UbiN, a novel Rhodobacter capsulatus decarboxylative hydroxylase involved in aerobic ubiquinone biosynthesis

Ubiquinone (UQ) is a lipophilic electron carrier that functions in the respiratory and photosynthetic electron transfer chains of proteobacteria and eukaryotes. Bacterial UQ biosynthesis is well studied in the gammaproteobacterium Escherichia coli, in which most bacterial UQ‐biosynthetic enzymes have been identified. However, these enzymes are not always conserved among UQ‐containing bacteria. In particular, the alphaproteobacterial UQ biosynthesis pathways contain many uncharacterized steps with unknown features. In this work, we identified in the alphaproteobacterium Rhodobacter capsulatus a new decarboxylative hydroxylase and named it UbiN. Remarkably, the UbiN sequence is more similar to a salicylate hydroxylase than the conventional flavin‐containing UQ‐biosynthetic monooxygenases. Under aerobic conditions, R. capsulatus ΔubiN mutant cells accumulate 3‐decaprenylphenol, which is a UQ‐biosynthetic intermediate. In addition, 3‐decaprenyl‐4‐hydroxybenzoic acid, which is the substrate of UQ‐biosynthetic decarboxylase UbiD, also accumulates in ΔubiN cells under aerobic conditions. Considering that the R. capsulatus ΔubiD‐X double mutant strain (UbiX produces a prenylated FMN required for UbiD) grows as a wild‐type strain under aerobic conditions, these results indicate that UbiN catalyzes the aerobic decarboxylative hydroxylation of 3‐decaprenyl‐4‐hydroxybenzoic acid. This is the first example of the involvement of decarboxylative hydroxylation in ubiquinone biosynthesis. This finding suggests that the C1 hydroxylation reaction is, at least in R. capsulatus, the first step among the three hydroxylation steps involved in UQ biosynthesis. Although the C5 hydroxylation reaction is often considered to be the first hydroxylation step in bacterial UQ biosynthesis, it appears that the R. capsulatus pathway is more similar to that found in mammalians.

. This biosynthetic pathway has been established using gammaproteobacterium Escherichia coli, and 15 proteins (UbiA to K, T to V, and X) are identified to carry out this synthesis [3].E. coli utilizes three flavin-containing monooxygenases (FMO), and UbiI, UbiH, and UbiF for the three aerobic hydroxylation reactions of C1, C5, and C6 positions, respectively (Fig. 1A).However, not all three FMO enzymes are conserved in alphaproteobacteria, suggesting that UQ biosynthesis might differ in these species.Using molecular phylogenetic analyses, Pelosi et al. [4] have classified UQ-biosynthetic FMOs into five groups, UbiF, H, I, L, and M, and alphaproteobacteria utilize UbiL, UbiM, or di-iron protein Coq7 for the three aerobic hydroxylation steps.The hydroxylating positions of Rhodospirillum rubrum UbiL and Coq7 as well as that of UbiM from the betaproteobacterium Neisseria meningitidis have been demonstrated [4].However, the correspondence between specific enzymes and their hydroxylation positions remains unknown in some alphaproteobacteria, including Rhodobacter capsulatus.
Rhodobacter capsulatus is a purple, nonsulfur facultative phototrophic alphaproteobacterium that can grow under aerobic-dark respiratory or anaerobiclight phototrophic conditions and uses UQ as its sole quinone source for both growth pathways.A related purple phototrophic alphaproteobacterium R. rubrum contains rhodoquinone (RQ) in addition to UQ. RQ is synthesized from UQ by RquA protein and is required under anaerobic growth conditions [5,6].Hence, UQ is an essential molecule for even anaerobic growth of R. capsulatus and R. rubrum, unlike E. coli that utilizes menaquinone (MK) under anaerobiosis [7].Available databases indicate that R. capsulatus possesses only two ubiL genes for aerobic UQ biosynthesis.Here, we refer to these two ubiL products as UbiL1 (WP_013066967.1) and UbiL2 (WP_013069027.1).We found that the co-expression of both ubiL1 and ubiL2 (ubiL1-L2) is not able to complement the aerobic growth of R. rubrum DubiL mutant cells.This lack of complementation of R. rubrum DubiL mutant cells suggests that R. capsulatus might have yet another hydroxylase activity involved in UQ biosynthesis.
Independently of the above-described experimental findings, a phylogenetically distant FMO (WP_ 013068139.1)has been detected in the genome of R. capsulatus (see below Fig. 2).Remarkably, this FMO is highly similar to salicylate hydroxylase enzymes of the naphthalene degradation pathway, even though the other enzymes of this pathway are not present in R. capsulatus genome.This observation suggests that this FMO might not be involved in naphthalene degradation and begs the question of whether it plays a role in UQ biosynthesis.Our results demonstrate that the FMO (WP_013068139.1) is responsible for aerobic C1 hydroxylation during UQ biosynthesis, and we name it UbiN.Moreover, we also analyzed the decarboxylation activity of UbiN because its amino acid sequence shows high similarity to a characterized salicylate hydroxylase, known to catalyze the decarboxylative hydroxylation of salicylate [8,9].Our overall results support that UbiN can catalyze the decarboxylative hydroxylation of polyprenylated 4hydroxybenzoic acid.Consequently, as the UQbiosynthetic decarboxylase UbiD is not always conserved among UQ-utilizing proteobacteria [3,[10][11][12], the decarboxylative hydroxylation carried out by an FMO elucidates the basis of aerobic UQ biosynthesis at least in some species containing UbiN.We conclude that the occurrence of this decarboxylative hydroxylation step implies that R. capsulatus UQ biosynthesis is carried out in the same order as mammalian UQ biosynthesis [13].

R. capsulatus ubiL1-L2 co-expression is unable to complement aerobic growth of R. rubrum DubiL
The hydroxylation positions of R. rubrum UbiL (C1,5) and Coq7 (C6) are known based on their expression in E. coli [4].We constructed R. rubrum DubiL and Dcoq7 strains for complementation assays with R. capsulatus genes.These KO strains can synthesize UQ under anaerobic conditions (Fig. 3A), but accumulate UQ-biosynthetic intermediates under aerobic conditions (Fig. 3B).These results agree with the determined hydroxylation positions of R. rubrum UbiL and Coq7.In order to define the hydroxylating positions of two known R. capsulatus FMOs UbiL1 and UbiL2, we expressed separately the ubiL1, ubiL2 and also together the ubiL1-L2 genes in R. rubrum DubiL or Dcoq7 mutant strains lacking the C1, C5hydroxylation or C6-hydroxylation capabilities, respectively (Fig. 3B), and tested their ability to overcome the R. rubrum UQ biosynthesis deficiency (Table 1 and Fig. S1).We found that the aerobic growth defect of R. rubrum Dcoq7 was restored by R. capsulatus ubiL1, indicating that it was expressed in R. rubrum and deduced that UbiL1 catalyzes C6 hydroxylation.Remarkably though, this was not the case for the R. capsulatus ubiL1 or ubiL2 expression, or even ubiL1-L2 co-expression in R. rubrum DubiL.In control experiments, all R. capsulatus genes complemented their cognate mutants under appropriate growth conditions (Fig. S2).Thus, R. capsulatus might have another UQ-biosynthetic hydroxylase besides the UbiL1 and UbiL2.Although R. capsulatus ubiL2 expression in R. rubrum was not known at this point unlike ubiL1 (but see below), these results imply the presence of another UQ-biosynthetic hydroxylase besides the UbiL1 and UbiL2 in R. capsulatus.[14].Initially, this FMO has been annotated as salicylate hydroxylase in some databases, due to its sequence similarity to the known salicylate hydroxylase NahG (Table 2 and Fig. S3).Moreover, as a Coq7 homolog was not detected in R. capsulatus with any homology search program, we inquired whether this FMO (thereafter named UbiN) could be a candidate UQ-biosynthetic hydroxylase.A phylogenetic tree of bacterial FMOs shows that UbiN is distant from any other known UQbiosynthetic FMOs (Fig. 2 and Fig. S4).

UbiN is involved in UQ biosynthesis
To test the participation of UbiN in UQ biosynthesis, ubiN, ubiN-L1, and ubiN-L2 from R. capsulatus were (co-)expressed in R. rubrum DubiL and Dcoq7 strains,

Host (R. rubrum)
Introduced R. capsulatus gene(s) respectively.As a result, both ubiN-L1 and ubiN-L2 co-expressions restored the aerobic growth and quinone contents of R. rubrum DubiL (Table 1, Fig. S1 and Fig. S5).In contrast, the ubiN expression alone was unable to restore the aerobic growth of both R. rubrum DubiL and Dcoq7 mutants.These results show that UbiN is involved in either C1 or C5 hydroxylation in UQ biosynthesis.
To determine the hydroxylating positions of R. capsulatus FMOs, we constructed R. capsulatus DubiL1, DubiL2, and DubiN strains, and their aerobic and anaerobic phototrophic growth phenotypes were examined (Table 3 and Fig. S6).The R. capsulatus FMO-gene KO strains DubiL1 and DubiN were unable to grow under aerobic conditions, while DubiL2 grew much slower.The growth defect of DubiN cells under aerobic conditions supports the involvement of UbiN in aerobic UQ biosynthesis, while the slow aerobic growth of DubiL2 cells suggests that another FMO seems to compensate for the UbiL2 activity.As controls, the KO and WT strains were cultured under anaerobic phototrophic conditions and then the cultures were diluted and transferred to aerobic conditions.All quinols in lipid extract were oxidized to quinones with FeCl 3 prior to injection to eliminate the reduced forms and simplify the chromatograms (Fig. 4A,B).LC-MS analyses of lipids extracted from cells exposed to aerobiosis showed that the DubiN cells accumulate 3-decaprenylphenol (DPP) and 3-decaprenyl-4-hydroxybenzoic acid (DHB) under aerobic conditions (Fig. 4B, orange trace).Considering that polyprenylphenol is a substrate of the first hydroxylase in UQ biosynthesis of E. coli and 3-polyprenyl-4hydroxybenzoic acid a substrate of UQ-biosynthetic decarboxylase UbiD (Fig. 1A), these accumulation patterns suggest that UbiN catalyzes both the decarboxylation and hydroxylation reactions during aerobic UQ biosynthesis (Fig. 1B).

DubiL1 and DubiL2 cells accumulate DMQ 10 and 4-HP 10 , respectively
Lipid extracts of DubiL1 and DubiL2 cells were also analyzed by LC-MS. Figure 4B shows demethoxyubiquinone-10 (DMQ 10 ) accumulation in DubiL1 cells, indicating the C6 hydroxylation deficiency.This result agrees with the complementation assay that shows that UbiL1 has the C6 hydroxylation activity.Remarkably, DubiL2 cells accumulated an intermediate (m/z 811.6) at a retention time of 10.3 min, and the ring structure of the oxidized intermediate was confirmed by LC-PDA (photodiode arrays detector) analyses (Fig. S7).Additionally, this intermediate was predominantly detected as an unoxidized form (m/z 813.6) from the lipid extract of DubiL2 when the FeCl 3 treatment was omitted (data not shown), indicating that the accumulated intermediate in DubiL2 cells corresponded to 3-decaprenyl-4-hydroxyphenol (4-HP 10 ).Indeed, a similar accumulation of 4-HP 8 in E. coli DubiI cells has been reported earlier [15].

UbiN catalyzes decarboxylative hydroxylation
Homology searches indicated that the UbiN sequence is more similar to salicylate hydroxylases than the UQbiosynthetic FMOs (Table 2 and Fig. S3).Salicylate hydroxylase converts salicylic acid into catechol via decarboxylative hydroxylation [8,9].Since the DubiN cells accumulate the non-decarboxylated intermediate, UbiN was therefore predicted to catalyze a decarboxylative hydroxylation reaction.To confirm the decarboxylation activity of UbiN, a R. capsulatus DubiD-X double mutant strain was constructed as UbiX synthesizes prenylated FMN required as cofactor for UbiD [16].The R. capsulatus ubiD and ubiX genes being in the same operon, the ubiD-X double KO strain was obtained using a single kanamycin-resistance cassette.As another control, the R. capsulatus DubiU-V double KO strain was also constructed as the UbiU and UbiV are anaerobic hydroxylases identified in E. coli and Pseudomonas aeruginosa [17,18].Both the R. capsulatus DubiD-X and DubiU-V cells show growth defects under anaerobic phototrophic conditions (Table 3 and Fig. S6).The DubiD-X and DubiU-V mutants as well as the WT cells as a control were grown aerobically and then transferred to anaerobic phototrophic conditions, and their lipid extracts were analyzed by LC-MS as before (Fig. 4C,D).Note that the accumulated intermediates have the same retention time and m/z as those seen with DubiN, and the aromatic structure of the intermediate accumulated in DubiD-X was confirmed by LC-PDA analyses.Comparison of the UV spectra of the intermediate to that of 4-hydroxy-3-methylbenzoic acid shows that R. capsulatus DubiD-X and DubiN accumulated the same intermediate, DHB (Fig. S7).In addition, no intermediate accumulation was detected in DubiD-X cells under aerobic conditions (Fig. 4C). Figure 5 shows the time course of UQ 10 and biosynthetic intermediates contents of DubiN and WT cells after For further confirmation of UbiN activities, the E. coli ubiD-H double KO and ubiH KO strains were constructed.ubiN was expressed in these KO strains under aerobic conditions and quinone contents were analyzed by HPLC-PDA.The KO strains transformed with the expression control vector accumulated 3-octaprenyl-4-hydroxybenzoic acid (OHB) and 3-octaprenylphenol (OPP), respectively, and UQ 8 was not detected in these KO strains (Fig. 6).As expected, ubiN expression restored UQ 8 biosynthesis in the ubiD-H KO strain.In addition, the ubiH KO strain was also rescued by the ubiN expression.decarboxylative hydroxylation of OHB and hydroxylation of OPP.

Discussion
Hydroxylating positions of R. capsulatus FMOs The enzyme diversity that occurs among three hydroxylation reactions for aerobic UQ biosynthesis in bacteria has been well described by Pelosi et al. [4], demonstrating that one enzyme may catalyze more than one hydroxylation reaction, especially when less than three UQ-biosynthetic monooxygenases are found in some bacteria, assuming the absence of additional candidates.Starting with this idea, we uncovered a new FMO clade, called UbiN, and experimentally confirmed the occurrence in R. capsulatus of three FMOs, UbiL1, UbiL2, and UbiN that are responsible for UQ biosynthesis.As expected, each of the corresponding KO strains accumulated different UQ intermediates, and we concluded that these FMOs have distinct roles in the aerobic UQ biosynthesis, while UbiL1 and UbiL2 belong to the same phylogenetic group.The accumulation patterns of the UQ intermediates in the mutant cells defined likely specific hydroxylating positions of these three FMOs.R. capsulatus DubiN cells accumulated both DPP and DHB, like the E. coli DubiH and DubiD cells accumulating 3ocataprenylphenol (OPP) and 3-octaprenyl-4hydroxybenzoic acid (OHB), respectively [19,20].As E. coli UbiH is considered to catalyze the C1 hydroxylation, R. capsulatus UbiN is assumed to catalyze the C1 hydroxylation.However, we note that accumulation of polyprenylphenol might not always represent the C1 hydroxylation deficiency as follows: the 3-ocataprenylphenol accumulation occurs in E. coli DubiB and DubiG as well as in DubiH cells, while UbiB and UbiG are not hydroxylases [21,22].To demonstrate firmly that UbiN is C1 hydroxylase, the hydroxylation ability of UbiN was shown by the complementation assays of R. rubrum DubiL and E. coli KO strains.Furthermore, our results show that UbiN catalyzes the decarboxylative hydroxylation of polyprenylhydroxybenzoates.Besides, UbiN also catalyzes the hydroxylation of polyprenylphenols, which is the decarboxylated intermediate (Fig. 1).If UbiN is not able to hydroxylate DPP, then this intermediate would have been accumulated even in the R. capsulatus WT cells under aerobic conditions because ubiD is expressed [23] and is functional (Fig. 5) under both aerobic and anaerobic conditions [23].
Rhodobacter capsulatus DubiL1 cells accumulated significant amounts of DMQ 10 , demonstrating the C6 hydroxylation activity of UbiL1.When R. rubrum Dcoq7 cells show the growth defect and DMQ accumulations under aerobic conditions (Fig. 3B), the ubiL1 expression restores the growth defect, further supporting the C6 hydroxylation activity of UbiL1.On the contrary, the R. capsulatus DubiL2 cells accumulated 4-HP 10 , reminiscent of E. coli DubiI cells, which have the C5 hydroxylation defect, and accumulate 4-HP 8 [15], indicating that UbiL2 is likely to catalyze the C5 hydroxylation.Additionally, the aerobic growth of R. rubrum DubiL cells was restored by the ubiN-L1 coexpression as well as the ubiN-L2 co-expression, possibly due to the residual C5 hydroxylation activity of UbiL1.Likewise, DubiL2 cells of R. capsulatus show slow growth as compared to WT cells under aerobic conditions, unlike the severe growth defects seen with the DubiN and DubiL1 cells under aerobic conditions.Interestingly, similar to the residual C5 hydroxylase activity seen in R. capsulatus UbiL, a residual C5 hydroxylation activity for E. coli C6 hydroxylase UbiF has also been reported [15].However, the molecular phylogenetic analysis based on amino acid sequences indicates that these two C6 hydroxylases are phylogenetically distant FMOs (Fig. 2).Thus, the hydroxylation positions of FMOs are not always correlated with their molecular phylogenetic classifications.

Remodeling the bacterial UQ-biosynthetic pathway
The UQ-biosynthetic pathway contains three hydroxylation steps, and the hydroxylating order was determined in E. coli, based on intermediate accumulation analyses.Primarily, the accumulated intermediates in the E. coli KO strains are regarded as direct substrates of the gene products.However, 4-HP 8 is not regarded as the substrate of UbiI, and the 4-HP 8 accumulation is considered to be due to 3-octaprenylphenol (OPP) hydroxylation activity of UbiH [15].This interpretation was based on 3-octaprenyl-5-methoxyphenol (OMP) being the original substrate of UbiH.In 1973, Young et al. [20] reported that OPP and trace amounts of OMP were accumulated in E. coli DubiH cells.Until UbiI was identified, UbiB has been annotated as the C5 hydroxylase; thus, OPP was not recognized as the original substrate of UbiH.Based on the idea that UbiN catalyzes the decarboxylative hydroxylation, it is assumed that the carboxylated C1 position is first hydroxylated and then decarboxylated sequentially, because the decarboxylation activity of FMO is initiated by electrophilic aromatic substitution [8,9].Hence, the C1 hydroxylation should be the first step of all three hydroxylation steps in R. capsulatus.This pathway agrees with the 4-HP accumulations since the decarboxylative hydroxylation step produces 4-HP.Although the C5 hydroxylation reaction is often considered to be the first hydroxylation step in bacterial UQ biosynthesis, it appears that the pathway in R. capsulatus is more similar to that found in mammalians [13].homologs are not essential for aerobic bacteria possessing UbiN homologs.This hypothesis is consistent with the absence of UbiD in some UQ-producing organisms.Indeed, it is impossible to conclude that all UbiN homologs are involved in the UQ biosynthesis, because substrate specificities of UQ-biosynthetic FMOs are inconsistent with their amino acid sequence similarity (Fig. 2).Moreover, some bacteria such as Zymomonas mobilis lack both ubiD and ubiN homologs, suggesting the presence of potential decarboxylation activity of C1 hydroxylating FMOs, or unknown decarboxylases.Furthermore, some genera of facultatively anaerobic phototrophs, such as Rhodomicrobium and Rhodovulum which possess UbiN homologs, lack ubiD homologs in their genomes.These phototrophs also lack any homolog of the candidate decarboxylase ubiZ [11].As our results show that the decarboxylation ability of UbiN is limited to aerobic conditions, such facultative anaerobes might possess another anaerobic decarboxylation system.Our next aim will be to investigate the anaerobic UQ biosynthesis in such species.

Gene-knockout and complementation assays
Rhodospirillum rubrum and R. capsulatus gene-knockout (KO) mutants were constructed by homologous recombination, and all constructed KO strains are listed in Table S1.To knockout coq7 gene in R. rubrum, coq7 gene fragment with 500 bp 5 0 upstream and 479 bp 3 0 downstream was PCR-amplified and ligated into pBluescript II KS(+) using XbaI and KpnI restriction sites.A kanamycin-resistance cassette from pHP45Ω-Km [26] and double-digested pBluescript-coq7 with PspOMI and BlpI were ligated after blunted by T4 DNA polymerase.The coq7::kan fragment obtained by XbaI and KpnI double digestion was inserted into the SphI site of the suicide vector pZJD29a [27] as blunt end ligation.To knockout ubi-genes, upstream and downstream regions of each gene were separately PCR-amplified.The fragment sets and a kanamycin-resistance cassette were combined at the HindIII site and then cloned into the suicide vector pZJD29c at the XbaI and KpnI sites.pZJD29c is a derivative of pZJD29a and constructed by removing the KpnI site of sacB gene.These plasmids were introduced into E. coli S17-1 kpir and used as donor strains for conjugation.Liquid cultures of the recipient and the donor stains were mixed on Sistr€ om's Medium A agar plate and then incubated for 24 h.The cell mixtures were plated on kanamycincontaining Sistr€ om's Medium A agar plate, and then, kanamycin-resistant recipients were selected.Replacement of the targeted genomic allele with a kanamycin-resistance cassette was verified by colony PCR and gentamicin sensitivity, showing the absence of pZJD29c sequence.The monooxygenase-gene KO strains were purified under anaerobic phototrophic conditions, while the R. capsulatus ubiU-V and ubiD-X KO strains were purified under aerobic conditions to avoid reversions, respectively.
The constructed R. rubrum KO strains were transformed with R. capsulatus FMO genes and used for the complementation assays.R. capsulatus FMO genes and empty vector (pRK415 [28]) as control were introduced via conjugation.pRK2013/E.coli HB101 was used as the mobilization helper [28].After conjugations, tetracyclineresistant recipients were selected, verified by colony PCR, and purified under anaerobic phototrophic conditions.
Escherichia coli ubiD KO mutants were constructed using BW25113, obtained from NBRP (NIG, Japan): E. coli, as a parental strain and by the same recombination strategy used for R. capsulatus.Deletions of ubiH were carried out for BW25113 and the ubiD KO strain.The exact ORF of ubiH was removed without any antibiotic resistance cassette not to affect the expression of ubiI, because ubiH and ubiI form a gene cluster.First, approximately 500 bp of upstream and downstream regions of the ubiH gene were separately PCR-amplified.Then, these two fragments were combined at the EcoRI site and then cloned into the suicide vector pZJD29c at the XbaI and BamHI sites.After the plasmid was introduced into BW25113 and the ubiD KO strain, singlecrossover integrant strains were selected using gentamicin resistance.The double-crossover stains were obtained by sacB counter selection.All transformations of E. coli were performed by electroporation, and all E. coli mutants were constructed and cultured under anaerobic conditions using AnaeroPack to prevent reversion.

Cloning and plasmid constructions
Rhodobacter capsulatus ubiN, ubiL1, and ubiL2 containing regions were separately PCR-amplified from genomic DNA.Primer sets were designed to produce fragments that include promoter regions and excluding other ORFs.Desired two-fragment pairs were combined by T4 DNA kinase and ligase for co-expression.The single gene fragments or the dual gene fragments were cloned into the pRK415 vector at the XbaI and KpnI sites.These plasmids were introduced into E. coli HB101 and sequenced.R. capsulatus ubiN was cloned into pTrcHis-TOPO vector (Invitrogen, CA, USA) for expression in E. coli.All primers and plasmids used in this study are listed in Tables S2 and S3, respectively.

In silico protein search and phylogenetic analyses
Using R. rubrum UbiL (WP_011391454.1) and Pseudomonas putida NahG (WP_011475386.1)sequences as query, BLASTP and DELTA-blast searches (BLAST 2.13.0+) were performed against the predicted proteome of R. capsulatus to detect FMOs on NCBI server.
FMO sequences used for the phylogenetic analysis were obtained from the 67 strains used in Pelosi et al. [4].First, all possible orthologues were obtained by BLASTP version 2.5.0 [29] using all FMO enzymes provided in Table S1 of Pelosi et al. [4] as query and an e-value threshold of 1e-10.Resultant sequences were screened for having FAD-binding motifs by HMMSEARCH version 3.1b2 [30] with PF01494 model and were aligned using MUSCLE version 3.8.31[31].After removing partial sequences and trimming ambiguous sites, the matrix composed of 200 sequences and 157 sites was used for phylogenetic analysis.Phylogenetic analysis was performed by PHYML-SMS web service [32,33].We used LG substitution matrix with supposing 4-category discretegamma-distributed rates with invariant sites and using empirical equilibrium frequencies (LG + G4 + I + F).The sequences in the phylogenetic tree were assigned to KEGG Orthology by BLASTKOALA web service [34].All FMOs used for the phylogenetic analyses are listed in Table S4.

Qualitative analysis for identification of UQbiosynthetic intermediates
Overnight anaerobic cultures of R. capsulatus FMO-gene KO strains and WT were inoculated into 5 mL of Sistr€ om's Medium A in 15 mL tubes for OD 630 = 0.05 and then cultured under anaerobic phototrophic conditions for 16 h.One milliliter aliquots of the cultures were diluted six times in 50 mL tubes with the medium and cultured under aerobic-dark conditions with shaking at 130 r.p.m. for 4 h.Overnight aerobic cultures of R. capsulatus DubiD-X, D-ubiU-V, and WT were inoculated into 5 mL medium in 50 mL tubes for OD 630 = 0.05 and then cultured under aerobic-dark conditions with shaking at 130 r.p.m. for 16 h.One milliliter aliquots of the cultures were diluted six times in 15 mL tubes with the medium and cultured under anaerobic phototrophic conditions for 4 h.The aerobic and anaerobic R. capsulatus cultures were centrifuged, and wet weights were determined.Cell pellets were stored at À30 °C until use.Lipid extractions were performed by adding 20 lL 2-propanol against 1 mg cell pellets, vortexing, and sonication.Lipid extracts were obtained through centrifugation and filtration by a 0.2 lm filter.Oxidation of quinols to quinones in lipid extracts was achieved by adding FeCl 3 to a final concentration of 2.4 mM.Four microliter aliquots were used for quinone contents analyses by HPLC-ESI-TOF-MS (Nanofrontier LD, Hitachi High-Technologies, Tokyo, Japan).All HPLC analyses were performed on a C18 reversed-phase column (Inertsil ODS-3, 2 lm, 2.1 9 5.0 mm, GL Sciences, Tokyo, Japan).A mixture of 80% methanol, 20% 2-propanol, and 0.1% formic acid was used as mobile phase at a flow rate of 0.2 mLÁmin À1 .UQ and their derivatives were mainly observed as [M + Na] + with our LC-MS system.
UQ-biosynthetic intermediates were identified by HPLC-PDA analyses on the same column as HPLC-MS analyses.One hundred microliter aliquots of oxidized lipid extracts were condensed into approx.Twenty microliter by N 2 gas spray and then 4 lL aliquots were used.A mixture of 90% methanol, 10% 2-propanol, and 0.1% formic acid was used as a mobile phase at a flow rate of 0.2 mLÁmin À1 .ptoluquinone, 4-hydroxy-3-methylbenzoic acid, and o-cresol were used as standards for the UV spectra.All standards were purchased from the Tokyo Chemical Industry (Tokyo, Japan).

Quantification of UQ and biosynthetic intermediates
The E. coli KO strains transformed with pTrcHis-ubiN and control strains were precultured under anaerobic conditions for 8 h with 2 mL LB medium (+glucose, +potassium phosphate buffer) in 4 mL tubes and then inoculated into 30 mL LB medium (+fructose, +IPTG, and +potassium phosphate buffer) in 200 mL flasks for OD 600 = 0.005 and grown aerobically for 16 h with shaking at 180 r.p.m.Lipid extractions were performed after adding 20 lL 2-propanol containing 5 lM UQ 10 as an internal standard per 1 mg cell pellets (wet weights).Two hundred and fifty microliter aliquots were dried using centrifugal evaporator and then dissolved in 25 lL 2-propanol containing 2.4 mM FeCl 3 .Concentrated lipid extracts were obtained by centrifugation through a 0.2 lm filter.Four microliter aliquots of concentrated extracts were used for quinone contents analyses.Quantification analyses were performed by HPLC-PDA (275 nm) using the same column and mobile phase used for the identification of UQ-biosynthetic intermediates in R. capsulatus.Twenty milliliter anaerobic phototrophic cultures of R. capsulatus DubiN and WT were diluted 12 times in 1 L flasks with the medium and started aerobic culture under dark conditions with shaking at 180 r.p.m. for 24 h.Twenty milliliter aliquots were sampled at 1, 2, 4, 6, 8 and 24 h.Lipid extractions were performed after adding 20 lL 2-propanol containing 5 lM diethoxyubiquione-10 (DEQ 10 ) as an internal standard per 1 mg cell pellets (wet weights).DEQ 10 was synthesized in house, essentially as described by Edlund [35].Four microliter aliquots of oxidized extracts were analyzed by UPLC-PDA (275 nm) on a C18 reversed-phase column (ACQUITY UPLC HSS T3, 1.8 lm, 2.1 9 5.0 mm, Waters).Mobile phase and linear gradient conditions are as follows: (A) methanol:H 2 O (3:7), (B) methanol, and (C) 2-propanol.All three solvents contained 0.1% formic acid.0-6 min: (A) 10%, (B) 54-18%, and (C) 36-72%.6-8 min: isocratic.8-10 min: isocratic, (A) 10%, (B) 9%, and (C) 81%.Flow rate was 0.4 mLÁmin À1 .Amounts of DHB and DPP in lipid extracts were estimated using 275 nm absorption of 4-hydroxy-3methylbenzoic acid and o-cresol in 2-propanol (0.1% formic acid), respectively.Quinone contents analyses of R. rubrum KO strains were performed following the same methods described above except for the growth time.R. rubrum KO strains and WT were grown under anaerobic phototrophic conditions for 3 days and then diluted and cultured under aerobic conditions for 36 h.

Fig. 5 .
Fig. 5. Time course analysis of UQ 10 and biosynthetic intermediates in DubiN cells.UQ 10 , DHB, and DPP were quantified in DubiN and WT cells after exposure to aerobiosis at 0, 1, 2, 4, 6, 8, and 24 h.Experiments were repeated four times (n = 4), and the error bars are standard deviations.

Fig. 6 .
Fig. 6. ubiN restored UQ 8 biosynthesis in E. coli DubiD-H and DubiH cells.(A-C) HPLC-PDA chromatograms of E. coli KO strains and WT transformed with pTrcHis-ubiN or pTrcHis-lacZ as a control, monitored at a wavelength of 275 nm.UQ 10 was used as an internal standard.(A) ubiN restored UQ 8 biosynthesis in DubiD-H double KO strain.DubiD-H/LacZ cells accumulated OHB.(B) ubiN restored UQ 8 biosynthesis in DubiH KO strain.DubiH/LacZ cells accumulated OPP.(C) WT has UQ 8 and MK 8 as major quinones.(D) UQ 8 of the strains was quantified.Experiments were repeated three times (n = 3), and the error bars are standard deviations.

2090FEBS
Open Bio 13 (2023) 2081-2093 ª 2023 The Authors.FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

Table 2 .
Percent identity matrix of FMOs created by MUSCLE 3.8.
FEBS Open Bio 13 (2023) 2081-2093 ª 2023 The Authors.FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

Table 3 .
Aerobic and anaerobic growth phenotypes of appropriate R. capsulatus KO strains.All stains were cultivated for 3 days.++, grow as WT; +, slow growth; À, not grow.