Evolution has selected ring methylations as critical features of prenylated quinones that serve as lipophilic electron and proton carriers in respiration and photosynthesis. Specifically, the presence of a methyl group on the ring carbon ortho of the prenyl chain, as is the case for menaquinone and ubiquinone (Fig. 1a, b), orientates the quinoid and prenyl moieties of these molecules perpendicular to each other (Himo et al. 1999;Seif Eddine et al. 2020). Conversely, absence of a methyl group at the same position in plastoquinone-9 or demethyl-menaquinone (Fig. 1a, c), results in the ring and prenyl chain of these quinones adopting a planar conformation (Himo et al. 1999;Seif Eddine et al. 2020). These conformational differences impact the positioning of the corresponding quinones within biological membranes and protein active sites (Himo et al. 1999;Seif Eddine et al. 2020;Braasch-Turi et al. 2022). The presence of ring methylations adjacent to the prenyl chain is also a major determinant of the redox properties of quinones. For instance, menaquinone possesses a lower midpoint redox potential than its demethylated counterpart (Rendon et al. 2015), and some facultative anaerobes leverage this chemistry by adjusting the methylation status of their quinone pool to the availability of redox substrates and oxygen levels (Unden 1988;Shestopalov et al. 1997;Søballe and Poole 1999;Bekker et al. 2007). The C-methyltransferases responsible for the methylations ortho of the prenyl chain act on quinols -i.e. the reduced versions of quinones (Fig. 1a, b)-and segregate into two phylogenetic clades (Fatihi et al. 2015;Stutts et al. 2023). Clade 1 orthologs, which encompass prokaryotic UbiE and mitochondrial COQ5 (Fig. 1a, b), are distributed throughout the tree of life with the notable exception of cyanobacteria. These enzymes are active towards both benzoquinol and naphthoquinol substrates (Fig. 1a, b), irrespective of the quinone profile of their native organism (Stutts et al. 2023). For instance, although mitochondrial COQ5 is dedicated to the biosynthesis of ubiquinone (a benzoquinone), recombinant versions of this enzyme can also efficiently methylate demethyl-menaquinol (a naphthoquinol) (Stutts et al. 2023). Similarly, UbiE C-methyltransferases from microorganisms that produce exclusively menaquinone can functionally replace UbiE for ubiquinone production when expressed in Escherichia coli (Stutts et al. 2023). Some facultative anaerobes that naturally produce menaquinone and ubiquinone exploit this catalytic promiscuity to synthesize both of these prenylated quinones using a single clade 1 enzyme (Fig. 1a, b) (Lee et al. 1997). Clade 2 quinol C-methyltransferases, on the other hand, are restricted to cyanobacteria and plastids, and display strict specificity for naphthoquinol substrates (Fig. 1a) (Stutts et al. 2023).
The emergence of such a stringent substrate recognition is thought to have been driven by the need for cyanobacteria Methyl-plastoquinone occurs naturally in non-photosynthetic and aerobic Nitrospirota, but is highly toxic to oxygenic phototrophs and plastids to harbor two vital quinones with diverging ring methylation patterns in ortho of their prenyl chains. The first quinone, a naphthoquinone called phylloquinone -more rarely, menaquinone in some taxa-is methylated at this position, while the second quinone, a benzoquinone called plastoquinone-9, is not (Fig. 1a, c). Preventing this specific ring methylation of plastoquinone-9 is of paramount importance for cyanobacteria and plastids, because methyl-plastoquinone-9 (Fig. 1c) is thought to behave as an antimetabolite of plastoquinone-9 (Stutts et al. 2023). The cytotoxicity of methyl-plastoquinone and its structural analogues is specific to oxygenic phototrophs (Latimer et al. 2024), and recent findings indicate that methyl-plastoquinone operates as the endogenous electron carrier in the respiratory chain of certain non-photosynthetic bacteria (Elling et al. 2025).
We undertook the present study with the aim to reconstruct the evolutionary history of clade 2 quinol C-methyltransferases and gain insight into the molecular determinants of these enzymes' substrate preference. While performing our initial sampling, we made the puzzling discovery that in extant taxa of early cyanobacterial lineages, some of these clade 2 enzymes were readily active on both naphthoquinol and benzoquinol substrates. We then expanded our investigation to refine the evolutionary model for the emergence of clade 2 quinol C-methyltransferases in oxygenic phototrophs.
Menaquinone-8, demethyl-menaquinone-8, and ubiquinone-8 standards were purified from E. coli extracts as previously described (Kim et al. 2008;Latimer et al. 2021). Demethyl-phylloquinone was purified from the leaf extracts of Arabidopsis demethyl-phylloquinol methyltransferase knockout line GABI_565F06 (Lohmann et al. 2006) using the same procedures as those described for phylloquinone analysis in Widhalm et al. (2012). Plastoquinone-9 was purified from Arabidopsis leaf extracts as described in Block et al. (2013). Methyl-plastoquinone-9 was obtained from extracts of Synechocystis sp. PCC 6803 Δsll1653 cells expressing Arabidopsis COQ5 minus its plastid-transit peptide (Stutts et al. 2023). Phylloquinone was from MP Biomedicals. Quinol standards were prepared by reduction of their corresponding quinone forms with NaBH 4 . Unless mentioned otherwise, other reagents were from Fisher Scientific.
Cyanobacterial orthologs of Synechocystis sp. PCC 6803 demethyl-phylloquinol methyltransferase (Sll1653) were mined from the GenBank protein database (National Center for Biotechnology Information; www.ncbi.nlm.nih.gov/ protein/) using BLASTp searches. Phylogenetic reconstructions were performed with NGPhylogeny.fr tool suite (Lemoine et al. 2019) and selecting the following workflow: MUSCLE (multiple alignment), Gblocks (removal of misaligned and divergent sequence regions), PhyML (tree inference via the maximum-likelihood method), Approximate likelihood-ratio test (branch support assessment), and Newick Display (tree rendering). Arabidopsis COQ5 and E. coli UbiE/MenG quinol C-methyltransferases were used as the outgroup. The accession numbers of the proteins used for phylogenetic reconstruction is provided in Table 1.
Structural modeling of Gloeobacter violaceus and Gloeobacter morelensis quinol C-methyltransferases were performed using AlphaFold 2.0 (Jumper et al. 2021). The G. violaceus model was predicted with a pLDDT confidence scores of 95.1. Lower confidence regions were located at the extreme N-terminus (70 > pLDDT > 50) and at the proposed dimer interface (90 > pLDDT > 70). The G. morelensis protein had a model prediction pLDDT score of 95.2 with a similar lower confidence region at the N-terminus. The interior Rossman-like fold maintained the highest confidence (pLDDT > 90) over its entirety for both predictions. Using the AlphaFold-generated models, experimentally determined structures of related proteins were identified using the DALI server (Holm et al. 2023). Of these structures, Saccharomyces cerevisiae Coq5 (PDB 4OBW; Dai et al. 2014) -DALI Z-score of 27.7 for G. violaceus and 27.4 for G. morelensis-was used as a guide to determine the probable orientation of the S-adenosyl methionine (AdoMet) ligand in the active sites of the Gloeobacter enzymes. These models were visualized using Pymol (PyMOL Molecular Graphics System, Version 3.0 Schrödinger, LLC).
E. coli gene UbiE and Synechocystis gene Sll1653 were directly PCR amplified from genomic DNA as previously described (Stutts et al. 2023). Cyanobacterial orthologs of Synechocystis Sll1653 were codon-optimized for expression in E. coli and synthesized by Genescript, USA. Arabidopsis COQ5 (At5g57300) and demethyl-phylloquinol methyltransferase (At1g23360) were amplified, minus their targeting µmole m -2 s -1 of PAR) at 22 °C for 13 days (photomixotrophic conditions) or 22 days (photoautotrophic conditions).
E. coli cultures (20 ml-50 ml) were grown in LB medium (25-37 °C) with 0.2% (w/v) arabinose as an inducer. Cells were harvested when cultures reached OD 600 ~ 0.7-1. Cell pellets were washed once with sterile water (4 °C) and resuspended cells (1 ml) were quantified by absorbance at 600 nm. Preparation of cell extracts and reverse-phase HPLC separation of prenylated quinones coupled to their quantification via diode array spectrophotometry was performed as described in Latimer et al. (2021). Demethylmenaquinone-8 (12.4 min) and menaquinone-8 (14 min) were monitored at 248 nm; ubiquinone-8 (9.5 min) was monitored at 275 nm. Ubiquinol-8, demethyl-menaquinol-8, and menaquinol-8 were fully oxidized into their quinone forms during extraction.
G. violaceus strain PCC 7421 from the American Type Culture Collection (#29082) was grown in BG-11 liquid medium (25 °C) with continuous illumination or in a 12 h photoperiod (30 µmol m -2 s -1 of PAR). Four-week-old cultures (25 ml) were harvested by centrifugation (2,200 × g for 5 min). G. violaceus strain VP3-01 from the Culture Collection of Autotrophic Organisms (#979) was harvested from four-week-old cultures on BG-11 solid medium. Colonies were resuspended in 750 µl of 0.9% (w/v) NaCl and then sequences, from leaf cDNAs as previously described (Stutts et al. 2023). For functional complementation assays in E. coli, these DNAs were cloned into EcoRI/XbaI-digested pBAD24 (Guzman et al. 1995) and introduced into E. coli Keio strain JW5581-1 ΔubiE778::kan (Baba et al. 2006). E. coli strains K12 and ΔubiE778::kan were transformed with empty pBAD24 to serve as positive and negative controls of quinone production, respectively. Cells were selected on LB plates containing kanamycin (50 µg/ml) and ampicillin (100 µg/ml). For the functional complementation assays in Synechocystis, gene Sll1653, its orthologs from G. morelensis, G. violaceus, S. sp. JA-2-3B'a, and their homolog from C. Obscuribacteriales were PCR-amplified from the corresponding pBAD24 constructs using gene-specific primers and sub-cloned into NdeI/BglII-digested pSynExp-2 expression vector (Sattler et al. 2003) using In-Fusion HD technology (Takara Bio, USA). These pSynExp-2 constructs were then introduced into Synechocystis Δsll1653::aadA knockout cells (Lohmann et al. 2006), and transformants were selected on BG-11 plates containing chloramphenicol (5 µg/ ml) and spectinomycin (15 µg/ml) at 22 °C in a 12-hour photoperiod (70 µmol m -2 s -1 of photosynthetically active radiation (PAR). For growth assays, clones were scooped directly from plates and were re-suspended in liquid BG-11 medium to a final OD 730 of 0.1. Serial dilutions were plated on solid BG-11 medium with (photomixotrophic conditions) or without (photoautotrophic conditions) 5 mM glucose. Plates were incubated in a 12-hour photoperiod (70 (a) Data are means of 3 measurements ± S.E. n.d., not detected 1 3
Clade 2 quinol C-methyltransferase orthologs were sampled from Gloeobacteria and Phycobacteria, the two groups that make the phylum of cyanobacteria, including at least one representative member from each major subclade as defined by Shih et al. (2013). The phylogenetic segregation of these proteins was consistent with their taxonomic group, indicating that these orthologs are unlikely to originate from recent events of horizontal gene transfer between Gloeobacteria and Phycobacteria (Fig. 2).
To verify the substrate selectivity of these enzymes, their cognate genes were codon-optimized, synthesized, and expressed in an E. coli ΔubiE quinol C-methyltransferase knockout, which lacks both menaquinone-8 and ubiquinone-8. In this complementation assay, expression of a monofunctional naphthoquinol C-methyltransferase restores exclusively the production of menaquinone-8. In contrast, expression of bifunctional naphthoquinol/benzoquinol C-methyltransferase restores the production of both menaquinone-8 and ubiquinone-8. Arabidopsis and pelleted by centrifugation (2,200 × g for 5 min). Cell pellets (~ 0.05 g) from strain PCC 7421 and strain VP3-01were resuspended in 75 µl of 0.9% (w/v) NaCl and then mixed with 200 µl of 100% ethanol and disrupted using a glass pestle in a Potter-Elvehjem tissue grinder. The grinder was washed twice with 100 µl of 100% ethanol and the washes were combined with the extract. Extractions of prenylated quinones from plate-grown Synechocystis cells were performed as described in Stutts et al. (2023). Prenylated benzoquinones and benzoquinols were immediately analyzed by reverse-phase HPLC as described in Block et al. (2013). Plastoquinone-9 (37.8 min) and methyl-plastoquinone-9 (45.7 min) were monitored by diode array spectrophotometry at 255 nm. Plastoquinol-9 (13.3 min) and methylplastoquinol-9 (15.1 min) were monitored fluorimetrically (290 nm excitation/330 nm emission). Prenylated naphthoquinones were analyzed via reverse-phase HPLC coupled to fluorimetric detection (238 nm excitation/426 nm emission) after in-line reduction as described in Widhalm et al. (2009). Retention times were 14.5 min for demethyl-phylloquinone and 17.3 min for phylloquinone. enzymes of Gloeobacter violaceus and Synechococcus sp. JA-2-3B'a did not discriminate between naphthoquinol and benzoquinol substrates (Fig. 2; Table 1).
This result apparently contradicts the previous findings that oxygenic phototrophs have evolved strictly monofunctional naphthoquinol C-methyltransferases in order to prevent cross-reactivity with plastoquinone-9 and subsequent poisoning of photosynthesis. Several scenarios could explain this surprising lack of catalytic stringency: i) G. violaceus and Synechococcus sp. JA-2-3B'a might be able to tolerate the presence of methyl-plastoquinone-9 or even use it as an alternative electron carrier to plastoquinone-9; (ii) these two taxa could prevent plastoquinone-9 from being methylated, or (iii) they could detoxify methyl-plastoquinone-9, for instance by recycling methyl-plastoquinone-9 into plastoquinone-9.
Isolates of Synechococcus sp. JA-2-3B'a are known for being contaminated by several other bacterial species (Allewalt et al. 2006) and were therefore not investigated here. We were able, however, to establish two independent cultures of G. violaceus corresponding to two pure isolates, strain PCC 7421 and strain VP3-01. After one month of cultivation, enough cells were obtained to permit the profiling of prenylated benzoquinones. After separation via reversephase chromatography, benzoquinones and benzoquinols from G. violaceus PCC 7421 and G. violaceus VP3-01 cell extracts were detected via diode array spectrophotometry and fluorometry, respectively (Fig. 3). Plastoquinone-9 and its reduced form, plastoquinol-9, were readily detected in both extracts (Fig. 3). In contrast, methyl-plastoquinone-9 and methyl-plastoquinol-9 were undetectable (Fig. 3). We concluded that G. violaceus has evolved a mechanism to either prevent the methylation of plastoquinone-9 or to eliminate methyl-plastoquinone-9.
To directly investigate the functionality of G. violaceus PCC 7421 and S. sp. JA-2-3B'a quinol C-methyltransferases in cyanobacteria, their encoding genes were subcloned into expression vector pSynExp-2. These constructs were then introduced into a Synechocystis demethyl-phylloquinol methyltransferase knockout (strain Δsll1653), and transformant clones were isolated using antibiotic selection. Δsll1653 clones expressing the bifunctional quinol Synechocystis sp. PCC 6803 demethyl-phylloquinol methyltransferases were selected as controls for strict naphthoquinol C-methyltransferase activity, while Arabidopsis COQ5, E. coli UbiE, and the UbiE ortholog of Candidatus Obscuribacterales in Vampirobrionia, the non-photosynthetic phylum closest to cyanobacteria (Soo et al. 2014), served as controls of bifunctionality (Fig. 2; Table 1). As expected, most Gloeobacteria and Phycobacteria orthologs displayed strict naphthoquinol C-methyltransferase activity (Fig. 2; Table 1). The non-detection of ubiquinone in these samples was not merely attributable to differences in the expression levels of the constructs, for should the corresponding enzymes have formed ubiquinone-8 in the same proportions relative to the menaquinone-8 levels measured in the positive controls, the expected amount of ubiquinone-8 would have been 3 to 57 times higher than the detection limit, depending on the constructs. Neither production of menaquinone-8 nor of ubiquinone-8 was detected for the Lyngbya ortholog indicating that the recombinant version of this enzyme was not functional in E. coli (Fig. 2; Table 1). This complementation assay revealed, however, two remarkable exceptions: When expressed in E. coli, the detected in the extract corresponding to the G. violaceus clone either (Fig. 5b). Low amounts of methyl-plastoquinone-9, representing ~ 14% of total prenylated benzoquinone pool, and traces of its reduced form (quinol) were detected in the cell extract corresponding to the Synechococcus sp. JA-2-3B'a construct (Fig. 5b). In marked contrast, methyl-plastoquinone-9 and methyl-plastoquinol-9 were the predominant benzoquinones (~ 72% of total prenylated benzoquinones) in the cells expressing the bifunctional enzyme from C. Obscuribacterales (Fig. 5b). These results demonstrate that the quinol C-methyltransferases from G. violaceus and S. sp. JA-2-3B'a are either not active (G.
C-methyltransferase from C. Obscuribacterales or its monofunctional naphthoquinol C-methyltransferase counterpart from Gloeobacter morelensis -the latter being most closely related to the G. violaceus ortholog (Fig. 2)-were isolated in parallel to serve as reference phenotypes. Clones harboring empty pSynExp-2 or native demethyl-phylloquinol methyltransferase (Sll1653) served as negative and positive controls of functional complementation, respectively. When scored via serial dilutions assays, either in photoautotrophic or photomixotrophic growth conditions, no significant difference in growth was observed between the cells harboring the G. violaceus and Synechococcus sp. JA-2-3B'a constructs and those harboring the constructs corresponding to the strictly monofunctional enzymes from G. morelensis and Synechocystis (Fig. 4). In contrast, cells harboring the C. Obscuribacterales construct displayed the pronounced growth retardation that is typically associated with the expression of bifunctional quinol C-methyltransferases in cyanobacteria (Fig. 4). Phylloquinone production was readily detected in the cognate cell extracts, verifying that these recombinant enzymes were indeed expressed and catalytically active (Fig. 5a). All the cyanobacterial constructs resulted in full complementation of demethyl-phylloquinol methyltransferase activity, while the residual amount of demethyl-phylloquinone represented ~ 28% of total prenylated naphthoquinones in the C. Obscuribacterales construct (Fig. 5a). As expected, plastoquinone-9 was the sole prenylated benzoquinone detected in the extracts of clones expressing the authentic monofunctional naphthoquinol C-methyltransferases of G. morelensis and Synechocystis (Sll1653). Remarkably, no methyl-plastoquinone-9 was of methyl-plastoquinone-9 in plastids, while a dual targeted monofunctional clade 2 counterpart would not be active on the ubiquinone precursor 2-methoxy-6-polyprenyl-1,4-benzoquinol in mitochondria (Fig. 1).
Here we identify two cyanobacterial clade 2 quinol C-methyltransferases that do not display absolute exclusion of prenylated benzoquinol substrates, a feature that until now was thought to typify this class of enzymes. Both quinol C-methyltransferases, however, retain the ability to discriminate plastoquinone-9. That such functionally hybrid enzymes occur in both Gloeobacteria (G. violaceus) and Phycobacteria (S. sp. JA-2-3B'a) suggests that partial discrimination of prenylated benzoquinol substrates was a trait present prior to the split between these two cyanobacterial lineages. We therefore propose that the G. violaceus and S. sp. JA-2-3B'a hybrid enzymes represent the remnants of a primeval version of quinol C-methyltransferase, which was transitioning from a fully bifunctional ancestor during the emergence of plastoquinone as a photosynthetic electron carrier (Fig. 6). In most present day cyanobacteria as well as in all their plastid orthologs (Stutts et al. 2023), clade 2 quinol C-methyltransferases have further evolved exclusive recognition of naphthoquinol substrates, likely to prevent moonlighting activity on plastoquinol-9 (Fig. 6). The low amount of methyl-plastoquinone-9 detected in Synechocystis cells expressing recombinant S. sp. JA-2-3B'a quinol C-methyltransferase (Fig. 5b) supports this evolutionary scenario. One should mention, however, that in this functional complementation assay S. sp. JA-2-3B'a violaceus) or marginally active (S. sp. JA-2-3B'a) with plastoquinol-9 as a substrate.
Clade 2 quinol C-methyltransferases of cyanobacteria and plastids are subjected to two stringent and seemingly opposite selection pressures. These enzymes must ensure efficient methylation of the carbon in ortho of the prenyl chain of naphthoquinols -most often phylloquinol or, in some red algae and some cyanobacteria, menaquinol (Mimuro et al. 2005;Yoshida et al. 2003)-but must not act on the prenylated benzoquinol, plastoquinol-9. This substrate discrimination is required because prenylated naphthoquinols must be methylated for optimal electron transfer in photosystem I (Lohmann et al. 2006;Fatihi et al. 2015), while methyl-plastoquinone-9 poisons the photosynthetic electron transport chain (Stutts et al. 2023). In sharp contrast, clade I quinol C-methyltransferases, which are near universally distributed throughout the prokaryotic and eukaryotic kingdoms, can act indiscriminately on prenylated benzoquinols and prenylated naphthoquinols. Prohibition of methylplastoquinone-9 formation thus explains why in plants the biosynthetic pathway of phylloquinone in plastids and that of the respiratory cofactor ubiquinone in mitochondria cannot operate via a single dual targeted quinol C-methyltransferase. In other words, a dual targeted bifunctional clade 1 quinol C-methyltransferase would result in the production Fig. 6 Proposed scenario for the evolution of quinol C-methyltransferase functionality. Cell and organelle icons were created in BioRender. https:// BioRender.com/kcp0xiq models (Wang et al. 2022), the Gloeobacter enzymes are predicted to have an additional N-terminal α-helix that caps the AdoMet binding site (Fig. 7b-d). The proximity of this α-helix to the active site suggests a possible role in controlling access to the substrate. These protein models indicate that none of the 17 differing residues were located directly in the predicted active site (Fig. 7b), therefore we proceeded to design a series of chimeric versions of the G. morelensis enzyme containing residues swaps from its G. violaceus counterpart. Such changes included T38I/L115V located in the vicinity of the predicted active site (construct A), and A156G/R168W/E171A predicted to introduce significant conformational shifts at the dimer interface (construct B). The R5Q/Q94R/D108A/T132A/R203Q swaps (construct C), which are predicted to result in marginal structural changes at the monomer surface, and the native enzymes from G. morelensis and G. violaceus served as controls. These cognate DNAs were then codon-optimized, synthesized, and separately introduced in the E. coli ΔubiE knockout. Although all constructs rescued menaquinone production, verifying that the corresponding enzymes were functional, none resulted in the detection of ubiquinone. It therefore appears that the difference in substrate recognition quinol C-methyltransferase was expressed under the control of a constitutive psbA2 promoter, and therefore that in its native environment this enzyme is likely to form even lower amounts -if any-of methyl-plastoquinone-9.
The quinol C-methyltransferases from G. violaceus (hybrid) and G. morelensis (monofunctional) share over 94% identity, the two enzymes differing by only 17 amino acids (Fig. 7a). We attempted to leverage this low divergence in primary sequence to identify the residues responsible for the difference in substrate recognition between the two enzymes. When superimposed with the yeast Coq5 enzyme (PDB:4OBW; Dai et al. 2014), the AlphaFoldpredicted structural models of G. violaceus and G. morelensis C-methyltransferases gave RMSDs of 1.4-1.5 Å Cα (Fig. 7b). As expected, these enzymes contain a Class-I AdoMet-dependent methyltransferase domain featuring a Rossmann-like α/β fold containing a seven stranded β-sheet with six parallel strands and the seventh anti-parallel to the others surrounded by five α-helices. The other three helices form a dimer interface in yeast Coq5 and are predicted to be conserved in the G. violaceus and G. morelensis enzymes, facilitating the formation of the same dimeric structures (Fig. 7c, d). Like the yeast and human COQ5 AlphaFold is not visible in the S. cerevisiae Coq5 protein used for X-ray diffraction, possibly because this region was disordered. (c) Ribbon diagram of the AlphaFold-predicted structure of the G. morelensis enzyme (purple) with all non-identical residues with G. violaceus shown as yellow spheres. Residues chosen for subsequent mutagenesis are labeled. (d) Ribbon diagram of the AlphaFold-predicted structure of the G. violaceus enzyme (blue) with all non-identical residues with G. morelensis shown as yellow spheres
Page 3 of 11