Golgi galactosyltransferase contains serine-linked phosphate

Transcription

1 Eur. J. Biochem. 169, (1987) 63 FEBS 1987 Golgi galactosyltransferase contains serine-linked phosphate Ger J. STROUS, Peter van KERKHOF, Robert J. FALLON and Alan L. SCHWARTZ Laboratory of Cell Biology, University of Utrecht, Medical School, Utrecht (Received March 26/June 30, 1987) - EJB In HeLa and HepG2 cells the Golgi complex enzyme galactosyltransferase became phosphorylated following incubation with 32Pi. Analysis on sodium dodecyl sulphate/polyacrylamide gel electrophoresis revealed incorporation of 32P into the mature 54-kDa form. This phosphorylation was independent of protein synthesis. Serine was identified as the sole phosphorylated amino acid; no radioactive phosphate was detected on N-linked oligosaccharide. The phosphate-labelled galactosyltransferase has the same turnover as [3 5S]methionine-labelled polypeptides (tljz = 20 h). Soluble enzyme, released by the cells, contained very little phosphate relative to that which remained cell-associated. Charge heterogeneity arising from phosphorylation contributes in part to the heterodispersed appearance of the enzyme on two-dimensional gels, as the degree of radioactive phosphate differs among the different iso-enzymes. The Golgi enzyme galactosyltransferase from different cell lines and tissues has been identified as an integral membrane protein and shown to have a relative molecular mass of [l -31. It is involved in galactosylation of N-linked oligosaccharide, and is secreted as a soluble enzyme of a slightly lower molecular size [3]. The enzyme is synthesized as two precursor polypeptides in the rough endoplasmic reticulum, N- and 0-glycosylated, and transported to the trans-golgi cisternae [I, 21. One of the intriguing questions of intracellular transport is how organelles discriminate between resident proteins and proteins in transit. It is generally thought that hydrophobic membrane-anchoring segments and cytoplasmic extensions play a role in targeting and sorting of proteins to different destinations. Molecular biological studies have shown that alterations of the cytoplasmic tail have diverse effects on the sorting of membrane and secretory proteins (4, 51. Although the cytoplasmic extension of galactosyltransferase is probably short [2], it may well play an important role in the unique localization of the enzyme. This is indicated by the fact that galactosyltransferase is secreted by the cells upon loss of the anchoring peptide [I]. Phosphorylation of membrane proteins may have profound effects on protein function. For example, phosphorylation of the EGF receptor results in a markedly altered affinity for its ligands [6]. In addition, immobilization of a polypeptide within the plasma membrane may result from phorphorylation [7]. Thus, in order to examine this potential post-translational modification of galactosyltransferase and the possibility that phosphorylation may be involved in membrane targeting of a Golgi-resident enzyme, the phosphorylation of galactosyltranferase was evaluated in HeLa cells. Correspondence to G. J. Strous, Laboratorium voor Cellulaire Biologie, Nicolaas Beetsstraat 22, NL-3511 HG Utrecht, The Netherlands Abbreviation. EGF, epidermal growth factor. Enzymes. Galactosyltransferase (EC ); Endo F, endo-bacetylglucosaminidase F (EC ). EXPERIMENTAL PROCEDURES Radioactive labelling of'hela cells Nearly confluent HeLa cells grown on 60-mm culture dishes were pulse labelled with [35S]methionine (60-70 pci/ ml; Ci/mmol, Amersham International, Amersham), chased with unlabelled methionine, and lysed in 1 O/O Triton X-100, 0.1 mm phenylmethanesulfonyl fluoride in phosphate-buffered saline. For phosphate labelling, cells were rinsed in Hepes-buffered saline (20 mm Hepes ph 7.4, 140mM NaCl) and preincubated for 40 min at 37"C, in phosphate-free minimal essential medium. Then 32Pi (carrierfree; Amersham International, Amersham) was added at 0.5 mci/ml and the incubation was continued at 37 C. After an appropriate time, the cells were either chased in the presence of regular minimal essential medium or lysed. To stop the pulse (chase) labelling, the cells were washed twice at 4 C with phosphate-buffered saline and thereafter solubilized in phosphate-buffered saline containing 1 % Triton X-I 00, 1 mm phenylmethanesulfonyl fluoride and 1 mm EDTA. EDTA was added to stop divalent-ion-dependent phosphatase activity as well as cell-free phosphorylation. The solution was clarified by centrifugation and prepared for immunoprecipitation with protein-a - Sepharose beads (Pharmacia). Immunoprecipitation and SDSlpolyacrylamide gel electrophoresis Immunoprecipitations were performed as previously described [I]. The antiserum was raised against soluble human milk galactosyltransferase. The enzyme was purified by affinity chromatography on a-lactalbumin-sepharose and acetylglucosamine-sepharose columns [8] and the antiserum was shown to be monospecific for galactosyltransferase [ 11. Aliquots of the material soluble in Triton X-100 were immunoprecipitated with either normal rabbit IgG (control) or rabbit anti-(human galactosyltransferase). The immune precipitates were isolated following centrifugation and washing at 4 C and the antigen and antibody were released

2 308 from the Sepharose beads by boiling in SDS-PAGE sample buffer. Gel electrophoresis was performed in 10% polyacrylamide gels in the presence of sodium dodecyl sulphate (SDS- PAGE). After electrophoresis, the gels were fluorographed [9] and the fluorograms were scanned with a microdensitometer (E. C. Apparatus, FL) within the linear range of the film and the microdensitometer. Two-dimensional gel electrophoresis was performed according the procedure of O'Farrell [lo] and employed ph ampholines and 10% polyacrylamide gels. As a standard, bovine serum albumin (PI 4.7) was added to the sample prior to isoelectric focussing. For detection of albumin the gels were stained with Coomassie brillant blue. Endo-P-acetylglucosaminidase F (Endo F) digestion of immunoprecipitated proteins was performed according to Elder and Alexander [ll] as modified by Breitfeld et al. [12]. A B GT hientqication of' phosphorylated amino acids The immunoprecipitated galactosyltransferase was separated on SDS-PAGE, the gel was dried and autoradiographed, and the galactosyltransferase band was cut out of the gel. After electro-elution, phosphate-labelled galactosyltransferase was hydrolyzed for 2 h at 110 C in 6 M HCI. The hydrolysate was lyophilized and applied to a thin-layer cellulose plate. Two-dimensional electrophoresis was performed using acetic acid/formic acid/h20 (78: 25: 897; v/v) in the first and acetic acid/pyridine/h20 (50 : 5 : 945; v/v) in the second direction [13]. Standards of phosphoserine, phosphothreonine, phosphotyrosine were co-chromatographed with the radioactive sample and visualized by staining with ninhydrin. The thin-layer plate was then exposed to X-ray film as described above. RESULTS HeLa cells were grown in the presence of 32Pi for 4 h and then chased for 16 h. Galactosyltransferase was isolated by immunoprecipitation after solubilization of the cells in nonionic detergent and in the presence of EDTA to prevent cellfree phosphorylation and phosphatase activity. Only the mature 54-kDa enzyme was labelled; no label was detected in the 44-kDa and 47-kDa precursor polypeptides (Fig. 1 A, lane 1). Similar results were obtained with HepG2 cells (data not shown). Absence of phosphate incorporation in the precursor polypeptides was also observed when cells were labelled for 60 min. Shorter labelling periods using phosphate did not result in incorporation of phosphate into protein, reflecting the time required to label the intracellular ATP pool. The phosphate-labelled galactosyltransferase has exactly the same apparent molecular mass as [3sS]methionine-labelled galactosyltransferase. This experiment shows that phosphate incorporation probably takes place in the Golgi complex, as the conversion of the precursors into mature galactosyltransferase is a Golgi event [2]. To define this further, cells were pre-incubated for 40 min in the presence of 100 pg/ml cycloheximide to stop protein synthesis, and then incubated in the presence of radioactive methionine or phosphate. As is apparent from Fig. 1 B (compare lanes 3 and 4), the synthesis of galactosyltransferase is completely arrested, but there is still labelling of phosphate in the 'mature' galactosyltransferase polypeptide (lanes 1 and 2). Under these conditions, there was approximately 70% less 32P radioactivity in galactosyltransferase as compared to galactosyltransferase in the control cells. As it takes about 40 min for the enzyme to migrate from Fig. 1. (A) Immunoprecipitations of galuctosyltransferose from He La cells and (B) effect of cycloheximide on phosphorylution. (A) cells were labelled in medium containing 0.5 mci/ml carrier-free 32Pi (lanes I and 2) or 60 kci/ml ["S]methionine (lanes 3 and 4) for 4 h. After an additional 16-h chase, galactosyltransferase was quantitatively immunoprecipitated from both the media (lanes 2 and 3) and from the cells (lanes 1 and 4) and analyzed by SDS-PAGE on 10% gels and fluorography. Markers on the left denote (top to bottom) 69,46 and 30-kDa. (B) Cells were labelled for 4 h with 32Pi (lanes 1 and 2) or [35S]methionine (lanes 3 and 4) in the presence (lanes 1 and 3) and absence (lanes 2 and 4) of 100 pg/ml cycloheximide and then lysed. The drug was added 40 min before the beginning of the label period. GT, galactosyltransferase the rough endoplasmic reticulum to the Golgi complex, and a certain lag time is needed to label the ATP pool, this again indicates the Golgi complex as the place where phosphorylation takes place. It also indicates that shortly after arrest of translation there is a reduction in the number of galactosyltransferase molecules serving as substrate for phosphorylation, most probably a result of the transit of galactosyltransferase substrate pool. Alternatively, the presence of a short-lived kinase may be involved. In order to consider the rate of turnover of the incorporated phosphate, while galactosyltransferase remains in the Golgi complex, cells were pulse-labelled and chased for periods up to 2 days. The stability of the Pi moiety was compared with the turnover rate of the galactosyltransferase polypeptide, labelled in parallel in the presence of [3sS]- methionine. As seen in Fig. 2, both 32P-labelled and methionine-labelled galactosyltransferase is released from HeLa cells. Quantification of the radioactive bands indicates that, if the amount of methionine label present in the cells after 16 h of chase is taken as loooh, total amounts of galactosyltransferase in media and cells remain constant at each time point. The rate of 32P-labelled galactosyltransferase release from the cells is comparable to that of the methionine label, but the amount of 32P released into the media is less than 50% of that to be expected on the basis of the release of methionine label. The main conclusion is that the P-galactosyltransferase linkage is stable during its stay in the Golgi compartment. However, it is not clear whether loss of 32P occurs during the process of release from the cells or whether

3 35s cells 32P medium cells medium , chase (h) Fig. 2. Time course of disappearance ofgalactosyltran.ferase from the cells. Cells were labelled in the presence of 32Pi or [35S]methionine for 4 h then chased. At the indicated times, the cells were washed and lysed in detergent. The amount of label in [35S]methionine-labelled galactosyltransferase is higher at 16 h than at 0 h of chase; this is probably caused by the presence of radioactive methionine during the start of the chase. If cells are chased for 1 h, radioactivity in galactosyltransferase is about the same as at 16 h of chase. Galactosyltransferase was immunoprecipitated and the immunoprecipitates were analyzed by SDS-PAGE EndoF.- C 0 u) E.- U GT First dimension Fig. 3. Autoradiogram of phosphorylated amino acids separated by high-voltage electrophoresis. After labelling of cells in the presence of 32Pi, galactosyltransferase was isolated by immunoprecipitation and purified by SDS-PAGE. The 54-kDa band was cut out of the gel, electro-eluted, and hydrolyzed for 2 h at 110 C in 6 M HCI as described in Methods. The dotted lines represent the positions of the ninhydrin-stained standards superimposed on the autoradiograph it is due phosphatase activity in the media. An exact comparison of the 32P- and the methionine-labelled galactosyltransferase is impossible, as the specific activity of labelled intracellular phosphate did not decrease at a rapid rate, when excess Pi was added to the medium, as compared to intracellular methionine during the chase period. The presence of galactosyltransferase in the Golgi complex is transient with an average half-life of about 19 h (i.e. the time for 50% of the whole cell pool to be secreted) [l]. In order to test whether the phosphate moiety is retained on the galactosyltransferase polypeptide, cells were pulse-labelled for 4 h in the presence of 32Pi and then chased for 16 h. As seen in Fig. 1 A, almost no 32P-labelled galactosyltransferase was detectable in the medium compared to the intracellularly present enzyme, (lane 2) while under the same conditions approximately 30% of cellular [35S]methionine-labelled galactosyltransferase is secreted in the medium as a 52-kDa polypeptide (lane 3). Two possible P-galactosyltransferase linkages were studied. The most likely link is to the polypeptide chain via threonine, serine or tyrosine. The other possibility is a Fig. 4. Endo F digestion of galactosyltransferase. Cells were labelled in the presence of [35S]methionine(lanes 1 and 2) or 32Pi (lanes 3 and 4) for 4 h. Galactosyltransferase was immunoprecipitated using protein-a - Sepharose and the washed beads carrying the antigenantibody complexes were divided into two equal aliquots. One aliquot was incubated with Endo F and the other without the enzyme. The samples were analyzed on SDS-PAGE. GT = galactosyltransferase carbohydrate-p bond. Characterization of 32P-containing amino acids following immunoprecipitation, purification by SDS-PAGE, hydrolysis and two-dimensional electrophoresis demonstrated that essentially all radioactivity co-migrated with phospho-serine (Fig. 3). In order to determine whether N-linked oligosaccharides also contain P, 32P-labelled galactosyltransferase was subjected to digestion with Endo F. As seen in Fig. 4, Endo F digestion caused a decrease in apparent M, of about 2000, but there was no reduction of radioactive label. As a control, [35S]methionine-labelled galactosyltransferase was treated with Endo F, resulting in exactly the same shift in molecular mass. Thus, galactosyltransferase does not contain P bound to N-linked oligosaccharide. Golgi-derived and soluble galactosyltransferase from HeLa cells and from human milk contains a series of differently charged molecules as determined by isoelectric

4 310 t ' 35s 32 P + Fig. 5. Two-dimensional gel electrophoresis of [35S]rnethionine- and 32P-labelled galactosyltransferase. Cells were labelled for 4 h and galactosyltransferase was isolated by immunoprecipitation. The ph gradient was between 3.5 and 10; the exact position of bovine serum albumin (PI 4.7) is indicated focussing with isoelectric points between 4 and 7 [3, 8, 141. Treatment with neuraminidase results in increase of the isoelectric points of the iso-protein, but the heterogeneity persists. Part of the heterogeneity is due to sulfation [3]. To establish whether phosphorylation also contributes to this heterogeneity we labelled the cells in the presence of 32Pi for 4 h and isolated galactosyltransferase by immunoprecipitation. Then galactosyltransferase was analyzed by two-dimensional gel electrophoresis. The occurrence of the different isoproteins at the different isoelectric points was determined by metabolically labelling galactosyltransferase in the presence of [35S]methionine. As shown in Fig. 5, 32P-labelled galactosyltransferase exhibits a charge distribution slightly different from the pattern of 35S-labelled galactosyltransferase. Two iso-proteins in particular are heavily phosphorylated and contain severalfold the amount of phosphate residues as compared to the other species. Thus, the amount of phosphate residues per polypeptide chain is not the same for each isoenzyme. Consequently, phosphorylation contributes to the negative charge of galactosyltransferase; in addition, it causes part of the charge heterogeneity observed after two-dimensional gel electrophoresis. DISCUSSION This paper describes for the first time the phosphorylation of a resident Golgi membrane protein. The mature 54-kDa galactosyltransferase molecule is phosphorylated in both HeLa and HepG2 cells. No label was detected in the two precursor polypeptides. As the phosphorylation does not require protein synthesis, it is likely that the majority of the phosphorylation occurs in the cis- or trans-golgi complex. This is consistent with the finding of Schwartz, who reported phosphorylation of the asialoglyco-protein receptor in HepG2 cells in the absence of protein synthesis [15]. Phosphorylation of this receptor starts at the precursor level, but most labelling occurs at the mature polypeptide stage. Galactosyltransferase is phosphorylated soon after biosynthesis and retains its phosphate moiety during its residency in the Golgi complex. The experiment with cycloheximide also indicates that phosphorylation continues until completion of post-translational modifications. The exact localization of the phosphate group has not been determined. Our results indicate that only serine is phosphorylated. No phosphorylation of N-linked oligosaccharide was detected. On the other hand, one cannot exclude the possibility that 0-linked oligosaccharides contain phosphate residues. This could be inferred from the fact that, although all iso-proteins contain phosphate, some contain severalfold more than others. There are numerous reports of serine phosphorylation of membrane proteins including cell-surface receptors : the EGF receptor [16], the insulin receptor [17], the transferrin receptor [18], the asialoglycoprotein receptor [15], the IgE receptor in RBL cells [19]. This phosphorylation reaction appears to be independent of ligand binding. In addition, phosphorylation of all these receptors takes place at the cytoplasmic face. Although galactosyltransferase is an intrinsic membrane protein by classic criteria [20], both immunocytochemical and biochemical experiments show that, although there is a cytoplasmic extension, it is rather short [2]. Krebs and Breavo [21] have shown that phosphorylation is a cytoplasmic event. If galactosyltransferase is phosphorylated at the cytoplasmic face, this phosphate-linked peptide cannot be present in the soluble form of the enzyme. Indeed, galactosyltransferase loses its phosphate almost completely upon release into the medium; only a very small amount of label remains associated with the secreted galactosyltransferase as is apparent from Fig. 1A. However, as the amount of radioactivity was minimal, it was not possible to analyse the nature of the phosphate linkage. The stability of the phosphate-serine linkage is remarkable. Protein phosphorylation-dephosphorylation plays a major role in hormonal control of enzyme activity (for a recent review by Cohen, see [22]). In addition, many plasma membrane receptors contain serine-phosphate linkages. In the case of the transferrin receptor the rate of phosphate group turnover (f1/2 = 30 min) appears inconsistent with a direct role in endocytosis and recycling (til2 = 12 min) [23]. However, the rate of turnover of the radiolabelled cellular ATP pool may overestimate the turnover rate of the labelled polypeptide. Nonetheless the phosphorylation-dephosphorylation of the phosphate residue may well be involved in directing receptors from the plasma membrane into CURL tubules and back to the cell surface [15, The role of a cytoplasmic phosphate moiety in galactosyltransferase is unclear. Considering the stability of the phosphate linkage and the fact that all iso-proteins seem to be phosphorylated, this group could potentially contribute to the specific localization of the enzyme in the Golgi complex. Recently, the deduced amino acid sequence of bovine galactosyltransferase has been deduced from a cdna clone [26]. However, the available clones do not extend to either the membrane-spanning or cytoplasmic portion of the enzyme. Cytoplasmic phosphorylation may potentially induce a conformational change in the membrane-spanning region or extracytoplasmic domain of a polypeptide as a result of alteration of distribution in the negative charge. Rees et al. [7] have reported that a high-affinity class of EGF receptors becomes immobilized in the plane of the plasma membrane coincident with receptor phosphorylation. Thus it is tempting to speculate that phosphorylation of serine residues on the cytoplasmic face might be responsible for locking proteins in a certain membrane environment. Examination of this issue will only be possible by direct comparison of the molecular structures of several Golgi glycosyltransferases.

5 31 1 We are grateful to Tom van Rijn and Maurits Niekerk for help with the photopraphs. The investigations were supported in part by a grant of the Foundation for Medical Research MEDIGON, The Netherlands ( ), by NATO (818/83), and by the National Science Foundation and National Institutes of Health. RJF is a Clinical Investigator of the National Heart, Lung and Blood Institute. ALS is an Established Investigator of the American Heart Association. REFERENCES 1. Strous, G. J. & Berger, E. G. (1982) J. Biol. Chem. 257, Strous, G. J., Van Kerkhof, P., Willemsen, R., Geuze, H. J. & Berger, E. G. (1983) J. Cell Biol. 97, Strous, G. J. (1986) CRC Crit. Rev. Biochem. 21, Rose, J. K. & Bergmann, J. E. (1983) CeN34, Guan, J., Machamer, C. E. & Rose, J. K. (1985) Cell 42, Fearn, J. C. & King, A. C. (1985) Cell 40, Rees, A. R., Gregoriou, M., Johnson, P. & Garland, P. B. (1984) EMBO J. 3, Gerber, A. C., Kozdrowski, I., Wyss, S. R. & Berger, E. G. (1979) Eur. J. Biochem. 93, Bonner, W. M. & Laskey, R. A. (1974) Eur. J. Biochem. 46, O Farrell, P. H. (1975) J. Biol. Chem. 250, Elder, J. H. & Alexander, S. (1982) Proc. Nut1 Acud. Sci. USA 79, Breitfeld, P. B., Rup, D. & Schwartz, A. L. (1984) J. Biol. Chem. 259, Hunter, T. & Sefton, B. M. (1980) Proc. Nail Acad. Sci. USA 77, Strous, G. J., Van Kerkhof, P., Willemsen R., Slot, J. W. & Geuze, H. J. (1985) Eur. J. Cell Biol. 36, Schwartz, A. L. (1984) Biochem. J. 223, Carlin, C. R., Phillips, P. D., Knowless, B. B. & Cristofalo, V. J. (1983) Nature (Lond.) 306, Kasuga, M., Zick, Y., Blith, D. L., Karlsson, F. A,, Haring, H. U. & Kahn, C. R. (1982) J. Biol. Chem. 257, Schneider, C., Sutherland, R., Newman, R. &Greaves, M. (1982) J. Biol. Chem. 257, Fewtrell, C., Goetze, A. & Metzger, H. (1982) Biochemistry 21, Helenius, A. & Simons, K. (1972) J. Biol. Chem. 247, Krebs, E. G. & Breavo, J. A. (1979) Annu. Rev. Biochem. 49, Cohen, P. (1985) Eur. J. Biochem. 151, Johnstone, R. M., Adam, M., Turbide, C. & Larrick, J. (1984) Can. J. Biochem. Cell Biol. 62, Geuze, H. J., Slot, J. W., Strous, G. J., Lodish, H. F. & Schwartz, A. L. (1983) Cell32, Harding, C., Heuser, J. & Stahl, P. (1983) J. Cell Biol. 97, Shaper, N. L., Shaper, J. H., Meuth, J. L., Fox, J. L., Chang, H., Kirsch, I. R. & Hollis, G. F. (1986) Proc. Nut1 Acad. Sci. USA 83,