|Home | About | Journals | Submit | Contact Us | Français|
Matching both the construction of a recombinant strain and the process design with the characteristics of the target protein has the potential to significantly enhance bioprocess performance, robustness, and reproducibility. The factors affecting the physiological state of recombinant Pichia pastoris Mut+ (methanol utilization-positive) strains and their cell membranes were quantified at the individual cell level using a combination of staining with fluorescent dyes and flow cytometric enumeration. Cell vitalities were found to range from 5 to 95% under various process conditions in high-cell-density fed-batch cultures, with strains producing either porcine trypsinogen or horseradish peroxidase extracellularly. Impaired cell vitality was observed to be the combined effect of production of recombinant protein, low pH, and high cell density. Vitality improved when any one of these stress factors was excluded. At a pH value of 4, which is commonly applied to counter proteolysis, recombinant strains exhibited severe physiological stress, whereas strains without heterologous genes were not affected. Physiologically compromised cells were also found to be increasingly sensitive to methanol when it accumulated in the culture broth. The magnitude of the response varied when different reporters were combined with either the native AOX1 promoter or its d6* variant, which differ in both strength and regulation. Finally, the quantitative assessment of the physiology of individual cells enables the implementation of innovative concepts in bioprocess development. Such concepts are in contrast to the frequently used paradigm, which always assumes a uniform cell population, because differentiation between the individual cells is not possible with methods commonly used.
Changes to the product quantity and quality as well as the robustness of bioprocesses can be triggered by a number of factors which affect the physiological state of the microorganisms. The expression of a foreign gene, the processing of the recombinant protein, and the exposure of the cell to metabolites, inductors, substrates, unfavorable environmental conditions, high cell density (HCD), and/or “aging” can all result in markedly different physiological responses (17).
The interpretation of bioprocess data often fails to reflect the actual state of individual cells within a population. It is based typically on performance characterization using concentration measurements and the simplifying assumption of uniform (“averaged”) performance of each cell. This practice (39) provides an example of the failure to distinguish between a homogenous population of equally compromised microorganisms and a heterogeneous population of cells each performing differently. An understanding of the state of the individual cells is critical in achieving high efficiency in recombinant protein production. Only by knowing the number of (vital) cells that express the target molecule at the highest possible rate can the number of such cells within the population be maximized and the proportion of dead cells undergoing lysis be kept to a minimum. Furthermore, lysed cells not only are “unproductive” but also may release undesirable host cell proteins and/or proteases causing product degradation (32, 33, 58).
Bacteria and yeasts can be readily differentiated by staining with appropriate fluorochromes and using epifluorescent microscopy. The combination of fluorescent staining and rapid counting using flow cytometry marked the advent of process monitoring, enabling the physiological state of both subpopulations and individual cells to be quantified and interpreted (22, 23, 37, 55). This particular technique is widely utilized in characterizing the physiology of mammalian cell cultures. However, its potential for monitoring processes with recombinant bacteria or yeasts has been reported in only a few studies (for example, for Escherichia coli, see references 24, 25, and 46, for Saccharomyces cerevisiae, see references 35 and 47, and for Pichia pastoris, see references 17, 27, 30, 44, and 59). The power of the technique lies in the use of specific, fluorescent dyes to discriminate between vital (i.e., active and viable) and compromised (or dead) cells. For instance, propidium iodide (PI), bis-(1,3-dibutylbarbituric acid)trimethine oxonol (BOX) (30), and ethidium bromide (EB) (34, 46) can be used to stain defective cells with permeabilized membranes and damaged membrane integrity, while 5(6)-carboxyfluorescein diacetate (cFDA) (26) is used to stain intact cells with nonspecific esterase activity.
The optimum pH value for Pichia growth is approximately 5 (6), but the pH value is commonly reduced to counteract proteolysis (9, 32) or to increase both the stability and productivity of heterologous proteins (5, 29, 56). However, a pH value of 4 or lower can compromise cells during the production process (27). Cells cultured at such a low pH can lose their vitality by any of the following mechanisms: (i) passive diffusion of nondissociated weak organic acids into the cell (43); (ii) changes in the structure of the cell wall (40); or (iii) active maintenance of the optimal intracellular pH at a level higher than that outside the cell (50). Furthermore, heterologous protein production may lead to severe stress on the physiology of host cells, resulting from various interrelated factors determined by the strain's molecular design, particularly codon usage (52), the number of gene copies (8, 12), and the promoters and/or chaperons used. The stress experienced by the cells during the production of a heterologous protein, which is usually detected as changes in cell staining properties, can occur at any stage during gene transcription and translation (4, 48), protein processing and folding (42, 49), and protein secretion by the cell (54).
Unfortunately, most of the published studies to date are concerned specifically with the effects of a single reporter gene expressed under the control of a specific promoter. Furthermore, the relationships among heterologous product formation, the metabolism of the host cell, and the respective strain construction are rarely described either systematically or in quantitative terms. Thus, the information available is insufficient to enable efficient development of bioprocesses and cloning strategies.
Data on subtle changes in the physiological states of four different recombinant P. pastoris strains are presented and interpreted in this paper. The effects of a “problematic” reporter (trypsinogen [TRP]) and a “harmless” reporter (horseradish peroxidase [HRP]) when combined with promoters differing in both strength and regulation are described. Comparisons of those effects with those of nonproducing cells without any of the heterologous genes are also included. Most interestingly, the combination of several factors affecting the vitality of recombinant cells was studied, and the changes resulting from the exclusion of any one of these factors were quantified.
Four recombinant Pichia pastoris strains and one nonrecombinant control were investigated in this study. The following two strains producing porcine trypsinogen and containing synthetic genes of porcine trypsinogen produced by GeneArt AG (Regensburg, Germany) were used: the P. pastoris X33 strain (i), using the AOX1 promoter transformed with a plasmid based on pPICZαA (Invitrogen, Carlsbad, CA), and the P. pastoris CBS 7435 strain (ii), using the d6* variant promoter (20) transformed with the unpublished plasmid pPpT2 of a synthetic AOX1 promoter (PAOX1) and a terminator, both synthesized by GenScript (Piscataway, NJ). Two strains producing horseradish peroxidase were used, in which the plasmid construct pPICZαB-HRP was integrated into the P. pastoris CBS 7435 strains (20), either with the AOX1 promoter (iii) or with the d6* promoter (iv). The wild-type X33 strain (v), not producing any recombinant protein, was used as a negative control. As both the X33 and CBS 7435 strains have the same genetic background (TU Graz, unpublished data) and physiological behavior, only the data on X33 as the negative control are presented.
The codon usage was optimized for P. pastoris. The selection of the transformants was based on the Zeocin (Invitrogen, Carlsbad, CA) resistance of the transformation vector. Each of the four selected recombinant strains showed a Mut+ (methanol utilization-positive) phenotype and contained the codon-optimized mating α-factor leader signal sequence of S. cerevisiae for extracellular protein secretion. The clones were screened in 96-deep-well microplates (monitoring product activity and growth) in minimal buffered medium with dextrose (BMD) (31) with (or without) methanol pulses every 12 h, which were provided to maintain induction (58). Finally, 0.5 to 1% of the clones evaluated in the first screening were characterized and used in the bioreactor cultivations. The stock cultures were conserved in 24% glycerol at −80°C.
All chemicals used in this study were of puriss grade p.a., purchased from Sigma-Aldrich (formerly Fluka, Buchs, Switzerland), unless stated otherwise. The buffered glycerol complex medium (BMGY) (31) used for the precultures contained 10 g glycerol, 10 g yeast extract, 20 g peptone, 100 mM potassium phosphate buffer (pH 6.0), 13.4 g yeast nitrogen base without amino acids, and 0.4 mg biotin per liter. The defined mineral medium used for both batch and fed-batch cultures contained 0.17 g CaSO4·2H2O, 2.86 g K2SO4, 0.64 g KOH, 2.3 g MgSO4·7H2O, 0.2 g EDTA, 7.23 g H3PO4, and 0.1 ml polypropylene glycol (PPG) per liter, which were all autoclaved, 4.35 ml filter-sterilized PTM1 solution, and 0.87 mg biotin per liter, which was added separately. The batch growth was achieved with 33 g glucose monohydrate per liter (Brenntag Schweizerhall, Switzerland).
In fed-batch cultures, one of the following feed solutions was used: (i) 784 g glucose monohydrate, 2.4 mg biotin, and 12 ml PTM1 solution per liter; (ii) 346 g glucose monohydrate, 452 g methanol, 2.4 mg biotin, and 12 ml PTM1 solution per liter; or (iii) 773 g methanol, 2.4 mg biotin, and 12 ml PTM1 solution per liter. The PTM1 stock solution (31) contained 5.0 ml 69% H2SO4, 3.84 g CuSO4, 0.08 g NaI, 3.0 g MnSO4·H2O, 0.2 g Na2MoO4·2H2O, 0.02 g H3BO3, 0.92 g CoCl2·6H2O, 20.0 g ZnCl2, and 65.0 g FeSO4·7H2O per liter.
A glycerol stock (1.8 ml) of each strain was thawed and used to inoculate a shake flask with 200 ml of BMGY medium. This first seed culture was grown over 24 h at 30°C and 150 rpm, during which the optical density at 600 nm (OD600) increased from 0.1 to 57. From this cell suspension, 60 ml was transferred into 600 ml of BMGY medium, with the objective of achieving an inoculum density of 10%. These second precultures were grown for 12 h at 30°C and 150 rpm, during which the optical density at 600 nm increased from approximately 6 to 30.
All cultivation processes were performed in a 14-liter stirred tank bioreactor (volumetric oxygen transfer coefficient [kLa], >125 h−1; Mavag, Neunkirch, Switzerland) at a constant temperature of 30°C, 18 liters per minute airflow (without any oxygen enrichment), 1,400 rpm agitator speed, and 0.5 bar overpressure. Ammonia solution (25%) and phosphoric acid (8.5%) were used to maintain the pH value at a constant value of 6 during biomass growth and at either 4 or 5 during the production phase.
The fed-batch process comprised a phase of biomass growth in both the batch and fed-batch modes and a production phase in the fed-batch mode. The process was initiated with a batch culture of 6 liters (working volume), containing 33 g of glucose monohydrate per liter (Fig. (Fig.1,1, phase A). The relative partial pressure of oxygen (pO2) in the medium was not regulated, which resulted in a continuous decrease in the pO2 level during the batch culture. After between 12 and 14 h, the pO2 value increased rapidly, at which point the exponential addition of glucose solution was initiated and continued over 13.5 h, consistent with the following function: f(t) = 4.66·e0.2·t g of carbon h−1 (Fig. (Fig.1,1, phase B). A glucose feed solution was then added at a constant rate of 34.5 g of carbon h−1 for 1 h, while the pH value was adjusted simultaneously. Within 45 min, the pH value was adjusted from pH 6 to the specific pH of the production phase and was then maintained at the final pH for 15 min (Fig. (Fig.1,1, phase B*).
A mixture of glucose and methanol was added during the subsequent production phase in one of the following modes: (i) a constant 43%/57% glucose/methanol mixture (these percentages being determined by the carbon content) added at a constant rate of 29.5 g of carbon h−1 for at least 45 h (Fig. (Fig.1,1, left, phase C) or (ii) two separate solutions added simultaneously at a constant feed rate of 12.6 g of carbon h−1 for pure glucose and a stepwise decreasing feed rate from 17.0 to 2.1 g of carbon h−1 for pure methanol (Fig. (Fig.1,1, right, phase C). The composition of the mixture was determined in accordance with the published data (2, 16). These data show that a proportion of methanol of approximately 10 to 20% in a mixture with glucose is sufficient to fully induce any enzyme of methanol catabolism and that alcohol oxidase is not repressed by glucose during the addition of carbon-limited substrate.
Chip-based separations were performed on LabChip 90 (Caliper Life Sciences, Hopkinton, MA) in combination with the HT protein express LabChip 90 assay kit and the protein assay software LabChip HT. All chips were prepared according to the protocol provided with the LabChip 90 assay kit. Samples (2 μl) were diluted in a sample buffer (with or without β-mercaptoethanol; total volume, 7 μl) containing a marker (HT protein express ladder) used for semiquantitative analysis. The samples were incubated for 5 min at 95°C. After spinning the tubes, the samples were diluted with demineralized water (35 μl) and loaded onto a chip filled with the gel-dye mixture. The separated proteins were detected by laser-induced fluorescence. All reagents, including the standard protein ladder containing different proteins with known concentrations and molecular masses, were obtained with the HT protein express LabChip 90 assay kit. Caliper LabChip 90 was controlled by LabChip HT software. GX/GXII software was used to collect and report the data.
The method of protein determination described above was verified using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), with silver staining. Polyacrylamide gels (15%) were prepared according to Laemmli's procedure (45). The supernatant of the sample (7 μl) was mixed with 1.75 μl Laemmli buffer (each 100 ml containing 3.75 g Tris, 10 g SDS, 35 ml glycerol, 5 mg bromophenol blue, and 5 ml β-mercaptoethanol, with 37% HCl added to achieve pH 6.8) and was then denatured for 5 min at 95°C. The molecular masses of horseradish peroxidase and trypsinogen are 38.8 kDa and 24.4 kDa, respectively.
Porcine trypsinogen was quantified using a trypsin activity assay. The cell-free supernatant from cultivations was assayed either directly or after incubation with 1 g enterokinase (EC 188.8.131.52; Sigma-Aldrich E-0632) per liter at 37°C and pH 8.2 for 4 h. The incubation mixture of 220 μl triethanolamine (TEA) buffer, 100 μl sample supernatant, and 10 μl enterokinase solution was mixed thoroughly. Whereas the activity of only the free proteases, including trypsin, was detected by direct measurement, trypsin or other proteases were released from their inactive precursors after enterokinase treatment. Inactive trypsinogen was quantified by subtracting the value for the activity of free proteases from the value obtained after sample activation. This spectrophotometric method using the chromogenic substrate N-α-benzoyl-l-arginine-p-nitroanilide (BApNA) was utilized for routine process monitoring (36). One unit of trypsin is defined as the amount of enzyme that hydrolyzes 1 μmol of BApNA to p-nitroaniline per minute at 25°C and pH 8.2. The release of p-nitroaniline was measured photometrically at 405 nm.
The activity of horseradish peroxidase (HRP) was determined as described in the literature (9). The activity of intracellular protein was measured in the supernatant of disintegrated cells, which required the cell pellets to be first washed three times in phosphate-buffered saline (PBS) and then dispersed by ultrasonication on ice (7 min, 75 W) (Sonoplus HD 3100, TT 13FZ; Bandelin, Zurich, Switzerland). The enzyme activities were measured in the supernatant after centrifugation (10 min at 14,000 rpm and 4°C) (5417C; Nethler, Hamburg, Germany).
The relative partial pressure of oxygen (pO2) in the medium, concentrations of both CO2 and O2 in the exhaust gas (extended process gas analyzer; Biospectra AG, Schlieren, Switzerland), pH value, temperature, reactor overpressure, and reactor weight were all monitored on-line. The biomass concentration was determined gravimetrically as cell dry weight (CDW). Samples were centrifuged for 5 min at 14,000 rpm (5417C; Nethler, Hamburg, Germany) in preweighed 2-ml Eppendorf tubes, which had been washed, recentrifuged, and subsequently dried to a constant weight at 105°C (Heraeus Instruments, Zurich, Switzerland). The concentrations of glucose, methanol, and other metabolites were determined by high-pressure liquid chromatography (HPLC) using an LC-20AB device equipped with the autosampler SIL-20A, thermostated column oven CTO-20A, and refractometer detector RID-10A (all produced by Shimadzu). The Aminex HPX-87H column, with an inside diameter (i.d.) of 300 by 7.8 mm (Bio-Rad, Munich, Germany), was run at 40°C, with a flow rate of 0.6 ml min−1 under isocratic conditions, with 2.5 mmol H2SO4 and an injection volume of 25 μl.
Cells stained with different fluorescent dyes were analyzed using a BD FACSCalibur 4CA flow cytometer (Becton Dickinson GmbH, Heidelberg, Germany) equipped with a 15-mW 488-nm argon laser, a 635-nm diode laser, and the appropriate FACSFlow sheath fluid. Samples were taken directly from the culture and were diluted with phosphate-buffered saline (PBS). The fluorescence emitted was collected in four optical channels, FL1 (515 to 545 nm), FL2 (564 to 606 nm), FL3 (>650 nm), and FL4 (>670 nm). BD CellQuest and BD FACSComp software were used in the cytometer, while the data were analyzed using FlowJo software (Tree Star Inc.).
The samples were diluted with PBS to an OD600 of 0.1. The cell suspension of P. pastoris was incubated in the BD measuring tube, with the fluorescent dye protected from light, and was then measured directly. The following four fluorescent dyes were each applied: (i) ethidium bromide (EB) (stock solution of 10 g liter−1, final concentration with cell suspension of 0.01 g liter−1, and 10 min of incubation time, triggered by FL3; Fluka AG); (ii) propidium iodide (PI) (stock solution in demi water of 1 g liter−1, final concentration with cell suspension of 0.01 g liter−1, and 5 min of incubation time, triggered by FL3; Molecular Probes); (iii) bis-(1,3-dibutylbarbituric acid)trimethine oxonol (BOX) (stock solution in demi water of 0.5 g liter−1, final concentration with cell suspension of 2 mg liter−1, and 10 min of incubation time, triggered by FL1; Molecular Probes); and (iv) 5(6)-carboxyfluorescein diacetate (cFDA) (stock solution in dimethyl sulfoxide [DMSO] of 2.5 g liter−1, final concentration with cell suspension of 5 mg liter−1, and 10 min at 37°C followed by 10 min of incubation on ice, with cells subsequently diluted with buffer containing 10 mM HEPES, 150 mM NaCl, and 100 mM EDTA, instead of PBS, triggered by FL1; Molecular Probes).
All samples were analyzed at a low flow speed of 12 ± 3 μl min−1 in the range of 300 to 600 particles per second and with the threshold set at the side light scatter (SSC). The samples were treated with 70% ethanol for 20 to 30 min to provide positive staining controls, thereby allowing detection of dead cells.
Physiological changes in recombinant P. pastoris were observed both across different high-cell-density fed-batch processes and during the course of each process. The processes which were compared differed only with respect to (i) the extracellularly secreted recombinant protein (reporter), (ii) the promoter used, and/or (iii) the process conditions, including the pH, pattern of carbon substrate addition, and extracellular methanol accumulation. Cells not containing foreign genes, and therefore, not producing any heterologous proteins under production conditions, were also evaluated in several control experiments.
The physiological state of cells was determined by flow cytometric enumeration of the cells stained with fluorescent dyes (Fig. (Fig.2)2) such as ethidium bromide (EB), propidium iodide (PI), bis-(1,3-dibutylbarbituric acid)trimethine oxonol (BOX), and 5(6)-carboxyfluorescein diacetate (cFDA). Conventional monitoring of the concentrations of the metabolites (including CO2 and O2 data obtained from off-gas analyses) (Fig. (Fig.3)3) complemented the physiological information gathered.
When cells were in an active, vital physiological state, 90 to 100% were stained with cFDA, and only 0 to 10% were stained with PI, EB, or BOX. These activity levels were independently confirmed on cells grown in batch cultures with glycerol in either mineral or complex media at pH values of 6, 5, or 4. Even nonproducing cells in stationary phase that were starved of carbon substrates for 24 h at 30°C showed an unchanged staining pattern after exposure to either 40 g liter−1 methanol or 1 g liter−1 trypsin (a serine protease). In addition, several positive-control experiments were performed with “healthy” cells extracted during exponential phase of batch culture (data not shown). These cells were treated with 70% ethanol for 20 min to ensure death. All control experiments confirmed the suitability of the staining method and demonstrated good discrimination between the subpopulations of active and compromised cells during the production process (Fig. (Fig.44).
Each of the fluorescent dyes tested exhibited a similar response to the change in the physiological state of the cells producing either porcine trypsinogen or horseradish peroxidase (Fig. (Fig.5).5). The trend in vitality loss was independent of the stain applied and confirmed by data collected under different conditions. The various slopes in the graphical representations illustrate how the loss of vitality proceeds (Fig. (Fig.5).5). Initially, the cells become stained with EB and/or BOX because of defects in the transport system, with the damage and resulting membrane permeabilization being indicated by the number of cells stained with PI. Consequently, the slope of the graph correlating EB (or BOX) with PI becomes steeper as the cells become less vital. Deviations from the direct proportionality between the number of cells stained with PI and those stained with EB or cFDA were observed when there was insufficient separation between compromised and active cells, as determined by flow cytometry analysis (Fig. (Fig.4).4). Standard deviations for all cell counts were below 3% (data not shown), both for triplicate samples and for samples from different fed-batch processes replicated under the same conditions.
Both autocatalytic activation and further degradation of the trypsinogen produced are triggered by the presence of trypsin in the culture supernatant. The (free) protease formed also causes physiological stress on cells. However, trypsinogen (24.4 kDa) and trypsin (23.5 kDa) can neither be separated nor distinguished on a standard gel (Fig. (Fig.6),6), necessitating the presence or absence of the active form (trypsin) to be established by protease activity assay. The production of recombinant porcine trypsinogen at the moderate pH of 5 was hampered by accelerated product degradation (autoproteolysis), which resulted in a severe decline in the production of protein observed shortly after induction. Degradation could be detected at about 7 h, with the highest product titer being reached at 10 to 13 h (Fig. (Fig.66 and and7).7). The common practice of inactivating proteolytic enzymes secreted in the culture broth to avoid product degradation was adopted. As a result, the pH during the production phase was decreased to a set point of 4 while maintaining the same profile of substrate addition (Fig. (Fig.1).1). Although this change in pH contributed to a reduced level of product degradation, the expression levels decreased by a factor of approximately 5 (Fig. (Fig.77).
Moreover, approximately 12 h after induction, methanol started to accumulate in the culture broth, increasing in concentration to 80 g liter−1 in the subsequent 37 h (Fig. (Fig.8,8, strain producing porcine trypsinogen by the AOX1 promoter at pH 4). The most pronounced changes in the physiological state of the cells were also detected under these culture conditions. At 30 h after induction, densities higher than 85 g liter−1 CDW were achieved, and more than 80% of the cells exhibited compromised membranes, defective transport systems, and reduced enzyme activity. The following three distinct phases were identified with respect to the specific growth (and death) rate using biomass data: 0.031 ± 0.005 h−1 at 0 to 10 h, 0.006 ± 0.002 h−1 at 10 to 32 h, and −0.006 ± 0.004 h−1 at 32 to 50 h after induction. This gradual decline in biomass growth, which continued until its total collapse, also correlated well with data on carbon dioxide production rates (Fig. (Fig.33).
Four different recombinant strains and two strains not containing foreign genes were cultured with each of two different modes of substrate addition at pH 4 or pH 5. Both the individual and combined effects of the target molecule, the strain construction, and the process conditions led to noticeable changes in vitality. These changes in physiology were quantified by comparing the differences in the percentages of cells stained with propidium iodide in pairs of processes. Comparisons were made between the following: (i) strains differing in respect to the reporter molecule only (Δr), (ii) strains producing the same heterologous protein but expressed under the influence of different promoters (Δp), (iii) cultures with and without methanol accumulation in the culture supernatants (Δm), and (iv) cultures grown at different pH set points during the production phase (Δh).The offsets (Δ) are shown in Fig. Fig.2,2, and the processes were characterized in Table Table11.
Severe physiological stress was observed as a result of the combined effects of high cell density, production of recombinant protein, and low pH. Vitality improved when any one of these stress factors, which impair the physiological state of cells when combined, was excluded (Fig. (Fig.2).2). Decreased vitality was experienced by cells secreting porcine trypsinogen at a pH value of 4 in high-cell-density culture (>80 g liter−1 CDW and >80% PI-stained cells) (Fig. (Fig.8)8) but not in continuous culture at a lower cell density under comparable cultivation conditions (<25 g liter−1 CDW and <3% PI-stained cells) (data not shown). The effects of pH on strains producing porcine trypsinogen were higher than on those producing horseradish peroxidase. However, cells expressing the porcine trypsinogen gene with a novel AOX1-derived promoter variant (d6*) were generally less compromised than those expressing the porcine trypsinogen gene with the ordinary AOX1 promoter. The positive effect of the optimized d6* promoter was not detected with horseradish peroxidase. This reporter affects the cell physiology less, even though it contains an active (cofactor-dependent) heme site, which presents challenges at both strain construction and protein production levels. Furthermore, physiological stress was not observed by any strain producing either porcine trypsinogen or horseradish peroxidase at a pH value higher than 4 and by any strains lacking foreign genes (Fig. (Fig.22).
Typically, valuable metabolic information used for conventional process monitoring is obtained from the carbon dioxide production rate (CPR), oxygen uptake rate (OUR), and the respiration quotient (RQ) (Fig. (Fig.3).3). However, information on cell vitality was not readily attainable from these analyses unless cells were fully compromised and did not show any metabolic activity, including that of CO2 production (Fig. (Fig.3,3, curve 1). Moreover, the effect of different promoters on cells producing porcine trypsinogen (Δp) and the different sensitivities to pH caused by the various promoters (Δh1 and Δh2) was not detected by exhaust gas analyses (Fig. (Fig.33).
To counter the accelerated cell death resulting from methanol accumulation in HCD fed-batch cultures (Fig. (Fig.2),2), the addition of methanol was reduced with time (Fig. (Fig.1,1, right). As the methanol added was immediately and completely utilized, the absence of methanol accumulation led to a vitality improvement of over 30% (Fig. (Fig.2,2, Δm1). Furthermore, the effects of heterologous protein production at low pH were separated from those of extracellular methanol concentration. More than 55% of compromised cells were still detected at pH 4, even when no methanol had accumulated. It can therefore be deduced that methanol accumulation itself contributes in part to the total decrease in vitality, which is also affected by other factors such as low pH and recombinant protein production.
In order to further understand the sensitivity of cells to methanol, several control experiments were completed. The percentage of PI-stained and BOX-stained cells was assessed at defined times in fed-batch or batch cultures under different pH conditions (Fig. (Fig.9).9). However, the physiological effects were less pronounced at methanol concentrations lower than 150 g liter−1 and, in some cases, could not be detected (data not shown). In control experiments using identical samples but without methanol addition, the percentage of stained cells was below 10% after 5 h of incubation. Furthermore, both cultures with more than 100 g liter−1 biomass cell dry weight and cultures with 10-times diluted biomass concentrations were spiked with methanol (Fig. (Fig.9),9), resulting in inhibited biomass growth over the entire 5 h of exposure in both cases. It can be concluded from these control experiments that the incubation conditions did not affect the cell staining properties and that the effects of extracellular methanol concentrations, rather than possible effects of higher methanol utilization or increased oxygen depletion, were simulated.
Cells which had already been induced by methanol and simultaneously produced the recombinant protein were more than three times more sensitive to methanol than those assayed before their induction (Fig. (Fig.9,9, left). Cells grown in batch culture at pH 4 exhibited more than twice the sensitivity to a higher concentration of extracellular methanol than cells grown at a pH of 5 or 6 for the entire batch growth phase of over 12 h (Fig. (Fig.9,9, right). However, no similar effect of pH on samples taken from carbon-limited fed-batch cultures was detected, where cells were exposed to a pH of 4 for less than 5 h (Fig. (Fig.8,8, sampling points a and b).
The results presented contribute to our understanding of the factors affecting cell “stress” in microorganisms that express foreign genes (17, 28, 48). An innovative application of flow cytometry enabled a more thorough understanding of the implications of the molecular biology for development of the biotechnological processes in high-cell-density fed-batch cultures. By comparing five different P. pastoris strains, it was possible to identify the individual and combined effects of the reporter molecule, the promoter, the culture history, and the process conditions on vitality at the single-cell level. The common practices, currently utilized in bioprocess monitoring and optimization, fail to provide information on the physiological heterogeneity within a population. Our approach offers significant opportunities for defining physiological states and adjusting both culture conditions and strain design for improved performance.
Typically, only one or two process strategies for each particular strain investigated are reported in the literature (5, 41, 59), and conclusions on the combination of factors affecting cell vitality are seldom possible. Although it is thought that the physiological effects attributed to individual factors are cumulative, it was found that a combination of several factors exerts a synergistic effect on cell vitality.
The combined impact of low pH and recombinant protein production on the physiological state of cells has reportedly caused a decrease in vitality of over 60% of the compromised cells (as was the case with human trypsinogen , porcine trypsinogen, and horseradish peroxidase). When cells were exposed to substances accumulated in the culture broth, severe physiological stress was generally detected in high-density fed-batch cultures but not in continuous cultures, despite the cultivation conditions being comparable (specific growth rate, pH, temperature, and substrate). Hewitt et al. (24, 25) observed a decrease in the viability of nonrecombinant E. coli of approximately 15% in fed-batch cultures and an absence of dead cells in batch and continuous cultures. They concluded that this phenomenon is due either to increased glucose limitation or to the accumulation of toxic by-products in fed-batch cultures. The higher vitality of cells in continuous cultivation can be attributed to the absence of accumulated substances that, unlike in fed-batch processes, are continuously washed out. In addition, fed-batch populations are heterogeneous, having cells of different ages and shapes, unlike continuous cultures, which are of a more uniform age. As aged cells are reported to be more sensitive to expression stress (17), vitality changes of between 0 and 15% observed in fed-batch processes could potentially be attributed to a simple aging phenomenon.
The magnitude of the physiological stress is highly dependent on the protein produced, as confirmed by the data reported here. The healthy physiological state of P. pastoris cultures expressing the reporter gene after induction by methanol (and with methanol used simultaneously for cell growth) may be threatened by either the toxicity of the accumulated methanol or intermediates formed during methanol catabolism (29, 41, 46). The observation that cells which had previously been exposed to (and simultaneously utilized) methanol are more sensitive to any additional stress factor introduced subsequently is consistent with previous results. Jungo et al. (38) found that Pichia cells previously exposed to an excess of methanol have already “learned” to synthesize the enzymes of the methanol catabolism. They are therefore again able to synthesize these enzymes immediately, increasing the formation and secretion of toxic intermediates, such as formaldehyde, hydrogen peroxide, and formiate.
Thus, the common concept of a toxicity threshold for extracellular methanol concentration may be questioned. However, production strategies under AOX1 promoter control of Mut+ strains usually require control of methanol addition to achieve residual (extracellular) concentrations ranging from 1 to 5 g liter−1 (10, 11, 19, 21, 57, 60). Concentrations of more than 3.5 g liter−1 are reported to inhibit biomass growth (60), while concentrations of more than 8 g liter−1 cause a decrease in productivity (57).
Both strain and process designs are still typically completed on a de novo protein-by-protein basis and, thus, in a highly time-consuming manner. The impact of strain construction on cell physiology in general and on the posttranslational processing of recombinant proteins in particular is less well understood. For example, the intracellular accumulation of the product, which can be detected by immunofluorescent staining, is considered to be the main reason for a vitality loss (as in human trypsinogen ). In contrast, product yields of both porcine trypsinogen and horseradish peroxidase in this study were determined to be five times higher at a pH value of 5 than at a pH value of 4. Additionally, at pH 5, the more productive cells also exhibited significantly enhanced vitality, and no intracellular product accumulation was detected. Thus, it can be concluded that overexpression alone does not always lead to defective protein processing, with accompanying cell stress and vitality loss.
Problematic or cell-toxic protein may not benefit from the strength of the natural Pichia assembly. However, the current availability of a variety of novel (synthetic) promoters (20) enables the patterns of product formation to be tightly controlled over the process time, consistent with the features of the particular reporter molecule. Thus, negative effects associated with problematic reporters may be reduced by using a suitable promoter. Interestingly, the synthetic d6* promoter (20) evaluated with porcine trypsinogen in the experiments reported caused measurable vitality improvements compared to the AOX1 promoter. However, the choice of promoter had only a minor impact on the strains producing the less problematic horseradish peroxidase. Further systematic investigations should therefore be completed, from which generic approaches suitable for specific product (protein) families can be developed.
There is a wide range of fluorescent dyes available, which enable assessment of both changes to the membrane structures and to the physiology of individual cells in different ways. However, staining by PI remains the gold standard for the determination of vitality (3, 7, 27, 41, 46, 59). The results obtained with using other vitality dyes (cFDA, EB, and BOX) usually correlate with those obtained with using PI. Physiological changes which are less pronounced can usually be detected only with more sensitive dyes (EB) and not with PI (46).
Each individual testing system, with its own combination of fluorescent dye, microorganism, medium matrix, and flow cytometry-measuring device, behaves differently. Therefore, even the routine use of well-established dyes is ineffective without an appropriate adjustment to the method, particularly with regard to the dye concentration, the incubation conditions, and the incubation time. As the physiological characteristics of cells can also change during any sample handling procedure, the reliability and reproducibility of the results should be confirmed with appropriate control experiments and at least triplicate analyses (1, 30, 51).
In addition to conventional monitoring, a combination of fluorescent staining and flow cytometric enumeration can provide information on vitality at the single-cell level. It can be used for either process monitoring, instantaneous process control (39), or “simply” for assessment and interpretation of the physiological state of cells (22). Microarray technology (18) or protein identification via mass spectrometry (14), which has been advanced due to the availability of the Pichia genome sequence (13), can deliver complementary information on the physiological state. These data are invaluable for enhancing understanding of the molecular basis and also for the engineering of strains with improved, customized properties.
Moreover, the opportunity to adapt flow cytometric devices for on-line use (15) may enable their implementation in advanced control systems for bioprocesses. As it is also desirable to combat unwanted, often irreversible, process changes in microbial cultivations in a timely manner (or even to prevent such changes), traditional microbial and analytical methods are of limited use. For instance, conventional vitality determination by CFU suffers from numerous experimental drawbacks, and consequently, not all active cells can be detected by this method (22). Moreover, the information required is provided with considerable delay, typically 24 h or more. In contrast, when using flow cytometry to assess the physiological state of cells, changes in product formation and the unwanted accumulation of metabolites were predicted several hours in advance (46), that is, before they actually occurred and prior to being detected analytically (Fig. (Fig.88).
Achievement of as high a proportion of active/vital cells as possible in a population and maintenance of productive cells over as long a time period as possible should be the objective of further improvements to strain design and cultivation processes. This segregated approach (27, 53), which implies that only part of the total biomass contributes to product formation, justifies the implementation of a vitality assessment as part of conventional process monitoring.
We kindly thank David Langenegger, Sandra Núñez, and Bindu Philip for their help in performing the long-term cultivations, Roger Fehér, Edith Lang, and Caspar Demuth for performing HPLC analyses, and Maggi Lussi Bell and Stella Cook for proofreading and editing the manuscript.
Published ahead of print on 14 May 2010.
†The authors have paid a fee to allow immediate free access to this article.