A Genetic Analysis of Neural Progenitor Differentiation

A Genetic Analysis of Neural Progenitor Differentiation

Neuron, Vol. 29, 325–339, February, 2001, Copyright 2001 by Cell Press A Genetic Analysis of Neural Progenitor Differentiation Daniel H. Geschwind,1...

1MB Sizes 0 Downloads 2 Views

Neuron, Vol. 29, 325–339, February, 2001, Copyright 2001 by Cell Press

A Genetic Analysis of Neural Progenitor Differentiation Daniel H. Geschwind,1,2,6,8 Jing Ou,1,2,9 Mathew C. Easterday,3,4,9 Joseph D. Dougherty,1,2 Robert L. Jackson,1,2,3,4 Zugen Chen,5 Heath Antoine,1,2 Alexey Terskikh,7 Irving L. Weissman,7 Stanley F. Nelson,5,6 and Harley I. Kornblum3,4 1 Neurogenetics Program 2 Department of Neurology 3 Departments of Molecular and Medical Pharmacology and Pediatrics 4 The Crump Institute for Molecular Imaging 5 Department of Human Genetics 6 Mental Retardation Research Center UCLA School of Medicine Los Angeles, California 90095 7 Departments of Pathology and Developmental Biology Stanford University School of Medicine Stanford, California 94305

Summary Genetic mechanisms regulating CNS progenitor function and differentiation are not well understood. We have used microarrays derived from a representational difference analysis (RDA) subtraction in a heterogeneous stem cell culture system to systematically study the gene expression patterns of CNS progenitors. This analysis identified both known and novel genes enriched in progenitor cultures. In situ hybridization in a subset of clones demonstrated that many of these genes were expressed preferentially in germinal zones, some showing distinct ventricular or subventricular zone labeling. Several genes were also enriched in hematopoietic stem cells, suggesting an overlap of gene expression in neural and hematopoietic progenitors. This combination of methods demonstrates the power of using custom microarrays derived from RDA-subtracted libraries for both gene discovery and gene expression analysis in the central nervous system. Introduction Neurons, astrocytes, and oligodendrocytes share a common progenitor (Williams et al., 1991; Reynolds and Weiss, 1992, 1996; Reynolds et al., 1992; Davis and Temple, 1994; Gage et al., 1995; Weiss et al., 1996; Cameron and McKay, 1998). However, with a few exceptions (Ferri and Levitt, 1995; Vicario-Abejon et al., 1995; Johe et al., 1996), the mechanisms whereby CNS progenitors are generated and the factors guiding them from multipotent undifferentiated cells to unipotent, terminally differentiated neurons and glia are unknown. It 8

To whom correspondence should be addressed (e-mail: [email protected] ucla.edu). 9 These authors contributed equally to this work.


is likely that CNS progenitors express populations of genes that inhibit their commitment toward specific lineages and other genes that allow for their response to inductive signals, while lineage-committed cells will express genes that induce and regulate their subsequent differentiation and development. Analogous to the field of hematopoiesis (Phillips et al., 2000; Weissman, 2000), an understanding of the gene expression patterns of CNS progenitor cells and their progeny is a critical step in forwarding our understanding of these processes (Lillien, 1998). Despite the enormous potential importance of the study of CNS stem cells, investigators have been hampered by several technical problems. First, CNS stem cells can be propagated in vitro but only in heterogenous cultures containing not only CNS stem cells but also numerous progenitors and more differentiated cells. Furthermore, there are no effective markers or combination of markers for murine CNS stem cells and other early progenitors, making isolation of pure stem cell populations challenging. Tissue and cellular heterogeneity, such as that existing in CNS progenitor cultures, presents an impediment to most subtraction or expression studies of molecular diversity within the nervous system (e.g., Geschwind, 2000). In the current study, we have largely overcome this impediment to perform a genomic-level expression analysis of neural progenitor differentiation that combines a powerful genetic subtraction technique, representational difference analysis (RDA), with cDNA microarray analysis (Welford et al., 1998), followed by downstream screening using in situ hybridization. We have identified many novel and known genes enriched in neural progenitors and have begun an initial characterization of the genetic programs and signaling pathways involved in CNS progenitor cell proliferation and function. This step-wise approach has allowed us to efficiently identify a large number of genes that were differentially expressed in a minority of cells within a heterogeneous population—an important technical advantage of the methods employed. The general approach of using subtracted arrays should therefore be readily applicable to many in vivo and in vitro paradigms of nervous system development, functioning, and degeneration. Results Neurosphere Cultures Are Enriched in Multipotent Neural Progenitors Cultured stem cells from various regions of the mammalian ventricular and subventricular zone form colonies called neurospheres that proliferate and are pluripotent under the proper growth factor stimulation (Reynolds et al., 1992; Reynolds and Weiss, 1992, 1996). From 3% to 4% of the cells within neurospheres are actually true stem cells in that they can give rise to all three neural lineages (Gritti et al., 1996). We reasoned that, under differentiating conditions for 24 hr, the proportion of

Neuron 326

Figure 1. The NS Condition Is Enriched in Pluripotent Progenitors (A) GFAP immunofluorescence for astrocytes. (B) ␤-tubulin III immunofluorescence for postmitotic neurons. (C) O4 immunofluorescence for oligodendrocytes and committed progenitors. (D) Composite overlayed montage of the pluripotent stem cell colony shown in (A)–(C), showing all three cell types in this colony. (E) The NS condition is graphed in blue, and the DC condition is in violet. The numbers in the “total” columns equal the total number of colonies, while those labeled “tripotent” include only those single cell–derived colonies that give rise to neurons, astrocytes, and oligodendrocytes.

multipotent and pluripotent progenitors within neurosphere cultures would decrease dramatically, allowing the identification of genes enriched in stem cells or neural progenitors by subtracting genes expressed in differentiated from undifferentiated cultures. Neurospheres derived from P0 mouse cortex were cultured under differentiating (Experimental Procedures) and nondifferentiating conditions (Kornblum et al., 1999). Both differentiated and undifferentiated cultures were dissociated into single-cell suspension and replated in the presence of human basic fibroblast growth factor (bFGF) in growth medium at densities giving rise to largely clonally derived spheres (Temple, 1989; Tropepe et al., 1999, 2000). Under these conditions, the undifferentiated neurosphere cultures (NS) produced 847 spheres of cells, while the cells that had been differentiated (DC) produced 98 spheres. These observations are consistent with the presence of larger numbers of proliferating, clonogenic progenitors in the NS conditions (Figure 1). Immunocytochemistry using neuron-, astrocyte-, and oligodendrocyte-specific markers (Experimental Procedures) was conducted to determine progenitor potentiality. The number and composition of spheres were determined in each condition and are summarized in Figure 2. Undifferentiated NS cultures produced far greater

numbers of pluripotent spheres than the DC conditions. The ratio of clones giving rise to all three neural lineages was 10:1, favoring the undifferentiated neurospheres (Figure 1), and 3.4% of cultured spheres were tripotent, consistent with previously published data (Gritti et al., 1996). RDA Subtraction An RDA subtraction was performed in which cDNA derived from stem cell cultures that had been differentiated for 24 hr (the DC cell condition) was subtracted from cDNA of sister cultures that were maintained as stem/ progenitor cell cultures throughout the 2 week culture period (the NS neurosphere condition). The subtraction was also carried out in the reverse direction (DC–NS), and the rounds were monitored by gel electrophoresis (data not shown; Lisitsyn and Wigler, 1993; Hubank and Schatz, 1994). Typically, RDA is conducted for four rounds, producing a subtracted but very redundant pool of a few highly differentially expressed products that are cut out as bands from a gel (Hubank and Schatz, 1994). Genes with smaller but significant differences in expression between the two comparison samples (⬍10fold) are usually not identified using standard RDA protocols (Hubank and Shatz, 1994). But, such genes have

Genetic Analysis of CNS Progenitors 327

Figure 2. Experimental Flow and Composite Microarray Image (A) Experimental flow. (B) cDNA microarray cohybridized with NS and DC amplicons. DC–NS clones are arrayed on the top left of each quadrant from rows 1–6, columns 1–24, and rows 7–12, columns 1–16, in the area surrounded in yellow, and the remainder are NS–DC clones (outlined in green) or control spots (light blue), as labeled. Two duplicate arrays are shown side by side, with the dyes labeling each amplicon reversed.

been identified in model systems such as cancer cell lines when fewer rounds of RDA are performed and the output is screened using microarrays (Welford et al., 1998). Here, we performed multiple rounds of RDA and arrayed 192 clones from each subtracted library derived from rounds two and three in a pilot experiment to determine the optimum number of rounds to subject to a large-scale microarray screen (Experimental Procedures). The subtracted libraries derived after two rounds of RDA balanced the optimal degree of subtraction and an acceptable level of redundancy; 72% of the clones were nonredundant (unique), including eight novel genes in round two, whereas only 36% unique clones, including one novel gene, were identified in this pilot screening of round three. Therefore, 3360 clones from

the round two libraries, 2208 from the NS–DC library, and 960 genes from the DC–NS library were gridded at high density to produce the array used in subsequent experiments. A flow diagram of the experimental approach is depicted in Figure 2A.

Microarray Hybridization and Analysis cDNA from NS and DC conditions was cohybridized onto the array, as depicted in Figure 2B. In this figure, two duplicate arrays are shown side by side, with the dyes labeling each amplicon reversed. In both hybridizations, the enrichment obtained after two rounds of RDA is apparent, as each subtraction direction labels preferentially with its appropriate cDNA. Controls consisting

Neuron 328

Figure 3. Data Visualization Confirms Differential Expression Patterns (A) Signal intensities representing background-subtracted gray scale values from experimental (NS versus DC, red) and control (DC versus DC, green) hybridizations are graphed. (B) A 3D contour plot of the upper left array quadrant from the control NS versus NS hybridization, with array locations plotted on the x and y axes and n-fold differential expression (Cy3:Cy5) plotted in the z dimension for each array location. Only a few spots representing noise demonstrate fold differences greater than 2. (C) A contour plot of the same array quadrant shown in (B) from an experimental NS (Cy3) versus DC (Cy5) hybridization. The extent of differential expression is far greater than in the control condition. The DC-enriched clones in the left of the array (rows 1–6, columns 1–24; rows 6–12, columns 1–16) show differential expression in the proper direction (NS:DC ⬍ 1.0), and those in the NS-enriched library on the remainder of the array show expression in the opposite direction (NS:DC ⬎ 1.0), confirming the visual impression in Figure 2. In both (B) and (C), violet indicates an NS:DC ratio ⬎ 1:1, and blue indicates a ratio of 1:1 or less.

of prokaryotic genes, empty vector, and G3PDH are arrayed in the marked area. The reproducibility of hybridizations onto this subtracted array was tested by cohybridization of differentially labeled cDNA samples from the same culture conditions. Figure 3 demonstrates the quality of the data obtained from two such hybridizations: NS versus NS and DC versus DC. The slope of both lines was 1.1, with correlations of 0.98, close to a line with a slope of 1.0 and correlation of 1.0 expected ideally. In the NS versus NS hybridization and DC versus DC hybridizations, the percentage of clones showing intensity ratios greater than 2-fold in the inappropriate direction were 0.72% and 1.5%. Since these false positives were random and filtered out by duplicate experiments, none of the spots derived from the DC–NS subtraction actually showed false differential expression by the criteria used. The coefficient of variance for multiple hybridizations to the same clone was generally low, ranging from 3.0% to 8.9% for signals from 75 elements representing nine clones present at five or more discrete locations on the array, consistent with reproducible hybridizations. Clones demonstrating significant differential expression were sequenced to identify them and to prioritize further characterization. Stringent criteria were used to provide a conservative estimate of clones with the highest degree of differential expression (Experimental Procedures). The average ratio of NS:DC enrichment of these clones by microarray analysis was 3.3-fold, with a range between 1.5 and 7.2. BLAST searches were done using sequence in the FASTA format via the NCBI BLAST web site (Experimental Procedures). Contigs or clusters representing individual genes were assembled using the SeqMan II program (DNAstar). A total of 354 clones met criteria for differential expression, but 132 clones had either double inserts or gave consistently poor sequence and were eliminated from further analysis. Eight of the double inserts were not eliminated,

because they were identified in more than one clone (Experimental Procedures). A total of 86 unique contigs representing 79 individual gene clusters were represented by the 232 differentially expressed sequenced clones from this array, 19 (25%) of which were without significant homology to known genes within gene and protein databases (novel). Of these 19 clusters, six (30%) lacked any significant expressed sequence tag (EST) homology. Gene clusters were divided into major functional categories based on homology with known genes and are listed in Table 1. Northern Blot Confirmation of Array Data Northern blotting was conducted on 20 of the most differentially expressed clones from the NS–DC condition and on five differentially expressed clones between the 20th and 50th percentile to confirm that the clones identified were not artifacts of the subtraction and microarray hybridization procedure. In all clones, the direction of differential expression was the same as that observed on the microarray, confirming their enrichment in starting NS messenger RNA (Figure 4). Even clones that were not among the most differentially expressed (near or below the 20th percentile) showed the same direction of enrichment on the microarray as on Northern blots, confirming the differential expression of clones in the subtracted libraries (Figure 4). Northern blotting was also conducted in a subset of NS- (n ⫽ 6) or DC-enriched (n ⫽ 2) clones using RNA from C17-2 cells, a pluripotent neural progenitor cell line generated by viral oncogene transformation of cerebellar progenitor cells (Snyder et al., 1992). These cells can differentiate into a variety of neurons and glia when transplanted into the appropriate environment (Snyder et al., 1992). The three novel (see Table 2) and three known NS-enriched genes studied (L3-PSP, cyclin D2, and ERCC-1) were also enriched in C17-2 cells, while the two DC-enriched genes were not highly expressed in C17-2 cells (data not shown).

Genetic Analysis of CNS Progenitors 329

Table 1. Known Genes and Homologs Identified Category Signaling

Cytoskeleton/process outgrowth

Immune related

Translation/protein trafficking

DNA binding/chromosome structure/transcription

Cell cycle



Novel, without significant homology

Gene or Homology a,b

r IGF-BP3 (NM012588) hL-3-Phosphoserine phosphatasea,b (NM004577) TOPK (MAPKK-like kinase) RAN GTPase maternal leucine zipper kinase (MELK) RAC GTPase-activating protein Rab-2 RagA Fyn protooncogene r cGMP–stimulated phosphodiesterase (RNU21101) stanniocalcin precursor rGTP binding protein (TC10) (AB031482) retinoic acid repressable homolog (RARG-1) GTP-nucleotide binding protein (Gnb2-RS1)a ATP binding cassette (ABGC2)a calpactin I heavy chain T-cadherinb tubulin ␣-6 thymosin ␤-4 cofilin 2b rNeuritinb (RNV88958) TCR-␤ locusb MHC class I region/VL-30a interferon-activated gene 203a,b T cell death–associated antigen EF4A EF2a hPrefoldin subunit 1 (MM30184) rRibosomal subunit S27 (metallopanstimulin 1) X59375.1 acidic ribosomal phosphoprotein ubiquitin-like/SMT3A,3B nuclear poly(A)-binding protein brain- and testes-enriched nuclear riboprotein D-like (JKTBP) hUbiquitin 2-like gene (NM013444) chaperonin subunit 5 (CCT5) chaperonin subunit 2 (CCT2)a histone H2A.Z (H2afz) GCN5 histone acetyltransferase (Gcn5) LimD1 homeobox homolog adapter-related protein complex AP-3,S1 cyclin D1a,b excision repair protein (ERCC1)a,b hcyclin D2a (NM001759) putative steroid dehydrogenase (KIK-1)b tricarboxylate carrier homolog (S70011) phosphoglycerate mutase type B subunit stearoyl-coenzyme A desaturase 2 cytochrome c oxidaseb glucose phosphate isomerase cystathione ␤-synthesase neutral amino acid transporter alchohol dehydrogenase 5 (adh5) h prostate tumor over-expressed gene 1 (AF237709) rMG87 (AF095741) translationally related transcript-1a hSmall fragment nuclease CGI-114 (NP056338)a ␣ enolase h mitogen-inducible gene 2 (Z24725) arsenite-translocating ATPase brain-specific TGF-␤-regulated sequence (S65035) 19 additional clones

Genbank Accession Number BF642825, BF642812 BF642826 AB041882 NM009391 NM010790 NM012025 MM21905 X85183 NM008054 BF642834 AF099098 BF642838 AB041566 NM008143 AF140218 NM007588 AB02210 NM009448 MMTHYB4 NM007688 BF718801, BF18802 MMAE000664 MMHC322F16, AF111103 NM008328 NM009344 NM013506 U89416 BF642804 BF642830 NM007475 AF067824 AF113011 NM016690 BF642814 MM037637 AB041570 NM016750 AF254441 NM013860 NM009681 NM007631 NM007948 BF642831 AF064635 BF718799 S63233 NM009128 MUSMTCG MUSGPI, BF642819 U43721 NM018861 NM007410 BF718800 BF642818 NM009429 BF642816 X552379.1 BF642821 AF039405.1 BF718803

Accession numbers after the gene name are the nonmouse homolog gene accession numbers (human or rat). The accession numbers starting with BF are the accession numbers for the putative mouse homolog whose sequence was determined in this study. a Clones enriched ⬎2.0-fold in HSC versus WBM. b Gene clusters comprised of more than one clone that represent different nonoverlapping regions of the gene.

Neuron 330

Figure 4. Northern Blot Confirmation of Differential Expression (A) Sample Northern blots from the gene clusters among the top 20% of differentially expressed NS–DC clones (n ⫽ 12) are depicted in the top row, as well as five clones between the 20th and 50th percentiles but still putatively enriched in NS cultures are depicted in the bottom row. (B) Two DC–NS clones randomly chosen are shown for comparison. In each blot, the first column contains 10 ␮g NS total RNA (NS) and the second column contains 10 ␮g DC total RNA (DC). The microarray spots are shown below each Northern for illustration.

In Situ Hybridization Reveals Distinct Expression Patterns Cellular proliferation within the CNS takes place within restricted neuroepithelial germinal zones, and we expected that genes important in CNS progenitor and stem cell functions would be expressed in these zones. Therefore, we performed screening in situ hybridization onto tissue sections from developing mouse at a variety of stages ranging from early embryonic day 13 to adult, using clones identified by microarray screening as being differentially expressed (Tables 2 and 3; Figures 5 and 6). These screening in situ hybridization studies demonstrated that virtually all of the genes studied, novel and previously identified, were expressed within the germinal zones of the developing mouse brain. Figures 5A–5F depict several examples of clones that demonstrated highly selective hybridization to germinal zones with little expression outside the neuroepithelium. In the adult,

germinal zone expression was most prominent within the rostral forebrain, a site of ongoing neuronal production (Luskin, 1993). Several of these ventricular zone (VZ)- or subventricular zone (SVZ)-enriched genes depicted in Figure 5 appear novel, including A1C5, A19G2, A16F10, and A20F10. A19G2, a novel gene with regions of high homology to a human ubiquitin-like gene product, and A1C5 (no homologies) showed restricted expression within the ventricular zone of the germinal neuroepithelium between E13 and P0 and very low levels of germinal zone and dentate gyrus expression in the adult (Figures 5C and 5E). The dentate is a region of stimulated neurogenesis in the adult, indicating the presence of progenitors in these regions (Parent et al., 1997; Eriksson et al., 1998; Gage et al., 1998; Kornack and Rakic, 1999). Of the 19 novel clones, 15 were studied by in situ hybridization (Table 2). Virtually all of the corresponding mRNAs were

Table 2. Summary of Novel Clones’ Expression Patterns Genbank Accession Number BF642824 BF642806 BF642823 BF642833 BF642817 BF642809 BF642825 BF642829 BF642837 BF642805 BF642822 BF642811 BF642832 BF642835 BF642807 BF642813 BF642836 BF642808 BF642810

Intensity of Germinal Zone Expression Contig (Clone ID) c

1 (A1C5) 5b,c (A19G2) 78 (A15D4) 29b,c (A16F10) 69 (A19D7) 54 (A18E1) 9 (A21B2) 23 (A3G4) 39 (A9C9) 47 (A14F11) 77 (A21B8) 59b (A20C3) 27b (A4G2) 35 (A7G8) 51 (A16E6) 60 (A23E6) 38 (A8G2) 53 (A18E1) 57 (A19H3)



⫹⫹⫹ ⫹⫹ ⫹⫹⫹ ⫹⫹ ⫹ ⫹⫹ ⫹⫹ ⫹⫹ ⫹⫹⫹ ⫹⫹ ⫹⫹ ⫹⫹⫹ ⫹ ⫹ ⫹ NA NA NA NA

⫹⫹ ⫹⫹⫹ ⫹⫹ ⫹⫹⫹ ⫹ ⫹⫹ ⫹⫹ ⫹⫹ ⫹⫹ ⫹ ⫹ ⫹ ⫹ ⫹ NA NA NA NA

Extent of Extragerminal CNS Expressiona

⫹ ⫹⫹ ⫹⫹⫹ ⫹ ⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹ ⫹⫹ ⫹ ⫹⫹⫹

Relative intensity of germinal zone expression is indicated as strong (⫹⫹⫹), moderate (⫹⫹), light (⫹), and absent. a Extent of extragerminal zone expression is graded from very limited focal expression (⫹), expression in more than one other brain area (⫹⫹), and relatively widespread expression (⫹⫹⫹). Clones from which in situ data is not available are indicated (NA). VZ, ventricular zone; SVZ, subventricular zone. b Also enriched in HSC relative to WBM. c Appears enriched in C17-2 cells by Northern analysis.

Genetic Analysis of CNS Progenitors 331

Table 3. Expression Pattern of Known Genes Intensity Germinal Zone Expression

Extent of Extragerminal CNS Expressiona

Gene Identity or Homology

Contig (Clone)

Genbank Accession Numbers



PGAM1 (phosphoglyxerate mutase)b CAL 1H-(calpactin 1 heavy chain)b hCyclin D2b TDAG (T cell death–associated gene)b Cyclin D1b h small fragment nucleaseb TOPKb hL3 phosphoserine phosphataseb MELK (maternal embryonic leucine zipper kinase)b SLC1A4 (neutral amino acid transporter) AP3 S1 (adaptor protein–related complex) TC10 (GTP binding protein) VL-30 ERCC-1 rMG87 CCT5 rIGFBP 3 GNB2-RS1 (guanine nucleotide binding protein)

87 30 26 15

(A1G6) (A8G4) A9H8 (A6G2)

AF283667 NM007585 NM001759 NM009344

⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹

⫹⫹ ⫹⫹ ⫹

⫹ ⫹ ⫹ ⫹

13 (A1G3) 67 (A20F10) 8 (A2D2) 17 (A10A3) 21 (A3F8)

NM007631 NP056338 AB041882.1 NM004577 NM010790

⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹ ⫹⫹⫹

⫹⫹ ⫹⫹ ⫹⫹ ⫹ ⫹⫹⫹

⫹ ⫹

61 (A5B10)




55 (A18H10)





40 (A11E2) 22 (A1A1) 50 (A1G11) 7 (A1A5) 2 (A1H1) 16 (A1H4) 75 (A15B6)

AB031482.1 AF111103.1 NM007948 AF095741.1 NM007637 NM012588 NM008143

⫹⫹ ⫹⫹ ⫹⫹ ⫹⫹ ⫹ ⫹ ⫹

⫹⫹ ⫹⫹ ⫹⫹ ⫹ ⫹ ⫹ ⫹

⫹⫹⫹ ⫹⫹⫹ ⫹⫹ ⫹ ⫹ ⫹⫹ ⫹⫹⫹

Relative intensity of germinal zone expression is indicated as strong (⫹⫹⫹), moderate (⫹⫹), light (⫹), and absent. Genbank accession numbers represent clone rat or human homolog. The new mouse Genbank IDs are listed in Table 1. a Extent of extragerminal zone expression is graded from very limited focal expression (⫹), expression in more than one other brain area (⫹⫹), and relatively widespread expression (⫹⫹⫹). VZ, ventricular zone; SVZ, subventricular zone. b Restricted neuroepithelial expression indicates potential progenitor markers.

expressed within the VZ or SVZ, and many were relatively restricted to these regions during development, as described above, suggesting roles in progenitor proliferation and/or differentiation. The relative distribution within these germinal zones was not homogenous. Several of the genes were highly expressed within the VZ (cells situated immediately adjacent to the ventricles) and showed little expression in the SVZ—a specialized proliferative zone that appears during late embryogenesis. However, this pattern was not observed with all clones. For example, A16F10, a novel gene with homology to an EST from an embryonic cancer cell line, had relatively high expression levels in the SVZ without significant VZ expression early in development. In the adult, expression is restricted to a narrow band of subependymal cells (Figures 5A and 5G–5J). Neurosphere expression was quite focal as well; only a subset of neurosphere cells expressed A16F10, consistent with enrichment in a subset of progenitors (Figures 5L and 5M). These observations support the hypothesis that progenitors within these two regions of the germinal epithelium are genetically distinct. Several known genes not previously associated with early neural precursor proliferation also showed restricted germinal zone expression. A summary of the expression patterns of the 18 known genes studied by in situ hybridization are listed in Table 3. Virtually all of these genes were expressed within germinal zones and about half demonstrated restricted germinal zone expression, with little labeling outside these regions. Figure 5 shows the expression patterns of two of these

clones. A1G11 is a known gene involved in cell cycle and DNA repair, the excision repair cross complementing gene (ERCC-1), while A1A9 shows high homology to the human gene L3 phosphoserine phosphatase (L-3-PSP; Collet et al., 1997) and thus is likely to be its mouse homolog. Clone A1A9, mL-3-PSP, showed expression restricted to the mouse telencephalic VZ and SVZ at E13 and was not expressed in the adjacent brain containing nonproliferating cells (Figure 5D). At later stages, its expression remained restricted to the VZ, and, accordingly, expression levels decreased, coincident with the diminution in VZ size during development. A1G11 (ERCC-1) and A9H8, which is homologous to the 3⬘ untranslated region of human cyclin D2 (data not shown), were also expressed in the dentate gyrus in the adult. The expression patterns of these known genes suggest important roles in neural progenitor proliferation that should be investigated in future studies. Another pattern identified in several clones, representing discrete gene clusters, was labeling of a small subset of cells at E13, during the time when neurons destined for the deep cortical layers have undergone or are undergoing their final mitoses. One example of this is A1E2, which corresponds to mIGFBP-3 and was expressed in a subset of germinal zone cells at E13 (Figures 6A and 6B). As cortical neurogenesis and migration proceeded, labeling became more abundant but remained restricted largely to the region of the cortical anlage containing layer 5 and 6 neurons (Figure 6B). In the adult, expression was restricted to a subset of hippocampal, thalamic, and deep cortical layer neurons.

Neuron 332

In contrast with NS–DC-enriched genes shown, the few DC-enriched genes studied by in situ hybridization labeled regions outside of the VZ. One of these, a gene of unknown function (L9), is shown in Figure 6C to illustrate this contrasting pattern. Highest levels of expression in the adult were observed in a laminar pattern within the cerebral cortex and in several midline subcortical structures but not the basal ganglia. No hybridization was detectable within the proliferative germinal zones. The pattern of expression of this class of genes is not consistent with a role in precursor proliferation but rather a role in specific postmitotic neuronal populations, as one would predict a priori based on its enrichment in DC cultures. Hematopoietic Stem Cell–Enriched Genes It has been reported that clonogenic cells contained within mouse neurospheres maintain the potential to reconstitute the bone marrow in sublethally irradiated mice (Bjornson et al., 1999), raising the issue of whether hematopoietic stem cells (HSC) and neural stem/progenitor cells could share a subset of common marker genes. To investigate the overlap of genes identified in this screen with HSC, the microarray was cohybridized with cDNA amplified from HSC and whole bone marrow (WBM). A subset of clones showing enrichment in HSC relative to WBM in addition to NS enrichment was identified (Table 1; Figure 7). Several of these clones discussed above, including A1A9 (L-3-PSP), A16F10, A1G11 (ERCC-1), A9H8 (cyclin D2), and A19G2, were also found to be largely restricted to germinal zones in the developing and adult CNS (Figure 5; data not shown), consistent with a role in progenitor function. Discussion We have performed a genomic-level expression analysis of neural progenitor proliferation and differentiation. We have used RDA and microarray technology coupled to anatomical screening to characterize gene expression patterns in pluripotent proliferating neural progenitor cells relative to cells at later stages of differentiation. Because of the highly ordered process of CNS development, the determination of where and when a gene is expressed has direct implications for the range of potential functions of that gene during development. The pro-

Figure 5. Expression of Identified Genes in the Germinal Zones of Mouse Brain (A–F) Bright-field images of film autoradiographs of the clones shown. The sections shown were taken from E13 (sagittal) and E17 (coronal) mouse embryos, P0 (coronal) pups, or adults (coronal) hybridized to 35S-labeled antisense cRNAs. ([E] and [F] omit the adult sections). Note the expression of all six mRNAs in the germinal zones adjacent to the ventricular systems (arrows in [A] and [E]). Some genes, as shown for A16F10 in (A), were also expressed in the adult hippocampus (hpc; arrow). (G–J) Emulsion-dipped autoradiograms demonstrating more detailed expression patterns of A16F10 at the ages shown. (G) Low-power dark-field image of an E13 sagittal section demonstrating silver grains in the striatal germinal zone (GZ), with limited expression elsewhere within the forebrain. The labeling on the external surface of the cortex (ctx) is extracerebral.

(H) Labeling in the subventricular (SVZ) portion of the germinal zone of the striatum and the septum (sept.) on E17. The lateral ventricle (Lat. Vent.) is identified for orientation. (I) Low-power image of labeling in the adult subependymal zone (arrow). Note the inhomogeneity of the labeled cells. (J) High-power bright-field view of subependymal labeling (arrow). (K) High-power bright-field image of hybridization using the A7C5 cRNA, demonstrating labeling in cells apparently lining the ventricular surface (arrow). (L and M) Low-power bright- (L) and dark-field (M) paired images of dipped autoradiograms from neurosphere sections hybridized to A16F10 cRNA. An individual neurosphere is outlined. Note the inhomogeneous distribution of labeled cells within the neurosphere. The scale bar in (M) ⫽ 5 mm for (A)–(D); 3.3 mm for (E) and (F); 480 ␮m for (G) and (I); 333 ␮m for (H), (L), and (M); and 83 ␮m for (J) and (K).

Genetic Analysis of CNS Progenitors 333

Figure 7. Some NS-Enriched Genes Are Also Enriched in Hematopoietic Stem Cells cDNA from flow-sorted HSC and WBM was amplified using SMART PCR (Clontech), labeled with different fluors, and cohybridized onto the neurodevelopmental array shown in Figure 2. In this false-color image of one array quadrant, red spots are enriched in HSC and green spots in WBM. Four of the gene clusters identified as NS enriched in this quadrant are marked on this figure.

Figure 6. Extragerminal Expression Patterns (A) Low-power images of film autoradiographs taken from sections hybridized to A1E2 (IGFBP-3) cRNA in sections taken on the ages shown. Note the relative lack of expression within the brain on E13, with greater expression in the deeper layers of the neocortex as well as the hippocampus on E17 and P0 (arrow). (B) Bright- (left) and dark-field (right) pairs of emulsion-dipped autoradiographs of slides hybridized to A1E2 cRNA. On E13, a few cells scattered through the SVZ were labeled (arrow). By E17 and through P0, numerous cells within the deeper layers of the developing neocortex were present, and no labeling was observed in the germinal zones. (C) Film autoradiographs of mouse sections hybridized to the antisense cRNA for B7A5—the LB9 gene isolated in the DC–NS subtraction. The sections are of the same ages as shown in (A). The scale bar in the lower part of the image ⫽ 5 mm for (A) and (C); 149 ␮m for the paired emulsion images for E13; and 372 ␮m for the E17 and P0 pairs.

cess of confirming and extending the initial microarray screening experiment using in situ hybridization to study the developmental expression patterns of selected clones thus represents an important way to begin to validate and extend microarray data. In situ hybridization demonstrated that several of these genes are expressed in subsets of cells within neural progenitor cultures, as well as in the germinal neuroepithelium in vivo. Further array analysis revealed that some are also enriched in clonogenic hematopoietic stem cells. These findings are consistent with a role for such genes in regulating precursor proliferation and potentiality or inhibiting differentiation and may indicate their potential for use as markers for stem cells in bone marrow or brain. Methodological Issues One of the major challenges to the study of the molecular basis of cellular diversity in the nervous system during development or normal functioning is the enormous level of cellular heterogeneity (McConnell, 1995; Geschwind, 2000), which leads to a high degree of RNA com-

plexity. Thus, methods for subtraction or expression comparisons that work well in model systems such as cancer cell lines or lymphoid lineage development do not necessarily translate well to the study of the mammalian nervous system. This is primarily because these methods depend upon hybridization. Low-abundance RNA species present in a small percentage of cells may be below the concentration necessary for efficient hybridization, as in a tester-driver hybridization for the case of a subtraction, or to array elements in an expression experiment (Geschwind, 2000). Thus, rare but nondifferentially expressed messages are not often subtracted, because of the low likelihood of hybridization with their complement in the other subtractive pool. RDA, by allowing multiple rounds, can eliminate this noise from nondifferentially expressed, low-abundance mRNAs and offers unprecedented ability to detect rare messages in model systems (Hubank and Shatz, 1994). However, RDA results in the identification of only five to ten genes when all four rounds are conducted (Hubank and Shatz, 1994). The combination of an abbreviated RDA with an efficient screening method, as performed here, provides a balance between subtraction efficiency and the preservation of output complexity. This demonstrates that such an approach can work in a heterogeneous CNS system, either in vitro or in vivo. One of the important methodologic principles used to optimize this technique for the CNS is the use of pilot experiments to empirically determine the best number of RDA rounds to array. Whereas two rounds worked well in the present system, for more complex tissues such as whole brain regions, three or four rounds of RDA are likely to be necessary to achieve an optimal level of subtraction (Geschwind et al., 1998a). Currently, the standard cDNA microarray experiment depends upon gridding down known cDNAs or ESTs (DeRisi et al., 1996; Schena et al., 1996) or genomic clones (Cheung et al., 1998; Geschwind et al., 1998b). This limits the analysis of gene expression to known genes, or ESTs, which currently account for an esti-

Neuron 334

mated 50%–60% of genes expressed in the nervous system (Unigene). Many genes, especially those of low abundance, remain unidentified. So, even in the postgenomic era, efficient methods for identifying and characterizing novel genes of functional importance in the nervous system, such as that presented here, remain of great utility. This approach has the advantage that known and unknown genes of more likely functional relevance to the system of interest are queried for changes in expression. A further practical demonstration of the relevance of the focused subtracted array approach described here can be derived from the percentage of array elements that are construed as “hits” (signals above a background threshold). In a typical array experiment, between 25% to 60% of the array elements are hits, whereas more than 90% were consistently above threshold on the NS–DC array used here. So, this subtracted array has the dual advantage that it contains genes enriched in a system of interest and that many of these genes are novel. The latter advantage will become less of a factor in the future, as genome annotation progresses and the majority of known genes are identified. The use of these custom neurosphere-derived arrays for expression analysis beyond subtraction screening presents another opportunity for the innovative application of microarray screening technology. Analysis of HSC and WBM on the neurodevelopmental subtracted array led to the identification of a number of genes enriched in neural progenitors and HSC, providing a valuable application of the NS-derived array. The use of RDA amplicons also demonstrates that PCR-amplified cDNA (prior to plateau) probes can be reliable for microarray studies and may provide a reasonable approach for the analysis of small samples. One disadvantage of our approach is that some of the novel clones may only represent the uncharacterized 3⬘ untranslated regions of known genes rather than novel genes. However, this is a problem with any 3⬘ EST-based array, which can be resolved by further study of individual genes, such as was done with the clone representing the putative mouse cyclin D2. In addition, two novel, nonoverlapping contigs (representing clones A19G2 and A16F10) had homology to ESTs within diverse regions of a putative rab-2 Unigene cluster (www.ncbi.nlm.nih.gov/UniGene/) but no homology to rab-2 cDNA. At this point, this may represent an idiosyncrasy of Unigene, since both of these two novel clones have high homologies to other nonoverlapping ESTs not contained within this Unigene cluster. This kind of discrepancy should be resolved as further Unigene builds continue. Known Genes Confirm Previous Studies and Implicate Potential Relevant Functional Pathways A number of the clones identified represent known genes or their probable mouse homologs, such as ERCC-1, mIGFBP3, cyclin D1, cyclin D2, thymosin ␤-4, mfyn, TOPK (MAPKK-like), a retinoic acid repressable protein (RARG-1), a maternal embryonic leucine zipper kinase (MELK), and the Rac GTPase-activating protein 1, as examples. The assertion that some of the previously undescribed genes may regulate progenitor prolifera-

tion and differentiation is supported by the identification of known genes from signaling pathways previously shown to be involved in neural progenitor proliferation (Learish et al., 2000; Takeuchi et al., 2000) or glial differentiation (Davy et al., 1999; Osterhout et al., 1999). Identifying cell cycle control and DNA repair elements in neural and glial progenitors has relevance beyond basic notions of proliferation in nonneural systems (Rakic, 1995; Haydar et al., 1999). Regulation of the cell cycle in neural precursors is a critical step in cerebral cortical regionalization and pattern formation and is likely important in the evolutionary emergence of new brain regions (Dehay et al., 1993; Rakic, 1995; Polleux et al., 1997; Kornack and Rakic, 1998). The specific restricted patterns of expression demonstrated by in situ hybridization provides further evidence against general or housekeeping roles for many of these NS-enriched cell cycle–related genes, such as cyclin D1 and cyclin D2. The restriction of cyclin D2 to specific embryonic brain regions is consistent with a specialized role in regional cell cycle control and neural development (Ross et al., 1996; Meyyappan et al., 1998). In the same vein, finding that a putative DNA repair enzyme–coding mRNA, ERCC-1, is expressed preferentially in neurospheres, germinal neuroepithelium, and HSC is especially intriguing, given that the absence of two DNA repair enzymes (Ligase 4 and ERCC-4) involved in the religation step results in defective neuroblast and progenitor lymphocyte development (Frank et al., 1998; Gao et al., 2000). Several of the genes identified may play roles in transcriptional regulation within CNS progenitors, such as GCN5, which links chromatin modification by histone acetylation to transcriptional regulation (Cheung et al., 2000; Sterner and Berger, 2000). GCN5 has also been shown to play an important role in Notch signaling (Kurooka and Honjo, 2000). Since Notch signaling is inhibitory for many differentiation programs, including mammalian CNS progenitors (de la Pompa et al., 1997), GCN5 may also play a more central role in the maintenance of the proliferative, undifferentiated CNS progenitor via direct transcriptional regulation rather than chromosome structure alterations. A Number of Novel Genes Expressed in Germinal Zones Were Identified Several novel genes without significant homologies to other genes demonstrated patterns of expression relatively restricted to the germinal epithelium during development. Others with similar expression patterns (e.g., L-3-PSP) are newly identified, but we did not categorize them as novel, since homologs exist in other species. To further increase their utility as potential markers at different stages in neural development, in situ hybridization was conducted with all novel clones. The majority of clones demonstrated clearly interpretable patterns of expression. A few clones did not produce reliable signals on several attempts, and further characterization must await cloning of the full-length cDNA. It is striking that despite the heterogeneity of the culture system used more than half of novel clones showed germinal zone enrichment, several of which demonstrate exclusive neuroepithelial expression. In addition, expression patterns within the developing neuroepithelium were not

Genetic Analysis of CNS Progenitors 335

uniform; two novel gene clusters were enriched in the subventricular zone. This discrimination between ventricular zone and subventricular zone fits with the emerging functional distinctions between these two contiguous neuroepithelial regions and is most consistent with expression by progenitors of differing potentiality (Doetsch et al., 1997; Reznikov et al., 1997; Lillien, 1998; Haydar et al., 2000). Finally, although this NS–DC subtraction yielded many genes enriched in neural progenitor cultures and germinal zones in vivo, some genes involved in the function of committed neurons were also identified, such as thymosin ␤-4, Rac GTPase, and stanniocalcin (Anand-Apte and Zetter, 1997; Zhang et al., 1998; Roth et al., 1999). Despite their known functions in committed cell types, such genes may also play separate roles in uncommitted or multipotent progenitors. From Genomics to Function This application of genomic-level screening methods to classic problems in neurobiology is a powerful combination likely to gain more widespread use as our knowledge of signaling pathways multiply and the need to rapidly determine a subset of genes for further study in a system of interest intensifies. Such an approach identifies more genes than any one laboratory can study in detail, but the results can provide important resources and research directions for others to follow. In this manner, the expression data and the novel clones identified here provide new resources for those studying progenitor biology and early neural development. Study at the cellular level will be necessary to characterize them in detail. The restricted pattern of expression in the developing nervous system, coupled with enrichment in proliferating stem cell cultures (NS) for some of the genes, is consistent with a role in progenitor cell function, lineage determination, or proliferation. In this regard, the recent demonstration that similarly derived neurosphere stem cell cultures can populate the bone marrow is relevant (Bjornson et al., 1999). Cohybridization of cDNA derived from hematopoietic stem cells and bone marrow onto the neurodevelopmental array demonstrates that several of the genes identified in this study are enriched in purified hematopoietic stem cells as well as CNS progenitor cultures. Several of the clones identified with expression largely restricted to progenitor zones in the embryo and adult were among those enriched in flow-sorted HSC (Tables 1 and 2). Thus, these genes are candidate markers for HSC and neural progenitors. Independent identification of overlapping genes in HSC and neural progenitor cultures suggests that an overlapping genetic program or signaling machinery is shared between these multipotent cell types (our unpublished data). A subset of these shared genes is likely to be important in maintaining proliferative capacity and pluripotent status while inhibiting differentiation in pluripotent progenitors.

flasks in the presence of 20 ng/ml recombinant bFGF (Sigma, St. Louis) in our growth medium containing DMEM/F12 supplemented with B27 (GIBCO–Life Sciences) and penicillin/streptomycin. bFGF was added twice weekly. After 2 weeks, the neurospheres were redissociated and passaged at the same density and repropagated for another 2 weeks (NS conditions). At the end of this second 2 week period, one-half of the cells were plated under differentiating conditions—attachment to poly-L-lysine-coated (10 ␮g/ml; GIBCO– Life Sciences) flasks in the presence of Neurobasal media (GIBCO) supplemented with B27—for 24 hr (DC condition). Sister cultures remained under the NS conditions for the same 24 hr. Floating cells were collected by centrifugation; attached cells were collected by treatment with 0.05% trypsin-EDTA (GIBCO–Life Sciences). Methods used for clonal analysis were adapted from those of Van der Kooy and colleagues (Tropepe et al., 1999), using conditions under which each of the stem cell colonies are derived from a single progenitor (Tropepe et al., 1999, 2000). Free-floating cells (NS) or DC were incubated in trypsin-EDTA, washed by centrifugation, passaged through a 30 ␮m mesh, and resuspended in growth medium supplemented with neurosphere-conditioned medium at the initial density of 2000 live cells/ml. At this density, clusters of cells were not observable after 12 or 24 hr, indicating that subsequent spheres were the progeny of individual progenitors. The conditioned medium was prepared by passing the growth medium from high-density neurosphere cultures through a 0.2 ␮m mesh (to exclude any cells). The conditioned medium was added to the normal growth medium at a 1:1 ratio. The clonal neurospheres were then propagated for 2 weeks and plated on poly-L-lysine-coated coverslips under differentiating conditions for 5 days. Triple-labeling immunohistochemistry was performed with antibodies directed against ␤-tubulin III (TuJ1) to identify neurons (Ciccolini and Svendsen, 1998), glial fibrillary acidic protein (GFAP) to identify astrocytes (Morrison and de Vellis, 1981), and O4 to identify immature and mature oligodendrocytes (Sommer and Schachner, 1981). Secondary antibodies were conjugated with FITC, Texas red, or AMCA to produce a green, red, or blue color, respectively. Each sphere was scored as being mono-, bi-, or tripotent, based on immunostaining. NS and DC RNA Preparation, Northern Blotting, and cDNA Synthesis RNA was extracted using acid phenol extraction (Trizol reagent; GIBCO–BRL), and polyA⫹ RNA was purified from total RNA, using oligo dT paramagnetic beads (PolyAtract–Promega) according to the manufacturer’s instructions. First-strand cDNA was synthesized for 1.5 hr at 42⬚C using 2 ␮g of starting polyA⫹ RNA, 10 ␮g Superscript II reverse transcriptase (200 ␮/ml; GIBCO–BRL), and standard buffers in a volume of 25 ␮l. Second-strand synthesis was completed as previously described (Geschwind et al., 1998a; Welford et al., 1998). For Northern blotting, 10 ␮g of total RNA from NS and DC cultures were electrophoresed on a 1.2% agarose/formaldehyde gel at 5 V/cM for 5 hr. RNA was transferred to a Hybond N membrane (Amersham) overnight by capillary action in 10⫻ SSC buffer and fixed to the membrane by baking for 2 hr at 80⬚C. Gel-purified gene-specific probes were labeled using the Strip-Ease protocol (Ambion). Hybridization and washing were done according to the manufacturer’s instructions (Ambion), and the Phosphor Imager system (Molecular Dynamics) was used to visualize and quantify probe signals.

Experimental Procedures

HSC and WBM Preparation and cDNA Synthesis All HSC were double sorted as described (Morrison and Weissman, 1994; Kondo et al., 1997). About 3 ⫻ 105 double-sorted HSCs from AKR/J mice and 6 ⫻ 105 WBM cells were used for total RNA isolation and SMART first-strand cDNA synthesis (Clontech). Sixteen cycles of PCR-amplification were conducted to amplify the HSC and WBM cDNA, prior to the plateau phase of PCR. Both HSC and WBM were treated identically to minimize potential bias introduced by PCR.

Cell Culture and Immunohistochemistry Neural progenitor colonies (neurospheres) were prepared from neonatal mouse neocortex as described previously (Kornblum et al., 1999; Tropepe et al., 1999). Cells were mechanically dissociated and suspended at a density of 100,000 cells/ml in 25 cm2 culture

Representational Difference Analysis Neural progenitor cultures (NS) were compared to cells differentiated (DC) for 24 hr. Amplicons were created using 1 ␮g of Dpn II–digested cDNA from NS and DC ligated to RBgl RDA adapters (RBgl24, 5⬘-AGCACTCTCCAGCCTCTCACCGCA-3⬘; RBgl12, 5⬘-GAT

Neuron 336

CTGCGGTGA-3⬘) under standard conditions (Hubank and Schatz, 1994). Each cDNA pool was independently PCR amplified for 16–18 cycles (prior to plateau phase) with adaptor-specific primers to create starting amplicons in the following reaction mixture: 2 ␮l cDNA/ RBgl ligation, 20 ␮l 10⫻ PCR buffer (Qiagen), 32 ␮l 25 mM MgCl2 (Qiagen), 2.5 mM dNTP mix, 20 ␮l RBgl24 primer (2 mg/ml), and 1 ␮l Taq DNA polymerase (Qiagen). Thermal cycling was performed (MJ Research) for 72⬚C ⫻ 5 min, 18–22 cycles of 95⬚C ⫻ 1 min, and 72⬚C ⫻ 3 min, followed by 72⬚C ⫻ 10 min. Twelve reactions were performed for each cDNA and pooled to produce adequate material for three rounds of RDA. DNA was purified by phenol/chloroform extraction, ethanol precipitated using sodium acetate, and resuspended in a total of 200 ␮l dH2O. Amplicons were compared on a 1.2% TAE agarose gel and digested with excess Dpn II to cleave RBgl adapters. Amplicons are separated from adapters using the Microcon 100 spin concentrator (Millipore) to make driver. The use of Microcon spin columns increased the yield and was more efficient than gel purification, which is typically used. One microgram was ligated to the JBgl adapters to form the first-round tester (JBgl24, 5⬘-ACCGACGTCGACTATC CATGAACA-3⬘; JBgl12, 5⬘-GATCTGTTCATG-3⬘). The RDA was performed for three rounds, as per Hubank and Shatz (1994), with modifications described by Welford et al. (1998), the most important of which is the elimination of the mung bean nuclease digestion step. Additionally, all primers were HPLC purified. JBgl-tester amplicon (40 ␮l; 0.01 mg/ml) was mixed with 40 ul of driver amplicon (0.5 mg/ml) and phenol/chloroform extracted, ethanol precipitated, rinsed two times in 70% ethanol, and dried completely. Buffer (4 ␮l) containing 3% 1 M EPPS (pH 8.0) (Sigma) and 3 mM EDTA was added to thoroughly resuspend the DNA, and the reaction was overlaid with 40 ␮l of mineral oil. The DNA was denatured at 96⬚C for 4 min, 1 ␮l of 5 M NaCl was added, and the mixture incubated for 24–48 hr at 67⬚C. The oil was removed, and the hybridization mixture was serially diluted in 5 ␮l of TE, 15 ␮l TE, and 80 ␮l of dH2O. The tester-driver hybridization mixture was differentially amplified using two successive PCR reactions without the intervening mung bean digestion step (Welford et al., 1998). The PCR reactions were purified and diluted 1/10 for the second round of PCR. Again, PCR reactions were done in duplicate for 16–18 cycles of 95⬚C ⫻ 1 min, 70⬚C ⫻ 3 min, followed by extension at 72⬚C ⫻ 10 min. Primary PCR (10 ␮l) was added to the secondary PCR reaction (as above) to a total volume of 100 ␮l per reaction. Round two was started by overdigestion of the PCR 2 products with dpn II to remove the JBgl and permit ligation of new adapters NBgl (NBgl24, 5⬘-AGGCAACTG TGCTATCCGAGGGAA-3⬘; NBgl12, 5⬘-GATCTTCCCTCG-3⬘). This NBgl-adapted cDNA served as round two tester, while the driver remained the same. Steps two and three were repeated using a higher driver-tester ratio (500:1) and the NBgl24 primer rather than the JBgl24 primer for PCR amplification. Round three hybridization was conducted with a 4000:1 driver to tester ratio, using JBgl primers for amplification. These round two and three ratios are lower than typically suggested (Hubank and Shatz, 1994), so as to facilitate the identification of genes within the lower ranges of differential expression. A detailed RDA and microarray screening protocol is available by request from the authors (D. H. G.). Library Construction Libraries were made from each direction of subtraction: neurospheres (NS) minus differentiated cells (DC) and DC minus NS. In brief, the RDA subtraction products (post-PCR two) were Dpn II–digested, shotgun cloned into BamH1-digested Bluescript sK II⫺, transformed into Escherichia coli, and plated at medium density on agar containing ampicillin at a concentration of 100 mg amp/ml. Microarray Fabrication Pilot experiments were conducted to optimize the number of unique products while maximizing subtraction efficiency. Clones (192 each from round two and round three RDA libraries) were randomly picked and arrayed as described below. (All subsequent array hybridization and image analysis was done as described below.) Fluorescently labeled cDNA from the NS and DC amplicons were cohybridized onto the array to identify differentially expressed clones. Approxi-

mately 70% of the round two clones were differentially expressed, as compared with more than 90% of the round three clones. To determine redundancy, the top 30 most differentially expressed clones from the NS–DC libraries from each round were sequenced. Clones occurring only once were considered unique, while those occurring two or more times were considered redundant. Compared with only 36% of round three clones sequenced, 72% of round two clones were unique (nonredundant). Thus, round two appeared to have the least redundancy and was chosen for large-scale arraying and analysis. Clones were gridded at 300 ␮m intervals, using a custom-built microarrayer onto an aminoalkyl silane–coated glass surface by capillary action. Negative control samples (prokaryotic genes and empty vector) and G3PDH were also gridded onto the array. The slides with the arrayed DNA fragments were hydrated briefly, snap dried, and UV cross-linked with 400 mJ. Prior to hybridization, the slides were denatured in 100⬚C water, rinsed in ethanol, and air dried. Microarray Hybridization and Analysis Equal amounts of between 1 to 2 ␮g of cDNA from starting amplicons or cDNA from NS and DC were separately labeled with either Cy3 or Cy5 dCTP by random hexamer labeling with Klenow fragment under standard conditions, except that dCTP was reduced to 50 ␮M, dCTP-Cy3 (or Cy5) was added to achieve a final concentration of 60 ␮M, and the reaction was conducted for 12–18 hr. When amplicons were used, care was taken to use PCR-amplified cDNA products obtained before the plateau phase of PCR, which was typically 14–16 cycles in these experiments. The probes were mixed, precipitated, resuspended in hybridization buffer, and cohybridized for 12–18 hr in a humidified chamber, rinsed, and dried. The slides were then scanned in a Genetic Microsystems 418 microarray scanner, with Cy3 and Cy5 signals from hybridization to G3PDH control spots normalized to a ratio of 1. In a typical experiment, ⵑ95% of NS–DC library clones generated signals ⬎2.0-fold above background and were considered hits. The images were analyzed using ImaGene 3.0 (Biodiscovery, Santa Monica) and IP lab (Scanalytics). Prior to quantitative analysis, hybridization signals for each probe were normalized using average Cy3/Cy5 intensity ratios of G3PDH. Hybridizations were done in duplicate and repeated with the fluorophores reversed. Quantified intensity data was downloaded into Genespring 3.0 (Silicon Genetics), a package containing advanced tools for data visualization and analysis. To prioritize clones for downstream analysis, the most differentially expressed clones were identified, showing significant differential expression using the following conservative criteria in at least two of three replicate experiments: (1) percent NS/(DC ⫹ NS) signal ratio in the top 20% of NS:DC expression ratios and (2) ratio of NS:DC signal ⬎1.5. Clones with double inserts identified by sequencing were eliminated from further analysis, except when two or more segments of the same gene were identified in different clones, which occurred for 8 of the 79 gene clusters identified. Sequencing and Analysis Differentially expressed clones were sequenced using M13F and M13R primers. To eliminate repeating clones and to avoid resequencing clones present at multiple locations on the array, 43 clones identified in the pilot experiment were hybridized onto the array. Contigs were assembled using Seqman II (Lasergene), and unique gene clusters were identified and redundant sequences eliminated. The term “gene cluster” refers to all of the sequences that are thought to represent a single unique gene analogous to Unigene clusters (NCBI). For seven separate gene clusters, clones representing two or more different nonoverlapping segments of the same gene were identified (Table 1). Protein databases, GenBank, CDS translations, PDB, SwissProt, PIR, and PRF, and the nucleotide sequence databases, GenBank, EMBL, DDBJ, dbEST, and PDB, were searched using Blastn and Blastx algorithms. In Situ Hybridization Screening in situ hybridization was performed using slides containing multiple samples derived from CD-1 mice. Expression patterns were determined by analysis of autoradiograms in most cases,

Genetic Analysis of CNS Progenitors 337

but, when available, developed slides were evaluated to provide cellular level resolution. Embryos, heads, or brains were frozen in isopentane and then sectioned on a cryostat (20 ␮m). One series of slides contained two sagittal sections through entire E13 embryos, two coronal sections through E17 heads, and two coronal sections through P0 brain. An additional series of slides contained two coronal sections through adult brain and sections through bFGF-derived cortical neurospheres. To obtain the neurosphere sections, cultures of neurospheres (grown under NS conditions) were centrifuged at low speed. The neurospheres were transferred to a microcentrifuge tube and frozen in M1 mounting medium (Lipshaw) and then sectioned on a cryostat at 20 ␮m thickness. After sectioning, all slides were postfixed in 4% paraformaldehyde, rinsed in phosphate buffer, and then stored at ⫺75⬚C until use. Hybridization was carried out as described previously (Kornblum et al., 1994), using [35S]riboprobes transcribed in the presence of T3 or T7 RNA polymerase using the appropriately restricted Bluescript plasmid containing the gene fragment of interest. The sense probe served as a control. After hybridization, sections were exposed to emulsioncoated film (␤-max, Amersham) for up to 7 days and then dipped in Kodak NTB2 emulsion, prior to development. Film images were digitized using the MCID image processing system (St. Catherines, Ontario), and slides were photographed with an Olympus Digital Camera attached to an upright microscope. All images were processed for size, labeling, and optimal visualization of actual hybridization, using Adobe Photoshop 5.0. All images of sections hybridized with a particular probe were processed to the same extent in order to maintain the ability to compare images from different-aged animals. Acknowledgments The authors thank M. Aldimassi and the UCLA DNA microarray facility for microarray fabrication, Dr. Evan Snyder for the gift of C17-2 cells, and the Mental Retardation Research Center Media Core for help with figures (Carol Gray). We acknowledge support from the National Institutes of Health (NIH), the National Institute of Mental Health (NIMH) grant MH60233 (D. H. G.), NIMH grant MH062800 (H. I. K. and D. H. G.), National Institute of Neurological Disorders and Stroke (NINDS) grant NS28383 (H. I. K.), and a National Institute of Child Health and Human Development (NICHD)/UCLA Child Health Research Center Award P30HD34610 (H. I. K.) and thank the Ron Shapiro Charitable Foundation for their generous support (D. H. G. and H. I. K.). M. C. E. is supported by the Training Program in Neural Repair (NS07449), J. D. D. is supported by a Howard Hughes Medical Institute Graduate Fellowship, and R. L. J. is supported by the NINDS training program in Neurorehabilitation (NS07479-08). Received October 26, 2000; revised January 22, 2001. References Anand-Apte, B., and Zetter, B. (1997). Signaling mechanisms in growth factor-stimulated cell motility. Stem Cells 15, 259–267. Arar, C., Ott, M.O., Toure, A., and Gacon, G. (1999). Structure and expression of murine mgcRacGAP: its developmental regulation suggests a role for the Rac/MgcRacGAP signalling pathway in neurogenesis. Biochem. J. 343, 225–230. Bjornson, C.R., Rietze, R.L., Reynolds, B.A., Magli, M.C., and Vescovi, A.L. (1999). Turning brain into blood: a hematopoietic fate adopted by adult neural stem cells in vivo. Science 283, 534–537. Cameron, H.A., and McKay, R. (1998). Stem cells and neurogenesis in the adult brain. Curr. Opin. Neurobiol. 8, 677–680. Cheung, V.G., Gregg, J.P., Gogolin-Ewens, K.J., Bandong, J., Stanley, C.A., Baker, L., Higgins, M.J., Nowak, N.J., Shows, T.B., Ewens, W.J., et al. (1998). Linkage-disequilibrium mapping without genotyping. Nat. Genet. 18, 225–230. Cheung, P., Tanner, K.G., Cheung, W.L., Sassone-Corsi, P., Denu, J.M., and Allis, C.D. (2000). Synergistic coupling of histone H3 phosphorylation and acetylation in response to epidermal growth factor stimulation. Mol. Cell 5, 905–915.

Ciccolini, F., and Svendsen, C.N. (1998). Fibroblast growth factor 2 (FGF-2) promotes acquisition of epidermal growth factor (EGF) responsiveness in mouse striatal precursor cells: identification of neural precursors responding to both EGF and FGF-2. J. Neurosci. 18, 7869–7880. Collet, J.F., Gerin, I., Rider, M.H., Veiga-da-Cunha, M., and Van Schaftingen, E. (1997). Human L-3-phosphoserine phosphatase: sequence, expression and evidence for a phosphoenzyme intermediate. FEBS Lett. 408, 281–284. Davis, A.A., and Temple, S. (1994). A self-renewing multipotential stem cell in embryonic rat cerebral cortex. Nature 372, 263–266. Davy, A., Gale, N.W., Murray, E.W., Klinghoffer, R.A., Soriano, P., Feuerstein, C., and Robbins, S.M. (1999). Compartmentalized signaling by GPI-anchored ephrin-A5 requires the Fyn tyrosine kinase to regulate cellular adhesion. Genes Dev. 13, 3125–3135. Dehay, C., Giroud, P., Berland, M., Smart, I., and Kennedy, H. (1993). Modulation of the cell cycle contributes to the parcellation of the primate visual cortex. Nature 366, 464–466. de la Pompa, J.L., Wakeham, A., Correia, K.M., Samper, E., Brown, S., Aguilera, R.J., Nakano, T., Honjo, T., Mak, T.W., Rossant, J., and Conlon, R.A. (1997). Conservation of the Notch signalling pathway in mammalian neurogenesis. Development 124, 1139–1148. DeRisi, J., Penland, L., Brown, P.O., Bittner, M.L., Meltzer, P.S., Ray, M., Chen, Y., Su, Y.A., and Trent, J.M. (1996). Use of a cDNA microarray to analyse gene expression patterns in human cancer. Nat. Genet. 14, 457–460. Doetsch, F., Garcia-Verdugo, J.M., and Alvarez-Buylla, A. (1997). Cellular composition and three-dimensional organization of the subventricular germinal zone in the adult mammalian brain. J. Neurosci. 17, 5046–5061. Eriksson, P.S., Perfilieva, E., Bjork-Eriksson, T., Alborn, A.M., Nordborg, C., Peterson, D.A., and Gage, F.H. (1998). Neurogenesis in the adult human hippocampus. Nat. Med. 4, 1313–1317. Ferri, R.T., and Levitt, P. (1995). Regulation of regional differences in the differentiation of cerebral cortical neurons by EGF familymatrix interactions. Development 121, 1151–1160. Frank, K.M., Sekiguchi, J.M., Seidl, K.J., Swat, W., Rathbun, G.A., Cheng, H.L., Davidson, L., Kangaloo, L., and Alt, F.W. (1998). Late embryonic lethality and impaired V(D)J recombination in mice lacking DNA ligase IV. Nature 396, 173–177. Gage, F.H., Ray, J., and Fisher, L.J. (1995). Isolation, characterization, and use of stem cells from the CNS. Annu. Rev. Neurosci. 18, 159–192. Gage, F.H., Kempermann, G., Palmer, T.D., Peterson, D.A., and Ray, J. (1998). Multipotent progenitor cells in the adult dentate gyrus. J. Neurobiol. 36, 249–266. Gao, Y., Ferguson, D.O., Xie, W., Manis, J.P., Sekiguchi, J., Frank, K.M., Chaudhuri, J., Horner, J., DePinho, R.A., and Alt, F.W. (2000). Interplay of p53 and DNA-repair protein XRCC4 in tumorigenesis, genomic stability and development. Nature 404, 897–900. Geschwind, D.H. (2000). Mice, microarrays and the genetic diversity of the brain. Proc. Natl. Acad. Sci. USA 97, 10676–10678. Geschwind, D.H., Loginov, M., Karrim, J., and Nelson, S. (1998a). Finding the differences between the developing cerebral hemispheres using RDA and DNA microarray technology. Soc. Neurosci. Abst. 24, 1003. Geschwind, D.H., Gregg, J., Boone, K., Karrim, J., PawlikowskaHaddal, A., Rao, E., Ellison, J., Ciccodicola, A., D’Urso, M., Woods, R., et al. (1998b). Klinefelter’s syndrome as a model of anomalous cerebral laterality: testing gene dosage in the X chromosome pseudoautosomal region using a DNA microarray. Dev. Genet. 23, 215–229. Gritti, A., Parati, E.A., Cova, L., Frolichsthal, P., Galli, R., Wanke, E., Faravelli, L., Morassutti, D.J., Roisen, F., Nickel, D.D., and Vescovi, A.L. (1996). Multipotential stem cells from the adult mouse brain proliferate and self-renew in response to basic fibroblast growth factor. J. Neurosci. 16, 1091–1100. Haydar, T.F., Kuan, C.Y., Flavell, R.A., and Rakic, P. (1999). The role

Neuron 338

of cell death in regulating the size and shape of the mammalian forebrain. Cereb. Cortex 9, 621–626. Haydar, T.F., Wang, F., Schwartz, M.L., and Rakic, P. (2000). Differential modulation of proliferation in the neocortical ventricular and subventricular zones. J. Neurosci. 20, 5764–5774. Hubank, M., and Schatz, D.G. (1994). Identifying differences in mRNA expression by representational difference analysis of cDNA. Nucleic Acids Res. 22, 5640–5648. Johe, K.K., Hazel, T.G., Muller, T., Dugich-Djordjevic, M.M., and McKay, R.D. (1996). Single factors direct the differentiation of stem cells from the fetal and adult central nervous system. Genes Dev. 10, 3129–3140. Kondo, M., Weissman, I.L., and Akashi, K. (1997). Identification of clonogenic common lymphoid progenitors in mouse bone marrow. Cell 91, 661–672. Kornack, D.R., and Rakic, P. (1998). Changes in cell-cycle kinetics during the development and evolution of primate neocortex. Proc. Natl. Acad. Sci. USA 95, 1242–1246. Kornack, D.R., and Rakic, P. (1999). Continuation of neurogenesis in the hippocampus of the adult macaque monkey. Proc. Natl. Acad. Sci. USA 96, 5768–5773. Kornblum, H.I., Chugani, H.T., Tatsukawa, K., and Gall, C.M. (1994). Cerebral hemidecortication alters expression of transforming growth factor alpha mRNA in the neostriatum of developing rats. Mol. Brain Res. 21, 107–114. Kornblum, H.I., Zurcher, S.D., Werb, Z., Derynck, R., and Seroogy, K.B. (1999). Multiple trophic actions of heparin-binding epidermal growth factor (HB-EGF) in the central nervous system. Eur. J. Neurosci. 11, 3236–3246. Kurooka, H., and Honjo, T. (2000). Functional interaction between the mouse notch1 intracellular region and histone acetyltransferases PCAF and GCN5. J. Biol. Chem. 275, 17211–17220. Learish, R.D., Bruss, M.D., and Haak-Frendscho, M. (2000). Inhibition of mitogen-activated protein kinase kinase blocks proliferation of neural progenitor cells. Brain Res. Dev. Brain Res. 122, 97–109. Lillien, L. (1998). Neural progenitors and stem cells: mechanisms of progenitor heterogeneity. Curr. Opin. Neurobiol. 8, 37–44. Lisitsyn, N., and Wigler, M. (1993). Cloning the differences between two complex genomes. Science 259, 946–951.

Rakic, P. (1995). A small step for the cell, a giant leap for mankind: a hypothesis of neocortical expansion during evolution. Trends Neurosci. 18, 383–388. Reynolds, B.A., and Weiss, S. (1992). Generation of neurons and astrocytes from isolated cells of the adult mammalian central nervous system. Science 255, 1707–1710. Reynolds, B.A., and Weiss, S. (1996). Clonal and population analyses demonstrate that an EGF-responsive mammalian embryonic CNS precursor is a stem cell. Dev. Biol. 175, 1–13. Reynolds, B.A., Tetzlaff, W., and Weiss, S. (1992). A multipotent EGFresponsive striatal embryonic progenitor cell produces neurons and astrocytes. J. Neurosci. 12, 4565–4574. Reznikov, K., Acklin, S.E., and van der Kooy, D. (1997). Clonal heterogeneity in the early embryonic rodent cortical germinal zone and the separation of subventricular from ventricular zone lineages. Dev. Dyn. 210, 328–343. Ross, M.E., Carter, M.L., and Lee, J.H. (1996). MN20, a D2 cyclin, is transiently expressed in selected neural populations during embryogenesis. J. Neurosci. 16, 210–219. Roth, L.W., Bormann, P., Bonnet, A., and Reinhard, E. (1999). betathymosin is required for axonal tract formation in developing zebrafish brain. Development 126, 1365–1374. Schena, M., Shalon, D., Heller, R., Chai, A., Brown, P.O., and Davis, R.W. (1996). Parallel human genome analysis: microarray-based expression monitoring of 1000 genes. Proc. Natl. Acad. Sci. USA 93, 10614–10619. Snyder, E.Y., Deitcher, D.L., Walsh, C., Arnold-Aldea, S., Hartwieg, E.A., and Cepko, C.L. (1992). Multipotent neural cell lines can engraft and participate in development of mouse cerebellum. Cell 68, 33–51. Sommer, I., and Schachner, M. (1981). Monoclonal antibodies (O1 to O4) to oligodendrocyte cell surfaces: an immunocytological study in the central nervous system. Dev. Biol. 83, 311–327. Sterner, D.E., and Berger, S.L. (2000). Acetylation of histones and transcription-related factors. Microbiol. Mol. Biol. Rev. 64, 435–459. Takeuchi, T., Misaki, A., Liang, S.B., Tachibana, A., Hayashi, N., Sonobe, H., and Ohtsuki, Y. (2000). Expression of T-cadherin (CDH13, H-Cadherin) in human brain and its characteristics as a negative growth regulator of epidermal growth factor in neuroblastoma cells. J. Neurochem. 74, 1489–1497.

Luskin, M.B. (1993). Restricted proliferation and migration of postnatally generated neurons derived from the forebrain subventricular zone. Neuron 11, 173–189.

Temple, S. (1989). Division and differentiation of isolated CNS blast cells in microculture. Nature 340, 471–473.

McConnell, S.K. (1995). Strategies for the generation of neuronal diversity in the developing central nervous system. J. Neurosci. 15, 6987–6998.

Tropepe, V., Sibilia, M., Ciruna, B.G., Rossant, J., Wagner, E.F., and van der Kooy, D. (1999). Distinct neural stem cells proliferate in response to EGF and FGF in the developing mouse telencephalon. Dev. Biol. 208, 166–188.

Meyyappan, M., Wong, H., Hull, C., and Riabowol, K.T. (1998). Increased expression of cyclin D2 during multiple states of growth arrest in primary and established cells. Mol. Cell. Biol. 18, 3163–3172. Morrison, R.S., and deVellis, J. (1981). Growth of purified astrocytes in a chemically defined medium. Proc. Natl. Acad. Sci. USA 78, 7205–7209. Morrison, S.J., and Weissman, I.L. (1994). The long-term repopulating subset of hematopoietic stem cells is deterministic and isolatable by phenotype. Immunity 1, 661–673. Osterhout, D.J., Wolven, A., Wolf, R.M., Resh, M.D., and Chao, M.V. (1999). Morphological differentiation of oligodendrocytes requires activation of Fyn tyrosine kinase. J. Cell Biol. 145, 1209–1218. Parent, J.M., Yu, T.W., Leibowitz, R.T., Geschwind, D.H., Sloviter, R.S., and Lowenstein, D.H. (1997). Dentate granule cell neurogenesis is increased by seizures and contributes to aberrant network reorganization in the adult rat hippocampus. J. Neurosci. 17, 3727–3738. Phillips, R.L., Ernst, R.E., Brunk, B., Ivanova, N., Mahan, M.A., Deanehan, J.K., Moore, K.A., Overton, G.C., and Lemischka, I.R. (2000). The genetic program of hematopoietic stem cells. Science 288, 1635–1640. Polleux, F., Dehay, C., Moraillon, B., and Kennedy, H. (1997). Regulation of neuroblast cell-cycle kinetics plays a crucial role in the generation of unique features of neocortical areas. J. Neurosci. 17, 7763– 7783.

Tropepe, V., Coles, B.L., Chiasson, B.J., Horsford, D.J., Elia, A.J., McInnes, R.R., and van der Kooy, D. (2000). Retinal stem cells in the adult mammalian eye. Science 287, 2032–2036. Vicario-Abejon, C., Johe, K.K., Hazel, T.G., Collazo, D., and McKay, R.D. (1995). Functions of basic fibroblast growth factor and neurotrophins in the differentiation of hippocampal neurons. Neuron 15, 105–114. Weiss, S., Dunne, C., Hewson, J., Wohl, C., Wheatley, M., Peterson, A.C., and Reynolds, B.A. (1996). Multipotent CNS stem cells are present in the adult mammalian spinal cord and ventricular neuroaxis. J. Neurosci. 16, 7599–7609. Weissman, I.L. (2000). Stem cells: units of development, units of regeneration, and units in evolution. Cell 100, 157–168. Welford, S.M., Gregg, J., Chen, E., Garrison, D., Sorensen, P.H., Denny, C.T., and Nelson, S.F. (1998). Detection of differentially expressed genes in primary tumor tissues using representational differences analysis coupled to microarray hybridization. Nucleic Acids Res. 26, 3059–3065. Williams, B.P., Read, J., and Price, J. (1991). The generation of neurons and oligodendrocytes from a common precursor cell. Neuron 7, 685–693. Zhang, K.Z., Westberg, J.A., Paetau, A., von Boguslawsky, K., Lindsberg, P., Erlander, M., Guo, H., Su, J., Olsen, H.S., and Andersson,

Genetic Analysis of CNS Progenitors 339

L.C. (1998). High expression of stanniocalcin in differentiated brain neurons. Am. J. Pathol. 153, 439–445. GenBank Accession Numbers The GenBank accession numbers for the entirely novel sequences reported in this paper are BF642824, BF642806, BF642823, BF642833, BF642817, BF642809, BF642825, BF642829, BF642837, BF642805, BF642822, BF642811, BF642832, BF642835, BF642807, BF642813, BF642836, BF642808, and BF642810. The GenBank accession numbers for the probable mouse homologs of previously identified genes reported here are as follows: insulin-like growth factor binding protein 3, BF642825 and BF642812 (two contiguous sequences); L 3 phosphoserine phosphatase, BF642826; cGMPstimulated phosphodiesterase, BF642834; GTP binding protein (TC10), BF642838; Neuritin, BF718801 and BF18802 (two sequences); Prefoldin subunit 1, BF642804; ribosomal subunit S27, BF642830; Ubiquitin 2–like gene, BF642814; Cyclin D2, BF642831; Tricarboxylate carrier, BF718799; glucose phosphate isomerase, BF642819; prostate tumor over-expressed gene 1, BF718800; MG87, BF642818; small fragment nuclease CGI-114, BF642816; mitogen inducible gene 2, BF642821; and brain-specific TGF-␤-regulated sequence, BF718803. Note Added in Proof While the current manuscript was in press, we performed parallel experiments using a 9k element more “standard” mouse array. Although the genes identified from the 9K array showed considerable overlap with those described in the current study, many of the genes identified here were not found using the 9K array (about 50%; data not shown). This further attests to the value of the current approach and its complementarity with more standard methods.