npgrj_nprot_2008-213 58..70

PROTOCOL

Engineering complex-type N-glycosylation in Pichia pastoris using GlycoSwitch technology Pieter P Jacobs1,2, Steven Geysens1,2, Wouter Vervecken1,2, Roland Contreras1,2 & Nico Callewaert1,3 1Unit

for Molecular Glycobiology, Department for Molecular Biomedical Research, VIB, Technologiepark 927, B-9052 Ghent (Zwijnaarde), Belgium. 2Department of Molecular Biology, Ghent University, Technologiepark 927, B-9052 Ghent (Zwijnaarde), Belgium. 3Laboratory for Protein Biochemistry and Biomolecular Engineering, Department of Biochemistry, Physiology and Microbiology, Ghent University, K.L.-Ledeganckstraat 35, B-9000 Ghent, Belgium. Correspondence should be addressed to N.C. ([email protected]).

© 2008 Nature Publishing Group http://www.nature.com/natureprotocols

Published online 18 December 2008; doi:10.1038/nprot.2008.213

Here we provide a protocol for engineering the N-glycosylation pathway of the yeast Pichia pastoris. The general strategy consists of the disruption of an endogenous glycosyltransferase gene (OCH1) and the stepwise introduction of heterologous glycosylation enzymes. Each engineering step results in the introduction of one glycosidase or glycosyltransferase activity into the Pichia endoplasmic reticulum or Golgi complex and consists of a number of stages: transformation with the appropriate GlycoSwitch vector, small-scale cultivation of a number of transformants, sugar analysis and heterologous protein expression analysis. If desired, the resulting clone can be further engineered by repeating the procedure with the next GlycoSwitch vector. Each engineering step takes B3 weeks. The conversion of any wild-type Pichia strain into a strain that modifies its glycoproteins with Gal2GlcNAc2Man3GlcNAc2 N-glycans requires the introduction of five GlycoSwitch vectors. Three examples of the full engineering procedure are provided to illustrate the results that can be expected.

INTRODUCTION Yeasts are widely used both by industrial and academic research laboratories for the production of heterologous proteins. In particular, the methylotrophic yeast Pichia pastoris is extensively used as protein production platform. The popularity of this particular yeast is attributable to its ability to produce foreign proteins at high levels, the simplicity of the techniques needed for its genetic manipulation and its capacity to perform many eukaryotic co- and post-translational modifications, including N-glycosylation1,2. However, therapeutic glycoproteins intended for parenteral use in humans are so far typically produced in mammalian cells because of their ability to modify proteins with mammalian complex-type N-glycan structures. Yeasts are unfavorable in this respect because they modify glycoproteins with nonhuman high mannose-type N-glycans2,3. These structures drastically reduce in vivo protein half-life4–7, may be immunogenic in man8,9 and hamper downstream processing (because of their sometimes extreme heterogeneity). N-glycosylation N-glycosylation is the attachment of oligosaccharides to specific asparagine residues within the consensus sequence Asn-X-Ser/Thr. For a review of the pathway, see ref. 10. Briefly, in eukaryotes, this process occurs cotranslationally, and the central step takes place at the luminal side of the endoplasmic reticulum (ER) membrane, involving the transfer of a Glc3Man9GlcNAc2 oligosaccharide to nascent polypeptide chains. This precursor structure is then further modified by a series of glycosidases and glycosyltransferases. The initial processing reactions take place in the ER. Following the removal of the three glucose residues by glucosidase I and II, one specific terminal a-1,2-mannose is removed by mannosidase I. These reactions are well conserved between most lower and higher eukaryotes. At this point, correctly folded Man8GlcNAc2 N-glycosylated proteins exit from the ER to the Golgi complex

58 | VOL.4 NO.1 | 2009 | NATURE PROTOCOLS

where the glycans undergo further species- and cell typespecific processing. In higher eukaryotes, the Man8GlcNAc2 structures coming from the ER are further trimmed by several a-1,2-mannosidases. The resulting Man5GlcNAc2 N-glycans are subsequently modified by the addition of a b-1,2-linked GlcNAc residue in a reaction catalyzed by GlcNAc transferase I (GnT-I), leading to the formation of ‘hybrid-type’ N-glycans. Upon removal of two mannoses by mannosidase II (Man-II), a second b-1,2-GlcNAc is added by GnT-II. Glycans with the resulting structure in which both corea-mannose residues are modified by at least one GlcNAc residue are called ‘complex type’ N-glycans. The addition of galactose and sialic acid residues is catalyzed by galactosyltransferases and sialyltransferases, respectively. Additional branching can be initiated by GnT-IV, GnT-V and GnT-VI10. In contrast to higher eukaryotes, N-glycan diversity in yeast is typically generated by the addition of mannose and mannosylphosphate residues. Yeasts do not further trim the Man8GlcNAc2 glycans that arrive at the Golgi from the ER. Instead, these structures are modified by the addition of an a-1,6-mannose residue (indicated in red in Fig. 1a) to the a-1,3-mannose of the trimannosyl core, a reaction catalyzed by Och1p11,12. This ‘initiating mannose’ is then further elongated by several (phospho) mannosyltransferases. The resulting mannan-type structures consist of a backbone of up to several dozen a-1,6-mannoses with short side branches, collectively known as the ‘outer chain.’ These N-glycans are often referred to as hyperglycosyl- or hypermannosyl-type structures3. Glycoengineering strategy The protocol presented here details our approach to engineer the P. pastoris N-glycosylation pathway. The general strategy to convert this yeast’s heterogeneous high mannose-type N-glycosylation to mammalian-type N-glycosylation (hybrid and complex-type


PROTOCOL Figure 1 | Glycoengineering strategy overview. High mannose a b HDEL receptor: COPI-mediated (a) Schematic outline of the procedure for P Cytoplasmic part retrieval to ER P Outer chain engineering the N-glycosylation pathway of Cytosol P. pastoris using GlycoSwitch plasmids. Each P P engineering step results in the introduction of one pGS-M8 Stalk region glycosidase or glycosyltransferase activity in the Golgi lumen α-1,3-Arm Pichia ER or Golgi complex. The construction of a Trimannosyl strain that modifies its glycoproteins with Catalytic domain core M8 WT HDEL biantennary complex-type N-glycans (e.g., Gal2Gn2M3) with terminal galactose requires the T. reesei Golgi-targeted glycosidase pGS-M5 pGS-Man-I or glycosyltransferase α-1,2-Mannosidase introduction of five GlycoSwitch plamids. In all figures in this presentation, the graphical ZEO R c EM7 P CYC1 T representation for glycans as advised by the TEF1 P ori Consortium for Functional Glycomics is used pGlycoSwitchM8 (http://www.functionalglycomics.org/static/ pGS-GnT-I Stop AOX1 terminator M5 consortium/Nomenclature.shtml): green circle, OCH1 fragment Stop Bst BI mannose; blue circle, glucose; blue square, Single homologous N-acetylglucosamine; yellow circle, galactose. recombination Only here in panel (a), red circles are used to pGS-Man-II BstBI OCH1 OCH1 promotor emphasize the a-1,6-linked polymannosyl Genome backbone of yeast-type hypermannosylated glycans, of which the synthesis is abolished by GalGnM3 GnM5 inactivation of the OCH1 gene. (b) Many OCH1 fragment Zeocin resistance Stop OCH1 Bst BI cassette ori Stop Bst BI glycosidases and glycosyltransferases are type II pGS-GnT-II pGS-GalT membrane proteins. Their N-terminal region OCH1 promotor AOX1 T OCH1 inactivation Primers to evaluate (cytosolic tail, transmembrane domain and part of Genome after OCH1 inactivation the luminal ‘stem’ region) is responsible for correct subcellular localization. Proper targeting of each Transform strain of interest with d introduced glycosylation enzyme was achieved by selected GlycoSwitch plasmid (Steps 1–22) fusing its catalytic domain (shown in red) with the N terminus of a yeast protein with a known Gn2M3 GalGnM5 subcellular localization (shown in green). Small-scale cultivation and Consequently, the introduced enzymes are in fact pGS-Man-II pGS-GalT protein preparation of hybrid proteins with a yeast N-terminal >20 single clones (Steps 23–35) localization domain. (c) Upon digestion of pGlycoSwitchM8 with BstBI and transformation in P. pastoris, this construct integrates at the OCH1 Protein analysis: N-glycan analysis: locus. This results in a short OCH1 fragment that SDS-PAGE, western blot pGS-GnT-II DSA-FACE (Step 36) (Step 37) does not result in the synthesis of a functional Och1 protein and a promotorless fragment that Gal2Gn2M3 cannot give rise to a functional protein. (d) Each GalGnM3 engineering step consists of four stages: (i) Glycoengineered strain Hybrid type Complex type transformation with the appropriate GlycoSwitch vector; (ii) small-scale cultivation of a number of transformants; (iii) N-glycan analysis; (iv) and heterologous protein expression analysis. If desired, the best clone in terms of N-glycan profile and protein expression level can then be further engineered by repeating the procedure with the next GlycoSwitch vector in line.

structures) consists of the disruption of an endogenous glycosyltransferase gene (OCH1) and the stepwise introduction of heterologous glycosidase and glycosyltransferase activities (Table 1). The protocol and GlycoSwitch vectors described here are an integration of previously published work by us13,14 and researchers

at Glycofi Inc.15,16 (now a division of Merck & Co Inc.) on the different steps in this engineering strategy. In the course of this previous research, the engineered strains have been used to produce Trypanosoma cruzi trans-sialidase13, influenza A virus hemagglutinin13, Trichoderma reesei a-1,2-mannosidase14, the kringle 3 domain

TABLE 1 | Glycosylation enzymes used in the GlycoSwitch technique. Enzyme Och1 Man-I GnT-I Man-II Gal-T Gal10 GnT-II Kre2 Mnn2

Gene name OCH1 MDS1 MGAT1 a-Man-II B4GALT1 Gal10 Mgat2 KRE2/MNT1 MNN2/TTP1

Activity a-1,6-Mannosyltransferase a-1,2-Mannosidase N-acetylglucosaminyltransferase a-1,3/6-Mannosidase b-1,4-Galactosyltransferase UDP-glucose 4-epimerase b-1,2-N-acetylglucosaminyltransferase a-1,2-Mannosyltransferase a-1,2-Mannosyltransferase

Organism P. pastoris T. reesei H. sapiens D. melanogaster H. sapiens S. pombe R. norvegicus S. cerevisiae S. cerevisiae

NATURE PROTOCOLS | VOL.4 NO.1 | 2009 | 59


PROTOCOL of human plasminogen15 and human IgG16. The combination of engineering steps we describe here is the one we settled on after many years of experimentation and provides B90% engineering efficiency at each step in shake flask cultures, as we demonstrate below for three murine cytokines. An outline of the procedure for engineering the N-glycosylation pathway of P. pastoris using our GlycoSwitch plasmids is provided in Figure 1a. The protocol results in the modification of glycoproteins with complex-type N-glycans, with the Gal2GlcNAc2Man3GlcNAc2 structure (Fig. 1a, indicated as Gal2Gn2M3) as the final product. The protocol does not detail procedures to further modify this structure, e.g., by sialylation. In vivo engineering for the sialylation of glycoproteins in Pichia has been reported17, which involves the impressive feat of copying the entire CMP-N-acetylneuraminic acid synthesis pathway, a CMP-N-acetylneuraminic acid transporter and a sialyltransferase into the yeast system. However, this approach involves the expression of several more heterologous genes in the already heavily engineered strains that produce Gal2GlcNAc2Man3GlcNAc2. Given this added complexity (with further potential issues in engineering efficiency and genetic stability under process conditions), we are at present investigating novel alternatives. This protocol does not resolve the issue of nonhuman type O-glycosylation in yeasts such as Pichia. This is a difficult engineering target because of the lethality of complete O-glycosylation shutdown in yeast18. Recent progress includes the inhibition of O-glycosylation with small-molecule drugs19. Although useful on a small scale, this method is unlikely to be applicable to large-scale industrial processes. Moreover, the inhibition of O-glycosylation is partial and yeast-type O-glycans are still present at least to some extent on expressed proteins. Further work is ongoing in our laboratory to manage this problem in yeasts in a robust, scaleable way. Subcellular targeting of enzymes. As N-glycosylation is a sequential process, where one enzyme produces the substrate for the next, correct subcellular targeting of the introduced proteins is of critical importance. As the a-1,2-mannosidase activity to convert Man8GlcNAc2 (M8, Fig. 1a) to Man5GlcNAc2 (M5, Fig. 1a) is needed close to ER–Golgi transport, this enzyme was targeted by fusing it C-terminally with the 4 amino acid sequence, HDEL (which labels soluble proteins for retrieval from the Golgi to the ER20,21 (Fig. 1b). Most of the glycosidases and glycosyltransferases catalyzing Golgi N-glycosylation reactions are type II membrane proteins. Their N-terminal region (cytosolic tail, transmembrane domain and part of the luminal ‘stem’ region) is responsible for correct subcellular localization. However, expression studies in Saccharomyces cerevisiae with full-length mammalian glycosyltransferases have shown that they are often mislocalized, indicating that mammalian Golgi retention signals are not necessarily functional in yeast22,23. Therefore, we replaced the N-terminal localization signal of each introduced glycosylation enzyme with the N terminus of a yeast protein with a known subcellular localization14. Consequently, the introduced enzymes are in fact hybrid proteins with a yeast N-terminal localization domain (Fig. 1b). Disruption of the OCH1 a-1,6-mannosyltransferase gene As the Och1p a-1,6-mannosyltransferase initiates the ‘outer chain,’ disruption of OCH1 is the first step in the engineering of the 60 | VOL.4 NO.1 | 2009 | NATURE PROTOCOLS

N-glycosylation pathway of P. pastoris14. This is accomplished by transformation with pGlycoSwitchM8 (GlycoSwitch is abbreviated as GS in the vector names in Fig. 1a). An efficient single homologous recombination event results in a short OCH1 fragment under the control of the OCH1 promoter that does not translate to a functional Och1 protein. A second OCH1 copy that results from this inactivation strategy cannot be transcribed because of the absence of a promoter. Any translation of mRNA that would result from cryptic promoter activity is mitigated by the presence of two in-frame stop codons in the construct (Fig. 1c). About half of the antibiotic-resistant clones obtained upon transformation will have an integration at the targeted site, which is much more than using classical double homologous-recombination knockout in this particular locus (where several hundred clones need to be screened) (data not shown). This feature makes the protocol suitable for routine engineering of any Pichia strain optimized previously for the secretion of particular glycoproteins. The resulting M8 strain has a much-reduced ability to modify glycoproteins (both endogenous and heterologous) with hyperglycosyl N-glycans. The main N-glycan species is Man8GlcNAc214. As this structure is the substrate for several endogenous glycosyltransferases besides Och1p, the N-glycan profile is not entirely homogeneous. However, the heterogeneity of glycoproteins produced in M8 strains is typically strongly reduced (i.e., ‘smearing’ on SDS–polyacrylamide gel electrophoresis (SDS-PAGE) due to hyperglycosylation is largely mitigated), which may be sufficient for some applications (e.g., when the only purpose of the engineering is to improve downstream processing efficiency). The next step involves the introduction of the HDEL-tagged a-1,2-mannosidase from the filamentous fungus T. reesei. This enzyme removes all terminal a-1,2-linked mannose residues from Man8GlcNAc2. The resulting M5 strain modifies its glycoproteins predominantly with Man5GlcNAc2 structures. As most endogenous glycosyltransferases are not able to act on this structure, the N-glycan profile of a typical M5 strain is very homogeneous. For easy conversion of any expression strain into a M5 strain, the a-1,2-mannosidase-HDEL, has been introduced into the och1 inactivation vector. The resulting combination vector is designated pGlycoSwitchM514. Maturation of N-glycans into hybrid- and complex-type structures. The first step in the maturation of N-glycans into hybrid- and complex-type structures is the addition of a b-1,2linked GlcNAc residue to the a-1,3-mannose of the trimannosyl core, a reaction catalyzed by GlcNAc transferase I (GnT-I). In our approach, we have fused the catalytic domain of human GnT-I to the N-terminal domain of S. cerevisiae Kre2p14, a glycosyltransferase with a known cis/medial Golgi localization24. Expression of the Kre2-GnT-I hybrid protein in a M5 strain results in a strain (GnM5) that modifies its glycoproteins predominantly with GlcNAcMan5GlcNAc2 N-glycans. The Kre2–GnT-I fusion construct is introduced by transformation with the vector pGlycoSwitchGnT-I. The vector pGlycoSwitchGalT drives the expression of a tripartite fusion protein composed of the catalytic domain of human b-1,4-galactosyltransferase 1, the entire Schizosaccharomyces pombe UDP-Gal 4-epimerase and the Golgi-localization domain of S. cerevisiae Mnn2p (strategy first developed for Escherichia coli25 and adapted for Pichia by researchers at GlycoFi Inc.)26. GalT

PROTOCOL


adds a galactose residue in b-1,4-linkage to the b-1,2-GlcNAc, using UDP-Gal as donor substrate. The epimerase supplies the Golgi complex with sufficient amounts of UDP-Gal, by converting UDP-Glc into UDP-Gal. A more simple construct consisting of the Kre2p N-terminal region fused to the catalytic domain of human b-1,4-galactosyltransferase 1 is also functional14, and rigorous comparison of different galactosyltransferase constructs is pending. The resulting GalGnM5 strain modifies its glycoproteins predominantly with hybrid-type GalGlcNAcMan5GlcNAc2 structures. Production of mammalian complex-type structures. The first step toward the production of mammalian complex-type structures is the introduction of mannosidase II (Man-II) activity. Man-II is responsible for the removal of both terminal a-1,3- and a-1,6-mannoses from GlcNAcMan5GlcNAc2 N-glycans. The presence of a terminal b-1,2-linked GlcNAc residue on the a-1,3-arm is essential for its activity. The catalytic domain of Drosophila melanogaster Man-II was fused to the Golgi-localization domain of S. cerevisiae Mnn2p15. Expression of this fusion protein in a GnM5 strain results in a GnM3 strain. The latter modifies its glycoproteins with GlcNAcMan3GlcNAc2 N-glycans. Expression of the Mnn2DmMan-II fusion protein in a GalGnM5 strain results in a strain that modifies its glycoproteins with GalGlcNAcMan3GlcNAc2 structures (GalGnM3). Introduction of Man-II is a difficult step in the engineering process, and although a workable solution has been found, further research may yield still better results: besides significantly influencing the growth characteristics (Box 1), the presence of this enzyme results in a much more heterogeneous N-glycan profile. The products of Man-II appear to be (non-natural) substrates for endogenous glycosyltransferases implicated in outer chain synthesis. Fortunately, this problem can be largely solved by introduction of GnT-II, which competes with these endogenous mannosyltransferases for the same substrate (i.e., product of ManII). The final step toward the production of biantennary complextype N-glycans with terminal galactose is the introduction of a GlcNAc transferase II (GnT-II) activity. This enzyme catalyzes the addition of a second b-1,2-linked GlcNAc residue to the free a-1,6mannose of the trimannosyl core. Similar to other researchers in the field15, we have fused the catalytic domain of rat GnT-II to the N-terminal part of S. cerevisiae Mnn2p. Transformation of a GnM3 strain with the vector pGlycoSwitchGnT-II results in a strain that

modifies its glycoproteins predominantly with GlcNAc2Man3GlcNAc2 structures, termed as Gn2M3. Similarly, transformation of a GalGnM3 strain results into a Gal2Gn2M3 strain, i.e., a strain that can synthesize Gal2GlcNAc2Man3GlcNAc2 N-glycans. Protocol outline It is recommended to start with a wild-type P. pastoris strain already optimized to express a glycoprotein of interest. We typically start from a GS115 wild-type strain transformed with a pPIC9-derived target protein expression plasmid to complement its histidine auxotrophy. Linearization of the expression vector in the HIS4 gene directs it to the same locus in the genome. As considerable clonal variation in expression levels occurs in P. pastoris, at least 10 (and preferably many more) transformants should be evaluated in small-scale expression experiments to identify a clone that expresses the protein at suitable levels. With subsequent glycoengineering steps in mind, it is recommended to set up a semi-highthroughput small-scale expression protocol that allows for the simultaneous analysis of at least 25 clones. The primary purpose of the small-scale cultivation step (Steps 23–38) is to (i) identify Pichia clones that have the desired N-glycosylation profile and (ii) evaluate the impact (if any) of this particular glycoengineering step on the production level of the protein of interest. The exact specifications of the applied small-scale production protocol depend on the strain that is being glycoengineered. The protocol should be adapted to the protein of interest and to the promoter used (e.g. AOX1, GAP, y) to drive the expression of the gene of interest. Therefore, it is recommended to set up a small-scale expression protocol before proceeding with the glycoengineering protocol. In our laboratory, the gene of interest is typically driven by the methanol-inducible AOX1 promoter. Two example smallscale production protocols are detailed in this article. An outline of the procedure for engineering the N-glycosylation pathway of P. pastoris using our GlycoSwitch plasmids is provided in Figure 1a,d. The user has the choice to engineer the strain for the production of either hybrid-type N-glycans (left-hand side of Fig. 1a) or complex-type glycans (right-hand side of Fig. 1a). The construction of a strain that modifies its glycoproteins with biantennary complex-type N-glycans with terminal galactose requires the introduction of five GlycoSwitch plamids. Each engineering step results in the introduction of one glycosidase or glycosyltransferase activity in the Pichia ER or Golgi complex and consists of a number of stages: transformation with the appropriate

BOX 1 | GROWTH CHARACTERISTICS OF GLYCOENGINEERED STRAINS Strain GS115 (his4) GS115mIL-10 M5mIL-10 GnM5mIL-10 GalGnM5mIL-10 GalGnM3mIL-10 Gal2Gn2M3mIL-10

Introduced enzyme None None Man-I GnT-I GalT Man-II GnT-II

% Conversion NA NA ND1 89.5% 84.5% 90.8% 95.5%

Doubling time (hours) 2.40 2.36 2.31 2.51 2.82 3.61 4.62

Maximum growth rate l (h1) 0.0048 0.0049 0.0050 0.0046 0.0041 0.0032 0.0025

NA, not applicable; ND, not determined. 1Could not be determined, as there is no way to reliably quantify the amount of substrate. GS115mIL-10 and M5mIL-10 have very similar growth rates and generation times, indicating that OCH1 disruption has no serious detrimental effects on the viability of the cells. Every additional glycoengineering step, however, results in an increase in generation time (B10%) and a decrease in growth rate. Especially Man-II and GnT-II introduction seem to place a serious burden on the cells as they result in a B22% decrease in growth rate. As a consequence, the most heavily engineered strain, Gal2Gn2Man3mIL-10, has a doubled generation time compared with the wild-type strain. For each strain, the processing efficiencies of the introduced enzymes were calculated by determining the ratio between the amount of product produced by a specific enzyme and the sum of the amount of product and the amount of substrate that was not processed. The relative amounts were calculated by determining the area under each peak in the DSA-FACE profiles. The processing efficiencies shown are the average efficiencies of all relevant strains.



PROTOCOL GlycoSwitch vector (Steps 1–22); small-scale cultivation of a number of transformants (Steps 23–38); N-glycan analysis (Steps 39–51); and heterologous protein expression analysis (Step 52). This procedure results in the identification of the best clone in terms of N-glycan profile and protein expression level. If desired, the selected clone can then be further engineered by repeating the procedure with the next GlycoSwitch vector in line (Fig. 1d). The transformation protocol described here (Steps 1–22) is derived from ref. 27 and makes use of electroporation. We prefer this protocol because it results in high-transformation efficiencies and does not involve digestion of the cell wall, of which mannoproteins form an important part. The glycoengineering efficiency can be evaluated by analyzing the endogenous mannoproteins of the yeast’s cell wall, and we provide a simple, high-throughput protocol herein (Steps 39–50). We use an adapted protocol developed by Peat et al. in 1961 (see ref. 28). The resulting crude cell wall extract is a mixture of mannoproteins and b-glucan. The protein-bound N-glycans are subsequently analyzed by DNA sequencer-assisted (DSA)–fluorophore-assisted

MATERIALS REAGENTS

. Antibiotics: Blasticidin S (Fluka), Zeocin (Invitrogen), Nourseothricin (Werner BioAgents), Geneticin G418 (Invitrogen), Hygromycin B (Calbiochem) (see Table 2) . Plasmid DNA (see Table 3). The listed GlycoSwitch vectors are available from the authors through a Material Transfer Agreement for specified academic research not sponsored by companies . Bacto Agar (Difco) . Bacto peptone (Difco) . Bacto yeast extract (Difco) . Biotin (Sigma) . BMGY (see REAGENT SETUP) . BMMY (see REAGENT SETUP) . Citric acid (Calbiochem) . Deionized water (dd-H2O) . DTT (Sigma) . Glucose monohydrate (Merck) . Glycerol (Biosolve) . HEPES (Sigma) . Methanol (Biosolve) . NaCl (Merck) . Restriction enzymes: AvrII (New England Biolabs); BsiWI (New England Biolabs); BstBI (New England Biolabs); PmeI (New England Biolabs); SapI (New England Biolabs) . Sorbitol (Sigma) . Yeast nitrogen base (YNB) without amino acids (Difco) . YPD media and plates (see REAGENT SETUP) EQUIPMENT . 24-well culture plates (X25 UNIPLATE 10000 24 PP; Whatman) . Airpore tape (Qiagen) . Autoclave . Baffled shake flask, 1 liter . Centrifuge (Sorvall) . CryoTube vials, 1 ml (Nunc) . Eppendorf Safe-Lock microcentrifuge tubes . Electroporation cuvettes, 2 mm (Bio-Rad) . Falcon 50-ml conical tubes (BD) . Gene Pulser electroporator (Bio-Rad) . Incubator shaker at 30 1C . Microcentrifuge tube closures (Sherlock tube closures; USA Scientific) . Nucleospin Extract-II kit (Macherey-Nagel) . Oven at 37 1C


carbohydrate electrophoresis (FACE) (see below). Loading of the crude cell wall extracts on PVDF membranes results in binding of the mannoproteins to the membrane. As the PVDF membranes do not bind glucans, contaminating cell wall b-glucans can be easily removed (and prevented from interfering with the subsequent N-glycan profiling). Alternatively, the glycoprotein mix present in the growth medium can be analyzed. In our laboratory, N-glycosylation analysis (Step 51) is routinely performed by DSA-FACE as described in general previously29 (see Box 2). Other methods of N-glycan analysis, such as matrix-assisted laser desorption/ ionization–time of flight mass spectrometry, can be used as well, although they are less robust and quantitation can be problematic. Heterologous protein expression analysis (Step 52) is performed by SDS-PAGE, western blotting and/or ELISA, depending on the protein. On the basis of the N-glycan and protein expression analysis data, one clone is selected that can be subject to a second glycoengineering round. This procedure can be repeated until the desired glycostrain has been created (Fig. 1d).

. Oven at 30 1C . Safe-Lock Eppendorf tubes, 2 ml (Eppendorf) . Spectrophotometer (able to measure 600 nm) . SpeedVac vacuum dryer (Savant) . Tabletop centrifuge . Vortex REAGENT SETUP 1 M potassium phosphate buffer, pH 6.0 Dissolve 23 g of K2HPO4 and 118 g of KH2PO4 in 1,000 ml of deionized water and confirm that the pH ¼ 6.0 ± 0.1 (if the pH needs to be adjusted, add phosphoric acid or KOH). Autoclave for 20 min at 121 1C. The shelf life for this solution is 41 year at room temperature (21 1C). 13.4% (wt/vol) YNB without amino acids Dissolve 67 g of YNB without amino acids in deionized water to make a final volume of 500 ml. Heat the solution to dissolve the YNB completely. Filter-sterilize. The shelf life for this solution is 41 year when stored at 4 1C. 5003 biotin Dissolve 20 mg of biotin in 100 ml of deionized water and filtersterilize. Stored at 4 1C, the shelf life for this solution is 1 year at room temperature. BMGY medium Mix 1% (wt/vol) yeast extract, 2% (wt/vol) peptone, 1.34% (wt/vol) YNB without amino acids, 100 mM potassium phosphate buffer (pH 6.0), 4105% (wt/vol) biotin and 1% (vol/vol) glycerol. For 500 ml, dissolve 5 g of yeast extract and 10 g of peptone in 400 ml of deionized water and autoclave for 20 min at 121 1C. Cool to room temperature and add 50 ml of sterile 1 M potassium phosphate buffer (pH 6.0), 50 ml of sterile 13.4% YNB without amino acids, 1 ml of 500 biotin and 5 ml of sterile 100% glycerol. The shelf life for this solution is B2 months at room temperature. BMMY medium Mix 1% (wt/vol) yeast extract, 2% peptone, 1.34% (wt/vol) YNB without amino acids, 100 mM potassium phosphate buffer (pH 6.0), 4105% (wt/vol) biotin and 1% (vol/vol) methanol. For 500 ml, dissolve 5 g of yeast extract and 10 g of peptone in 400 ml of deionized water and autoclave for 20 min at 121 1C. Cool to room temperature and add 50 ml of sterile 1 M potassium phosphate buffer (pH 6.0), 50 ml of sterile 13.4% YNB without amino acids, 1 ml of 500 biotin and 5 ml of 100% methanol. The shelf life for this solution is B1 month at room temperature. YPD medium Mix 1% (wt/vol) yeast extract, 2% (wt/vol) peptone and 2% (wt/vol) glucose monohydrate. Dissolve 5 g of yeast extract, 10 g of peptone and 10 g of glucose monohydrate in deionized water to make a final volume of 500 ml. Autoclave for 20 min at 121 1C. The shelf life for this solution is approximately 1 month at room temperature.

PROTOCOL BOX 2 | DSA-FACE N-GLYCAN PROFILING


Details of the protocol for this method have been published elsewhere29. To prepare samples for FACE on capillary DNA-sequencers, N-glycans are released from the glycoproteins by treatment with peptide: N-glycosidase F (PNGase F). Subsequently, the released N-glycans are derivatized with the fluorophore 8-aminopyrene-1,3,6-trisulfonate (APTS) by reductive amination. After removal of excess APTS, the labeled N-glycans are analyzed with an ABI 3130 DNA sequencer. N-glycans of bovine RNase B and a maltodextrose ladder are included as references. DSA-FACE reagents ABI 310 running buffer or ABI 3130 running buffer (Applied Biosystems) Ammonium acetate (Merck) APTS (Molecular Probes) Citric acid (Calbiochem) DMSO (Aldrich) DTT (Sigma) EDTA, dihydrate (Vel) Exoglycosidases Jack Bean a-mannosidase (Sigma) Trichoderma reesei a-1,2-mannosidase (expressed and purified in our laboratory and available upon request) b-N-Acetylhexosaminidase (Prozyme) b-Galactosidase (Prozyme) GeneScan-500 LIZ size standard (when using ABI3130) or GeneScan-500 ROX size standard (when using ABI310) (Applied Biosystems) HCl (Merck) Iodoacetic acid (Sigma) Maltodextrin ladder, APTS-labeled (prepared in the author’s laboratory and available upon request) Methanol (Biosolve) NaCl (Merck) NaCNBH3 (Acros) ! CAUTION Only small quantities are used, but this reagent is hazardous: vapors (HCN) formed in the acidic medium are used for N-glycan derivatization. Use in a well-ventilated hood. PNGase F (New England Biolabs) Polyvinylpyrrolidone 360 (Sigma) POP-6 (when using ABI310) or POP-7 (when using ABI3130) Performance Optimized Polymer (Applied Biosystems) RNase B N-glycans, APTS-labeled (prepared in the author’s laboratory and available upon request). Sephadex G10 (Amersham) Sodium acetate (Sigma) Tris (Invitrogen) Urea (Merck) DSA-FACE equipment Adhesive tape for 96-well plates (Millipore) Autoclave Capillary (36 cm for ABI 310) or capillary array (36 cm for ABI 3130; Applied Biosystems) Centrifuge (Eppendorf 5810R) Freezer at 20 1C Genescan (ABI 310) or GeneMapper (ABI 3130) software (Applied Biosystems) Genetic analyzer (ABI 310 or ABI 3130; Applied Biosystems) Incubation oven at 37 and 50 1C Microcentrifuge (Eppendorf 5417C) Micropipettes (1,000, 200, 50, 10 and 2.5 ml) Multiscreen-ImmobilonP (Millipore) Multiscreen-Durapore 96-well filtration plates (Millipore) Multiscreen column loader system, 100 ml (Millipore) PCR thermocycler PCR tubes Reaction plates, 96-well (Applied Biosystems) Refrigerator at 4 1C Vacuum manifold for filtration plates (Millipore) Vortex Water bath YPD plates Mix 1% (wt/vol) yeast extract, 2% (wt/vol) peptone, 2% (wt/vol) glucose monohydrate and 1.5% (wt/vol) agar. Dissolve 5 g of yeast extract, 10 g of peptone, 10 g of glucose monohydrate and 7.5 g agar in deionized water to make a final volume of 500 ml. Autoclave for 20 min at 121 1C. Cool the mixture while stirring until below 50 1C, add the appropriate antibiotics (Table 2) and pour plates. Store plates at 4 1C for no longer than 1 month.

Plasmid DNA preparation All GlycoSwitch vectors and vectors for overexpression of mouse interleukin-10 (mIL-10), mGM-CSF and mIL-22 were prepared from cultures of E. coli MC1061 using standard plasmid purification kits that yield sequencing-grade, low-salt preparations of the plasmids (such as those that are obtained from Qiagen or Machery-Nagel). The plasmid concentration was estimated through absorbance measurements at 260 nm. Plasmid DNA is stored frozen at 20 1C. NATURE PROTOCOLS | VOL.4 NO.1 | 2009 | 63

PROTOCOL PROCEDURE Pichia transformation: vector linearization 1| Digest 5–10 mg of plasmid DNA with the appropriate restriction enzyme (Table 3) using enzyme/substrate ratios recommended by the manufacturer. Complete linearization is best ascertained through analyzing an aliquot of the restriction digest mixture on an agarose gel. One band of the length corresponding to the linearized plasmid should be observed.

TABLE 2 | Antibiotics. Antibiotic Blasticidinw Zeocinw Nourseothricinz G418z Hygromycin Bw wPurchased


zPurchased

Final concentration (lg ml1) 500 100 100 350 150

as pre-dissolved stock solutions. as powders, which are dissolved in dd H2O as 1000 stock solutions just prior to use.

2| Desalt the mixture. We typically use Macherey–Nagels Nucleospin Extract-II kit. Contaminants, such as salts and enzymes, are removed by a simple washing step. Pure DNA is finally eluted under low ionic strength conditions. m CRITICAL STEP It is essential that the DNA mixture is ‘salt-free,’ as salt causes arching during the electroporation step. 3| Evaporate the mixture to dryness (SpeedVac; this typically takes about 30 min without heating), and the DNA is resuspended in 10 ml of ultrapure water. Pichia transformation: preparation of competent Pichia cells 4| Transfer 10 ml of sterile YPD medium (see REAGENT SETUP) to a 50-ml Falcon tube. m CRITICAL STEP Work aseptically throughout this procedure. 5| Inoculate the tube with one colony from the P. pastoris clone of interest and grow overnight at 250 r.p.m. and 30 1C. 6| Next day, measure the OD600 of the overnight preculture (an absorbance of 1 in a 1-cm cuvette at 600 nm is about 2107 cells per ml). 7| Transfer 250 ml of sterile YPD medium to a 1-liter baffled shake flask. 8| Inoculate the shake flask with X ml of the pre-culture and grow overnight to an OD600 of 1.3–1.5. X can be calculated from the following formula: XOD2y ¼ 2501.4, where OD is the OD600 of the preculture, y is the number of generations the culture will be grown, 250 is the volume of the culture (in ml) and 1.4 is the desired OD600 value. The GS115 wild-type P. pastoris strain has a generation time of B2 h. However, some glycoengineering steps result in an increase in doubling time of the resultant strain (this is discussed in more detail in Box 1). 9| When the shake flask culture has reached the desired OD600 value, centrifuge the cells at 1,500g for 5 min at 4 1C. 10| Resuspend the cell pellet in 100 ml of YPD, 20 ml of HEPES buffer, pH 8, to which 2.5 ml of 1 M DTT has been added. Transfer the mixture back to the 1-liter baffled flask and shake for 15 min at 30 1C. m CRITICAL STEP Prepare 1 M DTT freshly. 11| Add 125 ml of ultrapure water and centrifuge at 1,500g for 5 min at 4 1C. m CRITICAL STEP Keep the cells on ice during all subsequent manipulations. 12| Resuspend the cell pellet in 250 ml of ice-cold, sterile ultrapure water. Centrifuge the cells at 1,500g for 5 min at 4 1C and discard the supernatant. 13| Resuspend the cell pellet in 125 ml of ice-cold, sterile ultrapure water. Centrifuge the cells at 1,500g for 5 min at 4 1C and discard the supernatant. 14| Resuspend the cell pellet in 20 ml of sterile, ice-cold 1 M sorbitol. Centrifuge the cells at 1,500g for 5 min at 4 1C and discard the supernatant. 15| Resuspend the cells in 500 ml of sterile, ice-cold 1 M sorbitol. The cells are now electrocompetent and should be used as soon as possible. m CRITICAL STEP The purpose of all these washing steps is to ensure that the cells are ‘salt-free’ (as salt causes arcing during the electroporation step) while suspending them in an osmotically stabilizing solution. Pichia transformation: transformation 16| Mix 80 ml of the competent cells from Step 15 with the linearized DNA from Step 3 and transfer them to an ice-cold 0.2-cm electroporation cuvette. Equilibrate on ice for 5 min. 17| Pulse the cells according to the parameters suggested for yeast by the manufacturer of the electroporation instrument being used. 18| Immediately add 1 ml of ice-cold 1 M sorbitol. 19| Transfer the cells to a sterile 15-ml tube and incubate without shaking at 30 1C for 1–2 h. 64 | VOL.4 NO.1 | 2009 | NATURE PROTOCOLS

PROTOCOL TABLE 3 | GlycoSwitch vectors. Vector

Short name

Localization signal

Promoter

Doch1 Doch1 Man-I; T. reesei; AA 25-523 GnT-I; H. sapiens; AA 103-445

NA NA C-terminal HDEL tag

NA NA GAP

Kre2p; S. cerevisiae AA 1-100

GAP

Zeocin

GAP

AvrII

9

pGlycoSpGS-GnT-I-HIS witchGnT-I-HISa

GnT-I; H. sapiens; Kre2p; S. cerevisiae AA 103-445 AA 1-100

GAP

Histidine

GAP

DraIII

9

pGlycoSwitchGalT/1

pGS-GalT

GalT; H. sapiens; Mnn2p; S. cerevisiae AA 44-398 AA 1-46 UDP-Gal 4epimerase; S. pombe; full length

GAP

Nourseothricin

GAP

AvrII

12

pGlycoSwitchGalT/2

pGS-GalT

GalT; H. sapiens; Mnn2p; S. cerevisiae AA 44-398 AA 1-46 UDP-Gal 4epimerase; S. pombe; full length

AOX1

Nourseothricin

AOX1

PmeI

12

pGlycoSwitchMan-II/1

pGS-Man-II

Man-II; D. melanogaster; AA 74-1108

Mnn2p; S. cerevisiae AA 1-36

GAP

G418

GAP

AvrII

12

pGlycoSwitchMan-II/2

pGS-Man-II

Man-II; D. melanogaster; AA 74-1108


AOX1

Hygromycin AOX1 B

PmeI

12

pGlycoSwitchGnT-II

pGS-GnT-II

GnT-II; R. norvegicus; AA 88-443


GAP

Hygromycin GAP B AOX1

SapI BsiWI

12

pGlycoSwitchM8 pGS-M8 pGlycoSwitchM5 pGS-M5


pGlycoSwitchGnT-I

pGS-GnT-I

Selection

Integration Linearization locus (restriction enzyme) Zeocin OCH1 BstBI Blasticidin OCH1 BstBI

References

Glycosyltransferase

9 9

AA, amino acid; NA, not applicable. The above-mentioned GlycoSwitch vectors are available from the authors, through a Material Transfer Agreement for specified academic research not sponsored by companies. 1Allows zeocin-based selection for multiple copy integration of the recombinant protein-construct.

20| Plate 10, 50 and 200 ml of this cell suspension on YPD plates containing the appropriate antibiotic. ? TROUBLESHOOTING 21| Incubate for 2–3 d at 30 1C until colonies appear. 22| Isolate 420 single clones by picking individual colonies and streaking them on YPD plates containing the appropriate antibiotic. Small-scale cultivation 23| Small-scale cultivation can either be done by the 50-ml Falcon tube method (option A) or the 24-well plate method (option B). (A) 50-ml Falcon tube method (i) Grow each isolated single clone from Step 22 in a 50-ml Falcon tube containing 10 ml of BMGY medium at 30 1C while shaking (250 r.p.m.). (ii) After 48 h of growth, centrifuge the cultures at 3,000g for 5 min at room temperature. (iii) Resuspend the cell pellets in 10 ml of BMMY medium. (iv) To maintain induction, spike the cultures every 12 h with 100 ml of 100% methanol (1% final concentration). In this way, cultivate for another 48 h at 30 1C (or shorter or longer, depending on the predetermined optimum for your specific protein). (v) After the desired induction time, measure the OD600 and harvest the cultures by centrifugation (3,000g for 10 min at room temperature). Remove and retain the supernatant as well as the cell pellets. ’ PAUSE POINT Freeze the supernatant and the cell pellets at 20 1C until further use.


PROTOCOL


(B) 24-well plate method (i) Aseptically inoculate each isolated single clone from Step 22 in one well containing 2 ml of BMGY medium. (ii) Seal the plate with Airpore tape. (iii) Incubate at 30 1C, while shaking (250 r.p.m.). (iv) After 48 h of growth, centrifuge at 3,000g for 10 min at room temperature. (v) Resuspend the cell pellets in 10 ml of BMMY medium. (vi) To maintain induction, spike the cultures every 12 h with 50 ml of 100% methanol (1% final concentration). Using this method, cultivate for another 48 h at 30 1C (or shorter or longer, depending on the predetermined optimum for your specific protein). (vii) At the end of the induction phase, measure the OD600 and harvest the cultures by centrifugation (3,000g for 10 min at room temperature). Remove and retain the supernatant as well as the cell pellets. ’ PAUSE POINT Freeze the supernatant and the cell pellets at 20 1C until further use. Cell wall mannoprotein extraction 24| Wash the cell pellets from Step 23A(v) or 23B(vii) once with 1 ml of 0.9% NaCl in a 2-ml Safe-Lock Eppendorf tube (2,800g, 1 min). 25| Wash the cells once with 1 ml of dd-H2O (2,800g, 1 min). 26| Resuspend the pellet in 1.5 ml of 20 mM Na-citrate buffer, pH 7.0. 27| Autoclave at 125 1C for 90 min. m CRITICAL STEP To avoid opening of the tubes, they should be locked with microcentrifuge tube closures. We use Eppendorf Safe-Lock microcentrifuge tubes because they do not melt/deform when subjected to Step 27. 28| When the samples are at room temperature, resuspend the cellular debris by vortexing. 29| Centrifuge at X13,500g for 5 min at room temperature. 30| Collect the supernatant in a 15-ml tube and add 4 volumes (B6 ml) ice-cold methanol to precipitate the mannoproteins. 31| Stir the mixture overnight at 4 1C, preferably by placing the samples on a rotating wheel. 32| Pellet the mannoproteins by centrifugation (3,220g, 4 1C, 15 min). 33| Remove and discard the supernatant fraction and wash the pellet with 0.5 ml of methanol (3,220g, 4 1C, 15 min). 34| Dry the pellet at 37 1C for 1 h. 35| Dissolve the pellet in 50–100 ml of dd-H2O. N-glycan analysis 36| Perform N-glycan analysis of the samples from Step 23A(v), Step 23B(vii) or Step 35 by DSA-FACE as described in ref. 29 (reagent/ equipment requirements are summarized in Box 2). If the protein of interest is secreted in the growth medium, use 500–1,000 ml of the supernatant fraction from Step 23A(v) or 23B(vii). In case of mannoprotein N-glycan analysis, use 50–100 ml from Step 35. Protein expression analysis 37| To evaluate whether the performed glycoengineering step has affected the expression level of the protein of interest, perform either SDS-PAGE, western blot and/or ELISA. To be able to draw well-founded conclusions, differences in OD600 at the end of the induction phase should be taken into account. Clone preservation 38| After having identified the best clone in terms of N-glycan profile and heterologous protein expression level, grow it overnight in 5 ml of YPD medium containing the appropriate antibiotics. 39| Make 1 ml of aliquots containing 20–30% glycerol. ’ PAUSE POINT Store glycerol stocks at 80 1C. Strains can be recovered for at least 1 year after freezing, and probably after much longer. Strain startup from preserved clones stored at 80 1C 40| Upon thawing, plate the clones from Step 39 on YPD plates containing all appropriate antibiotics. 41| Isolate 5–10 single clones. 42| Grow these clones according to the 50-ml Falcon tube method (Steps 23A) or the 24-well plate method (Steps 23B). 43| Perform N-glycan analysis according to Step 36. 44| Evaluate the protein expression level according to Step 37. 45| On the basis of the results from Steps 43 and 44, choose one clone to work with further. 66 | VOL.4 NO.1 | 2009 | NATURE PROTOCOLS

PROTOCOL

? TROUBLESHOOTING Step 20: loss of construct by out-recombination When introducing the glycoengineering constructs through single homologous recombination, as is done here and as is common practice for Pichia transgenesis, a direct repeat is created in the genome. While homologous in- and out-recombination is a rare event in Pichia under normal cultivation conditions, stresses on the cells like electroporation and prolonged storage on agar plates at 4 1C appears to somewhat induce out-recombination. We sometimes observe such loss of previously introduced constructs through out-recombination upon introduction of the next GlycoSwitch vector, but fully engineered clones are always identified. A more stringent selection for ‘good’ transformants can be obtained by plating the electroporation mixture on YPD plates containing all previously used antibiotics: the simultaneous use of multiple antibiotics has no detrimental effects on the viability of the glycoengineered strains. When storing a glycoengineered strain on plate for longer than 2 weeks, the appropriate antibiotics should be included. However, addition of antibiotics to the culture medium of both small-scale and large-scale protein expression cultivations is not necessary.

Figure 2 | SDS-PAGE analysis of medium proteins from mIL-10-producing strains. Strains were grown according to Step 23 in the protocol. Proteins produced by equivalent amounts of yeasts cells were TCA-precipitated and loaded on gel. The predominant protein produced by these strains is mIL-10 (indicated with an arrow). The hyperglycosylation of the GS115mIL10produced protein is clearly visible. A small amount of nonglycosylated mIL-10 is produced by each strain. Proteolytic degradation of mIL-10 seems to increase with further engineering of the strains. Glycoengineering does not severely decrease mIL-10 yields. GS115 ctrl ¼ GS115 wild-type strain not producing any heterologous protein.

15 mI L-1 M5 0 mI L-1 0 Gn M5 mI L-1 Ga 0 lGn M5 mI Ga L-1 lGn 0 M3 mI Ga L-1 l2G 0 n2 M3 mI L-1 0

GS 1

5c

For each of the three proteins, five glycoengineered strains were constructed: M5, GnM5, GalGnM5, GalGnM3 and Gal2Gn2M3, according to the protocol outlined above. All introduced glycosylation genes are controlled by the constitutive GAP promoter. So far, this has consistently yielded the best results. Nevertheless, most GlycoSwitch vectors exist in an AOX1 version as well.

trl

ANTICIPATED RESULTS The workflow presented in Figure 1a,d allows engineering of the N-glycosylation pathway of any wild-type P. pastoris strain. The construction of a strain that modifies its glycoproteins with Gal2GlcNAc2Man3GlcNAc2 N-glycans requires the consecutive integration of five GlycoSwitch vectors into the Pichia genome. Each of these plasmids contains a different dominant antibiotic resistance marker for selection. As a consequence, it is essential that the starting strain is still sensitive to all five antibiotics: blasticidin, zeocin, hygromycin, geneticin and nourseothricin. Some other combinations of engineering enzyme–selection marker are available (see Table 3). We typically start with the GS115 wild-type strain because its histidine auxotrophy provides us with an additional selection maker that allows selection for integration of a pPIC9-derived vector that drives the production of a protein of interest. We here provide three examples of the full engineering procedure, with proteins spanning almost two orders of magnitude difference in expression level. 1. mIL-10 is an B18.5-kDa homodimeric glycoprotein that is modified by the addition of one N-glycan structure near the N terminus. Pichia-produced mIL-10 appears as a heavily smeared band on SDS-PAGE (Fig. 2). Removal of all N-glycans by treatment with PNGase F showed that this smearing was due to hyper-N-glycosylation. Production levels in P. pastoris are relatively low: 5–10 mg liter1. 2. Mouse granulocyte-macrophage colony-stimulating factor (mGM-CSF) is a 124 amino-acid monomeric glycoprotein, with a predicted molecular weight of B14 kDa. It has two potential N-glycosylation sites and both sequons can be modified by the Pichia N-glycosylation machinery. Production levels in P. pastoris are in the range of 200 mg liter1 (ref. 30 and unpublished results, P.P.J., R.C., N.C.). 3. mIL-22 is a 146 amino-acid glycoprotein, with a theoretical molecular weight of 16.5 kDa. However, the presence of up to three N-glycans increases the actual molecular weight of the protein by several kDa (depending on the N-glycans attached). Production in P. pastoris yields several hundred mg of mIL-22. We have been able to purify B100 mg liter1 of this cytokine from a mIL-22-producing M5 strain (Jacobs et al., manuscript in preparation).

GS 11


TIMING Steps 1–20, transformation: 3 d (hands-on time: 4 h) Step 21, time needed for colonies to appear on plates: 2–3 d (hands-on time: 0 h) Step 22, isolation of single clones: 2 d (hands-on time: 30 min) Steps 23–35, small-scale cultivation and cell wall/secreted protein preparation: 4 d (hands-on time: 2–3 h) Step 36, N-glycan analysis by DSA-FACE: 3–4 d (hands-on time: 1 d) Step 37, protein analysis: 1–2 d (during N-glycan analysis) Total: 3 weeks per engineering step The slower doubling time of some glycoengineered strains (Box 1) does not negatively impact the timeline for introducing GlycoSwitch vectors, as the yeast growth steps in the protocol are not the bottlenecks.

50 37 25 20 15 10

Nonglycosylated mIL-10 Smearing (hyperglycosylation) mIL-10 degradation products


a

b

Glucose units 10

5

5 1a

1,100

Glucose units

15

10

15 1b

900 0

0

M10 M11 M13 M14 1,400 0 3a

2,400 0

4a

1,800 0

5a

1,800 0 A 230

B

6a

800 Relative fluorescence units

Figure 3 | DSA-FACE profiles for mIL-10. (a) N-glycan profiles of the proteins present in unpurified growth medium after small scale (Step 23) cultivation of these strains. (b) DSAFACE N-glycan profiles of mIL-10 purified from 250-ml shake flask cultures. Electropherograms 1a and b show the results for a maltodextrose reference. Electropherograms 2 through 8 show the results for N-glycans, as follows: electropherogram 2b, GS115mIL-10 (typical wild-type P. pastoris profile); electropherograms 3a and b, M5mIL-10 (the predominant peak is Man5GlcNAc2); electropherograms 4a and b, GnM5mIL-10 (the predominant peak is GlcNAcMan5GlcNAc2); electropherograms 5a and b, GalGnM5mIL-10 (the predominant peak is GalGlcNAcMan5GlcNAc2); electropherograms 6a and b, GalGnM3mIL-10 (the main peaks are GalGlcNAcMan3GlcNAc2 and GalGlcNAcMan4GlcNAc2); electropherograms 7a and b, Gal2Gn2M3mIL-10 (the main peak is Gal2GlcNAc2Man3GlcNAc2); electropherograms 8a and b, reference N-glycans from bovine RNase B (Man5-9GlcNAc2 (M5–M9)).

Relative fluorescence units


PROTOCOL

M9

M12

2b M15

3b

0 700

4b

0 3,800

5b

0 1,300

6b

0

0

7b

7a 800

1,700

0

0

M5 Mouse interleukin-10 M5 8b 8a M6 M6 2,100 M8 M8 M7 M7 The DSA-FACE N-glycan profiles of the M9 M9 3,400 0 0 proteins present in unpurified growth medium after small-scale cultivation of the mIL-10-producing strains (Steps 23–29) are shown in Figure 3a. DSA-FACE N-glycan analysis on mIL-10 purified from 250-ml shake flask cultures is shown in Figure 3b. In electropherogram 2b (Fig. 3), the N-glycan profile of GS115-produced mIL-10 is shown. The predominant peaks are Man10GlcNAc2 and Man11GlcNAc2. Very large N-glycan species are often difficult to detect by DSA-FACE because the high resolving power separates the myriad of isomers, each of which have very low abundance. Therefore, N-glycan species larger than 20 residues are hard to detect by DSA-FACE (this also holds true for routine matrix-assisted laser desorption/ionization–time of flight mass spectrometry because of poor ionization efficiency of high-MW glycans). Electropherogram 3a (Fig. 3) shows the total N-glycan pool on medium proteins from an M5mIL-10 culture; electropherogram 3b (Fig. 3) shows the N-glycans present on mIL-10 produced by this strain. Disruption of OCH1 and simultaneous overexpression of an ER-targeted a-1,2-mannosidase efficiently abolishes hyperglycosylation and reduces the N-glycan pool to one predominant species, Man5GlcNAc2. Electropherogram 4a (Fig. 3) shows the total N-glycan pool on medium proteins from a GnM5mIL-10 culture; electropherogram 4b (Fig. 3) shows the N-glycans attached to purified GnM5mIL-10. The main peak is GlcNAcMan5GlcNAc2. A small fraction of Man5GlcNAc2, however, was not modified with a terminal GlcNAc residue. The effect of the introduction of a galactosyltransferase is shown in electropherograms 5a and b (Fig. 3). In both profiles, the predominant peak is GalGlcNAcMan5GlcNAc2. However, small amounts of GlcNAcMan5GlcNAc2 and Man5GlcNAc2 are present on the purified protein. Electropherogram 6a (Fig. 3) shows the total N-glycan pool on medium proteins from a GalGnM3mIL-10 culture; electropherogram 6b (Fig. 3) shows the N-glycan profile of mIL-10 produced in the GalGnMan3 strain. The observed heterogeneity is due to (i) incomplete processing of several intermediate structures (the peaks corresponding to Man5GlcNAc2, GalGlcNAcMan5GlcNAc2 and GlcNAcMan3GlcNAc2 in electropherogram 6b of Fig. 3 are the result of nonquantitative substrate-to-product conversion by GnT-I, Man-II and GalT, respectively) and (ii) the synthesis of N-glycan intermediates that can also serve as substrates for endogenous glycosyltransferases. It seems that GalGlcNAcMan3GlcNAc2 can be modified by one or more endogenous a-1,2-mannosyltransferases resulting in GalGlcNAcMan4GlcNAc2 (as was shown by in vitro treatment with a-1,2-mannosidase; data not shown). A similar observation was made by researchers of GlycoFi Inc.15. The most efficient solution to this problem would probably be to identify the glycosyltransferases responsible and knock them out. This should become feasible when the much-anticipated P. pastoris genome is published. As an alternative, correctly sub-Golgi-localized GnT-II is able to prevent the addition of an additional a-1,2-linked mannose residue by competing with the endogenous glycosyltransferase for the same N-glycan structure, GlcNAcMan3GlcNAc2 (Fig. 3, compare electropherograms 6a and 7a)15. The N-glycan profile of mIL-10 produced in the most heavily engineered strain, Gal2Gn2Man3mIL-10, is shown in electropherograms 7a and b. The predominant peak is Gal2GlcNAc2Man3GlcNAc2. Galactosylation, however, is incomplete as evidenced by the presence of Gal1GlcNAc2Man3GlcNAc2 and GlcNAc2Man3GlcNAc2 N-glycans. However, it is most likely that in vitro polishing steps will be necessary to achieve 490% homogeneous N-glycan profiles, as has been reported by researchers at GlycoFi Inc.31,16.

mGM-CSF and mIL-22 As was shown for mIL-10, mGM-CSF-producing and mIL-22-producing M5, GnM5 and GalGnM5 strains produce relatively homogeneous N-glycan profiles consisting predominantly of Man5GlcNAc2, GlcNAcMan5GlcNAc2 and GalGlcNAcMan5GlcNAc2 structures, respectively (Figs. 4a and 5a). However, introduction of Man-II and GnT-II results in more heterogeneous N-glycan profiles. The GalGnM3mGM-CSF and Gal2Gn2M3mGM-CSF strains produce GalGlcNAcMan3GlcNAc2 and Gal2GlcNAc2Man3GlcNAc2 as the predominant N-glycans, respectively, 68 | VOL.4 NO.1 | 2009 | NATURE PROTOCOLS

PROTOCOL G nM 5m G M -C SF

M 5m G M -C SF

G S1 1

G al G nM 5m G M -C SF G al G nM 3m G M -C SF G al 2G n2 M 3m G M -C SF

ct rl

15

G S1

Relative fluorescence units

but a significant portion of their N-glycomes is composed of intermediate and high mannose structures (Figs. 4a and 5a). This is due to interfering endogenous mannosyltransferases and nonquantitative conversion, mainly by GalT and to a minor extent by Man-II and GnT-II. Man-II seems to produce oligosaccharides that are substrates for endogenous glycosyltransferases implicated in outer chain

22

22

Gn M5 mI L-

M3 mI L-

n2

l2G

Ga

Ga lGn M3 mI L-

22

Ga lGn M5 mI L-

trl

GS 11 5c

22

M5 mI L-

22

5m IL-

11

GS

GS 1

15

ctr

l

22

Glucose units Figure 5 | DSA-FACE and SDS-PAGE analysis of a b medium proteins from mIL-22-producing strains. 15 10 5 (a) N-glycan profiles of the proteins present in 1 1,000 unpurified growth medium after small-scale _ _ _ _ + + + cultivation of these strains (according to 75 0 50 Step 23 in the protocol). Electropherogram 1 M10 37 M9 2 * * shows the results for a maltodextrose reference. M11 * 450 25 M12 Electropherograms 2 through 8 show the results 20 0 for N-glycans of the glycoengineered expression 15 3 strains, as follows: electropherogram 2, GS115mIL800 10 22 (typical wild-type P. pastoris profile); 0 electropherogram 3, M5mIL-22 (the predominant 4 peak is Man5GlcNAc2); electropherogram 4, 1,500 GnM5mIL-22 (the predominant peak is 0 GlcNAcMan5GlcNAc2); electropherogram 5, GalGnM5mIL-22 (the predominant peak is 5 3,600 GalGlcNAcMan5GlcNAc2); electropherogram 6, 0 GalGnM3mIL-22 (the main peaks are GalGlcNAcMan3GlcNAc2 and _ _ _ 6 + + + + 75 1,600 GalGlcNAcMan4GlcNAc2); electropherogram 7, 50 37 Gal2Gn2M3mIL-22 (the main peak is 0 * * * * #[ #[ 25 Gal2GlcNAc2Man3GlcNAc2); electropherogram 8, 7 20 430 reference N-glycans from bovine RNase B 15 (Man5–9GlcNAc2 (M5–M9)). (b) Proteins produced 0 by equivalent amounts of yeast cells (B30107; M5 10 8 corresponds to an OD600 of B15) were TCAM6 2,400 M8 M7 precipitated and loaded on a 15% SDS-PAGE gel. M9 0 Native samples are indicated with ‘’; samples deglycosylated with PNGase F are indicated with ‘+’. The band marked with an asterisk is PNGase F. The smear marked with ‘#’ is believed to be the result of interfering endogenous mannosyltransferases and incomplete processing by the introduced enzymes; the bracket ([) indicates the extent of smearing. These gels indicate that glycoengineering does not severely decrease mIL-22 yields. Relative fluorescence units


G S1 1

5

ct rl

5m G M -C SF

Figure 4 | DSA-FACE and SDS-PAGE analysis of Glucose units a b medium proteins from mGM-CSF-producing strains. 15 10 5 (a) N-glycan profiles of the proteins present in 1 1,000 unpurified growth medium after small-scale cultivation of these strains (according to 0 – – + – + – + 75 Step 23 in the protocol). Electropherogram 1 M9 50 M8 2 shows the results for a maltodextrose reference. 37 390 M10 * * * Electropherograms 2 through 8 show the results M11 25 0 for N-glycans of the glycoengineered expression 20 strains, as follows: electropherogram 2, 3 1,800 15 GS115mGM-CSF (typical wild-type P. pastoris 0 profile); electropherogram 3, M5mGM-CSF 10 (the predominant peak is Man5GlcNAc2); 4 120 electropherogram 4, GnM5mGM-CSF (the 0 predominant peak is GlcNAcMan5GlcNAc2); electropherogram 5, GalGnM5mGM-CSF (the 5 1,000 predominant peak is GalGlcNAcMan5GlcNAc2); 0 electropherogram 6, GalGnM3mGM-CSF (the 880 main peaks are GalGlcNAcMan3GlcNAc2 and 6 + – + – + – + GalGlcNAcMan4GlcNAc2); electropherogram 7, 440 75 50 Gal2Gn2M3mGM-CSF (the main peak is 0 37 * * * * Gal2GlcNAc2Man3GlcNAc2); electropherogram 8, 7 25 reference N-glycans from bovine RNase B 240 20 (Man5-9GlcNAc2 [M5-M9]). (b) Proteins produced 0 15 by equivalent amounts of yeast cells (B30107; M5 corresponds to an OD600 of B15) were TCA8 M6 M8 10 2,000 M7 precipitated and loaded on a 15% SDS-PAGE gel. M9 0 Native samples are indicated with ‘’; samples deglycosylated with PNGase F are indicated with ‘+.’ The band marked with an asterisk is PNGase F. The arrow indicates non-N-glycosylated mGM-CSF. These gels indicate that glycoengineering process does not severely decrease mGM-CSF yields.


PROTOCOL synthesis. Especially in the case of mIL-22, this causes some smearing on SDS-PAGE (Fig. 5b). As can be concluded from Figures 4b and 5b, the glycoengineering process has no severe negative effect on mGM-CSF and mIL-22 yields.


Conclusion These results indicate that more extensively engineered strains produce a more heterogeneous array of glycoforms due to (i) incomplete processing of several intermediate N-glycan species and (ii) some interference by endogenous mannosyltransferases. Both phenomena appear to be strongly influenced by growth conditions (compare Fig. 3a and b). Therefore, we are at present working on optimizing fermentation conditions to retain the more homogeneous N-glycan profiles, obtained in small-scale culture, after upscaling.

ACKNOWLEDGMENTS We thank Annelies Van Hecke for technical assistance and Dr. Bennet Cohen and Dr. David Bramhill for fruitful discussions. This work was supported by the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen) and Research Corporation Technologies (Tucson, AZ, USA). COMPETING INTERESTS STATEMENT The authors declare competing financial interests (see the HTML version of this article for details). Published online at http://www.natureprotocols.com/ Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/ 1. Cereghino, G.P., Cereghino, J.L., Ilgen, C. & Cregg, J.M. Production of recombinant proteins in fermenter cultures of the yeast Pichia pastoris. Curr. Opin. Biotechnol. 13, 329–332 (2002). 2. Cereghino, J.L. & Cregg, J.M. Heterologous protein expression in the methylotrophic yeast Pichia pastoris. FEMS Microbiol. Rev. 24, 45–66 (2000). 3. Dean, N. Asparagine-linked glycosylation in the yeast Golgi. Biochim. Biophys. Acta. 1426, 309–322 (1999). 4. Chitlaru, T. et al. Modulation of circulatory residence of recombinant acetylcholinesterase through biochemical or genetic manipulation of sialylation levels. Biochem J. 336, (Part 3): 647–658 (1998). 5. Lee, S.J. et al. Mannose receptor-mediated regulation of serum glycoprotein homeostasis. Science 295, 1898–1901 (2002). 6. Weigel, P.H. Galactosyl and N-acetylgalactosaminyl homeostasis: a function for mammalian asialoglycoprotein receptors. Bioessays 16, 519–524 (1994). 7. Winkelhake, J.L. & Nicolson, G.L. Aglycosylantibody. Effects of exoglycosidase treatments on autochthonous antibody survival time in the circulation. J. Biol. Chem. 251, 1074–1080 (1976). 8. Ballou, C.E. Isolation, characterization, and properties of Saccharomyces cerevisiae mnn mutants with nonconditional protein glycosylation defects. Methods Enzymol. 185, 440–470 (1990). 9. Yip, C.L. et al. Cloning and analysis of the Saccharomyces cerevisiae MNN9 and MNN1 genes required for complex glycosylation of secreted proteins. Proc. Natl. Acad. Sci. USA 91, 2723–2727 (1994). 10. Kornfeld, R. & Kornfeld, S. Assembly of asparagine-linked oligosaccharides. Annu. Rev. Biochem. 54, 631–664 (1985). 11. Nagasu, T. et al. Isolation of new temperature-sensitive mutants of Saccharomyces cerevisiae deficient in mannose outer chain elongation. Yeast 8, 535–547 (1992). 12. Nakayama, K., Nagasu, T., Shimma, Y., Kuromitsu, J. & Jigami, Y. OCH1 encodes a novel membrane bound mannosyltransferase: outer chain elongation of asparagine-linked oligosaccharides. Embo J. 11, 2511–2519 (1992). 13. Callewaert, N. et al. Use of HDEL-tagged Trichoderma reesei mannosyl oligosaccharide 1,2-a-D-mannosidase for N-glycan engineering in Pichia pastoris. FEBS Letts. 503, 173–178 (2001). 14. Vervecken, W. et al. In vivo synthesis of mammalian-like, hybrid-type N-glycans in Pichia pastoris. Appl. Environ. Microbiol. 70, 2639–2646 (2004).


15. Hamilton, S.R. et al. Production of complex human glycoproteins in yeast. Science 301, 1244–1246 (2003). 16. Li, H. et al. Optimization of humanized IgGs in glycoengineered Pichia pastoris. Nat. Biotechnol. 24, 210–215 (2006). 17. Hamilton, S.R. et al. Humanization of yeast to produce complex terminally sialylated glycoproteins. Science 313, 1441–1443 (2006). 18. Girbach, V. & Strahl, S. Members of the evolutionary conserved PMT family of protein O-mannosyltransferases form distinct protein complexes among themselves. J. Biol. Chem. 278, 12554–12562 (2003). 19. Kuroda, K. et al. Efficient antibody production upon suppression of Omannosylation in the yeast Ogataea minuta. Appl. Environ. Microbiol. 74, 446–453 (2008). 20. Chiba, Y. et al. Production of human compatible high mannose-type (Man5GlcNAc2) sugar chains in Saccharomyces cerevisiae. J. Biol. Chem. 273, 26298–26304 (1998). 21. Pelham, H.R., Hardwick, K.G. & Lewis, M.J. Sorting of soluble ER proteins in yeast. EMBO J. 7, 1757–1762 (1988). 22. Krezdorn, C.H. et al. Human beta 1,4 galactosyltransferase and alpha 2,6 sialyltransferase expressed in Saccharomyces cerevisiae are retained as active enzymes in the endoplasmic reticulum. Eur. J. Biochem. 220, 809–817 (1994). 23. Yoshida, S. et al. Expression and characterization of rat UDP-Nacetylglucosamine: alpha-3-D-mannoside beta-1,2-Nacetylglucosaminyltransferase I in Saccharomyces cerevisiae. Glycobiology 9, 53–58 (1999). 24. Lussier, M., Sdicu, A.M., Ketela, T. & Bussey, H. Localization and targeting of the Saccharomyces cerevisiae Kre2p/Mnt1p alpha 1,2-mannosyltransferase to a medial-Golgi compartment. J. Cell Biol. 131, 913–927 (1995). 25. Chen, X., Zhang, W., Wang, J., Fang, J. & Wang, P.G. Production of alphagalactosyl epitopes via combined use of two recombinant whole cells harboring UDP-galactose 4-epimerase and alpha-1,3-galactosyltransferase. Biotechnol. Prog. 16, 595–599 (2000). 26. Bobrowicz, P. et al. Engineering of an artificial glycosylation pathway blocked in core oligosaccharide assembly in the yeast Pichia pastoris: production of complex humanized glycoproteins with terminal galactose. Glycobiology 14, 757–766 (2004). 27. Cregg, J.M. & Russel, K.A. In Pichia Protocols Vol. 103. (eds. Higgins DR & Cregg JM) 27–39 (Humana Press, Totowa, NJ, 1998). 28. Peat, S., Whelan, W.J. & Edwards, T.E. Polysaccharides of Baker’s yeast. IV. Mannan. J. Chem. Soc. 1, 29–34 (1961). 29. Laroy, W., Contreras, R. & Callewaert, N. Glycome mapping on DNA sequencing equipment. Nat. Protoc. 1, 397–405 (2006). 30. Sainathan, S.K. et al. PEGylated murine granulocyte-macrophage colonystimulating factor: production, purification, and characterization. Protein Expr. Purif. 44, 94–103 (2005). 31. Choi, B.K. et al. Recombinant human lactoferrin expressed in glycoengineered Pichia pastoris: effect of terminal N-acetylneuraminic acid on in vitro secondary humoral immune response. Glycoconj J. 25, 581–593 (2008).