A native phosphoglycolate salvage pathway of the synthetic autotrophic yeast Komagataella phaffii

Abstract Synthetic autotrophs can serve as chassis strains for bioproduction from CO2 as a feedstock to take measures against the climate crisis. Integration of the Calvin–Benson–Bassham (CBB) cycle into the methylotrophic yeast Komagataella phaffii (Pichia pastoris) enabled it to use CO2 as the sole carbon source. The key enzyme in this cycle is ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) catalyzing the carboxylation step. However, this enzyme is error prone to perform an oxygenation reaction leading to the production of toxic 2-phosphoglycolate. Native autotrophs have evolved different recycling pathways for 2-phosphoglycolate. However, for synthetic autotrophs, no information is available for the existence of such pathways. Deletion of CYB2 in the autotrophic K. phaffii strain led to the accumulation of glycolate, an intermediate in phosphoglycolate salvage pathways, suggesting that such a pathway is enabled by native K. phaffii enzymes. 13C tracer analysis with labeled glycolate indicated that the yeast pathway recycling phosphoglycolate is similar to the plant salvage pathway. This orthogonal yeast pathway may serve as a sensor for RuBisCO oxygenation, and as an engineering target to boost autotrophic growth rates in K. phaffii.


Introduction
The Calvin-Benson-Bassham (CBB) cycle is responsible for the vast majority of carbon fixed on our planet.The k e y enzyme and rate limiting step of this cycle is ribulose-1,5-bisphosphate carbo xylase/o xygenase (RuBisCO), which catalyzes the carbon dioxide fixation to ribulose-1,5-bisphosphate. With an av er a ge turn over number of around 3 s −1 RuBisCO is a very slow enzyme.Plants ther efor e pr oduce huge amounts of this protein, with a Ru-BisCO content of around 30% to 50% of soluble protein (Feller et al. 2008 ).In addition, RuBisCO gets easily inhibited by a range of other sugar phosphates .T hese inhibitors can be released by the action of specific RuBisCO activ ases (P arry et al. 2008(P arry et al. , Hauser et al. 2015 ) ).
Another aspect which reduces the efficiency of the RuBisCO protein is the fact that it also reacts with oxygen instead of CO 2 whic h r educes the carbon fixation r ate .T he rate of this unfa vorable oxygenation reaction varies between different types of Ru-BisCO.In general, the RuBisCO protein family can be subdivided into 4 types where type I and II are the predominant forms in organisms performing the CBB cycle .Type II proteins ha ve often higher turnover numbers compared to type I proteins but lack their higher specificity to CO 2 leading to an inv erse corr elation between the specificity and the turnover rate (Tcherkez et al. 2006 ).
Even though various organisms harboring the CBB cycle perform carbon concentrating mechanisms to improve their carbon ca ptur e efficiency (Keeley and Rundel 2003, Raven et al. 2008, Yeates et al. 2008, Cameron et al. 2013, Wang et al. 2015 ), the oxygenation reaction is always present to some degree.Oxygenation of ribulose 1,5-bisphosphate produces one molecule 3phosphogl ycer ate (3-PG) and one molecule 2-phosphoglycolate.2phosphoglycolate is reported to be toxic in plants because it can inhibit at least two k e y enzymes of the central carbon metabolism, triose phosphate isomerase and phosphofructokinase (Hall et al. 1987, Flügel et al. 2017 ).Ther efor e, differ ent types of organisms hav e e volv ed differ ent pathways for r ecycling 2-phosphogl ycolate (Fig. 1 ).In plants, most algae and c y anobacteria this is called the C2 cycle (Fig. 1 , red line).It begins with the dephosphorylation of 2-phosphogycolate, follo w ed b y the oxidation into gly oxylate.Glyoxylate is then transaminated to glycine using either serine or glutamate as an amino group donor.Two molecules of glycine ar e then conv erted to serine in the mitoc hondria, r eleasing ammonia and carbon dioxide .T he serine is then transported back to the per oxisomes, wher e it donates its amino gr oup to another gl yoxylate molecule .T he deamination pr oduct, hydr oxypyruv ate, is then transported to the cytosol and reduced to glycerate.Finally, gl ycer ate is phosphorylated to 3-phosphogl ycer ate, whic h allows it to re-enter the CBB c ycle (Bauw e et al. 2010 ).The C2 cycle in plants consumes 3.5 ATP and 2 NADH equivalents and releases 0.5 molecules of CO 2 per RuBisCO o xygenation (Walk er et al. 2016 ).In c y anobacteria tw o different pathw ays r esponsible for the salv a ge of 2-phosphogl ycolate hav e been identified besides the C2 cycle.In one of these pathwa ys , the gl ycer ate pathway, 2 molecules of gl yoxylate ar e conv erted to tartr onate semialdehyde, further r educed to gl ycer ate and phosphorylated ending up in 3-PG avoid- ing the release of ammonia compared to the C2 cycle (Fig. 1 , violet line).The other pathway catalyzes the complete oxidation of 2phosphoglycolate to CO 2 via oxalate and formate in combination with harvesting of energy in the form of N ADH (Fig. 1 , y ello w line) (Eisenhut et al. 2006(Eisenhut et al. , 2008 ) ).In the c hemolithotr ophic bacterium Cupriavidus necator the gl ycer ate pathway is the main contributor to the recycling of 2-phosphoglycolate but another pathway was identified, the malate cycle which fully oxidizes glycolate to CO 2 (Fig. 1 , blue line) (Claassens et al. 2020 ) while harvesting energy in the form of 2 molecules NADH per glycolate oxidation.
All organisms which naturally perform the CBB cycle perform at least one of the possible routes for recycling of 2phosphogl ycolate.Latel y the CBB cycle was successfully integrated in the two model organisms Escherichia coli (Gleizer et al. 2019 ) and Komagataella phaffii (Gassler et al. 2020 ) .In both appr oac hes no specific pathway for the phosphoglycolate recycling was integrated.Cells were able to grow efficiently with CO 2 as sole carbon source leading to the assumption that the strains came up with a phosphogl ycolate salv a ge pathway based on native enzymes.Here we characterize this native phosphoglycolate salvage pathway in the synthetic autotrophic K. phaffii strain and show the robustness of yeast's metabolism in responding to a synthetic pathway and its side reactions.

Are synthetic autotrophs sensiti v e to oxygen?
In our pr e vious pa per, w e sho w ed that a synthetic autotrophic K. phaffii strain was able to grow on CO 2 as sole carbon source and methanol as energy source .T his was ac hie v ed by bloc king methanol assimilation via the xylulose-5-phosphate cycle as well as by the integration of six genes of the CBB cycle (Gassler et al. 2020 ).The k e y enzyme of the integrated CBB cycle is the Ru-BisCO enzyme which fixes the carbon dioxide to ribulose-1,5-bisphosphate, but also catalyzes the reaction with oxygen which reduces the efficiency of the CBB cycle for growth on CO 2 .Therefore, w e w anted to evaluate the effect of the oxygen concentration on growth rate of this strain.Bioreactor cultivations with oxygen concentrations in the inlet gas between 2.5% and 20% were performed and growth and dissolved oxygen were monitored.
Different inlet oxygen concentrations resulted in a difference in the growth profile (Fig. 2 A).The tw o lo w er oxygen concentrations of 5% and 10% resulted in faster growth compared to the higher oxygen le v els in the inlet gas .T he lo w est oxygen concentration of 2.5% resulted in slightly lo w er gro wth compared to 5 and 10%.A similar picture was observed in an engineered version (Gassler et al. 2022 ) of this strain which enables it to r eac h faster autotr ophic gr owth.Her e the only difference in terms of oxygen sensitivity was a slightl y incr eased oxygen demand of this strain ( Supplementary Figure 1 A), with an optimum oxygen concentration around 15%.
Man y or ganisms performing the CBB cycle have come up with carbon concentr ating mec hanisms to locall y incr ease CO 2 concentr ation to r educe the r ate of oxygenation.If these mechanisms ar e lac king in the synthetic autotr ophic K. phaffii , how does it cope with the oxygenation reaction?While reducing the oxygen concentr ation impr ov ed the gr owth r ate of the synthetic autotr ophic yeast, it can also gr ow efficientl y at ambient oxygen le v els .T his suggests that this synthetic autotroph is able to perform all reactions necessary to recycle 2-phosphoglycolate.

A single gene deletion blocks phosphogl ycola te salvage
Recentl y, we engineer ed this synthetic autotrophic K. phaffii for lactic acid production from CO 2 .To prevent the lactic acid consumption, we deleted CYB2 encoding for an L-lactate cytoc hr omec oxidor eductase whic h oxidizes lactate to pyruv ate in the mitoc hondrial intermembr ane space.Besides decr easing lactic acid reassimilation as intended, this knockout strain secreted glyco-Figure 2. Bior eactor cultiv ations using differ ent oxygen le v els in the inlet air to test their influence on gr owth.(A) Diamonds: offline OD 600 measurements and solid line online OD probe to monitor growth, (B) calculated growth rates and (C) dissolved oxygen concentrations of 5 different fermentations using oxygen concentrations in the inlet air from 2.5% to 20% v/v.Cultivations were performed at 30 • C with a constant stirrer speed of 300 rpm and a CO 2 concentration in the inlet air of 5%.
Figure 3. Cultivations using glycolate as sole carbon source to demonstrate that glycolate can be metabolized by K. phaffii and is suitable for further 13 C tracer experiments (A) Growth and (B) glycolate concentrations in the supernatant of the synthetic autotrophic K. phaffii str ain (par ental str ain) and CYB2 ov er expr ession str ains using weak or medium strength promoters.Cultivations were performed at 30 • C and ambient CO 2 concentrations in the atmosphere.Solid lines indicate the mean and shades the standard deviation of three biological replicates.late into the medium under autotrophic conditions (Baumschabl et al. 2022 ).As shown in Fig. 1 , glycolate is one the intermediate metabolites in the phosphoglycolate salvage pathwa ys , therefore these r esults pr ompted us to hypothesize that there is a pathway responsible for the recycling of the formed phosphoglycolate where the oxidoreductase Cyb2 is involved.
To identify the possible route of 2-phosphoglycolate salvage formed by the oxygenation reaction of RuBisCO, we designed 13 C tracer experiments with fully 13 C labeled glycolate.Prior to the 13 C tracer experiment, in order to make sure that glycolate can be metabolized by the cells and is suitable as a tracer metabolite, the synthetic autotrophic yeast strain and variants of this strain ov er expr essing CYB 2 under the control of a medium or weak promoter wer e cultiv ated using gl ycolate as the onl y carbon source.None of the tested strains were able to grow on glycolate (Fig. 3 A), but all strains could assimilate it.When CYB2 was overexpressed, a higher assimilation rate was observed (Fig. 3 B).These results confirmed that glycolate is assimilated in the metabolism and 13 C fr om gl ycolate will ther efor e be incor por ated demonstr ating that glycolate can be used as a tracer metabolite for the 13 C-labeling experiments.

Identification of the nati v e phosphoglycolate salv age pa thway
Glycolate can be efficiently assimilated by the autotrophic K. phaffii strain which is a prerequisite for its use as a tracer metabolite to further investigate the route of phosphoglycolate salvage.
Ther efor e, thr ee differ ent str ains, the par ental autotr ophic str ain, a CYB2 knoc k out str ain and the CYB2 ov er expr ession str ain using the medium strength promoter, were cultivated on fully labeled 13 C glycolate for 48 hours.Samples were taken after 3, 24, and 48 hours for the determination of the isotopologue distribution of v arious intr acellular metabolites using GC-TOFMS.If a metabolite w as inv olv ed in the phosphogl ycolate salv a ge pathw ay, it w ould show a decrease in the "M0" isotopologue ( 12 C only) (blue bars in Fig. 4 ) and an increase in the higher mass isotopologues, showing that 13 C fr om labeled gl ycolate is incor por ated into the r espectiv e metabolite.Metabolites with the highest 13 C content are supposed to be located at the start of a pathway.When inter pr eting the labeling data, it had to be considered that all strains produce 12 C glycolate via oxygenation of ribulose 1,5-bisphosphate catalyzed by the expressed RuBisCO, which can reduce the incorporation of 13 C atoms.
In the parental strain pronounced incorporation of 13 C atoms into serine and glycine was observed already after 3 hours (Fig. 4 A), indicating that most of the glycolate is recycled via glycine and not oxidized in the malate cycle .T he 13 C content of both metabolites increased until the end of cultivation.Neither pyruv ate nor hydr oxypyruv ate could be analyzed with the applied GC-TOFMS method, hence the metabolic step following serine had to be deduced from labeling patterns of adjacent metabolites.Significant incor por ation of 13 C into PEP, 2-PG, 3-PG and glycer ate was observ ed but to a lo w er extent compared to serine and glycine.Ribose 5-phosphate sho w ed only a small fraction of 13 C labeled isotopologues.Metabolites related to the TCA cycle such as malate and fumarate sho w ed a similar labeling degree as the phosphorylated hydro xycarbo xylic acids.Since malate and aspar- tate sho w ed similar labeling patterns, the labeling pattern of both metabolites pr obabl y originates fr om oxaloacetate and not from the malate c ycle.Tw o additional fr a gments of serine wer e e v aluated: the bac kbone fr a gment (BB) corr esponding to the amino acid backbone of serine containing the C1 and C2 carbon atom only and the decarboxylated fr a gment (DC) whic h corr esponds to the decarboxylated serine molecule containing the C2 and C3 carbon atom.The two fr a gments sho w ed significantl y differ ent labeling patterns.BB had a m uc h higher fraction of the fully labeled isotopologue M2 compared to DC which sho w ed a high fraction of M1, i.e. one labeled carbon atom.This finding indicates that especially at the beginning of the cultivation the methylene group of M-THF, which is used for the synthesis of serine from glycine, had a high 12 C content.At later stages of the cultivation, the fraction of the M2 isotopologue of the DC fr a gment incr eased indicating that labeled glycine is used for M-THF synthesis.
In the str ain ov er expr essing CYB2 the labeling pattern of the metabolites was similar but in general a higher incor por ation of 13 C into the metabolites was observed (Fig. 4 B).Already after 3 hours of cultivation on 13 C glycolate, glycine and serine were nearl y full y labeled.The 13 C content was reduced at the next sampling time point (24 h), because nearly all 13 C glycolate was consumed and mor e 12 C gl ycolate pr oduced fr om RuBisCO was present ( Fig. S2 ).In 2-PG a higher incorporation of 13 C was observ ed compar ed to 3-PG as well as gl ycer ate, whic h sho w ed similar labeling patterns.Again ribose 5-phosphate, the CBB intermediate e v aluated her e, sho w ed onl y a small fr action of 13 C labeled isotopologues.In addition, high 13 C contents can be found in the TCA metabolites malate and fumarate as well as in aspartate and glutamate, deriving from oxaloacetate.
In contrast to the other two strains, the strain harboring the deletion of CYB2 resulted in only minor incorporation of 13 C into glycine and serine (Fig. 4 C).All other analyzed metabolites did not show any significant incor por ation of 13 C-labeled carbon.As glycine and serine are weakly labeled in this strain, we hypothesized an alternative gene to CYB2 catalyzing the same reaction to be present.One candidate gene was DLD1 which shows similarities to C. necator gl ycer ate dehydr ogenase being r esponsible for the oxidation of glycolate to gly oxylate.Ho w ever, an overexpression of this gene did not result in any significant glycolate consumption ( Fig. S3 ).
The tracer experiments confirmed that CYB2 plays a major role in the recycling of 2-phosphoglycolate produced by the oxygenation reaction of RuBisCO.Cyb2 is annotated as an L-( + )-lactatecytoc hr ome c oxidor eductase.Giv en the structural similarity of L-lactate and glycolate, the results can be interpreted such that Cyb2 also oxidizes glycolate to glyoxylate, taking over the role of glycolate oxidase and transferring the electrons to cytochrome c (Cunane et al. 2002 ).To e v aluate if the natural activity of Cyb2 could lead to a bottleneck in phosphoglycolate recycling and ther efor e may limit the autotrophic growth rates, we cultivated CYB2 ov er expr ession str ains using a medium and weak promoter under autotrophic conditions.We included also a cytosolic version of this protein to test if cytosolic location helps to boost the gr owth r ate.None of the tested constructs could incr ease gr owth under autotrophic conditions showing the native expression levels of CYB2 is sufficient for the cells to recycle the produced glycolate (Fig. 5 ).Deletion of the mitochondrial signal induced more str ess and r educed gr owth r ate e v en mor e in the cytosolic version of CYB2 .

The CYB2 deletion strain can serve as RuBisCO oxygenation test platform
Deletion of CYB2 abolished the conversion of glycolate into glycine and serine and instead led to secretion of glycolate.We hypothesized that cyb2 K. phaffii strains with an intact CBB cycle could be used as RuBisCO oxygenation biosensor strains where the levels of glycolate secretion per biomass (glycolate/DCW) would correlate with the levels of oxygenation reaction of the chosen Ru-BisCO enzymes.Four different RuBisCO proteins were tested in the autotrophic CYB2 deletion strain and growth as well as glycolate production were monitored (Fig. 6 ).The type II RuBisCO proteins of T. denitrificans and Hydrogenovibrio marinus resulted in fastest growth and similar glycolate production.The RuBisCO protein of Gallionella sp. could onl y bar el y facilitate gr owth on CO 2 in our synthetic autotrophic strain.In addition, only small amounts of glycolate were produced.Switching to the type I RuBisCO of C. necator, which was tested together with the co-expression of its corr esponding RuBisCO activ ase CBBX r esulted in the lo w est glycolate production per biomass (Fig. 6 B). C. necator RuBisCO led, ho w e v er, to a lo w er gro wth rate than the best performing Ru-BisCO variants .T he lo w er specific glycolate production supports the concept that the cyb2 strain can serve as an indicator for the specificity of each RuBisCO (Table 1 ).Here, it should be noted that a higher specificity does not necessaril y r eflect a mor e efficient growth on CO 2 , but rather a lo w er sensitivity of the CBB cycle to oxygen.
Also, a PHO13 knoc k out str ain was included in this series of experiments .T his gene was the best BLAST hit of K. phaffii genes against Arabidopsis thaliana PGLP1, a 2-phosphoglycolate phosphatase .T he deletion of this potential phosphoglycolate phosphatase resulted in a reduced growth rate compared to its par ental str ain.Gl ycolate pr oduction was also r educed, but both wer e not completel y abolished, indicating that mor e than one phosphatase is acting on 2-phosphoglycolate in K. phaffii .

Str ain gener a tion
The host strain used in this study was the synthetic autotroph K. phaffii strain published by Gassler et al. (Gassler et al. 2020 ).The assembly of all DNA repair templates and over-expression cassettes were done using the Golden Pi CS cloning system (Prielhofer et al. 2017 ) and transformed using the CRISPR-Cas9 system (Gassler et al. 2019 ).Switching of the RuBisCO protein was done by deleting first the RuBisCO protein of the parental strain follo w ed by the integration of the new RuBisCO protein.Successfully transformed clones were verified by colony PCR.All strains used in this work are given in Table 2 .

Shake flask cultiv a tions
All strains were tested using shake flask cultures.As a first step a pr ecultur e on YPG (yeast extract 10 g L −1 , soy peptone 20 g L −1 , Figure 6.Testing the capabilities of the deletion of CYB2 as oxygenation biosensor.Autotrophic K. phaffii strains using 4 different RuBisCO proteins, and a pho13 variant strain of the T. denitrificans RuBisCO strain were tested for (A) growth and (B) the ratio of glycolate to dry cell weight at the last time point.Cultiv ations wer e performed at 30 • C and 5% CO 2 in the atmospher e .Solid lines indicate the means , and shades (for A) and error bars (for B) the standard deviation of 3 biological replicates .T he gycolate/biomass yield of Gallionella sp.RuBisCO might be inter pr eted car efull y as the r espectiv e K. phaffii strain has a severe growth impairment on CO 2 .gl ycer ol 20 g L −1 ) was performed overnight at 30 • C and 180 rpm.Afterw ar ds optical density was measured at 600 nm and the number of cells needed to inoculate the main culture (OD of 1, 4 or 20, r espectiv el y) was harvested, washed twice with water and resuspended in phosphate buffered Yeast Nitrogen Base Without Amino Acids (YNB) (3.4 g L −1 , pH 6, 10 g L −1 (NH 4 ) 2 SO 4 as nitrogen source and 0.5% (vol/vol) methanol at the beginning as energy source).For the experiments using glycolate as carbon source ad-ditionally 0.5 g L −1 (for the pre-tests) or 1 g L −1 glycolic acid were added.All experiments using glycolate as carbon source were incubated at 30 • C, 180 rpm and ambient CO 2 concentrations.All experiments where cells w ere gro wn autotrophically w ere incubated at 30 • C, 180 rpm, and 5% CO 2 in the atmosphere.For all experiments, methanol concentr ations wer e measur ed on the next day and adjusted to 1%.Afterw ar ds samples w ere taken regularly and optical density, methanol and glycolic acid concentrations wer e measur ed.Water was added to corr ect the cultur e volume for e v a por ation.

HPLC measurements
HPLC measurements to determine methanol and glycolic acid concentr ations, wer e done according to an already published w orkflo w (Baumschabl et al. 2022 ).

Dry cell weight determination
Glass vials were incubated at 105 • C for at least 24 hours, cooled down in the desiccator and weighed.Ten mL of culture were harvested b y centrifugation, w ashed twice with w ater, tr ansferr ed into the pre-weighed glass vials and incubated at 105 • C until the cells wer e full y dried.After cooling down in the desiccator, the glass vials including the dried cells were weighed again and the dry cell weight was calculated.

Labeling experiments
To determine the route of the phosphogl ycolate salv a ge pathway, 13 C labeling experiments were performed.The protocol was similar to the other shake flask cultivations using a starting OD of 20 but 13 C labeled glycolic acid was used (Merck product no.604011) for all 13 C cultures and unlabeled glycolic acid with a natural isotopologe distribution for all nat C cultures.Samples were taken at 2, 24, and 48 hours: 500 μL for the determination of optical density, the r emaining gl ycolic acid as well as methanol concentration, and 3 mL for metabolic sampling & GC-TOFMS isotopologue analysis.

Metabolic sampling & GC-TOFMS isotopologue analysis
The metabolomics w orkflo w w as performed as described b y Mitic et al. ( 2023 ) based on the method of Mairinger et al. 2015 .
In brief, the cell suspension was r a pidl y quenc hed in a 4-fold volume of 60% methanol, 125 mmol L −1 TRIS-HCl, 55 mmol L −1 NaCl at pH 8.2 and -27 • C (Mattanovich et al. 2017 ) and filtered through a cellulose acetate filter after vortexing the mixture for 4 s.After washing the cells on the filter with 60% methanol, the filters wer e stor ed at -70 • C. The consecutive boiling ethanol extraction was performed with 4 mL 75% ethanol at 85 • C.After centrifugation the supernatants with the extracted metabolites were e v a por ated until complete dryness with a vacuum centrifuge befor e r econstitution in 1 mL H 2 O.
For the GC-T OFMS measurements , two methods were applied to cover all needed metabolites in the linear range of the instrument.Automated just-in-time derivatization prior to sample injection was employed to stabilize metabolites and reduce their boiling point.Ethoximation follo w ed b y trimethylsilylation was combined with splitless injection and chemical ionization for the analysis of the phosphorylated metabolites and sugar compounds as well as other intracellular metabolites of low abundance (Mairinger et al. 2015 ).Tert -butylsilylation was combined with 1:50 split injection and electron ionization for the analysis of organic acids and amino acids, a method which offers some positional information due to specific fr a gmentation patterns of the amino acids (Zamboni et al. 2009 ).The methods used for specific metabolites and samples are listed in Supplementary information Table S1 .
For the e v aluation of 13 C incor por ation into the metabolites isotopologue distribution analysis was conducted.The extracted ion c hr omatogr ams of all carbon isotopologues of a target analyte wer e integr ated (e v aluated mass/c har ge r atios listed in (Mitic et al. 2023 )).The peak areas were corrected with the software ICT correction toolbox v.0.04 for the contribution of other heavy isotopes except 13 C, as well as the contribution of the natural 13 C abundance of the derivatization agent.The carbon isotopologue fr actions wer e calculated as follows: Carbo n iso to po log ue f ractio n (1) n = number of carbon atoms in the metabolite, A i = ICT corrected peak area of isotopologue i, i.e .an isotopologue containing i numbers of 13 C atoms For some metabolites such as the amino acids multiple fragments and/or adducts of both measurement methods display the carbon distribution.For data e v aluation, the fr a gment or adduct was chosen based on the trueness of the nat C control sample of the r espectiv e sequence (c hosen method/fr a gment/adduct listed in Supplementary information Table S1 ).As the labeling experiments were conducted in biological duplicates , the displa yed error bars stem from the standard deviation of duplicates multiplied with the correction factor 1.253314 to compensate for the small sample size (Roesslein et al. 2007 ).For the nat C control samples av er a ge and standard deviation of all data are displayed.

Bioreactor cultiv a tions
Bior eactor cultiv ations wer e performed using 1.0 L DASGIP reactors (Eppendorf).Fermentations were performed using YNB with 10 g L −1 (NH 4 ) 2 SO 4 as the nitrogen source, buffered at pH 6 using 100 mmol L −1 phosphate buffer, 0.5% (v ol/v ol) methanol at the beginning as energy source, at 30 • C and constant stirrer speed of 300 r pm.Differ ent oxygen concentr ations of 5, 10, 15 and 20%, r espectiv el y in the inlet gas flo w w ere used to vary the available amount of oxygen, and CO 2 concentrations were k e pt constant at 5%.
Fr om an ov ernight YPG pr ecultur e the amount of cells to inoculate 500 mL of culture with an OD of 1 wer e harv ested, washed twice with water and tr ansferr ed into the bioreactors .T he first sample was taken on the next day and methanol concentrations were adjusted to 1% (vol/vol).Afterwards samples were taken daily, and optical density and the methanol concentrations were measur ed.Ev ery other day the methanol concentr ations wer e adjusted to 1% (v ol/v ol).

Discussion
Pr e viousl y w e w er e able to engineer the methylotr ophic yeast K. phaffii to use CO 2 as its sole carbon sour ce b y the integration of the CBB cycle .T he carbon fixation step in this cycle is catalyzed by the RuBisCO enzyme .T his enzyme also tends to perform the oxygenation reaction resulting in a loss of carbon fixation efficiency.In this study we c har acterized our synthetic autotr ophic str ains in terms of oxygen tolerance and tried to identify the pathway used for phosphoglycolate regeneration in this yeast strain.
Varying the oxygen concentration in the inlet air sho w ed a significant influence on the gr owth r ate of the autotrophic yeast str ains, indicating an optim um dissolv ed oxygen concentr ation of around 15% (Fig. 2 C and Supplementary Figure 1 C).Reducing the oxygen le v els helped the cells to gr ow faster, pr obabl y because of a reduction in the oxygenation reaction of RuBisCO.Howe v er, at least 5% oxygen in the inlet air are needed for efficient growth.Otherwise not enough oxygen is available for sufficient methanol oxidation providing energy for the cells, resulting in reduced gr owth r ates as seen for the cultivations at 2.5% oxygen (Fig. 2 ).In the faster growing engineered version of this strain the optimal oxygen concentration is increased, probably due to the faster accumulation of biomass and ther efor e higher ener gy demand of the cells.A similar effect of a reduction of the oxygen concentration is also reported in microalgae (Kazbar et al. 2019 ), whereas in contrast most plants do not grow better under reduced oxygen conditions (Tisserat et al. 2002 ).
To get a deeper understanding of the pathways used to recycle the 2-phosphoglycolate produced via the oxygenation reaction of RuBisCO, a 13 C tracer experiment using fully labeled glycolate was performed.The metabolites which sho w ed early incorporation of 13 C atoms were glycine and serine .T his indicates that most of the phosphoglycolate is recycled and not fully oxidized to CO 2 .The further flux of the label is pr obabl y via deamination of serine to pyruv ate, whic h is the br anc h point wher e the label enters the TC A cycle .Evaluation of the labeled metabolites led to the conclusion that glycolate is metabolized through pyruvate-PEP-2-PG and 3-PG, the latter reentering the CBB cycle (Fig. 4 D).The CYB2 ov er expr ession str ain sho w ed a significantly higher incorporation of label into 2-PG compared to 3-PG and glycerate supporting the proposed route and dismissing the route present in plants via serine-hydr oxypyruv ate and gl ycer ate.CYB2 also serv es as the major responsible gene for the oxidation of glycolate to glyoxylate, as the deletion of it blocked the phosphoglycolate recycling pathway almost completely and led to the secretion of glycolate, while in the native and overexpression strains no glycolate was secreted.These results indicate that native CYB2 expression is already sufficient to enable a phosphogl ycolate salv a ge pathway rate allowing efficient growth under autotrophic conditions.
As the deletion of CYB2 leads to higher glycolate secretion, we tested whether a cyb2 strain can be used as a sensor strain for the oxygenation rate.We ov er expr essed differ ent RuBisCO pr oteins and calculated the glycolate/biomass yield as an indicator for the specificity of RuBisCO for CO 2 over oxygen.As expected, the choice of RuBisCO is a k e y factor for efficient autotrophic growth in the synthetic autotrophic K. phaffii .In general, the type II RuBisCO proteins led to faster growth compared to the type I pr otein tested.Onl y the type II pr otein fr om a Gallionella sp., for which the highest turnover number so far was reported (Davidi et al. 2020 ), was not able to facilitate growth on CO 2 in our synthetic autotrophic strain.Since the strain using this RuBisCO protein bar el y gr ows, the r esulting gl ycolate pr oduction has to be taken with a grain of salt.Overall, the oxygenation test platform clearl y r e v ealed the differ ence of specificity between type I and II RuBisCO proteins.Both type II proteins enabling growth on CO 2 resulted in an approximately 5 times higher glycolate production compared to the type I protein tested.This fits well to specificity data measured using in vitro assays (Table 1 ).Testing different RuBisCO proteins with varying oxygen concentrations might giv e mor e information about the robustness of the biosensor strain as well as the relationship between specificity and glycolate yield.
Similar to plants the recycling of 2-phosphoglycolate in the synthetic autotrophic K. phaffii releases 0.5 mol of previously captured carbon in the form of CO 2 and additionally 0.5 mol ammonia per mol 2-phosphoglycolate.Both lead to a loss of invested energy.In addition, at least one molecule of ATP is needed during the r ecycling pr ocess in the step fr om pyruv ate to PEP.Ho w e v er, the route via pyruvate and PEP is slightly more energy efficient compared to the hydroxypyruvate route since only one ATP contrary to one NADH has to be invested.In plants, the introduction of the gl ycer ate pathway (K ebeish et al. 2007 ) or the malate cycle (Maier et al. 2012, South et al. 2019 ) into c hlor oplasts could impr ov e the crop yields, making these pathways interesting engineering targets to improve the growth rate of the synthetic autotrophic K. phaffii.
With this work we could pr ov e that balancing the available amount of oxygen can help to impr ov e gr owth of the synthetic autotrophic K. phaffii .We were also able to solve the open question of how synthetic autotrophs can deal with the by-product of the oxygenation reaction and show how versatile a cell's metabolism can be in responding to ne wl y formed substances and their toxic effects, such as 2-phosphoglycolate from the RuBisCO side reaction.As a next step, other more efficient glycolate salvage pathways like the gl ycer ate pathway could be integrated into our strains to e v aluate if they can further impr ov e autotr ophic gr owth r ates.

Figure 4 .
Figure 4. Carbon isotopologue distribution of intracellular metabolites resulting from tracer experiments using fully labeled 13 C glycolate as carbon source to r e v eal the phosphoglycolate salvage pathway in the synthetic autotroph K. phaffii strain.Results of (A) parental strain, (B) CYB2 medium ov er expr ession str ain and (C) CYB2 deletion str ain.M0 to M5 corr esponds to the number of 13 C atoms pr esent in the metabolite .T he numbers in the blue bars indicate the isotopologue fraction of the M0 isotopologue of the r espectiv e sample .T he n umber of biological re plicates of the nat C samples is indicated by 'n = ' in the bars; 2 biological replicates were performed for all 13 C labeled samples.Error bars represent the standard deviation.Serine BB denotes the amino acids backbone fragment including only the C1 and C2 carbon atom of serine.Serine DC denotes the decarboxylated molecule with the C1 carbon atom being cleaved off.For the parental strain all three, for the ov er expr ession str ain the first two and for the deletion str ain the last two time points were measured.Data of the missing bars could not be evaluated as they did not match the quality criteria.(D) Phosphoglycolate salv a ge pathway in the synthetic autotrophic K. phaffii strain.Red arrows indicate the proposed pathway based on the measured metabolites' carbon isotopologue distribution.Abbr e viations: PEP: phosphoenolpyruv ate, 2-PG: 2-phosphogl ycer ate, 3-PG: 3-phosphogl ycer ate , R5P: ribose 5-phosphate , Ru1,5BP: ribulose 1,5-bisphosphate, CBB: Calvin-Benson-Bassham cycle, M-THF: methylene tetr ahydr ofolate , TC A: tricarboxylic acid cycle, CoA: coenzyme A.

Figure 5 .
Figure 5. Cultivation of the CYB2 ov er expr ession str ains compar ed to their parental strain to test whether the natural expression level of CYB2 is a bottleneck for autotrophic growth.Cultivations were performed at 30 • C and 5% CO 2 concentrations in the atmosphere.Solid lines indicate the means, shades the standard deviation of the biological replicates.

Table 1 .
Ov ervie w of kinetic parameters of the RuBisCO proteins used in this study (measured using in vitr o assays).K M : Mic haelis Menten constant for CO 2 .k cat : turnover number for CO 2 .S c/o : CO 2 : O 2 specificity factor.

Table 2 .
Ov ervie w of the strains used in this study.