Reliable high-throughput screening with Pichia pastoris by limiting yeast cell death phenomena


*Corresponding author. Tel.: +43 316 873 8422/4074; fax: +43 316 873 8430/9302, E-mail address:


Comparative screening of gene expression libraries employing the potent industrial host Pichia pastoris for improving recombinant eukaryotic enzymes by protein engineering was an unsolved task. We simplified the protocol for protein expression by P. pastoris and scaled it down to 0.5-ml cultures. Optimising standard growth conditions and procedures, programmed cell death and necrosis of P. pastoris in microscale cultures were diminished. Uniform cell growth in 96-deep-well plates now allows for high-throughput protein expression and screening for improved enzyme variants. Furthermore, the change from one host for protein engineering to another host for enzyme production becomes dispensable, and this accelerates the protein breeding cycles and makes predictions for large-scale production more accurate.


The methylotrophic yeast Pichia pastoris represents a powerful host for recombinant protein production [1]. However, the enterobacterium Escherichia coli was still considered as a preferred microorganism for protein engineering, the generation of mutant libraries of heterologous proteins and in many cases also for studies of cell physiology [2]. Applying E. coli as a host, both expression of proteins as well as screening for desired or even unknown enzyme activities has been scaled-down to microscale, making these processes amenable for automatisation in a high-throughput manner [3–5]. This development can be regarded as a committed step for the successful high-throughput expression of designated open reading frames derived from genome data [6], protein engineering by directed evolution [7], and production of therapeutics [8].

The drawback of prokaryotic protein expression is the lack of a eukaryotic machinery, which obviates the heterologous expression of correctly folded and processed proteins from a variety of eukaryotic genes in E. coli. The importance of this inadequacy is emphasized by two recent developments, the publication of draft sequences of eukaryotic genomes (including the human one), and the emerging market of engineered eukaryotic proteins.

The development of microscale cultivation systems for eukaryotes, ranging from yeasts as Saccharomyces cerevisiae[6] to animal cells [9], complies with the requirements for large-scale expression of genomic sequences. One article presents a microscale cultivation procedure for the host organism P. pastoris[10], but its value for comparative high-throughput screening for protein engineering has to be doubted due to the lack of data addressing standard deviations of expression levels across the whole plate.

Protein engineering through rational and/or random approaches offers a way to optimize enzymes rapidly. During the last few years, directed evolution techniques have been developed and used to solve ever more enzyme engineering problems [11]. The generation of large mutant gene libraries which are screened for variants with improved properties is enabled through a variety of methods, such as the use of degenerate oligonucleotides, chemical mutagenesis, passage of DNA through mutagenic strains, error-prone PCR as well as DNA-shuffling and recombination practices [12]. So it is equitable to say that the creation of mutant libraries almost becomes routine, whereas the outcome of directed evolution experiments critically depends on the means of expressing and screening the mutant libraries [13]. Actually, the most essential experimental effort of directed evolution is to devise, validate and implement a suitable screen, following the first law of random mutagenesis: “you get what you screen for”[7]. Many industrially important enzymes have been successfully improved in E. coli. Still, the need for functional expression of proteins that are not sufficiently well expressed in this host is a serious bottleneck that has kept several enzymes out of the evolutionist's reach [7,14].

The success of P. pastoris for the production of heterologous eukaryotic enzymes is mainly due to its capability of introducing posttranslational modifications, correct folding of eukaryotic proteins and growth at high cell densities, respectively [15]. Nevertheless, the creation of multiple variants through, e.g., directed evolution experiments and expression of the corresponding mutants can hardly be done in this yeast. Circumstantial cultivation, high oxygen demand and handling procedures of P. pastoris have limited its applicability for high-throughput experiments. In addition to simplifying the handling procedures and downscaling of recombinant protein production in P. pastoris, we optimised cell viabilities in 96 deep-well plate cultures in order to achieve high levels of active recombinant protein with a low deviation in expression levels from well to well. Only viable cells contribute to high-level protein expression and provide information about protein properties, which is predictive for enzyme variants, produced under controlled conditions on a large scale.

Apoptosis in yeast is one form of programmed cell death and its regulators and effectors have been intensely studied [16]. Apoptotic cells cannot maintain their full physiological activity, but are still not necrotic. Although several inducers have been well characterized, the effect of varying glucose concentrations in culture media on yeast apoptosis is still unclear. In this study, we describe the development of a reliable high-throughput expression and screening system for recombinant protein production in P. pastoris by employing apoptosis and cell necrosis as markers for mandatory media optimisations.

2Materials and methods

2.1Microbial cloning and reporter strains

Standard molecular-biology procedures were performed according to [17]. E. coli XL-1 blue (Stratagene, La Jolla, CA, USA) was used for all E. coli cloning experiments.

P. pastoris strains GS115 (his4) and X-33 (both Invitrogen, Carlsbad, CA, USA) were used as hosts for the yeast experiments. The strain P. pastoris GS115 PamHNL5-a37 [18] expressing the secreted enzyme hydroxynitrile lyase (HNL) from Prunus amygdalus, PaHNL5, the strain P. pastoris GS115 PpD1-17 [19] with the gene for expression of intracellular HbHNL and P. pastoris GS115 PaHNL5_L1Q [20] which shows an optimized N-terminal sequence for high level expression, were used as reporter strains for protein expression. An additional reporter plasmid pPICZalphaB-HRP-WT for the expression of secreted horseradish peroxidase (HRP) was obtained as a kind gift of Frances H. Arnold (Caltech, USA). This plasmid was transformed into the strain X-33 (see 2.3), yielding the new strain P. pastoris X-33 pPICZalphaB-HRP-WT.

2.2Chemicals and media

Unless otherwise stated, all chemicals were purchased from Carl Roth GmbH (Karlsruhe, Germany) and Becton, Dickinson and Company (Franklin Lakes, NJ, USA), respectively. Sterile water was purchased from Fresenius Kabi (Graz, Austria).

Unless specifically mentioned, all culture media and ingredients were prepared according to the protocol from the Pichia protein expression Kit (Invitrogen, Carlsbad, CA, USA). All shake-flask cultivations were carried out in wide-necked, baffled shake flasks covered with two layers of cotton cloth.

Yeast cultures were either grown in YPD medium (1% w/v yeast extract, 2% w/v peptone and 2% w/v glucose), minimal dextrose (MD) medium (1.34% Yeast Nitrogen Base YNB, 4 × 10−5% biotin and 0.2%, 1%, 2% or 3% glucose), minimal methanol (MM) medium (1.34% YNB, 4 × 10−5% biotin and 0.5% methanol), buffered MD (BMD) medium (containing 200 mM potassium phosphate buffer, pH 6.0) or buffered MM media with doubled (BMM2 containing 1% methanol)) or tenfold (BMM10 containing 5% methanol) concentration of methanol compared to MM. An additional “H” indicates the addition of 0.004% histidine to ensure growth of histidine-auxotrophic strains, as e.g., BMDH for P. pastoris GS115. Media for plates were solidified by addition of agar to 1.5% w/v.

Transformed X-33 cells were selected on YPD agarplates supplemented with 100 μg ml−1 Zeocin® (Cayla, Toulouse, France).

2.3Transformation of P. pastoris and selection for positive clones

P. pastoris was transformed using the standard electroporation protocol according to the “Pichia Expression Kit” (Invitrogen). Plasmid DNA (10 μg) was linearized using a restriction enzyme, e.g., BglII or SacI (both purchased from NEB, Beverly, MA, USA) for integration of the expression cassette into the genome of P. pastoris and desalted via dialysis using nitrocellulose filters (0.0025 μm, Millipore, Billerica, MA, USA) against sterile water for 60 min at room temperature.

After transformation, aliquots of 100 μl were plated on YPD agarplates supplemented with 100 μg ml−1 Zeocin and incubated for 2 d at 30 °C.

The presence of the expression cassette in the genome of P. pastoris was confirmed by colony PCR. Zeocin-resistant clones were replated on selective media for colony PCR analysis. A single colony was resuspended in 100 μl sterile water, heated to 95 °C for 5 min and centrifuged at top speed in a tabletop centrifuge for 1 min the supernatants, 10 μl served as template for a 50 μl reaction, containing 0.2 mM dNTPs, 1× reaction buffer (Qiagen, Hilden, Germany), 1.2 U HotStar Taq polymerase (Qiagen), 200 nM each of the primers PP5AOX1 (5′ GACTGGTTCCAATTGACAAGC 3′) and 3TTAOX1rev1 (5′ GGTGCCTGACTGCGTTAGC 3′). The following program was used for PCR: 15 min at 95 °C, 30 cycles with 30 s at 95 °C, 1 min at 55 °C and 2.5 min at 72 °C, followed by a final extension step of 10 min at 72 °C. The identity of the resulting PCR products was verified by DNA sequencing.

2.4Choice of cultivation conditions in micro-scale deep-well plates

P. pastoris GS115 and X-33 were grown in shake flasks in MDH to an OD600 of 2. The wells of the deep-well plate (Bel-Art Scienceware, Pequannock, NJ, USA) were then filled with 200, 300, 400, 500, 600, 700 and 800 μl of the cultures, always spaced by wells with 500 μl of sterile medium, covered and shaken with 340 rpm at 28 °C and 80% humidity. After 30 min, 4, 12 h, 1, 2, 3, 4 and 5 d, respectively, the covers and wells were checked for liquid spillings and volume loss in wells in the periphery of the plates. Furthermore, the different culture volumes were assessed for pelleted cells semi-quantitatively by eye.

2.5Optimization of glucose concentration and protein expression

Deep-well plates were filled with 250 μl BMD, BMDH or YPD (glucose concentrations were 0.2%, 1%, 2% and 3%) and inoculated from individual colonies of GS115 and GS115 PamHNL5-a37. The cultures were grown under standard conditions (28 °C, 340 rpm and 80% humidity) for 60 h. Then 250 μl BMM2 were added. For further feeding with methanol, 50 μl BMM10 were admixed 70, 82 and 106 h after inoculation. At 24 h after the last induction (i.e., a total cultivation time of 130 h) the cells were harvested by centrifugation at 4000 rpm and 4 °C for 10 min in an Eppendorf 5810R centrifuge, and the supernatant of each well was assayed for HNL activity in UV-microplates (Greiner Bio-one GmbH, Kremsmünster, Austria).

2.6Determination of residual glucose and ammonium in deep-well plate cultures

Glucose concentration measurements were performed with the “Glucose UV” kit (Dipromed, Weigelsdorf, Austria) using the hexokinase method. The assay was adapted to microplates: 190 μl of the provided reagent were combined with 10 μl sample, mixed and incubated at room temperature for 15 min. The absorbance at 340 nm was measured with a platereader (Spectramax 384plus, Molecular Devices, Sunnyvale, CA, USA) and offset against a calibration curve that was generated with known glucose concentrations. Samples represented the supernatant of briefly centrifuged culture aliquots withdrawn at given time-points from representative wells throughout a deep-well plate. If necessary, these samples were diluted in order to meet the criteria of linearity within the calibration curve. The determinations of ammonium in the culture media in deep-well plates were performed with the Spectroquant Ammonium-test (Merck, Darmstadt, Germany).

2.7Determination of necrotic cells

A correlation between the ratio of OD600-value to cell number was set up as follows: following determination of OD600 of cultures grown for 60 h, the cell number was counted under the microscope (Laborlox S, Leitz) in a Thoma chamber (0.1 mm depth, 0.0025 mm2). This revealed a cell number of 1.5 × 108 per OD600= 1. For this calculation factor, no significant dependence on glucose concentration was observed up to a concentration of 3%.

0.5 mg Propidium iodide (P-4170, Sigma, St. Louis, MO, USA) were dissolved in 10 ml sterile PBS (0.1 M phosphate buffer, pH 7.2, supplemented with 0.12 M NaCl), stored under protection from light at 4 °C and used within 30 min. The cells were cultured at various glucose concentrations according to the mentioned protocol, and after 60 h a theoretical 108 cells were gently pelleted through centrifugation at room temperature with 1500 rpm for 5 min in a tabletop centrifuge. After washing 3 times in 0.5 ml PBS, the pellets were softly resuspended in 200 μl of propidium iodide-solution, transferred to fluorescence microtiter plates (FIA plates black TC, F-form, Greiner Bio-one GmbH, Kremsmünster, Austria), incubated for 10 min at room temperature in the dark and finally measured for fluorescence intensity in a Spectramax Gemini XS fluorescence platereader (Molecular Devices, Sunnyvale, CA, USA). The excitation wavelength was 536 nm and the emission wavelength 617 nm with a cutoff filter set to 610 nm. Prior to all measurements, a calibration curve was created by mixing living cells (cultivated in deep-well plates in BMD 1% for 20 h) with dead cells (=living cells, incubated at 75 °C for 10 min) in ratios to yield always a calculated total of 108 cells, with a fraction of 0%, 10%, 25%, 30%, 50%, 75% and 100% dead cells.

For the determination of fully-viable cells, an aliquot of the culture containing a calculated number of 108 cells was withdrawn after cultivation in media with different glucose concentrations for 60 h. The cells were diluted to a theoretical 103 and 102 cells ml−1, and 50 μl as well as 100 μl of these dilutions were plated on MD and YPD agarplates, respectively. After 2 d at 30 °C, the colonies were counted and the values were compared to the calculated number of cells.

2.8Investigation of apoptotic phenomena

For assaying apoptotic phenomena, the TdT-mediated dUTP nick end labelling (TUNEL)-test was employed [16]. A calculated number of 8 × 107 cells was taken from representative wells of cultures incubated in media with different glucose concentrations (0.2%, 1%, 2% and 3%) for 60 h. Cells were treated as described in [21,22], except for the application of zymolyase 20T (from Arthrobacter luteus, Seikagaku Corp., Tokyo, Japan) for spheroplasting.

To obtain positive controls, 4 h prior to further processing (i.e., after 56 h of incubation) different concentrations of H2O2 (10, 100, 300, 600 and 1000 mM) were added to some culture wells grown as mentioned above. Suitable concentrations for the individual cultivation conditions were applied as positive control for further tests. As negative controls, a calculated number of 8 × 107 cells were grown in deep-well plates for 20 h in BMD 1% glucose. The calculation of apoptotic cells in relative numbers was determined in three independent windows of the corresponding slides.

2.9Additional reporter systems and the corresponding activity assays

The HNL standard assay employing benzaldehyde cyanohydrin (mandelonitrile), as a substrate was adapted to a 96-well microtiter plate format: 130 μl of 0.1 M citrate-phosphate buffer, pH 5.0, were combined with 20 μl of supernatant. After addition of 50 μl racemic mandelonitrile solution (80 mg racemic mandelonitrile, dissolved in 0.003 M citrate-phosphate buffer, pH 3.5, 15 mM final concentration), the absorbance at 280 nm was tracked over 5 min in the platereader (Spectramax Plus384, Molecular Devices, Sunnyvale, CA, USA). For the comparison of microtiter plate results with the standard assay in 1-cm cuvettes, an ε280 for benzaldehyde of 1.376 l mM−1 cm−1 and a correlation factor was used to calculate the actual specific activity in U (=μmol product min−1) ml−1.

For determination of horseradish peroxidase (HRP)-activity, a protocol from Morawski et al. [23] was adopted. The absorption was measured at 405 nm in a Spectramax 384plus platereader (Molecular Devices, Sunnyvale, CA, USA) after incubation of the culture supernatants with the substrate ABTS (2,2′-azino-bis-(3-ethylbenzthiazoline-6-sulfonic acid)) and H2O2 for 30 s in the dark.

For the assay of intracellular HNL, P. pastoris GS115 PpD1-17 cultures were centrifuged, the supernatant was discarded and for cell lysis under non-denaturing conditions 15 μl of “Y-PER reagent” (Pierce, Rockford, IL, USA) were added to the cell pellets in the wells. The plate was incubated with gentle shaking for 30 min at room temperature. After a centrifugation step for 40 min at 4000 rpm, the supernatant of each well was assayed for HNL-activity in a UV-microplate as described above for the secreted (R)-HNL.

2.10Scale up

P. pastoris fermentations were carried out similar to the protocols described in [24], with the notable exception that glucose was used instead of glycerol: at the end of the glucose feed, a methanol feed was started aiming at keeping the methanol concentration (off-line methanol analysis) around 1% by adjusting the feed rate. The pH-value was set at 5.5 and controlled by sulfuric acid and ammonia, respectively. Dissolved oxygen was kept above 30%, primarily controlled by stirrer speed, and backed up by aeration rate. The temperature was set to 30 °C. Cell dry weights were determined as described in [25].

3Results and discussion

3.1Microscale cultivation system for P. pastoris

Possible difficulties one inevitably encounters when using microscale expression and high-throughput screening facilities include: (a) differential growth and variations in gene expression by identical clones in separate wells of microtiter plates; (b) differential growth and/or gene expression of different clones from gene variant libraries; (c) differential stress exposure across the plate [3]. The goal is the development of a system in which every individual culture of a 96-well plate shows uniform growth and expression.

In previous studies [5,26], deep-well plates with square-shaped wells and a planar bottom were identified as optimal system to provide optimal oxygen supply [27] and to keep cells in suspension. Additionally, the covers of these plates (Bel-Art Scienceware, Pequannock, NJ, USA) contained inlets to seal each well against cross-contamination. In deep-well plates with round wells and a conical well-bottom the cells pelleted within minutes, even when shaken at the rotation limit of 400 per minute.

As first steps towards microscale cultivation, test runs under varying conditions were performed using P. pastoris strains GS115 and X-33 as reporters for growth behaviour. Standard cultivation parameters were: 340 rpm at 28 °C [24,28] with 80% humidity and a maximum filling capacity of 600 μl per well, because volumes above this limit tended to spill onto the cover.

Volume loss under these conditions was negligible within the cultivation time of three days.

3.2Optimization of the micro-scale cultivation and screening

The strong inducible AOX1-promoter was used for all expression experiments with reporter enzymes. This promoter is tightly repressed by most carbon sources like glucose or glycerol and highly induced by methanol when the C-sources are depleted.

For high-throughput screening the usually employed media change, from repressing growth medium to induction medium containing methanol as a sole carbon source had to be strictly avoided. Such manipulations cause a high risk of cross contamination and varying loss of cell material from individual cultures in deep-well plates, which is a major source of high standard deviations of enzyme activities. Moreover, this procedure is very laborious. We kept the P. pastoris-cells growing in 250 μl glucose-containing medium until glucose was depleted and then induced expression by adding methanol. In order to avoid stress to the cells by addition of pure methanol and to provide further nutrients for the expression phase, the methanol was applied not in a pure form but as a dilution in medium lacking glucose. For the first steps in the development of a microscale cultivation system for comparative screening of enzyme activities, we used P. pastoris GS115 PamHNL5-a37 [18] expressing a secreted hydroxynitrile lyase (HNL) as a reporter. Due to the experience with former shake flask experiments [18], the culture time for the initial growth phase under repressing conditions was set to 60 h to allow uniform glucose depletion in all individual wells of 96-well plates. With these basic settings, the most suitable parameters for both, uniform cell growth and expression levels, as well as the highest possible activity levels, were worked out for the reporter enzyme.

We expected that the specific activities of the reporter enzymes per ml of culture volume would increase with rising cell densities. For the evaluation of the ideal glucose concentration at microscale, 0.2%, 1%, 2% and 3% glucose were applied. Growth was performed in 250 μl BMD for 60 h and then 250 μl BMM2 were added for induction. Further 50 μl BMM10 were added after additional 10, 22 and 46 h, respectively. Twenty-four hours after the last induction (i.e., 70 h after the first one) the cells were pelleted and HNL activity in the supernatant was determined photometrically. Although the biomass was increased with higher glucose concentrations as expected (Fig. 1(a)), higher glucose concentrations in the medium than 1% resulted in reduced yield of active recombinant protein (Fig. 1(b)). Doubled concentrations of methanol in the induction media gave even higher specific activities, but the standard deviation in growth and expression levels was also significantly higher (data not shown). 1% Glucose in the initial culture media proved to be the most suitable carbon source concentration by resulting in the highest activity levels of the expressed reporter enzyme as well as the most homogenous expression level.

Figure 1.

Optical density at 600 nm after cultivation of P. pastoris GS115 PamHNL5-a37 for 60 h in media supplemented with glucose from 0.2% to 3% w/v (a). HNL activity with the substrate mandelonitrile in the culture supernatant after growth of P. pastoris GS115 PamHNL5-a37 in deep-well plates with varying glucose concentrations for 60 h and induction with methanol for 70 h (b).

3.3The physiological background of low specific activity at higher cell density

Residual glucose, repressing gene induction, as the most obvious explanation for this observation could be excluded. By three independently performed experiments, the time of glucose depletion was determined as accurately as possible for each of the applied concentrations. Cultures with 0.2% glucose consumed its C-source within 14 h. 1% Glucose was depleted after 23 h, 2% after 29 h, and for 3% glucose the cells needed 44 h. Therefore, promoter repression by residual glucose at the time of induction after 60 h of cultivation was not the reason for the lower activity levels of cultures grown at higher glucose concentrations. For all three glucose concentrations, a growth phase of 60 h prior to induction proved to be optimal, since there was sufficient time even for weakly inoculated wells to reach their maximal cell density by total consumption of the available glucose.

Granot et al. [29] have demonstrated that S. cerevisiae and Schizosaccharomyces pombe, when grown in water in the presence of glucose and in the absence of additional nutrients, undergo sugar-induced cell death (SICD) via apoptosis. We found sufficient amounts of residual nitrogen sources in all individual culture media after 60 h of cultivation, and moreover diluted methanol was added for induction of expression together with fresh nutrients lacking only the repressing C-source glucose. Thereby we also ruled out a possible repression of the AOX1-promoter upon nitrogen limitation [30].

Although SICD was explained by the absence of other nutrients than glucose [29], we thought that in our case where all nutrients were present, cell viability and thereby protein expression were possibly influenced by varying glucose concentrations.

The determination of the fraction of dead cells in the particular cultures after the initial growth phase (i.e., right before induction) was intended to enlighten the observed mystery. The following scenario would explain the events: if a considerable fraction of cells died off when grown on higher glucose concentrations, the overall activity of the expressed reporter enzyme would naturally be lower. The fraction of necrotic cells and the number of living cells (colony-forming units) were determined by propidium iodide staining and plating on agar plates, respectively. Therefore a microtiter plate method for high-throughput determination of necrotic P. pastoris-cells was established, based on a flow cytometric method described previously in the group of Mattanovich [25]. As a control for non-expressing strains, the wild-type strains X-33 and GS115 were treated in the same way as the reporter strain.

Experiments were performed in triplicate and by different experimenters. The results from the propidium iodide staining were somewhat puzzling. Although there was a clear increase in fluorescence with increasing glucose concentrations (except from 0.2% glucose, where the value exceeded the one of the 1% glucose cultures), the comparison with a calibration curve displayed the fraction of dead cells to be lower than 10% in all cases. All the more astonishing were the results of the cell plating experiment. For glucose concentrations other than 1%, lower expression rates were attended with decreased fractions of recultivable cells. However, the absolute number showed a large discrepancy between the sum of viable cells and necrotic cells, and the total number of analyzed cells (Fig. 2). The non-expressing control strains X-33 and GS115 showed the same results (data not shown). Colony counts did not differ significantly after plating on MD or agar plates, respectively.

Figure 2.

Cell viabilities of cultures grown for 60 h in media with varying glucose concentrations. The bottom bars show the percentage of recultivable cells after plating calculated numbers of cells. The fractions of necrotic cells (stained by propidium iodide) are presented by hanging bars. All numbers were calculated as an average from three independent experiments.

Experiments with cells grown in complex YPD media with different glucose concentrations revealed that for 2% and 3% glucose the discrepancy between cell density and reporter activity was still present, although less pronounced than with minimal media. However, for screening of transformants using auxotrophy selection markers such complex media cannot be used. Furthermore, the use of minimal media for screening purposes mimics the conditions of the enzyme production on a large scale, thereby facilitating the transferability of results between different scales.

The trend of highest reporter protein activity in media with limited cell density (1% glucose) was observed for shake-flask cultures as well (data not shown). This indicates that higher oxygen limitation in deep-well plates than in shake flasks cannot be solely responsible for the observations. There must be a considerable fraction of cells which is still not necrotic at higher glucose concentrations but on the other hand also physiologically not competent for recombinant protein expression anymore. We thought about apoptosis as a possible explanation for these observations.

3.4Suboptimal glucose concentrations in deep-well plate cultures cause apoptosis in P. pastoris

Apoptosis as a form of programmed cell death in yeast and its regulators and effectors have been intensely studied [16]. For the clarification of a possible participation of apoptotic phenomena in reduced and varying protein expression in a high-throughput expression system, the TUNEL-test was regarded as suitable method due to its high specificity and significance [21].

No reports about controlled, successful induction of apoptosis in P. pastoris were available. It was inevitable for the development of reliable protocols for our experiments to find out about conditions that induce apoptosis. One study already had described a possible occurrence of cell death in P. pastoris related to apoptosis [31]. The authors expressed the murine pro-apoptotic gene bax and showed that its product caused cell death, even though it was not clearly stated that apoptotic events took place. However, the TUNEL test results for Bax-expressing cells as well as negative control cells displayed unstained or only slightly stained cells, indicating that DNA cleavage was weak or even non-existing.

Concentrations of 0.6–1 mM H2O2 proved to induce apoptosis in S. cerevisiae during exponential growth sufficiently, whereas yeast cells in the stationary phase tolerated up to 500 mM H2O2. In both cases, the hydrogen peroxide was applied 3 h (2 generation times) prior to investigation for apoptosis markers (K.U. Froehlich, personal communication). Therefore, our P. pastoris reporter strains were grown for 56 h as described in media with different glucose concentrations. H2O2 concentrations of 10, 100, 300, 600 and 1000 mM were added to particular wells of all glucose concentrations and after a total of 60 h the cells were collected and treated as described in Section 2. At this time, there was yet no methanol added to the cultures, therefore intracellular H2O2 production due to metabolization of external methanol by alcohol oxidase can be excluded. For 0.2% and 1% glucose cultures 100 mM H2O2 and for 2% and 3% glucose 200 mM H202 showed the best induction effect, making them useful positive controls for apoptosis.

In the following, cultured cells including positive and negative control cells were examined for apoptosis-induced DNA damage employing the TUNEL-test. As Fig. 3 shows as a case in point, we found a direct connection between increasing glucose concentrations in deep-well culture media and increased numbers of apoptotic cells. The apoptosis tests were repeated twice to obtain a representative number for the fraction of apoptotic cells in the different culture samples. For cultures grown for 60 h in 0.2% glucose or 1% glucose, only 1% or less of the cells showed significant signs of apoptosis. The fraction of apoptotic cells ascended dramatically to ∼20% in the case of 2% and 25% at 3% initial glucose. Results from samples derived from shake-flask cultivation at different glucose concentrations resembled those from microscale experiments. These findings were not a result of heterologous protein production itself, which might cause stress for the host cells, because the untransformed P. pastoris strains X-33 and GS115 without foreign DNA showed the same behaviour.

Figure 3.

Results of the TUNEL-test which was applied to P. pastoris GS115 PamHNL5-a37 after growth for 60 h in media supplemented with 0.2% glucose (a), 1% glucose (b), 2% glucose (c) and 3% glucose (d), respectively. P. pastoris GS115 PamHNL5-a37 grown in media supplemented with 1% glucose for 20 h was used as a negative control (e). P. pastoris GS115 PamHNL5-a37 grown for 60 h in media with 1% glucose was induced for apoptosis by the addition of 100 mM H2O2 4 h prior to the test procedure and used as a positive control (f). The fraction of apoptotic cells was determined in three independent, randomly-chosen windows of the corresponding slides.

Recalling the puzzling gap between the number of dead cells from the propidium iodide staining and the number of living cells from plating experiments, its explanation became obvious: under unfavourable glucose concentrations a large fraction of the cells underwent apoptosis. It is reasonable to assume that this fraction of cells is physiologically not competent for recombinant protein expression. Therefore the yield of active reporter enzyme at higher glucose concentrations decreases and the standard deviation of expression levels from well to well increases. This phenomenon still remained unexplained. The media ingredients only differed in their proportion of glucose, and the step from 1% to 2% glucose has generally been regarded as minimal. A concentration of 2% glucose was even regarded as an optimal standard concentration for protein expression by P. pastoris in shake flasks (“Pichia Expression Kit”, Invitrogen) [1,15]. Limited oxygen supply for P. pastoris cells at higher densities in deep-well plates or shake flasks could be one cause for our observations.

With 1% glucose in the growth medium we found conditions causing a minimum of apoptotic and necrotic cells, which allows uniform expression of correctly-folded active enzyme at a high level. Employing apoptosis and necrosis as a marker for physiological competence of P. pastoris cells can be used to optimise culture conditions when online control is hardly possible.

3.5Verification of the screening protocol with further reporters

The validation of the usefulness of this screening system implies the need for a verification of the results using additional reporter systems. A second representative for secreted proteins, horseradish peroxidase (HRP), was examined for uniform expression levels across a whole 96-deep-well plate employing the developed protocol. A sensitive activity assay in microplate format for this enzyme expressed by S. cerevisiae was already available [23]. The plasmid harbouring the HRP-WT (wild-type) gene under the control of the AOX1-promoter was transformed into P. pastoris X-33, the transformants were tested for activity by oxidation of ABTS by the expressed HRP and H2O2. In addition, a single positive clone as well as P. pastoris X-33 as negative control were cultivated and tested according to our micro-scale fermentation protocol. The activity levels displayed a uniform expression level with a standard deviation lower than 15% (data not shown), confirming the reliability of the protocol.

The HbHNL of the tropical rubber tree Hevea brasiliensis, catalysing the same reaction as PaHNL but with opposite enantioselectivity, was chosen as a representative for intracellular proteins [19]. The active variant P. pastoris GS115 PpD1-17 and the untransformed strain GS115 (in presence of additional histidine in the medium) were cultivated following the established protocol. The cells were chemically lysed directly in the deep-well plate and after harvesting the fraction of soluble intracellular proteins by centrifugation, the same microtiter plate assay as for PaHNL was applied. Due to the massive overexpression of this protein, several dilutions were necessary to meet the criteria of the photometric test in terms of linearity. The activity levels of GS115 PpD1-17 showed uniformity with a low standard deviation across the whole plate, proving the applicability of the high-throughput screening for intracellular proteins as well. We were also pleased to see that the absolute numbers of activity were comparable to those obtained by shake-flask cultivation, indicating that the same amount of active protein was produced in the micro-scale process as in a 500-fold higher volume in shake flasks.

3.6Scale up: from 500 μl to fermentation on large scale

The ultimate goal of a micro-scale high-throughput screening for engineered proteins lies in a perfect translation of results to controlled fermenter scale. New enzyme variants with improved properties or enhanced production levels in micro-scale experiments often do not keep their promises after scale-up to fermentations under optimal conditions.

The correlation of enzyme activities in micro-scale, small scale and even large scale was clarified by cultivating P. pastoris GS115 PamHNL5-a37 and the improved expression variant P. pastoris GS115 PaHNL5_L1Q [20] in 2 ml deep-well plates (500 μl working volume), 2-l shake flasks (250 ml working volume) as well as in a fermenter with 10 l working volume. The most striking difference in operating procedures concerned the instrumental control of all-important parameters as pH-value, aeration or methanol concentration, which were continuously adjustable in the fermenter cultivations, but much less precise at smaller scale. Of course this causes differences in the absolute activity values reached in the fermenter as compared to those in deep-well plates and shake flasks. However, the ratio of the relative specific activities between the native PamHNL5-a37 and the improved secretion variant PaHNL5_L1Q showed the same trend in improvement of enzyme production from micro-scale to fermenter scale (Table 1). Summarizing, the trends we detected in deep-well plates held true in shake flasks as well as in 10 l fermentations. The results from 10 l fermentations can be scaled up to larger volumes without changes in the results as demonstrated for the expression of PaHNL by P. pastoris GS115 PamHNL5-a37 in a 4000 l fermenter (data not shown).

Table 1.  Comparison of activity levels of the strains P. pastoris GS115 PamHNL5-a37 and PaHNL5_L1Q after growth at different scales and normalization by their biomass
 HNL-activity/biomass (U g−1 cdw min−1)aImprovement factor
  1. The activity measurement and the biomass determination were performed 70 h after the first induction with methanol. Improvement factors represent the division of the values of the improved expression construct PaHNL5_L1Q by those of PamHNL5-a37.

  2. a1 U = 1 μmol benzaldehyde produced after 1 min, cdw = cell dry weight.

Microscale (500 μl)427.5104.44.1
Shake flask (250 ml)410.699.84.1
Fermenter (10 l)935.5208.54.5

The example for enhanced production rate given above can be translated to improved enzyme properties (e.g., the conversion of non-natural substrates [18]). However, we would like to note that the described procedures allow correct predictions for relative improvements of enzyme variants or expression strain variants. It was not our intention to predict absolute protein yields or optimal production conditions for large-scale fermentations, since this is strongly influenced by fermentation techniques. The findings strongly support the representativity of the new protocol for high-throughput analysis of enzyme improvements using P. pastoris on the scale of 96-deep-well plates.


The publication of more and more eukaryotic genome sequences, the availability of efficient production systems for eukaryotic proteins, as well as the improvement and extension of library production for directed-evolution experiments, called for high-throughput methods for eukaryotic hosts. We developed an efficient protocol for the parallel cultivation, expression and screening of thousands of protein variants in P. pastoris with low standard deviation in expression levels. The careful analysis of culture media composition showed that minor changes resulted in severe influences on cell physiology. Under unfavourable conditions, significant fractions of cells underwent apoptosis, which caused a bias in screening experiments for expression variants under screening conditions. The outcomes of such a screening were isolated variants that usually do not keep their promises after scale-up of enzyme production. Minimizing programmed cell death and necrosis in micro-scale culture procedures resulted in the first reliable high-throughput protocol for comparative screening of enzyme variants produced by P. pastoris. The output of screening new protein variants directly in the production host P. pastoris are new strains which can be directly scaled up for enzyme production on a large scale. Thus, the risk is minimized that important traits such as optimal codon usage for high-level expression get adapted to the screening host rather than to the production system. This provides relief to existing bottlenecks of this attractive host for high-level heterologous protein production as well as for protein engineering.

The comparison of cell viabilities with the fraction of apoptotic and necrotic cells can be used as a tool for optimisation of high-throughput growth and screening conditions.

The mechanisms for the observed apoptosis, induced by raising glucose levels in media for P. pastoris cultivation, remain to be elucidated.


We thank TIG, the Province of Styria, SFG and the City of Graz for financial support. The authors are indebted to Laura Leitner and Kai-Uwe Fröhlich for the apoptosis work. We thank Hannelore Mandl, Beate Pscheid and Franz Hartner for excellent technical support.