Address correspondence and reprint requests to M. Rao, Laboratory of Neurosciences, Gerontology Research Center, National Institute on Aging, Room 4E02, 5600 Nathan Shock Drive, Baltimore, MD 21224, USA. E-mail: firstname.lastname@example.org
To access and compare gene expression in fetal neuroepithelial cells (NEPs) and progenitor cells, we have used microarrays containing approximately 500 known genes related to cell cycle regulation, apoptosis, growth and differentiation. We have identified 152 genes that are expressed in NEPs and 209 genes expressed by progenitor cells. The majority of genes (141) detected in NEPs are also present in progenitor populations. There are 68 genes specifically expressed in progenitors with little or no expression in NEPs, and a few genes that appear to be present exclusively in NEPs. Using cell sorting, RT–PCR, in situ hybridization or immunocytochemistry, we have examined the segregation of expression to neuronal and glial progenitors, and identified several that appeared to be enriched in neuronal (e.g. CDK5, neuropilin, EphrinB2, FGF11) or glial (e.g. CXCR4, RhoC, CD44, tenascin C) precursors. Our data provide a first report of gene expression profiles of neural stem and progenitor cells at early stages of development, and provide evidence for the potential roles of specific cell cycle regulators, chemokines, cytokines and extracellular matrix molecules in neural development and lineage segregation.
Neural stem cells (NSCs) are self-renewing multipotent precursors that generate neurons, astrocytes and oligodendrocytes in the CNS (Gage et al. 1995; Weiss et al. 1996). Fetal NSCs have attracted attention not only for their role in normal development but also for their potential use for the treatment of neurodegenerative disorders (Studer et al. 1998; Sanchez-Pernaute et al. 2001; Kim et al. 2002). Our ability to regulate NSC differentiation and harness their potential for therapy is limited, however, and would be enhanced if we understood the process of NSC differentiation better. The genetic program for NSC differentiation may be similar to that of hematopoietic stem cell (HSC) differentiation where specific sets of genes regulate quiescence, self-renewal and differentiation (Geschwind et al. 2001; Terskikh et al. 2001), and either the ‘same stem cell genes’ or unique subsets of functionally similar genes may be used to regulate the neural development program. Alternatively, novel pathways regulating NSC differentiation may exist (Geschwind et al. 2001). Examination of gene expression patterns in NSC and comparison of genes expressed by NSCs with NSC-derived differentiated cells will probably promote our understanding of the process of NSC differentiation.
NSCs present in the developing neural tube, termed neuroepithelial cells (NEPs), provide a useful system for the analysis of neural development. NEPs are capable of self-renewal and generate differentiated cells by first producing intermediate precursor cells. Neuronal restricted precursors (NRPs) and glial restricted precursors (GRPs) are two such precursors that will undergo further maturation to finally generate neurons, astrocytes and oligodendrocytes (Kalyani et al. 1997; Mayer-Proschel et al. 1997). At embryonic day (E)10.5 (rats) the neural tube consists of a homogenous population of NEPs with the earliest differentiation occuring at E11 (Nornes and Das 1974; Altman and Bayer 1984) and is therefore a source of a highly enriched stem cell population. NEPs derived at this stage have (1) high telomerase activity; (2) expression of stem cell markers BCRP-1 (an ATP binding cassette transporter), Sox2 and Fz-9; and (3) lack of the expression of PSANCAM (embryonic cell neural adhesion molecule; E-NCAM), A2B5 and other markers characteristic of neurons and glia (Cai et al. 2002). During development, the percentage of NEPs rapidly declines and, at E14.5, NEPs comprise less than 10% of the neural tube population. The predominant cell populations present at this stage are NRPs and GRPs, which comprise about 80% of the cells (Kalyani et al. 1997; Mujtaba et al. 1999; Cai et al. 2002). The remaining population consists of some post-mitotic neurons and small numbers of endothelial cells and connective tissue elements (Cai, unpublished observations). E14.5 is therefore a useful stage at which neuronal and glial progenitors can be harvested in large number (Kalyani et al. 1997). Prospective identification of NRP-enriched and GRP-enriched cells can be accomplished by using cell surface markers, E-NCAM or A2B5, for selection. In addition, these partially differentiated neural progenitors can be distinguished from the NEPs by their unique antigen expression and differentiation ability (Kalyani et al. 1997). Comparing gene expression in two purified progenitor populations can be used to identify lineage-specific genes.
Multiple strategies have been used to compare gene expression (Noordewier and Warren 2001). We have chosen to use cDNA microarray, a newly developed technology that allows simultaneous assessment of the expression of potentially thousands of genes. This method has been used successfully to identify genes associated with specific biological functions (Geschwind et al. 2001; Noordewier and Warren 2001). However, the limited number of NEPs and their progeny in the developing neural tube limits the amount of RNA available, and many of the commercial arrays lack candidate genes known to be expressed on stem cells, thus limiting the choice of available arrays and comparative strategies that can be used. We reasoned, however, based on analogy to HSC development, that genes known to be involved in proliferation, cell cycle and apoptosis are likely to play critical roles in NSC development (Phillips et al. 2000; Weissman 2000; Sommer and Rao 2002). We therefore chose to use focused cDNA microarrays that contained large numbers of such genes that are also expressed during HSC differentiation, and enhanced the sensitivity of our approach with reverse transcription using gene-specific primers. We compared the expression of 497 known genes in NEPs and restricted precursor cells using validated focused microarrays and RT–PCR. We have identified 152 genes and 209 genes expressed on cells present at E10.5 and E14.5 respectively. These include genes involved in cell proliferation, chemokine and cytokine signaling, and extracellular matrix (ECM) molecules. There are 141 genes expressed on NEPs that continue to be expressed on neural progenitors, and 68 genes that are expressed at higher levels at E14.5 with little or no expression at E10.5. Comparison of the expression of selected genes expressed at high levels at E14.5 identifies some that are specifically expressed in either neuronal or glial progenitor populations. Thus, our data provide a first description of gene expression in NSCs during early development and suggest potential roles of cell cycle control, chemokines, cytokines and ECM molecules during neuro-glial differentiation.
Materials and methods
Experimentally naive, pregnant Sprague–Dawley rats at day 8 gestation were purchased from Harlan Sprague–Dawley (Indianapolis, IN, USA) and housed individually in standard cages at the National Institute on Aging (NIA), National Institutes of Health (NIH). They were maintained at a temperature of 22°C on a 12/12 h light/dark cycle with free access to food (NIH-07) and water. The rats were allowed to acclimatize to the vivarium until the day of the experiment. The experimental handling and experimental procedures were approved by the NIA Animal Care and Use Committee.
Antibodies and chemicals
A polyclonal antibody against PTEN was purchased from Cascade Bioscience (Winchester, MA, USA). Polyclonal anti-tenascin C (TnC) and monoclonal anti-A2B5 and anti-CD44 antibodies were from Chemicon International, Inc (Temecula, CA, USA). Polyclonal antibodies against α-catenin and epidermal growth factor receptor (EGFR) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Monoclonal antibody against p27Kip1 was from Transduction Laboratory (San Diego, CA, USA). GEArray™ Q series cDNA expression array filters were obtained from SuperArray Inc (Bethesda, MD, USA). Collagenase type I was from Worthington Biochemical Corporation and Dispase II was from Boehringer Mannheim (Mannheim, Germany). RNAlater™ was from Ambion, Inc. (Austin, TX, USA).
Isolation, cell culture and cell sorting of neural cells
E10.5 and 14.5 Sprague–Dawley rat embryos were used to isolate NEPs and neuronal progenitor cells as described previously (Mayer-Proschel et al. 1997). Briefly, the rat embryos were removed and placed in a Petri dish containing ice-cold phosphate-buffered saline (PBS) (Invitrogen, Carlsbad, CA, USA). The trunk segments of the embryos (last 10 somites) were first dissected, rinsed, and then transferred to fresh cold PBS. To obtain NEPs, the trunk segments from E10.5 embryos were further incubated with an enzyme solution containing 1 mg/mL collagenase type I and 2 mg/mL Dispase II prepared with a Ca2+/Mg2+-free Hank's balanced salt solution (Invitrogen) for 10 min at room temperature (22°C). The segments were then gently triturated with a Pasteur pipette in NEP basal medium (Kalyani et al. 1998) containing 10% chicken embryo extract (CEE) to release neural tubes free from surrounding somites and connective tissue. The isolated neural tubes derived from both E10.5 and E14.5 stages were stored in a RNAlater™ solution (Ambion) at 4°C for late RNA isolation.
For cultures of rat E10.5 cells, tissue culture dishes were coated with fibronectin (15 µg/mL; Sigma, St Louis, MO, USA) overnight at 4°C. For cultures of E14.5 neural tube cells, tissue culture dishes were first coated with a solution containing 13.3 µg/mL poly l-lysine (Sigma) for 1 h at room temperature followed by incubation with 15 µg/mL laminin (Invitrogen, Life Technologies, Carlsbad, CA, USA) overnight at 4°C. Before use, the dishes were rinsed with medium. Acutely dissected neural tubes at E10.5 and E14.5 stage were incubated with 0.05% trypsin–EDTA (Invitrogen) for 10 min at 37°C and then gently triturated with a Pasteur pipette in NEP basal medium with 10% CEE and basic fibroblast gowth factor (FGF) (20 ng/mL; Upstate Biotechnology, Lake Placid, NY, USA). The dissociated cells were plated on the coated dishes, maintained at 37°C and 5% CO2 in NEP basal medium containing basic FGF (20 ng/mL) with (for E10.5 cells) or without (for E14.5 cells) 10% CEE for 2 days, and were used for immunostaining.
Cell sorting was used to isolate NRP-enriched and GRP-enriched cells from a mixture of neuronal progenitor cells by a single label sorting. Briefly, the fresh trypsin–EDTA-dissociated cells (107/mL) from E14.5 neural tubes were immunostained with antibodies against E-NCAM (1 : 5, DSHB, Developmental Studies, Hybridoma Bank, Iowa City, IA, USA) and A2B5 (a cell surface marker for GRP, 1 : 200; Chemicon), respectively, in NEP medium for 1 h at 4°C, washed three times with 1 mL NEP medium, and then incubated in FITC-conjugated anti-mouse IgM secondary antibody (1 : 200; Jackson Immuno-Research, West Grove, PA, USA) for 30 min at 4°C. To discriminate between live and dead cells, propidium iodide (1 µg/mL) was added to the cell suspension. The cell concentration was adjusted to 5 × 106/mL (PBS with 5% fetal calf serum) and filtered through a 70-micron nylon cell strainer. These E-NCAM-positive NRP-enriched and A2B5-positive GRP-enriched cells were sorted through a FACSstarPlus cell sorter (Becton Dickinson Immunocytometry Systems, Franklin Lakes, NJ, USA) at a rate of 2500 cells/s. The cells were collected at 4°C and then stored at −80°C for subsequent experiments. Sorting purities were 96 ± 4% (n = 4) and 92 ± 2% (n = 4) for E-NCAM- and A2B5-positive cells, respectively.
The non-radioactive GEArray™ Q series cDNA expression array filters (MM-001 N, MM-002 N, MM-003 N, MM-005 N, MM-007 N, MM-009 N and MM-010 N; SuperArray Inc.) were used and hybridization procedures were as described by manufacturer. Total RNA from rat E10.5 and E14.5 neural tubes was isolated by using TRIzol® (Invitrogen). The biotin dUTP-labeled cDNA probes were specifically generated in the presence of a designed set of gene-specific primers using total RNA (4 µg per filter) and 200 U Moloney Murine Lukemia Virus (MMLV) reverse transcriptase (Promega, Madison, WI, USA). The array filters were hybridized with biotin-labeled probes at 60°C for 17 h. The filters were then washed twice with 2 × saline sodium citrate buffer (SSC)/1% sodium dodecyl sulfate (SDS) and then twice with 0.1 × SSC/1% SDS at 60°C for 15 min each. Chemilumilescent detection steps were performed by subsequent incubation of the filters with alkaline phosphatase-conjugated streptavidin and CDP-Star substrate. All cDNA microarray experiments were performed twice with new filters and RNA isolated at different times. The positive and negative spots were identified and matched by at least two investigators. Only the matched positive and negative results of two experiments are presented.
The cDNA was synthesized using 1 µg total RNA in the presence of Superscript II and oligo(dT)12–18 (both from Invitrogen). The PCR was performed in a 20-µL reaction solution containing 2 µL 10 × PCR buffer, 150 µmol MgCl2, 10 µmol dNTP, 20 pmol primer, 1 µL 10 × diluted cDNA and 1 U RedTag DNA polymerase (Sigma). The PCR conditions were as follows: 35 cycles of 94°C for 30 s, 55°C for 30 s and 72°C for 30 s, and final extension for 10 min at 72°C. Primer sequences are available upon request.
Lysate preparation and western blot
The fresh neural tubes derived from rat E14.5 stage were homogenized with 300 µL ice-cold lysis buffer containing 25 mm Hepes, pH 7.5, 300 mm NaCl, 1.5 mm MgCl2, 0.2 mm EDTA, 0.1% Triton X-100, 20 mmβ-glycerophosphate, 0.1 mm sodium orthovanadate, 0.5 mm dithiothreitol, 100 µg/mL phenylmethylsulfonyl fluoride and 2 µg/mL leupeptin, followed by sonication for 10 s on ice. The cellular extracts were then centrifuged for 15 min at 12 000 × g to remove debris. The supernatant was collected, aliquoted and stored at −70°C. Protein concentration was determined with BCA protein assay kit reagent (Pierce, Rockford, IL, USA).
For western blot, equal amounts of supernatant protein (40 µg/lane) were resolved on 4–12% SDS–polyacrylamide gels and electrophoretically transferred to Immopine membrane. The membrane blots were first blocked with 10% non-fat dry milk in PBST buffer containing 10 mm sodium phosphate, pH 7.4, 500 mm NaCl, and 0.01% Tween-20 and then incubated overnight with primary antibodies (TnC, 1 : 1000; CD44, 1 : 100; α-catenin, 1 : 1000; p27Kip1, 1 : 1000) in PBST buffer, with the addition of 1% bovine serum albumin (BSA) at 4°C. Immunoreactivity was detected by sequential incubation with horseradish peroxidase-conjugated secondary antibody (1 : 10 000; Jackson Immuno-Research) and Western Lightning™ (PerkinElmer Life Science, Boston, MA, USA).
Staining procedures for cultured cells or cryostat sections were as described previously (Mayer-Proschel et al. 1997). Briefly, the cultures of E14.5 mixtures or the sections were first incubated with either monoclonal anti-A2B5 or monoclonal anti E-NCAM in NEP medium for 30 min at room temperature, washed with 1 mL Dulbecco's modified Eagle's medium three times, and then reacted with FITC-conjugated anti-mouse IgM secondary antibody (1 : 200; Jackson Immuno-Research) for 30 min at room temperature. Following this procedure, cells were fixed with 4% paraformaldehyde for 30 min at room temperature. For double-staining, these stained sections or cultured cells were further incubated with primary antibody diluted to an appropriate concentration in blocking buffer (1% BSA, 10 mm sodium phosphate, pH 7.4, 500 mm NaCl and 0.01% Triton X-100) overnight at 4°C, followed by incubation with an Alexa Fluor® 568 donkey anti-goat or anti-rabbit IgG (1 : 500; Molecular Probes, Eugene, OR, USA) at room temperature for 1 h. Finally, the cultures or sections were incubated with 1 µm bisbenzimide (DAPI; Sigma) in PBST for 5 min to label all nuclei. The slides were examined under a fluroresence microscope. Images were obtained using an Olympus microscope with a digital camera attachment. Composites were prepared using Adobe Photoshop.
In situ hybridization
Section in situ hybridization with digoxygenin-labeled antisense riboprobe was performed as described previously (Liu et al. 2002). Briefly, antisense TnC RNA probe was generated by in vitro transcription (T3-T7 MaxiScript Kit; Ambion) in the presence of digoxigenin-11-UTP (Boehringer Mannheim). Cryostat sections of neural tube derived from E16.5 rat embryos were first air-dried and then fixed with freshly prepared 4% paraformaldehyde for 30 min. The slides were subject to digestion with proteinase K (50 µg/mL; InnoGenex, San Ramon, CA, USA) for 8–12 min and then refixation with a mixture of 4% paraformaldehyde and 0.1% glutaraldehyde (Sigma) for 30 min at room temperature. The slides were rinsed with diethylpyro-carbonate (DEPC)-PBS, and then hybridized with TnC probes at 65°C overnight. After that, the slides were washed with 2 × SSC twice at 65°C, blocked in blocking buffer (20% sheep serum, 2% BSA, 0.1% Triton X-100 in PBS) for 30 min, and incubated with anti-digoxygenin-AP-conjugated antibody (1 : 1000, InnoGenex) at 4°C overnight. The slides were washed and used to perform color reaction using nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyphosphate p-toluidine salt (NBT/BCIP) (Boehringer Mannheim) as substrate. The sections were examined and images were obtained using an Olympus microscope.
Membrane-based focused microarrays are reliable and reproducible
To assess candidate genes that are expressed by NSCs , we chose rat E10.5 neural tubes as a source of multipotent NEPs, based on the previously published data (Kalyani et al. 1997; Mayer-Proschel et al. 1997; Cai et al. 2002). The morphology, expression of stem cell markers Sox2, and RT–PCR analysis of stem cell markers are shown in Fig. 1(a). As expected, the cells at this stage are apparently homogenous as indicated by nearly uniform staining of Sox2, a transcription factor expressed in NEPs, (Kalyani et al. 1997; Mayer-Proschel et al. 1997; Cai et al. 2002) in sections, and the continued presence of nestin, Sox2 and the absence (or very low level, if any) of lineage markers glial fibrillary acidic protein (GFAP), PLP/DM20, platelet-derived growth factor (PDGF) receptor and MAP2, as assessed by RT–PCR (Fig. 1a). Acutely dissociated NEPs that expressed these characteristics were used for subsequent analysis.
GEArray™ Q series cDNA expression array filters were used for the analysis of gene expression. These membrane filters contain gene fragments (about 500 bp) that have been sequence verified. The region of cDNA used is selected to minimize potential cross-reactivity with related genes and for overall similarity in hybridization conditions. In the membrane format, each gene is reproduced in quadruplicate and controls are printed across the base of the membrane, allowing a rapid assessment of the evenness of hybridization and straightforward quantitation and normalization on the same array (Fig. 1b). Two house-keeping genes, β-actin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (indicated in Fig. 2b), one negative control, two positive controls and blanks were included on all membranes. A total of 497 genes (control genes excluded) present on seven different membranes were assessed. Several genes were present on more than one membrane as an independent internal control.
Genes chosen were classified into four categories: 58 genes related to apoptosis; 103 genes involved in cell cycle events; 229 genes for chemokines, cytokines and their receptors; and 107 genes for ECM molecules. A detailed gene list is available at website (www.superarray.com). Figure 1(b) shows image profiles of two independent array results performed by using membranes from two different batches and two preparations of total RNA from rat E10.5 neural tubes. As expected, all positive and house-keeping genes show a clear hybridization signal, and blank and negative controls show the absence of any hybridization signal. Figure 1(c) is a summary of results for all seven membranes. Of 180 positives in experiment 1, only one gene was not detected in experiment 2, indicating that such an array analysis is a reproducible and consistent. In the following data analysis, results from two independent experiments using membranes from different batches and RNA prepared from different samples will be presented.
Gene expression profiles in NEPs
NEPs were harvested from E10.5 rat embryos and RNA was isolated for microarray analysis. Before analysis, each batch of RNA was tested for the presence of stem cell markers and the absence of lineage-specific markers as described. RNA samples in which differentiation was not seen were then reverse transcribed to prepare cDNA probes for hybridization to membranes as described. Table 1 summarizes results of all experiments. Of 497 genes tested, expression of 152 genes was detected, representing about 30% of all examined genes. In order to confirm the gene expression pattern, 41 genes were selected randomly and their expression was analyzed by RT–PCR. The majority of genes (40 of 41; 98%) detected by array analysis could also be confirmed by RT–PCR. A subset of the PCR results is shown in Fig. 2. One gene that showed expression on microarray analysis but not by RT–PCR was EGFR (Fig. 2d). The failure to detect EGFR by RT–PCR could not be attributed to primer design as we could readily detect EGFR at later stages of development (Fig. 2d). The absence of EGFR is consistent with a previous report (Cai et al. 2002) and with immunocytochemical analysis of EGFR expression at E14.5. (Note absence of EGFR expression in the ventricular zone NEPs; Fig. 2e). It is likely that expression detected in the microarray is due to cross-hybridization with other members of the erb receptor family.
Table 1. Selected gene expression profile of rat E10.5 neural tubes
No. of detected genes
Total RNA from rat E10.5 neural tubes was harvested, reverse transcribed using gene-specific primers and used to probe cDNA microarrays. Pooled results from two independent experiments are presented. *EGFR was detected on array analysis but not on RT–PCR.
As can be seen in Table 1, a majority of the apoptosis-related genes present on the array were not detected in E10.5-derived NSCs. Of the 58 genes tested, expression of only eight genes could be seen. About 40% (40 of 103) of genes involved in cell cycle regulation were present at detectable levels in NEPs. All three sets of cyclin–CDK complexes in phase G1, which are needed for progression of the cell through cell cycle (Olashaw and Pledger 2002), were present: the D cyclins (D1, D2, D3) and CDK4; cyclin E and CDK2; and cyclin A and CDK2. The inhibitors of CDKs, p19 and p27Kip1, were not detected at this stage. The small signaling G protein, RhoC, kinases Mkk4 and c-src, and transcription factor 14-3-3 e, were also detected. Of the 229 cytokines, chemokines and their receptors that were present on the microarrays, we detected expression of 57. These included members of the tumor growth factor (TGF) superfamily (bone morphogenetic proteins, BMPs), FGFs, stem cell factor (SCF) and several chemokines. Forty-seven of 107 classified as ECM and ECM-related were present in stem cells as well. These included members of the matrix metalloproteinase (MMP) family, cadherins, integrins and laminins. Overall, the analysis provides an overview of the state of the NSC population. The cells are actively dividing and their cell division is probably regulated by the cytokines and chemokines present in the cellular environment for which receptors can be detected on the cell surface. Apotosis probably plays a small role at this stage, and expression of specific cadherins, integrins and laminins provides an indication of optimal substrates for NSC growth. The analysis also reveals the presence of several molecules not previously thought to play a role at this stage of development including osteopontin, NOS3, Scya 5, 6 and 7, and angiopoietin.
Comparison of gene expression patterns between E10.5 and E14.5 neural tube-derived cells
At E14.5, neural tubes contain NRPs (∼60%), GRPs (∼30%), some post-mitotic neurons and small numbers of NEPs (< 10%) (Kalyani et al. 1997; Mayer-Proschel et al. 1997; Cai et al. 2002). No astrocytes as assessed by RT–PCR for GFAP (Fig. 3e) or immunocytochemistry (Liu et al. 2002) can be detected at this stage. Ingrowing blood vessels, however, can be readily detected by PECAM immunoreactivity (data not shown). Given the preponderance of progenitor cells at this stage, we assumed that genes expressed would primarily reflect the characteristics of neuronal and glial progenitors. Some genes would be distinct from those expressed at E10.5 and perhaps be specific for the differentiated cell type, which may be uniquely involved in the process of differentiation, migration, survival or biasing cell fate at this stage of development. We reasoned that other genes would be expressed at both stages of development, and might be involved in common functions of cell growth and proliferation. Comparing expression patterns at these two stages of development would allow us to distinguish between these two classes of genes. To test this hypothesis, we harvested E14.5 neural tube cells and E10.5 NCSs and processed samples for microarray analysis. Before hybridization, samples were tested by examining the expression of neural cell adhesion molecule (NCAM), nestin and GFAP, as described earlier, and then used for cDNA probe preparation. Microarray analysis was performed in parallel for the comparison to ensure consistency. The results from the comparison are summarized in Table 2.
Table 2. Comparison of gene expression profiles of the rat E10.5 neural tubes with those of E14.5
No. of genes
detected in NSCs but not in progenitor cells
detected in both cell populations
detected in progenitors but not in the NSCs
Microarrays were probed with cDNA from either E10.5 neural tube stem cells or E15.5 neural tube progenitor cells in parallel.
Expression of 209 of 489 genes was detected when arrays were probed with cDNA from E14.5 neural tube cells. Among these 209 genes, 141 genes were originally identified as being expressed by E10.5 cells (see Table 1). These results indicate that most genes (73%) expressed at E10.5 continue to be expressed at E14.5 (coded black, Table 2).
Expression of 11 genes appeared restricted to E10.5 neural tube cells (Table 2). These 11 genes are bruce, Cdk2L, CKS1p9 (CDC28 protein kinase 1), Cyclin E, P55cdd (CDC20), MCM3 (P1 protein), MCM5 (CDC46), TRAF4, ID1 (DNA-binding protein inhibitor 1), MMP8 and MMP9. This result is consistent with previous observations that few stem cell-specific genes exist (Cai et al. 2002).
Several genes (68 of 209; 33%) were identified that were not expressed at E10.5 but were expressed at relatively high levels at E14.5 (Table 2). These included cytokines, chemokines and their receptors, as well as cell cycle regulatory genes.
Specificity of detection was confirmed by randomly selecting 82 (40%) of the genes for assessment by RT–PCR. The results obtained by RT–PCR indicate a 96% (79 of 82) concordance with the results from the array. A subset of these results (47 of 82) are shown in Figs 3(a)–(d). Additional confirmation was obtained by immunoblot for seven genes and immunocytochemistry for five of the genes. We detected expression for all seven genes identified by array analysis that were selected for examination (data not shown) and by immunocytochemistry (Fig. 4).
The immunocytochemistry and imunoblot data were in good concordance with the array analysis. PTEN, a tumor suppressor gene that has been shown negatively regulate neural stem/progenitor cell proliferation (Groszer et al. 2001) and cell size (Kwon et al. 2001) is a good example. Its expression at both stages was confirmed by RT–PCR although we noted increased levels at later stages of development (Figs 3b and 4c,c′). Phosphoinositide 3-kinase (PI3K), a partner of PTEN, was shown to have a similar expression pattern in our array analysis and by RT–PCR (Table 2 and Fig. 3b). PTEN protein expression was confirmed by immunoblot analysis as well. We detected an expected 60-kDa band of PTEN protein in cellular lysates from E14.5 cells (data not shown). Immunocytochemistry using the same antibody indicated that PTEN protein was distributed in both cell body and processes (Fig. 4c′).
Overall our results indicated that using gene-specific primers for reverse transcription and comparing expression across two stages of development allowed us to identify candidate stage-specific genes.
Segregation of expression to neuronal and glial subpopulations
The E14.5 neural tube contains at least two progenitor cell populations that can be distinguished from neuroepithelial stem cells and each other by the expression of polysialated NCAM (E-NCAM) and the expression of an epitope recognized by the A2B5 antibody (Kalyani and Rao 1998; Kalyani et al. 1998). Given the large number (68 of 209; 33%) of genes that appeared to be present at E14.5, but not at E10.5, we considered the possibility that a subset of these genes may be expressed on specific cell populations. To test this hypothesis we chose 24 genes and examined their expression in E-NCAM- or A2B5-positive cell populations (Fig. 5). Cells were harvested by live cell labeling and enriched by fluorescence-activated cell sorting. The cell purity for each subpopulation was greater than 90% in all experiments (Fig. 5a). Nine genes detected on both E-NCAM- and A2B5-positive populations include stromal cell derived factor 2 (SDF-2), MT1-matrix metalloproteinase 14 (MMP), FGF9, PDGFa, TGFBR1, Rb, cystatin-C, H-ras and caspase 9 (A in Fig. 5b). Ten candidate molecules appeared to be enriched in neuronal populations and five in glial populations (Fig. 5b). Genes that appeared to be differentially expressed were further tested by semiquantitative PCR, immunocytochemistry or in situ hybridization. (Figs 5–7). For semiquantitative PCR, cDNA derived from each population was prepared, adjusted to similar levels based on GAPDH PCR products, and then amplified at different cycle numbers as described in methods. Results from amplification to 27, 32 and 37 cycles are presented in Fig. 5(c) for seven of the genes. The results largely confirmed the array and initial PCR analysis. FGF11, CDK5 and α-catenin appeared to be enriched or selectively expressed in E-NCAM-positive sorted populations whereas chemokin CXC motif receptor (CXCR)-4, CD44 and Rho-C appeared to be enriched in A2B5-positive cells.
The distribution of TnC and CD44 in GRP-enriched cells was further confirmed by immunocytochemistry and in situ hybridization. As can be seen in Fig. 6, TnC expression could be detected by immunoblot analysis, confirming protein expression (Fig. 6i). In situ hybridization using a probe to TnC showed a characteristic distribution that appeared identical to the distribution of glial progenitors (Fig. 6e; Liu et al. 2002). In addition double-labeling experiments using cultured neural progenitor cells showed that TnC expression co-localized with A2B5 immunoreactive glial cells but not with E-NCAM immunoreactive neuronal precursors (Figs 6a–d). CD44, a second marker that appeared glial-specific by array analysis, also showed specificity by immunocytochemistry. CD44 expression did not overlap with the expression of E-NCAM, a neuronal precursor cell marker, in sections or in dissociated cell culture (Fig. 6h; also see Liu et al. 2002 for a detailed description of CD44 expression during development).
Expression of two genes, p27Kip1 and α-catenin, that appeared to be expressed specifically by neuronal populations were further examined (Fig. 7). Immunoblot analysis confirmed expression in neural tube progenitors, confirming the array and RT–PCR results. Staining of sections with p27Kip1 showed an absence of expression in ventricular zone stem cells (Fig. 7e) in E14.5 section and co-expression with E-NCAM-immunoreactive cells in culture and a largely non-overlapping expression with A2B5-immunoreactive cells. We noted, however, that once A2B5 cells were allowed to mature (growth factor withdrawal) then p27Kip1 was expressed by most cells (data not shown), suggesting that the specificity assessed by array and immunochemistry related to the stage of maturation of the cells.
α-Catenin expression was largely localized to E-NCAM-positive cells (Figs 7a and b), although occasional A2B5-positive cells showed co-expression (Figs 7c and d). However, most cells did not express α-catenin and in no experiment was co-expression significant (less than 5% in two independent experiments).
Overall our results suggest that cell type-specific markers that will distinguish between neuronal and glial progenitors exist, and that a microarray analysis using a candidate gene approach to compare gene expression in purified populations is one method of identifying such genes.
We have taken advantage of the temporal segregation of stem and progenitor cell development to compare the expression of selective cytokines, chemokines, receptors, ECM molecules and cell cycle regulators. Using focused microarrays with gene-specific primers for cDNA synthesis, we have identified genes that are expressed in neuroepithelial stem cells and progenitor cells. Our results show that the majority of genes detected in NEPs are also present in progenitor populations but NEPs express fewer genes than E-NCAM- and A2B5-positive progenitors. Purifying populations of neuronal and glial progenitors, and comparing candidate gene expression between, them reveals additional segregation of gene expression to specific progenitor populations.
Our rationale for choosing this subset of genes was based on our previous results and published data that suggest that extrinsic cues direct the differentiation of initially naive cells (Sommer and Rao 2002). Expression of these cues is spatially and temporally regulated, and includes cytokines, chemokines and ECM molecules that interact with each other in specific ways (Sommer and Rao 2002). We chose a focused array rather than an Affymetrix type large-scale array based on the limited amount of material available, and previous observations on the relatively low levels of expression of these genes in stem cells reducing the overall sensitivity of the array approach (Rao, unpublished observations). Further, analysis of several candidate genes that are expressed early in development showed that they were absent from the many of the commercial arrays available. The 15K array, for example, (Tanaka et al. 2000) did not contain approximately 30% of the candidate genes of interest to us (Luo, unpublished observations). The value of using a focused approach is seen in comparing this information with the two published reports of gene expression using a large-scale arrays (Geschwind et al. 2001; Terskikh et al. 2001). Although these latter studies identified several novel candidates that were specifically expressed in stem cells, other genes that are known to be expressed on stem cells and were present on the arrays were not detected, as levels were too low. Moreover, many of the genes that we have identified as being expressed on NSCs were not detected in these experiments. Thus the focused microarray approach complements the broad search using large-scale arrays and may reveal unexpected roles for known molecules, as well as detect expression of molecules expressed at low levels. The focused microarray approach may also be the better choice for quality control experiments in which samples need to be compared for reproducibility or for studying the effects of specific gene manipulations. A focused array that is cheap and reproducible also allows multiple direct comparisons of samples from different laboratories, which has been difficult given the cost of large-scale array experiments.
Our analysis of gene expression, however, illustrates not only the advantages but also the caveats of depending solely on the array analysis. For example, EGFR expression was detected in multiple experiments examining cytokine receptor expression in NSCs. Our previous results had indicated, however, that EGFR was not present and could not be detected by RT–PCR or immunocytochemistry (Cai et al. 2002). A repeat analysis by PCR and immunocytochemistry (Figs 1a and 2d,e) and in situ hybridization (data not shown) confirmed the absence of EGFR expression, indicating the importance of independent validation. Cross-reactivity of a related EGFR/erb family member is a likely explanation for this apparent contradiction as additional erb receptors E2, 3, 4 are present (Rao, unpublished data). Another example is TnC expression. The array (Table 2) and immunocytochemistry (Fig. 4) data consistently show that TnC is expressed both at E10.5 and E14.5. In addition, TnC was detected by immunocytochemistry as early as E10 in the mouse CNS (Kawano et al. 1995). However, our RT–PCR analysis (Fig. 3d) only detected its expression at E14.5 stage. This may result from amplifying different isoforms of TnC in our PCR conditions as there are five known members of the tenascin family (Joester and Faissner 2001). Analysis of approximately 30% of the total genes present on the blot by RT–PCR revealed a good concordance with blot results of expressed genes. However, we are cautious about interpreting the absence of detectable expression as evidence that a particular gene is not expressed until it is independently verified. RT–PCR is often more sensitive and detects expression at levels undetectable by the array approach. We suggest that one should refrain from interpreting absence of expression unless independent confirmation is available in the literature, or by in situ hybridization or immunocytochemistry.
The snapshot of NSCs revealed similarities and differences between these cells and HSCs. For example, we detect a subset of genes in NEPs that were reported to be expressed by HSCs. Some examples include KitL and cyclin D1. KitL together with KIT (steel factor receptor) and Flk2 tyrosine kinase receptor is crucial for HSC survival, renewal and differentiation (Mackarehtschian et al. 1995). Cyclin D1 was also detected on HSCs and mouse neurospheres but not in terminally differentiated neural cells (Terskikh et al. 2001). The expression of these common genes may reflect underlying common regulatory pathways for stem cell proliferation and self-renewal. However, we did not detect expression of CD70 and its receptor CD27, a marker for a subpopulation in bone marrow stem cells (Phillips et al. 2000; Wiesmann et al. 2000) in NSCs.
NSCs appear to be rapidly proliferating and express specific subsets of cell cycle regulators. Few if any G2–M transition genes, such as cyclin B and CDK1, were detected suggesting that the transition through this stage is short, consistent with previously published data (see review by (Yoshikawa 2000)). At early stages, negative regulators of proliferation (e.g. p27Kip1, p21cip, p18 and p19) were undetectable whereas specific subsets of cdks were expressed. PTEN, a tumor suppressor, was present in both stem and progenitor cells but was up-regulated as cells matured, consistent with published results showing a dramatic effect on proliferation and differentiation after loss of PTEN function (Lachyankar et al. 2000; Groszer et al. 2001).
Autocrine and paracrine factors have been thought to be important in early stem cell development as growth at high density obviates the need for any exogenous growth factor (Mujtaba et al. 1999). Several FGFs were detected, as were BMPs, both of which are known to have specific effect on NEP differentiation (Mabie et al. 1999; Mehler et al. 2000). Of interest was the absence of expression of FGR receptor (FGFR)3 at early stem cell stages and apparent expression at E14.5. FGFR3 is thought to negatively regulate cell proliferation and its late expression correlates with the increase in cell cycle time at this stage (Takahashi et al. 1999). Further experiments with FGFR3 mutant mice are in progress to examine cell cycle and differentiation effects. The expression of FGFR1 is consistent with its important role in transducing a proliferation signal for stem cells (Vaccarino et al. 1999a; Vaccarino et al. 1999b). The expression of FGFR4 is consistent with detection of its expression specifically at the stem cell stage in previous studies (Kalyani et al. 1999; Cai et al. 2002). It is important to note that several FGFs that belong to a specific subfamily that is thought to activate the FGFR4 receptor (Ornitz and Itoh 2001) are present in NEPs (FGF17, 21, 22). Some members of this subfamily may only activate this subtype (Xie et al. 1999). Of importance was our observation that leukaemia inhibitory factor (LIF) is not expressed at this stage in rodents and therefore is unlikely to be involved in maintenance of telomerase levels or the stem cell state. This is in contrast to data from human stem cells which indicate that LIF and receptor for leukaemia inhibitory factor (LIFR) may be critical in maintaining the proliferation of HSCs. PDGF and PDGF receptor were not expressed at detectable levels in NSC populations, suggesting that this important cytokine may act at later stages of neuronal and glial development (Pringle et al. 1991; Williams et al. 1997; Park et al. 1999).
Programmed cell death is critical for normal nervous system development and is regulated by Bcl2 and caspase family members. In the nervous system it has been suggested that apoptosis is regulated differentially at different stages in development with a transition from a caspase-9- to Bax- and Bcl-X (L)-mediated neuronal apoptosis (Kuan et al. 2000). Our results relating to the expression of apoptosis-related genes are consistent with these observations. Of the 58 apoptosis-related genes, including families of caspase, inhibitor of apoptosis protein (IAP), bcl, cida domain and tumour necrosis factor-linked death domain, only eight were found to be expressed in E10.5-derived NSCs. In particular fas and fas ligand were not present. Absence of Fas signaling is consistent with previous reports that Fas activation does not play an important role in regulating NCS number. It is noted that Bruce, a member of the IAP family, was present exclusively at E10.5 but not in neural progenitors at the E14.5 stage. Bcl2, a survival factor, was detected in E14.5 progenitors but not in E10.5 NEPs. The absence of bcl2 suggests that it is not required for survival of NSCs and is consistent with reports of bcl2 knockouts, which show a massive loss of neurons but not of stem cells (Motoyama et al. 1995; Roth et al. 2000). Caspase 9 was present at both stages and suggests that apoptotic death may require caspase 9 activation. Loss of caspases leads to a massive increase in stem cell numbers indicating that caspases may play an important developmental role (D'Sa-Eipper and Roth 2000; Roth et al. 2000). Overall our results suggest that selective activation of apoptotic pathway genes regulates cell number, and that bcl2 and fas do not play an important role at this stage.
Comparing expression of candidate genes in stem cells and progenitor cells revealed several families of genes that were expressed in all populations and some that were specific to a particular cell type. A detailed discussion of all genes is impossible, but some candidate genes bear mention (summarized in Table 3).
Table 3. Examples of specific genes in NSCs
A partial summary of candidate genes that are expressed by either stem cells, neuronal restricted progenitors or glial restricted progenitors was prepared from the larger data set. E, expressed; –, not detected in the array.
P55cdc and tumor necrosis factor receptor associated factor 4 (TRAF4) are two genes present at high levels in multipotent stem cells but virtually undetectable in neuronal and glial progenitors, suggesting a specific role for these molecules in early neural tube development. Indeed, a recent report by Regnier et al. (2002) confirms a high level of expression in the early neural tube and shows that loss of TRAF4 results in defects in neurulation. P55cdc (cdc20), a WD-repeat protein, directly binds and activates anaphase-promoting complex/cyclosome, a multicomponent, ubiquitin-protein ligase. Anaphase-promoting complex/cyclosome is required for the initiation of anaphase and exit from mitosis (Ohtoshi et al. 2000). In addition, P55cdc binds to, and is phosphorylated by, cyclin A, which is also expressed at high levels in NSCs, suggesting a possible pathway of cell cycle regulation in NSCs.
Exit from the cell cycle probably requires the accumulation and phosphorylation of p27Kip1. At this stage in development, it appears as if neuronal precursors have begun to exit the cell cycle whereas glial progenitors and stem cells are still actively dividing. Loss of p27Kip1 leads to an increase in proliferation and an increase in brain size, although no significant change in ventricular zone size, suggesting that p27Kip1 regulates progenitor but not stem cell exit from the cell cycle (Zindy et al. 1999; Levine et al. 2000).
Multiple roles of CDK5 in neuronal migration and synaptic transmission have been suggested (Dhavan and Tsai 2001; Smith and Tsai 2002). Our results suggest that even early in development its expression is enriched in neuronal precursors before the onset of neurotransmission or synaptogenesis, raising the possibility that it may regulate neuronal migration. Such a role with an interaction with the reeelin/dab1 pathway has been postulated recently (Ohshima et al. 2002).
Examination of chemokine expression revealed the presence of two receptor ligand pairs CXCR-4/SDF-1 and fractalkine/CX3CR-1. Both chemokines have been shown to be involved in migration, and loss of CXCR-4 in particular alters migration of oligodendrocyte precursors (Ma et al. 1998; Zou et al. 1998). The role of CX3CR-1–fractalkine in the CNS has not been well characterized but, based on its expression, we suggest that it may be involved in neural precursor cell migration, or survival of neurons and microglia as has been postulated previously (Boehme et al. 2000). Several cytokines known to be involved in macrophage migration are expressed at this stage. However, we believe that the expression we observed is unlikely to be due to macrophages. The available data suggest that macrophage infiltration take place somewhat later than the time we isolate cells and Mac-1 staining does not reveal microglia (Chamak et al. 1995; El-Nefiawy et al. 2002). The roles of cytokines in neural development need to be further examined.
Two extracellular molecules appear to be specifically expressed by glial progenitors but not by stem cells or neuronal cells at this early stage. These are TnC and CD44. The specificity of expression was confirmed by RT–PCR and immunocytochemistry (Figs 5 and 6). TnC KO's show an abnormal migration of oligodendrocytes (Kiernan et al. 1999; Garcion et al. 2001). Expression of CD44 was shown to be limited to astrocyte precursors and astrocytes during development (Alfei et al. 1999; Liu et al. 2002; Mayer-Proschel et al. 2002), and our recent results have suggested that CD44 expression may be used to identify an astrocyte precursor cell (Liu et al. 2002 and unpublished results). Identification of these two markers increases the number of glial-specific markers identified so far, and provides further evidence for a potential lineage relationship between astrocyte precursors and tripotential glial precursor cells.
Overall our results extend the number of genes known to be expressed in stem cells and down-regulated in precursor cells, as well as identify additional progenitor-specific genes. These results validate the focused microarray approach to analyzing candidate gene expression and suggest that genes identified in such a screen can be used for quality control to profile the state of stem and progenitor cells, and reveal the extent of differentiation in samples from different laboratories. Future experiments will be directed at constructing stem cell-specific arrays, based on the results of these and similar experiments, and further characterizing cell-type specific genes functionally and as potential stage-specific markers.
Yongquan Luo was supported by the NIA and the STAR grant. Haipeng Xue, Mahendra Rao, Francis Crest and Robert Wersto were supported by the NIA. Jingli Cai and Ying Liu were supported by the NIA and a graduate student fellowship from the Department of Neurobiology and Anatomy. MSR acknowledges the contributions of Dr S. Rao that made undertaking this project possible.