Regulation of mouse embryonic stem cell neural differentiation by retinoic acid.
Journal: 2009/June - Developmental Biology
ISSN: 1095-564X
Abstract:
Pluripotent mouse embryonic stem cells (ESCs) derived from the early blastocyst can differentiate in vitro into a variety of somatic cell types including lineages from all three embryonic germ layers. Protocols for ES cell neural differentiation typically involve induction by retinoic acid (RA), or by exposure to growth factors or medium conditioned by other cell types. A serum-free differentiation (SFD) medium completely lacking exogenous retinoids was devised that allows for efficient conversion of aggregated mouse ESCs into neural precursors and immature neurons. Neural cells produced in this medium express neuronal ion channels, establish polarity, and form functional excitatory and inhibitory synapses. Brief exposure to RA during the period of cell aggregation speeds neuronal maturation and suppresses cell proliferation. Differentiation without RA yields neurons and neural progenitors with apparent telencephalic identity, whereas cells differentiated with exposure to RA express markers of hindbrain and spinal cord. Transcriptional profiling indicates a substantial representation of transit amplifying neuroblasts in SFD cultures not exposed to RA.
Relations:
Content
Citations
(28)
References
(98)
Drugs
(2)
Chemicals
(5)
Organisms
(2)
Processes
(3)
Anatomy
(8)
Similar articles
Articles by the same authors
Discussion board
Dev Biol 328(2): 456-471

Regulation of mouse embryonic stem cell neural differentiation by retinoic acid

INTRODUCTION

In vitro differentiation of embryonic stem cells (ESCs) has attracted wide interest as an experimental system for investigating developmental pathways and mechanisms. In addition, the isolation of human ESCs [Thomson et al., 1998] and human induced pluripotent stem cells [Takahashi et al., 2007; Park et al., 2008] has raised the possibility that in vitro differentiation may provide a novel source of cells for tissue replacement or repair [Murry and Keller 2008]. Therapeutic use of ESCs will require robust and reliable methods for producing specific neural cell types. Early work on mouse ESC in vitro differentiation was performed in serum-supplemented medium [Doetschman et al., 1985]. These experiments found that aggregation of cells into embryoid bodies, combined with exposure to retinoic acid (RA), enhanced the efficiency of ESC conversion to a neural phenotype [Bain et al., 1995; Fraichard et al., 1995; Strübing et al, 1995]. Aggregation alone in the presence of serum favours differentiation into non-neural cell types including cardiac cells [Bain et al., 1996], whereas addition of 0.5 to 1 μM RA suppresses non-neural differentiation and instead results in a high proportion of cells becoming neurons or astrocytes [Bain et al., 1995]. Neurons produced in this way acquire axonal and dendritic polarity, form functional synapses, and include a mixture of excitatory cells that release glutamate as their transmitter and inhibitory cells that use either GABA or glycine [Strübing et al, 1995; Finley et al., 1996].

Because serum contains a large number of factors that might influence the differentiation process, a number of groups have investigated the in vitro conversion of ESCs into neurons, or neural precursors, under serum-free growth conditions [Okabe et al., 1996; Wiles and Johansson, 1999; Finley et al., 1999; Tropepe et al., 2001; Ying et al., 2003; Watanabe et al., 2005; Bouhon et al., 2005]. In addition, modifications to the original differentiation procedures have been devised with the goal of enhancing production of specific neural phenotypes including dopaminergic neurons [Kawasaki et al., 2000; Lee et al., 2000], motorneurons [Wichterle et al., 2002], cerebellar neurons [Salero and Hatten, 2007] and oligodendrocytes [Brüstle et al., 1999; Liu et al., 2000]. Many of these studies have used media or media supplements with proprietary composition, or they employed serum or cell-conditioned media [Kawasaki et al., 2000; Barberi et al., 2003], which makes it difficult to evaluate the specific requirements for efficient ESC growth and/or differentiation [Cai and Grabel, 2007]. Moreover, it is generally recognized that a more comprehensive comparison of the differentiated cell phenotypes produced by these different in vitro induction procedures is desirable [Glaser and Brustle, 2005].

A goal of our work has been to simplify the protocol required for in vitro neural induction while preserving cell survival and eliminating exposure to exogenous retinoids. Here we describe a serum-free, retinoid-free, growth medium supporting robust neural differentiation with insulin, transferrin and BSA as the only exogenous protein constituents. Neurons derived in this medium exhibit many characteristics of those induced by retinoic acid, but transcriptional profiling revealed substantial differences in gene expression between retinoid-free versus retinoid-exposed cell populations that was confirmed by electrophysiology and immunofluorescence.

METHODS

ES Cell Culture

Murine ESCs were propagated independent of feeder cells in 25 cm tissue culture flasks that had been coated with gelatin (0.1% from bovine skin in sterile water; Sigma). The D3 and CE3 ESC lines were obtained from Dr. David Gottlieb [Adams et al., 2002]; the B5 line was obtained from Dr. Andras Nagy [Hadjantonakis et al., 1998]. The growth medium for dividing ESCs was Dulbecco’s Modified Eagle Medium (DMEM; Life Technologies), which was supplemented with 20% calf serum, nucleosides (30μM adenosine, cytodine, uridine, guanidine and 10μM thymidine; Sigma), leukemia inhibitory factor (LIF, 1000 units ml ESGRO, murine; Life Technologies) and 2-mercaptoethanol (0.1mM; Sigma). When cells reached 80 to 90% confluence the medium was removed, the flask was rinsed with divalent-free Earl’s balanced salt solution (EBS; Life Technologies), and incubated for 1 min with protease XXIII (1 mg ml; Sigma) in divalent-free EBS. Detached cells were triturated gently, collected by centrifugation (70 × g for 5 min), and then resuspended in 4 ml of complete growth medium. A 1:4 dilution of this cell suspension was used to seed a new flask. All cultures were kept at 37°C in a 5% CO2 humidified air incubator.

Neural Differentiation

To initiate neural differentiation, ESCs were passaged from the flask with protease XXIII in divalent-free EBS and collected by centrifugation. Cells were resuspended in serum-free differentiation medium (Supplementary Table 1) supplemented with nucleosides (see above) and with 4 mM glutamine, and seeded at a density of 1.5 to 2 × 10 per 5 ml in 5 cm non-adhesive bacteriological petri plates that had been coated with 0.15% agarose (Sigma, Type II-A). Under these conditions, cells remained suspended and formed aggregates that increased in size to form embryoid bodies. Every 2d the aggregates were provided with fresh medium as follows. Medium and aggregates were transferred from the Petri dish to a 5 ml round-bottomed tube using a transfer pipette. Fresh medium was added to the empty Petri dish while the aggregates were allowed to settle in the tube for 5–10 min at room temp. Old medium was removed from the tube. The aggregates were resuspended with fresh medium from the Petri dish, transferred back into the Petri dish and returned to the incubator. For some experiments 500 nM all-trans RA (Sigma) was added to the medium from d4–8. After 10–12d, aggregates were dissociated into a single cell suspension and plated onto 35 mm plastic tissue culture dishes or glass cover slips that were coated with a mixture of poly-DL-ornithine (200 μg ml; Sigma) and mouse laminin (3 μg ml; Gibco). Aggregates were collected in a 5 ml round bottomed tube and incubated for 5 min at room temperature in divalent-free EBS containing protease XXIII (1 mg ml). Aggregates were rinsed 2 times with EBS containing 0.1% BSA (Sigma) and 0.1% ovomucoid (Sigma), and then once with SFD medium (Supplementary Table 1). Aggregates were triturated with a fire-polished Pasteur pipette in a total volume of 1–2 ml of SFD medium. Approximately 2–3 fold more cells were obtained from aggregates maintained in SFD medium alone for 12 d than from cultures that were exposed to RA (see Results). For plating, cells were suspended in SFD medium at a density of 3 to 6 × 10 cells ml and dispensed onto coated dishes or cover slips. Cells differentiated in SFD medium alone were plated at lower density (~150 cells mm) than RA-treated cells (300 cells ml) to compensate for differences in proliferation rate (see Results). Cells were allowed to settle and attach for 1 hr, then gently rinsed once with SFD medium before adding 0.5 to 1 ml of SFD medium supplemented with 0.25 mM glutamine. Cultures were fed every 2 to 3 d by partial medium replacement. For some experiments, division of non-neuronal cells was inhibited by addition of cytosine arabinoside (10 μM; Sigma) several days after plating. All inductions were performed within 10 passages of thawing from low passage number frozen stocks.

Electrophysiology

Recordings from cells with neuronal morphology were performed as described previously [Bain et al., 1995; Finley at al., 1996]. Briefly, cultures were perfused at a rate of 1–2 ml/min with Tyrode’s solution: 150 mM NaCl, 4 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 10 mM glucose, and 10 mM HEPES, adjusted to pH 7.4 with NaOH. Pipettes for whole cell recording were pulled from boralex glass capillaries. The internal solution used to record Na and K currents contained 140 mM KCH3SO3, or 140 mM K-glucuronate, 10 mM NaCl, 5 mM MgCl2, 200 μM EGTA and 10mM HEPES, adjusted to pH 7.4 with KOH. For recording agonist-evoked currents, pipettes were filled with 140 mM Cs-glucuronate, 10 mM EGTA, 1 mM ATP (Mg salt), 0.3 mM GTP (tris salt), and 10 mM HEPES, adjusted to pH 7.4 with CsOH. The open tip resistance ranged from 1 to 5 MOhm. Drug solutions were applied to the cells using an array of microcapllary tubes. The time constant for exchange of the external solution was 30–50 msec. Excitatory and inhibitory agonists were dissolved in 160 mM NaCl, 2mM CaCl2, 0.5 μM tetrodotoxin (TTX, Sigma) and 10 mM HEPES, adjusted to pH 7.4 with NaOH.

Current was recorded with an Axopatch 200A amplifier (Axon Instruments), filtered at 1 to 5 KHz (−3 db, 4 pole Bessel), and digitized at 10 to 20 KHz. Current traces were corrected for leak and capacity using scaled current evoked by an 8 to 10 mV hyperpolarizing step from the holding potential. Membrane potentials were corrected for the junction potential between the internal solution and the Tyrode’s solution in which seals were formed. This potential was − 10 mV for pipettes containing Cs-glucuronate. All experiments were performed at room temperature.

Immunofluorescence

ESC aggregates or monolayer cultures of differentiated cells were rinsed with Tyrode’s solution and incubated for 15–20 min at room temperature in 0.1 M Na Phosphate, pH 7.4, containing 4% paraformaldehyde and for some experiments 0.1% glutaraldehyde. After three rinses with Tris-buffered saline, pH 7.4, aggregates were equilibrated with 30% sucrose, collected in an inverted pyramidal mold (Polysciences) and frozen on dry ice. Aggregates frozen in sucrose were bonded to the chuck with O.C.T. compound (Tissue-Tek), cryosectioned (6 μm) and mounted on Snowcoat X-tra slides (Surgipath). Mounted sections and fixed, rinsed monolayer cultures were incubated for 30 min at room temperature with blocking solution (BS): PBS containing 1% normal goat serum, 0.02% sodium azide, and 0.1% Triton X-100; and then incubated overnight at 4 °C (or in some cases for 2hr at room temperature) with primary antibodies diluted in BS. Samples were rinsed three times with PBS and incubated for 1h at room temperature with fluorescent conjugated secondary antibodies diluted in BS. They were then rinsed three times with PBS and examined under epi-illumination with appropriate filters. Primary and secondary antibodies and dilutions are reported in Supplementary Table 2. Images were acquired with a Nikon Eclipse E600 microscope with 10x, 40x and 60x objectives (0.3, 0.8 and 1.0 N.A., respectively). Image analysis and quantification was performed with MetaMorph software (Universal Imaging). Most figure panels show sections that did not exceed the field of view of the camera (Photometrics CoolSNAP ES).

RT-PCR

Total RNA was isolated from differentiated ESC cultures using the RNeasy kit (Qiagen) and quantified by optical density. Reverse transcription and PCR were performed according to the manufacturer’s protocol using the OneStep RT-PCR Kit (Qiagen) and 50ng of RNA. In brief, reverse transcription for 30 min at 50 °C was immediately followed by PCR activation at 95 °C for 15 min, then 30 cycles of denaturation at 94 °C for 1 min, annealing at 50 °C for 1 min and extension at 72 °C for 1 min. Final extension was at 72 °C for 10 min. The following primers were used:

glycine receptor (GlyR)
GlyR α1forward: 5′-TGTACATCCCCAGCCTGCTCATCGTCATC-3′
reverse: 5′-GAGGCGGGTTAGTGGTGTTGTTGTTGTTGG-3′
GlyR α2forward: 5′-CACCTTGCCCCAGTTTATTTTGAA-3′
reverse: 5′-GATGCCCTGGAACCTGAACTCTGTGTAG-3′
GlyR α3forward: 5′-CTTTCGGACACTAGTTTCTGGATTTTAC-3′
reverse: 5′-AGTTAGCCCCCTTCTCATTAGCA-3′
GlyR α4forward: 5′-AGGGGCCAACTTCCATGAGGTGAC-3′
reverse: 5′-CAAGGCAGCAAATACAAAGAGCAGGCATACAGC-3′
GlyR βforward: 5′-CTCATTGCCTGCCTGCTCTTCGGGTTTG-3′
reverse: 5′-GATGCGCTTGGCTGCTGTTGGGATGA-3′
GAPDHforward: 5′-GAGGCCGGTGCTGAGTATGTC-3′
reverse: 5′-TCCACCACCCTGTTGCTGTAG-3′

Microarray Analysis

Total RNA was isolated using Trizol reagent (Life Techonologies) from replicate cultures of undifferentiated ESCs and from cells aggregated in suspension 12d in SFD medium, with or without exposure to RA (d4–d8), then dissociated and maintained 6d in plated cultures. RNA from each sample was used to generate biotinylated-fragmented cRNA probes using a Bioarray High Yield RNA transcript labelling kit (Enzo Biochem, New York). Hybridized Affymetrix gene chips (MOE430v2), which represent approximately 39,000 transcripts, were washed and stained according to the manufacturer’s protocols, then scanned in the Affymetrix GeneChip array scanner by the Siteman Cancer Center Multiplexed Gene Analysis Core at Washington University Medical School. Array images were analyzed using Affymetrix Microarray Suite version 5.0. Gene expression values were normalized to the median intensity array before model-based analysis using dChip software [Li and Wong, 2001]. Using undifferentiated ESCs as a baseline, genes were identified as enriched in differentiated cells if they satisfied both of the following criteria: 1) a ~2-fold increase in mean normalized intensity level (1.2 fold 90% lower confidence bound, see [Li and Wong, 2001; Ramalho-Santos et al., 2002]) and 2) an absolute difference in mean normalized intensity level >100. This second criterion sets a baseline for expression in the differentiated population to screens out genes with very low expression in both undifferentiated and differentiated populations. Genes identified as common among all three populations had mean normalized intensity values >100 and were identified as “present” by MAS 5.0 in at least one of the replicate samples from each population. Gene Function Enrichment analysis was performed in dChip. Hierarchical clustering based on population Gene Ontology term representation was performed with GOourmet software [Doherty et al., 2006; 2008].

BrdU Labelling

For cell proliferation analysis, bromodeoxyuridine (BrdU, Sigma) was added to aggregates at 10μg/ml every 4h for 24h. Aggregates were harvested at 0h, 1h, 2h, 4h, 6h, 8h and 24h after BrdU treatment, and fixed for 20–30 min at room temperature with 4% paraformaldehyde in 0.12M sodium phosphate, pH 7.0. Aggregates were cryosectioned, postfixed in 4% paraformaldehyde in 0.12M sodium phosphate, pH 7.0, and treated with 2N HCl for 10 min at 37°C. For cultured cells, BrdU was applied for 2 hr prior to fixation with 4% paraformaldehyde in 0.1M sodium phosphate, pH 7.0, for 10–15 min. Immediately before staining, cultured cells were post-fixed for 15 min at room temperature with 4% paraformaldehyde in 0.12M sodium phosphate, pH 7.0, then treated with 2N HCl for 30 min at 37°C. Cultures and slide-mounted sections were neutralized by room temperature incubation for 20 min in borax solution: 55% 0.2M boric acid + 45% 0.05 M borax. After 3 washes with PBS, samples were blocked with PBS containing 1% goat serum and 0.1% Triton X-100 for 30 min, then incubated for 1–2h at room temperature with mouse anti-BrdU (1:400, Boehringer Mannheim). Goat anti-mouse Cy3-conjugated secondary antibodies were used to detect BrdU-positive cells. Cell nuclei were counterstained with Hoechst33342 (2.5μM). Counts were performed to quantify the number of cells that incorporated BrdU as a percentage of Hoechst 33342 stained nuclei (23–25 aggregate sections or 34 plated culture fields for each time point).

Statistical Analysis

Unless otherwise specified, data are expressed as mean ± sem. Statistical analysis was performed with SigmaStat (Systat Software) or with the appropriate functions in Excel (Microsoft). Significance was assigned for p values < 0.05.

ES Cell Culture

Murine ESCs were propagated independent of feeder cells in 25 cm tissue culture flasks that had been coated with gelatin (0.1% from bovine skin in sterile water; Sigma). The D3 and CE3 ESC lines were obtained from Dr. David Gottlieb [Adams et al., 2002]; the B5 line was obtained from Dr. Andras Nagy [Hadjantonakis et al., 1998]. The growth medium for dividing ESCs was Dulbecco’s Modified Eagle Medium (DMEM; Life Technologies), which was supplemented with 20% calf serum, nucleosides (30μM adenosine, cytodine, uridine, guanidine and 10μM thymidine; Sigma), leukemia inhibitory factor (LIF, 1000 units ml ESGRO, murine; Life Technologies) and 2-mercaptoethanol (0.1mM; Sigma). When cells reached 80 to 90% confluence the medium was removed, the flask was rinsed with divalent-free Earl’s balanced salt solution (EBS; Life Technologies), and incubated for 1 min with protease XXIII (1 mg ml; Sigma) in divalent-free EBS. Detached cells were triturated gently, collected by centrifugation (70 × g for 5 min), and then resuspended in 4 ml of complete growth medium. A 1:4 dilution of this cell suspension was used to seed a new flask. All cultures were kept at 37°C in a 5% CO2 humidified air incubator.

Neural Differentiation

To initiate neural differentiation, ESCs were passaged from the flask with protease XXIII in divalent-free EBS and collected by centrifugation. Cells were resuspended in serum-free differentiation medium (Supplementary Table 1) supplemented with nucleosides (see above) and with 4 mM glutamine, and seeded at a density of 1.5 to 2 × 10 per 5 ml in 5 cm non-adhesive bacteriological petri plates that had been coated with 0.15% agarose (Sigma, Type II-A). Under these conditions, cells remained suspended and formed aggregates that increased in size to form embryoid bodies. Every 2d the aggregates were provided with fresh medium as follows. Medium and aggregates were transferred from the Petri dish to a 5 ml round-bottomed tube using a transfer pipette. Fresh medium was added to the empty Petri dish while the aggregates were allowed to settle in the tube for 5–10 min at room temp. Old medium was removed from the tube. The aggregates were resuspended with fresh medium from the Petri dish, transferred back into the Petri dish and returned to the incubator. For some experiments 500 nM all-trans RA (Sigma) was added to the medium from d4–8. After 10–12d, aggregates were dissociated into a single cell suspension and plated onto 35 mm plastic tissue culture dishes or glass cover slips that were coated with a mixture of poly-DL-ornithine (200 μg ml; Sigma) and mouse laminin (3 μg ml; Gibco). Aggregates were collected in a 5 ml round bottomed tube and incubated for 5 min at room temperature in divalent-free EBS containing protease XXIII (1 mg ml). Aggregates were rinsed 2 times with EBS containing 0.1% BSA (Sigma) and 0.1% ovomucoid (Sigma), and then once with SFD medium (Supplementary Table 1). Aggregates were triturated with a fire-polished Pasteur pipette in a total volume of 1–2 ml of SFD medium. Approximately 2–3 fold more cells were obtained from aggregates maintained in SFD medium alone for 12 d than from cultures that were exposed to RA (see Results). For plating, cells were suspended in SFD medium at a density of 3 to 6 × 10 cells ml and dispensed onto coated dishes or cover slips. Cells differentiated in SFD medium alone were plated at lower density (~150 cells mm) than RA-treated cells (300 cells ml) to compensate for differences in proliferation rate (see Results). Cells were allowed to settle and attach for 1 hr, then gently rinsed once with SFD medium before adding 0.5 to 1 ml of SFD medium supplemented with 0.25 mM glutamine. Cultures were fed every 2 to 3 d by partial medium replacement. For some experiments, division of non-neuronal cells was inhibited by addition of cytosine arabinoside (10 μM; Sigma) several days after plating. All inductions were performed within 10 passages of thawing from low passage number frozen stocks.

Electrophysiology

Recordings from cells with neuronal morphology were performed as described previously [Bain et al., 1995; Finley at al., 1996]. Briefly, cultures were perfused at a rate of 1–2 ml/min with Tyrode’s solution: 150 mM NaCl, 4 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 10 mM glucose, and 10 mM HEPES, adjusted to pH 7.4 with NaOH. Pipettes for whole cell recording were pulled from boralex glass capillaries. The internal solution used to record Na and K currents contained 140 mM KCH3SO3, or 140 mM K-glucuronate, 10 mM NaCl, 5 mM MgCl2, 200 μM EGTA and 10mM HEPES, adjusted to pH 7.4 with KOH. For recording agonist-evoked currents, pipettes were filled with 140 mM Cs-glucuronate, 10 mM EGTA, 1 mM ATP (Mg salt), 0.3 mM GTP (tris salt), and 10 mM HEPES, adjusted to pH 7.4 with CsOH. The open tip resistance ranged from 1 to 5 MOhm. Drug solutions were applied to the cells using an array of microcapllary tubes. The time constant for exchange of the external solution was 30–50 msec. Excitatory and inhibitory agonists were dissolved in 160 mM NaCl, 2mM CaCl2, 0.5 μM tetrodotoxin (TTX, Sigma) and 10 mM HEPES, adjusted to pH 7.4 with NaOH.

Current was recorded with an Axopatch 200A amplifier (Axon Instruments), filtered at 1 to 5 KHz (−3 db, 4 pole Bessel), and digitized at 10 to 20 KHz. Current traces were corrected for leak and capacity using scaled current evoked by an 8 to 10 mV hyperpolarizing step from the holding potential. Membrane potentials were corrected for the junction potential between the internal solution and the Tyrode’s solution in which seals were formed. This potential was − 10 mV for pipettes containing Cs-glucuronate. All experiments were performed at room temperature.

Immunofluorescence

ESC aggregates or monolayer cultures of differentiated cells were rinsed with Tyrode’s solution and incubated for 15–20 min at room temperature in 0.1 M Na Phosphate, pH 7.4, containing 4% paraformaldehyde and for some experiments 0.1% glutaraldehyde. After three rinses with Tris-buffered saline, pH 7.4, aggregates were equilibrated with 30% sucrose, collected in an inverted pyramidal mold (Polysciences) and frozen on dry ice. Aggregates frozen in sucrose were bonded to the chuck with O.C.T. compound (Tissue-Tek), cryosectioned (6 μm) and mounted on Snowcoat X-tra slides (Surgipath). Mounted sections and fixed, rinsed monolayer cultures were incubated for 30 min at room temperature with blocking solution (BS): PBS containing 1% normal goat serum, 0.02% sodium azide, and 0.1% Triton X-100; and then incubated overnight at 4 °C (or in some cases for 2hr at room temperature) with primary antibodies diluted in BS. Samples were rinsed three times with PBS and incubated for 1h at room temperature with fluorescent conjugated secondary antibodies diluted in BS. They were then rinsed three times with PBS and examined under epi-illumination with appropriate filters. Primary and secondary antibodies and dilutions are reported in Supplementary Table 2. Images were acquired with a Nikon Eclipse E600 microscope with 10x, 40x and 60x objectives (0.3, 0.8 and 1.0 N.A., respectively). Image analysis and quantification was performed with MetaMorph software (Universal Imaging). Most figure panels show sections that did not exceed the field of view of the camera (Photometrics CoolSNAP ES).

RT-PCR

Total RNA was isolated from differentiated ESC cultures using the RNeasy kit (Qiagen) and quantified by optical density. Reverse transcription and PCR were performed according to the manufacturer’s protocol using the OneStep RT-PCR Kit (Qiagen) and 50ng of RNA. In brief, reverse transcription for 30 min at 50 °C was immediately followed by PCR activation at 95 °C for 15 min, then 30 cycles of denaturation at 94 °C for 1 min, annealing at 50 °C for 1 min and extension at 72 °C for 1 min. Final extension was at 72 °C for 10 min. The following primers were used:

glycine receptor (GlyR)
GlyR α1forward: 5′-TGTACATCCCCAGCCTGCTCATCGTCATC-3′
reverse: 5′-GAGGCGGGTTAGTGGTGTTGTTGTTGTTGG-3′
GlyR α2forward: 5′-CACCTTGCCCCAGTTTATTTTGAA-3′
reverse: 5′-GATGCCCTGGAACCTGAACTCTGTGTAG-3′
GlyR α3forward: 5′-CTTTCGGACACTAGTTTCTGGATTTTAC-3′
reverse: 5′-AGTTAGCCCCCTTCTCATTAGCA-3′
GlyR α4forward: 5′-AGGGGCCAACTTCCATGAGGTGAC-3′
reverse: 5′-CAAGGCAGCAAATACAAAGAGCAGGCATACAGC-3′
GlyR βforward: 5′-CTCATTGCCTGCCTGCTCTTCGGGTTTG-3′
reverse: 5′-GATGCGCTTGGCTGCTGTTGGGATGA-3′
GAPDHforward: 5′-GAGGCCGGTGCTGAGTATGTC-3′
reverse: 5′-TCCACCACCCTGTTGCTGTAG-3′

Microarray Analysis

Total RNA was isolated using Trizol reagent (Life Techonologies) from replicate cultures of undifferentiated ESCs and from cells aggregated in suspension 12d in SFD medium, with or without exposure to RA (d4–d8), then dissociated and maintained 6d in plated cultures. RNA from each sample was used to generate biotinylated-fragmented cRNA probes using a Bioarray High Yield RNA transcript labelling kit (Enzo Biochem, New York). Hybridized Affymetrix gene chips (MOE430v2), which represent approximately 39,000 transcripts, were washed and stained according to the manufacturer’s protocols, then scanned in the Affymetrix GeneChip array scanner by the Siteman Cancer Center Multiplexed Gene Analysis Core at Washington University Medical School. Array images were analyzed using Affymetrix Microarray Suite version 5.0. Gene expression values were normalized to the median intensity array before model-based analysis using dChip software [Li and Wong, 2001]. Using undifferentiated ESCs as a baseline, genes were identified as enriched in differentiated cells if they satisfied both of the following criteria: 1) a ~2-fold increase in mean normalized intensity level (1.2 fold 90% lower confidence bound, see [Li and Wong, 2001; Ramalho-Santos et al., 2002]) and 2) an absolute difference in mean normalized intensity level >100. This second criterion sets a baseline for expression in the differentiated population to screens out genes with very low expression in both undifferentiated and differentiated populations. Genes identified as common among all three populations had mean normalized intensity values >100 and were identified as “present” by MAS 5.0 in at least one of the replicate samples from each population. Gene Function Enrichment analysis was performed in dChip. Hierarchical clustering based on population Gene Ontology term representation was performed with GOourmet software [Doherty et al., 2006; 2008].

BrdU Labelling

For cell proliferation analysis, bromodeoxyuridine (BrdU, Sigma) was added to aggregates at 10μg/ml every 4h for 24h. Aggregates were harvested at 0h, 1h, 2h, 4h, 6h, 8h and 24h after BrdU treatment, and fixed for 20–30 min at room temperature with 4% paraformaldehyde in 0.12M sodium phosphate, pH 7.0. Aggregates were cryosectioned, postfixed in 4% paraformaldehyde in 0.12M sodium phosphate, pH 7.0, and treated with 2N HCl for 10 min at 37°C. For cultured cells, BrdU was applied for 2 hr prior to fixation with 4% paraformaldehyde in 0.1M sodium phosphate, pH 7.0, for 10–15 min. Immediately before staining, cultured cells were post-fixed for 15 min at room temperature with 4% paraformaldehyde in 0.12M sodium phosphate, pH 7.0, then treated with 2N HCl for 30 min at 37°C. Cultures and slide-mounted sections were neutralized by room temperature incubation for 20 min in borax solution: 55% 0.2M boric acid + 45% 0.05 M borax. After 3 washes with PBS, samples were blocked with PBS containing 1% goat serum and 0.1% Triton X-100 for 30 min, then incubated for 1–2h at room temperature with mouse anti-BrdU (1:400, Boehringer Mannheim). Goat anti-mouse Cy3-conjugated secondary antibodies were used to detect BrdU-positive cells. Cell nuclei were counterstained with Hoechst33342 (2.5μM). Counts were performed to quantify the number of cells that incorporated BrdU as a percentage of Hoechst 33342 stained nuclei (23–25 aggregate sections or 34 plated culture fields for each time point).

Statistical Analysis

Unless otherwise specified, data are expressed as mean ± sem. Statistical analysis was performed with SigmaStat (Systat Software) or with the appropriate functions in Excel (Microsoft). Significance was assigned for p values < 0.05.

RESULTS

Neural Differentiation

In previous work [Finley et al., 1999] we demonstrated relatively efficient neural differentiation when ESC aggregates were maintained for 8d in serum-free Neurobasal Medium plus the B27 supplement, which contains 20 different ingredients including retinyl acetate [Brewer et al., 1993]; however, generation of neurons was dramatically curtailed when complete B27 supplement was replaced with B27 that lacked retinyl acetate [Finley et al., 1999]. To investigate the effects of retinoids on ESC neural differentiation, we developed a modification of earlier serum-free induction media [Wiles and Johansson, 1999; Tropepe et al., 2001; Ying et al., 2003] that lacks exogenous retinoids and avoids the use of proprietary growth supplements or media formulations (Supplementary Table 1).

To initiate induction, feeder-free ESC monolayers were detached with protease, rinsed extensively with balanced salt solution, suspended in serum-free differentiation medium (SFD; Supplementary Table 1), and dispersed into non-adherent Petri dishes. Under these conditions the cells formed floating aggregates that gradually enlarged by cell division. Expression of Oct3/4, which defines the undifferentiated state [Niwa et al., 2000], progressively declined (Supplementary Fig. 1), while there was a gradual increase in cells expressing proteins associated with neurons or neural precursors including nestin [Lendahl et al., 1990] and β-3-tubulin [Caccamo et al., 1989] (Supplementary Fig. 1). Cells maintained exclusively in SFD medium acquired neural properties over a time course of 10–12d, whereas addition of 0.5 μM RA from d4–d8 speeded differentiation, as evidenced by a higher proportion of β-tubulin-expressing cells at earlier time points. Aggregates maintained in suspension for 12d were dissociated to single cells and plated onto laminin plus poly-DL-ornithine coated tissue culture plastic in SFD medium with reduced glutamine to avoid conversion to excitotoxic glutamate [Newcomb et al., 1997]. Immunofluorescent staining 3–6d after plating revealed a high proportion of β-3-tubulin positive neurons, as well as glial fibrillary acidic protein (GFAP) positive astrocytes, in cultures induced with SFD medium alone, or with addition of RA during d4–d8 day as aggregates (Fig. 1A–E).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f1.jpg
Neural differentiation with and without retinoic acid

Immunofluorence for β-3-tubulin (green) (A, B) and GFAP (red) (C, D), together with Hoechst 33342 nuclear counterstaining (blue) in plated cultures differentiated in SFD medium without RA (A, C) or with RA (B, D). Scale bars: 25 μm. (E) Plots of β-3-tubulin immunofluorescent cells as a percent of Hoechst stained cell nuclei (mean ± sem) within the first 3d or d4–d6 after plating. Double immunofluorescence for GAP43 (red) and MAP2 (green) in cells induced without RA (F, G, H) or with RA (I, J, K), fixed 1d (F, I), 3d (G, J), or 6d (H, K) after plating. Scale bar: 25 um. (L) Percent GAP43/MAP2 colocalization (mean ± sem). Cultures from RA-treated inductions (solid bars) included a higher proportion of β-3-tubulin-positive neurons and established a polarized distribution of axonal and somatodendritic markers earlier than cells differentiated in SFD medium alone (open bars). * p < 0.01; # p < 0.05.

Polarity

Mature neurons exhibit distinct axonal and somatodendritic compartments; the degree of polarity increases with neuronal maturation [Caceres et al., 1986; Dotti et al., 1988]. We visualized the axonal marker growth associated protein 43 (GAP43) and the somatodendritic marker microtubule-associated protein 2 (MAP2) at several time points after plating. On d1 in culture, both GAP43 and MAP2 were observed within the same neurites of cells induced with or without exposure to RA (Fig. 1F, I, L). Segregation into separate axonal and dendritic compartments was well advanced by d3 after plating in cells induced with RA (Fig. 1J, L). In cells induced without RA, however, colocalization of axonal and dendritic markers declined more gradually, remaining higher than for RA treated cells at d6 after plating (Fig. 1H, L). Full segregation of GAP43 and MAP2 required 7–10d in cultures differentiated without RA (Supplementary Fig. 2). Thus, cells exposed to RA establish polarity more rapidly than cells induced in SFD medium alone, consistent with the idea that neurons induced with RA are more mature, or mature more rapidly, than those which differentiate without RA.

Transmitter Phenotype

To evaluate neurotransmitter phenotypes we used immunofluorescence with antibodies to the inhibitory neurotransmitter γ-aminobutyric acid (GABA) and to the GABA synthetic enzyme, glutamic acid decarboxylase (GAD), as markers for GABAergic neurons [Chang and Gottlieb, 1988]. Cells induced without RA generated a higher proportion of cells that were double-labelled for β-3-tubulin and either GABA (37.2 ± 2.6%, n=27) or GAD (42.8 ± 2.7%, n=30) than cells induced with RA (24.4 ± 1.5%, n=22) and (22.6 ± 1.2%, n=30), respectively (Fig. 2A–D, G). Essentially all GAD or GABA positive cells were also positive for β-3-tubulin regardless of RA exposure. Double immunofluorescence with anti-GAD and anti-GABA demonstrated co-staining in the majority of cells, although the anti-GAD signal was indistinguishable from background in a small percentage of GABA-positive cells (Supplementary Fig. 3). Consistent with the lower proportion of GAD- and GABA-positive neurons, immunofluorescence for the vesicular glutamate transporter 1 (VGLUT1), a marker for excitatory glutamatergic neurons [Bellochio et al., 2000; Takamori et al., 2000], was more prevalent in cultures from RA-treated aggregates (Fig. 2E, F, arrows). Because anti-vGLUT1 only weakly labels the cell soma, however, we did not undertake to quantify the proportions of positive cells.

An external file that holds a picture, illustration, etc.
Object name is nihms96388f2.jpg
Transmitter phenotypes and electrophysiological properties

Double immunofluorescence for β-3-tubulin (green) and either glutamic acid decarboxylase (GAD; A, B) or GABA (red) (C, D) in cells induced without RA (A, C), or with RA (B, D). Double immunofluorescence for β-3-tubulin (green) and vesicular glutamate transporter 1 (VGLUT1) (red), a marker for glutamatergic neurons, in cells induced without (E) or with RA (F). A higher proportion of cells induced with RA was glutamatergic (arrows). Scale bar: 25 um. (G) Percent of β-3-tubulin-positive neurons co-labelled with anti-GABA or anti-GAD in cells induced with SFD medium alone (open bars) or with exposure to RA (solid bars). Cells induced without RA generated a higher proportion of GABAergic neurons than cells induced with RA. * p < 0.01. (H) Whole-cell currents evoked by a voltage step from −80 to 0 mV. (I) Peak inward sodium and outward potassium current densities in neurons differentiated in SFD medium alone (open bars) or with RA (solid bars). (J) Whole-cell currents evoked by 100 μM kainate, GABA, glycine and by 40 μM NMDA in cells induced without (top) or with (bottom) RA. (K) Plots of agonist-evoked current density. All four agonists evoked current in neurons derived from ESCs exposed to RA; however, neurons generated in SFD medium alone failed to express functional glycine receptors. (L) Spontaneous inhibitory synaptic currents recorded at 0 mV and excitatory synaptic currents recorded at −80 mV in neurons differentiated in SFD medium alone. (M) Reverse transcription-polymerase chain reaction (RT-PCR) analysis of the expression of glycine receptor subunits, α1, α2, α3, α4, and β in cells induced without RA or with RA. S: cells induced in SFD medium alone; R: cells induced in SFD medium with RA from d4–d8 in suspension. Glyceraldehyde-3-phosphate dehydrogenase (lane G) was used as a control. Cells induced with RA expressed all of the glycine receptor subunits except α4, whereas only the β subunit and α2 embryonic subunit were detected in cells induced without RA exposure.

Electrophysiology

Whole-cell patch clamp recordings were used to compare the physiological characteristics of neurons induced without or with exposure to RA. Nearly all cells with neuronal morphology expressed voltage-gated sodium and potassium currents, when analyzed 7–14d after plating (Fig. 2H, I). In addition, currents were evoked in the cells by brief applications of GABA, acting on GABAA receptors, and by kainic acid and NMDA, which are agonists for two different glutamate receptor subtypes (Fig. 2J). Functional synaptic transmission between neurons generated in SFD medium was detected after ~3 weeks in culture (Fig. 2L). Interestingly, the inhibitory neurotransmitter glycine failed to elicit detectible current in the majority of neurons induced in SFD medium without exposure to RA (Fig. 2J, K). By contrast, neurons exposed to RA during induction in SFD medium expressed functional glycine receptors, as previously described for differentiation with serum [Bain et al. 1995].

The difference in glycine-evoked current between cells induced with RA or without RA led us to examine the expression of glycine receptor subunits by RT-PCR. Glycine receptors comprise pentameric subunit complexes that include α and β subunits. Only α subunits generate functional homo-oligomeric glycine receptors; β subunits do not form receptors by themselves [Kuhse et al., 1993]. Thus far, four α subunits and one β subunit have been identified. In adult rats the α1 and α3 subunits are highly expressed in the spinal cord and brain stem, and are also found in the superior and inferior colliculi and in some regions of the thalamus and hypothalamus [Malosio et al., 1991]. The α2 subunit is expressed throughout most of the embryonic CNS including cortex, thalamus and hippocampus [Akagi et al., 1991], but declines sharply after birth with more restricted expression in adults, notably in the retina, and in auditory brain stem nuclei [Piechotta et al., 2001]. The α4 subunit is a component of embryonic glycine receptors in the spinal cord, sympathetic ganglia and dorsal root ganglia [Harvey et al., 2000]. Cells induced with RA expressed all of the subunits except α4, whereas cells induced without RA expressed only small amounts of α2 and the β subunits (Fig. 2M). These results are consistent with our electrophysiological recordings. Together they suggest that cells induced without RA may be physiologically less mature than cells induced with RA, or may correspond to cell types that do not express functional glycine receptors in vivo.

Transcriptional Profiles

To gain a broader perspective on phenotypic differentiation we used microarray analysis to test for genome-wide transcriptional differences between cells induced with SFD medium alone or with exposure to RA from d4–d8 in suspension (Fig. 3A). For comparison, we also isolated RNA from replicate plates of undifferentiated ESCs. Probes were hybridized to Affymetrix Mouse genome arrays (MOE430v2). Array scans were analyzed with Affymetrix MAS 5.0 software and dChip, a model-based program for pooling and statistical comparison across multiple chip data sets [Li and Wong, 2001].

An external file that holds a picture, illustration, etc.
Object name is nihms96388f3.jpg
Transcriptional Profiles

(A) Mouse ESCs were maintained in suspension for 12d in serum free differentiation medium alone (SFD), or with exposure to RA from d4–d8 (RA). mRNA for microarray analysis was prepared from differentiated cells 6 days after plating on laminin plus poly-ornithine, and from feeder-free undifferentiated mouse ESCs (ES) maintained in DMEM plus serum and LIF. (B) Venn diagram showing shared and distinct genes among undifferentiated ESCs (ES, blue), cells induced without RA (SFD, green) and cells induced with exposure to RA (RA, red). (C) Genes specifically enriched in ES, in SFD, in both SFD and RA, and in the RA cell populations. Replicate microarray heat plots for the 50 genes with the highest fold enrichment (lower bound) in each population ordered from highest (top) to lowest (bottom). Genes specifically enriched in neurons (green), astrocytes (yellow), or oligodendrocytes (red) are highlighted. A black dot in the colored box indicates a match with the specific probe set identified by Cahoy et al. (2008) as cell type enriched. Bar plots illustrate a weighted index of cell type-specific expression for all of the genes in each population.

A common pool of 2496 genes were elevated in cells differentiated with SFD medium alone and in cells that were exposed to RA, including many encoding for known neuronal proteins (Fig. 3B, C; Supplementary Data File #1). In addition, 557 genes were enriched only in cells that underwent differentiation in SFD medium without RA, whereas 1202 genes were enriched only in cultures from inductions that had been exposed to RA. Fig. 3C presents the 50 genes most enriched in the undifferentiated (ES) or the differentiated (SFD, RA) cell populations, sorted from highest to lowest fold enrichment (lower bound). Based on a recent transcriptome analysis of neurons, astrocytes and oligodendrocytes [Cahoy et al., 2008] many of the genes enriched in our differentiated cell populations (SFD, RA) are specific for particular neural cell types. A high proportion of genes enriched in astrocytes are specifically up regulated in cells differentiated with SFD medium alone, whereas many neuron-specific genes are selectively up regulated in cells that received brief exposure to RA (Fig. 3C).

Genes enriched both in ESCs and in cultures differentiated with SFD medium alone (449) included many associated with cell division, such as cyclin A2 and D1, geminin, and wee 1. By contrast, only a few genes were enriched in common in both undifferentiated ESCs and in cultures differentiated with RA (107). The 5320 genes common among all three cell populations included many “house-keeping” proteins such as clathrin, β and γ actin, Na+/K+ ATPase subunits and 27 different ribosomal proteins; however, comparison with our other enrichment profiles revealed that many of the genes identified as specifically enriched in one or two out of the three cell populations were also included. Thus, a substantial number of these genes may be expressed within all three populations, but show significant up-regulation in one or two out of the three. The Venn diagram in Figure 3B lists the number of genes enriched in each category but also shows the percentage that is unique to that category. For example, of the 557 genes enriched in cultures differentiated in SFD medium alone, 38.6% were only represented in that category while the remaining 61.4% met the criteria for inclusion in the central pool of 5320 genes expressed to some extent in all three populations. Complete lists of all genes in each category are provided in the Gene Table Supplementary Data File.

Consistent with our evidence for the relative immaturity of cultures induced with SFD medium alone, the genes enriched in this population included many associated with early neural precursors, such as neurogenin1, neurod1, Dlx1, Pax6 and Eomes. In contrast, the list of genes specific to inductions that involved exposure to RA included components of mature neurons, such as synaptic molecules, receptors, channels and neurotransmitters. To visualize these differences in functionality we used gene ontology (GO) terms [Ashburner et al., 2000]. Fig. 4A highlights GO terms with significant prevalence among genes specifically enriched in SFD medium alone (green), in RA-treated cells (red), or genes enriched in both differentiated cell populations (yellow) relative to ESCs. Genes associated with brain development, neural differentiation, chromatin modification and Wnt receptor signalling were enriched in the cultures differentiated with SFD medium alone, whereas cultures from inductions that had been exposed to RA exhibited a higher proportion of genes associated with mature neurons (ion transport, pre- and post-synaptic membrane, neurotransmitter secretion)(Fig. 4A; Supplementary Tables 3–5).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f4.jpg
Profile Comparison

(A) Fractional representation of gene sets grouped by Gene Ontology terms. Boxes next to each Gene Ontology term indicate significance for genes specifically enriched in cells induced with SFD medium alone (green), for genes enriched in cells induced with SFD medium plus exposure to RA (red), or genes that were enriched in both populations relative to undifferentiated mES cells (yellow). Note that seven of the Gene Ontology terms (small GTPase-mediated signalling, membrane, vesicle-mediated transport, golgi-apparatus, protein transport, synapse, and ATP biosynthetic process) displayed significant representation both in the RA (red) and in the SFD and RA (yellow) categories, albeit each gene was only assigned to one category (see Table S3–5 for quantitation). (B) Hierarchical clustering based on the relative frequency of Gene Ontology term representation in data sets derived from tissue-derived stem cells, transit amplifying cell types and various differentiated cell populations presented as a dendrogram. Fractional numbers indicate the degree of dissimilarity, calculated as 1-Pearson’s coefficient of similarity, between two profiles at the depicted branch point. Genes specifically enriched in cells induced with SFD plus RA exposure (RA, 1) cluster with profiles of mature tissues, including hippocampus (3) and cerebellum (6); whereas genes enriched in the SFD alone cell cultures relative to RA-treated cells (SFD vs RA, 26) clustered most closely with transit-amplifying cell populations (25, 27, 28). Genes specifically enriched in SFD medium alone (SFD, 24) are intermediate to the transit amplifying and tissue-derived stem cell populations. By contrast, genes enriched in RA-treated cells relative to cells differentiated without RA (RA vs SFD, 2) clustered together with the genes specific for RA-treated cells (RA, 1) and with genes enriched in SFD alone (SFD vs ES, 5), or in RA-treated cells (RA vs ES, 4), relative to the undifferentiated ES cells. (C) Segregation of fully differentiated tissues, tissue-specific stem cell populations, and transit amplifying cell types based on fractional representation of membrane and nucleus GO terms.

Profile Comparison

To compare our transcriptional profiles quantitatively with previous expression profiling studies we used GOurmet software which allows for platform-independent analysis based on GO terms [Doherty et al., 2006; 2008]. Hierarchical clustering revealed that genes specifically enriched in RA-treated cells (RA) cluster with profiles of mature brain tissue (Supplementary Fig. 4) whereas genes specifically enriched in SFD medium alone exhibit the greatest similarity to the profile of E14.5 retinal neuroblasts [Fortunel et al., 2003], a transit-amplifying cell population [Doherty et al., 2008]. Inclusion of profiles from non-neuronal cell types further highlights the distinction between stem, transit-amplifying and mature expression patterns. The dendrogram in Fig. 4B illustrates the relationships among 28 different profiled cell populations including 3 profiles of neural stem cells (numbers 17, 18 and 19) and 7 other tissue-derived stem-like populations (numbers 14–16, 20–23), 3 profiles from transit-amplifying cell populations (numbers 25, 27, 28), and 9 profiles from mature tissues (numbers 3, 6–13; see Supplementary Materials for complete information on each population profile). In addition to the genes specific to SFD alone and to RA-treated cells, the dendrogram in Fig. 4B also displays lists that include genes upregulated in both of the differentiated cell populations (number 5, SFD vs ES; number 4, RA vs ES) relative to undifferentiated ESCs.

To focus on the differences between cells differentiated with and without RA, we also identified 1366 genes enriched in SFD alone cultures relative to RA-treated cells (number 26, SFD vs RA). These genes, which are likely to be expressed by cells present in the cultures differentiated with SFD alone but absent from RA-treated cultures, were found to cluster most closely with the transit-amplifying populations. In contrast, 1883 genes enriched in RA-treated cultures relative to SFD alone (number 2, RA vs SFD) clustered together with genes that were specific to RA-treated cells (number 1, RA). Taken together, our data suggests that cultures differentiated with SFD medium alone include neurons as well as a substantial representation of cells that are relatively immature and retain a high proliferative capacity, typical of transit-amplifying neuroblasts, whereas cultures induced with a period of exposure to RA are relatively more mature and appear to retain fewer precursor cells.

Cell Proliferation

Neural differentiation, both in vivo and in vitro, requires cells to exit the cell cycle and become postmitotic [Ohnuma and Harris, 2003]. Our microarray analysis revealed higher expression of early neural progenitor genes in cultures induced in SFD medium without RA as compared to undifferentiated cells and those induced with exposure to RA. This led us to evaluate the fraction of proliferating cells by incubation with bromodeoxyuridine (BrdU) [Gratzner, 1982]. ESC aggregates were maintained for 12d in SFD medium alone or with addition of RA from d4–d8. Sister cultures received BrdU for periods ranging from 1–24h immediately prior to harvesting at the end of the 12d induction. Cell aggregates induced with or without RA contained more BrdU positive cells as BrdU exposure time increased; however, the proportion of labelled cells in SFD medium alone was consistently higher than in cultures that had received RA (Fig. 5A–H). After 24h of BrdU treatment, almost 50% of cells in EBs induced without RA showed positive BrdU signal, whereas only 25% of cells in EBs induced with RA were labelled with BrdU (Fig. 5I). We next compared proliferation on d1–4 after plating by adding BrdU to plated cultures 2h prior to fixation. Cultures induced with SFD medium alone showed a high proportion of BrdU positive cells (~41%) on d1, and lower proportions on subsequent days. However, cells induced with RA showed much lower BrdU labelling (~5–8%) at all time points (Fig. 5J–L). Collectively, our results suggest that SFD medium alone supports a higher level of continued neural precursor proliferation, whereas exposure to RA promotes exit from the cell cycle and speeds neuronal differentiation and maturation.

An external file that holds a picture, illustration, etc.
Object name is nihms96388f5.jpg
Cell Proliferation

ESC aggregates maintained for 12d in SFD medium alone (A–D) or with exposure to RA from d4–d8 in suspension (E–H) were incubated with BrdU (20 μg/ml) for 2h (A, E); 4h (B, F); 8h (C, G) or 24h (D, H). (I) Percent of Hoechst stained nuclei (blue) in 12d aggregate cultures that were immunopositive for BrdU (red) plotted as a function of BrdU exposure time. Incorporation of BrdU in dissociated cells 4d after plating from aggregates maintained 12d in SFD medium alone (J) or with RA exposure from d4–d8 (K). (L) Percent of immunopositive cells following 2h incubation with BrdU evaluated in different cultures 1–4d after plating. Scale bars: 25 μm Cells exposed to RA displayed lower BrdU labelling both in floating aggregates and in plated cultures, indicating a lower rate of proliferation.

Neural Progenitors

Neurons in the developing CNS are generated from at least three types of progenitor cells, neuroepithelial cells, basal or intermediate progenitors, and radial glial cells, which express different molecular markers, [Doetsch 2003; Goldman, 2003]. Neuroepithelial cells express the intermediate filament nestin [Lendahl et al., 1990] and its posttranslational modifications recognized by the antibodies RC1 and RC2 [Malatesta et al., 2003]. Radial glia also express nestin as well as the transcription factor Pax6 [Gotz and Barde, 2005; Heins et al., 2002]. Unlike neuroepithelial cells and radial glia, basal progenitors express the T box transcription factor Eomes, also known as Tbr2 [Englund et al., 2005]. Expression of these progenitor-associated genes persists in proliferative neural precursors in the adult brain [Hodge et al., 2008]. Previous studies have highlighted the conversion of ESCs into radial glia [Liour and Yu, 2003; Liour et al., 2006; Bibel et al., 2004; Bouhon et al., 06], but have not addressed the production of basal progenitors. Our transcriptional profiles indicated that expression of both Pax6 and Tbr2 was significantly up regulated in cells induced with SFD medium alone, although Pax6 (but not Tbr2) was expressed at lower levels in RA-treated cells as well as undifferentiated ESCs.

To visualize the distribution of these progenitor cell markers we performed double immunofluorescence on ESCs induced in SFD medium without or with exposure to RA (Fig. 6). After 12d in suspension, many cells induced in SFD medium alone expressed only Pax6 (45 ± 5%) or only Tbr2 (13 ± 3%), although a small proportion of cells was found to be immunoreactive for both Pax6 and Tbr2 (10 ± 2%) (n=7; Fig. 6A). Approximately 36 ± 3% of cells induced with exposure to RA expressed Pax6, but Tbr2 was not detected in RA treated cells (n=5; Fig. 6B). At 3h after plating, cells expressing both nestin and RC2 were observed in cultures induced with RA or without RA (Fig 6C, D). Pax6 also was expressed in cells induced with RA or without RA, and many of the Pax6-positive cells coexpressed vimentin or RC2 (Fig. 6E–H). Tbr2 immunofluorescence was observed only by cells induced in SFD medium without RA, and some of the Tbr2-positive cells were immunoreactive for vimentin or RC2 (Supplementary Fig. 5A, B). Furthermore, the neural precursor population displaying strong vimentin immuno-fluorescence was distinct from neurons labelled with antibodies to β-3-tubulin (Supplementary Fig. 5C). Together, these data suggest that exposure to RA alters the specification of neural precursors, blocking the production of basal progenitors.

An external file that holds a picture, illustration, etc.
Object name is nihms96388f6.jpg
Neural Progenitors

ESC aggregates maintained for 12d in SFD medium alone (A) contained cells immunopositive for Pax6 (red), a marker expressed in radial glia, or for Tbr2 (green), which is expressed in basal progenitor cells. In contrast, RA-treated aggregates lacked Tbr2-positive cells (B) Scale Bar: 50 μm. (C–H) After dissociation, plating, and 3h in culture, numerous cells induced without (C, E, G) or with (D, F, H) exposure to RA were co-labelled with antibodies to RC2 and nestin (C, D), or with Pax6 and RC2 (E, F) or vimentin (G, H), suggesting that the majority of Pax6-positive cells are likely to be radial glia. Scale Bar: 25 μm.

Neuronal Phenotype Specification

Previous work suggests that exposure to RA can promote caudalization and suppress rostral neural structures [Maden, 2007; Wichterle et al., 2002; Bouhon et al., 2006]. To infer the positional identity of differentiated ESC, we compared our upregulated gene lists with a spatial database of gene expression from the mouse genome informatics (MGI) website, which contains comprehensive information about expression patterns [Smith et al., 2007]. As illustrated in figure 7, genes expressed in telencephalon, diencephalon and mesencephalon were more prevalent in cells differentiated in SFD medium alone relative to cultures treated with RA (Fig. 7A), whereas genes expressed in rhombencephalon and spinal cord predominated in cells differentiated with RA exposure (Fig. 7B). Interestingly, genes involved with regulating transcription were among the transcripts most highly enriched in each cell population (Fig. 7A, B).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f7.jpg
Positional Identity

(A) Replicate microarray heat plots for the 50 genes with differential anatomical expression patterns that showed the highest fold enrichment (lower bound) in cells differentiated in SFD medium alone relative to differentiation with exposure to RA, and (B) the 50 genes with differential anatomical expression that were most enriched in RA-treated cells relative to cells induced with SFD medium alone. Genes expressed in telencephalon (T, blue), diencephalon (D, green), mesencephalon (M, yellow), rhombencephalon (R, orange), and spinal cord (S, rose) in the mouse genome informatics anatomical expression database are indicated. Bar plots illustrate a weighted index of gene expression for each anterior-posterior (AP) anatomical domain. The AP index distributions were significantly different for genes enriched in the two differentiated cell populations (Chi-square, p<0.0001). T, D and M genes were enriched in cells induced with SFD medium alone, whereas R and S genes were enriched in cells differentiated with exposure to RA (z-test comparison of proportions, p ≤ 0.023). Genes specifically enriched in neurons (N, green), astrocytes (A, yellow), or oligodendrocytes (O, red) are highlighted. Plots show a weighted index of cell type-specific expression for all of the genes in each population, which were significantly different in (A) and (B) (Chi-square, p=0.0007). Genes expressed in astrocytes were enriched in cells differentiated with SFD medium alone (z-test comparison of proportions, p < 0.001).

To gain further insight into the positional identity of ESCs induced with RA or without RA we used antibodies to transcription factors associated with the differentiation of specific neuronal populations. In cultures induced with SFD medium for 10–12d followed by 4d adherent culture, 35 ± 3% of cells coexpressed β-3-tubulin and Tbr1 (Fig. 8A), a transcription factor expressed in the dorsal telencephalon (pallium)[Hevner et al., 2001; Remedios et al., 2007] and restricted to glutamatergic neurons [Hevner et al., 2001]. Accordingly, nearly all cells that strongly expressed the vesicular glutamate transporter1 (VGLUT1), were positive for Tbr1 (Fig. 8B) whereas GABAergic neurons, expressing glutamic acid decarboxylase (GAD), were uniformly negative for Tbr1 (Fig. 8C). Instead, many of the GAD-positive neurons expressed Islet1 (Fig. 8C), which is a LIM homeodomain transcription factor expressed in the spinal cord and ventral regions of the telencephalon, diencephalon, mesencephalon, and rhombencephalon [Ericson et al., 1995; Pfaff et al., 1996]. Islet1-positive cells become motorneurons in mesencephalon, rhombencephalon and spinal cord [Pfaff et al., 1996; Simon et al., 1994] but not in telencephalon and diencephalon [Ericson et al., 1995]. In cultures induced with SFD medium alone, 30 ± 3% of β-3-tubulin-positive cells expressed Islet1 and over 80% of cells immunopositive for Islet1 and β-3-tubulin also expressed GAD (Fig. 8C), which is prevalent in the subpallium/ventral telencephalon [Muzio et al., 2002; Yun et al., 2001]. Diencephalic Islet1-positive neurons can be distinguished from telencephalic Islet1-positive neurons by their expression of LIM homeodomain transcription factor, Lim1 [Ericson et al., 1995; Fotaki et al., 2006]. In the cultures differentiated with SFD medium alone, approximately 20% of cells were immunopositive for Lim1/2; however, none of the Islet1-positive neurons were co-labelled for Lim1 (Supplementary Fig. 6). These data suggest that Islet1-positive neurons in cultures differentiated with SFD medium alone acquire ventral telencephalic identity in the absence of RA. A few cells in cultures induced with SFD medium alone expressed Pax2 (3 ± 0.5%) or Engrail1 (1 ± 0.4%) which are typical markers for midbrain [Davis and Joyner, 1988; Matsunaga et al., 2000], whereas no cells were found to be positive for Hoxb4, a marker for the caudal hindbrain and spinal cord [Graham et al., 1988; Nordstrom et al., 2006] (Supplementary Fig. 7).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f8.jpg
Neuronal Phenotype Specification

(A–C) In plated cultures differentiated with SFD medium alone, distinct β-3-tubulin-positive (blue) neuron populations were immuno-labelled with the dorsal pallial marker Tbr1 (Ar, arrows, red) or with the subpallial marker Islet1 (Ag, arrowheads, green). Scale Bar: 50 μm. (B) Cells strongly immunopositive for the glutamatergic marker VGLUT1 (green) were co-labelled with Tbr1 (red), (C) whereas GAD-positive GABAergic cells (blue) co-labelled with Islet1 (green), but not Tbr1 (red). Scale Bar: 25 μm. (D–F) In plated cultures differentiated with brief exposure to RA, ~22% β-3-tubulin-positive neurons (blue) co-labelled with Evx1/2 (red) and Lim1/2 (green)(Dr and Dg, arrows), markers for V0 ventral spinal neurons, whereas ~18% co-expressed markers of dI4 and dI6 dorsal spinal neurons Pax2 (red) and Lim1/2 (green)(Er and Eg, arrows). (F) Most Pax2-positive (blue) neurons were co-labelled for GAD (red) as well as Lim1/2 (green). Scale Bar: 25 μm. (G) Quantitation of immunopositive cells differentiated with SFD medium alone (open bars), or with exposure to RA (solid bars); 21–35 fields each condition, 462 ± 13 cells/field. The dorsal pallial marker Tbr1 and subpallial marker Islet1 together account for ~65% of all β-3-tubulin-positive neurons in SFD alone cultures. In cultures prepared from RA-treated aggregates, ~40% of β-3-tubulin-positive neurons co-labelled with Pax2 and Lim1/2 typical of dI4/dI6 neurons, or with Evx1/2 and Lim1/2 characteristic of V0 neurons.

In contrast to differentiation with SFD medium alone, which led cells to adopt anterior phenotypes, in cultures induced with exposure to RA, almost 90% of β-3-tubulin positive cells were immunopositive for Hoxb4 (Supplementary Fig. 7), consistent with acquisition of caudal positional identity for the hindbrain or spinal cord [Gould et al., 1997; Nordstrom et al., 2006]. To investigate subtype identity of neurons produced following exposure to RA, we used antibodies to transcription factors with regionally restricted expression patterns along the dorsoventral axis of the spinal cord [Briscoe et al., 2000; Helms and Johnson, 2003]. Evx1/2 is expressed in specific dorsal (dI1) and ventral (V0) interneuron populations [Briscoe et al., 2000; Pierani et al., 1999] that can be distinguished by co-expression of Lim1/2 in ventral Evx1/2-positive cells [Pierani et al., 1999], whereas intervening dorsal interneuron populations (dI4, dI5) co-express Pax2 and Lim1/2 [Gross et al., 2002]. Among β-3-tubulin positive cells induced in SFD medium with exposure to RA, 22 ± 0.8% were positive for both Evx1/2 and Lim1/2 (Fig. 8D arrows), suggesting that they acquired V0 fate; 18 ± 1.2% expressed both Pax2 and Lim1/2, but not Evx1/2 (Fig. 8E arrows), suggesting that they are dI4 or dI6 neurons. In the developing spinal cord, dorsal Pax2-positive neurons are GABAergic [Cheng et al., 2004]. Most Pax2-positive, Lim1/2-positive neurons derived from ESCs with exposure to RA expressed the GABAergic neuronal marker, GAD (Fig. 8F arrows). A few β-3-tubulin-positive cells (~10%) were immunopositive for transcription factors that are expressed by dorsal interneuron subtype dI1 (Evx1) [Pierani et al., 1999] or dI2 (Brn3a and Lim1/2) [Muller et al., 2002], or ventral interneuron subtype V1 (En1) (Ericson et al., 1997; Matise and Joyner, 1997), V3 (Nkx2.2) [Briscoe et al., 1999], or motorneurons (Islet1) [Pfaff et al., 1996] (Fig. 8G). Together, these data suggest that a large proportion of neurons (~60–70%) induced with SFD medium alone acquire anterior neuronal identity, whereas neurons induced with exposure to RA express molecular markers with regionally restricted expression patterns in the spinal cord (Supplementary Fig. 8).

Neural Differentiation

In previous work [Finley et al., 1999] we demonstrated relatively efficient neural differentiation when ESC aggregates were maintained for 8d in serum-free Neurobasal Medium plus the B27 supplement, which contains 20 different ingredients including retinyl acetate [Brewer et al., 1993]; however, generation of neurons was dramatically curtailed when complete B27 supplement was replaced with B27 that lacked retinyl acetate [Finley et al., 1999]. To investigate the effects of retinoids on ESC neural differentiation, we developed a modification of earlier serum-free induction media [Wiles and Johansson, 1999; Tropepe et al., 2001; Ying et al., 2003] that lacks exogenous retinoids and avoids the use of proprietary growth supplements or media formulations (Supplementary Table 1).

To initiate induction, feeder-free ESC monolayers were detached with protease, rinsed extensively with balanced salt solution, suspended in serum-free differentiation medium (SFD; Supplementary Table 1), and dispersed into non-adherent Petri dishes. Under these conditions the cells formed floating aggregates that gradually enlarged by cell division. Expression of Oct3/4, which defines the undifferentiated state [Niwa et al., 2000], progressively declined (Supplementary Fig. 1), while there was a gradual increase in cells expressing proteins associated with neurons or neural precursors including nestin [Lendahl et al., 1990] and β-3-tubulin [Caccamo et al., 1989] (Supplementary Fig. 1). Cells maintained exclusively in SFD medium acquired neural properties over a time course of 10–12d, whereas addition of 0.5 μM RA from d4–d8 speeded differentiation, as evidenced by a higher proportion of β-tubulin-expressing cells at earlier time points. Aggregates maintained in suspension for 12d were dissociated to single cells and plated onto laminin plus poly-DL-ornithine coated tissue culture plastic in SFD medium with reduced glutamine to avoid conversion to excitotoxic glutamate [Newcomb et al., 1997]. Immunofluorescent staining 3–6d after plating revealed a high proportion of β-3-tubulin positive neurons, as well as glial fibrillary acidic protein (GFAP) positive astrocytes, in cultures induced with SFD medium alone, or with addition of RA during d4–d8 day as aggregates (Fig. 1A–E).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f1.jpg
Neural differentiation with and without retinoic acid

Immunofluorence for β-3-tubulin (green) (A, B) and GFAP (red) (C, D), together with Hoechst 33342 nuclear counterstaining (blue) in plated cultures differentiated in SFD medium without RA (A, C) or with RA (B, D). Scale bars: 25 μm. (E) Plots of β-3-tubulin immunofluorescent cells as a percent of Hoechst stained cell nuclei (mean ± sem) within the first 3d or d4–d6 after plating. Double immunofluorescence for GAP43 (red) and MAP2 (green) in cells induced without RA (F, G, H) or with RA (I, J, K), fixed 1d (F, I), 3d (G, J), or 6d (H, K) after plating. Scale bar: 25 um. (L) Percent GAP43/MAP2 colocalization (mean ± sem). Cultures from RA-treated inductions (solid bars) included a higher proportion of β-3-tubulin-positive neurons and established a polarized distribution of axonal and somatodendritic markers earlier than cells differentiated in SFD medium alone (open bars). * p < 0.01; # p < 0.05.

Polarity

Mature neurons exhibit distinct axonal and somatodendritic compartments; the degree of polarity increases with neuronal maturation [Caceres et al., 1986; Dotti et al., 1988]. We visualized the axonal marker growth associated protein 43 (GAP43) and the somatodendritic marker microtubule-associated protein 2 (MAP2) at several time points after plating. On d1 in culture, both GAP43 and MAP2 were observed within the same neurites of cells induced with or without exposure to RA (Fig. 1F, I, L). Segregation into separate axonal and dendritic compartments was well advanced by d3 after plating in cells induced with RA (Fig. 1J, L). In cells induced without RA, however, colocalization of axonal and dendritic markers declined more gradually, remaining higher than for RA treated cells at d6 after plating (Fig. 1H, L). Full segregation of GAP43 and MAP2 required 7–10d in cultures differentiated without RA (Supplementary Fig. 2). Thus, cells exposed to RA establish polarity more rapidly than cells induced in SFD medium alone, consistent with the idea that neurons induced with RA are more mature, or mature more rapidly, than those which differentiate without RA.

Transmitter Phenotype

To evaluate neurotransmitter phenotypes we used immunofluorescence with antibodies to the inhibitory neurotransmitter γ-aminobutyric acid (GABA) and to the GABA synthetic enzyme, glutamic acid decarboxylase (GAD), as markers for GABAergic neurons [Chang and Gottlieb, 1988]. Cells induced without RA generated a higher proportion of cells that were double-labelled for β-3-tubulin and either GABA (37.2 ± 2.6%, n=27) or GAD (42.8 ± 2.7%, n=30) than cells induced with RA (24.4 ± 1.5%, n=22) and (22.6 ± 1.2%, n=30), respectively (Fig. 2A–D, G). Essentially all GAD or GABA positive cells were also positive for β-3-tubulin regardless of RA exposure. Double immunofluorescence with anti-GAD and anti-GABA demonstrated co-staining in the majority of cells, although the anti-GAD signal was indistinguishable from background in a small percentage of GABA-positive cells (Supplementary Fig. 3). Consistent with the lower proportion of GAD- and GABA-positive neurons, immunofluorescence for the vesicular glutamate transporter 1 (VGLUT1), a marker for excitatory glutamatergic neurons [Bellochio et al., 2000; Takamori et al., 2000], was more prevalent in cultures from RA-treated aggregates (Fig. 2E, F, arrows). Because anti-vGLUT1 only weakly labels the cell soma, however, we did not undertake to quantify the proportions of positive cells.

An external file that holds a picture, illustration, etc.
Object name is nihms96388f2.jpg
Transmitter phenotypes and electrophysiological properties

Double immunofluorescence for β-3-tubulin (green) and either glutamic acid decarboxylase (GAD; A, B) or GABA (red) (C, D) in cells induced without RA (A, C), or with RA (B, D). Double immunofluorescence for β-3-tubulin (green) and vesicular glutamate transporter 1 (VGLUT1) (red), a marker for glutamatergic neurons, in cells induced without (E) or with RA (F). A higher proportion of cells induced with RA was glutamatergic (arrows). Scale bar: 25 um. (G) Percent of β-3-tubulin-positive neurons co-labelled with anti-GABA or anti-GAD in cells induced with SFD medium alone (open bars) or with exposure to RA (solid bars). Cells induced without RA generated a higher proportion of GABAergic neurons than cells induced with RA. * p < 0.01. (H) Whole-cell currents evoked by a voltage step from −80 to 0 mV. (I) Peak inward sodium and outward potassium current densities in neurons differentiated in SFD medium alone (open bars) or with RA (solid bars). (J) Whole-cell currents evoked by 100 μM kainate, GABA, glycine and by 40 μM NMDA in cells induced without (top) or with (bottom) RA. (K) Plots of agonist-evoked current density. All four agonists evoked current in neurons derived from ESCs exposed to RA; however, neurons generated in SFD medium alone failed to express functional glycine receptors. (L) Spontaneous inhibitory synaptic currents recorded at 0 mV and excitatory synaptic currents recorded at −80 mV in neurons differentiated in SFD medium alone. (M) Reverse transcription-polymerase chain reaction (RT-PCR) analysis of the expression of glycine receptor subunits, α1, α2, α3, α4, and β in cells induced without RA or with RA. S: cells induced in SFD medium alone; R: cells induced in SFD medium with RA from d4–d8 in suspension. Glyceraldehyde-3-phosphate dehydrogenase (lane G) was used as a control. Cells induced with RA expressed all of the glycine receptor subunits except α4, whereas only the β subunit and α2 embryonic subunit were detected in cells induced without RA exposure.

Electrophysiology

Whole-cell patch clamp recordings were used to compare the physiological characteristics of neurons induced without or with exposure to RA. Nearly all cells with neuronal morphology expressed voltage-gated sodium and potassium currents, when analyzed 7–14d after plating (Fig. 2H, I). In addition, currents were evoked in the cells by brief applications of GABA, acting on GABAA receptors, and by kainic acid and NMDA, which are agonists for two different glutamate receptor subtypes (Fig. 2J). Functional synaptic transmission between neurons generated in SFD medium was detected after ~3 weeks in culture (Fig. 2L). Interestingly, the inhibitory neurotransmitter glycine failed to elicit detectible current in the majority of neurons induced in SFD medium without exposure to RA (Fig. 2J, K). By contrast, neurons exposed to RA during induction in SFD medium expressed functional glycine receptors, as previously described for differentiation with serum [Bain et al. 1995].

The difference in glycine-evoked current between cells induced with RA or without RA led us to examine the expression of glycine receptor subunits by RT-PCR. Glycine receptors comprise pentameric subunit complexes that include α and β subunits. Only α subunits generate functional homo-oligomeric glycine receptors; β subunits do not form receptors by themselves [Kuhse et al., 1993]. Thus far, four α subunits and one β subunit have been identified. In adult rats the α1 and α3 subunits are highly expressed in the spinal cord and brain stem, and are also found in the superior and inferior colliculi and in some regions of the thalamus and hypothalamus [Malosio et al., 1991]. The α2 subunit is expressed throughout most of the embryonic CNS including cortex, thalamus and hippocampus [Akagi et al., 1991], but declines sharply after birth with more restricted expression in adults, notably in the retina, and in auditory brain stem nuclei [Piechotta et al., 2001]. The α4 subunit is a component of embryonic glycine receptors in the spinal cord, sympathetic ganglia and dorsal root ganglia [Harvey et al., 2000]. Cells induced with RA expressed all of the subunits except α4, whereas cells induced without RA expressed only small amounts of α2 and the β subunits (Fig. 2M). These results are consistent with our electrophysiological recordings. Together they suggest that cells induced without RA may be physiologically less mature than cells induced with RA, or may correspond to cell types that do not express functional glycine receptors in vivo.

Transcriptional Profiles

To gain a broader perspective on phenotypic differentiation we used microarray analysis to test for genome-wide transcriptional differences between cells induced with SFD medium alone or with exposure to RA from d4–d8 in suspension (Fig. 3A). For comparison, we also isolated RNA from replicate plates of undifferentiated ESCs. Probes were hybridized to Affymetrix Mouse genome arrays (MOE430v2). Array scans were analyzed with Affymetrix MAS 5.0 software and dChip, a model-based program for pooling and statistical comparison across multiple chip data sets [Li and Wong, 2001].

An external file that holds a picture, illustration, etc.
Object name is nihms96388f3.jpg
Transcriptional Profiles

(A) Mouse ESCs were maintained in suspension for 12d in serum free differentiation medium alone (SFD), or with exposure to RA from d4–d8 (RA). mRNA for microarray analysis was prepared from differentiated cells 6 days after plating on laminin plus poly-ornithine, and from feeder-free undifferentiated mouse ESCs (ES) maintained in DMEM plus serum and LIF. (B) Venn diagram showing shared and distinct genes among undifferentiated ESCs (ES, blue), cells induced without RA (SFD, green) and cells induced with exposure to RA (RA, red). (C) Genes specifically enriched in ES, in SFD, in both SFD and RA, and in the RA cell populations. Replicate microarray heat plots for the 50 genes with the highest fold enrichment (lower bound) in each population ordered from highest (top) to lowest (bottom). Genes specifically enriched in neurons (green), astrocytes (yellow), or oligodendrocytes (red) are highlighted. A black dot in the colored box indicates a match with the specific probe set identified by Cahoy et al. (2008) as cell type enriched. Bar plots illustrate a weighted index of cell type-specific expression for all of the genes in each population.

A common pool of 2496 genes were elevated in cells differentiated with SFD medium alone and in cells that were exposed to RA, including many encoding for known neuronal proteins (Fig. 3B, C; Supplementary Data File #1). In addition, 557 genes were enriched only in cells that underwent differentiation in SFD medium without RA, whereas 1202 genes were enriched only in cultures from inductions that had been exposed to RA. Fig. 3C presents the 50 genes most enriched in the undifferentiated (ES) or the differentiated (SFD, RA) cell populations, sorted from highest to lowest fold enrichment (lower bound). Based on a recent transcriptome analysis of neurons, astrocytes and oligodendrocytes [Cahoy et al., 2008] many of the genes enriched in our differentiated cell populations (SFD, RA) are specific for particular neural cell types. A high proportion of genes enriched in astrocytes are specifically up regulated in cells differentiated with SFD medium alone, whereas many neuron-specific genes are selectively up regulated in cells that received brief exposure to RA (Fig. 3C).

Genes enriched both in ESCs and in cultures differentiated with SFD medium alone (449) included many associated with cell division, such as cyclin A2 and D1, geminin, and wee 1. By contrast, only a few genes were enriched in common in both undifferentiated ESCs and in cultures differentiated with RA (107). The 5320 genes common among all three cell populations included many “house-keeping” proteins such as clathrin, β and γ actin, Na+/K+ ATPase subunits and 27 different ribosomal proteins; however, comparison with our other enrichment profiles revealed that many of the genes identified as specifically enriched in one or two out of the three cell populations were also included. Thus, a substantial number of these genes may be expressed within all three populations, but show significant up-regulation in one or two out of the three. The Venn diagram in Figure 3B lists the number of genes enriched in each category but also shows the percentage that is unique to that category. For example, of the 557 genes enriched in cultures differentiated in SFD medium alone, 38.6% were only represented in that category while the remaining 61.4% met the criteria for inclusion in the central pool of 5320 genes expressed to some extent in all three populations. Complete lists of all genes in each category are provided in the Gene Table Supplementary Data File.

Consistent with our evidence for the relative immaturity of cultures induced with SFD medium alone, the genes enriched in this population included many associated with early neural precursors, such as neurogenin1, neurod1, Dlx1, Pax6 and Eomes. In contrast, the list of genes specific to inductions that involved exposure to RA included components of mature neurons, such as synaptic molecules, receptors, channels and neurotransmitters. To visualize these differences in functionality we used gene ontology (GO) terms [Ashburner et al., 2000]. Fig. 4A highlights GO terms with significant prevalence among genes specifically enriched in SFD medium alone (green), in RA-treated cells (red), or genes enriched in both differentiated cell populations (yellow) relative to ESCs. Genes associated with brain development, neural differentiation, chromatin modification and Wnt receptor signalling were enriched in the cultures differentiated with SFD medium alone, whereas cultures from inductions that had been exposed to RA exhibited a higher proportion of genes associated with mature neurons (ion transport, pre- and post-synaptic membrane, neurotransmitter secretion)(Fig. 4A; Supplementary Tables 3–5).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f4.jpg
Profile Comparison

(A) Fractional representation of gene sets grouped by Gene Ontology terms. Boxes next to each Gene Ontology term indicate significance for genes specifically enriched in cells induced with SFD medium alone (green), for genes enriched in cells induced with SFD medium plus exposure to RA (red), or genes that were enriched in both populations relative to undifferentiated mES cells (yellow). Note that seven of the Gene Ontology terms (small GTPase-mediated signalling, membrane, vesicle-mediated transport, golgi-apparatus, protein transport, synapse, and ATP biosynthetic process) displayed significant representation both in the RA (red) and in the SFD and RA (yellow) categories, albeit each gene was only assigned to one category (see Table S3–5 for quantitation). (B) Hierarchical clustering based on the relative frequency of Gene Ontology term representation in data sets derived from tissue-derived stem cells, transit amplifying cell types and various differentiated cell populations presented as a dendrogram. Fractional numbers indicate the degree of dissimilarity, calculated as 1-Pearson’s coefficient of similarity, between two profiles at the depicted branch point. Genes specifically enriched in cells induced with SFD plus RA exposure (RA, 1) cluster with profiles of mature tissues, including hippocampus (3) and cerebellum (6); whereas genes enriched in the SFD alone cell cultures relative to RA-treated cells (SFD vs RA, 26) clustered most closely with transit-amplifying cell populations (25, 27, 28). Genes specifically enriched in SFD medium alone (SFD, 24) are intermediate to the transit amplifying and tissue-derived stem cell populations. By contrast, genes enriched in RA-treated cells relative to cells differentiated without RA (RA vs SFD, 2) clustered together with the genes specific for RA-treated cells (RA, 1) and with genes enriched in SFD alone (SFD vs ES, 5), or in RA-treated cells (RA vs ES, 4), relative to the undifferentiated ES cells. (C) Segregation of fully differentiated tissues, tissue-specific stem cell populations, and transit amplifying cell types based on fractional representation of membrane and nucleus GO terms.

Profile Comparison

To compare our transcriptional profiles quantitatively with previous expression profiling studies we used GOurmet software which allows for platform-independent analysis based on GO terms [Doherty et al., 2006; 2008]. Hierarchical clustering revealed that genes specifically enriched in RA-treated cells (RA) cluster with profiles of mature brain tissue (Supplementary Fig. 4) whereas genes specifically enriched in SFD medium alone exhibit the greatest similarity to the profile of E14.5 retinal neuroblasts [Fortunel et al., 2003], a transit-amplifying cell population [Doherty et al., 2008]. Inclusion of profiles from non-neuronal cell types further highlights the distinction between stem, transit-amplifying and mature expression patterns. The dendrogram in Fig. 4B illustrates the relationships among 28 different profiled cell populations including 3 profiles of neural stem cells (numbers 17, 18 and 19) and 7 other tissue-derived stem-like populations (numbers 14–16, 20–23), 3 profiles from transit-amplifying cell populations (numbers 25, 27, 28), and 9 profiles from mature tissues (numbers 3, 6–13; see Supplementary Materials for complete information on each population profile). In addition to the genes specific to SFD alone and to RA-treated cells, the dendrogram in Fig. 4B also displays lists that include genes upregulated in both of the differentiated cell populations (number 5, SFD vs ES; number 4, RA vs ES) relative to undifferentiated ESCs.

To focus on the differences between cells differentiated with and without RA, we also identified 1366 genes enriched in SFD alone cultures relative to RA-treated cells (number 26, SFD vs RA). These genes, which are likely to be expressed by cells present in the cultures differentiated with SFD alone but absent from RA-treated cultures, were found to cluster most closely with the transit-amplifying populations. In contrast, 1883 genes enriched in RA-treated cultures relative to SFD alone (number 2, RA vs SFD) clustered together with genes that were specific to RA-treated cells (number 1, RA). Taken together, our data suggests that cultures differentiated with SFD medium alone include neurons as well as a substantial representation of cells that are relatively immature and retain a high proliferative capacity, typical of transit-amplifying neuroblasts, whereas cultures induced with a period of exposure to RA are relatively more mature and appear to retain fewer precursor cells.

Cell Proliferation

Neural differentiation, both in vivo and in vitro, requires cells to exit the cell cycle and become postmitotic [Ohnuma and Harris, 2003]. Our microarray analysis revealed higher expression of early neural progenitor genes in cultures induced in SFD medium without RA as compared to undifferentiated cells and those induced with exposure to RA. This led us to evaluate the fraction of proliferating cells by incubation with bromodeoxyuridine (BrdU) [Gratzner, 1982]. ESC aggregates were maintained for 12d in SFD medium alone or with addition of RA from d4–d8. Sister cultures received BrdU for periods ranging from 1–24h immediately prior to harvesting at the end of the 12d induction. Cell aggregates induced with or without RA contained more BrdU positive cells as BrdU exposure time increased; however, the proportion of labelled cells in SFD medium alone was consistently higher than in cultures that had received RA (Fig. 5A–H). After 24h of BrdU treatment, almost 50% of cells in EBs induced without RA showed positive BrdU signal, whereas only 25% of cells in EBs induced with RA were labelled with BrdU (Fig. 5I). We next compared proliferation on d1–4 after plating by adding BrdU to plated cultures 2h prior to fixation. Cultures induced with SFD medium alone showed a high proportion of BrdU positive cells (~41%) on d1, and lower proportions on subsequent days. However, cells induced with RA showed much lower BrdU labelling (~5–8%) at all time points (Fig. 5J–L). Collectively, our results suggest that SFD medium alone supports a higher level of continued neural precursor proliferation, whereas exposure to RA promotes exit from the cell cycle and speeds neuronal differentiation and maturation.

An external file that holds a picture, illustration, etc.
Object name is nihms96388f5.jpg
Cell Proliferation

ESC aggregates maintained for 12d in SFD medium alone (A–D) or with exposure to RA from d4–d8 in suspension (E–H) were incubated with BrdU (20 μg/ml) for 2h (A, E); 4h (B, F); 8h (C, G) or 24h (D, H). (I) Percent of Hoechst stained nuclei (blue) in 12d aggregate cultures that were immunopositive for BrdU (red) plotted as a function of BrdU exposure time. Incorporation of BrdU in dissociated cells 4d after plating from aggregates maintained 12d in SFD medium alone (J) or with RA exposure from d4–d8 (K). (L) Percent of immunopositive cells following 2h incubation with BrdU evaluated in different cultures 1–4d after plating. Scale bars: 25 μm Cells exposed to RA displayed lower BrdU labelling both in floating aggregates and in plated cultures, indicating a lower rate of proliferation.

Neural Progenitors

Neurons in the developing CNS are generated from at least three types of progenitor cells, neuroepithelial cells, basal or intermediate progenitors, and radial glial cells, which express different molecular markers, [Doetsch 2003; Goldman, 2003]. Neuroepithelial cells express the intermediate filament nestin [Lendahl et al., 1990] and its posttranslational modifications recognized by the antibodies RC1 and RC2 [Malatesta et al., 2003]. Radial glia also express nestin as well as the transcription factor Pax6 [Gotz and Barde, 2005; Heins et al., 2002]. Unlike neuroepithelial cells and radial glia, basal progenitors express the T box transcription factor Eomes, also known as Tbr2 [Englund et al., 2005]. Expression of these progenitor-associated genes persists in proliferative neural precursors in the adult brain [Hodge et al., 2008]. Previous studies have highlighted the conversion of ESCs into radial glia [Liour and Yu, 2003; Liour et al., 2006; Bibel et al., 2004; Bouhon et al., 06], but have not addressed the production of basal progenitors. Our transcriptional profiles indicated that expression of both Pax6 and Tbr2 was significantly up regulated in cells induced with SFD medium alone, although Pax6 (but not Tbr2) was expressed at lower levels in RA-treated cells as well as undifferentiated ESCs.

To visualize the distribution of these progenitor cell markers we performed double immunofluorescence on ESCs induced in SFD medium without or with exposure to RA (Fig. 6). After 12d in suspension, many cells induced in SFD medium alone expressed only Pax6 (45 ± 5%) or only Tbr2 (13 ± 3%), although a small proportion of cells was found to be immunoreactive for both Pax6 and Tbr2 (10 ± 2%) (n=7; Fig. 6A). Approximately 36 ± 3% of cells induced with exposure to RA expressed Pax6, but Tbr2 was not detected in RA treated cells (n=5; Fig. 6B). At 3h after plating, cells expressing both nestin and RC2 were observed in cultures induced with RA or without RA (Fig 6C, D). Pax6 also was expressed in cells induced with RA or without RA, and many of the Pax6-positive cells coexpressed vimentin or RC2 (Fig. 6E–H). Tbr2 immunofluorescence was observed only by cells induced in SFD medium without RA, and some of the Tbr2-positive cells were immunoreactive for vimentin or RC2 (Supplementary Fig. 5A, B). Furthermore, the neural precursor population displaying strong vimentin immuno-fluorescence was distinct from neurons labelled with antibodies to β-3-tubulin (Supplementary Fig. 5C). Together, these data suggest that exposure to RA alters the specification of neural precursors, blocking the production of basal progenitors.

An external file that holds a picture, illustration, etc.
Object name is nihms96388f6.jpg
Neural Progenitors

ESC aggregates maintained for 12d in SFD medium alone (A) contained cells immunopositive for Pax6 (red), a marker expressed in radial glia, or for Tbr2 (green), which is expressed in basal progenitor cells. In contrast, RA-treated aggregates lacked Tbr2-positive cells (B) Scale Bar: 50 μm. (C–H) After dissociation, plating, and 3h in culture, numerous cells induced without (C, E, G) or with (D, F, H) exposure to RA were co-labelled with antibodies to RC2 and nestin (C, D), or with Pax6 and RC2 (E, F) or vimentin (G, H), suggesting that the majority of Pax6-positive cells are likely to be radial glia. Scale Bar: 25 μm.

Neuronal Phenotype Specification

Previous work suggests that exposure to RA can promote caudalization and suppress rostral neural structures [Maden, 2007; Wichterle et al., 2002; Bouhon et al., 2006]. To infer the positional identity of differentiated ESC, we compared our upregulated gene lists with a spatial database of gene expression from the mouse genome informatics (MGI) website, which contains comprehensive information about expression patterns [Smith et al., 2007]. As illustrated in figure 7, genes expressed in telencephalon, diencephalon and mesencephalon were more prevalent in cells differentiated in SFD medium alone relative to cultures treated with RA (Fig. 7A), whereas genes expressed in rhombencephalon and spinal cord predominated in cells differentiated with RA exposure (Fig. 7B). Interestingly, genes involved with regulating transcription were among the transcripts most highly enriched in each cell population (Fig. 7A, B).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f7.jpg
Positional Identity

(A) Replicate microarray heat plots for the 50 genes with differential anatomical expression patterns that showed the highest fold enrichment (lower bound) in cells differentiated in SFD medium alone relative to differentiation with exposure to RA, and (B) the 50 genes with differential anatomical expression that were most enriched in RA-treated cells relative to cells induced with SFD medium alone. Genes expressed in telencephalon (T, blue), diencephalon (D, green), mesencephalon (M, yellow), rhombencephalon (R, orange), and spinal cord (S, rose) in the mouse genome informatics anatomical expression database are indicated. Bar plots illustrate a weighted index of gene expression for each anterior-posterior (AP) anatomical domain. The AP index distributions were significantly different for genes enriched in the two differentiated cell populations (Chi-square, p<0.0001). T, D and M genes were enriched in cells induced with SFD medium alone, whereas R and S genes were enriched in cells differentiated with exposure to RA (z-test comparison of proportions, p ≤ 0.023). Genes specifically enriched in neurons (N, green), astrocytes (A, yellow), or oligodendrocytes (O, red) are highlighted. Plots show a weighted index of cell type-specific expression for all of the genes in each population, which were significantly different in (A) and (B) (Chi-square, p=0.0007). Genes expressed in astrocytes were enriched in cells differentiated with SFD medium alone (z-test comparison of proportions, p < 0.001).

To gain further insight into the positional identity of ESCs induced with RA or without RA we used antibodies to transcription factors associated with the differentiation of specific neuronal populations. In cultures induced with SFD medium for 10–12d followed by 4d adherent culture, 35 ± 3% of cells coexpressed β-3-tubulin and Tbr1 (Fig. 8A), a transcription factor expressed in the dorsal telencephalon (pallium)[Hevner et al., 2001; Remedios et al., 2007] and restricted to glutamatergic neurons [Hevner et al., 2001]. Accordingly, nearly all cells that strongly expressed the vesicular glutamate transporter1 (VGLUT1), were positive for Tbr1 (Fig. 8B) whereas GABAergic neurons, expressing glutamic acid decarboxylase (GAD), were uniformly negative for Tbr1 (Fig. 8C). Instead, many of the GAD-positive neurons expressed Islet1 (Fig. 8C), which is a LIM homeodomain transcription factor expressed in the spinal cord and ventral regions of the telencephalon, diencephalon, mesencephalon, and rhombencephalon [Ericson et al., 1995; Pfaff et al., 1996]. Islet1-positive cells become motorneurons in mesencephalon, rhombencephalon and spinal cord [Pfaff et al., 1996; Simon et al., 1994] but not in telencephalon and diencephalon [Ericson et al., 1995]. In cultures induced with SFD medium alone, 30 ± 3% of β-3-tubulin-positive cells expressed Islet1 and over 80% of cells immunopositive for Islet1 and β-3-tubulin also expressed GAD (Fig. 8C), which is prevalent in the subpallium/ventral telencephalon [Muzio et al., 2002; Yun et al., 2001]. Diencephalic Islet1-positive neurons can be distinguished from telencephalic Islet1-positive neurons by their expression of LIM homeodomain transcription factor, Lim1 [Ericson et al., 1995; Fotaki et al., 2006]. In the cultures differentiated with SFD medium alone, approximately 20% of cells were immunopositive for Lim1/2; however, none of the Islet1-positive neurons were co-labelled for Lim1 (Supplementary Fig. 6). These data suggest that Islet1-positive neurons in cultures differentiated with SFD medium alone acquire ventral telencephalic identity in the absence of RA. A few cells in cultures induced with SFD medium alone expressed Pax2 (3 ± 0.5%) or Engrail1 (1 ± 0.4%) which are typical markers for midbrain [Davis and Joyner, 1988; Matsunaga et al., 2000], whereas no cells were found to be positive for Hoxb4, a marker for the caudal hindbrain and spinal cord [Graham et al., 1988; Nordstrom et al., 2006] (Supplementary Fig. 7).

An external file that holds a picture, illustration, etc.
Object name is nihms96388f8.jpg
Neuronal Phenotype Specification

(A–C) In plated cultures differentiated with SFD medium alone, distinct β-3-tubulin-positive (blue) neuron populations were immuno-labelled with the dorsal pallial marker Tbr1 (Ar, arrows, red) or with the subpallial marker Islet1 (Ag, arrowheads, green). Scale Bar: 50 μm. (B) Cells strongly immunopositive for the glutamatergic marker VGLUT1 (green) were co-labelled with Tbr1 (red), (C) whereas GAD-positive GABAergic cells (blue) co-labelled with Islet1 (green), but not Tbr1 (red). Scale Bar: 25 μm. (D–F) In plated cultures differentiated with brief exposure to RA, ~22% β-3-tubulin-positive neurons (blue) co-labelled with Evx1/2 (red) and Lim1/2 (green)(Dr and Dg, arrows), markers for V0 ventral spinal neurons, whereas ~18% co-expressed markers of dI4 and dI6 dorsal spinal neurons Pax2 (red) and Lim1/2 (green)(Er and Eg, arrows). (F) Most Pax2-positive (blue) neurons were co-labelled for GAD (red) as well as Lim1/2 (green). Scale Bar: 25 μm. (G) Quantitation of immunopositive cells differentiated with SFD medium alone (open bars), or with exposure to RA (solid bars); 21–35 fields each condition, 462 ± 13 cells/field. The dorsal pallial marker Tbr1 and subpallial marker Islet1 together account for ~65% of all β-3-tubulin-positive neurons in SFD alone cultures. In cultures prepared from RA-treated aggregates, ~40% of β-3-tubulin-positive neurons co-labelled with Pax2 and Lim1/2 typical of dI4/dI6 neurons, or with Evx1/2 and Lim1/2 characteristic of V0 neurons.

In contrast to differentiation with SFD medium alone, which led cells to adopt anterior phenotypes, in cultures induced with exposure to RA, almost 90% of β-3-tubulin positive cells were immunopositive for Hoxb4 (Supplementary Fig. 7), consistent with acquisition of caudal positional identity for the hindbrain or spinal cord [Gould et al., 1997; Nordstrom et al., 2006]. To investigate subtype identity of neurons produced following exposure to RA, we used antibodies to transcription factors with regionally restricted expression patterns along the dorsoventral axis of the spinal cord [Briscoe et al., 2000; Helms and Johnson, 2003]. Evx1/2 is expressed in specific dorsal (dI1) and ventral (V0) interneuron populations [Briscoe et al., 2000; Pierani et al., 1999] that can be distinguished by co-expression of Lim1/2 in ventral Evx1/2-positive cells [Pierani et al., 1999], whereas intervening dorsal interneuron populations (dI4, dI5) co-express Pax2 and Lim1/2 [Gross et al., 2002]. Among β-3-tubulin positive cells induced in SFD medium with exposure to RA, 22 ± 0.8% were positive for both Evx1/2 and Lim1/2 (Fig. 8D arrows), suggesting that they acquired V0 fate; 18 ± 1.2% expressed both Pax2 and Lim1/2, but not Evx1/2 (Fig. 8E arrows), suggesting that they are dI4 or dI6 neurons. In the developing spinal cord, dorsal Pax2-positive neurons are GABAergic [Cheng et al., 2004]. Most Pax2-positive, Lim1/2-positive neurons derived from ESCs with exposure to RA expressed the GABAergic neuronal marker, GAD (Fig. 8F arrows). A few β-3-tubulin-positive cells (~10%) were immunopositive for transcription factors that are expressed by dorsal interneuron subtype dI1 (Evx1) [Pierani et al., 1999] or dI2 (Brn3a and Lim1/2) [Muller et al., 2002], or ventral interneuron subtype V1 (En1) (Ericson et al., 1997; Matise and Joyner, 1997), V3 (Nkx2.2) [Briscoe et al., 1999], or motorneurons (Islet1) [Pfaff et al., 1996] (Fig. 8G). Together, these data suggest that a large proportion of neurons (~60–70%) induced with SFD medium alone acquire anterior neuronal identity, whereas neurons induced with exposure to RA express molecular markers with regionally restricted expression patterns in the spinal cord (Supplementary Fig. 8).

DISCUSSION

We have demonstrated efficient neural conversion of mouse ESCs using serum- and retinoid-free differentiation (SFD) medium that contains few exogenous proteins and have used this system to compare the molecular and phenotypic properties of cells induced to differentiate with or without exposure to RA. Treatment with RA promotes neural maturation and alters the proportion of neurons expressing specific phenotypic properties, such as GABA production or glycine receptor subunit expression. Our microarray analysis identified a large number of neural genes that are expressed by cells that differentiate from mouse ESCs whether or not they are treated with RA. In addition, our results showed that cultures induced with SFD medium alone are selectively enriched in genes involved in early neurogenesis, whereas cultures induced with RA are enriched for genes associated with mature neurons. Together with our BrdU labelling studies, these results demonstrate that SFD medium permits the differentiation of neurons but also supports the continued proliferation of neural precursors, whereas exposure to RA restricts precursor proliferation and promotes maturation. Our results further suggest that cells induced with SFD medium alone acquire anterior neural identity whereas cultures induced in SFD medium with brief exposure to RA adopt posterior neural identity.

RA and neural differentiation

In serum-containing cultures, RA is essential to induce efficient neural differentiation from ESCs [Bain et al., 1995; 1996; Glaser &amp; Brüstle, 2005; Rohwedel et al., 1999]. Previous studies suggest that neural differentiation can proceed without RA under serum-free conditions [Okabe et al., 1996; Wiles and Johansson, 1999; Finley et al., 1999; Tropepe et al., 2001; Ying et al., 2003; Watanabe et al., 2005; Bouhon et al., 2005], although the use of proprietary media and supplements makes it difficult to rule out a cryptic retinoid inducer in many studies. Importantly, recent work suggests that ESCs cultured in PBS alone for 4h attain neural identity as indicated by expression of Sox1 and nestin [Smukler et al., 2006]. However, ESCs induced in PBS without any exogenous factors underwent massive cell death. We have demonstrated that our SFD medium devoid of exogenous proteins, aside from insulin, transferrin and BSA, supports neural conversion of ESCs without compromising cell growth and survival.

Consistently, our experiments indicate that cells differentiated with brief exposure to RA lose stem cell identity, adopt neural fate, and acquire mature axonal/dendritic polarity more rapidly than cells induced with SFD medium alone. Previous work suggests that RA promotes neural differentiation by inducing proneural genes, such as Sox1, Sox6, Brn2 and neurogenin1, and suppressing antineural genes [Maden, 2007]. For example, a major component of ESC neuralization by RA is thought to involve the induction of Wnt signalling inhibitors, secreted frizzled-related protein 2 (Sfrp2) and dickkopf homolog 1 (Dkk1) [Aubert et al., 2002; Verani et al., 2007, Watanabe et al., 2005]. Our gene expression profiles indicate that Sfrp2 is up regulated in cells induced in SFD medium with or without exposure to RA. In addition, Sox6, Brn2 and neurogenin2 are all upregulated in cells differentiated in SFD medium regardless of exposure to RA and the inhibitor of Wnt and FGF signalling pathways Tmem46, also known as Shisa2 [Furushima et al., 2007], is selectively enriched in cultures differentiated in SFD medium without RA.

The relative maturity of ESCs differentiated with exposure to RA may reflect the conversion of neural progenitor cells to postmitotic neurons. Thus, neurons induced with exposure to RA undergo their final mitosis at an earlier time point than cells induced with SFD medium alone. We also observe, however, that although neurons induced with SFD medium alone are less mature than those differentiated with RA during the first few days in adherent culture, they gradually become mature in longer-term culture, suggesting they will eventually “catch up” with the level of maturity attained by RA treated cells. Another possibility, which we cannot exclude, is that RA directly induces genes associated with neuronal maturation or with the mature state. During neurogenesis, after exiting the cell cycle, young neurons initiate formation and outgrowth of neurites that ultimately become axons and dendrites. Studies in cell culture, have shown that treatment with RA increases neurite outgrowth from a variety of central and peripheral neurons [Clagett-Dame et al., 2006]. Therefore, it is possible that exposure to RA speeds the acquisition of polarity as a secondary effect to the promotion of neurite formation.

Proliferation vs differentiation

Mouse ESCs divide rapidly and without apparent limit in vitro [Suda et al., 1987], but division rate decreases under conditions promoting differentiation [Stead et al., 2002]. Cell proliferation and differentiation are tightly coordinated by cell cycle controls involving the formation of complexes between cyclins and cyclin-dependent kinases (CDKs), a process inhibited by cyclin-dependant kinase inhibitors (CKIs) [Burdon et al., 2002]. RA promotes differentiation and inhibits cell proliferation by inducing cell cycle arrest in a variety of different embryonic cells including ESCs [Lin et al., 2005]. Within 72h of RA treatment, mouse ESCs express the tumor suppressor gene p53, reducing expression of Nanog, one of the genes required for maintenance of pluripotency, and increase expression of p21, the CKI gene that enhances ESC differentiation [Lin et al., 2005]. Our cultures, whether induced with or without exposure to RA, showed reduced expression of cyclin E1 that is highly active in undifferentiated ESCs [Stead et al., 2002] and showed increased expression of cyclin-dependent kinase inhibitors 1B (p27) and 2D (p19), which independently inhibit cell cycle progression in the CNS. Cultures induced with SFD alone showed increased expression of cyclin D2 and Cdk4, which together form a complex that is found in neural precursors and is increased upon initiation of ESC differentiation [Savatier et al., 1996]. Our results show that cultures induced with or without RA include postmitotic neurons and proliferating cells; however, cultures induced with SFD alone exhibit more proliferation. Thus, RA mediated suppression of cyclin genes and induction of CKI genes are likely to be one mechanism by which RA accelerates cell cycle exit and promotes ESC neural differentiation.

Positional and phenotypic identity

Nieuwkoop proposed that neurons are initially specified as forebrain but can become gradually posteriorized by caudalizing signals [Nieuwkoop, 1952]. Our data suggests that a large proportion of ESCs induced in SFD medium alone exhibit forebrain identity, but that brief exposure to RA results in acquisition of features characteristic of hindbrain and spinal cord, consistent with the Nieuwkoop model. Another recent study [Gaspard et al., 2008] supports the proposal that serum-free ESC differentiation yields predominantly telencephalic identity (see also [Bouhon et al., 2006], with ventral GABAergic differentiation being dependent on autocrine sonic hedgehog signalling [Gaspard et al., 2008]. In contrast, other serum-free conditions involving proprietary media were reported to result in a high percentage of ESC-derived neurons adopting midbrain identity [Kawasaki et al., 2000], with efficiency of telencephalic differentiation being enhanced by exogenous addition of the Wnt antagonist Dkk1 [Watanabe et al., 2005], suggesting that the baseline conditions may include unknown caudalizing factors that must be suppressed or overcome for acquisition of more anterior neural phenotypes.

RA suppresses expression of anterior genes and causes caudalization of neural tissues in vivo [Maden, 2007]. Consistent with this, our results show that genes highly expressed in hindbrain and spinal cord, including 14 members of the Hox gene family [Nolte and Krumlauf, 2007], are enriched in cultures induced with exposure to RA, whereas genes prevalent in more anterior structures were lacking. For example, Tbr2, which is highly expressed in basal progenitors of dorsal forebrain [Englund et al., 2005], was dramatically enriched in cells differentiated in SFD medium alone relative to cells that were exposed to RA, whereas both differentiated cell populations express molecular markers for neuroepithelial and radial glial cells that generate neurons both in the cortex and in spinal cord [Malatesta et al., 2003]. Moreover, a substantial proportion of cells induced with exposure to RA exhibited transcription factor expression patterns that were consistent with spinal cord identity. Dorso-ventral patterning in spinal cord depends on RA and gradients of sonic hedgehog, which specifies ventral phenotypes, and of bone morphogenetic protein (BMP) family members, originating from the roof plate, that promote dorsal phenotypes [Jessell, 2000]. Importantly, about 40% of mES cells induced with exposure to RA expressed transcription factors characteristic of V0 or of dI6 cell populations that lie adjacent to each other midway between the dorsal and ventral extremes in vivo (Supplementary Fig. 8). Approximately ~22% co-expressed Evx1 and Lim1/2 (V0), whereas 18% expressed Pax2 and Lim1/2 together with GAD (dI4 and dI6), suggesting that specification of caudal neuronal subtypes by RA parallels that observed in vivo (see also [Wichterle et al., 2002]). In contrast to this interpretation, Bibel et al. [2004] suggested that selection of highly proliferating ESCs followed by aggregation in serum-containing medium and exposure to RA yielded radial glia that differentiated into cells with physiological and morphological characteristics of glutamatergic neurons from cerebral cortex. However, when we induced ESCs in serum containing medium with exposure to RA, although most of the neurons were glutamatergic [Finley et al., 1996], many cells expressed the caudal marker Hoxb4 while none expressed the cortical cell marker Tbr1 that is prevalent with induction in SFD medium in the absence of RA (Supplementary Fig. 7). Thus, it seems likely that RA caudalizes differentiating ESCs regardless the presence or absence of serum, or selection for rapidly dividing cells.

Our study shows that SFD culture medium induces robust neural conversion of ESCs and efficient specification of anterior neural fate, and reproduces the in vivo effect of RA on neural differentiation. Development of methods for reprogramming somatic cells has opened up the possibility of generating customized patient-specific stem cells that can be used for tissue repair and cell replacement [Park et al., 2008]. Maximizing the therapeutic potential of stem cells will require identifying conditions that allow for efficient commitment to particular cell phenotypes. Previous studies have shown that ESCs can differentiate into specific neural types by sequential exposure to extracellular matrix proteins [Goetz et al., 2006] and signalling molecules [Salero, 2007; Wichterle, 2002] associated with their development in vivo. The restricted number of identified ingredients in SFD medium should aid in efforts to elucidate the signals necessary for differentiation of additional neural cell types.

RA and neural differentiation

In serum-containing cultures, RA is essential to induce efficient neural differentiation from ESCs [Bain et al., 1995; 1996; Glaser &amp; Brüstle, 2005; Rohwedel et al., 1999]. Previous studies suggest that neural differentiation can proceed without RA under serum-free conditions [Okabe et al., 1996; Wiles and Johansson, 1999; Finley et al., 1999; Tropepe et al., 2001; Ying et al., 2003; Watanabe et al., 2005; Bouhon et al., 2005], although the use of proprietary media and supplements makes it difficult to rule out a cryptic retinoid inducer in many studies. Importantly, recent work suggests that ESCs cultured in PBS alone for 4h attain neural identity as indicated by expression of Sox1 and nestin [Smukler et al., 2006]. However, ESCs induced in PBS without any exogenous factors underwent massive cell death. We have demonstrated that our SFD medium devoid of exogenous proteins, aside from insulin, transferrin and BSA, supports neural conversion of ESCs without compromising cell growth and survival.

Consistently, our experiments indicate that cells differentiated with brief exposure to RA lose stem cell identity, adopt neural fate, and acquire mature axonal/dendritic polarity more rapidly than cells induced with SFD medium alone. Previous work suggests that RA promotes neural differentiation by inducing proneural genes, such as Sox1, Sox6, Brn2 and neurogenin1, and suppressing antineural genes [Maden, 2007]. For example, a major component of ESC neuralization by RA is thought to involve the induction of Wnt signalling inhibitors, secreted frizzled-related protein 2 (Sfrp2) and dickkopf homolog 1 (Dkk1) [Aubert et al., 2002; Verani et al., 2007, Watanabe et al., 2005]. Our gene expression profiles indicate that Sfrp2 is up regulated in cells induced in SFD medium with or without exposure to RA. In addition, Sox6, Brn2 and neurogenin2 are all upregulated in cells differentiated in SFD medium regardless of exposure to RA and the inhibitor of Wnt and FGF signalling pathways Tmem46, also known as Shisa2 [Furushima et al., 2007], is selectively enriched in cultures differentiated in SFD medium without RA.

The relative maturity of ESCs differentiated with exposure to RA may reflect the conversion of neural progenitor cells to postmitotic neurons. Thus, neurons induced with exposure to RA undergo their final mitosis at an earlier time point than cells induced with SFD medium alone. We also observe, however, that although neurons induced with SFD medium alone are less mature than those differentiated with RA during the first few days in adherent culture, they gradually become mature in longer-term culture, suggesting they will eventually “catch up” with the level of maturity attained by RA treated cells. Another possibility, which we cannot exclude, is that RA directly induces genes associated with neuronal maturation or with the mature state. During neurogenesis, after exiting the cell cycle, young neurons initiate formation and outgrowth of neurites that ultimately become axons and dendrites. Studies in cell culture, have shown that treatment with RA increases neurite outgrowth from a variety of central and peripheral neurons [Clagett-Dame et al., 2006]. Therefore, it is possible that exposure to RA speeds the acquisition of polarity as a secondary effect to the promotion of neurite formation.

Proliferation vs differentiation

Mouse ESCs divide rapidly and without apparent limit in vitro [Suda et al., 1987], but division rate decreases under conditions promoting differentiation [Stead et al., 2002]. Cell proliferation and differentiation are tightly coordinated by cell cycle controls involving the formation of complexes between cyclins and cyclin-dependent kinases (CDKs), a process inhibited by cyclin-dependant kinase inhibitors (CKIs) [Burdon et al., 2002]. RA promotes differentiation and inhibits cell proliferation by inducing cell cycle arrest in a variety of different embryonic cells including ESCs [Lin et al., 2005]. Within 72h of RA treatment, mouse ESCs express the tumor suppressor gene p53, reducing expression of Nanog, one of the genes required for maintenance of pluripotency, and increase expression of p21, the CKI gene that enhances ESC differentiation [Lin et al., 2005]. Our cultures, whether induced with or without exposure to RA, showed reduced expression of cyclin E1 that is highly active in undifferentiated ESCs [Stead et al., 2002] and showed increased expression of cyclin-dependent kinase inhibitors 1B (p27) and 2D (p19), which independently inhibit cell cycle progression in the CNS. Cultures induced with SFD alone showed increased expression of cyclin D2 and Cdk4, which together form a complex that is found in neural precursors and is increased upon initiation of ESC differentiation [Savatier et al., 1996]. Our results show that cultures induced with or without RA include postmitotic neurons and proliferating cells; however, cultures induced with SFD alone exhibit more proliferation. Thus, RA mediated suppression of cyclin genes and induction of CKI genes are likely to be one mechanism by which RA accelerates cell cycle exit and promotes ESC neural differentiation.

Positional and phenotypic identity

Nieuwkoop proposed that neurons are initially specified as forebrain but can become gradually posteriorized by caudalizing signals [Nieuwkoop, 1952]. Our data suggests that a large proportion of ESCs induced in SFD medium alone exhibit forebrain identity, but that brief exposure to RA results in acquisition of features characteristic of hindbrain and spinal cord, consistent with the Nieuwkoop model. Another recent study [Gaspard et al., 2008] supports the proposal that serum-free ESC differentiation yields predominantly telencephalic identity (see also [Bouhon et al., 2006], with ventral GABAergic differentiation being dependent on autocrine sonic hedgehog signalling [Gaspard et al., 2008]. In contrast, other serum-free conditions involving proprietary media were reported to result in a high percentage of ESC-derived neurons adopting midbrain identity [Kawasaki et al., 2000], with efficiency of telencephalic differentiation being enhanced by exogenous addition of the Wnt antagonist Dkk1 [Watanabe et al., 2005], suggesting that the baseline conditions may include unknown caudalizing factors that must be suppressed or overcome for acquisition of more anterior neural phenotypes.

RA suppresses expression of anterior genes and causes caudalization of neural tissues in vivo [Maden, 2007]. Consistent with this, our results show that genes highly expressed in hindbrain and spinal cord, including 14 members of the Hox gene family [Nolte and Krumlauf, 2007], are enriched in cultures induced with exposure to RA, whereas genes prevalent in more anterior structures were lacking. For example, Tbr2, which is highly expressed in basal progenitors of dorsal forebrain [Englund et al., 2005], was dramatically enriched in cells differentiated in SFD medium alone relative to cells that were exposed to RA, whereas both differentiated cell populations express molecular markers for neuroepithelial and radial glial cells that generate neurons both in the cortex and in spinal cord [Malatesta et al., 2003]. Moreover, a substantial proportion of cells induced with exposure to RA exhibited transcription factor expression patterns that were consistent with spinal cord identity. Dorso-ventral patterning in spinal cord depends on RA and gradients of sonic hedgehog, which specifies ventral phenotypes, and of bone morphogenetic protein (BMP) family members, originating from the roof plate, that promote dorsal phenotypes [Jessell, 2000]. Importantly, about 40% of mES cells induced with exposure to RA expressed transcription factors characteristic of V0 or of dI6 cell populations that lie adjacent to each other midway between the dorsal and ventral extremes in vivo (Supplementary Fig. 8). Approximately ~22% co-expressed Evx1 and Lim1/2 (V0), whereas 18% expressed Pax2 and Lim1/2 together with GAD (dI4 and dI6), suggesting that specification of caudal neuronal subtypes by RA parallels that observed in vivo (see also [Wichterle et al., 2002]). In contrast to this interpretation, Bibel et al. [2004] suggested that selection of highly proliferating ESCs followed by aggregation in serum-containing medium and exposure to RA yielded radial glia that differentiated into cells with physiological and morphological characteristics of glutamatergic neurons from cerebral cortex. However, when we induced ESCs in serum containing medium with exposure to RA, although most of the neurons were glutamatergic [Finley et al., 1996], many cells expressed the caudal marker Hoxb4 while none expressed the cortical cell marker Tbr1 that is prevalent with induction in SFD medium in the absence of RA (Supplementary Fig. 7). Thus, it seems likely that RA caudalizes differentiating ESCs regardless the presence or absence of serum, or selection for rapidly dividing cells.

Our study shows that SFD culture medium induces robust neural conversion of ESCs and efficient specification of anterior neural fate, and reproduces the in vivo effect of RA on neural differentiation. Development of methods for reprogramming somatic cells has opened up the possibility of generating customized patient-specific stem cells that can be used for tissue repair and cell replacement [Park et al., 2008]. Maximizing the therapeutic potential of stem cells will require identifying conditions that allow for efficient commitment to particular cell phenotypes. Previous studies have shown that ESCs can differentiate into specific neural types by sequential exposure to extracellular matrix proteins [Goetz et al., 2006] and signalling molecules [Salero, 2007; Wichterle, 2002] associated with their development in vivo. The restricted number of identified ingredients in SFD medium should aid in efforts to elucidate the signals necessary for differentiation of additional neural cell types.

Acknowledgments

Supported by grants from the NIH (NS045023 and NS30888, JEH; {"type":"entrez-nucleotide","attrs":{"text":"DK079798","term_id":"187521152","term_text":"DK079798"}}DK079798, JCM; GM39746, RWM). We are grateful to Tim Wilding and Aileen Lu for technical assistance and to David Gottlieb, Paul Gray and Kristen Kroll for helpful comments on the manuscript. We thank the Alvin J. Siteman Cancer Center at Washington University School of Medicine and Barnes-Jewish Hospital in St. Louis, Mo., for the use of the Multiplexed Gene Analysis Core. The Siteman Cancer Center is supported in part by an NCI Cancer Center Support Grant #P30 CA91842

Department of Cell Biology &amp; Physiology, Washington University Medical School, 660 South Euclid Avenue, St. Louis, MO 63110
Department of Pathology &amp; Immunology, Washington University Medical School, 660 South Euclid Avenue, St. Louis, MO 63110
Program in Developmental Biology, Washington University Medical School, 660 South Euclid Avenue, St. Louis, MO 63110
Address all correspondence to: Dr. James E. Huettner, Department of Cell Biology and Physiology, Washington University School of Medicine, 660 South Euclid Avenue, St. Louis, MO 63110, 314-362-6624 phone, 314-362-7463 fax, ude.ltsuw@rentteuhj
Publisher's Disclaimer

Abstract

Pluripotent mouse embryonic stem cells (ESCs) derived from the early blastocyst can differentiate in vitro into a variety of somatic cell types including lineages from all three embryonic germ layers. Protocols for ES cell neural differentiation typically involve induction by retinoic acid (RA), or by exposure to growth factors or medium conditioned by other cell types. A serum-free differentiation (SFD) medium completely lacking exogenous retinoids was devised that allows for efficient conversion of aggregated mouse ESCs into neural precursors and immature neurons. Neural cells produced in this medium express neuronal ion channels, establish polarity, and form functional excitatory and inhibitory synapses. Brief exposure to RA during the period of cell aggregation speeds neuronal maturation and suppresses cell proliferation. Differentiation without RA yields neurons and neural progenitors with apparent telencephalic identity, whereas cells differentiated with exposure to RA express markers of hindbrain and spinal cord. Transcriptional profiling indicates a substantial representation of transit amplifying neuroblasts in SFD cultures not exposed to RA.

Abstract

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Footnotes
Collaboration tool especially designed for Life Science professionals.Drag-and-drop any entity to your messages.