| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
STEM CELL GENETICS AND GENOMICS |
a Developmental Genomics and Aging Section, Laboratory of Genetics, National Institute on Aging, National Institutes of Health, Baltimore, Maryland, USA;
b Institute for Stem Cell Research, Ospedale San Raffaele, Milan, Italy
Key Words. Embryonic stem cells • Principal component analysis • Neural commitment • Neural differentiation • Microarray
Correspondence: Minoru S.H. Ko, M.D., Ph.D., Developmental Genomics and Aging Section, Laboratory of Genetics, National Institute on Aging, National Institutes of Health, Baltimore, Maryland 21224, USA. Telephone: 410-558-8359; Fax: 410-558-8331; e-mail: kom{at}mail.nih.gov
Received July 23, 2005;
accepted for publication December 9, 2005.
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResults and DiscussionDisclosuresReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionDisclosuresReferences |
|---|
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionDisclosuresReferences |
|---|
Undifferentiated 129.3 (129/SvEv) ESCs were cultivated on gelatin-coated 100-mm dishes in DMEM supplemented with 15% FBS, 2 mM glutamine, 1 mM sodium pyruvate, 0.1 mM nonessential amino acids, 0.1 mM ß-mercaptoethanol, and 103 units/ml of leukemia inhibitory factor (LIF). For chemical neural differentiation of ESCs, EBs were grown for 4 days without LIF at 5 x 106 cells per 100-mm bacteriological dish and then treated for 4 days with 1 µM RA [12, 13]. EBs were collected at 0 hours (4–), 24 hours (4–/1+), 49–51 hours (4–/2+), 71–72 hours (4–/3+), and 95–96 hours (4–/4+) after RA treatment. In the text and figures, these cells were referred to as EB(4–), EBs cultured for 4 days without RA; EB(4–/1+), EB(4–/2+), EB(4–/3+) or EB(4–/4+), EB(4–) cultured with RA for 1, 2, 3 or 4 days, respectively. Total RNAs were extracted from these EBs as well as undifferentiated ESCs (referred to as ES_129a in Figs. 1
, 2, 3 and supplemental online Figs. 3, 4, 6).
|
|
|
|
) [14]. Briefly, undifferentiated 129 ESCs (2.5 x 106 cells) were cultured in gelatin-coated T25 flasks in ES standard medium with LIF for 24 hours. ESCs were plated at 1 x 104 cells per cm2 onto gelatin-coated 60 mm-dishes and were cultured as a monolayer in N2B27, which is a 1:1 mixture of DMEM/F-12 supplement with modified N2 and neurobasal medium supplement with B27 without vitamin A (Gibco, Grand Island, NY, http://www.invitrogen.com) [14]. ESCs were harvested every 24 hours for total RNA extraction. Undifferentiated ESCs cultured as described above are referred to as ES_129b or N2 d0 in Figures 1, 2, 3 and supplemental online Figures 3, 4, 6. Time-course data are referred to as N2 d1, N2 d2, N2 d3, N2 d4, N2 d5, and N2 d6; ESCs were cultured in the N2B27 medium as monolayer culture for 1, 2, 3, 4, 5, or 6 days, respectively. RNAs from undifferentiated R1 ESCs and trophoblast stem (TS) cells were kindly provided by Drs. Tilo Kunath and Janet Rossant [7]. Adult NS cells and neural differentiated cells derived from them were prepared as described before [15]. In brief, cells isolated from the forebrain periventricular zone were cultured to form neurospheres. The neurospheres were dissociated into single cells and passaged to a new culture dish. After 1 day in culture, cells that divided and formed doublets were harvested (NS1). After 5 days in culture, cells that formed neurospheres were harvested (NS5). By culturing in differentiation-promoting conditions, NS5 cells were turned into differentiated neuronal cells (DCs).
Preparation of Fluorescence-Labeled Target
We carried out gene expression profiles of ESCs, TS cells, NS cells, and terminally differentiated neural cells derived from NS cells and embryonic day 12.5 (E12.5) placenta. All samples were labeled with Cy3-CTP and were compared via a Cy5-CTP–labeled Universal Reference (Stratagene, La Jolla, CA, http://www.stratagene.com).
Two to 6 µg of total RNA or universal mouse reference RNA (Stratagene) was labeled with Cy3 dye or Cy5 dye, respectively, using a fluorescent linear amplification kit or a low RNA input fluorescent linear amplification kit (Agilent Technologies, Palo Alto, CA, http://www.agilent.com) [16]. Besides using undifferentiated cells and treated cells described above, we used cRNA from E12.5 placenta [16] as targets for microar-ray analysis. Three or two biological samples were used for hybridizations (supplemental online Fig. 2).
Hybridization on NIA Mouse 22K Microarray v2.0
Each cRNA target from stem cells or differentiated cells was combined with a single universal reference cRNA target into a hybridization reaction on a NIA mouse 22K microarray v2.0 slide (Dev2 [Agilent Technologies]), which was modified from the NIA mouse 22K microarray v1.0 [16]. The Dev2 chip has a larger number of genes than the Dev1 chip. The microarrays are constructed with more than 20,000 gene features derived from the NIA gene index [17]. We will use the term "gene" instead of gene feature in descriptions of microarray data. A complete list of the microarray’s annotated gene content can be found at the NIA mouse cDNA project Web site (http://lgsun.grc.nia.nih.gov/cDNA/cDNA.html).
Microarray Data Analysis
The intensity of each gene feature per array was extracted from scanned microarray images using Feature Extraction 5.1.1 software (Agilent Technologies) as described previously [16]. Text output was processed using an application developed in-house to perform analysis of variance (ANOVA) analysis [16, 18] (http://lgsun.grc.nia.nih.gov/ANOVA/; see also http://lgsun.grc.nia.nih.gov/ANOVA/help.html for the details of ANOVA). All samples labeled with Cy3 dye were compared via a Cy5-labeled universal mouse reference. Statistical significance was determined using the false discovery rate (FDR = 10%) method [18, 19]. Pairwise mean comparisons were done using t statistics and FDR = 10%. Further data processing, including scatter plots, hierarchical clustering, and principal component analysis (PCA), was also performed through NIA microarray analysis tool. The PCA is a multivariate analysis technique that finds major patterns in data variability. In mathematical terms, it finds eigenvalues and corresponding eigenvectors (i.e., principal components [PCs]). The most important are the first few PCs that explain the most about observed variance; the rest are mostly random fluctuations. Thus, by plotting data in the first two or three PC coordinates, we can reduce dimensionality of the data without losing much information ([17, 20, 21]; see http://lgsun.grc.nia.nih.gov/ANOVA/help.html for the details of PCA). TIGR Multiple Experiment Viewer [22] was used for k-means clustering and template/pattern matching. The template/pattern matching algorithm can identify a group of genes, the expression patterns of which match to a selected template gene ([23]; see http://lgsun.grc.nia.nih.gov/ANOVA/help.html for the details of template/pattern matching). GenMAPP and MAPP-Finder [24, 25] were used for a functional annotation analysis with Gene Ontology (GO) terms [26]. To be conservative on the characterization of cells by GO terms, we used a statistical test (FDR) that was more stringent than a statistical test used by the original paper [25]. As the gene database for GenMAPP and MAPPFinder, Mm_Std_20040411 was used. With extensive validation by quantitative real-time polymerase chain reaction, we have previously shown that our microarray results with a stringent statistical test have been confirmed for essentially all genes tested [16].
|
RESULTS AND DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionDisclosuresReferences |
|---|
Although both ES-N2B27 and EB-RA started from undifferentiated ESCs and became neural progenitors, we found a marked difference in gene expression patterns between these two induction systems (Fig. 1). Consistent with known involvement of RA in caudalization of brain during development, many homeobox (Hox) genes, which are known to be involved in hindbrain and spinal cord development, were more highly expressed in EB-RA than in ES-N2B27 cells (Fig. 1A). Similarly, notable differences between EB-RA and ES-N2B27 cells were observed in a number of signaling pathways, including fibro-blast growth factor (FGF)- and Wnt-signaling (supplemental online Figs. 3, 4), suggesting that the different types of neural cells were induced.
The global expression profile also suggests that ES-N2B27 cells are a less heterogeneous differentiation system than EB-RA cells. First, mesoderm markers, such as T/Brachyury and Mesp1, were upregulated at EB-RA (EB 4-) but immediately downregulated after RA treatment, whereas only T/Brachyury was only slightly upregulated (1.4-fold change compared with day 0) at day 3 and day 4 in the ES-N2B27 induction system (Fig. 1B). Primitive endoderm markers, such as Gata4 and Gata6, were upregulated in EB-RA cells during RA treatment but showed no significant change in ES-N2B27 (Fig. 1B). Furthermore, PCA, which reduces data with high dimensionality into a limited number of dimensions, or PCs [17, 20, 21], clearly showed that the EB-RA system contained heterogeneous cell populations as indicated by the inclusion of GO terms such as "blood vessel development," "muscle development," and "neurogenesis" (Fig. 1C). In contrast, the ES-N2B27 system contained mainly neural development-related terms such as "central nervous system development" and "neurogenesis" (Fig. 1D). For simplicity, we focused on N2B27-treated ESCs in the following analysis, but analysis of the EB-RA system can be found in supplemental online Figs. 3 and 4.
Three Phases of Neural Differentiation in ES-N2B27 Monolayer Cultures
Several lines of evidence indicate that during the 6-day neural induction period by N2B27 medium there are two major transitions of global gene expression patterns, which split the induction period into three phases. These transitions seem to roughly correlate to in vivo cell differentiation steps: from inner cell mass (ICM) to primitive ectoderm (PrEc) and from PrEc to neuroectoderm.
First, pairwise comparison of gene expression patterns between neighboring days of culture showed that, after initial dramatic changes, expression pattern changes increased and peaked again between day 3 and day 4 (Fig. 2A). This is most likely not caused by medium changes, because medium changes were performed at day 2 and day 4. This notion is further supported by the fact that hierarchical clustering analysis also separated the induction period into three phases (Fig. 2B).
Second, in ES-N2B27, the expression level of Oct3/4 decreased precipitously after day 3, whereas the gene expression of Pou3f2/Brn2, which is required for establishing the neural cell lineage [27], increased after day 3 (Fig. 2C). This is consistent with a recent report showing a correlation between Oct3/4 downregulation and the onset of neural differentiation [27]. The expression patterns of other neural-related genes, such as Ascl1/Mash1, Zic1, Hoxd4, and Pax6, which were identified by a template-matching algorithm [22, 23] with p < .001 and more than 0.5 slope value (supplemental online Table 1), were also dramatically upregulated after day 3 (Fig. 2C). These data indicate that neural cell lineage may be established after day 3.
Third, Fgf5, a marker for PrEc, was upregulated at day 1, peaked at day 2, and then downregulated at day 4 in ES-N2B27 cells (Fig. 2D). Espl1/PRCE, which was recently identified as a marker for early PrEc [28], preceded the Fgf5 expression (Fig. 2D). The expression patterns of these genes indicate that PrEc-like cells existed during this period, suggesting that ESCs in N2B27 monolayer culture differentiate toward a neural progenitor–like state via PrEc-like cells, just as with in vivo early development.
Global Expression Profiles of Stem Cells
Our original intent was to obtain compendium gene expression profiles of a variety of stem cells. We therefore first carried out gene expression profiles of adult NS cells (NS1 and NS5) [15], terminally differentiated neural (DC) cells derived from NS cells, two EC cell lines (F9 and P19), RA-treated P19 cells (P19-RA; a well known neural differentiation system), TS cells [29], and E12.5 placenta. Initial PCA result using these cells revealed that a first PC, PC1, appeared to correlate to neural differentiation from ESCs, whereas PC2 appeared to correlate to differentiation into placenta (Fig. 3A). To test this idea further, we introduced ES-N2B27 data into these new data sets.
By combining ES-N2B27 time-course data with these data sets, we first identified 16,908 genes for their differential expression in these cells and then analyzed these data by PCA. Of 16,908 genes, PC1 represented 7,012 genes (41.5%), PC2, 2,986 (17.7%), and PC3, 372 genes (2.2%). As expected, ES-N2B27 cells were plotted along PC1 and were progressively shifted toward adult neural progenitors from undifferentiated ESCs (Fig. 3B). This indeed supports the notion that the PC1 axis, which accounts for 3,889 genes (23.0%) out of 16,908 significant genes in the positive direction, reflects the neural commitment and differentiation from ESCs and a specific direction from ES to neural fate. In fact, many known genes involved in neurogenesis, including Ascl1/Mash1, Zic1, Emx2, Hes5, and Pax6, were plotted in the positive direction of PC1. Therefore, this PC1 seems to capture the trend that the expression of these 3,890 genes increases during neural differentiation of ESCs (supplemental online Table 2). Furthermore, the positions of cell types in the PC1 axis seem to indicate the degree of neural commitment/differentiation, as represented by the gradual shift of cell positions in ES-N2B27 toward NS1, NS5, and DC (Fig. 3B).
A Path to Neural Fate
Figure 4 summarizes neural cell differentiation along the PC1 axis in the PCA (i.e., a developmental path to neural fate from ESCs). We identified at least five distinctive cell types: undifferentiated ESCs, PrEc-like cells, ES-derived neural progenitor cells, adult brain-derived NS cells, and neural cells differentiated from NS. These cell types seem to mimic in vivo commitment and differentiation of neural cells by representing ICM, PrEc, neuroectoderm, neural progenitors, and neurons/glia. Of course, the validity of such a notion needs to be tested by comparing the expression profiles of in vivo neural tissues microdissected from mouse embryos at various stages with the current microarray data sets. GO annotations of these genes, grouped by k-means clustering analysis (supplemental online Fig. 5), depicted specific cellular processes, which provide useful insights into the function of each cell type (Fig. 4; supplemental online Tables 3 and 4). For example, expression of MAPKKK cascade and helicase genes decreased during ESC differentiation, suggesting that these activities are associated with the maintenance of undifferentiated ESCs. However, different sets of genes that belong to the MAPKKK cascade were upregulated in ES-derived neural progenitors and adult NS cells. High expression of MAPKKK cascade and other interesting signaling pathway genes seems to be required for neural progenitors derived from ESCs. Among the upregulated genes in ES-derived neural progenitors (Fig. 2C; supplemental online Table 1) were well known genes such as Pax6 and Zic1, but most genes were uncharacterized. These genes can provide candidates for specific markers of these cell types, such as neural progenitors.
The localization of expression profiles along one of the PCs defined by the PCA of global gene expression patterns indicates that the status of cells (i.e., their level of commitment and differentiation) can be defined by a scale on PC axis [17]. In the current global gene expression analysis, the PC1 axis represents the neural path, which was scaled by the average expression levels of approximately 4,000 genes (Fig. 3B). In other words, by monitoring and calculating the average expression levels of these approximately 4,000 genes, cell expression profiles can be plotted along the neural differentiation path, and the relative locations of cell types provides information about their levels of neural commitment/differentiation (Fig. 3B). This notion provides the following novel insights into the cells’ potential for differentiation.
First, it became evident that undifferentiated P19 EC cells are more committed to differentiation than are F9 EC cells, because they were positioned in the middle of the PC1 axis (Fig. 3; this can be seen more clearly in the PCA of supplemental online Fig. 6). The position in PC1 axis indicates that undifferentiated P19 cells are already at the stage of PrEc, which is further supported by the higher expression level of Fgf5—PrEc marker—in P19 cells than in F9 EC cells or in RA-treated P19 cells (supplemental online Fig. 3). This suggests that, although P19 cells retain the ability to differentiate into other cell types such as muscle [30], they more easily differentiate into neuronal cells than ESCs or F9 cells. In fact, overexpression of some neural-related genes, such as Sox1, Sox6, NeuroD2, and Mash1, in P19 cells is sufficient to differentiate toward neurons without RA [31–33]. In contrast, the position of F9 cells in PC1 axis indicates that these cells might have a developmental potential greater than P19 cells although they are often referred to as a nullipotent cell line [34]. Indeed, the expression levels of pluripotency-related genes such as Oct3/4 and Nanog in F9 cells were comparable with those in ESCs [35].
Second, it also became clear that neural cells differentiated from ESCs do not reach as high a level of commitment/differentiation as do NS cells derived from adult brain (Fig. 3). In other words, neural progenitors obtained from ESCs and EC cells, at their end point of differentiation, were positioned on the way to neural progenitors from adult brain. This result seems to reflect different cell status between embryonic and adult cells. This may also be related to the fact that adult neural progenitors can mainly differentiate into astrocytes in vitro (~72% as glial fibrillary acidic protein–positive cells) [36], whereas ES-derived neural progenitors in ES-N2B27 differentiate into neurons (~60% as Tau-positive cells) [14]. This obviously has important implications for the therapeutic use of ES-derived neural cells. It remains to be seen whether this reflects a difference in developmental potential between these cells. Nonetheless, the possibility of having a scale to measure the commitment/differentiation status provides a critical tool for guiding ESC differentiation into specific cell types. Further analysis, including differentiation systems into other cell types, should be able to define other paths as well as to refine the neural path.
|
ACKNOWLEDGMENTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionDisclosuresReferences |
|---|
|
DISCLOSURES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionDisclosuresReferences |
|---|
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionDisclosuresReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  Z. Yan, Z. Wang, L. Sharova, A. A. Sharov, C. Ling, Y. Piao, K. Aiba, R. Matoba, W. Wang, and M. S. H. Ko BAF250B-Associated SWI/SNF Chromatin-Remodeling Complex Is Required to Maintain Undifferentiated Mouse Embryonic Stem Cells Stem Cells, May 1, 2008; 26(5): 1155 - 1165. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M.-S. Tsai, S.-M. Hwang, K.-D. Chen, Y.-S. Lee, L.-W. Hsu, Y.-J. Chang, C.-N. Wang, H.-H. Peng, Y.-L. Chang, A.-S. Chao, et al. Functional Network Analysis of the Transcriptomes of Mesenchymal Stem Cells Derived from Amniotic Fluid, Amniotic Membrane, Cord Blood, and Bone Marrow Stem Cells, October 1, 2007; 25(10): 2511 - 2523. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| STEM CELLS | THE ONCOLOGIST | CME | ALPHAMED PRESS JOURNALS |
