W. Min Jou, Department of Molecular Biology, University of Gent and Flanders Interuniversity Institute for Biotechnology, K. L. Ledeganckstraat 35, B-9000 Gent, Belgium. Fax: + 32 92 64 87 99, E-mail: Willy.MinJou@dmb.rug.ac.be
The A/Victoria/3/75 (H3N2-subtype ) hemagglutinin (HA) gene was engineered for expression in Pichia pastoris as a soluble secreted molecule. The HA cDNA lacking the C-terminal transmembrane anchor-coding sequence was fused to the Saccharomyces cerevisiaeα-mating factor secretion signal and placed under control of the methanol-inducible P. pastoris alcohol oxidase 1 (AOX1) promoter. Growth of transformants on methanol-containing medium resulted in the secretion of recombinant non-cleaved soluble hemagglutinin (HA0s). Remarkably, the pH of the induction medium had an important effect on the expression level, the highest level being obtained at pH 8.0. The gel filtration profile and the reactivity against a panel of different HA-conformation specific monoclonal antibodies indicated that HA0s was monomeric. Analysis of the N-linked glycans revealed a typical P. pastoris type of glycosylation, consisting of glycans with 10–12 glycosyl residues.
Mice immunized with purified soluble hemagglutinin (HA0s) showed complete protection against a challenge with 10 LD50 of mouse-adapted homologous virus (X47), whereas all control mice succumbed. Heterologous challenge with X31 virus [A/Aichi/2/68 (H3N2-subtype)], resulted in significantly higher survival rates in the immunized group compared with the control group. These results, together with the safety, reliability and economic potential of P. pastoris, as well as the flexibility and fast adaptation of the expression system may allow development of an effective recombinant influenza vaccine.
Influenza is a well-known viral pathogen, which causes epidemics and pandemics in humans and animals. In humans, an influenza infection is normally restricted to the upper respiratory tract, and often results in severe morbidity, many hospitalizations, and even mortality. Two types of antigenic variation of the influenza virion have been recognized, namely ‘drift’ and ‘shift’, which are the underlying causes of recurrent epidemics and pandemics, respectively [1–4]. Because of this frequent and unpredictable antigenic variability, the design of a universal influenza vaccine has so far not been realized. Current prophylaxis against influenza is via parenteral vaccination with inactivated or subunit virus grown in embryonated chicken eggs, despite their potentially serious limitations as a host system [5–9]. A number of alternative approaches have been pursued to achieve protective immunity including peptide vaccination, DNA vaccination and recombinant vaccinia virus technology [10–13]. However, concerns about safety, pre-existing immunity in people, immune responses against the vector itself and, in some instances, an insufficient protective immune response, have limited the usefulness of these approaches. The use of purified, recombinant influenza membrane proteins appear to be a promising alternative [14–18].
Three membrane proteins are present on the influenza virion: hemagglutinin (HA), neuraminidase and the M2 protein. Although neuraminidase, a type-II membrane glycoprotein, displays similar amounts of antigenic variation to HA, antibodies specific to this protein, do not neutralize influenza infection, even though they do minimize viral spread. The conserved influenza M2 protein is only present in minute amounts on the viral surface and is therefore incapable of eliciting neutralizing immunity .
HA, a homotrimeric class I membrane glycoprotein, is quantitatively the major surface protein of influenza virus and the major antigen against which neutralizing antibodies are elicited . Therefore, recombinant HA is a very favorable antigen as a candidate influenza vaccine. HA mediates the attachment of the virus to the target cell through specific binding with sialic acid-containing determinants and, following internalization, the release of the viral content into the attacked cell [19, 20]. HA-specific antibodies are protective as a result of their ability to prevent virus attachment and penetration of the host cell, or presumably by interfering with the low-pH-induced conformational change of the HA molecule needed for fusion [21–24]. Because of the immune selection pressure, HA is the viral component which is most important in antigenic drift.
The HA monomer is synthesized as a single polypeptide chain which undergoes post-translational cleavage at two sites: the N-terminal signal sequence is removed and, depending on the host cell and virus strain, the molecule is cleaved, with the removal of one or more intervening residues, resulting in two polypeptide chains called HA1 (36 kDa) and HA2 (27 kDa), linked via a disulfide bridge [25–27]. A C-terminal stretch of hydrophobic amino acids anchors HA to the viral membrane and, though not essential for secretion, this sequence plays a major role in the trimerization process .
The use of transformed mammalian, avian or insect cells for expression is a serious constraint, both practically and economically, for production of an influenza vaccine. Therefore, the use of a recombinant organism, such as yeast that is easy and economical to grow on a large scale in a fermentor, could be a major advantage for production of an influenza HA vaccine. Recently, we described partial protective immunity in mice with soluble influenza neuraminidase expressed by Pichia pastoris. Here we report the expression, characterization and immuno-protective potential of glycosylated, secreted influenza H3-subtype HA in P. pastoris. The purified protein, administered in mice in combination with proper adjuvants, is capable of eliciting a fully protective antibody response in mice against a lethal viral challenge.
Materials and methods
Plasmids, yeast strain and culture conditions
The P. pastoris expression plasmid pPIC9 (Invitrogen, San Diego, CA, USA) and the A/Victoria/3/75 HA expression plasmid pSV23IVHAs (previously named pUR250HA8 ) were used to construct the influenza HA secretion vector for P. pastoris. pPIC9 contains the promoter and terminator signals from the methanol-inducible P. pastoris alcohol oxidase 1 (AOX1) gene, the α-mating factor prepro-secretion signal from Saccharomyces cerevisiae, and the HIS4 auxotrophic selection marker for transforming P. pastoris GS115 (his4) (Invitrogen). This strain was used as a host cell for the production of HA0s. Shake-flask cultures were grown in YPGly (1% yeast extract, 2% bacto peptone, 2% glycerol) and induced in YPNM (1% yeast extract, 1% bacto peptone, 1% yeast nitrogen base, 2% methanol), buffered as described below.
Construction of the expression plasmid and the recombinant P. pastoris strain
In order to fuse the HA gene directly downstream of the last GluAla codons from the S. cerevisiaeα-mating factor prepro-secretion signal sequence, pPIC9 was modified as follows: a PCR fragment was generated by using pPIC9, the AOX1 sequencing primer (Invitrogen) and the primer 5′ GAATTCTACGCCGGCTTCAGCCT 3′ (NaeI) in a 30 cycle reaction with Pfu DNA polymerase (Boehringer, Mannheim, Germany). The resulting 372 bp PCR fragment was digested with EcoRI and BamHI and ligated into the BamHI-EcoRI-opened pPIC9 plasmid, in which the resident NaeI site had been previously disrupted by insertion of the 5′ GTCGAC 3′HindII recognition sequence. The resulting plasmid was named pPP1 m and contains a unique NaeI restriction site, allowing gene fusion to the last GluAla codons from the S. cerevisiaeα-mating factor prepro-secretion signal.
To excise the mature HA-coding sequence, an ApaI restriction site was introduced into pSV23IVHAs using site-directed mutagenesis  with the oligonucleotide 5′ CTGTCTGGTTTGGGCCCAAGACC 3′, resulting in pSV23IVHAsm. The HA-containing SalI-BamHI fragment from this vector was ligated into the similarly opened pUC18 plasmid , resulting in pUCIVHAsm. This plasmid was linearized with BamHI, end-filled with T4 DNA polymerase and dNTPs, and ligated with the previously hybridized oligonucleotides 5′ TAAGCGGCCGCG 3′ and 5′ CGCGGCCGCTTA 3′ (NotI). A clone with the NotI site downstream (relative to the HA cistron) of the introduced TAA termination codon was chosen for further use, and named pUCIVHAsim. This plasmid was linearized with ApaI, flush-ended with T4 DNA polymerase in the presence of dNTPs, and subsequently digested with NotI. The resulting HAs gene contains an open reading frame starting with the Gln17 codon and ending with the Ile516 codon of the A/Victoria/3/75 influenza HA. This fragment was ligated into the NaeI-NotI opened pPP1 m plasmid, resulting in the P. pastoris expression vector pPP1IVHAsi.
SalI-linearized pPP1IVHAsi was used to transform P. pastoris GS115 by electroporation according to the manufacturer’s instructions (Invitrogen). Individual P. pastoris transformants were grown in 5 mL of YPGly for 24 h at 28 °C under continuous shaking. Twenty-four hours after inoculation, the cultures were induced by resuspending the cells collected by centrifugation in an equal volume of YPNM buffered with 100 m m potassium phosphate pH 6.0 and incubated for another 24 h. The proteins in the culture supernatants were precipitated with trichloroacetic acid and tested in a Western blot experiment for the presence of HA0s using rabbit polyclonal anti-HA IgG . One positive transformant was selected for further use and named GSIVHAs.
Detection of HA0s from P. pastoris GSIVHAs
ELISA, SDS/PAGE (10% acrylamide gels) and Western blotting were performed as described previously . Silver-staining of the 10% polyacrylamide gels was according to Morrissey . Protein concentrations were determined according to Bradford , using BSA as a standard.
Purification of HA0s
Purification of HA0s was performed by means of an FPLC apparatus (Pharmacia Biotech, Upsala, Sweden) equipped with an LCC-501 plus controller and an UV-MII optical unit. Unless stated otherwise, chromatography materials were purchased from Pharmacia Biotech. A 5-mL preculture of P. pastoris GSIVHAs was used to inoculate 4 × 400 mL YPGly in 2-L shake flasks. At A600 12–15, cells were collected by centrifugation (5 min, 5000 g, 20 °C), resuspended in half a volume of YPNM, 60 m m Tris/HCl pH 8.2 (final pH 8.0) and incubated for another 24 h.
After induction, cells were removed by centrifugation (6000 g, 20 min, 4 °C) and the culture medium was filtered through a paper prefilter (Millipore, Bedford, MA, USA). The filtrate was applied onto a Q-HyperD column (16 × 40 mm) (BioSepra, Villeneuve la Garenne, France), pre-equilibrated with 10 m m NaCl,10 m m Tris/HCl pH 8.0, using a flow rate of 10 mL·min−1. Under these conditions HA0s does not bind to the Q-HyperD anion exchanger, but this purification step does remove impurities, mainly of a nonprotein nature. The HA0s-containing flow-through of the Q-HyperD column was adjusted to 20% ammonium sulfate saturation, filtered through a 0.45-µm filter (Millipore) and applied onto a Phenyl HP-Sepharose column (26 × 50 mm – XK 26; Pharmacia Biotech), pre-equilibrated with 20% ammonium sulfate, 50 m m Tris/HCl pH 8.0 (buffer A), using a flow rate of 5 mL·min−1. After loading, the column was washed with one bed volume of the same buffer, and bound proteins were eluted with a three-step gradient followed by a linear gradient using 20 m m Tris/HCl pH 8.0 (buffer B) and a flow rate of 4 mL·min−1. The three-step elution was performed each time with 20 mL of 20%, 40% and 60% of buffer B, after which a linear gradient to 100% buffer B over a total volume of 100 mL was applied, followed by another 100 mL of buffer B only. Ten-millilitre fractions were collected starting from the middle of the linear gradient to the end of the elution. Typically, HA0s was found only in seven fractions, which were pooled and diluted with an equal volume of 10 m m Tris/HCl pH 8.5 and filtered through a 0.22-µm filter (Millipore). Subsequent flow rates were 1 mL·min−1. The HA0s-containing pool was loaded onto a 1-mL MonoQ column, equilibrated with 10 m m NaCl, 10 m m Tris/HCl pH 8.5. Elution was performed in 1-mL fractions by application of a linear NaCl concentration gradient from 10 m m to 1000 m m in 20-mL total volume, in the same buffer. HA0s-containing fractions were pooled and diluted 10-fold with 10% ammonium sulfate in 25 m m Tris/HCl pH 8.0 and loaded onto a 1-mL phenyl resource column. The column was washed with 5 mL of the same buffer, and eluted using 20 column volumes ammonium sulfate using a linear concentration gradient from 10% in 25 m m Tris/HCl pH 8.0, to 0% in 10 m m Tris/HCl pH 8.0. One-millilitre fractions were collected. HA0s eluted as a single peak and was more than 99% pure as analyzed by a silver-stained 10% polyacrylamide gel. HA0s was dialyzed against NaCl/Pi using a Centricon-10 device (Amicon, Danvers, MA, USA), and supplemented with glycerol up to 50% final concentration for storage at −20 °C.
Gel filtration analysis of HA0s
Five hundred microlitres of induced GSIVHAs culture supernatant was loaded onto a Superose-12 HR 10/30 gel filtration column (Pharmacia Biotech) which was run at flow rate at 0.5 mL·min−1 in NaCl/Pi. HA0s was followed in the collected fractions by ELISA. The column was calibrated with 50 µg of ferritin (440 kDa), bromelain-cleaved hemagglutinin (BHA; 200 kDa), aldolase (158 kDa), BSA (67 kDa) and ribonuclease A (13.7 kDa) in separate runs.
N-glycan analysis of HA0s
N-linked oligosaccharides were isolated by denaturing 100 µg of purified HA0s in 0.5% SDS and 1% β-mercaptoethanol at 100 °C for 5 min, followed by overnight N-glycosidase F (PNGase F) digestion at 37 °C in 50 m m sodium phosphate pH 7.5, supplemented with 1% NP-40 . A 2-µg aliquot was retained for SDS/PAGE analysis. Oligosaccharides and proteins were precipitated by adding 4 volumes of −20 °C-chilled acetone. After 10 min centrifugation at 13 000 g in an Eppendorf centrifuge, the supernatant was discarded and the pellet was extracted twice with 200 µL of −20 °C-chilled 60% methanol. The extracted glycans were lyophilized and covalently labeled with 8-aminonaphtalene-1,3,6-trisulfonate (ANTS) at the reducing end, and subsequently separated by 25% PAGE . The electrophoretic band patterns were visualized by UV illumination and photographed on black-and-white film (Polaroid, ISO 3000/36; Sigma Chemical Co., St Louis, MO, USA) through a red filter.
N-terminal amino acid sequence analysis of HA0s
Determination of the amino terminus of purified HA0s (by courtesy of Dr J. Vandekerckhove, Department of Medical Protein Chemistry, University of Gent and Flanders Interuniversity Institute for Biotechnology) was performed by automated Edman degradation, on a model 470 A gas-phase sequencer coupled to a model 120 A on-line phenylthiohydantoin amino acid analyzer (Applied Biosystems, Foster City, CA, USA).
Characterization of HA0s with different HA-specific monoclonal antibodies
HC59 was a gift of Drs J. Skehel and A. Douglas (Division of Virology, National Institute for Medical Research, London) and was prepared against A/Port Chalmers/1/73; it reacts also with A/Victoria/3/75 and binds to the native and to the low-pH-induced HA conformations. Monoclonal antibody IIF4, kindly provided by Dr F. Kostolansky (Institute of Virology, Bratislava), was raised against A/Dunedin/4/73 (H3N2-subtype). IIF4 is a cross-reactive HA2-specific mAb, reacting with the trimeric form of HA in its conformation at pH 5 [38–40]. LC89, a low-pH-induced HA-specific mAb, a gift of Dr S. Wharton (Division of Virology, National Institute for Medical Research, London, UK), was raised against detergent-extracted X31 HA in the fusion pH conformation . mAb LMBH5 was isolated in our laboratory and recognizes a fairly conserved epitope on the H3-subtype HA near the receptor-binding pocket. It recognizes the native structure of A/Victoria/3/75 HA, but binding is enhanced after low pH treatment. In a sandwich ELISA, reactivity of LMBH5 with A/Victoria/3/75 HA is only observed at low pH . LMBH6 binds to the membrane proximal end of trimeric H3-subtype HA . Reactivity of purified HA0s with these mAbs was determined in a sandwich ELISA as described .
Recombinant viruses between the WHO H3N2-subtype reference strains A/Aichi/2/68 (X31), A/Victoria/3/75 (X47) and A/PR/8/34 (H1N1) were used (kindly provided by Drs J. Skehel and A. Douglas). Viruses were grown in the allantoic fluid of 10-day-old, embryonated chicken eggs. X47 and X31 were adapted to mice by a series of lung passages. Mouse-adapted X47 virus was grown in the allantoic fluid of 10-day-old, embryonated chicken eggs and purified by clarifying the infected fluid using sequential centrifugation steps at 4 °C for 20 min at 10 000 g; virus was pelleted from the supernatant by 16 h centrifugation using the same conditions; the resuspended pellet (in NaCl/Pi) was centrifuged again in a Beckman ultracentrifuge at 100 000 gfor 2 h in an SW28 rotor. The resulting X47 virus pellet was resuspended in 5 mL NaCl/Pi, aliquoted and stored at −70 °C. Mouse-adapted X31 virus was available as an infectious mouse lung homogenate, stored at −70 °C in separate batches.
Immunization and challenge of mice
Seven-week-old, pathogen-free, female Balb/c mice (Charles River, Sulzfeld, Germany) were used for vaccination and challenge experiments. The animals were housed in a temperature-controlled environment at 22–24 °C with 12 h day-night cycles, and received food and water ad libitum. Mice were immunized three times subcutaneously in the hind leg at 3-week intervals with 200 µL doses of 1 µg HA0s plus adjuvant. In the first immunization 25 µg Salmonella typhimurium monophosphoryl lipid A, 25 µg trehalose-6,6-dimycolate, 1% squalene and 0,1% Tween-80 (Ribi adjuvant; Ribi Immunochem Research, Hamilton, MT, USA) was used, while for the two booster injections 25 µg monophosphoryl lipid A and 25 µg muramyl dipeptide (Sigma Chemical Co) were used. Control mice received adjuvant only. One week prior to the first immunization and 2 weeks after each immunization, blood samples were taken from a tail vein, sera prepared and stored at −20 °C before determination of the X47-and X31-specific titer in ELISA . Briefly, Maxisorp 96-well (Nunc, Roskilde, Denmark) were coated with purified X47 or X31 virus (2500 HAU·mL−1) dissolved in NaCl/Pi. After blocking with BSA, a serial dilution of pooled mouse serum was applied. Virus-specific serum IgG was detected with an alkaline phosphatase conjugated goat (anti-mouse IgG) IgG (Sigma Chemical Co.). Three weeks after the second boost injection, mice were challenged intranasally with 10 LD50 of mouse-adapted X47 or X31. Lethality and morbidity were monitored for 14 days after infection on the basis of rectal temperatures and body weights.
Expression of HA0s from P. pastoris
In order to evaluate the potential of the methylotrophic yeast P. pastoris as an expression host for secretion of recombinant influenza HA, the influenza HA expression cassette pPP1IVHAsi was constructed ( Fig. 1). In this plasmid the coding sequence for the anchor-free HA of the A/Victoria/3/75 (H3-subtype) influenza strain is fused to the heterologous α-mating factor prepro-secretion signal from S. cerevisiae. Proteolytic processing of this secretion signal is determined by the endogenous Kex2p protease activity and a Ste13p-type dipeptidyl aminopeptidase, releasing HAs. The P. pastoris AOX1 promoter and terminator sequences were used for transcriptional control. SalI-linearized pPP1IVHAsi was transformed to P. pastoris strain GS115 (his4) by electroporation and transformants were selected on histidine-free minimal medium. Southern analysis of randomly picked transformants revealed the presence of a single copy of transforming DNA in the his4 locus of the host genome (data not shown).
After 24 h induction with methanol, HA0s could be visualized by Western blotting and appeared as a double band with an apparent molecular mass of 75–78 kDa ( Fig. 2). We previously observed a dramatic effect of the pH of the extracellular medium on the outcome of the oligomeric and antigenic state of HA0s expressed in insect cells [34, 43]. Therefore, P. pastoris strain GSIVHAs was induced in sodium phosphate-buffered YPNM medium at different pH values. ELISA and Western blot revealed a pronounced effect on the expression level of HA0s ( Fig. 2). A pH 8.0-buffered medium resulted in the highest HA0s secretion level and was maintained for purification purposes (1 µg·mL−1 based on ELISA experiments using homologous BHA as a standard).
Purification and molecular characterization of HA0s from P. pastoris
For purification purposes GSIVHAs cultures were scaled up to 400 mL and induction was performed at pH 8.0 with Tris/HCl. HA0s was purified to over 99% homogeneity by standard chromatographic techniques ( Table 1; Fig. 3). Typically, 300–500 µg HA0s was recovered from 800 mL of induced yeast-culture supernatant in shake flasks. As this yield was obtained under normal laboratory conditions, it should be possible to obtain considerable higher expression levels by optimized fermentation procedures, allowing cell densities of A600 = 200–400 instead of 10–15. In Western blots under reducing and nonreducing conditions, HA0s showed the same electrophoretic mobility of 75–78 kDa, indicating that no cleavage between HA1 and HA2 had occurred (data not shown). PNGase F treatment of HA0s increased its electrophoretic mobility, resulting in a single polypeptide band of ≈58 kDa, consistent with the molecular mass calculated from the amino acid sequence of the mature, uncleaved anchor-free HA0s polypeptide ( Fig. 3). The first six N-terminal amino acids of the purified protein were determined by automated Edman degradation. This revealed complete Kex2p cleavage but only partial Ste13p-like processing of the α-mating factor prepro-secretion signal. HA0s appeared to be an equimolar mixture of proteins with N-terminal GluAlaGluAlaGlnAsp (Kex2p cleavage only), GluAlaGlnAspLeuPro (partial Ste13p-like processing), and GlnAspLeuProGlyAsn (N-terminus of HA0s) sequence. A similar incomplete maturation has been observed for recombinant influenza neuraminidase from P. pastoris.
Table 1. Flow scheme of the purification of HA0s from methanol-inducedP. pastorisGSIVHAs culture supernatant.
Total protein (mg)
Total protein content was determined according to Bradford .
HA0s was determined by ELISA as compared to X47 BHA.
Phenyl Sepharose HP
To determine the chain length of the N-glycans, 100 µg of purified HA0s was subjected to PNGase F digestion. The released sugars were precipitated, labeled with fluorogenic ANTS and loaded on a 25% polyacrylamide gel. The carbohydrates were homogeneous, consisting of N-glycans with 10–12 glycosyl residues, presumably (N-acetylglucosamine)2Man8–10( Fig. 4). This is in agreement with the reported average size of P. pastoris N-linked oligosaccharides [44, 45].
Determination of the oligomeric structure and conformation of HA0s
Size-exclusion chromatography and conformation-specific mAbs were used to examine the oligomeric state of yeast-expressed HA0s. The Superose-12 gel filtration profile revealed a single peak, eluting with an estimated m of 80–90 kDa, which is indicative of a monomeric state ( Fig. 5A). Cross-linking experiments with purified HA0s, supported this finding. Using various cross-linking agents under different conditions, no evidence for the presence of trimeric HA0s was obtained (data not shown).
Additional evidence for the monomeric assembly of HA0s was obtained from gel filtration experiments in combination with recognition by different HA-specific mAbs. As shown in Fig. 5B, HA0s is recognized very well by mAbs HC59 and LMBH5 and IIF4. Binding with the mAb IIF4 was enhanced after trypsin treatment, whereas binding with the trimer-specific mAb LMBH6  was not observed. mAbs HC59 and LMBH5 are reactive against the head domain of HA; IIF4 is a HA2-specific mAb, reacting with monomeric and trimeric HA, both only in the low-pH induced conformation [38–40, 43]. Reactivity of HA0s with these mAbs indicated that the HA0s appeared to be, at least partially, folded into a native HA conformation. From these observations we conclude that HA0s secreted from P. pastoris at pH 8.0 is monomeric and is at least partially correctly folded, resembling the monomeric HA0s fraction secreted from insect cells . Therefore, yeast-derived HA0s is an attractive recombinant vaccine candidate, and was assessed for its protective efficacy against an influenza virus challenge in a mouse model.
Protective immunity after HA0s immunization of mice against a homologous and heterologous viral challenge
To determine whether immunization with HA0s could elicit a protective immune response, an immunization-challenge study was performed in mice. Two groups of 11 Balb/c mice were immunized subcutaneously with three doses of 1 µg HA0s, given at 3-weekly intervals. Two control groups received adjuvant only. Viral HA-specific antibody responses in serum samples before immunization and 2 weeks after each immunization were determined by ELISA. Three immunizations with adjuvant only, did not result in a virus-specific immune response. After the second and, more clearly, after the third injection, HA0s vaccinees had obtained an important seroconversion against both X47 and X31 viruses ( Fig. 6). This serum response is also indicates that HA0s derived from yeast has the correct antigenic structure.
Three weeks after the third immunization, control mice and vaccinees were challenged with 10 LD50 of either mouse-adapted homologous X47 or heterologous X31. In the homologous viral challenge set-up, all control mice suffered from progressive loss of body weight and drop of rectal temperature, and eventually died within 8 days post challenge. HA0s vaccinees, although showing some signs of illness between day 2 and 6 after the viral challenge, were completely protected from the X47 challenge ( Fig. 7.A). The X31 challenge did not provoke complete lethality in the control group: 4/11 mice survived the infection ( Fig. 7.B). Still, this number is significantly lower than the 10/11 surviving mice in the HA0s-vaccinated group. Furthermore, mice from both groups became ill, shown by the loss of body weight and the decreased rectal temperature, but the morbidity of surviving mice from the control group was more severe and sustained than mice from the vaccinated group. Thus, P. pastoris-expressed HA0s, administered as an adjuvant vaccine, has the potential to induce a protective immune response in mice against a lethal, influenza-virus infection.
The methylotrophic yeast P. pastoris has been developed as a host for the efficient production of heterologous proteins . Interest in this eukaryotic methylotropic organism as a heterologous protein production system has grown, since it has the potential of high-level expression and rapid growth to very high cell densities in inexpensive media . Jabbar and Nayak (1987) reported the (low) expression of truncated, hyperglycosylated, and mainly cell-associated, A/WSN/33 immuno-reactive influenza HA in S. cerevisiae without, however, demonstrating its applicability in vivo. Here we demonstrate that P. pastoris is capable of expressing a soluble form of the influenza A virus HA with near native antigenic structure. Furthermore, purified HA0s, although monomeric, was capable of eliciting a protective antibody response against a lethal influenza challenge in mice.
The expression level of anchor-free HA from P. pastoris was markedly affected by the pH of the induction medium, with pH 8.0 being optimal. The influence of the extracellular pH on trimerization and transport of HA in a mammalian cell expression system has been reported previously . The monomeric structure of the purified HA0s is not unexpected, in view of the observations by Singh et al. who demonstrated that recombinant, anchor-free X31 [A/Aichi/68 HA (H3-subtype)] was secreted as a monomer from CV-1 cells . Anchor-free A/Japan/305/57 HA (H2), on the other hand, formed monomeric, trimeric and higher order complexes in the Golgi complex or in secretory vesicles, when expressed in the same cells. Vanlandschoot et al. reported on the expression, in insect cells, of aggregated and monomeric soluble A/Victoria/3/75 HA0s at pH 6.0, and monomeric HA0s at pH 8.0, with a minor trimeric fraction being observed under both conditions [34, 43]. Although we did not analyze the oligomeric state of the low-pH-expressed HA0s in P. pastoris, it seems unlikely that aggregated HA0s molecules of a similar size (> 1.5 × 106 Da) could traverse the yeast cell wall.
Analysis of the N-linked carbohydrates showed the presence, predominantly, of (N-acetylglucosamine)2Man8–10 residues ( Fig. 4). This result is in agreement with the reported average 8–14 mannose residues added post-translationally by P. pastoris, and is in striking contrast with the observation of the rather exceptional hyperglycosylated nature of soluble recombinant neuraminidase containing N-glycans with 30–40 mannose residues, from the same organism . Although the molecular mechanisms determining the outcome of the glycosylation pattern of a glycoprotein in a particular eukaryotic host organism remain enigmatic, one might speculate that the folding kinetics play a role. Glycoproteins that spend longer in the early exocytic vesicles might be more susceptible as as substrate for glycosyltransferase activity. Recognition of P. pastoris-secreted HA0s by a panel of mAbs implies that at least part of the molecule is correctly folded.
The amino terminus of HA0s was incompletely processed by the STE13-like dipeptidyl aminopeptidase A (DPAPA) activity from P. pastoris, resulting in an approximately equimolar mixture of fully, partially and unprocessed HA0s. Incomplete maturation by DPAPA of recombinantly expressed proteins has often been observed in yeast [29, 50]. The presence of a spacer peptide has been reported to improve the cleavage by Kex2p at the dibasic site of the α-factor leader sequence, but this advantage may be outweighed by incomplete removal of the GluAla dipeptides . Immunization of mice with purified HA0s in combination with adjuvant resulted in a strong virus-specific antibody response. A drawback of most soluble antigens is the requirement of an adjuvant to elicit immunity. Monophosphoryl lipid A was chosen in our experiments because it has already been successfully evaluated in clinical trials [52, 53].
The serum response obtained was fully protective against a homologous X47 challenge and partially against a heterologous X31 challenge. Furthermore, immunized mice showed less severe symptoms of infection than control mice. This partial cross-protection in mice is not completely unexpected, since we previously described the isolation of a cross-neutralizing mAb after immunization with the homologous HA0s expressed in COS cells and challenge with the X31 virus .
Until now, only inactivated vaccines, isolated from embryonated chicken eggs, have gained widespread use against influenza. However, their effectiveness is compromised by the frequent antigenic drift of the influenza virus, the limited supplies of high-quality eggs and the susceptibility of the eggs to avian influenza infection, and the danger of variant selection in the avian host. We here describe a alternative method to produce an influenza vaccine, based on the synthesis of recombinant HA0s in P. pastoris. This fermentable, methylotrophic yeast has proven to be an excellent and cost-effective host for the production of heterologous proteins . Although membrane-bound HA has already been successfully expressed in Sf9 insect cells  and has been shown to elicit protective antibodies , the expression of protective secreted HAs from a monocellular organism has so far not been reported. We have shown that the immunogenic potential of yeast-derived HA0s as described here, may be appropriate for the development of an easily adaptable, safe and economic alternative to the currently used influenza vaccine. Furthermore, being a recombinant expression system, it may be possible to improve its protective properties by genetic engineering. For example, one could produce an HA0s molecule with altered or omitted variable and immuno-dominant epitopes, which might result in an antigen with broadened protection potential. Although the mouse model used here is commonly accepted to evaluate experimental influenza vaccines, the results described should only be regarded as an initial proof of principle.
Finally, the flexibility and the potential speed associated with a yeast expression system, may prove to be indispensable at the time of an emerging pandemic. Indeed, it has been calculated that the outbreak of a new influenza pandemic will not allow sufficient time and number of embryonated eggs to produce vaccines for a large population at risk. Alternative antiviral medication will probably not be available in sufficient amounts. The H5N1 influenza subtype recently isolated from patients in Hong Kong, although apparently under control by now, is a reminder of the always looming, potential global threat . In this case, even the production of vaccine in embryonated eggs was compromised because the seed virus killed the embryos. In the USA, an influenza pandemic vaccination policy is being worked out for distribution of vaccines according to a predetermined priority system [56, 57]. The availability of a flexible influenza vaccine-producing system and that could be scaled up quickly may be the key step that would allow implementation of such a selective vaccination plan. A comparative study with conventional influenza vaccines and different viral strains will be necessary to be able to further evaluate the potential of the described influenza virus HA expression system.
The authors are indebted to Drs A. Douglas, J. Skehel, S. Wharton and F. Kostolansky for generously providing different mAbs and virus strains. Dr J. Vandekerckhove is acknowledged for amino-terminal sequence analysis of HA0s, M. Vandecasteele for critically reading the manuscript and W. Drijvers for artwork. W.M. held a fellowship from the Vlaams Instituut voor de Bevordering van het Wetenschappelijk-technologisch Onderzoek in de Industrie. Research was supported by the Vlaams Actiecomité voor Biotechnologie, the Fonds voor Geneeskundig Wetenschappelijk Onderzoek and the Nationale Loterij.
*Present address: Department of Clinical Biology, Microbiology and Immunobiology, Gent University Hospital, De Pintelaan 185, B-9000 Gent, Belgium.