Genomic Organization and Expression Pattern of Mouse Neuroglycan C in the Cerebellar Development*

Neuroglycan C (NGC) is a membrane-spanning chondroitin sulfate proteoglycan with an epidermal growth factor module that is expressed predominantly in the brain. Cloning studies with mouse NGC cDNA revealed the expression of three distinct isoforms (NGC-I, -II, and -III) in the brain and revealed that the major isoform showed 94.3% homology with the rat counterpart. The NGC gene comprised six exons, was approximately 17 kilobases in size, and was assigned to mouse chromosome band 9F1 by fluorescence in situhybridization. Western blot analysis demonstrated that, although NGC in the immature cerebellum existed in a proteoglycan form, most NGC in the mature cerebellum did not bear chondroitin sulfate chain(s), indicating that NGC is a typical part-time proteoglycan. Immunohistochemical studies showed that only the Purkinje cells were immunopositive in the cerebellum. In the immature Purkinje cells, NGC, probably the proteoglycan form, was immunolocalized to the soma and thick dendrites on which the climbing fibers formed synapses, not to the thin branches on which the parallel fibers formed synapses. This finding suggests the involvement of NGC in the differential adhesion and synaptogenesis of the climbing and parallel fibers with the Purkinje cell dendrites.

Proteoglycans are a group of proteins that bear covalently bound sulfated glycosaminoglycan chains. They are located on the cell surface and in extracellular spaces in various animal tissues including the central nervous system, and it is now believed that they play pivotal roles in the development, maintenance, and aging of tissues via cell-cell and cell-substratum interactions.
In the vertebrate central nervous system, there are many species of proteoglycans with different structural features (1,2). This could be due to the existence of a large number of cell types that constitute many neuronal circuits in the brain and to their multiple roles in various cellular events including mitogenesis, migration, differentiation, axonal outgrowth, and synaptogenesis (3)(4)(5). Of these neural proteoglycans, some are transmembrane chondroitin sulfate proteoglycans such as NG2 (6) and receptor-type protein tyrosine phosphatase /␤ (7,8). Both have been reported to exist mainly in the developing mammalian brain and are involved in the signal transduction of growth factors (9 -12) as well as cell-substratum interaction.
Recently, we found a new transmembrane chondroitin sulfate proteoglycan, named neuroglycan C (NGC), 1 that is expressed in the brain, especially at the surfaces of neuronal cells, but not in other tissues in rats (13) and humans (14). NGC cDNA cloned from rat brain libraries encodes a membrane protein composed of a signal peptide (30 amino acids) and a core protein of mature NGC (514 amino acids). The core protein is divided into five structurally different domains: an N-terminal domain to which chondroitin sulfate chain(s) may be attached, a cluster of 9 acidic amino acids, a cysteine-containing domain with an EGF-like motif, a membrane-spanning segment, and a C-terminal cytoplasmic domain of 95 amino acids with two potential phosphorylation sites for protein kinase C. Although the precise physiological functions are not known at present, considering that NGC is expressed exclusively in the brain, especially in the immature brain where the neuronal circuits are actively being formed, this proteoglycan may play some roles in neuronal circuit formation.
In this paper, we report that the localization of NGC changes in a developmentally regulated manner in the mouse cerebellum and that the structure of NGC changes from the proteoglycan form to the nonproteoglycan form as the cerebellar development proceeds. Based on these experimental results, we discuss how NGC is involved in the neuronal circuit formation in the cerebellum. Additionally, we describe the isolation of cDNA clones encoding three distinct isoforms of mouse NGC, organization, and chromosomal assignment of the mouse NGC gene.

EXPERIMENTAL PROCEDURES
cDNA Cloning of Mouse NGC-NGC phage clones were isolated from a ZAP cDNA library derived from the brain of a 2-month-old C57BL/6 strain mouse using a rat NGC cDNA (13) as a probe. Bacteriophages were plated at 5 ϫ 10 4 plaque-forming units/95 ϫ 135-mm dish and approximately 5 ϫ 10 5 plaques were screened. Positive plaques were purified by two additional rounds of screening and then the excision of the pBluescript phagemid from the ZAP vector (Stratagene, La Jolla, CA) was carried out. Both complementary strands of the mouse NGC cDNA were sequenced using BcaBEST Dideoxy Sequencing kit, dCTP version (TaKaRa, Osaka, Japan).
Isolation and Characterization of Mouse NGC Genomic DNA-A mouse NGC cDNA was used as a probe to screen a 129SVJ mouse liver genomic library (Stratagene). Seven positive clones with different XbaI patterns were isolated from approximately 5 ϫ 10 5 plaques. The clones were mapped by restriction enzyme digestion and Southern hybridization. Restriction fragments that included all exon sequences and intron/ exon splice junctions were subcloned into pBluescript phagemid (Stratagene) for DNA sequence analysis. Both complementary strands of the * This work was supported in part by grants-in-aid for scientific research from the Ministry of Education, Science, Culture and Sports of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF133700.
Chromosomal Assignment of NGC Gene by Fluorescence in Situ Hybridization (FISH)-Chromosome preparations were made from spleen lymphocytes from a male BALB/c mouse using the thymidine synchronization and bromodeoxyuridine release technique for delineation of replicated R and Q bands (15). Two genomic pBluescript subclones (17.5 kilobases) covering exons 1-5 were used as a probe for FISH. FISH was carried out according to published procedures (16,17) with slight modification. Briefly, the hybridization mixture consisted of approximately 10 g/ml each biotin-labeled probe, 50% formamide, 4ϫ SSC, 10% dextran sulfate, 2 mg/ml bovine serum albumin, 0.5 mg/ml salmon sperm DNA, 0.5 mg/ml E. coli tRNA, and 40 g/ml mouse Cot-1 DNA (Life Technologies, Inc.). The numbering of chromosome bands is according to Nesbitt and Francke (18).
Identification of Isoforms in the Cerebellum-To investigate which isoforms are expressed in the cerebellum, reverse transcriptase-mediated polymerase chain reaction (RT-PCR) was carried out using total RNAs from the cerebella of 5-day-old and adult mice.
In addition, to examine whether NGC-II mRNA exists in the cerebellum, 5Ј-ATTTGGGGCGGGAAACCATA-3Ј (nucleotides Ϫ98 to Ϫ79, based on the first nucleotide of the fragment common to NGC-I and -II cDNAs being nucleotide 97) and 5Ј-TTATCATGGACAGCAGG-GGA-3Ј (complementary sequence corresponding to nucleotides 478 -497) were used as primers to amplify the cDNA fragment ranging from exons 1Ј to 2. The PCR products were subcloned into pBluescript phagemid (Stratagene). In each case, 12 clones were picked up, and both complementary strands of the DNAs were sequenced using a BcaBEST Dideoxy Sequencing kit, dCTP version (TaKaRa).
Western Blot Analysis-Cerebral and cerebellar tissues from 5-, 10-, 15-, and 20-day-old, and adult C57BL/6 strain mice were processed for Western blot analyses using an anti-rat NGC antiserum (14) by the method of Matsui et al. (19). In brief, tissues were homogenized in phosphate-buffered saline containing protease inhibitors, and the homogenates were solubilized by boiling in the presence of 2% SDS. The samples were resolved by SDS-polyacrylamide gel electrophoresis on a 3% stacking gel and a 6% separating gel before and after digestion by protease-free chondroitinase ABC (EC 4.2.2.4; Seikagaku Corp., Tokyo, Japan). They were transferred electrophoretically onto a polyvinylidene difluoride membrane. Immunoreactive bands on the membranes were detected using a Vectastain elite ABC kit (Vector Laboratories, Burlingame, CA).
Immunohistochemistry-For immunohistochemistry, 5-day-old, 10day-old, and adult C57BL/6 strain mice were used. The cerebella were fixed with 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) and embedded in paraffin. The 5-m-thick sagittal sections were immunostained with an anti-rat NGC antiserum (14) as the primary antibody and with peroxidase-conjugated antirabbit IgG (Biomedical Technologies, Inc., Stoughton, MA) as the second antibody. Preimmune rabbit serum was used as control. Some sections were stained using the antiserum preabsorbed with a recombinant NGC, which was used as an antigen for production of the NGC antiserum (14). Vibratome sections (50 m) were treated with 0.1 IU/ml protease-free chondroitinase ABC for 3 h to remove chondroitin sulfate chains before they were immunostained.
For electron microscopy, the cerebella were fixed with 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2). After immunostaining, the 50-m-thick sagittal sections were postfixed with 1.3% OsO 4 in 0.1 M phosphate buffer (pH 7.2) and subjected to further sectioning to obtain ultrathin sections.

RESULTS
Cloning of the Mouse NGC cDNA-When a cDNA library (500,000 clones) of the mouse brain was screened for the NGCspecific sequence using a rat cDNA (13) as a probe, more than 400 colonies gave a positive signal above background. Plasmid DNAs were prepared from 20 individual colonies, digested with EcoRI and XhoI, and then the identity of the plasmid DNAs was confirmed by Southern hybridization with the rat cDNA used as a probe. The isolated plasmids had a cDNA insert ranging from 2.1 to 2.6 kilobases in size. Most of these cDNA inserts covered the full length of the coding region predicted from the sequence of the rat NGC cDNA.
DNA sequencing of these 20 clones suggested that there existed three isoforms (designated NGC-I, -II, and -III) of NGC in the mouse brain ( Fig. 1). Of the 20 clones, sixteen, two, and two encoded NGC-I, NGC-II, and NGC-III, respectively. The core protein of the major isoform, NGC-I, was composed of 539 amino acid residues and had a multidomain structure identical to rat and human NGCs: a signal peptide, an N-terminal domain to which chondroitin sulfate chains might be attached, an acidic amino acid cluster, an EGF-like domain, a membranespanning segment, and a C-terminal cytoplasmic domain with two putative phosphorylation sites for protein kinase C (Fig. 1). The homologies of the amino acid sequence of mouse NGC-I were 94.3% (Fig. 2) and 85.5% with rat and human NGCs (14), respectively.
As shown in detail below, only the N-terminal portion of pro-NGC, including the signal peptide, was structurally different between NGC-I and NGC-II, and NGC-III had a peptide insertion (after Asn 491 ) composed of 27 amino acid residues in the cytoplasmic domain of NGC-I (Fig. 1).
Cloning of the Mouse NGC Genomic DNA-The NGC gene had a size of approximately 17 kilobases and was comprised of six exons (Fig. 3). There were two alternative splicing sites. First, exon 1 covered the 5Ј-untranslated region and a part of the coding region common to NGC-I and -III mRNAs, and exon 1Ј covered them unique to NGC-II mRNA. Second, exon 5 coded an amino acid sequence (VRKFCDTPRVSSPHARALAHYD-NIVCQ) unique to NGC-III, which represents a part of the cytoplasmic domain with a new putative phosphorylation site The structural organization of the NGC core protein is proposed from the deduced amino acid sequence. The putative signal sequence is shown as a black box, the cluster of acidic amino acids is shaded with coarse dots, the transmembrane domain is diagonally striped, and the insertion is shaded with fine dots. In the horizontally striped box, the amino acid sequence is different from that in NGC-I and III. The potential site of glycosaminoglycan attachment is indicated by a solid vertical bar, and potential sites of phosphorylation by protein kinase C in the cytoplasmic domain are indicated by open triangles.
(underlined) for protein kinase C. No introns existed between exons 1Ј and 2 nor between exons 4 and 5. Other exon/intron junctions followed the GT/AG rule (Table I).
Exon 2 encoded the chondroitin sulfate attachment domain, the acidic amino acid cluster, and a part of the EGF-like module containing two cysteine residues. Exon 3 coded the remaining part of the EGF-like module containing four cysteine residues and the transmembrane domain. The cytoplasmic domain was encoded by exons 4 and 6 in the cases of NGC-I and -II and exons 4, 5, and 6 in the case of NGC-III.
Chromosomal Mapping of the Mouse NGC Gene-In 85 cells examined, 142 double-dot signals were observed by FISH using the mouse genomic DNA of NGC as a probe. These signals were specifically located in region F1 of mouse chromosome 9 (Fig.  4). Other locations did not show any double-dot signals. When mouse NGC-I cDNA was used as a probe, FISH signals were observed in the same region (data not shown). Based on these results, the mouse NGC gene was assigned to chromosome 9 at band F1.
Expression of Three Splice Variants of NGC in the Mouse Cerebellum-To confirm the expression of three splice variants coding NGC-I, -II, and -III in the mouse cerebellum, RT-PCR was performed using total RNA preparations of the cerebella from 5-day-old and adult mice. When the part of the NGC cDNA corresponding to exons 2-6 was amplified by RT-PCR, only two products of 408 and 489 bp were observed in both the 5-day-old and adult samples (Fig. 5A). The DNA sequence analyses of these products showed that the smaller one, which was the major product, represented the corresponding part of cDNAs of NGC-I and II, and that the larger one represented the corresponding cDNA fragment of NGC-III with the short peptide insertion.
When the parts of the NGC cDNAs corresponding to exons  TGCAGgtaac tgtagGTGTA 1-2 and exons 1Ј-2 were amplified by RT-PCR, only one product was detected, with a size of 460 bp in the former case (Fig. 5B) and 596 bp in the latter (Fig. 5C), in both the 5-day-old and adult samples. These findings indicate that the three splice variants of NGC are actually expressed in both the immature and mature cerebella of the mouse. Western Blot Analysis of NGC in the Mouse Cerebellum-Developmental change in the amount of NGC in the mouse cerebellum was examined from day 5 to adulthood by Western blotting using an antiserum raised against the recombinant polypeptide representing the chondroitin sulfate attachment domain of rat NGC (14). NGC was recognized as a smear, which is characteristic of proteoglycan, in homogenates of the 5-dayold mouse cerebellum (Fig. 6A). The average molecular mass of the smear was 150 kDa, the same as that of rat NGC (13). Digestion of the homogenates with chondroitinase ABC produced a relatively narrow band with a molecular mass of 120 kDa. The amounts of the 120-kDa band in the chondrotinasetreated samples decreased gradually with the cerebellar development from day 5 to day 15 and reached an adult level at around 20 days when the cerebellum is almost matured.
Unexpectedly, a significant amount of the 120-kDa band was visible in the cerebellar homogenates of 20-day-old and adult mice even without the chondroitinase treatment. The migra-tion position of this band was not altered by the chondroitinase digestion, showing that NGC occurs in a nonproteoglycan form without chondroitin sulfate chains in the mature cerebellum. In other words, NGC is a part-time proteoglycan.
In contrast, most NGC in the mouse cerebrum existed in a proteoglycan form with a molecular mass of 150 kDa at various postnatal days from 5-day-old to adulthood (Fig. 6B). The NGC content of the cerebrum increased with the postnatal development up to around day 15 and then decreased gradually reaching about the half-level of the peak at adulthood.
Localization of NGC in the Mouse Cerebellum-When the cerebellum of the 5-day-old mouse was immunostained with the anti-NGC antiserum, the Purkinje cell layer was stained diffusely (Fig. 7A). The reaction was very weak. At day 10, the reaction products were observed on the cell bodies and on the first and second (arrowheads) dendrites of the Purkinje cells ( Fig. 7B). It is noteworthy that the spiny branchlets of the Purkinje cell dendrites were totally immunonegative. Compared with the dendrites, the soma was stained weakly. In the adult mouse cerebellum, the reaction products were observed in spots on the large dendrites of the Purkinje cells (Fig. 7, C and F, arrowheads), whereas the soma was hardly stained (Fig.  7C, arrows). No immunoproducts were detected on the sections treated with either the preimmune serum or the preabsorbed antiserum (not shown).
Treatment of sections from the cerebellum of 10-day-old mice with chondroitinase ABC strengthened the immunostaining on the cell bodies of the Purkinje cells but not on the dendrites (Fig. 7, B and D). The same result was obtained on the Purkinje cell soma at day 5 (not shown). However, in adult mice, the Purkinje cell soma was no more deeply stained even after treatment with chondroitinase.
At day 10, the diverging points of the Purkinje cell dendrites were markedly stained (Fig. 7, B and D, arrowheads). In contrast, the immunopositive spots were not concentrated on the diverging points in the adult mouse cerebellum (Fig. 7, C and F,  arrowheads).
Electron microscopic observation showed that the cell membrane (large arrows), synaptic junctions (asterisks), and membranes of endoplasmic reticula (small arrows) were immunoreactive in the cerebellar Purkinje cells of 5-day-old mice (Fig.  8A). The cell membrane and the membranes of endoplasmic reticula were not immunostained in the other types of cell, such as granule cells and Bergmann glial cells. In the adult mouse cerebellum, the membrane of hypolemmal cisternae (arrow-heads) in the large dendrites of the Purkinje cells immunoreacted strongly with the anti-NGC antiserum (Fig. 8, B and C). The membrane of endoplasmic reticula (small arrows) and cell membrane (large arrows) were also immunostained.

DISCUSSION
Three Predicted Isoforms of NGC-In the present work, we showed that three splice variants were expressed from a single gene in the mouse brain (Fig. 3). RT-PCR demonstrated that these variants were actually expressed in both the immature and mature cerebella of the mouse (Fig. 5). The major variant coded a protein (designated NGC-I) composed of 539 amino acid residues, namely the mouse counterpart of NGC that we originally isolated from the developing rat brain (13). The other variants coded NGC isoforms designated as NGC-II and -III.
To determine the translation initiation site for NGC-II, we completely sequenced the first intron between exon 1 and 1Ј. However, the consensus sequence (A/G)NNATG(A/G) for the initiation of protein synthesis (20) could not be found in this region. Instead, we noticed that the third methionine (125th) in NGC-I was conserved in rat and human NGCs. The initial amino acid of NGC-II may be the third methionine of NGC-I.
NGC-III has a peptide insertion, which neither NGC-I nor NGC-II has in their cytoplasmic domain (Figs. 1 and 3). We are preparing antibodies in the rabbit specific to the peptide composed of the 27 amino acid residues to examine the developmental changes in the amount and location of this isoform in the brain. Preliminary immunohistochemical experiments suggested that one of the peptide antibodies stained, but very faintly, the Purkinje cells in the cerebellum, suggesting that NGC-III is actually expressed as a protein.
The predicted structure of NGC-III is very interesting in terms of protein phosphorylation. NGC-I has two putative phosphorylation sites for protein kinase C in its cytoplasmic domain, and NGC-III has an extra site in the peptide insert. A recombinant peptide representing the cytoplasmic domain of NGC-I can be phosphorylated well by protein kinase C isolated from the brain. 2 Whether the extra site in the NGC-III is actually phosphorylated in vitro/vivo and has some effects on signal transduction system in the cerebellum is open for further investigation.
Proposed Functions of NGC in the Cerebellar Development-The cerebellum is composed of several types of neurons. Of these cells, only Purkinje cells were stained with the antiserum against the chondroitin sulfate attachment domain of NGC. Purkinje cells form synapses with two major fiber systems: the climbing fiber system largely originated from the inferior olivary complex and the parallel fiber system from the cerebellar granule cells. At almost the same stage of the cerebellar development, the climbing fibers selectively adhere to and form synapses with the large stems of the Purkinje cell dendrites, whereas the parallel fibers exclusively form synapses with the spiny branchlets of the Purkinje cell dendrites (21). The developmental change in the localization of NGC on the Purkinje cells correlates well with synaptogenesis of the climbing fiber system with the Purkinje cells; NGC deposition and climbing fiber's synapse buttons are found on the soma of the Purkinje cells at day 5, on both the soma and the large dendrites at day 10, and on the large dendrites in the mature cerebellum (Fig.  7). These findings may suggest that NGC expressed on the Purkinje cells mediates the adhesion and/or synaptogenesis of the climbing fibers with the Purkinje cells. Alternatively, NGC may inhibit the adhesion and/or synaptogenesis of the parallel fibers with the large dendrites of the Purkinje cells.
It is of interest that NGC acts as a typical part-time proteoglycan in the Purkinje cell differentiation; NGC exists in a proteoglycan form with chondroitin sulfate chains on immature Purkinje cells and in a nonproteoglycan form on the large dendrites of mature Purkinje cells. Neither the biological significance nor molecular mechanism for the developmental change in the NGC structure is clear at this moment. However, it may be that the proteoglycan form of NGC on the Purkinje cells exerts more of an attractive effect on climbing fiber adhesion or more of a repulsive effect on parallel fiber adhesion.
As shown in Fig. 7, NGC is concentrated at the hypolemmal cisternae of the Purkinje cell dendrites in the mature cerebellum. The hypolemmal cisterna supplies membrane to the plasma membrane. Strong NGC immunoreactivity is observed on the diverging points of the Purkinje cell dendrites at day 10 ( Fig. 7), where membrane supply is needed. Considering these findings, NGC may be related to membrane supply system, especially in latter stage of the cerebellar development.
Recently, Ozaki et al. (22) have suggested that neuregulin-␤, a member of the neural EGF family, increases the expression of NR2C, which is a subunit of NMDA receptors by innervation of the glutamatergic mossy fibers to the internal granule cells in the cerebellum. Likewise, by innervation of the climbing fibers to the Purkinje cells, NGC, another member of the neural EGF family, may regulate the expression of some receptors in the cerebellum.
Location of the NGC Gene on Mouse Chromosomes-We have assigned the human NGC gene to chromosome 3p21.3 (14) and the mouse NGC gene to chromosome 9F1 (Fig. 4). The human chromosome 3p21.3 is known to be syntenic to the distal segment of mouse chromosome 9 on the basis of comparative gene mapping (23,24). The present result reconfirmed the syntenic association between human chromosome 3p21.3 and mouse chromosome 9F1. A computer search revealed that there had been reported no mouse mutants whose responsible genes are supposed to map to chromosomal region 9F1.