A genome-reduced Corynebacterium glutamicum derivative discloses a hidden pathway relevant for 1,2-propanediol production

Background 1,2-propanediol (1,2-PDO) is widely used in the cosmetic, food, and drug industries with a worldwide consumption of over 1.5 million metric tons per year. Although efforts have been made to engineer microbial hosts such as Corynebacterium glutamicum to produce 1,2-PDO from renewable resources, the performance of such strains is still improvable to be competitive with existing petrochemical production routes. Results In this study, we enabled 1,2-PDO production in the genome-reduced strain C. glutamicum PC2 by introducing previously described modifications. The resulting strain showed reduced product formation but secreted 50 ± 1 mM d-lactate as byproduct. C. glutamicum PC2 lacks the d-lactate dehydrogenase which pointed to a yet unknown pathway relevant for 1,2-PDO production. Further analysis indicated that in C. glutamicum methylglyoxal, the precursor for 1,2-PDO synthesis, is detoxified with the antioxidant native mycothiol (MSH) by a glyoxalase-like system to lactoylmycothiol and converted to d-lactate which is rerouted into the central carbon metabolism at the level of pyruvate. Metabolomics of cell extracts of the empty vector-carrying wildtype, a 1,2-PDO producer and its derivative with inactive d-lactate dehydrogenase identified major mass peaks characteristic for lactoylmycothiol and its precursors MSH and glucosaminyl-myo-inositol, whereas the respective mass peaks were absent in a production strain with inactivated MSH synthesis. Deletion of mshA, encoding MSH synthase, in the 1,2-PDO producing strain C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) improved the product yield by 56% to 0.53 ± 0.01 mM1,2−PDO mMglucose−1 which is the highest value for C. glutamicum reported so far. Conclusions Genome reduced-strains are a useful basis to unravel metabolic constraints for strain engineering and disclosed in this study the pathway to detoxify methylglyoxal which represents a precursor for 1,2-PDO production. Subsequent inactivation of the competing pathway significantly improved the 1,2-PDO yield. Supplementary Information The online version contains supplementary material available at 10.1186/s12934-024-02337-w.


Introduction
The C3-diols 1,2-and 1,3-propanediol (PDO) are important building blocks and are widely used in the polymer, food, cosmetic and drug industry [1,2].For 1,3-PDO and 1,2-PDO a global market size of around 1.4 billion and 0.4 billion US dollars is expected in the next years [2,3].Both diols are mainly produced from fossil fuels, but bio-based production processes, utilizing renewable resources, are favorable to tackle the concerns of climate change and the limited availability of fossil resources.Notably, first commercial microbial processes for 1,3-PDO production are readily available [4].Although several natural 1,2-PDO producers are known and established microbial systems have been extensively engineered for its production, a sustainable process at the industrial level is still missing [2].
Methylglyoxal is a key intermediate in the 1,2-PDO production route, but also represents a potent cytotoxic compound which reacts with arginine, lysine, and cysteine residues and might lead to protein inactivation [23,24].Consequently, methylglyoxal is in many organisms detoxified to d-lactate by a GloAB-mediated glyoxalase system with the tripeptide glutathione (GSH).d-lactate is further oxidized to pyruvate which is subsequently channeled back into the central carbon metabolism (Fig. 1) [23,24].A characteristic of Actinobacteria, such as C. glutamicum, is the absence of GSH biosynthesis.However, in these organisms the low molecular weight thiol and non-enzymatic antioxidant mycothiol (MSH) is synthesized for detoxification instead [25].MSH is synthezised from the intermediates myo-inositol and N-acetyl-glucosamine via glucosaminyl-myo-inositol by five enzymatic steps encoded by the genes mshA, mshA2, mshB, mshC and mshD [26].However, no GloAB homologs have been identified in C. glutamicum and the mechanism for detoxification of methylglyoxal is elusive so far.Besides MSH as antioxidant, the transcriptional regulator OxyR plays an important role in the oxidative stress response in C. glutamicum.Under unstressed conditions this master regulator represses genes encoding enzymes such as catalase.As a result, an oxyR deletion mutant shows increased resistance to H 2 O 2 [27].
The concept of genome reduction to improve physiological characteristics and create optimized hosts for the industrial environment has been applied to many organisms [28] as well as to C. glutamicum [29,30].In a systematic top-down approach all genes of the C. glutamicum genome were ranked by their relevance for growth in minimal medium with glucose as sole carbon and energy source.The proposed gene clusters were individually deleted and the genome-reduced strains (GRS) were screened for their growth phenotype [29].In a follow-up study the deletions of the most promising genomic regions were combined in a stepwise manner, leading to the pre-chassis strains PC1 and PC2, with 8.5% and 12.6% reduced genomes, respectively.While both of these strains grew like the wildtype, the finally engineered chassis strain C1 (13.4% genome-reduction) showed a growth deficit on acetate due to an unwanted mutation in the promoter of the ramA gene, which was repaired yielding the final genome-reduced derivative C1* [30].Recently, some intermediate strains of the aforementioned genome-reduction approach were utilized to screen for improved heterologous cutinase secretion in C. glutamicum [31].
In this study, we introduced known genetic modifications to enable 1,2-PDO production [22] into the genome-reduced strain C. glutamicum PC2 [30] and its parental strain GRS [29] (in later studies also called CR099).Only in the PC2 genetic background, we observed the formation of d-lactate as byproduct, which was attributed to the lack of the respective d-lactate dehydrogenase and pointed to a yet unknown pathway relevant for 1,2-PDO production.Further analysis indicated that in C. glutamicum methylglyoxal, the precursor for 1,2-PDO synthesis, is converted with MSH by a glyoxalase-like system to lactoylmycothiol and further to d-lactate which is channeled back into the central carbon metabolism on the level of pyruvate (Fig. 1).

Reconstruction of the 1,2-PDO production pathway in the C. glutamicum strains GRS and PC2
In a former study, it was shown that the deletion of hdpA and ldh and heterologous expression of the E. coli genes mgsA, yqhD and gldA led to the highest production of 1,2-PDO with C. glutamicum reported yet [22].To analyze the production capacity of the genome-reduced strains C. glutamicum PC2 [30] and its parental strain GRS [29], the named genetic modifications were introduced into both strains.The resulting strains C. glutamicum GRSΔhdpAΔldh and PC2ΔhdpAΔldh, each carrying the plasmid pEKEx3-mgsA-yqhD-gldA were cultivated in modified CGXII minimal medium with glucose as sole carbon and energy source [22].Compared to GRS, the strain with PC2 genetic background showed a similar growth rate (0.20 ± 0.01 vs. 0.22 ± 0.01 h -1 ), accumulated less biomass (max.CDW 6.60 ± 0.38 g L -1 vs. 8.25 ± 0.36) and exhibited a decreased glucose consumption (178 mM ± 7 vs. 126 ± 1 mM consumed) (Fig. 2A).While C. glutamicum GRSΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) produced up to 44 ± 3 mM 1,2-PDO (Y P/S = 0.25 ± 0.01 mol mol -1 ), only 10 ± 2 mM (Y P/S = 0.08 ± 0.02 mol mol -1 ) were secreted by C. glutamicum PC2ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) (Fig. 2B).Interestingly, the PC2-based strain produced up to 50 ± 1 mM lactate whereas for the GRS-derivative no lactate was determined in the supernatant (Fig. 2B).With the applied HPLC method it was not possible to distinguish between l-and d-lactate.However, since in both strains the ldh gene encoding l-lactate dehydrogenase is deleted, we speculated that d-lactate is secreted by C. glutamicum PC2ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA).Notably, compared to the GRS strain, the genome-reduced variant PC2 additionally lacks the dld gene encoding a quinone-dependent d-lactate dehydrogenase [32], which might be a reasonable explanation for the observed phenotype.

Impact of dld on the production of 1,2-propanediol in C. glutamicum
To clarify if the dld gene has a yet unknown effect on 1,2-PDO production in C. glutamicum, we deleted the gene in C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA).The resulting strain C. glutamicum ΔhdpAΔldhΔdld(pEKEx3-mgsA-yqhD-gldA) and its parental strain were cultivated in shaking flasks with modified CGXII medium and glucose as sole carbon and energy source.The additional deletion of dld led to a slightly decreased biomass concentration and reduced Fig. 1 Overview of the 1,2-PDO pathway introduced into C. glutamicum [22], including the proposed bypass from methylglyoxal to d-lactate and pyruvate via lactoylmycothiol in blue.Black arrows represent native pathways; dotted arrows indicate more than one reaction; green and disrupted arrows represent heterologously expressed proteins and deletions of gene sequences of mentioned proteins, respectively.Abbreviations: TCA, tricarboxylic acid cycle; DHA, dihydroxyacetone; DHAP, dihydroxyacetone phosphate; GAP, glyceraldehyde 3-phosphate; UDP-GlcNAc, uridine diphosphate-N-acetylglucosamine; 1-l-Ins-1-P, 1-l-myo-inositol 1-phosphate; GldA, glycerol dehydrogenase (from E. coli); HdpA, dihydroxyacetone phosphate phosphatase; MgsA, methylglyoxal synthase (from E. coli); Dld, quinone-dependent d-lactate dehydrogenase; LldD, quinone-dependent l-lactate dehydrogenase; Ldh, NAD-dependent l-lactate dehydrogenase; YqhD, aldehyde reductase (from E. coli); MshA, glycosyltransferase; "GloA", proposed lactoylglutathione lyase homolog in C. glutamicum; "GloB", proposed hydroxyacylglutathione hydrolase homolog in C. glutamicum glucose consumption (Fig. 3A and B).Further, C. glutamicum ΔhdpAΔldhΔdld(pEKEx3-mgsA-yqhD-gldA) secreted up to 24 ± 1 mM lactate and produced three times less 1,2-PDO compared to the parental strain (59 ± 12 mM as compared to 20 ± 7 mM; Fig. 3C and D).The lactate concentrations in the culture supernatants determined via HPLC were confirmed and identified to be d-lactate by a specific enzyme assay (Figure S1).These findings are in agreement with the results obtained with the chassis strain PC2, confirming the relevance of d-lactate dehydrogenase for 1,2-PDO production and hint towards a not yet known pathway in C. glutamicum which causes a degradation of the main product or an intermediate of the production pathway.

Inactivation of mycothiol biosynthesis improves 1,2-PDO production
We speculated that methylglyoxal might be converted to d-lactate by a MSH-dependent system similar to the glyoxalase system acting with GSH.To investigate this, we identified several targets potentially involved in the detoxification of methylglyoxal in C. glutamicum: deletion of the mshA gene encoding a glycosyltransferase to avoid MSH biosynthesis [25]; inactivation of the Xiclass mycothiol S-transferase [33] by deletion of the gene cg1426 (NCgl1216/mstX); deletion of the putative gloA and gloB homologs cg1073 as well as cg0071, cg1482 and cg1856, which were identified by annotations in the databases CoryneRegNet 7 [34] and Kyoto Encyclopedia of Genes and Genomes (KEGG) [35].These deletions were introduced individually into C. glutamicum ΔhdpAΔldh, followed by the transformation of the newly constructed strains with the plasmid pEKEx3-mgsA-yqhD-gldA.The resulting strains were cultivated in shaking flasks with modified CGXII minimal medium and glucose as sole carbon and energy source.Neither inactivation of the genes putatively encoding GloAB homologs nor MstX led to a significant change in formed biomass, glucose consumption or product yield (p > 0.1) compared to the parental strain C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) which showed a Y P/S of 0.34 ± 0.06 mM 1,2−PDO mM glucose −1 after 72 h of fermentation (Fig. 4  and S2).In contrast, deletion of the mshA gene resulted in a 20% reduced biomass formation after 72 h and impaired glucose consumption, however, improved the Y P/S significantly (p < 0.01) (Fig. 4 and S2).C. glutamicum ΔhdpAΔldhΔmshA(pEKEx3-mgsA-yqhD-gldA) showed a Y P/S of 0.53 ± 0.01 mM 1,2−PDO mM glucose −1 which is 56% higher compared to C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) (Fig. 4), while no lactate was formed.Because the mshA deletion abolishes MSH biosynthesis (see below), the improved Y P/S of C. glutamicum ΔhdpAΔldhΔmshA(pEKEx3-mgsA-yqhD-gldA) might be a result of increased metylglyoxal availability and/or also of an overall elevated oxidative stress level.To investigate this effect, we deleted and overexpressed oxyR, encoding the master regulator of the oxidative stress response [27], in C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) and analyzed the impact on growth and 1,2-PDO production.Overexpression of oxyR, increasing the sensitivity to oxidative stress [27], led to a strongly decreased Y P/S of 0.05 ± 0.01 mM 1,2−PDO mM glucose −1 whereas deletion of oxyR did not significantly impact 1,2-PDO production after 72 h (p > 0.1 regarding the yield), but inhibited growth and glucose uptake (Figure S3).Since the deletion of mshA increased the Y P/S strongly, these results may indicate that in C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) a significant portion of methylglyoxal is rerouted to the central carbon metabolism with MSH as reaction partner.

Discussion
Genome-reduced strains have been engineered to improve properties of microbes for industrial application, to study the physiological role of gene sets and to indentify essential genes to eventually design minimal genomes with streamlined functions [28,36].This concept has also been applied to the industrially relevant bacterium C. glutamicum which yielded a strain genealogy varying in the degree of genome reduction [30].This genomereduced strain library was recently utilized to screen for strains with improved heterologous secretion of cutinase [31].In this study, we harnessed the genome-reduced derivative C. glutamicum PC2 [30] and its parental strain GRS [29] to unravel metabolic bottlenecks for the production of the industrial relevant bulk chemical 1,2-PDO.
The reconstruction of the previously described genetic modifications [22] to enable 1,2-PDO production in C. glutamicum PC2 led to reduced product formation accompanied by the accumulation of lactate.In C. glutamicum three enzymes are known, which are responsible for synthesis and utilization of lactate.The NADH-dependent l-lactate dehydrogenase LdhA, encoded by ldh, which is mainly responsible for l-lactate formation under excess of NADH [22,37].Further, C. glutamicum harbors quinone-dependent l-and d-lactate dehydrogenases (LldD and Dld), which are specific for the respective enantiomer and essential for its utilization [32,38].C. glutamicum PC2ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) lacks LdhA and due to the genome reduction Dld, as well.The applied enzyme assay identified accumulating lactate to be the d-enantiomer.These results suggested the presence of a metabolic route relevant for 1,2-PDO synthesis and with d-lactate as intermediate which was interrupted by the inactive Dld in the PC2 background.Further, no other genes encoding proteins obviously related to the described production pathway could be identified in the PC2 background (Table S2).However, the PC2-derivative showed higher accumulation of lactate and less propanediol compared to the wildtype background harboring the same genetic modification in addition to dld deletion (50 ± 1 mM vs. 10 ± 2 mM lactate and 24 ± 1 mM vs. 20 ± 7 mM propanediol; Figs.2B and 3C and D).
Methylglyoxal is a precursor for 1,2-PDO synthesis but it also represents a cytotoxic intermediate leading to protein inactivation and oxidative stress [23,24].In E. coli and other organisms methylglyoxal is detoxified with GSH, forming the intermediate S-lactoylglutathione, which is further metabolized via the glyoxalase system GloAB to d-lactate [23,24,39].In actinobacteria, such as C. glutamicum, the low molecular weight thiol MSH is present instead of GSH [25], however, a glyoxalaselike system based on MSH has not been identified so far.Recently, in the actinobacterium Streptomyces coelicolor a Ni 2+ -activated and MSH-dependent glyoxalase I enzyme was described [40].However, BLASTx analysis [41] did not identify a homolog in C. glutamicum (data not shown).Also the deletion of the annotated gloA and gloB homologs cg1073 as well as cg0071, cg1482 and cg1856 in C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) did not influence growth, glucose consumption and product formation.Therefore, the identification of the glyoxylase system remains elusive and has to be elucidated in future studies.It was shown that overexpression of mshA, encoding a glycosyltransferase catalyzing the first step of MSH synthesis, and deletion of the gene mstX, encoding a MSH transferase, affected the robustness of C. glutamicum towards methylglyoxal [33,42].Taken together, these results indicate that, in analogy to GSH-dependent systems, methylglyoxal in C. glutamicum is detoxified with MSH via S-lactoylmycothiol by an unknown glyoxalase-like system yielding d-lactate.Notably, LC-MS-Qtof-based analysis identified major mass peaks characteristic for lactoylmycothiol, MSH and glucosaminyl-myo-inositol in cell extracts of the wildtype, C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) and its derivative with inactivated d-lactate dehydrogenase.For lactoylmycothiol, the exact mass and isotope pattern matches the expected sum formula C 20 H 34 N 2 O 14 S, and, although alternative structures cannot be excluded, the observed fragmentation pattern is in accordance with the proposed lactoylmycothiol structure.Moreover, in C. glutamicum ΔhdpAΔldhΔmshA(pEKEx3-mgsA-yqhD-gldA), with inactive MSH synthase, lactoylmycothiol and both precursors were not detectable.These findings support the presence of the MSH-dependent detoxification The secretion of up to 24 ± 1 mM of d-lactate by C. glutamicum ΔhdpAΔldhΔdld(pEKEx3-mgsA-yqhD-gldA) indicated an increased rerouting of methylglyoxal back into the central carbon metabolism in the parental strain with active Dld.Consequently, to improve the precursor availability, we inactivated MshA which significantly improved 1,2-PDO production.C. glutamicum ΔhdpAΔldhΔmshA(pEKEx3-mgsA-yqhD-gldA) showed a Y P/S of 0.53 ± 0.01 mM 1,2−PDO mM glucose −1 which is 56% higher compared to C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) and represents the highest reported value for C. glutamicum.Also in E. coli inactivation of the methylglyoxal detoxification system by deletion of gloA proved to be beneficial for 1,2-PDO production from glucose [43].
It should be noted, that the strain lacking mshA showed reduced growth and glucose consumption (Figure S2) which might indicate an impaired stress tolerance.In the closely related Mycobacterium tuberculosis MSH is essential for survival whereas for M. smegmatis it is not [44].For E. coli GSH is dispensable, however, GSH-deficiency results in an increased sensitivity towards oxidative stress [45].Also in C. glutamicum MSH synthesis is not essential but its absence was shown to increase the sensitivity against oxidative stress and some toxic compounds indicated by reduced growth and cell viability [26,42,46].Notably, a MSH-deficient C. glutamicum strain, lacking mshC, possessed impaired growth in bioreactor cultivations at a pO 2 of 30%, whereas at a pO 2 level of 20% this growth defect was abolished [46].The relevance of a proper functioning oxidative stress response for 1,2-PDO production is also indicated by the result that overexpression of oxyR in C. glutamicum ΔhdpAΔldh(pEKEx3-mgsA-yqhD-gldA) strongly diminished 1,2-PDO production.OxyR is the master regulator of the oxidative stress response in C. glutamicum [27] and under unstressed conditions this transcriptional regulator is acting as a repressor of its target genes (e.g.katA encoding catalase).Deletion of oxyR resulted in an increased resistance of the C. glutamicum mutant to hydrogen peroxide whereas its overexpression completely inhibited growth in the presence of H 2 O 2 [47].Therefore, a carefully adjusted oxygen transfer into the culture broth might be crucial for scale-up of fermentation processes based on MSH-deficient 1,2-PDO production strains.

Conclusion
Libraries with genome-reduced strains provide a valuable tool to identify novel targets for metabolic engineering.In this study, the genetic background of the genomereduced C. glutamicum PC2 strain was the basis to disclose the pathway for detoxification of methylglyoxal with MSH to d-lactate and its relevance for 1,2-PDO production.Subsequent inactivation of this competing pathway significantly improved 1,2-PDO production in C. glutamicum.Regarding the detoxification of methylglyoxal, the low molecular weight thiol MSH in C. glutamicum, seems to play a role comparable to GSH in other organism, such as E. coli.However, the enzymatic machinery catalyzing the detoxification process with MSH has to be identified in future studies.

Microorganisms, media and cultivation conditions
All C. glutamicum strains and plasmids used in this work are listed in Table 1.Escherichia coli DH5α [48] was used as host for plasmid construction and was grown in lysogeny broth (LB) or in 2x TY complex medium [49] at 37 °C and 120 rpm.For precultivation of C. glutamicum, LB medium with 2% (w/v) glucose was applied and the main cultivations were carried out in modified CGXII medium [22].C. glutamicum was cultivated aerobically in 500 mL baffled shaking flasks with 50 mL medium at 30 °C on a rotary shaker at 120 rpm.For induction of gene expression, 1 mM isopropyl-β-D-thiogalactopyranosid (IPTG) and, if appropriate, kanamycin (25 or 50 µg mL − 1 ) and spectinomycin (100 µg mL − 1 ) were supplemented to the medium.For solid medium plates 18 g agar-agar L − 1 were added to the liquid medium.

Recombinant DNA work and construction of C. glutamicum deletion mutants
Oligonucleotides (Table S1) were purchased either from Eurofins MWG Operon (Ebersberg, Germany), metabion international AG (Planegg, Germany) or Sigma-Aldrich Chemie GmbH (Steinheim, Germany).The construction of plasmids (Table 1) was carried out as described in [9].Purification and isolation of PCR products, plasmids and genomic DNA was conducted via "NucleoSpin® Gel and PCR Clean-up", "NucleoSpin® Plasmid" and "NucleoSpin® Microbial DNA", respectively, as recommended by MACHEREY-NAGEL GmbH & Co. KG (Düren, Germany).Enzymes were all purchased from New England Biolabs GmbH [(NEB), Frankfurt am Main, Germany] and used as recommended by the manufacturer.For cloning of pVWEx1-oxyR, pVWEx1 [52] was linearized via BamHI and the oxyR gene was PCR-amplified with primers #16 and #17 from genomic C. glutamicum DNA by Phusion® High-Fidelity DNA Polymerase.pVWEx1 and the insert were fused via Gibson assembly [53].To construct pK19mobsacB [50] derivatives the plasmid was linearized via SmaI and the two homologous flanks for recombination were amplified via PCR, respectively.E.g. for the deletion of the mshA gene the primer pairs #19 + 20 and #21 + 22 (Table S1) were used.Eventually, the flanks and linearized pK19mobsacB were fused via Gibson assembly [53].For subsequent transformation of E. coli DH5α the calcium chloride method was conducted [49].Positive clones were identified by colony PCR applying Quick-Load R® Taq 2X Master Mix, validated via restriction digestion with appropriate enzymes and eventually the correct sequence of the inserts was confirmed via Sanger sequencing by Microsynth Seqlab GmbH (Göttingen, Germany).
For transformation and markerless gene deletions in C. glutamicum GRS and PC2 derivatives the procedure of [22] was applied: Electrocompetent cells [54] were transformed via electroporation [6] and, in case of gene deletions via pK19mobsacB, a subsequent two-step homologous recombination procedure followed [6].All WTΔhdpAΔldh derivatives were engineered as described in [9]: Electrocompetent cells [55] were transformed via electroporation [56] and the subsequent two-step homologous recombination was carried out as described previously [50].Successful gene deletions were confirmed via colony PCR as described above with the respective primer pairs listed in Table S1.

Determination of biomass
OD 600 (optical density at 600 nm) was measured with a spectrophotometer (either V-1200 of VWR International, Darmstadt, Germany or ULTROSPEC® 10 of Biochrom, Holliston, MA, United States) by diluting the samples into an OD 600 range of 0.05 to 0.3.Cell dry weight (CDW) was calculated via a correlation factor of either 0.25 g L − 1 (V-1200) or 0.23 g L − 1 (ULTROSPEC® 10) per OD 600 of 1.The growth rates (h − 1 ) were determined by linear regression in a semi-logarithmic blot and maximizing the coefficient of determination (R 2 ) in the exponential growth phase.At the given time points 1 mL samples were taken, centrifuged (10 min, 16,000 or 21,300 x g) and the resulting supernatants were stored at -20 °C for further analysis.

GC-MS and HPLC analysis
The GC-MS-based quantification of 1,2-PDO and the HPLC measurement of glucose and lactate shown in Fig. 1 was conducted as described before [22].Elsewhere, glucose, 1,2-PDO, lactate and acetol were determined via a modified HPLC method as described in [9], employing a 1260 Infinity II system (Agilent Technologies, Waldbronn, Germany), equipped with a Hi-Plex H column (7.7 × 300 mm, 8 μm) and a Hi-Plex H guard cartridge (3 × 5 mm, 8 μm) kept at 30 °C.As mobile phase 5 mM sulfuric acid (H 2 SO 4 ) in water was applied with a flow rate of 0.4 mL min − 1 , while signal acquisition was conducted via refractive index detector (RID) hold at 30 °C.

Metabolomics via LC-MS
For LC-MS analyses cell extracts were prepared as previously described for C. glutamicum [57].From cultures grown as described above for HPLC analysis, cells were harvested via centrifugation (5 min, 3,000 x g, 4 °C), washed once with 0.5 volumes of saline (9 g L − 1 sodium chloride in water) and again centrifuged.The cell pellet was resuspended with boiling water to adjust a concentration of approximately 125 g CDW L − 1 .This suspension was incubated three times in a bath with boiling water for 2 min with intermittent vortexing of 5 s.The cell extracts were collected by centrifugation (10 min, 20,000 x g, 4 °C), aliquoted and stored at -80 °C for further analysis.
LC-MS analyses were performed on a bio-inert 1290 series UHPLC system which was interfaced to a Q-TOF mass spectrometer (G6546AA, Agilent Technologies) via a dual Agilent jet stream electrospray ion source.MassHunter LC/MS Data Acquisition (version 10.1) and Qualitative Analysis software (version 10.0) from Agilent Technologies was used for data acquisition and data evaluation, respectively.The mass spectrometer was operated in low mass range (m/z 1700), Gas temp 325 °C, VCap 4000 V, Nozzle Voltage 2000 V.The instrument was auto tuned and calibrated according to manufacturer´s recommendations using ESI-L tuning mix (Agilent Technologies).
Collision-induced dissociation (CID) mass spectra were acquired in targeted MS/MS mode.Precursor ions were isolated by the quadrupole using an isolation width of m/z 4 and fragmented inside the collision cell by applying collision energies in the range of 10, 20, and 30 V and nitrogen as collision gas.Fragment ions were detected from m/z 100-1000.For reference mass correction a solution of purine (5 µM) and hexakis-(2,2,3,3tetrafluoropropoxy)phosphazine (2.5 µM) in acetonitrile/ water, 95/5 (v/v) % was continuously introduced through the second sprayer of the dual ion source at a flow rate of 10 µL min -1 using an isocratic HPLC pump (G7110B, Agilent Technologies) equipped with a 1:100 splitting.

Enzyme assay specific for d-lactate
For the specific, enzymatic determination of d-lactate the "D-LACTIC ACID (D-LACTATE) (Rapid) Assay Kit (K-DATE)" from Megazyme Ltd. (Wicklow, Ireland) was purchased and applied as recommended by the manufacturer.The assay was conducted in a 96-well format with a Tecan Spark and SparkControl V 2.3 (Tecan Group Ltd., Männedorf, Switzerland), while for quantification an external calibration curve ranging from 0 to 0.15 g L − 1 was prepared with a d-lactic acid standard solution.

Determination of significance
The significance (p-value) was calculated via "two-sample t-test assuming equal variances" in Excel 2016 (Microsoft Corporation, Redmond, Washington, USA).

Table 1
Strains and plasmids used in this study R resistance gene; q Quantity