Respiratory reoxidation of NADH is a key contributor to high oxygen requirements of oxygen-limited cultures of Ogataea parapolymorpha

Abstract While thermotolerance is an attractive trait for yeasts used in industrial ethanol production, oxygen requirements of known thermotolerant species are incompatible with process requirements. Analysis of oxygen-sufficient and oxygen-limited chemostat cultures of the facultatively fermentative, thermotolerant species Ogataea parapolymorpha showed its minimum oxygen requirements to be an order of magnitude larger than those reported for the thermotolerant yeast Kluyveromyces marxianus. High oxygen requirements of O. parapolymorpha coincided with a near absence of glycerol, a key NADH/NAD+ redox-cofactor-balancing product in many other yeasts, in oxygen-limited cultures. Genome analysis indicated absence of orthologs of the Saccharomyces cerevisiae glycerol-3-phosphate-phosphatase genes GPP1 and GPP2. Co-feeding of acetoin, whose conversion to 2,3-butanediol enables reoxidation of cytosolic NADH, supported a 2.5-fold increase of the biomass concentration in oxygen-limited cultures. An O. parapolymorpha strain in which key genes involved in mitochondrial reoxidation of NADH were inactivated did produce glycerol, but transcriptome analysis did not reveal a clear candidate for a responsible phosphatase. Expression of S. cerevisiae GPD2, which encodes NAD+-dependent glycerol-3-phosphate dehydrogenase, and GPP1 supported increased glycerol production by oxygen-limited chemostat cultures of O. parapolymorpha. These results identify dependence on respiration for NADH reoxidation as a key contributor to unexpectedly high oxygen requirements of O. parapolymorpha.


Introduction
Microbial biotechnology offers promising options for replacing petrochemically produced chemicals with sustainable bio-based alternatives (Weusthuis et al. 2011, Thorwall et al. 2020. Microbial production of ethanol as a transport fuel, based on plant carbohydrates as renewable feedstocks, is already applied on a large scale. The 87 Mtonnes of ethanol produced worldwide in 2020 (Annual World Fuel Ethanol Production 2020) were almost exclusively produced with the yeast Saccharomyces cerevisiae (Jansen et al. 2017, Favaro et al. 2019. In well-established 'first-generation' ethanol processes, this yeast ferments sugars, predominantly derived from corn starch or sugar cane, to ethanol with high productivity, titers, and yields (Basso et al. 2011). Use of genetically engineered pentose-fermenting S. cerevisiae strains for conversion of lignocellulosic hydrolysates, generated from agricultural residues such as corn stover and sugar cane bagasse, is currently being explored at industrial scale (Jansen et al. 2017).
Economic viability of yeast-based ethanol production requires low processing costs and near-theoretical product yields on carbohydrate feedstocks, which are only possible in the absence of respiration (Jansen et al. 2017, Favaro et al. 2019. Industrial ethanol fermentation is performed in large tanks that readily become and remain anoxic due to vigorous carbon-dioxide production by fermenting yeast cells. The popularity of S. cerevisiae for application in these processes is related to its high fermentation rates, innate ethanol tolerance, tolerance to low pH, ability to grow and ferment in the absence of oxygen, and amenability to modern genomeediting techniques (Thomas and Ingledew 1992, Della-Bianca et al. 2013, Lopes et al. 2016.
Saccharomyces cerevisiae grows optimally at approximately 35 • C (Laman Trip and Youk 2020). Fermentation at higher temperatures is industrially attractive as it could reduce cooling costs and, potentially, enable a higher productivity. Additional benefits of thermotolerance may be gained by integrating enzymecatalyzed polysaccharide hydrolysis and fermentation of the released mono-and di-saccharides in a single unit operation (simultaneous saccharification and fermentation, SSF; Althuri et al. 2018). In addition to simplifying industrial processing, SSF could prevent inhibition of hydrolytic enzymes by released monosaccharides (Costa et al. 2014, Althuri et al. 2018. Moreover, use of thermotolerant yeasts in high-temperature SSF processes can enable a reduction of the required dose of fungal hydrolases, thereby further improving process economy. This advantage is especially relevant for second-generation bioethanol production, in which physically and/or chemically pretreated lignocellulosic plant biomass is hydrolyzed to monomeric sugars by hydrolases with high temperature optima (typically 50-80 • C; Alvira et al. 2010, Shirkavand et al. 2016.
Since supra-optimal temperatures potentially affect all proteins in a cell (Cruz et al. 2012), substantial extension of the temperature range of S. cerevisiae by metabolic engineering may prove to be an elusive target. Indeed, elegant adaptive laboratory evolution and metabolic engineering studies aimed at improving thermotolerance of S. cerevisiae, e.g. by engineering its sterol composition, enabled only modest improvements of its maximum growth temperature (Caspeta et al. 2014, Caspeta and Nielsen 2015, Li et al. 2021. Use of naturally thermotolerant, facultatively fermentative yeasts such as Ogataea sp. (Hansenula sp.; Kurtzman 2011) and Kluyveromyces marxianus, with temperature maxima of up to 50 • C (Hong et al. 2007, Kurylenko et al. 2014, appears to offer an attractive alternative. Like the large majority of yeast species, these yeasts readily ferment sugars under oxygen-limited conditions (Visser et al. 1990, Merico et al. 2007). However, like many other Saccharomycotina yeasts whose evolutionary history did not involve the whole-genome duplication (WGD) event that shaped the genomes of S. cerevisiae and closely related species (Wolfe and Shields 1997), Ogataea parapolymorpha and K. marxianus, cannot grow in the complete absence of oxygen (Visser et al. 1990, Blomqvist et al. 2012.
Fast anaerobic growth of S. cerevisiae requires supplementation of anaerobic growth media with sources of sterols and unsaturated fatty acids (UFAs; Stier 1953, 1954). Omission of UFAs from growth media leads to a drastically reduced growth rate (Dekker et al. 2019), which reflects involvement of the oxygen-consuming cytochrome-b5 9-desaturase Ole1 (Stukey et al. 1989) in UFA synthesis. Similarly, a strict sterol requirement of anaerobic S. cerevisiae cultures reflects involvement of multiple mono-oxygenases in sterol biosynthesis (Henneberry and Sturley 2005). Because of these biosynthetic oxygen requirements, media for anaerobic cultivation of S. cerevisiae are routinely supplemented with ergosterol and Tween 80, an oleate ester that serves as UFA source (Andreasen and Stier 1954). Additional oxygen requirements of S. cerevisiae for synthesis of biotin, nicotinate, pantothenate, and thiamine (Wightman and Meacock 2003, Perli et al. 2020, Wronska et al. 2021 generally go unnoticed in laboratory studies due to their routine inclusion in synthetic media (Perli et al. 2020). In most non-Saccharomyces yeasts, pyrimidine synthesis imposes an additional biosynthetic oxygen requirement, as it depends on a mitochondrial, respiration-coupled Class-II dihydroorotate dehydrogenase (DHODase, Ura9). In contrast, S. cerevisiae only harbors a cytosolic, respiration-independent Class-I-A DHODase (Ura1), which uses fumarate as electron acceptor (Nagy et al. 1992, Wolfe and Shields 1997, Langkjaer et al. 2003, Riley et al. 2016. Kluyveromyces sp. contain both Ura1 and Ura9 orthologs, whose expression is regulated in response to oxygen availability . Development of metabolic engineering strategies for eliminating oxygen requirements of non-Saccharomyces yeasts requires elucidation of underlying oxygen-and/or respiration-dependent biochemical reactions. To investigate oxygen requirements of the thermotolerant yeast O. parapolymorpha, its physiological and transcriptional responses to oxygen limitation were studied in chemostat cultures and compared to recent literature data on K. marx-ianus and S. cerevisiae . Based on the results of these experiments, co-feeding of acetoin was used to explore the impact of cytosolic NADH oxidation on the physiology of O. parapolymorpha in oxygen-limited cultures. In additional the genome of O. parapolymorpha was searched for orthologs of genes implicated in the (in)ability of other yeasts to grow anaerobically and, in particular, in the production of glycerol as 'redox sink' for reoxidation of NADH formed in biosynthetic reactions. Glycerol metabolism in O. parapolymorpha was further investigated in a mutant strain in which key genes involved in mitochondrial, respiration-linked NADH oxidation were deleted. Metabolic engineering of redox metabolism in O. parapolymorpha was explored by expressing S. cerevisiae genes involved in glycerol production.

Strain maintenance
Ogataea parapolymorpha CBS11895 (DL-1) was obtained from the Westerdijk Fungal Biodiversity Institute (Utrecht, The Netherlands). For propagation and maintenance, cultures were grown on yeast extract-peptone-dextrose (YPD) medium (10 g l -1 Bacto yeast extract, 20 g l -1 Bacto peptone, and 7.5 g l -1 glucose) in an Innova shaker incubator (New Brunswick Scientific, Edison, NJ) set at 30 • C and 200 rpm. YPD was prepared by autoclaving (20 min at 121 • C) a solution of yeast extract and peptone and then aseptically adding a separately autoclaved (20 min at 110 • C) concentrated glucose solution. Fro−en stock cultures, prepared from exponentially growing cultures by addition of glycerol to a final concentration of 30% (v/v), were aseptically stored at −80 • C.

Molecular biology techniques
PCR amplification for cloning was performed with Phusion High Fidelity polymerase (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer's instructions. DreamTaq polymerase (Thermo Fisher Scientific) was used for diagnostic PCR with yeast genomic DNA, isolated with the LiAc/SDS method (Lõoke et al. 2011) from overnight cultures on YPD, as template. Desalted or PAGE-purified oligonucleotide primers (Sigma-Aldrich, St. Louis, MO) are listed in Table S1 (Supporting Information). PCR-amplified DNA fragments were analyzed by gel electrophoresis and, when required, purified from agarose gels with a Zymoclean Gel DNA Recovery Kit (Zymo Research, Irvine, CA). Prior to purification, template plasmid DNA was removed by FastDigest DpnI digestion (Thermo Fisher Scientific). Alternatively, DNA fragments were purified with a GenElute PCR Clean-Up Kit (Sigma-Aldrich). Gibson assembly with the NEBuilder HiFi DNA Assembly Master mix (New England Biolabs, Ipswich, MA), was performed with a down-scaled reaction volume of 5 μl and a total incubation time at 50 • C of 1 h. The GenElute Plasmid Miniprep kit (Sigma-Aldrich) was used for plasmid isolation from overnight cultures of Escherichia coli XL1-Blue, which was used for plasmid amplification and storage.

Plasmid construction
Plasmids used in this study are described in Table 1. Promoter and terminator sequences of OpPMA1 and OpTEF1, which encode plasma-membrane H + -ATPase and translation elongation factor EF-1α, respectively, were chosen based on high transcript levels across a range of specific growth rates (Juergens et al. 2020). Promoter and terminator fragments were defined as regions 800 bp upstream and 300 bp downstream, respectively, of coding sequences. For targeted integration into a genetic locus, 500 bp Juergens et al.

Strain construction
Yeast strains used in this study are described in Table 2. Ogataea parapolymorpha strains were transformed by electroporation of freshly prepared electrocompetent cells (Juergens et al. 2018b). Transformants were selected on YPD agar containing hygromycin B (300 μg ml -1 ) or nourseothricin (100 μg ml -1 ). Strains IMX2119, IMX2587, and IMX2588 were constructed with the split-marker integration approach (Fairhead et al. 1996), with approximately 480-bp overlapping homology sequences for marker recombination and genome integration. The natNT2 split-marker fragments for integration of an ScGPP1 expression cassette into the OpGBU1 locus were amplified from pUD1069 (ScGPP1) with primer pairs 15192/15194 and 15196/15197, yielding two integration fragments with a homologous sequence overlap. Similarly, hph splitmarker fragments for integration of an ScGPD2 cassette were constructed by amplification from pUD1082 (ScGPD2) with primer pairs 15740/15741 and 15742/15736. Correct integration of the split-marker fragments at the OpGBU1 locus was verified by diagnostic PCR with primers 15192, 15197, 15233, and 15234 and integration at the OpSGA1 locus with primers 15894, 15748, and 15895.

Bioreactor cultivation
Chemostat cultures of O. parapolymorpha strains were grown in 2-l bioreactors (Applikon Biotechnology, Delft, The Netherlands) with a working volume of 1.2 l, operated at a dilution rate of 0.1 h -1 , at pH 6, at 30 • C, and at a stirrer speed of 800 rpm. Oxygen-limited chemostat cultures were sparged at a rate of 0.5 l min -1 (0.4 vvm) with a mixture of N 2 and air that contained 840 ppm O 2 , and aerobic cultures with air (21 × 10 4 ppm O 2 ). Cultures were fed with a synthetic medium with vitamins and with urea as nitrogen source (Luttik et al. 2000), supplemented with 7.5 g l -1 glucose (aerobic cultures) or 20 g l -1 glucose (oxygen-limited cultures) and 0.2 g l -1 Pluronic 6100 PE antifoam (BASF, Ludwigshafen, Germany). An 800-fold concentrated solution of the anaerobic growth factors Tween 80 (polyethylene glycol sorbitan monooleate; Merck, Darmstadt, Germany), ergosterol (≥ 95% pure; Sigma-Aldrich) in ethanol was prepared and added to sterile media as described previously (Dekker et al. 2019), but with a 5-fold lower Tween 80 concentration. Concentrations of Tween 80, ergosterol and ethanol in reservoir media of oxygen-limited cultures were 84 mg l -1 , 10 mg l -1 , and 0.67 g l -1 , respectively. Tween 80 was omitted from media for aerobic cultivation to prevent excessive foaming. Where indicated, a filter-sterilized acetoin solution was added to a concentration of 2.0 g l -1 . Before autoclaving, bioreactors were checked for gas leakage by submersion in water while applying a 0.3 bar overpressure. Bioreactors were equipped with Fluran tubing and Viton O-rings and the glass medium reservoir was equipped with Norprene tubing and continuously sparged with pure nitrogen gas to minimize oxygen entry. Inocula for bioreactor cultures were prepared by harvesting an exponentially growing 100-ml shakeflask culture on synthetic medium with glucose by centrifugation (5 min at 4000 x g) and washing the biomass once with sterile demineralized water. Oxygen-limited chemostat cultures were started from aerobic bioreactor batch cultures on synthetic medium containing 1.5 g l -1 glucose. When CO 2 production in these batch cultures had reached a maximum and started to decline, chemostat cultivation was initiated by applying a constant medium feed rate and continuous effluent removal. Chemostat cultures were assumed to have entered steady state when, at least 5 volume changes after a change in growth conditions, the biomass concentration and specific carbon dioxide production rate differed by less than 10% over three samples separated by at least one volume change.
Aerobic bioreactor batch and chemostat (D = 0.1 h -1 ) cultures of O. parapolymorpha CBS11895 and IMX2167 (Table 5) were grown at 30 • C and at pH 5 in 2-l bioreactors (Applikon Biotechnology) with a working volume of 1.0 l, on a synthetic medium (Verduyn et al. 1990) containing 7.5 g l -1 glucose and 20 g l -1 glucose, respectively, and ammonium sulfate as nitrogen source.

Analytical methods
Off-gas analysis, biomass dry weight measurements, optical density measurements, metabolite HPLC analysis of culture supernatants, and correction for ethanol evaporation in bioreactor experiments were performed as described previously ). Rates of substrate consumption and metabolite production were calculated from glucose and metabolite concentrations in steady-state cultures, analyzed after rapid quenching of culture samples (Mashego et al. 2003). Recoveries of carbon and degree of reduction (Roels 1980) were calculated based on concentrations of relevant components in medium feed, culture samples, and inand out-going gas streams. For organic compounds that only contain carbon, hydrogen, and/or oxygen, degree of reduction (γ ) represents the number of electrons released upon complete oxidation to CO 2 , H 2 O and or H + . These oxidized compounds are assigned a γ of zero, which yields defined values of γ for H, C, and O of 1, 4, and 2, respectively and for positive and negative charge of −1 and 1, respectively. To simplify construction of degree-ofreduction balances, the nitrogen source is generally assigned γ = 0 which, with NH 4 + as nitrogen source, implies that γ = 3 for N.
Calculations were based on an estimated degree of reduction and carbon content of yeast biomass (Lange and Heijnen 2001).

Genome sequencing and assembly
Cells were harvested from an overnight culture on YPD by centrifugation (5 min at 4000 x g) and genomic DNA was isolated with the Qiagen genomic DNA 100/G Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. MinION genomic DNA libraries (SQK-LSK108, Oxford Nanopore Technologies, Oxford, UK) were prepared using the 1D genomic DNA by ligation and the SQK-LSK108 library was sequenced on an R9 chemistry flow cell (FLO-MIN107). Base calling was performed with Albacore v1.1.5 (Oxford Nanopore Technologies), reads were assembled using Canu v1.4 (Koren et al. 2017), and the resulting assembly was polished with Pilon v1.18 (Walker et al. 2014

RNA extraction, isolation, sequencing, and transcriptome analysis
Biomass samples from batch and chemostat cultures were directly sampled into liquid nitrogen to prevent mRNA turnover (Piper et al. 2002). Processing of samples for storage at −80 • C and RNA isolation were performed as described previously . Batch cultures were sampled when, in the exponential growth phase, approximately 25% of the initially supplied glucose had been consumed (Juergens et al. 2020 et al. 2012) was used to perform differential gene expression and genes with fewer than 10 reads per million in all conditions were eliminated from subsequent analysis. Counts were normalized using the trimmed mean of M values (TMM; Robinson and Oshlack 2010) method and the dispersion was estimated using generalized linear models. Differential expression was calculated using a log ratio test adjusted with the Benjamini-Hochberg method. Absolute log 2 fold-change values (> 2), false discovery rate (< 0.5), and P-value (< .05) were used as significance cut-offs. Gene set analysis (GSA) based on gene ontology (GO) terms with Piano (v2.4.0; Väremo et al. 2013) was used for functional interpre-tation of differential gene expression profiles. Interproscan (Jones et al. 2014) was used to assign GO terms to the genome annotation of O. parapolymorpha. Co-ortholog groups of genes were generated with ProteinOrtho5 (Lechner et al. 2011) as implemented in the funannotate pipeline and used to homogenize GO terms for coortholog groups as described previously . GSA was done with Piano (v2.4.0; Väremo et al. 2013) and gene statistics were calculated with Stouffer, Wilcoxon rank-sum test and reporter methods as implemented in Piano. Consensus gene level statistics were obtained by P-value and rank aggregation and considered significant when absolute log 2 fold-change values > 1. ComplexHeatmap (v2.4.3;Gu et al. 2016) was used to visualize differentially expressed genes. To interpret the GO-term based GSA between three yeast species in response to oxygen limitation, hierarchical clustering (complete method and Euclidian distance) in R (R Core Team 2021) was performed on GO-terms from biological process category. Clustering was based on the number of overlapping distinct directionality P-values in the three yeast species with a significance P-value cut-off of .01.

Sequence homology searches
Saccharomyces cerevisiae protein sequences were used as queries to search whole-genome sequences of 16 Ogataea species, K. marxianus, Candida arabinofermentans, and Brettanomyces bruxellensis with tblastn (blast.ncbi.nlm.nih.gov; Camacho et al. 2009). Significance was based on alignment criteria, with an e-value of < 10 -7 , > 70% alignment coverage and > 50% nucleotide identity. Blast results were mapped to a subtree of selected yeast species in the phylum Ascomycota (Shen et al. 2020) using Treehouse (Steenwyk and Rokas 2019) to subset the phylogenetic tree.

Oxygen requirements of O. parapolymorpha in oxygen-limited chemostat cultures
Chemostat cultivation enables analysis of impacts of different process parameters at a fixed specific growth rate, which in ideally mixed, steady-state chemostat cultures equals the dilution rate (D, h -1 ). Oxygen requirements of the wild-type O. parapolymorpha strain CBS11895 (DL-1; Suh and Zhou 2010) were quantitatively assessed by comparing its physiology under two aeration regimes in glucose-grown chemostat cultures operated at D = 0.1 h -1 (Table 3). Results from this analysis were compared with data that were previously obtained, under the same cultivation conditions, with S. cerevisiae CEN.PK113-7D and K. marxianus CBS6556 .
In fully aerobic chemostat cultures sparged with air (0.5 l min -1 ), growth of O. parapolymorpha was glucose limited and sugar dissimilation occurred exclusively via respiration, as indicated by a respiratory quotient (RQ) close to 1 ( Table 3). The apparent biomass yield on glucose in aerobic cultures was approximately 10% higher than previously reported (Verduyn et al. 1991) due to co-consumption of ethanol, which was used as solvent for ergosterol. When cultures were instead sparged with a mixture of N 2 and air (0.5 l min -1 , oxygen content 840 ppm), the apparent biomass yield on glucose in steady-state cultures was 4-fold lower than in the aerobic cultures (0.15 g g -1 and 0.59 g g -1 , respectively, Table 3). A high residual glucose concentration (15.9 g l -1 ) indicated that growth in these cultures was limited by oxygen rather than by glucose. Respiro-fermentative glucose dissimilation by the oxygen-limited cultures was evident from an RQ of 10.7 and a specific ethanol-production rate of 4.8 mmol (g biomass) -1 h -1 .
In contrast to results that were previously obtained with K. marxianus and S. cerevisiae , a further reduction of the oxygen content of the inlet gas to below 0.5 ppm caused wash-out of the O. parapolymorpha chemostat cultures.
Saccharomyces cerevisiae can grow anaerobically in synthetic media supplemented with sterols, a UFA source and a standard vitamin solution also used for aerobic cultivation (Andreasen and Stier 1953, 1954, Dekker et al. 2019. Based on UFA and sterol contents of aerobically grown S. cerevisiae biomass, the minimum oxygen-uptake rate required for synthesis of these lipids at a specific growth rate of 0.10 h -1 were estimated at 0.01 mmol O 2 (g biomass) -1 h -1 (Dekker et al. 2019). The biomass-specific oxygenconsumption rate of 0.60 mmol O 2 (g biomass) -1 h -1 observed in oxygen-limited cultures of O. parapolymorpha (Table 3) was 60fold higher than this estimate. Based on the assumption that oxygen-limited cultures predominantly used oxygen for respiration, oxygen-uptake and ethanol-production rates indicated that approximately 3% of the glucose consumed by these cultures was respired. Under the same oxygen-limitation regime, S. cerevisiae and K. marxianus showed specific oxygen-consumption rates below 0.25 mmol (g biomass) -1 h -1 , RQ values above 50 and very low residual glucose concentrations (Table 3; Dekker et al. 2021).
Oxygen-limited cultures of O. parapolymorpha showed an over 10-fold lower biomass-specific rate of glycerol production than similar cultures of K. marxianus and S. cerevisiae (0.02 versus 1.12 and 0.45 mmol g (biomass) -1 h -1 , respectively, Table 3). In multiple yeast species, glycerol production plays a key role during anaerobic and oxygen-limited growth by enabling reoxidation of surplus NADH formed in biosynthetic reactions (Scheffers 1966, van Dijken and Scheffers 1986, Weusthuis et al. 1994, Bakker et al. 2001, which, under aerobic conditions, is achieved by mitochondrial respiration (Bakker et al. 2001). In fully anaerobic cultures of S. cerevisiae and in severely oxygen-limited K. marxianus cultures, 7-12 mmol glycerol was formed per gram of yeast biomass . At D = 0.1 h -1 , reoxidation of an equivalent amount of NADH by respiration would require an uptake rate of 0.4-0.6 mmol O 2 g (biomass) -1 h -1 , which corresponds well with the observed oxygen consumption rates of the oxygen-limited O. parapolymorpha chemostat cultures (Table 3).
In some yeasts, an insufficient capacity for glycerol production has been linked to an inability to grow under severe oxygen limitation. Scheffers (1963Scheffers ( , 1966 showed that this phenomenon, which he labeled the Custers effect, no longer occurred when cultures were supplemented with acetoin. Similarly, in S. cerevisiae, NADHdependent reduction of acetoin by the 2,3-butanediol dehydrogenase Bdh1 (Gonzalez et al. 2000) restores fermentation of glycerolnegative strains (Björkqvist et al. 1997). Presence of an ScBdh1 ortholog in the predicted proteome of O. parapolymorpha CBS11895 (HPODL_00988; Figure S1, Supporting Information) indicated that this enzyme activity also occurs in O. parapolymorpha.
Addition of acetoin to oxygen-limited chemostat cultures of O. parapolymorpha led to an increase of the steady-state biomass concentration from 0.62 to 1.57 g l -1 . A higher rate of ethanol production, a higher biomass yield on oxygen and a higher RQ (Table 4) indicated that acetoin addition led to a more fermentative metabolism. Although the biomass-specific ethanol production rates in acetoin-supplemented cultures (5.9 mmol g x -1 h -1 , Table 4) approached those of oxygen-limited cultures of S. cerevisiae CEN.PK113-7D grown at the same dilution rate (7.5 mmol g x -1 h -1 , Table 3), almost half of the glucose in the cultures remained unused. In addition, biomass-specific rates of acetoin consumption (0.97 mmol g x -1 h -1 ) in the O. parapolymorpha cultures were much higher than the rates estimated to be required for reoxidation of Table 3. Physiological parameters of chemostat cultures (D = 0.1 h -1 , 30 • C) of O. parapolymorpha, K. marxianus, and S. cerevisiae, grown on glucose under aerobic (21 × 10 4 ppm O 2 in inlet gas, 7.5 g l -1 glucose in feed medium) or oxygen-limited (840 ppm O 2 in inlet gas; 20 g l -1 glucose in feed medium) conditions. Data for K. marxianus and S. cerevisiae were obtained from a previous study . Growth media were supplemented with ergosterol and Tween 80, except for media for aerobic cultures of O. parapolymorpha, from which Tween 80 was omitted to prevent excessive foaming. Data are represented as mean ± standard deviation of data obtained from replicate chemostat cultures. Negative and positive biomass-specific conversion rates (q) represent consumption and production rates, respectively, with subscript x denoting biomass dry weight. B.D.: below detection limit (concentration < 0.  Table 4. Physiological parameters of glucose-grown, oxygen-limited chemostat cultures (D = 0.1 h -1 , 30 • C) of O. parapolymorpha strains expressing the S. cerevisiae glycerol pathway genes ScGPP1 and/or ScGPD2 or supplemented with 2.0 g l -1 acetoin. Data are represented as mean ± standard deviation of data obtained from replicate chemostat cultures. Negative and positive biomass-specific conversion rates (q) represent consumption and production rates, respectively, with subscript x denoting biomass dry weight. B.D.: below detection limit (concentration < 0.1 mM) and (-): not applicable due to co-consumption of ethanol added as ergosterol solvent. 92.0 ± 6.1 93.5 ± 3.1 102.9 ± 0.4 106.0 ± 3.1 95.6 ± 1.1 NADH generated in biosynthetic reactions. This observation suggested that 2,3-butanediol dehydrogenase activity in O. parapolymorpha not only reoxidized NADH formed in biosynthetic reactions but also NADH derived from sugar dissimilation. As a consequence, it would compete for NADH with alcohol dehydrogenase. Furthermore, we cannot exclude that, in these cultures, 2,3butanediol dehydrogenase also used NADPH generated in the oxidative pentose-phosphate pathway or other NADP + -dependent oxidative processes or reactions. This notwithstanding, these results clearly implicated a limited capacity for NADH reoxidation as a key factor in the unexpectedly large oxygen requirements of O. parapolymorpha.

Absence of orthologs of S. cerevisiae glycerol-3P phosphatase in Ogataea species
To study the molecular basis for the near absence of glycerol formation in oxygen-limited cultures of O. parapolymorpha, we investigated presence of orthologs of S. cerevisiae GPD1/2 and GPP1/2 in genomes of Ogataea species. These genes encode isoenzymes that catalyze the two key reactions of the S. cerevisiae glycerol pathway, NAD + -dependent glycerol-3P dehydrogenase and glycerol-3P phosphatase, respectively (Albertyn et al. 1992, Norbeck et al. 1996. A homology search in translated wholegenome sequences of 16 Ogataea species (Shen et al. 2020) revealed clear Gpd orthologs, but no Gpp orthologs (Fig. 1). In this respect, Ogataea yeasts resembled the phylogenetically related genus Brettanomyces (syn. Dekkera; Fig. 1), whose representatives are known to exhibit a Custers effect (Wijsman et al. 1984, Galafassi et al. 2013. In the absence of glycerol-3P phosphatase, NAD + -dependent glycerol-3P dehydrogenase can still contribute to glycerolipid synthesis (Athenstaedt et al. 1999) and participate in the glycerol-3P shuttle for coupling oxidation of cytosolic NADH to mitochondrial respiration (Larsson et al. 1998, Overkamp et al. 2000, Rigoulet et al. 2004; Fig. 1).

Transcriptional responses of O. parapolymorpha to oxygen limitation
Responses of O. parapolymorpha to oxygen limitation were further explored by transcriptome analyses on aerobic and oxygenlimited chemostat cultures. The resulting transcriptome data were first used to refine the genome annotation of a de novo assembled genome sequence of O. parapolymorpha CBS11895 obtained from long-read sequence data (see Data availability). Transcriptional responses of O. parapolymorpha to oxygen limitation were compared to those of S. cerevisiae and K. marxianus ) grown under the same aeration regimes. A global comparison at the level of functional categories indicated large differences in the transcriptional responses of these three yeasts to oxygen limitation ( Fig. 2A). Of genes for which orthologs occur in all three species (Fig. 2B), only very few showed a consistent cross-species transcriptional response to oxygen limitation ( Fig. 2C and D). At first glance, these different transcriptional responses suggested a completely different wiring of their oxygenresponsive transcriptional regulation networks. Based on functional categories, the only shared global transcriptional responses of O. parapolymorpha, K. marxianus, and S. cerevisiae were a downregulation, in the oxygen-limited cultures, of genes involved in the metabolism of nonglucose carbon sources (GO categories fattyacid metabolic process, tricarboxylic acid cycle, transmembrane transport, metabolic process, and lipid metabolic process; Fig. 2A). These responses are in line with the requirement for oxygen in the dissimilation of nonfermentable substrates and for a key role of the tricarboxylic acid cycle in respiratory glucose metabolism. However, in addition to oxygen availability, different glucose and ethanol concentrations in chemostat cultures of the tested yeast strains (Table 3) may have had a strong impact on transcript profiles. For example, in comparisons of glucose-limited and glucosesufficient chemostat cultivation regimes, hundreds of S. cerevisiae genes were shown to exhibit an at least 2-fold difference in transcript level (Meijer et al. 1998, Boer et al. 2003, Tai et al. 2005. In view of this intrinsic limitation of the chemostat-based transcriptome studies, analysis was focused on genes and pathways that were previously implicated in biosynthetic oxygen requirements of yeasts. Sterol biosynthesis requires molecular oxygen and, under anaerobic conditions, S. cerevisiae can acquire ergosterol from the media. In contrast to S. cerevisiae, which showed downregulation of genes associated with sterol metabolism in oxygen-limited cultures, GO-term enrichment analysis showed upregulation of genes associated with this process in O. parapolymorpha and K. marxianus ( Fig. 2A). Kluyveromyces marxianus and several other pre-WGD yeast species lack a functional sterol-import system , Tesnière et al. 2021). Upregulation of sterol synthesis genes in oxygen-limited, sterol-supplemented cultures ( Fig. 2A), as well as absence of clear orthologs of the S. cerevisiae AUS1 and PDR11 sterol-importer genes in its genome, suggested that the same holds for O. parapolymorpha.
A recent study confirmed that OpURA9, which encodes the respiratory-chain-linked Class-II DHODase of O. parapolymorpha, complements the uracil auxotrophy of ura1 S. cerevisiae under aerobic, but not under anaerobic conditions . OpURA9 showed higher transcript levels in oxygen-limited cultures than in aerobic cultures, while its K. marxianus ortholog KmURA9 showed the reverse response (Fig. 3). This observation is consistent with the presence and absence of a respirationindependent Class-I-A DHODase in K. marxianus and O. parapolymorpha, respectively .
The importance of glycerol production in oxygen-limited cultures of S. cerevisiae and K. marxianus was evident from an upregulation of GPP1 (Fig. 3), for which no ortholog was found in O. parapolymorpha (Fig. 1). Lack of a transcriptional response of the single GPD ortholog to oxygen limitation are consistent and further supports the notion that O. parapolymorpha does not use glycerol formation as a redox sink during oxygen-limited growth.

Glycerol production in aerobic cultures of an O. parapolymorpha strain lacking mitochondrial glycerol-3P dehydrogenase and alternative NADH dehydrogenases
Presence of orthologs of S. cerevisiae GPD1/2 and GUT2 in the O. parapolymorpha genome suggested possible involvement of a glycerol-phosphate shuttle (Larsson et al. 1998) in respiratory oxidation of cytosolic NADH. To investigate whether elimination of systems for mitochondrial, respiratory oxidation of NADH would affect glycerol production by O. parapolymorpha, we studied growth and product formation in strain IMX2167. In this strain, OpGUT2 and the genes encoding three cytosol-and matrix-facing alternative mitochondrial NADH dehydrogenases were deleted, while leaving the Complex-I NADH dehydrogenase complex intact (Juergens et al. 2021). In aerobic chemostat cultures grown at D = 0.1 h -1 , conversion rates of strain IMX2167 were not substantially different from those of the wild-type strain CBS11895 (Table 5). Apparently, as observed in aerobic cultures of corresponding mutant strains of S. cerevisiae (Bakker et al. 2001), an ethanol-acetaldehyde  Table 5. Physiological parameters of glucose-grown aerobic bioreactor-batch and chemostat cultures (30 • C) of wild-type O. parapolymorpha and strains carrying null mutations in genes involved in mitochondrial oxidation of NADH. Batch culture data were derived from analyses on samples taken during the exponential growth phase. Data on aerobic chemostat cultures were derived from a separate study (Juergens et al. 2021). Chemostat cultures (D = 0.1 h -1 ) and batch cultures were grown on 20 g l -1 glucose and 7.5 g l -1 glucose, respectively. Data are represented as mean ± standard deviation of data obtained from replicate cultures. Negative and positive biomass-specific conversion rates (q) represent consumption and production rates, respectively, with subscript x denoting biomass dry weight. B.D.: below detection limit (concentration < 0.1 mM) and N.D.: not determined.
In aerobic batch cultures, strain IMX2167 grew slower than the wild-type strain CBS11895 (0.26 h -1 and 0.36 h -1 , respectively, Table 5). Glycerol production by strain IMX2167 suggested that, while , and S. cerevisiae (scer) to oxygen limitation. Aerobic (regime 1, 21 × 10 4 ppm O 2 in inlet gas, and 7.5 g l -1 glucose in feed medium) and oxygen-limited (regime 2, 840 ppm O 2 in inlet gas; 20 g l -1 glucose in feed medium) chemostat cultures were grown at D = 0.1 h -1 and 30 • C. Data for K. marxianus and S. cerevisiae were obtained from a previous study . (A) Gene-set enrichment analysis showing GO-terms overrepresented among genes showing a transcriptional response to oxygen limitation (regime 2 versus regime 1) in at least two of the three yeast species. Distinct directionalities calculated with Piano (Väremo et al. 2013) are indicated as distinct-directional down (pdddn), mixed-directional down (pmddn), nondirectional (pnd), mixed-directional up (pmdup), and distinct-directional up (pddup). Hierarchical clustering was based on degree of overrepresentation. Data on all enriched GO-terms for biological processes are shown in Figures  lacking an ScGPP1 ortholog, O. parapolymorpha contains an alternative glycerol-3-phosphatase. Annotation of the newly assembled genome sequence of strain CBS11895 yield 24 genes annotated with the GO-term 'phosphatase activity' (GO:0016791). While all 24 were transcribed (log Counts per million (CPM) > 3.5), none showed significantly higher (log FC > 2) transcript levels in strain IMX2167 than in the wild-type strain CBS11895 (Table S2 and Figures S5 and S6,Supporting Information).

Engineering of glycerol metabolism in O. parapolymorpha
Based on the absence of orthologs of S. cerevisiae ScGPP1 in genomes of Ogataea sp. (Fig. 1B), we investigated whether expression of ScGPP1 in O. parapolymorpha supported glycerol production by oxygen-limited cultures. An expression cassette in which the coding region of ScGPP1 was expressed from the OpPMA1 promoter (Juergens et al. 2020) was integrated into the genome of O. parapolymorpha CBS11895. The resulting strain IMX2119 showed a 9-fold higher biomass-specific rate of glycerol formation in oxygen-limited cultures than the wild-type strain (0.18 mmol (g biomass) -1 h -1 and 0.02 mmol (g biomass) -1 h -1 , respectively, Table 4 and Fig. 4A). A further increase of the glycerol production rate to 0.22 mmol (g biomass) -1 h -1 was observed when ScGPP1 expression was combined with integration of a cassette in which ScGPD2 was expressed from the OpTEF1 promoter (Juergens et al. 2020; strain IMX2588; Table 4 and Fig. 4A). Integration of only the ScGPD2 cassette (strain IMX2587) did not result in a significantly higher rate of glycerol production in oxygen-limited cultures than observed for the wild-type strain ( Table 4).
The higher biomass-specific rates of glycerol production by the ScGPP1 and ScGPP1/ScGPD2 expressing O. parapolymorpha strains coincided with higher biomass yields on oxygen under oxygenlimited conditions (0.24 and 0.32 g biomass mmol O 2 -1 , respectively, versus 0.17 g biomass mmol O 2 -1 for the wild-type strain; Table 4 and Fig. 4B). A larger contribution of alcoholic fermentation to glucose dissimilation was also concluded from the RQ val- . Ogataea parapolymorpha washed out under regime 3. Data for K. marxianus and S. cerevisiae were obtained from a previous study  ues of strains IMX2119 andIMX2588 (14.4 and17.4, respectively), which were higher than those of corresponding oxygen-limited cultures of the wild-type strain (RQ of 10.7, Table 4). For fully anaerobic chemostat cultures of S. cerevisiae CEN.PK113-7D grown at D = 0.1 h -1 , a biomass-specific rate of glycerol production of 0.67 mmol (g biomass) -1 h -1 was reported ( Fig. 4A; Geertman et al. 2006). An even higher rate of glycerol production (1.1 mmol (g biomass) -1 h -1 ) was reported for strains of another S. cerevisiae lineage grown under these conditions (Weusthuis et al. 1994, Nissen et al. 2000. Assuming that biomass composition and biosynthetic pathways in S. cerevisiae CEN.PK113-7D and O. parapolymorpha CBS11895 lead to a similar net generation of NADH, the glycerol production rate of the ScGPP1/ScGPD2 expressing O. parapolymorpha strain IMX2588 remained approximately 4-fold lower than needed for reoxidation of all NADH generated in biosynthesis. A limiting capacity of the engineered glycerol pathway was further indicated by the residual glucose concentrations in oxygen-limited cultures of Data are represented as mean ± standard deviation of data obtained from independent chemostat cultures of each strain. (A) Biomass-specific glycerol production rates versus biomass-specific oxygen consumption rates. The dashed line depicts a stoichiometric relationship between glycerol production and oxygen consumption in S. cerevisiae cultures, based on the assumption that, for NADH reoxidation, consumption of one mol O 2 corresponds to production of 2 moles of glycerol (Weusthuis et al. 1994). (B) Biomass yields on oxygen versus biomass-specific rates of glycerol production. strain IMX2588, which were higher than in acetoin-supplemented cultures of the wild-type strain CBS11895 (Table 4).

Discussion
This study revealed a surprisingly high oxygen requirement in oxygen-limited cultures of the facultatively fermentative yeast O. parapolymorpha (previously Hansenula polymorpha; Kurtzman 2011) relative to those previously reported for the pre-WGD yeasts Kluyveromyces marxianus, K. lactis, and Candida utilis (Cyberlindnera jadinii; Weusthuis et al. 1994, Kiers et al. 1998). Very low glycerol-production rates and a strong impact of acetoin co-feeding to oxygen-limited cultures identified reoxidation of NADH, formed in biosynthetic reactions, as a key contributor to the large oxygen requirement of O. parapolymorpha. A large oxygen requirement for fermentative growth ('Custers effect'; Wikén et al. 1961), absence of glycerol production and a stimulating effect of acetoin on oxygen-limited growth were previously observed in Brettanomyces (Dekkera) yeasts (Custers 1940, Wikén et al. 1961, Scheffers 1966, Wijsman et al. 1984. The Custers effect in B. bruxellensis was attributed to absence of glycerol-3P phosphatase activity in cell extracts (Wijsman et al. 1984) and lack of an ortholog of the S. cerevisiae GPP1/GPP2 genes (Tiukova et al. 2013). The genera Ogataea and Brettanomyces both belong to the Pichiacaea family (Shen et al. 2016). Our observations on O. parapolymorpha, combined with the absence of clear ScGPP1/ScGPP2 orthologs in genomes of other Ogataea species, provide an incentive for further studies into the occurrence, regulatory basis and ecophysiological significance of a Custers effect in Pichiacaea. In view of its fast growth in synthetic media (Juergens et al. 2018a) and its accessibility to genome-editing techniques (Juergens et al. 2018b, Gao et al. 2021, O. parapolymorpha offers an interesting experimental platform for such studies.
Ogataea parapolymorpha is applied in aerobic industrial processes for production of heterologous proteins (Stasyk 2017) and, based on its thermotolerance and natural ability to metabolize dxylose, is under investigation as a potential platform organism for second-generation ethanol production (Kurylenko et al. 2014). In anaerobic industrial applications of Saccharomyces yeasts such as beer fermentation, introduction of a brief aeration phase enables yeast cell to synthesize and intracellularly accumulate sterols and UFAs, which are then used during the subsequent anaerobic fermentation phase (Casey et al. 1984, Meyers et al. 2017. The large oxygen requirements of O. parapolymorpha observed in this study imply that such a strategy is not feasible for this yeast. Elimination of the Custers effect in O. parapolymorpha is, therefore, a priority target for development of industrial ethanol-producing strains. Formation of glycerol in aerobic cultures of strain IMX2167, in which genes encoding key enzymes of respiratory NADH oxidation, including mitochondrial glycerol-3-phosphate dehydrogenase (OpGut2), were deleted, suggested that the O. parapolymorpha genome may harbor a gene encoding a glycerol-3-phosphatase. Alternatively, glycerol formation in this strain may reflect activity of another pathway for glycerol production (e.g. involving DHAP phosphatase, Fig. 1). Laboratory evolution of wild-type and engineered O. parapolymorpha strains under oxygen-limited conditions and resequencing of evolved strains (Mans et al. 2018) may contribute to a better understanding of glycerol production in this yeast.
Expression of S. cerevisiae GPP1 and GPD2 enabled increased rates of glycerol formation and a higher biomass yield on oxygen in oxygen-limited cultures of O. parapolymorpha (Table 4). However, glycerol production rates were lower than observed in anaerobic cultures of S. cerevisiae (Fig. 4A) and a large fraction of the glucose fed to the cultures remained unused. These results indicated that the in vivo capacity of NADH reoxidation via heterologously expressed Gpp1 and Gpd2 was insufficient to fully replace the role of mitochondrial respiration in the reoxidation of NADH generated in biosynthetic reactions. Increased expression of GPP1 and GPD2, possibly combined with expression of a glycerol exporter and/or laboratory evolution under oxygen-limited conditions can be explored to further enhance glycerol production in O. parapolymorpha. Alternatively, expression of heterologous pathways for NADH-dependent reduction of acetyl-CoA to ethanol (Medina et al. 2010) or NADH oxidation via a pathway involving ribulose-1,5bisphosphatase and phosphoribulokinase (Guadalupe- Medina et al. 2013, Papapetridis et al. 2018 can be explored. In oxygen-limited cultures of O. parapolymorpha that were cofed with acetoin, incomplete glucose consumption occurred despite rates of acetoin conversion that were 2-fold higher than glycerol production rates in anaerobic S. cerevisiae cultures (Table 3 and Fig. 4). This result suggests that, in this yeast, not only the capacity for reoxidation of NADH generated in biosynthesis but also for NADH generated in glycolysis may be limited. This hypothesis can be tested by laboratory evolution under oxygen-limited conditions or, alternatively, by overexpression of key enzymes of pyruvate decarboxylase and/or alcohol dehydrogenase.
Predicted stoichiometric oxygen requirements for sterol synthesis and pyrimidine synthesis of O. parapolymorpha are small in comparison with those for NADH reoxidation. However, their physiological impacts can be augmented when key enzymes involved in these processes have a low affinity for oxygen. Absence of orthologs of the S. cerevisiae Aus1 and Pdr11 sterol transporters indicates that, similar to other pre-WGD yeasts (Seret et al. 2009), O. parapolymorpha is probably unable to import sterols. Due to the incompletely resolved role of cell wall proteins in sterol import in S. cerevisiae (Alimardani et al. 2004), functional expression of a heterologous system for sterol import in O. parapolymorpha may not be a trivial challenge. Alternatively, it may be explored whether, as shown in S. cerevisiae and K. marxianus (Wiersma et al. 2020, expression of a heterologous squalene-tetrahymanol cyclase, which synthesizes the sterol surrogate tetrahymanol, can support sterol-independent growth of O. parapolymorpha. Genomesequence data indicate that pyrimidine synthesis in O. parapolymorpha depends on a respiratory-chain-linked dihydroorate dehydrogenase (OpUra9), thus rendering pyrimidine biosynthesis in this yeast oxygen dependent Jeffries 1998, Gojković et al. 2004). As previously explored in Scheffersomyces stipitis, expression of the soluble fumarate-coupled DHODase from S. cerevisiae (Ura1; Shi and Jeffries 1998) or, alternatively, of recently described respiration-independent orthologs of Ura9  may be applied to bypass this oxygen requirement.
This study illustrates how rigorous standardization of oxygenlimited cultivation regimes (Mooiman et al. 2021) enables quantitative comparisons and physiological analysis of oxygen requirements of facultatively fermentative yeasts. We recently showed that enabling synthesis of a sterol surrogate sufficed to eliminate oxygen requirements of oxygen-limited K. marxianus cultures . By demonstrating that oxygen requirements of O. parapolymorpha are much larger as well as more complex, the present study underlines the relevance of further comparative physiology studies on oxygen requirements across yeast and fungal species. Such studies are not only of fundamental scientific interest but should help to unlock the full potential of non-Saccharomyces yeasts for application in anaerobic industrial processes.

Supplementary data
Supplementary data are available at FEMSYR online.

Data availability
Numerical data presented in the figures in this work are available at https://figshare.com/s/283842c2a2a9a847e0bf. Raw sequencing data are available from NCBI (www.ncbi.nlm.nih.gov/geo/) under BioProject PRJNA717220.

Code availability
Codes used to generate the results obtained in this study are archived in a Gitlab repository (https://gitlab.tudelft.nl/rortizme rino/opar_anaerobic).