Journal of Applied Microbiology ISSN 1364-5072

ORIGINAL ARTICLE

Production of recombinant human erythropoietin from Pichia pastoris and its structural analysis
E. C elik1,2, P. C alk1, S.M. Halloran3 and S.G. Oliver2
1 Chemical Engineering Department, Industrial Biotechnology and Metabolic Engineering Laboratory, Middle East Technical University, Ankara, Turkey 2 Faculty of Life Sciences, The University of Manchester, Michael Smith Building, Oxford Road, Manchester, UK 3 Biotechnology Institute, Ankara University, Bes evler, Ankara, Turkey

Keywords AOX1 promoter, erythropoietin, expression, factor Xa, glycan, his-tag, MALDI, Pichia pastoris. Correspondence Pnar C alk, Department of Chemical Engineering, Middle East Technical University, 06531, Ankara, Turkey. E-mail: pcalik@metu.edu.tr

Abstract Aims: To design and investigate a recombinant expression system producing a therapeutically important glycoprotein, human erythropoietin (rHuEPO), by Pichia pastoris. Methods and Results: EPO cDNA was cloned into pPICZaA for expression under control of AOX1 promoter and fused, on the amino-terminal end, with a polyhistidine tag for rapid purication. A target site for factor Xa protease was also introduced, such that cleavage in vitro produced a mature form of rHuEPO having the native N- and C-termini. RHuEPO was characterized as to the extent and nature of N-linked glycosylation using matrix-assisted laser desorption ionization time-of-ight mass spectrometry and western blotting. The rHuEPO produced was approximately 30 kDa. All three N-linked glycosylation sites were occupied dominantly by Man17(GlcNAc)2. N-glycanase-treated rHuEPO puried but not digested with factor-Xa-protease, showed a spectral peak centered about m z 20400 Da. Conclusions: The native polypeptide form of human EPO (c. 18 kDa) was obtained for the rst time in P. pastoris expression system, after afnity purication, deglycosylation and factor-Xa-protease digestion. The amount of sodium dodecyl sulfate used prior to deglycosylation was found to be crucial in determining the dominant form of glycan in glycoproteins. Signicance and Impact of the Study: The novel approaches to protein expression and purication system and structural analysis presented, would be important especially for therapeutic proteins expressed in P. pastoris.

2007 0271: received 21 February 2007, revised and accepted 17 April 2007
doi:10.1111/j.1365-2672.2007.03448.x

Introduction Human protein-based therapeutic drugs are derived from natural biomolecules in the body that have a specic physiological role, such as cellular messengers (hormones), structural components (cytoskeleton), mediators of the cell metabolism (enzymes) and major components of the immune response (antibodies and lymphokines). Erythropoietin (EPO) is a glycoprotein hormone that chiey regulates the production of red blood cells, and is produced primarily by the kidney in the adult and by the liver during fetal life (Jacobson et al. 1957; Zanjani et al.
2084

1977). EPO is composed of 165 amino acids in its mature form, having an average molecular weight of 30 kDa. Three complex type N-glycans located at asparagine (Asn) residues at positions 24, 38 and 83, and a mucin-type Oglycan located at Ser-126 comprise about 40% of the total mass of EPO. The glycosylation of EPO is quite essential to its properties as a hormone, the rst and foremost of these being prolongation of its biological half-life; however, is not required for receptor binding (Narhi et al. 1991). Decreased production of EPO, because of kidney failure, results in anaemia. The use of recombinant human EPO (rHuEPO) has been approved by the U.S.

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

E. C elik et al.

Human erythropoietin from P. pastoris

Food and Drug Administration (FDA), and is now widely used for the treatment of anaemia associated with renal failure, cancer, prematurity, chronic inammatory disease and human immunodeciency virus infection (Jelkmann 1992). The large quantities of the hormone required to satisfy clinical demand are currently met by recombinant expression in mammalian cells, namely Chinese hamster ovary (CHO) cells (Egrie 1990). However, there are signicant disadvantages to the use of mammalian cell cultures with respect to efciency and cost (Fernandez and Hoefer 1999), and there is a strong need to nd alternative systems for the more efcient production of EPO. Production of EPO in bacterial hosts has been reported for Escherichia coli (Leehuang 1984; Bill et al. 1995) and for Bacillus brevis (Nagao et al. 1997), and in a eukaryotic host the bakers yeast, Saccharomyces cerevisiae (Elliott et al. 1989). However, prokaryotes do not possess the ability to glycosylate proteins, while even microbial eukaryotes can effect that process; although, admittedly, S. cerevisiae hyperglycosylates proteins (Walker 1998). Drosophila melanogaster Schneider-2 cells were also recently employed (Kim et al. 2005) in an attempt to achieve a simpler eukaryotic expression system, as an alternative to CHO cells, for production of rHuEPO; the N-glycan structures of rHuEPO so produced were investigated. Pichia pastoris has become popular as a useful alternative to S. cerevisiae, which is the most commonly used and studied yeast for production of heterologous proteins. There are several reasons that account for the popularity of the P. pastoris expression system: its ability to produce foreign proteins at high levels, either extracellularly or intracellularly; its facility in performing many post-translational modications (e.g. glycosylation without the hyperglycosylation of S. cerevisiae, correct disulde bond formation and proteolytic processing); the availability of the alcohol oxidase I (AOX1) promoter (known to be one of the strongest and most tightly regulated eukaryotic promoters) for controlled gene expression; the ability to stably integrate expression plasmids at specic sites in the P. pastoris genome in either single or multiple copies; its ability to grow to a very high cell density in bioreactors (Cereghino and Cregg 1999, 2000). The engineering of P. pastoris has been reported very recently, to produce humanized glycoproteins and it has been shown that a recombinant rat EPO secreted by this particular host was able to raise haematocrit levels in mice whereas the wild-type P. pastoris had much less activity (Hamilton et al. 2006). In this study, an expression system for production of the therapeutically important glycoprotein, recombinant erythropoietin, by P. pastoris was designed and investigated. The cDNA of EPO was fused, on the amino-terminal end, with a polyhistidine tag for rapid purication using

immobilized metal afnity chromatography (IMAC) purication and also fused with a target site for the factor Xa protease, in which cleavage produces a mature form of rHuEPO, strictly having the native N- and C-termini. The recombinant product was then characterized as to the extent and nature of N-linked glycosylation using matrixassisted laser desorption ionization time-of-ight mass spectrometry (MALDI-ToF MS) and western blotting. Materials and methods Construction of the plasmid pPICZaA::epo EPO cDNA was amplied from the E. coli host strain (clone ID IOH44362; Invitrogen, Carlsband, CA, USA) carrying the plasmid pENTRTM221 with a homo sapiens epo ORF (NCBI accession number NM_000799) and antibiotic resistance gene to kanamycin. Primers were designed to amplify the cDNA of EPO by polymerase chain reaction (PCR) from pENTRTM221, and an EcoRI restriction site, 6xHis tag sequence (18 bp) and factor XA recognition sequence (12 bp) were added to the 5 end of epo sequence during amplication. Two forward primers EPO-F3-1 (5CACCATATTGAAGGGAGAGCCCCACCACGCCTCATC3) and EPO-F3-2 (5GGAATTCCACCATCACCATCACCATATTGAAGGGAG3) were designed instead of a single long primer, as an addition of 36 bases was required at the 5-end of epo sequence. The reverse primer EPO-R3 (5CCACGCTCTAGATTAGTCCCCTGTCCTGC3) was designed such that, at the 3-end of the epo sequence, a stop codon and XbaI restriction site were present. Primers were synthesized by MWG Biotech (Ebersberg, Germany). The shuttle vector pPICZaA (Invitrogen) was propagated in E. coli TOP10 (Invitrogen) chemically competent host cells, grown in low-salt LuriaBertani (LSLB) medium containing (in g l)1) tryptone, 10; yeast extract, 5; NaCl, 5; Zeocin, 0025 (Invitrogen) and puried. The DNA fragment (EcoRI restriction site + 6xHis tag + factor Xa recognition sequence + epo + stop codon + XbaI restriction site; scheme in Fig. 1) puried from PCR amplication and the puried vector pPICZaA were double-digested with EcoRI and XbaI (Roche, Mannheim, Germany). Puried digestion products were used in ligation reaction (Sambrook and Russell 2001) and 5-ll aliquot of the reaction product was used to transform E. coli TOP10 cells, according to the manufacturers instructions. Plasmid isolated from Zeocin-resistant colonies was evaluated by restriction enzyme digestion analysis, by PCR amplication using the primers 5AOX1 (5GACTGGTTCCAATTGACAAGC3) and 3AOX1 (5GCAAATGGCATTCTGACATCC3) (Invitrogen), and by DNA sequencing in ABI 377 uorescent sequencer (Applied Biosystems, Foster City, CA, USA).
2085

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

Human erythropoietin from P. pastoris

E. C elik et al.

Recombinant protein production

pENTR221::EPO EPO-F3-1 EPO EPO-F3-2 EPO-R3 1st PCR Factor Xa EPO 2nd PCR Stop xba I site EPO-R3

2 x His

EcoR I site

6 x His

Factor Xa

EPO

Stop

xba I site

Digestion and ligation reactions

xba I
c-myc epitope 6 x His 3' AOX1 Stop

EcoR I
-factor

6 x His Factor Xa EPO Stop

5' AOX1

AO X1

TT

Bam H I

pPICZA::epo
41 kb

Sac I Bgl II
PUC ori
cy TT c1

RHuEPO-producing P. pastoris E17 strain was streaked onto YPD solid medium containing (in g l)1) peptone, 20; yeast extract, 10; glucose, 20; agar, 20; Zeocin, 0100, and incubated for 48 h at 30C. A single colony was inoculated into 5 ml of YPD + 100 lg ml)1 of Zeocin medium, and grown shaking at 225 rev min)1, 30C overnight. The seed culture was then inoculated into buffered complex glycerol medium (BMGY) containing (in g l)1) yeast extract, 10; peptone, 20; yeast nitrogen base (YNB), 134; biotin, 4 10)5; glycerol, 10 and 01 mol l)1 of potassium phosphate buffer pH 60 and growth allowed to proceed until OD600 was about 6 (Cx  25 g l)1). Cells from the medium were collected by centrifugation and the pellet resuspended in BMMY production medium (BMGY medium containing 5 ml l)1 of methanol instead of glycerol) such that OD600 was about 1. Protein production was carried out in batch cultures using bafed V = 250-ml Erlenmeyer asks containing VR = 50 ml of production medium. Recombinant protein production was continued for 72 h in total, and every 24 h methanol was added to the medium to a 5 ml l)1 nal concentration. At the end of 72 h, the medium was centrifuged at 13000 g for 10 min and the cell pellet was discarded. Purication of rHuEPO Concentration and desalting of the production medium was achieved by ultraltration under nitrogen gas (55 psi, 38 bar) at 4C using Amicon 400-ml stirred pressure cells (Millipore, Bedford, MA, USA) with regenerated cellulose ultraltration membranes having MWCO of 10 kDa (Millipore). The polyhistidine-tagged rHuEPO was puried from the concentrated medium using cobalt-based metal afnity resins (BD Talon; BD Biosciences, Palo Alto, CA, USA). Deglycosylation of puried rHuEPO For digestion of the recombinant glycoprotein, to release its N-linked glycans, 50200 lg of puried rHuEPO dissolved in 45 ll of reaction buffer (10 mmol l)1 of TrisHCl, pH 80) and 25 ll of denaturation solution [02% of sodium dodecyl sulfate (SDS), 01 mol l)1 of b-mercaptoethanol] was added. After denaturation at 100C for 5 min, 25 ll of 15% NP-40 and 5 mU of N-Glycanase (Glyko, Novato, CA, USA) were added to the reaction mixture and incubated overnight at 37C. Factor Xa digestion After deglycosylation, approximately 10 lg of rHuEPO was digested at 25C for 16 h using 1 U of factor Xa

P TE

Figure 1 Schematic representation of epo amplication, integration of specic recognition sites by two-step polymerase chain reaction (PCR) using degenerate primers, and construction of pPICZaA::epo plasmid. The EcoRI and XbaI sites used in ligation of the insert to the vector, the SacI site used to linearize the plasmid before transformation and the primers used for sequencing are also shown. There are 894 nucleotides between the 5AOX1 primer and EPO-R3 primer, and 708 nucleotides between the EPO-F3-2 primer and 3AOX1 primer.

Transfection of Pichia pastoris Puried pPICZaA::epo plasmid was digested with SacI (Roche), and after verifying complete digestion by agarose gel electrophoresis, the plasmid was used for transfection of P. pastoris wild-type host strain X-33 from Invitrogen using lithium chloride transformation method. Selection of the potential rHuEPO-producing strain After transfection according to a method described by Burke et al. (2000), genomic DNA was isolated from yeast colonies resistant to 02 g l)1 of Zeocin and characterized by PCR and southern blotting (Sambrook and Russell 2001) using the alkaline phosphatase-labelled epo insert (512 bp) as the probe, and further by western blot analysis for selection of the true transformants.

2086

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

Zeo

cin

5' AO X1

F1

P EM7

E. C elik et al.

Human erythropoietin from P. pastoris

protease (Qiagen, Valencia, CA, USA), which targets the Ile-Glu-Gly-Arg sequence. The protease was removed using Xa removal resin provided with the enzyme, according to the manufacturers instructions. Cell and protein concentrations Cell concentrations based on dry weights were measured with a ultraviolet-visible (UV-Vis) spectrophotometer (Thermo Spectronic, Hekios-a, Waltham, MA, USA) using a calibration curve obtained at 600 nm. Total protein concentrations were determined using the Bradford (1976) assay. SDS-PAGE and western blotting SDS-polyacrylamide gel electrophoresis (PAGE) (Laemmli 1970) was performed in a 12% separating gel with a 5% stacking gel using the Mini-PROTEAN-3 apparatus (BioRad, Hercules, CA, USA). Proteins were either silver-stained or transferred from gel to Immobilon-P polyvinylidene uoride (PVDF) nylon membranes (Millipore) by electroblotting. Immunodetection was achieved by using monoclonal antihuman EPO antibody (R&D Systems, Minneapolis, MN, USA) as primary antibody and antimouse IgG horseradish peroxidaselinked whole antibody (Amersham Biosciences, Uppsala, Sweden) as secondary antibody. The bands were visualized using SuperSignal West Pico chemiluminescent substrate (Pierce, Rockford, IL, USA) with exposure to Biomax MR lm (Eastman Kodak, Rochester, NY, USA). Mass spectrometry analysis Protein identication was made by producing a tryptic digest of protein corresponding to rHuEPO excised from Coomassie Brilliant Blue G-250-stained SDS-PAGE gel, and then peptide mass ngerprints were determined on a Q-ToF mass spectrometer using electrospray ionization (Waters-Micromass, Milford, MA, USA). The molecular weights of the glycans, of intact rHuEPO and of deglycosylated rHuEPO were analysed by the use of a MALDI-LR (Waters-Micromass) instrument. All spectra were generated using a pulsed nitrogen gas laser (337 nm) in positive-ion mode with built-in delayed extraction. Accelerating voltage was 15 kV. Deglycosylated protein was precipitated by the addition of ice-cold eth ster et al. 1997). The glycans, dissolved in the anol (Ku supernatant, were recovered and dried by centrifugal evaporator. All samples were desalted by drop dialysis prior to analysis (Gorisch 1988). Exactly 3 ll of

10 mg ml)1 of matrix dissolved in 50% acetonitrile and 01% triuoroacetic acid (TFA) solution, was mixed with 1 ll of c. 10 pmol ll)1 of sample and 1 ll of this mixture was spotted on target plate and air dried (dried droplet technique; Karas and Hillenkamp 1988). Exactly 05 ll of ethanol was used for recrystallization on the plate, in case of glycan samples. Sinapinic acid matrix was used for protein samples and Super-DHB (Sigma, St Louis, MO, USA) matrix for glycan samples. For detection of proteins, linear mode with a low mass gate of 1000 Da and for glycans, reectron mode with a low mass gate of 800 Da was used. Man9(GlcNAc)2 (Sigma) oligosaccharide and intact protein cytochrome c were used as external molecular weight standards. Spectra were generated from the sum of 100200 laser pulses and mass determinations were made by nding the peak centroid of a smoothed signal (Savitzky-Golay algorithm) after background subtraction. Results Construction of the plasmid pPICZaA-epo Plasmid extracted from E. coli IOH44362 cells was used in PCR amplication of EPO cDNA, using the primer set EPO-F3-1 and EPO-R3. The puried PCR product was used as template in the second set of PCR amplications using the primer set EPO-F3-2 and EPO-R3. The aim of this reaction was to elongate the 5-end of the epo gene (Fig. 1). EcoRI and XbaI were used in double digestion of the insert and vector DNA (pPICZaA) for correctly orienting the insert into the vector in-frame with the a-factor secretion signal. The putative recombinant plasmids were transformed into E. coli TOP10. Twelve single colonies were randomly selected for characterization of the transformed plasmid. Six of the 12 colonies were shown to have an insert having the expected size when digested with the same enzymes used to clone the insert. PCR products were amplied from these six clones using the EPO-F3-2 and EPO-R3 primers, and all six colonies produced products of the expected 528-bp size. Plasmids from three of the six clones were selected for DNA sequencing using 5AOX1 and 3AOX1 primers. One of the three clones was found to have the exact sequence when analysed by Basic Local Alignment Search Tool (BLAST) sequence alignment. The sequencing primer set was sufcient to overlap and conrm the nucleotide sequence of the insert, as well as the a-factor signal sequence in the 5-end of the insert. The colony that gave the exact match of sequence was named as E. coli pPICZaA-EPO and its plasmid was named as pPICZaA::epo (Fig. 1).
2087

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

Human erythropoietin from P. pastoris

E. C elik et al.

894 bp 1000 800 600

708 bp

Figure 2 Agarose gel electrophoresis view for verication of epo insertion to Pichia pastoris genome. Polymerase chain reaction (PCR) amplication performed using P. pastoris genomic DNA as template and primer set 5AOX1 and EPO-R3, lane 1; EPO-F3-2 and 3AOX1, lane 2; Hyperladder I (Bioline), lane M.

showing the most intense band in the western analysis (data not shown) was selected for its production potential and was named P. pastoris E-17. Recombinant protein expressed and puried from this strain was then digested with trypsin and the peptides were analysed using a Q-ToF mass spectrometer with an electrospray ionization source to generate a peptide mass ngerprint. The predicted fragment products of the trypsin digestion reaction of rHuEPO are shown in Fig. 3. The peptide fragments underlined in the gure were those identied by Q-ToF MS, based on a comparison of the calculated monoisotopic masses of the fragments with the measured masses obtained from the mass spectrum, typically accurate to 78 signicant digits. The peptides identied, have sequence coverage of 26% (i.e. 26% of the protein sequence was matched to the peptides detected in MS). More importantly, the peptide fragments detected included peptides close to both the N- and C-termini ends of the protein, indicating that expression is of the complete epo open reading frame. Production, purication and obtaining the native rHuEPO Cell concentrations of the BMMY medium-grown yeast during the production phase of P. pastoris cultivation reached 8 g cell dry weight l)1 at the end of 72 h, and the rHuEPO produced under these conditions was >5 mg l)1, as calculated from Bradford assay of the afnity-puried supernatant, assuming 100% purication yields. The rst step of the purication process after cell growth and expression of the secreted recombinant protein was the concentration and desalting of the supernatant from the P. pastoris production medium. Then, the concentrated sample was mixed with cobalt-based metal afnity resins and binding was carried out overnight to assure completion. The puried rHuEPO (Fig. 4a, lane 1; b, lane 1) was then deglycosylated. To obtain a concentration of 5 lg of glycoprotein ll)1 reaction buffer for the N-glycanase digestion, a buffer exchange was required and ttingen, ultraltration spin columns (Sartorius AG, Go Germany) with a 10-kDa cut-off were used. The deglycos-

Transformation of Pichia pastoris and selection of the potential rHuEPO production strain In order to obtain integration of the pPICZaA::epo plasmid into the P. pastoris genome at the AOX1 locus, the plasmid was linearized by SacI digestion in its AOX1 promoter region. The digestion product was puried and transformed into P. pastoris, Zeocin-resistant single colonies being obtained at a frequency of 160 CFU lg)1 of DNA. Genomic DNA, isolated from these colonies, was characterized by PCR and southern blotting. Integration of epo gene into the P. pastoris genome at the AOX1 promoter site was veried by PCR (Fig. 2a) and southern blotting using the epo insert (512 bp) as the probe. All 12 colonies generated PCR products of the expected sizes and ve of these were conrmed by southern blotting, giving a band with the same size as the positive control (data not shown). These ve clones were then grown for protein expression and the recombinant proteins from clones were puried by metal afnity chromatography, run in SDS-PAGE and transferred to a membrane for immunoblotting. The strain

Figure 3 Erythropoietin (EPO) amino acid sequence. The bold letters (r and k) in the 165 amino acid EPO sequence indicate the expected trypsin digestion sites. (Trypsin specically hydrolyses the carboxyl-terminal of arginine and lysine residues which do not precede proline residues.) The underlined peptide fragments are the ones detected in the Q-time-of-ight mass spectrometry analysis. The star (*) after asparagine (n) residue indicates the glycosylation sites in EPO.
2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

2088

E. C elik et al.

Human erythropoietin from P. pastoris

(a) 1 2 M1 3 4 M1

(b) 3 2 1 M2

43 kDa 34 kDa 37 kDa 26 kDa 25 kDa 20 kDa 17 kDa 18 kDa 15 kDa

Figure 4 (a) Silver-stained sodium dodecyl sulfate-polyacrylamide gel electrophoretic (SDS-PAGE) view of recombinant human erythropoietin (rHuEPO) produced by Pichia pastoris E17. Lane 1, polyhistidine tag afnity-puried rHuEPO from P. pastoris E17 production medium; lane 2, puried and factor Xa-digested rHuEPO; lane 3, puried and deglycosylated rHuEPO; lane 4, puried, deglycosylated, and factor Xa-digested rHuEPO; M1, PageRuler Prestained Protein Ladder (Fermentas, Germany). (b) Western blot view of rHuEPO produced by P. pastoris E17. Lane 1, polyhistidine tag afnity-puried rHuEPO from P. pastoris E17 production medium; lane 2, puried and deglycosylated rHuEPO; lane 3, puried, deglycosylated and factor Xa-digested rHuEPO; M2, positions of protein size marker (Presicion Plus Dual-Colour Prestained Protein Marker; BioRad, Hercules, CA, USA).

ylated rHuEPO can be seen in Fig. 4a (lane 3) and b (lane 2), as having an apparent molecular weight of c. 20 kDa. After removal of the 6xHis tag and the factor Xa recognition sequence by the factor Xa protease, the digest mixture is passed through a special resin that efciently removes the protease. This eluate is then passed through the metal afnity resin to remove proteins not digested by factor Xa protease and the peptides containing the 6xHis tag. The nally puried, deglycosylated and factor Xa-digested rHuEPO, corresponding to the wild-type polypeptide was analysed by SDS-PAGE (Fig. 4a, lane 4) and western blotting (Fig. 4b, lane 3), and showed an apparent molecular mass of the expected size around 18 kDa. The relatively lighter bands seen in silver-stained gel electrophoresis view are the incompletely digested proteins, which can be removed with prolonged digestions, and unspecically bound proteins in polyhistidine tag afnity purication, which can be removed by step elution. Structural analysis RHuEPO glycosylation by the P. pastoris expression system was characterized by MALDI-ToF MS. Neutral glycans produce predominantly [M + Na]+ ions during MALDI ionization (Harvey 1999). The most intense isotopic peak cluster seen in the mass spectrum of the glycans of rHuEPO shown in Fig. 5a is m z 3204, which corresponds to Man17(GlcNAc)2, and is denoted as M17. The isotopes in the cluster are separated by 1 unit on the m z axis, z = 1 and m = 3204 Da, which is practically the same as the calculated [M + Na]+ ion mass, 32036 Da. Glycosylation is known to be a post-translational modi-

cation with considerable microheterogeneity, meaning that a glycoprotein is synthesized as a group of glycoforms (Papac et al. 1996). So, as expected, there are also
(a) 100 M17 3204

M10 M9 1906 2068 M11 2230

(b) 100

M10 M9 2068 1906

M11 2230

0 1800 2000 2200 2400 2600 m/z 2800 3000 3200

Figure 5 Matrix-assisted laser desorption ionization (MALDI) mass spectra of glycans of recombinant human erythropoietin (rHuEPO) (released by N-glycanase treatment) on a Super-DHB matrix detected in [M + Na]+ form. Final concentration of sodium dodecyl sulfate (SDS) used in protein denaturation before deglycosylation reaction: (a) 001%; (b) 01%; M9, Man9(GlcNAc)2; M10, Man10(GlcNAc)2; M11, Man11(GlcNAc)2; M17, Man17(GlcNAc)2.

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

2089

Human erythropoietin from P. pastoris

E. C elik et al.

peaks present in the mass spectrum of relatively lower mass, corresponding to other high mannose-type glycans, particularly M9, M10 and M11. The presence of SDS, used for the purpose of unfolding the glycoprotein substrate so that the N-glycanase has greater access, had the unwanted effect of suppressing the ionization of the dominant glycan form in MALDI spectral analysis. When the nal concentration of SDS in the deglycosylation reaction was 01%, which accorded to the manufacturers recommendations and has been used in many other studies, the glycans M911 were detected but M17 was never seen (Fig. 5b). However, when 10-fold less SDS was used in the deglycosylation reaction (001% SDS of nal concentration), the most spectrally intense glycan detected was M17 (Fig. 5a). Although in both cases drop dialysis was carried out for 45 min to remove any low molecular weight substances including SDS, the results show that the higher mass glycan form was nonetheless undetectable by MALDI-ToF MS. MALDI MS analysis of the N-glycanase-treated rHuEPO puried from the medium of the recombinant P. pastoris (not digested with factor Xa protease) shows a spectral peak centered about m z 20400 (Fig. 6). Production of singly charged (z = 1) ions of intact proteins is typical in MALDI ionization, and so this m z value agrees quite well with the 20 kDa apparent molecular weight of the band detected in a western blot loaded with N-glycanase-treated rHuEPO (Fig. 4b, lane 2). A relatively low intensity peak in the spectrum, present at m z 29700 (Fig. 6), would correspond well with a small amount of intact rHuEPO glycoprotein that had not been deglycosyl-

100

10 022

9273
%

0 10 000 15 000 20 000 25 000 m/z 30 000 35 000

Figure 7 Matrix-assisted laser desorption ionization time-of-ight (MALDI-ToF) mass spectrum of the native recombinant human erythropoietin (rHuEPO), deglycosylated and factor Xa protease-digested rHuEPO (m z = 9273) on sinapinic acid matrix, detected in [M + 2H]2+ form. Because of incomplete digestion reaction with factor Xa protease, the [M + 2H]2+ form of only deglycosylated rHuEPO is also detected (m z = 10 022).

ated with N-glycanase treatment (Fig. 4, lane 1). The spectral peak at m z 35740 (Fig. 6) likely corresponds to the mass of the N-glycanase (PNGase F) itself, which has a reported molecular weight of 35 500 Da (Tarentino et al. 1985). Thus, the enzyme was detected with 07% error and therefore served as an internal mass calibrant. Mass spectrum of the deglycosylated and factor Xadigested rHuEPO is shown in Fig. 7. The major ionization form of the protein with the sinapinic acid matrix used, was the [M + 2H]2+ form, thus z = 2 with a peak at m z position of 92 kDa, which means m is around 184 kDa. Discussion

100

20 376

35 740

26 650 0 20 000 25 000 m/z

29 690

30 000

35 000

Figure 6 Matrix-assisted laser desorption ionization time-of-ight (MALDI-ToF) mass spectrum of the deglycosylated recombinant human erythropoietin (rHuEPO) (m z = 20 376) on sinapinic acid matrix. Because of incomplete reaction, the glycosylated rHuEPO (m z = 29 690) and partially deglycosylated rHuEPO (m z = 26 650) were also detected. The peak at m z 35 740 is the N-glycanase enzyme left from deglycosylation, which also serves as an internal mass calibrant. All the molecules detected are in singly charged [M + H]+ form.

In this study, the coding sequence of the therapeutically important human glycoprotein EPO was fused with a polyhistidine tag to enable easy, rapid purication of an expressed protein product, and this construct was cloned into P. pastoris. Factor Xa protease substrate consensus sequence on the amino terminus of the wild-type epo sequence was also included in the cassette, to enable easy removal of all polypeptide elements that are not part of the native EPO sequence, and used as part of the nal step of in vitro maturation. The P. pastoris strain nally selected (E17) reached 8 g cell dry weight l)1 at the end of the 72-h production period. This is much lower than the high cell density fermentation claimed for P. pastoris, which is 130 g cell dry weight l)1 (Wegner 1990). This is probably because of uncontrolled nature of shake-ask experiments compared with bioreactors, where methanol and oxygen levels can be tightly controlled. Controlled environment of a bioreactor is preferred because methanol metabolism utilizes oxygen at a high rate and excess

2090

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

E. C elik et al.

Human erythropoietin from P. pastoris

methanol is toxic to the cells (Cregg et al. 2000). The rHuEPO produced under these conditions was >5 mg l)1, while reported recombinant protein production capacities by P. pastoris cells vary from 1 to 1000 mg l)1 (Cereghino and Cregg 2000). On the other hand, rEPO secreted from other nonmammalian expression systems reported were 003 mg l)1 (Elliott et al. 1989), 18 mg l)1 (Kim et al. 2005) and 20 mg l)1 (Hamilton et al. 2006). The results suggest that, through further optimization of medium and bioreactor operation conditions, much higher cell concentrations (and therefore, higher amounts of recombinant protein) could be obtained. Moreover, purication protocols can be optimized to obtain higher yields. Figure 1 shows the synthesis of a recombinant construct that will produce a protein secreted by P. pastoris not having the native sequence because it possesses elements (polyhistidine tag) used to enable its rapid purication and isolation. The C-terminal polyhistidine tag already present in the pPICZaA vector was not employed because its removal using known methods would not achieve the goal of producing a mature recombinant protein that was the native form of EPO, but would leave extra amino acids. For this reason, a factor Xa recognition site was placed immediately at the N-terminal end of EPO, which enabled use of this protease to digest the non-native N-terminal region of EPO after its secretion and afnity purication. The shift in the size of intact rHuEPO when digested with factor Xa protease can be seen from lanes 1 and 2 in Fig. 4a. Moreover, the nally puried, deglycosylated and factor Xa-digested rHuEPO, corresponding to the wild-type polypeptide, when analysed by SDS-PAGE and western blotting, showed an apparent molecular mass of the expected size, 18 kDa. The constructed expression plasmid thus provides the potential of large-scale production of recombinant human EPO and its processing to give the native form; more advantageous compared with the studies, where rat EPO cDNA cloned directly into pPICZaA for expression in recombinant P. pastoris (Hamilton et al. 2006), and where rEPO expressed in Drosophila S2 cells, could not be obtained in native form following afnity purication, as the fusion construct including the polyhistidine tag was not digestable. Because of the fact that the factor Xa protease hydrolyses the polypeptide on the carboxyl terminal side of the four-residue specic recognition sequence (Ile-Glu-GlyArg), it is a better strategy to place the factor Xa protease recognition sequence on the amino-terminal side of any wild-type polypeptide sequence when the protease is intended to be used to affect the nal step in maturation. This makes it possible to leave no residues at the N-terminus, following digestion, that are not part of the native sequence. On the other hand, the C-terminal end of EPO

was as it should be in the wild-type and, therefore, required no further treatment. Moreover, even if the factor Xa protease recognition sequence was placed at the Cterminal end of EPO, because of hydrolysation of the sequence on the carboxyl-terminal, there would have been four amino acid residues left instead of the six histidine residues. Remnants of a few residues following recombinant production and maturation processes may not signicantly affect the biological function or physiological role of a very large protein; however, smaller proteins (and, certainly, peptides) with signicant biological roles can have their functions abolished or altered by the presence of such remnant amino acids. EPO is a modestly sized therapeutic protein, and so these residues may have a signicant impact on its function or stability, including halflife. Another convenience of using factor Xa protease cleavage is that its recognition sequence (Ile-Glu-Gly-Arg) is relatively short, which makes it easy for its DNA sequence to be integrated as part of the PCR primers. EPO is a naturally heavily glycosylated (40% of total mass) protein, and the glycan structure is important for its structure, function and stability (biological half-life) (Higuchi et al. 1992). RHuEPO produced by P. pastoris E17 in this study is also glycosylated by the post-translational biosynthetic processes of the yeast host, but these glycosylating modications differ from those in human cells. The difculty in glycoprotein analysis is that a large number of techniques are required for a full characterization of the structure (Anumula 2000) and there are many ways to characterize the glycosylation of glycoproteins, including their structure. MALDI-ToF MS and electrospray ionization mass spectrometry (ESI-MS) are the most sensitive and often used methods of mass measurement analysis because of their soft ionization techniques (Morelle and Michalski 2005). MALDI-ToF MS is the pre-eminent technique for glycan analysis because it is more sensitive compared with ESI-MS (Dell and Morris 2001); unlike many other forms of analysis, spectra can be obtained from unmodied glycans (Harvey 1999), and only picomole amounts of oligosaccharide sample are ster et al. 1997). Therefore, the information needed (Ku involving the amount of N-linked glycans and the level of mannosylation of rHuEPO were simply determined by analysing the products of the deglycosylation reaction using MALDI-ToF MS. All three N-linked glycosylation sites were found to be occupied dominantly by Man17(GlcNAc)2. It has been shown that there is a linear relationship between the signal produced by the [M + Na]+ ion from neutral glycans and the amount applied to the target, irrespective of the structure (Harvey 1993, 2005), unlike the case for peptides and glycopeptides, where signal intensity depends on the proton afnity of the compound (Naven and Harvey
2091

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

Human erythropoietin from P. pastoris

E. C elik et al.

1996; Harvey 2005). Therefore, Man17(GlcNAc)2 (32 kDa) can be said to be the dominant form of Nlinked glycosylation in rHuEPO, as the highest signal in Fig. 5a is obtained from M17. Pichia pastoris is known to produce oligosaccharides with 818 mannose residues as the major modication component (Hirose et al. 2002). Further, with the use of other wild-type P. pastoris strain (NRRL-Y11430) producing rHuEPO, 913 mannose residues were detected (Hamilton et al. 2006). The RHuEPO produced by P. pastoris E17 had a molecular weight of c. 30 kDa, and the deglycosylated form was c. 20 kDa, which were detected both by western blotting and MALDI-ToF MS. As EPO has three possible sites for Nglycosylation, it is most probable that each N-glycan has a molecular weight of about 3132 kDa, as derived from the molecular weight differences between the glycosylated(c. 30 kDa; Fig. 4b, lane 1 and Fig. 6) and deglycosylated(c. 205 kDa; Fig. 4b, lane 2 and Fig. 6) rHuEPO, thus also proving that Man17(GlcNAc)2 should be the dominant form of N-linked glycosylation in rHuEPO. Moreover, the second low-intensity peak in Fig. 6 is most likely the partially deglycosylated rHuEPO, which has lost one of its N-linked glycans, as the molecular weight difference between the intact protein of 297 kDa and that of 266 kDa peak corresponds to the molecular weight of a single N-linked glycan, perhaps M17. The molecular weight difference between the deglycosylated rHuEPO (c. 205 kDa) observed in MALDI and of the native protein (c. 18 kDa) is attributable to the amino acid residues on the N-terminal portion of the polypeptide that had not been removed by factor Xa protease digestion, as shown in Fig. 4. On the other hand, from MALDI MS spectrum of the native polypeptide of rHuEPO in Fig. 7, the mass was veried to be around 184 kDa, calculated from the [M + 2H]2+ peak at m z position of 92 kDa, which means z = 2 and m = 184 kDa. This result is in agreement with the ndings of Stanley and Poljak (2003), where they analysed deglycosylated EPREX (commercial rHuEPO, expressed in mammalian cells), and also correlates well with the SDS-PAGE and western data in Fig. 4. Moreover, the larger peak at m z position of 10 kDa is likely the [M + 2H]2+ form of rHuEPO which has been incompletely digested with factor Xa protease, thus when multiplied by the charge, z = 2, mass of 20 kDa corresponds well with the 20-kDa peak in Fig. 6. Having detected the multiply charged molecules is advantageous, as these values are more accurate than those of the singly biger et al. charged ions because of better statistics (Stu 2005), where the charge on the molecule is dependent on the choice of the matrix and its interaction with the analyte.

Interestingly, the amount of SDS used prior to deglycosylation was found to be crucial in determining the dominant form of glycan. This is probably because of suppression of ionization by SDS, which cannot be easily removed after the deglycosylation reaction. Moreover, according to the calculations, M17 should be the dominant form of glycan present in rHuEPO, because the molecular weight difference between the glycosylated- and deglycosylated-rHuEPO is more than 9 kDa and when divided between three possible sites for N-glycosylation, M911 (with molecular weights between 19 and 22 kDa) would not be expected. This result demonstrates the importance of considering the possibility that the absence of expected spectral peaks in MALDI MS might be because of effects of ion suppression that cannot be entirely overcome by standard sample preparation or clean-up. Thus, the already complicated glycosylation proles of many glycoproteins could be determined incorrectly because of the presence of ionization inhibitors like SDS or other salts. The O-linked glycans of yeasts are relatively small (i.e. Man16 in S. cerevisiae); they cannot be digested by enzymatic methods and, moreover, the chemical methods employed for digestion may cause a peeling reaction that destroys the oligosaccharide species (Gemmill and Trimble 1999). Thus, little has been reported concerning the O-glycans of P. pastoris, and they were not analysed in this study. Furthermore, their role in functioning and stability of EPO in human has not been dened and is, perhaps, not crucial. Acknowledgements The authors thank Dr Jill Wishart and Dr Nianshu Zhang for their helpful discussions. Technical services provided by the Biomolecular Analysis Facility and the Sequencing Facility at The University of Manchester and by Ankara University, Biotechnology Institute, Proteomics Unit for MALDI TOF analysis are gratefully acknowledged. This work was supported by the Middle East Technical University Research Fund Project (BAP 2005-03-04-02) and the infrastructure used in the experiments by SPO (Turkey) Grants 2001K121030. E. C elik was awarded scholarships by the Scientic and Technical Research Council of Turkey (TUBITAK-B_ IDEB; 2211 and NATO-A2). References
Anumula, K.R. (2000) High-sensitivity and high-resolution methods for glycoprotein analysis. Anal Biochem 283, 1726.

2092

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

E. C elik et al.

Human erythropoietin from P. pastoris

Bill, R.M., Winter, P.C., McHale, C.M., Hoddges, Y.M., Elder, G.H., Caley, J., Flitsch, S.L., Bicknell, R., et al. (1995) Expression and mutagenesis of recombinant human and murine erythropoietins in Escherichia coli. Biochim Biophys Acta 1261, 3543. Bradford, M.M. (1976) A rapid and sensitive for the quantitation of microgram quantities of protein utilizing the principle of protein dye binding. Anal Biochem 72, 248254. Burke, D., Dawson, D. and Stearns, T. (2000) Methods in Yeast Genetics. New York: Cold Spring Harbor Laboratory Press. Cereghino, G.P.L. and Cregg, J.M. (1999) Applications of yeast in biotechnology: protein production and genetic analysis. Curr Opin Biotech 10, 422427. Cereghino, G.P.L. and Cregg, J.M. (2000) Heterologous protein expression in the methylotrophic yeast Pichia pastoris. FEMS Microbiol Rev 24, 4566. Cregg, J.M., Cereghino, J.L, Shi, J.Y. and Higgins, D.R. (2000) Recombinant protein expression in Pichia pastoris. Mol Biotech 16, 2352. Dell, A. and Morris, H.R. (2001) Glycoprotein structure determination by mass spectrometry. Science 291, 23512356. Egrie, J. (1990) The cloning and production of recombinant human erythropoietin. Pharmacotherapy 10, 3S8S. Elliott, S., Gifn, J., Suggs, S., Lau, E.P.L. and Banks, A.R. (1989) Secretion of glycosylated human erythropoietin from yeast directed by the a-factor leader region. Gene 79, 167180. Fernandez, J.M. and Hoefer, J.P. (1999) Gene Expression Systems: Using Nature for the Art of Expression. San Diego: Academic Press Inc. Gemmill, T.R. and Trimble, R.B. (1999) Overview of N- and O-linked oligosaccharide structures found in various yeast species. Biochim Biophys Acta General Subjects 1426, 227237. Gorisch, H. (1988) Drop dialysis time course of salt and protein exchange. Anal Biochem 173, 393398. Hamilton, S.R., Davidson, R.C., Sethuraman, N., Nett, J.H., Jiang, Y.W., Rios, S., Bobrowicz, P., Stadheim, T.A., et al. (2006) Humanization of yeast to produce complex terminally sialylated glycoproteins. Science 313, 1441 1443. Harvey, D.J. (1993) Quantitative aspects of the matrix-assisted laser-desorption mass-spectrometry of complex oligosaccharides. Rapid Comm Mass Spec 7, 614619. Harvey, D.J. (1999) Matrix-assisted laser desorption ionization mass spectrometry of carbohydrates. Mass Spec Rev 18, 349450. Harvey, D.J. (2005) Structural determination of N-linked glycans by matrix-assisted laser desorption ionization and electrospray ionization mass spectrometry. Proteomics 5, 17741786. Higuchi, M., Oheda, M., Kuboniwa, H., Tomonoh, K., Shimonaka, Y. and Ochi, N. (1992) Role of sugar chains in the expression of the biological activity of human erythropoietin. J Biol Chem 267, 77037709.

Hirose, M., Kameyama, S. and Ohi, H. (2002) Characterization of N-linked oligosaccharides attached to recombinant human antithrombin expressed in the yeast Pichia pastoris. Yeast 19, 11911202. Jacobson, L.O., Goldwasser, E. and Fried, W. (1957) Role of the kidney in erythropoiesis. Nature 179, 633634. Jelkmann, W. (1992) Erythropoietin: structure, control of production, and function. Physiological Rev 72, 449489. Karas, M. and Hillenkamp, F. (1988) Laser desorption ionization of proteins with molecular masses exceeding 10000 daltons. Anal Chem 60, 22992301. Kim, Y.K., Shin, H.S., Tomiya, N., Lee, Y.C., Betenbaugh, M.J. and Cha, H.J. (2005) Production and N-glycan analysis of secreted human erythropoietin glycoprotein in stably transfected Drosophila S2 cells. Biotech Bioeng 92, 452461. ster, B., Wheeler, S.F., Hunter, A.P., Dwek, R.A. and Ku Harvey, D.J. (1997) Sequencing of N-linked oligosaccharides directly from protein gels: in-gel deglycosylation followed by matrix-assisted laser desorption ionization mass spectrometry and normal-phase high-performance liquid chromatography. Anal Biochem 250, 82101. Laemmli, U.K. (1970) Cleavage of structural proteins during assembly of head of bacteriophage-T4. Nature 227, 680. Leehuang, S. (1984) Cloning and expression of human erythropoietin cDNA in Escherichia coli. P Natl Acad Sci Biol 81, 27082712. Morelle, W. and Michalski, J.C. (2005) The mass spectrometric analysis of glycoproteins and their glycan structures. Curr Anal Chem 1, 2957. Nagao, M., Inoue, K., Moon, S.K., Masuda, S., Takagi, H., Udaka, S. and Sasaki, R. (1997) Secretory production of erythropoietin and the extracellular domain of the erythropoietin receptor by Bacillus brevis: afnity purication and characterization. Biosci Biotechnol Biochem 61, 670674. Narhi, L.O., Arakawa, T., Aokii, K.H., Elmore, R., Rohde, M.F., Boone, T. and Strickland, T.W. (1991) The effect of carbohydrate on the structure and stability of erythropoietin. J Biol Chem 266, 2302223026. Naven, T.J.P. and Harvey, D.J. (1996) Effect of structure on the signal strength of oligosaccharides in matrix-assisted laser desorption ionization mass spectrometry on time-ofight and magnetic sector instruments. Rapid Commun Mass Spectr 10, 13611366. Papac, D.I., Wong, A. and Jones, A.J.S. (1996) Analysis of acidic oligosaccharides and glycopeptides by matrix-assisted laser desorption ionization time-of-ight mass spectrometry. Anal Chem 68, 32153223. Sambrook, J. and Russell, D. (2001) Molecular Cloning: A Laboratory Manual, 3rd edn. New York: Cold Spring Harbor Laboratory Press. Stanley, S.M.R. and Poljak, A. (2003) Matrix-assited laser-desorption time-of-ight ionization and high performance liquid chromatography-electrospray ionization mass spectral analyses of two glycosylated recombinant epoetins. J Chromatogr B 785, 205218.

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094

2093

Human erythropoietin from P. pastoris

E. C elik et al.

biger, G., Marchetti, M., Nagano, M., Reichel, C., Gmeiner, Stu G. and Allmaier, G. (2005) Characterisation of intact recombinant human erythropoietins applied in doping by means of planar gel electrophoretic techniques and matrix-assisted laser desorption ionisation linear time-of-ight mass spectrometry. Rapid Commun Mass Spectr 19, 728742. Tarentino, A.L., Gomez, C.M. and Plummer, T.H. (1985) Deglycosylation of asparagine-linked glycans by peptide: Nglycosidase F. Biochemistry 24, 46654671.

Walker, G.M. (1998) Yeast: Physiology and Biotechnology. New York: John Wiley & Sons Inc. Wegner, G. (1990) Emerging applications of the methylotrophic yeasts. FEMS Microbiol Rev 7, 279283. Zanjani, E.D., Poster, J., Borlington, H., Mann, L.I. and Wasserman, L.R. (1977) Liver as the primary site of erythropoietin formation in the fetus. J Lab Clin Med 89, 640.

2094

2007 The Authors Journal compilation 2007 The Society for Applied Microbiology, Journal of Applied Microbiology 103 (2007) 20842094