Recombinant production of a hard‐to‐express membrane‐bound cytochrome P450 in different yeasts—Comparison of physiology and productivity

Abstract Cytochrome P450s comprise one of the largest protein superfamilies. They occur in every kingdom of life and catalyse a variety of essential reactions. Their production is of utmost interest regarding biotransformation and structure‐function elucidation. However, they have proven hard to express due to their membrane anchor, their complex co‐factor requirements and their need for a redox‐partner. In our study, we investigated and compared different yeast strains for the production of the plant cytochrome P450 chalcone 3‐hydroxylase. To our knowledge, this is the first study evaluating different yeasts for the expression of this abundant and highly significant protein superfamily. Saccharomyces cerevisiae and three different strains of Pichia pastoris expressing chalcone 3‐hydroxylase were cultivated in controlled bioreactor runs and evaluated regarding physiological parameters and expression levels of the cytochrome P450. Production differed significantly between the different strains and was found highest in the investigated P. pastoris MutS strain KM71H where 8 mg P450 per gram dry cell weight were detected. We believe that this host could be suitable for the expression of many eukaryotic, especially plant‐derived, cytochrome P450s as it combines high specific product yields together with straightforward cultivation techniques for achieving high biomass concentrations. Both factors greatly facilitate subsequent establishment of purification procedures for the cytochrome P450 and make the yeast strain an ideal platform for biotransformation as well.


| INTRODUCTION
Cytochrome P450s, sometimes referred to as "Nature's most versatile catalysts" (Gillam, 2007) form one of the biggest protein superfamilies and are present in all kingdoms of life (Werck-Reichhart & Feyereisen, 2000). They are responsible for the catalysis of uncountable reactions, ranging from detoxification over hormone synthesis to carbon source assimilation (Guengerich, Waterman, & Egli, 2016;Werck-Reichhart & Feyereisen, 2000). The core of these enzymes consists of an iron porphyrin group, which classifies them as hemoproteins. Interestingly, aside from this characteristic feature cytochrome P450s often share only little homology regarding their amino acid sequences, sometimes below 16% (Werck-Reichhart & Feyereisen, 2000). This makes recombinant expression of cytochrome P450s for structure-function elucidation inevitable. Moreover, their recombinant production can be of interest for biotransformations, as these enzymes are able to specifically hydroxylate complex hydrocarbons (Girvan & Munro, 2016).

However, recombinant expression of eukaryotic cytochrome
P450s is challenging due to a variety of reasons: (1) They have complex co-factor requirements: Incorporation of the heme group into their core is essential for catalytic activity; (2) a redox partner protein, cytochrome P450 reductase, is required for catalytic activity; (3) eukaryotic cytochrome P450s are usually membrane bound, therefore they are either engineered as soluble enzyme and expressed in prokaryotic hosts as for instance Escherichia coli or they are natively expressed in eukaryotic hosts such as yeasts, mainly Saccharomyces cerevisiae. Sudhamsu et al., 2010;Wagner et al., 2008).
We wanted to express the cytochrome P450 as a full-length protein for future structure elucidation. This seemed particularly interesting as only one cytochrome P450 has been crystallized including its native membrane anchor so far, although this region is not very well conserved (Monk et al., 2014;Nelson, 2006). As E. coli lacks inner organelles which makes most of the prokaryotic strains unsuitable for membrane protein production, we chose yeasts for CH3H expression (Rosano & Ceccarelli, 2014).
Interestingly, up to now, many more studies regarding cytochrome P450 expression have been conducted in S. cerevisiae rather than in Pichia pastoris (Hausjell, et al., 2018a). Therefore, we were curious if indeed baker's yeast is a superior host for recombinant production of these enzymes. We cultivated one S. cerevisiae strain and three different strains of P. pastoris (KM71H, GS115 and SMD1168H; Table 1) expressing CH3H in controlled bioreactor runs and analysed strain specific parameters as well as CH3H production levels. All three P. pastoris strains are frequently employed for recombinant protein expression (Ahmad, Hirz, Pichler, & Schwab, 2014). Strains KM71H and GS115 were chosen as they had been employed successfully in several cases for membrane protein production before (Byrne, 2015).
However, most frequently strain SMD1163 was used as a host for the production of membrane proteins (Byrne, 2015). As this strain is no longer commercially available, we decided to investigate its close relative, strain SMD1168H, instead.
KM71H is a methanol utilizer slow (MutS) strain carrying only the alcohol oxidase 2 (aox2) gene. In contrast GS115 as well as SMD1168H both carry the aox1 and aox2 gene, characterizing them as methanol utilizer plus (Mut+) strains. SMD1168H is often favored for production of recombinant protein as it is protease deficient, avoiding degradation of target protein (Ahmad et al., 2014).
One study already investigated the expression of a membranebound protein (catechol-O-methyltransferase) in a MutS and Mut+ strain, however, only in shake flasks, where the results strongly depended on the carbon source present (Pedro et al., 2015). We decided to shed more light on the suitability of Mut+ and MutS strains for membrane protein expression by performing controlled bioreactor runs of strains expressing the bitopic membrane protein CH3H.
To our knowledge, this is the first study comparing different yeasts as expression hosts for cytochrome P450s. In finding the most suitable expression host, our main criteria were high space-time-yields and high product titres, as we ultimately wanted to establish a purification procedure for the membrane bound cytochrome P450, which can be rather cumbersome as low recovery yields are an issue and stability can be challenging.

| S. cerevisiae
The coding sequence of DvCH3H was synthesized codon-optimized for expression in yeast (Genscript, Piscataway, NJ, USA). The primers (

| Cultivations
All cultivations consisted of a batch and fed-batch phase for biomass generation, followed by an induction phase for production of target protein. Samples were taken in the beginning of the batch, the end of the batch and start of the fed-batch, before induction and every 24 h during induction. All samples were analysed regarding dry cell weight as well as metabolites in the supernatant.

Batch and fed-batch phase
The cultivation was carried out in an InforsHT Labfors 5 fermenter equipped with a 5 L glass-vessel (Infors, Bottmingen, Switzerland). All process parameters were logged at least every 30 s and controlled by the process information management system Lucullus (Securecell, Schlieren, Switzerland). 1.8 L Delft Media (Jensen et al., 2014) with 20 g/L glucose (supplemented with 0.002 g α-L-Tyrpthophan and 0.002 α-L-Histidine per gram of glucose) were inoculated with 200 mL preculture. The temperature was kept constant at 30 C, the reactor was stirred at 1400 rpm and aerated at 2 vvm to avoid oxygen limitation. an increase in dissolved oxygen and yielded 1.5 g/L biomass. Thereafter a fed-batch phase was started at a constant q s,glu using feed forward strategy according to equation 1 with F being the feed rate in g/ h, q s being the specific substrate uptake rate in g/g/h, X being the biomass concentration in g/L, V being the reactor volume in L and W being the amount of substrate per gram of feed in g/g.
We fed with a 450 g/L glucose feed (supplemented with 0.002 g α-L-Trypthophan and 0.002 α-L-Histidine per gram of glucose) at a q s,glu of 0.4 g/g/h for 48 h reaching a biomass concentration of 10 g/L. The feed was pumped through a preciflow peristaltic pump (Lambda Laboratory Instruments, Brno, Czech Republic) controlled by the process information management system and its addition was monitored gravimetrically.

Induction phase
Induction was performed by addition of copper sulfate to a final concentration of 0.5 mM. In addition, a pulse of hemin-chloride to a final concentration of 5 μM was added. The feed-rate was turned down from 0.4 to 0.05 g/g/h to avoid stressing the cells. Induction for production of target protein lasted 120 h. were adjusted using a soft-sensor tool (Wechselberger, Sagmeister, & Herwig, 2013).

Induction phase
For induction, a methanol adaption pulse (pure methanol supplemented with 12 mL/L PTM1) was added to the cultivations reaching a final concentration of 0.5 % (v/v) in the bioreactor. In addition, a pulse of hemin-chloride to a final concentration of 5 μM was added. After complete metabolization of the first pulse, a second pulse of 1 % (v/v) was added for determination of the maximum specific methanol uptake rate (q s,meoh,max ). Thereafter a methanol feed (400 g/L Methanol, 4 mL/L PTM1, 0.3 mL/L Struktol J650) was started. We fed at a constant q s,meoh of 0.03 g/g/h until an overall induction time (starting with the adaption pulse) of 120 h was reached.

| Analysis of dry cell weight and metabolites
Dry cell weight, optical density at 600 nm (OD 600 ) as well as metabolite concentrations in the supernatant were analysed offline during the cultivation. For analysis of the cell dry weight, 1 mL culture broth was transferred into pre-dried and pre-weighed 2 mL plastic tubes and centrifuged (4500 g, 4 C, 10 min

| Analysis of CH3H expression levels
The

| Data analysis
For determination of statistical significance of differences between results, non-parametric Kruskal-Wallis tests were performed due to the small sample (usually three measurements).
Significance of differences was determined at a 95 % confidence level.

| Cultivations of S. cerevisiae
We first investigated expression of CH3H in S. cerevisiae as, in general, higher cytochrome P450 expression titres were found compared with studies in P. pastoris. However, to our knowledge there is no study, where the same cytochrome P450 has been expressed in both hosts, which makes comparisons difficult . Strain INVSc1 was chosen together with the CUP1 promoter system, where production of target protein is induced upon addition of copper (Butt & Ecker, 1987). The yeast cells were first cultivated in batch and fed-batch mode for biomass generation and subsequently induced by the addition of copper sulfate to a final concentration of 0.5 mM.

| Physiological parameters
Physiological parameters were calculated and are summarized in Table 2.
The strain was cultivated on glucose as source of carbon, as there was no growth detectable when glycerol was used as C-source in a defined medium (data not shown).
In the batch phase, high ethanol and glycerol production (up to 7 g/L and 2 g/L, respectively) were observed, coming hand in hand with  Wojtczak, 1996). However, even during the uninduced fed batch, where glucose was fed in limiting amounts, ethanol and glycerol were produced, although at a lower rate. This nevertheless resulted in concentrations of up to 39 g/L and 12 g/L, respectively. Due to this vast production of metabolites the biomass never exceeded levels of 14.6 g/L dry cell mass. Moreover, it has been reported that concentrations of 4-6 v/v % or 32-47 g/L ethanol strongly induce stress response in the cells, similar to a heat shock response (Piper, 1995) and it has also been shown that viability decreased below 80 % within one hour after a 5 v/v % (or 39 g/L) pulse of ethanol (Birch & Walker, 2000). None of these factors seemed ideal, however, we still wanted to investigate expression of CH3H.

| Production of cytochrome P450
Productivity was investigated on a western blot, with strep-tagged protein standards, where a specific product yield of 0.7 ± 0.1 mg CH3H per gram biomass was found, which, at a biomass concentration of 14.6 g/L, corresponds to a volumetric titre of 9.6 ± 1.9 mg CH3H per litre of cultivation broth. In literature the highest T A B L E 2 Physiological parameters of

S. cerevisiae INVSc1 expressing CH3H
Physiological parameters of S. cerevisiae INVSc1 expressing CH3H during a cultivation consisting of a batch and fed-batch phase followed by an induced fed-batch. Standard deviations were calculated from triple measurements and calculated by error propagation for rates and yields. Abbreviations: q s,Glu : specific glucose uptake rate; q P, EtOH : specific ethanol production rate; q P,Gly : specific glycerol production rate;  . In comparison, the titre achieved for CH3H expression in S. cerevisiae (53 pmol/mg protein) seemed to be at the lower end and was perhaps at least partially a result of the high stress put on the cells by the high ethanol levels.
All of these results led us to conclude that S. cerevisiae was not a suitable host for the production of CH3H.

| Cultivations of P. pastoris KM71H, GS115, SMD1168H
Three different P. pastoris strains (KM71H, GS115, SMD1168H; Table 1) were transformed for expression of the plant P450 under control of the alcohol oxidase 1 (AOX1) promoter. All were cultivated analogously and to investigate effects from recombinant protein production, also an empty strain ("KM71H empty") was fermented identically: A batch was carried out on glycerol for biomass generation followed by a glycerol fed-batch with the same purpose. Thereafter methanol was pulsed to allow the cells to adapt to the different C-source. A second pulse of this substrate was applied afterwards allowing determination of the maximum methanol uptake rate q s,meoh,max , similar to previous studies from our group (Dietzsch, Spadiut, & Herwig, 2011a;Dietzsch, Spadiut, & Herwig, 2011b). Subsequently a methanol fed-batch was started to induce production of CH3H.

| Physiological parameters
All four strains were investigated regarding characteristic physiological parameters during the cultivation. The carbon dioxide evolution rate (CER) was calculated and is plotted over time in Figure 1 exemplarily for one MutS and one Mut+ strain. Based on the CER, biomass and substrate concentration, strain specific parameters during the cultivation were calculated and are summarized in Table 3.
While on glycerol all strains show similar behaviour, there are striking differences once they are grown on methanol. The adaption time to methanol (time from pulse addition to the maximum of the CER) differs strongly when comparing KM71H with SMD1168H and GS115, the one in KM71H being approximately by a factor ten longer.
This again results from the differences in the methanol utilization of the strains. It goes hand in hand with the maxima of the specific methanol uptake rates, where the Mut+ strains GS115 and SMD1168H exhibit higher values than the MutS strain. As they have two AOX enzymes, they can convert more methanol per biomass per hour.
When comparing the empty KM71H strain with the producing one, it can be stated that the maximum methanol uptake rate is slightly lower in the producing one. We believe this is a result from the additional burden of recombinant expression of the target protein. The same was found before by our group when a native KM71H strain was compared with a horseradish-peroxidase expressing one (Dietzsch et al., 2011b).

F I G U R E 1 Carbon dioxide evolution rate (CER) over time during the cultivation of Mut+ strain GS115 (A) and MutS strain KM71H (B).
The CER during the methanol pulses is zoomed out (right). The adaption time was calculated from addition of the pulse until the maximum of the CER was reached, indicated by the star. The different phases during the cultivation are numbered chronologically: The glycerol batch phase (1) was followed by a glycerol fed-batch (2). Thereafter, the adaption pulse of 0.5 % methanol was added (3) and a second methanol pulse with 1 % was applied after complete metabolization of the first (4). After that, a methanol fed-batch was started (5) which lasted for 72 h

| Production of cytochrome P450
Motivated by the promising results of the cultivations of the different P. pastoris strains, we investigated whether we could detect differences in recombinant protein production in the studied strains.
Indeed, we found striking differences in specific product titres when we investigated CH3H expression on western blots ( Figure A1. Additional file 1: TIFF, Western Blot).
Strain KM71H clearly produced the most target protein with 8.0 ± 0.6 mg P /g X , followed by GS115 (4.5 ± 1.2 mg P /g X ) and SMD1168H (1.9 ± 0.3 mg P /g X ). It is the only MutS strain among the investigated and we have two hypotheses for this behaviour. KM71H in contrast to GS115 and SMD1168H only expresses AOX2 when fed with methanol but not AOX1. We believe that this results in less stress for the organism, as for the other strains also AOX1 has to be expressed at high levels. It has been reported that when grown on methanol, AOX can make up 20-30% of total protein (Zhang et al., 2009). This could make a crucial difference especially during the production of hard-toexpress proteins such as the membrane bound CH3H. Another difference between Mut+ and MutS strains is the hydrogen peroxide concentration during methanol metabolization, which is elevated in Mut+ strains. Hydrogen peroxide is deleterious to cytochrome P450s (Guengerich, Martin, Sohl, & Cheng, 2009), which is why MutS strains might be more suitable for their production.
Among the Mut+ strains, GS115 outcompeted SMD1168H in regard to CH3H production. This was most likely due to the fact that in the cultivation of strain SMD1168H methanol accumulation was observed. Although we only fed at a specific methanol uptake rate of 0.03 g/g/h, which is below half of the maximum specific uptake rate, we observed methanol accumulation 48 h after starting the methanol fed-batch. We turned down the feed rate to 0.01 g/g/h, however, such low feed rates only deliver energy close the maintenance level which results in neither growth nor productivity (Capone, Horvat, Herwig, & Spadiut, 2015). We hypothesize that the methanol accumulation in strain SMD1168H could be caused by its vacuolar aspartyl protease deficiency (Table 1). It has previously been shown that yeasts lacking this gene show increased apoptosis in the presence of hydrogen peroxide (Alugoju, Janardhanshetty, Subaramanian, Periyasamy, & Dyavaiah, 2018), which is produced during the first step of methanol metabolization.
Within this study we investigated expression of CH3H on a methanol-only feed in the three different P. pastoris strains. Another option instead of feeding methanol alone would be the application of mixed feeds of methanol together with glycerol or other C-sources. This has especially often been applied for MutS strains, as they grow slower on methanol (Krainer et al., 2012). For Mut+ strains studies with mixed feeds showed the same or even lower productivity (Jahic, Veide, Charoenrat, Teeri, & Enfors, 2006;Jungo, Marison, & von Stockar, 2007). This led us to conclude that such process conditions could potentially increase productivity in KM71H but most likely would not change or even reduce productivity in the Mut+ strains.

| Comparison of P. pastoris and S. cerevisiae as expression hosts
In Figure 2, the different hosts are compared in regard to their physiology and productivity.
Regarding physiological parameters in the cultivations of P.
pastoris and S. cerevisiae, it is obvious that P. pastoris is favorable when T A B L E 3 Physiological parameters for the P. pastoris strains. Physiological parameters of three P. pastoris strains expressing CH3H as well as one empty strain. Standard deviations were calculated from triple measurements and calculated by error propagation. Differences in the maximum specific methanol uptake rates were determined significant between Mut+ and MutS strains at a confidence level of 95 %. In the SMD1168H cultivation strong methanol accumulation occurred after 48 h of induction. Therefore, the feed-rate was reduced. However, the biomass concentration continuously decreased as a result of feeding close to the maintenance level. This resulted in an extremely low biomass yield and high standard deviation thereof. A star in the table marks respective parameters. Abbrevations: μ max,gly : maximum growth rate on glycerol; q s,max,gly : maximum specific glycerol uptake rate; Y X/S : biomass yield on the respective substrate Y CO2/S : CO 2 yield on the respective substrate; C-Balance: carbon balance; t adapt : adaption time to methanol; q s,adapt : specific methanol uptake rate during adaption time; q s,max,meoh : maximum methanol uptake rate it comes to conversion of C-source into biomass as higher biomass concentrations can be achieved and the biomass yield during the batch and fed-batch phase is clearly elevated compared with S.
Regarding productivity it can be stated that even the lowest product yield in P. pastoris found in strain SMD1168H, is more than twice as high as the product yield found in S. cerevisiae ( Figure 2F). The obtained results for expression of CH3H, strongly argue for employment of P. pastoris as expression hosts for cytochrome P450s, since both, biomass concentrations and specific product yields were higher in comparison with S. cerevisiae.
In general the cytochrome P450 yield found in our study is higher compared with values reported so far in literature which range around 50 to 400 pmol/mg protein for cytochrome P450 expression in yeasts . The product yield we found of 8.0 mg CH3H per gram dry cell weight corresponds to 600 pmol/mg protein, which is 1.5-times as high as the highest expression yield reported (400 pmol/mg (Haudenschild, Schalk, Karp, & Croteau, 2000)). Nevertheless, these results are hard to compare as recombinant production of cytochrome P450s has not often been studied in controlled bioreactor cultivations but rather in shake-flask experiments.
Although the product yield is already remarkably high, there could be room for improvement by optimization of other process parameters such as temperature, growth rate or induction time, which were kept at the same level in this study for all strains to allow comparability.
Finally, in spite of these promising production results, it has to be kept in mind that the protein needs to be purified for structure-function elucidation. This can be quite challenging as the protein has to be extracted out of the membrane by detergents while retaining its native conformation (Prive, 2007). Usually numerous different conditions, including various detergents and concentrations thereof, buffer compositions, incubation times and temperatures as well as protein concentrations are screened to find the most advantageous method (Newby et al., 2009). Considering the number of experiments needed, the high biomass and product yield found in the CH3H expressing KM71H strain provide an optimal starting point.

| CONCLUSION
Summing up, we recommend the yeast P. pastoris for the production of cytochrome P450s. It is Crabtree negative and thus allows more straightforward cultivations up to much higher biomass concentrations. Among the tested P. pastoris strains, the MutS phenotypic KM71H clearly was the best suited for production of CH3H. We believe that this host could be suitable for the expression of many F I G U R E 2 Comparison of the different yeast strains regarding physiology and productivity. Yields during the batch phase (A), fedbatch phase (B) and induced fed-batch phase (C) are shown as well as the obtained dry cell weight concentrations (D), product titres (E) and product yields (F) at the end of the cultivations."I" indicates S. cerevisiae strain INVSc, "G" P. pastoris strain GS115, "S" P. pastoris strain SMD1168H and "K" P.pastoris strain KM71H. "+" is written for strains expressing CH3H, "-"for empty strains. For A), (B) and (C) only the errors on the C-balances (calculated by error propagation) are shown for better readability. Errors on the respective yields can be found in Tables 2 and 3. For (D), (E), and (F) errors were calculated from at least three measurements. Differences in productivity were determined as significant at a 95 % confidence level eukaryotic, especially plant-derived, cytochrome P450s as it combines high specific product yields together with the capability of cultivations to high cell densities. It thus provides optimal prerequisites for either biotransformation by cytochrome P450 enzymes or establishment of purification procedures for the protein allowing structure function elucidation.