Testican-3: a brain-specific proteoglycan member of the BM-40/SPARC/osteonectin family


Address correspondence and reprint requests to Dr. Ursula Hartmann, Center for Biochemistry, Medical Faculty, University of Cologne, Joseph-Stelzmann-Str. 52, D-50931 Cologne, Germany. E-mail: ursula.hartmann@uni-koeln.de


The testicans are a three-member family of secreted proteoglycans structurally related to the BM-40/secreted protein acidic and rich in cystein (SPARC) osteonectin family of extracellular calcium-binding proteins. In vitro studies have indicated that testicans are involved in the regulation of extracellular protease cascades and in neuronal function. Here, we describe the biochemical characterization and tissue distribution of mouse testican-3 as well as the inactivation of the corresponding gene. The expression of testican-3 in adult mice is restricted to the brain, where it is located diffusely within the extracellular matrix, as well as associated with cells. Brain-derived testican-3 is a heparan sulphate proteoglycan. In cell culture, the core protein is detected in the supernatant and the extracellular matrix, whereas the proteoglycan form is restricted to the supernatant. This indicates possible interactions of the testican-3 core protein with components of the extracellular matrix which are blocked by addition of the glycosaminoglycan chains. Mice deficient in testican-3 are viable and fertile and do not show an obvious phenotype. This points to a functional redundancy among the different members of the testican family or between testican-3 and other brain heparan sulphate proteoglycans.

Abbreviations used
EC domain

extracellular calcium-binding domain


expressed sequence tag

FS domain

follistatin-like domain


matrix metalloproteinase


sodium dodecyl sulphate


secreted protein acidic and rich in cystein


saline-sodium citrate


thyroglobulin domain

The testicans form a subgroup within the BM-40/SPARC/osteonectin family of modular extracellular proteins (Hartmann and Maurer 2001) and have also been referred to as the SPOCKs [secreted protein acidic and rich in cystein (SPARC)/Osteonectin CWCV and Kazal-like domains proteoglycan; Charbonnier et al. 1998]. Their modular structure is characterized by an N-terminal testican-specific domain followed by the follistatin-like (FS) and extracellular calcium-binding (EC) domains characteristic of the BM-40 family. Towards the C-terminus they contain a thyroglobulin-like domain (TY) and a novel sequence (domain V), which includes two potential glycosaminoglycan attachment sites.

Testican-1 was originally isolated as a proteolytic fragment from human seminal plasma carrying chondroitin sulphate and heparan sulphate chains (Bonnet et al. 1993). In the mouse, testican-1 is predominantly expressed in the nervous system during embryonic development and its expression correlates with periods of neuronal migration and axonal growth (Charbonnier et al. 2000). In adult mice, expression is restricted to the brain where it is located at post-synaptic densities (Bonnet et al. 1996). In search for homologous members of the BM-40 family, we identified cDNAs that were most similar to testican-1 and designated the corresponding proteins testican-2 and -3 (Vannahme et al. 1999; Hartmann and Maurer 2001).

In adult mice, testican-2 shows a broader tissue distribution than testican-1. Although the proteoglycan can be found primarily in the nervous system, it is also deposited in lung and especially in endocrine glands. In most organs, testican-2 is present throughout the extracellular space or localized at the cell surface, indicating that it may interact with membrane-bound molecules. (Schnepp et al. 2005). In contrast to testican-1, which carries both heparan and chondroitin sulphate chains, testican-2 is a pure heparan sulphate proteoglycan (Schnepp et al. 2005).

Functional information on testicans is still scarce, but there are clear indications that these may serve in the regulation of extracellular protease cascades. When expression cloning was used to identify regulators of pro-matrix metalloproteinase (MMP)-2 processing, mediated by membrane-type (MT) 1-MMP, a splice variant of testican-3, called N-Tes, was found (Nakada et al. 2001). Indeed, it was shown that testican-1 and testican-3 inhibit pro-MMP-2 activation by either MT1-MMP or MT3-MMP and that testican-2 abrogates the inhibition by the other family members (Nakada et al. 2003). Similarly, testican-1, but not testican-2, inhibits cathepsin L, (Bocock et al. 2003; Pavsic et al. 2008). Testican-1 was also found to inhibit attachment and neurite outgrowth in cultures of N2A neuroblastoma cells (Marr and Edgell 2003), While testican-2 is able to inhibit neurite outgrowth from primary cerebellar cells (Schnepp et al. 2005). Yet, inactivation of testican-1 in mice did not lead to any obvious phenotype (Röll et al. 2006).

In this study, we describe the biochemical characterization of mouse testican-3, its tissue distribution and the localization of the secreted protein in cell culture. Further, we describe the generation of mice deficient in testican-3 which, however, are viable and fertile and do not show any obvious morphological phenotype.

Materials and methods


All animal procedures were approved by local government authorities (Bezirksregierung, permit no. 20.11.245). Testican-3 deficient mice were backcrossed onto the C57Bl/6N background. Wild-type mice were purchased from Charles River, Sulzfeld, Germany. In the analysis of testican-3 deficient mice, gender-matched wild-type littermates were used as controls. Adult mice were 6 weeks.

Isolation of cDNA- and genomic clones

A mouse brain cDNA library (ICR outbred strain, Harlan Sprague Dawley, newborns; Stratagene, La Jolla, CA, USA) was screened with a 32P-labelled fragment derived from a human EST clone (ImaGenes, Berlin, Germany). Positive plaques were excised and rescreened. One positive plaque remaining after several rounds of rescreening was in vivo excised, yielding a partial testican-3 cDNA in the pBluescript vector. The complete sequence of the cDNA was identified by rapid amplification of cDNA ends (RACE) using the SMART RACE cDNA amplification KIT (Takara Bio Europe/Clontech, Saint-Germain-en-Laye, France) and the following primers: 5′-CAG CAC TGC CCT ACA CTG CCA TGG C-3′ and 5′-CGG AGA TCA GTA ACA TTC AGA AGC G-3′. RACE products were cloned into the pCRII vector (Life Technologies, Darmstadt, Germany) and sequenced.

A cDNA-fragment covering the complete coding region of testican-3 was amplified by PCR and used to screen a murine genomic 129/ola cosmid library (Nr. 121; ImaGenes). Several clones covering all exon-intron boundaries were further characterized by restriction fragment analysis.

RNA preparation, northern blotting and RT-PCR

Total RNA was isolated using the guanidinium isothiocyanate method (Chomszynski and Sacchi 1987). Poly(A) + RNA was prepared using the Qiagen Oligotex mRNA kit (Hilden, Germany). For northern analysis, 3 μg of mRNA was fractionated on a 1% denaturing agarose gel, blotted and hybridized overnight at 65°C with 32P-labelled cDNA probes specific for testican-3 or glyceraldehyde phosphodehydrogenase gene (GAPDH). Blots were washed three times in 0.5X SSC, 0.1% sodium dodecyl sulphate (SDS) for 15 min at 65°C and exposed overnight on a phosphoimager photoplate (Molecular Dynamics, Sunnyvale, CA, USA). For semiquantitative RT-PCR, 200 ng of total RNA were transcribed into cDNA and amplified using the following primers: 5′-AAC TAT GAC CTG CTA TTG GAC-3′ (testcan-3 forward), 5′-CAT TCA CGA AAA TCT CCA C-3′ (testican-3 reverse), 5′-TCA CTG TGC CTG AAC TTA CC-3′ (beta-tubulin forward) and 5′-GGA ACA TAG CCG TAA ACT GC-3′ (beta-tubulin reverse). The chosen conditions were as follows: 30 min 42°C (reverse transcription) followed by a 3 min denaturation step of 95°C and 30 cycles of 30 s denaturation at 95°C, 30 s annealing at 50°C and 1 min elongation at 72°C. Tubulin served as a housekeeping control. All reactions were performed in the presence or absence (negative control) of reverse transcriptase.

Recombinant expression of the full-length testican-3 protein

Based on the RACE construct a fragment coding for domains I–V of the protein was cloned into the eukaryotic expression vector pCEP-Pu vector 3′ of a sequence coding for the BM40 signal peptide and a His6-Myc tag followed by a factor X cleavage site. Plasmids were transfected into the human embryonic kidney cell line EBNA-293 (Life Technologies). Conditioned serum-free media of EBNA-293 cells were collected, passed over a Ni-NTA column (Qiagen, Hilden, Germany) and testican-3 was eluted with a linear gradient of 0.005–0.25 M imidazol, in 50 mM sodium phosphate, pH 8.0, containing 0.3 M sodium chloride. The testican-3 containing fractions were pooled and after dialysis against 50 mM Tris-HCl, pH 7.4, submitted to ion exchange chromatography on a Mono Q column (GE Healthcare Europe, Glattbrugg, Switzerland). The column was eluted with a linear NaCl gradient (0–1 M) and testican-3 detected in two peaks representing the core protein (apparent molecular weight 60 kDa) and the proteoglycan form (apparent molecular weight 70–150 kDa) respectively.

Preparation of specific antibodies against mouse testican-3

A mixture of equal amounts of the proteoglycan and core protein form of recombinant mouse testican-3 was used to immunize a rabbit. The resulting antiserum was passed through a column of CNBr-Sepharose (GE Healthcare Europe) to which recombinant testican-3 had been coupled and the antibodies bound to testican-3 were eluted with 3 M KSCN, 50 mM Tris-HCl, pH 7.4, followed by rapid dialysis against 0.15 M NaCl, 50 mM Tris-HCl, pH 7.4.

Immunoprecipitation and western blot analysis

Tissue samples were dissected and briefly homogenized with a Dounce homogenizer in a fivefold volume of ice-cold 0.15 M NaCl, 50 mM Tris-HCl, pH 7.4, containing 2 mM EDTA, 1% NP-40 and the Complete protease inhibitor cocktail (Roche, Mannheim, Germany). After incubation for 30 min on ice, tissue residues were removed by centrifugation for 30 min at 21 000 g and 4°C. For detection, the testican-3 protein was first enriched by immunoprecipitation using the affinity-purified rabbit antibody followed by protein A agarose (Roche) and the immunoprecipitates dissolved in SDS-sample buffer. SDS polyacrylamide gel electrophoresis was performed on 4–15% gradient gels, followed by electrophoretic transfer to a nitrocellulose membrane (Protran NC, Schleicher & Schüll, Bioscience, Dassel, Germany). Proteins were detected with the affinity-purified rabbit polyclonal antibody against testican-3. Detection was with the appropriate secondary antibodies coupled to horseradish peroxidase and the enhanced bioluminescence (ECL) kit (GE Healthcare Europe).

Glycosaminoglycan chain analysis of tissue-derived testican-3

Tissue extracts were digested with specific enzymes to determine the nature of the glycosaminoglycans attached to testican-3. Aliquots corresponding to extracts from 50 to 200 mg tissue wet weight were digested. Heparan sulphate was degraded by simultaneous digestion with 0.1 mIU heparinase I (Sigma, Taufkirchen, Germany) and 0.5 mIU heparinase III (Sigma) in 50 mM NaCl, 50 mM Tris-HCl, 4 mM CaCl2, pH 7.5, at 37°C overnight. Dermatan sulphate and chondroitin sulphate were cleaved by digestion with 30 mIU chondroitinase ABC (Seikagaku, Tokyo, Japan) in 50 mM NaCl, 50 mM sodium acetate, 50 mM Tris-HCl, pH 7.5, at 20–24°C for 4 h with an equal amount of enzyme added again after 2 h. Combined digestions were performed sequentially, first in the heparinase buffer and then after adding sodium acetate to a concentration of 50 mM.

Immunofluorescence detection of testican-3

Brains from adult mice were embedded in TissueTek (Sakura, Staufen, Germany) and 7 μm thick sagittal cryosections prepared. These were air-dried at 20–24°C for 1 h, fixed for 10 min in 2% paraformaldehyde, washed three times in phosphate buffered saline (PBS) and blocked for 1 h with 1.5% normal goat serum in PBS containing 0.1% Triton-X-100. The sections were incubated overnight with an affinity-purified rabbit polyclonal antibody against testican-3, followed by a 1 h incubation with an Alexa 488-coupled secondary antibody (Life Technologies).

Analysis of testican-3 distribution in cultures of transfected cells

The cDNA construct covering testican-3 domains I–V in pCEP-Pu was used to stably transfect HT1080 cells, These were grown to confluency followed by an incubation in serum-free medium for 24 h. The supernatant was collected and mixed with SDS-sample buffer. The dish was washed 3x with PBS at 20–24°C and the cells were incubated in 20 mM EDTA, 1 mM phenyl methylsulfonyl fluoride, 10 μg/mL leupeptin, and 10 μg/mL soybean trypsin inhibitor in PBS over night at 4°C. Cells were collected by centrifugation in the same buffer, suspended in SDS-sample buffer and the lysate centrifuged to remove cell debris. The remaining matrix was washed 3x with PBS, covered with SDS-sample buffer and scraped off. The samples were separated by SDS-PAGE and analysed by western blot. In addition to antibodies against testican-3, we used antibodies against fibronectin (MAB1932, Millipore, Schwalbach, Germany) and the Golgi protein GM-130 (610823, BD Transduction Laboratories, Heidelberg, Germany) as markers for the extracellular matrix and the cellular fraction respectively.

Statistical analysis

Data were analysed by Student's t-test using SPSS (IBM, Cologne, Germany) and displayed as means ± SEM (n = 3). p-values of < 0.05 were considered significant.


Organization of the murine testican-3 protein and the Ticn-3 gene

The complete sequence of the cDNA (GenBank accession number AJ278998) was cloned and determined by (RACE) using a 550 bp fragment derived from a mouse brain cDNA library. An open reading frame coding for a putative protein of 436 amino acids is preceded by a short 5′-UTR and followed by 1.5 kb of 3′ UTR containing a canonical polyadenylation sequence. Sequence comparison showed that the mouse protein is 89.9% identical to human testican-3 over its complete length. The overall domain structure is identical to testican-1 and testican-2 (Fig. 1). The N-terminus, a hydrophobic stretch of 23 amino acids, is predicted to function as a signal peptide. The mature protein with a calculated molecular mass of 46790 Dalton can be further subdivided into five domains. The N-terminal region (residues 24–88) does not show any homology to known proteins except testican-1 and testican-2. It is followed by a cysteine-rich domain (residues 89–199) homologous to follistatin (FS). A third domain (residues 200–315) is homologous to the extracellular calcium-binding (EC) domain of BM-40, characterized by two Ca2+-binding EF-hand motifs that were shown to be functional in testican-1 (Kohfeldt et al. 1997) and -2 (Vannahme et al. 1999). Following the EC-domain, a thyroglobulin-like (TY) domain (residues 316–381) and a C-terminal domain (residues 382–436), unique to testicans, is found. The latter contains two putative glycosaminoglycan attachment sites at Ser 387 and Ser 392.

Figure 1.

Structure-based alignment of the domains within the testican family. Identical amino acids are underlayed with light grey. Numbers above the sequences correspond to cysteines forming disulphide bonds shown in white font underlayed with black. Calcium coordinating residues are underlayed with dark grey. Asterisks above the sequences indicate potential glycosaminoglycan attachment sites. Vertical bars indicate predicted signal peptide cleavage sites (SignalP 3.0 Server). FS, follistatin-like domain; EC, extracellular calcium-binding domain, TY, thyroglobulin-like domain.

A set of genomic cosmid clones covering the whole Ticn-3 gene was obtained after screening a murine genomic mouse library. The gene structure was elucidated by analysis of these cosmid clones and a genomic contig (NT_039461), originating from chromosome 8. The murine Ticn-3 gene spans at least 406 kb and consists of 12 exons (Fig. 2). Each domain of the testican-3 core protein in encoded by one or more exons and the domain borders coincide with splice sites.

Figure 2.

Structure of the Ticn-3 gene. Schematic representation of the Ticn-3 gene and the relationship between exon-intron structure and protein domains. The alternatively spliced tripeptide EVE is encoded by exon 3. The lower scheme gives the codon phases. SP, signal peptide; TCAN, testican-specific domain I; FS, follistatin-like domain; EC, extracellular calcium-binding domain, TY, thyroglobulin-like domain; V, C-terminal domain V.

In the testican-1 gene, we could identify an alternatively spliced microexon coding for the three amino acids EVE in the testican-specific domain I (Röll et al. 2006). The sequence coding for EVE was conserved in testican-3 (Fig. 1). Analysis of genomic sequences indicated that also in the Ticn-3 gene the three amino acids EVE are encoded by an independent exon. Screening of the EST database led to the identification of 26 ESTs for murine testican-3, of which 22 corresponded to the cloned testican-3 cDNA, whereas four lacked EVE. Thus, alternative splicing of the microexon occurs also in testican-3. The exon is conserved in the human testican-3 gene (not shown). For human testican-3, 100 ESTs are present in the database with 94 containing the microexon and 6 lacking EVE.

Expression of mouse testican-3 is mainly restricted to the brain

In northern blot analysis of a variety of adult mouse tissues a transcript of about 3.1 kb could be detected only in brain mRNA, whereas all other tissues analysed showed no expression of testican-3 mRNA (Fig. 3a). Semi-quantitative RT-PCR revealed that earlier during post-natal life the protein is also expressed in other organs like heart and lung but to a much lesser extent. A faint band was also present in the thymus of adult mice (Fig. 3b). An affinity-purified antibody was used to assess the tissue distribution of testican-3 in newborn and adult mice (Fig. 4a and b). Immunoprecipitation of testican-3 followed by SDS-PAGE and immunoblot analysis revealed a polydisperse band in brain extracts but not in extracts from any other tissue analysed. The apparent molecular weight of 80–130 kDa was similar to that of the proteoglycan form of the recombinant protein (Fig. 4a and b, right lanes). A band corresponding to the 60 kDa core protein of recombinant testican-3 was missing in the tissue extract, indicating that native testican-3 occurs mainly as a proteoglycan.

Figure 3.

Distribution of testican-3 mRNA in adult mouse tissues. (a) Northern hybridization was done with 3 μg of poly(A)+ RNA from different mouse tissues. The blot was hybridized with a 32P labelled fragment of the coding region of testican-3 cDNA (upper panel) or with glyceraldehye phosphodehyrogenase GAPDH (lower panel) as a loading control. (b) Semiquantitative RT-PCR was performed on RNA samples from mouse organs at different postnatal (P) days.

Figure 4.

Distribution of testican-3 protein in mouse tissues. Tissues from newborn (a) and adult (b) mice were extracted and testican-3 immunprecipitated. (c): Brain extracts were immunoprecipitated with the testican-3 antibody and subsequently treated with chondroitinase (Ch′ase) ABC or with a mixture of heparinase (H′ase) I and III or with all enzymes. The samples were separated on 4–15% sodium dodecyl sulphate (SDS) polyacrylamide gels under reducing conditions and the immunoblots developed with the testican-3 antibody. Molecular masses of marker proteins are given at the left. The core protein (CP) and the proteoglycan form (PG) of testican-3 are indicated to the right, asterisks indicate immunglobulin heavy chains derived from the precipitating antibody. Testican-specific domain I (TCAN)-3 rec, recombinant testican-3 protein, serving as positive control.

Brain-derived testican-3 is a heparan sulphate proteoglycan

Next, we wanted to verify if the polydisperse band detected in immunoblots of the tissue-derived testican-3 indeed represents the proteoglycan form and determine the nature of the glycosaminoglycan chains. We performed digestions with chrondroitinase ABC or a mixture of heparinase I and III or both (Fig. 4c). Treatment of testican-3 immunoprecipitated from adult mouse brain with heparinase I/III resulted in a shift towards a lower molecular weight with a sharp band appearing at around 66 kDa. Treatment with chondroitinase ABC gave just a slight shift of the proteoglycan smear. Upon digestion with both enzymes, almost all testican-3 ran as a distinct 60 kDa core protein band. Apparently, the tissue form of testican-3 is a nearly pure heparan sulphate proteoglycan. Also, smaller forms of testican-3 were detected which probably represent degradation products. The core protein of recombinant testican-3 was digested with peptide N-glycosidase F, an endoglycosidase cleaving N-glycosidically linked glycans. In contrast to the recombinant testican-2 core protein, which has been shown to be N-glycosylated (Schnepp et al. 2005) and was here used as positive control, for the testican-3 core protein almost no shift was seen in SDS-PAGE (Figure S2a). The same was true for brain-derived testican-3, which was treated with chrondroitinase ABC and heparinase I and III to generate the core protein prior to the incubation with N-glycosidase F (not shown). Analysis of O-glycans was performed using MALDI mass spectrometry and revealed the presence of several mucin-type O-glycans (Figure S2b).

Testican-3 shows a widespread distribution within the extracellular matrix of the brain

Immunofluorescence staining revealed the presence of testican-3 in most regions of the brain although the expression was low (Fig. 5). In the forebrain, the striatum, the thalamus and to a lesser extent the cortex was stained. A weak staining could also be detected in the CA3 region of the hippocampus. In the hindbrain, only the Purkinje cells in the cerebellum showed a weak signal. The staining was in part diffuse, indicating that testican-3 is deposited within the extracellular matrix of the brain. In addition, some, but not all cells were stained for testican-3. This could either reflect intracellular pools or the association of the proteoglycan with the cell surface of certain subtypes of neurons.

Figure 5.

Distribution of testican-3 in adult brain determined by immunofluorescence microscopy. Immunostainings were done on sections of either wildtype or testican-3 deficient mice using the testican-3 antibody followed by a fluorochrome-coupled secondary antibody. Co, cortex; hc, hippocampus; st, striatum; th, thalamus; cb, cerebellum. Bar, 50 μm.

Recruitment of testican-3 to the extracellular matrix is blocked by the glycosaminoglycan chains in cultures of HT1080 cells

A cDNA-fragment spanning the complete coding region of mature mouse testican-3 was expressed in HT1080 cells, which do not endogenously express testican-3. Cells were grown to confluency and incubated in serum-free medium for 24 h. Supernatant, extracellular matrix and cell fractions were collected and separated by SDS-PAGE followed by immunoblotting with the affinity-purified testican-3 antibody. In the cell lysate, only a band with the size of the core protein was visible. Among the secreted testican-3 forms, only the core protein was found in the matrix fraction, whereas in the supernatant the core protein and the proteoglycan form were present (Fig. 6). The results indicate that only the testican-3 core protein can be retained in the extracellular matrix. In contrast, attachment of glycosaminoglycans to the core protein blocks its recruitment into the matrix.

Figure 6.

Distribution of testican-3 forms in cell culture. HT1080 cells were stably transfected with a plasmid encoding mouse testican-3 carrying an N-terminal His6-Myc-factor X tag. Cells were grown to confluency, cultured in serum-free medium for 24 h and the supernatant collected. Cells were incubated in phosphate buffered saline (PBS) containing EDTA over night at 4°C and collected. The matrix was scraped off the dish and lysed in sodium dodecyl sulphate (SDS) sample buffer. The distribution of testican-3 forms was analysed by reducing SDS PAGE on 4–15% gels and detected by immunoblotting with the testican-3 antibody. As a control for the isolation procedure, the different samples were probed with antibodies against fibronectin and the Golgi protein GM130. The molecular masses of marker proteins are given at the left, proteoglycan form (PG) and the core protein (CP) are indicated at the right.

Testican-1 deficient mice lack an overt phenotype

The mouse Ticn3 gene was inactivated by homologous recombination in embryonic stem cells with a targeting vector carrying a neomycin cassette (Figure. S1a). Insertion of the vector into the Ticn3 locus should interrupt the reading frame in exon 2, which contains the translation start site and codes for the signal peptide and part of the N-terminal domain unique for testicans. Three correctly targeted clones were identified by Southern blot analysis of genomic DNA using an external or internal probe (not shown). For two of the three clones, germ line transmission was obtained. Heterozygous breeding produced wildtype, heterozygous and mutant offspring (Figure S1b) in a normal Mendelian ratio (not shown). The null mutation was confirmed on northern blots (Figure S1c) and by immunoprecipitation followed by western blot analysis (Figure S1d), using an affinity-purified antibody against testican-3.

Testican-3 deficient mice showed no obvious morphological or behavioural abnormalities, were fertile and had normal life spans. Since the expression is mainly restricted to the brain, we focused further studies on this tissue. Brains of mutant animals had an identical gross morphology to those of wildtype littermates. Similarly, a light microscopic comparison of wildtype and mutant brains showed no differences (Fig. 7). The distribution, organization and numbers of the major brain cell types was also not altered, as judged by qualitative and quantitative immunohistochemistry/immunofluorescence detection using markers for neurons (neuron-specific enolase, NeuN) or glial cells (glial fibrillary acidic protein, GFAP) (Figures S3 and S4 respectively).

Figure 7.

Analysis of brain morphology and histology of testican deficient mice in adult and developing testican-3 deficient mice. (a) No difference could be detected in the gross morphology, length or weight of brains derived from 6-week-old testican-3 deficient mice compared to wild-type littermates (n ≥ 3). (b) Nissl stainings of sagittal (upper panel) and coronal (lower panel) sections of 6-week-old mouse brains do not reveal any obvious histological difference. (c) No obvious alterations could be detected between testican-3 deficient mice and wild-type littermates in Nissl stained sections of E (embryonic day) 18.5 (horizontal sections, upper panel) and E15.5 (sagittal sections, lower panel) mouse embryos. Scale bars represent 1000 μm.


We have identified and cloned the cDNA for murine testican-3. The mature protein shows the same domain organization as testican-1 and testican-2. An N-terminal domain unique for testicans is followed by paired follistatin-like (FS) and extracellular calcium-binding (EC) domains, a thyroglobulin-like (TY) domain and a C-terminal domain with two glycosaminoglycan attachment sites. The paired FS and EC domains are the signature of the BM-40/SPARC/osteonectin family. Testicans comprise the only members of that family which form proteoglycans as shown earlier for testican-1 (Bonnet et al. 1993) and -2 (Vannahme et al. 1999). The gene coding for testican-3 spans 406 kb and consists of 12 exons. The exon-intron structure with 12 exons is similar to that of Ticn-1 (Röll et al. 2006), whereas the Ticn-2 gene is composed of 11 exons (own unpublished data).

Exon 3 of Ticn-3 is a microexon coding for the three amino acids EVE, conserved among the testicans. We could earlier show that in Ticn-1 this exon is alternatively spliced and that transcripts with and without EVE are present in brain of adult mice (Röll et al. 2006). In the EST database 26 ESTs covering the region surrounding exon 3 of the murine and 100 ESTs covering that part of the human testican-3 gene are present. Although most of these (22/94) contain exon 3, some (4/6) lack the sequence, indicating that alternative splicing also occurs in testican-3 from both species. Microexons are present in a variety of eucaryotic genes and the preferred length is a multiple of three bp (Volfovsky et al. 2003). Examples include the gene for the neural cell adhesion protein NCAM where the insertion of a 30 bp exon is associated with a developmentally regulated loss of the stimulation of neurite outgrowth (Saffell et al. 1994).

The tissue distribution of testican-3 in mouse is similar to that of testican-1 (Bonnet et al. 1996) in that it is mainly restricted to the nervous system. In humans both proteoglycans are more widely expressed (Marr et al. 1997; NCBI Unigene EST profile viewer). After immunoprecipitation and subsequent western blot, testican-3 derived from brain extracts of newborn or adult mice was detected as a polydisperse band with an apparent molecular weight of 70–130 kDa. In contrast, in samples of recombinant testican-3 an additional band of an apparent molecular weight of about 60 kDa is present, representing the core protein. Treatment of brain-derived testican-3 with heparanase I and III led to a shift of the polydisperse band to the size of the core protein, whereas chontroitinase ABC had no effect. The results reveal that native testican-3 is a pure heparan sulphate proteoglycan. The value of about 60 kDa for the core protein is considerably higher than the 46790 Da calculated from the amino acid sequence of the recombinant protein. This may be because of irregular migration behaviour upon SDS-PAGE or post-translational modifications. For testican-2, we have already demonstrated the usage of an N-glycosylation site (Schnepp et al. 2005). For testican-3, neither N-glycosylation nor oligosaccharide O-glycosylation sites are predicted by the NetNGlyc 1.0 or NetOGlyc 3.1 (Julenius et al. 2005) software. We could confirm the lack of N-liked glycans in testican-3 by digestion with peptide N-glycosidase F. In contrast, despite the lack of predicted attachment sites, several mucin-type O-glycans could be demonstrated by mass spectrometry, which may account for the difference between the calculated and the apparent molecular mass.

On tissue sections a diffuse immunostaining for testican-3 was seen in some regions of the brain, indicating that testican-3 is a component of the brain extracellular matrix. Further, some but not all cells were stained for testican-3. The brain extracellular matrix, contributing 20% of the brain volume, is unique in its composition. It mainly consists of hyaluronan, chondroitin sulphate proteoglycans of the lectican family, tenascins and link proteins, whereas fibrous components like collagens or fibronectin are missing (reviewed in Rauch 2004). The components are secreted both by neurons and glia cells (Faissner et al. 2010) and are located either in the interstitial zones of the neuropil or in specialized structures of dense organized matrix, called perineuronal nets, which surround neurons and ensheath synapses (Celio et al. 1998). The diffuse immunostaining for testican-3 indicates that the proteoglycan is in part deposited dispersed in the neuropil. The more discrete cellular staining could indicate that it forms part of the perineuronal nets or that it binds to cells. We can also not exclude that the cellular staining reflects not yet secreted intracellular pools of testican-3. Testican-3 was mainly restricted to the grey matter, whereas in the main fibre tracts that contain glial but not neuronal cell somata, no testican-3 was detected.

Our results on the distribution of testican-3 protein are in agreement with situ hybridization data available from public resources.1 The ‘Allen Mouse Brain Atlas’ (http://mouse.brain-map.org/) contains in situ hybridization for testican-3 on adult brains. It shows a widespread expression by brain neurons. Further, a heat map representing the expression levels in different brain regions reflects our immunofluorescence results with the highest expression of testican-3 in the striatum and the thalamus and a lower expression in the cortex, the CA4 region of the hippocampus and the Purkinje cells of the cerebellum. Generally, low level of expression in agreement with our results at the protein level indicates that testican-3 is of low abundance. This is also reflected by that testican-3 cannot be detected by direct western blot analysis but requires prior concentration by immunoprecipitation. Similar in situ hybridization data are available from the ‘Brain Gene Expression Map’ (http://www.stjudebgem.org). This database also provides results from brains of younger mice (E11, E15, P7), revealing a higher expression in the CNS at earlier timepoints. This is also seen in western blots of immunoprecipitated testican-3 from mouse brains of different ages, with a more intense signal from brains of newborn mice (Fig. 4). The restriction to nervous tissues is also reflected in data from E14.5 mice in the ‘EMAGE gene expression database’ (http://www.emouseatlas.org/emage/).

In transfected HT1080 cells recombinant testican-3 was produced in two forms: a proteoglycan form and one representing the core protein. Analysis of the supernatant and extracellular matrix fractions of the cell cultures revealed that the core protein was present in both compartments, whereas the proteoglycan was restricted to the supernatant. This indicates that the glycosaminoglycan chains block the integration into the matrix. Related experiments showed that the integration of the human testican-3 core protein into the extracellular matrix is mediated by the N-terminal part of the protein. (Nakada et al. 2001).

Several members of the BM-40/SPARC/osteonectin family have been implicated in the establishment of neuronal connectivity. Testican-1 and testican-2 were shown to modulate attachment of neuronal cells and inhibit neurite outgrowth in vitro, respectively (Marr and Edgell 2003; Schnepp et al. 2005). SC1/hevin contributes to the correct neuronal positioning in the brain (Gongidi et al. 2004). Recently, it was shown that SPARC/BM-40/osteonectin regulates synaptic functions by preventing the maturation of cholinergic terminals (Albrecht et al. 2012). Yet, mice deficient in testican-1 or SC1 do not show an obvious phenotype (McKinnon et al. 2000; Röll et al. 2006). To elucidate the in vivo function of testican-3 we inactivated the gene in mouse. Testican-3 deficient mice developed normally were viable and fertile, and it did not show any obvious morphological or behavioural abnormalities. This may be explained by a functional redundancy within the BM-40/SPARC/osteonectin family. It can also not be excluded that the function of testican-3 is not because of the core protein but to the glycosaminoglycan chains. In that case also other brain heparan sulphate proteoglycans could account for a redundancy. Considering the pronounced homology between the three testicans and the functional similarities between testican-1 and testican-3, we expect that in vivo evidence for the role of these proteoglycans may be obtained through the generation of mice with double or triple deficiencies.


This study was supported by the “Deutsche Forschungsgemeinschaft” (HA 2263/2-2), the “Köln Fortune” program of the Medical Faculty of the University of Cologne and the Cologne Excellence Cluster on Cellular Stress Responses in Aging-associated Diseases (CECAD), University of Cologne. We thank Drs. Patrik Maurer and Christian Vannahme for contributions to early phases of the study.

Conflict of interest

The authors declare no conflict of interest.

  1. 1

    In situ hybridization data. Cited data on the in situ hybridization of testican-3 were taken from the following public databases: Allen Mouse Brain Atlas. Website: ©2012 Allen Institute for Brain Science. Allen Mouse Brain Atlas (AMBA) [Internet]. Available from: http://mouse.brain-map.org/. (Lein et al. 2007). Brain Gene Expression Map (BGEM) (http://www.stjudebgem.org) (Magdaleno et al. 2006). EMAGE gene expression database (http://www.emouseatlas.org/emage/) (Richardson et al. 2009). Detailed data on testican-3 from AMBA can be found at: http://mouse.brain-map.org/experiment/show/70634298 (ISH SPOCK3 coronal) and http://mouse.brain-map.org/experiment/show/583547 (ISH SPOCK 3 sagittal). Data from BGEM are available from: http://www.stjudebgem.org/web/view/probe/viewProbeDetails.php?id&#x00A0;=&#x00A0;1115. Data from EMAGE are available from EMAGE: 12674.