Process for sialylating glycoproteins Field of the invention The present invention relates to genetic engineering, and provides a novel and efficient process for sialylating glycoproteins. In the process said proteins are incubated with recombinant yeast cells expressing a sialyltransferase enzyme as a fusion protein with Hsp150h carrier in the porous cell wall.

Background of the invention Most pharmaceutical proteins desired to be manufactured in cultured cells by genetic engineering are secretory proteins, which carry sialylated N-glycans in their authentic form. The host cell, as well as cell culture conditions, have profound effects on glycosylation of the protein product (Andersen and Goochee, 1994; James et al., 1995). The host cell may lack sialyltransferases, like yeast, fungal and insect cells, or shed sialidase activity which desialylates the protein product after secretion (Gramer et al., 1995). The sialylation of recombinant mammalian proteins performed by mammalian cells is often incomplete, or not authentic (Curling et al., 1990; Goochee et aL, 1991; Patel et al., 1992; Maiorella et al., 1993). For instance, cultured CHO and BHK-21 cells lack a functional a2,6-sialyltransferase and add only a2,3-sialic acid to glycoproteins, whereas human and mouse cells decorate glycoproteins with both residues (Lee et al., 1989). Lack of terminal sialic acid residues may or may not affect the biological activity of the protein, but the clearance rate from the blood stream is dramatically increased. Proteins with exposed terminal galactose residues are removed from the circulation, severely compromising the effect of the pharmaceutical proteins (Fukuda et al., 1989; Sareneva et al., 1993; Szkudlinski et al., 1993).

Authentic sialyltransferases are type 11 transmembrane glycoproteins anchored to the Golgi membrane. The catalytic ectodomain faces the Golgi lumen and a short N-terminal segment the cytoplasm (Field and Wainwright, 1995) (see

Fig. 1). We have constructed a Saccharomyces cerevisiae strain, which ex- presses the ectodomain of rat liver a2,3-sialyltransferase (ST3Ne) as a fusion <BR> <BR> <BR> protein (Mattila et al., 1996). Glycosyltransferases like many mammalian sec- retory proteins are retained in the endoplasmic reticulum (ER) in yeast (Krez- dorn et al., 1994). Thus we linked ST3Ne to the C-terminus of the Hsp1505 polypeptide, which consists of the 321 N-terminal amino acids of the natural secretory glycoprotein Hsp150 of yeast (Russo et al., 1992). The Hspl 50A polypeptide has been shown to promote proper folding and secretion compe- tence of several foreign proteins in S. cerevisiae (Simonen et aL, 1994 and 1996; Makarow, US patent No. 5,677,172). The 162 N-terminal amino acids of the Hsp150 protein also promote folding of foreign proteins and confer them secretion competence. Thus the Hsp150A carrier is defined as an N-terminal fragment of the translation product of the HSP150 gene, having 162 to 321 amino acid residues. The Hsp150A-ST3Ne fusion protein was efficiently sec- reted in enzymatically active form, but adhered tightly to the yeast cell wall (see Fig. 1). Incubation of the recombinant yeast cells with N-acetyllactos- amine and CMP-Neu5Ac resulted in a2,3-sialylation of the disaccharide, demonstrating that the substrates as well as the product penetrated the cell wall efficiently (Mattila et al., 1996). Previously, heterologous proteins have been immobilized in the yeast cell wall e. g. by a fragment of the yeast a-ag- glutinin protein (Schreuder et al., 1993), whereas in here, the fusion protein as such is bound to the cell wall.

Summary of the invention In this invention we show that, surprisingly, also whole glycoproteins can pe- netrate easily the yeast cell wall and reach the ST3N activity. Several desialyl- ated mammalian glycoproteins were shown to be sialylated by incubating them with the whole living recombinant yeast cells, which express a sialyltransferase enzyme as a fusion protein with Hsp150h carrier in the porous cell wall.

In this work whole Saccharomyces cerevisiae cells expressing the catalytic ectodomain of rat liver a2,3-sialyltransferase (ST3Ne) in the porous cell wall were used to complement sialylation of glycoproteins with native conforma- tions. Incubation of the yeast cells with desialylated fetuin, prothrombin and transferrin resulted in transfer of Neu5Ac from CMP-Neu5Ac to the terminal galactose residues of the protein-bound N-glycans. The Ka values of the yeast cell wall-borne enzyme for CMP-Neu5Ac, N-acetyllactosamine and lacto-N-tetraose were similar to those of recombinant ST3Ne produced in insect cells and of authentic rat liver ST3N. The recombinant yeast strain provides an inexpensive and self-perpetuating source of ST3N activity for sialylation of glycoproteins.

Detailed description of the invention In the following the process of the invention is illustrated by the experimental procedure of the sialylation of bovine plasma fetuin, bovine prothrombin and human transferrin by incubating said proteins with CMP-Neu5Ac and whole living recombinant S. cerevisiae cells expressing the catalytic ectodomain of rat liver a2,3-sialyltransferase (ST3Ne).

Brief description of the drawings Figure 1. Authentic sialyltransferases are located in the Golgi membrane (GM). The catalytic ectodomain faces the lumen and a short N-terminal se- quence the cytoplasm (CP). We expressed the ST3N catalytic ectodomain in yeast cells as a fusion protein, its N-terminus linked to the C-terminus of the Hsp150A polypeptide (wavy line). In yeast cells the fusion protein is transport- ed to the exterior of the yeast plasma membrane (PM), but remains bound to the cent watt (CW).

Figure 2. Transfer of [14C] Neu5Ac to asialofetuin and asialomucin by whole yeast cells expressing Hsp150A-ST3Ne in the cell wall. (A) Duplicate samples of yeast strain H626 were incubated for 1-4 hours with 100000 cpm of CMP-

[14C] Neu5Ac and asialofetuin (closed circles), with completely sialylated fetuin (open circles), or with asialomucin (diamonds). Parental yeast cells (H23) were incubated with asialofetuin (open squares). All incubations were in the presen- ce of 10 mM NaN3. (B) Asialofetuin was incubated as above with NaN3 (closed circles), or with 40 mM DTT (closed squares), or with 4% glucose (open cir- cles), or with 4% glucose and 40 mM DTT (open squares). TCA-precipitated radioactivity is plotted against incubation time.

Figure 3. MALDI-TOF mass Spectra of the N-glycans. Oiigosaccharides liberated from asialofetuin (A), or from asialofetuin incubated for 4 h (B) or 8 h (C) with H626 cells. In (A), signals were assigned to nonsialylated triantennary glycans (m/z 2008, [M+H] + ; m/z 2030, [M+Na) +) and nonsialylated biantennary glycans (m/z 1665, [M+Na] +). In (B) and (C), monosialylated biantennary (m/z 1956, [M+Na] + ; m/z 1978, [M-H+2Na] +); monosialylated triantennary (m/z 2321, [M+Na] + ; m/z 2343, [M-H+2Na] +); disialylated triantennary (m/z 2612, [M+Na] +; m/z 2634, [M-H+2Na] + ; m/z 2656, [M-H+3Na] +) and trisialylated triantennary glycans (m/z 2903, [M+Na] + ; m/z 2925, [M-H+2Na] +) were detec- ted. The asterisk (*) designates matrix adducts typical to this mode of analysis (Nyman et al., 1998).

Materials and methods Strains and media. S. cerevisiae strains H23 (Mata his3-11,15 leu2-3, 112 trp 1-1 ade2-1 canl-100 hsp150 : : URAæ and H626 (Mata his3-11, t5 leu2- 3,172 ade2-1 cany-100 hsp750 : : URA3 TRP1 : : HSP150A-ST3Ne) (Mattila et al., 1996) were grown at 24°C in YPD medium consisting of 1% yeast extract (Oxoid Ltd., UK), 2% bacto peptone (Difco, Detroit, MI) and 2% glucose (BDH Pharmaceuticals Ltd., UK), or in synthetic complete (SC) medium (Simonen et al., 1994) lacking tryptophane for selection.

Sialyltransferase assays. Duplicate samples of 5 x 107 whole living yeast cells were incubated in 70, lit of 50 mM imidazole buffer, pH 7 (glycoproteins), or of 50 mM Tris-maleate buffer, pH 6.7 (oligosaccharides), either with 0.2

nmol (100000 cpm) of CMP- ['4C] Neu5Ac (294 mCi/mmol, Amersham Interna- tional, Buckinghamshire, UK) or with saturating concentrations of unlabelled CMP-Neu5Ac (Sigma, St. Louis, MO), and varying amounts of the acceptor substrates, in a shaker at 24°C. For determination of intracellular plus extra- <BR> <BR> <BR> cellular ST3N activity, 5 x107 cells were lysed mechanically with glass beads (Mattila et al., 1996) prior to the sialyltransferase assay. Recombinant ST3Ne (0.34 mU) produced in Sf9 cells (Calbiochem-Novabiochem, La Jolla, CA) was incubated at 37°C in 20/il of 50 mM MOPS buffer, pH 7.5, containing 1% BSA, and 0.06 nmol of CMP- [C] Neu5Ac (28500 cpm) or saturating concen- trations of unlabelled CMP-Neu5Ac, and different amounts of the acceptor substrate. After sialylation, the proteins were precipitated with 20% TCA for 30 min on ice and collecte on filters for scintillation counting. Oligosaccharides were applied on columns of Dowex AG 1 (acetate form, BioRad, Hercules, CA) and Dowex 50 (H+form, Fluka, Switzerland). Neutral oligosaccharides were eluted with 4 ml of water and sialyloligosaccharides with 20 mi of 0.5 M acetic acid (Renkonen et al., 1991), and subjected to scintillation counting.

Fetuin, asialofetuin, asialoprothrombin, asialomucin, lacto-N-tetraose and N- acetyllactosamine were from Sigma, St. Louis, MO, transferrin from the Fin- nish Red Cross, Helsinki, Finland, and prothrombin from ICN, Aurora, Ohio.

Desialylation of prothrombin and transferrin was performed in 0.025 M H2SO4 for 1 h at 80°C (Spiro, 1960), whereafter the preparations were neutralized and ultrafiltrated using Centricon devices (cut off 30 kD) to remove the re- leased Neu5Ac residues and the ions.

Isolation of N-glycans. After sialylation, the proteins were desalted on rever- sed-phase HPLC and the fractions dried in a vacuum centrifuge. The samples were dissolved in 10 jul of 20 mM sodium phosphate, pH 7.2, containing 1% SDS, and boiled for 3 minutes. After cooling, 75 ul of 20 mM sodium phos- phate, pH 7.2,10 ul of 10% OGP and 1 U of N-glycosidase F (Boehringer Mannheim GmbH, Germany) were added and the mixtures were incubated at 37°C for 3 days. Proteins and detergents were removed using a BondEiute C18-column (Analytichem International, CA, USA). The samptes were diluted to 300 ul with water prior to loading onto the column, and the glycans were

eluted with 1.5 mi of water. Buffer salts were removed by drop-dialysis against water on VSWP 02500 membranes (Millipore, Bedford, MA, USA) (Börnsen et al., 1995).

Mass spectrometry and CD spectroscopy. Matrix-assisted laser desorption/ ionization time-of-flight (MALDI-TOF) mass spectrometry was performed on a Biflex instrument (Bruker-Franzen Analytik, Bremen, Germany) equipped with a nitrogen laser operating at 337 nm. N-glycans were analyzed in the linear positive ion delayed extraction mode by using 2,4,6-trihydroxyaceto- phenone (THAP, 3 mg/ml in acetonitrile/20 mM aqueous diammonium citrate, 1: 1) as matrix. Samples were prepared by mixing 0.5 pi of drop-dialysed oli- gosaccharide solution (see isolation of N-glycans) and 0.5 PI of THAP matrix on the target plate, and immediately dried under vacuum to produce a thin homogenous matrix (Papac et al., 1996). This mode of analysis causes little loss of Neu5Ac residues, and may be used for rapid quantitation of both sialy- lated and neutral oligosaccharides in a mixture (Nyman et aL, 1998). The mass spectra were externally calibrated with Dextran 5000 (Fluka BioChemica, Switzerland). CD spectroscopy was performed using a JASCO J-720 spectro- polarimeter. The spectra were recorded in the far UV region (190-259 nm).

Each spectrum was the mean of five scans obtained with a time constant of 1 s and a speed of 20 nm/min.

Kinetic studies. The K, determinations were according to Lineweaver-Burke. <BR> <BR> <BR> <P>To determine the activity of Hsp150A-ST3Ne, H626 cells (5 x 107) were incu- bated with 0.4 mM lacto-N-tetraose and 10 mM CMP-Neu5Ac. The superna- tant was subjected to ion exchange chromatography over the Dowex columns and the eluates were concentrated and chromatographed over a Superdee Peptide PC 3.2/30 column (Pharmacia, Sweden) (Nyman et al., 1998). The oligosaccharides were quantitated against external lacto-N-tetraose (Sigma) and NeuNAc (Sigma).

Results Transfer of ['4C] Neu5Ac to N-glycans of asialoglycoproteins.

Authentic bovine plasma fetuin (Mw 48 kD) carries three N-glycans (Spiro, 1973; Rice et al., 1990). To study whether asialofetuin could penetrate the yeast cell wall and be sialylated by the ST3N activity, fetuin was desialylated by mild acid hydrolysis, and incubated with CMp_ [14 C] Neu5Ac and the recom- binant yeast cells. Sodium azide was included in the reaction mixture to block intracellular protein transport to the cell wall or medium. Samples were remo- ved at different times, the cells were pelleted and the supernatants subjected to precipitation with trichloro acetic acid (TCA). Scintillation counting showed that the precipitated radioactivity increased with time (Fig. 2A, closed circles).

The precipitated radioactivity was confirmed by mass spectrometry to be fetuin with covalently linked [14C] Neu5Ac residues (see below). Very little radioactivity remained cell-associated, showing that the sialylated protein and CMP-['4C]- Neu5Ac did not adhere to the cells. When completely sialylated fetuin was incubated with the yeast cells (Fig. 2A, open squares), or when asialofetuin was incubated with the parental yeast cells lacking the HSP150A-ST3Ne gene (Fig. 2A, open circles), no radioactivity could be precipitated. Authentic fetuin has, in addition to N-glycans, two O-glycans which have two Neu5Ac resi- dues each (Spiro, 1973). Sialylation by Hsp150A-ST3Ne was specific for N- glycans, since no [14qNeu5Ac could be transferred to asialomucin, which carries only O-glycans (Fig. 2A, diamonds).

The sialylation of asialofetuin was enhanced when the incubation with the recombinant yeast cells was carried out in the presence of dithiotreitol (DTT) (Fig. 2B, closed squares). DTT apparently increases the porosity of the cell wall of S. cerevisiae by reducing disulphide bridges, thus increased efficiency of sialylation was likely to result from enhanced penetration of asialofetuin into the yeast cell wall. Next we allowed Hsp150A-ST3Ne to be synthesized and transported to the cell wall during the assay by omitting sodium azide and adding glucose to the reaction mixture. This resulted in doubling of [14C]-sialy- lation of asialofetuin (Fig. 2B, open circles). When the recombinant yeast cells

were incubated in the absence of sodium azide with both DTT and glucose, ['4C]-sialylation of asialofetuin was further increased (Fig. 2B, open squares).

Also larger proteins like bovine prothrombin (Mw 73.6 kD) and human transfer- rin (Mw 79.6 kD) could be [t4C]-sialylated, after removal of their Neu5Ac resi- dues, by the recombinant yeast cells (Table 1). No radioactivity could be TCA- precipitated when the completely sialylated forms of these proteins were used as the acceptor substrate (Table 1).

Efficiency of sialylation. Next we quantitated the degree of sialylation of the asialoglycoproteins. Two of the three N-glycans of fetuin are triantennary and one is biantennary, and thus asialofetuin has 8 terminal galactose residues <BR> <BR> <BR> <BR> (Spiro et al., 1973; Rice et al., 1990). Asialofetuin and saturating concentra- tions of unlabeled CMP-Neu5Ac were incubated for 4 h in the presence of NaN3 with the recombinant yeast cells expressing Hsp150A-ST3Ne. The cells were removed by pelleting, and the N-glycans were released by N-glycosida- se F digestion and analysed by MALDI-TOF mass spectrometry (Fig. 3). In 4 hours, 31.7% of the terminal galactose residues were sialylated (Fig. 2B).

When a parallel cell suspension was pelleted after the 4 h incubation, and the supernatant incubated for another 4 h with a fresh batch of yeast cells and CMP-Neu5Ac (Fig. 3C), 55.3% of the galactose residues were sialylated, demonstrating that both completely and incompletely desialylated protein bound N-glycans could be sialylated. When the incubation was prolonge to 16 h, Neu5Ac was bound to 61.3% of the terminal galactose residues (not shown). A similar overnight incubation of asialotransferrin resulted in sialylation of 41.5% of the exposed galactose residues. To ensure that desialylation of fetuin, prothrombin and transferrin by mild acid hydrolysis had not denatured the proteins, we subjected them to circular dichroism spectroscopy. The spectra of the sialylated and desialylated preparations were superimposable (data not shown).

Kinetic properties of Hsp150A-ST3Ne. Finally we compared the kinetic pro- perties of yeast cell wall Hsp150A-ST3Ne and recombinant ST3Ne produced in insect cells. The K", values for asialofetuin, lacto-N-tetraose and N-acetyllac-

tosamine were similar (Table 2A). The relative ratios VmaX/K"demonstrated that both enzyme preparations preferred lacto-N-tetraose (type 1: Gales1- 3GIcNAc) over N-acetyllactosamine and asialofetuin (type 2: Galß1-4GlcNAc) (Table 2B). The K, 1, value of Hsp150#-ST3Ne for CMP-Neu5Ac was similar to those reported by others for recombinant ST3Ne from insect cells and authen- tic isolated rat liver ST3N (Table 2B). A one liter overnight culture containing 18 g (dry weight) of yeast cells contained 117 mU of ST3N activity (lacto-N- tetraose as acceptor).

According to the data presented here, whole living Saccharomyces cerevisiae cells expressing the catalytic ectodomain of rat ST3N as a Hsp150A-ST3Ne fusion protein in the porous cell wall provide a convenient and inexpensive source of the transferase. The transferase needs not to be purified for use, and neither has the sialylated protein product to be separated from the trans- ferase. Purified transferases have a limited lifetime, whereafter activity is lost, whereas our recombinant yeast cells provide a self-perpetuating source of the enzyme.

Table 1. Transfer of ['4C] Neu5Ac to prothrombin and transferrin. The sialyl- ation assay was performed in the presence of NaN3 as described in the Le- gend of Figure 1. [14C]Neu5Ac(cpm)TimeProtein-bound (h) Prothrombin Asialoprothrombin Transferrin Asialotransferrin 2 292 3856 428 994 4 657 9016 341 1885

Table 2. Kinetic parameters of various ST3N preparations. The Km values and the relative Vmax/Km values are given for the indicated acceptor substrates.

(The highest value for V,,,/K, is 100). In (A) CMP-Neu5Ac was available in saturating concentrations, and the protein or sugar substrate in varying con- centrations. In (B), lacto-N-tetraose was available in saturating concentra- tions, and varying concentrations of CMP-Neu5Ac up to 10 mM were mixed with a constant amount of CMP- [C] Neu5Ac. Enzyme preparations: Hsp150A-ST3Ne in the wall of yeast cells; recombinant ST3Ne from insect cells (rST3Ne) ; authentic ST3N from rat liver.

(A) Asialofetuin Lacto-N-tetraose N-acetyllactosamine Km(µM)Vmax/KmKm(µM)Vmax/KmKm(µM) Hsp150A-ST3N 42.7 100 717.7 5.9 rST3Ne 34.4 10. 9 43.0 100 1014 11.3 (B) Km tM) for Reference CMP-Neu5Ac Hspl 50A-ST3Ne 55.0 this work Williamsetal.,1995rST3Ne74.1 ST3N 57.3 Gross et al., 1989

References Andersen, D. C. and Goochee, C. F. (1994). Curr. Opin. Biol. 5: 546-549.

Börnsen, K. O., Mohr, M. D. and Widmer, H. M. (1995). Rapid Commun. Mass Spectrom. 9: 1031-1034.

Curling, E. M. A., Hayter, P. M., Baines, A. J., Bull, A. T., Gull, K., Strange, P. G. and Jenkins, N. (1990). Biochem. J. 272: 333-337.

Field, M. C. and Wainwright, L. J. (1995). Glycobiology 5: 463-472.

Fukuda, M. N., Sasaki, H., Lopez, L. and Fukuda, M. (1989) Blood 73: 84-89.

Goochee, C. F., Gramer, M. J., Andersen, D. C., Bahr, J. B. and Rasmussen, J. R. (1991) Bio/Technology 9: 1348-1355.

Gramer, M. J., Goochee, C. F., Chock, V. Y., Brousseau, D. T. and Sliwkowski, M. B. (1995) Bio/Technology 13: 692-698.

Gross, H. J., Krause, J. M., Paulson, J. C., Feeney, R. E. and Brossmer, R.

(1989). Biochemistry 28: 7386-7392.

James, D. C., Freedman, R. B., Hoare, M., Ogonah, O. W., Rooney, B. C., La- rionov, O. A., Dobrovolsky, V. N., Lagutin, O. V. and Jenkins, N. (1995). Bio/ Technology 13: 592-596.

Krezdorn, C. H., Kleene, R. B., Watzele, M., Ivanov, S. X., Hokke, C. H., Kamer- ling, J. P. and Berger, E. (1994). Eur. J. Biochem. 220: 809-817.

Lee, E. U., Roth, J. and Paulson, J. C. (1989). J. Biol. Chem. 264: 13848- 13855.

Maiorella, B. L., Winkelhake, J., Young, H., Moyer, B., Bauer, R., Hora, M., Andya, J., Thomson, J., Patel, T. and Parekh, R. (1993). Bio/Technology 11: 387-392.

Makarow, US patent No. 5,677,172.

Mattila, P., Joutsjoki, V., Kaitera, E., Majuri, M.-L., Niittymaki, J., Saris, N., Maaheimo, H., Renkonen, O., Renkonen, R. and Makarow, M. (1996). Glyco- biology 6: 8: 851-859.

Nyman, T. A., Kalkkinen, N., Tolu, H. and Helin, J. (1998). Eur. J. Biochem.

253: 485-493.

Papac, D. I., Wong, A. and Jones, A. J. S. (1996). Anal. Chem. 68: 3215-3223.

Patel, T. P., Parekh, R. B., Moellering, B. J. and Prior, C. P. (1992). Biochem. J.

285: 839-845.

Renkonen, O., Niemela, R., Leppanen, A., Maaheimo, H., Seppo, A., Penttilä, L. and Vikman, A. (1991). Glycoconjugate J. 8: 368-375.

Rice, K. G., Rao, N. B. N. and Lee, Y. C. (1990). Anal. Biochem. 184: 249-258.

Russo, P., Kalkkinen, N., Sareneva, H., Paakkola, J. and Makarow, M. (1992).

Proc. Natl. Acad. Sci. USA 89: 3671-3675.

Sareneva, T., Cantell, K., Pyhala, L., Pirhonen, J. and Julkunen, I. (1993). J.

Interferon Res. 13: 267-269.

Schreuder, M. P., Brekelmans, S., van den Ende, H. and Klis, F. M. (1993).

YEAST 9: 399-409.

Simonen, M., Jamsa, E. and Makarow, M. (1994). J. Biol. Chem. 269: 19: 13887-13892.

Simonen, M., Vihinen, H., Jamsa, E., Arumae, U., Kalkkinen, N. and Makarow, M. (1996). YEAST 12: 457-466.

Spiro, R. G. (1960). J. Biol. Chem. 235: 10: 2860-2869.

Spiro, R. G. (1973). Adv. Protein Chem. 27: 349-467.

Szkudlinski, M. W., Thotakura N. R., Bucci, I., Joshi, L. R., Tsai, A., East-Pal- mer, J., Shiloah, J. and Weintraub, B. D. (1993). Endocrinology 133: 1490- 1503.

Williams, M. A., Kitagawa, H., Datta, A. K., Paulson, J. C. and Jamieson, J. C.

(1995). Glycoconjugate J. 12: 755-761.