Evolution of the cytochrome bd oxygen reductase superfamily and the function of CydAA’ in Archaea

Abstract

Cytochrome bd-type oxygen reductases (cytbd) belong to one of three enzyme superfamilies that catalyze oxygen reduction to water. They are widely distributed in Bacteria and Archaea, but the full extent of their biochemical diversity is unknown. Here we used phylogenomics to identify three families and several subfamilies within the cytbd superfamily. The core architecture shared by all members of the superfamily consists of four transmembrane helices that bind two active site hemes, which are responsible for oxygen reduction. While previously characterized cytochrome bd-type oxygen reductases use quinol as an electron donor to reduce oxygen, sequence analysis shows that only one of the identified families has a conserved quinol binding site. The other families are missing this feature, suggesting that they use an alternative electron donor. Multiple gene duplication events were identified within the superfamily, resulting in significant evolutionary and structural diversity. The CydAA’ cytbd, found exclusively in Archaea, is formed by the co-association of two superfamily paralogs. We heterologously expressed CydAA’ from Caldivirga maquilingensis and demonstrated that it performs oxygen reduction with quinol as an electron donor. Strikingly, CydAA’ is the first isoform of cytbd containing only b-type hemes shown to be active when isolated from membranes, demonstrating that oxygen reductase activity in this superfamily is not dependent on heme d.

Introduction

The predominance of oxygen in our atmosphere determines the bioenergetic importance of oxygen as an electron acceptor and the prevalence of aerobic respiratory chains. There are only three enzyme superfamilies capable of acting as terminal respiratory oxygen reductases—heme-copper oxygen reductases, alternative oxidases and cytochrome bd-type oxygen reductases (cytbd) [1]. While enzymes from this superfamily have been characterized from a number of Bacteria, their role in archaeal respiration has not yet been determined. Archaeal aerobic respiratory chains share some similarities with bacterial respiratory chains, however they often differ in their composition of respiratory enzymes and are adapted to use different co-factors. Complexes that are typically involved in bacterial respiration such as succinate-quinone oxidoreductases [2], cytochrome bc₁ complexes [3], and heme-copper oxygen reductases have been previously characterized in Archaea [4, 5], while NADH:quinone oxidoreductases (complex I) and alternative complex III are either rare or absent from this domain [6]. The presence of cytbd has been noted in archaeal genomes [7], metagenomes and metaproteomes [8–11] but, no functional member of the cytbd superfamily in Archaea has ever been demonstrated.

Cytochrome bd-type oxygen reductase is a respiratory enzyme that converts oxygen to water using three hemes, unlike the heme-copper oxygen reductases that have two hemes and a copper in the active sites [1]. Purified cytbd accepts electrons from quinols using a low-spin heme b₅₅₈ and transfers these electrons to a di-heme active site containing two high-spin hemes. In some of the characterized cytochrome bd enzymes, these active site hemes were shown to be heme b₅₉₅ and heme d, but some other isoforms were shown to contain only hemes b. Those cytochrome bd family members that contain only hemes b are usually referred to as cyanide insensitive oxidases (CIO) or cytochrome bb’–type oxygen reductase, and have been identified in Pseudomonas aeruginosa, Bacillus subtilis and others [12–15]. It is unclear whether the presence of only hemes b is physiologically relevant but it has been suggested that these enzymes are less sensitive to inhibition by cyanide [13]. There is no sequence signature that distinguishes those enzymes in the superfamily that only contain heme b, and no CIO has ever been isolated and characterized.

The canonical cytochrome bd oxygen reductases contain a minimum of two subunits, CydA and CydB, but often have additional “auxiliary” subunits [16–18] such as CydX, a single-transmembrane subunit that is associated with cytochrome bd-I from E. coli that has been implicated in the stability of the enzyme [19]. Cytochrome bd-type oxygen reductases (cytbd) have a high affinity for oxygen [20] and the previously characterized cytbds have been associated with roles in oxygen detoxification, respiratory protection of nitrogenases and as part of sulfide oxidizing respiratory chains [21–25]. Cytochrome bd catalytic turnover generates a proton motive force by translocation of protons using a conserved proton channel from the cytoplasm to the site of oxygen reduction located near the periplasmic side (electrically positive) of the membrane [26]. Yet, cytochrome bd is not as energetically efficient as the heme-copper oxygen reductases that pump protons in addition to translocating “chemical” protons from the cytoplasm to the periplasmic active site [27, 28]. Expression of cytochrome bd has often been associated with microoxic conditions where a high-affinity oxygen reductase would be required [29].

In this work, we used phylogenomics to determine the diversity and distribution of this high-affinity oxygen reductase in Archaea and Bacteria. We determined that there are three distinct families of cytbd—one of which contains the quinol binding characteristics present in the structures of cytbd from Escherichia coli and Geobacillus thermodenitrificans [30–32] and two which do not—and discussed their evolutionary relationships. The distribution of these families even within Archaea involve significant variation and include the two distinct isoforms CydAB and CydAA’, the latter of which appears to have been created by gene duplication. We evaluate the relative distribution of the CydAA’ and CydAB within the domain Archaea and consider the likely role of CydAA’ variants within their ecological context. In addition, we show that the CydAA’ from Caldivirga maquilingensis is a highly active oxygen reductase with unique biochemical and structural characteristics. This combined phylogenomic and experimental approach has significantly expanded our knowledge of the evolutionary and biochemical diversity within the superfamily, which has important implications for the role of the cytbd superfamily in novel respiratory pathways.

Results and discussion

Diversity of cytochrome bd-type oxygen reductases

The molecular structures of cytbd from Escherichia coli and Geobacillus thermodenitrificans have been determined and showed that cytbd typically has two conserved subunits CydA and CydB, along with a third subunit, CydX or CydS which is a single-transmembrane subunit. The cydX or cydS genes are not well conserved or always found along with cydA and cydB in the genome [30–33]. Of the two main subunits, CydA is better conserved in all known cytbd while CydB is very divergent and is hypothesized to have evolved at faster rates than CydA [34]. The CydA subunit is composed of nine transmembrane helices and contains almost all the conserved amino acids known to be important for catalyzing oxygen reduction and proton translocation, including the ligands for three hemes and the residues forming a proton channel [35, 36]. The first four helices of CydA contain all of the amino acids that form the active site. These include the proton channel and ligands to bind the active site heme b₅₉₅ and heme d (these ligands have only been verified in the isoforms containing hemes b and d, and not the ones containing only hemes b). While H19 is a ligand to heme d (in E. coli cytbd), R9 appears to help stabilize the heme b₅₉₅ [31]. The other five helices (V–IX) form the quinol binding site in the biochemically characterized bd-type oxygen reductases and include the ligands to heme b₅₅₈, the point of entry for electrons from quinols [30–32]. With this structural framework in mind, we performed a sequence analysis of cytochrome bd-type oxygen.

An analysis of 24,706 genomes available in the Genome Taxonomy Database (release89) [37, 38], revealed the presence of 17938 CydA homologs. Of these, 13,018 genomes contained at least one CydA homolog, suggesting that this enzyme family is widely distributed and important (Supplementary Table1). Phylogenomic analysis of CydA homologs revealed 15 clades of CydA that could be distinguished on the basis of unique sequence characteristics. (Supplementary Fig. S1, Supplementary Tables 1–3). Four of these clades contain the features that are considered part of the quinol binding site—for e.g., conserved residues Lys252 and Glu257 (E. coli CydA numbering) while the remaining do not. We inferred that the former four CydA clades were quinol:O₂ oxidoreductases and we named them qOR1, qOR2, qOR3 and qOR4a. While CydA of the families qOR1, qOR2 and qOR3 associate with CydB to form CydAB, CydAA’ is formed by the co-association of two distinct CydA clades, qOR4a and qOR4b. qOR4b does not possess quinol binding site characteristics and is likely the result of a gene duplication event. Sequence analysis of CydA sequences from all 15 CydA clades demonstrated that 2 of the remaining clades are missing quinol binding sites and instead contain a number of heme c binding motifs (CxxCH) (Supplementary fig. S1, Supplementary multiple sequence alignments MSA1, MSA2). We named these enzymes OR-C1a and OR-C1b because of the presence of heme c binding motifs. The ‘a’ and ‘b’ attachment to the names signifies that they likely co-associate to form one enzyme OR-C1, just like qOR4a-CydA and qOR4b-CydA’. Eight of the remaining CydA clades were related and named OR-N1, OR-N2, OR-N3a, OR-N3b, OR-N4a, OR-N4b, OR-N5a and OR-N5b. OR-N is named for Nitrospirota because of predominance of these enzymes in that phylum (Supplementary Figs. S1, S2 Supplementary multiple sequence alignments MSA1, MSA2, MSA3). Their close relationship is also supported by the likely structure of the proteins of which they are a part and their genomic context (Fig. 1). We have attempted to develop a nomenclature for the cytochrome bd-type oxygen reductase family that can be easily expanded upon. We designate three large families of cytbd—qOR, OR-C and OR-N—based on their phylogenetic placement, presence/absence of biochemical signatures such as quinol and heme c binding site features, genomic operon context and taxonomic origin. We have designated subfamilies numerically starting from 1 and attached an ‘a’ or ‘b’ subscript if it is likely that two CydA subfamilies co-associate to form one enzyme. Most of the ‘a’-type subfamilies include the proton channel residues E99 and E107 or one of the two (E. coli numbering) while ‘b’-type subfamilies mostly do not. It appears that ‘a’ and ‘b’-type subfamilies are also the result of multiple independent gene duplication events within this superfamily. We will discuss the unusual number of gene duplication events within the cytbd superfamily and the OR-C and OR-N families later in the text and in Supplementary Material but begin with the quinol oxidizing qOR family. Sequences from this family contain all the amino acids that were previously identified as forming the heme ligands, proton channel, oxygen reduction site and quinol binding site [30–32, 36].

Fig. 1. Families within the cytochrome bd-type oxygen reductase superfamily.

The cytochrome-bd type oxygen reductase superfamily is divided into three families—qOR, OR-C and OR-N, primarily defined by the presence of the quinol binding site in the first, the presence of heme c binding site in the second and the abundance of OR-N enzymes in Nitrospirota. The above schematic represents the various subfamilies within each family that are defined by the phylogenetic clustering shown in Supplementary Fig. S1. The operon context and putative complex arrangement of each CydA-containing enzyme is also shown with a reference protein accession number and source microorganism. The potential gene duplication events are highlighted in yellow. A legend is also provided to mark the related conserved domains in the same colors and redox co-factors such as hemes and iron-sulfur clusters.

Evolution of the quinol-oxidizing cytochrome bd-type oxygen reductases (qOR)

To explore the evolutionary relationship between the qOR1, qOR2, qOR3 and qOR4a subfamilies, which are true orthologs, we generated a maximum likelihood phylogenetic tree using RAxML and IQ-TREE with the OR-C and OR-N family sequences as outgroup (Fig. 2, Supplementary Fig. S3). This topology is also similar to that inferred by the phylogenetic inference software, MrBayes (Supplementary Fig. S4). Sequence features can be identified to distinguish these families and to validate the above identified monophyletic clades as meaningfully distinct; some of which are outlined below while the remaining features are mentioned in Supplementary Table 3. All CydA sequences from the qOR1 subfamily have seven amino acids between the two conserved glutamates (ex. Glu99 and Glu107 such as in Escherichia coli CydA [26, 39]) in the proton channel, while CydA sequences from qOR2, qOR3 and qOR4a typically have six amino acids between the two conserved glutamates (ex. Glu101 and Glu108 in qOR3-subfamily cytbd from Geobacillus thermodenitrificans [30]). This insertion/deletion has been hypothesized to lead to a reversal in the position of hemes from the qOR1-bd in Escherichia coli to the qOR3-bd in Geobacillus thermodenitrificans although further research is required to establish whether the reversal of heme positions is universal (further discussion on the insertion/deletion in the proton channel and the Q-loop is included in Supplementary Material). Sequence features that distinguish the qOR4a-subfamily cytbd are insertions between helices V and VI, as well as insertions in helix VIII (Supplementary alignment MSA1, Supplementary Table 3). Conserved tyrosines (Tyr115 and Tyr117 in Geobacillus thermodenitrificans CydA) are present in qOR2 and qOR3 families but not in the qOR4a-subfamily, which is consistent with the close evolutionary relationship between the qOR2 and qOR3-subfamilies observed in the tree topology. Other conserved sequence features that are unique to each family are listed in Supplementary alignments MSA1 and MSA3 and Supplementary Table 3.

Fig. 2. Phylogeny of quinol-oxidizing cytochrome bd-type oxygen reductases.

Protein sequences of cytbd subunit I, CydA were extracted from a taxonomically diverse set of genomes and metagenomes from IMG, filtered with UCLUST using a percentage identity cut-off of 0.6 and aligned using MUSCLE. The multiple sequence alignment, MSA4 was used to infer a phylogenetic tree using RAxML. The RAxML tree topology was similar to that inferred by IQ-tree. The CydA sequences that do not contain the quinol binding site, from the OR-C and OR-N subfamilies as well as qOR4b, were used as the outgroup. At least four monophyletic clades of typical CydA sequences that contain the O₂- and quinol binding site could be defined—qOR1, qOR2, qOR3 and qOR4a. The long branch within the qOR1 clade comprises a number of cytbd that are highly similar to enzymes from this clade but are missing the proton channel. Subunit I of CydAA’ is from the qOR4a-subfamily while subunit II or CydA’ is from the qOR4b-subfamily.

Comparing the CydA phylogenetic tree and the distribution of cytbd across Archaea and Bacteria provides some insight into the relative age of these families. The qOR1 subfamily, which includes the Escherichia coli enzyme, appears to be the most widely distributed with enzymes in over 60 bacterial phyla, (Supplementary Table 4) but it is only sparsely distributed in Archaea. In fact, there are only a very few representatives in Euryarchaeota and Asgardarchaeota (Fig. 3). It is only widely distributed in Halobacterota, whose oxidative metabolism is expected to have evolved relatively late [40] (Fig. 3). This strongly suggests that the qOR1 subfamily is the oldest of the extant families and that cytbd likely originated in Bacteria. While CydA from the qOR2 subfamily is also fairly well-distributed and found in over 20 bacterial phyla, the qOR3-subfamily enzymes are almost exclusive to the Firmicutes and Firmicutes_I phyla with a few enzymes in Archaea. The qOR4a-subfamily enzymes appear to be specific to the Archaea (Supplementary Tables 2, 4). A close evolutionary relationship between the qOR2 and qOR3-subfamilies is suggested by CydA tree topology and identifiable sequence characteristics. However, other trees we inferred (~2 out of 30 trees) have modelled a closer relationship between the qOR2 and qOR1 subfamilies than between qOR3 and qOR1 (Supplementary Fig. S5). This alternative topology might suggest that qOR2 evolved from qOR1 before qOR3, but a majority of our tree topologies suggest that qOR2 and qOR3 are more closely related to each other than they are to qOR1. Therefore, we do not infer anything about the relative ancestry of qOR2 and qOR3. The qOR1 subfamily has seven amino acids between the conserved glutamates in the proton channel, while the qOR4a, qOR2 and qOR3 subfamilies consistently have six amino acids. Lastly, enzymes from the qOR4a, qOR2 and qOR3 subfamilies are almost completely absent from Proteobacteria. This suggests that the qOR2, qOR3 and qOR4a subfamilies diverged from the qOR1 family, before Proteobacteria diverged from other bacterial phyla. While our dataset and phylogenetic analysis is consistent with the above discussion, it must be noted that many lateral gene transfers have been observed within the cytochrome bd-type oxygen reductase [1] that complicate evolutionary analysis. In addition, while our dataset has a large number of species, it is ~3% of the number of bacterial OTUs in the Global Prokaryotic Census [41]. While the phyla estimated to have the greatest global diversity [41] are well represented in the GTDB, our analysis is still limited by the number of available genomes, especially for rare taxa and this represents an additional constraint on the analysis we have presented here.

Fig. 3. Distribution of cytochrome bd-type oxygen reductases in Archaea.

Concatenated gene alignments were made from the archaeal genomes in GTDB using Anvi’o. A phylogenetic tree was made from the concatenated gene alignments using FastTree. All CydA sequences were extracted from GTDB genomes using BLAST with an e-value of 1e⁻¹. The sequences were then filtered to remove CydA sequences without characteristics of the quinol binding site and then classified using a Hidden Markov Model (HMM)-based classifier trained to identify the families—qOR1, qOR2, qOR3 and qOR4a. CydA sequences from each family were then mapped back to each species, and visualized along with the species tree on the iTOL server. Most phyla of the domain Archaea were distinguished by color and a few classes of the phylum Crenarchaeota were labelled to emphasize the presence of CydAA’. It is clear that CydAA’ is almost exclusive to the order Thermoproteales and Desulfurococcales.

As mentioned above, most CydA subfamilies are widely distributed within Bacteria and Archaea, but the qOR4a-subfamily is unique in having sequences that belong only to Archaea. In addition, the qOR4a-subfamily is unique in having a completely different subunit II (CydA’), while the qOR1-, qOR2- and qOR3- subfamily members appear to have CydB homologs as their subunit II. CydB is either not homologous to CydA’ or is evolutionarily distant. The unique ancestry of the qOR4a-subfamily enzymes that are specific to Archaea raises a question about its distribution within that domain.

Distribution of cytochrome-bd type oxygen reductases in Archaea

To investigate the distribution of cytbd within Archaea and to contextualize the evolution of CydA within archaeal evolution, we mapped the presence of qOR4a, qOR1, qOR2 and qOR3 subfamilies of CydA onto a phylogenetic tree of all Archaea, using a concatenated gene alignment made from the archaeal genomes in GTDB [37] using Anvi’o [42], (Fig. 3). It is clear from this representation that most of the qOR4a-subfamily or CydAA’ belong to the class Thermoprotei within the phylum Crenarchaeota with a few CydAA’ in Nitrosphaeria, Thermoplasmatota and Archaeoglobi. Within the Thermoprotei, almost all members of the order Thermoproteales and family Acidilobaceae contain CydAA’ (Supplementary Table 5).

To place CydAA’ into an ecological context we looked at their environmental distribution (Supplementary Table 6). Microbes containing CydAA’ are largely found in thermal environments, including anoxic sediment from solfataric fields, hot springs and deep-sea vents, suggesting that CydAA’ might only be utilized by thermophiles, such as organisms from the genera Vulcanisaeta, Caldivirga, Thermofilum, Pyrobaculum and Thermocladium [43–47] (Supplementary Table 6). Within Yellowstone National Park (YNP), a number of these genera are found in hypoxic, sulfur/iron-rich ecosystems, although Pyrobaculum and Thermofilum have also been found in more oxygenated environments. It has been suggested that members of the Thermoproteales that are found in aerobic environments and have a heme-copper oxygen reductase are more likely to be using aerobic respiration as their primary energetic pathway [43]. This is consistent with what we observe in Pyrobaculum. Of the 8 Pyrobaculum genomes in the GTDB database, P. aerophilum, P. oguniense and P. caldifontis have a heme-copper oxygen reductase and are capable of respiring oxygen [48, 49], but are missing CydAA’. The three genomes that have CydAA’, but are missing a heme-copper oxygen reductase (P. neutrophilum, P. islandicum and Pyrobaculum sp001189275) are not capable of aerobic respiration and appear to be strict anaerobes [48]. This is suggestive of an adaptation based on oxygen availability in the environment, i.e., those Pyrobaculum which have heme-copper oxygen reductases respire oxygen and conserve energy using the more energetically efficient oxygen reductase [28], while those that have CydAA’ live in environments with less oxygen and use this high-affinity oxygen reductase for oxygen scavenging [20].

Expression of the cydAA’ genes have been demonstrated in the hot springs and sulfur-rich/iron-rich environments within YNP, using RT-PCR [10] and metatranscriptomics (Table 1 [50]). While the hot springs were typically hypoxic and sulfur-rich, the iron oxide mats had higher oxygen concentration at the surface and had <0.3 μM concentration of O₂ within 1 mm. Nitrosphaeria, Acidilobaceae, Thermoproteales and Thermoplasmatota expressed cydAA’ in these environments, however it is not clear whether these microorganisms were exposed to high O₂ concentrations. In fact, the Acidilobaceae are expected to be found in the middle and bottom layers of this mat where O₂ concentrations are lower [10, 51]. All of the above observations are consistent with the presence of cydAA’ in microaerobic and hypoxic environments. The obvious question that needed to be addressed is whether CydAA’ actually functions as an oxygen reductase. This was accomplished by biochemically characterizing the CydAA’ from C. maquilingensis.

Table 1.

cydAA’ is expressed in many environments. ProteinGene expression is estimated based on read counts in metatranscriptomes.

Genome_ID	Genome_Name	Locus_Tag	Reads	Best BLAST Hit NCBI	Organism	IMG Study ID	References for IMG study iD
3300003719	Ferric microbial mat communities from Yellowstone National Park, Wyoming, USA - One Hundred Spring Plain (OSP_B) (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0040881_1063211	246	ESQ23817.1	Acidilobus sp. OSP8	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040881_1110491, Ga0099831_11874701	106 42	AMD31390.1	Acidilobus sp. 7A	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040881_1111231	27	WP_081246098.1	Thermoproteus sp. CP80	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040881_1134581	25	WP_066793619.1	Caldivirga sp. MU80	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040881_1142191, Ga0040881_1205091	2174 12	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040881_1202971	5	KUO89797.1	Caldivirga sp. CIS_19	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040881_1228021, Ga0040881_1244931	8 5	KUO80886.1	Vulcanisaeta sp. JCHS 4	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300003723	Hypersaline microbial mat communities from Yellowstone National Park, Wyoming, USA - Beowulf (BE_B) (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0040873_1002501	4784	AMD31390.1	Acidilobus sp. 7A	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040873_1003761	1921	EQB65596.1	Thermoplasmatales archaeon I-plasma	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040873_1011521	2174	KUO93148.1	Thermocladium sp. ECH_B	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040873_1013971	707	PMP91269.1	Caldisphaera sp.	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040873_1028371	244	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300003709	Hypersaline microbial mat communities from Yellowstone National Park, Wyoming, USA - Grendel (GRN_C) (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0040875_1014081	163	AMD31390.1	Acidilobus sp. 7A	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040875_1034841	11	KUO93148.1	Thermocladium sp. ECH_B	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300007166	Iron oxide microbial mat communities from Yellowstone National Park, Wyoming, USA - BED_top_diel_T=1 metaT (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0099835_1035811	192	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099835_1427011, Ga0099835_1431661	4 3	KUO93148.1	Thermocladium sp. ECH_B	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099835_1456681	2	PMP93815.1	Nitrosphaera sp.	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099835_1510461	17	KUO89797.1	Caldivirga sp. CIS_19	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300007164	Iron oxide microbial mat communities from Yellowstone National Park, Wyoming, USA - BED_top_diel_T=3 metaT (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0099836_1305913	60	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099836_1428691	2	WP_066793619.1	Caldivirga sp. MU80	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099836_1428931, Ga0099836_1472421	3 3	KUO93148.1	Thermocladium sp. ECH_B	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300007168	Iron oxide microbial mat communities from Yellowstone National Park, Wyoming, USA - BED_top_diel_T=7 metaT (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0099838_1606901	3	KUO93148.1	Thermocladium sp. ECH_B	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099838_1754862	40	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300007161	Iron oxide microbial mat communities from Yellowstone National Park, Wyoming, USA - BED_top_diel_T=8 metaT (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0099839_1320821	2	NAZ28310.1	Caldivirga sp.	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099839_1444841, Ga0099839_1444842, Ga0099839_1513781	19 4	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300007574	Iron oxide microbial mat communities from Yellowstone National Park, Wyoming, USA - ECH_B_top_diel_T=1 metaT (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0099840_1194611, Ga0099840_1228901	6	KUO93148.1	Thermocladium sp. ECH_B	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099840_1208591, Ga0099840_1293341 Ga0099840_1307751,	29 27 7	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099840_1212061	16	AMD31390.1	Acidilobus sp. 7A	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300007486	Iron oxide microbial mat communities from Yellowstone National Park, Wyoming, USA - ECH_B_top_diel_T=5 metaT (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0099841_1138611	5	WP_117355195.1	Acidilobus sp. 7A	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099841_1140711, Ga0099841_1141101	23 12	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099841_1256151	7	NAZ28310.1	Caldivirga sp.	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0099841_1293981	15	KUO93148.1	Thermocladium sp. ECH_B	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300007575	Iron oxide microbial mat communities from Yellowstone National Park, Wyoming, USA - ECH_C_top_diel_T=7 metaT (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0099844_1215141, Ga0099844_1308701	10 10	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300021851	Metatranscriptome of extremophilic microbial mat communities from Yellowstone National Park, Wyoming, USA - CONBC_RNA (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0187838_10229991	154	WP_014289867.1	Pyrobaculum ferrireducens	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0187838_11515921	155	KUK06933.1	Archaeoglobus fulgidus	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
3300037625	Metatranscriptome of soil microbial communities from Old Woman Creek estuary, Ohio, United States - Aug_M1_C1_D5_B (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0401895_061319_2_361	–	HGD34321.1	Candidatus Korarchaetoa archaeon	Study ID: Gs0135942	Data used with permission from PI
3300003730	Thermal spring microbial communities from Beowulf Spring, Yellowstone National Park, Wyoming, USA - Beowulf (BE_D) (Metagenome Metatranscriptome) (*) (MER-FS) (assembled)	Ga0040879_1097101	63	NAY81522.1	Thaumarchaeota archaeon	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)
		Ga0040879_1273451	7	NAZ28310.1	Caldivirga sp.	Study ID: Gs0067861	D’haeseleer, P. et al. Stand. Genome Sci. (2017)

Expression of cydAA’ in the environment estimated using publicly available metatranscriptomes. cydA homologs from the metatranscriptomic data available on the IMG website were extracted using a BLASTP cutoff of 1e-5. The short fragments found in metatranscriptomic data were matched with the full corresponding protein sequence based on the best hit in the NCBI database.

Partial purification and spectroscopic characterization of the CydAA’ from C. maquilingensis

The cydAA’ operon from Caldivirga maquilingensis consists of two genes—cydA and cydA’. There are no additional subunits encoded within the operon corresponding to cydX/cydY or cydS, which are associated, respectively, with E. coli cydAB and G. thermodenitrificans cydAB. Homologues of these subunits are not apparent in the C. maquilingensis genome. We cloned the operon into the pET22b vector and expressed it in an Escherichia coli strain in which both bd-I and bd-II were deleted (CBO - C43, ΔcydA ΔappB) [17]. The enzyme, cytochrome bb’ oxygen reductase from Caldivirga was engineered to have numerous different tags—6xHistidine, FLAG, GST and GFP. None of these tags were successful, either because of a poor yield of protein or because of the inability of the affinity-tagged proteins to bind to columns with their corresponding epitopes. A GFP-tagged protein was used to verify the expression of CydAA’ from Caldivirga maquilingensis in E. coli. The presence of the protein could be observed by following the fluorescence of the protein under UV light. Since subunit II was tagged with GFP, it confirms the presence of subunit II in the preparation (Supplementary Fig. S6). In addition, the purified protein was verified by mass spectrometry with many peptides recovered from subunit I. (Supplementary Fig. S7). Gel electrophoresis of a partially purified CydAA’ shows two bands of the sizes expected for CydA and CydA’ (Supplementary Fig. S8).

A UV-visible spectrum of CydAA’ in the reduced-minus-oxidized state reveals the absence of the heme d absorbance peak. The presence of heme b₅₉₅ is also not apparent in the spectrum since the maxima at 595 nm and the Soret peak at 440 nm are also missing. This could indicate that heme b₅₉₅ is low spin in this preparation. Hemes were extracted from CydAA’ of Caldivirga maquilingensis as described previously [52]. Only b-type hemes are present in the enzyme (Fig. 4). This was verified by analyzing the hemes in the protein by LC-MS (data not shown).

Fig. 4. Biochemical characteristics of CydAA’ from Caldivirga maquilingensis.

A. and B. UV-visible spectra of cytochrome bd-type oxygen reductase purified from Escherichia coli and Caldivirga maquilingensis, respectively. C. Pyridine hemochrome spectra of CydAA’ from Caldivirga maquilingensis revealing the absence of heme d in the partially purified enzyme. D. Oxygen reductase activity of CydAA’ from C. maquilingensis shows that it is highly active and cyanide insensitive. It is sensitive to Aurachin C1-10, a quinol binding site inhibitor that also inhibits E. coli cytochrome bd.

CydAA’ from Caldivirga maquilingensis has oxygen reduction activity

The oxygen reduction activity of CydAA’ was tested using a Clark electrode, with reduced coenzyme Q1 (reduced using DTT) as the electron donor. (Table 2, Fig. 4) At 37 °C the specific activity is ~330 e^-/s (/heme b). While this is not as high as the activity of E. coli bd at the same temperature (over 1000 e^-/s), the enzymatic activity is substantial, particularly considering the fact that the source of the enzyme is a thermophilic organism whose growth is optimum at 65 °C. The oxygen reductase activity of CydAA’ was insensitive to the presence of 250 µM KCN, a concentration of cyanide that would completely inhibit the activity of heme-copper oxygen reductases [1]. Since CydAA’ was expressed in the bd-deletion mutant, E. coli strain CBO, the only other potential oxygen reductase in this preparation is bo₃ ubiquinol oxygen reductase [53] so the lack of cyanide sensitivity confirms that our purification protocol has separated the two enzymes. The enzyme is also susceptible to Aurachin AC1-10, a known inhibitor of cytochrome bd at concentrations as low as 250 nM [54]. We did not test for other possible functions for CydAA’ such as catalase activity [55] or peroxidase activity [56].

Table 2.

Oxygen reduction activity of E. coli CydAB and C. maquilingensis CydAA’ in the presence of 350 µM coenzyme Q1 and 5 mM DTT.

Protein	Oxygen reduction activity (e-/s)
Caldivirga maquilingensis cytbb’ (CydAA’)	333 $\pm$ 20
Escherichia coli cytbd (CydAB)	1065 $\pm$ 73

We previously noted that cydAA’ is typically found in organisms that perform sulfur-based chemistry such as sulfur reduction and sulfate reduction (Supplementary Table 6) and use DMSO-reductase like enzymes that contain molybdopterin as a cofactor [10]. Combining the above observation with the demonstrated oxygen reductase activity of CydAA’, it is likely that the role of CydAA’ is to detoxify oxygen to protect oxygen-sensitive enzymes involved in sulfur metabolism. This is similar to its expected role in Desulfovibrio [24] and in the protection of nitrogenases during the process of nitrogen fixation [22]. While its primary role may be that of oxygen detoxification, the presence of conserved residues corresponding to the proton channel in qOR4a-cydA implies that CydAA’ is likely to contribute to the generation of proton motive force, by proton translocation. The extent of its contribution would be limited by the prevailing oxygen concentration.

It is striking that oxygen reduction is conserved in the qOR4a-subfamily despite the replacement of CydB with CydA’ and therefore, it is worth considering the similarities and differences between E. coli CydAB and C. maquilingensis CydAA’.

Structural differences between CydAB and CydAA’ inferred from homology models

To aid in the understanding of differences between CydAA’ and other CydAB, we used multiple sequence alignments (Supplementary Figs. S9, S10) and structural models of CydA and CydA’ from Caldivirga maquilingensis (Supplementary Fig. S11, pdb files are available in supplementary material). The most drastic difference between the E. coli and C. maquilingensis enzymes is the absence of CydB. CydB in E. coli was shown to contain the oxygen diffusion channel [31, 32] and an additional proton channel leading to heme d, bound to subunit I [30–32]. In C. maquilingensis the second subunit is CydA’ is 26% similar to CydA. Only the first two helices are well conserved between these subunits in C. maquilingensis whereas other qOR4a-type CydA and qOR4b-type CydA’, such as in A. fulgidus are similar in the first 4 helices. To substitute for the proton channel that exists in CydB, conserved residues in CydA’ such as T71, T74 and H126 might form a different proton channel. CydA’ probably hosts an oxygen diffusion channel to substitute for the loss of the one in CydB but it is not possible to tell from the sequence alignment or structural model where in the subunit this might be. Interestingly, CydA’ retains H19, which has been implicated as a ligand to heme d and heme b₅₉₅ in E. coli and G. thermodenitrificans cytbd respectively. This suggests that an additional heme might bind to the CydA’ subunit but we cannot verify or refute this from our protein preparation. A number of mutations are observed around the binuclear-active site in subunit I, substituting the polar residues in qOR3-CydA with non-polar residues in qOR4a-CydA (Supplementary Fig. S11) that might affect the midpoint potential of the heme or the proton-coupled electron transfer mechanisms.

Characteristics of the OR-C and OR-N families of the cytochrome bd superfamily

As mentioned earlier, a phylogenomic analysis of CydA homologs revealed two new families, OR-C and OR-N that share the first four helices containing the oxygen reduction site. The CydA subunit of OR-C bd-type oxygen reductases typically has eight transmembrane helices and an extended C-terminal periplasmic portion that binds hemes c (Supplementary Table 7), strongly suggesting that a cytochrome c could be an electron donor to this family. Adjacent to the OR-C1a-type cydA in the genome is OR-C1b, which also has 8 transmembrane helices. The OR-N3a/b, -N4a/b and -N5a/b family CydA typically have 10 helices while the OR-N2 and -N1 have 14 transmembrane helices. OR-N enzymes have been previously noted in Nitrospira [57] and Chloroflexi (N5a/b) [58]. They were recently shown to be expressed in manganese oxidizing autotrophic microorganism, Candidatus Manganitrophus noduliformans (N2/N1) from the phylum Nitrospira [59] and is implicated in oxygen reduction. Greater details on the OR-C and OR-N families, including distribution, alignments and conserved amino acids are found in Supplementary Material. The OR-C and OR-N families are widely distributed in Bacteria. OR-C is present only in Bacteria, while OR-N has very few representatives in Archaea (Supplementary Table 2, Supplementary Table 4).

A phylogenetic tree of all CydA clades suggest that the OR-C and OR-N families are more closely related to qOR4b than the other qOR subfamilies (Fig. 2, Supplementary Fig. S1). There are also conserved sequence features that suggest that the OR-C and OR-N families are more closely related to the qOR3 and qOR4a families than the qOR1 family including the deletion in the proton channel between the conserved glutamates (E101 and E108 like in G. thermodenitrificans cytbd), and the presence of nearby conserved tyrosines (Y123 and Y125 in G. thermodenitrificans cytbd). (Supplementary Table 3). This suggests that the OR-C and OR-N families diverged from either of these two families and evolved after the qOR reductases. The evolutionary analysis within this superfamily is complicated by the high number of independent gene duplication events. It appears that qOR4b, OR-N5b, OR-N3b, OR-N4b and OR-C1b subfamilies were the result of gene duplication events (Supplementary Fig. S1). In fact, OR-3a and OR-3b CydA share 50% amino acid sequence similarity. In addition, the presence of OR-N1-type and OR-N2-type cydA in the same operon in some Bacteria and the extent of similarity between them (up to 40%) suggest that they were part of yet another gene duplication. The importance of gene duplication in protein evolution and functional diversification is well known [60]. The nature this process has taken in the cytbd is interesting—a majority of the CydA paralogs have maintained the architecture associated with oxygen reduction and all of them have maintained the H19 ligand to the active site heme d (as per the E. coli structure). In addition, all the above-mentioned duplication events appear to have resulted in a complex of multiple CydA-like proteins with the possible exception of OR-N1 and OR-N2. OR-N2 is often found in operons without another CydA-like protein (Fig. 1). H19 and heme d are found near the interface of subunit I and subunit II in the CydAB structures. The complete conservation of these features with a change in their interacting partner is suggestive of the process of duplication and interface evolution recently investigated in hemoglobin [61]. Future work on the biochemical and structural characterization of the various cytbd families will help us develop insight into the driving forces behind the evolution of this superfamily. Presently, it is clear that the defining characteristic of the cytbd superfamily is the di-heme oxygen reduction site found in the first four helices of CydA homologs. Our analysis suggests that the bd protein scaffold was diversified multiple times to perform O₂ chemistry in unique environments, possibly to function with different electron donors.

Conclusions

The cytbd superfamily is one of two oxygen reductase superfamilies that are widely distributed in Bacteria and Archaea. In the current work we have demonstrated the large diversity of this superfamily using phylogenomics. In addition, we biochemically characterized the CydAA’ from C. maquilingensis showing that CydAA’ is a robust oxygen reductase. The isolated CydAA’ contained only b hemes and no heme d. Hence, C. maquilingensis CydAA’ is a bb’-type oxygen reductase and is the first enzyme from the qOR4a-subfamily to be purified and demonstrated to be a functional oxygen reductase. Finally, we demonstrate that significant diversification of CydA has occurred with the conserved oxygen reduction site being adapted to multiple functions within various ecological niches.

Materials and methods

Phylogenomic analysis of cytochrome bd sequences

To investigate the diversity of the cytochrome bd oxygen reductase superfamily, we performed a large-scale analysis of CydA protein sequences. We first queried the NCBI and IMG databases with BLASTP using an e-value of 1e-³ to generate a database of CydA sequences that had at least some of the conserved amino acids identified in CydA sequences previously [36]. We then used the database of 13738 sequences, whose diversity is reflected in MSA1, and filtered it with a sequence cut-off of 50% to generate the tree in Fig. 2. The parameters used for the various tree-generating algorithms are detailed below. From this tree, we extracted seven CydA sequences from various families, listed below, to probe the full extent of the cytochrome bd family diversity within the GTDB database.

In order to reconcile the protein phylogeny of cytochrome bd oxygen reductases with species taxonomy, we identified and mapped all cytbd to their respective species within the GTDB database release89 [37]. All CydA protein sequences were extracted from GTDB genomes using BLAST [62] by querying with seven CydA sequences (YP_001729719.1-qOR1, MBA3917267.1-qOR2, Ga0063455_1000025672-qOR3, WP_012186727.1-qOR4b, WP_012186728.1-qOR4a, WP_014321499.1-OR-C, TKB90292.1-OR-N). The resulting 67,521 sequences were filtered for sequences >300 amino acids and for unique accession ids to obtain 18,212 sequences. These were then aligned using muscle [63] using the optional maxiters cut-off of 2. The alignment was visualized on Jalview [64] and sequences were filtered by manually deleting the CydA sequences without characteristics of the quinol binding site or the proton channel, as well as some sequences which did not align well with the rest of the sequences. This filtration step was used to remove 369 sequences but also resulted in the loss of 95 CydA sequences from within qOR1 that appear to have lost the proton channel. We added these 95 sequences back to the database after the HMM-based classification step and labelled them as qOR1*. Of the remaining sequences 274 sequences, 200 were cytochromes c that were homologous to the C-terminal portion of the OR-C1a subfamily. The characterization of the remaining divergent 74 sequences is beyond the scope of this work. The filtered set of 17,843 CydA sequences were then classified using a Hidden Markov Model (HMM)-based classifier trained to identify the families—qOR1, qOR2, qOR3, qOR4a, OR-C and OR-N. The HMMs for those subfamilies and families are available on Github (https://github.com/ranjani-m/cytbd-superfamily). The presence or absence of cytbd in each species was tabulated and is available as Supplementary Table 2. The all Archaea species tree used to analyze the distribution of cytochrome bd oxygen reductases in Archaea was generated using Anvi’o [42]. A multiple sequence alignment was created by extracting all ribosomal proteins from archaeal genomes using the HMM source Archaea_76. This alignment was then used to generate a phylogenetic tree using FastTree as per Anvi’o’s default settings. This tree was annotated using the data available in Supplementary Table 4 on the iTOL server [65].

The protein phylogeny of cytbd sequences was inferred using sequences of cytbd subunit I, CydA. These were extracted from a taxonomically diverse set of genomes and metagenomes from IMG [66], filtered with UCLUST [67] using a percentage identity cut-off of 0.5 and aligned using MUSCLE. The multiple sequence alignment was used to infer a phylogenetic tree using RAxML [68] on the CIPRES Science Gateway [69] with the PROTGAMMA model of rate heterogeneity, DAYHOFF matrix specification and a bootstrap analysis with 100 iterations. IQ-trees [70] were inferred using a Gamma model of rate heterogeneity, DAYHOFF substitution matrix and 1000 ultrafast bootstrap replicates while MrBayes [71] trees were inferred using an equal rate of variation across sites and the Jones substitution matrix. MCMC analysis was run for the default 4 chains and 1,000,000 generations.

Preparation of construct for of cytochrome bd oxidase from Escherichia coli

The genes encoding the bb’ oxygen reductase (Gene Object ID: 641276193-4) from C. maquilengensis were PCR amplified using primers purchased from Integrated DNA Technology. The genes were cloned into pET22b (Invitrogen) using 5′ NdeI and 3′ XhoI cut sites. The vector was engineered to use EGFP, GST or FLAG tags alternatively. The tag was added to subunit II in case of EGFP and FLAG; a tag on both subunit I and II was attempted for the His-tag and GST tag. The expression vector, along with pRARE (Novagen) was then transformed into (CBO ΔcydBΔappC::kan) for protein expression.

Cell growth and protein purification

A single colony was inoculated into 5 ml of LB (yeast extract and tryptone were purchased from Acumedia and NaCl from Sigma-Aldrich) with 100 µg/ml Ampicillin and incubated with shaking at 37 °C. The following day, the 5 ml culture was inoculated in 300 ml LB with 100 µg/ml Ampicillin and grown overnight at 37 °C. On the third day, 10 ml of the secondary culture was inoculated into 24 2 L flasks containing 1 L LB with 100 µg/ml Ampicillin, each. The flasks were incubated at 37 °C while shaking at 200 rpm, until the OD600 of the culture reached 0.6. Gene expression was then induced using 0.5 mM IPTG, the temperature was lowered to 30 °C, and the culture was incubated for 8 h or overnight.

The fully-grown cultures were then pelleted by spinning down at 8000 rpm for 8 min, in 500 ml centrifuge bottles. The harvested cells were then resuspended in 100 mM Tris-HCl, 10 mM MgS04, pH 8 with DNaseI and a protease inhibitor cocktail from Sigma. The cells were then homogenized using a Bamix Homogenizer, and passed through a Microfluidizer cell at 100 psi, three times, to lyse the cells. The soluble fraction of the lysate was then separated from the insoluble by spinning down the lysate at 8000 rpm. Membranes were extracted from the soluble fraction by centrifuging the soluble fraction at 42,000 rpm for 4 h.

Membranes were resuspended in 20 mM Tris, 300 mM NaCl, pH 8 and then solubilized with 1% DDM or 1% SML. The solubilized membranes were spun down at 42,000 rpm for 45 min to remove unsolubilized membranes. The supernatant was stirred with Ni-NTA resin for 1 h and then loaded onto a column. The flow through was shown to contain the CydAA’ because of its poor affinity for the nickel column. The flow through was then diluted in buffer to contain 50 mM salt and then loaded onto a DEAE column equilibriated with 20 mM Tris, pH 8, 0.05% DDM. An elution gradient was run between 0 and 500 mM NaCl and CydAA’ was partially purified from the fraction with higher absorbance at A_412nm, corresponding to the soret peak for heme b and used for assays and spectroscopy. This is similar to the first step for purification of cytbd from Escherichia coli [33]. If the elute from the Ni-NTA resin was subject to two steps of purification, first with the Q-sepharose column with the same gradient as above and then the DEAE-Sepharose, a more pure cytbd preparation was obtained but the yield was drastically reduced.

UV-visible spectroscopy

Spectra of the protein were obtained using an Agilent DW-2000 Spectrophotometer in the UV-visible region. The cuvette used has a pathlength of 1 cm. The oxidized spectrum was taken of the air-oxidized protein. The enzyme was reduced with dithionite to obtain a reduced spectrum.

Collection of pyridine hemochrome spectra and heme analysis

For the wildtype or mutants enzymes, 35 μl of the enzyme solution was mixed with an equal volume of 40% pyridine with 200 mM NaOH. The oxidized spectra was measured in the presence of ferricyanide and the reduced in the presence of dithionite. The values of heme b were calculated according to the matrix suggested in [52]. The concentration of heme d was estimated using the extinction coefficient ε_{(629–670nm)} = 25 mM⁻¹ cm⁻¹.

Measurement of oxygen reductase activity

Oxygen reductase activity was measured using the Mitocell Miniature Respirometer MT200A (Harvard Apparatus). 5 mM DTT and 350 µM Q1 were used as electron donors to measure oxygen reduction by C. maquilingensis CydAA’ and E. coli cytochrome bd. 150–250 µM KCN was used to test the cyanide sensitivity of the enzymes.

Structural modelling of CydAA’ from Caldivirga maquilingensis

Sequences of subunit I from Geobacillus thermodenitrificans and Caldivirga maquilingensis were aligned using a larger alignment comprising many hundreds of bb’ sequences made with the software MUSCLE. This alignment was used as to create a model of subunit I from Caldivirga using the Geobacillus subunit I as a template on the Swiss Model server. A model of subunit II was also created using subunit I as a template. The model was then visualized using VMD 1.9.2beta1 [72].

Compliance with ethical standards

Conflict of interest

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41396-021-01019-4.