Bioreactor-scale cell performance and protein production can be substantially increased by using a secretion signal that drives co-translational translocation in Pichia pastoris.

Abstract

Pichia pastoris (Komagataella spp.) has become one of the most important host organisms for production of heterologous proteins of biotechnological interest, many of them extracellular. The protein secretion pathway has been recognized as a limiting process in which many roadblocks have been pinpointed. Recently, we have identified a bottleneck at the ER translocation level. In earlier exploratory studies, this limitation could be largely overcome by using an improved chimeric secretion signal to drive proteins through the co-translational translocation pathway. Here, we have further tested at bioreactor scale the improved secretion signal consisting of the pre-Ost1 signal sequence, which drives proteins through co-translational translocation, followed by the pro region from the secretion signal of the Saccharomyces cerevisiae α-factor mating pheromone. For comparison, the commonly used full-length α-factor secretion signal, which drives proteins through post-translational translocation, was tested. These two secretion signals were fused to three different model proteins: the tetrameric red fluorescent protein E2-Crimson, which can be used to visualize roadblocks in the secretory pathway; the lipase 2 from Bacillus thermocatenulatus (BTL2); and the Rhizopus oryzae lipase (ROL). All strains were tested in batch cultivation to study the different growth parameters obtained. The strains carrying the improved secretion signal showed increased final production of the proteins of interest. Interestingly, they were able to grow at significantly higher maximum specific growth rates than their counterparts carrying the conventional secretion signal. These results were corroborated in a 5L fed-batch cultivation, where the final product concentration and volumetric productivity were also shown to be improved.

Introduction

The methylotrophic yeast Pichia pastoris (Komagataella spp.) has become one of the most popular yeasts for the production of industrially relevant proteins [1–5], most of which are secreted into the extracellular media to facilitate subsequent downstream processing. Its popularity has arisen due to several benefits including the capacity to grow at high cell densities at a bioreactor scale [6,7], the low cost of the culture media, and the availability of several inducible promoters that enable the decoupling of growth and protein production [8–10]. The best known inducible promoter in P. pastoris is the strong and tightly regulated alcohol oxidase 1 promoter (P_AOX1), which is strongly repressed in the presence of glucose and glycerol, and becomes strongly induced with methanol [10,11]. In addition, P. pastoris typically shows low levels of endogenous protein secretion, a property that facilitates downstream processing of the protein of interest. Notably, the high similarity between P. pastoris and Saccharomyces cerevisiae permits the sharing of protocols and genetic elements including secretion signals.

In order to achieve secretion in yeast, it is necessary to add a secretion signal at the N-terminus of the protein of interest. For P. pastoris, the most often used secretion signal is derived from the α-factor mating pheromone (α-MF) of S. cerevisiae [12]. This secretion signal contains two parts: a 19-amino-acid signal sequence (the pre region) that directs translocation from the cytoplasm to the lumen of the endoplasmic reticulum (ER), followed by a 66-amino-acid pro region that is recognized by the transmembrane Erv29 receptor to mediate rapid ER export through packaging into COPII vesicles that are sent to the Golgi [13,14]. In addition, at the end of the pro region, there is a signal that allows the secretion signal to be removed from the heterologous protein by the Golgi-localized peptidases Kex2 and Ste13 [15]. Although the α-MF secretion signal has been shown to drive efficient secretion of multiple heterologous proteins, variable levels of secretion have been reported among different proteins [16]. Therefore, several attempts to improve secretion efficiency have been made [17], many of them focused on modifying the α-MF secretion signal [18–20].

Recently, we identified two major limitations of the α-MF secretion signal [21]. The first is that the signal sequence (pre-α-MF) drives proteins through post-translational translocation. Hence, if a protein intended for secretion folds prematurely, it could fail to enter the ER and become retained in the cytoplasm or even block the translocons [22]. To avoid this problem, pre-α-MF was replaced by the S. cerevisiae Ost1 signal sequence (pre-Ost1), which has been reported to drive efficient co-translational translocation [21,23,24]. The second limitation is that the pro-region of the α-MF secretion signal is prone to aggregation in the ER. For this, the solution was to replace leucine at position 42 by serine; this mutation reduced protein aggregation in the ER and increased secretion efficiency. The resulting chimeric secretion signal consisted of the pre-Ost1 signal sequence followed by the Ser42 variant of the α-MF pro region [21].

Here, we have studied the effect of the different translocation pathways by using either the pre-Ost1 or the pre-α-MF signal sequence, in each case fused to the Ser42 variant of the α-MF pro region. This was carried out using three different model proteins produced under control of the methanol-inducible P_AOX1: E2-Crimson, a far-red fluorescent protein that allows the visualization of intracellular protein trafficking, and two microbial lipases, Bacillus thermocatenulatus lipase 2 (BTL2) and Rhizopus oryzae lipase (ROL). The two lipases are structurally different and can give a broad indication about the effect of using pre-Ost1 on cellular physiology and protein production levels. BTL2 is a lipase of industrial interest that has been reported to fold rapidly, and its secretion can be dramatically affected by the choice of secretion signal [25]. ROL is a lipase and a promising industrial biocatalyst [26] that we have used as a reporter protein in previous studies. This enzyme is structurally different from the other model proteins and has been shown to trigger a strong stress response and to cause a significant metabolic burden even at relatively low production levels in P. pastoris [27,28], making it an appealing choice to test whether the pre-Ost1 signal sequence can reduce cellular stress/metabolic burden in yeast.

Furthermore, this comparison was studied in both batch and fed-batch cultivations, i.e. under bioprocess-relevant conditions, aiming to study how a secretion signal that replaces pre-α-MF with pre-Ost1 affects fermenter cultures in terms of product yields and cellular physiology. Although the levels of secretion varied between proteins, all strains carrying the pre-Ost1 signal sequence showed improved final protein production compared with those carrying the pre-α-MF signal sequence, and this effect was preserved across cultivation scales from shake flasks to 5L bioreactors.

Materials and methods

Strains and plasmids

P. pastoris strains were derived from the strain X33 (Thermofisher Scientific, Waltham, MA, USA). Plasmids were created by In-Fusion cloning (Clontech, Mountain View, CA, USA), and primers required for the plasmid construction were acquired from Integrated DNA Technologies (IDT), CA, USA. Expression of the genes encoding E2-Crimson, BTL2, and ROL was driven by the methanol-inducible promoter P_AOX1, while expression of msGFP-HDEL, a gene encoding the monomeric superfolder Green Fluorescent Protein variant with the ER retention signal HDEL, was driven by the constitutive KAR2 promoter [21,23]. The model proteins were fused to either the full-length α-MF secretion signal consisting of the pre-α-MF signal sequence followed by the α-MF pro region with the Ser42 mutation [21], or a chimeric secretion signal consisting of the pre-Ost1 signal sequence followed by the same α-MF pro region. msGFP-HDEL was fused to the pre-Kar2 signal sequence. Genetic engineering procedures were designed and recorded by the use of SnapGene software (Insightful Science, San Diego, CA, USA). All the plasmids used were described previously [21] and are available at Addgene, except for the ROL expression vector [29]. The BTL2 gene was codon optimized for P. pastoris and synthetized by GenScript and the plasmid sequence is available in a previous publication [21]. The GenBank accession numbers for the native forms of BTL2, ROL, and E2-Crimson are CAA64621, GQ502721, and AMO27221, respectively.

Plasmids were linearized with the PmeI restriction enzyme (New England Biolabs, MA, USA), and then transformed by electroporation using 100 ng of linear DNA, an amount that favored integration of a single copy of the expression cassette into the genome. Thereafter, screening was performed to detect positive colonies. Strains were grown and selected in a Yeast Extract - Peptone - Dextrose (YPD) rich medium, supplemented with either the antibiotic Zeocin (100 μg/mL) (InvivoGen, San Diego, CA, USA) or G418 (500 μg/mL) (InvivoGen, San Diego, CA, USA). For each strain, 8 independent clones were tested, 3 were selected for further shake flask-scale characterization, and one was finally selected for further cultivation experiments. Single-copy integration was confirmed by droplet digital PCR [30], after purification of genomic DNA, using primers that flanked the integration locus. E2-Crimson, BTL2, and ROL single-copy integration at the AOX1 locus was confirmed using primers 5′-GAAATAGACGCAGATCGGGAAC-3′ and 5′-GAAGGTAGACCCATGGGTTGTTG-3′, supplied by IDT, CA, USA. The pre-Kar2-msGFP-HDEL construct was integrated at the HIS4 locus by linearizing the plasmid with Sail (New England Biolabs, MA, USA), and single-copy integration was verified using primers 5′-GCTCTAGCCAGTTTGCTGTCCAAAC-3′ and 5′-GGATGTTAGATGCCGGTTAGATC-3′.

YPD cultures for each strain were used to prepare 15% glycerol stock cell suspensions with final optical density (OD) of 60 at 600 nm wavelength (λ) and stored at −80°C.

Cultivation media

YPG pre-culturing medium contained per L: 10 g yeast extract (Merck, Kenilworth, NJ, USA), 20 g peptone (Merck, Kenilworth, NJ, USA), and 20 g glycerol (PanReac AppliChem, Darmstadt, Germany) and was supplemented with 100 μg/mL of Zeocin. The basal salt synthetic medium for methanol batch and fed-batch cultivations contained per L distilled H₂O: H₃PO₄ (85%), 26.7 ml.; CaSO₄, 0.93 g; K₂SO₄, 18.2 g; MgSO₄·7H₂O, 14.9 g; KOH, 4.13 g; glycerol, 40 g; biotin solution (200 mg/L), 4 mL; trace salts solution, 10 mL; and antifoam agent (A6426, Sigma-Aldrich Co., St. Louis, MO, USA), 0.5 mL. The trace salts solution contained per liter: CuSO₄·5H₂O, 6.0 g; NaI, 0.08 g; MnSO₄·H₂O, 3.0 g; Na₂MoO₄·2H₂O, 0.2 g; H₃BO₃, 0.02 g; CoCl₂, 0.5 g; ZnCl₂, 20.0 g; FeSO₄·7H₂O, 65.0 g; biotin, 0.3 g; and concentrated H₂SO₄, 5 mL. Chemicals were purchased from Merck, NJ, USA. The biotin and trace salts solutions were sterilized separately by filtration with a 0.22 μm pore filter (4433, PALL Life Sciences, New York, USA) or a 0.2 μm pore filter (17764-ACK0, Sartorius Stedim, Göttingen, Germany), respectively.

Pre-cultures

Duplicate Erlenmeyer flask cultures (150 mL working volume) were prepared as follows: 150 mL of YPG medium containing 100 μg/mL of Zeocin were inoculated with thawed cryo-stocks at an OD₆₀₀ of 0.2 and incubated overnight at 30°C and 160 rpm (Infors Multitron Shaker, 25 mm shaking diameter, Infors HT, Basel, Switzerland). The pre-culture was then centrifuged at 6,371 rcf for 10 min, and the harvested cells were resuspended in sterile H₂O in a final volume of 100 mL and directly inoculated into the reactor. Methanol batches were inoculated at an initial OD₆₀₀ of 2, while fed-batches were inoculated at initial OD₆₀₀ of 1.

Bioreactor cultivation set-up and operational conditions

Methanol batch cultures

Batch cell cultures were carried out in a 2 L bioreactor controlled by the ez-Control system (Applikon Biotechnology B.V., Delft, The Netherlands) under the following cultivation conditions: working volume 1 L, 30°C, pH 5.5 controlled by adding 30% (v/v) NH₄OH, 1 vvm air flow and dissolved O₂ set-point at 30% air saturation controlled in cascade with agitation (from a minimum of 500 to a maximum of 700 rpm). Before starting the cultivation, 10 g/L of methanol were added, and samples were taken at different time points until the methanol was fully consumed. In Table 1, q_{p MAX} was calculated using the Y_P/X × μ_MAX. All methanol batch cultivations were performed in duplicate.

Table 1. Methanol batch bioprocess parameters.

Summary of maximum specific growth rates (μ_MAX), specific methanol (MetOH) consumption rates (q_s), specific protein production rates (q_{p MAX}), protein/biomass yields (Y_(p/x)), and biomass/methanol yields (Y_(x/s)) (±SD) during batch cultivations performed with the reference and the three model proteins producing strains. Biomass concentrations used in calculations were measured as dry cell weight (DCW). These parameters were calculated from cultivations performed in duplicate. E2-Crimson levels were measured in arbitrary units (A.U.) of fluorescence, and BTL2 and ROL levels were measured in units of lipase activity (U.A.).

	X33	E2-Crimson		BTL2		ROL
	control	pre-α-MF	pre-Ost1	pre-α-MF	pre-Ost1	pre-α-MF	pre-Ost1
Final Protein titer *	-	143±10	551±13	4.01±0.32	8.53±0.47	9.37±0.10	12.60±0.33
μMAX (h−1)	0.097±0.002	0.071±0.001	0.086±0.002	0.062±0.002	0.088±0.001	0.045±0.001	0.084±0.001
qs (gMetOH)/gDCW·h)	0.296±0.011	0.249±0.015	0.258±0.002	0.205±0.001	0.259±0.007	0.16±0.010	0.242±0.012
qp MAX (protein/gDCW·h)	-	2670±130	12130±770	68.1±6.2	187.4±19.1	120.5±2.5	270.5±16.3
Y(p/x) (protein/gDCW)	-	37600±2500	141200±11200	1098±112	2130±168	2677±40	3150±130
Y(x/s) (gDCW/gMetOH)	0.33±0.01	0.28±0.02	0.33±0.01	0.30±0.01	0.34±0.01	0.28±0.02	0.34±0.02

^*Protein titer units: E2-Crimson (A.U./mL); BTL2: (U.A./mL); ROL: (U.A./mL).

Fed-batch cultures

Fed-batch cultures were carried out in a 5 L bioreactor controlled by a Braun Biostat B system (Braun Biotech, Melsungen, Germany) under the following cultivation conditions: initial volume 2 L, 30°C, pH 5.5 controlled by adding 30% (v/v) NH₄OH during the glycerol batch phase and 5 M KOH during the transition phase and methanol induction phase, 2 L/min air flow and dissolved O₂ set at 30% air saturation controlled in cascade with agitation (from 900 to 1200 rpm during the glycerol batch phase and 1200 rpm during the transition phase and methanol induction phase), and air O₂ enrichment when required to maintain the culture at 30% O₂.

The glycerol batch phase was first started with 40 g/L of glycerol; when the glycerol was completely consumed, as detected by a sudden increase in the dissolved O₂ concentration (DO), a 5 h transition phase began. Finally, the methanol induction phase started, in which methanol acted simultaneously as the sole carbon source and inducer. The strategy was a non-limited fed-batch culture (MNLFB). A predictive-PI control strategy was implemented to maintain a constant methanol concentration of 3 g/L [31].

In the transition phase, successive solutions of 250 mL of 50% (v/v) glycerol, and 1 L of pure methanol complemented with 10 mL of trace salts solution and 4 mL of biotin solution, were added as carbon sources. Transition phase started with 300 μL/min glycerol addition for 2 h. Thereafter, the proportion of methanol addition was gradually increased using 160 μL/min glycerol and 100 μL/min methanol for the first hour, 100 μL/min glycerol and 100 μL/min methanol for the second hour, and 65 μL/min glycerol and 100 μL/min methanol for the third hour. During the methanol induction phase, methanol was the only carbon source. Glycerol solution and methanol were added using automatic microburettes S1 (Crison Instruments S.A., Alella, Spain). The microburette system was sterilized beforehand with 1 M HCl, 70% (v/v) ethanol, 1 M NaOH, and sterile water. After transition, the N source was changed from 30% NH₄OH to NH₄Cl (100 g NH₄Cl diluted in 500 mL sterile H₂O), because of interference of the NH₄OH with the methanol sensor. NH₄Cl solution was also complemented with 10 mL of trace salts solution and 4 mL of biotin solution. The NH₄Cl solution flow rate was directly linked to methanol addition by previously calculated NH₄Cl/methanol requirements (0.12 g/g) [32]. The pH was controlled after this point by 5 M KOH addition. All cultivations ended when the culture reached a final biomass concentration of around 70 g/L of Dry Cell Weight (DCW), always below the maximal working volume (4 L). All the state variables and specific rates were calculated as previously described [33]. E2-Crimson fed-batch cultivation was performed in duplicate to obtain biological replicates for the fluorimetric and cytometry analyses shown in Figures 5 and 6.

Figure 5. Intracellular E2-Crimson fluorescence in fed-batch cultivations.

(A) Evolution of intracellular E2-Crimson fluorescence during fed-batch cultivation with the different signal sequences. The letters represent glycerol batch phase (B), transition phase (T), and methanol induction phase (I). The data represent the amount of fluorescence per 1 unit of OD₆₀₀. Error bars represent differences between biological replicates. (B) Images of intracellular fluorescence with the different signal sequences at the end of each fed-batch cultivation. Projected confocal Z-stacks of E2-Crimson fluorescence (red) were merged with differential interference contrast (DIC) images of the cells. Scale bar, 5 μm.

Figure 6. Effects of the signal sequences on cell size, ROS, and cell viability in fed-batch cultivations.

Results are shown from fed-batch cultivations expressing E2-Crimson with either the pre-α-MF or the pre-Ost1 signal sequence (Figure 4A) as well as the X33 reference fed-batch cultivation (Figure S2). The letters and numbers represent glycerol batch phase (B), transition phase (T), and the time of methanol induction (numbers). (A) Average cell diameter (± SD). (B) Percentage of cells showing ROS levels (± SD). (C) Percentage of viable cells (± SD). Error bars represent differences between biological replicates.

Methanol measurement

Methanol concentration was monitored on-line using an immersed sensor in the culture (Raven Biotech, Vancouver, BC, Canada), equipped with a Figaro TGS-822 sensing element (Figaro USA Inc., Glenview, USA) [34,35]. Methanol concentration was corroborated and/or corrected in the culture by HPLC off-line analyses, as reported elsewhere [34]. Relative standard deviation (SD) was below 5%.

Biomass analysis

Biomass concentration was measured as both DCW per L culture broth and OD₆₀₀. Samples were first centrifuged at 2,420 rcf for 2 min (Espresso Microcentrifugue, Thermo Electron Corporation, Waltham, MA, USA), to collect the biomass in a pellet. Pellets were washed and centrifuged twice in a 1 g/L citric acid and 9 g/L NaCl solution, and filtered through previously weighed 0.45 μm pore glass fiber filters (APPF04700, Merck Millipore Ltd, Burlington, MA, USA) by applying vacuum. Finally, filters were dried at 105°C for 24 h. OD₆₀₀ and DCW measurements were performed in triplicate. Relative SD was about 5%. In methanol batch experiments, cell growth was monitored by OD₆₀₀ and then converted to DCW with a correlation set as 1 unit DCW per 0.48 × OD₆₀₀.

Quantification of heterologous protein titers

Lipase activity assay

Enzyme activity assays were used to quantify protein secretion in the case of the BTL2 and ROL lipases. Activity was measured in triplicate using a lipase colorimetric assay, as previously described in [21]. Briefly, 0.5 mL of the suitably diluted culture supernatant from each strain was mixed with 0.5 mLTris-HCI buffer (200 mM, pH 7.25 for ROL or 100 mM, pH 7.25 for BTL2). The mixture was placed in a thermostatically controlled cuvette, at 30°C for ROL or 50°C for BTL2, pre-incubated for 2 min to reach the correct temperature. Then, 0.3 mL of substrate (Lipase assay reagent solution from Roche Diagnostics (Mannheim, Germany, Cat No. 11821792), containing 1,2-O-dilauryl-racglycero-3-glutaric-(methylresorufin)-ester, was added to the mixture and absorbance measured at 580 nm for 7 min in a Specord 200 Plus Spectrophotometer (Analytic Jena, Jena, Germany). Data used to determine the activity were taken from mins 3 to 5. The increased absorbance per s was used to calculate the activity, with 1 activity unit defined as the amount of enzyme needed to hydrolyze 1 μmol ester bond per min. The relative SD of the method was <5%.

E2-Crimson fluorescence assay

Both extracellular and intracellular fluorescence were analyzed. The former was measured from the supernatant obtained after centrifugation at 2,420 rcf in a microcentrifuge for 2 min. Intracellular fluorescence was measured from the solid fraction after centrifugation under the same conditions. The pellet corresponding to an OD₆₀₀ of 1 in 1 mL was first washed with phosphate-buffered saline (PBS) and then resuspended in 1 mL of PBS. Samples were stored at −20°C until the time of analysis.

100 μL each of the extracellular and intracellular fractions were transferred to a 1 cm quartz cuvette, and fluorescence was measured in duplicate in a Varian Eclipse Fluorescence Spectrophotometer (Agilent Technologies, Santa Clara, CA, USA). The excitation and emission wavelengths were 570 nm and 636 nm, respectively. Fluorescence intensity was measured in the same arbitrary units for all samples. Fluorescence values were measured in triplicate and relative SD was about 5%.

Fluorescence microscopy

Images for Figures 3B and 5C were captured as Z-stacks using an SP5 confocal microscope (Leica, Wetzlar, Germany) equipped with a 1.4-NA/63x oil objective. Images for Figures 1B and 3C were captured as Z-stacks using an Olympus Fluoview 1000 equipped with an UPIansApo/60x oil objective. The Z-stacks were average projected, and the brightness and contrast were adjusted evenly in all images. A Gaussian blur filter was used to smooth the red and green signals. Images were processed with ImageJ software (https://imagej.nih.gov/ij/).

Figure 3. Intracellular E2-Crimson fluorescence in a methanol batch cultivation.

(A) Evolution of intracellular E2-Crimson fluorescence during batch cultivation with the different signal sequences. The data are plotted in arbitrary units (A.U.) and represent the amount of fluorescence per 1 unit of OD₆₀₀. Error bars show standard deviations. (B) Images of intracellular fluorescence with the different signal sequences at the end of each batch cultivation containing the ER-targeted GFP-HDEL. Left: projected confocal Z-stacks of E2-Crimson fluorescence (red). Center: projected confocal Z-stacks of the GFP-HDEL. Right: The previous images merged. Scale bar, 5 μm.

Figure 1. Effects of the different signal sequences in shake-flask cultures.

(A) Box plot representing extracellular protein secretion in shake-flask cultures for E2-Crimson, BTL2, and ROL, comparing secretion signals containing the pre-α-MF and pre-Ost1 signal sequences. Eight clones were evaluated for each strain. Box plots for E2-Crimson and BTL2 were obtained from [21]. The activity/fluorescence values were normalized by setting the value for the pre-α-MF strain to 1. Plots were generated with GraphPad Prism. Each box extends from the 25^th to 75^th percentiles, with the internal line representing the median. Individual data points are shown as dots, and the whiskers mark the minimum and maximum values. (B) Fluorescence images of ER-targeted GFP-HDEL are shown. In the “Control” sample, no model protein was produced. Scale bar, 5 μM.

Flow cytometry analysis

Cell size and viability were measured as described by elsewhere [30,36]. Intracellular reactive oxygen species (ROS) formation was measured using the dihydroethidium (DHE) fluorochrome. Samples were first sonicated and diluted in the same way as when measuring cell size and viability. After dilution, 0.2 mL of DHE solution (12.5 μg/mL in filtered PBS) was added to 1 mL cell suspension. Samples were incubated at 30°C for 30 min with agitation and finally centrifuged at 11,360 rcf for 3 min and resuspended in 1 mL filtered PBS. In addition, positive and negative controls were prepared, in which the former was the same strain containing 20% (v/v) ethanol and the latter lacked DHE.

Results and Discussion

The pre-Ost1 signal sequence improves protein secretion and increases the maximum growth rate

We have previously shown that the pre-Ost1 signal sequence can drive passenger proteins through co-translational translocation into the ER lumen [21,23]. In shake flask cultures, this effect increased the final secreted protein titer in comparison with the commonly used pre-α-MF signal sequence [21]. However, it remained to be seen whether this improvement observed in small cultures could be scaled to the fermenter level, i.e., to bioprocess-relevant conditions, and applied to diverse proteins. For this purpose, we selected the BTL2 lipase from B. thermocatenulatus and E2-Crimson, which were previously examined in small-scale cultures [21], as well as the mature form of a R. oryzae lipase (ROL). These proteins were fused to either a conventional secretion signal containing the pre-α-MF signal sequence or an improved secretion signal containing the pre-Ost1 signal sequence. Eight independent clones were evaluated for each strain, and a representative clone was selected for subsequent batch and fed-batch cultivations.

As shown in Figure 1A, all the proteins were secreted more efficiently in shake flask cultures when the strains carried the improved secretion signal. However, the magnitude of the effect was protein dependent. Extracellular protein production was 6-fold higher for E2-Crimson, almost 3-fold higher for BTL2, and just 30% higher for ROL. For each strain, co-expression of GFP-HDEL was carried out to assess whether the heterologous proteins failed to cross the ER membrane and therefore blocked the ER translocons [21]. As shown in Figure 1B, when the three model proteins were fused to the conventional secretion signal containing pre-α-MF, GFP-HDEL accumulated in the cytosol, indicating that they had clogged the ER translocons. In contrast, when E2-Crimson and BTL2 were fused to the improved secretion signal containing pre-Ost1, GFP-HDEL gave a typical ER pattern similar to that seen in the control strain, indicating that the ER translocons were not blocked and therefore the proteins had presumably traversed the ER membrane. When ROL was fused to the improved secretion signal containing pre-Ost1, GFP-HDEL exhibited an uncharacteristic ER pattern that likely reflected an ER stress response, but once again there was very little GFP-HDEL fluorescence in the cytosol. These observations support the hypothesis that the pre-Ost1 signal sequence can improve the ability of multiple passenger proteins to traverse the ER membrane, but that the downstream effect on secretion depends on other properties of the passenger protein.

To further study the benefits of using the pre-Ost1 signal sequence, batch cultivations with an initial methanol concentration of 10 g/L were performed for all strains, including the X33 control strain. Cell growth, methanol concentration, and extracellular protein concentration are represented in Figure 2, and related bioprocess parameters are listed in Table 1.

Figure 2. Time course of methanol batch cultivations.

The average values for biomass (OD₆₀₀), methanol concentration (MetOH), and protein production (±SD) over time are represented for each cultivation. Seven batches performed in duplicate are shown, one for each model protein (E2-Crimson, BTL2, and ROL) and signal sequence (pre-α-MF and pre-Ost1) plus the X33 reference control strain.

Heterologous protein production has often been related to a reduction in the specific growth rate (μ) and biomass yield [37]. In particular, this correlation has been previously observed for ROL [27,38]. Such a negative effect was significantly diminished when the secretion signal contained pre-Ost1 instead of the pre-α-MF signal sequence (Table 1). Moreover, use of pre-Ost1 notably increased both the protein titers and the maximum specific protein production rate (q_p). As would be expected, the methanol specific consumption rate (q_s) was also higher with use of pre-Ost1 in agreement with the increase of μ_MAX (Table 1).

E2-Crimson accumulates in the ER when using the pre-Ost1 signal sequence in a methanol batch cultivation

To track the possible intracellular accumulation of a model protein during a methanol batch cultivation, E2-Crimson production by the pre-Ost1-E2-Crimson and pre-α-MF-E2-Crimson strains was monitored over time by measuring intracellular fluorescence levels (Figure 3A). In both strains, intracellular fluorescence increased during the cultivation period but, surprisingly, was significantly higher when E2-Crimson was fused to pre-Ost1. Insight into this phenomenon came from capturing cell images by fluorescence microscopy at the end of the batch cultivation (Suppl. Figure S1). With the pre-α-MF-E2-Crimson strain, fluorescent rings were visible in some cells. This effect was previously observed for shake-flask cultures and, because GFP-HDEL accumulated in the cytosol in those cultures, it was inferred that E2-Crimson became trapped during passage into the ER and blocked the translocons [21]. After a batch cultivation of the pre-α-MF-E2-Crimson strain, a similar cytosolic accumulation of GFP-HDEL was apparent, suggesting that E2-Crimson was once again becoming trapped during passage into the ER (Figure 3B). With the pre-Ost1-E2-Crimson strain, fluorescent rings were also visible after a batch cultivation (Suppl. Figure S1), but co-localized with GFP-HDEL in a normal ER pattern (Figure 3B), suggesting that the ER translocons were not clogged and that E2-Crimson accumulated inside the ER. A likely interpretation is that under methanol batch conditions, the pre-Ost1 signal sequence continued to drive translocation across the ER membrane, but the resulting high level of E2-Crimson in the ER created a new bottleneck at the level of ER export.

Secretion of model proteins in fed-batch cultivations is enhanced by using the pre-Ost1 signal sequence

Fed-batch cultivations were next performed to test culture behaviour under bioprocess-like conditions, i.e. at high cell densities. The methanol level was maintained at 3 g/L. Evolution of biomass and the concentrations of the secreted model proteins during the fed-batch cultivations are shown in Figure 4, and an X33 reference cultivation is shown in Suppl. Figure S2. The bioprocess parameters are listed in Table 2. Each cultivation was run until the biomass concentration reached around 70 g/L of DCW.

Figure 4. Time course of fed-batch cultivations.

Cell growth (left graphs) and fluorescence or lipolytic activity (right graphs) are plotted overtime for the different signal sequences for (A) E2-Crimson, (B) BTL2, and (C) ROL. The letter symbols represent glycerol batch phase (B), transition phase (T), and methanol induction phase (I). A.U., arbitrary units; U.A., units of lipase activity. Error bars in E2-Crimson correspond to the differences between biological duplicates and error bars in BTL2 and ROL represent the difference between technical triplicates.

Table 2. Fed-batch bioprocess parameters.

Summary of specific growth rates (μ), specific methanol (MetOH) consumption rates (q_s), specific protein production rates (q_p), protein/biomass yields (Y_(p/x)), biomass/methanol yields (Y_(x/s)), volumetric productivities, and specific productivities during fed-batch cultivations with the reference (X33), E2-Crimson, BTL2, and ROL producing strains. Biomass concentrations used in calculations were measured as dry cell weight (DCW). E2-Crimson levels were measured in arbitrary units (A.U.) of fluorescence, and BTL2 and ROL levels were measured in units of lipase activity (U.A.).

	X33	E2-Crimson		BTL2		ROL
	control	pre-α-MF	pre-Ost1	pre-α-MF	pre-Ost1	pre-α-MF	pre-Ost1
Final protein titer*	-	879.6	3096.2	107.0	259.9	192.4	291.9
μmean (h−1)	0.081	0.050	0.050	0.047	0.072	0.026	0.040
qs (gMetOH)/gDCW·h)	0.30	0.20	0.19	0.18	0.25	0.11	0.14
qp (protein/gDCW·h)	-	740.0	2357	81.6	255.4	74.9	155.0
Y(x/s) (gDCW/gMetOH))	0.28	0.25	0.25	0.26	0.27	0.25	0.27
Y(P/X) (Total protein/gDCW)	-	13620	44962	1739	3487	2863	3873
Volumetric productivity (Total protein/(L·h))	-	19465	66250	2035	6047	2681	4906
Specific productivity (Total protein/(gDCW·h))	-	253	901	28.7	77.6	37.0	63.7

^*Protein titer units: E2-Crimson (A.U./mL): BTL2: (U.A./mL): ROL: (U.A./mL).

In contrast with the observation in E2-Crimson batch cultivations, the mean specific growth rate was maintained similarly when producing E2-Crimson with either the pre-Ost1 or pre-α-MF signal sequence (Figure 4A) at this methanol concentration set-point. However, when comparing secreted protein titers, extracellular fluorescence was 3.5-fold higher with the pre-Ost1-E2-Crimson strain, similar to the value obtained in batch (3.8-fold higher). Moreover, volumetric productivity, specific productivity and Y_P/X were, respectively, 3.4-fold, 3.6-fold, and 3.3-fold higher with the pre-Ost1-E2-Crimson strain compared to the pre-α-MF-E2-Crimson strain (Table 2).

For the BTL2 cultivations, the maximum targeted biomass concentration was reached in a shorter cultivation time with the pre-Ost1-BTL2 strain than with the pre-α-MF-BTL2 strain (Figure 4B), yielding a higher mean μ (Table 2). Regarding protein secretion, extracellular lipase activity was increased 2.4-fold when using pre-Ost1, while volumetric productivity and specific productivity were 3.0-fold and 2.7-fold higher, respectively (Table 2).

For the ROL cultivations, the pre-Ost1-ROL strain gave a 1.5-fold increase in the final product titer compared to the pre-α-MF-ROL strain (Figure 4C). Because the growth rate of the pre-Ost1-ROL strain was significantly higher, the volumetric productivity and specific productivity were almost doubled (Table 2).

Overall, the pre-Ost1 signal sequence yielded increased protein productivities due to higher protein titers combined with faster growth rates that allowed earlier termination. Nonetheless, the growth rates in the fed-batch cultivations were lower than in the bioreactor batch cultures, particularly for the strains expressing ROL. This reduced growth might be related to cellular stress due to extended time under inducing conditions, high cell densities, and high-level expression of proteins with cytotoxic properties.

pre-Ost1 efficiently improves protein translocation but its accumulation at the ER level triggers ER stress

Building on the results with the methanol batch cultivations, the strains expressing E2-Crimson were examined during the fed-batch cultivations. In the transition phase, when addition of methanol started, intracellular fluorescence started to increase for both the pre-α-MF-E2-Crimson and the pre-Ost1-E2-Crimson strains. However, in the induction phase, intracellular fluorescence evolution was different with the two signal sequences despite similar growth of the two production strains during this phase (Figure 5A). In the pre-α-MF-E2-Crimson fed-batch cultivation, intracellular fluorescence increased during the first 12 h of methanol feeding and then remained stable until the end of the cultivation. In contrast, in the pre-Ost1-E2-Crimson cultivation, intracellular fluorescence initially accumulated faster, but then decreased after 10 h of methanol feeding. These results agree with fluorescence microscopy images (Figure 5B), which revealed intracellular fluorescent rings similar to those seen after the methanol batch cultivations indicating ER accumulation of E2-Crimson, and also confirmed that intracellular fluorescence was notably higher for the pre-α-MF-E2-Crimson strain at the end of the fed-batch cultivation.

Average cell size, ROS accumulation and cell viability during fed-batch cultivations were also measured for the pre-α-MF-E2-Crimson and pre-Ost1-E2-Crimson strains as well as the X33 reference strain (Figure 6). These parameters are good indicators of the physiological state of the cells. It has been reported that in methanol fed-batch cultivations of recombinant P. pastoris, metabolic stress increased while cell viability often decreased [39–41]. Intracellular protein accumulation is one of the main causes of metabolic stress, which can compromise cellular viability [41–43], and cell size and ROS levels are metabolic stress indicators [44,45]. In the present analysis, cell size tended to increase during the cultivation period for all the strains (Figure 6A). However, such increases were 10~20% smaller in the X33 reference strain than in the strains expressing E2-Crimson (depending on the strain and cultivation time after induction). In addition, in the pre-Ost1-E2-Crimson strain cultivation, cells were 5~23% smaller (depending on the cultivation time) than in the pre-α-MF-E2-Crimson strain. ROS levels also increased over time in all of the strains, and this effect was greater in the strains expressing E2-Crimson (Figure 6B). Specifically, in the pre-α-MF-E2-Crimson strain, the percentage of cells with high ROS levels gradually increased until reaching 33% of the population at the end of the fed-batch cultivation. Interestingly, with the pre-Ost1-E2-Crimson strain, the percentage of cells with high ROS levels transiently reached higher levels than with the pre-α-MF-E2-Crimson strain after the transition phase and at the beginning of the induction phase, but started to decrease 15 h after, reaching significantly lower values than those observed for the pre-α-MF-E2-Crimson strain towards the end of the induction phase (Figure 6B). Finally, cell viability was statistically significant higher in the X33 reference strain, where cell viability was maintained at over 90% throughout the fermentation, than in E2-Crimson-producing strains, and slightly higher in the pre-Ost1-E2-Crimson strain than in the pre-α-MF-E2-Crimson strain, in which the final viability was 10% lower at the end of the fermentation (Figure 6C). Taken together, these results indicate that E2-Crimson production leads to accumulation of this protein inside the ER and subsequently stresses the cells, probably upregulating the ER protein degradation machinery. The improved secretion signal apparently enabled cells to adapt better to stress after the initial response, somewhat alleviating the stress during long fed-batch cultivations. However, they adapted to the stress conditions too late, resulting in a failure to increase growth rate and support higher specific production rates throughout the later stages of the induction phase. This phenomenon bears further investigation.

To this end, transcriptional analysis of P. pastoris fed-batch cultivations should lead to the understanding of the initial stress response and adaptation to the dynamic conditions along the fed-batch phase, thereby supporting the identification of physiological limitations in the later phases [46], as a basis for subsequent design of cell engineering strategies. Even if efficient protein translocation across the ER is achieved by using the pre-Ost1 signal sequence, this study suggests that very high expression levels could create a new bottleneck at the level of ER export. A potential strategy to circumvent this issue would be to overexpress the Erv29 protein [47], which serves as a receptor recognizing an ER export signal in the pro region of the secretion signal [48,49]. Thus, further improvements may be possible due to continued rational engineering of the P. pastoris secretory pathway.

Conclusions

These observations show that the pre-Ost1 signal sequence can improve heterologous protein secretion at a bioreactor scale by driving the proteins through co-translational translocation instead of post-translational translocation. Thus, pre-Ost1 has great potential to replace pre-α-MF at an industrial scale. Efficient co-translational translocation appears to reduce the metabolic burden for some model proteins, probably by ensuring that a larger fraction of the protein molecules enter the secretory pathway correctly. Further studies should provide insights into the physiological impact of this secretion strategy. To exploit fully the power of the improved secretion signal, additional cellular engineering may be needed to overcome bottlenecks that can appear downstream of the translocation event, particularly under bioprocess-type (fed-batch) conditions.

HIGHLIGHTS.

• pre-Ost, an engineered signal sequence for improved protein secretion in yeast

• Bioreactor-scale evaluation of pre-Ost’s full potential in Pichia pastoris

• The evaluation was carried out with three different model proteins

• The new signal sequence can easily outperform the pre-α-MF in Pichia pastoris

• The new signal sequence enhances cell performance in bioreactor cultures

List of abbreviations

BTL2

Bacillus thermocatenulatus lipase 2

DCW

dry cell weight

GFP-HDEL

GFP with the ER retention signal HDEL

μ

growth rate

msGFP

monomeric superfolder variant of Green Fluorescent Protein

pre-α-MF

signal sequence from the mating α-factor of S. cerevisiae

pre-Ost1

signal sequence from the Ost1 protein of S. cerevisiae

q_s

methanol consumption rate

q_p

production rate

ROL

Rhizopus oryzae lipase

ROS

reactive oxygen species