Stem Cell-derived Neural Stem/Progenitor Cell Supporting Factor Is an Autocrine/Paracrine Survival Factor for Adult Neural Stem/Progenitor Cells*

Recent evidence suggests that adult neural stem/pro-genitor cells (ANSCs) secrete autocrine/paracrine factors and that these intrinsic factors are involved in the maintenance of adult neurogenesis. We identified a novel secretory molecule, stem cell-derived neural stem/ progenitor cell supporting factor (SDNSF), from adult hippocampal neural stem/progenitor cells by using the signal sequence trap method. The expression of SDNSF in adult central nervous system was localized to hippocampus including dentate gyrus, FLAG epitope tag at the predicted signal sequence cleavage site and constructed rat SDNSF-N (cid:3) FLAG fusion protein. The primer sequences for rat FLAG-SDNSF construction were 5 (cid:3) -GACTAGTCATGGCATCC-CTGCAGCTGCTCAGAGGTCCCTTCCTGTGTGTTCTGCTCTGGGCC-TTTTGTGTTCCTGGTGCCAGGGCCGACTACAAAGACGATGACGA-CAAGCAGGAGCATGGGGCTGGTGTCCACC-3 (cid:3) and 5 (cid:3) -CTACTGCA-GCGACTTGGCAAACTCT-3 (cid:3) . HEK293T cells and COS7 cells were transfected with these expression vectors pEF-SDNSF-C (cid:3) FLAG/His, pEF-SDNSF-N (cid:3) FLAG, or pEF-His-FLAG in serum-free DMEM/Ham ’ s F-12 medium (1:1) with N2 supplement using CellPhect (Amersham Biosciences) and were treated with lysis buffer (150 m M NaCl, 1% Triton X-100, 10 m M Tris-HCl, 1 m M EDTA, 0.2 m M phenylmethylsulfonyl fluoride, 10 (cid:1) g/ml aprotinin). The conditioned medium from these transfected HEK293T cells were filtered and concentrated with Centricon YM-10 (Millipore, Bedford, MA). SDNSF proteins were purified from the concentrated conditioned medium using Ni-NTA-agarose (Qiagen, Valencia, CA). According to the manufacturer ’ s protocol with the slight modification of buffer concentrations as below, the concentrated conditioned mediums from pEF-SDNSF-C (cid:3) FLAG/His and from pEF- His-FLAG transfected HEK 293T cells cultures were purified under native conditions. We mixed 1 ml of the 50% Ni-NTA slurry to 2 ml concentrated conditioned medium and 2 ml of 2 (cid:1) lysis buffer (40 m M imidazole, 100 m M were counted. To assess the differentiation potentials, the newly formed neurospheres were replated onto poly- L -ornithine/laminin-coated Labtek 4-well chamber slides (Nunc) at the density of 10 – 30 spheres/well and cultured in differentiation medium for further 6 days. Under confocal laser microscope or fluorescent microscope, positively stained cells were quantified at least 20 fields systematically across the coverslips from three to four independent experiments of parallel cultures. Analytical Procedures and Data Analysis — Group changes were as- sessed using one-way ANOVA. When statistical differences were obtained at the p (cid:6) 0.01 level between groups, post hoc comparisons were made using the Fisher least squares difference (LSD) test.

Studies carried out in the last few decades have revealed the potential for lifelong neurogenesis in the adult mammalian central nervous system (CNS). 1 At present, discrete regions of the adult brain, the subventricular zone of the forebrain and the subgranular layer of hippocampal dentate gyrus, are known to mediate adult neurogenesis (1)(2)(3)(4), but little is known about signaling that maintains the pool of self-renewing stem cells within these regions. Primary culture of adult neural stem/progenitor cells (ANSCs) is a potent tool to investigate signals controlling adult neurogenesis. ANSCs can be isolated and expanded by means of epidermal growth factor (EGF) and/or basic fibroblast growth factor (FGF-2). Cycling cells can maintain the properties of self-renewal and multipotency (5)(6)(7)(8). The fate of ANSCs is under tight environmental control, and various extrinsic factors to promote lineage commitment have been described (9 -11). ANSCs grafted into the neurogenic or non-neurogenic regions give rise to neurons in a site-specific manner, i.e. only within the neurogenic regions (7,(12)(13)(14). Therefore, we anticipate that cell fate is tightly regulated by specific signaling molecules within neurogenic region or "stem cell niche." However, the maintenance of the stem cell phenotype must also be an important attribute of the stem cell niche, and in vitro culture has provided evidence that stem cells themselves may produce autocrine/paracrine factors that facilitate proliferative self-renewal (7). Glycosylated cystatin C (15) and insulin-like growth factor-I (IGF-I) (16) are two such essential autocrine/paracrine molecules that have been identified as cofactors of FGF-2 and EGF, respectively. To further explore autocrine/paracrine signaling within the stem cell niche, therefore, we used the signal sequence trap method (17,18) to isolate additional novel secretory molecules from ANSCs. Using this cDNA screening method, we have efficiently isolated many secreted molecules with a wide variety of functions (19 -25). In this report, we describe the cloning and characterization of a novel molecule, stem cell-derived neural stem/progenitor cell survival factor (SDNSF), from a cDNA library from rat adult hippocampal NSCs (9). SDNSF is secreted via the classical vesicular export pathway and provides trophic support for ANSC in the absence of mitogenic growth factors. In addition, SDNSF maintains the self-renewal potential and multipotency of ANSCs in the absence of FGF-2. SDNSF-treated ANSCs can proliferate in response to FGF-2 and produce neuronal and glial cell population after differentiation treatment.

EXPERIMENTAL PROCEDURES
Animals and Transient Forebrain Global Ischemia Model-Fisher 344 rats, Sprague-Dawley rats, and C57BL/6 mice were maintained and used for experimentation according to the guidelines of the Kyoto University Animal Research Committee and Stanford University Animal Facility. The animals were deeply anesthetized, sacrificed with sodium pentobarbital, and then dissected immediately or fixed by intracardiac perfusion with 4% paraformaldehyde. For cell cultures, the brains were dissected according to the cell culture protocols as below. For RNA extraction, the tissues are immediately frozen in liquid nitrogen. For in situ hybridization, 50-m-thick coronal sections were prepared from adult female Fisher 344 rat brains after paraformaldehyde perfusion.
Transient global ischemia was induced on male Sprague-Dawley rats (300 -350 g; Charles River Laboratories, Wilmington, MA) by bilateral common carotid artery occlusion and induced hypotension using the modified two-vessel occlusion method (26). In brief, the animals were anesthetized with 1.5% isoflurane, 68.5% nitrous oxide, and 30% oxygen and monitored from the femoral artery with PE-50 catheter (427410; Becton Dickinson, San Diego, CA). After exposure of the right jugular vein and both common carotid arteries, 150 IU/kg heparin was intravenously injected, and blood was quickly withdrawn via the jugular vein. When the mean arterial blood pressure became 30 mm Hg, both common carotid arteries were clamped with surgical clips. Mean arterial blood pressure was maintained at 30 -35 mm Hg for 5 min. After ischemic treatment, the clips were removed, and the blood was reinfused. Body temperature was monitored with a rectal probe and controlled at 37°C. Sham-operated animals underwent exposure of vessels without blood withdrawal or clamping of carotid arteries. All animals were treated in accordance with Kyoto University Animal Research Committee guidelines, Stanford University guidelines, and the animal protocol approved by Stanford University's Administrative Panel on Laboratory Animal Care.
Cell Culture-ANSCs were isolated from adult rat hippocampi and cultured as described previously (6,27). Briefly, hippocampi from adult female Fisher 344 rats were enzymatically dissociated with a papain (2.5 units/ml; Worthington, Freehold, NJ), dispase II (1 unit/ml; Roche Applied Science), and DNase I (250 units/ml; Worthington) solution. A whole digested tissue was then suspended in 50% Percoll solution and fractionated by centrifugation for 10 min at 20,000 ϫ g. Fractionated cells were washed free of Percoll and plated onto poly-L-ornithine/ laminin-coated dishes in DMEM/Ham's F-12 medium (1:1) containing 10% FCS medium for 24 h, and then the medium was replaced with serum-free growth medium consisting of DMEM/Ham's F-12 medium (1:1) supplemented with N2 supplement (Invitrogen) and 20 ng/ml recombinant human FGF-2 (Genzyme, Cambridge, MA). To follow the proliferating single NSC, NSCs were infected by replication-deficient GFP-expressing recombinant retrovirus, LZRS-CAMut4GFP (28). To promote differentiation, the growth medium was replaced with DMEM/ F-12 containing 0.5% FCS, 0.5 M all-trans-retinoic acid (differentiation medium), and ANSCs were cultured for 6 days. For primary culture of neurons and astrocytes, Fischer 344 rat E18 hippocampi for neurons and P2-4 hippocampi for astrocytes were dissected and dissociated by treatment with 0.25% trypsin and trituration with a fire-polished Pasteur pipette. The cells were plated onto poly-L-lysine-coated dishes and cultured in DMEM/Ham's F-12 medium (1:1) supplemented with N2 supplement for neurons and in DMEM with 10% FCS for astrocytes (29). The astrocytes were cultured for 3-6 weeks with two to three passages and 5 mM glutamate treatment to eliminate neurons. Rat embryonic NSCs were isolated from E14 striata described previously with slight modification (30). Dissected striata were enzymatically digested similarly to rat adult NSC isolation and triturated with firepolished Pasteur pipette, and the cells were plated onto noncoated Nunc four-well dishes in the same growth medium as adult NSCs in the presence of FGF-2 at a concentration of 20 ng/ml. COS-7 cells, HEK293T cells, rat glioma C6 cells (31), mouse neuroblastoma N18 cells (32), and human glioblastoma U251 cells (33) were cultured in DMEM containing 10% FCS. Mouse embryonic stem (ES) cell line, R1, were cultured on a feeder layer of irradiated mouse embryonic fibroblasts with daily changed medium containing knockout DMEM, penicillin-streptomycin, 100 M ␤-mercaptoethanol, 2 mM L-glutamine, 100 mM nonessential amino acids (Invitrogen), 15% FCS (HyClone, Logan, UT), and 1,000 units/ml leukemia inhibitory factor (Chemicon, Temecula, CA) (34).
RNA Analysis and SDNSF cDNA Cloning-To isolate novel secretory or membrane proteins from neural stem cells, cDNA library was constructed from cultured ANSCs. Poly(A) ϩ RNA was extracted from ANSCs with TRIzol reagent (Invitrogen) and Oligotex-dT30 Super (Roche Applied Science). The construction of the cDNA library of NSCs and screening by signal sequence trap using yeast was carried out as described previously (18). To obtain full-length SDNSF cDNAs from poly(A) ϩ RNA of ANSC cells and rat brain, 5Ј-and 3Ј-rapid amplification of cDNA ends methods with the Marathon kit (Clontech) were performed. From this sequence data of 5Ј and 3Ј-rapid amplification of cDNA ends methods, the coding sequence of rat SDNSF was confirmed of RT-PCR using a set of primer pairs: 5Ј-GCGTCAGGGGGACGCAG-CTGG-3Ј and 5Ј-GTCAGCTCCGATTGCACAAATACTTGA-3Ј. Coding sequences of mouse and human were obtained from adult mouse brain total RNA and human heart total RNA (Clontech, Palo Alto, CA) with RT-PCR using two sets of primer pairs based on rat sequence data and on expressed sequence tag data from a data base search: 5Ј-GTGCGG-AGAAAAGCGTCCCAG-3Ј and 5Ј-TCCATTTTATTGTCAGATAGCCA-GAGTTCA-3Ј, 5Ј-TGGTGAGGCCCGAGGCGTT-3Ј and 5Ј-TCTTGGGT-ACGTCTTTATCAGCAGCAT-3Ј. The Northern analysis was performed essentially as described (35). Rat SDNSF and rat GAPDH (459 -1001 of GenBank™ accession number M17701) cDNAs were labeled with [␣-32 P]dCTP by random priming. The filter for Northern analysis was prepared from 4 g of poly(A) ϩ RNA prepared as above from following cells and tissues: ANSCs, adult rat brain, heart, lung, liver, spleen, kidney, testis, and skeletal muscle. The filter was hybridized using Quick-Hyb solution (Stratagene, La Jolla, CA). Blotting was analyzed using an image analyzer (BAS 2000, Fuji Film, Tokyo, Japan).
For in situ hybridization, we used riboprobes for the in situ detection of mRNAs derived from rat SDNSF as described previously (36). One-kb cDNA fragments including coding region and 3Ј-noncoding region of the rat SDNSF were generated and subcloned into pGEM-T vector (Promega, Madison, WI). Digoxigenin-labeled sense and antisense rat SDNSF probes were generated with SP6 and T7 polymerases, respectively, using the digoxigenin RNA labeling kit (Roche Applied Science).
For RT-PCR, total RNA was extracted using Trizol reagent from undifferentiated ANSCs, differentiated ANSCs, embryo NSCs, C6, N18, U251 cells; rat whole embryo at 8.5 days postcoitus (dpc) and 10.5; embryonic and postnatal rat brain from 12.5, 14.5, 16.5, and 18.5 dpc and P2, P4, and P7. Equal amounts of samples were subjected to RT-PCR as previously described (25). Each reaction was standardized against a GAPDH control to permit comparison between samples in each PCR. cDNA was generated SuperScript First-strand Synthesis System for RT-PCR (Invitrogen) and amplified by the gene-specific primers for rat SDNSF as above and for rat GAPDH: 5Ј-TGCATCCT-GCACCACCAACT-3Ј and 5Ј-CGCCTGCTTCACCACCTTC-3Ј.
For semiquantitative RT-PCR (37), total RNA was extracted from ischemic hippocampi at two time points, postoperative day 1 and 7 with sham operated control (n ϭ 6/time point), and synthesis of cDNA was performed as above using 1 g of total RNA from each sample. Then cDNA was amplified in 50 l of PCRs that contained 1.5 mM MgCl 2 , 0.2 mM dNTP mixture, 2.5 units of Taq DNA in PCR buffer (Invitrogen), and 0.5 M gene-specific primers using i-Cycler™ (Bio-Rad) between 22 and 40 cycles using SDNSF and GAPDH primers as above. The resultant PCR products were electrophoresed in 1.5% agarose gel and stained with ethidium bromide, the fluorescent bands were scanned, and the volume density of SDNSF was then quantified using NIH Image 1.62. These conditions produced amplicons within the linear exponential phase of the PCR curve. The quantification of SDNSF were normalized with GAPDH by dividing time point band density by that of its matched PCR.
Computer Analysis and Data Base Search-Sequencing was performed with an automated sequencer (model 377A; Applied Biosystems, Foster City, CA). Handling of all the nucleotide and amino acid sequence data and the construction of hydrophobicity profiles were performed with GenetyxMac Version 9.0 (Software Development, Tokyo, Japan). Sequence alignment was executed with ClustalX (38). BLAST and FASTA searches were performed at www.ncbi.nlm.nih.gov/blast/ blast.cgi and www.fasta.genome.ad.jp/ideas/fasta/fasta_nr-aa.html, respectively. A motif search was done at www.motif.genome.ad.jp/. The nucleotide sequences reported in this paper has been submitted to the GenBank™/EBI Data Bank with accession numbers AF475282 (rat SDNSF), AF475283 (mouse SDNSF), and AF475284 (human SDNSF).
Protein Analysis and Immunological Methods-Rat SDNSF was recloned into pEF-6V5-His (Invitrogen) with swapping the original V5 epitope to the FLAG epitope tag at the C terminus (pEF-His-FLAG) to generate FLAG fusion protein, SDNSF-CЈ FLAG/His. We also inserted FLAG epitope tag at the predicted signal sequence cleavage site and constructed rat SDNSF-NЈ FLAG fusion protein. The primer sequences for rat FLAG-SDNSF construction were 5Ј-GACTAGTCATGGCATCC-CTGCAGCTGCTCAGAGGTCCCTTCCTGTGTGTTCTGCTCTGGGCC-TTTTGTGTTCCTGGTGCCAGGGCCGACTACAAAGACGATGACGA-CAAGCAGG AGCATGGGGCTGGTGTCCACC-3Ј and 5Ј-CTACTGCA-GCGACTTGGCAAACTCT-3Ј. HEK293T cells and COS7 cells were transfected with these expression vectors pEF-SDNSF-CЈ FLAG/His, pEF-SDNSF-NЈ FLAG, or pEF-His-FLAG in serum-free DMEM/Ham's F-12 medium (1:1) with N2 supplement using CellPhect (Amersham Biosciences) and were treated with lysis buffer (150 mM NaCl, 1% Triton X-100, 10 mM Tris-HCl, 1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin). The conditioned medium from these transfected HEK293T cells were filtered and concentrated with Centricon YM-10 (Millipore, Bedford, MA). SDNSF proteins were purified from the concentrated conditioned medium using Ni-NTA-agarose (Qiagen, Valencia, CA). According to the manufacturer's protocol with the slight modification of buffer concentrations as below, the concentrated conditioned mediums from pEF-SDNSF-CЈ FLAG/His and from pEF-His-FLAG transfected HEK 293T cells cultures were purified under native conditions. We mixed 1 ml of the 50% Ni-NTA slurry to 2 ml concentrated conditioned medium and 2 ml of 2ϫ lysis buffer (40 mM imidazole, 100 mM NaH 2 PO 4 , pH 8.0, 1 M NaCl) at 4°C for 60 min, loaded this mix into a column, washed with 20 mM of imidazole buffer (20 mM imidazole, 100 mM NaH 2 PO 4 , pH 8.0, 500 mM NaCl) twice, then washed with 40 mM of imidazole buffer (40 mM imidazole, 100 mM NaH 2 PO 4 , pH 8.0, 500 mM NaCl) twice, and eluted in elution buffer (250 mM imidazole, 100 mM NaH 2 PO 4 , pH 8.0, 500 mM NaCl). After this purification step, the eluted samples and the concentrated conditioned medium were electrophoresed in 15% SDS-polyacrylamide gel, and the purity of SDNSF in eluted solution was determined from the densitometric analysis of silver-stained SDS-polyacrylamide gel and NIH Image 1.62 (National Institutes of Health, Bethesda, MD). The purified SDNSF protein was dialyzed with Slide-A-Lyzer (Pierce) overnight against phosphate-buffered saline at 4°C. The protein concentrations were determined with Coomassie protein assay reagent (Pierce). We measured endotoxin/lipopolysaccharide activity by Toxicolor LS-6 (Seikagaku Corporation, Tokyo, Japan), and the concentration of endotoxin/lipopolysaccharide turned out to be under the detection limit. The similarly treated conditioned medium from pEF-His-FLAG transfected HEK293T cells transfected by pEF-His-FLAG was used as a control in the following SDNSF survival and proliferation assays. SDS-PAGE and Western analysis were carried out with these cell lysates and conditioned medium samples following ECL (Amersham Biosciences) Western blot protocols with using mouse 1:5,000 monoclonal anti-FLAG antibody (Sigma) and 1:5,000 anti-mouse IgG-horseradish peroxidase antibody (Sigma).
The 45 Ca 2ϩ binding assay (40) and mobility shift assay for calcium binding (41) were performed as described. The membranes of the 45 Ca 2ϩ binding assay were analyzed using an image analyzer as above. Deglycosylation assay and lectin blot analysis was performed as described (42,43).
NSCs Viability Assays and Neurosphere Assay-To test the SDNSF effects on ANSCs and mouse ES R1 cells, purified SDNSF was added into the medium, and cell survival and proliferational activities were assessed. ANSCs were plated onto Nunc 96-well or six-well plates at the density of 1,000 cells/cm 2 and cultured for 6 days in the serum-free growth medium minus FGF-2. SDNSF or FGF-2 were added at concentrations of 0, 0.1, 1, 10, 100, and 500 ng/ml. As a control, the same volume of similarly treated HEK293T cell conditioned medium as purified SDNSF was added in the control group medium. To test the possibility of SDNSF as a cofactor for FGF-2, ANSCs were cultured in FGF-2 ϩ growth medium with 100 ng/ml of SDNSF. Undifferentiated mouse ES cells were placed in Costar ultra low cluster 96-well plates (Corning, Acton, MA) at the density of 1,000 cells/cm 2 and cultured in medium with 100 ng/ml of SDNSF or same volume of control solutions and without leukemia inhibitory factor for 6 days. At DIV6 cell survival and/or proliferation effects were estimated by using premix WST-1 kit (Takara, Shiga, Japan), which is added to the growth medium and measures the number of viable cells and cell viability by detecting the cleavage of tetrazolium salts and mitochondrial enzyme activity as A 450 nm . Cell proliferation activity was estimated by BrdU uptake for 3 h at DIV4, which is shown as A 492 nm , using Cell Proliferation ELISA and by BrdU (colorimetric) kit (Roche Applied Science) according to the manufacturer's protocol; also by following clonal expansion of single GFP-labeled ANSC, SDNSF effects on proliferational activity were estimated. To test the self-renewal and differentiation potential of SDNSF-treated ANSCs, after ANSCs were cultured in the SDNSF ϩ / FGF-2 Ϫ medium (100 ng/ml of SDNSF without FGF-2) for 6 days, those ANSCs were replated on noncoated Nunc 48-well plates at the density of 2,000 cells/cm 2 and grown for 6 days in growth medium containing 20 ng/ml of FGF-2, and the number of neurospheres were counted. To assess the differentiation potentials, the newly formed neurospheres were replated onto poly-L-ornithine/laminin-coated Labtek 4-well chamber slides (Nunc) at the density of 10 -30 spheres/well and cultured in differentiation medium for further 6 days. Under confocal laser microscope or fluorescent microscope, positively stained cells were quantified at least 20 fields systematically across the coverslips from three to four independent experiments of parallel cultures.
Analytical Procedures and Data Analysis-Group changes were assessed using one-way ANOVA. When statistical differences were obtained at the p Ͻ 0.01 level between groups, post hoc comparisons were made using the Fisher least squares difference (LSD) test.

Isolation of Human, Rat, and Mouse SDNSF cDNA Clones-
Yeast transformants (3 ϫ 10 6 ) from ANSC cDNA library were screened, and 450 positive clones were obtained by the signal sequence trap method as described previously (18). Nucleotide sequences of all positive clones were determined. Of 29 independent clones, 27 clones were identical or homologous to sequences reported in rat or other mammals and two clones were novel. Examples of isolated known secretory or surface molecules, which bear secretory signal sequences include CD44, CD164 (MGC-24), chemokine CX3C, neural adhesion molecule F3, seizure-related gene (SEZ-6), CNS myelin membrane glycoprotein M6, growth/differentiation factor-15/macrophage-inhibiting cytokine-1, transforming growth factor-␤2, serpentine receptor Cyt28, neurotrimin, podocalyxin, protocadherin 7, and cystatin C. The full-length cDNA for rat SDNSF was isolated from rat ANSC and rat brain cDNA library. The sequence of the full-length rat SDNSF encodes a 145-amino acid protein (Fig. 1A). The translation start site methionine was assigned at nucleotide position 80 because of the presence of an N-terminal signal sequence and the compatibility of sequences immediately upstream with the Kozak consensus initiation sequence (44).
Homology searches on DNA data bases revealed partial cDNA fragments of mouse and human SDNSF, whose fulllength sequences were integrated and confirmed on the product of RT-PCR from mouse brain and human heart poly(A) ϩ RNA. The deduced amino acid sequences of the mouse and human SDNSF were compared with the rat SDNSF (Fig. 1A). The mouse and human SDNSF have 91.0% (132 of 45) and 85.5% (124 of 145) identity with rat SDNSF in their amino acid sequences, respectively. Data base searches also revealed that Caenorhabditis elegans gene NP 505967, which had been reported from the genetic analysis of developmental arrest and longevity in C. elegans (45), had 44.8% (65 of 145) identity, and the CG17271 gene product of Drosophila melanogaster had 39.3% (57 of 145) identity with rat SDNSF (Fig. 1A). These data base searches suggest that SDNSF genes are highly conserved.

SDNSF Is Secreted via the Classical Secretory Pathway-
The deduced amino acid sequence has an N-terminal hydrophobic domain, which is presumed to be the signal peptide, and has no other hydrophobic regions that can serve as a transmembrane domain. Among the known motifs, EF hand motifs were recognized (Fig. 1); however, there were no known Golgi or endoplasmic reticulum retention signals at its C terminus such as KDEL, HDEL, or HDEF (20,21). These data base search results suggested that SDNSF was a secreted protein.
To determine the validity of this signal sequence, HEK293T cells and COS7 were transfected with SDNSF plasmids, which were tagged with inserted FLAG at the predicted cleavage site (pEF-SDNSF-NЈFLAG) or with FLAG/His at C terminus (pEF-SDNSF-CЈ FLAG/His). The expression of these tagged SDNSF proteins was examined with Western blot analysis, and their subcellular localization was analyzed with immunofluorescent staining and immunoelectron microscopy. In both of the cell lysates and their concentrated conditioned mediums, tagged SDNSF proteins were detected at 19 kDa on the blotted membrane with anti-FLAG antibody (Fig. 2A). The slight size difference on blotted membrane between SDNSF-NЈ FLAG or SDNSF-CЈ FLAG/His in electrophoresis might come from the different tags inserted at the different positions in synthetic SDNSF protein.
To visualize the subcellular localization of SDNSF, the transfected COS7 cells with the SDNSF-CЈ FLAG/His construct were stained with anti-FLAG antibody. Under a confocal laser microscope, the staining profile exhibited a perinuclear distribution, which is a typical pattern of endoplasmic reticulum staining (Fig. 2B). Furthermore, the immunoelectron micro-scope using anti-FLAG antibody revealed that the expressed fusion proteins were densely packed in the nuclear envelope, endoplasmic reticulum, Golgi apparatus, and secretory vesicles (Fig. 2C). The subcellular localization pattern of sorted SDNSF was in accordance with classic vesicular pathway. According to the results shown above, we conclude that the predicted signal peptide of SDNSF functioned normally in the intracellular sorting process and that SDNSF is a secreted protein via classic vesicular pathway.
To obtain the purified SDNSF protein, overexpressed SDNSF-CЈ FLAG/His were treated with Ni-NTA-agarose, and eluted samples were analyzed after SDS-polyacrylamide gel electrophoresis and silver staining with using NIH Image 1.62. Intensity of the band for SDNSF was compared with the total bands visualized (Fig. 2D). After Ni-NTA-agarose treatment, the major fractions of bovine serum albumin were removed, and the ratio of SDNSF to the total proteins was 88.4% in elute 3 (Fig. 2D). From the amount of SDNSF protein in elution 1 through 4, the recovery rate were estimated as 51%. In the following experiments, 88.4% of the obtained SDNSF protein was treated as purified SDNSF protein.
Although the SDNSF amino acid sequence has two EF hand motifs, SDNSF could not show the calcium binding activity with 45 Ca 2ϩ blotting assay (40) and Ca 2ϩ -dependent mobility shift assay (41). SDNSF has no predicted N-glycosylation site and deglycosylation assay (42), and lectin blot assay (43) did not reveal any detectable glycosylation (data not shown).
SDNSF Is Localized at Dentate Gyrus, CA3, and Subiculum in Adult Rat Brain-To examine the systemic distribution of SDNSF, the poly(A) ϩ RNA from various rat tissues were tested FIG. 1. Amino acid sequences of SDNSF. A, alignment of deduced amino acid sequences of rat (r), mouse (m), and human (h) SDNSF with two similar gene products, NP 505967 from C. elegans and CG17271 from D. melanogaster. Identical residues are indicated with asterisks (*), and homologous residues are indicated with colon (:) or period (.). The full-length nucleotide sequences of rat SDNSF, mouse SDNSF, and human SDNSF were previously registered in the GenBank TM 47 EBI Data Bank with accession numbers AF475282, AF475283, and AF475284, respectively. As known motifs, two EF hand motifs are found in SDNSF sequences. Both motifs, EF hand 1 and EF hand 2, are shown with square boxes. B, alignment of EF hand domains of SDNSF and those of calmodulin with an EF hand consensus motif. The highly conserved residues are shaded.
with Northern blot analysis, and it was revealed that every tested organ expressed the major transcript of 1.8 kilobases (Fig. 3).
To localize the distribution of SDNSF in the adult brain, especially for the relationships with neurogenesis, in situ hybridization was performed. The positive signals for SDNSF probe that includes the 3Ј-untranslated region were detected in hippocampus (Fig. 4, A and C) and subiculum (Fig. 4D). In hippocampus, CA3 and dentate gyrus have positively stained cells. In CA3 SDNSF-positive cells were localized at its stratum oriens, and in dentate gyrus, they were distributed at the polymorphic cell layer also called hilus. Some immunoreactive cells are localized at the subgranular zone of granule cell layer, where the neurogenesis persists throughout the life (46). In another major region of adult neurogenesis, the subventricular zone of lateral ventricles, we could not detect positive signals for SDNSF expression (Fig. 4E). According to these in situ hybridization data, we can expect that the expression of SDNSF is related with ANSC and/or neurogenesis-related neurons in hippocampal dentate gyrus.
To observe the correlation of SDNSF with neurogenesis, the expression of SDNSF in several neural cell types and developing brain tissue was examined with RT-PCR. The SDNSF transcript was moderately or highly expressed in developing brains, especially late embryonic stages, primary hippocampal neurons, embryonic NSCs, and neuroblastoma N18 cells, glioblastoma U251 cells, and ANSCs (Fig. 5A). The signal levels in primary astrocytes and C6 glioma cells were low (Fig. 5A). Therefore, we assumed that SDNSF was expressed not specifically to nervous system, but their distribution in the CNS might be restricted to several cell types such as NSCs and neuronal cells and less astroglial cells.
Further, the expressions of SDNSF in the induced adult neurogenesis were also tested. It is known that ischemic stress to the brain induces neurogenesis in hippocampal dentate gyrus (47). We performed forebrain ischemic treatment on adult rat and induced hypoxic stress on hippocampus (26). Semiquantitative RT-PCR results showed on postoperative days 1 and 7 that the SDNSF transcripts were highly up-regulated (Fig. 5, B and C). (FGF-2 Ϫ growth medium) had a significant effect on ANSC viability. The purified recombinant SDNSF protein was added to FGF-2 Ϫ growth medium at concentrations of 0.1, 1, 10, 100, and 500 ng/ml when ANSCs were plated at a density of 1,000 cells/cm 2 . At DIV6, premix WST-1 assay was performed to measure the number of viable cells or cell viability by detecting the cleavage of tetrazolium salts and mitochondrial enzyme activity. It was shown that SDNSF improved ANSCs viability in a dose-dependent manner (Fig. 6A). SDNSF at the concentration of 100 ng/ml has a statistically significant difference on viability of ANSCs (Fig. 6B). To see whether this difference of viability came from proliferation activity of SDNSF, mitogenic activity was examined by tracking single cell proliferation via BrdU incorporation. In addition, the progeny from ANSCs were tracked by marking single ANSCs with replication-deficient GFP-expressing recombinant retrovirus. ANSCs treated with 100 ng/ml of SDNSF did not proliferate as much as FGF-2treated ANSC, and the sizes of the colonies in SDNSF group were similar to that of the control group (Fig. 6C). BrdU ELISA assay did not reveal an increased uptake of BrdU in the ANSCs treated with SDNSF (Fig. 6D). To test the possibility of SDNSF as a cofactor for FGF-2, ANSCs were cultured in FGF-2 ϩ growth medium with 100 ng/ml of SDNSF; however, mitogenesis was not enhanced by the addition of SDNSF (data not shown). Therefore, we conclude that SDNSF improves ANSC viability in the absence of FGF-2. To test similar effects on embryonic stem cells, the mouse ES cell line R1 was used; however, the addition of SDNSF did not show significant improvement of ES cell survival nor proliferation (Fig. 6E).
Under normal conditions, FGF-2 withdrawal induces apoptosis or differentiation of ANSCs (6,9,10); however, the addition of SDNSF without FGF-2 improved the survival of ANSCs as we have shown above. To determine whether SDNSF support maintain the multipotency of ANSCs, we test the response of the SDNSF-treated ANSC to FGF-2 and check the formation of neurosphere (30) and phenotype of descendent cells. The formation of neurospheres was used as an indicator for the abundance of stem-like cells (30).
ANSCs that had been treated with 100 ng/ml of SDNSF without FGF-2 for 6 days could form neurospheres. The number of neurospheres/well was smaller than that from the FGF-2-treated group but significantly larger than that from the control group (Fig. 7A). The neurospheres were subsequently treated with 0.5 M retinoic acid, and phenotypes of the cells from the neurospheres were examined. The neurospheres from SDNSF-treated ANSCs could differentiate into three different cellular types: neurons, astrocyte, and oligodendrocytes (Fig. 7,  B and C). A neuronal marker, Tuj-1-positive cells were seen in FIG. 7. Assessment of self-renewal and multipotency of SDNSF-treated ANSCs. A, the number of neurospheres formed in FGF-2 ϩ growth medium after treatment with SDNSF, FGF-2, or control. SDNSF-treated ANSCs can make neurospheres, although the number of sphere is less than those from FGF-2-treated ANSCs but more than those from control ANSCs. The data are expressed as the means Ϯ S.D. (n ϭ 8). Statistical analysis was performed by one-way ANOVA followed by a post hoc Fisher LSD test. Significant differences between results for SDNSF, FGF-2, and control are indicated by an asterisk (p Ͻ 0.01). B, differentiation profiles of the ANSC that were treated with SDNSF in the absence of FGF-2. The neurospheres formed in FGF-2 ϩ growth medium after treatment with SDNSF, FGF-2, or control were differentiated in the differentiation medium. Percentage of Tuj1 ϩ cells and GFAP ϩ cells from neurospheres were shown with their mean values Ϯ S.D. (n ϭ 10). Statistical analysis was performed by one-way ANOVA followed by a post hoc Fisher LSD test. Significant differences versus control are indicated by an asterisk (p Ͻ 0.01). C-E, immunofluorescence images of the differentiated ANSCs cells from the differently treated ANSCs neurospheres. Tuj-1 ϩ cells were labeled with Cy3. Similarly, NG-2 were with fluorescein isothiocyanate and GFAP with Cy5. ANSCs that had been treated with SDNSF (C), FGF-2 (D), or control (E) for 6 days were stimulated with FGF-2 for the neurosphere formation and then were differentiated in the presence of retinoic acid. 6.2 Ϯ 2.0%. Glial marker, GFAP-positive cells were 5.9 Ϯ 2.1%, and oligodendrocyte marker, NG-2-positive cells were 1.2 Ϯ 0.6% (Fig. 7B) from the SDNSF-treated neurospheres. From FGF-2 or control CM-treated neurospheres, Tuj-1-positive cells were 8.8 Ϯ 3.3% or 1.3 Ϯ 0.9%, GFAP-positive cells were 4.8 Ϯ 2.1% or 7.2 Ϯ 2.4%, and NG-2-positive cells were 2.3 Ϯ 1.1% or 0.9 Ϯ 0.5% (Fig. 7, B, D, and E). From these results, the surviving cells in SDNSF-treated ANSCs still retained selfrenewal potentials and maintained multipotency to differentiate into neuronal and glial phenotypes as has been seen in FGF-2-treated ANSCs. DISCUSSION ANSC conditioned medium is known to provide both trophic and mitogenic support for ANSCs, and previous analysis of the ANSC conditioned medium led to the characterization of glycosylated cystatin C as a cofactor for FGF-2 (15). By analyzing roles of IGF-1 in EGF-dependent ANSCs, IGF-1 was also identified as a cofactor of EGF (20). Our attempt here was to identify additional autocrine/paracrine molecules from FGF-2dependent ANSCs by selecting molecules bearing the secretory signal sequence. By using the signal sequence trap method on a cDNA library from ANSCs, we isolated a previously unknown secreted protein, SDNSF, in addition to several known molecules bearing signal sequences.
An interesting characteristic of SDNSF is an autocrine/paracrine effect on ANSCs themselves. Unlike glycosylated cystatin C or IGF-I, the addition of SDNSF into FGF-2-treated cultures did not modify the proliferation activity of FGF-2, and SDNSF itself did not display mitogenic activity. However, our data indicate that a high concentration of SDNSF improves the viability of ANSCs in the absence of FGF-2. In addition, SDNSF-treated ANSCs were shown to retain stem-cell characteristics, i.e. neurosphere formation as an indirect indicator of self-renewal and multipotency and also retinoic acid-induced differentiation assay. Therefore, we conclude that SDNSF has two effects on ANSCs: (a) to prevent ANSC cell death and (b) to maintain stem cell characteristics. These two effects on ANSCs have been observed in mitogenic growth factors, FGF-2 and EGF. However, SDNSF is unlikely to possess mitogenic activity and also unlikely to be cofactor for other mitogenic growth factors. As for the effect on maintenance of stem cell characteristics, the ability of SDNSF was lower than that of FGF-2 (Fig. 7A). Although SDNSF and FGF-2 (14.41 kDa) have similar molecular weights, SDNSF needs a higher concentration to effect the survival of ANSCs. Considering these characteristics of SDNSF, we could hypothesize that SDNSF delayed cell death and the differentiation processes that usually proceed in the absence of mitogenic growth factors.
The expression analyses provide us another conjecture on the role of SDNSF. SDNSF was expressed in every organ tested (Fig. 3A), but in CNS the transcript is restricted to cells with neuronal and immature neural phenotypes, including neuroblastoma and glioblastoma cells (Fig. 3B). In situ hybridization revealed that the in vivo expression of SDNSF transcripts are localized at CA3 stratum oriens and hilus in dentate gyrus (Fig.  4, A and C). Several positively stained cells were localized at the subgranular zone of pyramidal cell layer (Fig. 4C) where the active adult neurogenesis exists (36). Considering the distribution of expression and the results of NSC viability assays, SDNSF could be a molecular component in neurogenic microenvironment in adult NSC. Another area where SDNSF is expressed is the subiculum, which is a pivotal structure between the hippocampus and parahippocampal regions such as the entorhinal cortex. However, the functional properties of subiculum are not fully understood, and its role is ill defined in spatial navigation and mnemonic processing. From the analy-sis of SDNSF transcript, SDNSF can be detected from E8.5, one of the earliest time points we can isolate neural progenitor cells (48), and also can be detected highly in the late embryonic period, when the more neural progenitors are enrolled in CNS (48,49) (Fig. 5A). Besides ANSCs, embryonic NSCs, primary neurons, neuroblastoma cells, and glioblastome cells also expressed SDNSF and less in astrocyte or in glioma cells (Fig.  5A). This suggests that SDNSF expression is seen more in CNS progenitor population and neuronal lineage, which is correspondent with in situ hybridization data. Further studies using ischemically treated hippocampus showed that the tissues that have more neurogenesis (47) have more expression of SDNSF transcripts. Although we cannot estimate SDNSF instructing neurogenesis or the expression of SDNSF results from activated neurogenesis, available data suggest that SDNSF has a correlation with neurogenesis and may play a role in maintaining the pool of ANSCs within the neural stem cell microenvironment. The biological roles and mechanisms of SDNSF effects remain unclear, and further analyses of SDNSF, including identification of its receptors and related signal transduction networks, are required. Because SDNSF is a secreted protein and is expressed in several brain tumor cell lines (Fig. 5A), SDNSF could be a novel circulating tumor marker for neuroblastoma or glioblastoma. Because there are no available neural tumor markers secreted in circulating blood or cerebrospinal fluid, it is worth studying further to test SDNSF availability.
Another unique feature of SDNSF is its motif structure. Generally, it is rare that molecules within the EF hand superfamily have a signal peptide. Most of the EF hand molecules with secretory signal sequences usually also retain signals to the organelle, and so they are not secreted extracellularly. Calumenin (20) and FKBP23 (21) are localized to the lumen of the endoplasmic reticulum, and Cab45 (42) is localized to the Golgi lumen. The only exception among those EF hand family members with secretory signal sequences is BM-40, also known as SPARC (50), which is known to be secreted. One of the essential roles of EF hands in BM-40 is suspected to be Ca 2ϩ binding for folding and secretion of BM-40 (50). Although amino acid sequences around the two EF hand motifs of SDNSF are highly conserved from Drosophila to human, the importance of EF hands in SDNSF remains unknown, and the Ca 2ϩ binding capacity remains unconfirmed. In addition, the function of EF hand proteins and involvement of Ca 2ϩ in the regulation of ANSCs viability are still unknown.
The unique characteristic of SDNSF is its survival effect on ANSCs without proliferation activity, which was not reported previously for other growth factors or cofactors. This effect of SDNSF might be able to explain why the quiescent ANSCs could persist and maintain the multipotency in adult CNS, and utilizing SDNSF might help the therapeutic application of ANSC for diverse neurological entities by preventing further loss of adult neurogenesis.