Вы находитесь на странице: 1из 4

J BT,865-868,460 d'Anjou 4/12/97 1:57 pm Page 865

Biotechnology Techniques, Vol 11, No 12, December 1997, pp. 865–868

11111
2
3
A model-based feeding strategy
4
5
for fed-batch fermentation of
6
7
recombinant Pichia pastoris
8
9 M.C. d’Anjou and A.J. Daugulis*
10111 Department of Chemical Engineering, Queen’s University, Kingston, Ontario, Canada K7L 3N6
1
2 A two stage, exponential feeding strategy with mixed glycerol/methanol substrate was used in a fed-batch recombi-
3 nant Pichia pastoris fermentation. The feeding strategy was developed using a simple model based on mass balances,
4 Monod-type growth kinetics, and constant specific heterologous protein production rate. The model accurately
5 predicted cell growth, and demonstrated the usefulness of a rational, model-based approach for improving the produc-
6 tivity of recombinant P. pastoris fermentation.
7
8
9 Introduction gies (Brierley et al., 1990; Loewen et al., 1997). To this
20111 The methylotrophic yeast Pichia pastoris has been devel- point, much of the work done with this system is
1 oped into an effective expression system of heterologous preliminary and highly empirical resulting in subop-
2 proteins for bench and pilot-scale fermentation. A timal and often disappointing expression levels of
3 number of features of Pichia give it decided advantages heterologous proteins.
4 as an expression host over some more widely used
5 systems including: frequent high to very high expres- Significant enhancements to the productivity of recom-
6 sion levels; a lack of endotoxins which make produced binant fermentations in better characterized expression
7 protein suitable for therapeutic uses; correct folding and systems such as Escherichia coli and S. cerevisiae have been
8 secretion of eukaryotic proteins into the growth achieved by developing models accurately describing
9 medium; fewer problems of hyperglycosylation in Pichia system behaviour and using the models to improve
30111 systems compared to Saccharomyces cerevisiae; well devel- fermentation protocols to optimize heterologous protein
1 oped high cell density fermentation protocols; and, a expression (Baheri et al., 1997; Patkar et al., 1993). The
2 strong, tightly regulated promoter for the alcohol success of these techniques in recombinant E. coli and
3 oxidase gene (AOX1) which has been well characterized S. cerevisiae suggests that a rational, model-based
4 and exploited by a number of P. pastoris expression approach to developing an improved fermentation
5 vectors. A more detailed description of the features of protocol for recombinant P. pastoris systems is
6 this system are provided in two recent reviews (Cregg warranted. A P. pastoris strain that has been genetically
7 et al. 1993; Romanos, 1995). modified to produce and secrete sea raven anti-freeze
8 protein (SR-AFP) is used as a model system to demon-
9 However, despite its promise as a superior expression strate the use of a model-based feeding strategy in fed-
40111 system, the levels of expression of different heterologous batch recombinant P.pastoris fermentation. Fed-batch
1 proteins in P. pastoris vary over a wide range. A number mass balance equations, based on Monod-type kinetics
2 of molecular biology methods are being developed to and a constant specific heterologous protein production
3 improve on poor expression levels and overall produc- rate, were used to develop a feeding strategy and to
4 tivity (Cregg et al. 1993; Romanos, 1995). Enhance- predict biomass growth and protein production. A fed-
5 ments to process productivity in recombinant P.pastoris batch fermentation was carried out on a 10 l scale
6 fermentation can also be made by optimizing the according to the predicted feeding strategy to verify the
7 fermentation medium (Siegel and Brierley, 1989), and accuracy of the model.
8 improving the fermentation methodology. Examples of
9 the latter include substrate feeding strategies (Rodriguez Materials and methods
50111 Jimenez et al., 1997), oxygen supplementation Organism and growth
1 combined with feed back control on aeration to allow The Pichia pastoris strain GS115 (his4) was transformed
2 higher cell densities while avoiding oxygen limitation with plasmid pPIC9 containing the gene encoding for
3111 (Chen et al., 1997), and mixed-substrate feeding strate- SR-AFP according to the methodology presented in

© 1997 Chapman & Hall Biotechnology Techniques · Vol 11 · No 12 · 1997 865


J BT,865-868,460 d'Anjou 4/12/97 1:57 pm Page 866

M.C. d’Anjou and A.J. Daugulis

11111 Loewen et al. (1997). The pPIC9 plasmid contains the anti-rabbit IgG; detection was by enhanced chemilu-
2 HIS4 gene for selection of his+ clones, and integrates minescence; and, quantifying of protein concentrations
3 by homologous recombination into the AOX1 gene site by visual inspection of band intensities relative to
4 when linearized with BglII. This transformation known standards.
5 produces an AOX1-deficient clone characterized by a
6 MutS (Methanol utilization slow) his+ phenotype which Results and discussion
7 secretes SR-AFP into the fermentation broth. Expression of heterologous proteins using this system
8 relies on the action of the strong, tightly regulated
9 Fed-batch fermentation was carried out in a 10 l promoter for the AOX1 gene. Regulation of this
10111 Chemap fermenter under the following conditions: promoter is due to combined glycerol/ glucose repres-
1 temperature, 30°C; impeller speed, 750 rpm; aeration, sion/derepression and methanol induction mechanisms.
2 15 l/min; pH, 5.5 measured by an Ingold pH probe and Expression of gene sequences downstream of the
3 maintained at that level by the addition of 5 M KOH. promoter occurs when low levels of methanol are present
4 Dissolved oxygen was measured using an Ingold steril- in the absence of repressing levels of glycerol or glucose.
5 izable polarographic electrode. A 20% (v/v) methanol The recombinant Pichia pastoris clone used in this study
6 feed and a 50% (v/v) glycerol feed were supplied during exhibits a MutS phenotype due to the AOX1 gene being
7 the fed-batch phase by means of calibrated peristaltic transplaced by the SR-AFP gene. However, the strain
8 pumps with speed controllers. Inocula and fermentation continues to have a limited ability to metabolize
9 media are as follows: methanol due to the presence of the transcriptionally
20111 BMGY: yeast extract, 10 g/l; meat peptone, 20 g/l; weak AOX2 gene. The maximum specific growth rate
1 100 mM potassium phosphate buffer pH 6.0; yeast on methanol alone of MutS strains lies between
2 nitrogen base without amino acids, 13.4 g/l; biotin, 0.01–0.04 h–1 (Brierley et al., 1990). As a consequence,
3 400 mg/l; glycerol, 10 ml/l. growth using methanol as the sole carbon source will
4 MGY: yeast nitrogen base without amino acids, result in efficient induction of the heterologous protein
5 13.4 g/l; biotin, 400 mg/l; glycerol 10 ml/l. but very slow cellular growth rates, and low overall
6 Fermentation medium: glycerol 50 g/l; (NH4)2SO4, productivity. Growth in the presence of excess glycerol
7 20 g/l; KH2PO4, 12 g/l; MgSO4·7H2O, 4.7 g/l; or glucose will rapidly increase biomass but repress the
8 CaCl2·2H2O, 0.36 g/l; plus trace elements: expression of the desired protein. However, a fed-batch
9 CaSO4·5H2O, 0.2 mM; KI, 1.25 mM; MnSO4·4H2O, fermentation using a mixed glycerol/methanol feed
30111 4.5 mM; Na2MoO4·2H2O, 2 mM; H3BO3, 0.75 mM; where the glycerol is fed at limiting rates and the
1 ZnSO4·7H2O, 17.5 mM, FeCl3·6H2O, 44.5 mM; pH methanol is fed to excess should lead to conditions
2 adjusted to 5.5 with 5 M KOH. where the cell mass is allowed to increase relatively
3 quickly while conditions favorable to the expression of
4 Inoculum cultures were started from MGY streak plates the heterologous protein are maintained (Loewen et al.,
5 in 10 ml of BMGY grown for 24 h at 30°C in a 50 mL 1997).
6 centrifuge tube at 250 rpm. 1 ml of the BMGY culture
7 was transferred to two 300 ml MGY inocula cultures It is therefore necessary to develop a fermentation
8 grown at 30°C at 250 rpm in 1 l Erlenmeyer flasks for protocol where the cells are grown on glycerol to a high
9 48 h. These cultures were used to inoculate 6l of fermen- cell density relatively quickly while ensuring that no
40111 tation medium in the 10 l Chemap fermenter. glycerol is allowed to accumulate in the growth
1 medium. An exponentially increasing glycerol feeding
2 Cell dry weight was estimated from the optical density rate matched to the growth of the cell biomass will
3 measured at 650 nm (1 OD ≈ 0.3 g dry wt./l). Methanol allow this to occur. Methanol need only be present to
4 and glycerol concentrations in the feed solutions and induce translation of the cloned gene and should play
5 the fermentation broth were measured by HPLC using no role in cell growth of the MutS strain and can be
6 a Sugar Pak I (Waters Millipore) column and a 50 mg neglected in modeling cell growth. The feeding rate
7 EDTA/l mobile phase 0.5 ml/min. Protein levels were profile during the production period was determined
8 estimated from Western blots according the following using simple mass balances based on a Monod-type
9 procedure: electrophoresis on a 15% SDS-PAGE con- kinetic model. The fed-batch mass balances are based
50111 taining 0.1 mM sodium phosphate and 4 M urea, at pH on a constant cell yield on glycerol and neglecting the
1 6.8; blotting onto a nylon membrane (PVDF, NEN); small contribution of consumed methanol to cell
2 incubation with rabbit anti-SR-AFP antiserum and growth. The model equations and the feeding rate which
3111 subsequently with horseradish peroxidase-linked goat allows exponential growth with a constant specific

866 Biotechnology Techniques · Vol 11 · No 12 · 1997


J BT,865-868,460 d'Anjou 4/12/97 1:57 pm Page 867

A model-based feeding strategy for fed-batch fermentation of recombinant Pichia pastoris

11111 growth rate are presented below: where X is cell biomass rate, it was expected that the system would become
2 (g dry wt./l); V is the fermentation volume (l); F is the oxygen-limited as the cell density became large. As
3 glycerol feed rate (l/h); m is the specific growth rate oxygen limitation is detrimental to heterologous protein
4 (h–1); S is the glycerol concentration in the growth expression in P. pastoris (Romanos, 1995), a second stage
5 medium (g/l); So is the glycerol concentration in the of slower growth (m = 0.01 h–1) was used to decrease
6 feed medium (g/l); YX/S is the cell yield on glycerol the culture’s oxygen requirement, and to allow a usable
7 (gX/gS); t and to are the run time and the run time at concentration of the product protein to accumulate, thus
8 the start of the fed-batch production period respectively facilitating downstream separation and purification. The
9 (h); mmax is the maximum specific growth rate on glyc- second stage of feeding was to be initiated when the
10111 erol (h–1); P is the heterologous protein concentration dissolved oxygen level reached 30% of saturation.
1 in the growth medium (gP/l); and qP is the specific
2 protein production rate (gP/gX·h). In our work, a batch growth phase was followed by a
3 24 h induction period of glycerol starvation and
4 d(XV) = m(XV) – FX methanol addition at a rate of 1.6 g/h. Though not
5 dt necessarily required, this induction phase was used to
6 d(SV) = FS – Y m(UV) ensure that the entire system was exposed to conditions
0 x/s
7 dt which induce expression of the heterologous protein.
8 d(PV) = q (XV) The methanol feed rate was kept constant throughout
p
9 dt the run such that the methanol concentration in the
20111 dV = F(t) growth medium was at inducing but non-inhibitory
1 dt levels (0.1–0.2%). Following the induction period, a
2 m = mmaxS glycerol feed following an exponential profile was started
3 S + Ks to initiate the higher growth rate production phase
4 F(t) = mV(t0)X(t0) exp[m(t – t0)] (stage 1). The glycerol feed rate was adjusted manually
5 S0Yx/s and discretely at regular intervals to reflect the predicted
6 exponential feed rate. The dissolved oxygen level was
7 Provided that the selected specific growth rate is signif- monitored and once the target of 30 % of saturation
8 icantly lower than the maximum and that no inhibitory was reached, the feed rate was lowered to initiate the
9 or otherwise unexpected effects occur due to the pres- second production phase (stage 2). Figure 1 shows the
30111 ence of methanol, this simple model should predict dissolved oxygen trace as well as the predicted and
1 cellular growth with acceptable accuracy. The necessary actual glycerol feed rates.
2 kinetic parameters were determined from batch and
3 continuous culture experiments (data not shown). As Figure 2 shows predicted and actual cell density and
4 well, the specific rate of heterologous protein produc- product protein concentration. Glycerol and methanol
5 tion, qP, was determined from these experiments and
6 assumed to be constant for the given conditions. These
7 kinetic parameters and the conditions at the start of the
8 fed-batch production period are presented as Table 1.
9
40111 Initially, a specific growth rate of 0.07 h–1 was chosen
1 as being sufficiently far removed from the maximum
2 specific growth rate (even if methanol does have some
3 inhibitory effect). However, even at this low growth
4
5
6 Table 1 Fed-batch model parameters
7
Maximum specific growth rate, mmax (h–1) 0.25
8 Cell yield on glycerol, YX/S (g dry wt/gS) 0.42
9 Monod saturation constant, KS (g/l) 0.005
50111 Feed glycerol concentration, So (g/l) 950
1 Initial volume, V(to) (l) 6.84
Initial cell density, X(to) (g dry wt/l) 16.2 Figure 1 Predicted and actual feed rate profile and
2 dissolved oxygen trace for fed-batch fermentation showing
Specific SR-AFP production rate, qP (ngP/gX· h) 5.1
3111 transition to reduced feed rate at 30% of saturation.

Biotechnology Techniques · Vol 11 · No 12 · 1997 867


J BT,865-868,460 d'Anjou 4/12/97 1:57 pm Page 868

M.C. d’Anjou and A.J. Daugulis

11111 the cell density is maximized prior to induction and


2 maintained at low growth conditions in the production
3 phase. As confirmation, we have also observed (data not
4 shown) that once the maximum cell density had been
5 reached no additional SR-AFP was produced despite
6 methanol being present in inducing concentrations.
7
8 Conclusion
9 A simple model was developed to predict growth and
10111 heterologous protein expression in recombinant Pichia
1 pastoris. The model was also used to develop a rational
2 two-stage exponential feeding strategy for a mixed-
3 substrate, fed-batch fermentation of a MutS strain. A
4 two-stage approach was necessary to avoid oxygen limi-
5 Figure 2 Predicted and experimental profiles for cell tation at higher cell densities and to allow the product
6 growth and protein production in fed-batch fermentation protein to accumulate to levels facilitating separation
7 using two stage feeding strategy. and purification. The model predicted cell growth with
8 reasonable accuracy, but the assumption of constant
9 specific heterologous protein production rate may have
20111 levels were evaluated off-line by HPLC as the run accounted for discrepancies between the model and
1 proceeded to ensure that no repressing levels of glyc- experimental data. It has been demonstrated that
2 erol had built-up, and that methanol was present at feeding strategies can be developed using a very simple
3 levels which induced heterologous protein expression model in order to improve process productivity in
4 but were non-inhibitory to growth. Glycerol was never fed-batch recombinant P. pastoris fermentations without
5 detected in the fermentation broth. Methanol was the need for complex model development and predic-
6 always present, but in levels (0.02–0.08%) much lower tive/feed-back compensated adaptive controller mecha-
7 than would be expected assuming that it was not being nisms.
8 consumed by the yeast. This is not entirely unexpected
9 as simultaneous utilization of methanol and some other References
30111 carbon source by methylotrophic yeast is well docu- Baheri, HR, Roesler, WJ, Hill, GA. (1997), Biotech. Techniques
1 mented (Egli et al., 1986). 11: 47–50.
Brierley, RA, Bussineau, C, Kosson, R, Melton, A, and Siegel,
2 RS. (1990), Ann. N.Y. Acad. Sci. 589: 350–362.
3 Cell growth was predicted by the model with reason- Chen, Y, Cino, J, Hart, G, Freedman, D, White, C, and Komives,
4 able accuracy. The contribution of methanol to cell EA. (1997), Process Biochem. 32: 107–111.
5 growth becomes more significant at lower growth rates Cregg, JM, Vedvick, TS, and Raschke, WC. (1993),
6 in methylotrophic yeast (Eggeling and Sahm, 1981). Bio/Technology. 11: 905–910.
Thus, the actual cell density is slightly higher than Eggeling, L, and Sahm, H. (1981). Arch. Microbiol. 130: 362–365.
7 Egli, T, Bosshard, C, and Hamer, G. (1986). Biotech. Bioeng. 28:
8 predicted by the model. Protein production was less 1735–1741.
9 well described, suggesting that the assumption of a Loewen, MC, Liu, X, Davies, PL, and Daugulis, AJ. (1997), Appl.
40111 constant specific expression rate may not be entirely Microbiol. Biotech. In press.
1 valid. The data suggest that the specific expression level Patkar, A, Seo, JH, and Lim, HC. (1993), Biotech. Bioeng. 41:
is higher at the higher growth rate and reduced during 1066–1074.
2 Rodriguez Jimenez, E, Sanchez, K, Roca, H, and Delgado, JM.
3 the slow growth period. This would preclude the use (1997), Biotech. Techniques 11: 461–466.
4 of more traditional approaches to fermentation opti- Romanos, M. (1995), Current Opinion in Biotech. 6: 527–533.
5 mization such as high cell density fermentation in which Siegel, RS, and Brierley, RA. (1989), Biotech. Bioeng. 34: 403–404
6
7
8 Received 22 August 1997;
9 Revisions requested 15 September 1997;
50111 Final Revisions received 30 September 1997;
1 Accepted 1 October 1997
2
3111

868 Biotechnology Techniques · Vol 11 · No 12 · 1997

Вам также может понравиться