Structure and Cellular Distribution of Mouse Brain Testican ASSOCIATION WITH THE POSTSYNAPTIC AREA OF HIPPOCAMPUS PYRAMIDAL CELLS*

The complete deduced primary structure of mouse brain testican has been established from cDNA cloning. The cDNA encodes a polypeptide of 442 amino acids belonging to the proteoglycan family. The mouse brain testican core protein is 95% identical to its human tes- ticular counterpart. In situ hybridization investigations revealed that mouse testican mRNA is mainly present in a subpopulation of pyramidal neurons localized in the CA3 area of the hippocampus. An immunocytochemical approach, with antibodies directed against an overex-pressed chimeric antigen, produced in bacterial sys- tems, showed that testican is associated with the postsynaptic region of these pyramidal neurons. Testi- can includes several putative functional domains related to extracellular or pericellular proteins associated with binding and/or regulatory functions. On the basis of its structural organization and its occurrence in postsynaptic areas, this proteoglycan might contribute to various neuronal mechanisms in the central nervous system. Proteoglycans (PGs) 1 in mechanisms such cell adhesion, cell and cell a 2 (cid:51) SSC, 0.05% SDS solution at room temperature for 15 min and subsequently four times (15 min each) with a 0.1 (cid:51) SSC, 0.1% SDS solution at 50 °C (1 (cid:51) SSC: 150 m M NaCl, 150 m M sodium citrate, pH 7.4). The blot was exposed to Hyperfilm-MP (Amersham Corp.) at (cid:50) 70 °C for 72 h. Isolation of Mouse cDNA Clones— An oligo(dT)- and random-primed adult Balb/c mouse brain (cid:108) gt11 cDNA library (Clontech Laboratory Inc.) was screened with human probe B. Approximately 2 (cid:122) 10 5 recombinant plaques were screened according to standard techniques. The hybridization with the digoxygenin-11-dUTP randomly labeled probe (20 ng/ml) was carried out in 50% formamide, 5 (cid:51) SSC, 0.02% SDS,

The complete deduced primary structure of mouse brain testican has been established from cDNA cloning. The cDNA encodes a polypeptide of 442 amino acids belonging to the proteoglycan family. The mouse brain testican core protein is 95% identical to its human testicular counterpart. In situ hybridization investigations revealed that mouse testican mRNA is mainly present in a subpopulation of pyramidal neurons localized in the CA3 area of the hippocampus. An immunocytochemical approach, with antibodies directed against an overexpressed chimeric antigen, produced in bacterial systems, showed that testican is associated with the postsynaptic region of these pyramidal neurons. Testican includes several putative functional domains related to extracellular or pericellular proteins associated with binding and/or regulatory functions. On the basis of its structural organization and its occurrence in postsynaptic areas, this proteoglycan might contribute to various neuronal mechanisms in the central nervous system.
Proteoglycans of the nervous system have been reviewed recently Lander, 1993;Oohira et al., 1994). Chondroitin sulfate and keratan sulfate proteoglycans are mainly found in the extracellular space, whereas heparan sulfate proteoglycans are predominantly located on the cell surface (Margolis and Margolis, 1989). A developmentally regulated expression of PGs in various areas of the central nervous system has been described (Krueger et al., 1992;Rauch et al., 1992;Maeda et al., 1992;Carey et al., 1992;Stipp et al., 1994). PGs play a pivotal role in the growth and morphology of neurons (Cole and McCabe, 1991;Oohira et al., 1991;Faissner et al., 1994). GAGs have been shown to modulate axonal outgrowth and dendrite elongation (Damon et al., 1988;Lafont et al., 1994). The protein cores of PGs have also been proposed to participate in these events (Perris and Johansson, 1990;Iijima et al., 1991).
We have recently characterized a human testicular PG, named "testican" on the basis of its first identification in testes (Alliel et al., 1993). The analysis of a mouse multiple tissue Northern blot with human testican cDNA probes revealed that hybridizing transcripts are restricted to mouse brain. In order to define a murine model suitable for functional characterization of testican in brain, the complete primary structure of mouse brain testican was deduced from cDNA cloning experiments. The human and mouse testican protein cores were compared, and their multidomain structure is discussed and connected with potential biological functions. In situ hybridization showed that testican transcripts are mainly detected in pyramidal cells of the hippocampus. Immunohistochemical localization studies further indicated that testican is localized in the postsynaptic area of these neurons. A contribution of testican to a postsynaptic complex of these cells is suggested.

MATERIALS AND METHODS
Human Testican cDNA Probes-Two probes, spanning coding regions and portions of the 5Ј and 3Ј ends of human testican cDNA, were used in this study (see Fig. 2a). Probe A is a 889-bp EcoRI/Msc1 restriction fragment encompassing the 5Ј end noncoding region and encoding the 152 N-terminal amino acids. Probe B is an EcoRI/EcoRI restriction fragment encoding the 120 C-terminal amino acids and including 310 bp of the 3Ј end noncoding region.
Antibodies-A restriction fragment of human testican cDNA coding for the 120 C-terminal amino acids was subcloned at the EcoRI site in pGEX-2T expression vector (Pharmacia Biotech, Inc.). Plasmid clones were grown in Escherichia coli strain CAG 1139. A clone, containing the inserted fragment and producing a fusion protein of M r 38 kDa with Schistosoma japonica glutathione S-transferase attached at its N-terminal end, was selected. This fusion protein was purified by affinity chromatography on glutathione-Sepharose 4B (Pharmacia). A polyclonal antiserum, PTC9, directed against this hybrid protein was raised in rabbits and purified by three successive adsorptions on a pGEX-2Twild clone protein extract conjugated to CNBr-activated Sepharose 4B.
Isolation of Mouse cDNA Clones-An oligo(dT)-and random-primed adult Balb/c mouse brain gt11 cDNA library (Clontech Laboratory Inc.) was screened with human probe B. Approximately 2⅐10 5 recombinant plaques were screened according to standard techniques. The hybridization with the digoxygenin-11-dUTP randomly labeled probe (20 ng/ml) was carried out in 50% formamide, 5 ϫ SSC, 0.02% SDS, * This work was supported by the Centre National de la Recherche Scientifique and by the Institut National de la Santé et de la Recherche Médicale (Contrat de Recherche Externe 910714). 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.
Subcloning, Sequencing, and Computer Analysis-Phage DNA of positive clones was prepared from liquid cultures using the Sephaglas PhagePrep kit (Pharmacia). The EcoRI fragments were subcloned into the EcoRI site of plasmid pUC18 and used either as probes for further screenings or subcloned at the EcoRI site into M13 mp18 for sequencing experiments. DNA sequencing was performed by the dideoxy chain termination method (Sanger et al., 1977) using T7 DNA-polymerase Sequenase II (Amersham) with [␣-35 S]dATP. Sequencing compressions were resolved using dITP. Sequences were verified from overlapping clones, by sequencing both strands of DNA using M13 universal primers or cDNA specific primers.
Nucleic acid sequence comparison between mouse and human testicans was performed following the DIAGON program (Staden, 1982), using the CITI-2 (Paris, France) sequence analysis software.
In Situ Hybridization-Sagittal cryostat sections (16 m) of 15-dayold mouse brain (Swiss OF1 strain) were collected at Ϫ20°C on 0.5% gelatin, 0.05% chrome alum-coated slides and fixed for 10 min in 4% paraformaldehyde dissolved in phosphate-buffered saline. The sections were pretreated and hybridized under the experimental conditions previously reported (Roussel et al., 1991). Mouse cDNA probes were labeled to high specific activity (0.6 -0.8 ϫ 10 9 dpm/g) with 50 Ci (1.85 MBq) of [␣-35 S]dATP by the multiprime DNA labeling system (Boehringer Mannheim). After development of the x-ray film (5 days of exposure), the slides were dipped in Ilford K5 emulsion, exposed at 4°C for 21 days, and treated as usual (Roussel et al., 1991) for observation under bright or epipolarization light illumination with a LEICA DMRB microscope.
Optic and Electron Microscopic Immunocytochemistry-The procedure was performed essentially as described previously (Roussel et al., 1987). 100-m-thick cross-sections, taken at different brain levels were incubated for 20 min in Tris-buffered saline containing 5% calf serum. Afterwards, an appropriate volume of polyclonal antiserum (1:500 final dilution) was added. The sections were incubated for 2 h and then washed exhaustively with Tris-buffered saline. They were then treated for 1 h with horseradish peroxidase-conjugated secondary antibodies (either goat (anti-rabbit IgG) IgG (1:100) or sheep (anti-rabbit IgG) Fab fragments (1:500)) diluted in Tris-buffered saline. After a final washing step, the peroxidase activity was revealed with hydrogen peroxide and either 4-chloro-1-naphtol (optic microscopy) or diaminobenzidine (electron microscopy) as chromogens. Irrelevant antibodies were used in parallel as control. The sections were viewed with a LEICA DMRB optical microscope fitted with interferential contrast optics. For electron microscopic examination, the sections were additionally fixed with 1% osmic acid in 0.1 M phosphate buffer, pH 7.4, and then dehydrated in graded ethanols and embedded in Spurr's resin as usual (Roussel et al., 1987). Ultrathin sections were observed with a Philips EM 420 microscope.

Northern Blot Analysis, Molecular Cloning, and Characterization of Mouse Brain cDNA
Hybridization of a mouse multiple tissue Northern blot with human testican cDNA probe B revealed both a minor 2.3-kb hybridizing band and a major transcript of 5.2 kb, restricted to the lane containing mouse brain poly(A) ϩ RNA (Fig. 1).
Three positive clones, MB4, MB6, and MB8, were selected during a first screening of the gt11 mouse brain cDNA library, using human testican probe B. The two largest EcoRI restriction fragments from these mouse cDNA clones, which hybridized either with human probe A or B, were used as selective probes during a second screening of the gt11 mouse brain library. Four independent clones (CSA4, CSA5, CSA6, and CSA10) were selected. The larger 3Ј EcoRI restriction fragment in the CSA5 clone did not hybridize with any human probe used in this study; it allowed us to select two additional clones (TCS5-1 and TCS5-2) during a last screening (Fig. 2a).

Nucleotide and Deduced Amino Acid Sequences of Mouse Brain Testican
Comparison with Human Testican-The nucleotide sequence of a 4343-bp composite mouse brain testican cDNA, established from nine overlapping cDNA clones and coding for the complete deduced protein core of mouse testican is shown in Fig. 2b. The sequence around the first ATG, encountered at position 134 in the longest open reading frame, fulfills the criteria of eucaryote initiation sites (Kozak, 1991). This putative initiation codon is followed by a nucleotide sequence coding for a stretch of hydrophobic amino acids consistent with a signal peptide (von Heijne, 1986), which is identical to the N-terminal end of human testican. However, 98 nucleotides upstream from the initiation codon, mouse testican cDNA differs from its human counterpart by a seven-nucleotide deletion that has been observed in four independent clones (MB4, CSA4, CSA6, and CSA10). These observations support the assumption that the putative initiator is used both in mouse and human testican transcripts. In addition, nucleotide 10 was identified as a G or C in CSA6 and CSA10, respectively.
The initiation codon is followed by a 1326-bp open reading frame, which codes for a polypeptide of 442 amino acids. The mouse nucleotide coding sequence and its deduced protein sequence are 90 and 95% identical to those of human testican, and 60% of the amino acid substitutions are conservative (Fig.  3a). Serine residues 386 and 391, whose counterparts are the GAG attachment sites in human testican, are conserved. The main difference between both testicans is a Glu-Val-Glu tripeptide insertion at positions 63-65 in mouse brain testican.
A dot matrix plot comparing human and mouse testican cDNAs indicated that transcript homologies extend beyond the coding sequence both in the 5Ј and 3Ј regions (Fig. 3b). A putative polyadenylation site (ATTAAA), conserved in both mouse and human transcripts, was identified at position 1953. The use of this polyadenylation site may give rise to minor mRNA, with a shortened 3Ј region.
In Situ Hybridization-Two EcoRI restriction fragments from the MB6 clone (Fig. 2a) were used as probes. MB6-M is a 472-bp fragment covering the middle part of the translated region. MB6-T is a 932-bp fragment that codes for the 120 C-terminal amino acids and encompasses also the 3Ј noncoding region of the clone. Serial parasagittal to sagittal brain sections (Fig. 4), hybridized with both testican cDNA fragments, clearly show that the silver grains are mostly concentrated over the CA3 area of the hippocampus. The CA1, CA2, and CA4 areas are less labeled or not labeled. The cortex, thalamus, and hypothalamus display a faint labeling, whereas the white matter areas (corpus callosum, fimbria) of the cerebrum and the striatum show only a background distribution of grains (Fig. 4,  a-d). In the cerebellum, only the Purkinje cell layer shows a low accumulation of silver particles (not shown). At higher magnification, examination of the labeled CA3 area shows that the silver particles are mainly distributed over the cell bodies of pyramidal neurons (Fig. 4, i and j).  2. Cloning, nucleotide, and deduced amino acid sequences of mouse testican. a, the establishment of a 4343-bp composite cDNA sequence encoding mouse testican was established by the analysis of nine cDNA overlapping clones. A and B denote the human testican cDNA probes used in this study; M and T in the MB6 clone denote the probes used for the in situ hybridization experiments. The shaded box corresponds to the coding region. E corresponds to restriction enzyme sites for EcoRI. Scale bar, 0.5 kb. b, the deduced amino acid sequence is shown below the nucleotidic sequence. Single-letter code is used, and numbers indicate the positions of nucleotide or amino acid residues starting at the initiation codon. The first 21 amino acids fit with the von Heijne (1986) consensus for a leader peptide. The putative cleavage site is shown by an arrow. Cytidine 10, in boldface type, was identified as a guanosine in CSA6. A putative polyadenylation signal (nucleotides 1953-1958), also described in human testican, is in italic type. These sequence data are available from EMBL under accession number X92864.
Immunocytochemistry-The spatial distribution of testican in 15-day-old mouse brain structures was investigated with antiserum PTC9 directed against the 120 C-terminal amino acids of human testican. Low magnification revealed that diffuse immunolabeling is present along the hippocampus, whereas the cerebral cortex, thalamus, and hypothalamus display only a faint punctate labeling, and the cerebrum white matter is nonreactive (not shown). High magnification of the hippocampus CA3 area allowed us to notice a diffuse cytoplas-mic staining in the pyramidal cells and an intense punctate labeling, mostly located in the stratum lucidum, which contains synapses of the dendrites emanating from the stained neighboring pyramidal neurons (Fig. 5, a and b). In the cerebellum, a faint punctate labeling limited to the molecular layer is observed (Fig. 5c). Immunoreactive dots are evenly distributed over the cortex (Fig. 5d). Electron microscopic immunocytochemistry revealed a clear-cut postsynaptic localization of testican in the brain (Fig. 6). The labeling is concentrated on FIG. 3. Comparison between mouse and human testicans. a, comparison between mouse and human testican deduced amino acid sequences. Lines between residues refer to identity, and dots refer to homologous amino acids (A/G/S/T, D/E/N/Q, I/L/V, Y/F/W, and R/K/H). A probable processing site, based upon statistical data, is indicated by an arrow. Putative glycosaminoglycan attachment sites are indicated by arrowheads. b, dot matrix plot analysis of mouse and human testicans nucleic acid sequences. The analysis was plotted by the DIAGON program (Staden, 1982), with a score of 8 and a window size of 9. Numbering for mouse and human cDNA testicans begins at nucleotide 1 and 295, respectively. The comparison extends over 3200 nucleotides. Solid lines represent the coding regions of human (horizontal) and mouse (vertical) testican. the postsynaptic densities. Some reactivity, likely due to testican molecules carried from their site of synthesis to their target postsynaptic site, is also present in the surrounding cytoplasmic fluid. The presynaptic area and the synaptic cleft of these synapses do not react with the antiserum.

DISCUSSION
The composite sequence, established from nine mouse brain cDNA clones, encodes a 442-amino acid polypeptide highly homologous to the human testican protein core. Structural conservation around the so far identified GAG attachment sites suggests that mouse brain testican may be substituted by chondroitin sulfate and heparan sulfate, as shown for testican from the human male reproductive organs (Bonnet et al., 1992). Neither any other serine-glycine pairs, matching the consensus sequence corresponding to GAG attachment sites (Bourdon, 1990), nor NXS or NXT N-linked glycosylation consensus sequences are present within mouse or human testican deduced protein sequences. The Northern blot analysis was indicative of a brain-restricted distribution of testican transcripts in mice; however, reverse transcriptase-polymerase chain reaction also allowed their identification in testes. 2 Human transcripts appeared more broadly distributed, but, among 16 tested tissues, brain is one of the three tissues where testican transcripts were mostly detected. 3 The high conservation of mouse and human polypeptidic structures led us to consider that functional elements reside within the testican's protein core. Such conserved structures are encountered in proteins involved in basic cell biological processes. The distribution of the proline residues, which account for 5.4% of the total amino acids, is not random. The 182 PCPCLPEPEPLKP peptide contains 25% of the total proline residues clustered in a region corresponding to less than 3% of the protein. The human counterpart of this peptide, despite a variation in the proline residue distribution, shows the same features. The tetrapeptide GKSL is present at positions 172-175 and in the palindromic peptide LSKGKSL (amino acids 326 -332). In mouse testican, the supernumerary tripeptide 63 EVE is inserted in a pseudopalindromic structure, 57 WNR-FRDEVEDDYFRNW. The corresponding peptide in human testican allows us to determine a WNRFXD(1-4)DXFRNW palindromic sequence (X is a nondefined amino acid, and 1-4 denotes the number of spacing amino acids) whose biological significance, if any, is not known yet. Testican contains a cluster of glutamic acid and aspartic acid residues, which account for 40% of the 63 C-terminal residues. This acidic region includes a stretch of 9 consecutive acidic amino acids. A similar stretch is also encountered in other PGs: in the ␤-amyloid precursor protein (Kang et al., 1987), in claustrin (Burg and Cole, 1994), in versican (Zimmermann and Ruoslahti, 1989), and in brevican (Yamada et al., 1994). Such a stretch may be involved in the binding of cationic substances. The QKLSK pentapeptide, located in the upstream vicinity of the CWCV domain, is highly reminiscent of the NKISK sequence, which has been shown to be of importance for the binding of decorin to fibronectin (Schmidt et al., 1991).
The mouse brain testican, as its human testicular counterpart (Alliel et al., 1993), is a multidomain protein that shares substantial similarities with proteins involved in adhesion, migration, and cell proliferation. Some of these domains have been identified in proteins involved in neural development, synaptogenesis, and synaptic transmission. The counteradhesive neural molecules QR1 and SC1, either associated with neurite outgrowth and synapse formation (Guermah et al., 1991) or expressed during neural postnatal development (Johnston et al., 1990), contain testican's osteonectin-like domains. A 45-amino acid sequence related to Kazal-type protease inhibitor domains (Apostol et al., 1993) exists as a single copy in QR-1 and SC-1, and is tandemly repeated in follistatin (Hemmati-Brivanlou et al., 1994). Besides its implication for the gonadal function, follistatin-activin interaction has been suggested to contribute to the maintenance of nerve survival (Schubert et al., 1990), to be involved in the control of neurosecretory neuron activity (Sawchenko et al., 1988), and to participate in neural differentiation (Hashimoto et al., 1992). A wide protease inhibitory effect has been demonstrated for a recombinant rat agrin (Biroc et al., 1993), a protein that contains nine Kazaltype domains (Ferns and Hall, 1992) and plays a role in acetylcholine receptors clustering around the neuromuscular junctions. Agrin, which can be substituted by GAG chains (Tsen et al., 1995), has been proposed to be an interactive partner of neural cell adhesion molecules that may regulate a variety of FIG. 5. Immunolocalization of testican in mouse brain. Transverse vibratome sections of 21-day-old mouse brain immunostained with PTC9 antiserum (1:500). Diffuse and punctate immunolabeling is observed in the area of the pyramidal neurons of the hippocampus. In the CA3 area the punctate labeling can be correlated to the spines of the dendrites (arrows) originating from pyramidal neurons (a, b). In other brain regions, i.e cerebellum (c) and cortex (d), only a faint punctate labeling (arrows) is noticed. In cerebellum (c), the staining is clearly limited to the molecular layer (ML). P, Purkinje cells; GL, granular layer. Scale bars, 50 m. cell adhesion processes, including synaptogenesis. A 46-amino acid motif, the CWCV domain, has been identified in various proteins associated to the basement membrane and epithelial cell surface (Alliel et al., 1993). It is of interest to notice that this interspecies-conserved domain in the IGF-BP family (Shimasaki et al., 1991) is assumed to contribute to growth factor binding (Huhtala et al., 1986). Heparin-like GAGs have been proposed to act as local regulators of IGF/IGF-BP complexes (Arai et al., 1994). IGFs are known to influence neurotransmission in the hippocampus and regulation of synaptic transmission occurs through modulation of transmitter receptors in the postsynaptic membrane (Araujo et al., 1989). Thus testican is a molecule whose structure contains peptidic motifs and can be substituted by polysaccharides encountered in contributors to extracellular matrix organization and in regulators of neuralization and neurotransmission.
The spatial immunolabeling distribution of testican in mouse brain correlates well with the in situ hybridization results. Glial cells appear to be devoid of testican. Among the neuron-enriched brain regions, testican is mainly distributed on the postsynaptic densities of neurons occurring mostly in the CA3 area. The staining is concentrated at the postsynaptic side of the plasma membrane of the dendrite. This observation and structural features of testican suggest that this PG is likely to interact with membrane-bound components at the postsynaptic nerve terminals. Testican contains neither an hydrophobic domain, except for the N-terminal putative signal peptide, nor a C-terminal sequence that could be glypiated (Gerber et al., 1992). The ability of the core protein of some brain PGs to interact with neurons via neural cell adhesion molecules has been recently proposed as a regulatory mechanism for cell adhesion and cell migration in the central nervous system (Grumet et al., 1993;Friedlander et al., 1994;Maurel et al., 1994;Grumet et al., 1994). The testican transcripts and protein core are mainly distributed within the pyramidal cells in the hippocampus CA3 region where glypican is also present . On the basis of its structure, mimicking a set of proteins sharing diverse binding abilities, it is FIG. 6. Immunoelectron microscopic localization of testican. Testican localization in the postsynaptic portion of synapses present in the stratum lucidum of the hippocampus (a, b), cortex (c, d), and molecular layer of the cerebellum (e) of a 21-day-old mouse brain after incubation with PTC9 antiserum (1:500).
The labeling (open arrows) is strong on the postsynaptic (Po) density and faint in the immediate cytoplasmic fluid. No labeling is present in the synaptic cleft. The corresponding presynaptic areas (Pr), clearly identified by the presence of synaptic vesicles, are also negative; not all the postsynaptic densities are immunostained (black arrow). Scale bars, 0.5 m.
tempting to hypothesize that testican modulates the binding of ligands to target cells. It may play a crucial role in specialized neurons, in a brain area where receptors for growth factors have been identified (Fayen et al., 1992;Marks et al., 1991;Breese et al., 1991).
Among the proteoglycan protein cores so far depicted, none can be structurally related to testican. This molecule can thus be considered as a new member of the brain proteoglycan family. The spatial distribution of testican in a restricted area of the brain, together with its appearance in the postsynaptic region, is an argument for its potential contribution to receptor activity, neuromodulation, synaptic plasticity, or even neurotransmission.