PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Alcohol Clin Exp Res. Author manuscript; available in PMC Aug 1, 2013.
Published in final edited form as:
PMCID: PMC3374055
NIHMSID: NIHMS343458
Metabolic Biomarkers of Prenatal Alcohol Exposure in Human Embryonic Stem Cell-derived Neural Lineages
Jessica A. Palmer, MS,1,2* Ashley M. Poenitzsch, MS,1* Susan M. Smith, PhD,3^ Kevin R. Conard, MS,2 Paul R. West, PhD,2 and Gabriela G. Cezar, PhD1,2
1Department of Animal Sciences, University of Wisconsin-Madison, 1675 Observatory Drive, Madison, WI
2Stemina Biomarker Discovery, Inc., 504 South Rosa Road, Madison, WI
3Department of Nutritional Sciences, University of Wisconsin-Madison, 1415 Linden Drive, Madison, WI
^Corresponding Author: Susan Smith, Ph.D., Professor, Department of Nutritional Sciences, University of Wisconsin-Madison, 1415 Linden Drive, Madison, WI 53706, Tel: (608) 263-4316, Fax: (608) 262-5860, suesmith/at/nutrisci.wisc.edu
Gabriella G. Cezar, D.V.M., Ph.D., Assistant Professor, Department of Animal Sciences, University of Wisconsin-Madison, 1675 Observatory Drive, Madison, WI 53706; Tel: (608) 263-4307; Fax: (608) 262-5157; ggcezar/at/wisc.edu
*These authors contributed equally to this manuscript.
Background
Fetal alcohol spectrum disorders (FASD) are a leading cause of neurodevelopmental disability. The mechanisms underlying FASD are incompletely understood, and biomarkers to identify those at risk are lacking. Here, we perform metabolomic analysis of embryoid bodies and neural lineages derived from human embryonic stem (hES) cell to identify the neural secretome produced in response to ethanol exposure.
Methods
WA01 and WA09 hES cells were differentiated into embryoid bodies, neural progenitors or neurons. Cells along this progression were cultured for four days with 0%, 0.1% or 0.3% ethanol. Supernatants were subjected to C18 chromatography followed by ESI-QTOF-MS. Features were annotated using public databases and the identities of four putative biomarkers were confirmed with purified standards and comparative MS/MS.
Results
Ethanol treatment induced statistically significant changes to metabolite abundance in human embryoid bodies (180 features), neural progenitors (76 features) and neurons (42 features). There were no shared significant features between different cell types. Fifteen features showed a dose-response to ethanol. Four chemical identities were confirmed; L-thyroxine, 5’-methylthioadenosine, and the tryptophan metabolites L-kynurenine, and indoleacetaldehyde. One feature with a putative annotation of succinyladenosine was significantly increased in both ethanol treatments. Additional features were selective to ethanol treatment but were not annotated in public databases.
Conclusions
Ethanol exposure induces statistically significant changes to the metabolome profile of human embryoid bodies, neural progenitors and neurons. Several of these metabolites are normally present in human serum, suggesting their usefulness as potential serum FASD biomarkers. These findings suggest the biochemical pathways that are affected by ethanol in the developing nervous system and delineate mechanisms of alcohol injury during human development.
Keywords: human embryonic stem cells, neural progenitors, fetal alcohol syndrome, 5-methylthioadenosine, L-kynurenine, indoleacetaldehyde, LC-MS, metabolomics
Fetal alcohol spectrum disorders (FASD) are a leading cause of neurodevelopmental disability and are associated with deficits in learning, motor skills and executive functioning. The basis of alcohol injury to neurons includes altered proliferation, differentiation, and synaptogenesis and is incompletely understood (reviewed in Miller 2006). Deeper understanding of the mechanisms underlying alcohol’s toxicity could facilitate prevention, earlier detection or repair the resulting damage. However, implementation of such strategies is challenged by the limited availability of diagnostic biomarkers that could identify at-risk pregnancies and those with FASD.
The human embryo and the regulatory mechanisms directing its development are challenging to study. Human embryonic stem (hES) cells permit this study using an in vitro model. hES cells are pluripotent, self-renewing cells isolated from human embryos (Thomson et al., 1998). They give rise to all differentiated cell types in the human body and recapitulate embryonic development in vitro, including the faithful replication of neurodevelopment and differentiation of hES cells into neural precursors, neurons and glia (Reubinoff et al., 2000; Zhang et al., 2001). Because they function similarly to in vivo counterparts (Ben-Hur et al., 2004; Zhang et al., 2001), hES cell-derived models are an ideal research tool to elucidate biochemical mechanisms underlying disease, including alcohol’s effects on human neural development.
The human metabolome is the comprehensive set of endogenous small molecules. It is a product of health or disease states and can be measured on every level of complexity e.g., tissues, biofluids and cells (Fiehn, 2002). Abnormal metabolite production in response to environmental, dietary or disease insult provides mechanistic insights into these atypical processes. Metabolite composition of a biological sample is very sensitive to environmental effects (Lindon et al., 2003). We have used hES cells to both discover biomarkers and establish predictive models of chemically-induced developmental toxicity (Cezar et al., 2007; West et al., 2010).
Specific changes to metabolites identified in response to alcohol exposure could serve as biomarkers for FASD. Such biomarkers also outline the mechanisms underlying alcohol’s pathogenicity, leading to novel treatment opportunities. The analysis of biomarkers in serum and amniotic fluid is an established approach to predict birth defects and premature births. The presence of fatty acid ethyl esters (FAEEs) in meconium is employed to identify alcohol exposure during pregnancy. However, FAEEs can be present in the meconium of unexposed infants (Chan et al., 2003) and do not provide information on the timing, duration and dosage of exposure (Litten et al., 2010). Additionally, FAEEs do not provide information on the damage that may have occurred specifically to the developing brain and the mechanism of injury. Thus, there is a significant need for reliable biomarkers that can be used for FASD prevention, diagnosis and treatment.
This study tested the hypothesis that alcohol exposure induces statistically significant changes to the abundance of small molecule metabolites in human neural progenitors and neurons. To test this, we treated hESC-derived embryoid bodies, neural progenitors and neurons with 0, 0.1% or 0.3% ethanol. We employed liquid chromatography coupled with electrospray-ionization time-of-flight mass spectrometry (LC-ESI-QTOF-MS) to identify and measure changes to secreted metabolite abundance in response to alcohol exposure. Here, we describe the metabolome of hES cell derivatives exposed to ethanol during three stages of neurogenesis.
Ethanol treatment
WA01 and WA09 hES cells, obtained from WiCell Research Institute (Madison, WI), were exposed to 0%, 0.1% and 0.3% ethanol for four days at one of three stages of neurogenesis: days 0–4 (embryoid bodies, EBs), days 17–21 (neural progenitors) and days 28–32 (neurons) (Zhang et al., 2001) (Outlined in Figure 1). Ethanol doses were selected to mimic fetal exposure to binge-like low and high ethanol levels, respectively. Each cell line, time point and treatment included five biological replications of three samples per treatment per replication. Additionally at each time point, media controls were included containing appropriate media exposed to the plate conditions without cells.
Figure 1
Figure 1
Outline of hES cell differentiation and treatment protocol used in this study.
Neural Differentiation
Stage 1: Embryoid Body Formation
hES cells grown in mTeSR1 media (Stem Cell Technologies, Inc., Vancouver, BC) were dispersed following treatment with dispase (2 mg/ml; Invitrogen, Carlsbad, CA) and grown in suspension. EBs were cultured in EB formation media, containing Iscoves's Modified Dulbecco's Medium (1×; Invitrogen), supplemented with 15% heat inactivated certified fetal bovine serum (Invitrogen) for four days in 90 mm low-binding dishes (Nalge Nunc International, Rochester, NY). EB formation media was changed every other day for neural progenitor and neuron differentiation experiments. For treatment, ~50 EBs were placed in each well of 24-well low-binding microplates (Nalge Nunc International) 24 hours after formation. During treatment, media and compound were replaced daily.
Stage 2: Neurospheres
On day 5 of suspension culture, media was switched to neural precursor media, which consisted of DMEM/F12 (1:1) supplemented with 1% N2 Supplement, 0.1 mM MEM non-essential amino acids (NEAA), 1% Insulin-Transferrin-Selenium-X Supplement, 20 ng/ml FGF2, 20 ng/ml EGF (all from Invitrogen), 2 µg/ml heparin, 0.5 µM retinoic acid, and 60 µM putrescine (all from Sigma-Aldrich, St. Louis, MO). Neurospheres were maintained in neural precursor media for 11 days, with media changed every other day.
Stage 3: Neural Progenitors
On day 17 of neural differentiation, neurospheres were dissociated into a single cell suspension using 0.25% trypsin-EDTA (Invitrogen) containing 2% chick serum (Sigma-Aldrich). The cells were plated on pre-coated poly-L-ornithine and laminin 24 well plates (BD Biosciences, San Jose, CA) at 2×106 cells/well. Neural progenitors were maintained in neural progenitor media, which consisted of DMEM/F12 (1:1) supplemented with 1% N2 supplement, 0.2 mM MEM-NEAA, 500 ng/ml sonic hedgehog (R&D Systems, Minneapolis, MN), 100 ng/ml FGF8 (Invitrogen), and 10 ng/ml FGF4 (Invitrogen). Media was changed every other day during routine maintenance. Treatment began on day 18 and media was replaced and fresh compound added daily during the exposure period. Treated neural progenitors were fixed on day 22 for nestin and caspase-3 analysis.
Stage 4: Neural Differentiation
Beginning on day 22, neural progenitors were differentiated into neurons. Cells were cultured in neural differentiation media consisting of neurobasal medium supplemented with 1% N2 supplement, 0.1 mM MEM-NEAA, 0.5 mM L-glutamine, 1 µg/ml laminin, 20 ng/ml BDNF, 10 ng/ml GDNF (all from Invitrogen), 1 µM cAMP, 30 µM serotonin (both from Sigma-Aldrich), and 50 ng/ml sonic hedgehog. Starting on day 28, neurons were treated for 4 days and media was replaced daily. Neurons were fixed on day 32 for caspase-3, synaptogenesis and differentiation immunocytochemistry analysis.
Immunocytochemistry
Immunostaining of paraformaldehyde-fixed cells was performed using standard techniques. Primary antibodies included anti-nestin (MAB5326. Millipore, Billerica, MA), anti-β-III-tubulin (T8660, Sigma-Aldrich), anti-synaptophysin (ab52636, Abcam, Cambridge, MA), anti-SV2A (sc-28955, Santa Cruz Biotechnology, Santa Cruz, CA), anti-tyrosine hydroxylase (AB152, Millipore), and anti-cleaved caspase-3 (9661, Cell Signaling Technology, Danvers, MA). Secondary antibodies were conjugated with Alexa Fluor 488 (A11001, Invitrogen) or Alexa Fluor 594 (A21207, Invitrogen). For synaptophysin and SV2A, Triton X-100 was omitted to maintain the integrity of membrane structures. Positive staining of post-mitotic neurons was defined as cells that co-expressed β-III-tubulin and an additional neuronal marker: synaptophysin, SV2A, or tyrosine hydroxylase. At least 300 cells/sample and 6 samples/treatment were counted. Statistical analysis was performed using a one-way ANOVA with statistical significance defined as p ≤ 0.05.
Sample Preparation
On the fourth day of treatment, spent media was collected from each sample, quenched with LC-MS grade acetonitrile (final acetonitrile concentration = 40%), and stored at −80°C until preparation. For mass spectrometry analysis, equal amounts of quenched sample and LC-MS grade water were added to a Microcon centrifugal filter with a molecular weight cut-off of 3 kDa (Millipore). Each sample was centrifuged at 13,000 × g for three hours at 4°C. The flow-through was retained and dried using a vacuum concentrator and stored at −80°C. Samples were reconstituted in 50 µl of 0.1% formic acid in water prior to mass spectrometry analysis.
LC-ESI-QTOF mass spectrometry
Liquid chromatography-electrospray ionization-quadrupole time of flight mass spectrometry (LC-ESI-QTOF-MS) was performed at Stemina Biomarker Discovery (Madison, WI) using an Agilent QTOF LC-MS system consisting of a 1200SL HPLC system and a G6520AA QTOF high resolution mass spectrometer (Agilent Technologies, Wilmington, DE) capable of exact mass MS and MS/MS ion fragmentation. The flow through of each sample was analyzed using a 2.1 × 50mm Zorbax 1.8 um C18-SB column (C18, Agilent Technologies). Ultra performance liquid chromatography (UPLC) analysis was performed using a 20 minute gradient starting with ratio of 95% 0.1% formic acid in water (Solvent A) to 5% 0.1% formic acid in acetonitrile (Solvent B) and ending with 100% Solvent B at a flow rate of 500 µl/min. Electrospray ionization was employed using a dual ESI source, with an Agilent isocratic pump continuously delivering an internal mass reference solution into the source at approximately 0.01 ml/min for mass correction. For positive-ion ESI the reference solution consisted of 7H-purine (m/z 121.0509) and HP-0921 (m/z 922.0098). The reference solution for negative-ion ESI contained trifluoroacetic acid (TFA, m/z 112.9856) and the TFA adduct of HP-921 (m/z 1033.9881). The mass range of the instrument was set at 70–1600 Da. The sample order was randomized and 5 µl of each sample was injected. A solvent blank (0.1% formic acid) was run after every ten samples. All three time points were analyzed with positive-ion ESI. Additionally, negative-ion ESI was performed on the EB experiment samples. Data acquisition was performed with Agilent MassHunter Acquisition software using high-resolution exact mass conditions.
Data analysis of LC-ESI-QTOF metabolomics
Prior to data analysis, the total ion chromatogram (TIC) of each sample was carefully inspected for quality and reproducibility of the MS signal. If a sample's TIC abundance deviated by more than 25% from the median across the LC-MS gradient, the LC-MS analysis was repeated for that sample. The data was deisotoped and then converted into the open source mzData format. Data analysis was performed using the open source statistical programming and analysis software, R. The XCMS package (Smith et al., 2006) was used to analyze the LC-ESI-QTOF-MS resulting files using the Centwave algorithm for peak peaking (Tautenhahn et al., 2008). Retention time deviations across EB LC-MS samples were corrected using retcor loess regression and the obiwarp method for neural precursor and neuron samples. After retention time correction, the features were grouped using the density based functions in XCMS. After the grouping function was performed features missing in LC-MS samples were iteratively integrated using m/z and retention time windows based on the range of the feature group. Contaminant ions were removed by comparing spent media extracts with blank extraction samples. Statistical significance of individual mass features was performed under the null hypothesis that no difference exists in feature abundance between control and treated samples. Differential small molecule metabolites, or features, were determined using a complete block design ANOVA with the model "log2(abundance) ~ treatment + replicate + cell line." Features were considered statistically significant if they had a p-value ≤ 0.05 in the treatment factor of the ANOVA model. Additionally, to be considered authentic, each feature was required to show a statistically significant alteration in both cell lines evaluated. The extracted ion chromatogram (EIC) of each statistically significant feature was then visually evaluated to confirm an observable difference between treated and control samples and to reduce the inclusion of spurious results.
For feature annotation, the neutral exact mass of each feature was queried against the public searchable databases METLIN (http://metlin.scripps.edu), The Human Metabolome Database (http://www.hmdb.ca), and the Kyoto Encyclopedia of Genes and Genomes (www.genome.jp/kegg/) for candidate identities. Measured mass features were considered a putative match to a small molecule present in the databases if their exact masses were within 20 ppm of the annotated database molecule (0.00002 × mass). Confirmation of statistically significant annotated features was carried out using tandem mass spectrometry (MS/MS).
Confirmation of candidate biomarkers by LC-ESI-QTOF-MS/MS
Analytical grade chemical standards for thyroxine, 5'-methylthioadenosine, L-kynurenine, and indoleacetaldehyde were purchased from Sigma-Aldrich for comparative mass spectrometry. Chemical references were evaluated using identical sample preparation and chromatographic methods employed in the analysis of the original samples. Additionally, the three original samples with the highest abundance for each feature of interest were re-prepped and analyzed for comparison with the standard. Chemical references were dissolved in the appropriate basal media at three concentrations: 1 mM, 0.1 mM and 0.01 mM. Additionally, a 0.1 mM solution was prepared for each standard in 0.1% formic acid, with the exception of thyroxine, which was solubilized in a 0.1 mM solution of 50/50 methanol and dichloromethane. Data acquisition for targeted MS/MS was performed by inputting the m/z value of the precursor ion of interest into the acquisition software with a value for the collision-induced dissociation (CID) collision energy and expected retention time range. A putative annotation was considered confirmed if the retention time, measured exact mass and fragmentation pattern of the metabolite reasonably match these data from chemical references. If published MS/MS data was available, the fragmentation pattern of the sample must additionally match the published MS/MS spectra.
Metabolomics
Embryoid Bodies
In culture medium from human embryoid bodies, we identified 3246 total features using C18 chromatography followed by positive ionization ESI-QTOF. Of these features, 510 were considered secreted molecules, as defined by having at least a 10% increase in abundance in the cell samples compared to the media controls. This standard was used in defining secreted molecules for all data sets. In 0.1% and 0.3% ethanol treated EBs, 70 and 62 secreted features were statistically significant, respectively, from control-treated cells (p<0.05, Supplemental Table 1). There were a total of 9 significant features in common between the two ethanol treatments (Table 1). One of the most challenging aspects of metabolomics is that >50% of the features identified do not have annotation in public databases yet. Of these 9 features, two had molecular weights consistent with a putative annotation of diethylphosphate and heptacarboxylporphyrin III. Three were tentatively assigned identities as tripeptides and four additional small molecules were unannotated. Several of these identities remain putative given the unavailability of a commercially available chemical standard. We confirmed the identity of 5'-methylthioadenosine (MTA, exact mass 297.0895) using MS/MS (Figure 2, Table 2). MTA participates in polyamine synthesis and methionine metabolism. MTA levels were significantly decreased in 0.3% ethanol-treated EBs (p=0.0275). MTA was also decreased in 0.1% ethanol-treated EBs, although the reduction was not statistically significant (p=0.1251).
Table 1
Table 1
Metabolome features significantly altered by both 0.1% and 0.3% ethanol treatment.
Figure 2
Figure 2
Identification and abundance of 5’-methylthioadenosine (MTA) in ethanol-treated embryoid bodies. A. Extracted ion chromatogram (EIC) in positive ionization mode compares the abundance of MTA in individual sample media plotted against its characteristic (more ...)
Table 2
Table 2
Summary of MS/MS Confirmed Molecules at Various Differentiation Time Points.
Using C18 chromatography and negative ionization ESI-QTOF, 1009 total features were measured of which 295 were secreted. Of the secreted features, 21 and 27 were significantly altered in 0.1% and 0.3% ethanol-treated EBs, respectively, when compared to controls (p<0.05, Supplemental Table 2). Two secreted features were significantly altered in both ethanol treatment groups, and based on their exact mass and retention time they were putatively identified as the methylated purine 7-methylinosine (exact mass 283.1060) and succinyladenosine (exact mass 383.1075, Table 1). Succinyladenosine levels were significantly increased in both 0.1% and 0.3% ethanol-treated EBs compared with untreated controls (p=0.0380 and 0.0387, respectively). Succinyladenosine is a biomarker for adenylosuccinate lyase deficiency but it has also been measured in the cerebrospinal fluid of children without this condition, using HPLC-negative ion ESI mass spectrometry on a C18 column (Krijt et al., 1999). Unfortunately, no chemical standard was commercially available for chemical identity confirmation. Using negative ionization ESI-QTOF MS/MS, we also confirmed the identity of thyroxine (exact mass 776.6867; Figure 3, Table 2). Thyroxine was significantly increased in 0.1% ethanol-treated EBs compared with controls (24% increase, p=0.0011).
Figure 3
Figure 3
Identification and abundance of thyroxine in ethanol-treated embryoid bodies (EBs). A. EIC in negative ionization mode compares the abundance of thyroxine in individual sample media plotted against its characteristic retention time following separation (more ...)
Neural Progenitors
At the neural progenitor differentiation stage, we measured 927 total features with C18 chromatography and positive ionization ESI-QTOF. Nearly half of the total features were considered secreted (n=434). In comparison to untreated neural progenitors, 56 and 20 features were significantly different in 0.1% and 0.3% ethanol treatment groups, respectively (p<0.05, Supplemental Table 3). Between the two ethanol treatment groups, there was one significant feature in common, which had a decreased abundance in ethanol treatment (Table 1). It could not be assigned an identity and had neutral mass of 746.0645 Da. In addition, L-Kynurenine (exact mass 208.0848), a tryptophan metabolite, was significantly increased in 0.1% ethanol-treated neural progenitors (p=0.0200). The identity of kynurenine was confirmed using MS/MS (Figure 4).
Figure 4
Figure 4
Identification and abundance of L-kynurenine in ethanol-treated neural progenitors. A. EIC in positive ionization mode compares the abundance of kynurenine in individual sample media plotted against its characteristic retention time following separation (more ...)
Neurons
In terminally differentiated neurons, 1268 total features were measured by C18 chromatography and positive ionization ESI-QTOF. The total number of secreted features was 582. In 0.1% and 0.3% ethanol-treated neurons, 23 and 19 features, respectively, were statistically significant when compared with control samples (p<0.05, Supplemental Table 4). Three significant features were common to both ethanol treatment groups (p<0.0001 for all vs. control neurons, Table 1). The EIC of the primary feature is shown in Figure 5. These three features were tentatively identified as a tripeptide containing cys-his-lys (exact mass 386.1734), and its respective sodium (exact mass 408.1558) and potassium adducts (exact mass 424.1308). The amino acid sequence order could not be determined. We confirmed the identity of indoleacetaldehyde (IAA, exact mass 159.06842), another metabolite in the tryptophan pathway, with MS/MS (Figure 6). IAA was significantly increased in the media of 0.1% ethanol-treated neurons compared to controls (p=0.0493). IAA levels were also increased in 0.3% ethanol-treated EBs compared to controls, although the increase was not statistically significant (p=0.47).
Figure 5
Figure 5
Identification and abundance of an unknown feature with an exact mass of 386.1734 in ethanol-treated neurons. EIC in positive ionization mode compares the feature’s abundance in individual sample media plotted against its retention time following (more ...)
Figure 6
Figure 6
Identification and abundance of indoleacetaldehyde (IAA). A. Metabolomics EIC of IAA in ethanol-treated neurons comparing abundance to retention time of the compound in various sample types. The solid lines are control samples, dotted lines are 0.1% ethanol (more ...)
Immunocytochemistry
To determine if ethanol exposure affected the differentiation of these neural populations, we examined the expression of neural markers at the end of the four-day alcohol exposure period. For the neural progenitors, we found no alcohol-related differences in their expression of nestin, an intermediate filament protein expressed in mitotic neural progenitors (Table 3). Control, 0.1% and 0.3% ethanol treatments all exhibited greater than 95% nestin expression in the progenitor cultures, confirming their neural progenitor identity. For the neurons, we observed no significant differences in the percentage of cells that co-expressed β-III-tubulin, demarcating neuronal populations, and tyrosine hydroxylase, a marker of dopaminergic neurons (Table 3, Supplemental Figure 1). Similarly, there were no significant differences in the percentage of cells that co-expressed β-III-tubulin and synaptophysin or SV2A; the latter are both integral membrane glycoproteins present on presynaptic vesicles.
Table 3
Table 3
Distribution of Neural and Apoptosis Markers*
We found modest ethanol effects on apoptosis of human neural cultures (Table 3, Supplemental Figure 2). For the neural progenitors there was no statistically significant difference in the percentage of cleaved caspase-3-positive cells in controls and those treated with 0.3% ethanol (p=0.19); there was a non-significant trend to reduced number of cleaved caspase-3-positive cells following treatment with 0.1% ethanol (p=0.06). In the neurons (D32), we detected a statistically significant decrease in the percentage of caspase-3-positive cells in 0.1% ethanol-treated neurons compared to controls (p=0.03), but not in those treated with 0.3% ethanol (p=0.10).
In this study, hES cells were differentiated into neuronal lineages and exposed to either 0, 0.1 or 0.3% ethanol at three developmental stages, EBs, neural progenitors and neurons, to model the effects of ethanol exposure on early human brain development. This in vitro model was coupled with metabolomics to test the hypothesis that ethanol exposure induces statistically significant changes to endogenous small molecules. We also sought to define specific small molecules and biochemical pathways that are altered in alcohol-induced developmental neurotoxicity, leading to FASD. Neural differentiation of hES cells is an important model to study the earliest effects of alcohol on human development given that the timing of neuronal and glial differentiation from hES cells is similar to lineage allocation in vivo (Zhang, 2006). Alcohol exposure during embryogenesis and early neural differentiation, a period when pregnancy may be unknown, leads to a higher incidence of craniofacial defects and mental disabilities (Ernhart et al., 1987; Guerri, 2002). There is a significant need to identify the biochemical pathways that play a causal role in alcohol-induced developmental neurotoxicity. Our study confirms that the exposure of hES cell-derivatives to ethanol results in significant changes to the abundance of human endogenous metabolites.
No significant features were shared between the three cell types following ethanol treatment. This likely reflects the disparate maturation states of these differentiating cells. Approximately half of the features showed similar responses at both ethanol concentrations, although the magnitude of change did not always achieve statistical significance. Other features, such as T4 and kynurenine, were increased at 0.1% ethanol but had decreased abundance at 0.3%. This differential response likely reflects the cells’ adaptive responses to the higher ethanol level over the four-day culture period (e.g. Buck and Harris, 1991; Yakovleva et al. 2011). This interpretation is endorsed by the finding that 0.1% ethanol treatment produced appreciably more significant features (154) than did 0.3% ethanol (112), of which 15 were significant in both exposures. Many individuals with FASD experience chronic alcohol exposure and biomarkers that reflect those adaptive mechanisms are clinically relevant.
Embryoid bodies are comprised of the three primary germ layers and thus model the earliest stages of differentiation. We identified statistically significant changes in multiple developmentally important metabolites between 0, 0.1% and 0.3% ethanol-treated EBs. The abundance of 5’-methylthioadenosine (MTA) was significantly decreased in 0.3% ethanol-treated EBs compared to untreated controls. MTA is produced in all mammalian cells during the synthesis of polyamines putrescine, spermidine and spermine from decarboxylated S-adenosylmethionine (SAM), and its levels directly reflect the rates of polyamine synthesis (Avila et al., 2004; Sufrin et al., 1989). Inhibition of polyamine synthesis impedes embryo growth (Fozard et al. 1980; Mendez 1989) whereas exogenous polyamines protect against embryotoxicants (Chirino-Galindo et al., 2009). Both acute and chronic ethanol exposure suppress polyamine synthesis and its dysregulation was identified as a potential mechanism contributing to FASD (Desiderio et al., 1987; Sessa and Perin, 1997; Shibley et al., 1995; Thadani et al., 1977). The decline in MTA is consistent with those reports and thus secreted MTA may be a useful biomarker for alcohol’s toxicity. MTA also supplies the sulfur atom for methionine synthesis within the methionine salvage pathway. Reductions in MTA might also contribute to the depleted methionine and SAM levels observed in chronic ethanol consumption (Finkelstein et al., 1974; Walcher and Miller, 2008). Importantly, because MTA is produced from SAM, its alteration may reflect a larger disruption of methyl group homeostasis within ethanol-exposed cells, an interpretation affirmed by the increased folic acid and polyglutamate content found in ethanol-treated EBs and neural progenitors (Supplemental Table 1 and 2). Methyl group supplements in the form of choline, betaine and SAM can prevent or reverse ethanol’s developmental toxicity both in vitro and in vivo (Seyoum and Persaud, 1994; Thomas et al. 2009, 2010). To our knowledge there are no previous reports specifically linking MTA to ethanol exposure or FASD. Its metabolic links with polyamine and methyl metabolism make MTA an excellent potential biomarker in addition to offering mechanistic insights into FASD.
We additionally confirmed thyroxine as a putative biomarker of prenatal ethanol exposure during the embryogenesis stage. Thyroxine abundance was significantly increased in 0.1% ethanol-treated EBs compared to untreated controls. Thyroxine (T4) is one of the two hormones known as thyroid hormone secreted by the thyroid gland. In these cultures T4 likely originated from the serum in the culture medium. Thyroxine signaling through its nuclear receptors is critical for normal central nervous system development and contributes to multiple processes including cell migration, dendrite/axon outgrowth and synaptogenesis. A precise balance of thyroid hormone is necessary for proper brain development and both deficiency and hyperthyroidism can cause significant and irreversible brain dysfunction (Bernal and Nunez, 1995; Lauder, 1977; Pharoah et al., 1971). Our studies show a 24% increase in exogenous thyroxine levels, suggesting an ethanol-induced perturbation to thyroxine metabolism and/or its uptake by the early embryo. This is consistent with prior demonstrations of altered thryroid hormone metabolism in animal models and infants with FASD (Cudd et al. 2002; Hannigan and Bellisario 1990; Kornguth et al. 1979). In rat models, prenatal alcohol exposure (PAE) also alters the levels of iodothyronine deiodinase III and thyroid hormone receptor alpha-1 (Shukla et al., 2010, 2011). Thus the significant increase in thyroxine detected herein suggests that ethanol exposure is likely to impair proper thyroid hormone metabolism in the early embryo, corroborating a proposed mechanistic pathway that ethanol may produce neurodevelopmental disorders by negatively affecting maternal-fetal hormonal homeostasis. Our study is the first to indicate a differential uptake or processing of maternal thyroxine by human embryos as a result of PAE. It is plausible to speculate that this detrimental effect may be mediated by functional impairment of the thyroid hormone-specific transporters Oatp1c1 (organic anion transporting polypeptide 1c1), monocarboxylate transporter (MCT) 8 and MCT10. Mutations to MCT8 actually increase the extracellular abundance of thyroxine in vivo, producing severe psychomotor retardation (Visser et al. 2008; Grijota-Martinez et al., 2011). There are currently no published studies examining the direct effects of ethanol exposure on thyroxine-specific transporters, which will be a focus of future studies.
Ethanol exposure at two different stages of neurogenesis produced statistically significant changes to tryptophan metabolism. These changes are summarized in Figure 7. These differences were not attributed to trivial effects of ethanol upon differentiation states, as the cellular type distribution between treatments did not differ. Tryptophan is metabolized primarily along the kynurenine pathway and kynurenine levels were increased in neural progenitors treated with 0.1% ethanol, as were the levels of the tryptamine catabolite indoleacetaldehyde in neurons exposed to 0.1% ethanol. Dysregulation of tryptophan metabolism is reported in several models of alcohol exposure. Kynurenine levels are increased in human plasma following acute ethanol consumption (Badawy et al. 2009) and tryptophan levels are significantly reduced in the plasma, brain and liver of rat pups following gestational ethanol exposure (Lin et al. 1990). Serum levels of kynurenine are also perturbed in several cognitive disorders such as Down syndrome, bipolar disorder and postpartum depression but had not been previously described in FASD. Kynurenine has additionally been identified in hES cells treated with valproate, a known inducer of neurodevelopmental defects (Cezar et al. 2007). Kynurenine was recently reported to activate the aryl hydrocarbon xenobiotic receptor (AhR) with an apparent Kd ~4 µM (Optiz et al. 2011). AhR activation adversely affects brain development and behavior (Latchney et al. 2011; Seo et al. 1999). Our finding of elevated Kyn in ethanol-treated neural progenitors is a novel insight into the mechanisms underlying FASD, and confers this metabolite a translational role as a candidate biomarker of FASD.
Figure 7
Figure 7
Summary of ethanol-induced changes in tryptophan metabolism in human ES cells. Black font indicates metabolites that are detected in media of ethanol-treated human EBs, neural precursors, and neurons. Asterisk indicates those metabolites whose levels (more ...)
Our study is the first to indicate a direct effect of ethanol exposure on the human endogenous metabolite IAA. Tryptophan is critical for normal neurogenesis as the precursor to the neurotransmitter serotonin (5HT) and its availability has a direct effect on 5HT synthesis (Fernstrom and Wurtman, 1971). Development of the serotonergic system occurs earlier than other neurotransmitter systems and its sensitivity to disruption by PAE is well documented (Druse et al., 1991; Maier et al., 1996; Sari et al., 2010). Elevated IAA decreased the activity of tryptophan hydroxylase, the rate-limiting enzyme in serotonin synthesis, by as much as 33% (Nilsson and Tottmar, 1987). Thus, elevations of kynurenine and IAA detected here could be biomarkers for a metabolic diversion of tryptophan away from serotonin synthesis and toward these alternate metabolic pathways and may explain the reduced tryptophan and 5HT content of the ethanol-exposed fetal brain (Maier et al., 1996; Sari et al., 2010). In fact, an increase in the IAA-related 5-hydroxylindoleacetic acid (5-HIAA) was measured in the urine of rats chronically exposed to ethanol (Bonner et al., 1993) and IAA itself is amenable to detection in urine, which renders it a translational application in biofluids. Alternatively, we note that kynurenine also lies along the synthetic pathway of nicotinamide from its tryptophan precursor. Ethanol oxidation places a high requirement for niacin and NADPH reducing equivalents, and one possibility is that ethanol’s effects on cellular redox potential might be driving these changes in tryptophan metabolites. Whether one or both of these hypotheses underlie increased kynurenine levels in ethanol-exposed neural progenitors, serving as the cause of serotonergic dysfunction, will be elucidated in future mechanistic studies. Nonetheless, it is noteworthy that significant perturbations to the tryptophan pathway and its metabolites were consistent across different timelines of neural differentiation, neural precursors and neurons.
In addition to the four confirmed metabolites with marked changes resulting from ethanol exposure, we identified multiple unknown compounds with statistically significant changes in abundance. Further experiments are required to determine the chemical formula of such compounds by ion fragmentation and other analytical platforms such as NMR. The fact that statistically significant small molecules are not annotated in public databases does not preclude their potential application as candidate diagnostic biomarkers for FASD. Given the highly sensitive, high resolution nature of the mass spectrometric detection system employed in this study, a measurable endpoint (exact mass) is determined which can be used for continued chemical structure determinations. An important limitation to the lack of a chemical annotation is that these unannotated small molecules cannot be mapped to a biochemical pathway for mechanistic elucidation.
No changes in the overall cell differentiation ability of hES cells were measured in response to 0.1% or 0.3% ethanol (Table 3), at least for the culture conditions employed in this study and for the relatively short ethanol exposure period. There was a non-significant trend for an increased percentage of neurons expressing synaptophysin and SV2A following the ethanol treatment of mature neurons. Together with the multiple changes in the levels of tryptophan metabolites, these data may indicate an overall trend of increased synaptogenesis in response to ethanol exposure. In a finding that seemed counterintuitive, we found modest non-significant decreases in apoptosis frequency following ethanol treatment of neural precursors and neurons. Prock and Miranda (2007) reported a similar resistance to apoptosis in ethanol-treated mouse neural progenitors, a population substantially similar to those studied herein. Our findings endorse their interpretation that apoptosis in response to ethanol may be a differentiation state-specific response. Further study is required to understand why this population is resistant to apoptosis upon exposure to ethanol.
In conclusion, exposing hES cell-derived EBs, neural progenitors and neurons to ethanol identified marked changes in biochemical pathways and metabolites critical for proper neurodevelopment. Our findings not only provide novel candidate diagnostic biomarkers for FASD, but also corroborate independent studies that have proposed alterations to methionine, methyl group, thyroid hormone and tryptophan metabolism as potential mechanisms underlying FASD. Although some of these metabolites, i.e. kynurenine, have been described as putative biomarkers for other neurodevelopmental disorders, it is the integration of several of these endogenous small molecules from independent biochemical pathways, that may be used as a specific fingerprint test set or specific fingerprint for the diagnosis of FASD in biofluids. Metabolomics is able to measure simultaneous changes to multiple unrelated biochemical pathways, providing not only a biochemical signature specific to FASD but also mechanistic insight into the effects of PAE on early human embryogenesis. The identification of differential biomarkers in our in vitro model prior to analysis of biofluids allows for a more targeted approach in an in vivo model, which has multiple confounding factors, i.e. maternal-fetal interactions. Biofluids are complex mixtures of systemic byproducts influenced by both endogenous and exogenous factors e.g., genetics, diet and environment. Using our strategy, we can design specific analytical protocols to examine biofluids based on our confirmed metabolites. Our future efforts will examine whether molecules detected here are subject to similar changes in in vivo models and children diagnosed with FASD. The biochemical pathways revealed by metabolomics of hES cell derivatives may also present new targets for the identification and development of novel therapies for FASD.
Supplementary Material
Supp Fig S1-2 & Table S1-S4
Acknowledgments
Supported by NIH awards AA16958 (G.G.C.) and MERIT award AA11085 (S.M.S.).
  • Avila MA, García-Trevijano ER, Lu SC, Corrales FJ, Mato JM. Methylthioadenosine. Int J Biochem Cell Biol. 2004;36:2125–2130. [PubMed]
  • Badawy AA, Doughrty DM, Marsh-Richard DM, Steptoe A. Activation of liver tryptophan pyrrolase mediates the decrease in tryptophan availability to the brain after acute alcohol consumption by normal subjects. Alcohol Alcohol. 2009;44:267–271. [PMC free article] [PubMed]
  • Ben-Hur T, Idelson M, Khaner H, Pera M, Reinhartz E, Itzik A, Reubinoff BE. Transplantation of human embryonic stem cell-derived neural progenitors improves behavioral deficit in Parkinsonian rats. Stem Cells. 2004;22:1246–1255. [PubMed]
  • Bernal J, Nunez J. Thyroid hormones and brain development. Eur J Endocrinol. 1995;133:390–398. [PubMed]
  • Bonner AB, Brien S, Preedy VR. The urinary excretion of tryptophan and tryptophan metabolites in the chronic ethanol-fed rat. J Pharm Pharmacol. 1993;45:81–85. [PubMed]
  • Branchey L, Shaw S, Lieber CS. Ethanol impairs tryptophan transport into the brain and depresses serotonin. Life Sci. 1981;29:2751–2755. [PubMed]
  • Buck KJ, Harris RA. Neuroadaptive responses to chronic ethanol. Alcohol Clin Exp Res. 1991;15:460–470. [PubMed]
  • Cezar GG, Quam JA, Smith AM, Rosa GJM, Piekarczyk MS, Brown JF, Gage FH, Muotri AR. Identification of small molecules from human embryonic stem cells using metabolomics. Stem Cells Dev. 2007;16:869–882. [PubMed]
  • Chan D, Bar-Oz B, Pellerin B, Paciorek C, Klein J, Kapur B, Farine D, Koren G. Population baseline of meconium fatty acid ethyl esters among infants of nondrinking women in Jerusalem and Toronto. Ther Drug Monit. 2003;25:271–278. [PubMed]
  • Chirino-Galindo G, Baiza-Gutman LA, Barrera-Escorcia E, Palomar-Morales M. Polyamines protect rat embryo in vitro from high glucose-induced developmental delay and dysmorphogenesis. Birth Defects Res B Dev Reprod Toxicol. 2009;86:58–64. [PubMed]
  • Cudd TA, Chen WJ, West JR. Fetal and maternal thyroid hormone responses to ethanol exposure during the third trimester equivalent of gestation in sheep. Alcohol Clin Exp Res. 2002;26:53–58. [PubMed]
  • Desiderio MA, Sessa A, Perin A. Polyamines and diamine oxidase activity in maternal, embryonal, and fetal tissues of rat after chronic ethanol consumption. Biochem Biophys Res Commun. 1987;142:843–848. [PubMed]
  • Druse MJ, Kuo A, Tajuddin N. Effects of in utero ethanol exposure on the developing serotonergic system. Alcohol Clin Exp Res. 1991;15:678–684. [PubMed]
  • Ernhart CB, Sokol RJ, Martier S, Moron P, Nadler D, Ager JW, Wolf A. Alcohol teratogenicity in the human: a detailed assessment of specificity, critical period, and threshold. Am J Obstet Gynecol. 1987;156:33–39. [PubMed]
  • Fernstrom JD, Wurtman RJ. Brain serotonin content: physiological dependence on plasma tryptophan levels. Science. 1971;173:149–152. [PubMed]
  • Fiehn O. Metabolomics – the link between genotypes and phenotypes. Plan Mol Biol. 2002;48:155–171. [PubMed]
  • Finkelstein JD, Cello JP, Kyle WE. Ethanol-induced changes in methionine metabolism in rat liver. Biochem Biophys Res Commun. 1974;61:525–531. [PubMed]
  • Fozard JR, Prat ML, Prakash NJ, Grove J, Schechter PJ, Sjoerdsma A, Koch-Weser J. L-Ornithine decarboxylase: an essential role in early mammalian embryogenesis. Science. 1980;208:505–509. [PubMed]
  • Grijota-Martínez C, Díez D, Morreale de Escobar G, Bernal J, Morte B. Lack of action of exogenously administered T3 on the fetal rat brain despite expression of the monocarboxylate transporter 8. Endocrinology. 2011;152:1713–1721. [PubMed]
  • Guerri C. Mechanisms involved in central nervous system dysfunctions induced by prenatal ethanol exposure. Neurotox Res. 2002;4:327–335. [PubMed]
  • Hannigan JH, Bellisario RL. Lower serum thyroxine levels in rats following prenatal exposure to ethanol. Alcohol Clin Exp Res. 1990;14:456–460. [PubMed]
  • Kornguth SE, Rutledge JJ, Sunderland E, Siegel F, Carlson I, Smollens J, Juhl U, Young B. Impeded cerebellar development and reduced serum thyroxine levels associated with fetal alcohol intoxication. Brain Res. 1979;177:347–360. [PubMed]
  • Krijt J, Kmoch S, Hartmannová H, Havlícek V, Sebesta I. Identification and determination of succinyladenosine in human cerebrospinal fluid. J Chromatogr B Biomed Sci Appl. 1999;726:53–58. [PubMed]
  • Latchney SE, Lioy DT, Henry EC, Gasiewicz TA, Strathmann FG, Mayer-Pröschel M, Opanashuk LA. Neural precursor cell proliferation is disrupted through activation of the aryl hydrocarbon receptor by 2,3,7,8-tetrachlorodibenzo-p-dioxin. Stem Cells Dev. 2011;20:313–326. [PMC free article] [PubMed]
  • Lauder JM. The effects of early hypo- and hyperthyroidism on the development of rat cerebellar cortex. III. Kinetics of cell proliferation in the external granular layer. Brain Res. 1977;126:31–51. [PubMed]
  • Lindon JC, Holmes E, Nicholson JK. So what’s the deal with metabonomics? Anal Chem. 2003;75:384A–391A. [PubMed]
  • Litten RZ, Bradley AM, Moss HB. Alcohol biomarkers in applied settings: recent advances and future research opportunities. Alcohol Clin Exp Res. 2010;34:955–967. [PubMed]
  • Maier SE, Chen WJ, West JR. Prenatal binge-like alcohol exposure alters neurochemical profiles in fetal rat brain. Pharmacol Biochem Behav. 1996;55:521–529. [PubMed]
  • Mendez JD. Polyamines and human reproduction. In: Bachrach U, Heimer Y, editors. The Physiology of Polyamines. vol. 1. Boca Raton, FL: CRC Press; 1989. pp. 23–28.
  • Miller MW. Brain Development. New York: Oxford University Press; 2006.
  • Nilsson GE, Tottmar O. Effects of biogenic aldehydes and aldehyde dehydrogenase inhibitors on rat brain tryptophan hydroxylase activity in vitro. Brain Res. 1987;409:374–379. [PubMed]
  • Opitz CA, Litzenburger UM, Sahm F, Ott M, et al. An endogenous tumour-promoting ligand of the human aryl hydrocarbon receptor. Nature. 2011;478:197–203. [PubMed]
  • Pharoah PO, Buttfield IH, Hetzel BS. Neurological damage to the fetus resulting from severe iodine deficiency during pregnancy. Lancet. 1971;1(7694):308–310. [PubMed]
  • Prock TL, Miranda RC. Embryonic cerebral cortical progenitors are resistant to apoptosis, but increase expression of suicide receptor DISC-complex genes and suppress autophagy following ethanol exposure. Alcohol Clin Exp Res. 2007;31:694–703. [PMC free article] [PubMed]
  • Reubinoff BE, Pera MF, Fong CY. Embryonic stem cell lines from human blastocysts: somatic differentiation in vitro. Nat Biotechnol. 2000;18:399–404. [PubMed]
  • Santillano DR, Kumar LS, Prock TL, Camarillo C, Tingling JD, Miranda RC. Ethanol induces cell-cycle activity and reduces stem cell diversity to alter both regenerative capacity and differentiation potential of cerebral cortical neuroepithelial precursors. BMC Neurosci. 2005;6:59. [PMC free article] [PubMed]
  • Sari Y, Hammad LA, Saleh MM, Rebec GV, Mechref Y. Alteration of selective neurotransmitters in fetal brains of prenatally alcohol-treated C57BL/6 mice: quantitative analysis using liquid chromatography/tandem mass spectrometry. Int J Dev Neurosci. 2010;28:263–269. [PMC free article] [PubMed]
  • Seo BW, Sparks AJ, Medora K, Amin S, Schantz SL. Learning and memory in rats gestationally and lactationally exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) Neurotoxicol Teratol. 1999;21:231–239. [PubMed]
  • Sessa A, Perin A. Ethanol and polyamine metabolism: physiologic and pathologic implications: a review. Alcohol Clin Exp Res. 1997;21:318–325. [PubMed]
  • Seyoum G, Persaud TV. In vitro effect of S-adenosyl methionine on ethanol embryopathy in the rat. Exp Toxicol Pathol. 1994;46:177–181. [PubMed]
  • Shibley IA, Jr, Gavigan MD, Pennington SN. Ethanol's effect on tissue polyamines and ornithine decarboxylase activity: a concise review. Alcohol Clin Exp Res. 1995;19(1):209–215. [PubMed]
  • Shukla PK, Sittig LJ, Ullmann TM, Redei EE. Candidate Placental Biomarkers for Intrauterine Alcohol Exposure. Alcohol Clin Exp Res. 2010;35:559–565. [PMC free article] [PubMed]
  • Sittig LJ, Shukla PK, Herzing LB, Redei EE. Strain-specific vulnerability to alcohol exposure in utero via hippocampal parent-of-origin expression of deiodinase-III. FASEB J. 2011;25:2313–2324. [PubMed]
  • Smith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak G. XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal Chem. 2006;78:779–787. [PubMed]
  • Sufrin JR, Spiess AJ, Kramer DL, Libby PR, Porter CW. Synthesis and antiproliferative effects of novel 5'-fluorinated analogues of 5'-deoxy-5'-(methylthio)adenosine. J Med Chem. 1989;32:997–1001. [PubMed]
  • Tautenhahn R, Böttcher C, Neumann S. Highly sensitive feature detection for high resolution LC/MS. BMC Bioinformatics. 2008;9:504. [PMC free article] [PubMed]
  • Thadani PV, Slotkin TA, Schanberg SM. Effects of late prenatal or early postnatal ethanol exposure on ornithine decarboxylase activity in brain and heart of developing rats. Neuropharmacology. 1977;16:289–293. [PubMed]
  • Thomas JD, Abou EJ, Dominguez HD. Prenatal choline supplementation mitigates the adverse effects of prenatal alcohol exposure on development in rats. Neurotoxicol Teratol. 2009;31:303–311. [PMC free article] [PubMed]
  • Thomas JD, Idrus NM, Monk BR, Dominguez HD. Prenatal choline supplementation mitigates behavioral alterations associated with prenatal alcohol exposure in rats. Birth Defects Res A. 2010;88:827–837. [PMC free article] [PubMed]
  • Thomson JA, Itskovitz-Eldor J, Shapiro SS, Waknitz MA, Swiergiel JJ, Marshall VS, Jones JM. Embryonic stem cell lines derived from human blastocysts. Science. 1998;282:1145–1147. [PubMed]
  • Visser WE, Friesema EC, Jansen J, Visser TJ. Thyroid hormone transport in and out of cells. Trends Endocrinol Metab. 2008;19:50–56. [PubMed]
  • Walcher BN, Miller RR., Jr Ethanol-induced increased endogenous homocysteine levels and decreased ratios of SAM/SAH are only partially attenuated by exogenous glycine in developing chick brains. Comp Biochem Physiol C Toxicol Pharmacol. 2008;147:11–16. [PubMed]
  • West PR, Weir AM, Smith AM, Donley EL, Cezar GG. Predicting human developmental toxicity of pharmaceuticals using human embryonic stem cells and metabolomics. Toxicol Appl Pharmacol. 2010;247:18–27. [PubMed]
  • Yakovleva T, Bazov I, Watanabe H, Hauser KF, Bakalkin G. Transcriptional control of maladaptive and protective responses in alcoholics: a role of the NF-κB system. Brain Behav Immun. 2011;25(Suppl 1):S29–S38. [PMC free article] [PubMed]
  • Zhang SC, Wering M, Duncan ID, Brustle O, Thomson JA. In vitro differentiation of transplantable neural precursors from human embryonic stem cells. Nat Biotechnol. 2001;19:1129–1133. [PubMed]
  • Zhang SC. Neural subtype specification from embryonic stem cells. Brain Pathol. 2006;16:132–142. [PubMed]