Beyond alcohol oxidase: the methylotrophic yeast Komagataella phaffii utilizes methanol also with its native alcohol dehydrogenase Adh2

ABSTRACT Methylotrophic yeasts are considered to use alcohol oxidases to assimilate methanol, different to bacteria which employ alcohol dehydrogenases with better energy conservation. The yeast Komagataella phaffii carries two genes coding for alcohol oxidase, AOX1 and AOX2. The deletion of the AOX1 leads to the MutS phenotype and the deletion of AOX1 and AOX2 to the Mut– phenotype. The Mut– phenotype is commonly regarded as unable to utilize methanol. In contrast to the literature, we found that the Mut– strain can consume methanol. This ability was based on the promiscuous activity of alcohol dehydrogenase Adh2, an enzyme ubiquitously found in yeast and normally responsible for ethanol consumption and production. Using 13C labeled methanol as substrate we could show that to the largest part methanol is dissimilated to CO2 and a small part is incorporated into metabolites, the biomass, and the secreted recombinant protein. Overexpression of the ADH2 gene in K. phaffii Mut– increased both the specific methanol uptake rate and recombinant protein production, even though the strain was still unable to grow. These findings imply that thermodynamic and kinetic constraints of the dehydrogenase reaction facilitated the evolution towards alcohol oxidase-based methanol metabolism in yeast.


INTRODUCTION
Aerobic utilization of the C1 compound methanol is an exceptional capability of certain microbial species and it is a result of adaptation to specific niche environments where methanol is present. One such rather large environment is the Phyllosphere, the areal surface of plants and the associated decaying plant matter. The metabolism and decomposition of the cell wall constituent pectin serves as the source of methanol (Fall and Benson 1996;Galbally and Kirstine 2002;Kawaguchi et al. 2011;Vorholt 2012). These microorganisms, termed methylotrophs are phylogenetically very diverse and span across the domains. They comprise of Gram-positive and Gram-negative bacteria as well as yeasts. These microbes have developed different metabolic adaptations to achieve utilization of methanol and other C1 compounds as an energy and carbon source (Anthony 1982). The first step of methanol metabolism is the oxidation of methanol to formaldehyde, which represents a central key metabolite in most but not all of the known methylotrophic pathways. At this point the pathway partitions into the assimilation of formaldehyde into the biomass and dissimilation to CO 2 for generating energy (Anthony 1982;Dijkhuizen, Levering and de Vries 1992;Yurimoto, Kato and Sakai 2005;Khadem et al. 2011). In yeasts these two pathways are also spatially separated into different compartments, the cytosol and peroxisome (van der Klei et al. 2006;Rußmayer et al. 2015).
The formation of formaldehyde from methanol is accomplished by different enzymes depending on the organism. Gramnegative methylotrophic bacteria have evolved a pyrroloquinoline quinone (PQQ) depended alcohol dehydrogenase (Adh) while their Gram-positive counterparts such as the thermophilic Bacillus methanolicus have an NAD + dependent methanol dehydrogenase (Mdh) for the same purpose (Arfman et al. 1992;Dijkhuizen, Levering and de Vries 1992;Yurimoto, Kato and Sakai 2005;Krog et al. 2013). NAD + dependent Mdhs have a low activity towards methanol (Krog et al. 2013;Ochsner et al. 2014) and the Gibbs free energy of methanol oxidation by Mdh is unfavorable at mesophilic temperatures compared to either Aox or the PQQ dependent Adh reaction (Whitaker et al. 2015), which may explain why the use of Mdhs evolved in thermophilic bacteria. Methylotrophic yeasts such as Komagataella phaffii (Pichia pastoris) or Ogataea polymorpha depend on an alcohol/methanol oxidase (Aox/Mox) to convert methanol to formaldehyde (Yurimoto, Kato and Sakai 2005;Yurimoto, Oku and Sakai 2011). This reaction implies two disadvantages: (i) Alcohol oxidase produces H 2 O 2 which necessitates the localization of the enzyme to the peroxisome where the harmful H 2 O 2 can be safely degraded without harming the cell (van der Klei et al. 2006). (ii) As the electrons are directly transferred from methanol to O 2 they do not pass the electron transport chain, so that the ATP yield and subsequently the biomass yield (Y X/S ) are reduced (Sheehan et al. 1988;Whitaker et al. 2015). A working hypothesis is that yeasts developed Aox as an alternative, thermodynamically feasible methanol oxidizing reaction with faster kinetics at lower methanol concentrations at the expense of energy efficiency, as they do not have PQQ enzymes available and evolved in mesophilic environments.
Komagataella phaffii contains two genes coding for alcohol oxidase. Based on the presence of intact AOX genes and the ability to utilize methanol, three methanol utilization (Mut) phenotypes of K. phaffii are defined. The wild type phenotype is Mut + , a aox1 strain is called Mut S and a aox1 aox2 strain is called Mutfor methanol utilization positive, slow and negative, respectively. The Mutphenotype is generally regarded as incapable of utilizing methanol (Cregg et al. 1989;Sreekrishna et al. 1989;Chiruvolu, Cregg and Meagher 1997). However, in a recent study we found evidence that the Mutphenotype facilitates a low but significant rate of methanol oxidation although no growth on methanol was supported (Zavec, Gasser and Mattanovich 2020). Similarly, Singh and Narang (2020) suggested that there might be some residual Aox activity or an Aoxindependent pathway in Mutstrains that leads to formaldehyde formation. Therefore, we set out to investigate whether there is any additional, Aox independent pathway for methanol oxidation in K. phaffii. Mutstrains have a residual specific methanol uptake rate (q MeOH ) of about 4 mg g -1 h -1 which is about 2% of a wild type Mut + and 10% of a Mut S strain (Zavec, Gasser and Mattanovich 2020). We hypothesized that native Adhs may elicit a side reaction on methanol and are responsible for the low but significant methanol uptake in Mutstrains.

Generation of ADH deletion and overexpression strains
To generate the adh2, adh900 and adh2 adh900 strains the previously described K. phaffii strain CBS2612 aox1 aox2 (Mut -) with or without human serum albumin (HSA) overexpression were used as parents (Zavec, Gasser and Mattanovich 2020). First the adh2 ::loxP-hphMX-loxP strain was created using a split marker cassette carrying a hygromycin resistance cassette already described (Nocon et al. 2014). For adh900 a new split marker cassette was designed carrying a geneticin resistance (Gasser et al. 2013). The three new strains are found in Table 1. The K. phaffii strains X-33 and X-33 Adh2KO were added as a comparison (Nocon et al. 2014). The deletion strains were selected on YPD with either 200 μg mL -1 hygromycin or 500 μg mL -1 gentamicin or a combination of both. The strains were verified by PCR amplification and sequencing of the PCR amplicons. The ADH overexpression strains were created by amplifying the ADH2 and ADH900 genes from CBS2612 Mutgenomic DNA. The fragments were assembled by overlap extension PCR to obtain the coding sequence without BsaI and BbsI restriction sites needed for Golden Gate cloning (Prielhofer et al. 2017). The created plasmids BB3aZ pGAP ADH2 cycTT and BB3aZ pGAP ADH900 cycTT were linearized with AscI (New England Biolabs) and transformed into the Mutstrain, selected on YPD with 25 μg mL -1 Zeocin creating Adh2OE and Adh900OE strains. Transformation was done by electroporation (Gasser et al. 2013).
The Mut -P AOX1 vHH strain carrying a single copy of the vHH expression construct from our previous study was transformed with BB3aK pAOX1 ADH2 cycTT and BB3aK pFLD1 ADH2 cycTT (linearized with AscI) and selected on YPD 25 μg mL -1 Zeocin and 500 μg mL -1 gentamicin creating the strains Mut -P AOX1 vHH P AOX1 Adh2 and Mut -P AOX1 vHH P FLD1 Adh2. As a comparison, CBS2612 Mut S was also transformed with the pPM2pZ30 pAOX1 αMFvHH CycTT vector carrying a codon optimized variable region of a camelid antibody (vHH) fused to a Saccharomyces cerevisiae α-mating-type secretion signal sequence (Zavec, Gasser and Mattanovich 2020). Selection was done on YPD with 25 μg mL -1 Zeocin. Prior to bioreactor cultivation the strains were screened, an average producer was selected and a working cell bank for bioreactor cultivations was prepared as described previously (Zavec, Gasser and Mattanovich 2020).
This study This study

Cell free extracts for alcohol dehydrogenase assays
The alcohol dehydrogenase activity in cell free extracts was assayed by washing 2 mL of a liquid overnight culture on YPD at 25 • C with 1 mL PBS and resuspending it in 500 μL cell lysis buffer with glass beads. The modified lysis buffer consisted of 20 mM HEPES, 420 mM NaCl, 1.5 mM MgCl 2 , 10% glycerol, 1 SIGMAFAST TM Protease Inhibitor Cocktail Tablet per 50 mL (Sigma-Aldrich GmbH) (Karaoglan, Karaoglan and Inan 2015). The cultures were lysed by bead beating (FastPrep-24, MP Biomedicals, Inc.) for 3 × 20 s at 6 m s -1 with 1-minute cooling on ice in-between steps. After the lysis step, the cultures were centrifuged, and the supernatant was transferred to a fresh microcentrifuge tube and centrifuged again at 13 200 rpm for 30 min at 4 • C to remove any carried over cell debris. After the second centrifugation step the supernatant was stored at −20 • C till use.

Alcohol dehydrogenase activity assay
Prior to activity measurement, the protein concentration of the cell free extracts was measured by Pierce TM BCA Protein Assay (Thermo Scientific, Inc.) and adjusted to a common concentration for all samples. Then 20 μL cell free extracts were added to the reaction buffer and equilibrated for 10-15 minutes before addition of 1 M of ethanol as a substrate. The total end volume was 300 μL. The absorbance measurements of NADH at 340 nm were done in a 96 well plate using a microplate reader (Tecan Group Ltd.). The reaction buffer consisted of 100mM MOPS; 5mM MgSO 4 ; 2mM NAD + at pH 8.9 modified from Ochsner et al. (2014) and the activity was calculated in mU mg -1 as described elsewhere (Müller et al. 2015).

Bioreactor cultivations
Bioreactor cultivation experiments were carried out in a DASGIP R Parallel Bioreactor System (Eppendorf AG). The cultivations consisted of (i) a batch phase, (ii) a feed phase and (iii) a methanol only phase where there was no other carbon source available except methanol. The batch medium used was BSM with 40 g L -1 glycerol as a carbon source followed by a 50% glucose feed (Mellitzer et al. 2014). Depending on the cultivation the pH was set to 5.0 or 5.5 by addition of 12.5% or 25% NH 4 OH and 10% phosphoric acid. The glucose feed rate was controlled gravimetrically by a custom balance controlled script. The glucose feed was run for 24 h at a feed rate of 2.9 g h -1 or 3.39 g h -1 depending on the cultivation to increase the biomass concentration before measuring the methanol metabolism associated parameters in the methanol feed phase. We applied a similar method for measuring the methanol uptake rate based on methanol pulses as already published (Dietzsch, Spadiut and Herwig 2011). Methanol was pulsed up to 1.5% (v/v) at the beginning of the glucose feed to induce and adapt the culture. After the glucose feed finished a second methanol pulse to 1.5% (v/v) was applied and the methanol concentration was measured by HPLC (Shimadzu, Corp.) at intervals to assess the q MeOH . Cell dry weight (CDW) was determined prior to the methanol shot as described before (Zavec, Gasser and Mattanovich 2020). The cultivation of the vHH expressing strains was performed with strategy B and strategy D for the Mut S comparison (Zavec, Gasser and Mattanovich 2020). Strategy B is divided into three phases: (i) batch, (ii) methanol-glucose co-feed, (iii) methanol feed phase. In the batch phase 300 mL BSM with 40 g L -1 glycerol was used, followed by a 50% (w/w) glucose feed at 5.8 g L -1 for 25 h. A methanol pulse was applied at the start of the methanolglucose co-feed phase. A 50% (v/v) methanol feed was started to keep the methanol concentration at a target of 1.0% to 1.5% (v/v) till the end of the cultivation. Strategy D consisted of four phases: (i) batch, (ii) glycerol feed phase, (iii) methanol-glucose co-feed and (iv) methanol feed phase. The growth limiting 100% methanol feed was increasing at a rate of f(x) = 0.028x + 0.6. Every cultivation was done in duplicates and the reported data is the average of the two repeats. Where indicated, OTR and heat of reaction were calculated as described in Zavec, Gasser and Mattanovich (2020).

Quantification of the recombinant protein
Quantification of the secreted recombinant protein in the culture supernatant was done by the LabChip GX/GXII System (PerkinElmer) using the consumables Protein Express Lab Chip (760499, PerkinElmer) and Protein Express Reagent Kit (CLS960008, PerkinElmer) according to the supplier´s instructions.

C-Methanol labeled bioreactor cultivation
The 13 C-methanol labeling experiments were carried out in the bioreactor in a similar manner. The (i) batch phase and (ii) glucose feed phase were carried out as described earlier. The glucose feed rate was increased to 5.8 g L -1 to achieve a biomass concentration of approximately 100 g L -1 CDW. The pH was controlled at 5.5 by 12.5% NH 4 OH and 10% phosphoric acid. Two hours prior to the end of the glucose feed phase, the gassing was changed to synthetic air without CO 2 (20% O 2 , 80% N 2 ). At the end of the glucose feed phase 150 mL of the reactor volume was removed so that approximate 325 mL was left in each reactor. At this point a 50% methanol pulse with 13 C isotope labeled methanol or 12 C methanol as the unlabeled control was added and a subsequent feed was started to keep the methanol concentration between 1.0% and 1.5% (v/v). An HPLC sample was taken right after the pulse and later approximately every 24 h. Biomass samples were used to determine CDW and the 13 C/ 12 C biomass isotope ratio, and the supernatant was analyzed for the 13 C content in the secreted recombinant proteins. A CO 2 trap consisting of 1M NaOH was used to capture the reactor exhaust gas over a period of 24 h.
Metabolite sampling was done one hour after the methanol pulse and then approximately every 24 hours. Quenching was done as described before (Mattanovich et al. 2017). Briefly, the sampling port was flushed and immediately a 2 mL sample was taken and quenched in 8 mL quenching solution (60% methanol, 125 mM TRIS-HCl, 55 mM NaCl, pH 8.2; T −27 • C). Then 500 μL of the quenched mixture was filtered through a 0.45 μm cellulose acetate filter (Sartorius Stedim Biotech GmbH) and washed with 10 mL quenching solution at −27 • C. The filter was immediately stored at −70 • C until metabolite extraction.

Intracellular metabolite sample preparation and 13 C labeling measurements
For the 13 C/ 12 C metabolite ratio analysis the quenched samples were extracted with boiling ethanol (Neubauer et al. 2012;Rußmayer et al. 2015). 4 mL of 75% ethanol at 85 • C was added to quenched and frozen biomass samples. The samples were vortexed for 20 s and transferred to a water bath at 85 • C for 3 minutes. The sample were vortexed again after 1.5 minutes and at the end of the incubation period for 10 s each followed by rapid cooling on dry ice, avoiding freezing of the ethanol solution. The cooled samples were centrifuged for 10 minutes at 4000 g and −20 • C and decanted. The ethanolic supernatant was vacuum dried and stored until use at −70 • C.
The 13 C labeling patterns of free intracellular metabolites were analyzed via gas chromatography chemical ionizationtime of flight mass spectrometry (GC-CI-TOFMS) according to Mairinger et al. and Chu et al. with minor modifications (Chu et al. 2015;Mairinger et al. 2015). Measurements were carried out with an Agilent 7890B gas chromatograph combined with an Agilent 7200B QTOFMS system showing a mass accuracy of < 5 ppm. Prior to analysis, a two-step derivatization based on ethoximation and subsequent silylation was performed online on a GERSTEL DualRail MultiPurposeSampler (MPS2, GERSTEL, Germany). Isotope interference correction for the contribution of heavy isotopes from the derivatization agents and the native metabolite itself was performed using the software Isotope correction toolbox (ICT) developed by Jungreuthmayer et al. (2015). M/z ratios and mass extraction windows of the fragments or adducts used for data evaluation were first chosen as described in Mairinger et al. (2015), but needed to be adapted for some metabolites due to matrix interferences or saturation effects (

Protein identification and peptide profiling by liquid chromatography-electrospray ionization-mass spectrometry
The supernatants containing a total amount of 30 μg protein each were S-alkylated with iodoacetamide and further digested with Sequencing Grade Modified Trypsin (Promega Corp.). An aliquot of 5 μg of the peptide mixture was analyzed using a Dionex Ultimate 3000 system directly linked to a Q-TOF instrument (maXis 4G ETD, Bruker GmbH) equipped with the standard ESI source in the positive ion, data dependent acquisition mode, DDA mode ( = switching to MSMS mode for eluting peaks). MS scans were recorded (range: 150-2200 m/z, spectra rate: 1 Hz) and the six highest peaks were selected for fragmentation (CID mode). Instrument calibration was performed using ESI calibration mixture (Agilent Inc.). For separation of the peptides a Thermo BioBasic C18 separation column (5 μm particle size, 150 × 0.320 mm) was used. A gradient from 97% solvent A and 3% solvent B to 62.5% solvent B in 45 min was applied, followed by a 15 min gradient from 62.5% solvent B to 95% solvent B at a flow rate of 6 μL/min at 32 • C. Solvent A: 65 mM ammonium formate buffer, pH 3.0; Solvent B: 80% Acetonitril (VWR LLC; BDH Prolabo) and 20% solvent A. DataAnalysis 4.0 (Bruker GmbH) was used for peptide evaluation.
For each of the analyzed peptides the 13 C/ 12 C ratio was calculated using line spectra intensities of one specific charge state after normalization according to the number of carbon atoms present (cysteine carbamidomethylation considered). Error values for the presence of other heavier isotopes such as for nitrogen, oxygen, hydrogen and sulfur were calculated from the theoretical isotopic patterns (IsotopePattern; Bruker GmbH) and used for the correction of the measured and normalized ratio of the monoisotopic mass to the other isotopomers of each of the peptides. For each sample, an average 13 C % value was calculated considering all of the six analyzed peptides (Table S1, Supporting Information).

Biomass and CO 2 isotope ratio by elemental analysis isotope ratio mass spectrometry
The biomass samples were kept frozen at −70 • C and were washed twice with PBS to remove any residual 13 C methanol before analysis. The captured CO 2 in the form of sodium carbonate/bicarbonate was precipitated with ethanol. About 200 mL of the NaOH capture solution was mixed with 800 mL of absolute ethanol and cooled on ice until the sodium carbonate/bicarbonate precipitated. The precipitate was filtered, and vacuum dried to remove any residual ethanol from the precipitation. The biomass and sodium carbonate/bicarbonate isotope ratios were determined with elemental analysis isotope ratio mass spectrometry (EA-IRMS) performed by Imprint analytics GmbH, Austria (Gassler et al. 2020).

Adh2 is the major ethanol dehydrogenase of K. phaffii
Alcohol dehydrogenases are notoriously promiscuous enzymes that generally show low specificity towards a specific alcohol (Verduyn et al. 1988;Sealy-Lewis and Fairhurst 1995;Krog et al. 2013). Komagataella phaffii encodes six alcohol dehydrogenases (Valli et al. 2016), among which ADH2 and ADH900 are the two most highly transcribed genes (Ata et al. 2018). Activity of cell free extracts against ethanol was used to confirm the successful deletion of the active alcohol dehydrogenases in our test strains, as the activity against methanol in cell free extracts was too low to directly measure the effect of ADH2 and ADH900 deletion. When ADH2 was deleted, a substantial reduction of ethanol dehydrogenase activity by 93% was observed, further deletion of ADH900 resulted in almost complete loss of activity ( Table 2). The Mut -AdhKO double deletion strain still had a residual activity of 8.0 mU mg -1 but this represented only 0.6% of the initial Mutactivity. Thus, ADH900 only represents a marginal activity compared to ADH2. Taken together, this confirms that irrespective of the Mut phenotype, ADH2 is mainly responsible for the ethanol dehydrogenase activity.

Native alcohol dehydrogenases cause methanol uptake in Mutstrains
The Mut -AdhKO strain (Mut -adh2 adh900) that had nearly no Adh activity towards ethanol was tested for its specific methanol uptake (q MeOH ) in the bioreactor. Additionally, the Mut -Adh2KO (Mut -adh2) and Mut -Adh900KO (Mut -adh900) single deletion strains and a strain overexpressing human serum albumin (Mut -P AOX1 HSA) in which we first observed methanol depletion (Zavec, Gasser and Mattanovich 2020) were used as comparison. Biomass was grown in glucose batch and fed batch cultures and induced with methanol during the limited glucose feed. Then a second methanol pulse of 1.5% (v/v) was applied and q MeOH was measured by following the MeOH concentration by HPLC over time. Marked differences in dissolved oxygen and CO 2 exhaust-gas concentrations were observed between the different strains indicating different degrees of methanol oxidation (Fig. 1, Table 3). By deletion of both ADH2 and ADH900 in the Mutbackground strain (AdhKO) q MeOH was reduced to 0.7 mg g -1 h -1 . According to the sterile bioreactor control published earlier (Zavec, Gasser and Mattanovich 2020) this value can be entirely explained by evaporation of methanol from the reactor medium by aeration and agitation. In contrast, the Mut -P AOX1 HSA had a measured q MeOH of 5.1 mg g -1 h -1 which is consistent with previous observations (Zavec, Gasser and Mattanovich 2020). The single ADH deletion strains show that ADH900 had no measurable effect on q MeOH , which is consistent with the ethanol dehydrogenase data where the ADH900 deletion only had a marginal effect.
Overexpression of ADH2, but not of ADH900, increases specific methanol uptake rate Two strains overexpressing ADH2 and ADH900 with the constitutive glyceraldehyde 3-phosphate dehydrogenase (GAP) promoter (Adh2OE and Adh900OE) were generated and cultivated in the bioreactor as described above to determine the effect of the individual Adh overexpression on q MeOH . After the glucose feed phase the CDW reached an average concentration of 75.1 g L -1 (Adh2OE) and 74.9 g L -1 (Adh900OE). Then methanol was pulsed batch-wise to 10 g L -1 and q MeOH was measured after 4.1 and 20.1 h (Table 4). After 4.1 h the methanol concentration was still in the range of 8g L -1 and thus not limiting. Therefore, we consider these data more reliable than q MeOH calculated over 20.1 h where the methanol concentration was already in a sub saturation range below 0.4% (v/v). ADH2 overexpression clearly shows an increase of q MeOH over the wild type and of the ADH900 overexpression strain. The ADH900 overexpression strain has a q MeOH which is approximately at the same level as in the Mutstrain and the Adh900KO.

While most methanol is oxidized to CO 2 in Mutstrains, some is assimilated to primary metabolites and heterologous protein
To determine the fate of methanol in the K. phaffii Mutphenotype and to confirm the hypothesis that Adh2 and potentially Adh900 are responsible for methanol utilization, an experiment with 13 C isotope labeled methanol was carried out. For this purpose, we used the already described Mut -P AOX1 HSA (Zavec, Gasser and Mattanovich 2020) as it provided the additional possibility to look at the labeling of a secreted recombinant protein.
The second strain was Mut -AdhKO, which was not capable of metabolizing methanol according to the previous experiments and served as a biological control. Each strain was cultivated in two repeats with 13 C methanol and two parallels with 12 C methanol as unlabeled control.
The CDW at the end of the glucose feed phase before the addition of methanol was similar for all 8 reactors at 99.3 ± 2 g L -1 . After the glucose feed phase, either 12 C or 13 C methanol was added to the cultivation media, and kept at a concentration between 0.8% and 1.8% (v/v). In the methanol feed phase, the Mut -AdhKO strain showed near zero CO 2 in the exhaust gas while dissolved oxygen in the culture rose to nearly 100%, indicating that no substrate was oxidized in this phase. The Mut -P AOX1 HSA strain, in contrast maintained an exhaust CO 2 level of 0.8% and a lower dissolved oxygen level (Fig. 1). The Mut -P AOX1 HSA strain led to lower DO and exhaust O 2 concentrations and higher exhaust CO 2 concentrations compared to the Mut -AdhKO strain. Elemental analysis of the isotope ratios showed that the exhaust CO 2 was highly enriched up to 79% with 13 C isotope in the Mut -P AOX1 HSA strain with wild-type ADH2 expression ( Fig. 2A). In the Mut -AdhKO, CO 2 emission was substantially reduced and had lower 13 C enrichment of 4.6%. The exhaust CO 2 observed in the bioreactor cultivations of the Mutstrain is therefore derived from oxidation of methanol by K. phaf-fii´s native Adh2. The methanol dissimilation ratio (q CO2/ q MeOH ) shows the relative amount of carbon flux through the dissimilatory pathway used for NADH and subsequently for ATP generation. Interestingly, the dissimilation ratio of the Mut -P AOX1 HSA strain is higher compared to the Mut S strain, as determined in a later experiment (Table 5).
The Mut -P AOX1 HSA biomass was slightly but significantly enriched with 13 C compared to the unlabeled control (Fig. 2B). Table 2. Alcohol dehydrogenase activity (mU mg -1 ) against ethanol in cell free extracts of K. phaffii with standard errors and sample size in parenthesis.

Strain name
Gene deletions Adh overexpression q MeOH (mg g -1 h -1 ) Mut -P AOX1 HSA aox1 aox2 none 5.1 Mut -AdhKO aox1 aox2 adh2 ::HphR adh900 ::KanMX none 0.7 Mut -Adh900KO aox1 aox2 adh900 ::KanMX none 5.5 Mut -Adh2KO aox1 aox2 adh2 ::HphR none 0.9 Mut -Adh2OE aox1 aox2 P GAP ADH2 7.7 Mut -Adh900OE aox1 aox2 P GAP ADH900 5.6 Mut -P FLD1 Adh2 aox1 aox2 P FLD1 ADH2 9.2 Mut -P AOX1 Adh2 aox1 aox2 P AOX1 ADH2 12.7 The Mut -AdhKO strain showed no enrichment at all. This shows that the Mutphenotype is still capable of assimilating methanol although it cannot grow due to the lack of an alcohol oxidase, while the Mut -AdhKO strains lose this ability completely. This was further confirmed by the increasing enrichment of 13 C in the secreted protein which reached up to 3.5% at the end of the cultivation ( Fig. 2C; Table S1, Supporting Information). Over the course of the cultivation, three samples for metabolite analysis were taken at 1, 25, 52 h after the first methanol addition. The 13 C enrichment in metabolite pools increased consecutively and was highest at the last sampling point. In particular, some key metabolites of the assimilatory methanol metabolism were highly 13 C labeled as for instance sedoheptulose-7P (25%), fructose-6P (25%) and ribose-5P (11%). The lower glycolysis metabolites 3-phosphoglycerate (20%), 2-phosphoglycerate (19%) and the TCA cycle metabolite malate (16%) were also highly labeled ( Table 6). As the lower glycolysis and TCA cycle are the source of amino acid precursors, this explains the 13 C enrichment of the secreted recombinant protein.

Overexpression of ADH2 increases recombinant protein production
As shown in our previous study, the Mutstrain is able to produce recombinant proteins when cultivated on methanol alone. This was quite unexpected as it was assumed that Mutstrains cannot utilize methanol alone, so that they are typically cultivated with a co-substrate. The data presented here make it obvious, however, that methanol oxidation by K. phaffii´s native Adh2 enzyme results in a low but steady energy generation. To test whether alcohol dehydrogenase really impacts protein production, we transformed a Mutstrain producing a camelide antibody fragment (Mut -P AOX1 vHH) with two ADH2 overexpression constructs. ADH2 overexpression was under control of two methanol responsive promoters with different expression strength, namely the P AOX1 and P FLD1 . The resulting strains Mut -P AOX1 vHH P AOX1 ADH2 and Mut -P AOX1 vHH P FLD1 ADH2 were tested in a bioreactor cultivation with strategy B as previously described (Zavec, Gasser and Mattanovich 2020). In addition, a Mut S strain expressing VHH was cultivated as reference.  Error bars denote the standard deviation. 13 CH3OH fed cultures are represented in orange, and 12 CH3OH fed controls in blue. 13 C enrichment is seen in all samples of the Mutstrain while samples of the AdhKO strain do not enrich any 13 C above the native content, as in the 12 C control samples. Table 5. Comparison of the specific methanol and oxygen uptake rates (q O2 ) and specific CO 2 evolution rate (q CO2 ), respiratory quotient (RQ) and methanol dissimilation ratio.  As can be seen in fig. 3, overexpression of ADH2 increases the productivity of the Mutproduction strain significantly. Compared to the parent strain Mut -P AOX1 vHH (Zavec, Gasser and Mattanovich 2020), the specific productivity (q P ) in phase 3 is increased by 2.6 fold and approaches the q P of the Mut S strain.
The Mut S strain in phase 4 has an average q P of 205 μg g -1 h -1 .
The ADH2 overexpressing strains show a slightly higher q P at 234 and 237 μg g -1 h -1 for the P AOX1 Adh2 and P FLD1 Adh2 overexpression, respectively. Notably, ADH2 overexpression increased q MeOH further, to a maximum of 12.7 mg g -1 h -1 (Table 3), and it increased methanol dissimilation even further compared to Mut -P AOX1 HSA towards the 100% mark (Table 5). In terms of productivity these strains are on par with the industry standard Mut S strain for methanol induced recombinant protein production, however, at much lower oxygen demand and heat output (Table 7).

AOX deficient K. phaffii can still utilize methanol
Methanol oxidation in methylotrophic yeasts is accomplished by alcohol oxidases (Aox) (Couderc and Baratti 1980;Cregg et al. 1989). Our previous study with K. phaffii lacking both AOX genes, however, revealed that a low level of methanol depletion of an unknown cause still occurred in these Mutstrains even though there was no observable growth (Zavec, Gasser and Mattanovich 2020). One possible candidate for methanol oxidation in these strains are alcohol dehydrogenases, which are promiscuous enzymes with rather low substrate specificity and are generally more active towards higher alcohols (Verduyn et al. 1988;Sealy-Lewis and Fairhurst 1995;Krog et al. 2013;Zhang et al. 2017). We used this feature to detect and determine the presence of active alcohol dehydrogenases in cell free extracts of Adh deleted K. phaffii strains. In accordance with the data published by Karaoglan et al. (2020), deletion of both ADH2 and ADH900 completely abolished the dehydrogenase activity towards ethanol. Strikingly, when the strains lacking ethanol dehydrogenase activity were tested for q MeOH in bioreactors, methanol depletion was abolished to a level where it cannot be discriminated from evaporation anymore. On the other hand, deletion of ADH900 in the Mutor the Mut -adh2 background strain did not show any additional reduction of q MeOH . The effect of adh2 and adh900 on methanol was reflecting the dehydrogenase activity measurements, where we observed that ADH900 accounts for only 6% of the total activity towards ethanol although expression levels of both Adh genes are similar and high on methanol (Prielhofer et al. 2015). In the same sense ADH900 overexpression did not produce any distinct increase in q MeOH , suggesting that Adh900 is indeed much less active towards methanol (and ethanol) than Adh2.
The observed effects of both deletion and overexpression of ADH2 make it obvious that this is the gene responsible for the observed methanol depletion. Thus K. phaffii can in fact utilize methanol by oxidation with the Adh2 enzyme. q MeOH of the ADH2 overexpression strains responds to the strength of the promoter used. From P GAP , to P FLD and P AOX1 the q MeOH increased by 1.5-, 1.8, and 2.5-fold compared to the control (Table  3). This represents 5% of the methanol uptake rate reported for Mut + and 30% of that of Mut S strains. The reported values range as high as 240-250 mg g -1 h -1 for Mut + (Barrigon, Valero and Montesinos 2015;Tomàs-Gamisans, Ferrer and Albiol 2018) and up to 62 mg g -1 h -1 for Mut S strains (Dietzsch, Spadiut and Herwig 2011). So even though Adh2 was overexpressed with the same promoter as Aox1 this leads only to a minor increase of q MeOH compared to the wild type and the Mut S . In terms of q MeOH Adh2 cannot substitute quantitatively for Aox1 and Aox2.

Alcohol dehydrogenase may have had an auxiliary role in the evolution of yeast methylotrophy
Methanol metabolism is divided into the assimilatory and the dissimilatory pathways. In K. phaffii the assimilatory pathway is localized in the peroxisome (Rußmayer et al. 2015) and the dissimilatory in the cytosol (Yurimoto, Kato and Sakai 2005;van der Klei et al. 2006;Vanz et al. 2012;Rußmayer et al. 2015). The 13 C labeling experiments confirmed that the CO 2 formation observed with the Mutstrains is indeed sourced from methanol. Furthermore, metabolite labeling reveals that at least a small amount of carbon is assimilated into the biomass and can be found both in the recombinant protein and the biomass.
Hypothetically, using an alcohol dehydrogenase instead of an oxidase could enable yeasts to utilize methanol in a more efficient way. The additional NADH yield from methanol oxidation by an Adh would increase the ATP yield per methanol unit (Sheehan et al. 1988) and would therefore decrease the needed flux through the dissimilatory pathway. This would change the balance between the assimilatory and dissimilatory flux toward assimilation and increase the biomass yield (Y X/S ). Why did the pathway evolve in such a suboptimal way that came with three disadvantages, despite alcohol dehydrogenases being ubiquitous in yeasts? It was energetically unfavorable, produced H 2 O 2 as a byproduct and necessitated the localization of the pathway to the peroxisomes by the duplication of numerous genes from the pentose phosphate pathway. The answer might be in the notoriously low activity of Adhs towards methanol (Krog et al. 2013;Ochsner et al. 2014) and the unfavorable Gibbs free energy of the Adh reaction at mesophilic temperatures compared to either Aox or the PQQ dependent Adh reaction (Whitaker et al. 2015). Additionally, microbes seem to be selected towards faster growth rates rather than Y X/S so that an increased rate of the reaction even at low methanol concentrations may have an advantage over optimal ATP generation (Anthony 1982). Evidence from the bacterial domain suggest that temperature might be the decisive factor here. Bacteria lack peroxisomes to shield them from the harmful H 2 O 2 , so evolution of an intracellular alcohol oxidase was restricted. Instead, two Adhs evolved in bacterial methylotrophs, the PQQ dependent Adh and the NAD + dependent Adh. Reduced PQQ yields less ATP upon oxidation than NADH and therefore the PQQ dependent Adh is less efficient than the NAD + dependent counterpart. However, the Gibbs free energy change upon PQQ reduction is larger (Anthony 1982;Sheehan et al. 1988;Whitaker et al. 2015). So far, only thermophilic Gram-positive methylotrophs have been found to use NAD + dependent Mdh (Nazina et al. 1988;Arfman et al. 1989;Krog et al. 2013;Visser et al. 2013). This might be due to thermodynamic constraints of this strategy at low temperatures and methanol concentrations, and improved kinetics at higher temperatures. The mesophilic lifestyle of yeast might have directed evolution towards an Aox based reaction. This perspective suggests that the development of mesophilic synthetic methylotrophs based on methanol dehydrogenases is inherently problematic, which is reflected by the difficulty to engineer a strain capable of growth on methanol as a single carbon source (Müller et al. 2015;Whitaker et al. 2017;Bennett et al. 2018Bennett et al. , 2020Meyer et al. 2018;Chen et al. 2020). In our case even a natural methylotroph already adapted to methanol utilization was unable to grow with an Adh as the initial step of methanol oxidation. Very recently, Espinosa et al. (2020) observed minor growth of S. cerevisiae on methanol when co-fed with yeast extract, and referred this to its native Adh activity.
The Adh2 present in K. phaffii is active towards methanol and in the Mutbackground strain it facilitates a q MeOH that is beyond the reported non-growth associated maintenance energy (NGAME) of 2.81 mg g -1 h -1 (Tomàs-Gamisans, Ferrer and Albiol 2018). This could have served as the source for evolution towards methylotrophy, providing initially an additional energy source via methanol dissimilation. In the Mutstrain more than 80% of methanol is oxidized to CO 2 (a ratio that even increases to 100% when ADH2 is overexpressed). In the presence of alcohol oxidase (Mut + strain), only 50-80% of methanol is dissimilated to CO 2 (Jordà et al. 2012(Jordà et al. , 2013Vanz et al. 2012;Tomàs-Gamisans, Ferrer and Albiol 2018). According to the published model of yeast methanol metabolism, formaldehyde needs to diffuse into the cytosol to be dissimilated after being formed in the peroxisome. In some perspective the dissimilated formaldehyde is an overflow of the peroxisomal assimilation pathway (Douma et al. 1985;van der Klei et al. 2006). Our data suggest that the colocalization of methanol oxidation with the assimilatory pathway has been a key driver to evolve growth on methanol. While Adh2 is a cytosolic enzyme (Karaoglan, Karaoglan and Inan 2015;Valli et al. 2020) both Aox1 and Aox2 are peroxisomal matrix proteins. Contrary, in the Mutstrain formaldehyde is formed in the cytosol where its oxidation is located which may explain the predominance of dissimilation in these strains (Fig. 4).

ADH2 overexpression is a useful tool to enhance recombinant protein production in Mutstrains
Although protein synthesis and production is an energy demanding process microbes evolved to prioritize it even under severe energy restrictions and starvation. This enables them to respond to changes in the environment by replacing enzymes and remodeling the metabolism to gain, for example, access to alternative energy and carbon sources (Jewett et al. 2009). This may explain why the Mutstrain is capable of producing recombinant proteins while having a severely restricted q MeOH and why the rather modest increase in q MeOH by the ADH2 overexpression has such a positive impact on both recombinant protein titers and productivity. Notably, even with ADH2 overexpression, the Mutstrains are still unable to grow on methanol alone.
ADH2 overexpression increases productivity of the Mutstrains to the level of the Mut S strain, an industry standard, but the q P profile is different. While q P steadily increases for the Mut S strain during the cultivation and peaks at the end, the Mut -ADH2 overexpressing strains start at a much higher q P , than peak around the midpoint and decrease towards the end, forming a bell shape. Thus, using the Mut -ADH2 overexpressing strains the final titer is already reached earlier in the cultivation, shortly after 92 h. In conclusion the Mut -ADH2 overexpression strain can produce recombinant proteins at the same level as Mut S while still retaining the benefits of low oxygen uptake and heat output.

CONCLUSION
The combined evidence gathered here has made us rethink the long-lasting concept that the Mutstrains of K. phaffii (and other methylotrophic yeasts) lacking AOX1 and AOX2 cannot oxidize methanol and that methanol loss is due to evaporation (Cregg et al. 1989;Looser et al. 2015). We showed that methanol metabolism in these strains is active, relying on the promiscuous activity of the Adh2 enzyme. Carbon from methanol gets incorporated into metabolites, biomass and recombinant protein. Overexpression of ADH2 has a significant positive effect on q MeOH compared to the Mutstrain and q MeOH is well above the reported NGAME for methanol, yet biomass growth cannot be observed. The wild type K. phaffii strains exhibit multiple times higher q MeOH , suggesting that evolution of a peroxisomal alcohol oxidase was necessary for yeast cells to compete for resources and achieve a competitive growth rate. Finally, we highlighted the potential application of ADH2 overexpression for recombinant protein production in an industrial scenario.

SUPPLEMENTARY DATA
Supplementary data are available at FEMSYR online.

ACKNOWLEDGMENTS
EQ BOKU VIBT GmbH is acknowledged for providing fermentation and mass spectrometry equipment. We kindly acknowledge the technical support by Philipp Tondl and Viktoria Kowarz. We are thankful for the scientific and technical advice from Hannes Russmayer and Corinna Rebnegger.