STEM CELLS AND DEVELOPMENT 15:232–244 (2006) © Mary Ann Liebert, Inc.
Original Research Report Massively Parallel Signature Sequencing Profiling of Fetal Human Neural Precursor Cells JINGLI CAI,1,6 SOOJUNG SHIN,1,6 LYNDA WRIGHT,2 YING LIU,1 DAIXING ZHOU,5 HAIPENG XUE,1 IRINA KHREBTUKOVA,5 MARK P. MATTSON,1 CLIVE N. SVENDSEN,2 and MAHENDRA S. RAO3,4
ABSTRACT We have examined gene expression in multipotent neural precursor cells (NPCs) derived from human fetal (f) brain tissue and compared its expression profiles with embryonic stem (ESC) cells, embryoid body cell (EBC), and astrocyte precursors using the technique of massively parallel signature sequencing (MPSS). Gene expression profiles show that fNPCs express core neural stem cells markers and share expression profiles with astrocyte precursor cells (APCs) rather than ESC or EBC. Gene expression analysis shows that fNPCs differ from other adult stem and progenitor cells in their marker expression and activation of specific functional networks such as the transforming growth factor (TGF) and Notch signaling pathways. In addition, our results allow us to identify novel genes expressed in fNPCs and provide a detailed profile of fNPCs. INTRODUCTION
C
ELL DIFFERENTIATION involves a complex interplay of transcriptional activation and repression regulated by extrinsic and cell autonomous pathways. What has become clear is that the overall response of a cell to a particular signal depends not only on the particular signal but also the state of the cell. The exposure to a signal can result in diametrically opposing effects depending on their developmental stages and the presence of other signals. For example, bone morphogenetic proteins (BMPs) may act on neuroepithelial cells to generate the neural crest or act at a later developmental time to generate dorsal interneurons. It can promote cell death or cell proliferation (1). Likewise, the effect of the protein sonic hedgehog (Shh) on cell proliferation is dependent on the exposure of a cell to fibro-
blast growth factor (FGF) (2) and the effect of ciliary neurotrophic factor (CNTF) on differentiation of neural precursors is dependent on previous exposure to platelet-derived growth factor (PDGF) (3). Therefore, it is important to determine the cell state and the context of exposure to a particular factor to predict its effects with confidence. Determining the state of a cell requires a global approach, and several large-scale analytical techniques have been developed. These include microarray analysis, where the expression of large numbers of genes in the order of thousands can be assessed rapidly and repeatedly; serial analysis of gene expression (SAGE), where expressed sequence tags (ESTs) are enumerated to determine relative levels of expression; shotgun library sequencing, where transcripts are directly sequenced and counted; and massively parallel signature sequencing
1Stem Cell Biology Unit/Laboratory of Neurosciences, National Institute on Aging Intramural Research Program, National Institutes of Health, Baltimore, MD 21224. 2Waisman Center, University of Wisconsin, Madison, WI 53705. 3Corporate Research Laboratories, Invitrogen Corporation, Carlsbad, CA 92008. 4Johns Hopkins School of Medicine, Baltimore, MD 21287. 5Solexa, Inc., Hayward, CA 94545. 6Both authors contributed equally to this work.
232
GLOBAL PROFILE OF HUMAN NEURAL STEM AND PROGENITOR CELLS (MPSS), where EST tags are analyzed as in SAGE but to a greater extent with greater speed. Each of these techniques relies on the assumption that gene expression reflects protein expression and that sufficient tags/genes can be assessed to obtain a reasonable profile of cell state. Each method has its advantages and disadvantages (4). MPSS technology is robust and highly sensitive to identify the transcript with a single copy per cell (5). Our previous experiments have suggested that 80% of transcripts can be mapped to the human genome and independent analysis have suggested that expression levels greater than 3 transcripts per million (tpm) is predictive of detectable expression (6,7). Critical to such an analysis, however, is the ability to obtain relatively pure populations of well-characterized cells and to compare expression among related populations to provide a method of determining cell type and tissue-specific expression. In this study, we chose fetal neural precursor cells (fNPCs)(1). They represent a widely studied multipotent precursor cell population and they also have been used in cell therapy. Especially for cell therapy, characterizing cells and getting profiles of global gene expression is necessary. Astrocyte precursors (APCs) are the subpopulation of differentiated neural stem cells that are restricted in their differentiation to astrocyte lineage. We chose APCs as a comparison population because several studies have indicated that adult neural stem cells (NSCs) may express some glial markers and some astrocyte precursors are likely to be stem cells. We reasoned that fNPCs and APCs would be closely related and both similarities and differences would be of importance. In addition, the fNPC profile was compared to a stem cell lineage of human embryonic stem (ESC) cells and to the universally differentiated population of embryoid body (EBC) cells.
RNA isolation and library preparation for MPSS Total RNAs were isolated using Trizol according to manufacturer’s protocol (Invitrogen), DNase-treated, and their mRNA purified. The mRNA was processed according to the MPSS protocol, as outlined in the previous publications (5,12), with some modifications. Briefly, the mRNA was reverse-transcribed and the cDNA was digested with Dpn II. The cDNA of the last Dpn II site and the downstream 16 bases were cloned into a Megaclone vector. The resulting library was amplified and loaded onto microbeads. About 1.2 million microbeads were loaded into each flow cell and the signature sequences were determined by a series of enzymatic reactions as outlined in the above publications. the abundance for each signature was converted to tpm for the purpose of comparisons among samples.
MPSS signature classification and annotation To simplify the MPSS data analysis, multiple tags corresponding to the same Unigene ID for each dataset were combined into one tpm count as follows: the sum of tpm for signatures of class 1, 2, 3, 22, 23 if any found, if not the sum of class 4, and if still nothing the sum of class 5 tags. Tag classifications are as follows: class 1, forward strand, poly(A) signal, poly(A) tail, 3-most; class 2, forward strand, poly(A) signal, 3-most; class 3, forward strand, poly(A) tail, 3-most; class 4, forward strand, no poly(A) info, 3-most, class 5, forward strand, no poly(A) info, not 3-most; class 22, unknown orientation, poly(A) signal, last before signal; class 23, unknown orientation, poly(A) tail, last before tail. The resulting signature database was used to annotate the data from the experiments.
RT-PCR amplification MATERIALS AND METHODS Cell culture
cDNA was synthesized by using a reverse transcription kit (Superscript II, Invitrogen) according to the manufacturer’s instruction. cDNA template (0.5 l) was used in a 50-l reaction volume with the RedTaq DNA polymerase (Sigma). The cycling parameters were: 94°C, 1 min; 55°C, 1 min; 72°C, 1 min, for 30 cycles. The PCR cycles were preceded by an initial denaturation of 3 min at 94°C and followed by a final extension of 10 min at 72°C. The primer sequences are available upon request.
Human cortical progenitors (fNPCs) were isolated from the embryonic central nervous system (CNS) and induced to proliferate as free-floating neurospheres as detailed previously (8–10). Neurospheres were passaged every 14 days by sectioning of neurospheres into 200m sections that were seeded into fresh growth medium at a density of or equivalent to 200,000 cells/ml. APCs were derived from human NPCs (ScienCell Research Laboratories) as described in Liu et al. (11). Human NPCs were maintained in human NPC medium (ScienCell Research Laboratories) supplemented with 1% fetal bovine serum (FBS). After 20 days in culture, almost all cells became CD44 and did not express neuronal or oligodendrocyte markers. These cells were used for further analysis.
Immunostaining
233
Immunohistochemistry was performed as the following. Briefly, human cortex neurospheres were fixed with 2% paraformaldehyde and embedded in O.C. T. Blocks were cut on a cryostat to obtain 8-m sections. Cell cultures were fixed in 2% paraformaldehyde for half an hour and washed three times with phosphate-buffered saline (PBS). In general, sections or cells were blocked in blocking buffer, con-
CAI ET AL. sisting of 5% goat serum, 1% bovine serum albumin (BSA), 0.1% Triton X-100 in PBS, for 1 h, and then incubated in the primary antibody at 4°C overnight. Appropriately coupled secondary antibodies were used for single and double labeling. All secondary antibodies were tested for cross reactivity and nonspecific immunoreactivity. Slides were mounted with FluorSave (Calbiochem) and cover-slipped. The antibodies against following proteins were used: Nestin (Rat-401, 1:5, DSHB), Sox2 (1:500; R&D systems), glial fibrillary acidic protein (GFAP, 1:500; Dako), III tubulin (1:2000; Sigma), S100 (1:200; Sigma), CD44 (1:25; generous gift from Dr. Sherman), fetoprotein (AFP, 1:500; Sigma), muscle actin (1:50; DAKO), and Oct-3/4 (1:200, Santa Cruz). Bis-benzamide (Dapi; 1:1,000, Sigma) was used to identify the nuclei. Images were captured on an Olympus fluorescence microscope.
RESULTS Characterization of fNPCs, APCs, ESC cells, and EBC cells using phenotype marker expression To assess the quality of human NSCs prior to subjecting them to MPSS analysis, we examine the expression of known NSC markers and markers of differentiated populations in NSCs generated from human fetal tissue (fNPCs). Characteristics of fNPCs have been described by Svendsen and colleagues (10). fNPCs were grown as neurospheres and expressed most NSC markers, such as Nestin, Sox2, Sox1, Brn1, and Prominin1 (Figs. 1A, 1B, 1G). In addition to stem cell markers, markers for more differentiated progenitors were also observed, which indicated that the neurospheres included more differentiated progenitor populations and differentiated cells as well. These marker included neuronal markers such as III tubulin, MAP2, and Neurofilament (NF) (Figs. 1C, 1G), astrocyte markers GFAP, S100, and CD44 (Figs. 1D, 1G), and oligodendrocyte precursor markers Olig1 and Olig2 (Fig. 1G). Human APCs were chosen as a comparison set because they are differentiated progenies of neural stem cells. Previous studies showed that astrocytes and APCs comprised the largest percentage of contaminating cells in fNPC cultures (ref. 10 and data not shown). By subtracting gene expression of APCs from that of fNPCs, we can discern the NSC-derived gene expression from that of APCs from the acquired profile of fNPC samples. APCs were prepared as described before (11). They were defined by CD44 expression (Fig. 1E) and distinguished from other precursor cells by their ability to differentiate into astrocytes (Fig. 1F), but not into oligodendrocytes or neurons under conditions where other precursor populations readily differentiate. In accordance with the homogenous immunocytochemistry results, gene expression of APCs showed astrocyte-related genes such as CD44, GFAP,
and S100 and the absence of other lineage markers like MAP2, NF, Olig1, and Olig2 (Fig. 1G). Human ESC cells and ESC cell-derived EBC cells served as a control population for a stem cell population and a differentiated cell population, respectively. As shown in Fig. 1H, ESC cells expressed the pluripotent markers Oct4 and Sox2. EB cells included some mesodermal cells (Fig. 1I) and endodermal cells (Fig. IJ).
MPSS data acquisition and verification In theory, for any given sample, MPSS can identify almost every expressed transcript by capturing, sequencing, and counting the representing tag (signature). In this study, signature sequence tags of 17/20 bp in length were generated to a depth of greater than 1.9 million tags for each sample. The total set of mapped and unmapped signatures is provided in Supplementary Table 1 (available at ftp://ftp. lynxgen.com/bioinfo/). For the comparison among cell populations, gene abundance was shown as tpm, which is the normalized value by scaling the recorded transcript number for each signature to the number per million of total transcripts. Only replicate signatures were considered so that a signature had to be present in at least two runs across all samples. These totaled 10,040, 12,217, 12,007, and 9,327 for fNPC, APC, ESC, and EBC, respectively. tpm values were reported for genes identified in at least two runs, and no single tags were included. The validity of the results was assessed by examining the expression of known genes that are expressed in fNPCs. From the literature, genes shown to be expressed in NPCs were selected, and their MPSS values were examined. The result showed that fNPCs had consistent expression value for examined genes such as Nestin, GAP43, Sox-2, Sept8, Lrrn1, Notch1, MCM2, SNIP1, CNTFR, and CST3 (Table 1). In addition, Sox1 expression was also detectable, albeit at relatively low levels (15 tpm). As predicted by the initial analysis of fNPC samples by immunocytochemistry and RT-PCR MPSS confirmed contamination of fNPC samples with more differentiated cell types. The expression of astrocyte markers GFAP and SLC1A3, oligodendrocyte precursor marker PDGFR, radial glial cell marker CBX8, and neuronal markers DCX and NCAM1 could all be detected in the MPSS data set. The APC sample was much more homogeneous and expressed far fewer nonastrocyte-related genes. Therefore, the MPSS analysis appeared to be robust and sensitive enough to identify lineage-specific markers, even those that are expressed at relatively low levels.
Overall transcriptome complexity of fNPCs and APCs and comparison with human ESC cell and EBC cell data sets To examine overall difference between cell populations, we generated scatter plots of the logarithmic val-
234
GLOBAL PROFILE OF HUMAN NEURAL STEM AND PROGENITOR CELLS ues of MPSS data from each sample and examined the correlation coefficient in four cell populations (Fig. 2A). As shown in Fig. 2B, fNPCs were more similar to APCs with a correlation coefficient value of 0.74. The next clos-
est correlation of fNPC gene expression pattern was with ESC cells (correlation coefficient value of 0.705). The overall similarity relationship was the same if only more abundant genes were used in the analysis (tpm 100) or
FIG. 1. Most of fNPCs express NSC markers, although some differentiated cells are also present. The expression patterns of the neural markers are distinguishable between fNPCs and APCs. (A,B) Almost every cell in the fNPC neurospheres is Nestin(A, green) and Sox2- (B, red) positive. (C) Subpopulation of the fNPC spheres are positive for III tubulin (red) and GFAP (green). (D) Most fNPC express S100 (red). (E,F) Almost all of APCs express CD44 (E, red) and GFAP (F, green) when they are differentiated into astrocytes. (H) Undifferentiated ESCs expressed Oct4 (red) and Sox2 (green). (I,J) EBCs differentiated from ESCs contained muscle actin-positive mesodermal cells (green) and AFP-positive endodermal cells (red). Dapi was used for nuclear staining to show all cells in the fields. (G) RT-PCR analysis of fNPCs and APCs shows that these two cell populations have distinct gene expression patterns. fNPCs are a heterogeneous progenitor cell population expressing neuronal, glial, and stem cell markers, whereas APCs show a commitment to the astrocytic lineage with expression of astrocytic markers (CD44, GFAP) and the absence of neuronal (MAP2, NF) or oligodendrocytic markers (Olig1, Olig2).
235
CAI ET AL. TABLE 1. Unigene Undifferentiated Hs.527971 Hs.134974 Hs.202526 Hs.518438 Hs.533017 Hs.163244 Hs.304682 Hs.495473 Hs.129966 Hs.3260 Hs.477481 Hs.471951 Hs.488293 Hs.490817 Hs.165950 Differentiated Hs.514227 Hs.481918 Hs.74615 Hs.387258 Hs.533317 Hs.34780 Hs.503878
tpm
OF THE
KNOWN MARKER GENES
Gene symbol
tpm (fNPC/APC)
NES GAP43 Sox1 Sox2 Sept8 Lrrn1 CST3 NOTCH1 CNTFR PSEN1 MCM2 SNIP1 EGFR VIPR2 FGFR4
1,920/1,543 1,484/265 15/7 719/705 113/15 79/4 8,248/524 210/75 174/6 131/48 104/50 28/0 28/4 8/0 5/0
GFAP SLC1A3 PDGFR CBX8 VIM DCX NCAM1
2,321/1,611 694/150 17/0 22/2 4,432/10,451 136/30 381/102
sion could be identified we mapped all the sequences onto the cytobands of the human chromosomes. As shown in Fig. 3A, the numbers of genes expressed in these four cell populations are different. APCs and ESC cells expressed genes compared to fNPCs and EBC cells. However, the expressed genes are distributed evenly on each chromosome so that there are no significant differences in percentage of expressed genes on each chromosome among fNPCs, APCs, ESC cells, and EBC cells (Fig. 3B). The overall pattern of fNPC, APC, ESC cell, and EBC cell gene expression indicated that fNPCs have a unique pattern of gene expression that serves to distinguish fNSCs from other cell populations. Furthermore, the components that gave rise to this difference do not appear to be focused to specific locations on chromosome but rather distributed throughout the genome and 70% of all genes have the expression level below 50 tpm.
Other stem cell gene expression in fNPCs
if the entire data set was used. The fNPC gene expression pattern appeared quite distinct from that of other differentiated populations including EBC cells (correlation coefficient value of only 0.252). (Fig. 2C, 2D). We then evaluated the complexity of the signature sequences generated from each sample. The cumulative tpm (with cutoff at 3 tpm) of gene distributions for human fNPCs, APCs, ESC cells, and EBC cells are shown in Table 2. A total of 9,542, 10,455, 11,214, and 8,565 distinct tags were detected in fNPCs, APCs, ESC cells, and EBC cells, respectively. Thus, the complexity of gene expression in fNPCs is comparable to that seen in other populations examined by MPSS (unpublished data) and, overall, the distribution of transcripts based on abundance was strikingly similar. Only about 1% of genes are expressed at a level greater than 1,000 tpm, greater than 70% of genes are expressed at less than 50 tpm, and greater than 40% of all signatures are present at a level of 10 tpm or lower. As previously reported for other cell samples, examination of the most abundant genes showed that the top 200 genes included ribosomal, mitochondrial, and housekeeping genes whereas growth factors, transcription factors, and regulators of gene expression were expressed in the low tpm value. A list of the top 100 genes of fNPCs and APCs is shown in Table 3. To determine if any hot or cold spots of gene expres-
We showed above that overall gene expression of fNPCs overlaps with ESC cells. Next, we checked for the genes expressed by other stem cells such as germ cells, hematopoietic stem cells (HSCs), and mesenchymal stem cells (MSCs). Commonly used markers for ESC cells, germ cells, HSCs, and MSCs were selected from a literature search and their expressions in fNPCs were examined (Table 4). As expected, fNPCs expressed generally recognized NSC markers of nestin, sox1, sox2, and prom1 but did not express most other stem cell markers. Though fNPCs shared an overall gene profile with ESC cells, it did not express ESC cell-specific genes such as Pou5f1 (Tpm 0 vs. 658), Nanog (Tpm 0 vs. 16), or Lefty1 (Tpm 0 vs. 72). Likewise, fNPCs did not have genes for HSCs (CD34, VEGFR1, Thy1), MSCs (CD55, CD105), or germ cells (Twist, KCNQ2), which indicates fNPCs are a unique population that can be distinguished from other stem cell populations.
Differential pattern of signaling pathways in fNPCs To scrutinize the global gene expression difference observed between fNPCs and the other population, we selected genes related to signal pathways such as Notch, epidermal growth factor (EGF), FGF, TGF, Wnt, leukemia inhibitory factor (LIF), tumor necrosis factor (TNF), and cell cycle pathway, and then analyzed their expression profiles. Each population had a distinct pattern for almost every defined signal pathway (Table 5). For example, fNPCs used a distinct set of TGF- signaling molecules that differed from the other three cell populations. Fetal NPCs lacked TGFB1, TGFBI but expressed TGFB2. In the case of Notch signaling, fNPCs and APCs expressed a high amount of Notch 1 (tpm 210, 75) and low amount of Notch 3 (tpm 0, 32), whereas ESC
236
GLOBAL PROFILE OF HUMAN NEURAL STEM AND PROGENITOR CELLS
FIG. 2. Global comparison of fNPCs, APCs, ESC cells and EBC cells transcriptome. (A) Correlation coefficient values of samples indicate the overall similarity of the two populations that are compared. (B–D) Scatter plots of the genes detected by MPSS show that fNPCs are more similar to APCs than to ESC cells or EBC cells. (B) Log value of fNPCs versus log value of HAPC. (C) Log value of fNPCs versus log ESC cells. (D) Log value of fNPCs versus log value of EBC cells.
cells and EBC cells expressed more Notch 3 (tpm 213, 177) than Notch 1 (tpm 0, 5). Also, fNPCs seemed likely to be responsive to EGF and FGF mitogens because the EGF and FGF receptors were present. In contrast, ESC cells may depend more on FGF than EGF because they did not express high levels of EGF receptor. It seems reasonable to speculate that fNPCs represent a unique pop-
TABLE 2.
tpm CUMULATION
fNPCs
IN
ulation compared to not only its differentiated progenies but to other stem cell populations as well.
Genes specifically expressed in fNPCs Cell membrane receptors and extracellular matrix molecules: One purpose of this research was to identify
fNPCS, APCS, ESC CELLS,
APCs
AND
EBC CELLS
ESC cells
EBC cells
MPSS (tpm)
No. of tag
(%)
No. of tag
(%)
No. of tag
(%)
No. of tag
(%)
10,000 5,000 1,000 500 100 50 10 3 1
6 25 155 311 1,600 2,923 6,828 9,542 10,040
0.06 0.25 1.54 3.10 15.94 29.11 68.01 95.04 100
9 27 131 241 1055 1921 5932 10455 12217
0.07 0.22 1.07 1.97 8.64 15.72 48.56 85.58 100
7 26 117 233 1289 2584 7956 11214 12007
0.06 0.22 0.97 1.94 10.74 21.52 66.26 93.40 100
13 41 142 246 1218 2302 6162 8565 9327
0.14 0.44 1.52 2.64 13.06 24.68 66.07 91.83 100
237
CAI ET AL. TABLE 3.
TOP 100 tpm GENES
IN
fNPCS
AND
APCS
Only in fNPCs
Both in fNPCs and APCs
Only in APCs
Ribosomal proteins
RPL10A, RPL32, RPL7A, RPL8, RPS27A, RPS29 RPS9 (7)
RPL14, RPL29, RPN2, RPS28, UBA52 (5)
Mitochondrial protein Structural proteins
None (0)
Membrane proteins and ECM proteins
CSPG3, GPR56, PCDHGC3, PON2, PTPRZ1, TNC, TTYH1 (7) APLP2, CST3, IGFBP2, LAPTM4B, SCD, UBB (5)
RPL10, RPL11, RPL13, RPL13A, RPL15, RPL22, RPL23, RPL26, RPL27A, RPL28, RPL3, RPL30, RPL35, RPL37A, RPL4, RPL41, RPL5, RPL6, RPL9, RPLP1, RPLP2, RPS11, RPS17, RPS18, RPS19, RPS2, RPS20, RPS3, RPS3A, RPS8 (30) 16S rRNA, COX1, COX2, CYTB, ND1, ND3, ND4, ND6 (8) ACTB, ACTG1, CFL1, CTNNA1, GFAP, K-ALPHA-1, MGC8685, MYL6, NES, TUBB, VIM (11) CD63, LAMR1 (2)
Metabolic proteins
Signal transduction proteins
Nuclear proteins Others Unknown genes Total
PMP2 (1)
GNAI2, HLA-A, HLA-B, MARCKSL1, MRC2, NGFRAP1, PTN, (7) HOP (1)
CALM2, CLU, FTH1, FTL, GAPD, PKM2, UBC (7)
EEF1A1 (2), GNAS, PTMA, PTMS, TMSB10, TMSB4X (7)
12S rRNA (1) ANK2, MAP1B, SEPT11, TUBB3 (4) CD81, COL1A1, SPARC (3) ATP5A1, APT5B, ATP5O, CCNI, CKB, PPIA, TPT1 (7) ANXA2, EEF1G, EEF2, HSPCA, MAGED1, NNAT, PSAP, SRP14 (8) NAP1L1 (1)
SMOC1 (1) Hs.536051 (1)
H3F3A, HMGB1, HNRPA1, NSEP1 (4) None Hs.485090 (1)
C20orf149 (1) 0
30
70
30
Total tpm of top 100 fNPC genes: 457,258/1,096,639 41.70%. Total tpm of top 100 APC genes: 496,190/951,561 52.14%.
distinct characteristics for NSCs. Until now, no single, exclusive NSC marker has been described. This is particularly true for cell-surface markers that could serve not only as markers for identification but could also be used to isolate NSCs. Therefore, we examined the data to identify potential cell-surface molecules that could serve as markers of fNPCs. The list of potential markers is provided and confirmation of a selected subset by RT-PCR is shown (Fig. 4). Aquaporin 4 is a member of the osmoreceptor class that regulates body water balance and mediates water flow within the CNS. It was highly expressed in fNPCs (tpm 1,042) but the expression levels in APCs, ESC cells, or EBC cells was very low or ab-
sent (tpm 0, 0, 29, respectively). Similarly, CNTFR expression was high in fNPCs, low in APCs and absent in ESC cells or EBC cells. Cell adhesion molecules with homology to L1CAM (CHL1) have been shown to play a role in neurite outgrowth and neuronal cell survival (13). In this study, the expression of CHL1 was high in fNPCs but low or absent in APCs, ESC cells, or EBC cells (tpm 82 vs. 1, 3, and 0, respectively). Neuroligin3, a member of the family of proteins that meditates cellto-cell interactions between neurons (14), was high in fNPCs as compared to APCs, ESC cells, or EBC cells (tpm 578 vs. 25, 0, and 0, respectively). Likewise, Seizure-related 6 homolog like SEZ6L (tpm 264 vs. 47, 21, and 0,
238
GLOBAL PROFILE OF HUMAN NEURAL STEM AND PROGENITOR CELLS
A
Expressed genes located to chromosomes
1400 1200
Number of genes
1000 NPC APC ESC EBC
800 600 400 200 0 1
2
3
4
5
B
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 Human chromosome
X
Y Mito
Percentage of genes expressed on chromosomes
12.00 10.00
Percentage
8.00
NPC APC ESC EBC
6.00 4.00 2.00 0.00 1
3
5
7
9
11 13 15 Human chromosomes
17
19
21
X
Mito
FIG. 3. Distribution of genes expressed in fNPCs, APCs, ESC cells, and EBC cells mapped to chromosomes. More genes are expressed in APCs and ESCs compared to fNPCs or EBCs, but these genes are evenly distributed so that there are no significant differences among the distribution patterns. (A) Plot of gene numbers expressed on each chromosome. (B) Plot of percentages of total detected genes expressed on each chromosome.
respectively), anion transfer membrane protein family member SLCO1C (tpm 142 vs. 0, 0, and 0, respectively), and proteoglycan BCAN (tpm 1,252 vs. 11, 0, and 0, respectively) were differentially expressed in fNPCs and other populations examined. Although we have highlighted differences in cell-surface molecules, differences were detected in all other classes of molecules examined. Prox1, for example, a member of the homeobox transcription factor family, which has been shown to control progenitor cell proliferation (15), was also expressed highly in fNSCs. Its levels in other populations were low in ESC cells, EBC cells and APCs (tpm 207 vs. 5, 0, and 0, respectively). A de-
tailed list of genes whose expression was higher fNSCs compared to APCs and ESC cells at levels of more than five-fold is provided in Supplementary Table 2 (available at ftp://ftp.lynxgen.com/bioinfo/). Overall the data confirmed that the MPSS data set is a rich source of markers that are differentially expressed in stem and differentiated cell populations.
Novel genes expressed in fNPCs An advantage of the MPSS methodology is that it is possible to identify a novel gene in the cell population analyzed without any apriori knowledge of that gene. As
239
CAI ET AL. TABLE 4.
EXPRESSION LEVELS
NSC markers
UniGene no.
Nestin Sox2 Sox1 Notch1 FGFR4 Glut1 ABCG2 Prom1 MCM2 GJA1 EGFR
Hs.527971 Hs.518438 Hs.202526 Hs.495473 Hs.165950 Hs.473721 Hs.480218 Hs.479220 Hs.477481 Hs.74471 Hs.488293
Other stem cell markers
UniGene no.
LIN28 DNMT3B CLDN6 SFRP2 GAL LIN41 IMP-2 LECT1 GABRB3 GYLTL1B LEFTY1 LEFTY2 SCGB3A2 KIT CKMT1 FOXD3 CER1 Germ cell markers Numb Twist AREG (amphiregulin) TIAL1 SIAH1 PJA1 PHC2 PIWIL2 PEG3 KCNQ2 GDF3
OF
NSC
AND
OTHER STEM CELL MARKERS Other stem cell markers
fNPC (tpm) 1920 719 15 210 5 15 N.D. 6 104 104 28
ZFX ZFY GATA6 GAGE5 SYCP3 DDX4 (VASA) ESC and germ cells POU5F1 (Oct3/4) ZFP42 (Rex1) UTF1 IFITM1 (Fragilis-1) IFITM2 (Fragilis-2) STELLAR Nanog HSCs CD34 CD133 (AC133) VEGFR1 (Flt1) KDR (VEGFR2) CDCP1 C-Kit (CD117) Thy-1 (CD90) CD38 Runx1 Runx2 Runx3 MSCs CD54 CD55 CD105 (ENG) CD166 (ALCAM) CD13 CD44 Thy-1 (CD90) Other SC markers ITGA6 Keratin 15 p63 (CKAP4) MSX2 HoxB4 PDX-1 (IPF1) M cadherin (CDH15)
fNPC (tpm)
Hs.86154 Hs.251673 Hs.533779 Hs.481022 Hs.278959 Hs.517750 Hs.35354 Hs.421391 Hs.302352 Hs.86543 Hs.278239 Hs.520187 Hs.483765 Hs.479754 Hs.425633 Hs.552589 Hs.248204
0 0 0 0 0 N.D. 42 0 0 0 0 0 0 0 0 N.D. 0
Hs.509909 Hs.66744 Hs.270833
11 0 N.D.
Hs.501203 Hs.295923 Hs.522679 Hs.524271 Hs.528649 Hs.201776 Hs.161851 Hs.86232
93 6 81 80 1 87 0 N.D.
an example, we screened for novel genes (based on tpm values) whose expression is high in fNPCs and low or absent in APCs, ESC cells, or EBC cells populations. The entire list is available in Supplementary Table 2 and a subset is shown in Fig. 5. Results were confirmed using RT-PCR for 13 genes (Fig. 5 and Table 6).
IN THE
fNPCS
UniGene no.
fNPC (tpm)
Hs.370424 Hs.522845 Hs.514746 Hs.278606 Hs.506504 Hs.223581
41 11 0 N.D. N.D. N.D.
Hs.249184 Hs.335787 Hs.458406 Hs.458414 Hs.174195 Hs.528118 Hs.329296
0 0 0 31 N.D. N.D. 0
Hs.374990 Hs.479220 Hs.507621 Hs.479756 Hs.476093 Hs.479754 Hs.134643 Hs.479214 Hs.149261 Hs.122116 Hs.170019
0 6 0 0 8 0 0 N.D. 20 N.D. N.D.
Hs.515126 Hs.527653 Hs.76753 Hs.150693 Hs.1239 Hs.502328 Hs.134643
N.D. 0 0 50 N.D. 70 0
Hs.133397 Hs.80342 Hs.74368 Hs.89404 Hs.532669 Hs.32938 Hs.148090
14 N.D. 64 0 0 N.D. N.D.
In addition, the EST values of each clone were imported from the National Cancer Institute (NCI) EST database and overall information from the Go ontology data base are shown in Table 6. For all but Hs.8379, EST counts in brain tissue were high, attesting independently that these genes are relatively brain specific. Hs.46627,
240
GLOBAL PROFILE OF HUMAN NEURAL STEM AND PROGENITOR CELLS TABLE 5. Pathway Notch
DIFFERENCES AMONG fNPCS
Genes
ADAM10 DLK1 NOTCH1 FHL1 BLOC1S1 HES1 JAG1 PSEN1 NOTCH3 DVL2 Cell Cycle ATM CHEK2 CUL1 CUL3 CUL4B CCNC CCND3 CCNG1 CDK4 CDK7 CDKN1A CDKN1B CDKN2A E2F2 MCM3 NBS1 PRC1 RAD50 RB1 RBL2 UBE3A ABL1 CDC6 CCNB1 CCND1 CDK2 CDK6 E2F1 E2F4 MCM2 MCM5 MCM6 MTBP PCNA TFDP2 EGF EGFR GAB1 RAF1 MAP2K2
AND THE
fNPCs
APCs
ESC cells
EBC cells
17 0 210 205 598 140 138 131 0 5 67 160 64 72 68 31 86 28 125 39 270 71 32 71 403 32 308 42 29 92 109 56 47 81 3 39 20 14 16 104 52 22 0 86 23 28 95 133 106
10 93 75 44 271 196 22 48 32 40 17 4 11 15 8 18 20 63 186 18 4 26 50 6 74 15 301 61 16 42 10 44 2 74 49 29 69 10 12 50 0 11 0 119 66 4 9 41 302
112 70 0 0 58 84 19 12 213 23 9 8 37 19 4 97 11 429 114 86 22 39 0 2 543 85 130 280 4 17 95 27 33 188 280 29 18 0 54 62 0 51 0 167 122 4 0 35 128
11 121 5 118 132 125 19 65 177 23 1 17 25 38 42 113 87 95 86 53 594 14 80 0 105 13 50 111 7 50 25 37 5 204 240 25 33 0 23 38 0 31 0 97 51 24 2 128 218
OTHER CELL POPULATIONS Pathway TGF-
Wnt
LIF
Apoptosis and TNF
FGF
241
IN
VARIOUS SIGNAL PATHWAYS
Genes
fNPCs
APCs
ESC cells
EBC cells
BMP2 ID2 ID4 SMAD1 Sox4 ARAF TGFB2 FOS JUN BMP1 BMP7 COL1A1 GDF15 SMAD2 SMAD4 SAMD6 SERPINE1 TGFB1 TGFBI DKK3 FZD3 LRP5 LRP6 DVL2 FZD2 WNT4 CNTFR PIAS1 SH3KBP1 SOCS4 LIFR BIRC5 CFLAR CASP3 TRAF2 TRAF4 TANK CRADD BAD BAX TNFRSF10B TNFRSF19 FGF1 FGF2 FGFR3 FGFR1 FGF14 HSPG2 SPRY1
94 96 33 113 649 0 426 85 51 1 73 0 9 3 15 0 0 4 0 267 13 55 54 5 0 0 174 29 27 28 31 80 33 49 51 152 48 8 102 73 113 13 19 41 57 352 0 8 0
7 267 80 39 323 17 61 293 7 29 52 1550 16 8 21 3 98 58 411 1305 22 33 30 40 16 0 6 13 9 2 17 2 6 22 6 41 5 9 81 70 34 26 0 6 44 396 0 14 68
35 104 0 14 117 0 10 15 0 2 56 3982 90 46 43 0 47 4 265 66 37 42 33 23 0 1 0 40 20 2 1 29 0 132 86 25 4 0 0 39 93 0 0 72 79 649 0 31 27
69 222 5 76 42 15 131 7 0 60 5 68889 256 1 36 48 499 79 5393 211 0 48 14 23 4 10 0 22 98 12 32 24 3 3 34 4 0 19 39 70 125 1 0 8 67 136 0 227 0
CAI ET AL.
FIG. 4. RT-PCR confirmation of known genes specific to fNPCs. All of the genes are organized into two groups: one group of cell membrane or matrix proteins and the other of cytoplasmic proteins and transcription factors.
Hs.491287, Hs.517235, and Hs.66187 showed a very restricted pattern of expression based on EST analysis with their expression reported less than 11 (10, 11, 7, and 5, respectively) out of the 31 tissues examined. The EST profile of Hs.66187 in particular showed a highly restricted expression pattern so that more than half of all contributing ESTs came from brain. When EST counts were broken down by developmental stages, 10 out of 13 clones showed restricted expression to embryonic or juvenile stages of development at a time when stem and progenitor cells are much more abundant (as compared to adult tissue). Overall the results show that the MPSS data set is a rich source for identifying novel (know and unknown) markers of stem cell and progenitor cells.
DISCUSSION The overall objective of this study was to profile human NSCs and show that specific gene markers can be identified by comparing fNPCs with both a more differentiated cell population (an astrocyte precursor cell) as
well as more immature populations such as ESC cells and EBC cells. We chose MPSS, which is a large-scale and sensitive method, to compare the whole transcriptome based on our previous success with this methodology (6,16). Our results showed that fNPCs differ from APCs, ESC cells, or EBC cells. Multiple genes not known to be expressed by APCs or NSCs could be identified. We were able to confirm the expression of a subset by PCR or by database comparisons. Several novel progenitor cell markers could also be identified, and signaling pathways that are active at different stages of development could be determined. Global analysis of gene expression confirmed that fNPCs maintained as undifferentiated populations contain stem cells and that markers previously identified as markers of NSCs continue to be expressed in these propagated populations. These included Sox1, Notch1, Cystatin-c, GAP43, Sept8, and Lrrn1. In addition, the expression of LIF receptors, EGF receptors, and FGF confirmed the importance of these molecules in the propagation of stem cells, and pathway analysis of these signaling molecules indicates that these pathways are active in NSC cultures.
FIG. 5. RT-PCR confirmation of novel genes specific to fNPCs. RT-PCR analysis shows that these unknown genes identified by MPSS are specifically expressed in the fNPC population, but not in APCs.
242
GLOBAL PROFILE OF HUMAN NEURAL STEM AND PROGENITOR CELLS TABLE 6.
UNKNOWN GENES IDENTIFIED
Unigene ID
tpm (fNPC/APC)
EST countsa
Hs.508108
100/3
6
13q14.2
Hs.10414 Hs.386989 Hs.22920 Hs.126856 Hs.500643 Hs.46627 Hs.51483 Hs.492187 Hs.342478 Hs.517235 Hs.66187 Hs.8379
891/40 111/9 72/1 278/0 131/15 180/10 187/2 174/2 243/1 327/15 182/3 51/0
34 15 21 30 43 41 6 9 23 5 45 0
1q32.1 17p12 20p12 5p15.32 17q25.3 20q13.33 12q13.13 8q21.3 11p14.2 21q21 2 and 5 2p25
aEST
Chromosome location
BY
MPSS Possible function or suggested molecule Heterogeneous nuclear ribonucleoprotein A1 Like Hypothetical protein FLJ10748 DKFZP566o084 protein Chromosome 20 open reading frame 103 Similar to CG4502-PA Ring finger protein 157 Chromosome 20 open reading frame 58 Hypothetical protein MGC17301 Hypothetical protein DKFZp761D112 Similar to Riken cDNA 1110018M03 Chromosome 21 open reading frame 49 Clone 23700 mRNA sequence Clone CS0DJ001YJ05 of T Cells
counts in human brain.
However, we noted that, in addition to stem cells, these cultures contain a substantial fraction of differentiated cells. Markers of neurons, astrocytes, and early markers of oligodendrocyte differentiation could be detected. These were confirmed by immunocytochemistry and RT-PCR, suggesting that data sets such as these must be used with caution when assessing NSC-specific markers. Nevertheless, the fraction of NSCs present is sufficiently large that differential expression (when compared to other populations) may allow one to identify cell type-specific markers. We chose to compare with APCs because these constituted the largest contaminating population and with ESC cells, another class of stem cells, and with other differentiated cells (EBC cells) for which verified MPSS datasets were available (Supplementary Table 3, available at ftp://ftp.lynxgen.com/bioinfo/). To determine further if a stemness phenotype existed, we examined marker expression of other stem cell populations. As is shown in the results (Table 4), no co-expression of stem cell markers was seen. Likewise, comparison with other data sets suggested that fNPCs are distinct from ESC cells in marker expression, cytokine and extracellular matrix response, expression of telomerase activity, and cell cycle control. Global pair-wise comparison suggested that fNSCs are more similar to APCs than to ESC cells. Overall the results clearly show that stem cell populations can be readily distinguished from one another, maintenance in culture does not obscure the phenotype sufficiently to allow cells to revert to some base stem stage and that stem cells from a particular lineage appear more similar to differentiated cells in that lineage than to other stem cell populations. Comparison with APCs identified multiple markers that could be used to distinguish between these popula-
tions. These included known markers as well as previously unidentified markers. A subset of these markers were tested by RT-PCR and the difference in gene expression was confirmed (Fig. 4). We anticipate that the present data set will readily allow us to define a small subset of markers that can be used to access the differentiation status of examined neural precursor cells. Two other groups have identified NSCs and examined gene expression by somewhat different methods (17,18). In both cases, those authors noted the difficulty in getting pure populations of cells and used somewhat different strategies than ours to identify NSC-specific genes. We have not directly compared our data set to those generated by these investigators because our previous studies have cautioned against comparison across different species and different platforms. We have found (unpublished data) that using the same sample and using an array format and MPSS to generate data showed approximately 60–70% equivalence (depending on the stringency of the cut-offs used). However, if crossspecies comparisons of samples using identical culture methods are performed, then results are poorer. Less than 40% similarity was seen when mouse ESC cells and human ESC cells were compared, even when the same assay platform (MPSS) was used. Comparison of selected gene expression, however, does provide substantial confirmation, and we suggest that similar case-by-case comparisons are worthwhile, but global comparisons across different assay platforms and cross species are probably not useful in determining similarity. In conclusion, we have used MPSS to obtain a profile of human fetal-derived neural progenitors and have compared their gene expression to those of ESC cell, EBC cells, and APCs. The profile indicated that fNPCs were composed
243
CAI ET AL. of cells including NSCs, differentiated progenitor cells into neurons as well as glial cells and their profiles were unique enough to be discerned from other three cell populations. In addition, the comparison of fNPCs to APCs, ESC cells, and EBC cells made it possible to identify novel genes that are differentially expressed in fNPCs.
ACKNOWLEDGMENTS This work was supported by the Intramural Research Program of the National Institutes of Health, National Institute on Aging, CNS Foundation, and the Packard Foundation. We thank all members of our laboratories for constant stimulating discussions. M.S.R. acknowledges the contributions of Dr. S. Rao that made undertaking this project possible. We thank Q Therapeutics for providing data on astrocyte precursor cells.
REFERENCES 1. Panchision DM and RD McKay. (2002). The control of neural stem cells by morphogenic signals. Curr Opin Genet Dev 12:478–487. 2. Kalyani AJ, D Piper, T Mujtaba, MT Lucero and MS Rao. (1998). Spinal cord neuronal precursors generate multiple neuronal phenotypes in culture. J Neurosci 18:7856–7868. 3. Park JK, BP Williams, JA Alberta and CD Stiles. (1999). Bipotent cortical progenitor cells process conflicting cues for neurons and glia in a hierarchical manner. J Neurosci 19:10383–10389. 4. Pollock JD. (2002). Gene expression profiling: methodological challenges, results, and prospects for addiction research. Chem Phys Lipids 121:241–256. 5. Brenner S, M Johnson, J Bridgham, G Golda, DH Lloyd, D Johnson, S Luo, S McCurdy, M Foy, M Ewan, R Roth, D George, S Eletr, G Albrecht, E Vermaas, SR Williams, K Moon, T Burcham, M Pallas, RB DuBridge, J Kirchner, K Fearon, J Mao and K Corcoran. (2000). Gene expression analysis by massively parallel signature sequencing (MPSS) on microbead arrays. Nature Biotechnol 18:630–634. 6. Miura T, Y Luo, I Khrebtukova, R Brandenberger, D Zhou, RS Thies, T Vasicek, H Young, J Lebkowski, MK Carpenter and MS Rao. (2004). Monitoring early differentiation events in human embryonic stem cells by massively parallel signature sequencing and expressed sequence tag scan. Stem Cells Dev 13:694–715. 7. Audic S and JM Claverie. (1997). The significance of digital gene expression profiles. Genome Res 7:986–995. 8. Ostenfeld T, E Joly, YT Tai, A Peters, M Caldwell, E Jauniaux and CN Svendsen. (2002). Regional specification of rodent and human neurospheres. Brain Res Dev Brain Res 134:43–55.
9. Svendsen CN, MG ter Borg, RJ Armstrong, AE Rosser, S Chandran, T Ostenfeld and MA Caldwell. (1998). A new method for the rapid and long term growth of human neural precursor cells. J Neurosci Methods 85:141–152. 10. Wright LS, J Li, MA Caldwell, K Wallace, JA Johnson and CN Svendsen. (2003). Gene expression in human neural stem cells: effects of leukemia inhibitory factor. J Neurochem 86:179–195. 11. Liu Y, SS Han, Y Wu, TM Tuohy, H Xue, J Cai, SA Back, LS Sherman, I Fischer and MS Rao. (2004). CD44 expression identifies astrocyte-restricted precursor cells. Dev Biol 276:31–46. 12. Brenner S, SR Williams, EH Vermaas, T Storck, K Moon, C McCollum, JI Mao, S Luo, JJ Kirchner, S Eletr, RB DuBridge, T Burcham and G Albrecht. (2000). In vitro cloning of complex mixtures of DNA on microbeads: physical separation of differentially expressed cDNAs. Proc Natl Acad Sci USA 97:1665–1670. 13. Hillenbrand R, M Molthagen, D Montag and M Schachner. (1999). The close homologue of the neural adhesion molecule L1 (CHL1): patterns of expression and promotion of neurite outgrowth by heterophilic interactions. Eur J Neurosci. 11:813–826. 14. Nam CI and L Chen. (2005). Postsynaptic assembly induced by neurexin-neuroligin interaction and neurotransmitter. Proc Natl Acad Sci USA 102:6137–6142. 15. Dyer MA, FJ Livesey, CL Cepko and G Oliver. (2003). Prox1 function controls progenitor cell proliferation and horizontal cell genesis in the mammalian retina. Nature Genet 34:53–58. 16. Reinartz J, E Bruyns, JZ Lin, T Burcham, S Brenner, B Bowen, M Kramer and R Woychik. (2002). Massively parallel signature sequencing (MPSS) as a tool for in-depth quantitative gene expression profiling in all organisms. Brief Funct Genomic Proteomic 1:95–104. 17. Kornblum HI and DH Geschwind. (2001). Molecular markers in CNS stem cell research: hitting a moving target. Nature Rev Neurosci. 2:843–846. 18. Abramova N, C Charniga, SK Goderie and S Temple. (2005). Stage-specific changes in gene expression in acutely isolated mouse CNS progenitor cells. Dev Biol 283:269–281.
Address reprint requests to: Dr. Soojung Shin National Institute on Aging Laboratory of Neurosciences 333 Cassell Drive, Room 406A Baltimore, MD 21224 E-mail:
[email protected] Received November 15, 2005; accepted December 3, 2005.
244
