NZ266045A - Saccharomyces cerevisiae strain defective in o-mannosylation - Google Patents

Description

New Zealand No. 266045 International No. PCT/SE94/00421 TO BE ENTERED AFTER ACCEPTANCE AND PUBLICATION Priority dates: 07.05.1993; Complete Specification Filed: 06.05.1994 Classification:^) C12N1/16; C12N15/81 Publication date: 24 June 1997 Journal No.: 1417 new zealand patents act 1953 COMPLETE SPECIFICATION Title of Invention: Yeast strain and methods for expressing heterologous proteins in yeast Name, address and nationality of applicant(s) as in international application form: PHARMACIA AB, a Swedish company of S-171 97 Stockholm, Sweden New Zealand No. International No. 266045 PC7/SE94/00421 NEW ZEALAND PATENTS ACT 1 953 complete specification Title of Invention: Yeast strain and methods for expressing heterologous proteins in yeast Name, address and nationality of applicant(s) as in international application form: PHARMACIA AB, a Swedish company of S-171 97 Stockholm, Sweden 1 2660 45 yeast strain and methods for expressing heterologous proteins in yeast.

Technical field: The invention concerns a novel yeast (Saccharomyces cerevisiae) strain that performs O-glycosylation of at least one serine and/or at least one threonine residue at a reduced level. The invention also provides novel methods for the production of 10 heterologous proteins, primarily eukaryotic such as mammalian proteins, in yeast cells.

Technical background: In fungi O-glycosylation is initiated in the endoplasmatic 15 reticulum and consists of the transfer of a single mannose residue from dolichol-phosphate-mannose (Dol-P-Man) to the nascently secreted protein (Haselbeck and Tanner, FEBS lett. 158 (1983) 335-338). In the yeast S. cerevisiae further additions of 3-4 mannose residues using GDP-Man as donor can occur in the Golgi system 20 leading to maximal O-glycosyl chain lengths of 5 mannose residues (Sharma et al., Eur. J. Biochem. 46 (1974) 35-41). Linkages between mannose residues at positions 1, 2 and 3 are a 1 -2, while al -3 linkages exist between mannose residues at positions 3, 4 and 5. While the in vivo attachment sites for O-glycosyl chains on 25 yeast proteins have not been defined, sites on some heterologous proteins secreted by S. cerevisiae are known. Human granulocyte-macrophage colony-stimulating factor (hGM-CSF) is partially O-glycosylated in yeast (Ernst et al., Eur. J. Biochem. 203 (1992) 663-667). O-glycosylated hGM-CSF species appear to carry either 30 an extended chain of up to 5 mannoses on serine-9, or single mannose residues simultaneously on serine-9 and threonine-10.

Thus, serine-9 is the principal O-glycosyl attachment site in hGM-CSF. In yeast-secreted human insulin-like growth factor (hIGF-l) about 50% of the secreted protein carries a dimannosyl chain on 35 threonine-29 (Gellerfors et al., J. Biol. Chem. 19 (1989) 11444-49). Serine and threonine attachment sites have also been identified in experiments in vitro using synthetic peptides as substrates for WO 94/26873 PCT/SE94/00421 yeast mannosyltransferases (Bause and Lehle, Eur. J. Biochem. 101 (1979) 531-40; Strahl-Bolsinger and Tanner, Eur. J. Biochem. 196 (1991) 185-90; Lorenz et al., Eur. J. Biochem. 205 (1992) 1163-67). Like in hGM-CSF and hIGF-l the O-glycosylated serine or threonine 5 residues in synthetic peptides may be directly flanked at their N-terminal by a proline residue. The glycosylation patterns in mammalian systems and in S. cerevisiae differ significantly, particularly with regard to O-glycosylation attachment sites (serine and threonine) and types of carbohydrates involved. 10 Recently, Strahl-Bohlsinger et al (Yeast 8 (1992) S489, the 16th International Conference on Yeast Genetics and Molecular Biology, Vienna (Austria), August 15-21, 1992) described a gene in S. cerevisiae, whose mutation reduces the in vivo O-glycosylation by about 50%, while the in vitro O-glycosylation in the mutant 15 extract is absent. The mntl mutation leads to shortened O-glycosyl chains consisting of only 2 mannose residues; in this mutant a specific a 1-2 mannosyltransferase is defective (Hausler et al., Proc. Natl. Acad. Sci. USA 89 (1992) 6846-50). The mntl mutation is also known as kreZ and renders S. cerevisiae resistant to the 20 action of K1 killer toxin (Hausler et al., Proc. Natl. Acad. Sci. USA 89 (1992) 6846-50). In mnnl mutants, a defective a 1-3 mannosyltransferase does not add the terminal 1 -3 linked mannose residues in 0-, and N-glycosyl chains (Ballou et al., Proc. Natl.

Acad. Sci. USA 88 (1991) 3209-12). Other mutations defective in 25 the synthesis of glycosylation precursors, mannose, GDP-Man and Doi-P-Man, also affect both O-or N-glycosylation. Mutations of this type include defects in SEC59, DPMI, SEC53 and PMI genes encoding dolichol kinase, dolichol phosphate mannose synthase, phosphomannomutase and phosphomannose isomerase, respectively. 30 At present it is not clear, if O-glycosylation in yeast is essential for viability. Only mutations abolishing 0- and N-glycosylation completely (SEC59, DPMI, SEC53 and PMI) lead to lethality (Heller et al., Proc. Natl. Acad. Sci. USA 89 (1992) 7013-16; Orlean et al., J. Biol. Chem. 163 (1988) 17499-507; Kepes and Schenkman, J. Biol. 35 Chem. 163 (1988) 9155-61; Smith et al., Mol. Cell. Biol. 12 (1992) 2924-30).

WO 94/26873 PCT/SE94/00421 An Article in J. Protein Chem. 9:95-104 (1990) by Elliot et al. describes the purification and properties of yeast-produced hlGF-l.The hlGF-1 protein has been mutated to reduce O-glycosylation. Site-directed mutagenesis was used to convert Thr 29 to Asn29 5 which reduced but not eliminated IGF-I glycosylation. It thus describes alterations in the IGF-1 protein itself, and not the host cell.

EP 314096 concerns yeast mutants that are defective in the addition of outer N-glycosyl chain^EP 276846 discloses the 10 biological activities of various forms of GM-CSF lacking sites for N- and/or O-glycosylation. This patent describes mutants of the expressed protein GM-CSF, and not the host mutants.

Problems solved by the invention: The unique glycosylation patterns of yeast strains have in many cases a negative impact on the yield obtained of the desired form of heterologously produced proteins. Moreover extra precautions during purification must be applied in order to remove undesired forms of the produced protein.

Figure 1 A and B shows reversed phase HPLC chromatograms.

Objectives of the invention: One objective of the invention is to provide S. cerevisiae 25 expressed proteins, in particular heterologous proteins, that are minimally O-glycosylated at one or more particular serine and/or threonine attachment site(s). An additional objective is to provide a production method for these proteins and a means (a mutated S.cerevisiae strain) for the method. In addition the invention will 30 provide S. cerevisiae homologous proteins that are minimally O-glycosylated at a serine or threonine site.

The invention: The S. cerevisiae strain of the invention is characterized in 35 that it is defective in initiating O-mannosylation of the hydroxyl group of at least one residue selected from serine and threonine in a protein expressed by the strain. This is probably due to a defect .WO 94/26873 PCT/SE94/00421 in the recognition of serine, or threonine acceptor sites, or in the mannosyltransferase activity required to glycosylate these sites. A serine and/or threonine specific O-mannosyltransferase may be missing or altered in activity. Other aspects of glycosylation are 5 similar to wild-type glycosylation, e.g glycosylation performed by strain YE-465. Thus, further attachment of saccharide units to a monomannosylated serine or threonine residue and/or N-glycosylation and/or in vivo synthesis of necessary carbohydrate intermediates are in most cases normal as determined by the 10 procedures given in the experimental section.

Particularly important modes of the invention are transformed S. cerevisiae strains comprising an expression vector containing a gene for a heterologous, in particular mammalian, protein, such as hGM-CSF or hIGF-l. It follows that in these modes 15 of the invention, the mutation causing decreased O-glycosylation is in a yeast chromosomal gene.

A second main aspect of the invention is a method for producing a protein, often a heterologous, in particular mammalian, protein comprising culturing the inventive S. cerevisiae strain and 20 recovering the protein from the fermentation medium or from the cells. It is conceivable that conventional S. cerevisiae culturing methods and conventional recovering methods may be used.

Best Mode at the priority date: The greatest advantage based on today's knowledge for the inventive S. cerevisiae strains and production method is obtained for proteins containing one or more serine or threonine residue(s) that are recognized as O-glycosylation sites by wild-type strains of S.cerevisiae (containing a vector expressing such a protein) but 30 not by mammalian systems such as in humans. In other words the most preferred S. cerevisiae strains of the invention contain an expression vector (e.g. plasmid) for a mammalian, in particular human, protein that is not O-glycosylated by its original species but O-glycosylated by for instance YE-449 containing the 35 appropriate expression plasmid for the protein. The most preferred strain has the same mutation as the Ml 95 strain (see below). The most preferred heterologous protein is hIGF-l.

.WO 94/26873 ' PCT/SE94/00421 EXPERIMENTAL PROCEDURES 5 Strains and growth conditions: The starting strain for mutagenesis was S. cerevisiae YE-465 which consists of host strain YE-449 (Biogen) (MATcc leuZ ura3-52 prbl-1122 pep4-3 cii°) carrying plasmid p539/12 (Gellerfors et al., J. Biol. Chem. 264 (1989) 11444-49). p539/12 is an expression plasmid for hIGP-! 10 that leads to the secretion of biologically active hIGF-l into the growth medium of yeast transformants.

Mutagenesis: Ethyl methane sulfonate (EMS) was used to mutagenize strain YE-465. Treatment of a log-phase culture for 120 min and 180 min with 2.5% EMS yielded killing rates of 81 % 15 and 91%, respectively. The 180 min culture was plated for single colonies on minimal medium. Single colonies were inoculated with a needle into 2 ml production medium containing 4% casamino acids in reagent tubes and grown for 3 days at 30°C. One ml of the culture supernatant was analysed by concanavalin A blotting, or 20 immunoblotting using an anti-hlGF-l antibody.

To lose the expression plasmid in putative mutants they were grown non-selectively in YPD medium (Sherman et al., Methods in Yeast Genetics (Cold Spring Harbor Lab.) Cold Harbor, N.Y. (1986) 163), followed by the analysis of single isolates for the Ura" phenotype. While mutants are named "M", cured mutants were given the designation "CM". Cured mutants (and YE-449 as control) were retransformed with p539/12, or with a secretion vector for hGM-CSF, pER545/4. pER545/4 is a derivative of pER562 (Ernst et al., Bio/Technology 5 (1987) 831-34). Transformants were first grown 30 in selective minimal medium (Sherman et al., Methods in Yeast Genetics (Cold Spring Harbor Lab.) Cold Spring Harbor N.Y. (1986) 164); this culture was used to inoculate production medium containing 4% casamino acids (Ernst et al., Bio/Technology 5 (1987) 831 -34) and the culture was grown for 2-3 days at 30°C to 35 an ODgoonm=10- Cells were removed by centrifugation; for hIGF-l expressing transformants the culture supernatant was concentrated 25-fold by Trichloro acetic acid (TCA) precipitation .WO 94/26873 PCT/SE94/00421 prior to electrophoresis followed by concanavalin A and immunoblotting procedures.

Blotting procedures: Glycoproteins in the culture medium were identified by concanavalin A blotting (Clegg, Anal. Biochem. 127 5 (1982) 389-94). For this purpose 20 of the TCA-concentrated medium was separated by SDS-PAGE (17,5% gel), transferred to nitrocellulose and stained as described. For immunoblottings proteins were transferred to membranes (Immobilon-P, Millipore) and reacted with the monoclonal anti-hlGF-l antibody 5B3 diluted 10 1:1000, followed by treatment with peroxidase-coupled goat anti-mouse IgG antibody diluted 1:2000 (Jackson Immuno Research, USA). For the detection of hGM-CSF a rabbit polyclonal anti-hGM-CSF antibody (kindly supplied by Glaxo Institute for Molecular Biology, Geneva) was used as first antibody (diluted 1:100), 15 followed by alkaline phosphatase-coupled goat anti-rabbit IgG antibody (Dianova, Germany) diluted 1:5000.

Chromatography procedures: Reversed phase high performance liquid chromatography (HPLC) was performed with a diphenyl silica analytical column equipped with a C4-silica precolumn. The 20 monomeric hlGF~l forms were eluted with an acetonitrile-trifluoro acetic acid (TFA) gradient containing 0.1% (v/v) TFA. The hIGF-l forms were eluted at around 25 % (v/v) acetonitrile.

Other procedures: Standard procedures were used for genetic analyses and crosses of mutants to laboratory strains SI 50-2B 25 (MATa leu2-3,7 12 ura3-52 trpl-289 his3-l) and BJ1991 (MATa ura3-52 Ieu2 trpl prbl-1122 pep4-3 gal2) (Sherman et al., Methods in Yeast Genetics (Cold Spring Harbor Lab.) Cold Spring Harbor, N.Y. (1986) 17-27). Hygromycin B-sensitivity was analyzed on antibiotic gradient plates (gradient from 0 to 200 //g/ml). 30 Resistance/sensitivity to killer toxin K1 was assayed as previously described (Sherman et al., Methods in Yeast Genetics (Cold Spring Harbor Lab.) Cold Spring Harbor N.Y. (1986) 57-60) using strain RC1777 (MATa ade his4C (KIL-kl)).

WO 94/26873 PCT/SE94/00421 results Screening for O-alvcosvlation mutants: Out of 600 mutagenized isolates 8 putative mutants were isolated. To exclude the possibility that the mutagenesis had affected the hIGF-l gene 5 or other genetic material present on the expression plasmid we obtained derivatives of the mutants that had lost their plasmid during non-selective growth. When these cured mutants were transformed with the expression plasmid, an identical phenotype as in the original mutant was obtained. Two of the identified 10 mutants, Ml 95 and M38, were characterized in more detail, as described below.

Characteristics of Ml 95: Mutant strain Ml 95 secretes similar amounts of hIGF-l into the growth medium as compared to the parent production strain YE-465 according to SDS-PAGE and 15 immunoblotting. In contrast, we consistently detected lower concanavalin A-reactivity with hIGF-l secreted by Ml 95, than with hIGF-l secreted by YE-465 indicating reduced O-glycosylation at threonine-29. This phenotype was not due to mutations in the expression vector since the cured retransformed mutant (CM195) 20 also showed reduced hIGF-l O-glycosylation. SDS-PAGE also showed several minor bands of somewhat larger size than IGF-I. These bands represent material O-glycosylated on serine-69, adding to the evidence that the mutation in strain CM195 affects O-linked threonine glycosylation but not serine glycosylation. 25 In order to quantify the mutant phenotype we purified and analyzed hIGF-l secreted by strains Ml 95 and YE-465 by reversed phase HPLC. HPLC-results for culture media from strain YE-465 and Ml 95 are represented in Figure 1 (For legends see after Results). The chromatograms show four main peaks for monomeric hlGF-1 30 corresponding to O-glycosylated misfolded hlGF-1, non- glycosylated misfolded hlGF-1, O-glycosylated correctly folded hIGF-l, non-glycosylated correctly folded hIGF-l (i.e. the product). Misfolding of h!GF-l occurs due to improper disulfide formation from cysteine (Axelsson et al., Eur. J. Biochem. 206 (1992) 987-35 994). The results clearly demonstrate that the amount of 0- glycosylated hIGF-l is reduced and the amount of non-glycosylated hIGF-l is increased for Ml 95 compared to YE-465.

WO 94/26873 PCT/SE94/00421 In order to test if strain Ml 95 is also defective for protein 0-glycosylation at serine residues, we cured strain Ml 95 of its hlGF-I expression plasmid (resulting in strain CM195) and transformed an expression plasmid for hGM-CSF (pER545/4) into the cured 5 strain. No differences in the glycosylation pattern was found between a wild-type transformant (YE-449(pER545/4)) and the mutant transformant (CM195(pER545/4)) as revealed by SDS-PAGE followed by immunoblotting. This result suggests that serine glycosylation is not affected in Ml 95. On the other hand, N-10 glycosylation of hGM-CSF also does not seem to be affected in Ml 95, a finding that is confirmed for yeast invertase (see below).

Strain Ml 95 does not show any morphological abnormality, nor does it display temperature sensitivity, or sensitivity to high osmolarities in the growth medium. Ml 95 is as sensitive as the 15 parent YE-465 for killer toxin K1 and 5 mM vanadate. The only detectable phenotype other than its O-glycosylation defect that appears associated with strain M195 is an increased sensitivity for the aminoglycoside antibiotic hygromycin B. While strain YE-465 grows well in the presence of 60 pg/ml hygromycin B, Ml95 is 20 completely inhibited. The mnn9 mutation that affects N- glycosylation also is hygromycin B sensitive but resistant to 5 mM vanadate (Ballou et al., Proc. Natl. Acad. Sci. USA 88 (1991) 3209-12). To ascertain that the increased hygromycin B sensitivity in Ml 95 was associated with the observed glycosylation defect we 25 crossed Ml 95 to the wild-type strain SI 50-2B and examined the haploid progeny. Although many of the segregants of this cross failed to grow we were able to examine hIGF-l O-glycosylation in two hygromycin B sensitive haploid segregants by HPLC analysis and found reduced O-glycosylation in both strains. This finding 30 suggests that reduced O-glycosylation and hygromycin B-sensitivity are due to defects in the same gene.

Characteristics of M38: Secretion of hIGF-l by mutant strain M38 neither shows quantitative, nor qualitative differences compared to the parental strain YE-465; in particular hIGF-l is 35 modified by O-glycosylation to approximately the same extent in M38 and YE-465 as revealed by SDS-PAGE followed by immunoblotting. However, differences in the glycoprotein staining .

WO 94/26873 PCT/SE94/00421 9 pattern by concanavalin A demonstrate that several secreted glycoproteins are reduced in size in M38. In particular, a prominent protein of 35 kDa is missing; instead, a 27 kDa glycoprotein of equal intensity is detected.

In a wild-type strain of S. cerevisiae transformed with an hGM-CSF expressing plasmid immunoblotting on SDS-PAGE of secreted forms of hGM-CSF give bands at 14.5 kDa (unglycosylated), 15.5 kDa (O-glycosylated) and 50 kDa (N-glycosylated) (Ernst et al., Eur. J. Biochem 203 (1992) 663-67). To 10 assess if glycosylation of hGM-CSF is affected, we isolated a plasmid-free derivative of M38. This strain, CM38, was transformed with the hGM-CSF expression plasmid pER545/4 and the hGM-CSF secreted by this transformant was analyzed by immunoblotting. The unglycosylated 14.5 kDa form of hGM-CSF and 15 the heterogeneous N-glycosylated 50 kDa form of hGM-CSF occur both in CM38(pER545/4) and control strain YE-449(pER545/4). However, the O-glycosylated 15.5 kDa form of hGM-CSF is missing in CM38(pER545/4). This result suggests that the defect in M38 is different from the defect in Ml 95, although both defects affect 20 aspects of O-glycosylation in S. cerevisiae.

O-alvcosvlation of chitinase: The O-glycosylation of a homologous yeast protein, chitinase, which is extensively O-glycosylated, was examined in the yeast strains. In YE-465 chitinase appeared on a SDS-PAGE as a band of an approximate 25 molecular weight of 110 kDa. In the mutant strain Ml 95 the migration of chitinase was unchanged, indicating no difference in the O-glycosyation of this homologous protein.

N-alvcosvlation of invertase in mutant strains: To examine if any of the putative glycosylation mutants is deficient in N-30 glycosylation, we analyzed the homologous protein invertase, which is essentially only N-glycosylated. Extracts of the mutants were separated on a non-denaturing acrylamide gel and invertase activity was visualised by an activity stain. For two control strains (mnn9 and mnn7) invertase migrates further than for the 35 strain YE-465 and the mutants. By immunoblotting after SDS-PAGE on the same extracts, the defect in the mnn9 strains is clearly detected leading to a relatively homogeneous protein due to the • WO 94/26873 lack of outer glycosyl chains, but the production strain and the mutants all express invertase as a heterogeneous glycoform. This result indicates that in the putative mutants including Ml 95 and M38 N-glycosylation is not affected.

P1SCUSS1PN The mutant strains Ml 95 and M38 differ from other mutants that have been shown to be defective in O-glycosylation in S. cerevisiae. Because protein N-glycosylation of hGM-CSF and 10 invertase proceed normally in these mutants their genetic defect is different from "unspecific" mutations, such as mutations in SEC59, SEC53, DPMI or PMI, which affect 0-, as well as N-glycosylation. It has been reported recently that defects in the mntl gene lead to a specific shortening of all O-glycosyl chains to 15 two mannose residues (Hausler et al Proc. Natl. Acad. USA 89 (1992) 6846-50). However, in Ml 95 full-length O-glycosyl chains are observed in hIGF-l (although at a low frequency) and in hGM-CSF; in M38 O-glycosylation of hGM-CSF is defective. Thus defects in the mntl gene appear not to be the reason for the mutant 20 phenotype in Ml 95, or M38, although M38 is resistant to the K1 killer toxin, as has been reported previously for mntl strains. A mannosyltransferase has recently been purified from S. cerevisiae (Strahl-Bolsinger et al., Yeast 8 (1992) S489, the 16th International Conference on Yeast Genetics and Molecular Biology, 25 Vienna (Austria), August 15-21, 1992) based on in vitro 0-glycosylation assays using synthetic peptides (Strahl-Bolsinger and Tanner, Eur. J. Biochem. 196 (1991) 185-90). A gene corresponding to the transferase was isolated, the disruption of which leads to loss of in vitro mannosyltransferase activity and to 30 a reduction of in vivo O-glycosylation to about 50% (Strahl-Bolsinger et al., Yeast 8 (1992) S489, the 16th International Conference on Yeast Genetics and Molecular Biology, Vienna (Austria), August 15-21, 1992). In addition, the disruptant strain shows normal cytology, but forms multiple adherent clumps of 35 cells. Unlike the phenotype of this disruptant, strains Ml 95 and M38 do not form clumps during growth; Ml 95 has no morphological defects, while M38 tends to form rod-like, elongated cells at high —a WO 94/26873 W 11 temperature. Also, cell extracts of strains Ml 95 and M38 contain wild-type levels oi <n vitro O-glycosylation (Strahl-Bolsinger, unpublished requite ). This evidence indicates that neither gene known to affect J-glycosylation is mutated in mutants Ml 95 and 5 M38.

The genetic evidence obtained here and in previous studies strongly suggests that O-glycosylation in S. cerevisiae is a complex process that requires multiple cellular factors. The experments by Strahl-Bolsinger et al., (Yeast 8 (1992) S849, the 10 16th International Conference on Yeast Genetics and Molecular Biology, Vienna (Austria), August 15-21, 1992) indicate that not all of the mannosyltransferases that are active in vivo can be assayed in vitro, a result that clearly demonstrates the necessity of a genetic approach. The mutants isolated in the present study 15 may be defective in the recognition of serine, or threonine acceptor sites, or in the mannosyltransferase activity required to glycosylate these sites. Thus, Ml 95 may be defective in mannosylation of threonine residues (as in hlGF-1), but not serine residues (as in hGM-CSF). On the other hand M38 may be defective 20 in mannosylation of serine, but not threonine residues. Further analyses of the isolated mutants including gene cloning and their in vivo disruption promise to clarify details of the O-glycosylation process in S. cerevisiae.

Legends to Figure 1 25 Reversed phase HPLC chromatograms. (A) culture medium from Ml 95; (B) culture medium from YE-465. Four main peaks representing monomeric hIGF-l are visible in each chromatogram, from left (shorter retention time): O-glycosylated incorrectly folded hIGF-l (mismatched), nonglycosylated incorrectly folded 30 hlGF-1 (mismatched), O-glycosylated correctly folded hIGF-l and nonglycosylated correctly folded hIGF-l (desired product). There is a remaining small peak in position (3) for the Ml 95 strain (B). This peak might represent some remaining O-glycosylation on serine-29 of hIGF-l or possibly a completely different form of hIGF-l with 35 similar retention time. 12 266045

Claims (10)

1. Saccharomyces cerevisiae strain, characterized in that it is a transformant containing a heterologous gene for human insulin-like growth factor I (hIGF-I) to be expressed and is defective in O-mannosylation of one hydroxyl group in at least one residue selected from serine and threonine in hIGF-I expressed by the strain.

2. Saccharomyces cerevisiae strain according to claim 1, characterized in that the strain is the mutant strain M195 as decribed in the specification.

3. Saccharomyces cerevisiae strain according to claim 1, characterized in that the strain is defective in O-mannosylation of a free hydroxyl group of at least one serine residue.

4. Saccharomyces cerevisiae strain according to any one of claims 1-3, characterized in that the strain is defective in the O-mannosylation of a free hydroxyl group in at least one threonine residue.

5. Saccharomyces cerevisiae strain according to any one of claims 1-4, characterized in that (a) the expressed protein is derived from a mammalian species, arid (b) the O-glycosylation of the protein occurring directly at a hydroxyl group of at least one threonine and serine attachment site differs between the mammalian species and a wild-type strain of S. cerevisiae.

6. Saccharomyces cerevisiae strain as claimed in claim 5 wherein the mammalian species is homo sapiens and the protein is a protein that is not substantially O-glycosylated in homo sapiens.

7. Saccharomyces cerevisiae strain according to any one of claims 1-6, characterized in that the strain performs N-glycosylation that is essentially normal for Saccharomyces cerevisiae. —— 13 266 045

8. Saccharomyces cerevisiae strain according to any one of claims 1-7, ctiaracterized in that the strain performs essentially normal wild-type attachment of monosaccharide residues to an O-attached monomannosyl serine or O-

9. P-^ress for the production of a protein that normally is O- glycosylated by S. c r .oisiae, characterized in that a S. cerevisiae strain according to any one of claims 1-8 is used to express the protein. 10

10. Process according to claim 9, characterized in that the protein is human insulin-like growth factor I (hIGF-I) and that the strain carries a gene for said p rotein. 5 attached monomannosyl threonine residue. 15 END OF CLAIMS