Tagging Secretory and Membrane Proteins with a Tyrosine Sulfation Site TYROSINE SULFATION PRECEDES GALACTOSYLATION AND SIALYLATION IN COS-7 CELLS*

Sulfation of proteins on tyrosines is a late Golgi modi- fication that can be used to label proteins with [36Slsu1-fate for the analysis of post-Golgi transport. To extend the use of this modification to proteins not naturally sulfated, we fused a tyrosine sulfation site, the carboxyl- terminal nonapeptide of cholecystokinin precursor, to the carboxyl terminus of two normally unsulfated pro- teins: a,-proteinase inhibitor, a secretory protein, and subunit H1 of the asialoglycoprotein receptor, a type I1 membrane protein. The tagged proteins were efficiently sulfated in transfected COS-7 and Madin-Darby canine kidney cells. Specifically in COS-7 cells, the proteins were sulfated before they were galactosylated and sialylated and were converted to the mature forms with a half-time of approximately 23 min. This is in contrast to other cell types in which tyrosine sulfation was found to be virtually the last modification of the Golgi apparatus. Our results suggest that tyrosine sulfation occurs before the trans-Golgi in transfected COS-7 cells. To study individual steps of intracellular membrane transport, two approaches have been widely used in mammalian systems: the use of conditions that result in the reversible accumulation of transported markers in a specific compartment

To study individual steps of intracellular membrane transport, two approaches have been widely used in mammalian systems: the use of conditions that result in the reversible accumulation of transported markers in a specific compartment from which transport can be followed upon shift to permissive conditions, and the use of compartment-specific modifications that allow labeling of proteins in transit and following their fate in pulse-chase experiments. For the characterization of post-Golgi transport, proteins can be reversibly accumulated in the trans-Golgi network by reducing the temperature to 20 "C (Matlin and Simons, 1983;Saraste and Kuismanen, 1984). However, depending on the cell type this transport block may not be effective for longer than 1-2 h. As a result, proteins which are slowly transported from the endoplasmic reticulum through the Golgi are not significantly accumulated in this period of time.
The sulfation of proteins on either carbohydrates or tyrosine residues is a n established post-translational modification specific for the trans-Golgi (reviewed by Huttner (1988)). Sulfation is particularly suitable for pulse labeling of proteins because [35Slsulfate is available at high specific radioactivity, is rapidly * This work was supported by Grant 31-34008.92 from the Swiss National Science Foundation, by the Incentive Award of the Helmut Horten Foundation, and by the Roche Research Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
5 To whom correspondence should be addressed: Dept. of Biochemistry, Biozentrum, University of Basel, Klingelbergstrasse 70, CH-4056 Basel, Switzerland. Fax: 41-61-267-2148. converted to the active precursor phosphoadenosine phosphosulfate and imported into the lumen of the Golgi, but is not incorporated into nascent proteins via methionine or cysteine (Huttner, 1984). Sulfation is considered to be one of the last Golgi modifications, occurring even after sialylation of attached oligosaccharides (Baeuerle and Huttner, 1987). Labeling of sulfated proteins with [35Slsulfate has been used successfully to study the sorting of proteins of the regulated and constitutive secretory pathways into distinct vesicle populations (e.g. Tooze and Huttner (1990) and Grimes and Kelly (1992)).
However, most proteins are not sulfated. Relatively few sulfated proteins have been identified so far, mainly secretory proteins and only few membrane proteins (Huttner, 1988). In this study we have extended the use of sulfation for transport studies to proteins not naturally sulfated by introducing a tyrosine sulfation site by in vitro mutagenesis of the encoding DNA. Tyrosine residues are sulfated by the enzyme tyrosylprotein sulfotransferase which recognizes tyrosines in exposed protein domains containing acidic amino acids (Hortin et a l . , 1986;Rosenquist and Nicholas, 1993). We introduced a potential sulfation site into two proteins which are normally not labeled with [35S]sulfate: human a,-proteinase inhibitor (AlPi),l a secretory protein, and subunit H1 of the human asialoglycoprotein (ASGP) receptor, a type I1 membrane protein. We have chosen the carboxyl-terminal nonapeptide sequence of rat cholecystokinin precursor (proCCK) which is tyrosine sulfated in vivo (Adrian et al., 1986). When this peptide sequence was fused to the exposed carboxyl terminus ofAlPi or H1, the tagged proteins AIPiTS and HITS were efficiently sulfated in transfected COS-7 and Madin-Darby canine kidney (MDCK) cells. Surprisingly, sulfation of these proteins in COS-7 cells could be observed before galactosylation and sialylation, indicating that the relative compartmentalization of the corresponding transferases in transfected COS-7 cells is unusual.

8115
DNA Constructs-'ho complementary oligonucleotides were synthesized, CAGCGCAGAAGACTACGAATACCCATCTI'GAGCT and CAA-GATGGGTATTCGTAGTCTTCTGCGCTGGTAC, which encode the 9 carboxyl-terminal residues of rat cholecystokinin precursor followed by a stop codon and produce terminal protruding ends corresponding to a KpnI and a SacI site at the 5' and 3' end, respectively. Both oligonucleotides were 5'-phosphorylated with polynucleotide kinase.
A KpnI site was introduced into the cDNA of the ASGP receptor subunit H1 (Spiess et al., 1985) following codon 279 by polymerase chain reaction using the antisense primer GTTCGGTACCTGTCTCG-CAGACC (the KpnI site is underlined) in combinaGGi3Ka primer in the vector sequence. The amplified cDNA (from the Hind111 to the KpnI site) was ligated into the vector plasmid pGEM4. Analogously, a KpnI site was introduced into the cDNA of AlPi (variant MZ Brantly et al. (1988)) following codon 392 using the antisense primer CGCTGGTACand SacI and ligated with the annealed synthetic oligonucleotides. The new 3' ends of the cDNAs were verified by DNA sequencing. To avoid mutations potentially introduced by Taq polymerase, most of the amplified sequences were replaced by corresponding segments from nonamplified plasmids by restriction enzyme digestion and ligation. The cDNAs were subcloned into pECE (Ellis et al., 1986) for transient expression in COS-7 cells and into pK20 (with a Rous sarcoma virus promoter; provided by R. Gentz, Hoffmann-La Roche, Basel) for the production of stable MDCK cell lines. Stable lines expressing the H1 fusion protein were only obtained after nucleotides 10661278 of the original H1 cDNA (corresponding to most of its 3' untranslated sequence) had been added to the cDNA in the plasmid pK20.
Cell Culture and Dansfection-COS-7 cells and MDCK cells (strain 11) were grown in minimal essential medium with 10% fetal calf serum, supplemented with 2 IIIM L-glutamine, 100 unitdml penicillin, and 100 mg/ml streptomycin sulfate. For transient expression, wild-type and mutant cDNAs subcloned in the expression plasmid pECE were transfected according to Cullen (1987) into COS-7 cells in 10-cm plates. To ensure uniform expression within each experiment, the cells were trypsinized 1 or 2 days after transfection, mixed, and seeded into 35-mm wells for use on the following day (i.e. 40-72 h after transfection).
For stable transfection, MDCK cells were cotransfected with pK20 containing the cDNA and pSVneo at a ratio of 1 O : l using Polybrene and dimethyl sulfoxide according to Kawai and Nishizawa (1984). After 2 days, cells were split into selective medium containing 1 mg/ml G418sulfate. After 14 days, resistant colonies were isolated and screened for expression by immunoblot analysis using a polyclonal antiserum directed against the ASGP receptor.
Labeling with r5S]MethioninelCysteine"COS-7 cells expressing wild-type and mutant AlPi were washed with phosphate-buffered saline (PBS), starved for 45 min in methionine-free medium (Selectamine kit, Life Technologies, Inc.), labeled for 30 min in 0.5 ml of methioninefree medium supplemented with 100 pCi/ml [35Slmethionine, washed with PBS, and chased for different times a t 37 "C in complete medium containing excess methionine. The medium was collected, and the cells were washed twice with ice-cold PBS and lysed in 500 pl of lysis buffer (1% Triton X-100, 0.5% deoxycholate, 2 m~ phenylmethylsulfonyl fluoride in PBS). Chase media and cell lysates were immunoprecipitated with affinity purified rabbit anti-human AlPi antibody and protein A-Sepharose (Pharmacia, Uppsala, Sweden). The samples were analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. Quantitation was performed with a Molecular Dynamics 300A computing densitometer. COS-7 cells expressing wild-type and mutant H1 were starved in medium lacking methionine and cysteine for 45 min, labeled with 100 pCi/ml [35Slmethionine and [35Slcysteine for 1 h, and chased in complete medium containing excess methionine and cysteine. The cells were lysed and immunoprecipitated with a polyclonal rabbit antiserum raised against the purified ASGP receptor.
Labeling with P5SISulfate-Cells were grown in 35-mm wells, washed with PBS, and incubated in sulfate-free medium (containing one-tenth of the normal concentration of methionine and cysteine, 350 mglliter NaHC03, buffered with 20 m~ Hepes, pH 7.2) for at least 1 h a t 37 "C. The medium was then removed and replaced with 0.5 ml of the same medium supplemented with 1 mCi/ml [35Slsulfate. Labeling was performed either in the incubator at 37 "C or a t 20 "C in a water bath.
For pulse-chase experiments the cells were labeled for 7 min at 37 "C, quickly washed with PBS a t room temperature, and incubated a t 37 "C with 0.6 ml of chase medium (minimal essential medium containing excess sulfate) which had been pre-equilibrated in the 37 "C incubator. At the end of the chase, the plates were placed on ice, the medium was collected, and the cells were washed with ice-cold PBS and lysed in lysis The carboxyl-terminal nonapeptide of rat cholecystokinin precursor (proCCK, shown in bold with the two sulfatable tyrosines outlined) and two linker residues replace the carboxyl-terminal 2 amino acids ofAlPi inAIPiTS and the carboxylterminal 12 amino acids of H1 in HITS.
buffer. Chase media and cell lysates were processed for immunoprecipitation as described above.
Carbohydrate Analysis-For Endo F digestion, the protein A-Sepharose immunocomplexes were suspended in 100 pl of 0.1 M sodium phosphate, pH 6.8, containing 50 m~ EDTA, 1% P-mercaptoethanol, 0.1% SDS, and boiled for 5 min. After addition of Nonidet P-40 to a final concentration of 1%, samples were incubated with 0.25 units of Endo F for 3 h at 37 "C. For digestion with Endo H, the immunocomplexes were suspended in 100 pl of 100 n" sodium citrate, pH 6.0, 1% SDS, boiled for 5 min, and incubated with 2 milliunits of Endo H for 3 h a t 37 "C. For desialylation, the immunocomplexes were incubated without prior denaturation with 5 milliunits of neuraminidase in 50 pl of 0. l M sodium acetate, pH 5.6, for 3 h at 37 "C. For subsequent 6-galactosidase treatment, the sample was denatured by addition of 6 pl of 1% SDS and boiling for 5 min, supplemented with Triton X-100 to a final concentration of 1%, and incubated for 16-24 h a t 37 "C with 10 milliunits of 6-galactosidase. The samples were processed for SDS-polyacrylamide gel electrophoresis and fluorography. Control samples were treated identically except that no enzyme was added.
For ricin binding assays, immunocomplexes were supplemented with SDS to a final concentration of 0.2%, boiled for 5 min, diluted with 400 pl of binding buffer (100 n" Tris-HC1, pH 7.0, 150 n" NaC1, 1 m~ MgC12, 1 m~ CaC12), and centrifuged briefly. The supernatant was taken off the protein A-Sepharose beads and incubated with shaking for 2 h a t room temperature with 70 pl of a 50% (packed gel per volume) suspension of ricin-agarose. The agarose beads were pelleted for 5 min at 500 x g to separate free and ricin-bound protein. The supernatant was concentrated by precipitation in 75% acetone (with 80 pg of bovine serum albumin as a carrier) at -70 "C. After centrifugation for 5 min at 10,000 x g at 4 "C the pellets were redissolved in SDS-sample buffer. The pelleted agarose beads were washed twice with binding buffer, followed by boiling in SDS-sample buffer to elute bound protein.
Stoichiometry ofSulfation-The stoichiometry of sulfation of AIF'iTS was determined according to Huttner (1984). Transfected COS-7 cells were preincubated with medium lacking tyrosine for 30 min at 37 "C and labeled for 6 h in the same medium supplemented with 40 pCi/ml [3Hltyrosine. Labeled AIPiTS was immunoprecipitated from the medium and hydrolyzed in 0.2 M Ba(OH), at 110 "C for 20 h. The hydrolysate was neutralized with sulfuric acid and centrifuged. The supernatant was lyophilized, redissolved in 20 pl of electrophoresis buffer (7.8% acetic acid, 2.2% formic acid, pH 9), mixed with 100 pl of acetone, and centrifuged. The supernatant was dried, redissolved in electrophoresis buffer, and spotted onto a cellulose thin layer chromatography sheet (20 cm x 20 cm) 7 cm from the anionic edge. The sheet was wetted with the same buffer and run at 750 V for approximately 45 min to separate tyrosine and sulfotyrosine. t3H1Tyrosine and sulf~-[~H]tyrosine were quantified by liquid scintillation counting.

RESULTS
Fusion of a Potential Tyrosine Sulfation Site to AlPi and HI-In Fig. 1 the sulfated nonapeptide sequence of proCCK is shown. I t was chosen as a carboxyl-terminal tyrosine sulfation tag, because it is also carboxyl-terminal in proCCK, it contains two potentially sulfated tyrosines, and a synthetic peptide of this sequence has been shown to be a substrate for tyrosylprotein sulfotransferase (Niehrs et al., 1990a(Niehrs et al., , 1990b. The secretory protein AlPi has its carboxyl terminus exposed on the surface of the molecule as is evident from its crystal structure (Loebermann et al., 1984). Synthetic oligonucleotides corresponding to the nonapeptide sequence were fused to the cDNA with prosine Sulfation Sites 8117  and 8-10). Wild-type and mutant AlPi was immunoprecipitated from cell lysates (C) and media ( M ) . Alternatively, the cells were labeled for 2 h with ["?S]sulfate (lanes 4-7 and 11-14). A 5% aliquot of the total cell lysate and media (7') was directly analyzed, the rest was immunoprecipitated UP). Panel B, COS-7 cells expressing H1 and HITS were pulse-labeled for 1 h with ["S]methionine and [35S]cysteine and chased for 0, 3, or 5 h as indicated (lanes 1, 4, and 5 ) .
Alternatively, the cells were labeled for 2 h with [" SJsulfate (lanes 2, 3, 6 , and 7). A 5 6 aliquot of the total cell lysate and media (T) was directly analyzed, the rest was immunoprecipitated U P ) . Samples were analyzed by SDS-gel electrophoresis and fluorography.
of AlPi via a linker segment encoding Gly-Thr which replaced the last two codons. It has previously been shown that the terminal 3 amino acids of AlPi are not essential and can be deleted without affecting secretion (Brodbeck and Brown, 1992).
Subunit H1 of the human ASGP receptor is a single-spanning type I1 membrane protein. The carboxyl-terminal half of the protein forms a galactose binding domain homologous to a large family of calcium-dependent carbohydrate binding proteins (Drickamer, 1988;Geffen and Spiess, 1992). The homology does not extend beyond Cys-277 (which is involved in a disulfide bridge), suggesting that the carboxyl-terminal 14 residues are not essential for protein folding. The proCCK nonapeptide was fused to Thr-279 via the linker sequence Gly-Thr, thereby replacing the 12 terminal residues. The two fusion proteins carrying the tyrosine sulfation peptide were named HIT" and AIPiTS, respectively.

Expression and Sulfation of the Fusion Proteins in COS-7
Cells-In vivo sulfation of the fusion proteins was analyzed in transiently transfected COS-7 cells. AIPiTS, HITS, and the wild-type proteins were expressed in COS-7 cells, labeled with either ["Slmethionine or [35SJsulfate, immunoprecipitated, and analyzed by gel electrophoresis and fluorography (Fig. 2). After labeling for 30 min with [3sS]methionine, AIPiT" was detected as a 50-kDa protein (Fig. 2 A , lane 8 ) corresponding to the oligomannose glycosylated precursor form ofAlPi (lane 1 ). Following a 2-h chase, both labeled proteins were recovered in the medium with a somewhat reduced mobility reflecting maturation to the complex glycosylated forms (lanes 2,3,9, and 10). After incubation for 2 h with [35S]sulfate, wild-type AlPi was not labeled (lanes 5 and 7), whereasAIPiTS was efficiently labeled and could even be detected in a small aliquot of the total medium without immunoprecipitation (lanes 13 and 14). Radioactive AIPiTS could also be detected in the cells (lanes 11 and 12). The protein immunoprecipitated from the cells (lane 12) was predominantly of the mature form; however, a small fraction with a lower apparent molecular weight was also detectable (see below).
Wild-type H1 was expressed and labeled with [35S]methionine and [35Slcysteine, but did not incorporate [35Slsulfate (Fig. 2B, lanes 1 3 ) . The fusion protein HITS, however, was efficiently labeled with [355S]sulfate. While a considerable fraction of ["5Slmethionine/cysteine-labeled HITS was present as the oligomannose-glycosylated precursor form of 40 kDa after a 3-h chase (lane 5 ) , only the mature 46-kDa form was labeled with radioactive sulfate in this experiment (lane 6). Besides the proteoglycans (which accumulate on top of the gel), HIT" was the major sulfated protein synthesized in these transfected COS-7 cells, since it was easily detected in total cell lysates (lane 7).
To test whether the extension of AlPi at its carboxyl terminus by the tyrosine sulfation peptide affected its transport behavior in the secretory pathway, we determined the kinetics of secretion of [35S]methionine-labeled AlPi and AIPiTs. Transfected COS-7 cells were pulse-labeled for 30 min and chased for up to 2 h. Wild-type or mutant AlPi was immunoprecipitated from the medium and the cell lysate, and quantitated by gel electrophoresis, fluorography, and densitometric scanning. As shown in Fig. 3, secretion of AlPi and AIPiTS occurred with very similar kinetics: 50% secretion was observed after 40-45 min of chase. This suggests that the tyrosine sulfation tag did not significantly alter the transport rate of AlPi from the endoplasmic reticulum to the cell surface.
To determine the stoichiometry of sulfation, COS-7 cells ex-pressingAIPiTS were labeled for 6 h with [3H]tyrosine, secreted AIPiTS was immunoprecipitated, and hydrolyzed with Ba(OH),. Tyrosine and sulfotyrosine were separated by thin layer electrophoresis, and [3Hltyrosine and sulf~-[~H]tyrosine were quantified by liquid scintillation counting. In four experiments, 11.0 2 1.8% of labeled tyrosine residues were present in sulfated form (not shown). Since there is a total of 8 tyrosines in AIPiTS, this corresponds to approximately 1 sulfotyrosine/ molecule. This value suggests that in average only one of the two potentially sulfatable tyrosines in the proCCK tag sequence is modified. Tino Intracellular Forms of AIPiTS Are Sulfated in COS-7 Cells-The introduction of a tyrosine sulfation tag allowed the determination of the transport kinetics of AIPiTS specifically were labeled with ["Slsulfate for 1-7 min a t 37 "C, and AIPiTS was immunoprecipitated from the cell lysates. Panel C, cells were labeled with [""Slsulfate for up to 1'00 min at 20 "C, and AIPiTs was immunoprecipitated from the cell lysates ( C ) and the medium (M). Samples were analyzed by gel electrophoresis and fluorography. For better visualization of the initially labeled material, lanes 1 and 2 of panel C were exposed longer than the other lanes.

37°C
A B Tyrosine Sulfation Sites from the site of sulfation to the cell surface. Based on previously published studies, we expected tyrosine sulfation to occur in the trans-Golgi network. Transfected COS-7 cells were pulse-labeled with [35S]sulfate for 7 min a t 37 "C and chased for up to 35 min. Cell-associated and secreted labeled AIPiTS was immunoprecipitated and analyzed by gel electrophoresis and fluorography, as shown in Fig. 4A. The results show that a precursor form of lower apparent molecular weight was sulfated. During the chase, this material was converted to the larger, mature form of the protein, which was finally released into the medium. Conversion of the precursor to the mature form occurred with a half-time of approximately 2-3 min.
Quantification of similar experiments (by densitometric scanning of both cellular forms versus secreted AIPiTS) yielded the kinetics of transport from the intracellular site of sulfation to the cell surface, as shown in Fig. 5. After a lag time of approximately 10 min the protein began to appear in the medium. After approximately 17 min of chase, 50% of the labeled AIPiTS was secreted.
When cells were incubated with [35S]sulfate for very short periods a t 37 "C, the lower form ofAIPiTS was labeled first and was detected after 3 min of labeling (Fig. 4B). The higher, mature form appeared approximately 2 min later. Exocytic transport can be blocked specifically at the level of the trans-Golgi network by lowering the temperature to 20 "C (Matlin and Simons, 1983;Saraste and Kuismanen, 1984). When COS-7 cells expressing AIPiTS were incubated at 20 "C with [35S]sulfate (Fig. 4C), only the smaller labeled form was detected after 10 min of labeling (lane l ). After longer labeling periods, the higher molecular weight form appeared and increased in intensity with time (lanes 3,5, and 7). This confirms that the lower form corresponds to a transient precursor transport step, i.e. before exit from the trans-Golgi network.

Sulfation of AIPiTS in COS-7 Cells Occurs before Galactosylation and Sialylation-Human
AlPi contains three asparagine-linked, biantennary oligosaccharide side chains (Mega et al., 1980). Different carbohydrate modifications were thus likely to account for the two sulfated forms of AIPiTS. To analyze the structural difference between the precursor and the final form of sulfated AIPiTS, transfected COS-7 cells were labeled with [35S]sulfate for 25 min at 20 "C. Under these conditions, the two forms were recovered in approximately equal amounts (Fig. 6, lane 2). Both forms were resistant to Endo H treatment (lane 1 ). Since Endo H resistance is the result of medial-Golgi modifications, this indicates that sulfation occurred beyond the cis-Golgi. Deglycosylation with Endo F converted both forms to a single species of an apparent molecular mass of 44 kDa (lane 3 ) , demonstrating that the two forms differ within their carbohydrate moieties. Similarly, when COS-7 cells expressing AIPiTS were incubated with tunicamycin (3 pg/ml) to inhibit N-glycosylation and then labeled with [35Slsulfate for 25 min at 20 "C, only a single radioactive species of 44 kDa was immunoprecipitated corresponding to the unglycosylated protein (not shown).
Of the two forms of [35S]sulfated AIPiTS only the larger one was sensitive to neuraminidase (Fig. 6, lane 5), indicating that the final form, but not the precursor, was sialylated. However, desialylation did not convert the larger form to the smaller one, but resulted in a protein of intermediate electrophoretic mobility. The two species migrated with indistinguishable mobility only when the sample was treated with neuraminidase in combination with P-galactosidase (lane 6). This indicates that the precursor form was neither sialylated nor galactosylated. This was confirmed using the galactose-specific lectin ricin. Both the precursor as well as the final sulfated form of AIPiTS were recovered in the supernatant after incubation with ricin-agarose and centrifugation (lanes 7 and 8). After neuraminidase treatment, the desialylated final species was efficiently bound to ricin-agarose (lanes 9 and 1 0 ) and served as a positive control for the assay.

Sulfation of a Precursor of HITS Occurs in COS-7 Cells, but Not in MDCK Cells
-To test whether tyrosine sulfation of ungalactosylated and unsialylated AIPiTS is due to the particular substrate glycoprotein analyzed, or is a general property of COS-7 cells, we analyzed the sulfation of HITS in more detail in AIPiTS expressed in COS-7 cells was labeled with [3sSlsulfate for 25 min a t 20 "C and immunoprecipitated from the cell lysate. The immunoprecipitates were incubated with Endo H, Endo F, neuraminidase (Neu), neuraminidase and P-galactosidase (Neu + PGal), or no enzyme (-) as indicated (lanes 1-61. In addition, untreated (-) and neuraminidase-digested samples (Neu) were incubated with ricin-agarose, and bound and free protein separated by centrifugation into a pellet (P) and supernatant fraction (S), respectively (lanes 7-10). Samples were analyzed by gel electrophoresis and fluorography. COS-7 cells and compared it to the sulfation of HITS in a transfected MDCK cell line (Fig. 7). In transfected COS-7 cells, HITS showed the same characteristics as AIPiTS. Two species of different electrophoretic mobility were [35S]sulfated. After a 15min pulse with [35S]sulfate a t 20 "C, predominantly the faster migrating form was labeled (lane 1 ); with longer labeling times the proportion of the slower migrating form increased (lanes 2 and 5). During an additional incubation a t 37 "C in the presence of excess unlabeled sulfate, the faster form was completely converted into the slower one (lane 4 ) . Only the slower migrating form was sensitive to neuraminidase treatment which converted i t

into a species of intermediate mobility (lane 3 ) .
In contrast, in a stable MDCK cell line expressing HITS (MITS), only a single [35S]sulfated species was detected when the cells were labeled at 20 or at 37 "C ( Fig. 7, lanes 6-16). The labeled protein was sensitive to neuraminidase treatment (lane 8 ) confirming that the final, sialylated form of HITS is the substrate of tyrosine sulfation in MDCK cells. After 3 min of labeling a t 37 "C, i.e. at the earliest time point when sulfated HITS could be detected, the final form of HITS was found exclusively. Essentially the same was found for AIPiTS expressed in transfected MDCK cell lines (not shown). Sulfation of precursor forms of glycoproteins is therefore a specific property of COS-7 cells.

DISCUSSION
In this study, we have tagged two normally unsulfated proteins with a carboxyl-terminal tyrosine sulfation peptide. 5rosine sulfation is a post-translational modification of the late secretory pathway catalyzed by the enzyme tyrosylprotein sulfotransferase. Specific tyrosine residues are recognized in exposed protein segments typically rich in acidic residues (Hortin et al., 1986;Rosenquist and Nicholas, 1993). Synthetic peptides corresponding to identified tyrosine sulfation sites are sulfated by the enzyme in vitro (Niehrs et al., 1990a(Niehrs et al., , 1990b, indicating that these short sequences are sufficient for substrate recognition. In vivo, an artificial protein composed of repeated tyrosine sulfation sequences was shown to be stoichiometrically sulfated (Niehrs et al., 1992).
We fused the nonapeptide sequence of rat proCCK to the secretory protein AlPi and to the type I1 membrane protein H1 of the ASGP receptor. Both fusion proteins AIPiTS and HITS were efficiently sulfated in transfected COS-7 and MDCK cells, demonstrating that in different structural contexts the proCCK peptide can be a functional substrate of tyrosylprotein sulfotransferase in vivo. The sulfation tag facilitates the analysis of transport from the site of sulfation in the Golgi to the cell surface. After 7 min of pulse labeling, it takes -17 min for 50% ofAIPiTS to pass through this part of the secretory pathway in COS-7 cells. This value is similar to the half-time of secretion of -20 min determined for sulfated IgM in a hybridoma cell line (after a 5-min pulse; Baeuerle and Huttner (198711, and some-what higher than the half-times of -10 min determined for Drosophila yolk protein YP2 in mouse fibroblasts (after a 10min pulse ;Friederich et al. (1988)) and synaptophysin in PC12 cells (after a 5 min-pulse; RBgnier-Vigouroux et al. (1991)). When tyrosine sulfation ofYP2 was inhibited either by chlorate treatment or by mutation of the sulfated tyrosine to a phenylalanine, transport of unsulfated YP2 from the trans-Golgi to the cell surface was slowed down by 15-18 min compared to that of the sulfated protein (Friederich et al., 1988). In our system, however, we did not observe a significant difference in the overall rate of secretion of [35S]methionine-labeled AlPi and AIPiTS.
Tyrosine (and carbohydrate) sulfation is generally considered to be virtually the last modification in the Golgi, but in the absence of direct immunolocalization of tyrosylprotein sulfotransferase, this localization is based on indirect evidence. In hybridoma cells, sulfation of IgM occurs after galactosylation and sialylation (Baeuerle and Huttner, 1987). Tyrosylprotein sulfotransferase purified from bovine adrenal medulla is sialylated, suggesting a localization in or beyond the compartment of sialyltransferase (Niehrs and Huttner, 1990). pl,4-Galactosyltransferase and a2,6-sialyltransferase have been immunolocalized to the trans-cisternae of the Golgi apparatus and the trans-Golgi network (Roth and Berger, 1982;Strous et al., 1983;Slot and Geuze, 1983;Roth et al., 1985). In addition, the effect of brefeldin A on sulfation in PC12 cells is consistent with a localization of tyrosylprotein sulfotransferase in the trans-Golgi network (Rosa et al., 1992). In contrast, in 3T3 L1 adipocytes tyrosine sulfation was suggested to be localized to the medial-Golgi. This is based on the observation that sulfation of entactin is not affected by incubation with the ionophore monensin, while terminal glycosylation is blocked (Aratani and Kitagawa, 1988).
In our study, we observed that incompletely processed precursor glycoproteins are already sulfated in COS-7 cells. Both tagged proteins AIPiTS and HITS behaved identically. The sulfated precursor of AIPiTS differs from the mature form exciusively in the structure of the oligosaccharides. The precursor is endo H-resistant and insensitive to neuraminidase and p-galactosidase treatment, it is not recognized by the galactosespecific lectin ricin, and has the same electrophoretic mobility as the desialylated and degalactosylated mature form of the protein. As a result, the oligosaccharide structures of the precursor must be either (GlcNAc)z(Man),GlcNAc or (G~cNAc)~(M~~)~(GIcNAc)~ (assuming biantennary structures as determined for AlPi from serum). These carbohydrate structures are the products of mannosidase I1 and N-acetylglucosaminyltransferase 11, respectively, which are considered to be medial-Golgi enzymes (Dunphy et al., 1981;Dunphy and Rothman, 1983;Goldberg and Kornfeld, 1983).
Our results indicate that the relative compartmentalization of the three modification enzymes galactosyltransferase, sialyltransferase, and tyrosylprotein sulfotransferase is different in COS-7 cells than in other cell types. No intermediate forms between the precursor and the final sialylated form could be observed. Conversion of the precursor to the final form occurred in the time span of a few minutes. These two observations point to a transport step between a Golgi compartment where sulfation takes place and a subsequent compartment where both galactosyltransferase and sialyltransferase are predominantly active. Tyrosine sulfation thus appears to occur before the trans-Golgi. Consistent with this conclusion, secretion of [35S]sulfated AIPiTS from the cells was preceded by a lag period of approximately 10 min, which had not been observed in other cell types where only mature proteins are sulfated (Baeuerle and Huttner, 1987;Friederich et al., 1988;Regnier-Vigouroux et al., 1991).
The same fusion proteins which are sulfated as precursors in COS-7 cells are exclusively sulfated as the mature species in transfected MDCK cells. This shows clearly that the phenomenon observed in COS-7 cells is characteristic of that cell type rather than a property of the proteins analyzed. Different processing of secretory glycoproteins by different cells and tissues is considered to be due to qualitative and/or quantitative differences in the modification activities in the secretory pathway. Our results suggest that also the distribution of enzyme activities in the Golgi subcompartments can be different in different cell types.