Emergence of cytochrome bc complexes in the context of photosynthesis

Abstract

The cytochrome bc (cyt bc) complexes are involved in Q‐cycling; they oxidize membrane quinols by high‐potential electron acceptors, such as cytochromes or plastocyanin, and generate transmembrane proton gradient. In several prokaryotic lineages, and also in plant chloroplasts, the catalytic core of the cyt bc complexes is built of a four‐helical cytochrome b (cyt b) that contains three hemes, a three‐helical subunit IV, and an iron‐sulfur Rieske protein (cytochrome b ₆ f‐type complexes). In other prokaryotic lineages, and also in mitochondria, the cyt b subunit is fused with subunit IV, yielding a seven‐ or eight‐helical cyt b with only two hemes (cyt bc ₁‐type complexes). Here we present an updated phylogenomic analysis of the cyt b subunits of cyt bc complexes. This analysis provides further support to our earlier suggestion that (1) the ancestral version of cyt bc complex contained a small four‐helical cyt b with three hemes similar to the plant cytochrome b ₆ and (2) independent fusion events led to the formation of large cyts b in several lineages. In the search for a primordial function for the ancestral cyt bc complex, we address the intimate connection between the cyt bc complexes and photosynthesis. Indeed, the Q‐cycle turnover in the cyt bc complexes demands high‐potential electron acceptors. Before the Great Oxygenation Event, the biosphere had been highly reduced, so high‐potential electron acceptors could only be generated upon light‐driven charge separation. It appears that an ancestral cyt bc complex capable of Q‐cycling has emerged in conjunction with the (bacterio)chlorophyll‐based photosynthetic systems that continuously generated electron vacancies at the oxidized (bacterio)chlorophyll molecules.

Abbreviations

BCl

bacteriochlorophyll

BLAST

Basic Local Alignment Search Tool

Ca.

Candidatus (in bacterial names)

COG

Cluster of Ortologous Groups of proteins

cyt

cytochrome

EPR

Electronic Paramagnetic Resonance

ET

electron transfer

GOE

Great Oxygenation Event

LGT

lateral gene transfer

LUCA

Last Universal Common/Cellular Ancestor

PRC

photochemical reaction center

PSI

photosystem I

Introduction

The cytochrome bc complexes (electrogenic quinone:cytochrome c / plastocyanin oxidoreductases, hereafter cyt bc complexes) play key roles in eukaryotic respiration (mitochondrial complex III or cytochrome bc ₁ complex, hereafter cyt bc ₁ complex) and photosynthesis (cytochrome b ₆ f‐complex of chloroplasts, hereafter cyt b ₆ f complex), see Fig. 1. These enzymes catalyze electron transfer (ET) from diverse membrane quinols to high‐potential redox carriers (usually c‐type cytochromes) and couple these endergonic redox reactions with generation of the transmembrane difference in electrochemical potential of protons (the proton gradient) by a Q‐cycle mechanism (see Hauska et al. 1983, Berry et al. 2000, Mulkidjanian 2005, 2010, Cramer et al. 2011, Al‐Attar and de Vries 2012, Cramer and Hasan 2016, Esser et al. 2016, and Bhaduri et al. 2017 for reviews). As a result, one side of the membrane becomes positively charged (p‐side), whereas the other one charges negatively (n‐side) (Cramer and Knaff 1990).

Figure 1.

Two types of cytochrome bc complexes. Left panel, structure of the mitochondrial cyt bc ₁ complex [PDB entry 2FYN (Esser et al. 2006)]; right panel, structure of the cyanobacterial cyt b ₆ f complex [PDB entry 1VF5 (Kurisu et al. 2003)]. Typical operon structures of complexes are shown above their crystal structures. Cytochrome b ₆‐like parts (the four‐helical bundles) are colored orange, subunit IV‐like parts are colored brown, Rieske proteins are colored pink. Unrelated cytochromes c ₁ and f, which obtain electrons from the FeS cluster of the Rieske protein are colored yellow and green, respectively. The figure was produced with rasmol (Sayle and Milner‐White 1995).

In the beginning of the 1980s, both the functional similarity of the cyt bc ₁ and cyt b ₆ f complexes and their evolutionary relatedness have been demonstrated (Hauska et al. 1983, Widger et al. 1984). The four N‐terminal helices of the eight‐helical, large cytochrome b (cyt b) in the cyt bc ₁ complex were shown to be homologous to the cyt b ₆ (four transmembrane helices, hereafter small cyt b), whereas the three C‐terminal transmembrane helices of the large cyt b were found to resemble the subunit IV (PetD) of the cyt b₆f complex. In addition, the relatedness of the Rieske proteins in the mitochondrial and chloroplast cyt bc complexes was noted (Malkin and Aparicio 1975, Furbacher et al. 1996). The availability of the genomic data for diverse prokaryotes has shown that some prokaryotic lineages contain large cyts b, while others contain small cyts b and separate homologs of subunit IV (Schutz et al. 2000, Nitschke et al. 2010, Baymann et al. 2012, ten Brink et al. 2013). The cyt bc complexes with small and large cyts b are often referred to as cyt b ₆ f‐like and cyt bc ₁‐like complexes, respectively.

Starting from 1996, many crystal structures were obtained for cyt bc ₁ complexes from different organisms by X‐ray crystallography (Xia et al. 1997, Iwata et al. 1998, Zhang et al. 1998, Lange and Hunte 2002, Berry and Huang 2003, Esser et al. 2006, 2008, 2016, Berry et al. 2013). Although the number of subunits in cyt bc ₁‐complexes from different sources varies between 3 and 11, the catalytic cores, made of a large cyt b, Rieske protein and cytochrome c ₁ (cyt c ₁), are highly conserved (see Fig. 1). The cyt bc ₁ complex is an intertwined dimer. Each cyt b is a bundle of eight alpha‐helices that accommodates two (proto)hemes, whereby one heme (b _p) is closer to the p‐side of the membrane and the other heme (b _n) is closer to the n‐side of the membrane. From the p‐side of the membrane, the [Fe₂S₂] cluster‐carrying domain of the iron‐sulfur Rieske protein (the FeS domain) docks to the cyt b close to the heme b _p; the docking interface serves as the quinol oxidizing site P of the complex. Single membrane‐anchoring alpha‐helices of Rieske proteins are tilted in the membrane; each of them interacts with the cyt b of the other monomer, thus keeping the complex together. The heme‐carrying domains of the two cyts c ₁, which accept electrons from the FeS domains, are also anchored in the membrane by single alpha‐helices.

As discussed in several publications (Stroebel et al. 2003, Cramer et al. 2011, Cramer and Hasan 2016), crystal structures of cyt b ₆ f complexes of green plants and cyanobacteria (Kurisu et al. 2003, Stroebel et al. 2003, Cramer et al. 2011, Hasan and Cramer 2014, Cramer and Hasan 2016) partially overlap with the structures of cyt bc ₁ complexes (Xia et al. 1997, Iwata et al. 1998, Zhang et al. 1998, Lange and Hunte 2002, Berry and Huang 2003, Berry et al. 2004, 2013, Esser et al. 2006, 2008, 2016). Specifically, an almost exact overlap is seen for the four heme‐binding N‐terminal helices A–D of the large cyt b and the four helices of cyt b ₆, as well as for the FeS cluster‐carrying domains of the two complexes. This structure conservation reflects the key importance of these components for the catalysis. The four‐helical cyt b ₆ subunit, besides accommodating two b‐type hemes, carries an additional c‐type heme (denoted c _n or c _i) that is covalently bound to a cysteine residue of cyt b and does not have a counterpart in the cyt bc ₁ complex (Kurisu et al. 2003, Stroebel et al. 2003, Zhang et al. 2004, Alric et al. 2005, Cramer et al. 2006, Zatsman et al. 2006, Baymann et al. 2007, de Lacroix de Lavalette et al. 2009, Bergdoll et al. 2016, Zito and Alric 2016). The c _n heme is sandwiched, close to the n‐side of the membrane, between the cyt b and the three‐helical subunit IV; the iron atom of this heme is connected to the propionate of heme b _n by a water bridge. In our previous work, it was noted that the presence of an additional heme c _n seems to correlate with the presence not just of the binding cysteine residue, but of the whole, potentially heme c _n‐binding CxGG motif (Dibrova et al. 2013). In addition, the subunit IV of the cyt b ₆ f complexes binds single molecules of chlorophyll a and β‐carotene (Pierre et al. 1997, Zhang et al. 1999, Kurisu et al. 2003, Stroebel et al. 2003, Dashdorj et al. 2005, Baniulis et al. 2008). In the cyt b ₆ f complexes, the FeS domain of the Rieske protein delivers electrons to the cyt f, which, while carrying a c‐type heme, is structurally unrelated to the cyt c ₁ of the cyt bc ₁ complex (Martinez et al. 1994).

The cyt bc complexes appear to be functional dimers, capable of electron exchange between the monomers via the closely placed heme b _p, which contributes to the robustness of the enzyme mechanism (see Xia et al. 1997, Gopta et al. 1998, Mulkidjanian 2007, Swierczek et al. 2010, Cramer et al. 2011 and Sarewicz et al. 2016, for details).

Hence, the structural similarity between the cyt bc ₁ and cyt b ₆ f complexes is limited to the cyt b subunits and the iron‐sulfur Rieske protein (Stroebel et al. 2003, Cramer et al. 2011, Cramer and Hasan 2016). Because of that, Kramer et al. (2009) suggested referring to cyt bc ₁ and cyt b₆f complexes together as ‘Rieske/cyt b complexes’. Although we agree with this argumentation, we still mostly use here the more traditional term ‘cytochrome bc complex’ to address the members of the whole enzyme family.

The very presence of distinct types of cyt bc complexes in plants and animals prompted the question on the evolutionary relations between them. As initially only large cyts b were found in archaea, it has been speculated that the Last Universal Common Ancestor (LUCA), the common ancestor of bacteria and archaea, already contained a large cyt b, whereas the cyt b ₆ f‐type complexes with a short cyt b could have emerged after some later fission event(s) (Castresana et al. 1995, Schutz et al. 2000, Baymann et al. 2003, 2012, Nitschke et al. 2010, ten Brink et al. 2013).

In our previous work, we have applied phylogenomic analysis to the genes of cyt b subunits in the organisms with fully sequenced genomes (Dibrova et al. 2013). That analysis showed that the genes of cyts b fall into several independent subfamilies, which, supposedly, unify cyt bc complexes with similar functionalities. Small and large cyts b have been found both in bacteria and archaea. Based on the phylogenomic data and comparative structural analysis, we have suggested that the ancestral form of cyt bc complexes contained a small cyt b together with a subunit IV, and that the cyt bc complexes with large cyts b resulted from several independent fusions of small cyts b with respective subunits IV (Dibrova et al. 2013).

This evolutionary scenario has been recently challenged (Kao and Hunte 2014, Ducluzeau and Nitschke 2016). In the view of new important data (Bergdoll et al. 2016, de Almeida et al. 2016) and a dramatic increase in the number of sequenced genomes, we revisit here the evolutionary history of the cyt bc complexes and present an updated phylogenomic analysis of the cyt b subunits. We show that new genomic data provide further support to our earlier suggestion on the evolutionary primacy of cyt bc complexes with small cyts b.

Materials and methods

Selection of cyt b sequences

We based our analysis on the assignment of proteins from 711 genomes to Cluster of Ortologous Groups of proteins (COGs) database released in 2015 (Galperin et al. 2015). Our core selection of the genomes was obtained by removing a number of genomes from overrepresented phyla, specifically proteobacteria, firmicutes and actinobacteria – a single species per order was selected. The condensed sample contained 347 genomes.

The COG1290 (annotated as the ‘Cytochrome b subunit of the bc complex’) covers both the cyt b ₆ f part and subunit IV, thus before performing the alignment we manually fused these proteins (in those cases when it was possible).

In order to enrich the sample with new sequences from the clades described in our previous work, we ran blast (Basic Local Alignment Search Tool) on the NCBI nr database (Altschul et al. 1997) for two proteins: DAMO_0768 from Candidatus (Ca.) Methylomirabilis oxyfera (member of the clade G) and Psta_3849 from Pirellula staleyi DSM 6068 (member of the clade D). To obtain only closely related sequences, presumably belonging to the same clade, we analyzed the blast bit score graphs (see Appendix S1 in Supporting information for an example). As we wanted to analyze the genome neighborhoods of the corresponding genes, we selected only those records that could be traced back to the corresponding genome.

As several important species (Heliobacterium modesticaldum Ice1, Desulfotomaculum ruminis, Desulfitobacterium hafniense DCB‐2, Brevibacillus brevis NBRC 100599, Oceanobacillus iheyensis HTE831) were not included in the original sample, we manually added fused cyt b and subunit IV sequences from their genomes.

Phylogenetic tree construction

The final protein set in our sample contained 287 proteins. Their sequences were aligned with muscle (MUltiple Sequence Comparison by Log‐Expectation, Edgar 2004), truncated and poorly aligned protein sequences were removed, and the remaining sequences were re‐aligned. Conserved blocks were selected after manual adjustment of the alignment and used for phylogenetic tree analysis (144 positions were actually used for the tree construction), see Fig. S2 for the multiple alignment regions. Phylogenetic trees were constructed in mega 7 (Molecular Evolutionary Genetics Analysis Version 7.0, Kumar et al. 2016) using the neighbor‐joining method (Saitou and Nei 1987) with the JTT matrix for distance calculation (Jones et al. 1992) and uniformly distributed rates among sites. The number of bootstrap replications (Felsenstein 1985) for each tree was 100. The full version of the tree is presented in Fig. S1.

In order to obtain a shorter, printable version of the tree (Fig. 2), we manually collapsed clades which were already discussed in details in (Dibrova et al., 2013). This resulted in 167 sequences, for which the tree was reconstructed as described above.

Figure 2.

Phylogenetic tree for selected cytochrome b sequences (see Fig. S1 for the full version). Subfamilies from A to J are marked as on the previously published tree (Dibrova et al. 2013) and compressed. New subfamilies K, L, M, N and G′ are presented in more detail. The subfamily groups and tree leaves are colored according to the taxonomy of the respective organisms, as depicted at the bottom of the figure. Fading colors indicate those compressed clades that contain mostly sequences from a certain phylum and a small number of representatives of other taxonomic groups. Clades D and G include several different phyla, see Fig. S1. Each protein is indicated by its genomic locus tag, followed by the name of the source organism; in two instances, the proteins are labeled by their PDB codes (1VF5 and 1NTM). Bootstrap support was obtained from 100 replications, bootstrap values lower than 15% are not shown. Abbreviations are as follows: Heme c_n – additional heme‐binding site in the N‐terminal region of cytochrome b, H2 and H4 – second and fourth helices of cytochrome b, respectively, Q_P site – the quinone‐binding site in the subunit IV. In the operon structures, shown on the right, the four‐helical region of the cytochrome b (either small cytochrome b or the homologous part of the large cytochrome b) is colored orange, the subunit IV is colored dark red and the Rieske protein is colored pink. Complete color code for the operon structures is given in Fig. S1.

We used a desktop version of the cognat software (Klimchuk et al., unpublished) to obtain genome neighborhoods of all proteins included in the tree. This information was mapped on the tree (Figs 2 and S1).

Results and discussion

An updated phylogenetic tree of bacterial and archaeal cyt b sequences

With a larger set of genomes made available in the past several years, we performed a phylogenomic analysis of cyts b (see Appendix S1 for methods). We based our analysis on the most recent release of the COG database (Galperin et al. 2015) which contains a representative set of 711 complete prokaryotic genomes. We decided to focus on complete genomes, as they allow to perform phylogenomic analysis and to trace possible events of lateral gene transfer (LGT), gene duplication and gene loss. We also performed a large‐scale blast search (Altschul et al. 1997) in a search for new members of several subfamilies, which we described in the previous work (Dibrova et al. 2013), as well as new subfamilies, and inspected 68 archaeal genomes from additional taxa that had not been included into the aforementioned set. We built a phylogenetic tree for the long cyt b sequences and the corresponding sequences of the ‘short’ cyts b and subunits IV of the b ₆ f‐type complexes, see Appendix S1 for further details on the methods used.

A compressed version of the phylogenetic tree of cyts b is shown in Fig. 2 and the complete zoomable version is presented in Fig. S1. As has been shown previously (Baymann et al. 2012, Dibrova et al. 2013, ten Brink et al. 2013), cyt b sequences can be separated into several subfamilies, members of which not only show sequence similarity, but also, in many cases, share specific, functionally relevant traits. It is noteworthy that some of these subfamilies were not present in the previously published phylogenomic trees, which had been obtained with fewer genomes (Baymann et al. 2012, Dibrova et al. 2013, ten Brink et al. 2013), these novel clades are marked red. Figs 2 and S1 also contains information on the nature of potential ligands of the two hemes b, on the characteristic motif of the quinol‐binding site (in the sequences of long cyts b or subunits IV, respectively), and on the conservation of the c _n heme‐binding CxGG motif. In this motif, the cysteine residue covalently binds the c _n heme, whereas the first glycine residue, which is conserved in many membrane cyts b, provides space for the edge of the heme b _n (Tron et al. 1991, Yun et al. 1992, Berry and Walker 2008). The amide group of the second glycine residue stabilizes the water molecule that connects the iron atom of heme c _n with the propionic group of heme b _n; in addition, this glycine residue, owing to its small size, enables the linkage between the two hemes (Kurisu et al. 2003, Stroebel et al. 2003). In the previous work, we also traced the nature of redox‐proteins that accept electrons from the Rieske proteins; they were found to be quite different in different subfamilies (Dibrova et al. 2013). Upon updating the phylogenetic tree, we have not specifically addressed the nature of such electron acceptors in newly defined subfamilies of cyt bc complexes, although investigation of their diversity might, in principle, help in establishing relations between different subfamilies of cyt bc complexes.

From a bird's‐eye view, the tree in Figs 2 and S1 shows that many subfamilies of cyt bc complexes contain representatives of diverse taxonomic groups. Hence, cyt bc complexes are prone to the LGT, as already discussed at length in (Dibrova et al. 2013).

Only short cyt b sequences, together with the homologs of subunit IV, are found in the previously described subfamilies A (representing sequences from cyanobacteria), B and C (firmicutes), as well as in the archaeal subfamily N, which drastically expanded with the new genomes sequenced. The CxGG motif is present in all cyt b sequences of subfamilies A, B and C, but is absent in the subfamily N. In the sequences of subfamily N, only the first of the two glycine residues of the CxGG motif is conserved, whereas the cysteine and the second glycine residues are not conserved.

Only long cyt b sequences are found in the subfamily I, which contains the ‘paradigmal’ cyt bc ₁ complexes of α‐proteobacteria that are related to their mitochondrial counterparts. In addition, long cyt b sequences are present in the subfamily J (Chlorobi), in the newly described clade G′ of the subfamily G (where the genes of cyts b are found in large conserved operons that apparently code for protein complexes of an unknown function), and in archaeal subfamilies F, E and E′. The two glycine residues of the CxGG motif are found in all sequences of subfamily J, but only in some sequences in subfamilies I, G′, E and E′.

Several subfamilies contain both short and long cyt b sequences. With more genomes analyzed, the subfamily D was found to contain cyt bc complexes of Rubrobacter xylanophilus and Conexibacter woesei with short cytochromes; in these cases, the CxGG motif is replaced by a TLGS motif. The newly described subfamily L contains short bacterial sequences of cyts b with the CxGG motif, with the exception of Dacet_1797 protein from Denitrovibrio acetiphilus DSM 12809, which is ‘long’ and contains a YAGA motif that is supposedly unable to bind a heme. Another interesting feature of the subfamily L is that the likely quinol‐binding motif SxWF in subunits IV deviates from the typical PxWY motif.

The subfamily G shows the largest diversity with regard to the lengths of cyt b sequences and the taxonomy of host organisms. The cyt b sequences are mostly short; a distinct correlation between the length of the cyt b sequence and the presence of the CxGG motif is seen throughout the subfamily. We have previously noted the taxonomic diversity of the subfamily G (Dibrova et al. 2013), this subfamily has now been expanded and additionally includes proteins from Ca. Kryptonia, Ca. Saccharibacteria, Patescibacteria group, Ktedonobacteria and Nitrospinae. With even more genomes sequenced, some of the groups in the subfamily G, perhaps, will have to be upgraded to separate subfamilies, see e.g. the clade G′ in Figs 2 and S1.

Several subfamilies of cyts b show specific traits. As shown in Figs 2 and S1, cyanobacteria harbor two cyt b subfamilies, namely A and K, both colored bright green. While subfamily A contains omnipresent cyts b of the cyt b ₆ f complexes, the second cyanobacterial subfamily K contains short cyt b sequences lacking the CxGG motif and found only in some cyanobacterial genomes in addition to the cyts b of subfamily A. In this subfamily, the heme‐binding histidine in helix 2 is replaced with either glutamic acid or glutamine and, in addition, the heme b _n‐accommodating glycine residue in helix 1 is absent (Figs 2 and S1). The functions of these small cyts b, as well as the number of hemes bound by them, remain obscure. In most genomes that contain cyts b of the subfamily K, the only encoded subunits IV are those related to the cyt b ₆ f complexes. The only exceptions are genomes of Gloeobacter violaceus PCC 7421 and Chamaesiphon minutus PCC 6605. These two organisms encode a second cyt b that groups with the clade K and an additional subunit IV with a typical PEWY quinone‐binding motif (the gvip258–gvip257 pair in G. violaceus and Cha6605_5913–Cha6605_5912 in C. minutus). It cannot be excluded that these two cases witness gene duplication‐induced transitions from a ‘normal’ cyt b ₆ f complex to the proteins of the subfamily K with obscure function.

The newly described planctomycetal subfamily M contains short cyt b sequences with the CxGG motifs that should be able to bind a heme. In the cases of Ca. Brocadia fulgida and Ca. Jettenia caeni, we were able to identify subunit IV genes in the respective genomic neighborhoods, but for other genomes (Ca. Scalindua brodae, Ca. Kuenenia stuttgartiensis and Ca. Brocadia sinica) no such genes were found. This could be, however, due to incomplete genomic data for these organisms.

All studied genomes contain multiple genes of Rieske protein homologs, which, however, not always are located next to cyts b. For instance, Ca. B. fulgida carries three genes coding for Rieske proteins, two of which are located upstream of the cyt b genes belonging to the subfamilies G and D, whereas the third copy (accession number KKO21266.1) is encoded in a separate contig and could be involved in the functioning of cyts b from the subfamily M.

Finally, large proteins that contain cyt b domains form the subfamily H. The functions of these proteins are yet to be established.

The data in Figs 2 and S1 are compatible with our earlier suggestion on the evolutionary primacy of cyt bc complexes with small cyts b. Multiple, independent fusions with subunit IV would then produce the long cyt b sequences that are found in different subfamilies of the phylogenetic tree. As we have noted earlier, the linkers that connect the cyt b parts with the subunit IV parts in the long cyt b sequences are highly conserved within each subfamily but completely distinct between different subfamilies (Dibrova et al. 2013). This dissimilarity indicates an independent emergence of the linkers in different subfamilies. Otherwise, if these linker regions were initially present in the ancestral, large cyt b, one would expect an overall conservation of the respective stretches of amino acid residues among long cyt b sequences. This is because these stretches are functionally important, being involved in the formation of the quinone‐reducing site N of the cyt bc complexes (Berry et al. 2000). In addition, the primacy of small, heme c _n‐carrying cyts b is also supported by the partial conservation of the c _n heme‐binding CxGG motif in the sequences of many long cyts b, see Figs 2 and S1 and also (Stroebel et al. 2003).

In our previous work, we have noted that the suggested fusion of a short cyt b and subunit IV could be directly traced within the subfamily G [see Fig. S1 and Dibrova et al. (2013)]. One more such case is seen in the newly described subfamily L. This subfamily contains a clade with bootstrap value as high as 95, which includes three short cyt b sequences and one long cyt b sequence, all of them with an unusual tentative quinone‐binding motif S[A/G]WF (located either in the long cyt b sequence or in the sequences of subunits IV). The short sequences have the c _n heme‐binding motif CLGG, whereas the only long sequence has a YAGA motif in the respective position. We interpret these data as evidence of a fusion event within this clade of subfamily L. An alternative suggestion on the primacy of the long cyt b (Castresana et al. 1995, Schutz et al. 2000, Baymann et al. 2003, 2012, Nitschke et al. 2010, ten Brink et al. 2013, Kao and Hunte 2014) would imply that, within a separate subfamily of large cyts b with an uncommon S[A/G]WF motif in the tentative quinol‐binding site, (1) the cyt b gene split exactly in the same point as in other subfamilies of cyts b with the dominant P[E/D]W[F/Y] quinone‐binding motif, and (2) the CxGG motif independently appeared in the same position as in other subfamilies. Such coincidences seem to be extremely unlikely; see our previous work (Dibrova et al. 2013) for further arguments.

These considerations, based on the new data, buttress our earlier suggestion on the emergence of long cyt b sequences from several independent fusions of short cyt b sequences with their cognate subunits IV (Dibrova et al. 2013).

Modular assembly of ancestral cyt bc complexes

The identification of short, four‐helical cyt b as an evolutionary primal form (Dibrova et al. 2013) led to a rather straightforward scenario on the emergence of the ancestral cyt bc complexes from available protein modules, see Fig. 3. Indeed, a bundle of four alpha‐helices represents one of the most widespread protein folds (Berry et al. 2000); it is one of the few folds that are found both in water‐soluble and membrane proteins (Neumann et al. 2010). Binding of two hemes has been shown to stabilize the fold (Choma et al. 1994, Rojas et al. 1997, Koch and Schneider 2016). Membrane‐embedded four‐helical cytochromes usually serve as membrane anchors for large, protruding oxidoreductase subunits where a distal substrate‐specific binding site is connected by an electron‐transferring ‘wire’ of iron‐sulfur clusters with the membrane, such as e.g. in formate dehydrogenase (Jormakka et al. 2003) or Ni–Fe hydrogenase (Pandelia et al. 2012). By accommodating two electrons, the two hemes enable electron coupling between one‐electron carriers, such as iron‐sulfur centers, and two‐electron carriers, such as the membrane‐localized quinones (Lancaster et al. 2008).

Figure 3.

Evolutionary scenario of the emergence of cyt bc complexes. Cytochrome b ₆‐like parts (the four‐helical bundles) are colored orange, subunit IV‐like parts are colored brown, Rieske proteins are colored pink and the protruding oxidoreductase is colored gray. The figure was produced with rasmol (Sayle and Milner‐White 1995). See main text for further details.

It has been noted earlier that short cyts b which belong to the subfamily G co‐occur with large proteins possessing a NAD(P)‐binding domain and many conserved cysteine residues capable of binding FeS‐clusters and typical for oxidoreductases (Kartal et al. 2011, Dibrova et al. 2013, Kartal et al. 2013). In some planctomycetes, the genes encoding proteins of this family are found within the operons of cyt bc complexes (Kartal et al. 2011, 2013, Dibrova et al. 2013). De Almeida et al. (2016) have recently shown that in Ca. K. stuttgartiensis such protein complexes, which contain both the cyts b and large oxidoreductase subunits, are expressed at high levels. De Almeida et al. (2016) suggested a function for such enzyme complexes in ET from hydrazine to menaquinone‐7. Whatever the function of these enzymes, the cyt bc complexes with a large cytoplasmic oxidoreductase subunit anchored in the membrane by a small cyt b do exist. The updated phylogenomic tree in Fig. 2 and S1 shows the proximity of the genes of cyts b and similar oxidoreductase subunits in many other cases.

Hence, as discussed at length in Dibrova et al. (2013) and as shown in Fig. 3, the emergence of the first cyt bc complexes could involve a combination of a four‐helical, two‐heme cyt b (perhaps, with a cytoplasmic oxidoreductase subunit attached) and a membrane‐anchored Rieske‐type iron‐sulfur cluster. Rieske‐type proteins are not unique for cyt bc complexes but are found also in other enzyme complexes (Link 1999, Baymann et al. 2003, Lebrun et al. 2006, ten Brink et al. 2013); the ancestral Rieske protein could mediate the electron exchange between the hemes of cyt b and redox‐active components at the periplasmic surface of the cell membrane.

The unique part of cyt bc complexes with a small cyt b is the subunit IV that shapes the quinol‐oxidizing site P. This three‐helical membrane subunit does not show significant similarity to other known protein families, so its origin remains obscure.

Another unique feature, as compared with other dehydrogenases that are anchored by two‐heme cyts b, is the presence of an additional, covalently bound c _n heme in cyt b, from the n‐side of the membrane.

It is tempting to speculate that both these features could be acquired together, paving the way to the ancestral cyt bc complex. Indeed, the available structures of cyt b ₆ f complexes (Fig. 4) show that heme c _n is sandwiched between the cyt b and subunit IV, so that the hydrophobic heme surface make contacts with hydrophobic and aromatic residues of both subunits (Kurisu et al. 2003, Stroebel et al. 2003, Cramer et al. 2011, Hasan and Cramer 2014, Cramer and Hasan 2016). Our inspection of the structure of the cyt b ₆ f complex with the best resolution (PDB entry 4OGQ) (Hasan and Cramer 2014) showed that, in addition, the propionate residues of heme c _n are involved in a hydrogen‐bonded network that connects the N‐terminal and C‐terminal helices of cyt b ₆ with the N‐terminal helix of the subunit IV (Fig. 4). Hence, heme c _n, in addition to its function as a redox carrier, appears to participate in linking the cyt b ₆ with subunit IV. This function of the heme is extremely well conserved. In Fig. 4, the structure of the best resolved cyanobacterial cyt b ₆ f complex is superimposed on the structure of the eukaryotic cyt b ₆ f complex from Chlamydomonas reinhardtii. The orientation of all side chains that are involved in the hydrogen‐bonded network is exactly the same in both structures, which implies the presence of the same hydrogen‐bonded network involving the c _n heme in the eukaryotic cyt b ₆ f complex. Taking into account that separation of C. reinhardtii from cyanobacteria happened more than 1 billion years ago (Cramer et al. 2004), the strict conservation of the hydrogen‐bonded network between the c _n heme, cyt b and subunit IV points to the importance of this network.

Figure 4.

Structural comparison of cyanobacterial and eukaryotic cyt b ₆ f complexes. (A) Quinone‐reducing center N of the cyt b ₆ f complex from Nostoc sp. PCC 7120 [PDB: 4OGQ (Hasan and Cramer 2014)], resolution: 2.5 Å). The cytochrome b ₆ is shown in orange, the subunit IV is shown in dark red. Carbon atoms of the hemes are shown in cyan, carbon atoms of the quinone are in magenta; otherwise oxygen atoms are in red, nitrogen atoms are in dark blue and iron atoms are in ochre. All residues with side chain atoms at a distance of ≤4 Å from heme c _n are shown as sticks. The cyt b ₆ contributes Val30, Cys35, Leu41 (helix 1 in the foreground); Phe203, Ile206, Ile211 (helix 4 in the background); additionally, residues Lys24 and Arg207 of cyt b ₆ participate in the hydrogen bond network around the propionate of the heme c _n. The subunit IV (dark red) accommodates heme c _n with Phe40 and Val43, in addition Asn25 is involved in the hydrogen bond network. Hydrogen bonds are shown as black dashed lines between donor and acceptor atoms with distances ≤3.4 Å. (B) The same structure of the cyt b ₆ f complex from Nostoc sp. PCC 7120 superimposed with the structure of the cyt b ₆ f complex from Chlamydomonas reinhardtii [PDB 1Q90, resolution: 3.1 Å (Stroebel et al. 2003)]. The figure was produced with pyMOL (The PyMOL Molecular Graphics System, Version 1.7 Schrödinger, LLC).

The evolutionary primacy of the CxGG motif for the cyt bc complexes follows also from the conservation pattern of the heme b stabilizing amino acid motifs. Berry and Walker (2008) noted that in the four‐helical cyts b of succinate dehydrogenase, fumarate reductase and formate dehydrogenase, the aforementioned conserved, heme‐accommodating glycine residues in helices 1 and 3 are usually preceded by serines or threonines. In the available structures, these serine/threonine residues stabilize the hemes by making hydrogen bonds with the heme‐ligating histidines in the corresponding helices 2 and 4 (Berry and Walker 2008). The multiple alignment in Fig. S2, in agreement with the earlier alignments in Berry and Walker (2008) shows that the number of [T/S]G motifs varies from one to three in different lineages of cyt bc complexes, with the heme b _p‐stabilizing [T/S]G motifs being more conserved than those next to the heme b _n. Cyanobacteria and other organisms with the cytochrome c _n have the CxGG motif instead of the heme b _n‐stabilizing [T/S]G motif in helix 1. In Figs 2, S1 and S2, among 216 cyt b sequences without the CxGG motif, only four sequences have the TG motif in the position that corresponds to this CxGG motif. We consider this to be an additional indication that the common ancestor of cyt bc complexes already contained the CxGG motif and the heme c _n. Indeed, if the acquisition of the CxGG motif were a late event, then one would expect [T/S]G motifs to be as abundant in the corresponding position of helix 1 of the heme c _n‐lacking, large cyts b, as it happens with the [T/S]G motifs in the three other positions (Fig. S2). As this is not the case, we suggest that (1) the ancestral cyt bc complex already contained interacting hemes b _n and c _n and, accordingly, the unique CxGG motif (instead of the typical [T/S]G motif) in helix 1, and (2) the loss of the heme c _n correlated with the loss of the CxGG motif in different lineages.

In the separate archaeal clade that includes Ta1228 of Thermoplasma acidophilum DSM 1728, and proteins from Thermoplasma volcanium GSS1, Caldivirga maquilingensis IC‐167 and Picrophilus torridus DSM 9790, as well as a single bacterium P. staleyi DSM 6068 (Figs 2 and S2), the TG motif is present in helix 1 in the position of the CxGG motif. In all these cases, however, the [T/S]G motif is absent from the apposing helix 3 of cyt b, which makes these TG motifs the only ones to hold the heme‐coordinating histidine residues. We assume that in these few cases the [T/S]G motifs in helix 1 have been re‐acquired owing to the evolutionary pressure for stabilization of heme b _n.

Evolutionary clues to the Q‐cycling

The presence of the third heme in the primordial cyt b would be of crucial importance for the establishment of the Q cycle mechanism (Mitchell 1975b, 1976), see Fig. 5A. The Q‐cycle, which was shown to operate in the best studied cyt bc ₁ complexes of mitochondria/proteobacteria and cyt b ₆ f complexes of green plants, couples oxidation of membrane quinols to the generation of the proton gradient (see Crofts and Wraight 1983, Hauska et al. 1983, Berry et al. 2000, Mulkidjanian 2005, 2007, 2010, and Cramer et al. 2011, for reviews). In the Q‐cycle, a quinol molecule is oxidized in the catalytic center P that is shaped by the aforementioned highly conserved P[D/E]W[F/Y] motif of the p‐side loop of the cyt b (or of the subunit IV in case of the cyt b ₆ f‐type complexes). Upon quinol oxidation, one electron is accepted by the FeS cluster of the Rieske protein to be transferred to cyt c ₁ (or cyt f) and then to leave the complex, whereas the other electron, via heme b _p and heme b _n, crosses the lipid bilayer to reduce the quinone molecule in the catalytic center N close to the opposite side of the membrane. The oxidation of a quinol molecule in center P causes release of protons to the p‐side of the membrane, whereas the ultimate formation of a quinol molecule in center N results in binding of protons from the n‐side of the membrane.

Figure 5.

Q‐cycle mechanism. (A) The original Q‐cycle scheme of Peter Mitchell. The figure, modified from Mulkidjanian (2010), is based on Fig. 1 from Mitchell (1975a) and Fig. 3 from Mitchell (1976). The symbol (o) corresponds to center P, the symbol (i) marks center N. The additional donor of the second electron to center (i) is denoted ‘d’. Otherwise, the original Mitchell's notation was changed to the currently used one (see the text for details). (B) An ancestral Q‐cycle mechanism for the cyt b ₆ f‐type complexes with small, three‐heme cyt b. Cyanobacterial cyt b ₆ f complex is used as a model for the cyt b ₆ f‐type complex of anoxic bacteria, see the text for details. Cytochrome b ₆ is in green, subunit IV is in cyan, Rieske protein is in yellow and cytochrome f is in pink. Thick arrows, electron transfer steps upon quinol oxidation; red arrows, proton transfer steps. The dashed thin arrows indicate the fast electron equilibration between all six hemes of the two cytochromes b. The purple arrow indicates the transfer of the second electron to the SQ_N ^•− semiquinone from a pre‐reduced heme c _n. The heme c _n should be constantly reduced in the case of menaquinone‐oxidizing cyt b ₆ f‐type complexes such as those of Heliobacterium modesticaldum and Geobacillus stearothermophilus (Bergdoll et al. 2016), see the text for further details. The figure was produced with pyMOL (The PyMOL Molecular Graphics System, Version 1.7 Schrödinger, LLC).

Still, formation of a quinol in the center N requires not one, but two electrons. Peter Mitchell has realized this problem and suggested that the second electron is coming to the center N from a separate enzyme donor ‘d’ [e.g. the succinate dehydrogenase was suggested to perform this function in mitochondria (Mitchell 1975b, 1976), see Fig. 5A]. However, the suggested involvement of a separate enzyme as a second electron donor for the center N came in contradiction with the apparent ability of purified, liposome‐incorporated cyt bc ₁ complexes to pump – on their own – additional protons (Leung and Hinkle 1975). Accordingly, a scheme of a ‘modified Q cycle’ was put forward where a semiquinone Q_N ^•− was suggested to form in the center N after the first turnover of the center P, whereas the oxidation of the next quinol molecule in the center P resulted in the formation of a Q_NH₂ quinol (Garland et al. 1975, Crofts et al. 1983). According to the modified Q‐cycle mechanism, one quinol molecule Q_NH₂ gets formed in center(s) N per each two molecules of substrate quinol Q_PH₂ oxidized in center(s) P.

However, the stable semiquinone intermediate in center N, while found in cyt bc ₁ complexes of mitochondria and phototrophic α‐proteobacteria (Siedow et al. 1978, de Vries et al. 1980, Ohnishi and Trumpower 1980, de la Rosa and Palmer 1983), has not been detected in the plant cyt b ₆ f complexes. This is not totally surprising, because semiquinones, generally, are extremely unstable in the non‐polar membrane environment (Cramer and Knaff 1990). Therefore, in the cyt b ₆ f complex, it is believed that the first electron is stored on the c _n heme until the second quinol molecule is oxidized in center P and the second electron is delivered to heme b _n, after which the formation of a quinol proceeds in a concerted, two‐electron fashion (Cramer et al. 2006, 2011, Baymann et al. 2007, Cramer and Hasan 2016), see also Fig. 5B. It was suggested that, under conditions of physiological turnover, the c _n hemes in centers N can promptly obtain electrons from the redox carriers involved in the cyclic electron flow, such as ferredoxins (Kurisu et al. 2003, Stroebel et al. 2003, Cramer et al. 2004, Mulkidjanian 2010).

Investigation of the kinetics of the cyt bc complexes in preparations from phototrophic organisms showed that the flash‐induced turnover rate of pre‐oxidized enzymes was much slower than the rate measured under reducing conditions or during their steady‐state operation, both in cyt bc ₁ complexes of phototrophic bacteria (Takamiya et al. 1979, Bowyer and Crofts 1981, Gopta et al. 1998, Klishin et al., 2002) and in plant cyt b ₆ f complexes (Rich et al. 1992, Heimann et al. 1998). We have suggested that redox‐centers of centers N can get pre‐reduced under physiological conditions. Specifically, the electron carriers in center N, both in cyt b ₆ f and cyt bc ₁ complexes, appear to obtain electrons from the membrane quinol pool. Indeed, the structures of cyt bc complexes show centers N more accessible from the lipid bilayer than centers P (Xia et al. 1997, Kurisu et al. 2003, Stroebel et al. 2003). In addition, kinetic data for cyt bc ₁ complexes of phototrophic bacteria indicated that ubiquinol molecules, as formed in the photochemical reaction center, interacted with center(s) N faster than with center(s) P (Mulkidjanian et al. 1990, Mulkidjanian 2007). Then, when the electron carriers in center N are pre‐reduced (activated), oxidation of each quinol molecule in center P would lead to the formation of a quinol molecule in center N [see Mulkidjanian (2007, 2010) for the further details on the activated Q‐cycle mechanism].

In the case of plant cyt b ₆ f complexes, where electrons could be transiently stored on hemes c _n, the ‘activated’ state of the enzyme seems to decay in the darkness (Rich et al. 1992, Heimann et al. 1998), most likely because the midpoint redox potential of heme c_n is too low for it to stay reduced in an oxic environment. This feature may explain the initial slowness of the Q‐cycle in dark‐adapted, oxidized plant cyt b ₆ f complexes (Rich et al. 1992). Recently the midpoint redox potentials of the three hemes of small cyts b were estimated in the cyt b ₆ f‐like complexes of the menaquinone‐containing firmicutes, an anaerobe H. modesticaldum (subfamily B in Figs 2 and S1) and the microaerophile Geobacillus stearothermophilus (subfamily C in Figs 2 and S1) (Bergdoll et al. 2016). As in the case of cyt b ₆ f complexes of oxygenic organisms (Zhang et al. 2004, Alric et al. 2005, Zatsman et al. 2006, de Lacroix de Lavalette et al. 2009), the midpoint redox potential of heme c _n was found to be distinctly higher than that of the two b hemes. The reported midpoint redox potential of hemes c _n in H. modesticaldum and G. stearothermophilus (approximately −50 mV at pH 7.0) implies that, under anoxic conditions, these c _n hemes would stay constantly reduced owing to the electron exchange with the membrane menaquinol pool (which has midpoint redox potential of approximately −100 mV at pH 7.0). Oxidation of a menaquinol molecule via one of the centers N would lead to the eventual reduction of the both c _n hemes of the dimer as they have the highest midpoint potential; the electron exchange between all the six hemes of two cyts b in a cyt b ₆ f‐type complex is expected to proceed within hundreds of microseconds (Xia et al. 1997, Gopta et al. 1998, Mulkidjanian 2007, Swierczek et al. 2010, Cramer et al. 2011, Sarewicz et al. 2016). Such a pre‐reduction of two c _n hemes by the membrane menaquinol would correspond to a backward turnover of the Q‐cycle. A pre‐loaded (by two electrons) cyt b ₆ f‐type complex could then support two productive sequential turnovers yielding two quinol molecules (in each of its two centers N), see Fig. 5B. The total balance would be the same as in the ‘textbook’ Q‐cycle models, with two quinol molecules oxidized in centers P and one quinol molecule (in sum) formed in center(s) N. One of the advantages of the suggested back‐and‐forth Q‐cycling in cyt b ₆ f complexes is the lack of the need to stabilize the semiquinone in center N, which is consistent with the inability to detect a semiquinone signal in cyt b ₆ f complexes. Hence, the redox properties of the three hemes of cyt b ₆ f‐like complexes in menaquinol‐containing anaerobic bacteria, as determined by Bergdoll et al. (2016), appear to be tailor‐made for the Q‐cycle mechanism, which suggests that this mechanism may have emerged in cyt b ₆ f‐like complexes of anaerobes that contained ancient small, three‐heme cyts b.

Gradual oxygenation of the atmosphere after the Great Oxygenation Event (GOE, Lyons et al. 2014) apparently led to the replacement of menaquinone by the high‐potential plastoquinone in oxygen‐generating cyanobacteria and to an increase in the midpoint potentials of the redox components involved in their cyt b ₆ f complexes (Baymann and Nitschke 2010, Dibrova et al. 2013, Bergdoll et al. 2016). Therefore, in cyt b ₆ f complexes of chloroplasts, c _n hemes seem to get oxidized in the dark‐adapted, resting state, so that the ‘initiating’, slow enzyme turnovers appear to be needed to reduce the c _n hemes (Mulkidjanian 2010).

It is noteworthy that the heme c _n performs exactly the function that was attributed by Peter Mitchell to the hypothetical component ‘d’: it provides the second electron to accomplish the quinol formation in center N in one shot (Fig. 5A, B).

Emergence of cyt bc complexes in the context of photosynthesis

Thermodynamically, the Q‐cycle is driven by the exergonic transfer of the ‘first’ electron from the quinol in center P – via the mobile FeS domain – to further high‐potential electron acceptors, which are different in different subfamilies (Dibrova et al. 2013, ten Brink et al. 2013, Schmetterer 2016). Accordingly, the availability of high‐potential electron acceptors was a pre‐condition for the emergence of the Q‐cycle. Before the GOE some 2.4 billion years ago (Lyons et al. 2014), the Earth environments were highly reduced. The extent of the environmental redox poise could be inferred from the highly reduced state of the cytoplasm of modern cells of approximately −300 mV (Wald 1964, Mulkidjanian and Galperin 2007). This redox poise corresponds to the redox buffering capacity of such naturally abundant compounds as H₂, H₂S and SO₂ (Jelen et al. 2016), which were being produced by outgassing of the earth mantle throughout the entire geological history. Oxidizing chemical species could sporadically form in the primordial atmosphere (e.g. NO could be generated by lightning). Still, redox‐active sulfur species in the atmosphere and hydrosphere, such as H₂S and SO₂, would have efficiently quenched any oxidized species and prevented their interactions with primordial organisms. Accordingly, in the reduced environments, high‐potential electron acceptors could be constantly generated only by solar light. In photoactive materials, absorption of a light quantum may lead to a separation of electric charges and formation of electron vacancies (holes). Formation of such vacancies is independent of the redox poise of the environment; it may have continuously proceeded in the vicinity of living organisms. Indeed, as long as solar light, in all likelihood, was involved in prebiotic synthesis and selection of first biopolymers (Mulkidjanian et al. 2003, Powner et al. 2009, Sutherland 2016), the first life is believed to form in the vicinity of terrestrial geothermal systems which could deliver building blocks, such as organic molecules, as well as phosphorous and nitrogen compounds (Florovskaya 1978, Deamer et al. 2006, Ricardo and Szostak 2009, Mulkidjanian et al. 2012, Sponer et al. 2016, Sutherland 2016). Many mineral compounds deposited at the sites of geothermal activity, including the sulfides of zinc, manganese and cadmium, are photoactive (Schoonen et al. 2004). Such minerals are even capable of light‐driven generation of organic molecules with a high quantum yield, up to 80% in the case of production of formic acid by illuminated crystals of zinc sulfide (Henglein 1984, Schoonen et al. 2004, Guzman and Martin 2009). The primordial cells could, in principle, benefit from products of such an abiotic photosynthesis. Furthermore, many bacteria can deposit potentially photosynthesizing metal sulfide nanoparticles on their surface (Labrenz et al. 2000). Recently it was reported that biologically precipitated cadmium sulfide nanoparticles enabled the photosynthesis of acetic acid from carbon dioxide in a non‐photosynthetic bacterium Moorella thermoacetica (Sakimoto et al. 2016). However, while inorganic photosynthesis could provide primordial organisms with food, those organisms were unlikely to utilize the electron vacancies generated within mineral crystals. In case of sulfides, the light‐generated vacancies are promptly refilled by sulfide anions yielding elemental sulfur (see Schoonen et al. 2004, Mulkidjanian 2009, Mulkidjanian and Galperin 2009, and references therein).

Enzymes, however, could have had access to electron vacancies generated upon (bacterio)chlorophyll (BCl) based photosynthesis. As argued elsewhere (Mulkidjanian and Galperin 2009), BCl‐based photosynthesis may have developed from cellular ultraviolet (UV)‐protecting systems to replace abiotic photosynthesis in those organisms that moved away from metal‐rich geothermal settings, e.g. while invading the ocean. Upon BCl‐mediated photosynthesis, electron vacancies would be generated within biological membranes and would be accessible to membrane enzymes, specifically to the potential ancestors of cyt bc complexes.

Tice and Lowe (2004) have identified the remnants of phototrophic microbial communities in the 3.4‐Gyr‐old Buck Reef Chert, a 250‐ to 400‐m thick rock running along the South African coast. The BCl‐based photosynthesis, according to the current views, emerged within the bacterial lineage, after its separation from archaea (Hohmann‐Marriott and Blankenship 2011). Bacterial photosynthesis involves separation of charges at the so‐called ‘special’ pair of BCl molecules within the membrane‐embedded photochemical reaction center (PRC) [see Crofts and Wraight (1983) for a review]. As a result of such separation, an electron leaves the special pair to reduce low‐potential acceptors, e.g. iron‐sulfur proteins, as in photosystem I (PSI), at the opposite, n‐side of the membrane, whereas a high‐potential electron vacancy (hole) remains at the BCl moiety close to the p‐side of the membrane. In photosynthetic organisms, the vacancies are refilled by high‐potential electron carriers that, in turn, can be reduced by the Rieske proteins of respective cyt bc complexes. In the reduced primordial biosphere, there must have been an abundance of potential electron donors for the oxidized BCl molecules. Still, coupling of a PRC to a cyt bc complex capable of Q‐cycling had a unique advantage of generating additional membrane voltage for the ATP synthesis by utilizing the light‐generated redox gap between the membrane quinols and the light‐oxidized redox components. Per one charge translocated, as a result of the charge separation within a PRC, one more charge could be translocated across the membrane by a cyt bc complex (Crofts and Wraight 1983). Therefore, to outcompete other potential electron donors to the PRC, the ancestral cyt bc complexes should have been in direct vicinity of the PRC. In most modern phototrophic organisms, the cyt bc complexes are structurally coupled with PRCs (Hauska et al. 1983, Mulkidjanian et al. 2006, Nitschke et al. 2010), which guaranties that cyt bc complexes get the ‘first hand’ in filling the electron vacancy in the PRCs. In 2001, the ferredoxin:NADPH⁺ reductase, which accepts electrons from PSI, was purified together with the cyanobacterial cyt b ₆ f complex and was shown to be functionally coupled to it (Zhang et al. 2001). More recently, the isolation of ‘supersupercomplexes’ comprised of PSI and the cyt b ₆ f complex from C. reinhardtii was reported (Iwai et al. 2010, Minagawa 2016). Formation of such supercomplexes is also likely in diverse, still poorly studied prokaryotes that carry the genes for PRCs and cyt bc complexes within the same, apparently mobile, operons (Mulkidjanian et al. 2006, Nitschke et al. 2010). It is tempting to speculate that the Chlamydomonas PSI‐cyt b ₆ f supercomplex (Iwai et al. 2010, Minagawa 2016) is a descendant of ancient supercomplexes of PRCs with their cognate cyt bc complexes, which could resemble the enzymes of modern anaerobic phototrophic bacteria, such as H. modesticaldum (Nitschke et al. 2010, Sarrou et al. 2012, Kondo et al. 2015, Bergdoll et al. 2016).

The suggested emergence of the first b ₆ f‐like complexes within phototrophic membranes, as it has been already noted (Dibrova et al. 2013), also rationalizes the recruitment of a chlorophyll molecule and a carotenoid molecule as structural elements of the cyt b ₆ f complexes (Kurisu et al. 2003, Stroebel et al. 2003).

As long as more and more high‐potential electron acceptors became available with oxygenation of the atmosphere, the ability of cyt bc complexes to couple reduction of such acceptors with generation of proton gradient would become increasingly beneficial. The genes of the cyt bc complexes would then spread via the LGT to non‐phototrophic organisms. However, the functionality of laterally transferred cyt bc complexes in recipient cells would be limited by their ability to insert and covalently bind the c _n heme. The incorporation of the c _n heme in Chlamydomonas, plants and cyanobacteria requires a dedicated maturation system that operates from the n‐side of the membrane (Kuras et al. 2007, de Vitry 2011, Cline et al. 2016). This system is absent in other prokaryotic groups that contain heme c _n; which systems are responsible for the insertion and covalent binding of heme c _n in these organisms remains a mystery. On one hand, conservation of the CxGG motif indicates that heme c _n, in addition to the well‐studied Firmicutes (Yu and Le Brun 1998), appears to be inserted in the cyt bc complexes of members of the Planctomycetes and several other phyla represented in subfamilies G and L. On the other hand, the inability to insert and covalently bind heme c _n into the cyt bc complexes acquired via LGT may have been responsible for the loss of heme c _n in many lineages (Figs 2 and S1). The consequences of losing heme c _n could be dramatic, as this heme is involved in linking the small cyt b with its cognate subunit IV, as shown in Fig. 4. The loss of the c _n heme, therefore, would impose evolutionary pressure to link these two proteins via gene fusion, which might explain the existence of several independent subfamilies with linked, long cyt b sequences (Figs 2 and S1). In such cases, different linkers were used in different subfamilies (Dibrova et al. 2013) and the CxGG motif would get lost, although to a different extent in different subfamilies of bc, as documented in Figs 2 and S1.

A loss of the c _n heme would also have dramatic consequences for the Q‐cycle operation. In the cyt bc ₁ complexes of proteobacteria and, accordingly, of mitochondria, the absence of heme c _n is compensated by the ability of the enzyme to stabilize the SQ_N ^•− semiquinone to a high extent. The midpoint redox potential of the Q_N/SQ_N ^•− couple was estimated to be as high as +150 mV (Robertson et al. 1984, Berry et al. 2000), which provides sufficient driving force for the Q cycle and, in all likelihood, accounts for the constant presence of one stable, Electronic Paramagnetic Resonance (EPR)‐silent SQ_N ^•− semiquinone per a dimeric cyt bc ₁ complex (see Siedow et al. 1978, de la Rosa and Palmer 1983 and Mulkidjanian 2007, for details). Structural analysis of cyt bc ₁ complexes shows that the ubiquinone molecule is bound in the place of heme c _n, and not where the plastoquinone molecule is bound in the cyt b ₆ f complex (Stroebel et al. 2003, Cramer and Hasan 2016). Apparently, the loss of heme c _n in the ancestors of cyt bc ₁ complexes enabled moving the Q_N quinone closer to heme b _n; the large, partially water‐filled cavity, which was left from the heme c _n, facilitated stabilization of the polar SQ_N ^•− state, which made the Q‐cycling possible even in the absence of heme c _n (see Mulkidjanian 2007, 2010, Dibrova et al. 2013). For other subfamilies of cyt bc complexes, both with short and long cyt b sequences, no experimental evidence of Q‐cycling is available. The respective cyt bc complexes could be involved in ET not coupled to additional voltage generation.

Archaeal cyt bc complexes: endowment of the LUCA or a friendly gift from phototrophic bacteria?

Noteworthy, the idea of emergence of cyt bc complexes in the context of photosynthesis was discussed by Hauska more than 30 years ago (Hauska et al. 1983). Later, however, this idea has been dropped in favor of the view that this enzyme had been ancient and already present in the LUCA (Castresana et al. 1995, Schutz et al. 2000, Baymann et al. 2003, 2012, Nitschke et al. 2010, ten Brink et al. 2013). This view had been based on the finding of long cyts b in some archaea and the belief, now largely abandoned, that the presence of homologous genes in some archaea and some bacteria was an evidence of their presence in the LUCA (Castresana et al. 1995). As a result of this belief, many enzymes found in both archaea and bacteria had been initially attributed to the LUCA, which boosted the gene content of the LUCA beyond all possible limits. Thorough analysis of the first complete genomes, however, showed that most metabolic genes of archaea are of bacterial origin and had been acquired via the LGT (Koonin et al. 1997, Makarova et al. 1999). In fact, only about 70 ubiquitous genes are present in all free‐living organisms; these genes exist in clear‐cut bacterial and archaeal versions and therefore can be assumed to be present in the LUCA (Koonin 2000, 2003). No genes of cyt bc complexes are in this common set. For all other enzymes (and their genes), no straightforward attribution to the LUCA is possible based solely on their distribution; each time, case‐by‐case phylogenomic and structural analyses are needed.

Still, in spite of the apparent propensity of cyt bc complexes for the LGT and diverse evidence for their bacterial origin, presented in our previous work (Dibrova et al. 2013), Kao and Hunte (2014) recently again invoked the possibility of the emergence of the large cyt b in the LUCA. The authors argued that, in their view, ‘the archaeal clade of the cyt b tree contains no bacterial homologs with exception of few haloarchaeal sequences, which cluster with cyt b from Deinococcus–Thermus and Actinobacteria’ and apparently resulted from bacteria via horizontal gene transfer. The authors concluded that archaeal cyt b genes (with the exception of haloarchaeal ones) did not have bacterial origin and that cyt bc complexes with a large cyt b were present in the LUCA. As follows from Figs 2 and S1, haloarchaeal short cyt b sequences (subfamily N) have a sister relationship to the subfamily F with long cyts b. The archaeal subfamily E2 also contains bacterial sequences, which indicates either that cyt b genes can easily travel between bacteria and archaea or that this whole subfamily is prone to the long‐branch attraction, as argued previously (Dibrova et al. 2013).

As an additional argument, Table S1 shows the distribution of some respiratory enzymes in the available archaeal genomes. The presence of cyt bc complexes strongly correlates with the ability of their hosts to respire by reducing diverse high‐potential acceptors and also, in many cases, with the presence of terminal oxidases. There is also a less trivial correlation with the presence of enzymes that appear to participate in fatty acid metabolism in archaea (Dibrova et al. 2014). The fatty acid tails of bacterial lipids are known to serve as structural elements by filling cavities in many bacterial membrane enzymes (Hunte and Richers 2008). Specifically, several dozen lipid molecules were found to be associated with the cyanobacterial cyt b ₆ f complex (Hasan and Cramer 2014). More bulky isoprenoid tails of archaeal lipids not always could substitute for fatty acid chains, so that fatty acids that are found within archaeal membrane enzymes (Kolbe et al. 2000), most likely, indicate their bacterial origin. The archaeal hosts of such enzymes, while lacking full‐fledged fatty acid synthases, appear to be able to synthesize fatty acids, to stabilize their membrane enzymes of bacterial origin, by combing bacterial‐type enzymes of beta‐oxidation of fatty acids – working in the reverse mode – with archaea‐specific acetyl‐CoA C‐acetyltransferase (Dibrova et al. 2014). Therefore, the presence of these enzymes may serve as an indication of the presence of membrane enzymes of bacterial origin in the particular archaea (Dibrova et al. 2014). As shown in Table S1 in archaea, the presence of the genes that encode the enzymes of fatty acid metabolism strictly correlates with the presence of genes of cyt bc complexes.

Presence of the genes of cyt bc complexes in archaea also correlates with the presence of the genes of copper‐containing cytochrome oxidase subunits (Table S1). Copper‐dependent cytochrome oxidases were unlikely to evolve before the GOE: under anoxic, reducing conditions, there would be no free copper ions available for enzymes (Ochiai 1978).

Hence, Table S1 indicates that those archaea that are challenged by oxidized habitats, unlike those that dwell under anaerobic conditions, possess not just stand‐alone cyt bc complexes, but whole sets of enzymes that support respiration with high‐potential terminal electron acceptors. Bacteria would be natural sources of such enzyme sets. Indeed, the oxygenation of the environment was triggered by the appearance of the oxygen‐producing PSII in some phototrophic bacteria (Mulkidjanian and Junge 1997, Mulkidjanian et al. 2006, Hohmann‐Marriott and Blankenship 2011). Accordingly, these ancestral oxygen‐producing bacteria were the first to be challenged by oxygen. They could only survive by producing enzymes that could deactivate oxygen, preferably, by reducing it to water, as terminal oxidases do. Therefore, the emergence of respiratory machinery, most likely, happened in phototrophic bacterial communities and only later these enzymes could be acquired by archaea via LGT.

In a recent review, Ducluzeau and Nitschke (2016) noted that our previous paper (Dibrova et al. 2013) was the only one to challenge the emergence of the cyt bc complexes before the LUCA, whereas ‘a pre‐LUCA presence repeatedly arrived at over the last two decades (Castresana et al. 1995, Schutz et al. 2000, Lebrun et al. 2006, Ducluzeau et al. 2009, ten Brink et al. 2013) as well as most recently (Kao and Hunte 2014)’. We hope that the current work, which presents additional arguments in favor of the post‐LUCA origin of cyt bc complexes in the context of bacterial photosynthesis, by virtue of being the second article on the subject, would make our case at least two times stronger.

Conclusions

In this work, an analysis of a much larger set of bacterial and archaeal genomes than in the previous one (Dibrova et al. 2013) confirmed our previous conclusions on the origin of the cyt bc complexes within the bacterial lineage, most likely in relation to anoxygenic photosynthesis.

Generally, the updated phylogenetic tree suggests that the traditional categorization of cyt bc complexes into two major groups, namely the cyt b ₆ f‐like and cyt bc ₁‐like complexes, may be an oversimplification. Formally speaking, the subfamilies A to N should have the same phylogenetic ranking. Furthermore, because of the absence of conservation in the linker regions, unification of the cyt bc complexes with long cyts b into a separate group of the ‘cyt bc ₁‐type complexes’ might not be justified. The cyt bc complexes with long cyts b from different subfamilies, on the sequence level, are no more similar to each other than they are to the ‘cyt b ₆ f‐type complexes’ with short cyts b and separate subunits IV.

Further, application of the evolutionary method, along with the structural, biochemical and genetic data, to the study of bioenergetics proved to be a very powerful approach. It allowed highlighting some unresolved questions in the functioning of the membrane energy‐transducing complexes and also provided a road map to a better understanding of the peculiarities of these complexes in various organisms. In addition, this approach demonstrated its (predictive) power, allowing an insight into the mechanisms of enzymes that have not been studied yet.

Author's contributions

D. V. D. and A. Y. M. designed the study; D. V. D. performed the phylogenomic analysis; D. N. S. performed the structural analysis; Y. G. and A. Y. M. analyzed the results and wrote the manuscript. All authors have edited the manuscript and approved it prior to the submission.

Supporting information

Appendix. S1. Selection of additional proteins belonging to the clade D from the BLAST hits of DAMO_0768 protein (accession no. CBE67833) from Candidatus Methylomirabilis oxyfera.

Fig. S1. Phylogenetic tree of selected cytochromes b.

Fig. S2. Selected regions of the multiple sequence alignment of cytochromes b (fused with the subunit IV in case of short cytochromes b ₆, where possible).

Table. S1. Co‐occurrence of COGs of the suggested archaeal fatty acid biosynthesis (Dibrova et al. 2014) with COGs.