ABSTRACT
Signal regulatory protein-α (SIRPα) is a member of the Ig superfamily selectively expressed by neuronal and myeloid cells. The molecule mediates functional interactions with CD47/integrin-associated protein. Here we provide evidence for the tissue-specific glycosylation of neuronal and haematopoietic SIRPα. We demonstrate a major difference in the galactosylation of N-linked glycans isolated from neuronal (i.e. brain-derived) SIRPα as compared to myeloid (i.e. spleen-derived) SIRPα, with neuronal SIRPα almost completely lacking galactose. p4-galactosyltransferase assays demonstrated that this is most likely due to a low galactosylation capacity of the brain. In order to investigate the role of galactosylation of SIRPα in cellular interactions, soluble recombinant SIRPα glycoforms containing galactose (SIRPα-Fc) or lacking galactose (SIRPα(ΔGal)-Fc) were produced. Binding studies demonstrated superior binding of SIRPα(ΔGal)-Fc to cerebellar neurons and isolated lymphocytes. In contrast, SIRPα-Fc bound relatively strong to macrophages. These data show that the galactosylation of SIRPα determines its cellular binding specificity.
INTRODUCTION
Signal regulatory proteins (SIRPs) comprise a family of cell-surface glycoproteins selectively expressed by myeloid cells (macrophages, monocytes, granulocytes, dendritic cells) and neurons (Kharitonenkov et al., 1997; Timms et al., 1998; Adams et al., 1998; Veillette et al., 1998). SIRPs, also known as src homology 2 (SH2) domain-containing phosphatase substrate-1, brain immunoglobulin (Ig)-like molecule with a tyrosine-based activation motif, P84 or macrophage fusion receptor, form a subfamily within the Ig superfamily closely related to the antigen receptors Ig, TCR and MHC (Fujioka et al., 1996; Kharitonenkov et al., 1997; Sano et al., 1997; Saginario et al., 1998). SIRPα constitutes an extracellular region with three Ig-like domains, a single transmembrane domain, and a cytoplasmic tail containing two immunoreceptor tyrosine-based inhibitory motifs (ITIMs) (Fujioka et al., 1996; Kharitonenkov et al., 1997; Veillette et al., 1998; Sano et al., 1997; Comu et al., 1997; Brooke et al., 1998; Saginario et al., 1998). In addition to the SIRPα sequences cloned from rat, mouse and cattle, several human SIRP members have been identified (Kharitonenkov et al., 1997).
In fibroblasts, SIRPα has been shown to regulate signalling through receptor tyrosine kinases (RTKs) like the insulin receptor, epidermal growth factor receptor, and platelet derived growth factor receptor (Kharitonenkov et al., 1997; Takada et al., 1998), receptors which transduce extracellular signals regulating a diverse array of cellular responses including proliferation, differentiation and cell survival (Ullrich and Schlessinger, 1990). Upon binding of insulin or growth factors to RTKs, the tyrosines of the SIRPα ITIMs are phosphorylated and recruit/activate SH2 domain-containing phosphatases like SHP-1 or SHP-2, which may, in turn, positively or negatively affect signalling (Kharitonenkov et al., 1997; Ochi et al., 1997; Takada et al., 1998). The broadly expressed SHP-2 is required for activation of the mitogen-activated protein kinase cascade, which makes it a positive signal transducer for RTKs (Milarski and Saltiel, 1994; Noguchi et al., 1994; Kim et al., 1998). In contrast, SHP-1 is selectively expressed in haemopoietic cells and generally constitutes a negative regulator of function (Tonks and Neel, 1996; Neel and Tonks, 1997; Veillette et al., 1998). Furthermore, SIRPα negatively regulates cell activation induced by immunoreceptor tyrosine-based activation motif-containing immunoreceptors. A chimeric molecule consisting of the extracellular part of murine FcγRIIB and the transmembrane and cytoplasmic domains of human SIRPα was shown to inhibit IgE-induced mediator secretion and cytokine synthesis by mast cells (Liénard et al., 1999).
Besides their involvement in the regulation of signaling pathways, SIRPα has been demonstrated to participate in cell-cell communication by interaction with extracellular ligands. It has been demonstrated that SIRPα can stimulate the outgrowth of neurons in vitro (Sano et al., 1997). Furthermore, Saginario et al. (Saginario et al., 1998) showed that the extracellular domain of SIRPα could prevent fusion of macrophages. The discovery that a major ligand for mouse SIRPα is the integrin-associated protein (CD47/IAP) was recently reported (Jiang et al., 1999). Studies by Seiffert et al. (Seiffert et al., 1999) revealed that the extracellular regions of human SIRPα adhere to a number of primary haematopoietic cells and cell lines and they also identified human CD47/IAP as a prominent extracellular ligand.
The extracellular part of rat SIRPα contains 15 potential N-glycosylation sites and a large number of serines and threonines, which might become O-glycosylated (Fujioka et al., 1996; Sano et al., 1997; Adams et al., 1998; Saginario et al., 1998). A schematic representation of rat SIRPα is shown in Fig. 1. Several studies have provided evidence that SIRPα is strongly glycosylated (Fujioka et al., 1996; Kharitonenkov et al., 1997; Adams et al., 1998; Bartoszewicz et al., 1999). The difference in molecular mass observed between myeloid SIRPα (110 kDa) and neuronal SIRPα (85-90 kDa) suggest that SIRPα is tissue-specific glycosylated (Sano et al., 1989; Sano et al., 1990; Sano et al., 1997; Adams et al., 1998; Veillette et al., 1998).
The present study was conducted to investigate the tissue-specific glycosylation of SIRPα in more detail and to examine whether the differential glycosylation influences the binding capacity of SIRPα. We compared the glycans occurring on neuronal and myeloid SIRPα, isolated from rat brain and rat spleen, respectively, and found a major difference with respect to the extent of galactosylation. To analyze the possible significance of this tissue-specific glycosylation of SIRPα we produced soluble recombinant fusion proteins of the extracellular regions of SIRPα in Chinese hamster ovary (CHO) cells as well as in galactosylation-defective CHO cells (Lec8) (Deutscher and Hirschberg, 1986), which mimic splenic (i.e. haematopoietic) and brain (i.e. neuronal) SIRPα glycoforms, respectively. The proteins were analyzed for their capability to bind to different cells and tissues, which revealed that galactosylation influences the cellular binding-specificity of SIRPα?????
MATERIALS AND METHODS
Materials
CHO cells and Lec8 cells were obtained from the American Type Culture Collection (Rockville, Maryland, USA) and PC-12 cells from the European Collection of Cell Cultures (Salisbury, Wilts, UK). The NR8383 rat alveolar M? cell line was kindly donated by Dr R. J. Helmke (University of Texas, San Antonio, TX, USA). Media, sera and cell culture reagents were from Gibco-BRL (Paisley, Scotland). UDP-[3H]Gal (36.0 Ci/mmol) was purchased from New England Nuclear (Boston, MA, USA). The sugar nucleotide donor was diluted with unlabelled UDP-Gal (Sigma, St Louis, MO, USA) to give the desired specific radioactivity. N-acetylglucosamine (GlcNAc) and p-nitrophenyl-N-acetyl-1-thio-p-D-glucosaminide (GlcNAc-S-pNP) were purchased from Sigma. The pIgplus/ratSIRPα construct (pIgplus containing all three extracellular domains of rat SIRPα) was a kind gift of Dr D. L. Simmons (department of Neuroscience, SmithKline Beecham Pharmaceuticals, Harlow, Essex, UK). HiTrap protein A columns and protein A-sepharose were from Amersham Pharmacia (Uppsala, Sweden). Peptide:N-glycosidase F (PNGase F) was obtained from New England Biolabs (Beverly, MA, USA). Peroxidase-conjugated goat anti-human IgG (γ-chain specific), biotinylated goat anti-human IgG (γ-chain specific), rabbit anti-human IgG (Fab2), and peroxidase-labeled streptavidin were purchased from Sigma. The mAbs ED9 and MRC-OX41 (both mouse IgG1) were isolated, purified and biotinylated as descibed previously (Adams et al., 1998) and BF5 (mouse IgG1 against human CD4) was a generous gift of Dr J. Wijdenes (Diaclone Laboratories, Besançon, France). Murine sialoadhesin-Fc protein (the first three Ig-like domains of murine sialoadhesin fused to the Fc region of human IgG1) was produced as described by Vinson et al. (Vinson et al., 1996). Phycoerythrin-conjugated avidin STAR 4B was from Serotec (Oxford, UK). Alexa 488-conjugated avidin and Alexa 548-conjugated avidin were obtained from Molecular Probes (Eugene, OR, USA). The SuperSignal detection kit was from Pierce (Rockford, IL, USA).
Cells
NR8383 cells were cultured in RPMI 1640 containing 10% fetal calf serum (FCS), 2 mM glutamine, 100 i.u./ml penicillin and 100 μg/ml streptomycin. PC-12 cells were grown in RPMI 1640 supplemented with 10% horse serum, 5% FCS, 2 mM glutamine, 100 i.u./ml penicillin and 100 μg/ml streptomycin. CHO and Lec8 cells were grown as monolayers at 37°C in a humified atmosphere (95% air, 5% CO2) in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% FCS, 50 i.u./ml penicillin, and 50 μg/ml streptomycin. Lymphocyte suspensions were prepared from rat cervical lymph nodes by cutting the tissue into small fragments and flushing the released cells through a nylon gauze.
Affinity purification of SIRPα
Spleens and brain (cerebrum and cerebellum) from 60 rats (about 30 g) were homogenized separately in 50 ml 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA containing protease inhibitors (pepstatin, leupeptin, antipain and PMSF) using a Polytron homogenizer. After the addition of Triton X-100 (1%) and overnight stirring, the homogenates were centrifuged for 10 minutes at 800 g. Subsequently, the supernatant was incubated in 0.2% deoxycholate for 15 minutes and cleared by centrifugation for 30 minutes at 100,000 g. All manipulations were carried out at 0-4°C. Affinity purification, employing an ED9-sepharose column, and western blot analysis were performed as described previously (Adams et al., 1998).
Immunoprecipitation
Prior to immunoprecipitation, 107 cells were surface-labeled for 15 minutes at room temperature (RT) with 200 μg biotin (10 mg/ml solution in DMSO). Cells were washed with phosphate-buffered saline (PBS) and homogenized in 1.0 ml lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% NP40 containing pepstatin, leupeptin, antipain and PMSF). After centrifugation at 10,000 g in an Eppendorf centrifuge (30 minutes, 4°C), the lysate was precleared by incubation with 50 μl protein A-sepharose (10% slurry in lysis buffer) for 30 minutes at 4°C. After brief centrifugation, the supernatants were incubated with 10 μg SIRPα reactive mAb MRC-OX41, 10 μg CD4 reactive mAb BF5 (negative control), the recombinant SIRPα-Fc glycoforms (5 μg), or 5 μg murine sialoadhesin-Fc protein (negative control). Proteins were precipitated by incubation with 50 μl protein A-sepharose (10% slurry in lysis buffer) for 90 minutes at 4°C. Proteins were separated on a 10% polyacrylamide gel under reducing conditions and transferred to nitrocellulose by western blotting. Biotinylated proteins were detected by incubation of the membrane with peroxidase-labeled streptavidin followed by staining with chloronaftol and H2O2 or by the use of the SuperSignal detection kit (Pierce). PNGase F treatment prior to electrophoresis was performed according to the manufacturer’s instructions. Briefly, samples were denatured at 100°C for 10 minutes and subsequently incubated with 500 U of PNGase F at 37°C for 1 hour.
Monosaccharide composition analysis
The monosaccharide content of SIRPα purified from rat brain and rat spleen was analyzed by high-pH anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD). Oligosaccharides were released from the glycoproteins and converted into monosaccharides by trifluoroacetic acid hydrolysis (2 N TFA, 4 hours, 100°C) using 20 μg of the affinity-purified glycoproteins. After hydrolysis, the samples were lyophilized to remove the TFA and resuspended in 100 μl deionised water. The HPAEC system used for the analysis of the monosaccharides and the conditions for detection were as described previously (Van den Nieuwenhof et al., 1999b). The column was run isocratically in 0.2 M NaOH at a flow rate of 0.4 ml/minute.
Determination of galactosyltransferase activities
1 g of frozen rat brain or rat spleen was homogenized in 4 ml 50 mM sodium cacodylate, pH 7.2, using a Polytron homogenizer (3×30 seconds at maximum speed). The homogenate (5 ml) was centrifuged for 10 minutes at 600 g to obtain a debris-free homogenate, which was then centrifuged for 60 minutes at 100,000 g to yield a pellet consisting of total cellular membranes and a supernatant, which represented the soluble cell fraction. Subsequently, the cellular membrane fraction was resuspended in 50 mM sodium cacodylate, pH 7.2. NR8383 cells (2×106) and PC-12 cells (2×106) were harvested by brief centrifugation in an Eppendorf centrifuge and resuspended in 250 μl 50 mM sodium cacodylate, pH 7.2. All manipulations were carried out at 0-4°C. The p4-galactosyltransferase (β4-GalT) activities were assayed in the presence of 0.8% of Triton X-100 as previously described (Van den Nieuwenhof et al., 1999a). Protein concentrations were determined according to Peterson (Peterson, 1977) using bovine serum albumin (BSA) as a standard.
Stable expression of recombinant SIRPα-Fc in CHO cells and Lec8 cells
Soluble forms containing the extracellular domains of rat SIRPα were expressed as fusion proteins to human IgG-Fc in CHO cells and Lec8 cells. The extracellular domains of rat SIRP were amplified by polymerase chain reaction and subcloned into the pIGplus plasmid (R & D Systems, Abingdon, UK). CHO cells and Lec8 cells were transfected with the pIgplus/ratSIRPα construct by calcium phosphate precipitation. G-418 resistant clones were analyzed for SIRPα-Fc production by an enzyme-linked immunosorbent assay. Briefly, microtiter plates were coated with rabbit anti-human IgG (Fab2) (5 μg/ml) blocked with PBS/0.1% BSA, and incubated subsequently with various dilutions of supernatant derived from the transfected cells, followed by biotinylated ED9, and peroxidase-labeled streptavidin. Peroxidase activity was visualized with O-phenylenediamine dihydrochloride and H2O2 and the absorbance was measured. Clones secreting SIRPα-Fc were further expanded in selection medium (DMEM supplemented with 5% low IgG FCS, 50 i.u./ml penicillin, 50 μg/ml streptomycin and 500 μg/ml G418). Media derived from these clones were collected, pooled, stored at 4°C, and used as source for the purification of the SIRPα-Fc protein.
Purification and characterization of the SIRPα-Fc proteins
The SIRPα-Fc proteins were purified from the culture media by the use of HiTrap protein A columns. Bound material was eluted with 0.2 M glycine-HCl, pH 2.4, neutralized, dialyzed against PBS and storedat 4°C until further analysis. For characterization of the proteins, they were subjected to sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) (7.5%) and stained with Coomassie Brilliant Blue, or transferred to nitrocellulose by western blotting and stained with peroxidase-conjugated goat anti-human IgG. The monosaccharide content of SIRPα-Fc glycoproteins was analyzed as described above for SIRPα purified from rat brain and rat spleen.
Immunohistochemistry
Tissues were excised and frozen in liquid nitrogen. Immunohistochemistry was performed on acetone-fixed cryostat sections (8 μm) as reported earlier (Van den Berg et al., 1989) using a saturating concentration of 5 μg/ml SIRPα-Fc glycoforms and optimal concentrations of biotinylated ED9 and Alexa 548-avidin or biotinylated anti-human IgG and Alexa 488-avidin. Staining was evaluated using an immunofluorescence microscope (ECLIPSE E800, Nikon).
FACS analysis of SIRPα expression and SIRPα-Fc binding
SIRPα expression in NR8383 cells and PC-12 cells was analyzed as described previously (Adams et al., 1998). Cell binding of the recombinant SIRPα-Fc proteins was done as described below. Cells (1×106) were washed in PBS containing 0.1% BSA (PBS/BSA) and incubated for 60 minutes with 50 μl of a saturating concentration of SIRPα-Fc in PBS/BSA (20 μg/ml) on ice. After washing with PBS/BSA, cells were incubated with biotinylated anti-human IgG for 30 minutes on ice. After washing in PBS/BSA, they were incubated with phycoerythrin-conjugated avidin followed by washing and analysis using a FACScan (Becton-Dickinson) as described earlier (Adams et al., 1998).
RESULTS
Differential N-glycosylation of neuronal and myeloid SIRPα
Immunoblot analysis of affinity-purified SIRPα demonstrated that SIRPα from rat brain has an apparent molecular mass of 90 kDa. SIRPα purified from rat spleen was shown to migrate as two bands of 112 kDa and 52 kDa (Fig. 2A). In order to confirm the neuronal and myeloid origin of the SIRPα purified from brain and spleen, respectively, immunohistochemistry was performed. Staining of rat cerebellum with the mAb ED9 clearly showed that SIRPα is mainly expressed in neurons of the molecular layer and granular layer (Fig. 2B). In rat spleen, macrophages from the red and white pulp were labeled by ED9 (Fig. 2C). There is preliminary evidence that myeloid and neuronal SIRPα are differentially glycosylated (Veillette et al., 1998). In order to investigate this in more detail we compared SIRPα glycosylation in rat myeloid (NR8383) and neuronal (PC-12) cell lines. FACS analysis, after staining NR8383 cells and PC-12 cells with SIRPα reactive mAb MRC-OX41, demonstrated that both cell lines express surface SIRPα (Fig. 3A). Staining with SIRPα reactive mAb ED9 resulted in similar observations (data not shown). Immunoprecipitation of biotin-labeled cells with MRC-OX41 followed by western blotting and staining with streptavidin-peroxidase, showed that SIRPα from PC-12 cells has an apparent molecular mass of approximately 85-95 kDa (Fig. 3B, lanes 2 and 4) and that SIRPα from NR8383 cells migrated as a band of about 105-115 kDa (Fig. 3C, lanes 1 and 3). Immunoprecipitation with mAb BF5 (against human CD4) resulted in the precipitation of non-specific proteins only (Fig. 3B, lanes 1 and 3). PNGase F treatment of both proteins resulted in partial cleavage of glycans that led to the appearance of molecular forms of approximately 65 kDa (Fig. 3C, lanes 2 and 4), indicating that the difference in molecular mass is due to N-glycosylation.
Monosaccharide composition of the glycans derived from neuronal and myeloid SIRPα
The monosaccharide composition of the glycans derived from SIRPα proteins purified from brain and spleen were analyzed and calculated by comparing the profile with a mixture of monosaccharide standards consisting of fucose, galactose, galactosamine, glucosamine, glucose and mannose (100 ng each) (Table 1). SIRPα purified from rat brain was shown to contain only a small amount of galactose as compared to splenic SIRPα (Table 1). Furthermore, Table 1 demonstrates that N-acetylgalactosamine (GalNAc) is absent in both glycoforms, indicating the absence of O-glycosylation and N,N′-diacetyllactosediamine (lacdiNAc)-based N-glycans.
Galactosyltransferase activities in neuronal and myeloid tissues and cells
To reveal the cause of the difference in galactosylation of brain and haematopoietic SIRPα, rat brain, PC-12 cells, rat spleen and NR8383 cells were assayed for β4-GalT activity. In membranes prepared from rat spleen a much higher β4-GalT activity could be demonstrated compared to the activity found in the membrane-bound fraction prepared from rat brain (Table 2). NR8383 cells (rat alveolar M? cell line) were shown to express a relatively high β4-GalT activity whereas the β4-GalT activity detected in the neuronal cell line PC-12 was much lower (Table 2). These results indicate that the differences in SIRPα galactosylation (brain versus haematopoietic tissue) are due to differences in galactosylation capacity of the tissues involved.
Production of SIRPα-Fc proteins
In order to study the biological role of galactosylation of SIRPα, fusion proteins composed of rat SIRPα extracellular domains and human IgG Fc were produced in CHO cells, as well as in the galactosylation defective CHO mutant Lec8. Analysis of the produced recombinant proteins by SDS-PAGE showed that the chimeric protein produced in CHO cells (SIRPα-Fc) migrates as a diffuse band of 120 kDa whereas the chimeric protein produced in Lec8 cells has an apparent molecular mass of approximately 100 kDa. On western blots both chimeric proteins were recognized by anti-human IgG (Fig. 4). Monosaccharide composition analysis of the recombinant SIRPα-Fc proteins demonstrated that SIRPα-Fc produced in Lec8 cells was completely devoid of galactose and was therefore named SIRPα(ΔGal)-Fc (Table 3). The amount of fucose measured in both recombinant proteins most likely represents core fucose since known glycosylation patterns of CHO cells, as demonstrated for several proteins, are sialylation and core fucosylation but no antennae fucosylation (Sasaki et al., 1987; Spellman et al., 1989).
Cellular binding-specificity of SIRPα-Fc glycoforms
To reveal a possible difference in binding capacity between the two SIRPα-Fc glycoforms, rat cerebellum and lymph node sections were subjected to immunohistochemistry. In cerebellum sections, staining of the two glycoforms was mainly observed in the gray matter (molecular layer and granular layer) and typically shows a neuronal labeling pattern. An overall stronger staining, however, was observed with SIRPα(ΔGal)-Fc (Fig. 5). Staining of lymph node sections demonstrated that SIRPα(ΔGal)-Fc strongly labeled cells present in medulla, paracortex (T cell area), and follicle (B cell area) whereas SIRPα-Fc predominantly stained subcapsular sinus macrophages and medullary macrophages (Fig. 5). In order to investigate quantitatively whether binding to the cell surface occurred, isolated cells were analyzed for their interaction with the recombinant SIRPα-Fc glycoforms by FACS analysis. The results clearly demonstrate that SIRPα(ΔGal)-Fc has a higher avidity for lymphocytes than SIRPα-Fc, the former showing an approximately 5 times higher mean fluorescence intensity than the latter glycoform (Fig. 6). The two glycoforms were shown to have similar avidity for NR8383 cells (Fig. 6). To investigate whether there is a difference in ligand-specificity between the two glycoforms, SIRPα binding proteins from NR8383 cells were immunoprecipitated. Although there was no quantitative difference in the capacity of individual macrophage surface proteins to bind to SIRPα noted by FACS analysis, immunoprecipitation cleary demonstrated a qualitative difference. Proteins of about 130 kDa, 180 kDa, and 250 kDa were selectively recognized by SIRPα-Fc, and not by SIRPα(ΔGal)-Fc (Fig. 7). By contrast, proteins of approximately 95 kDa and 80 kDa were shown to be specifically precipitated by SIRPα(ΔGal)-Fc and not by SIRPα-Fc. Control precipitations with murine sialoadhesin Fc-protein did not identify any of these precipitated proteins (result not shown), indicating that all precipitated bands are specific.
DISCUSSION
In this study, we have investigated the tissue-specific glycosylation of neuronal and haematopoietic SIRPα and examined whether differential glycosylation affects the binding capacity of SIRPα. We demonstrate that a major difference in the glycosylation between rat SIRPα expressed in spleen (myeloid) and brain (neuronal) is the almost complete lack of galactose in the N-linked glycans of SIRPα isolated from brain. Similar observations have previously been made by Parekh et al. (Parekh et al., 1987) for rat Thy-1, another member of the Ig superfamily, isolated from brain and thymus. Their results show in detail the tissue-specific glycosylation patterns for Thy-1 and demonstrate that brain Thy-1 oligosaccharides carry no sialic acid and no (poly)lactosaminoglycans (Parekh et al., 1987). Analysis of the glycans of asialo-transferrin from human cerebrospinal fluid revealed that the major oligosaccharide structure found on this protein is a complex-type agalactodiantennary glycan in contrast to transferrin isolated from serum, which is galactosylated and sialylated (Hoffmann et al., 1995). Brain-specific glycosylation patterns have also been demonstrated for several other proteins (Shimizu et al., 1993; Hoffmann et al., 1994; Zamze et al., 1999).
Our present findings are supported by those from a recent study performed by Bartoszewicz et al. (Bartoszewicz et al., 1999) which revealed the presence of fucose, a trace of galactose, but no sialic acid and an unusually high content of oligomannosidic carbohydrate moieties on rat SIRPα expressed in brain. In that study, however, only a subset of rat brain SIRPα was analyzed that was isolated by affinity-chromatography with Galanthus nivalis agglutinin-agarose, which recognizes terminal mannose residues, thus excluding SIRPα that did not bind to this column. Oligomannosidic structures have been suggested to be important for cis-interactions between adhesion molecules that function in signal transduction in the brain (Schachner and Martini, 1995) and the same may be true for SIRPα. In the present study, we also show that the glycans derived from rat SIRPα? isolated from either brain or spleen, completely lack GalNAc. This strongly suggest that rat SIRPα carries no O-linked glycans, despite the presence of a large number of serine and threonine residues (Fujioka et al., 1996; Sano et al., 1997; Adams et al., 1998; Saginario et al., 1998) and no lacdiNAc-based N-glycans.
The glycosylation of a protein in a given tissue is determined by the expression levels and specificities of the glycosyltransferases in that tissue, as well as by the availability of the appropriate donor and acceptor structures. The present study provides for the first time enzymatic evidence for the low content of galactosylated glycoproteins in rat brain (Parekh et al., 1987; Bartoszewicz et al., 1999; this study) by demonstrating a low β4-GalT activity in this tissue. The existence of a gene family of β4-GalTs with related functions has been demonstrated by the combined work of several groups (Bakker et al., 1994; Almeida et al., 1997; Van Die et al., 1997; Sato et al., 1998; Lo et al., 1998). The best studied enzyme belonging to this family is β4-GalT I, which has been identified in several species including rat (Bendiak et al., 1993). β4-GalT I shows the highest intrinsic activity, and acts in vitro on any terminal p-GlcNAc found on oligosaccharides, N-linked glycans, O-linked glycans, polylactosaminoglycan chains, or glycolipid acceptors and, in addition can efficiently use free GlcNAc as acceptor (Blanken et al., 1982; Blanken et al., 1984; Almeida et al., 1997; Van Die et al., 1999). Its broad acceptor specificity and high activity makes β4-GalT I the major p4-GalT functioning in the galactosylation of glycoproteins and glycolipids in general. Furthermore, in the lactating mammary gland β4-GalT-I can interact with the milk protein α-lactalbumin to form the lactose synthase complex that yields lactose (Brew et al., 1968). High levels of β4-GalT I mRNA are found in many tissues, but, interestingly, only low levels are found in human brain (Lo et al., 1998; Sato et al., 1998). By contrast, normal mRNA levels of β4-GalT III, β4-GalT V and β4-GalT VI, were shown in human brain (Lo et al., 1998; Sato et al., 1998). In this study, we demonstrate that the low β4-GalT activities in rat brain and in PC-12 cells were hardly detectable with free GlcNAc. This indicate that β4-GalT I is hardly present, and suggest that the residual β4-GalT activities are due to the action of one or more other members of the p4-GalT family. Although, no evidence for the existence of rat orthologs of human β4-GalT II, β4-GalT III, β4-GalT IV or β4-GalT V are currently available, the murine ortholog of each human β4-GalT has been identified (Lo et al., 1998). Rat brain lactosylceramide synthase (β4-GalT VI) is not expected to contribute significantly to protein galactosylation (Nomura et al., 1998). Acceptor specificity studies on human β4-GalT V showed that this enzyme has a low intrinsic activity and hardly acts on N-glycans (Van Die et al., 1999). Thus, the low protein galactosylation capacity of the brain is probably the cause of the galactose deficiency in the N-glycans of SIRPα and other glycoproteins in this tissue. Clearly, this may have important consequences for interactions mediated by glycoproteins in the brain in general.
In order to investigate the role of galactosylation of SIRPα in cellular interactions, soluble recombinant SIRPα glycoforms containing galactose (SIRPα-Fc) or lacking galactose (SIRPα(ΔGal)-Fc) were produced. We show by immunohistochemistry that in rat cerebellum the molecular layer and granular layer were stained by both SIRPα glycoforms, with SIRPα(ΔGal)-Fc resulting in stronger labeling. A similar staining pattern was seen with anti-SIRPα mAb (Fig. 2B) and with anti-CD47 mAb (T. K. Van den Berg et al., unpublished), the latter suggesting that binding may well be mediated by CD47.
We also demonstrate in this study glycosylation-dependent differences in binding of SIRPα towards cells present in lymph node. SIRPα(ΔGal)-Fc was shown to bind various cells present in lymph node whereas SIRPα-Fc was shown to predominantly bind subcapsular sinus and medullary macrophages. These results may indicate that in addition to CD47, other ligands for SIRPα are present on macrophages, which most likely interact with the glycans present on SIRPα. Possible counter-receptors able to interact with the glycans of haematopoietic SIRPα are, for example, carbohydrate binding proteins with affinity for terminal galactose (galectins) or sialic acids (siglecs).
From the data presented in this study it can be concluded that tissue-specific glycosylation of rat SIRPα determines its cellular binding-specificity and might thereby affect its adhesive properties and involvement in several biological processes.
ACKNOWLEDGEMENTS
We are indebted to Mrs C. A. M. Koeleman, Mrs W. E. C. M. Schiphorst and Mrs A. van Tetering for expert technical assistance in this study. We also thank Dr R. J. Helmke (University of Texas, San Antonio, TX, USA), Dr D. L. Simmons (Department of Neuroscience, SmithKline Beecham Pharmaceuticals, Harlow, Essex, UK) and Dr J. Wijdenes (Diaclone Laboratories, Besançon, France) for their kind gifts of cells, cDNA and mAb.