Abbreviations used: CD, circular dichroism; DIG, digoxigenin; EC domain, extracellular Ca2+-binding domain; EST, expressed sequence tag; FS domain, follistatin-like domain; ORF, open reading frame; SDS, sodium dodecyl sulfate; SPARC, secreted protein, acidic and rich in cysteine; SSC, saline-sodium citrate; TY domain, thyroglobulin-like domain; UTR, untranslated region.
Address correspondence and reprint requests to Dr. P. Maurer at Institute for Biochemistry II, Medical Faculty, University of Cologne, Joseph-Stelzmann-Strasse 52, 50931 Cologne, Germany.
Abstract: We have screened a human cDNA library using an expressed sequence tag related to the BM-40/secreted protein, acidic and rich in cysteine (SPARC)/osteonectin family of proteins and isolated a novel cDNA. It encodes a protein precursor of 424 amino acids that consists of a signal peptide, a follistatin-like domain, a Ca2+-binding domain, a thyroglobulin-like domain, and a C-terminal region with two putative glycosaminoglycan attachment sites. The protein is homologous to testican-1 and was termed testican-2. Testican-1 is a proteoglycan originally isolated from human seminal plasma that is also expressed in brain. Northern blot hybridization of testican-2 showed a 6.1-kb mRNA expressed mainly in CNS but also found in lung and testis. A widespread expression in multiple neuronal cell types in olfactory bulb, cerebral cortex, thalamus, hippocampus, cerebellum, and medulla was detected by in situ hybridization. A recombinant fragment consisting of the Ca2+-binding EF-hand domain and the thyroglobulin-like domain of testican-2 showed a reversible Ca2+-dependent conformational change in circular dichroism studies. Testican-1 and -2 form a novel Ca2+-binding proteoglycan family built of modular domains with the potential to participate in diverse steps of neurogenesis.
In the CNS proteoglycans have attracted attention not only by their ability to contribute to structural properties of the interstitial brain matrix, but also by their influence on diverse aspects of neurogenesis. The local extracellular environment in which neuron and astrocyte progenitors migrate and differentiate offers trophic and guidance cues for neuronal development and maintenance. In addition to cell surface proteoglycans, soluble proteoglycans also may influence specific cell-cell and cell-matrix adhesion. For several proteoglycans a regulation of neuronal migration, neurite outgrowth, and axon guidance has been demonstrated, thus contributing to nervous system patterning (Oohira et al., 1994; Ruoslahti, 1996; Margolis and Margolis, 1997).
Based on the type of glycosaminoglycan chains attached to the protein core, proteoglycans can be divided into chondroitin/dermatan sulfate, heparan sulfate, keratan sulfate, or mixed-type proteoglycans. However, functions of proteoglycans are determined not only by the number, type, and length of the glycosaminoglycan chains, but also by distinct binding properties of the protein cores (Grumet et al., 1996). Most proteoglycans of the brain have been characterized by monoclonal antibodies, and the primary structures of their protein cores remain to be elucidated (Margolis and Margolis, 1993; Lander, 1993). The small leucine-rich proteoglycans encompass a class of secreted proteoglycans that bind growth factors and modulate their activity (Iozzo, 1997). The lectican/hyalectin family is a second well-described proteoglycan family that is built by a modular architecture and substituted by chondroitin sulfate and sometimes keratan sulfate chains. The family includes aggrecan, versican, brevican, and neurocan, all of which are characterized by their binding to hyaluronan (Iozzo and Murdoch, 1996; Ruoslahti, 1996).
Testican was isolated as a chondroitin sulfate/heparan sulfate proteoglycan from human seminal plasma (Bonnet et al., 1992). Its amino acid sequence was deduced from a testicular cDNA clone (Alliel et al., 1993) and shown to share homologous domains with the BM-40/secreted protein, acidic and rich in cysteine (SPARC)/osteonectin protein family (Alliel et al., 1992; Kohfeldt et al., 1997). In adult mice testican mRNA expression is restricted to the brain, and the protein is localized to postsynaptic densities (Bonnet et al., 1996).
In search for homologous members of the BM-40/SPARC/osteonectin family (Maurer et al., 1995), we identified an expressed sequence tag (EST) that was homologous but not identical to the known members and was most similar to testican. Using this EST we screened a human fetal brain cDNA library under low-stringency conditions and isolated the cDNA for testican and a novel cDNA encoding a homologue of testican. Northern blot data demonstrate that both cDNAs are prominently expressed in brain and thus define a novel proteoglycan family of the CNS.
MATERIALS AND METHODS
Isolation of cDNA clones
A human fetal brain cDNA library (Clontech, Palo Alto, CA, U.S.A.) was screened with a 385-bp 32P-labeled HindIII-ApaI fragment isolated from the human brain-derived EST H14434 (ATCC). Hybridization was carried out at 65°C in aqueous solution [0.5 M sodium phosphate (pH 7.2), 7% sodium dodecyl sulfate (SDS), 1 mM EDTA, 10% bovine serum albumin, and 0.2 mg/ml salmon sperm DNA]. Filters were washed under low-stringency conditions [40 mM sodium phosphate (pH 7.2) and 1% SDS at 60°C for 15 min]. Three positive clones in the final rescreen were in vivo excised, yielding cDNA clones in the pDR2 vector. Using the dideoxy method and an ALF automatic sequencer (Pharmacia), plasmids were sequenced on both strands with universal and internal primers. Nucleotide sequence analysis and homology searches in the dbEST database (Boguski, 1995) were performed with the programs of the GCG package.
Total RNA of various tissues was extracted from adult mice by the guanidinium-thiocyanate method (Chomczynski and Sacchi, 1987). Poly(A)+ RNA was prepared from the total RNA using the Qiagen Oligotex mRNA kit. Poly(A)+ RNA was carefully quantified by UV spectroscopy. Aliquots (3 μg) were electrophoresed on a 1% denaturing agarose gel, stained with ethidium bromide to verify equal loading, and blotted onto HybondXL. Hybridization with the 32P-labeled HindIII-ApaI fragment of testican-2 (see above) was performed overnight in 4 × saline-sodium citrate (SSC; 600 mM NaCl and 60 mM sodium citrate, pH 7.4), 1% milk powder, 1% SDS, 0.3 mg/ml salmon sperm DNA, and 10% dextran sulfate at 65°C. Blots were washed two or three times in 0.5 × SSC and 0.1% SDS at 65°C for 15 min and exposed overnight on a phosphoimager photoplate (Molecular Dynamics). The radiolabeled testican-1 cDNA probe was generated using the human testican-1 extracellular Ca2+-binding domain (EC domain) fragment (Kohfeldt et al., 1997).
In situ hybridization
In situ hybridizations were carried out using the following riboprobes: Human testican-2 EC domain (nucleotides 875-1,210) was amplified by PCR with primers 5′-GCCCGCTAGCCGAGACTTGCACCGGTCA-3′ and 5′-CAATGACTGAGCTCCTACCTCCAGAAGCAGAAGC-3′ using an annealing temperature of 60°C. A 145-bp fragment of 3′ untranslated region (UTR) of mouse testican-2 (nucleotides 1,485-1,630) was amplified with primers 5′-GATAGCCCTCGACGGCCTCGGC-3′ and 5′-GTCCCCAGTCCTCCCCCACTGG-3′ using an annealing temperature of 62°C. Both PCR products were ligated into the pCRII vector (Invitrogen), and sequences were confirmed by dideoxy sequencing. The antisense riboprobe was produced by EcoRV restriction of the pCRII plasmid containing human testican-2 EC domain and Sp6 polymerase transcription following the digoxigenin (DIG) RNA labeling protocol (Boehringer Mannheim). Sense riboprobe was obtained by NheI restriction and T7 polymerase transcription. For the mouse testican-2 probe, HindIII restriction/T7 polymerase transcription and XhoI restriction/Sp6 polymerase transcription, respectively, were used. The testican-1 EC domain cDNA (Kohfeldt et al., 1997) was subcloned into pBluescript (Stratagene). Antisense probes were produced by BamHI restriction and T3 polymerase transcription; sense probes were produced by NotI restriction and T7 polymerase transcription. Yield and concentrations of the probes were estimated according to the manufacturer’s protocol. The integrity of the riboprobes was checked on an agarose gel.
Dissected brains of adult mice (CD1 strain) were fixed in 4% paraformaldehyde in phosphate-buffered saline for 6 h. Dehydration was achieved in 50, 70, 90, and 100% ethanol, which was subsequently replaced by methylbenzoate and finally by Paraplast Plus (Sherwood). Tissues were then embedded in Paraplast Plus and cut into 7-μm sections. Two sections were placed on a glass slide at a time and were stored at 4°C. Sections were deparaffinized by xylol treatment followed by decreasing ethanol concentrations. After proteinase K digestion (2 μg/ml), sections were refixed in 4% paraformaldehyde and acetylated with acetic acid anhydride in 0.1 M triethanolamine. A further round of dehydration with ethanol preceded the prehybridization performed in 50% formamide, 1 × Denhardt’s solution (0.02% Ficoll 400, 0.02% polyvinylpyrrolidone, and 0.02% bovine serum albumin), 5% (wt/vol) dextran sulfate, 0.5 mg/ml salmon sperm DNA, 0.25 mg/ml yeast tRNA, and 4 × SSC (pH 7.0) containing 0.1% SDS. Riboprobes were dissolved at 2 ng/μl in 50 μl of the above hybridization solution. The sections on each slide were incubated overnight at 50°C with the antisense and the corresponding sense riboprobes, respectively. After two washing steps at room temperature with 50% formamide and 2 × SSC, detection of DIG label with the anti-DIG antibody (Boehringer Mannheim) was performed according to the manufacturer’s protocol.
Recombinant expression of testican-2-EC-thyroglobulin-like domain (TY domain)
A cDNA fragment spanning the EC and the TY domains of testican-2 (residues 198-377) were generated by PCR using the 5′-primer 5′-GCCCGCTAGCCGAGACTTGCACC and the 3′-primer 5′-GTTACTGACTCGAGGATTAGCGTCAGCCCCGAG. The primers introduced new restriction sites (NheI and XhoI) and a stop codon. The NheI/XhoI-restricted PCR product (536 bp) was purified and inserted between the same restriction sites of the eukaryotic expression vector pCEP-Pu/BM40 (Kohfeldt et al., 1997) to obtain the final expression vector pCEP-Pu/testican-2-EC-TY. The correct insertion and sequence were verified by cycle sequencing. Human embryonic kidney cells (293 EBNA; Invitrogen) were transfected with pCEP-Pu/testican-2-EC-TY and selected with puromycin as described (Kohfeldt et al., 1997). Conditioned serum-free media (2.1 L) were collected and passed over DEAE-Sepharose column equilibrated in 50 mM Tris-HCl, pH 8.6. The column was eluted with a linear NaCl gradient (0-1 M), and the recombinant protein eluted at 0.1-0.2 M NaCl. After concentration of the protein on a Source Q (Pharmacia) and elution in a small volume (1.5 ml), it was finally purified by gel filtration on Sephadex G-75 in 50 mM Tris-HCl (pH 7.4) and 150 mM NaCl. Edman degradation was performed on a protein sequencer (model-473A; Applied Biosystems) following the manufacturer’s protocol.
Circular dichroism (CD)
CD spectra were recorded in a Jasco model 715 CD spectropolarimeter at 25°C in thermostatted quartz cells of optical pathlength 1 mm. The molar ellipticity [θ] (expressed in degrees · cm2· dmol-1) was calculated on the basis of a mean residue molecular mass of 110 Da. The Ca2+ dependence of the CD spectrum was measured by addition of 2.5 mM CaCl2. Reversibility of the conformational change was tested by subsequent addition of 4 mM EDTA. A baseline with buffer (50 mM Tris-HCl and 150 mM NaCl, pH 7.4) was recorded separately and subtracted from each spectrum.
Molecular cloning of human brain cDNAs
Scrutiny of EST databases revealed a human brain EST with 35% similarity on the amino acid level to the follistatin-like domain (FS domain) of BM-40, indicating that it might be derived from a novel member of the BM-40/SPARC/osteonectin family of proteins. Using a fragment of this EST we probed a human brain cDNA library under low-stringency conditions and could isolate three individual clones. One clone encompassed an open reading frame (ORF) of 1,317 bp and was identical to human testican (Alliel et al., 1993). Both other clones encoded a novel cDNA with an ORF of 1,272 bp and differed only in the length of 5′ and 3′ UTRs (Fig. 1). The sequence of the EST is found within this ORF. A conserved translational start site context (Kozak, 1996) surrounds the putative ATG initiation codon. No consensus polyadenylation sequence is found within the isolated 3′ UTR, suggesting an mRNA >1.9 kb. The cDNA encodes a putative protein sequence of 424 amino acids with a calculated molecular mass of 46.8 kDa. Sequence comparison showed that this protein is 43% identical to testican over its complete length. We propose naming this gene testican-2 and reserve testican-1 for the gene described by Alliel et al. (1993).
The deduced amino acid sequence of human testican-2 can be subdivided into six regions (Fig. 2). The N-terminal portion is predicted to function as a signal peptide. It contains a hydrophobic stretch of 22 amino acids ending with a consensus signal peptidase cleavage site (Nielsen et al., 1997). The lack of hydrophobic transmembrane protein-like regions implies that the protein will be secreted into the extracellular space.
The mature protein of 402 amino acids can be further subdivided into five domains: The N-terminal region (residues 23-88) does not show any homologies to known proteins except testican-1. It is followed by a cysteine-rich domain (residues 89-192) homologous to follistatin. The crystal structure of the FS domain of BM-40 has recently been solved (Hohenester et al., 1997) and allows a structure-based alignment of testican-2 (Fig. 2). All 10 cysteine residues forming the structural core of disulfides are conserved. When modeled on the structure of BM-40, the two additional cysteines (Cys176 and Cys182) in testican-2 come into close vicinity and may form an additional disulfide bond.
The third domain is homologous to the EC domain of BM-40 (Hohenester et al., 1996). The EC domain is characterized by two Ca2+-binding EF-hand motifs (EF-1 and EF-2) that encradle an amphiphilic helix αA. All essential features of the EC domain are well conserved in testican-2 (Fig. 2). The connecting part between a helix αA and the EF-hand motifs is much shorter than in BM-40 and presumably assumes a different conformation. Following the EC domain a TY domain (residues 312-377) can be detected. Thyroglobulin domains are short domains stabilized by three conserved disulfide bonds and contain a characteristic CWCV tetrapeptide sequence (Molina et al., 1996). The C-terminal domain of testican-2 (residues 378-424) is unique to testicans and contains two putative glycosaminoglycan attachment sites. A comparison of this modular structure to those of other members of the BM-40 protein family is shown in Fig. 3.
Northern blot analysis of mRNA isolated from multiple adult mouse tissues revealed an expression restricted to brain and a weaker signal in lung and testis (Fig. 4A). Separate isolation of mRNA from the cerebellum and the remainder of the brain allowed us to show that transcription of the testican-2 gene occurs in both parts of the brain. A 6.1-kb transcript is the most prevalent testican-2 mRNA. Minor bands are detected at 4.4 and 2.6 kb and may be caused by usage of different polyadenylation sites (Fig. 4A). The specificity of the testican-2 probe was verified by dot blotting and cross-hybridization of the probes for testican-2 and testican-1. This pattern of tissue distribution was confirmed by RT-PCR (data not shown). In contrast to testican-2, testican-1 was specifically expressed only in brain of adult mice, consistent with results shown by Bonnet et al. (1996) (Fig. 4B).
In situ hybridization of brain sections revealed the presence of the mRNA for testican-2 in most regions of the brain: olfactory bulb, cerebral cortex, hippocampus, thalamus, cerebellum, pons, and medulla (Fig. 5). No difference could be detected between the two antisense riboprobes for testican-2. Control hybridizations with the sense probes did not show any signals.
In the olfactory bulb the most intense staining was seen in mitral cells, whereas periglomerular cells showed a weaker staining (Fig. 5B). External and internal plexiform layers did not label. Although granular cells in the olfactory bulb showed a strong background color, it could be seen at higher magnification that staining was diffuse, and we thus assume that granular cells do not transcribe testican-2. In the accessory olfactory bulb weak staining of some mitral cells could be detected.
In the cerebral cortex neurons were stained throughout the different layers (Fig. 5C). Strong transcription of the testican-2 gene was also detected in the hippocampus, with both dentate gyrus and hippocampus proper being labeled. Signals appeared first in pyramidal neurons of CA3 and CA4 regions, although the remaining pyramidal neurons and granular cells of the dentate gyrus also became clearly stained (Fig. 5D).
In cerebellum the Purkinje cells showed the most pronounced labeling, but dispersed cells were also detected in the molecular and in the granular layer (Fig. 5E). Testican-2 transcripts in cells of the molecular layer were attributed to stellate and basket cells. We assume that the large cells labeled in the granular cell layer are Golgi cells. The small granular cells were not stained, although as seen in the granular cells of the olfactory bulb, diffuse background staining was observed.
Pronounced labeling for testican-2 mRNA was also found in inferior and superior colliculi and in medullary nuclei. In the thalamus the anterodorsal nucleus was most prominently stained. Blood vessels, meninges, and choroid plexus did not show a signal for testican-2.
We have expressed and purified a fragment encompassing the EC and TY domain pair of testican-2 (residues 198-377) in mammalian cells. The signal peptide region of BM-40 (Kohfeldt et al., 1997), which directed the protein to the culture medium, had been connected to the EC-TY module pair and added an N-terminal APLA to the authentic testican-2 sequence. Edman degradation demonstrated a single N-terminal sequence APLAETXTGQD as expected. The testican-2 EC-TY module pair showed a distinct conformation, as demonstrated by CD spectroscopy in the far UV region (Fig. 6A). The spectra were characteristic for a protein with high α-helical content with a shoulder at 222 nm. On addition of Ca2+ the molar ellipticity at 222 nm decreased by 27%. Addition of excess EDTA resulted in a complete reversal of the Ca2+-induced effect (Fig. 6A). The decrease in molar ellipticity at 222 nm was used to monitor quantitatively calcium binding. A model describing binding of calcium to a single site could be fitted and resulted in an equilibrium dissociation constant of KD = 32 μM (Fig. 6B). Fitting a model with cooperative binding sites revealed no cooperativity (Hill coefficient nH = 1.04).
We have identified the cDNA coding for testican-2, a novel modular protein expressed in brain, lung, and testis. It shows the same domain organization as the proteoglycan testican(-1) (Alliel et al., 1993) and was hence named testican-2. Testican-1 was isolated from human seminal plasma as a proteoglycan carrying chondroitin sulfate and heparan sulfate chains (Bonnet et al., 1992). Both glycosaminoglycan attachment sites are conserved between testican-1 and testican-2. We therefore assume that testican-2 will also be secreted as a proteoglycan in vivo.
Testican-2 cDNA encodes 424 residues. Sequence comparisons and the recently solved x-ray structure of the homologous protein BM-40 (Hohenester et al., 1997) allowed us to deduce a detailed domain model for testican-2 comprising five independently folded protein modules: An N-terminal domain unique to testicans is followed by an FS module, a calcium-binding EC module, a TY module, and a C-terminal region encompassing two glycosaminoglycan attachment sites. FS and EC domains are characteristic for the BM-40/SPARC/osteonectin protein family (Lane and Sage, 1994; Hohenester et al., 1997).
From the structure-based alignment we predict that the FS and EC domains of testican-2 possess a folding similar to that of BM-40 (Fig. 2). BM-40 binds two Ca2+ with high affinity and cooperativity in its EC domain (Hohenester et al., 1997; E. Busch and P. Maurer, unpublished data). Here we show that Ca2+ induces a large conformational change in testican-2. The affinity for calcium is moderate (KD = 32 μM) and does not show any signs of cooperativity. This is readily explainable by the sequence of the EF-hand 2 of testican-2, where Tyr292 is found at a position normally occupied with an acidic residue or a serine (Drake et al., 1997) directly coordinating the Ca2+ (Fig. 2). This substitution presumably prevents Ca2+ binding to EF-hand 2 and caused EF-hand 1 to bind one Ca2+ with low affinity. Similar Ca2+-binding characteristics with noncooperative binding and a KD = 68 μM had been shown for testican-1 (Kohfeldt et al., 1997).
A collagen-binding site was localized in the EC domain of BM-40 and characterized at the atomic level (Maurer et al., 1995; Sasaki et al., 1998). In an Ala mutagenesis screening it was found that, among others, substitution of Ala for Asn173 drastically decreased the affinity of BM-40 to collagens (Sasaki et al., 1998). Asn173 of BM-40 corresponds to the aliphatic residues Leu217 in testican-2 and Ala216 in testican-1 (Fig. 2). This is obviously the reason why testican-1 does not bind to collagens (Kohfeldt et al., 1997), and the same is expected for testican-2.
TY domains are found in proteins such as nidogen-1 and -2, thyroglobulin, saxiphilin, insulin-like growth factor binding proteins, major histocompatibility class II invariant chains, the cell surface antigen EPG, and tumor-associated antigen GA-733 (for review, see Molina et al., 1996). It is interesting that insulin-like growth factor-binding proteins are involved in neurotransmission in the hippocampus (Werther et al., 1998).
In situ hybridizations on brain sections showed a widespread expression of testican-2 in many neuronal cell types. In main fiber tracts (corpus callosum, cerebral peduncles, and fimbria fornix) that contain glial cells but not neuronal cell somata, no expression of testican-2 was detectable (Fig. 5). Thus, we assume that testican-2 expression is mainly restricted to neuronal cells. In contrast, other members of the BM-40 family, including BM-40, SC1, and QR1, are products of astrocytes (Mendis et al., 1994, 1995; Casado et al., 1996; McKinnon and Margolskee, 1996). We could not detect significant differences in the spatial expression of testican-1 and testican-2, although the level of expression might vary within one type of neuron. We are currently investigating the expression of both testicans during brain development. During the course of our studies a gene termed KIAA0275 was identified in a random screening project of brain cDNAs (Nagase et al., 1996) and turned out to be identical to testican-2. In the EST database we could furthermore detect 87 human testican-2 ESTs. Fifty-eight ESTs (68%) were derived from brain tissues and included cerebellum, hippocampus, cortex, and substantia nigra. Twelve ESTs (14%) were derived from ovary or testes, and five (6%) were derived from retina. Prostate and kidney made up another 6% of the ESTs, with the remaining ones being derived from different tumor tissues, among them one neuroendocrine lung tumor. Thus, the spatial expression of testican-2 mRNA in the human and mouse seems to be very similar. Two sequence-tagged sites (accession nos. WI-6816 and Cda0fh11) derived from the 3′ UTR of human testican-2 and KIAA0275 (Nagase et al., 1996) have been independently mapped to chromosome 10q21-23 distinct from the localization of testican-1 on chromosome 5 (Charbonnier et al., 1998). The gene DFNB12 causing autosomal recessive deafness was mapped in this region (Chaib et al., 1996). Affected individuals have profound sensorineural hearing impairment, and it will be interesting to determine whether a neuronal defect caused by testican-2 may be involved.
Electron microscopic immunocytochemistry revealed a postsynaptic localization of testican-1 in hippocampus, cortex, and cerebellum. The labeling was concentrated on the postsynaptic densities, whereas presynaptic areas and synaptic clefts lacked testican-1 immunoreactivity (Bonnet et al., 1996). Agrin, a heparan sulfate proteoglycan containing FS domains, is also found at postsynaptic densities and is essential for the formation of synaptic structures at the neuromuscular junction and presumably also at interneuronal junctions (Gautam et al., 1996; Cohen et al., 1997; Ruegg and Bixby, 1998). It remains to be shown whether testican-2 protein is also localized in synaptic structures and whether it contributes to synaptic architecture and/or transmission.
The unique mosaic domain composition of testicans gives them the potential to modulate growth factor activity, cell adhesion, axonal growth, and guidance as well as synaptogenesis and synaptic transmission in the brain. The presence of a Ca2+-binding domain offers the exciting opportunity that the function of these proteoglycans may be regulated by extracellular Ca2+ concentration changes under normal and pathological conditions (Pumain and Heinemann, 1985; Jefferys, 1995; Maurer et al., 1996; Schäfer and Heizmann, 1996; Antanitus, 1998).
This work was supported by the Deutsche Forschungsgemeinschaft [project Ma 1932-1 and the Graduiertenkolleg (no. 106)] and the Köln Fortune Program of the Faculty of Medicine, University of Cologne. We thank Drs. C. Andressen, S. Arnhold, N. Smyth, and M. Plomann for critical reading of the manuscript and valuable discussions and E. Kohfeldt (Max-Planck-Institut for Biochemistry, Martinsried, Germany) for providing us with the testican-1 probe.