Intracellular Distribution of Lysosomal Sialidase Is Controlled by the Internalization Signal in Its Cytoplasmic Tail*

Sialidase (neuraminidase), encoded by the neu-1 gene in the major histocompatibility complex locus catalyzes the intralysosomal degradation of sialylated glycoconju-gates. Inherited deficiency of sialidase results in sialidosis or galactosialidosis, both severe metabolic disorders associated with lysosomal storage of oligosaccharides and glycopeptides. Sialidase also plays an important role in cellular signaling and is specifically required for the production of cytokine interleukin-4 by activated T lymphocytes. In these cells, neu-1 -encoded sialidase activity is increased on the cell surface, suggesting that a specific mechanism regulates sorting of this enzyme to the plasma membrane. We investigated that mechanism by first showing that sialidase contains the internalization signal found in lysosomal membrane proteins targeted to endosomes via clathrin-coated pits. The signal consists of a C-terminal tetrapeptide 412 YGTL 415 , with Tyr 412 and Leu 415 essential for endocytosis of the enzyme. We further demonstrated that redistribution of sialidase from lysosomes to the cell surface of activated lymphocytes is accompanied by increased reac-tivity of the enzyme with anti-phosphotyrosine antibodies. We speculate that phosphorylation of Tyr 412 results in inhibition of sialidase internalization in activated lymphocytes.

Sialidase (neuraminidase), encoded by the neu-1 gene in the major histocompatibility complex locus catalyzes the intralysosomal degradation of sialylated glycoconjugates. Inherited deficiency of sialidase results in sialidosis or galactosialidosis, both severe metabolic disorders associated with lysosomal storage of oligosaccharides and glycopeptides. Sialidase also plays an important role in cellular signaling and is specifically required for the production of cytokine interleukin-4 by activated T lymphocytes. In these cells, neu-1-encoded sialidase activity is increased on the cell surface, suggesting that a specific mechanism regulates sorting of this enzyme to the plasma membrane. We investigated that mechanism by first showing that sialidase contains the internalization signal found in lysosomal membrane proteins targeted to endosomes via clathrin-coated pits. The signal consists of a C-terminal tetrapeptide 412 YGTL 415 , with Tyr 412 and Leu 415 essential for endocytosis of the enzyme. We further demonstrated that redistribution of sialidase from lysosomes to the cell surface of activated lymphocytes is accompanied by increased reactivity of the enzyme with anti-phosphotyrosine antibodies. We speculate that phosphorylation of Tyr 412 results in inhibition of sialidase internalization in activated lymphocytes.
Lysosomal sialidase (neuraminidase), encoded by the neu-1 gene in the major histocompatibility complex locus catalyzes the hydrolysis of terminal sialic acid residues of oligosaccharides, glycoproteins, and glycolipids. In the lysosome, sialidase is associated with lysosomal carboxypeptidase A (also cathepsin A or protective protein), ␤-galactosidase and N-acetylgalactosamine-6-sulfate sulfatase in a multienzyme complex (1)(2)(3). Dissociation of the complex in vitro results in complete inactivation of sialidase, although activity can be restored after the reconstitution of the complex (2). These results suggest that association with the complex is required for sialidase to adopt its catalytically active conformation, but direct structural data are not available to support this mechanism. Inherited muta-tions in cathepsin A result in disruption of the complex and cause galactosialidosis, an autosomal recessive disease characterized by combined deficiency of sialidase, ␤-galactosidase, and cathepsin A (4). Another autosomal recessive disease, sialidosis, is caused by mutations directly affecting the lysosomal sialidase gene (5).
Various data obtained during the purification of sialidase from different tissues have demonstrated the existence of two pools of lysosomal sialidase, soluble and membrane-associated. Both forms appear to be encoded by a single gene, since they are absent from cultured cells of sialidosis patients (6,7). Using immunoelectron microscopy, we have demonstrated the presence of sialidase on lysosomal membranes and within the lumen of lysosomes, as well as on plasma membranes of transfected cells (8). In addition, activated T lymphocytes exhibit severalfold increase of sialidase activity on their cell surfaces (9,10). Lysosomal sialidase and the sialidase expressed on the cell surface are products of the same gene, since cell surface expression does not occur in T-cells obtained from SM/J or SM/B10 mice that have mutation in the neu-1-encoded sialidase (11). These data suggest that a specific mechanism exists for sorting newly synthesized sialidase to the plasma membrane and the lysosome, as well as for retention of sialidase on the plasma membrane.
Previous studies have described several pathways for sorting lysosomal membrane proteins and endocytosed receptors to the trans-Golgi network (TGN), 1 plasma membrane, and endosome. All these cargo proteins contain amino acid motifs with conserved tyrosine or dileucine residues recognized by adaptor protein (AP) complexes. These coat proteins form vesicles destined for the lysosome (12)(13)(14). Two adaptor complexes, AP1 and AP2, are associated with clathrin-coated vesicles derived from the Golgi and the plasma membrane, respectively. Both AP1 and AP2 contain -subunits (1 and 2, respectively) that recognize and bind to the internalization signals of their cargo proteins (15). The AP3 adaptor complex is required for the formation of lysosomal/endosomal vesicles from the TGN. Proteins targeted to these vesicles contain tyrosine-based internalization signals, and genetic deficiency of AP3 in the cells of patients suffering from Hermansky-Pudlak syndrome (HPS) results in increased surface expression of such proteins (16).
In this study we obtained direct evidence that sialidase is synthesized and transported to the lysosomes as a membranebound protein. We show that the C terminus of sialidase contains a tyrosine-based internalization signal represented by the tetrapeptide 412 YGTL 415 , which can be inactivated by mutating the essential Tyr 412 and Leu 415 residues.

MATERIALS AND METHODS
Purification of Lysosomal Membranes from Human Liver and Cultured COS-7 Cells-Human liver tissues obtained from the materials discarded during a reduced liver transplantation were homogenized in isotonic (0.25 M) sucrose buffered with 10 mM HEPES-NaOH (pH 7.4) and 1 mM EDTA (homogenization buffer) using a Teflon and glass Potter-Elvehjem homogenizer. The nuclei, cell debris, and heavy mitochondria were removed by a 10-min centrifugation at 3,000 ϫ g. A "light mitochondrial fraction" containing the lysosomes, mitochondria, peroxisomes, and partially microsomes was obtained by centrifugation of the supernatants at 16,000 ϫ g for 18 min.
Lysosomes were purified from the light mitochondrial fraction by the ultracentrifugation (180,000 ϫ g for 3 h in a Beckman VTi 65 vertical rotor) in a self-forming density gradient of metrizamide (OptiPrep, Nycomed) as described (17). Immediately after centrifugation each tube was divided into 15 fractions using a Beckman Tube slicer kit. Each fraction was assayed for activities of lysosomal membrane and matrix enzymes, marker enzymes for mitochondria, plasma membranes, peroxisomes, and Golgi as well as for the presence of the lysosomal membrane marker protein LAMP-2 by Western blot as described below. The gradient fractions containing the "light" lysosomes were pooled, diluted with 3 volumes of the homogenization buffer, and centrifuged for 20 min at 35,000 ϫ g to obtain the lysosomal pellet. For the purification of lysosomal membranes, the pellet was resuspended in ice-cold water containing a protease inhibitor mixture (Roche Molecular Biochemicals) and sonicated twice for 5 s at 50 W. Membranes were precipitated by a 1 h centrifugation at 100,000 ϫ g and washed several times with ice-cold water containing the protease inhibitor mixture (Roche Molecular Biochemicals). The same procedure was used for the purification of lysosomal membranes from COS-7 cells, but the glass and glass Potter-Elvehjem homogenizer was used for the homogenization of the cells and during the first centrifugation step the speed was reduced to 1000 ϫ g.
The lysosomal membranes were purified 150 -200-fold compared with the homogenate as determined by the increase of specific activity of the marker enzyme, ␤-glucosidase, and were not cross-contaminated by microsomal (marker enzyme, UDP-galactosyl transferase) or plasma (marker enzyme, alkaline phosphatase) membranes. The average yield (measured as the total activity of ␤-glucosidase recovered in the lysosomal fraction) was about 20 -30% as compared with the homogenate.
Construction of Sialidase Mutants-Sialidase substitution mutants were generated by site-directed mutagenesis using a overlap-extension polymerase chain reaction (18). First, using the pCMV-SIAL (19) vector as a template, we obtained polymerase chain reaction fragments AC with sense primer A and mutagenic antisense primer C and BD with mutagenic sense primer B and antisense primer D (Table I). Fragments AC and BD were linked by overlap polymerase chain reaction using primers A and D to obtain the product containing the desired mutation. In particular, TAT codon of Tyr 412 was changed to GCT, GGG codon of Gly 413 to GCG, and CTC codon of Leu 415 to GCC. KpnI/EcoRV restriction fragments containing the introduced mutations were subcloned into the pCMV-SIAL plasmid. The final constructs were verified by sequencing.
Solubilization of Sialidase Expressed in COS-7 Cells-COS-7 cells were co-transfected with the wild-type human sialidase and cathepsin A cDNA. 48 h after transfection cell aliquotes (ϳ6 million cells) were harvested, washed with PBS, resuspended in 1 ml of ice-cold 50 mM sodium acetate buffer (pH 5.2) and sonicated by 3 pulses, 5 s each at 50 W. After centrifugation for 1 h at 100,000 ϫ g, the supernatant (S1) was collected and the pellet was resuspended in 1 ml of 50 mM sodium acetate buffer (pH 5.2), containing 1% (w/v) of Zwitterionic detergent 3-12, sonicated and centrifuged as before. The supernatant (S2) was collected and the pellet (P2) was resuspended in 1 ml of 50 mM acetate buffer (pH 5.2). Buffer (S1) and detergent (S2) extracts as well as the homogenate of cell pellet (P2) were assayed for protein and for the activities of sialidase, ␤-galactosidase, ␤-glucosidase, and cathepsin A as described above.
Confocal Immunofluorescence Microscopy of Transfected COS-7 Cells and Cultured Skin Fibroblasts-Forty-eight hours after transfection with wild-type or mutant sialidase, COS-7 cells were treated for 40 min with LysoTracker Red DND-99 (Molecular Probes, Eugene, OR) dye, washed twice with ice-cold PBS, and fixed with 3% paraformaldehyde in PBS for 40 min. Cells were permeabilized by 0.3% Triton X-100, washed twice with PBS, and stained with rabbit anti-sialidase antibodies and fluorescein isothiocyanate-conjugated monoclonal antibodies against rabbit IgG. Alternatively cells were double stained with rabbit antisialidase antibodies and monoclonal antibodies against LAMP-2 (Washington Biotechnology Inc., Baltimore, MD).
Skin fibroblasts of normal controls and of a HPS type 2 patient (cell line 42) were treated with LysoTracker Red, fixed, permeabilized, and stained with primary anti-sialidase and anti-LAMP-2 antibodies similar to COS-7 cells. Slides were counterstained with Oregon Green 488-conjugated anti-rabbit IgG and Oregon Green 488-conjugated antimouse IgG antibodies (Molecular Probes, Eugene, OR). Slides were studied on a Zeiss LSM410 inverted confocal microscope (Carl Zeiss Inc., Thornwood).
Antibody Internalization Assay-Antibody internalization assay was performed as described (16). Normal and HPS-2 fibroblasts grown on glass coverslips were incubated for 15 min at 37°C in the presence of primary anti-sialidase antibodies diluted in Dulbecco's modified Eagle's medium, containing 10% (v/v) of fetal bovine serum and 25 mM HEPES (pH 7.4). Subsequently the cells were washed 3 times 5 min each with ice-cold PBS, fixed in 3% paraformaldehyde, blocked, permeabilized, and stained with Oregon Green 488-conjugated anti-rabbit IgG as described above.
Sialidase Activity Based Internalization Assay-For the purification of lysosomal membranes, 48 h after transfection, COS-7 cells from one T-25 flask were washed with Hank's balanced salt solution and then scraped in the presence of ice-cold homogenization buffer. Cells were homogenized by 20 strokes in glass and glass Potter-Elvehjem homogenizer and the procedure for purification of lysosomal membranes followed as described above for human liver. The only modification was the reduced speed (1000 g versus 3000 g) during the first centrifugation step.
To measure the sialidase activity on the outer cell surface, the cells seeded in 6-well culture dishes were washed several times with Hank's balanced salt solution and overlaid with 500 l of 20 mM acetate buffer (pH 5.2), 0.25 M sucrose, and 0.4 mM sialidase substrate, MufNANA. After 30 min of incubation, 200-l aliquots of medium were added to 1.8 ml of 0.4 M glycine buffer (pH 10.4), and the concentration of the product was measured as described (21). To measure the total amount of enzyme, the assay was performed in the presence of 0.3% Triton X-100.
Western Blotting of COS-7 Cells-Proteins were resolved by SDSpolyacrylamide gel electrophoresis according to the method of Laemmli (27) and electrotransferred to polyvinyldifluoride membrane. The sialidase and cathepsin A detection was performed with rabbit antibodies as described (8,28) and LAMP-2 detection with monoclonal anti-human LAMP-2 antibodies (Washington Biotechnology Inc.) using the BM Chemiluminiscence kit (Roche Molecular Biochemicals) in accordance with the manufacturer's protocol.
Lymphocyte Stimulation and Analysis-The peripheral blood mononuclear cells were purified from heparinized peripheral blood obtained from apparently healthy adult donors after informed consent. The blood was centrifuged over Ficoll-Hypaque (Amersham Pharmacia Biotech, Montreal, Quebec) density gradients and the peripheral blood mononuclear cells were collected as described (29). Cells were stimulated with concanavalin A (20 g/ml; Sigma) in RPMI 1640 containing recombinant human IL-2 (100 units/ml; Cetus), 10% fetal bovine serum, 2 mM L-glutamine, and antibiotics (100 units of penicillin and 100 g of streptomycin/ml) as described (30). After incubation at 37°C in humidified 5% CO 2 atmosphere for 3 days, the cells were collected for analysis. Sialidase activity on the surface of lymphocytes was measured as described for COS-7 cells. To measure the total amount of the enzyme the assay was performed in the presence of 0.3% Triton X-100.
For immunoprecipitation 15 ϫ 10 6 cells were placed on ice, washed twice with ice-cold PBS then lyzed by sonication in 1 ml of radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% (v/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS) supplemented with 5 g/ml leupeptin, 50 mM sodium orthovanadate, 1 mM EDTA, 1.5 mM MgCl 2 , 50 mM NaF, complete protease inhibitory mixture (ICN, Costa Mesa, CA), and 1 mM ␣-toluensulfonyl fluoride. The lysate was collected and centrifuged at 13,000 ϫ g for 10 min to remove the cell debris. 1.0 ml of lysate was incubated for 4 h with preimmune serum at a final dilution of 1/20. Then the pellet obtained from 300 l of Pansorbin cells (Calbiochem) was added and the resulting suspension was incubated for 2 h at 4°C, followed by centrifugation for 10 min at 13,000 ϫ g. Supernatants were incubated overnight with the antisialidase antibodies at a 1/100 final dilution, then for 2 h at 4°C with the pellet from 100 l of Pansorbin cells and precipitated as above. The pellet was washed three times with 1 ml of RIPA buffer. The antigens were eluted from the pellet by the addition of 100 l of a buffer containing 0.1 M Tris-HCl (pH 6.8), 4% (w/v) SDS, 20% (v/v) glycerol, 0.2 M dithiothreitol, and 0.02% (w/v) bromphenol blue. The proteins were denatured by boiling for 5 min and 50 l of each sample was subjected to SDS-PAGE according to Laemmli (27). After electrophoresis the proteins were electrotransferred from gels to nitro-cellulose membrane. The sialidase detection was performed with rabbit antibodies and phosphotyrosine detection with monoclonal anti-phosphotyrosine Ab-1 antibodies (NeoMarkers, Fremont, CA) using the Vectastain ABC-Amp kit (Vector Laboratories, Burlingame, CA) in accordance with the manufacturer's protocol.

The Distribution of Lysosomal Sialidase between Late Endosomal and Lysosomal Pools, and Its Pattern in AP3-deficient Cells Indicate That Sialidase Trafficking Resembles That of
Lysosomal Membrane Proteins-A light mitochondrial fraction prepared from human liver homogenates was subjected to ultracentrifugation in a self-forming gradient of metrizamide (17). A typical distribution was obtained for the marker enzymes for lysosomes, microsomes, and mitochondria. As expected, lysosomal marker enzymes were found in two major peaks, previously described as "light" (fractions 3-5) and "dense" (fractions 11-13) lysosomes (Fig. 1). Previous data have shown that light lysosomes are identical to late endosomes, i.e. mannose 6-phosphate receptor-positive organelles containing internal lipid vesicles and responsible in most cells for the majority of intracellular catabolism. Soluble lysosomal enzymes targeted by a mannose 6-phosphate receptor-dependent pathway (i.e. ␤-galactosidase, N-acetyl ␤-hexosaminidase) were equally distributed between the light and dense lysosomes (Fig. 1). In contrast, lysosomal proteins, targeted to this organelle via clathrin-coated pits (i.e. LAMP-1, LAMP-2, ␤-glucosidase, and acid phosphatase) were present predominantly in the light pool of lysosomes (Fig. 1). Both measurement of sialidase activity (Fig. 1) and Western blotting of sialidase (not shown) demonstrated a distribution of sialidase between secondary endosomes and lysosomes similar to that of lysosomal membrane proteins which trafficking involves adaptor complexes.
Other indications of the trafficking pattern of sialidase was obtained from comparative studies involving LAMP-2, which moves from the TGN to the lysosome via clathrin-coated pits. We examined targeting of sialidase and LAMP-2, when the normal pathway from the TGN to the lysosome, was perturbed by deficiency of an AP3 adaptor complex, as occurs in HPS, type 2. In the two patients with this genetic disorder of vesicle formation, the gene coding for the ␤-3A subunit of the heterotetrameric AP3 adaptor complex is mutated (16), and the patients' fibroblasts exhibit drastically reduced levels of AP3. Consequently, trafficking of lysosomal membrane proteins such as CD63, LAMP-1, and LAMP-2 to the plasma membrane is enhanced. This apparently results in a default mechanism when the normal TGN to lysosome trafficking, which requires AP3 recognition of tyrosine-based signals on cargo proteins destined for the lysosome, is impaired (16).
In any event, the AP3-deficient cells provide a model system in which LAMP-2 is misrouted to the plasma membrane, and in our experiments sialidase exhibited the same misrouting. In normal fibroblasts, both LAMP-2 and sialidase antibodies stained cytoplasmic perinuclear punctate structures which colocalized with lysosomal marker, LysoTracker Red dye ( Fig.  2A, NORM). In contrast, in AP3-deficient cells ( Fig. 2A, HPS-2) both antibodies showed mostly diffuse staining of the cell surface consistent with increased targeting of LAMP-2 and sialidase to the plasma membrane. In addition in AP3-deficient cells we detected the increased internalization of anti-sialidase antibodies (Fig. 2B) upon incubation of the cells with these antibodies for 15 min at 37°C. Previously similar results with AP3-deficient cells were obtained with antibodies to CD63, LAMP-1, and LAMP-2 (16). Along with the results of cellular fractionation experiments, these data suggest that sialidase and LAMP-2 use the same mechanism for lysosomal targeting.
The consensus internalization signal is a stretch of four or six amino acids containing an essential tyrosine and a bulky hydrophobic amino acid (phenylalanine, valine, leucine, or isoleucine) located three residues C-terminal from the essential tyrosine. It is abbreviated Tyr-X-X-⌽, where ⌽ stands for bulky hydrophobic residue.
Sequence alignment with the internalization motifs of 6 proteins (cation independent mannose 6-phosphate receptor, transferrin receptor, lysosomal acid phosphatase, LAMP-1, LAMP-2, and CD63) indicates that the C-terminal fragment of human lysosomal sialidase contains a bulky hydrophobic amino acid (Leu 415 ) two residues from the essential tyrosine (Tyr 412 ). Therefore if Tyr 412 -Leu 415 tail of sialidase would be exposed to the cytoplasm, it could represent the AP2 adaptor complex-binding motif.
Since the empirical algorithms for prediction of transmembrane domains in protein sequence did not provide a high score for such fragment in the C-terminal lobe of sialidase we have performed the experiments aimed to detect directly if sialidase is an integral membrane protein. We performed extraction of COS-7 cells transfected with the wild-type sialidase and cathepsin A cDNA first with 50 mM acetate buffer (pH 5.2), and then with the same buffer, containing 1% of Zwitterionic detergent 3-12 and assayed the percent of sialidase activity recovered in each fraction. We found (Fig. 3) that the majority (75-85%) of sialidase activity could not be extracted with buffer solution and required a detergent for solubilization.

FIG. 2. Immunohistochemical localization of sialidase and LAMP-2 in cultured skin fibroblasts of normal controls and the patients with Hermansky-Pudlak syndrome type 2.
A, skin fibroblasts of normal controls (NORM) and from HPS patients (HPS-2) were probed with lysosomal marker, LysoTracker Red, fixed, stained with rabbit anti-sialidase antibodies or monoclonal anti-LAMP-2 antibodies, as indicated on the figure, and counterstained with Oregon Green 488-conjugated secondary antibodies. B, skin fibroblasts of normal controls (NORM) and HPS patients (HPS-2) were allowed to internalize antibodies to lysosomal sialidase for 15 min at 37°C. Cells were fixed and counterstained with Oregon Green 488-conjugated secondary antibodies as above. All slides were studied on a Zeiss LSM410 inverted confocal microscope. Magnification ϫ600.
Site-directed Mutagenesis of the Sialidase C Terminus Identifies the Tetrapeptide YGTL as an Internalization Signal-To determine whether the conserved Tyr 412 and hydrophobic Leu 415 are involved in targeting of sialidase to the endosome we prepared sialidase mutants with alanine codons replacing the wild-type codons for these amino acids. In another mutant, we replaced with alanine the Gly 413 residue, presumably not important for endocytosis of sialidase.
The mutants, as well as wild-type sialidase, were expressed in COS-7 cells (20) together with human cathepsin A, which is necessary for expression of sialidase activity. The cell lysates were assayed for sialidase and cathepsin A activities 48 h after transfection (Fig. 4A). All transfected cells had similar cathepsin A activity suggesting the same transfection efficiency for all cells. Two of the expressed mutants, Y412A and L415A, had reduced sialidase activity, i.e. 30 -45% of normal (Fig. 4A), while the activity of the G413A mutant was 50 -70% of normal (Fig. 4A). In the cells transfected with wild-type or mutant sialidase, Western blotting of the expressed sialidase protein (Fig. 4B) revealed a 48 -46-kDa band previously identified as mature, active sialidase (8).
The internalization of wild-type and mutant sialidase was studied by immunohistochemical staining with anti-sialidase antibody. The wild-type sialidase expressed in COS-7 cells was found in endosomal-lysosomal compartment; anti-sialidase immunofluorescence was observed in punctate structures co-localizing with the lysosomal markers LysoTracker Red (Fig. 5) or LAMP-2 (not shown). The G413A mutant showed similar localization, suggesting that conservation of Gly 413 is not essential for internalization. In contrast, both Y412A and L415A mutants showed strong peripheral staining consistent with localization of the most of the expressed enzyme at the plasma membrane. In addition, diffuse cytoplasmic staining, partially overlapping with LysoTracker Red and possibly representing the Golgi compartment, was apparent in some cells. Thus, the presence of both Tyr 412 and Leu 415 in the C-terminal peptide of sialidase appears necessary for internalization.
To quantify the distribution of sialidase mutants targeted to the endosomal/lysosomal compartment and to the cell surface, we assayed sialidase activity of purified lysosomal and plasma membranes. COS-7 cells harvested 48 h after transfection were divided into two portions. From one portion, lysosomal membranes were purified by subcellular fractionation and density centrifugation using a self-forming metrizamide gradient; from the other portion, plasma membranes were purified using polylysine-coated beads. The sialidase activity present on purified plasma membranes of COS-7 cells transfected with wild-type sialidase or the sialidase G413A mutant was very low (Fig. 6A,  black bars). In contrast, a considerable amount of sialidase activity was found on plasma membranes purified from cells transfected with Y412A and L415A mutants (Fig. 6A, black  bars). The activity of a control plasma membrane enzyme, alkaline phosphatase, was similar for nontransfected cells and for cells transfected with wild-type or mutant sialidase (Fig.  6A, open bars). The opposite effect was observed for lysosomal membranes. Transfection of COS-7 cells with wild-type sialidase or the sialidase G413A mutant increased the sialidase activity present on the lysosomal membranes ϳ3-fold (Fig. 6B,  black bars). The sialidase activity on lysosomal membranes purified from cells transfected with the Y412A and L415A mutants was similar to that present on membranes purified from nontransfected control cells (Fig. 6B, black bars). The activity of the endogenous lysosomal membrane enzyme, ␤-glucosidase, was similar for all the cells studied (Fig. 6B, open  bars). The results suggest that the increased amount of siali-dase present on the plasma membranes of the Y412A and L415A mutants resulted from impaired endocytosis.
A significant increase of sialidase activity on the surface of COS-7 cells transfected with Y412A and L415A mutants as compared with nontransfected cells was also observed when we incubated intact cells in an isotonic buffer containing sialidase substrate (Fig. 7A, black bars). A small increase of sialidase activity was also detected for cells transfected with the wildtype sialdase DNA or the G413A mutant. However, when the same assay was performed in the presence of 0.3% Triton X-100, which permeabilizes cell membranes (Fig. 7A, open

FIG. 5. Immunohistochemical localization of sialidase mutants expressed in COS-7 cells. COS-7 cells co-transfected with cathepsin
A and wild-type (WT) or mutant sialidase cDNAs, as indicated on the figure, were probed with LysoTracker Red, fixed, stained with rabbit anti-sialidase antibodies, and counterstained with fluorescein isothiocyanate-conjugated monoclonal antibodies against rabbit IgG. Slides were studied on an inverted confocal microscope. Magnification ϫ600.

FIG. 6. Sialidase activity of the plasma (A) and lysosomal (B) membranes of COS-7 cells transfected with sialidase mutants.
Plasma and lysosomal membranes were purified 48 h after co-transfection of COS-7 cells with cathepsin A (CathA) and wild-type (WT) or mutant sialidase cDNA as described under "Materials and Methods." On average lysosomal membranes were purified 150 -200-fold compared with the homogenate as determined by the increase of specific activity of ␤-glucosidase, and were not cross-contaminated by microsomal (marker enzyme, UDP-galactosyl transferase) or plasma (marker enzyme, alkaline phosphatase) membranes. The average yield (total activity of ␤-glucosidase recovered in the lysosomal fraction) was about 20 -30% as compared with the homogenate. Plasma membranes were purified ϳ80-fold by following the specific activity of alkaline phosphatase and did not contain lysosomal membranes (marker enzyme ␤-glucosidase). The recovery of plasma membrane (total alkaline phospha- bars), sialidase activity was 3-fold higher for the cells transfected with wild-type sialidase or the G413A mutant, reflecting the increased amount of enzyme localized within the cells. For the L415A and Y412A mutants, the activity was the same in the absence or presence of Triton, suggesting that all sialidase in the intact COS-7 cells transfected with L415A and Y412A mutants was available for the substrate. These results not only confirmed the findings observed for the purified membranes, but also showed that sialidase is located on the outside surface of the cell membrane. Measurement of activity of the endogenous lysosomal enzyme, N-acetyl-␤-hexosaminidase (not shown), demonstrated the same 70 -80% latency for all the cells studied indicating that the cells remained intact during the assay and that the detected sialidase activity did not result from secretion or release of the expressed enzyme. Measurement of the cathepsin A activity (Fig. 7B) did not reveal any difference between the cells expressing the wild-type or mutant sialidase either in the absence or in the presence of Triton X-100.
Sialidase Is Co-localized with CTLA-4 and Can Be Tyrosinephosphorylated in Activated T Cells-The previously described increase of neu-1-encoded sialidase activity on the surface of activated lymphocytes (9, 10) makes these cells particularly interesting for studying sialidase internalization. The availability of antisialidase antibodies allowed us to follow the intracellular distribution of sialidase using confocal immunomicroscopy. In nonactivated cells, sialidase is localized within intracellular vesicles, which stain positively for the lysosomal marker LysoTracker Red (Fig. 8). Only weak expression of sialidase was observed on the cell surface. Upon stimulation with concanavalin A the fluorescence of the intracellular vesicles containing sialidase was reduced, whereas the cell periphery became brightly stained. This apparent translocation of sialidase from lysosomes to the cell surface was consistent with results of sialidase assays. Sialidase activity on the cell surface of intact cells incubated in an isotonic buffer containing substrate was ϳ4-fold higher for concanavalin A-activated lymphocytes (Fig. 9, black bars). Total sialidase activity assayed in the presence of 0.3% Triton X-100 was similar for both activated and nonactivated cells. The activity of a control endogenous lysosomal enzyme, N-acetyl-␤-hexosaminidase, was similar for activated and nonactivated cells both in the absence and in the presence of detergent (Fig. 9). We also did not detect any difference in the level of activities of cathepsin A (Fig. 9) or ␤-galactosidase (now shown), which form with sialidase a multienzyme complex.
Similar re-distribution from lysosomes to the cell surface upon activation of T cells was reported for another protein, CTLA-4 antigen. This glycoprotein of the Ig superfamily functions as a co-receptor for CD28, but plays an opposite role by down-regulating T cell activation (42). Confocal immunomicroscopy studies of lymphocytes double stained with anti-CTLA-4 antibody and LysoTracker and with anti-CTLA-4 and anti-sialidase antibody showed that sialidase and CTLA-4 colocalize in the lysosomes of nonactivated cells and undergo translocation to the plasma membrane in activated lymphocytes (Fig. 8).
The intracellular distribution of CTLA-4 in lymphocytes is controlled by a complicated mechanism involving phosphorylation of the essential tyrosine of the internalization signal. In activated T cells, this phosphorylation blocks AP2-mediated endocytosis and retains CTLA-4 on the plasma membrane (43)(44)(45)(46). To determine whether a similar type of regulation operates in the trafficking of sialidase we performed the immunoprecipitation of sialidase from the lysates of activated and nonactivated lymphocytes followed by the SDS-PAGE of the precipitated protein and Western blot using anti-sialidase and anti-phosphotyrosine antibodies. We observed that anti-phosphotyrosine antibodies reacted with sialidase (Fig. 10). The phosphorylation was significantly increased in activated compared with nonactivated cells. Although more experiments are needed to prove that Tyr 412 is the particular amino acid residue, which undergoes phosphorylation, these results suggest that the interaction of sialidase and CTLA-4 with its adaptor complex is inhibited via the same mechanism. DISCUSSION Like other lysosomal proteins, sialidase is synthesized as a precursor which undergoes modification. After cleavage of a signal peptide and glycosylation, sialidase becomes a 48.3-kDa mature active enzyme found in the lysosome and present in the multienzyme lysosomal complex. However, the exact mechanism by which the sialidase precursor is sorted has remained unclear until now. Since sialidase contains three Asn-linked oligosaccharide side chains, it could be mannose 6-phosphorylated and targeted to the lysosome via mannose 6-phosphate receptors, as many soluble lysosomal enzymes. However, the mannose phosphorylation of sialidase is not sufficient for targeting the enzyme to the lysosomes (47). Based upon the intracellular distribution of human sialidase expressed in COS-1 cells transfected with sialidase cDNA alone or co-transfected with sialidase and human cathepsin A cDNA, van der Spoel (47) suggested that the association of sialidase with cathepsin A occurs as early as in the endoplasmic reticulum and is required for proper sorting of the enzyme. In this scenario, sialidase would be transported to the lysosome via cathepsin A, which does obtain a robust mannose 6-phosphate recognition signal and binds to the mannose 6-phosphate receptor. In the absence of cathepsin A, sialidase is partially secreted and partially segregates to the endosomal compartment (47). This hypothesis is supported by data showing a significant increase of sialidase activity and immunoreactive material in lysosomes of cells co-transfected with sialidase and cathepsin A plasmids as compared with cells transfected with sialidase alone (48). However, this mechanism is not consistent with targeting of sialidase to the endosomal compartment observed in cultured fibroblasts of galactosialidosis patients lacking cathepsin A (8). In addition, sialidase was localized not only on the inner side of the lysosomal membrane and in the lysosomal lumen, but also on the plasma membrane and in small intracellular (possibly endocytic) vesicles (8). This suggests that sialidase may also be transported by a mechanism involving its targeting to the plasma membrane, followed by endocytosis.
In this study, we demonstrate a similarity in the distribution of sialidase and lysosomal membrane proteins that are targeted to the lysosome by adaptor complexes AP2 (acid phosphatase) or AP3 (LAMP-2). Similarly to lysosomal membrane enzyme, ␤-glucosidase, the majority of sialidase expressed in COS-7 cells could not be extracted with buffer solution, but required high (1%) concentration of detergent for solubilization (Fig. 3). We also found that both LAMP-2 and sialidase exhibit the same misrouting in AP3-deficient (HPS-2) cells, in which trafficking of lysosomal membrane proteins to the plasma membrane is enhanced. Both anti-sialidase and anti-LAMP-2 antibodies showed a diffuse staining in AP3-deficient cells instead of cytoplasmic perinuclear punctate structures detected in normal cells. Although these results do not prove that sialidase lacks internalization in AP3-deficient cells they suggest that sialidase and LAMP-2 use the same mechanism for lysosomal targeting.
We have demonstrated that the C-terminal tetrapeptide of sialidase 412 YGTL 415 represents the tyrosine-containing lysosomal targeting signal, and that Tyr 412 and Leu 415 amino acid residues are essential for rapid endocytosis of sialidase. Specifically, the sialidase substitution mutants Y412A and L415A were sorted to the plasma membrane but not internalized. Altogether our data suggest that lysosomal sialidase contains the internalization signal most probably capable of binding both AP2 and AP3 adaptor complexes.
The identified mechanism of sialidase targeting not only provides essential information about the enzyme's biogenesis but also extends our knowledge about several important physiological processes involving this enzyme. Several studies have shown that neu-1-encoded sialidase, in addition to its role in intralysosomal catabolism of sialylated glycoconjugates, is also involved in cellular signaling during the immune response. In particular, T cells require sialidase for both early production of IL-4 and the IL-4 priming of conventional T cells to become active IL-4 producers (49,50). During the activation of T cells, sialidase is expressed on the plasma membrane where it participates in desialylation of surface antigen-presenting molecules such as myosin heavy chain class I, required to render T cells responsive to antigen presenting cells (51), and G M3ganglioside, which modulates Ca 2ϩ immobilization and regulates IL-4 production (50). In addition, sialidase of T lymphocytes converts the group specific component or Gc protein into a factor necessary for the inflammation-primed activation of macrophages (10,52). T-cells derived from SM/J or B10.SM strains of mice, deficient in neu-1 sialidase (11), fail to convert Gc and synthesize IL-4, and B cells of these mice cannot produce IgG 1 and IgE after immunization with pertussis toxin (9,49,52). A significant increase of the sialidase activity and immunoreactivity on the surface of activated lymphocytes that we demonstrated in this study are in good agreement with the published data (49 -51). Since binding of sialidase to cathepsin A is required for activation it is possible to suggest that these two enzymes remain associated on the plasma membrane. However, our experiments did not reveal the increased cathepsin A activity on the surface of activated lymphocytes or COS-7 cells expressing sialidase Y412A and L412A mutants, targeted to the plasma membrane (Figs. 7 and 9). We suggest that sialidase may interact with cathepsin A precursor in the Golgy and stay associated with it on the root to the plasma membrane. In accordance with this hypothesis recent studies demonstrated the presence of enzymatically nonactive cathepsin A precursor on the plasma membrane (54).
Our studies of the promoter of the human sialidase gene have shown that its expression is potently induced by proinflammatory factors and is inhibited by curcumin and N-acetylcysteine, which inhibit inflammatory responses in different tissues. 2 However, it remains unclear how de novo synthesized sialidase can be retained on the surface of activated T cells. One possibility is that the interaction of sialidase with adaptor complexes and targeting to the lysosome can be blocked by phosphorylation of the essential tyrosine in the internalization signal. Such a mechanism has been recently described for cytotoxic T lymphocyte-associated antigen (CTLA-4), which is transiently expressed on the surface of activated T cells and is involved in their down-regulation (42).
The cytoplasmic tail of CTLA-4 contains the 165 YVKM signal responsible for its binding to AP2 and internalization (43). In activated T cells, phosphorylation of the essential Tyr 165 residue by the T-lymphocyte associated tyrosine Src family and Jak2 kinases prevents its interaction with the 2 subunit of AP2 and results in its expression on the cell surface (44 -46). Since the recognition sequence pattern for Src family kinases (Tyr(P)-hydrophylic-hydrophylic-(Leu/Ile/Pro) (53)) is present in the sialidase internalization signal, a similar mechanism may control the cellular distribution of both proteins. This is consistent with the observed cross-reaction of sialidase and anti-phosphotyrosine antibodies on Western blots of activated lymphocytes. Further experiments that should directly demonstrate the phosphorylation of the Tyr 412 residue by the mass spectral analysis of the material immunoprecipitated from the activated T cells with anti-sialidase antibodies are in progress in our laboratory. If proven, this hypothesis would explain the mechanism of sialidase involvement in the immune response.