Mitochondrial membrane transporters as attractive targets for the fermentative production of succinic acid from glycerol in Saccharomyces cerevisiae

Abstract Previously, we reported an engineered Saccharomyces cerevisiae CEN.PK113-1A derivative able to produce succinic acid (SA) from glycerol with net CO2 fixation. Apart from an engineered glycerol utilization pathway that generates NADH, the strain was equipped with the NADH-dependent reductive branch of the TCA cycle (rTCA) and a heterologous SA exporter. However, the results indicated that a significant amount of carbon still entered the CO2-releasing oxidative TCA cycle. The current study aimed to tune down the flux through the oxidative TCA cycle by targeting the mitochondrial uptake of pyruvate and cytosolic intermediates of the rTCA pathway, as well as the succinate dehydrogenase complex. Thus, we tested the effects of deletions of MPC1, MPC3, OAC1, DIC1, SFC1, and SDH1 on SA production. The highest improvement was achieved by the combined deletion of MPC3 and SDH1. The respective strain produced up to 45.5 g/L of SA, reached a maximum SA yield of 0.66 gSA/gglycerol, and accumulated the lowest amounts of byproducts when cultivated in shake-flasks. Based on the obtained data, we consider a further reduction of mitochondrial import of pyruvate and rTCA intermediates highly attractive. Moreover, the approaches presented in the current study might also be valuable for improving SA production when sugars (instead of glycerol) are the source of carbon.


Introduction
Succinic acid (SA) is an industrially relevant compound that can serve as a precursor in the synthesis of numerous speciality and commodity chemicals such as 1,4-butanediol, gammabutyr olactone, and tetr ahydr ofur an (Wer p y and P etersen 2004 ).Mor eov er, SA is utilized as a precursor in biopolymer production and as an ad diti ve in food and pharmaceutical products (Ahn et al. 2016 ).Curr entl y, SA is mostl y pr oduced fr om non-r ene wable fossil fuel deri vati ves, namely n-butane (Cok et al. 2014 ).Ho w ever, being an important constituent of cellular metabolism, SA can also be pr oduced fr om r ene wable carbon via microbial fermentation (Ahn et al. 2016 ).
The curr ent biotec hnological SA pr oduction pr ocesses r el y on fungal and bacterial cell factories (Ahn et al. 2016 ).In comparison to bacterial processes, the usage of fungal cell factories is advantageous because the latter can produce SA at low extracellular pH, thus enabling a m uc h simpler and c hea per pr oduct purification (Jansen andvan Gulik 2014 , Mancini et al. 2022 ).Among acid-tolerant fungal organisms, the yeast Saccharomyces cerevisiae is especially attractive because it allows extensive metabolic engineering endeavours aiming at maximum yield, productivity, and titre (Abbott et al. 2009 ).
Among the microbial metabolic pathways that can generate SA, the r eductiv e br anc h of the TC A cycle (rTC A pathwa y) is more attr activ e than the oxidative branch of the TCA cycle or the glyoxylate cycle because it allows net CO 2 fixation (e.g. via pyruvate carboxylation; Fig. 1 ) and thus the highest theoretical SA yield (Ahn et al. 2016 ).T he rTC A pathwa y r equir es 2 c ytosolic N ADH molecules to generate 1 SA molecule from 1 molecule of pyruvate (Fig. 1 ).This implies that the usage of glycerol and CO 2 as carbon sources enables a r edox-neutr al SA pr oduction with a 16.67% higher maxim um theor etical SA yield compar ed to the use of glucose and CO 2 (1.33 Cmol/Cmol compared to 1.14 Cmol/Cmol, respectiv el y) (Malubhoy et al. 2022 ).Mor eov er, twice the amount of CO 2 can be fixed per mol of SA when gl ycer ol is used as the substrate (Steiger et al. 2017 ).
Notabl y, man y wild-type S. cerevisiae strains, including the CEN.PK family, do not utilize glycerol as a sole carbon source in synthetic medium (Xiberras et al. 2019 ).The growth defect of CEN.PK could be overcome by ada ptiv e labor atory e volution and r e v erse engineering, as pr e viousl y demonstr ated (Oc hoa-Estopier et al. 2011, Ho et al. 2017 ).In fact, the r espectiv e CEN.PK str ain onl y r equir ed single point mutations in two genes ( GUT1 and UBR2 ) in order to become a gl ycer ol-utilizing str ain with a moder ate maximal gr owth r ate in synthetic gl ycer ol media (Ho et al. 2017 ).Using a different strain background, Klein et al. ( 2016 ) demonstrated that the expression of a heterologous aquaglyceroporin (e.g.Fps1 from Cyberlindnera jadinii ) also had a positive effect on gl ycer ol utilization in S. cerevisiae .
in S. cerevisiae is r espir atory.In fact, the endogenous gl ycer ol catabolic pathw ay (L-G3P pathw ay) in S. cerevisiae transfers a part of the electrons from glycerol to FADH 2 and subsequently to oxygen via the r espir atory c hain.By abolishing the FAD-dependent L-G3P pathway and establishing a heterologous so-called DHA pathway, the r espectiv e electr ons can be tr a pped in c ytosolic N ADH molecules (Klein et al. 2016 ).This pathway replacement is a k e y pr er equisite for r edox-neutr al SA pr oduction (Fig. 1 ).Efficient SA pr oduction (r eac hing up to 47% of the maxim um theor etical yield) fr om gl ycer ol and CO 2 has been ac hie v ed by equipping a S. cerevisiae DHA pathway deri vati ve of CEN.PK113-1A (also carrying a deleted GUT1 gene, a functional UBR2 allele, and the Fps1 from C. jadinii ) with a heterologous rTC A pathwa y and a heterologous SA exporter (Xiberras et al. 2020, Malubhoy et al. 2022 ).The bestperforming strain has also been demonstrated to achieve a slight net fixation of CO 2 during the phase of SA production (Malubhoy et al. 2022 ).
Even so, the results presented in the study of Malubhoy et al. ( 2022 ) indicated that a significant amount of carbon still entered the CO 2 -releasing o xidati ve TCA cycle affecting the total CO 2 net balance during the fermentation process in an unwanted manner.It is clear that oxTCA is necessary for ATP generation and the supply of precursors for biomass formation.Ho w ever, one might wonder whether the activity of this pathway exceeded the le v el r equir ed for maintenance and gr owth.It m ust be noted in this context that S. cerevisiae natur all y upr egulates its r espir atory metabolism when gl ycer ol is the sole carbon source (Xiberras et al. 2019 ).
During r espir atory metabolism, a part of the carbon is assumed to enter the mitochondria in the form of pyruvate, which is converted to acetyl-CoA by the pyruvate dehydrogenase complex (Fig. 1 ).Mor eov er, it is expected that carbon enters the mitochondria in the form of c ytosolic rTCA pathw ay intermediates such as oxaloacetate , malate , fumarate , and e v en succinate.In fact, these compounds are assumed to enter the mitochondria via different membr ane tr ansporters (see below).Inside the mitoc hondria, the additional C 4 compounds can either help in anabolic reactions and/or e v entuall y r esult in the formation of pyruvate from malate via the malic enzyme (Fig. 1 ), which eventually leads to the complete oxidation into CO 2 via the PDH complex and the o xidati ve br anc h of the TC A cycle .In that r egard, se v er al tr ansporters located in the inner mitoc hondrial membr ane (P almieri et al. 2006 , Ferramosca andZara 2021 ) represent promising targets in order to fine-tune the flux of carbon into the mitochondria.
The mitochondrial pyruvate carrier (MPC) complex is responsible for the uptake of cytosolic pyruvate into the mitochondria of S. cerevisiae , where it is dissimilated by the pyruvate dehydrogenase complex and the o xidati ve TCA cycle for energy production or utilized as a precursor in the biosynthesis of branched-chain amino acids (Herzig et al. 2012 ) (Fig. 1 ).Two functional variants of the MPC complex exist (Bender et al. 2015 ).The MPC FERM is a heterodimer formed by the Mpc1 and Mpc2 subunits responsible for pyruvate uptake during fermentative growth.The MPC OX is a heterodimer formed by the Mpc1 and Mpc3 subunits, and it is responsible for pyruvate uptake during growth on non-fermentable carbon sources.Among the two variants, the MPC OX has been shown to have a significantly higher pyruvate uptake activity, thereby promoting the dissimilation of pyruvate via the o xidati ve TCA cycle under the r espectiv e conditions (Bender et al. 2015 ).T hus , the MPC3 gene is considered to be a promising deletion target with the goal of increasing the supply of cytosolic pyruvate for SA production when cells grow on glycerol.In the study of Bender et al. ( 2015 ), the deletion of the MPC1 gene completely abol-ished the mitochondrial pyruvate uptake.Notably, this might not lead to complete growth impairment of our SA-pr oducing str ain equipped with the rTC A pathwa y due to the import of cytosolic C 4 compounds by other mitoc hondrial tr ansporters (as described below) and their potential conversion to mitochondrial pyruvate via the malic enzyme (Mae1) (Boles et al. 1998 ).T hus , MPC1 was also considered a promising target gene.
Sacc harom yces cerevisiae has two mitochondrial transporters which enable the net uptake of rTCA intermediates and SA into the mitochondria-Oac1 and Dic1 (Ferramosca and Zara 2021 ).Oac1 mediates oxaloacetate import by exchanging cytosolic oxaloacetate with mitochondrial sulfate (Palmieri et al. 1999 ), while Dic1 mediates malate and SA import by exchanging cytosolic malate or SA with mitochondrial phosphate (Palmieri et al. 1996 ) (Fig. 1 ).The main physiological role of these two transporters is ana pler otic , i.e .r eplenishing the mitoc hondrial TCA cycle intermediates spent in anabolic reactions (Ferramosca and Zara 2021 ).By deleting one of the two tr ansporters, the mitoc hondrial consumption of cytosolic oxaloacetate , malate , and SA could be dir ectl y pr e v ented.Indir ectl y, this ma y also tune down the o verall activity of the oxidative TCA cycle by limiting the process of ana pler osis.
Sfc1 is another mitochondrial transporter with affinity for rTCA intermediates, primaril y exc hanging cytosolic SA with mitoc hondrial fumarate (Palmieri et al. 1997 ).In contrast to Oac1 and Dic1, the activity of Sfc1 does not lead to w ar ds net mitochondrial consumption of C 4 dicarboxylic acids since it exchanges one C 4 molecule with another.Ho w e v er, in S. cerevisiae strains equipped with the rTC A pathwa y, the activity of Sfc1 in conjunction with cytosolic fumarate reductase and mitochondrial succinate dehydrogenase may lead to w ar ds a wasteful cycle in which the electr ons fr om c ytosolic N ADH ar e tr ansferr ed to mitoc hondrial FAD and end up in the r espir atory c hain (Fig. 1 ), whic h is counter pr oductive for the NADH-dependent cytosolic SA production.
The SDH complex plays an important role in the hypothesized mitochondrial consumption of cytosolic SA via Dic1 and in the hypothesized transfer of electrons into the respiratory chain via Sfc1 (Fig. 1 ).T hus , SDH1 deletion r epr esents an alternativ e a ppr oac h to tackle these issues by disrupting the mitochondrial conversion of SA into fumarate, and it may also serve as a control to better inter pr et the results obtained from the DIC1 and SFC1 deletions.
In this w ork, w e individually deleted the genes MPC1, MPC3, O A C1, DIC1 , and SFC1 encoding the aforementioned mitochondrial transporters as well as the SDH1 gene with an aim to increase SA production via the rTC A pathwa y in S. cerevisiae .We performed shak e-flask culti v ations of the r espectiv e deletion m utants and compared the effects of these deletions on succinic acid production, gl ycer ol consumption, biomass formation, and byproduct formation.

Strains, plasmids, and strain maintenance
The S. cerevisiae strains and plasmids used in this study are listed in Tables 1 and 2 , r espectiv el y.Yeast cells were routinely grown on solid YPD media, which contained yeast extract (10 g/L), peptone (20 g/L), glucose (20 g/L), and agar (20 g/L).Agar plates were cultivated in a static incubator at 30 • C. Media were supplemented with phleom ycin (20 mg/L), hygrom ycin B (300 mg/L), or nourseothricin (100 mg/L) for selection purposes when needed.Escherichia coli DH5 α cells were used for plasmid isolation and were routinely grown in lysogeny broth (LB) containing NaCl (10 g/L), yeast Table 1.Sacc harom yces cerevisiae strains used in this study.

Plasmid name Description Reference
pNatCre Used for the r emov al of loxP-ble-loxP disruption cassettes from the genome.Contains the Cr e-r ecombinase gene under the control of the GAL1 promoter and the nourseothricin resistance marker for selection.

Gueldener et al. ( 2002 ) pAG32
Template for the amplification of hphMX disruption cassettes.Goldstein and McCusker ( 1999 ) pUG74 Template for the amplification of loxP-natMX-loxP disruption cassettes.Hegemann and Heick ( 2011 ) extract (5 g/L), and peptone (10 g/L) adjusted to a pH of 7.5 with 2 M NaOH (Bertani 1951 ).For the selection and maintenance of plasmid-containing cells, ampicillin (100 mg/L) was added.Cultiv ations wer e performed on an orbital shaker at 250 rpm and 37 • C and plasmids were isolated using the GeneJET TM Plasmid Miniprep Kit (Thermo Fisher Scientific, Waltham, MA, USA).

Marker rescue via the Cre-loxP system
To r emov e the phleomycin resistance marker ( ble ) flanked by loxP sites from the genome of S. cerevisiae strains, the respective strains wer e tr ansformed with pNatCr e (Table 2 ) according to the lithium acetate method described by Gietz et al. ( 1995 ).Positive transformants were selected on a solid YPD medium supplemented with nourseothricin.A single colony of the transformed strain was inoculated into 3 mL of YPD medium supplemented with nourseothricin and grown at 30 • C and 200 rpm overnight.Cells were then centrifuged at 800 g for 5 min and washed 3 times in 5 mL of YPGal medium, which contained yeast extract (10 g/L), peptone (20 g/L), and galactose (20 g/L).The obtained pellet was resuspended in 3 mL of YPGal supplemented with nourseothricin (100 mg/L) and incubated at 30 • C and 200 rpm for 6-8 h.Afterw ar ds, an aliquot of the cell suspension (10 μL) was spread on a solid YPD medium and incubated at 30 • C for 2 da ys .Single colonies were streaked in a lawn on solid YPD media with and without phleomycin.In order to verify the loss of the ble marker via diagnostic PCR, genomic DN A w as isolated from those clones that were not capable of growing in the presence of phleomycin.
The pNatCre plasmid was subsequently removed from the verified clone by serial transfers in YPD medium lacking nourseothricin.

Gene deletion
Gene deletions were obtained by using disruption cassettes consisting of the phleomycin or nourseothricin resistance mark ers flank ed by lo xP sites ( loxP-ble-loxP or lo xP-natMX-lo xP, r espectiv el y) or the hygromycin resistance marker ( hphMX ) ( Supplementary Table S1 ).The disruption cassettes were amplified from pUG66, pUG74, or pAG32 plasmids, r espectiv el y, using the primer pairs listed in Supplementary Table S1 .The primers used for amplification contained at their 5' terminal end a 40-60 bp sequence complementary to the region immediately upstr eam or downstr eam of the start or stop codon of the gene to be deleted.Pr epar ativ e PCRs for amplification of disruption cassettes were performed using Phusion ® High-Fidelity DNA Pol ymer ase (Ne w England BioLabs, Fr ankfurt am Main, German y).PCR conditions were adapted to the guidelines of the manufactur er.PCR pr oducts wer e purified by using the GeneJET TM PCR Purification Kit (Thermo Fisher Scientific).Transformation of S. cerevisiae with linear disruption cassettes was performed according to the lithium acetate method described by Gietz et al. ( 1995 ) using 2 μg of DNA for one tr ansformation.Positiv e tr ansformants wer e selected on solid YPD media supplemented with the a ppr opriate antibiotic.

Isolation of genomic DNA from S. cerevisiae transformants and diagnostic PCR
Both corr ect integr ation of a disruption cassette into the genome of S. cerevisiae and successful marker r emov al wer e v erified by diagnostic PCR using OneTaq Quick-load DNA polymerase and buffer according to the manufacturer's guidelines (New England Biolabs).Genomic DNA from S. cerevisiae strains was isolated according to a modified pr otocol fr om Hoffman and Winston ( 1987 ).
Appr oximatel y 50 mg of cells were suspended in 200 μL of TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0).Subsequently, 300 mg of acid-washed glass beads (diameter of 0.425-0.6mm) and 200 μL of phenol:c hlor oform:isoamyl alcohol (25:24:1) wer e added.The tubes w ere v ortexed at maximum speed for 2 min and centrifuged at 15 700 g for 10 min.The aqueous phase (1 μL) was used as a template in 25 μL PCR reactions.Primers for diagnostic PCR were designed to bind within the disruption cassette and/or within the genomic DNA upstream and downstream of the sequence to be deleted.

Media and cultiv a tion conditions for the shak e-flask culti v a tions to test the production of succinic acid from glycerol
Synthetic medium containing glucose (20 g/L) and ammonium sulphate (5 g/L) as the carbon and nitrogen sources, respectiv el y, was used for all pr ecultur es and intermediate cultures.
The synthetic medium was pr epar ed according to Verduyn et al. ( 1992 ), containing 3 g/L KH 2 PO 4 , 0.5 g/L MgSO 4 sterilized vitamins were added.Final vitamin concentrations were: 0.05 mg/L D-( + )-biotin, 1 mg/L D-pantothenic acid hemicalcium salt, 1 mg/L nicotinic acid, 25 mg/L myo-inositol, 1 mg/L thiamine chloride hydrochloride, 1 mg/L pyridoxine hydrochloride, and 0.2 mg/L 4-aminobenzoic acid.The pH of the synthetic glucose medium was adjusted to 6.5 with 4 M KOH.All main experiments for assessing succinic acid production in shake flask batc h cultiv ations wer e performed in a synthetic medium containing 60 mL/L (75.6 g/L) gl ycer ol as the sole carbon source, urea (2.27 g/L) as the nitrogen source, and CaCO 3 (30 g/L), according to Malubhoy et al. ( 2022 ).To pr epar e this latter medium, an appropriate aliquot of a urea stock solution (113.5 g/L) was added to obtain a final concentration of 2.27 g/L.The pH of the CaCO 3buffered synthetic glycerol medium was adjusted to 6.0 with 4 M KOH and the medium was then filter-sterilized.CaCO 3 was added as a sterile powder only after inoculation, as described below.
For pr e-cultiv ation, cells fr om a single colon y wer e used to inoculate 3 mL of the synthetic glucose medium in a 10 mL glass tube and incubated at orbital shaking of 200 rpm and 30 • C overnight.
The pr e-cultur e was used to inoculate 10 mL of the same medium in a 100-mL Erlenmeyer flask (closed with a metal cap), adjusting an OD 600 of 0.2.This cultur e, her eafter r eferr ed to as intermediate culture, was cultivated under the same conditions for 24 h.The a ppr opriate cultur e volume fr om the intermediate cultur e (in order to later adjust an OD 600 of 0.2 in 100 mL of synthetic glycerol medium) was centrifuged at 800 g for 5 min and the supernatant was discarded.The cell pellet was then washed once by re-suspending the cells in the synthetic gl ycer ol medium.The cell suspension was centrifuged again and re-suspended in 100 mL of the same medium, adjusting to a final OD 600 of 0.2.Subsequently, the entire cell suspension was added to a sterile 500-mL Erlenmeyer flask already containing 3 g of sterile CaCO 3 .The Erlenmeyer flasks were closed with cotton plugs .T he main cultures were incubated by orbital shaking of 200 rpm and 30 • C. Samples for OD 600 determination and HPLC analysis were taken at regular time interv als.For OD 600 measur ements, samples wer e diluted in 0.2 M HCl, ensuring complete dissolving of the suspended CaCO 3 .

Metabolite analysis by HPLC
Samples of culture supernatants (1 mL) were first filtered through 0.2 mm Minisart RC membrane filters (Sartorius, Göttingen, Germany) and stored at −20 • C until analysis .T he concentrations of succinic acid, malic acid, gl ycer ol, and ethanol in cultur e media were determined using a Waters HPLC system (Eschborn, Germany) consisting of a binary pump system (Waters 1525), an injector system (Waters 2707), the Waters column heater module WAT038040, a r efr activ e index (RI) detector (Waters 2414), and a dual wavelength absorbance detector (Waters 2487).The samples were loaded on an Aminex HPX-87H cation exchange column (Bio-Rad, München, Germany) coupled to a Micro-guard column (Bio-Rad) and eluted with 5 mM H 2 SO 4 as the mobile phase at a flow rate of 0.6 mL/min and a column temperature of 45 • C. Volumes of 20 μL of sample were used for injection.Succinic acid and malic acid were detected using the dual-wavelength absorbance detector (Waters 2487) while gl ycer ol and ethanol were analysed with the RI detector (Waters 2414).The retention time for malic acid was 9.6 min, for succinic acid 11.9 min, for gl ycer ol 13.5 min, and for ethanol 21.6 min.Data were processed and analysed using the Breeze 2 software (Waters).All reported maximum SA yields were r egister ed at the time points when the HPLC measurements and the r espectiv e yield calculations could be done r eliabl y, i.e. between 96 and 144 h of cultivation.

Results
The deletion of MPC3 , DIC1 , and SDH1 genes improved succinic acid production in the DHA-SAT strain First, we individually deleted MPC3 , MPC1 , O A C1 , DIC1 , SFC1 , and SDH1 in a S. cerevisiae strain previously engineered for SA production from glycerol and CO 2 via the rTC A pathwa y.As the current study was conducted in parallel to the study of Malubhoy et al. ( 2022 ), we started here with the strain DHA-SAT (i.e. the strain without PYC2 ov er expr ession), whic h was first gener ated by the latter authors.All deletion m utants ar e listed in Table 1 and their construction is described in Material and Methods.Shake-flask cultures of strain DHA-SAT and the r espectiv e m utant str ains were performed in a synthetic gl ycer ol medium supplemented with CaCO 3 to analyse the effects of individual gene deletions on SA production.
As shown in Fig. 2 , the mpc3 strain exhibited a 20% increase in maximum SA yield compared to the DHA-SAT strain.The mpc3 strain also displayed a significant reduction in biomass accumulation compared to the baseline str ain.Inter estingl y, the MPC1 deletion did not lead to an impr ov ement in SA production (Fig. 2 ).Howe v er, the MPC1 deletion had a noticeable impact on growth as the mpc1 strain had a slightly extended lag phase and accumulated significantly less biomass than the DHA-SAT strain (Fig. 2 ).
As a next step, we investigated the strains with abolished Oac1 (oxaloacetate carrier) or Dic1 (malate and SA carrier) activity.Deletion of O A C1 led to a slight decrease in the maximum obtained SA titre; ho w ever, the maximum SA yield did not significantl y differ fr om the one obtained by the DHA-SAT strain (Fig. 3 ).The oac1 strain displayed an extended lag phase compared to the DHA-SAT strain, but both strains accumulated equal amounts of biomass by the end of the cultivation (Fig. 3 ).The deletion of DIC1 caused a slight increase in SA production ( ∼6% increase in maxim um SA yield compar ed to the DHA-SAT str ain).Unlike the O A C1 deletion, the deletion of DIC1 did not cause a prolonged lag phase (Fig. 3 ).Additionally, the DIC1 deletion significantly reduced the re-consumption of SA from the medium upon glycerol exhaustion and pr e v ented the formation of malate during this stage of cultiv ation.These r esults indicate that Dic1 indeed participates in the mitochondrial consumption of SA in the DHA-SAT strain, both before and after glycerol exhaustion.
The deletion of SFC1 did not cause a significant change in the phenotype ( Supplementary Fig. S1 ).The sfc1 str ain r eac hed a similar maximum SA yield and accumulated a similar amount of biomass as the DHA-SAT strain.
As expected, the deletion of SDH1 caused a significant impr ov ement in SA production; the sdh1 strain exhibited a ∼17% increase in SA yield over the DHA-SAT strain (Fig. 3 ).In contrast to all other deletion mutants tested, the sdh1 strain produced significantly less malate throughout the cultivation when compared with the DHA-SAT strain (Fig. 3 ).The sdh1 strain also exhibited an extended lag phase and accumulated less biomass than the DHA-SAT strain.

Assessment of the promising gene deletions (mpc3 , dic1 , and sdh1 ) in the DHA-SAT strain overexpressing PYC2
In our pr e vious work conducted by Malubhoy et al. ( 2022 ), we impr ov ed the performance of the DHA-SAT strain (higher SA yield and pr oduction r ate) by incr easing the activity of pyruvate carboxylase in the cytosol, which was ac hie v ed by integr ating an additional expression cassette for PYC2 .As a next step, we wanted to c hec k to what extent the SA pr oduction by the PYC2 oe str ain would benefit from the gene deletions that resulted in SA yield impr ov ements in the strain DHA-SAT (i.e.MPC3 , DIC1 , or SDH1 ).We constructed the r espectiv e deletion mutants and compared them with the isogenic PYC2 oe strain in shake flask cultivations in synthetic gl ycer ol media supplemented with CaCO 3 (Fig. 4 ).
Deletion of the MPC3 gene in the PYC2 oe strain background again had a clear positive impact on SA production, causing a ∼13% increase in the maximum SA yield (Fig. 4 ).The deletion of the MPC3 gene in the PYC2 oe strain negatively affected growth, although not as se v er el y as in the DHA-SAT strain (Fig. 4 ).Evidently, the mpc3 deletion str ain mana ged to dir ect additional pyruv ate into the rTC A pathwa y, despite the fact that the PYC2 oe strain already had a strong flux through this pathway.
The PYC2 oe strain is assumed to have higher concentrations of SA in the cytosol compared to the DHA-SAT strain.For this reason, we expected that the DIC1 deletion would have an e v en higher impact on SA production in the PYC2 oe str ain.In contr ast to our expectations, the DIC1 deletion did not impr ov e SA pr oduction in this strain background (Fig. 4 ).Ho w ever, the DIC1 deletion reduced the re-consumption of SA from the medium upon glycerol exhaustion, while also pr e v enting malate formation during this sta ge of cultiv ation (Fig. 4 ).These latter r esults r esembled the observ ations alr eady made in the DHA-SAT str ain bac kgr ound.
Deletion of the SDH1 gene in the PYC2 oe strain again positiv el y affected SA pr oduction at the expense of malate production (Fig. 4 ).A ∼19% increase in maximum SA yield was achieved in the sdh1 deletion m utant str ain compar ed to the r efer ence PYC2 oe str ain (Fig. 4 ).Mor eov er, the PYC2 oe -sdh1 str ain accum ulated less biomass than the PYC2 oe strain (Fig. 4 ).Ho w ever, this difference in gro wth w as not as pronounced as in the DHA-SAT str ain bac kground.
Lastly, we decided to combine the two most promising deletions ( mpc3 and sdh1 ) in the PYC2 oe strain to test whether they will cause a cum ulativ e effect on SA production.The resulting PYC2 oe -mpc3 sdh1 strain produced 45.5 g/L of SA (at 144 h) and r eac hed a maxim um SA yield of 0.66 g SA /g gl ycer ol (at 96 h).The yield impr ov ement corr esponded to a ∼27% increase over the PYC2 oe str ain (Fig. 5 ).Mor eov er, the PYC2 oe -mpc3 sdh1 str ain accum ulated the lo w est amounts of malate and biomass in comparison with the other isogenic strains tested.

Discussion
In biotec hnological SA pr oduction using micr obes , the rTC A pathway is the pr eferr ed metabolic r oute in order to maximize SA yield by fixing CO 2 (Ahn et al. 2016 ).To minimize production costs, low external pH is highl y favour able, and r obust fungal or ganisms such as S. cerevisiae and Pichia kudriavzevii represent highly suitable cell factories (Abbott et al. 2009, Tran et al. 2023 ).Due to ATP constraints, we assume that homofermentative SA production (fr om gl ycer ol) cannot be run anaer obicall y, whic h is in contrast to the w ell-kno wn ethanol production (from glucose) in yeast.Based on the following considerations, we hypothesize that a portion of the substr ate m ust be r espir ed in the mitochondria to provide the cells with sufficient ATP for SA export, cell growth, and maintenance.In fact, the net ATP yield for the fermentative SA formation fr om gl ycer ol (or glucose) is zer o, particularl y when the CO 2 fixation is carried out by the ATP-consuming pyruvate carboxylase (Fig. 1 ).On top of that, the export of SA r equir es energy to be thermodynamically feasible under industrially relevant conditions , i.e .low pH and high extr acellular pr oduct concentr ation (de Kok et al. 2012 ).In this context, the facilitated export of Figur e 2. T he SA-o v er pr oducing S. cerevisiae str ain DHA-SAT as well as the isogenic deletion str ains mpc3 and mpc1 cultiv ated in synthetic gl ycer ol medium using urea as the nitrogen source and buffered with 30 g/L of CaCO 3 (see composition in Materials and Methods).The cultivations were performed in 500-mL shake flasks filled with 100-mL medium.The initial pH of the medium was 6.0, prior to CaCO 3 addition.HPLC analysis was used to determine the concentrations of succinic acid, malic acid, and gl ycer ol in the culture supernatant.Biomass accumulation was recorded by optical density measurements at 600 nm (OD 600 ).Mean values and standard deviations (SD) were determined from three biological replicates.divalent SA anions is considered to be the least energetically expensive type of export thermodynamically feasible in such conditions (de Kok et al. 2012, Renduli ć et al. 2022 ).Ho w e v er, it r equir es the concurrent export of two protons via H + -ATP pumps to maintain cellular pH and c har ge homeostasis, thus costing two ATP per exported SA molecule (de Kok et al. 2012 ).For these reasons, efficient fungal-based SA production processes (including commercial pr oduction) ar e carried out in aer obic conditions (Abbott et al. 2009, Jansen et al. 2013, Ahn et al. 2016 ) to be able to generate the r equir ed ATP via mitochondrial respiration.
Although this reasoning makes clear that a certain activity of the o xidati v e TCA cycle is r equir ed for ATP pr oduction, it is questionable whether the flux through this pathway in our pr e viousl y constructed SA producers, DHA-SAT and PYC2 oe , exceeds the requir ed le v el.Compar ed to the situation on glucose, the activity of the r espir atory metabolism (i.e .the oxTC A cycle and the respir atory c hain) is indeed highl y upr egulated during the gr owth on r espir atory carbon sources suc h as gl ycer ol, as described by Xiberras et al. ( 2019 ).T hus , the mitochondrial activity under nonr epr essiv e conditions might be far too high, resulting in suboptimal SA yields.
Consistent with our assumptions, the deletion of MPC3 had a positive effect on SA yield (Figs 2 and 4 ) by reducing the influx of pyruvate into the mitochondria.As both SA-producing S. cerevisiae MPC3 deletion strains still displayed considerable biomass formation, our results suggest that an even stronger reduction in mitoc hondrial pyruv ate import could be sustainable, potentiall y r esulting in an e v en higher SA yield.The inter pr etation of our results obtained with the mpc1 mutant is rather puzzling.According to the liter atur e, the lac k of MPC1 is expected to cause the str ongest reduction in mitochondrial uptake of pyruvate because neither of the two active carrier complexes (MPC OX nor MPC FERM ) can be formed without the Mpc1 subunit (Bender et al. 2015 ).Our mpc1 mutant could grow in a synthetic glycerol medium, showing that it ob viousl y still gener ated sufficient mitoc hondrial pyruv ate to pr oduce ener gy and br anc hed-c hain amino acids, most likel y via the net uptake of rTCA intermediates coupled with the malic enzyme activity (Fig. 1 ).The slo w gro wth of the mpc1 strain during the early phase of cultivation (Fig. 2 ) and the increased levels of ethanol r egister ed at 72 h ( Supplementary Fig. S2 ) suggest that the cells indeed experienced a lack of mitochondrial pyruvate uptake as well as an accumulation of pyruvate in the cytosol.T hus , it Figur e 3. T he SA-o v er pr oducing S. cerevisiae strain DHA-SAT and the isogenic deletion mutants dic1 , oac1 , and sdh1 cultivated in synthetic gl ycer ol medium using urea as the nitrogen source and buffered with 30 g/L of CaCO 3 (see composition in Materials and Methods).The cultivations were performed in 500-mL shake flasks filled with 100-mL medium.The initial pH of the medium was 6.0, prior to CaCO 3 addition.HPLC analysis was used to determine the concentrations of succinic acid, malic acid, and gl ycer ol in the culture supernatant.Biomass accumulation was recorded by measuring optical density at 600 nm (OD 600 ).Mean values and SDs were determined from three biological replicates.was surprising that there was no increase in SA production compared to the baseline DHA-SAT strain.Perhaps the complete lack of mitochondrial pyruvate uptake triggered a strong rewiring of the central carbon metabolism to sustain gr owth, whic h consequently nullified the increased supply of cytosolic pyruvate for SA pr oduction.Hypotheticall y, a str ong upr egulation of alternative metabolic routes such as the PDH bypass, the glyoxylate cycle, the mitochondrial uptake of rTC A intermediates , and the malic enzyme reaction might have caused such a phenomenon.The remaining question would be why such a r e wiring did not take place in our mpc3 str ain.Ov er all, mor e r esearc h is r equir ed to better understand the interplay between the mitochondrial transport of pyruvate (or lack thereof) and the regulatory mechanisms which control the central carbon metabolism in glycerol-grown and SAov er pr oducing S. cerevisiae cells.
By deleting O A C1 or DIC1 , the current study attempted to limit the import of cytosolic rTCA intermediates into the mitochondria, thereby limiting the replenishment of the oxTCA cycle with C 4 compounds .Pro vided that the formation of acetyl-CoA is not ratecontrolling for the oxTCA, our a ppr oac h might reduce the amount of available mitochondrial oxaloacetate for the dissimilation of acetyl-CoA and thus the loss of carbon ending up in CO 2 .The observ ed pr olonged la g phase of the oac1 str ain, whic h occurr ed upon transferring the strain from the glucose-based preculture medium into the gl ycer ol-based main culture medium, may indeed be explained in this w ay.Ho w e v er, the oac1 str ain e v entuall y r eac hed similar gr owth r ates as the r efer ence str ain (DHA-SAT) and did not show any improvements in the SA yield.T hus , the activity of the remaining Dic1 transporter could have compensated for the lack of Oac1.In contrast to the deletion of O A C1 , the deletion of DIC1 in the DHA-SAT strain slightly improved the SA yield.The fact that the dic1 deletion strain did not display any impediments in cell growth might indicate that the impr ov ed SA production was the result of a reduced dissimilation of cytosolic malate and SA into CO 2 (via the mitochondrial malic enzyme and the PDH complex) rather than merely a limitation in anaplerotic pathwa ys .
Another inter esting observ ation based on our dic1 mutants (Figs 3 and 4 ) is that Dic1 ob viousl y participates in the consumption of extracellular SA and the buildup of malate in the period after gl ycer ol is consumed.The uptake of extracellular SA into the cytosol upon gl ycer ol exhaustion can be attributed to the Figur e 4. T he SA-o v er pr oducing S. cerevisiae str ain PYC2 oe and the isogenic mpc3 , dic1 , and sdh1 m utant str ains cultiv ated in synthetic gl ycer ol medium using urea as the nitrogen source and buffered with 30 g/L of CaCO 3 (see composition in Materials and Methods).The cultivations were performed in 500-mL shake flasks filled with 100-mL medium.The initial pH of the medium was 6.0, prior to CaCO 3 addition.HPLC analysis was used to determine the concentrations of succinic acid, malic acid, and gl ycer ol in the culture supernatant.Biomass accumulation was recorded by measuring optical density at 600 nm (OD 600 ).Mean values and SDs were determined from three biological replicates.dicarboxylate plasma membrane transporter Dct-02 (Renduli ć et al. 2022 ).Dic1 seems to import cytosolic SA into the mitochondria, where it is oxidized into fumarate via the SDH complex and further converted into malate via the fumarase Fum1.The results further suggest that the obtained mitochondrial malate is exported back into the cytosol (possibly also via Dic1), and subsequently into the medium via Dct-02.The complete process involving the activities of Dct-02, Dic1, the SDH complex, and Fum1 c hannels electr ons via FADH 2 into the r espir atory c hain and thus ma y pro vide the cells with an alternativ e source of ener gy upon depletion of gl ycer ol.
Ov er all, our attempts to limit the import of rTCA intermediates (C 4 ) into the mitochondria did not lead to improvements in SA yield that were as high as the ones obtained by limiting the mitoc hondrial pyruv ate uptake .T his ma y be partially explained by the fact that the standalone deletion of O A C1 can be compensated by DIC1 and vice v ersa.Notabl y, the double deletion of O A C1 and DIC1 was pr e viousl y r eported to disable the growth of S. cerevisiae on non-fermentable carbon sources in synthetic media due to a complete abolishment of ana pler osis (P almieri et al. 1999 ).T hus , future metabolic engineering endea vours ma y focus on simulta-neously tuning down the activities of Oac1 and Dic1, rather than completely abolishing them.
In order to ac hie v e optimal SA yields, one has to make sure that, in addition to carbon, unnecessary loss of electrons in respiration is avoided.One of the objectives of this study was to investigate the existence of a hypothetical succinate-fumarate shuttle (mediated by Sfc1), which was considered to transfer electrons from cytosolic NADH into the mitoc hondrial r espir atory c hain (via SDH).Our results sho w ed that this potential shuttle is not relevant, at least when the genetic constitution of the strain and the conditions used are considered.One reason might be that the fumarate concentr ation in mitoc hondria is not high enough to drive this shuttle.It might also be that other, more obvious, routes of oxidizing cytosolic NADH, such as the external mitochondrial NADH dehydrogenases Nde1 and Nde2 (Bakker et al. 2001 ), lead to a loss of electrons from cytosolic NADH and reduce SA yield.
Disruption of the mitochondrial SDH complex by knocking out one of its subunits r epr esents a common and straightforw ar d strategy to induce SA overproduction via the oxTCA cycle in S. cerevisiae and other yeasts (Raab et al. 2010, Ito et al. 2014, Gao et al. 2016, Xi et al. 2021 ).In strains exhibiting the rTC A pathwa y in Figur e 5. T he SA-o v er pr oducing S. cerevisiae str ain PYC2 oe and the isogenic sdh1 and mpc3 sdh1 m utant str ains cultiv ated in synthetic gl ycer ol medium using urea as the nitrogen source and buffered with 30 g/L of CaCO 3 (see composition in Materials and Methods).The cultivations were performed in 500-mL shake flasks filled with 100-mL medium.The initial pH of the medium prior to CaCO 3 addition was 6.0.HPLC analysis was used to determine the concentrations of succinic acid, malic acid, and gl ycer ol in the culture supernatant.Biomass accumulation was recorded by measuring optical density at 600 nm (OD 600 ).Mean values and SDs were determined from four biological replicates.the cytosol, the deletion of SDH1 can contribute to an increase in SA production in two ways: (i) by blocking the oxidation of mitochondrial SA and leading to a net SA export to the cytosol and, subsequently, to the culture medium, and (ii) by preventing mitochondrial dissimilation of the SA that is generated in the cytosol by the rTC A pathwa y and imported to mitochondria via the transporter Dic1 and/or Sfc1.In the current study, the sdh1 represented a useful control.In fact, the deletion of the two genes that encode mitochondrial succinate transporters ( DIC1 and SFC1 ) caused little to no impr ov ement in SA production (Figs 3 , 4 , and Supplementary Fig. S1 ), while the deletion of SDH1 resulted in a significant impr ov ement.Taken together, these r esults suggest that the additional SA obtained by deleting SDH1 rather originates from the o xidati ve than from the reductive SA production route.Still, this is not consider ed counter pr oductiv e for our efforts to increase the SA production via the CO 2 -fixing rTC A pathwa y because the SDH1 deletion is not supposed to increase the flux of carbon into the mitochondria.
Another consequence of the SDH1 deletion observed in the current study is a reduced accumulation of malate in the culture medium (Figs 3 and 4 ).In fact, the observ ed decr ease in malate yield corresponded to the observed increase in SA yield (in mol acid /mol gl ycer ol ), when comparing the PYC2 oe -sdh1 and PYC2 oe strains ( Supplementary Fig. S3 ).As the lack of Sdh1 abolished the mitoc hondrial conv ersion of SA into fumar ate , malate , and oxaloacetate (Fig. 1 ), the sdh1 deletion mutants must utilize alternative means to provide the acceptor for acetyl-CoA in the oxTCA, and thus oxaloacetate , malate , and/or fumar ate m ust be imported from the cytosol.The lower concentration of malate detected for the sdh1 deletion mutants equipped with the rTC A pathwa y could be an indication that these strains indeed make use of this opportunity.A r educed accum ulation of malate inside the cell can be considered a benefit for SA production via yeast, particularly since Dct-02 also exports malate alongside SA into the culture medium (Renduli ć et al. 2022 ).
In conclusion, we were able to further increase the maximum SA yield compared to our pr e viousl y published str ain by tar geting mitoc hondrial tr ansporters and/or the SDH complex.The highest impr ov ement in yield (27%) was ac hie v ed by combining the benefits of deleting MPC3 (impr ov ed cytosolic rTCA flux) with the benefits of deleting SDH1 (increase in cytosolic SA and a drop in cytosolic malate).To our knowledge, the obtained maximum SA yield (0.66 g/g) r epr esents the highest SA yield that was obtained thus far using yeasts (Table 3 ).Notably, this result was ac hie v ed using shake-flasks without the particular effort to optimize cultivation conditions.Known process engineering techniques could ther efor e be exploited to ac hie v e potentiall y e v en better k e y performance parameters with this strain.For example, nitrogen-limitation could be applied to uncouple growth from SA pr oduction, ther eby further reducing biomass formation and maximizing SA yield (Liu et al. 2021 ).Although the results have been obtained here with glycerol as a carbon source, the metabolic eng ineering strateg ies presented in our work might also support future efforts to improve the SA production from sugars.Apart fr om the dir ect impr ov ements shown her e, the r esults of the current study can guide us e v en further to w ar ds impr ov ements in SA production with yeast-based production processes.We consider the further reduction of mitochondrial pyruvate import combined with a further reduction in mitochondrial import of rTCA intermediates highl y attr activ e .T his might also be combined with the r eduction of electr on tr ansfer fr om c ytosolic N ADH to the mitoc hondrial electr on tr ansport c hain.In gener al, it will be c hallenging to predict the optimal expr ession le v els of gene targets in-volved, and a combinatorial approach is considered most promising.
ND: not determined.