SUMMARY
Most bacteria can generate ATP by respiratory metabolism, in which electrons are shuttled from reduced substrates to terminal electron acceptors, via quinone molecules like ubiquinone. Dioxygen (O2) is the terminal electron acceptor of aerobic respiration and serves as a co-substrate in the biosynthesis of ubiquinone. Here, we characterize a novel, O2-independent pathway for the biosynthesis of ubiquinone. This pathway relies on three proteins UbiT (YhbT), UbiU (YhbU) and UbiV (YhbV). UbiT contains an SCP2 lipid-binding domain and is likely an accessory factor of the biosynthetic pathway, while UbiU-UbiV are involved in hydroxylation reactions and represent a novel class of O2-independent hydroxylases. We demonstrate that UbiU-UbiV form a heterodimer, wherein each protein binds a 4Fe-4S cluster via conserved cysteines that are essential for activity. The UbiT, -U, -V proteins are found in α-, β-, γ-proteobacterial clades including several human pathogens, supporting the widespread distribution of a previously-unrecognized capacity to synthesize ubiquinone in the absence of O2. Together, the O2-dependent and O2-independent ubiquinone biosynthesis pathways contribute to optimize bacterial metabolism over the entire O2 range.
IMPORTANCE In order to colonize environments with large O2 gradients or fluctuating O2 levels, bacteria have developed metabolic responses that remain incompletely understood. Such adaptations have been recently linked to antibiotic resistance, virulence and the capacity to develop in complex ecosystems like the microbiota. Here, we identify a novel pathway for the biosynthesis of ubiquinone, a molecule with a key role in cellular bioenergetics. We link three uncharacterized genes of Escherichia coli to this pathway and show that the pathway functions independently from O2. In contrast, the long-described pathway for ubiquinone biosynthesis requires O2 as substrate. In fact, we find that many proteobacteria are equipped with the O2-dependent and O2-independent pathways, supporting that they are able to synthesize ubiquinone over the entire O2 range. Overall, we propose that the novel O2-independent pathway is part of the metabolic plasticity developed by proteobacteria to face varying environmental O2 levels.
INTRODUCTION
Since the oxygenation of the Earth’s atmosphere some 2.3 billion years ago, many organisms adopted dioxygen (O2) as a terminal electron acceptor of their energy-producing respiratory chains [1]. Indeed, oxygenic (aerobic) respiration has a superior energetic output compared to anaerobic respiration or fermentation, which are both O2-independent processes [1]. In fact, several microorganisms, including many important human pathogens, are facultative anaerobes that are able to adopt either an aerobic or an anaerobic lifestyle depending on the environmental conditions [2,3]. In the laboratory, bacteria are usually cultured and studied under fully aerobic or completely anaerobic conditions (absence of O2) [4], whereas natural habitats cover the entire range of O2 concentrations [5]. For instance, large O2 gradients are typically encountered in the human large intestine, in biofilms or in transition zones between oxic and anoxic environments [5]. Moreover, bacteria can experience rapid transitions between environments with vastly different O2 contents, as during the infection process of enteric pathogens that progress along the gastrointestinal tract [3].
To maximize their bioenergetic capacities according to the various levels of O2 encountered in their environment, bacteria modulate the composition of their respiratory chains, notably the quinone species and the terminal reductases [3,4,6]. Quinones are lipophilic redox molecules that fuel electrons to terminal reductases, which reduce O2 whenever available, or alternative electron acceptors for instance nitrate, dimethyl sulfoxide (DMSO), trimethylamine N-oxide [7]. Naphtoquinones (menaquinone (MK) and demethylmenaquinone (DMK)) and ubiquinone (UQ) are the two main groups of bacterial quinones. (D)MK and UQ differ by the nature of their head group and the value of their redox midpoint potential [8]. (D)MK are considered “anaerobic quinones” since they function primarily in anaerobic respiration whereas UQ is considered an “aerobic quinone” because its supplies electrons mostly to the reductases that reduce O2 [1,8,9]. Accordingly, UQ is the main quinone of the facultative anaerobe Escherichia coli in aerobic conditions, whereas the naphtoquinones are predominant in the absence of O2 [10,11], UQ being nevertheless present.
The biosynthesis of UQ requires a total of eight reactions to modify the aromatic ring of the precursor, 4-hydroxybenzoic acid (4-HB): one prenylation, one decarboxylation, three hydroxylation and three methylation reactions (Figure 1A) [12]. In addition to the enzymes that catalyze the various steps, three accessory factors - UbiB, UbiJ and UbiK – are also needed. UbiB has an ATPase activity [13] and we showed that UbiJ and UbiK [14,15] belong to a multiprotein UQ biosynthesis complex, in which the SCP2 domain (sterol carrier protein 2) of UbiJ binds the hydrophobic UQ biosynthetic intermediates [16]. This UQ biosynthetic pathway is under the dependence of O2 since all three hydroxylases - UbiI, UbiH and UbiF – use O2 as a cosubstrate (Figure 1A) [17,18]. We showed recently that other hydroxylases, UbiL, UbiM and Coq7, replace UbiI, UbiH and UbiF in some proteobacteria [19]. The six hydroxylases have in common their dependence to O2 and thus function in UQ biosynthesis only when sufficient O2 is available. Interestingly, Alexander and Young established 40 years ago that E. coli was able to synthesize UQ anaerobically [20], suggesting the existence of an O2-independent biosynthesis pathway, which is still uncharacterized.
In this study, we describe the O2-independent UQ biosynthetic pathway in E. coli and identify three essential components, the UbiT, UbiU and UbiV proteins, formerly called YhbT, YhbU and YhbV. We show that the O2-independent UQ biosynthetic pathway is widely conserved in proteobacteria. UbiT likely functions as an accessory factor in the O2-independent UQ biosynthetic pathway and we show that UbiU and UbiV are involved in at least one O2-independent hydroxylation reaction. Moreover, we demonstrate that both UbiU and UbiV bind a [4Fe-4S] cluster essential for activity, which identifies these proteins as prototypes of a new class of O2-independent hydroxylases. Our results highlight that many proteobacterial species use two different and complementary molecular pathways to produce UQ over the entire continuum of environmental O2.
RESULTS
4-HB is the precursor of UQ synthesized in anaerobic conditions
4-HB is the precursor of UQ synthesized in aerobic conditions [21]. Accordingly, an E. coli ΔubiC mutant impaired in 4-HB biosynthesis [22], is deficient in UQ8 and is complemented by addition of 4-HB to the growth medium [23]. In order to evaluate whether 4-HB is also the precursor of the O2-independent UQ biosynthetic pathway, we grew a ΔubiC strain anaerobically. The ΔubiC strain showed a diminished level of UQ8, which was partially recovered by supplementation with 4-HB (Figure 1B). Furthermore, we grew the ΔubiC strain in a medium supplemented with 13C7-4HB and we analyzed the labelling of biosynthesized UQ8 by HPLC-mass spectrometry (MS). In cells grown in aerobic or anaerobic conditions, the labelled form of UQ (13C6–UQ8, m/z= 750.4 for the NH4+ adduct) represented 98.3% and 97.3% (±0.2%) of the total UQ8 pool, respectively (Figure 1C, 1D). As expected, ΔubiC cells grown anaerobically with unlabeled 4-HB didn’t show any 13C6-UQ8 (Figure 1E). Altogether, these results establish that 4-HB is the precursor of the O2-independent UQ biosynthetic pathway.
Ubi enzymes, except hydroxylases, are common to the aerobic and anaerobic UQ biosynthesis pathways
The above result suggests that the UQ biosynthetic pathways decorate 4-HB with the same chemical groups irrespective of the presence of environmental O2. Thus, we evaluated whether the enzymes of the aerobic pathway are also involved in the O2-independent pathway by measuring the UQ8 content of knock-out (KO) strains grown in aerobic and anaerobic conditions. Deletion of ubiA, ubiE or ubiG abrogated UQ8 biosynthesis in both conditions whereas ΔubiB, ΔubiD or ΔubiX strains synthesized a limited amount of UQ8 but only in aerobic conditions (Figure 1F). In contrast, ubiJ and ubiK had no effect on UQ biosynthesis under anaerobic conditions (Figure 1F).
In aerobic conditions, the hydroxylation reactions are catalyzed by the flavin monooxygenases (FMOs) UbiF, UbiH and UbiI that use dioxygen as a co-substrate [17,18,24] (Figure 1A). We previously reported that cells deleted for a single O2-dependent hydroxylase (ΔubiF, ΔubiI or ΔubiH) were deficient in UQ, when cultured in the presence of air, but synthesized UQ in anaerobic conditions [17], consistent with the existence of an alternative hydroxylation system in the O2-independent pathway [20]. Indeed, we confirmed that all three FMOs are dispensable for the O2-independent UQ biosynthetic pathway since a triple mutant ΔubiF ΔubiI ΔubiH was deficient for UQ when grown in air, but synthesized wild-type (WT) level of UQ8 in anaerobic conditions (Figure 1G). Together, our results demonstrate that the O2-dependent and O2-independent UQ biosynthetic pathways share the enzymes involved in the prenylation (UbiA), decarboxylation (UbiX and UbiD) and methylation reactions (UbiE and UbiG), but differ by their hydroxylases and the accessory factors UbiJ and UbiK (Figure 1H).
Identification of three genes required for UQ biosynthesis in anaerobic conditions
To identify genes involved in the O2-independent UQ biosynthetic pathway, we cultivated anaerobically a collection of ∼ 200 E. coli strains that contained deletions covering multiple ORFs [25,26] and we analyzed their UQ8 content by HPLC-electrochemical detection (HPLC-ECD). We found a complete absence of UQ8 in strains ME4561, ME5034 and ME4746 that carry deletions encompassing, ubiE-ubiJ-ubiB-ubiD, ubiG, and ubiX, respectively (Table S1). Several other strains had a low UQ8 content and a poor growth in synthetic medium supplemented with glycerol and nitrate (SMGN). However, those strains showed better growth and higher UQ8 content in LB medium (Table S1). Thus we did not investigate them further, as a genetic defect affecting directly the O2-independent UQ pathway was unlikely. In contrast, ME4641 showed a profound UQ8 deficiency and robust anaerobic growth in LB and SMGN media (Table S1). Importantly, ME4641 had a WT UQ8 level when grown aerobically, suggesting that only the O2-independent pathway was altered (Figure 2A). ME4641 contains a deletion named OCL30-2 that covers 9 genes, 5 of them lacking an identified function (Figure 2B). To find the candidate gene involved in the anaerobic biosynthesis of UQ, we obtained 8 single gene KO strains from the Keio collection [27] and analyzed their quinone content after anaerobic growth (Figure 2C). The ΔyhbT and ΔyhbU strains were strongly deficient in UQ8. We then transduced the ΔyhbT and ΔyhbU mutations from the Keio strains into a MG1655 genetic background and also constructed the ΔyhbV strain, which was not available in the Keio collection. We found that all three strains had very low levels of UQ8 when grown in anaerobic conditions but showed normal levels after aerobic growth (Figure 2D, Table 1). In addition, the mutant strains showed a two-fold decrease in MK8 and a two-fold increase in DMK8 after anaerobic growth (Table 1). This effect might indirectly result from the UQ8 deficiency as it was also observed in the ΔubiG strain (Figure S1A).
Deletion of yhbT, yhbU or yhbV causes UQ8 deficiency specifically in anaerobic conditions
We then transformed the mutant strains with an empty vector or a vector carrying a WT allele of the studied gene. In yhb KO strains expressing the corresponding gene from the plasmid, we observed a complementation of the UQ8 deficiency (Table 1) and a normalization of the levels of octaprenylphenol (OPP), an early UQ8 biosynthetic intermediate (Figure 2E). The ∼3 fold elevation of OPP in the yhbT, -U, - V KO mutants suggested that the O2-independent UQ biosynthetic pathway was blocked downstream of OPP in these strains. In these experiments, no cross-complementation was observed - for example the plasmid with yhbV had no effect in ΔyhbT or ΔyhbU strains - suggesting the absence of redundancy in the function of each gene. We also measured a profound UQ8 deficiency when the ΔyhbT, ΔyhbU and ΔyhbV strains were grown anaerobically in various media (glycerol + DMSO, lactate + KNO3) (Figure 2F), showing that the UQ8 biosynthetic defect is not linked to a particular carbon source or electron acceptor. Altogether, our results demonstrate that the yhbT, yhbU and yhbV genes are part of the O2-independent UQ biosynthetic pathway, so we propose to rename them ubiT, ubiU and ubiV, respectively.
Complete absence of UQ biosynthesis in ΔubiT, ΔubiU or ΔubiV mutants grown under strict anaerobic conditions
As the ΔubiT, ΔubiU or ΔubiV strains still contained small amounts of UQ8 after growth in anaerobic conditions (Table 1, Figure 2F), we wondered how this UQ8 was synthesized. To verify if the O2-dependent pathway contributed to this synthesis, we inactivated the ubiH gene in the ΔubiU strain. Sensitive HPLC-MS detection established that UQ8 was completely absent in extracts from the ΔubiH ΔubiU cells (Figure 2G). This result supported that the residual UQ8 synthesized in ΔubiU cells originated from the O2-dependent pathway and suggested that our anaerobic media may contain traces amounts of O2 or that O2-dependent UQ biosynthesis may occur during the handling of cells, under normal atmosphere, prior to quinone extraction.
To eliminate the traces of O2, we took extra precautions in the degassing and inoculation of our media (see Material and Methods) and also added a reductant (L-Cysteine). In addition, we used the redox indicator resazurin to verify strict anaerobiosis during the entire culture. We also modified our sampling procedure to rapidly quench the anaerobic cells in ice-cold methanol, in order to prevent any O2-dependent UQ biosynthesis prior to quinone extraction. HPLC-MS analysis of extracts from cells cultivated and handled under such strict anaerobic conditions showed the nearly complete absence of UQ8 in ΔubiT, ΔubiU and ΔubiV strains (Figure 2H, 2I). In contrast, the UQ8 level of the WT strain (Figure 2I) was comparable to those we measured previously in WT cells cultivated in suboptimal anaerobic conditions (Table 1) [14], establishing that UQ biosynthesis occurred independently from O2. Together, our results show that ΔubiT, ΔubiU and ΔubiV cells are unable to synthesize UQ under strict anaerobic conditions unlike WT cells. Furthermore our results support that the low residual UQ content previously observed in ΔubiT, ΔubiU and ΔubiV strains (Table 1, Figure 2F) resulted from the function of the O2-dependent pathway.
ubiT,-U,-V strongly co-occur, exclusively in genomes with potential for UQ biosynthesis
We investigated the distribution of ubiT, -U and -V in a large genome dataset gathering 5750 genomes of bacteria and archaea. We found no evidence of genomes harboring matches for more than one of the three genes of interest outside of Alphaproteobacteria, Betaproteobacteria and Gammaproteobacteria (α-, β-, γ-proteobacteria) and three genomes of Acidithiobacillia (Table S2A). Interestingly, the three former classes are the only known so far to be able to produce UQ [1,8], consistent with a specific link between UbiT, -U, -V and UQ biosynthesis.
This was confirmed by analyzing in more details the distribution of ubiT, -U and -V and of three marker genes of the UQ biosynthetic pathway (ubiA, ubiE and ubiG) in 611 representative and reference genomes of α-, β-, γ-proteobacteria (Table S2B). We chose ubiA, -G, -E because they have a higher conservation compared to other genes like ubiC, -D, -X [28] and because they are part of the O2-dependent and O2-independent pathways (Figure 1H). 575 genomes had positive matches for ubiA, -G, -E and 589 for at least two of them. Regarding the distribution of ubiT, -U and -V, 221 genomes had matches for at least one of the three genes, and in 210 cases (95%), the three genes were present. Importantly, all genomes with ubiT, -U and -V harbored at least two of the three marker genes for the UQ biosynthetic pathway (Table S2B). In addition, we found that 22 out of the 29 proteobacterial orders analyzed had up to 50% genomes harboring a complete set of the ubiT, -U and -V genes (Figure 3A), demonstrating a wide taxonomic distribution. Overall, our analysis indicates a strong pattern of co-occurrence of ubiT, -U and -V, and demonstrates that they are uniquely found in genomes showing signs of UQ production.
We found ubiA, -G, -E and O2-dependent hydroxylase genes in 570 genomes, of which 204 also contained ubiT, -U and –V (Table S2B and Figure 3B). Only 3 species, Phaeospirillum fulvum, Magnetococcus marinus and Oxalobacter formigenes, seem to rely exclusively on the O2-independent pathway for UQ production as their genomes contain the ubiT, -U and -V genes but no O2-dependent hydroxylases (Figure 3B). We noticed that the ubiT, -U and -V genes are found next to other ubi genes in M. marinus (Figure S1B). Interestingly, P. fulvum and M. marinus were described to synthesize UQ in microaerobic conditions [29-31] and O. formigenes has been documented as an obligate anaerobe [32]. Together, our results show that the O2-independent UQ biosynthesis pathway is widespread in α-, β-, γ-proteobacterial orders and co-exists with the O2-dependent pathway in 98% of the cases.
We then looked into the relative positioning of ubiT, -U and -V in the 210 genomes harboring the three genes. We found 106 cases, covering α-, β-, γ-proteobacterial orders, where they were encoded next to each other (“3-genes loci”) and 82 cases of a 2-genes locus with the third gene being encoded elsewhere in the genome (Figure 3C). Evaluation of the genetic architecture of ubiT, -U and –V revealed that ubiU and ubiV were found exactly next to each other in 69% of the loci (all 3-genes loci and in 39/82 of the 2-genes loci) and that the three genes were encoded in three separate parts of the genome only in 21 cases (Figure 3C). Interestingly, as an additional support for their involvement in a same function, we found an example of a gene fusion between ubiT and ubiU in two genomes from Zymomonas mobilis strains (α-proteobacteria), which also contain an ubiV gene directly downstream of the fused gene (Figure 3C).
UbiT is an SCP2 protein and UbiU-V are required for the O2-independent hydroxylation of DMQ8
We then analyzed the sequences of the UbiT, UbiU and UbiV proteins. The major part of UbiT (amino acids 45-133, on a total of 174 aa in the E. coli protein) corresponds to a SCP2 domain (PFAM: PF02036, Figure S2). SCP2 domains typically form a hydrophobic cavity that binds various lipids [33] and our sequence alignment indeed showed the conservation of hydrophobic amino acids at several positions in the SCP2 domain of UbiT (Figure S2). Recently, we reported that UbiJ binds UQ biosynthetic intermediates in its SCP2 domain and organizes a multiprotein complex composed of several Ubi enzymes [16]. We propose that UbiT and its SCP2 domain may fulfill similar functions in the O2-independent UQ pathway, as UbiJ is required exclusively for the O2-dependent biosynthesis of UQ (Figure 1F).
UbiU and UbiV have ∼330 and 300 aa respectively, and contain an uncharacterized motif called peptidase U32 (PF01136) (Figure S3 and S4). Since only the hydroxylation reactions are uncharacterized in the O2-independent pathway (Figure 1H), we hypothesized that UbiU and UbiV might function in these steps. To test our hypothesis, we developed an in vivo assay based on the O2-independent conversion of labeled DMQ8 into labeled UQ8. This assay monitors the C6-hydroxylation and the subsequent O6-methylation (Figure 4A). ΔubiC ΔubiF cells grown aerobically with 13C7-4HB synthesized DMQ8, 73% of which was labelled with 13C6. Upon transfer to anaerobic conditions, the cells gradually converted a significant part of (13C6)-DMQ into (13C6)-UQ8 (Figure 4B). Inactivation of either ubiU or ubiV in ΔubiC ΔubiF cells did not perturb the accumulation of (13C6)-DMQ8 but prevented its conversion into (13C6)-UQ8 (Figure 4C-D). This result demonstrates that UbiU and UbiV are essential for the C6-hydroxylation reaction of the O2-independent UQ biosynthetic pathway.
UbiV contains a [4Fe-4S] cluster
To get insights into the potential presence of cofactors in UbiU and UbiV, we attempted to characterize them biochemically. UbiU was not soluble but we purified UbiV6His, which behaved as a monomer in solution (Figure S5A-B). UbiV was slightly pink-colored and had a UV-visible absorption spectrum with features in the 350-550 nm region (Figure 5A, dotted line), suggesting the presence of iron-sulfur (Fe-S) species [34-36]. We indeed detected substoichiometric amounts of iron and sulfur (0.2 Fe and 0.2 S/monomer), indicating a potential oxidative degradation of the Fe-S cluster during aerobic purification, as already observed with many other Fe-S proteins [37,38]. Consistent with this hypothesis, anaerobic reconstitution of the Fe-S cluster yielded a UbiV protein with 3.9 iron and 3.3 sulfur/monomer (Table 2) and with a UV-Vis spectrum characteristic of a [4Fe-4S]2+ cluster [39] (Figure 5A, solid line), that was affected by exposure to air (Figure S5C). The electron paramagnetic resonance (EPR) spectrum of the cluster reduced anaerobically displayed features characteristic of a [4Fe-4S]1+ cluster in the S = 1/2 state (Figure 5B) [38,40]. Overall, we conclude that, under anaerobic conditions, UbiV is able to bind one air-sensitive, redox-active [4Fe-4S] cluster.
Fe-S clusters are typically coordinated by cysteine residues [41,42] and we obtained evidence that the [4Fe-4S] cluster in UbiV is coordinated by four conserved cysteines arranged in a CXnCX12CX3C motif (Figure S4). Indeed, combinatorial elimination of C39, C180, C193, and C197 in double, triple and quadruple mutants resulted in proteins incapable of binding [Fe-S] clusters in vivo, as shown by the absence of absorption bands in the 350-550nm region of their UV-vis spectra (Figure S5D). Furthermore, after anaerobic reconstitution of the cluster, the Fe and S contents, and the absorbance at 410 nm were largely decreased in the double and triple mutants, and were undetectable in the quadruple mutant (Table 2 and Figure 5C). At last, we found that the mutation of C180 or C193 altered the function of UbiV in vivo (Figure 5D), suggesting that the [4Fe-4S] cluster is important for activity.
UbiU contains a [4Fe-4S] cluster and forms a complex with UbiV
We succeeded purifying UbiU after coexpressing it with UbiV6His (Figure S6A). The two proteins copurified in the form of a heterodimer (Figure S6A-B) that showed traces of Fe-S clusters (Figure 6A, dotted line), with substoichiometric amounts of iron and sulfide (0.4 Fe and 0.4 S/heterodimer). Reconstitution with iron and sulfide yielded a heterodimer with about 8 iron and 8 sulfur (Table 2) and a UV-visible spectrum characteristic of [4Fe-4S]2+ clusters (Figure 6A, solid line). The EPR spectrum of reduced UbiU-UbiV was also consistent with the presence of 2 different [4Fe-4S] clusters since it showed a composite signal, which reflects the presence of two different S=1/2 species (Figure 6B).
Four strictly conserved cysteines are also found in UbiU (Figure S3) and we hypothesized that they might bind the [4Fe-4S]. We eliminated these cysteines in pairs and purified heterodimers composed of WT UbiV6His and mutant UbiU (Figure S6C). After reconstitution, UbiU C169A C176A-UbiV and UbiU C193A C232A-UbiV had about half the iron as the WT heterodimer (Table 2), and their A280/A410 ratio were also diminished about two-fold (Figure 6C and Table 2). Altogether, our data clearly demonstrate that each protein of the heterodimeric UbiU-UbiV complex binds one [4Fe-4S] cluster and that the iron-chelating cysteines in UbiU are C169, C176, C193 and C232. Finally, an in vivo complementation assay demonstrated that C176 was important for the function of UbiU (Figure 6D).
Many U32 proteases display motifs of four conserved cysteines
The presence of Fe-S clusters in UbiU and UbiV, two U32 proteases family members, led us to evaluate the presence of conserved Cys motifs in other U32 proteins. Kimura et al. reported a phylogenetic tree of 3521 peptidase U32 domains which formed 12 groups, belonging to 10 protein families [43]. We extracted and aligned the sequences of the 10 protein families and found highly conserved 4-cysteines clusters (97-100% conservation) in eight of them (Figure 7), suggesting an important functional role for these residues. Only families PepU32#5 and PepU32#6 had respectively none, and two mildly (60-80%) plus three poorly conserved cysteines in their sequences (40-65%). The cysteine motifs for each of the eight families showed a high degree of conservation, and strikingly, most of them could even be aligned with each other, the CX6CX15CX3/4C patterns appearing recurrently (Figure 7). To be noted, UbiV had a slightly distinct motif, with the first cysteine occurring much upstream of the three others and outside of the U32 domain. Overall, our data suggest that most members of the U32 peptidase family may contain a 4Fe-4S cluster coordinated by conserved cysteines, similar to what we demonstrated for UbiU and UbiV.
DISCUSSION
The O2-independent UQ biosynthesis pathway is widespread in proteobacteria
The evidence for an O2-independent synthesis of UQ by E. coli was reported more than forty years ago [20], yet this pathway remained uncharacterized until now. Circumstantial evidence had been obtained that a few species – limited, to our knowledge, to E. coli [20], Rhodobacter sphaeroides [44], Paracoccus denitrificans [45], Halorhodospira halophila [46] – were able to synthesize UQ in anaerobic conditions, as demonstrated by biochemical measurements of the quinone content of cells grown anaerobically. Here, we demonstrate that the O2-dependent and O2-independent UQ biosynthesis pathways differ only by three hydroxylation steps (Figure 1), and we identify three genes, ubiT,-U,-V that are essential for the O2-independent biosynthesis of UQ in E. coli (Figure 2). The facts that the UbiT, -U, -V proteins are widespread in α, β, γ–proteobacterial clades (Figure 3) and co-occur with UbiA, -E, -G enzymes (Table S2B), reveal UbiT, -U, -V as key elements of the broadly-distributed, O2-independent UQ pathway. Overall, our data support that many proteobacteria have the previously-unrecognized capacity to synthesize UQ independently from O2.
Physiological possibilities offered by UQ biosynthesis over the entire O2 range
In our set of reference genomes, only three species (P. fulvum, M. marinus and O. formigenes) seem to rely exclusively on the O2-independent pathway for UQ production (Table S2B). Indeed, the vast majority of proteobacteria with the O2-independent UQ biosynthesis pathway also possess the O2-dependent hydroxylases of the aerobic pathway (207 out of 210) (Figure 3B). This result supports that both pathways confer physiological advantages, allowing production of UQ over the entire spectrum of O2 levels encountered by facultative aerobes.
So-called microaerobes, able to respire O2 in microaerobic conditions, are very abundant in Nature [5] and E. coli is known to respire nanomolar O2 concentrations [47]. To sustain O2 respiration in the microaerobic range, these organisms are equipped with high affinity O2 reductases [5,47,48]. These enzymes reduce efficiently the low levels of environmental O2 present at the cell’s membrane, leaving the cytoplasm devoid of any O2 [49]. Under such conditions, UQ - which is the main electron donor for the high affinity O2 reductases bdI and bdII of E. coli [9] - must therefore be synthesized via the O2-independent pathway. Overall, we believe that the O2-independent UQ biosynthesis pathway operates not only in anaerobic conditions but also in microaerobic conditions, in which UQ is likely crucial for bacterial physiology.
The O2-independent UQ biosynthesis pathway may also confer a significant advantage to facultative bacteria in case of a rapid transition from an anaerobic to an aerobic environment. Indeed, anaerobic biosynthesis of UQ will result in cellular membranes containing UQ at the time of the transition, allowing an immediate switch to the energetically favorable metabolism of O2 respiration. Our identification of the anaerobic UQ pathway provides the unique opportunity to selectively disrupt UQ biosynthesis depending on O2 levels and should foster new research on bacterial physiology in the microaerobic range. Indeed, apart from E. coli, which was thoroughly studied over the microaerobic range [4,49,50], details on bacterial physiology in microaerobiosis are scarce.
ubiT,U,V mutants and pathogenicity
In addition to bioenergetics per se, anaerobic and microaerobic respirations are thought to be important for pathogenicity [3,51]. Interestingly, homologs of UbiT, -U, -V have been linked to pathogenicity in several bacterial models. Indeed, the inactivation of ubiU-V homologs in Proteus mirabilis leads to a decreased infection of the urinary tract of mice [52] and to a diminished virulence of Yersinia ruckeri [53], a pathogen that develops in the gut of fish, an environment with a notoriously low O2 content. Furthermore, inactivation of PA3911 [54] (ubiT) and PA3912-PA3913 [55] (ubiU-V) in Pseudomonas aeruginosa, abolished nitrate respiration, the main anaerobic metabolism used by the bacterium in the lungs of cystic fibrosis patients [56,57]. Based on our results showing that the deletion of ubiT, ubiU or ubiV abrogates the O2-independent biosynthesis of UQ in E. coli, we suggest that the attenuation of the mutants discussed above results from their UQ deficiency in microaerobic / anaerobic conditions.
Proposed roles for UbiT, UbiU and UbiV
UbiT possesses an SCP2 domain, similar to UbiJ that we recently demonstrated to be an accessory factor that binds the hydrophobic UQ biosynthetic intermediates and structures a multiprotein Ubi complex [16]. Since UbiJ functions exclusively in the O2-dependent pathway, whereas UbiT is important only for the O2-independent pathway, we propose that UbiT may fulfill, in anaerobiosis, the same functions than UbiJ in aerobiosis. Whether UbiT is part of a complex and is able to bind UQ biosynthetic intermediates, will be addressed in future studies. Interestingly, PA3911, the homolog of UbiT in P. aeruginosa, was recently shown to bind phosphatidic acid [54], demonstrating an affinity of UbiT for lipid molecules.
UbiU and UbiV form a tight heterodimer suggesting that the proteins function together, as further supported by the fact that the ubiU and ubiV genes co-occur in genomes in 99% of the cases and that they are mostly found next to each other (Figure 3). We demonstrated that UbiU and UbiV are both required for the O2-independent C6-hydroxylation of DMQ and the accumulation of OPP in ΔubiU or ΔubiV mutants suggests that the two proteins may also function in C5-hydroxylation. We want to emphasize our recent demonstration that a single hydroxylase catalyzes all three hydroxylation steps in the O2-dependent UQ pathway of Neisseria meningitidis [19]. This result showed that three different enzymes are not necessarily required and opens the possibility that UbiU-V may in fact catalyze all three hydroxylation reactions of the O2-independent UQ biosynthesis pathway. Establishing the hydroxylase activity and the regioselectivity of UbiU-V will require the development of an in vitro assay, a challenging task given that the oxygen donor of the reaction is currently unknown and that the substrates are not commercial and highly hydrophobic. Of note, one of the O2-dependent hydroxylases was shown to hydroxylate DMQ0, a substrate analog with no polyprenyl side chain [58], suggesting that an in vitro assay for UbiU-V might be developed with soluble analogs.
Fe-S clusters in UbiU-V and other members of the U32 peptidase family
Up to now, members of the peptidase family U32 were not shown to bind Fe-S clusters or to contain any of the ~30 cysteine motifs found in well-characterized iron-sulfur proteins [59]. The expression, purification and spectroscopic characterization of UbiV and of the UbiU-UbiV heterodimeric complex clearly showed that each protein contains one [4Fe-4S] cluster (Figure 5 and Figure 6). Mutation of the candidate cysteine ligands, arranged in a CX6CX16CX38C motif in UbiU and in a CXnCX12CX3C motif in UbiV, disrupted Fe-S binding and abolished in vivo complementation, suggesting a crucial function of the [4Fe-4S] clusters in these proteins. The conservation of a CX6CX15CX3/4C motif in other U32 proteases supports that these proteins likely bind Fe-S clusters. This hypothesis should guide and stimulate investigations of U32 members, very few of which have currently an established molecular function ([60], http://www.ebi.ac.uk/merops/). Interestingly, RlhA, a member of the U32 proteases family involved in the C-hydroxylation of a cytidine on E. coli 23S rRNA, was recently shown to be connected to iron metabolism [43], corroborating our suggestion that RlhA is also an Fe-S protein.
In biological systems, Fe-S clusters are mainly known to be involved in electron transfer reactions, but also in substrate binding and activation, in transcription regulation, in iron storage, and as a sulfur donor [41,61-63]. The role of the Fe-S clusters in UbiU and UbiV is unknown at this stage. Our current working hypothesis is a role as an electron transfer chain between the substrate (the UQ biosynthetic intermediate to be hydroxylated) and an unidentified electron acceptor required for the activation of the substrate. Clearly, the Fe-S clusters of UbiU-V are distinct from the molybdenum cofactor present in molybdenum-containing hydroxylases, the only family currently known to catalyze O2-independent hydroxylation reactions [64]. Together, our results identify UbiU and UbiV as prototypes of a novel class of O2-independent hydroxylases and extend the framework of the chemically-fascinating O2-independent hydroxylation reactions.
MATERIALS and METHODS
Strain construction
Strains used in this study are listed in table S3. We obtained the collection of E. coli strains containing large and medium deletions from the National BioResource Project, National Institute of Genetics, Japan (http://www.shigen.nig.ac.jp/ecoli/pec/).
The ΔubiA∷cat, ΔubiD∷cat, ΔubiT∷cat and ΔubiV∷cat mutations were constructed in a one-step inactivation of ubi genes as described [65]. A DNA fragment containing the cat gene flanked with a 5⍰ and 3⍰ region bordering the E.coli ubi genes was amplified by PCR using pKD3 as a template and oligonucleotides 5wanner and 3wanner (Table S3). Strain BW25113, carrying the pKD46 plasmid, was transformed by electroporation with the amplified fragment and catR colonies were selected. The replacement of chromosomal ubi by cat gene was verified by PCR amplification in the catR clones. Mutations (including ubiU∷kan from the keio strain) were introduced into MG1655 strains by P1 vir transduction [66], selecting for the appropriate antibiotic resistance. The antibiotic resistance cassettes were eliminated when needed using plasmid pCP20 as described [67].
Plasmid construction
All plasmids generated in this study were verified by DNA sequencing. The yhbU, yhbT and yhbV inserts (UniProtKB: P45527, P64599, P45475) were obtained by PCR amplification using E. coli MG1655 as template and the oligonucleotides pairs yhbU5/yhbU3, yhbT5/yhbT3 and yhbV5/yhbV3, respectively (Table S3). Yhb inserts were EcoRI-SalI digested and inserted into EcoRI-SalI-digested pBAD24 plasmids, yielding the pBAD-yhbU, pK-yhbT or pBAD-yhbV plasmids, respectively.
To create a plasmid expressing the ubiV (yhbV) ORF as C-terminally His-tagged protein, the ubiV gene was amplified using pET-22-UbiV-FW (introducing the NdeI site) and pET-22-UbiV-RV (introducing the XhoI site) as primers and pBAD-yhbV as template. The NdeI and XhoI digested amplicon was ligated to NdeI and XhoI digested pET-22b(+) plasmid to obtain pET-22-UbiV.
The plasmid pETDUET-UbiUV containing UbiU in MCS1 and UbiV in MCS2 was obtained as follows. ubiU was amplified from pBAD-yhbU using pETDUET-UbiU-FW (introducing the NcoI site) and pETDUET-UbiU-RV (introducing the EcoRI site) as primers. The NcoI and EcoRI digested amplicon was ligated to NcoI and EcoRI digested pETDUET-1-plasmid to obtain pETDUET-UbiU. The ubiV gene was then cloned from pET-22-UbiV into the MSC2 of pETDUET-UbiU by PCR amplification with pET-22-UbiV-FW and pETDUET-UbiV-RV (introducing the C-terminal His6-tag) as primers. The NdeI and XhoI digested amplicon was ligated to NdeI and XhoI digested pETDUET-UbiU to obtain pETDUET-UbiUV.
A hexahistidine-tag was fused at the N-terminal extremity of UbiV to create pBAD-UbiV6His. The ubiV6His gene was obtained by PCR amplification (Phusion High-Fidelity DNA Polymerase) using pBAD-UbiV as a template and 6HisV5 (introducing the NcoI site) and 6HisV3 (introducing the HindIII site and the DNA sequence of 6His-tag) as primers. The NcoI/HindIII digested amplicon was cloned into the NcoI/HindII-digested pBAD plasmid.
Variants of UbiV and UbiU were constructed using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs) according to the manufacturer’s specifications. The plasmids (pET-22b-UbiV, pETDUET-UbiUV, pBAD-yhbU or pBAD-yhbV) were used as templates in conjunction with the appropriate primers for each respective amino acid substitution.
Culture conditions
E. coli strains were grown at 37 °C in lysogeny broth (LB) medium or in synthetic medium (SM) containing either 0.4% glycerol (w/v), 0.4% lactate (w/v) or 0.2% glucose (w/v) as carbon sources. Autoclaved SM medium was supplemented with 0.5% casaminoacids (w/v) and with 1/100 volume of a filter-sterile solution of 1mM CaCl2, 200 mM MgCl2, 1% thiamine (w/v) [68]. Ampicillin (50 mg/L), kanamycin (25 mg/L), and chloramphenicol (25 mg/L) were added from stocks (1000X solution sterilized through 0.22 µm filters and stored at −20°C), when needed. When needed, 0.02% arabinose was added to induce the expression of genes carried on pBAD and pK plasmids. External electron acceptors like KNO3 (100 mM) or dimethylsulfoxide (DMSO, 50 mM) were added to SM medium for anaerobic cultures. Anaerobic cultures were performed in Hungate tubes containing 12 mL medium deoxygenated by argon (O2<0.1 ppm) bubbling for 25 min before autoclave (in case of LB medium, 0.05% antifoam (Sigma) was added). Hungate tubes were inoculated through the septum with 100 µL of overnight precultures taken with disposable syringe and needles from closed Eppendorf tubes filled to the top. Aerobic cultures were performed in Erlenmeyer flasks filled to 1/10 of the maximal volume and shaken at 180 rpm.
For the initial screen, we grew ME strains anaerobically in SMGN (SM medium supplemented with glycerol and nitrate). Strains that presented a severe growth defect or a low UQ8 content were subsequently grown anaerobically in LB medium.
Cultures were cooled down on ice before transferring 5-10 mL volumes into 15 mL Falcon tubes for centrifugation at 3200 g, 4°C, 10 min. Cell pellets were washed in 1 mL ice-cold PBS and transferred to pre-weighted 1.5 mL Eppendorf tubes. After centrifugation at 12000 g, 4°C, 1 min and elimination of supernatant, the cells wet weight was determined (∼10-20 mg) and pellets were stored at −20°C prior to quinone extraction. We note that these steps were conducted under normal atmosphere and allowed limited O2-dependent UQ biosynthesis in cells grown anaerobically. Thus, modifications (detailed below) were adopted in additional experiments conducted under strict anaerobic conditions.
Culture under strict anaerobic conditions and cells quenching
LB medium was supplemented with 100 mg/L L-cysteine (adjusted to pH 6 with NaOH) and 2.5 mg/L resazurin. The medium was distributed in Hungate tubes and was deoxygenated by argon (O2<0.1 ppm) bubbling for 45 min at 60°C. The resazurin was initially purple, then quickly turned to pink, and eventually became colorless. The Hungate tubes were sealed and autoclaved. Two sequential precultures were performed in order to dilute the UQ present in the initial aerobic inoculum. The first preculture was performed overnight and used Eppendorf tubes filled to the top and inoculated with cells grown aerobically on LB agar. The second preculture was performed for 8 hours in Hungate tubes and was used to inocculate Hungate tubes that were subsequently incubated overnight at 37°C. Disposable syringes (1 mL) and needles were flushed 5 times with argon prior to inoculating 50 µL of preculture through the septum of the Hungate tubes. The resazurin remained colorless at all steps of the culture, indicating that the medium in the Hungate tubes was strictly anaerobic. At the end of the culture, the Hungate tubes were cooled down on ice for 45 min and 2 mL medium was sampled through the septum with argon-flushed syringes (2 mL) fitted with needles. The cells were immediately quenched by transfer to −20°C precooled glass tubes containing 6 mL methanol, 0.5 mL glass beads (0.5 mm diameter) and 20 mM KCl. The tubes were homogenized by vortex for 30 seconds and kept at −20°C prior to quinone extraction. In parallel, we also centrifuged 2 mL of culture from the Hungate tubes in order to determine the weight of the cells and normalize the UQ content of the quenched cells that was subsequently measured.
For the experiments conducted under strict anaerobic conditions (Figure 2H, 2I), we used LB medium instead of MSGN, since nitrite - produced during the anaerobic respiration of nitrate in MSGN medium - is able to oxidize resazurin [69].
Lipid extraction and quinone analysis
Quinone extraction from cell pellets was performed as previously described [17].
Quinone extraction from cells quenched in methanol was slightly adapted from [9]. Briefly, 4 µL of a 10 µM UQ10 solution was added as internal standard to the cells-methanol mixture. Then 4 mL petroleum ether (boiling range 40–60 °C) was added, the tubes were vortexed for 30 sec and the phases were separated by centrifugation 1 min, 600 rpm. The upper petroleum ether layer was transferred to a fresh glass tube. Petroleum ether (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 lipid extracts were resuspended in 100 µL ethanol and a volume corresponding to 1 mg of cells wet weight was analyzed by HPLC-electrochemical detection-mass spectrometry (ECD-MS) with a BetaBasic-18 column at a flow rate of 1 mL/min with mobile phases composed of methanol, ethanol, acetonitrile and a mix of 90% isopropanol, 10% ammonium acetate (1 M), 0.1% TFA: mobile phase 1 (50% methanol, 40% ethanol and 10% mix), mobile phase 2 (40% acetonitrile, 40% ethanol, 20% mix). Mobile phase 1 was used in MS detection on a MSQ spectrometer (Thermo Scientific) with electrospray ionization in positive mode (probe temperature 400°C, cone voltage 80V). Single ion monitoring (SIM) detected the following compounds: OPP (M+NH4+), m/z 656.0-656.8, 5-10 min, scan time 0.2 s; DMQ8 (M+Na4+), m/z 719-720, 6-10 min, scan time 0.2 s; 13C6 –DMQ8 (M+Na+), m/z 725-726, 6-10 min, scan time 0.2 s; UQ8 (M+ 4+), m/z 744-745, 6-10 min, scan time 0.2 s; UQ8 (M+Na+), m/z 749-750, 6-10 min, scan time 0.2 s; 13C6-UQ (M+Na+), m/z 755.0-756, 6-10 min, scan time 0.2 s; UQ10 (M+NH4+), m/z 880.2-881.2, 10-17 min, scan time 0.2 s. MS spectra were recorded between m/z 600 and 900 with a scan time of 0.3 s. UV detection at 247 nm was used to quantify DMK8 and MK8. ECD, MS and UV peak areas were corrected for sample loss during extraction on the basis of the recovery of the UQ10 internal standard and were then normalized to cell’s wet weight. The peak of UQ8 obtained with electrochemical detection was quantified with a standard curve of UQ10 [17]. The absolute quantification of UQ8 based on the m/z= 744.6 signal at 8 min (Figure 2H, I) was performed with a standard curve of UQ8 ranging from 0.5 to 150 pmoles UQ8 (the detection limit was around 0.1 pmole).
Anaerobic 13C6-UQ8 biosynthesis activity assay
ΔubiC ΔubiF cells containing or not the additional ΔubiU or ΔubiV deletions were grown overnight in MS medium supplemented with 0.2% glucose. This preculture was used to inoculate at OD600=0.1, 100 mL of fresh medium supplemented with 10 µM 13C7-4HB. The culture was grown at 37°C, 180 rpm until OD600=1, at which point 100 µM 4HB was added. The cells were pelleted by centrifugation at 3200 g, 4°C, 10 min and suspended in 100 mL MSGN medium. A 10 mL aliquot was taken for quinone extraction (aerobiosis, Figure 4B-D) and the rest of the culture was placed at 37°C in an anaerobic bottle with a two- port cap fitted with plastic tubing used to inject argon (O2<0.1 ppm) throughout the experiment in order to create and maintain anaerobiosis. After 5 min bubbling, a 10 mL sample was taken corresponding to 0 min anaerobiosis, then samples were taken every 30 minutes and analyzed for quinone content.
Overexpression and purification of proteins
Overexpression and purification E. coli wild-type UbiV and variants
The pET-22b(+) plasmid, encoding wild-type UbiV or variants, were co-transformed with pGro7 plasmid (Takara Bio Inc.) into E. coli BL21 (DE3) competent cells. Single colonies obtained from transformation were grown overnight at 37°C in LB medium supplemented with ampicillin (50 µg/mL) and chloramphenicol (12.5 µg/mL). 10 mL of preculture was used to inoculate 1 L of LB medium with the same antibiotics, and the bacteria were cultured further at 37 °C with shaking (200 rpm). At an OD600 of 1.2, D-arabinose was added to the cultures at a final concentration of 2 mg/mL. At an OD600 of 1.8, the culture was cooled in an ice-water bath, and isopropyl 1-thio-β-D-galactopyranoside (IPTG) was added at a final concentration of 0.1 mM. Cells were then allowed to grow further at 16 °C overnight. All subsequent operations were carried out at 4°C. Cells were harvested in an Avanti® J-26XP High-Performance centrifuge from Beckman Coulter with a JLA-8.1000 rotor at 5,000 × g for 10 minutes. The cell pellets were resuspended in 5 volumes of buffer A (50 mM Tris-HCl, pH 8.5, 150 mM NaCl, 15% (v/v) glycerol, 1 mM DTT) containing Complete™ Protease Inhibitor Cocktail (one tablet per 50 mL) (Roche) and disrupted by sonication (Branson Digital Sonifier, amplitude 40% for 10 min). Cells debris were removed by ultracentrifugation in an Optima™ XPN-80 ultracentrifuge from Beckman Coulter with a 50.2 Ti rotor at 35,000 x g for 60 min. The resulting supernatant was loaded onto a HisTrap FF Crude column (GE Healthcare) pre-equilibrated with buffer A. The column was washed with 10 column volumes of buffer B (50 mM Tris-HCl pH 8.5, 150 mM NaCl, 15% (v/v) glycerol, 1 mM DTT, 10 mM imidazole) to remove non-specifically bound E. coli proteins then eluted with a linear gradient of 10 column volumes of buffer B containing 500 mM imidazole. Fractions containing WT UbiV or variants were pooled and phenylmethylsulfonyl fluoride was added at a final concentration of 1mM. The proteins were then loaded on a HiLoad 16/600 Superdex 75 pg (GE Healthcare) pre-equilibrated in buffer C (50 mM Tris-HCl pH 8.5, 25 mM NaCl, 15% (v/v) glycerol, 1 mM DTT). The purified proteins were concentrated to 30-40 mg/mL using Amicon concentrators (30-kDa cutoff; Millipore), aliquoted, frozen in liquid nitrogen and stored at −80 °C. Overall, a high yield of 150 mg UbiV/L culture was obtained.
Overexpression and purification of UbiU/V complex and variants
The overexpression of wild-type UbiU/V complex or variants in E. coli BL21 (DE3) competent cells were performed following the same protocol as for UbiV. The expression of these proteins was induced by addition of IPTG to a final concentration of 0.05 mM. Wild-type UbiU/V complex or variants were purified with the same procedure as for UbiV, with the exception that the proteins were loaded on the HiLoad 16/600 Superdex 75 pg with buffer A.
[Fe-S] cluster reconstitution
The [Fe-S] cluster(s) reconstitution of holo-UbiV and holo-UbiU/V were conducted under anaerobic conditions in an Mbraun LabStar glove box containing less than 0.5 ppm O2. Classically, a solution containing 100 µM of as-purified UbiV or UbiU/V complex, was treated with 5 mM DTT for 15 min at 20°C and then incubated for 1 hour with a 5-fold molar excess of both ferrous ammonium sulfate and L-cysteine. The reaction was initiated by the addition of a catalytic amount of the E.coli cysteine desulfurase CsdA (1-2% molar equivalent) and monitored by UV-visible absorption spectroscopy. The holo-UbiV or holo-UbiU/V complexes were then loaded onto a Superdex 75 Increase 10/300 GL column (GE Healthcare) pre-equilibrated with buffer C or A, respectively, to remove all excess of iron and L-cysteine. The fractions containing the holo-proteins were pooled and concentrated to 20-30mg/mL on a Vivaspin concentrator (30-kDa cutoff).
Quantification Methods
Protein concentrations were determined using the method of Bradford (Bio-Rad) with bovine serum albumin as the standard. The iron and acid-labile sulfide were determined according to the method of Fish [70] and Beinert [71], respectively.
UV-Vis spectroscopy
UV-Visible spectra were recorded in 1 cm optic path quartz cuvette under aerobic conditions on a Cary 100 UV-Vis spectrophotometer (Agilent) and under anaerobic conditions in a glove box on a XL-100 Uvikon spectrophotometer equipped with optical fibers.
EPR spectroscopy
EPR spectra of frozen solutions were recorded on a Bruker Continuous-Waves (CW) X-Band ELEXSYS E500 spectrometer operating at 9.39 GHz, equipped with an SHQE cavity cooled by an helium flow cryostat ESR 900 Oxford Instruments under non-saturating conditions and using the following parameters: a microwave power in the range 2 to 10 mW and a modulation of the magnetic field at 100 kHz with a modulation amplitude of 0.6 mT. Holo UbiV or holo UbiU-UbiV complex were treated with 10-fold molar excess of dithionite to reduce the Fe-S cluster. Each solution was introduced into EPR quartz tubes in a glove box and frozen with liquid nitrogen before the EPR measurements.
Genome datasets
The protein sequences from 5750 complete genomes (“extended dataset”) were downloaded from the NCBI Refseq database (bacteria and archaea, last accessed in November 2016, Table S2A). A representative set of complete genomes from a monophyletic group of bacteria that potentially harbor the ubiquinone biosynthesis pathway was also created: “Reference” and “Representative” genomes were downloaded from the NCBI Refseq database for 204 Alphaproteobacteria, 103 Betaproteobacteria, 303 Gammaproteobacteria (last accessed in November 2018). In addition to these 610 genomes, the genome of Phaeospirillum fulvum (99.5% estimated completeness according to CheckM, http://gtdb.ecogenomic.org/genomes?gid=GCF_900108475.1) was included (Table S2B).
HMM protein profiles creation
An initial set of protein sequences (“curated set”) was retrieved from genomes manually and from a publication [72], to cover the diversity of UQ-producing organisms. The curated set included 48 pairs of YhbU and YhbV from 10 alpha, 19 beta and 19 gamma-proteobacteria, 17 sequences for YhbT, 64, 181, 69 and 189 sequences for UbiA, MenA, UbiG and UbiE respectively (Table S4). Then, for each gene family these sequences were aligned with Mafft (v7.313, “linsi”) [73], and each alignment trimmed at its N-ter and C-ter extremities based on the filtering results of BMGE (BLOSUM 30) [74]. The core of the alignments were kept as is, and Hidden Markov Model (HMM) profiles were created directly from the trimmed alignments using the Hmmbuild program (Hmmer suite, version 3.1b2) [75].
To ensure a more sensitive search, and good delineation between homologs, phylogenetic curation was used, as YhbU and YhbV are known to be part of the larger U32 protease gene family [43]. A search using the YhbU and YhbV HMM profiles was performed with Hmmsearch (Hmmer suite) on the extended 5750 genomes dataset, and sequences with an i-evalue (“independent e-value”) lower than 10E-20 and a coverage of the profiles higher than 90% were selected. The 4212 sequences obtained were de-replicated using Uclust from the Usearch program suite (80% identity level) [76]. A phylogenetic tree was built by maximum-likelihood with the IQ-Tree program (best evolutionary model) based on the alignment (Mafft: “linsi”, BMGE with BLOSUM30) of the 480 selected sequences including all curated YhbU and YhbV sequences [73,74,77]. YhbU and YhbV proteobacterial sequences formed two separate monophyletic groups, giving credit to our curated set (100% and 80% UF-Boot support respectively). The other sequences that formed a large monophyletic group of bacterial sequences were categorized as “U32 proteases” (98% UF-Boot, https://doi.org/10.6084/m9.figshare.7800614.v1). The 98 sequences from this U32 proteases group (Table S4) were used to re-create an HMM profile as described above, and served as an outgroup for the profile search.
For YhbT, the first profile obtained from the curated set of sequences was used altogether with the YhbU, YhbV and U32 proteases profiles for a search in the 611 proteobacteria genomes dataset. A second profile was created from YhbT sequences (YhbT2, Table S4) that were co-localizing with YhbU and YhbV hits (10E-20 i-evalue and 80% profile coverage). The two YhbT profiles matched complementary sets of sequences and therefore were both used for annotating YhbT in genomes.
A similar approach was taken in order to identify the six known aerobic hydroxylases. 51 Coq7, 73 UbiF, 80 UbiH, 58 UbiI, 24 UbiL and 32 UbiM sequences were extracted manually and from publications [72] [19](Table S4, “version 1”) to serve as a reference, annotated set of sequences. Profiles were created as described above. To ensure their specificity, we ran the HMM profiles against our 5570 genomes dataset and selected the sequences that had an i-evalue lower than 10E-20 and a coverage of the profiles higher than 90%. We built two phylogenetic trees as described above: one for Coq7, and another one for UbiFHILM, which are known to be part of the large FMO protein family [19]. In the latter case, we de-replicated the 1619 sequences obtained for the FMO protein family before performing the alignment, alignment filtering, and tree reconstruction steps (using Uclust at the 60% identity level). The Coq7 tree obtained showed our reference Coq7 sequences covered the whole diversity of retrieved sequences, suggesting that they all could be bona fide Coq7 (https://doi.org/10.6084/m9.figshare.7800680). The FMO tree showed a monophyletic group containing all reference FMO ubiquinone hydroxylases, forming sub-groups for the different homologs (UbiFHILM) in Proteobacteria (https://doi.org/10.6084/m9.figshare.7800620). Further, a large set of sequences formed an outgroup consisting of sequences from various clades of bacteria, a lot being found outside of Proteobacteria, robustly separated from the ubiquinone hydroxylases. We split this large clade into four sub-trees, and extracted the corresponding sequences to obtain four new HMM profiles (as described above), to be used for precise discrimination between ubiquinone hydroxylases and other members of the FMO family (“AlloFMO_1” to “AlloFMO_4” in Table S4, “version 2”). FMO ubiquinone hydroxylases sub-trees were also used to re-design improved HMM profiles for UbiFHILM (36, 168, 198, 139, and 65 sequences respectively, see Table S4 “version 2”).
Evaluation of genomic distributions with HMMER and MacSyFinder
Two MacSyFinder models were created to i) search sequences of interest in genomes using Hmmer, and ii) investigate their genetic architecture [78]. A first model was created to focus only on YhbTUV-related genes. In this model, the YhbTUV components were defined as “mandatory”, and U32 protease as “accessory”. A second, more comprehensive model “Ubi+YhbTUV” was designed to list the families corresponding to the 9 profiles obtained (UbiA, MenA, UbiE, UbiG, 2 YhbT, YhbU, YhbV and U32 proteases). For both models, the two YhbT profiles were set as “exchangeable”. The parameter “inter_gene_max_space” was set to 5 in the “YhbTUV” model, and 10 in the “Ubi+YhbTUV” model. MacSyFinder was set to run HMMER with the options “--i-evalue-select 10E-20” and “--coverage-profile 0.8”. Independently of their genetic context, sequences corresponding to selected HMMER hits were listed for all profiles in all genomes analyzed in order to establish the genomic distribution for each of the protein families of interest. When several profiles matched a sequence, only the best hit (best i-evalue) was considered.
U32 proteases sequence analysis
We retrieved from the UniProt-KB database 3460 protein sequences of U32 proteases that were categorized in 12 families by Kimura et al. [43], and created a FASTA for each of these families. 50 sequences from different families could not be retrieved as they had been deleted from the Uniprot-KB database (46 “obsolete”) or were not found based on the published accession numbers. As the “RlhA1” and “RlhA2” families mostly corresponded to two domains from the same protein sequences that had been split, we put whole sequences all together into a single fasta file for sequence analysis of the overall “RlhA” family. For each of the 10 family, sequences were de-replicated at the 80% identity level with Uclust, in order to limit any potential taxonomic sampling bias, and sequences were aligned (Mafft, linsi). The alignments were visualized in Jalview [79] and used to create the logo sequences. Images of alignments were created using the ESPript webserver (http://espript.ibcp.fr/ESPript/ESPript/) [80].
COMPETING INTERESTS
The authors declare that they have no competing interests.
ACKNOWLEDGEMENTS
This work was supported by the Agence Nationale de la Recherche (ANR), ANR Blanc (An)aeroUbi ANR-15-CE11-0001-02 to FP. We thank Amélie Amblard for technical assistance, Louis Givelet for preliminary bioinformatic analyses, Barbara Schoepp-Cothenet for providing accession numbers to sequences of UbiA, -G, -E and for critical reading of the manuscript, and the GEM team at TIMC for discussions and suggestions. CDTV, ML and MF acknowledge support from the French National Research Agency (Labex program DYNAMO, ANR-11-LABX-0011). CC was funded by the Grenoble Alpes Data Institute, supported by the French National Research Agency under the “Investissements d’avenir” program (ANR-15-IDEX-02). We thank the National Bioresource Project, National Institute of Genetics for providing ME strains from the medium and large deletions E. coli collection.
Footnotes
addition of three new panels to figure 2 (I-H) addition of a new paragraph in the "results" section describing the quinone content of strains grown under strict anaerobic conditions
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵