O-Glycosylation of C-terminal Tandem-repeated Sequences Regulates the Secretion of Rat Pancreatic Bile Salt-dependent Lipase*

Amino acid sequences rich in Pro, Glu, Ser, and Thr (PEST) are common to rapidly degraded proteins (Rogers, S., Wells, R. & Rechsteiner, M. (1986)Science 234, 364–368). On pancreatic bile salt-dependent lipase (BSDL), PEST sequences are present in the C-terminal region of the enzyme to which is associated theO-glycosylation. We have postulated that theO-glycosylation of BSDL may contribute to mask PEST sequences and to trigger the secretion of this enzyme instead of its delivery into a degradative pathway (Bruneau, N., and Lombardo, D. (1995) J. Biol. Chem. 270, 13524–13523). To further examine the role of the O-linked glycosylation on BSDL metabolism, rat pancreatic BSDL cDNA was stably transfected into two Chinese hamster ovary (CHO) cell lines, the CHO K1 wild-type line and the O-glycosylation defective CHO ldlD line. In these latter cells, O-glycosylation can be reversibly modulated by culture conditions. Results indicate that the rate of BSDL synthesis by transfected CHO K1 or CHO ldlD cells reflects, independently of culture conditions, the amount of mRNA specific for BSDL present in these transfected cells. Nevertheless, the rate of secretion of the enzyme depends upon cell culture conditions and increases with the cell capability to O-glycosylate C-terminal tandem-repeated sequences. Immunoprecipitation experiments performed on cell lysates suggested that a rapid degradation of BSDL occurred particularly when transfected CHO ldlD cells were cultured under non-permissive conditions. We further showed that BSDL secreted by CHO ldlD cells grown under non-permissive conditions that normally prevent O-glycosylation incorporated galactose and was reactive with peanut agglutinin, which recognizes the core structure ofO-linked glycans. We concluded that the BSDL expressed by CHO ldlD cells grown under non-permissive conditions was rapidly degraded but a fraction of the enzyme was allowed toO-glycosylate and consequently was secreted.

The bile salt-dependent lipase (BSDL, 1 EC 3.1.1.-) secreted by the pancreas plays an essential role in the intestinal processing of cholesterol and lipid soluble vitamins (1). This enzyme, as suggested by its presence in various tissues, may serve different functions depending upon its location (2)(3)(4). BSDL is found in the pancreatic secretions of all species examined up to now, from fish to human (5,6). All sequenced BSDLs have a site for N-linked glycosylation on Asn-187 (7) close to the Ser-194 involved in substrate catalysis (8). This N-linked glycosylation has been shown to be cotranslational and essential for the secretion and expression of a fully active enzyme by pancreatic cells (9). The size of this protein varies by species as a consequence of the number of C-terminal tandem repeated sequences and because the amount of glycosylation varies from species. The human enzyme possesses 16 of these tandem repeated sequences (10) and is the largest (100 kDa). The rat BSDL has only four of these repeats (7), which are absent on the smaller salmon enzyme (5). Each tandemly repeated sequence presents a potential site for O-linked glycosylation and is rich in proline (P), glutamic acid (E), serine (S), and threonine (T) (7,10). Proteins that are rapidly degraded within eukaryotic cells frequently contain such sequences, referred to as PEST regions (11). On BSDL, PEST regions coexist with the C-terminal cluster of O-linked oligosaccharides (7,10), and are absent in sequences of other known pancreatic enzymes. Hanson et al. (12) showed that deletion of the C-terminal region of BSDL including the PEST region did not affect the activity of the recombinant enzyme. The truncated BSDL presented the same dependence on bile salts as the native enzyme (12). However, compared with the full-length enzyme, a truncated variant lacking the C terminus starting from Leu-519 showed an increased susceptibility to proteolysis, suggesting that Cterminal repeats may regulate proteolytic degradation of BSDL (13). We have shown recently that human (14) and rat pancreatic BSDL (15), in contrast to the other secretory pancreatic enzymes, are associated with intracellular membranes during their secretory process. This association with intracellular membranes, which involves a multimeric folding complex including a protein immunologically related to the glucoseregulated protein of 94 kDa (Grp94), could be essential for the O-glycosylation of the C-terminal tail of BSDL. Consequently, we have postulated that O-glycosylation of the C-terminal repeats that include PEST sequences may contribute to divert BSDL away from a possible degradation route (15). To test this * This work was supported in part by Grant 6122 from the Association pour la Recherche sur le Cancer (ARC, Villejuif, France), financial help from the Conseil Général des Bouches-du-Rhône (Marseille, France), and the W. W. Smith Charitable Trust (Philadelphia, PA). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  hypothesis and to further examine the role of the glycosylation on BSDL metabolism, rat pancreatic BSDL cDNA was stably transfected into Chinese hamster ovary (CHO) cells defective for O-linked glycosylation (ldlD line) (16) and into wild-type CHO cells (K1 line).

EXPERIMENTAL PROCEDURES
Materials-Monensin and peanut agglutinin immobilized on agarose (PNA-agarose) were from Sigma. Brefeldin A, sialidase, and leupeptin came from Boehringer (Mannheim, Germany). Phenylmethylsulfonyl fluoride, benzamidine, and ␤-phenyl propionate were from Fluka (Buchs, Switzerland). Ham's F-12 medium, RPMI medium, fetal calf serum, and Geneticin (G418) were from Life Technologies Inc. [  Antibodies-Polyclonal antibodies raised using purified secretory rat BSDL were obtained in our laboratory and isolated by affinity chromatography on protein A-Sepharose (17). These antibodies were able to immunoprecipitate non-glycosylated BSDL obtained by in vitro translation using pancreatic mRNA and rabbit reticulocytes (data not shown). Dot blots were performed using antibodies specific for BSDL as primary antibodies and the BM chemiluminescence Western blotting kit (Boehringer) was used to detect the antigen-antibody complexes.
Protein and Enzyme Assays-Proteins were routinely assayed with the bicinchoninic acid method (Pierce) using bovine serum albumin as standard. The activity on 4-nitrophenyl hexanoate was measured at 410 nm and pH 7.4 in a thermostated cell at 30°C as described elsewhere (18).
Cell Culture-The Chinese hamster ovary cell lines CHO K1 (wildtype) and CHO ldlD were supplied by the American Type Culture Collection (Rockville, MD; ATCC designation CCL 61 and SD 1401). The CHO ldlD cells were used with the kind permission of Dr. M. Krieger (Massachusetts Institute of Technology, Cambridge, MA). Each CHO cell line was maintained in 5% CO 2 at 37°C in Ham's F-12 nutrient mixture (Life Technologies, Inc.) supplemented with 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, 0.1% (v/v) fungizone, and fetal calf serum (FCS). The CHO ldlD cell line is defective in O-linked glycosylation when grown in culture medium supplemented with 0.5% (v/v) FCS (non-permissive conditions) (16). The reversion of the defect can be obtained by incubating these cells in medium supplemented with 10% (v/v) FCS (permissive conditions) (16).
cDNA Probes-The cDNA specific for actin was obtained by RT-PCR using specific primers (Life Technologies, Inc.). The Grp94 cDNA was a generous gift of Dr. A. S. Lee (University of Southern California, Los Angeles, CA). The cDNA probe (0.5 kb) specific for BSDL was obtained by reverse transcription and polymerase chain reaction (20). The 1-kb cDNA probe for the neomycin phosphotransferase was obtained by digestion of the pMAMneo vector by SmaI and BglII. Purified cDNA probes were 32 P-labeled by random priming using [␣-32 P]dCTP (NEN Life Science Products) and the random primed DNA-labeling kit (Life Technologies, Inc.) to a specific radioactivity of approximately 4 ϫ 10 8 cpm/g. RNA Isolation and Northern Blot Analysis-Total RNAs were extracted from cells using the method of Chirgwin (21). Thirty g of these total RNAs suspended in 10 mM phosphate buffer (pH 7.0), glyoxal (18%), and dimethyl sulfoxide (50%) were subjected to electrophoresis on 1% agarose gel. The RNAs were then transferred to nylon membranes (Biodine A Pall-Bio-support, Portsmouth, United Kingdom). Prehybridization, hybridization, and washing conditions were as described previously (20).
For mRNA quantitation, total RNAs were dotted using a Bio-Dot microfiltration apparatus (Bio-Rad) onto a nitrocellulose membrane (BA 83 type, Schleicher & Schuell, Dassel, Germany) in decreasing rank amount from 30 to 0.45 g/well. Prehybridization and hybridization with specific cDNA probes were performed as above. The amount of specific mRNA was estimated by scanning the autoradiogram as described previously (9). Densitometric data were plotted against the mass of total RNA; from the slope of each regression line, it was possible to determine the relative amount of specific mRNA (in arbitrary units) in the mixture of total RNA.
Expression Plasmid Vectors-The rat pancreatic BSDL cDNA inserted into the pECE-1 plasmid, a mammalian cell expression vector utilizing the simian virus 40 early promoter, was used with the kind permission of Dr. W. J. Rutter (University of California, San Francisco, CA). This plasmid referred to as pBSDL was described previously by Han et al. (7). Expression plasmid pMAMneo, which contains the neomycin phosphotransferase gene conferring the resistance to neomycin analogue Geneticin (G418), was obtained from CLONTECH (Palo Alto, CA).
Stable Expression of the Rat Pancreatic BSDL-Stable transfection of CHO K1 and CHO ldlD cells was performed using a lipopolyaminemediated transfection procedure according to the manufacturer (Transfectam ® reagent, Promega). The selection of stable clones was performed for 3 weeks in selection Ham's F-12 medium with G418 (400 g/ml medium); the surviving cells were then trypsinized and cloned by end-limiting dilution method. The stably transfected clones were maintained under the same selection conditions and tested for the BSDL activity in the culture medium. Finally, positive clones were trypsinized and plated in 100-mm dishes. Transfected CHO ldlD cells were progressively accustomed to media containing 0.5% FCS (unpermissive conditions) for at least 24 h before use.
Inhibitor Treatment-Confluent transfected CHO cells were treated with drugs affecting protein traffic. For this purpose, the drug was added from a stock solution (monensin, 10 mM in ethyl alcohol; brefeldin A, 1 mg/ml in PBS: ethyl alcohol (1:1 by volume)) to dishes at the appropriate final concentration. After the required time of incubation, the cell culture medium was withdrawn and saved for further analyses.The corresponding cell layer was washed twice with PBS, scraped with a rubber policeman, and lysed in a 10 mM Hepes (pH 7.4) buffer (lysis buffer, 200 mM NaCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 1.5% Triton X-100, 10 g/ml leupeptin, 2 mM benzamidine, 2 mM phenylmethylsulfonyl fluoride, 2 mM soybean trypsin inhibitor, and 2 mM ␤-phenyl propionate). The homogenate was then clarified by centrifugation (20 min, 10,000 ϫ g, 4°C). The BSDL activity was determined in the cell culture medium and the cell lysate.
Pulse-chase Protocol-After 45 min of starvation of CHO cells in methionine-free RPMI medium, dishes were pulse-labeled with [ 35 S]methionine (20 or 40 Ci/ml). The pulse medium was removed after the required time, followed by two quick washes with the chase medium, and cells were chased in Ham's F-12 medium with the required amount of FCS for different time intervals. The chase was stopped by aspirating the medium and cells were washed twice with ice-cold PBS without CaCl 2 and MgCl 2 . Cells were then lysed in lysis buffer, and cell lysates were clarified. Supernatant were used for immunoprecipitation. When indicated, cell-free chase medium was also used for immunoprecipitation.
Immunoprecipitation of BSDL-One ml of clarified cell lysate or cell-free medium was incubated overnight at 4°C with 25 g of antibodies to rat BSDL. Protein A (10 mg), prewashed three times with the lysis buffer, was added to antibody-antigen complexes and incubated for 4 h at 4°C under agitation. The antigen-antibody-Protein A complexes were recovered by centrifugation (10,000 ϫ g, 15 min, 4°C). The final pellet was then washed twice with the washing buffer (10 mM Tris/HCl (pH 7.4) buffer, 5 mM EDTA, and 0.5% Triton X-100), twice with the washing buffer supplemented with 1 M NaCl and 0.1% SDS, and finally twice with 10 mM Tris/HCl (pH 7.4), 5 mM EDTA buffer. The pellet was then transferred into the SDS-PAGE Laemmli's sample buffer, warmed for 2 min at 95°C, centrifuged, and electrophoresed on SDS-PAGE. When required, a two-cycle immunoprecipitation procedure slightly modified from Doolittle et al. (22) was used. Gels were stained with Coomassie Blue R250 and destained as above, immersed (30 -60 min) in Amplify (Amersham), and autoradiographed using Kodak Bio Max films.
Isolation of the C-terminal Peptide of BSDL-The basic protocol for the isolation of the O-linked C-terminal tail of BSDL has been extensively described by Mas et al. (23). Briefly, immunoprecipitated BSDL was radiolabeled with [ 3 H]DFP on the serine residue (Ser-194) involved in the catalytic site (8). This residue is close to Asn-187, which is the unique site for N-linked glycosylation of BSDL (7,10). When required the protein was desialylated with sialidase (10 milliunits in 50 mM sodium acetate (pH 5.0) buffer, 3 h, 37°C). Because there is no methionine residue between Asn-187 and Ser-194, a cyanogen bromide cleavage of radiolabeled BSDL allows us to separate the radiolabeled Nglycosylated peptide from the C-terminal O-glycosylated peptide that comprises PEST sequences (23,24). The N-and O-linked glycopeptides were separated on PNA-agarose column equilibrated and eluted with a sodium phosphate (pH 7.4) buffer, 1 mM each CaCl 2 and MgCl 2 (PBS) containing 0.5% Triton X-100. After elution of wash-out fractions, which contained the N-glycosylated radioactive peptide, the column was washed with PBS (ϩ 0.1% Triton X-100). Finally, bound O-glycosylated material was eluted with 0.3 M lactose in PBS buffer and detected by dot blot as described above.

Secretion of Bile Salt-dependent Lipase by Transfected CHO
Cells-The wild-type CHO K1 and the O-glycosylation-defective CHO ldlD cells were stably transfected with the full-length cDNA of the rat BSDL. Two clones expressing BSDL activity were selected from each cell line and used for these studies. They were referred to as the CHO K1-3B and the CHO ldlD-6B clone, respectively. No enzyme activity above background was expressed in either non-transfected or pMAMneo transfected (control) cell lines. Northern blot analyses were used to assess the mRNA abundance in both CHO K1-3B and CHO ldlD-6B clones. The BSDL mRNA level in stably transfected CHO K1-3B cells was higher than in CHO ldlD-6B cells and was absent in control CHO cells (Fig. 1A). The size of this mRNA (2.0 kb) was of the expected size (7). The 300-base pair cDNA probe for actin hybridized with a transcript of the expected size (2.0 kb), indicating no mRNA degradation. We next attempted to determine the relative amount of mRNA transcribed in CHO K1-3B and CHO ldlD-6B clones. As shown on Fig. 1B, quantitation of the Northern dot represented in the lower panel indicated that the abundance of mRNA was approximately 4 -5 times higher in CHO K1-3B than in CHO ldlD-6B cells (the amount of dotted RNA was normalized with the actin probe). The lower expression of BSDL cDNA observed in CHO ldlD-6B cells compared with CHO K1-3B appeared rather specific because parallel quantitation of Grp94 mRNA and neomycin phosphotransferase mRNA taken to represent endogenously and exogenously expressed mRNAs, respectively (25), were identical in both cell lines. It is worth noting that the amount of mRNA specific for BSDL was identical in CHO ldlD-6B cultured under permissive and non-permissive conditions (data not shown). The time-dependent synthesis of BSDL was estimated by pulse-labeling of CHO cells with [ 35 S]methionine and immunoprecipitation performed on cell lysates, after an adequate period of chase. The band corresponding to BSDL was quantitated by scanning (9), the dark intensity (pixel) at each time of chase was subtracted from that at time 0, giving the amount of neo-synthesized BSDL with time ( Fig. 2). Slopes thus obtained suggested that the BSDL synthesis was 5 times faster in CHO K1-3B than in CHO ldlD-6B cells. These slopes, normalized for the amount of mRNA specific for BSDL, gave the translational rate (amount of protein produced per min and unit of mRNA), which was, within experimental errors, identical in CHO K1-3B cells and in CHO ldlD-6B cells cultured under permissive or non-permissive conditions. These results showed that culture conditions did not affect either transcriptional or translational processes in transfected CHO cells.
The time dependence of the secretion of BSDL was next examined. For this purpose, CHO K1-3B and CHO ldlD-6B cells were pulse-labeled by [ 35 S]methionine and chased for time intervals up to 240 min. Aliquots of cell culture medium were collected and immunoprecipitated. As shown in Fig. 3A, after a pulse of 15 min, BSDL was detected in the culture medium of CHO K1-3B cells after only 5 min of chase. CHO ldlD-6B cells also secreted BSDL; however, the enzyme was detectable by immunoprecipitation after a 30-min chase. Quantitation of the precipitated material by scanning (Fig. 3B) indicated that the rate of BSDL secretion by CHO K1-3B cells was about 50 times faster than that of CHO ldlD-6B grown under permissive conditions. The latter clone appeared to secrete 4-fold less BSDL when cultured under non-permissive conditions. The time dependence of BSDL secretion by either CHO K1-3B and CHO ldlD-6B cells was also determined from the enzyme activity recorded in culture medium. For this purpose, CHO K1-3B and CHO ldlD-6B cells were washed in PBS and incubated with fresh medium. At time intervals up to 4 h, aliquots of the medium were collected and assayed for BSDL activity. The activity of BSDL increased with time in the culture medium of both cell lines, but the rate of BSDL secretion by CHO K1-3B cells (36.6 Ϯ 1.3 milliunits/h/mg of cell protein) was approximately 30 times higher than that of CHO ldlD-6B cells grown under permissive conditions (1.2 Ϯ 0.2 milliunits/h/mg of cell protein). Additionally, as above, and consistent with the estimates made by scanning, the rate of secretion was approximately 5-fold slower (0.22 milliunits/h/mg of cell protein) when CHO ldlD-6B cells were cultured under non-permissive conditions. These data showed that, even though the rates of BSDL translation in CHO K1-3B and CHO ldlD-6B clones were similar, the rate of secretion of the enzyme significantly differed between these clones. Overall, the rate of BSDL secretion by CHO ldlD-6B cells was dependent upon the culture, permissive versus non-permissive, conditions. Taken together, these data suggested that post-translational modification of BSDL by Oglycosylation could regulate the rate at which the enzyme is secreted.
BSDL Degradation in CHO ldlD-6B Cells-As shown above, the rate of BSDL synthesis, which reflects the abundance of BSDL mRNA in each clone, was independent of the culture conditions. Nevertheless, the rate of secretion of the enzyme by CHO ldlD cells clearly depends upon culture conditions. Therefore, it was tempting to address the question of whether, in CHO ldlD cells grown under non-permissive conditions, BSDL was partly subjected to intracellular degradation. The CHO ldlD-6B cells grown to confluence, were pulsed for 30 min with [ 35 S]methionine. After a 3-h chase, cell culture medium and cell lysate were immunoprecipitated. The CHO K1-3B cells were pulsed for 5 min and chased for 20 min before immunoprecipitation. The immunoprecipitated material was then separated by SDS-PAGE and autoradiographed. The material immunoprecipitated from the cell lysate and culture medium of CHO K1-3B cells (data not shown) displayed an electrophoretic migration corresponding to a doublet at 72-75 kDa, compatible with the M r of the rat BSDL (9, 26 -28). The cell culture medium of CHO ldlD-6B cells, independently of culture conditions, also contained material associated with a M r ϳ 75 kDa. In contrast, analysis of the CHO ldlD-6B cell lysates indicated that the radioactive material migrating at approximately 75 kDa was largely depleted; instead, predominant bands with M r values of 43-46 and 55 kDa were detected. An immunoprecipitation of cell lysate performed immediately after a short pulse (Յ5 min) also displays sub-bands at about 60 and 55 kDa (data not shown). This, a priori, indicated that BSDL expressed in transfected CHO ldlD-6B cells may be rapidly degraded and that only a fraction of non-degraded enzyme can be secreted. Therefore, we wished to assess the relative amounts of BSDL (75 kDa) and the possible degradation products (43-46 and 55 kDa). For this purpose, CHO ldlD-6B cells grown in 0.5% FCSsupplemented medium were pulse-labeled for 15 min with suggested that degradation of BSDL may occur in the transfected CHO ldlD cells.
In previous studies, we have shown that the ionophore monensin and brefeldin A inhibited the BSDL secretion; the ionophore also induced the retention of the enzyme in a membrane compartment of AR 4 -2J pancreatic cells (15). These drugs, which impair trans-Golgi to secretory vesicles and the endoplasmic reticulum to Golgi vesicular traffic, respectively, also inhibited the secretion of BSDL by either CHO K1-3B and CHO ldlD-6B cells. They also induced a concomitant increase of the BSDL activity within cell lysate of both transfected cell lines. The increase in activity was higher in CHO K1-3B cells supplemented with monensin and brefeldin A (170 Ϯ 10% and 310 Ϯ 20%, respectively) than in CHO ldlD-6B cells grown under non-permissive conditions (130 Ϯ 5% and 140 Ϯ 4%, respectively). However, these data suggested that brefeldin A and monensin may have blocked the degradative process of BSDL. Pulse-chase experiments were performed to analyze the effects of monensin and brefeldin A on BSDL present in cell culture medium and in cell lysate. As expected, these drugs led to increased retention of BSDL in CHO K1-3B cells and decreased the secretion of the enzyme (Fig. 4, upper panel). For CHO ldlD-6B cells, the signal detected after immunoprecipitation of the culture medium was also significantly decreased by monensin and brefeldin A treatment (Fig. 4, lower panel). In lysates of latter cells, grown under non-permissive conditions, multiple signals were detected after immunoprecipitation, and accumulation of material at 75 kDa, corresponding to undegraded BSDL, was evident (arrow). Intense bands presenting with M r of approximately 55 and 43-46 kDa (arrowheads), lower than that of full-length BSDL, were also detected. As with the CHO K1-3B cells, brefeldin A and monensin led to intracellular accumulation and decreased secretion of BSDL. When compared with the amount of undegraded BSDL, it appeared that the relative quantity of low M r material was much higher in lysate of CHO ldlD-6B cells cultured in the absence than in the presence of drugs. The same pattern was obtained with CHO ldlD-6B cells cultured under permissive conditions (data not shown), whereas a minute amount of this material was detected in CHO K1-3B cell lysate (Fig. 4, upper  panel). Hence, brefeldin A and monensin induced the intracellular accumulation of undegraded enzyme. These data indicated that low M r bands can be generated by a breakdown process of BSDL. Although unlikely, it is possible that low M r material resulted from a co-precipitation of proteins associated with BSDL (15).
To overcome the possible co-precipitation of proteins associated with BSDL, lysate from either CHO K1-3B and CHO ldlD-6B cells was subjected to a two-step immunoprecipitation procedure (22) with polyclonal antibodies to rat BSDL. After the two-step immunoprecipitation, BSDL can be specifically detected in cell lysate of CHO K1-3B cells, as observed after the one-step immunoprecipitation procedure, whereas in CHO ldlD-6B cell lysate the low M r material, although decreased, can still be precipitated. Therefore, it is likely that the low M r material immunoprecipitated in CHO ldlD-6B cells with antibodies to rat BSDL reflects some degradation of BSDL.
Glycosylation of BSDL Expressed in CHO ldlD-6B Cells-At this point, we needed to take into account one fundamental fact, which concerns the ability of CHO ldlD cells to glycosylate the fraction of BSDL to be secreted. We first examined the incorporation of galactose and mannose in BSDL expressed under permissive conditions. For this purpose, CHO ldlD-6B cells were cultured in glucose-depleted RPMI, and then glycoproteins were metabolically radiolabeled with [ 14 C]galactose or [ 14 C]mannose (10 Ci/ml for 4 h). At the end of the incubation time, cell-free medium was withdrawn and cells were extensively washed with fresh medium, harvested, lysed, and clarified as above described. Cell-free medium and cell lysate (1 ml each) were subjected to immunoprecipitation and analyzed by SDS-PAGE. As shown in Fig. 5, radioactive material corresponding to BSDL can be precipitated either after mannose or galactose labeling. Immunoprecipitation obtained from cell lysate indicated the presence of three bands migrating at 68, 72, and 75 kDa; within this heterogeneous material, the 72 kDa was dominantly radiolabeled by [ 14  spond to different glycoforms of BSDL. These data demonstrated that CHO ldlD-6B cells glycosylated BSDL when grown under permissive conditions. The incorporation of mannose strongly suggested that the N-linked structure was present on BSDL (9); nevertheless, and because galactose may be incorporated in N-and O-glycans, we next investigated the possible presence of O-linked structures on BSDL. CHO ldlD-6B cells were consequently cultured in glucose-depleted RPMI in the presence of [ 14 C]galactose, and at the end of the incubation cell-free medium and cell lysate were obtained as above. The cell-free medium and the cell lysate (500 l) were adjusted to pH ϭ 5 with dilute HCl then treated overnight with 0.5 unit of sialidase at 37°C. During this time, PNA-agarose was extensively washed with a 10 mM Tris/HCl (pH 7.4) buffer containing 0.3% lactose, then with the same buffer without lactose, and then with the lysis buffer. Finally, 500 l of suspended PNAagarose beads were mixed together with the sialidase-treated cell lysate (adjusted to pH 7.4 with dilute NaOH) or cell-free medium and incubated overnight at 4°C on a roller. At the end of the incubation, the mixture was centrifuged for 15 min at 5000 ϫ g to separate beads from the unbound material present in the supernatant. Beads were washed at least twice with the lysis buffer and twice again with a buffer containing 10 mM Tris/HCl (pH 7.4) and EDTA (5 mM). The adsorbed material was then eluted with 100 l of a buffer containing 0.1 M Tris/ HCl (pH 7.4), 2% SDS, and 10% glycerol, boiled 2 min, and centrifuged for 15 min at 10,000 ϫ g. The eluted material found in the supernatant was then 10-fold diluted with a buffer containing 0.1 M Tris/HCl (pH 7.4), 3% Triton X-100, and 0.15 M NaCl. The presence of BSDL in unbound and bound material was then investigated by immunoprecipitation, SDS-PAGE, and autoradiography. A typical result of this experiment is displayed on Fig. 6. Clearly, in both cell lysate and cell-free medium, BSDL partitioned between unbound (lane 1) and bound (lane 2) fractions obtained on PNA-agarose beads. The presence of [ 14 C]galactose-labeled BSDL within the fraction bound to PNA-agarose strongly suggested the presence of O-linked structures on the BSDL expressed by CHO ldlD-6B cells. Nevertheless, the presence of BSDL in unbound fractions suggested that the O-glycosylation of BSDL was not complete or that the sialidase treatment was not totally efficient (PNA recognizes Gal␤1-3GalNAc-O-T/S structures). To substantiate this result, [ 14 C]galactose-labeled BSDL was immunoprecipitated and subjected to mild alkaline treatment according to Carlsson et al. (29). At the end of the ␤-elimination, BSDL was precipitated with trichloroacetic acid (20% w/v) and centrifuged (15,000 ϫ g, 4°C, 30 min). Under these conditions, Ϸ58% of the starting radioactivity was found in the supernatant. This result suggested that O-linked glycans, represented by labile material, had incorporated some 60% of the [ 14 C]galactose that labeled BSDL. The remaining [ 14 C]galactose coprecipitates with BSDL and was likely incorporated into N-linked glycans. However, when BSDL was labeled with [ 14 C]mannose and treated under the same conditions, the material released represented less than 20% of the starting radioactive material. In another set of experiment, [ 14 C]mannose or [ 14 C]galactose-labeled BSDL was immunoprecipitated and boiled in the presence of 0.5% SDS and 5% ␤-mercaptoethanol and finally treated with 20 units/ml PNGase F (37°C, 18 h), precipitated with trichloroacetic acid, and centrifuged as above. Under these conditions, approximately 63% of mannose and 40% of galactose radioactivity was released. In controls where PNGase F was omitted, more than 80% of the radioactive material was found in the trichloroacetic acid pellet, which corresponded to the trichloroacetic acid precipitation yield (90% as determined by [ 35 S]methionine-labeled BSDL). Taken together, these data suggested that mannose was, as expected, mainly incorporated in N-linked structures. Additionally, galactose partitioned between N-(approximately 40%) and O-glycosylation (approximately 60%). Obviously, binding to PNA and galactose incorporation are good evidence for the O-glycosylation of BSDL. Therefore, we next investigated the glycosylation of BSDL expressed by the CHO ldlD-6B cells grown under non-permissive conditions. For this purpose, CHO ldlD-6B cells were incubated for 4 h with [ 14 C]galactose in glucose-depleted RPMI after adapting them for 24 h to non-permissive conditions. Labeled proteins present in the cell-free medium were then examined before and after immunoprecipitation. Three proteins present in the cell-free medium were detected at approximately 75, 50, and 34 -36 kDa (Fig. 7, lane 1); only BSDL (75 kDa, arrow) and trace of the lower M r protein were immunoprecipitated by antibodies specific for rat BSDL (lane 2). The same data were obtained when cells were incubated with [ 14  BSDL of cell-free medium was immunoprecipitated (Fig. 8, lane 1) or treated with sialidase (0.5 unit) and subjected to the PNA-agarose fractionation as described above. The unbound and bound fractions were then analyzed on SDS-PAGE and autoradiography. Results of a typical experiment are given in Fig. 8 (lanes 2 and 3). As shown in this figure, radioactive material corresponding to BSDL (M r ϭ 75 kDa, arrow; the low M r material may be a degradation product) was detected in the bound fraction but could not be visualized in the unbound fraction. These data indicated that BSDL was totally adsorbed on the PNA-agarose beads and consequently appeared O-glycosylated. Therefore, we can suggest that the fraction of BSDL secreted by CHO ldlD-6B cells, grown under conditions that did not normally allow glycosylation of proteins, was indeed O-glycosylated.
Glycosylation of C-terminal Repeats of BSDL Expressed in CHO ldlD-6B Cells-Although it is known that the O-linked glycosylation of BSDL strictly locates within the C-terminal tandem-repeated sequences that comprise PEST sequences (23,24), we next attempted to demonstrate that the O-glycosylation of BSDL expressed by CHO ldlD-6B cells was effectively located on the C-terminal tail of the enzyme. For this purpose, we applied the procedure described by Mas et al. (23) to characterize the O-glycosylated peptide that carries the J28 epitope of the oncofetal variant of BSDL (30). Therefore, BSDL expressed by CHO ldlD-6B cells was immunoprecipitated, labeled with [ 3 H]DFP, desialylated, and cleaved by cyanogen bromide. Peptides were then fractionated on the PNA-agarose column. Alternatively, BSDL was metabolically labeled with [ 14 C]galactose. Under these conditions, the labeling with [ 3 H]DFP was omitted. As shown on Fig. 9A, after elution of unbound fractions, which include the [ 3 H]DFP-labeled N-glycopeptide, a non-radioactive material detected by dot blot was eluted with lactose. This material could not be detected when the desialylation step was omitted. When BSDL was labeled with [ 14 C]galactose (Fig. 9B), the radioactivity partitioned between unbound (Ϸ40%) and bound material (Ϸ60%). These data suggested that galactose was associated with the O-glycosylated peptide interacting with PNA. Mas et al. (23) have shown that the sequence of this peptide corresponded to that of the C-terminal domain of BSDL. Therefore, as expected, CHO ldlD-6B cells O-glycosylate the C-terminal tandem-repeated sequences of BSDL.

DISCUSSION
The structures of the rat and human BSDL genes were determined a few years ago (31,32). The organization of these genes indicated the presence of 11 exons interrupted by 10 introns. Each exon may encode an unique structural or functional domain of the enzyme. The largest exon, number 11, encodes the signal sequence for degradation (11) (PEST sequences) and the mucin-like tandem repeated sequences (33). The function of this domain, the size of which varies by species, remains unknown, but it may function to dictate intracellular processing and transport of BSDL (33). Recent studies (14,15) allowed us to postulate that the C-terminal domain of BSDL may be responsible for the interaction of the enzyme with a membrane folding complex including a 94-kDa protein (p94) immunologically related to Grp94. The release of BSDL from membranes occurs once terminal sugars are added to glycans, and the Grp94-related protein was assumed to assist the proper sorting of BSDL from the endoplasmic reticulum (ER) to the trans-Golgi (15). Because no intracellular degradation of  (23). The peptides were desialylated and separated on a PNA-agarose column equilibrated and eluted with PBS ϩ 0.5% Triton X-100, washed with PBS ϩ 0.1% Triton X-100, and finally the O-glycosylated peptide was eluted with PBS containing 0.3 M lactose. The radioactivity was determined in each fraction (Ⅺ), and bound material eluted by lactose was detected by dot blot using antibodies specific for BSDL and quantitated by scanning (E). B, BSDL expressed by CHO ldlD-6B cells was metabolically labeled with [ 14 C]galactose and immunoprecipitated. BSDL was then treated with cyanogen bromide; peptides were desialylated and analyzed as in A. The 14 C radioactivity was determined in every eluted fraction (‚). BSDL occurred in pancreatic cells upon treatment with drugs that affect secretion (9), the association of BSDL with microsome membranes by means of the Grp94-related protein may be important for the O-glycosylation of the C-terminal tandemrepeated sequences of BSDL which includes PEST sequences (15). Thus, we hypothesized that the O-glycosylation of this C-terminal domain can divert BSDL from entering into a degradative route.
To investigate this possibility, two CHO cell lines, the glycosylation-defective ldlD line and the wild-type K1 line, were transfected with the full-length cDNA of the rat BSDL. The CHO-ldlD cell line is markedly deficient in UDP-Gal and UDP-GalNAc 4-epimerase activity, leading to a defect in O-linked glycosylation when these cells are grown in 0.5% FCS-supplemented medium (16). The abnormal glycosylation of proteins could be completely reversed by providing the cells with exogenous sources of galactose and N-acetylgalactosamine (free sugars or 10% FCS) (34). In this study, we showed that the rate of secretion of BSDL by CHO ldlD-6B cells was 1/30 to 1/50 that observed with CHO K1-3B cells. The difference was even more significant when CHO ldlD-6B cells were grown under nonpermissive conditions. This large difference cannot be explained by modification of translational rate, although the amount of BSDL mRNA present in CHO ldlD-6B cells was 5-fold lower than that of CHO K1-3B. Therefore, it is possible that the low amount of BSDL secreted by CHO ldlD-6B cells can reflect the intracellular degradation of BSDL as a result of its O-linked glycosylation defect. We have further examined the behavior of BSDL within CHO K1-3B and CHO ldlD-6B cells by pulse-chase experiments followed by immunoprecipitation of cell lysates with antibodies to BSDL. The results suggested that low M r material can be precipitated; however, it increased slowly with time, whereas that of full-length BSDL decreased rapidly. Moreover, low M r material was detectable after one-and two-step immunoprecipitation procedures performed with CHO ldlD-6B cell lysate. This material was even present after a short pulse-chase experiment, but was absent in CHO K1-3B cell lysate. Treatment of transfected cells with brefeldin A and monensin resulted in an accumulation of BSDL in CHO K1-3B and in CHO ldlD-6B cells. Consequently, rapid degradation of BSDL expressed in CHO ldlD-6B cells grown under conditions that prevent O-glycosylation was strongly suspected. The site for BSDL degradation remains unknown; however, the effect of brefeldin A, which blocked the degradation of the protein, suggested the involvement of a post-ER compartment. Monensin also blocked the degradative process of BSDL, implying that a post-Golgi compartment could be involved. Therefore, the diversion of BSDL to a post-Golgi compartment for degradation can be compared with that of apolipoprotein (apo) E and B, which are secreted by hepatocytes. Ye et al. (35) reported that apoE may be degraded in a post-Golgi compartment by lysosome proteases and cytosol Ca 2ϩdependent cysteine proteases. In addition, apoB degradation occurred in a post-ER location, which is associated with a Golgi membrane fraction via the action of cysteine proteases (36). Similarly, BSDL could also be degraded in lysosomes or in cytosol. Degradation of misfolded ER luminal proteins by the cytosolic ubiquitin-proteasome pathway is now well documented (37). Nevertheless, one may consider that BSDL is a PEST sequence-containing protein. It is supposed that PEST sequences target proteins for degradation in the cytosol, possibly by calpains (38). However, many reports have suggested that PEST regions may constitute the recognition sites for ubiquitination and degradation by the 26 S proteasome (39). PEST motifs are often conditional proteolytic signals, and there are a number of way to expose or mask them to proteolytic factors (39). Therefore, a model for BSDL degradation could be a route to cytosol that involves the diversion of post-Golgi vesicles to an intracellular location in which the contents of the vesicles are degraded. It is known that the targeting of secretory vesicles is sensitive to the state of cargo. Consequently, when the C-terminal domain of BSDL is not O-glycosylated, the protein could be included in vesicles diverted from secretion and targeted to a post-trans-Golgi compartment for degradation; this route is invalid when BSDL is normally O-glycosylated. Conceivably, the membrane complex to which BSDL is associated during its transport from ER to a post-Golgi compartment (14,15) may be involved in the routing of the enzyme either to secretion or to degradation according to the glycosylation state of the enzyme. Accordingly, when BSDL is fully glycosylated, it is released from membrane (15) and routed to secretion. Otherwise, BSDL could be kept associated with the complex and targeted to degradation.
Because O-linked glycosylation can influence the rate at which a glycoprotein is secreted (16,40), this would provide a clear explanation as to why the secretion rate of BSDL increased with the cell ability to O-glycosylate in the following rank order: CHO ldlD-6B cells grown under non-permissive conditions Ͻ CHO ldlD-6B cells grown under permissive conditions Ͻ CHO K1-3B cells. We further showed that BSDL secreted by CHO ldlD-6B cells was O-glycosylated; the protein incorporated alkaline-sensitive galactose on C-terminal tandem-repeated sequences and carried out Gal␤1-3GalNAc-O-T/S structures, which were recognized by the PNA lectin. Incorporation of mannose also demonstrated the presence of an N-linked structure (9). The fraction of BSDL secreted by CHO ldlD-6B cells cultured under non-permissive conditions had also incorporated galactose and was recognized by the PNA lectin. Therefore, these data indicated that O-linked glycans are present on BSDL secreted by CHO ldlD-6B cells independently of culture conditions.
In conclusion, CHO ldlD-6B cells, when grown under conditions that should prevent O-glycosylation, secreted a low amount of BSDL. This protein was apparently rapidly degraded intracellularly, but a small fraction of BSDL was able to be O-glycosylated and secreted. Taken together, these data suggest that the rate of secretion of BSDL depends upon the ability of the cell to O-glycosylate C-terminal repeats of the protein.