, 1:250), phospho-TH(Ser40) (Cell Signaling Technologies, Beverly, MA, http://www.cellsignal.com; 2791, 1:250), dopamine transporter (DAT) (Chemicon, MAB 369, 1:50), aromatic amino acid decarboxylase (AADC) (Pel-Freez, P40401-0, 1:200), Synapsin (Chemicon, MAB355, 1:100), Synaptophysin (Chemicon, MAB5258-20UG, 1:250), Tau (Chemicon, MAB361, 1:200), vesicular monoamine transporter 2 (VMAT2) (Chemicon, AB1767, 1:500), SSEA-1 (DSHB, MC-480, 1:5), SSEA-3 (DSHB, MC-631, 1:5), SSEA-4 (DSHB, MC-813-70, 1:5), Tra-1-60 (Chemicon, MAB4360, 1:100), Tra-1-81 (Chemicon, MAB4381, 1:100), glial fibrillary acidic protein (GFAP; Sternberger, SMI21, 1:100), dopamine β-hydroxylase (DβH) (Chemicon, AB1536 1:100), and Peripherin (Chemicon, AB1530, 1:100). The SSEA-1, -3, -4 monoclonal antibodies developed by Davor Solter and Barbara Knowles were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences. Embedding of suspension aggregates, sectioning, staining, and counting of DAPI-stained nuclei were performed as described [15].

6-Hydroxydopamine Treatment of Differentiated Aggregates

6-OHDA experiments were performed on cell aggregates differentiated for 1 month in DMEM/N2. 6-OHDA (Sigma) was prepared in 0.2 mg/ml ascorbic acid (Sigma) with 1× PBS, and DMEM/N2 aggregates were exposed to 0.2 mg/ml ascorbic acid (negative control), 10 mM or 1 mM 6-OHDA, or 10 mM and 1 mM 6-OHDA plus 100 mM dopamine (Sigma) for 10 minutes. Aggregates were washed extensively and incubated in MedII/FGF2 medium for a 5-hour recovery. Aggregates were fixed and stained as whole-mount preparations for βIII tubulin and TH.

Focused cDNA Array

GEArray Q series cDNA array filters (HS-601; SuperArray, Frederick, MD, http://www.superarray.com) were probed nonradioactively with biotin dUTP-labeled cDNA, according to the manufacturer's protocol. Total RNA was prepared from BG01 suspension aggregates differentiated in DMEM/N2 for 6 weeks using the Trizol reagent (Invitrogen), and 4 μg RNA was used to make a labeled cDNA probe for each filter. Hybridizations were detected by chemiluminescence and exposure to x-ray film.


Electrophysiology was performed as described previously [18]. Whole-cell recordings were made from cells with neuronal morphology (visible neurites) on the stage of an inverted phase-contrast microscope using standard electro-physiological techniques using a potassium gluconate–based internal solution. Glutamate was applied via a large-bore pipette positioned immediately in front of the cell under study, which was continuously perfused with a physiological saline.

Reverse Transcription–Polymerase Chain Reaction and Real-Time Reverse Transcription–Polymerase Chain Reaction

Primers and probes used for polymerase chain reaction (PCR) are listed in Table 1. RNA was isolated with the Trizol reagent (Invitrogen) and treated with DNase I (Promega, Madison, WI, http://www.promega.com). First-strand cDNA was generated using a Superscript first-strand synthesis kit (Invitrogen), according to the manufacturer's protocols. A total of 2.5 μg DNase I treated RNA was used in each cDNA synthesis, in a total volume of 60 μl. The synthesis reaction was heat inactivated and diluted to 200 μl, such that 5 μl of template, or the equivalent of 62.5 ng RNA, was used in each 25-μl PCR reaction. Real-time PCR was performed in triplicate using master mixes (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com) for the TaqMan system or SYBR green incorporation and an ABI Prism 7700 detector. Mock reverse transcriptase minus cDNAs were used as negative controls for each primer set and were all negative. The thermal parameters were 50°C for 2 minutes and 94°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. The specificity of amplification of products detected with SYBR green was demonstrated by melting curve analyses as well as digestion at internal restriction sites and electrophoresis. Standard curves were used to determine the amplification efficiency of each primer set, and the REST software (http://www.gene-quantification.info/) [19] was used to determine relative gene expression from cycle crossing point data and statistical significance using a pair-wise fixed reallocation randomization test. These comparisons factor in primer efficiencies and normalization to parallel glyceraldehyde-3-phosphate dehydrogenase (GAPDH) reactions.

Table Table 1.. Primers used in reverse transcriptase–polymerase chain reaction
  1. a

    Abbreviations: AADC, aromatic amino acid decarboxylase; DAT, dopamine transporter; DβH, dopamine β-hydroxylase; TH, tyrosine hydroxylase; VMAT2, vesicular monoamine transporter 2.

GenePrimers F (top), R (bottom)Probe, or size, restriction site, cut products
SOX1CACAACTCGGAGATCAGCAA171 bp, BssHII digestion: 92,78
NURR1CCCAGTGGAGGGTAAACTCA151 bp, EcoRI digestion: 94,56
AADCCTCGGACCAAAGTGATCCAT212 bp, SacI digestion:45,166
GIRK2AGCAAGGTTTCTGGTGCCTA129 bp, BamHI digestion: 63,65
DβHGCAGATCTCGTGGTGCTCT146 bp, BamHI digestion: 83,62
CHATCGTGGACAACATCAGATCG147 bp, HindIII digestion: 88,58
VGLUT1ACCCTGCTCCTCTCTGTCCT142 bp, PstI digestion: 109,32
VGLUT2GCGTCAAGTAGAGGCGACAT149 bp, EcoRV digestion: 48,100
VGLUT3GAGCTGCGCTCAGTTGATAA133 bp, HincII digestin: 55,77
GAD67TTTGTGAGCCAAAGAGAAAAGA150 bp, EcoRI digestion: 97,52
GFAPTCATCGCTCAGGAGGTCCTT353 bp (Vercovi, Exp Neurol(99) 156:71–83)

The following standard conditions were used with endpoint reverse transcription (RT)-PCR analysis: 25-μl reactions using Taq DNA polymerase (Invitrogen) and cDNA prepared as above; a first denaturation step of 95°C for 1 minute, followed by 35-cycle reactions of 95°C for 30 seconds, 54°C for 1 minute, and 72°C for 2 minutes; agarose electrophoresis; and detection with SYBR green staining. For the LMX1B expression comparison, captured electrophoretic images were compared and normalized to GAPDH using ImageJ version 1.31 (http://rsb.info.nih.gov/ij/upgrade/).

Evoked Release of Dopamine and HPLC

Adherent cultures were depolarized with 56 mM KCl/Hanks' balanced salt solution (HBSS) for 15 minutes or HBSS as a negative control, and metabisulfite and orthophosphoric acid were added to stabilize the samples [20]. HPLC detection of dopamine was performed at the Neurochemistry Analytical Core Laboratory, John F. Kennedy Center, Vanderbilt University, Nashville, TN (http://www.mc.vanderbilt.edu/root/vumc.php?site=neurosci&doc=697). Briefly, samples were mixed with the internal standard dihydroxybenzylamine, and catecholamines were extracted by adsorption to solid Al203, washed, and deadsorbed with 0.1 N acetic acid. Samples were injected into a HPLC system consisting of a Waters Model 515 pump, Waters 717+ Autosampler, and an Antec Electrochemical Detector. A calibration curve run along with the unknown samples was used to calibrate the instrument.

Transplantation into 6-OHDA–Lesioned Rats

Adult Sprague-Dawley rats (Harlan, IN) were lesioned unilaterally by injection of 4 μl of 6-OHDA (Sigma; 2 mg/ml in 0.2 mg/ml ascorbic acid [Sigma]/PBS) over 4 minutes into the left medial forebrain bundle (coordinates: AP, −4.3 mm; L, −1.5 mm; D, −8.8 mm; with a 1-mm correction for Dura depth). Lesioning was verified by assessing the rotational response of the animals to amphetamine. Two and 4 weeks after lesioning, a subcutaneous injection of 5 mg/kg amphetamine (Sigma) was administered and rotations were assessed in an automated rotometer (AccuScan, Columbus, OH). Rats showing significant ipsilateral rotations (>3 rpm) were used for implantations. Differentiated suspension aggregates were dissected into pieces (∼103 to 2 × 103 cells per piece), and 1 to 10 pieces were implanted in a 1- to 5-μl volume over 5 minutes per rat using a Kopf Stereotaxic frame and Hamilton syringe. Cell clumps were implanted into the lesioned striatum using the following injection coordinates: AP, +0.9; L, −2.7; D, −6.0. The rats were given daily injections of cyclosporin A (10 mg/kg) starting on the day before cell implantation and for a period of 8 weeks before euthanasia and collection of whole brains by cardiac perfusion with 4% PFA. Animal research protocols (#A2002-10120-0) were reviewed and approved by the University of Georgia, Athens, GA, and experiments were conducted according to institutional guidelines. A coarse section of the fixed rat brain encompassing the injection site was isolated using a razor blade and brain matrix, dehydrated through graded alcohols, and permeabilized with dimethylsulfoxide before embedding in a 3:1 mixture of polyethylene glycol 1,450:1,000. Microtome sections of the brain 12- and 20-μm thick were obtained, and hematoxylin and eosin staining was used to locate the implant site. Verification that human cells were found used a modified in situ hybridization detection method [21] with two Biotin-tagged oligonucleotides to the human genome-specific Alu repeat sequence. Immunohistochemical characterization of the surviving implanted cells was performed using a sequential free-floating protocol, and sections were transferred among standard permeabilization, blocking, and antibody-containing solutions in a multiwell tray format with watchmakers forceps. Fluorescently conjugated secondary antibodies were detected using epifluorescence microscopy, whereas horseradish peroxidase–conjugated secondary antibodies were visualized with a diaminobenzidine chromogenic reaction.


Collagenase/Trypsin Passaging and SSEA-4 Enrichment of hESCs

The BG01 and BG03 hESC lines [16] were used in this study and are listed on the NIH registry. Until the time of this study, these cells had been maintained exclusively by manual microdissection of individual undifferentiated colonies (microdissection passaging). Because of the ability to selectively passage morphologically undifferentiated cells, microdissection passaging is currently the most appropriate method to maintain long-term cultures of undifferentiated hESCs and may contribute to the maintenance of a normal karyotype [22]. However, this approach is laborious, and scaling up cultures for experiments is difficult. Therefore, we tested several enzymatic cell dissociation methods for maintaining and expanding BG01 cells. After cell dissociation with collagenase and trypsin, undifferentiated BG01 cells were enriched by immunomagnetic-bead cell sorting using a monoclonal antibody against SSEA-4, a cell-surface antigen that is robustly expressed on pluripotent hESCs [23, 24]. Flow cytometric analysis of a representative experiment detected SSEA-4 expression on 85% of the starting population of cells, whereas 99.2% of the cells expressed SSEA-4 after immunomagnetic enrichment, with 60.7% of cells in the flow-through being SSEA-4+. Cultures enriched for SSEA-4–expressing cells (Fig. 1A) grew as colonies that strikingly resembled those of mouse ES cells and other hESC lines passaged with trypsin [25] and exhibited the characteristic profile of the following pluripotent markers expressed by hESCs: SSEA-1, SSEA-3+, SSEA-4+, Tra-1-60+ (Figs. 1B–1E, respectively), Tra-1-81+ (not shown), and OCT-4+ (Fig. 1F). In addition, aggregates of SSEA-4+ cells allowed to differentiate in a serum-containing medium formed cells expressing ectodermal (Nestin, Sox1), endodermal (Amylase, AFP), and mesodermal (Cardiac actin) markers, suggesting that the SSEA-4–enriched cells could form lineages of the three embryonic germ layers (not shown) and maintained their pluripotency. These hESCs also expressed the neural progenitor markers nestin (Fig. 1F) and vimentin (Fig. 1G). Expression of vimentin has been detected in the H1 hESC line by RT-PCR and immunocytochemistry [26, 27], whereas RT-PCR has detected nestin expression in some lines but not others [2729]. It is possible that nestin is expressed in at least a subset of the cells within most other established hESC lines.

Figure Figure 1..

Culture and neural differentiation of hESCs. (A): Collagenase/trypsin-passaged and SSEA-4–enriched BG01 hESCs. BG01 cells were SSEA-1(B), SSEA-3+(C), SSEA-4+(D), Tra-1-60+(E), OCT-4+ and Nestin+(F), and Vimentin+(G). (H): βIII tubulin and (I) TH immunostaining of plated MedII/FGF2 differentiations. (J): Merged image of (H, I). TH (K), vesicular monoamine transporter 2 (L), and merged (M) immunostainings of plated MedII/FGF2+ differentiations show cell body staining. Scale bars = 100 μm (A, H–J) and 50 μm (B–G, K–M). Abbreviations: DAPI, 4′,6′-diamidino-2-phenylindole; FGF2, fibroblast growth factor 2; hESC, human embryonic stem cell; TH, tyrosine hydroxylase.

To produce the numbers of cells required for these studies in a timely fashion, we used cultures of BG01 cells maintained by collagenase/trypsin dissociation and SSEA-4 enrichment unless otherwise noted. Because of the possibility of accumulating aneuploidies as well as spontaneously differentiated cells in enzymatically passaged BG01 cultures, these cells were not used beyond approximately 20 passages after SSEA-4 enrichment or 30 total passages with collagenase/trypsin. Chromosome counting indicated that under these culture conditions, up to 50% of cells had an abnormal karyotype after a total of 33 passages with collagenase/trypsin (I. Nasonkin, unpublished data). Key experiments (derivation and proliferation of neural progenitors in DMEM/N2, the generation of large networks of TH+ cells but rare DβH+ cells in suspension aggregates, and evoked release of dopamine) were confirmed using karyotypically normal BG01 and BG03 cells maintained by passaging as clumps of cells, either with microdissection passaging or disaggregation using EDTA-free trypsin.

Neural Differentiation of hESCs in a Serum-Free Minimal Medium

A summary of the neural and dopaminergic differentiation observed in these experiments is outlined in Table 2. We performed differentiation experiments using variations of two basic conditions: 50% MedII-conditioned medium plus FGF2 (DMEM/F12+N2+MedII+FGF2) and DMEM plus N2 supplement (minimal medium). Experiments were typically analyzed after 1 month in suspension, and both of these conditions supported the differentiation of large networks of TH+ neurons. Because initial survival of cell aggregates was lower in DMEM/N2 conditions, we also derived cell aggregates into MedII/FGF2 for 3–5 days, followed by 1 month in minimal medium, which also generated large networks of TH+ neurons. We used minimal DMEM/N2 conditions as a base to assess the role of additional factors on neural differentiation. Finally, for some analyses, differentiated aggregates were plated to adherent culture for approximately 1–2 weeks in either MedII/FGF2 or Neurobasal medium supplemented with B27, serum, BDNF, and GDNF, because minimal DMEM/N2 conditions did not support effective attachment of differentiated aggregates to adherent culture.

Table Table 2.. Summary of neural differentiation experiments
  • a

    aTypical 1-month suspension differentiations.

  • b

    bInitial high cell death and survival of aggregates larger than approximately 150 μm.

  • c

    cMinimal conditions seem to support neural precursor and neuronal differentiation at the expense of other cell types.

  • d

    dHigher degree of nonneural differentiation such as cysts.

  • e

    eRare TH+ cells by immunostaining.

  • f

    fHigh proportion of peripherin+ neurons indicative of neural crest–derived peripheral differentiation.

  • g

    Abbreviations: BMP, bone morphogenic protein; DMEM, Dulbecco's modified Eagle's medium; FGF2; fibroblast growth factor 2; TH, tyrosine hydroxylase.

TreatmentaGrowthNeural differentiationTH+neurons
MedII/FGF2 for 3–5 days, then DMEM/N2++++++High
DMEM/N2 + serum+++++dRaree
DMEM/N2 + BMP4PoorPoor
DMEM/N2 + BMP4 + serum+++PoorRare
DMEM/N2 + BMP4 days 5–9++++++Highf

We initially tested the ability of the collagenase/trypsin–passaged and SSEA-4–enriched BG01 cells to differentiate in serum-free conditions in MedII/FGF2 medium [17]. MedII-conditioned medium has been shown previously to promote neural differentiation from mouse, rhesus monkey, and human embryonic stem cells [15, 18, 30]. Whole hESC colonies were removed from the feeder layer and cultured in suspension. Characteristic folds and rosettes of neural precursors were observed after 5–10 days of culture, as observed in differentiations performed from microdissection-passaged hESCs [15]. Cell aggregates were plated on polyornithine/laminin-coated chamber slides 2 or more weeks after derivation and cultured for an additional 5–7 days before immunostaining. Stained cultures were highly enriched for nestin+ neural precursor rosettes and large networks of βIII tubulin+ (Fig. 1H) neurons. Most of these neurons also expressed TH (Figs. 1I, 1J). Scoring of isolated βIII tubulin–expressing neurites in merged images showed that approximately 75% (69 of 90, n = 5 fields) were TH+/βIII tubulin+. This was strikingly different from our previous differentiations from microdissection-passaged hESCs [15], in which suspension cultures were plated after only approximately 1 week of culture and previous reports [29, 31, 32], in which TH+ neurons were rare. MedII seemed to enhance rather than induce neuronal differentiation, because significant differentiation to TH+ and VMAT2+ neurons also occurred in the same medium without added MedII (Figs. 1K–1M).

The ability of BG01 cell aggregates to differentiate into neurons in serum-free suspension culture led us to test the role of the added MedII and FGF2 in promoting early neural lineage formation. In these experiments, hESC aggregates were cultured in DMEM/N2. Unlike FGF2/MedII differentiations, aggregates incubated in DMEM/N2 exhibited a very high level of obvious cell death through their first approximately 2 weeks, indicating that MedII/FGF2 contributed significantly to cell survival. This was consistent with our previous results, indicating that MedII provided a cell survival/proliferation activity rather than a neural inducing factor [15]. Only hESC aggregates that were initially larger than approximately 150 μm were viable and proliferated in the minimal medium, suggesting a community effect in the delivery of essential growth factors and signaling within differentiating aggregates. After differentiation for 2 weeks in DMEM/N2, aggregates seemed to be comprised largely of neural precursor rosettes/neurectoderm structures (Fig. 2A). As suspension aggregates were cultured further, there appeared to be a gradual loss of this distinct morphology, from approximately 2–4 weeks, possibly indicating a shift away from neural progenitor proliferation to neuronal differentiation (not shown). However, persistence of neural precursor rosettes could be detected even after 4 weeks of differentiation. Sectioning, followed by toluidine blue or DAPI staining (Fig. 2B), demonstrated that at 2 weeks, cell aggregates cultured in DMEM/N2 were comprised of distinctly organized regions of neural precursor rosettes and nonrosette regions. Counts of DAPI-stained nuclei (Fig. 2B, inset) indicated that rosette neural progenitor structures comprised 39.4 ± 12.3% (14334/36663 nuclei, n = 11 sections) of the cells. The nonrosette regions were demonstrated by whole-mount analysis and counting of anti-HuC/D and DAPI-stained overlayed images to contain 45.5 ± 7.2% HuC/D+ early postmitotic neurons (445/984 cells, n = 5 fields). DMEM/N2 aggregates were also dense with βIII tubulin+ neuronal extensions (Fig. 2C) and TH+ neurons (Fig. 3A). The rosettes exhibited a characteristic structure with a core of tightly packed proliferating neural precursor cells, of which 7.3 ± 4.4% (30 of 376, n = 7 fields) exhibited condensed mitotic chromosomes when DAPI-stained nuclei in 1-μm confocal optical sections were counted. Rosette cells expressed nestin (Figs. 2E, 2G) and vimentin (Figs. 2F, 2G), whereas expression of HuC/D, a marker of early postmitotic neurons [33], was first observed in the differentiating cells surrounding the rosettes (Figs. 2H–2J). Double immunostaining of plated cultures with HuC/D and phospho-HistoneH3, a mitotic marker [34], was used to confirm the postmitotic status of the neurons associated with neural precursor rosettes, with no phospho-H3+/HuC/D+ cells being observed from >500 counted HuC/D+ cells (Figs. 2K–2M). Expression of synapsin (Fig. 2N) and synaptophysin (Fig. 2O) was detected in plated neurites, suggesting the formation of synaptic complexes. In addition to the analysis of the cell aggregates with immunocytochemistry, the expression of general neuron markers as well as markers of neurotransmitter phenotypes was determined by RT-PCR analysis and a focused microarray screen. We analyzed gene expression in BG01 DMEM/N2 suspension aggregates after 6 weeks of differentiation using a focused array of 266 human genes, selected to represent different human stem cell populations [35, 36]. We compared gene expression in hESCs and in differentiated aggregates (Fig. 4) and found 14 transcripts that were upregulated in the differentiated cells. Many of these genes have known or presumed function during neural development and differentiation, including BMP signaling (BMPR2), FGF signaling (FGF11, FGFR1, FGFR2), WNT signaling (FZD3), neurogenic functions (CXCR4, DLK1, VEGF), and neurotrophin signaling (NTRK2). Of the 11 SOX-family transcription factors present on the array, only SOX1, 2, 3, and 4, which exhibit neural tube/progenitor expression or function, were detected. Common markers of neuronal cell function were also upregulated such as neurofilaments (INA, NEFL), MAP2, and NCAM1. The expression of FGF11 confirmed that the differentiated aggregates contained neuronal progenitors. In a previous analysis of rat central nervous system (CNS) progenitors, it was found that FGF11 expression was activated after neuronal precursors appeared within the CNS, and cell sorting of the progenitors showed FGF11 expression exclusively within the E-NCAM+ neuronal progenitor population [37]. In addition, the focused array contained more than 22 markers of differentiated nonneural lineages representing endoderm, mesoderm, and nonneural ectoderm. Expression of most of these markers (20 of 22) was not detected (Fig. 4), confirming enriched neural differentiation in these aggregates. A previous characterization of the sensitivity of similar focused microarrays showed a 96% correspondence between the results of the arrays and RT-PCR analysis [37]. This shows that these focused micro-arrays are quite sensitive because of the use of gene-specific primers in making the cDNA probe. The overall pattern of expression in BG01 hESCs using this array was similar to that reported previously [35]. Transcripts that were upregulated in hESCs were CER1, FGF2, DNMT3B, FOXM1, FZD7, ITGA6, PDGFA, POU5F1, and TERF1. RT-PCR analysis of DMEM/N2 suspension aggregates at 4 weeks detected expression of choline acetyltransferase, vesicular glutamate transporters 1, 2, and 3, and the vesicular inhibitory amino acid transporter (not shown), which are markers of cholinergic, glutaminergic, and GABAergic/glycinergic neurons, respectively. The expression of GAD67 was not detected by immunostaining or RT-PCR analysis, suggesting that few γ-aminobutyric acid (GABA)–producing neurons were present. The capacity for glial differentiation was demonstrated by the expression of GFAP (Fig. 2P, inset). This analysis suggested that a range of neural lineages could be generated in this system.

Figure Figure 2..

Neural differentiation of DMEM/N2 suspension aggregates. (A): Suspension DMEM/N2 aggregates after 2 weeks. Inset shows a higher magnification indicating a central cavity (*), surrounded by the radial organization of the neuroepithelia. (B): Three-μm plastic section of DMEM/N2 aggregates stained with toluidine blue or DAPI (inset), showing neural precursor rosettes (arrows) and nonrosette regions (*). (C): Whole-mount βIII tubulin immunostaining and 1-μm confocal optical section of suspension DMEM/N2 aggregates. Rosette area is indicated (*). (D): Whole-mount DAPI staining and 2-photon, 1-μm optical confocal section of DMEM/N2 aggregates. Condensed chromosomes in the core regions of rosettes (arrows) and two adjacent rosettes (dashed lines) are indicated. (E): Nestin expression in neural rosette cells. The arc of a rosette is indicated (dashed lines). (F, G): Coexpression of Vimentin and nestin in neural rosette cells. (H–J):Whole-mount HuC/D immunostaining and DAPI staining of DMEM/N2 aggregates. (H): DAPI staining (in grayscale) showing a neural precursor rosette (dashed oval). (I, J): HuC/D was expressed in the cells immediately surrounding the rosette structures. (K–M): Immunostaining of plated DMEM/N2 aggregates demonstrated that rosette-associated (dashed oval) early neurons were postmitotic, with no double-positive (K) phospho-HistoneH3 and (L) HuC/D cells detected (M). Synapsin (N), synaptophysin (Synapt.) (O), and GFAP (P) expression in plated cultures is shown. (P): Inset, reverse transcription–polymerase chain reaction of GFAP expression in DMEM/N2 suspension aggregates. (Q, R): DMEM/N2 differentiation-derived neurons plated in MedII/FGF2 possess the physiological characteristics of central nervous system neurons. (Q): Leak-subtracted current (I) traces evoked by a family of increasingly depolarizing voltage (V) commands (–50, −30, −10, +10 mV) from a holding potential of −70 mV are shown superimposed. Inward and outward currents characteristic of sodium and delayed-rectifier potassium currents were evoked in 9 of 10 cells. (R): Inward membrane current and an increase in noise evoked by application of 1 mM glutamate (indicated by horizontal bar); holding potential was −70 mV. Similar currents were evoked in 10 of 10 cells. Scale bars = 100 μm (A, B, C, N), 50 μm (B inset, H–J, K–M), and 25 μm (D, E, F, G, P, O). Abbreviations: DAPI, 4′,6′-diamidino-2-phenylindole; DMEM, Dulbecco's modified Eagle's medium; GFAP, glial fibrillary acidic protein.

Figure Figure 3..

BMP-4 and serum affect neural and dopaminergic differentiation. BG01 hESC aggregates were differentiated under different conditions and examined after 1 month. (A–C): BMP-4 inhibits neuronal differentiation of hESCs. (A): DMEM/N2 aggregates and parallel cultures containing (B) 10 ng/ml BMP-4 or (C) 10 ng/ml BMP-4 and 10% fetal calf serum generated 180, 18, and ∼300 viable aggregates 11 days after derivation, respectively. Aggregates were immunostained with βIII tubulin and TH, demonstrating dense neuronal networks (A) and nearly complete inhibition of neuronal differentiation (B). An example of maximal neuronal differentiation in +BMP-4 conditions is shown, with other aggregates exhibiting no neurons. (C): Recovery and enhanced overall aggregate viability, but neuronal differentiation was not restored. (D–F): Addition of BMP-4 to DMEM/N2 differentiations from days 5 through 9 induced peripheral neuronal differentiation. (D): Rare peripherin+ cells in DMEM/N2 differentiations. (E): High proportions of βIII tubulin+/TH+ cells generated with d5-9 BMP-4 treatment, but (F) large proportions of peripherin+ cells were induced. (G–I): Serum inhibits dopaminergic differentiation. Differentiations in 10% serum showed an increase in nonneural differentiation (G, H) but still generated a large number of βIII tubulin+ neurons. Only rare TH+ neurons were observed. (I): RT–polymerase chain reaction for LMX1B expression in DMEM/N2 and 10% serum conditions. Scale bars = 100 μm (A–C, G) and 50 μm (D–F, H). Abbreviations: DMEM, Dulbecco's modified Eagle's medium; hESC, human embryonic stem cell; RT, reverse transcription; TH, tyrosine hydroxylase.

Figure Figure 4..

Focused array of gene expression in BG01 hESC and DMEM/N2 suspension aggregates differentiated for 6 weeks. The spots and names of the transcripts that were upregulated in each condition are indicated (arrows). The bottom row shows the indicated control cDNAs. Marker expression in DMEM/N2 differentiations: SOX genes expressed: SOX1,2,3 (neural tube/progenitors), SOX4 (differentiating neural progenitors, heart, B cells). SOX genes not detected: SOX5,6,10,13,15,17,18 (chondroblasts, neural crest, kidney, ovary, embryonic artery, testis, definitive endoderm, heart). Markers of nonneural lineages that were not detected: AFP, MYH11, CDH3,5,15, FABP4,6, GATA4, GCG, INSRR, ISL1, KRT8,14,15,17, MYH6, MYL4, NKX2.5, PECAM1, TNC (yolk sack, liver, smooth muscle, placenta, mammary gland, vascular, endothelia, muscle, adipose, enterocyte, heart, gut, epithelia, pancreas, kidney, islet, platelets, endothelial cells, mesenchyme, cartilage, bone). Abbreviations: DMEM, Dulbecco's modified Eagle's medium; hESC, human embryonic stem cell.

To physiologically verify the phenotype of hESC-derived neurons, whole-cell voltage-clamp recordings were made from DMEM/N2 differentiations plated to adherent culture in MedII/FGF2 medium. Depolarizing voltage commands from a negative holding potential evoked rapid inward sodium currents and delayed outward potassium currents (n = 9 of 10 cells; Fig. 2Q). Application of the excitatory and inhibitory neurotransmitters glutamate (Fig. 2R) and GABA (not shown) evoked rapidly desensitizing membrane currents consistent with the expression of ionotropic glutamate and GABA receptors (n = 10). Therefore, these neurons expressed the voltage- and ligand-gated ion channels that would allow them to generate action potentials and receive synaptic information.

Early Exposure to BMP-4 Antagonizes Neuronal Differentiation and Later Exposure Induces Peripheral Neurons

To demonstrate that the cell aggregates cultured in minimal medium would respond to extracellular factors, we tested the effect on neural differentiation of early or late exposure to BMP-4. We tested the ability of BMP signaling to antagonize the formation of neuronal lineages in hESC aggregates cultured in minimal medium. BMPs are a potent inhibitor of neural development and are known to induce nonneural ectoderm at the expense of neural ectoderm [3840]. We performed differentiations from parallel dishes of BG01 hESCs using three conditions: DMEM/N2 medium alone, DMEM/N2 + BMP-4, and DMEM/N2 + BMP-4 + FCS. Addition of 10 ng/ml BMP-4 to the DMEM/N2 minimal medium led to an approximately 10-fold reduction in aggregate viability and nearly completely blocked the formation of βIII tubulin+ and βIII tubulin+/TH+ neurons compared with aggregates in DMEM/N2 (Figs. 3A, 3B). Addition of serum to BMP-4–containing differentiations improved aggregate viability but did not restore the neural differentiation observed in DMEM/N2 conditions (Fig. 3C). These observations suggest that BMP-4 blocked neural lineage formation from the hESCs and instead stimulated the formation of nonneural serum-dependent cell types when cells were exposed to BMP-4 from day 1 of the differentiation. This was consistent with the known role of BMP-4 as an antagonist of neural lineage formation in Xenopus embryos and mouse ES cells [40]. In later stages of neural development, BMP signals induce the formation of neural crest cells from the dorsal crest of the neuroepithelium [4143]. To examine the response of hESC differentiations to a later BMP signal, we added 10 ng/ml BMP-4 to DMEM/N2 differentiations from days 5 through 9, followed by culture in DMEM/N2 until 1 month after derivation. Unlike an early BMP signal, late exposure to BMP-4 did not affect the viability of aggregates. Whole-mount immunostaining using antibodies to TH, βIII tubulin, and peripherin, a marker of neural crest–derived peripheral neurons [44, 45], detected a high proportion of βIII tubulin+/TH+ cells (Fig. 3E) but also a large number of peripherin+ cells (Fig. 3F), indicating the presence of neural crest–derived neurons. In contrast, only rare peripherin+ neurons were found in aggregates differentiated in DMEM/N2 (Fig. 3D), demonstrating that most of the βIII tubulin+/TH+ neurons represented a neural tube/CNS lineage.

We also examined the effect that addition of serum would have on differentiation within this system. In aggregates differentiated in DMEM/N2 plus 10% serum, large networks of βIII tubulin+ neurons could still be detected after 1 month despite an increased amount of nonneural differentiation, such as cysts, compared with aggregates in DMEM/N2. The proportion of TH+ neurons was greatly reduced in DMEM/N2 plus serum compared with DMEM/N2 (Figs. 3G, 3H). This indicated that although effective neuronal differentiation was possible in serum, factors present in these conditions may inhibit presumptive dopaminergic differentiation. Consistent with this, the midbrain dopaminergic marker LMX1B [46, 47] was expressed at elevated levels in DMEM/N2 compared with serum-containing conditions (Fig. 3I).

Neurons in the Cell Aggregates Express Multiple Markers Characteristic of Dopamine Neural Precursors and Neurons

Because large networks of TH+ neurons were generated in DMEM/N2 conditions, but not in the presence of added serum, we examined gene expression in these conditions using multiple neural and dopaminergic markers. Real-time PCR was performed and gene expression was compared in differentiated aggregates after 4 weeks using GAPDH–normalized relative gene expression ratios (Table 3). Expression of SOX1 [48, 49] and MAP2 confirmed the presence of neural precursors and differentiated neurons, respectively, in both conditions. Higher expression of SOX1 in serum-free conditions and MAP2 in serum-containing conditions suggested that there was a bias toward proliferation of neural precursors in serum-free conditions and differentiation to neurons under the influence of serum. Several transcription factors that are involved in the specification of the midbrain dopaminergic lineage, EN1, NURR1, PITX3, and LMX1B, were all expressed at higher levels in serum-free conditions, at approximately 5.8-, 1.8-, 2.2-, and 1.5-fold, respectively [46, 47, 5056]. The difference in expression of NURR1 and PITX3 was statistically significant (p = .025 and .001, respectively), whereas the difference in EN1 (p = .081) expression was not significant in this analysis. The comparison of LMX1B expression was performed by densitometry of end-point RT-PCR (Fig. 3I). Analysis of markers of differentiated dopaminergic neurons demonstrated expression of TH, AADC, VMAT2, and the DAT in both conditions. Only AADC (p = .001) showed significant elevated expression in serum-free conditions, with VMAT2 not being significantly different and TH and DAT showing elevated expression in serum-containing conditions. The GIRK2 channel protein is a marker of A9 dopaminergic neurons [57], which are the major dopamine neuron subtype depleted in Parkinson's disease [5860]. GIRK2 was expressed approximately 7.1-fold higher in serum-free conditions (p = .001). Expression of DβH, a more specific marker for other catecholaminergic neurons, was upregulated in 10% serum (p = .001). This expression analysis suggested formation of lineages expressing these markers in both conditions, with elevated expression of dopaminergic transcription factors and some markers of differentiated neurons in DMEM/N2 conditions. However, because this was a population-wide analysis, we also performed immunostaining to determine the relative distribution of neurons expressing some of these markers.

Table Table 3.. Real-time reverse transcription–polymerase chain reaction comparison of gene expression in aggregates differentiated in serum or serum-free conditions
  • a

    Aggregates were differentiated for 1 month.

  • a

    aPrimer efficiencies.

  • b

    bTriplicate reactions.

  • c

    cRelative expression ratio (serum-free/serum), normalized with parallel GAPDH controls.

  • d

    dExample of one set of GAPDH control reactions, n = 3. Overall ratio was 0.95, n = 12.

  • e

    eStatistically significant.

  • f

    fDetermined by end-point reverse transcription–polymerase chain reaction and densitometry.

  • h

    Abbreviations: AADC, aromatic amino acid decarboxylase; CP, threshold crossing point; DAT, dopamine transporter; DβH, dopamine β-hydroxylase; EFF, primer efficiencies; SD, standard deviation; TH, tyrosine hydroxylase; VMAT2, vesicular monoamine transporter 2.

GeneEFFaMean CPbSDMean CPbSDRatiocpvalue
LMX1Bf     1.5f 

To quantify the proportion of neurons in DMEM/N2 differentiations that expressed TH, aggregates were plated in adherent culture in MedII/FGF2 medium. Extensive networks of TH+ neurons were observed (Figs. 5A, 5B) at a far greater abundance than reported previously [15, 29, 31, 32]. Scoring of isolated βIII tubulin+ neurites in overlayed images showed that 73.9 ± 10.5% (46 of 64, n = 3 fields) were TH+/βIII tubulin+ (Figs. 5A, 5B; panels 1, 2). To support the formation of mature neuron cell types, DMEM/N2 aggregates were plated in medium containing GDNF, BDNF, and 5% serum, a formulation known to support the survival of mouse ES cell–derived dopaminergic neurons [61]. Counting of cell bodies demonstrated that TH+ neurons comprised 63.8 ± 4.6% (689 of 1,085 cells, n = 3 wells) of the MAP2+ population, whereas VMAT2+ neurons comprised 94.9 ± 2.9% (317 of 334 cells, n = 3 wells) of the MAP2+ population. Figure 5C shows an example of the most highly differentiated TH+ neurons observed in these cultures, exhibiting a cell body, an approximately 580-μm dendritic extension and spines, and presumed growth cone. Additional immunostaining analysis demonstrated expression of additional markers of the dopaminergic phenotype in DMEM/N2 differentiations. Coexpression of the βIII tubulin, TH, VMAT2, and DAT proteins was demonstrated in aggregates in DMEM/N2 suspension cultures (Fig. 5D), which is similar to what was seen in MedII/FGF2 suspension aggregates (Fig. 5E). Coexpression of TH and active phospho-TH(Ser40) (Fig. 5F) [62], expression of the panneuronal marker TAU [63] and AADC (Fig. 5G), and coexpression of TH and DAT (Figs. 5H, 5I) were also demonstrated. Although RT-PCR analysis had detected DβH message in DMEM/N2 aggregates, expression was significantly lower than in 10% serum conditions. We used immunostaining to detect DβH-expressing cells in 4-week suspension aggregates. Only rare DβH+ cells were detected in DMEM/N2 aggregates from trypsin-passaged BG01 cells, as well as from microdissection-passaged BG01 (Fig. 5J) and BG03 (Fig. 5K) cells that were differentiated with an initial 5 days in MedII/FGF2 followed by 1 month in DMEM/N2. These differentiations, as well as microdissection-passaged BG01 and BG03 that were differentiated in only DMEM/N2, also generated large networks of βIII tubulin+/TH+ neurons (Fig. 6A). We made several additional observations during the course of these experiments. Unlike embryoid body differentiations in serum, very few cysts were formed in embryoid bodies in serum-free conditions. Occasionally, pigmented epithelial cells were generated (Fig. 6B), similar to that observed in stromal cell–mediated differentiations of primate ES cells [13], although this was not a common event. RT-PCR and protein expression analyses therefore demonstrated the presence of the developmental and cellular factors that specify the midbrain dopaminergic lineage in suspension aggregates and mediate dopamine biosynthesis, vesicle loading, and dopamine reuptake after neurotransmitter release.

Figure Figure 5..

Expression of dopaminergic markers in differentiated aggregates. Representative low-magnification images of DMEM/N2 aggregates plated in MedII/FGF2 medium and immunostained with (A) βIII tubulin and (B) TH. Boxed regions are shown at higher magnification in 1 and 2, with βIII tubulin, TH, and merged panels. (C):A highly differentiated MAP2+/TH+ neuron observed when DMEM/N2 aggregates were plated in medium containing GDNF, BDNF, and 5% serum. Insets are higher magnifications of the indicated regions showing (left to right) presumed growth cone, connections to other MAP2+ neurons, and cell body. The length of the extension from cell body to growth cone was ∼580 μm. (D): Whole-mount four-color immunostaining of DMEM/N2 and (E) MedII/FGF2 suspension aggregates. βIII tubulin (visualized with a secondary antibody labeled with Alexafluor 350), TH (Alexafluor 594), VMAT2 (Alexafluor 647), and DAT (Alexafluor 488) images were merged to the cyan, magenta, yellow, and black channels of a CMYK image, respectively. Coexpression in cell bodies is indicated by white staining (arrows). (F–I): Immunostaining for markers of dopaminergic neurons in DMEM/N2 aggregates plated in MedII/FGF2 medium. (F): TH and phospho-TH(Ser40). (G): Aromatic amino acid decarboxylase and TAU. (H, I): DAT and TH. (J, K): Only rare DβH+ cells were detected in suspension aggregates. DβH (arrow) and TH (J) and DβH (arrow) and βIII tubulin (K) immunostaining of microdissection-passaged BG01 and BG03 suspension aggregates, respectively. Scale bars = 100 μm (A, B), 50 μm (1, 2, C, F, G, J, K), 25 μm (E, H, I), 10 μm (D), and 5 μm (C insets). Abbreviations: BDNF, brain-derived neurotrophic factor; DAPI, 4′,6′-diamidino-2-phenylindole; DAT, dopamine transporter; DMEM, Dulbecco's modified Eagle's medium; FGF2, fibroblast growth factor 2; GDNF, glial-derived neurotrophic factor; MAP2, microtuble-associated protein 2; TH, tyrosine hydroxylase; VMAT2, vesicular monoamine transporter 2.

Figure Figure 6..

(A): Whole-mount TH and βIII tubulin immunostaining of BG03 suspension aggregates differentiated in DMEM/N2. (B): Pigmented cells observed in some BG01 DMEM/N2 suspension aggregates. (C): Differentiation of BG01 cells maintained on human keloid fibroblasts. Scale bars = 50 μm (B) and 25 μm (A, C). Abbreviations: DMEM, Dulbecco's modified Eagle's medium; TH, tyrosine hydroxylase.

The use of MEF feeder layers to support hESC culture will add regulatory complexity, because new clinical products derived using these feeder layers will be considered xenotransplants. Although others have demonstrated the maintenance of hESCs on human feeder cells [6466] or in a feeder-free environment [2, 4], it has not been determined whether hESCs grown under these conditions can differentiate to TH+ neurons. We therefore differentiated collagenase/trypsin BG01 cells that had been maintained on a layer of human keloid fibroblasts (I.L., unpublished data) as DMEM/N2 aggregates and demonstrated that a high proportion of TH+ neurons were also generated under these conditions (Fig. 6C). Therefore, hESCs that retain appropriate developmental potential may be able to be derived and maintained on human feeder layers, avoiding stringent xenotransplantation regulations.

6-OHDA is a catecholamine neurotoxin that is taken up by dopaminergic cells expressing DAT and noradrenergic neurons [67]. To examine whether the TH+ neurons present in DMEM/N2 suspension aggregates were sensitive to 6-OHDA, we exposed aggregates to 10 mM or 1 mM 6-OHDA for 10 minutes, followed by a 5-hour recovery in MedII/FGF2 medium, so that degenerating cells could be visualized. Exposure to 6-OHDA led to widespread ablation of TH+ neurons, which were rarely intact but exhibited disrupted and punctuated staining (Figs. 7A, 7B, 7D, 7E). βIII tubulin+/TH neuronal extensions and nonneuronal DAPI-stained nuclei appeared intact. Ascorbic acid–treated controls (not shown) were comparable with untreated aggregates (e.g., Fig. 3A). We used a 100- or 10-fold (Figs. 7C, 7F) excess of dopamine to compete with 6-OHDA uptake. This protected TH+ neurons from ablation, indicating that the cells express functional dopamine or norepinephrine transporters.

Figure Figure 7..

Sensitivity to 6-OHDA and evoked release of catecholamines. (A–F): 6-OHDA selectively ablates TH+ neurons in DMEM/N2 suspension aggregates. βIII tubulin and TH whole-mount immunostaining of DMEM/N2 aggregates exposed to 10 mM (A, D) or 1 mM (B, E) 6-OHDA. (C, F): Addition of 100-mM dopamine to 10-mM and 1-mM (not shown) 6-OHDA exposures protected TH+ neurons from ablation. (G): HPLC traces demonstrating evoked release of 2175 pg/ml DA, 4475 pg/ml adrenaline (A), and 3404 pg/ml NA per 106 cells from DMEM/N2 differentiations plated in medium containing 5% serum, GDNF, and BDNF in response to depolarization with 56-mM KCl. These catecholamines were not detected in parallel cultures treated with HBSS. The elution times for adrenaline, NA, and DA were 7.13, 7.84, and 17.77 minutes, respectively. DHBA was used as an internal standard. The amplitude of electrochemical detection (mV) is shown for the HBSS and KCl samples. Scale bars = 100 μm (A–C) and 50 μm (D–F). Abbreviations: 6-OHDA, 6-hydroxydopamine; BDNF, brain-derived neurotrophic factor; DA, dopamine; DHBA, 3,4-dihydroxybenzyamine; DMEM, Dulbecco's modified Eagle's medium; GDNF, glial-derived neurotrophic factor; HBSS, Hanks' balanced salt solution; NA, noradrenaline; TH, tyrosine hydroxylase.

The release of dopamine in response to depolarization is a key indicator of the functional capacity of ES cell–derived neurons to synthesize dopamine, load it into vesicles, and release it in response to neurophysiological cues [11]. BG01 hESCs were expanded by passaging as clumps with EDTA-free trypsin, followed by differentiation in serum-free suspension culture for 3 days in MedII/FGF2 and then DMEM/N2 until 1 month after derivation. Differentiated aggregates were then plated to adherent culture in medium containing GDNF, BDNF, and 5% serum, generating ∼ 6.4 × 106 cells per 35-mm dish after 2 weeks. From these cultures, HPLC analysis detected evoked release of 2,175 pg/ml dopamine per 106 cells in response to a K+-depolarizing stimulus (Fig. 7G). Release of 4,475 pg/ml adrenaline and 3,404 pg/ml noradrenaline per 106 cells could also be unambiguously resolved as peaks by HPLC. The evoked release of dopamine and other catecholamines was also detected in plated differentiations of microdissection-passaged hESCs, as well as collagenase/trypsin-passaged and SSEA-4–enriched BG01 hESCs, although in some experiments, only dopamine, but not adrenaline or noradrenaline, was detected (not shown). This indicates there is some variability in the proportions of these lineages that can be generated using these techniques. Although rare DβH+ cells were detected in suspension aggregates, plating to adherent culture in medium containing BDNF, GDNF, and serum clearly supported the differentiation of adrenergic and noradrenergic lineages, as well as dopaminergic neurons.

Survival and Differentiation of Transplanted Human ES-Derived Neurons in the Striatum of a Rat Parkinson's Disease Model

To determine whether TH+ neurons derived from hESCs could survive engraftment, we transplanted differentiated aggregates into the striatum of rats with a unilateral lesion in the substantia nigra, which ablates the dopaminergic neurons projecting to the striatum. Aggregates that had been differentiated in MedII/FGF2 for 3 weeks, or DMEM/N2 for 1 month, were implanted into 8 and 15 rats, respectively, and rats were euthanized 8 weeks after implantation. Surviving cells were detected histologically (Fig. 8A) in 6 of 8 rats implanted with MedII/FGF2-differentiated aggregates and 11 of 15 implanted with DMEM/N2 aggregates. Biotinylated-human Alu repeat in situ hybridization probes were used as a lineage marker to confirm the presence of human cells (Fig. 8B). The implants varied considerably in size and the degree of cell survival, and one obvious teratoma containing cartilaginous structures and glandular epithelium was observed in a rat implanted with a DMEM/N2 cell population, indicating that some residual pluripotent cells may persist under these differentiation conditions. Survival of presumed neural rosettes was detected in some implants (Fig. 8A), and the expression of nestin (Figs. 8B, 8C) was also detected in many of the surviving implants. In some cases, regions of MAP2 expression were observed, and the expression of Ki67 indicated that proliferation was still occurring (data not shown). In implants of DMEM/N2-differentiated aggregates, we were able to detect the survival of rare TH+ cells in two animals and a more numerous survival of TH+ cells in a third (Figs. 8D, 8E). The data suggest that after an 8-week period, neural progenitors, but not large numbers of differentiated neurons, can survive and proliferate following implantation of the cell aggregates.

Figure Figure 8..

Differentiation of cell clumps implanted into the lesioned rat brain. (A–C): Sections of implants from MedII/fibroblast growth factor 2–differentiated aggregates. (A): Hematoxylin and eosin staining of a section with a surviving implant. Putative neural precursor rosettes are indicated (arrows). (B): Implant detection using human-specific Alu probes and a human-specific nestin antibody. (C): Nestin expression in a large implant. (D): Chromogenic detection of surviving TH+ cells derived from a DMEM/N2 transplant 8 weeks after implantation. The border of the graft is indicated (black dashed lines). (E): Higher magnification of the boxed region (white dashed) in (D). Scale bars = 250 μm (A), 100 μm (B, C, D), and 50 μm (E). Abbreviations: COR, cortex; DMEM, Dulbecco's modified Eagle's medium; STR, striatum; TH, tyrosine hydroxylase.


The generation of midbrain dopaminergic neurons for cell transplantation therapy of Parkinson's disease is one of the first clear objectives in the clinical application of hESCs [1]. We report here the differentiation of hESCs to form neurons expressing markers of the midbrain dopaminergic lineage in a serum-free suspension culture system. The resulting neurons coexpressed multiple markers of the dopamine neurotransmitter phenotype and demonstrated functional characteristics expected of neurons. Two independent hESC lines, when cultured on mouse or human feeder layers, displayed a similar response to these differentiation conditions. This represents a simple, robust, and potentially scalable platform for the large-scale derivation of dopaminergic neurons, a key step in development of a cell-based product to treat PD.

The neuronal lineages formed by this method of differentiation expressed SOX1, 2, and 3, which are specific markers of neural progenitors, and multiple transcription factors that are involved in the specification of the midbrain dopaminergic lineage. Expression of EN1, NURR1, PITX3, and LMX1B was upregulated in serum-free conditions, indicating that differentiation of hESC aggregates in a minimal medium was sufficient for the specification of this lineage and suggesting that appropriate signaling to induce and support dopamine neuron differentiation exists in these aggregates. Also, long-term differentiation in large cellular aggregates may provide a low-oxygen environment, a factor that has previously been shown to significantly influence the specification and differentiation of mouse ES cells to dopaminergic neurons [68]. These variables are likely to have contributed to the generation of much higher proportions of TH+ neurons than we or others have observed previously from hESCs. We also demonstrated expression of markers of the differentiated dopaminergic phenotype in serum-free differentiations, including phospho-TH(Ser40), AADC, VMAT-2, DAT, NTRK2 (BDNF receptor), and GIRK2. We showed that most TH+/βIII tubulin+ neurons generated in DMEM/N2 conditions were representative of CNS lineages and not neural crest–derived peripheral neurons. Only rare peripherin+ neurons were detected in DMEM/N2 suspension differentiations, but addition of BMP-4 during days 5 through 9 induced peripheral neuronal differentiation, as measured by an increase in the proportion of peripherin-expressing neurons. These conditions still generated networks of TH+ neurons, confirming the importance of identifying TH+/peripherin neurons [45]. Similarly, although DβH mRNA was detected in DMEM/N2 suspension differentiations by real-time RT-PCR, expression was significantly lower than in serum-containing conditions. Only rare DβH+ neurons were detected by immuno-staining of multiple DMEM/N2 differentiations, suggesting that a large proportion of TH+ neurons generated in suspension aggregates were dopaminergic. TH+/βIII tubulin+ neurons in suspension aggregates were ablated by 6-OHDA but were protected from ablation by an excess of dopamine. This provided functional evidence that most TH+ neurons in suspension aggregates were dopaminergic or noradrenergic. Although few cells in the aggregates expressed DβH, differentiation to significant proportions of adrenergic and noradrenergic neurons occurred when suspension aggregates were plated into BDNF, GDNF, and serum, because adrenaline and noradrenaline, as well as dopamine, were released by depolarized cultures grown in these conditions. Together with the response of patch-clamped plated neurons to glutamate and GABA, the expression of synaptic components and the capacity of these neural populations to respond to neurophysiological stimuli, a functional neuronal phenotype was demonstrated. The proportions of other noncate-cholaminergic neural lineages generated under these conditions also still needs to be ascertained, although the presence of cholinergic, glutaminergic, and GABAergic neurons was suggested by the expression analysis.

Transplantable neural cells derived from hESCs have been reported previously [29, 32]. However, the frequency of TH+ neurons generated with these previous in vitro differentiation approaches was low, and no engrafted TH+ cells were detected after transplantation to newborn mice. We report here one of the first examples of detection of surviving TH+ cells after transplantation in vivo. We successfully transplanted TH+ cells to the striatum of the lesioned adult rat brain as cell clumps and could detect implanted cells 8 weeks after implantation. However, poor viability of differentiated neurons after transplantation may be the primary reason why relatively few TH+ cells were detected in these experiments, despite our efforts to minimize cell death by avoiding dissaggregation to single cells for implantation. Apoptosis of 90%–95% of implanted neurons has also been observed in clinical transplants of fetal neural tissue [1, 6]. To counter this problem, it may be possible to identify a window of development in vitro, in which abundant neural precursors committed to the dopaminergic fate are present and can be transplanted. Conversely, transplanted hESC-derived neural progenitors may require a longer time to differentiate after transplantation than required with mouse ES cell–derived progenitors [10].

The differentiation system outlined in this study models several events that occur during embryonic development. The formation of early neural progenitor cells could be inhibited by BMP signaling, a signal that can direct early embryonic cells to a nonneural ectodermal fate, whereas cells that had progressed to a neural progenitor stage and exhibited neural tube-like characteristics [15, 32] were apparently induced to form peripheral neurons by a later addition of BMP. Rosette structures formed during the differentiation of hESC aggregates were reminiscent of the early neural tube in that they formed a tightly packed radial array, exhibited mitosis at the central core, differentiated to HuC/D+ early postmitotic neurons as they exited this structure, and exhibited ultrastructural characteristics of the neural tube (T.S. and J. McDonald, unpublished data). These features suggest that many early aspects of human neurogenesis may be accessible to study using hESCs and show that the cell types produced in our differentiations respond to developmental factors, as predicted from previous studies. Similarly, as with mouse and primate ES cells, hESCs could also differentiate effectively to TH+ neurons when cocultured with the PA6 stromal cell line [69]. This differentiation approach reflected many of the features of our suspension differentiation system, including differentiation to neural progenitors and neurons that express the key transcription factors NURR1, LMX1B, and PITX3, as well as multiple markers of the dopaminergic phenotype. Cell fate in this differentiation could also be altered by BMP-4 or serum, dopamine was released in response to depolarization, and survival of TH+ cells implanted in the rat brain was demonstrated. Expression of a similar profile of neural markers was detected in PA6 coculture differentiations [69] and in our suspension differentiations using the same cDNA array. This included CXCR4, FGFR1, FGFR2, DLK1, NTRK2, NCAM1, NEFL, MAP2, INA, FZD3, and VEGF. This also indicates significant overlap in the types of lineages generated in both conditions; however, higher expression of markers such as ACTA2, ACTG2, AFP, CTNNB1, CDH1, KRT8, IL6ST, and IGF1R in PA6 differentiations suggests that there are also differences in the derived lineages. The similarities in these differentiation outcomes may suggest that the signals that enable differentiation to TH+ neurons provided by PA6 coculture are also provided within differentiating aggregates in a minimal medium in the absence of exogenous neural-inducing factors.


We have developed a simple culture system for the differentiation of hESCs to enriched neuronal populations of cells, including those of the midbrain dopaminergic lineage, characterized the expression of a variety of neuronal and dopaminergic markers, and demonstrated the functionality expected of differentiating neurons. This differentiation system could provide a simple experimental model for developing optimal cultures of midbrain dopaminergic populations suitable for implantation studies in animal models of PD and possible therapeutic applications.

Note Added in Proof

Recent additional reports have also demonstrated the differentiation of hESCs to dopaminergic neurons. These studies induced differentiation by coculture with stromal cell layers: Perrier AL, Tabar V, Barberi T et al. Derivation of midbrain dopamine neurons from human embryonic stem cells. Proc Natl Acad Sci U S A 2004;101:12543–12548; and Buytaert-Hoefen KA, Alvarez E, CR Freed. Generation of tyrosine hydroxylase positive neurons from human embryonic stem cells after coculture with cellular substrates and exposure to GDNF. Stem Cells 2004;22:669–674.


We thank current and former members of the BresaGen Cell Therapy Programs in Adelaide and Athens, Ray Johnson for performing HPLC analysis, Clifton Baile and Diane Hartzel and the Animal Facility of the University of Georgia, Animal and Dairy Science Department for performing rat implantations, and Mahendra Rao for critically reading the manuscript. This work was supported by BresaGen Inc. and the Augusta Chapter of the American Legion (to N.A.L.). hESC characterization work was also supported by NIH grant R24DK063689 (to B.G.C., awarded to BresaGen Inc.).