Rational Engineering of Non-Ubiquinone Containing Corynebacterium glutamicum for Enhanced Coenzyme Q10 Production

Coenzyme Q10 (CoQ10) is a lipid-soluble compound with important physiological functions and is sought after in the food and cosmetic industries owing to its antioxidant properties. In our previous proof of concept, we engineered for CoQ10 biosynthesis the industrially relevant Corynebacterium glutamicum, which does not naturally synthesize any CoQ. Here, liquid chromatography–mass spectrometry (LC–MS) analysis identified two metabolic bottlenecks in the CoQ10 production, i.e., low conversion of the intermediate 10-prenylphenol (10P-Ph) to CoQ10 and the accumulation of isoprenologs with prenyl chain lengths of not only 10, but also 8 to 11 isopentenyl units. To overcome these limitations, the strain was engineered for expression of the Ubi complex accessory factors UbiJ and UbiK from Escherichia coli to increase flux towards CoQ10, and by replacement of the native polyprenyl diphosphate synthase IspB with a decaprenyl diphosphate synthase (DdsA) to select for prenyl chains with 10 isopentenyl units. The best strain UBI6-Rs showed a seven-fold increased CoQ10 content and eight-fold increased CoQ10 titer compared to the initial strain UBI4-Pd, while the abundance of CoQ8, CoQ9, and CoQ11 was significantly reduced. This study demonstrates the application of the recent insight into CoQ biosynthesis to improve metabolic engineering of a heterologous CoQ10 production strain.


Identification of Accumulating Compounds in the Parent Strains
The chromatograms of the lipid extracts from strains UBI401, UBI405, UBI412, and UBI413 in our previous publication showed multiple peaks that remained unidentified [20] ( Figure 1A). Using LC-MS, we tried to identify these compounds.
First, we suspected that the two main peaks observed in wild type (WT) and UBI401 corresponded to MK 8 (H 2 ) and MK 9 (H 2 ), which had previously been described in corynebacteria. The mass spectra of the compounds eluting at 11 and 14.4 min in WT cells showed ions corresponding to H + and NH 4 + adducts of MK 8 (H 2 ) (m/z = 719.6 and 736.6, Figure S1A) and MK 9 (H 2 ) (m/z = 787.6 and 804.6, Figure S1B). The UV spectra (not shown) were also characteristic of naphthoquinone species. Single ion monitoring (SIM) of the NH4 + adducts showed that MK 8 (H 2 ) and MK 9 (H 2 ) were indeed present in the lipid extracts of all strains, albeit in various amounts ( Figure S1C,D).
Strain UBI405 expresses the decaprenyl synthase gene ddsA from P. denitrificans and the ubiA gene from E. coli, which encodes the polyprenyl transferase that prenylates 4-HBA. In comparison to UBI401, UBI405 showed several new peaks, two of them eluting late at 19 and 25.3 min and four others eluting early between 3.5 and 7 min ( Figure 1A). The compounds eluting at 19 and 25.3 min displayed UV spectra characteristic of naphthoquinone species (not shown) and their mass spectra showed ions corresponding to H + and NH 4 + adducts of MK 10 (H 2 ) (m/z = 855.7 and 872.7, Figure S2A) and MK 11 (H 2 ) (m/z = 923.8 and 940.8, Figure S2B). SIM of the NH 4 + adducts showed that MK 10 (H 2 ) ( Figure 1B) and MK 11 (H 2 ) ( Figure S2C) were absent in WT and strain UBI401 but present in the extracts of strains UBI405, UBI412, and UBI413, in agreement with the presence of ddsA in those later strains. These results show that expressing ddsA in C. glutamicum allows the synthesis of unnatural decaprenyl dihydromenaquinone as expected, but they also demonstrate poor specificity of DdsA from P. denitrificans since we also observed undecaprenyl dihydromenaquinone, MK 11 (H 2 ) ( Figure S2B,C). Together, our data show that strains UBI405, UBI412, and UBI413 synthesize four isoprenologs of MK(H 2 ), ranging from MK 8 (H 2 ) to MK 11 (H 2 ) (Figures S1 and S2), with MK 10 (H 2 ) and MK 11 (H 2 ) being the most abundant in strains UBI412 and UBI413 ( Figure S2D). Interestingly, the abundance of MK 8 (H 2 ) and MK 9 (H 2 ) increased in strain UBI401 compared to WT ( Figure S2D), validating the engineering aimed at increasing FPP supply and flux in the shikimate pathway.
Strains UBI412 and UBI413 showed several new peaks compared to the other strains ( Figure 1A) and LC-MS analysis identified two series of compounds: polyprenylphenols (nP-Ph) eluting between 5 and 12 min and menaquinones [8][9][10][11] ) eluting between 9.5 and 23 min ( Figures S4 and S5). We detected NH 4 + adducts of polyprenylphenol composed of 9, 10, and 11 isoprene units at 7, 8.9, and 11.5 min, respectively ( Figure S4A-C). The corresponding SIM showed the presence of these molecules only in strains UBI412 and UBI413 ( Figure 1C and Figure S4D,E), consistent with the expression of UbiD and UbiX, allowing for decarboxylation of nP-HB from strain UBI405 into nP-Ph. We could not obtain an unambiguous detection of octaprenylphenol (8P-Ph, m/z = 656.5) because a co-eluting compound at 5.5 min exhibited a prominent signal at m/z = 654.6 (data not shown).
The compounds eluting at 9.7, 12.7, 16.5, and 22.2 min in strains UBI412 and UBI413 corresponded to fully unsaturated menaquinones 8-11 with mass spectra displaying characteristic H + and NH 4 + adducts ( Figure S5). Finally, strain UBI413 that expresses all the enzymes of the CoQ pathway was shown to produce CoQ 10 (in agreement with our previous results [20], Figure 1E) and also CoQ 8 , CoQ 9 , and CoQ 11 ( Figure S6). It is worth noting that the peaks corresponding to CoQ [8][9][10][11] were barely detectable in the 275 nm absorbance chromatogram (black arrows on Figure 1A), whereas those corresponding to nP-Ph were more prominent (blue arrows on Figure 1A). Since the molar absorption coefficient of 8P-Ph is about five-fold lower than that of CoQ 8 [34], 8-11P-Ph are certainly more abundant than the corresponding CoQ [8][9][10][11] in strain UBI413, denoting that the late steps of the CoQ pathway do not function optimally in strain UBI413.
In conclusion, we have now assigned all the peaks displayed in the 275 nm chromatograms of the lipid extracts of the strains previously published. These results suggested to us several options to increase CoQ 10 biosynthesis in C. glutamicum, namely, (i) favor the accumulation of decaprenyl compounds over those with chains composed of 8, 9, or 11 prenyl units (ia-deletion of endogenous ispB, ib-screen for ddsA with higher specificity) and (ii) increase the overall efficiency of the CoQ pathway by expression of "accessory proteins".

Deletion of ispB Diminishes Formation of 8-and 9-Isoprenologs
Due to native polyprenyl diphosphate synthase activity in C. glutamicum, presumably encoded by ispB [32], isoprenologs with a chain length of eight and nine prenyl units appeared in all quinone extracts in addition to the desired 10-isoprenologs ( Figure 1F). When deletion of ispB was previously attempted in C. glutamicum, it remained unsuccessful as it is likely essential for menaquinone biosynthesis [32]. Here, we performed deletion of ispB with simultaneous replacement by ddsA from P. denitrificans in the strain UBI4, which was successful, yielding the viable strain UBI5. This strain and its isogenic parent strain were grown by microcultivation and examined for quinone content.
As the ECD chromatograms of UBI4 and UBI5 quinone extracts revealed, the latter strain synthesized no MK 8 (H 2 ) and MK 9 (H 2 ) at all and instead, to a much lesser extent, MK 10 (H 2 ) and MK 11 (H 2 ) were detected (Figure 2A,B), which serves as an indirect proof that IspB catalyzes octa-and nonaprenyl diphosphate formation in C. glutamicum. The MK 11 (H 2 ) accumulation was a result of DdsA from P. denitrificans being unspecific regarding polyprenyl diphosphate synthesis activity, as observed previously in strain UBI413 ( Figure S2C). Since menaquinone has vital functions in the cell, the influence of these drastic changes on growth behavior was monitored. Surprisingly, no significant difference in growth between both strains was observed in glucose minimal medium as the growth curves almost resembled the growth curve of the WT ( Figure S7). Hence, in the next step, both strains were transformed with the plasmids necessary for CoQ 10 production, pRG_Duet2-ddsA Pd -ubiA, pEC-XT99A-ubiDIBX, and pEKEx3-ubiGHEF, resulting in the strains UBI4-Pd (=UBI413) and UBI5-Pd, followed by cultivation in shake flasks and quinone extract analysis.

Screening of Different Decaprenyl Diphosphate Synthases
The expression of ddsA from P. denitrificans led to the formation of undesired 11isoprenologs; thus, two other ddsA genes from A. tumefaciens and R. sphaeroides were screened for their by-product formation as these bacteria are among the best and most relevant CoQ 10 production hosts [35,36]. The three genes were expressed in separate strains from the plasmids pRG_Duet2-ddsA Pd -ubiA, pRG_Duet2-ddsA At -ubiA, and pRG_Duet2-ddsA Rs -ubiA, together with those coding for the late CoQ pathway proteins, resulting in the strains UBI4-Pd, UBI4-At, and UBI4-Rs. The strains were cultivated in shake flasks and analyzed for their quinone content. Metabolites 2022, 12, x FOR PEER REVIEW 6 of 21 As can be seen from the ECD chromatograms from extracts of UBI4-Pd and UBI5-Pd, fewer peaks appeared for UBI5-Pd ( Figure 2C). SIM chromatograms of NH4 + adducts display that the missing peaks comprised, i.a., CoQ9 (m/z = 812.6, Figure 2D), CoQ8 (m/z = 744.5, Figure S8A), and MK9(H2) (m/z = 804.6, Figure S8C) as expected. However, levels of CoQ10 (m/z = 880.7, Figure 2E,F), CoQ11 (m/z = 948.8, Figure S8B), and MK10(H2) (m/z = 872.7, Figure S8D) remained unchanged between the strains. Overall, the replacement of ispB with ddsA abolished the synthesis of unwanted 8-and 9-isoprenologs without any impact on cells' growth, but it did not increase CoQ10 production.

Screening of Different Decaprenyl Diphosphate Synthases
The expression of ddsA from P. denitrificans led to the formation of undesired 11-isoprenologs; thus, two other ddsA genes from A. tumefaciens and R. sphaeroides were screened for their by-product formation as these bacteria are among the best and most relevant CoQ10 production hosts [35,36]. The three genes were expressed in separate strains from the plasmids pRG_Duet2-ddsAPd-ubiA, pRG_Duet2-ddsAAt-ubiA, and pRG_Duet2-ddsARs-ubiA, together with those coding for the late CoQ pathway proteins, resulting in the strains UBI4-Pd, UBI4-At, and UBI4-Rs. The strains were cultivated in shake flasks and analyzed for their quinone content.
To estimate the differences in the formation of isoprenologs with different side chain lengths, ratios of peak areas from mass spectrometry analysis were calculated for several compounds ( Table 1). The CoQ10/CoQ11 ratio was close to 1 in UBI4-Pd, revealing that DdsA from P. denitrificans is rather unspecific regarding polyprenyl diphosphate synthesis activity. On the other hand, extracts of UBI4-At contained almost no CoQ11 at all, resulting in a ratio of 145.4 ± 12.4. UBI4-Rs had a comparatively lower CoQ10/CoQ11 ratio of 7.6 ± 0.05, but a higher CoQ10/CoQ9 ratio than the other strains, which indicates that DdsA from R. sphaeroides synthesizes less nonaprenyl diphosphate as a side product than the other DdsA enzymes. This also reflected in the ratio of the intermediate prenylphenols To estimate the differences in the formation of isoprenologs with different side chain lengths, ratios of peak areas from mass spectrometry analysis were calculated for several compounds ( Table 1). The CoQ 10 /CoQ 11 ratio was close to 1 in UBI4-Pd, revealing that DdsA from P. denitrificans is rather unspecific regarding polyprenyl diphosphate synthesis activity. On the other hand, extracts of UBI4-At contained almost no CoQ 11 at all, resulting in a ratio of 145.4 ± 12.4. UBI4-Rs had a comparatively lower CoQ 10 /CoQ 11 ratio of 7.6 ± 0.05, but a higher CoQ 10 /CoQ 9 ratio than the other strains, which indicates that DdsA from R. sphaeroides synthesizes less nonaprenyl diphosphate as a side product than the other DdsA enzymes. This also reflected in the ratio of the intermediate prenylphenols 10P-Ph/9P-Ph (Table 1). The production of CoQ 10 was not significantly improved, which might be due to a metabolic bottleneck downstream of the DdsA reaction. The results demonstrate that although all three DdsAs perform the same reaction, they differ in their precision to elongate FPP by a fixed number of seven IPP units. By this metric, DdsA enzymes of A. tumefaciens and R. sphaeroides are superior to that of P. denitrificans. Table 1. Ratios of relative peak areas from mass spectrometry analysis and CoQ 10 biomass yields, titers, and volumetric productivities for the strains UBI4-Pd, UBI4-At, and UBI4-Rs.

Expression of ubiJK Alleviates a Major Bottleneck
As recently shown, the E. coli enzymes UbiI-G-H-E-F, that catalyze the steps from prenylphenol to CoQ, form a soluble multiprotein complex with the accessory factors UbiJ and UbiK [33,37]. To find out if this complex might form in a heterologous environment, C. glutamicum in this case, and improve the flux in the late CoQ pathway, the genes ubiJ and ubiK from E. coli were integrated into the genome of UBI4 under control of the strong promoter of actA [38], yielding strain UBI4JK. Equipped with the necessary plasmids for CoQ 10 biosynthesis (pRG_Duet2-ddsA Pd -ubiA, pEC-XT99A-ubiDIBX, and pEKEx3-ubiGHEF), quinones of the strains UBI4-Pd and UBI4JK-Pd were extracted after shake flask cultivation and subjected to LC-MS analysis as described above.
Regarding growth of the strains UBI4 and UBI4JK without plasmids, no comparative experiment was performed; hence, the direct influence of ubiJK expression on growth was not evaluated. For strains UBI4-Pd and UBI4JK-Pd, however, no significant difference in growth rate or final biomass formation was observed (data not shown). The CoQ 10 /10P-Ph ratio, although not reflecting the stoichiometry of the two molecules, can serve as an indicator for the flux efficiency between the early pathway intermediate 10P-Ph and the final product CoQ 10 ( Table 2). While in UBI4-Pd the ratio was 0.3 ± 0.1, indicating a rather low flux, expression of ubiJK in UBI4JK-Pd increased the ratio in favor of CoQ 10 production (1.5 ± 0.2). Consequently, the biomass yield, titer, and volumetric productivity increased around four-fold ( Table 2). The improved CoQ 10 production was also well visible in the ECD chromatograms where the peak corresponding to CoQ 10 increased in the extract of UBI4JK-Pd ( Figure 3A). The SIM chromatograms for the NH 4 + adduct of CoQ 10 underline the difference more clearly ( Figure 3B). The increased flux in the late CoQ pathway also reflected in increased CoQ 8 and CoQ 9 levels ( Figure S9A,B), while MK 9 (H 2 ) and MK 10 (H 2 ) levels remained almost unchanged ( Figure S9C,D). In line with results showing that E. coli ubiJ and ubiK mutants contain no or a reduced amount of CoQ 8 [33,39], our data demonstrate that the expression of ubiJK from E. coli is also important for efficient CoQ production in a heterologous host such as C. glutamicum. Table 2. Ratios of relative peak areas from mass spectrometry analysis and CoQ 10 biomass yields, titers, and volumetric productivities for the strains UBI4-Pd, UBI4JK-Pd, UBI5-Pd, and UBI6-Pd.

Combinatorial Approach for Maximized CoQ 10 Production
To study the combined effect of ispB replacement and ubiJK expression, the ∆ispB::P tuf -ddsA replacement was performed in UBI4JK, resulting in the strain UBI6. It was transformed with the plasmids pRG_Duet2-ddsA Pd -ubiA, pEC-XT99A-ubiDIBX, and pEKEx3-ubiGHEF, resulting in strain UBI6-Pd, followed by shake flask cultivation and LC-MS analysis of extracts. Compared to UBI4JK-Pd, the additional ∆ispB::P tuf -ddsA replacement had no significant effect on CoQ 10 content, titer or volumetric productivity (Table 2). However, as observed for UBI5-Pd, CoQ 8 , CoQ 9 , and MK 9 (H 2 ) amounts decreased severely (data not shown), which reflected in a CoQ 10 /CoQ 9 ratio of 38.6 ± 1.9, even higher than for UBI5-Pd (14.4 ± 5.5). The deletion of endogenous ispB and the expression of ubiJK significantly improved the CoQ 10 production and reduced the accumulation of side products in UBI6-Pd when compared to the initial strain UBI4-Pd.

Combinatorial Approach for Maximized CoQ10 Production
To study the combined effect of ispB replacement and ubiJK expression, the ΔispB::Ptuf-ddsA replacement was performed in UBI4JK, resulting in the strain UBI6. It was transformed with the plasmids pRG_Duet2-ddsAPd-ubiA, pEC-XT99A-ubiDIBX, and pEKEx3-ubiGHEF, resulting in strain UBI6-Pd, followed by shake flask cultivation and LC-MS analysis of extracts. Compared to UBI4JK-Pd, the additional ΔispB::Ptuf-ddsA replacement had no significant effect on CoQ10 content, titer or volumetric productivity (Table 2). However, as observed for UBI5-Pd, CoQ8, CoQ9, and MK9(H2) amounts decreased severely (data not shown), which reflected in a CoQ10/CoQ9 ratio of 38.6 ± 1.9, even higher than for UBI5-Pd (14.4 ± 5.5). The deletion of endogenous ispB and the expression of ubiJK significantly improved the CoQ10 production and reduced the accumulation of side products in UBI6-Pd when compared to the initial strain UBI4-Pd.
We have shown that DdsA from A. tumefaciens and R. sphaeroides are more specific than DdsA from P. denitrificans towards the formation of 10P-HB compared to 9P-HB and 11P-HB (Table 1). In a final combinatorial approach, strains UBI6-At and UBI6-Rs were constructed for additive benefits and compared to UBI6-Pd. As observed previously, the strains with ddsA from A. tumefaciens and R. sphaeroides had an improved CoQ10/CoQ11 ratio of around 3.5 (Table 3). However, the ratio was not as high as for the strains UBI4-At and UBI4-Rs (Table 1), likely because of the chromosomal expression of the additional ddsA from P. denitrificans inserted in the ΔispB locus. Regarding the CoQ10/CoQ9 ratio, UBI6-Pd and UBI6-Rs surprisingly shared a high ratio of around 40 compared to only 5.1 ± 0.4 for UBI6-At (Table 3). This indicates that the chromosomal expression of ddsA from P. denitrificans had less influence on CoQ9 production than on CoQ11 production. CoQ10 content, titer, and volumetric productivity were all twice as high in UBI6-Rs as for the other strains with values of 126.9 ± 10.7 µg g −1 CDW, 1.21 ± 0.12 mg L −1 , and 16.8 ± 1.7 µg L −1 h −1 , respectively. The CoQ10/10P-Ph ratio of UBI6-Rs was 60% higher than for the other We have shown that DdsA from A. tumefaciens and R. sphaeroides are more specific than DdsA from P. denitrificans towards the formation of 10P-HB compared to 9P-HB and 11P-HB (Table 1). In a final combinatorial approach, strains UBI6-At and UBI6-Rs were constructed for additive benefits and compared to UBI6-Pd. As observed previously, the strains with ddsA from A. tumefaciens and R. sphaeroides had an improved CoQ 10 /CoQ 11 ratio of around 3.5 (Table 3). However, the ratio was not as high as for the strains UBI4-At and UBI4-Rs (Table 1), likely because of the chromosomal expression of the additional ddsA from P. denitrificans inserted in the ∆ispB locus. Regarding the CoQ 10 /CoQ 9 ratio, UBI6-Pd and UBI6-Rs surprisingly shared a high ratio of around 40 compared to only 5.1 ± 0.4 for UBI6-At (Table 3). This indicates that the chromosomal expression of ddsA from P. denitrificans had less influence on CoQ 9 production than on CoQ 11 production. CoQ 10 content, titer, and volumetric productivity were all twice as high in UBI6-Rs as for the other strains with values of 126.9 ± 10.7 µg g −1 CDW, 1.21 ± 0.12 mg L −1 , and 16.8 ± 1.7 µg L −1 h −1 , respectively. The CoQ 10 /10P-Ph ratio of UBI6-Rs was 60% higher than for the other two strains. The relative peak areas for 10P-Ph ( Figure S10) were in a similar range for all of them despite the higher CoQ 10 content of UBI6-Rs, indicating a pull effect for the intermediate 10P-Ph. This would be favorable if the flux from 10P-Ph to CoQ 10 was further optimized in future strain engineering. To visualize the difference caused by the combinatorial approach, the ECD chromatograms and SIM chromatograms for the NH 4 + adduct of CoQ 10 are displayed in Figure 4. Compared to the initial strain UBI4-Pd, UBI6-Rs extracts contain much fewer and/or lower peaks, while the peak for CoQ 10 has become the most prominent one. Nevertheless, the ECD chromatograms also show that MK 10 and MK 10 (H 2 ), eluting at 16.5 and 19 min, accumulated to considerable amounts as well. While the CoQ 10 productivity reached here is still not competitive with productivity in native hosts such as R. sphaeroides [36], we provided a rational metabolic engineering approach in a non-native host, resulting in significantly higher CoQ 10 production and lower by-product formation.

Influence of Growth Phase and Medium on CoQ 10 Production
The best strain, UBI6-Rs, was cultivated for a time-resolved analysis of CoQ 10 production ( Figure 5). The cells grew with a specific growth rate of 0.13 h −1 in the first 24 h and the biomass reached its peak at 32 h with an optical density at 600 nm (OD 600 ) of 62 (15.5 g L −1 CDW). The kinetics of growth and CoQ 10 production largely overlapped with the CoQ 10 content reaching 95 µg g −1 CDW after 40 h and remaining around that level until the end of cultivation. Notably, the CoQ 10 /CoQ 11 ratio was rather low at the beginning with 0.6 but increased steadily during exponential growth to a maximum of 3.3 at 32 h ( Figure 5). The strain UBI6-Rs carries the ddsA gene from P. denitrificans that is expressed constitutively in the ispB locus, and the ddsA gene from R. sphaeroides, expressed from the vector pRG_Duet2 upon induction by isopropyl-β-D-1-thiogalactopyranoside (IPTG) at the beginning of the cultivation. As shown before, expression of ddsA from R. sphaeroides promotes higher CoQ 10 production with lower accumulation of CoQ 11 than ddsA from P. denitrificans, which explains the low CoQ 10 /CoQ 11 ratio at the beginning and its increase over time. Table 3. Ratios of relative peak areas from mass spectrometry analysis and CoQ 10 biomass yields, titers, and volumetric productivities for the strains UBI6-Pd, UBI6-At, and UBI6-Rs. In addition, UBI6-Rs was cultivated in a BioLector microcultivation system in CGXII medium (same as before) and WSCH medium. Metabolites 2022, 12, x FOR PEER REVIEW 9 of 21 two strains. The relative peak areas for 10P-Ph ( Figure S10) were in a similar range for all of them despite the higher CoQ10 content of UBI6-Rs, indicating a pull effect for the intermediate 10P-Ph. This would be favorable if the flux from 10P-Ph to CoQ10 was further optimized in future strain engineering. To visualize the difference caused by the combinatorial approach, the ECD chromatograms and SIM chromatograms for the NH4 + adduct of CoQ10 are displayed in Figure 4. Compared to the initial strain UBI4-Pd, UBI6-Rs extracts contain much fewer and/or lower peaks, while the peak for CoQ10 has become the most prominent one. Nevertheless, the ECD chromatograms also show that MK10 and MK10(H2), eluting at 16.5 and 19 min, accumulated to considerable amounts as well. While the CoQ10 productivity reached here is still not competitive with productivity in native hosts such as R. sphaeroides [36], we provided a rational metabolic engineering approach in a non-native host, resulting in significantly higher CoQ10 production and lower byproduct formation.

Influence of Growth Phase and Medium on CoQ10 Production
The best strain, UBI6-Rs, was cultivated for a time-resolved analysis of CoQ10 production ( Figure 5). The cells grew with a specific growth rate of 0.13 h −1 in the first 24 h and the biomass reached its peak at 32 h with an optical density at 600 nm (OD600) of 62 (15.5 g L −1 CDW). The kinetics of growth and CoQ10 production largely overlapped with the CoQ10 content reaching 95 µg g −1 CDW after 40 h and remaining around that level until the end of cultivation. Notably, the CoQ10/CoQ11 ratio was rather low at the beginning After CoQ 10 production was successfully improved, the strain UBI6-Rs was cultivated with a hydrolysate of the alternative feedstock wheat side stream concentrate (WSCH) [40]. In a sustainable circular economy, side streams of industrial production processes can provide excellent alternative feedstocks for microbial production containing macro-and micronutrients [41]. Here, WSCH was supplemented with the nitrogen source ammonium sulfate and the buffer 3-(N-morpholino) propanesulfonic acid (MOPS) for cultivation of UBI6-Rs in a microcultivation system, CGXII minimal medium was used for comparison. LC-MS analysis confirmed that the CoQ 10 concentration in WSCH medium before cultivation was below the detection limit of 1 nM. In the microcultivation system used for this comparison, the CoQ 10 content, titer, and volumetric productivity in CGXII minimal medium were lower than in shake flasks ( Table 3). The CoQ 10 content obtained in microcultivation with WSCH medium was about 40% of that in microcultivation with CGXII medium ( Table 3). As microcultivation with WSCH medium supported growth to a higher biomass concentration, the titer and volumetric productivity were about 55% of the values obtained in microcultivation with CGXII medium (Table 3). Moreover, by-product and intermediate formation in WSCH were mostly lower compared to CGXII medium as indicated by the CoQ 10 /CoQ 11 and CoQ 10 /10P-Ph ratios (Table 3).
( Figure 5). The strain UBI6-Rs carries the ddsA gene from P. denitrificans that is expressed constitutively in the ispB locus, and the ddsA gene from R. sphaeroides, expressed from the vector pRG_Duet2 upon induction by isopropyl-β-D-1-thiogalactopyranoside (IPTG) at the beginning of the cultivation. As shown before, expression of ddsA from R. sphaeroides promotes higher CoQ10 production with lower accumulation of CoQ11 than ddsA from P. denitrificans, which explains the low CoQ10/CoQ11 ratio at the beginning and its increase over time.  Table 3. The numbers next to the grey data points indicate the CoQ10/CoQ11 ratios of relative peak areas from mass spectrometry analysis. Values and error bars represent means and standard deviations of 3 independent cultivations.
After CoQ10 production was successfully improved, the strain UBI6-Rs was cultivated with a hydrolysate of the alternative feedstock wheat side stream concentrate (WSCH) [40]. In a sustainable circular economy, side streams of industrial production processes can provide excellent alternative feedstocks for microbial production containing macro-and micronutrients [41]. Here, WSCH was supplemented with the nitrogen source ammonium sulfate and the buffer 3-(N-morpholino) propanesulfonic acid (MOPS) for cultivation of UBI6-Rs in a microcultivation system, CGXII minimal medium was used for comparison. LC-MS analysis confirmed that the CoQ10 concentration in WSCH medium before cultivation was below the detection limit of 1 nM. In the microcultivation system used for this comparison, the CoQ10 content, titer, and volumetric productivity in CGXII minimal medium were lower than in shake flasks ( Table 3). The CoQ10 content obtained in microcultivation with WSCH medium was about 40% of that in microcultivation with CGXII medium ( Table 3). As microcultivation with WSCH medium supported growth to a higher biomass concentration, the titer and volumetric productivity were about 55% of the values obtained in microcultivation with CGXII medium (Table 3). Moreover, byproduct and intermediate formation in WSCH were mostly lower compared to CGXII medium as indicated by the CoQ10/CoQ11 and CoQ10/10P-Ph ratios (Table 3).
To conclude, although C. glutamicum is not a natural producer of any ubiquinone, we were able to identify metabolic bottlenecks in the initial metabolically engineered producer UBI4-Pd and to optimize the strain for seven-fold increased CoQ10 content and eight-fold increased CoQ10 titer with considerably lower by-product formation.  Table 3. The numbers next to the grey data points indicate the CoQ 10 /CoQ 11 ratios of relative peak areas from mass spectrometry analysis. Values and error bars represent means and standard deviations of 3 independent cultivations.
To conclude, although C. glutamicum is not a natural producer of any ubiquinone, we were able to identify metabolic bottlenecks in the initial metabolically engineered producer UBI4-Pd and to optimize the strain for seven-fold increased CoQ 10 content and eight-fold increased CoQ 10 titer with considerably lower by-product formation.

Discussion
The biosynthesis of CoQ is complex and has not been fully elucidated after 60 years of research, making its transfer to an organism without CoQ biosynthesis challenging. Previously, we set the foundation for heterologous CoQ 10 production in C. glutamicum by expression of genes, fulfilling the minimum requirements for CoQ biosynthesis [20]. In this study, our goal was to identify metabolic bottlenecks by LC-MS analysis and to alleviate them by genetic engineering. The strategy comprised increasing flux by expression of accessory factor genes ubiJ and ubiK, deleting the native polyprenyl diphosphate synthase IspB, and expressing the best decaprenyl diphosphate synthase to concentrate isoprenologs production towards CoQ 10 .
In E. coli, UbiJ and UbiK induce the formation of a Ubi complex with UbiI-G-H-E-F that catalyzes the reactions of the late CoQ pathway [37]. Here, expression of ubiJK increased CoQ 10 production four-fold. It is likely that, although not proven biochemically, this may have led to the formation of a Ubi complex in the heterologous host C. glutamicum. While CoQ 10 production was substantially improved, 10P-Ph was still abundant in the strain UBI6-Rs, indicating that flux through the putative Ubi complex did not reach its full potential. One reason might be that the E. coli Ubi proteins we expressed in C. glutamicum are not well suited to modifying CoQ intermediates with a decaprenyl side chain because they naturally operate on compounds with an octaprenyl chain. Therefore, expressing Ubi proteins from a bacterium that naturally produces CoQ 10 could improve the flux between 10P-Ph and CoQ 10 . Another reason might lie in the subunit stoichiometry of the Ubi complex. It has been shown in E. coli that synthesis rates of proteins that belong to a multiprotein complex are proportional to the subunit stoichiometry of their corresponding complexes [42] in order to save cellular resources and to avoid protein aggregation and misfolding [43]. Not only do the synthesis rates of the Ubi complex-associated proteins UbiK-J-I-G-H-E-F differ from each other [42], which is indicative of heterogeneous stoichiometry, but also UbiK and UbiJ were found to associate in a heterotrimeric UbiK 2 -UbiJ 1 complex [33] and several Ubi proteins interacted with themselves [37]. This is in contrast to our cloning strategy in which ubi genes were expressed in artificial polycistronic operons and lacked any regulation, leading to disproportional protein abundances in relation to the Ubi complex stoichiometry. Expression fine-tuning would be a way to change expression levels of the single components of the Ubi complex, e.g., by the use of a promoter library [44], adjustment of transcriptional initiation rates using artificial ribosome binding sites [45], introduction of multiple gene copies into the genome [31,46], or changing the order of genes in the polycistronic operons [47]. In the natural CoQ 10 producer R. sphaeroides, metabolic bottlenecks were identified to be UbiE, UbiH, and UbiG. Three different bottleneck elimination strategies were tested, among which fusion of UbiE and UbiG and localization of the fused protein onto the membrane via pufX linker gave the best results with a titer of 108.5 mg L −1 after 96 h of cultivation, a CoQ 10 content of 8.9 mg g −1 CDW, and a volumetric productivity of 1.13 mg L −1 h −1 [16]. However, it is questionable if this strategy would be effective in our case due to differences in the molecular mechanism of CoQ biosynthesis between E. coli and R. sphaeroides enzymes. A multienzyme complex has not been confirmed for R. sphaeroides CoQ biosynthesis, and proteins homologous to E. coli UbiI, UbiG, UbiH, UbiE, and UbiF exist in R. sphaeroides, but none to the accessory factors UbiJ and UbiK according to Protein BLAST analysis. It should be mentioned that overexpression of ubiJ and ubiK in E. coli led to decreased CoQ 8 content and four-fold and two-fold increases in 8P-Ph and 2-decaprenyl-3-methyl-6-methoxy-1,4-benzoquinol levels, respectively, presumably as a consequence of sequestration of these CoQ 8 intermediates [37]. Thus, expression strength of ubiJ and ubiK seems to influence the flux from 10P-Ph to CoQ 10 quite dramatically and should be adjusted accordingly.
Moreover, an increase in MK 10 and MK 10 (H 2 ) amounts was observed with expression of ddsA from R. sphaeroides in both background strains UBI4 and UBI6. We propose that Cgl0472 is a menaquinone oxidoreductase that reduces a double bond in the chain of MK, resulting in MK(H 2 ) ( Figure 1F), because it shares 51% sequence identity with menaquinone oxidoreductase MenJ from Mycobacterium tuberculosis [48]. The accumulation of fully unsaturated menaquinones might be caused by the inhibition of Cgl0472; however, this requires further investigation. Irrespective of the fact that the growth of C. glutamicum was not influenced by the different chain length of menaquinone, MK 10 and MK 10 (H 2 ) are competing by-products to CoQ 10 and should be limited in favor of increased CoQ 10 production. In E. coli, the competitive MK biosynthesis was blocked by the deletion of 1,4-dihydroxy-2-naphthoate (DHNA) octaprenyltransferase gene menA, ensuring that octaprenyl diphosphate would only be used to prenylate the CoQ intermediate 4-HBA, which led to increased CoQ 8 content by 81% [49] and squalene content by 18% [50]. In C. glutamicum, however, MK is the only natural isoprenoid quinone and therefore vital. The downregulation of menA in C. glutamicum might allow flux into MK biosynthesis to be lowered without impacting growth. Analogously, in R. sphaeroides, the competitive carotenoid biosynthesis was downregulated, resulting in 28% increased CoQ 10 production, since the complete disruption of carotenogenesis impaired both growth and CoQ 10 production [17]. Nevertheless, it could be challenging to determine the right balance to lower the flux into MK biosynthesis as much as possible while maintaining growth and, thus, CoQ 10 productivity. Alternatively, specific MenA inhibitors can be used to reduce MenA activity. Several drugs, among them an allylaminomethanone class of compounds, have been identified to inhibit MenA of Mycobacterium tuberculosis, acting as demethylmenaquinone (DMK) mimics [51,52]. The benefit of this approach is that different inhibitors can be tested and the optimal dose can be found comparatively fast, rendering the need for menA expression fine-tuning superfluous, if the inhibitor is not costly and can be used at a larger scale as well. Furthermore, quorum sensing provides another alternative to plain downregulation of menA. Being able to maintain regular expression of menA in the early growth phase and reducing it with increasing cell density would prevent growth deficits caused by menaquinone deficiency and reduce menaquinone accumulation. Liu et al. adapted the ComQXPA-P srfA quorum sensing system of Bacillus subtilis to C. glutamicum such that P srfA -controlled transcription of an hfq-sRNA complementary to a target gene was activated with high cell density, leading to silencing of the target gene by its complementary sRNA. In addition, a library of synthetic P srfA promoters was established to modulate the expression of the hfq-sRNA [53], allowing for optimized control over menA expression.
In order to improve precursor supply for CoQ 10 production, overexpression of the MEP pathway genes dxs and idi is a common way to increase flux towards IPP and DMAPP and has been shown to increase patchoulol production in an engineered C. glutamicum strain [29]. Other strategies aim at the supply and distribution of the molecules of the MEP pathway entry point, glyceraldehyde 3-phosphate and pyruvate [54], e.g., by the modification of central carbon metabolism [55], CRISPRi-mediated repression [56], and increase in the NAD(P)H pool [57]. A different kind of approach is membrane engineering that involves the expression of proteins with membrane-bending properties and the overall increase in membrane synthesis to expand the membrane surface area and storage capacity for CoQ 10 . In E. coli, the monoglucosyldiacylglycerol synthase Almgs was overexpressed to induce the formation of membrane stacks or tubules and intracellular membrane vesicles, and the genes plsB and plsC were overexpressed to increase glycerophospholipid biosynthesis, which synergistically increased β-carotene production 2.9-fold [58]. However, media composition and cultivation conditions are potent factors as well and should be considered to improve productivity. For CoQ 10 production, strategies such as controlling a low sucrose concentration during fed-batch fermentation of A. tumefaciens [59] or the cultivation of R. sphaeroides under phosphate limitation [60] proved to be very effective. In this study, the standard minimal medium CGXII for C. glutamicum [19] was used. Since it was designed for the production of amino acids, it contains a high concentration of nitrogen that should be tuned down in case of production of the nitrogen-free CoQ 10 . The reduction of nitrogen to 10% and of glucose to 50% increased the production of N-methylphenylalanine by C. glutamicum and reduced by-product formation [24]. As there are numerous CGXII components, and macro and trace elements, statistical methods such as response surface methodology help to find optimized conditions by using the proper design of experiments as was demonstrated for glutamate production in C. glutamicum [61]. Media optimization can also be employed to generate high cell densities in cultures, which is especially interesting for cell-bound products such as CoQ 10 . In a recent study, lignocellulose-derived acetate was utilized as a sole carbon source and as acid pH titrant, while urea was fed as a nitrogen source. By dynamical adaptation of the C/N feeding ratio, a maximal cell dry weight of 80.2 g L −1 was achieved [62].
With respect to alternative feedstocks that are not competitive with food or feed, C. glutamicum has been employed and engineered for many different substrates. Here, we demonstrated the successful production of CoQ 10 from a wheat side stream-based hydrolysate that has been utilized previously for the production of 5-aminovalerate [40] and L-2-hydroxyglutarate [22]. As well as the wheat side stream, access to numerous monomeric and polymeric carbon sources has been established, e.g., xylose, arabinose, mannose, starch, lignocellulose, N-acetylglucosamine, and alginate, which can be derived from hydrolysates of sustainable second generation feedstocks such as spent sulfite liquor, Miscanthus biomass, brown seaweed, corn straw, rice straw, and shrimp waste [63][64][65][66].
Taken together, we demonstrated how CoQ 10 production can be established in the non-ubiquinone containing organism C. glutamicum and optimized substantially by applying current knowledge about CoQ biosynthesis to establish its efficient production. Although the achieved CoQ 10 content is not yet competitive with natural producers such as R. sphaeroides, our strain holds the potential for further improvements with regards to metabolic engineering, media and cultivation conditions, and cell density. Its wellestablished genetic tools, systems metabolic engineering, and insights into sustainable production processes make C. glutamicum an attractive organism for the production of high value-added compounds such as CoQ 10 .

Bacterial Strains and Growth Conditions
All bacterial strains used in this study are listed in Table 4. E. coli DH5α [67] was used for plasmid construction, E. coli S17-1 [68] was used for transfer of suicide vectors by trans-conjugation prior to chromosomal gene replacements, C. glutamicum UBI4 [20] was used for strain construction. Pre-cultures of E. coli and C. glutamicum were inoculated from fresh LB or BHI agar plates and cultivated in lysogeny broth (LB) and brain heart infusion (BHI) medium at 37 • C and 30 • C in non-baffled and baffled shake flasks on a rotary shaker at 180 rpm and 120 rpm, respectively. When appropriate, kanamycin (25 µg mL −1 ), spectinomycin (100 µg mL −1 ), and tetracycline (5 µg mL −1 ) were added to the media and plates. For production experiments, C. glutamicum cells from pre-cultures were washed with TN buffer pH 6.3 (10 mM Tris-HCl, 150 mM NaCl) and inoculated to an OD 600 of 1 in 50 mL CGXII minimal medium [19] in 500 mL shake flasks. When specified, cultivations were performed in a BioLector microcultivation system (m2p-labs, Baesweiler, Germany) in 3.2 mL FlowerPlates at 1100 rpm and 30 • C and with filling volumes of 1 mL. The minimal medium was supplemented with 40 g L −1 glucose as sole carbon source, 1 mM IPTG to induce gene expression of ddsA from pRG_Duet2 [69] and all genes from pEC-XT99A [70] and pEKEx3 [71], 0.25 µg mL −1 anhydrotetracycline (ATc) to induce gene expression of ubiA from pRG_Duet2, and respective antibiotics. For cultivation in WSCH medium, 80% (v/v) hydrolysate (from 190 g L −1 WSC, [40]) was supplemented with inducers and antibiotics as described above, 20 g L −1 ammonium sulfate and 42 g L −1 MOPS to a final glucose concentration of 33.2 g L −1 . OD 600 was measured using a V-1200 spectrophotometer (VWR, Radnor, PA, United States). After 72 h, cells were harvested and stored at −20 • C. Table 4. Bacterial strains used in this study.

Molecular Genetic Techniques and Strain Construction
Standard molecular genetic techniques were performed as described [72]. Competent E. coli cells were prepared with the RbCl method and transformed by heat shock [72]. Competent C. glutamicum cells were prepared in NCM medium, with the exception of DL-threonine, according to an optimized transformation protocol [73]. Cells were transformed using electroporation at 2.5 kV, 200 Ω, and 25 µF. PCR amplification was performed with Phusion High-Fidelity DNA polymerase according to the manufacturer (New England Biolabs, Hitchin, UK). All plasmids are listed in Table 5 and were constructed via Gibson Assembly [74], using DNA fragments created with the primers specified in Table 6.  [20] pK19mobsacB Kan R , pK19 oriV Ec , sacB, lacZα, E. coli/C. glutamicum shuttle vector for construction of insertion and deletion mutants in C. glutamicum [75] pK19mobsacB-∆actA:ubiJK pK19mobsacB with a construct for deletion of actA (cg2840) and insertion of ubiJ and ubiK from E. coli under control of the native actA promoter This work pK19mobsacB-∆ispB:P tuf -ddsA Pd pK19mobsacB with a construct for deletion of ispB (cg0559) and insertion of ddsA from P. denitrificans under control of C. glutamicum promoter P tuf This work The ddsA genes were amplified from genomic DNA from A. tumefaciens C58 and R. sphaeroides ATH 2.4.1 (DSM 158); pRG_Duet2 was restricted with BamHI for insertion of ddsA and with NheI for insertion of ubiA as described [20]. The pK19mobsacB plasmids were constructed in two steps by amplification of the flanking regions of actA and ispB from genomic DNA of C. glutamicum ATCC 13,032 and restriction of pK19mobsacB with BamHI. In the second step, the plasmids pK19mobsacB-∆actA and pK19mobsacB-∆ispB were restricted in the newly generated restriction sites BamHI and BcuI, respectively, between the flanking regions. The genes ubiJ and ubiK were amplified from genomic DNA from E. coli K-12 MG1655; P tuf -ddsA Pd was amplified from pSH1-ddsA Pd , which was constructed before by restriction of pSH1 [76] with BamHI and amplification of ddsA from genomic DNA of P. denitrificans. Correct sequences were confirmed by sequencing of inserts. Gene replacements were performed by using the suicide vector pK19mobsacB and two-step homologous recombination as described [77]. Transfer of the vectors by trans-conjugation using S17-1 as donor strain and selection of the mutants was conducted as described [19]. Successful replacements were verified by PCR and sequencing with the primers specified in Table 6.

Quinone Extraction and Analysis
Pellets of C. glutamicum cells (10-25 mg) were suspended in cold PBS buffer in Eppendorf tubes. Cells were centrifuged at 13,000 rpm for 2 min at 4 • C, the supernatant was eliminated, and the wet weight of the pellet was determined. Glass beads (100 µL), 50 µL of 0.15 M KCl, and a volume of 2 mM MK 7 solution (used as an internal standard, Sigma-Aldrich) proportional to the wet weight (2 µL/mg) were added to cell pellet. Quinone extraction was performed by adding 0.6 mL of methanol, vortexing for 10 min, then adding 0.4 mL of petroleum ether (boiling range 40-60 • C) and vortexing for 3 min. The phases were separated by centrifugation at 1 min, 5000 rpm. The upper petroleum ether layer was transferred to a fresh tube. Petroleum ether (0.4 mL) was added to the glass beads and methanol-containing tube, and the extraction was repeated. The petroleum ether layers were combined and dried under nitrogen. The dried samples were stored at −20 • C and were resuspended in 100 µL ethanol. Aliquots corresponding to 2 mg of cells' wet weight were analyzed by reversed-phase HPLC with a C18 column (Betabasic-18, 5 mm, 4.6 × 150 mm; Thermo Scientific) at a flow rate of 1 mL/min with a mobile phase composed of 25% isopropyl alcohol, 20% ethanol, 45% methanol, and 10% of a mix of 90% isopropyl alcohol/10% ammonium acetate (1 M)/0.1% formic acid. Hydroquinones present in samples were oxidized with a precolumn 5020 guard cell set in oxidizing mode (E, +650 mV). Quinones were monitored by in-line UV detection (247 and 275 nm), by electrochemical detection (ECD) with an ESA Coulochem III electrochemical detector equipped with a 5011A analytical cell (E1, −650 mV; E2, +650 mV), and by mass spectrometry (MS) with an MSQ Plus spectrometer. The flow was divided after the diode array detector with an adjustable split valve (Analytical Scientific Instruments) to allow simultaneous EC ( [78].