Genomic Analysis of Melioribacter roseus, Facultatively Anaerobic Organotrophic Bacterium Representing a Novel Deep Lineage within Bacteriodetes/Chlorobi Group

Melioribacter roseus is a moderately thermophilic facultatively anaerobic organotrophic bacterium representing a novel deep branch within Bacteriodetes/Chlorobi group. To better understand the metabolic capabilities and possible ecological functions of M. roseus and get insights into the evolutionary history of this bacterial lineage, we sequenced the genome of the type strain P3M-2T. A total of 2838 open reading frames was predicted from its 3.30 Mb genome. The whole proteome analysis supported phylum-level classification of M. roseus since most of the predicted proteins had closest matches in Bacteriodetes, Proteobacteria, Chlorobi, Firmicutes and deeply-branching bacterium Caldithrix abyssi, rather than in one particular phylum. Consistent with the ability of the bacterium to grow on complex carbohydrates, the genome analysis revealed more than one hundred glycoside hydrolases, glycoside transferases, polysaccharide lyases and carbohydrate esterases. The reconstructed central metabolism revealed pathways enabling the fermentation of complex organic substrates, as well as their complete oxidation through aerobic and anaerobic respiration. Genes encoding the photosynthetic and nitrogen-fixation machinery of green sulfur bacteria, as well as key enzymes of autotrophic carbon fixation pathways, were not identified. The M. roseus genome supports its affiliation to a novel phylum Ignavibateriae, representing the first step on the evolutionary pathway from heterotrophic ancestors of Bacteriodetes/Chlorobi group towards anaerobic photoautotrophic Chlorobi.


Introduction
Members of the phylum Chlorobi, also known as green sulfur bacteria, are strictly anaerobic obligately photoautotrophic Bacteria, which obtain electrons for anaerobic photosynthesis from sulfide and other reduced sulfur compounds [1;2]. The specific light-harvesting complexes, or chlorosomes, of these bacteria are large structures, consisting of bacteriochlorophylls and carotenoids [3]. Other characteristic properties of Chlorobi are their ability to fix nitrogen via molybdenum-containing nitrogenase [4], and to fix CO 2 via the reverse tricarboxylic acid (TCA) cycle [5;6]. Bacteria from the phylum Chlorobi occupy a narrow environmental niche in anoxic aquatic or terrestrial environments where both sulfide and light occur, such as chemocline regions in stratified lakes [1]. Phylogenetically cultured representatives of Chlorobi phylum form one distinct monophyletic group that shares a common root with the Bacteroidetes [7]. Currently, Bacteroidetes and Chlorobi are recognised as two different phyla, but they are closely related in phylogenetic trees based on 16S rRNA as well as other gene sequences [8;9;10]. Conserved indels in several conserved proteins (FtsK, UvrB and ATP synthase a subunit) also strongly indicate that these two phyla shared a common ancestor distinguished from other bacteria [10;11].
Unlike Chlorobi, bacteria from the Bacteroidetes phylum (previously known as the Cytophaga-Flavobacteria-Bacteroides) are metabolically diverse chemoorganotrophes that are able to grow on a variety of complex biopolymers, such as cellulose, chitin and agar [12;13]. Bacteroidetes inhabit diverse habitats including the oral cavity and gastrointestinal tract of humans, where they represent one of the major components of its microbiome. Some species are parasites or sybmionts of humans, animals, algae and protozoa. The free-living Bacteroidetes inhabit soils, fresh and marine water, sediments, and a number of other mostly mesophilic environments [13].
Recently, two non-photosynthetic bacteria related to the Chlorobi phylum were described, -the Ignavibacterium album strain Mat9-16 T [14] and the Melioribacter roseus strain P3M-2 T [15]. The strains were isolated from microbial mats developing in streams of hydrothermal water at Yumata, Japan, and the Tomsk region of Russia, respectively. Described as moderately thermophilic, facultatively anaerobic heterotrophic bacteria, these two species and related environmental clones form a separate deep branch within the Bacteriodetes/Chlorobi group. Thereby, the novel class Ignavibacteria within the Chlorobi phylum was proposed [14]. Subsequent phylogenetic analysis, as well as chemotaxonomic and physiological studies of M. roseus and I. album suggested that these two organisms may represent a novel phylum Ignavibacteriae [15]. The complete genome sequence of I. album was recently published [16], and the analysis indicated that I. album is capable of organoheterotrophy under both oxic and anoxic conditions.
Here, the complete genome sequence of M. roseus strain P3M-2 T is reported. A whole genome analysis and metabolic reconstruction not only provides insight into the lifestyle of M. roseus, revealing its potential for polysaccharide degradation and catabolic versatility, but also supports the phylum level classification of Ignavibacteriae.

Genome Sequencing
A M. roseus genomic DNA sample was isolated according to Perevalova et al. [17]. The genome was sequenced on a Roche GS FLX genome sequencer using the Titanium protocol for a shotgun genome library using titanium chemistry. The GS FLX run (1/4 plate) resulted in the generation about 114 Mb of sequences with an average read length of 403 bp. The GS FLX reads were assembled into 42 contigs by a GS De Novo Assembler (Roche). The contigs were oriented into scaffolds, and the complete genome sequence was obtained upon the generation and sequencing of appropriate PCR fragments. The assembly of the genome at sites with repeated elements was verified by PCR amplification and sequencing of these regions.

Genome Annotation and Analysis
The rRNA genes were identified with RNAmmer [18]. Transfer RNA (tRNA) genes were located with tRNAscan-SE [19]. Proteincoding genes were identified by the GLIMMER gene finder [20]. Whole-genome annotation and analysis were performed with the AutoFACT annotation tool [21]. Clusters of regularly interspaced repeats (CRISPR) were identified by CRISPR Finder [22]. Signal peptides were predicted with SignalP v. 3.0 [23]. The annotated genome sequence of M. roseus has been deposited in the GenBank database under accession number CP003557.

Proteome Composition and Phylogenetic Analysis
A comparison analysis was performed for the predicted proteins encoded by the M. roseus genome using BLASTP to query against the NCBI non-redundant database. Those proteins that had a significant BLAST hit (e-value: 1e-05) were recorded and divided into two different categories: proteins that had a putative assigned function, and those that matched to hypothetical proteins. Proteins that did not have a significant BLAST hit were designated as species-specific proteins.

Detection of Hydrolytic Activities in vitro
The presence of endoglucanases and their approximate molecular weights were determined by zymography. Culture broths of M. roseus, grown on carboxymethyl cellulose (CMC) under aerobic and anaerobic conditions, were centrifuged at 12,1006g for 15 minutes. The pellets were rinsed twice with 50 mM Tris-HCl pH 7.7 following by centrifugation at 12,1006g for 15 minutes. Extracellular proteins, associated with the outer membrane, were washed from the cell surface, using solutions of 9 M Urea, 1 M NaCl or 1% SDS, which were added to the pellets and incubated for 1 hour at room temperature. On completion, 50 mM Tris-HCl pH 7.7 was added to each wash solution in the volume ratio 9:1. The resulted solutions of membrane-free proteins were applied to the polyacrylamide gel during zymography.
Briefly, 7 ml of 46stock solution of lysis buffer (200 mM Tris-HCl, pH 6.8; 4% SDS; 0.01% bromophenol blue; 40% glycerol) were added to 20 ml of protein solutions. 7.5% polyacrylamide gel containing 0.2% CMC (sodium salt, medium viscosity, Sigma) was prepared and SDS-PAGE was performed without boiling the sample before application to the gel. After electrophoresis, gels were rinsed from SDS in water and incubated in 50 mM Tris-HCl pH 7.7 at 52uC for 1.5 h. Gels were stained using Congo Red solution as follows: 0.1% aqueous solution of Congo Red (Sigma) for 30 minutes (staining), 1 M NaCl 2 times per 30 minutes (destaining) and 5% acetic acid (fixing). White bands reflecting active endoglucanases were observed on the blue background.

Results and Discussion
General Features of the M. roseus Genome Melioribacter roseus has a single circular chromosome of 3,300,414 bp without any extrachromosomal elements. The GC content of the genome is 41%, -an average between the values characteristic for Bacteroidetes and Chlorobi phyla [11]. A single 16S-23S-5S rRNA operon and 45 tRNA genes coding for all of the 20 amino acids were identified. The chromosome displays two clear GC skew transitions that likely correspond to the DNA replication origin and terminus; the likely origin of replication was identified in the 303 bp region between the genes dnaA and dnaN (Fig. S1).
Annotation of the genome sequence revealed 2,838 potential protein-coding genes with an average length of 1,068 bp, covering 92% of the genome, of which 2133 (75%) can be functionally assigned. The functions of the remaining 705 genes could not be predicted from the amino acid sequences; of them, 272 genes are unique to M. roseus with no significant similarity to any known sequences.
The genome contains three CRISPR loci contained 16, 20 and 21 repeats units. The first CRISPR locus is linked to an operon, containing cluster of CRISPR-associated (cas) genes. In particular this operon includes cas1, cas2, and cas3 genes. The latter gene encodes a large protein with helicase and DNAse activities that specify the type I CRISPR-Cas system [27]. The second CRISPR locus is associated with extra copies of genes coding for Cas1 and Cas2 proteins, while the third CRISPR is not linked to cas genes. CRISPR elements are widespread in bacterial genomes, and are considered an ancient antiviral defence system in the microbial world. The spacer regions are supposed to be derived from extrachromosomal elements like viruses [28], but none of the M. roseus spacers matches known plasmids or phages. Regarding the cas genes, the first CRISPR system of M. roseus is related to those found in Firmicutes, particularly in thermophilic cellulosolytic bacteria of the genus Caldicellulosiruptor, while cas genes of the second locus are closely related to those from Bacteriodetes. Therefore, one or both of the M. roseus CRISPR/Cas systems may have been independently horizontally acquired and followed by the loss of duplicated cas genes.
Phylogenetic Placement of Ignavibacteriae as Revealed by Whole Genome Analysis The first cultivated representative of Ignavibacteriae, I. album, was originally placed within the phylum Chlorobi, and was classified as the novel class Ignavibacteria [14]. Subsequent phylogenetic analysis of M. roseus, based on rRNA and RecA phylogenies suggested that M. roseus and I. album may be classified into the novel phylum Ignavibacteriae within Bacteroidetes/Chlorobi group [15].
To help further resolve this phylogenetic placement, a taxonomic distribution analysis was performed [29] by comparing each predicted protein in the M. roseus genome against the NCBI non-redundant database (excluding proteins from I. album). For each M. roseus protein with a match, the taxonomic identity of its top BLASTP hit was determined and the total number of proteins in M. roseus that have their closest match to microbes belonging to a given phylum was counted, as shown in Figure 1. If M. roseus is closely related to bacteria in a particular phylum, it is expected that a majority of M. roseus proteins would have close matches to proteins in bacteria belonging to that phylum. However, this study found that M. roseus proteins best matches ones in several phylumlevel bacterial lineages, primarily Bacteriodetes (30%), Proteobacteria (13%), Chlorobi (12%), Firmicutes (10%), and Caldithrix abyssi (16%). The latter bacterium is a moderately thermophilic anaerobe isolated form a deep-sea hydrothermal chimney, and on 16S rRNA phylogenetic trees this isolate is very distant from any recognised bacterial phyla [30]. These data support, on a wholegenome scale, that M. roseus belongs to a separate phylum. Comparative genome analysis confirmed the close relationship of M. roseus and I. album, since the I. album proteome accounted for 56% of the best BLASTP hits of M. roseus proteins.
In order to resolve the phylogenetic position of Ignavibacteriae, genomic information of M. roseus and I. album was used alongside those from other complete bacterial genomes to perform phylogenetic analysis based on concatenated ribosomal proteins (Fig. 2). The results are consistent with rRNA phylogenies [15] and show that Ignavibacteriae forms a separate lineage branching close to the root of Chlorobi phylum after the divergence of Bacteriodetes and Chlorobi. On the phylogenetic tree of concatenated ribosomal proteins ( Fig. 2 and Fig. S2 for a larger dataset), C. abyssi appeared to be affiliated with the group comprising Fibrobacteres, Bacteriodetes and Chlorobi, which may explain above mentioned high fraction of C. asyssi among the best hits for M. roseus proteins.
Some other particular features support the separation of Ignavibacteria from Chlorobi. The most notable is an absence in the M. roseus genome of genes involved in photosynthesis, including fmoA encoding the Fenna-Matthews-Olson antenna protein [31], a frequently used phylogenetic marker of Chlorobi. The M. roseus genome did not include nif genes, encoding MoFe-nitrogenase and its assembly proteins, which are highly conserved and universally present in all Chlorobi, except for the aerobic photoheterotroph Candidatus 'Thermochlorobacter aerophilum', being the deepest lineage within Chlorobi [32]. The absence of genes for photosynthetic machinery and nitrogen fixation correlates with facultatively anaerobic heterotrophic lifestyle of M. roseus. The ability of M. roseus to grow aerobically correlates with the presence of genes encoding superoxide dismutase and catalase, which are involved in the detoxification of the by-products of oxygen respiration. Another notable feature of M. roseus is its flagellar motility consistent with the presence of genes encoding flagellation proteins, absent in Chlorobi [15].

Polysaccharide Degradation and Utilisation
M. roseus was isolated from a microbial mat, covering the wooden surface of a chute, under the flow of hot water (46uC) coming out of a deep oil exploring well [15]. One of the hallmarks of M. roseus is its ability to efficiently hydrolyse different plant polysaccharides, including microcrystalline cellulose, carboxymethyl cellulose, starch, dextrin, dextran, lichenan, agarose and xylan [15]. A number of low molecular weight carbohydrate substrates (glucose, mannose, rhamnose, arabinose, cellobiose, xylose, trehalose, ribose, maltose, fructose, lactose, and glycogen) can also support its growth [15]. These characteristics indicate that the metabolic versatility of M. roseus is similar to that of Bacteroidetes, which are well-known as active carbohydrate utilisers. This is reflected in M. roseus genome encoding extensive set of extracellular and intracellular enzymes responsible for polysaccharide degradation.
A search for carbohydrate-active enzymes (CAZy) [33] revealed more than 100 genes that were classified into 31 different families of glycoside hydrolases (GHs), carbohydrate esterases (CEs), and polysaccharide lyases (PLs) (Tables S1). A number of predicted carbohydrate-active enzymes contain carbohydrate binding modules (CBMs), facilitating their interaction with polysaccharide substrates (Fig. 3). Most of the CBM-containing enzymes also possess a C-terminal Por sorting domain, which is associated with the sorting of proteins to the outer membrane and the covalent modification in Fibrobacteres and Bacteroidetes. In addition to glycoside hydrolases, the M. roseus genome encodes several dozen glycoside transferases that could be involved in the degradation and synthesis of polysaccharides. Many of the M. roseus CAZymes are predicted to have signal peptides, indicating that these enzymes are targeted outside of the cytoplasm.
One of the most distinct features of M. roseus is its ability to digest microcrystalline and carboxymethyl cellulose in both aerobic and anaerobic conditions. Cellulose utilisation by the most well-studied microorganisms involves the activities of multiple enzymes, including endo-b-1,4-glucanases, cellobiohydrolases (also called exo-b-1,4-glucanases), and b-glucosidases, which act synergistically to convert crystalline cellulose to glucose [34]. Examination of the M. roseus genome identified many genes encoding possible cellulolytic enzymes (Table S1). These include possible endo-beta-1,4-glucanases belonging to glycoside hydrolase families GH5 (Mros0504, 0753, 0960, 2625) and GH9 (Mros0757, 2241, 2626, 2837). The calculated molecular weights of putative endoglucanases (Table 1) are in good agreement with the positions of bands detected by zymography of the cell-bound protein fraction with carboxymethyl cellulose. The results shown in Fig. 4 indicate the presence of two or three enzymes with molecular weights between 55 and 65 kD: Mros0757 and Mros2241 being the likely candidates. Notably, none of the putative cellulase genes contain CBMs of families 2 or 3, which are associated with binding to crystaline cellulose, and only a single GH5 family enzyme (Mros1693), which is predicted to encode endo-beta-1,4-mannanase, contains a CBM6 domain enabling binding to cellooligosaccharides, laminarin, xylo-oligosaccharides and beta-1,4, beta-1,3 mixed glucans [35]. None of the cellulase genes contain dockerin and scaffoldin domains, confirming the absence of cellulosomal structures in M. roseus.
Genes encoding proteins with a similarity to known cellobiohydrolases, which are typically members of the GH6 or GH48 families in bacteria [36;37], were not detected. However, it is known that several cellobiohydrolases belong to families GH5 and GH9, which were found in the M. roseus genome. Beta-glucosidases belonging to the GH3 (Mros0754 and others) and GH1 (Mros0638) families, as well as cellobiose phosphorylases of the GH94 family (Mros1576, 1712 etc.), were also found in the genome.
Most aerobic microorganisms secrete a set of cellulases, containing carbohydrate-binding modules that serve to bind the catalytic domains to insoluble substrates, outside the cell (free cellulase mechanism) while anaerobic microorganisms produce large multi-enzyme complexes on their outer surface (cellulosomes), although not many organisms have been found to use it so far [38]. Several microorganisms possess their own mechanism, like aerobic Cytophaga hutchinsonii and anaerobic Fibrobacter succinogenes, which do not produce free cellulases and cellulosomes [39]. A possible mechanism of C. hutchinsonii cellulose degradation, proposed by Wilson [39], involves the action of the protein complex, located on the surface of the outer membrane, which is able to remove individual cellulose molecules from cellulose fibres and transport them through the outer membrane into the periplasmic space, where they would be cleaved by endoglucanases. Extracellular endocellulases might function by producing cellulose ends for binding to the outer membrane component of the cellulose-degrading machinery. This mechanism resembles one used by Bacteroidetes (sus operon) for starch degradation [40]. A similar mechanism was proposed for F. succinogenes; however it uses different outer membrane proteins for binding, removing and transporting cellulose molecules [39;41]. Taking into account the fact that the majority of M. roseus endoglucanases lacking CBMs have homologs in C. hutchinsonii and F. succinogenes, while the cellulolytic activity was associated with the cell surface (Fig. 4), it is proposed that the cellulolytic mechanism of M. roseus resembles that of C. hutchinsonii and F. succinogenes.
M. roseus possesses genes encoding a wide range of glycoside hydrolases that are active against various polysaccharides used by this organism for growth. The genome encodes GHs for the hydrolysis of xylan (Mros2090, 2091 etc.), mannanan (Mros0511, 1693), lichenan (Mros0976), and agarose (Mros0970). Some of the hemicellulases contain CBM6 domains and CBM4/CBM9 domains binding to amorphous [42] and crystalline [43] cellulose, among other substrates. In particular, GH10 hydrolase Mros2091, containing a CBM4/CBM9 domain and a signal peptide enabling its export from the cell, may be active towards xylan. This gene is a part of a cluster encoding several enzymes involved into xylan metabolism: beta-xylosidases (Mros2087-2088), glucuronoarabinoxylan endo-1,4-beta-xylanase (Mros2090), and alpha-N-arabinofuranosidase (Mros2093). All of the enzymes, except for Mros2087, contain signal peptides indicating their extracellular operation, while Mros2088 also contains a CBM6 domain.
The utilisation of starch may be enabled by several amylases that are sporadically distributed in the genome (Mros0648, 1330, 1936 etc.), as well as by a sus operon (Mros0764-0758), which is common among Bacteroidetes and has been well characterised for Bacteroides thetaiotaomicron [40]. One enzyme, alpha-amylase of GH13 family (Mros0758), was predicted to be secreted out of the cell and carry a CMB48 domain. This CBM is found in a range of enzymes that act on branched substrates: isoamylase, pullulanase and branching enzyme, suggesting that Mros0758 may have broad substrate specificity. Two other enzymes, the GH13 family alphaamylase (Mros0762) and the GH31 family alpha-glucosidase (Mros0763), probably, act in the periplasm for the complete hydrolysis of maltooligosaccharides and maltose. This operon also encodes the outer membrane proteins SusC, SusD and SusE (Mros0761-0759) involved in binding starch on the cell surface and the transport of maltose and maltoolisaccharides across the outer and inner membranes. Another interesting point is that three genes of enzymes possibly involved in cellulose hydrolysis (endobeta-1,4-glucanases Mros0753 and Mros0757, and beta-glucosidase Mros0754) are located immediately downstream of the sus operon. Thus, it is possible that M. roseus may use SusCDE also for cellulose binding and the transport of cellooligosaccarides and cellobiose into the cell. A similar hypothesis was made for the cellulose utilisation mechanism of Cytophaga hutchinsonii [44], which does not grow on starch. Several polysaccharide lyases and carbohydrate esterases were identified, and most of them were predicted to act as pectate lyases (Mros0560 etc.) and pectin methylesterases (Mros0879 and Mros2353). Some enzymes contained a signal peptide and carbohydrate binding modules, indicating their extracellular localisation, although the growth of M. roseus on pectin was not confirmed [15]. The possible explanations might be that (1) M. roseus use pectin esterases to release cellulose and hemicelluloses for their further hydrolysis, as it is known for a highly-specialised cellulolytic bacterium, Fibrobacter succinogenes, [41] which hydrolyses a wide range of polysaccharides while metabolising only cellulose, or/and (2) the synthesis of pectin esterases might be induced by another component, like galacturonic acid or galactose, as shown for Neurospora crassa [45]. One gene encoding PL6, a putative alginate lyase, was also found (Mros1361). Another explanation could be simply that M. roseus is able to grow on particular types of pectin other then one tested in [15].
The genome of M. roseus contains two genes (Mros0583 and Mros0752) that are predicted to encode proteins with CBM4/9 domains, but which have no apparent catalytic activity. The function of these proteins is unclear. Notably, the product of gene Mros0752 contains a signal peptide, suggesting the extracellular operation of this protein.
The M. roseus genome encodes several dozen glycoside transferases, which are involved in the synthesis of intra-and extracellular polysaccharides. Two putative GT5 family glycogen synthases (Mros0053, Mros1119) are necessary for the synthesis of intracellular glycogen and/or extracellular alpha-glucan. GT2 enzymes (Mros1224, 1401, 1750  Analysis of the M. roseus genome reveals the presence of genes encoding type IV pili genes (pilT, pilQ, pilB, pilC). Although a complete set of genes necessary to produce pili is probably not present since genes pilA and pilD were not found, these pilin proteins may be used by M. roseus to adhere to cellulose in a manner similar to how Escherichia coli and other gram negative bacteria adhere to solid substrates [47]. The reliance of M. roseus on membrane-bound adherence proteins may also be correlated to its protein secretion systems. M. roseus appears to encode the Secdependent and type II secretion pathways. In other bacteria such as the plant pathogen Xanthomonas campestris, the Sec-dependent and type II systems are used to secrete hydrolytic enzymes, like cellulases, chitinases, and pectate lyases, in order to facilitate its pathogenic lifestyle [48]. Given that M. roseus requires adherence to cellulose for efficient hydrolysis, this pathway may play a role in cell adherence.

Central Metabolism
M. roseus did not grow photo-or chemoautotrophically and is able to ferment polysaccharides producing H 2 , acetate and CO 2 as the main products of fermentation, although traces of lactate were also detected [15]. The genome of M. roseus contains a complete set of genes encoding enzymes of the Embden-Meyerhof (EM) pathway of glucose metabolism. Pathways for the utilisation of other sugars are linked to the EM pathway (Fig. 5). For example, xylose, a major constituent of hemicellulose, is metabolised by the action of a putative xylose isomerase (Mros2028) and a xylulose kinase (Mros2029), that enter into the non-oxidative branch of the pentose phosphate pathway (PPP). The non-oxidative branch of the PPP uses ribulose phosphate 3-epimerase (Mros2501), ribose-5-phosphate isomerase (Mros0016 and Mros0521), transketolase (Mros0693 and Mros2334), and transaldolase (Mros1726) to produce fructose-6-phosphate and glyceraldehyde-3-phosphate, entering into the EM pathway. Besides the metabolism of pentoses, the non-oxidative PPP allows the production of intermediates necessary for nucleic acid synthesis (Fig. 5).
The pyruvate generated in the EM pathway is oxidised to acetyl-CoA by pyruvate:feredoxin oxidoreductase producing reduced ferredoxin, or by the pyruvate dehydrogenase complex (Mros0859-857), while lactate dehydrogenase may account for the production of lactate. Acetyl-CoA is converted to acetate with the concomitant production of ATP either via a single-step reaction by an acetyl-CoA synthetase (Mros_1085 and Mros_2210) or a twostep reaction by phosphate acetyltransferase (Mros2846) and acetate kinase (Mros0094 and 1926). In the presence of external electron acceptors, acetyl-CoA may be completely oxidised via the TCA cycle, which is encoded by the M. roseus genome (Fig. 5). However, unlike all Chlorobi, including photoheterorophic species 'Ca. T. aerophilum' [32], M. roseus lacks the ATP-dependent citrate lyase gene that makes the reductive TCA cycle incomplete, unless this function is performed by another enzyme. The key enzymes of the Calvin-Benson-Bassham, 3-hydroxypropionate, and Wood-Ljungdahl CO 2 -fixing pathways were also not identified.
In the presence of oxygen or other external electron acceptors, M. roseus was capable of respiratory growth on acetate [15], which is a principal fermentation product in microbial mats [49]. Acetate can be activated to acetyl-CoA which then could be carboxylated by pyruvate:ferredoxin oxidoreductase to form pyruvate, or it could enter the oxidative TCA cycle.
Three hydrogenases of the [FeFe]-family are encoded by the M. roseus genome. The catalytic subunits (Mros0634, Mros2482 and Mros2488) contain the three characteristic sequence signatures of the active site, H-cluster [50;51], and additional Nterminal domain, which accommodate 4Fe-4S clusters. These hydrogenases are similar to heterotrimeric [FeFe] hydrogenase from T. maritima [52]. The catalytic subunits are parts of operons comprising two accessory subunits, encoded by genes Mros0635-636, Mros2481-2480, and Mros2487 (fused). The T. maritima enzyme oxidises NADH and ferredoxin simultaneously in a synergistic fashion to produce H 2 . It represents a class of bifurcating [FeFe] hydrogenases in which the exergonic oxidation of ferredoxin is used to drive the unfavourable oxidation of NADH to produce H 2 [52]. All of the components of M. roseus hydrogenases contain no recognisable signal peptides or transmembrane helices, suggesting their operation in the cytoplasm. During the anaerobic fermentation of sugars by M. roseus in the EM pathway and the subsequent oxidation of pyruvate to acetate, both NADH and reduced ferredoxin are generated. Activity of the bifurcating [FeFe]-family hydrogenases in M. roseus will couple the  oxidation of reduced ferredoxin and NADH to the reduction of protons to H 2 , which is detected as one of the fermentation products.

Respiratory Electron Transfer Chains
The genome of M. roseus encodes all of the major components of the electron transfer chain necessary for energy generation via oxidative phosphorylation, namely, the proton-translocating NADH-dehydrogenase complexes, membrane-bound succinate dehydrogenase (complex II), isoprenoid quinones and the alternative complex III (ACIII) quinol oxidase. The resulting transmembrane proton gradient may be used for ATP generation by the encoded membrane F 0 F 1 -type ATP synthase.
The M. roseus genome contains two main clusters encoding most of the subunits of a type I NADH dehydrogenase, ndhABCD-HIJKLMN (Mros2379-2369), and ndhABCDEFJKLMN (Mros0160-0155, 0138-0135, and 0132). The third locus comprises the genes ndhGHI (Mros2032-2034). The ndh genes from the first and the third clusters are closely related to those found in Chlorobi while close homologs of the subunits encoded by the second cluster were found in various bacterial lineages. Together, the three clusters encode a full set of the 14 ndh genes required for the assembly of NADH:quinone oxidoreductase. However, the absence of genes encoding the NuoEFG subunits that are known to form the dehydrogenase domain involved in NADH binding and oxidation [53] in the first cluster suggest the possibility that it encodes an alternative membrane-bound proton-transporting complex accepting electrons from donor besides NADH (e.g. reduced ferredoxin).
The first NDH cluster is linked to the genes Mros2380-2382, encoding membrane-bound nicotinamide nucleotide transhydrogenase. These enzymes couple the transfer of reducing equivalents between NAD(H) and NADP(H) to the inward translocation of protons across a membrane in the reaction NADH+NAD-P + +H + out = NAD + +NADPH+H + in [54]. The resulting NADPH can be used in biosynthesis reactions. Therefore, this enzyme consumes the proton electrochemical gradient generated by NADH dehydrogenase.
The succinate dehydrogenase complex II is encoded by a threegene operon (Mros0459-457) and consists of an iron-sulfur subunit (Mros0457), a flavoprotein subunit (Mros0458) and a transmembrane cytochrome b subunit (Mros0459). The organisation of this complex is typical for Bacteroidetes and Chlorobi.
In most bacteria, complex III, which is also called cytochrome bc 1 , mediates electron transfer between complex I or II and complex IV by accepting electrons from reduced quinones and reducing cytochrome c. However, inspection of the M. roseus genome revealed no cytochrome bc 1 genes; instead the so-called alternative complex III (ACIII) is encoded. This complex, initially discovered in Rhodothermus marinus [55], is structurally different from the cytochrome bc1 complexes, even though it performs the same function. As in some other organisms [56;57], in M. roseus the genes encoding ACIII (Mros0045-0040) and cytochrome c oxidase genes (Mros0039-0033) are clustered in the genome. The same genomic context has been reported for ACIII of R. marinus, which is able to directly interact with oxygen reductase, but not for ACIII transferring electrons to periplasmic acceptors [56;57]. Thus, it is likely that the ACIII complex is involved in aerobic rather than anaerobic respiratory processes. This assumption is further supported by the fact that a class I soluble c551/c552 cytochrome (Mros0033), encoded upstream of the cytochrome c oxidase, represents a common component of the aerobic respiratory electron transfer chains that shuttle electrons between complex III and terminal oxidases, suggesting that ACIII functions in respiration. The subunits of R. marinus ACIII can be divided in two groups according to the proposed function [57]: membrane attachment and quinone interaction modules (ActCDFG), and electron transfer modules (ActABE). The actA gene (Mros0045) codes for a cytochrome protein containing four haem C binding motifs. The catalytic subunits ActB and ActC are encoded by genes Mros0044 and Mros0043, respectively. The cytochrome C family protein coded by gene actE (Mros0040) contains a single CXXCH motif, and the conserved motif MPA present in other cytochromes [58]. The Mros0042 gene codes for a protein predicted to have three transmembrane helices and a single CXXCH motif in the C-terminal part, and it is thought that the encoded protein is a fusion between ActD and ActE. Finally, the actF gene (Mros0041) codes for another integral membrane protein, predicted to have nine transmembrane helices. A homolog of the R. marinus gene actG was not found in the M. roseus genome.

Rnf Complex
Analysis of the M. roseus genome revealed an operon containing a set of six genes (rnfCDGEAB, Mros1482-1478, 1476) that are related to the potential membrane-bound electron transport complex (Rnf) found in various bacteria. Two other genes, Mros1477, encoding a putative lipoprotein, and Mros1475, encoding a protein with an unknown function, are apparently components of the rnf operon, which suggests that they may be responsible for related functions. The complex was first discovered in the Rhodobacter capsulatus [59] where it transports electrons from NADH to ferredoxin, which donates electrons to nitrogenase [60]. It was further proposed that the same complex runs in an opposite direction, acting as an energy-conserving ferredoxin:NAD+ oxidoreductase complex [61;62]. This complex catalyses the oxidation of reduced ferredoxin coupled with the reduction of NAD + , a process which is coupled with the electrogenic pumping of Na + or H + ions across the membrane out of the cell. The ion gradient can be used for ion (Na + /H + )-dependent symporters to take up sugars and amino acids into the cell, and/or by membrane ATP sythase for ATP generation. Thus, the Rnf complex represents a unique type of primary ion pump and an evolutionary mechanism of energy conservation. The proteins of the Rnf complex of M. roseus, as well as the gene order of the operon, are most similar to those found in Firmicutes, while another variation rnfBCDGEA is found in Bacteroidetes and Chlorobi [62].

Terminal Oxidoreductases of Aerobic and Anaerobic Respiratory Pathways
M. roseus prefers aerobic conditions as indicated by the fact that at optimal conditions the doubling time was 30 minutes for aerobic growth and 2.5 hours for anaerobic growth [15]. Terminal oxygen reductases in M. roseus are represented by two proton translocating cytochrome c oxidases and a quinol oxidase bd complex. The quinol oxidase bd complex is encoded by the gene cluster Mros0842-0843. The first gene, Mros0843, or cydA, encodes subunit I, which is predicted to comprise nine transmembrane helices, and two b-type haems. Mros0842, or cytB, encodes the second subunit comprising eight transmembrane regions and one d-type haem. Both proteins are most similar to subunits of the bd complex of the delta proteobacteria of the order Myxococcales, and its homologs are also present in phototrophic Chlorobi.
The first cytochrome c oxidase-encoding cluster is composed of seven genes (Mros0039-0033), which are organised as a putative operon. The Mros0038 and Mros0035 genes encode the catalytic subunits CoxI and CoxII, respectively, while Mros0037 and Mros0038 encode the transmembrane subunits CoxIII and CoxIV, respectively. Protein sequences and gene order are similar to that of delta-proteobacteria of the orders Desulfovibriales and Desulfuromonadales, suggesting that this cluster was laterally transferred between Ignavibacteriae and a common ancestor of Desulfovibriales and Desulfuromonadales [15].
Another putative membrane-bound cytochrome c oxidase complex is encoded by the eight gene operon Mros1511-1518. This cluster encodes the cbb3-type cytochrome c oxidase, composed of fused subunits I and II (Mros1513), III (Mros1515) and IV (Mros1514). Other genes in this cluster encode cytochrome c oxidase accessory proteins and a subunit of the E1-E2 ATPase family. M. roseus cbb3-type cytochrome c oxidase clearly belongs to type C [63], found in Bacteroidetes and Chlorobi. The cbb3-type oxidases are widespread among bacteria and usually have a very high affinity for O 2 [64]. For example, the human pathogen Campylobacter jejuni has a branched aerobic electron transport chain, with a cbb3-type oxidase of high O 2 affinity (Km = 40 nM) and a quinol oxidase of low (Km = 800 nM) O 2 affinity. The latter complex was mainly expressed at high O 2 conditions, while cbb3type oxidase was significantly induced in a low O 2 environment [65]. Similarly, in Rhodobacter capsulatus, the activity of cbb3-type oxidase is high under microaerobic conditions, but low under fully aerobic or under fully anaerobic growth conditions [66]. Enzymes closely related to M. roseus cbb3-type cytochrome c oxidase were found in aerobic and microaerophilic Bacteroidetes bacteria, for example in Cytophaga hutchinsonii. Therefore, this oxidase may allow respiration of M. roseus under microaerophilic and even aerobic conditions.
Under anaerobic conditions M. roseus is able to grow by fermentation or anaerobic respiration. Fe(III), arsenate and nitrite may be used as electron acceptors, while elemental sulfur, thiosulfate, sulfite and nitrate did not stimulate growth [15]. Terminal reductases of anaerobic respiratory pathways are represented in the M. roseus genome by the nitrite-reductase complex, membrane-bound oxidoreductase complexes of the Complex Iron-Sulfur Molybdopterine (CISM) family and multihaem c-type cytochromes.
Putative dissimilatory nitrite-reductase complex of NrfAH-type consists of the tetrahaem cytochrome c electron transfer subunit NrfH (Mros1099) and a catalytic pentahaem c type cytochrome subunit NrfA (Mros1098). The Mros1100 gene, encoding a putative electron transfer protein with five transmembrane helices, is located in the same operon.
The gene cluster predicted to encode cytochrome c nitrous oxide (N 2 O) reductase is located close to the nrfHA gene. This cluster comprises genes encoding the catalytic subunit NosZ (Mros1104), accessory proteins NosL (Mros1105), NosD (Mros1106), NosF (Mros1107) and NosY (Mros1108), cytochrome C (Mros1103), and a putative iron-sulfur protein (Mros1109). Transcription of this operon may be controlled by the BadM/Rrf2 family transcriptional regulator encoded by the Mros1101 gene. The actual physiological role of the predicted nitrous oxide reductase in M. roseus remains to be investigated since the ability of the bacterium to reduce N 2 O was not reported.
The genome analysis revealed two gene clusters encoding putative membrane bound molybdopterin oxidoreductases of the Psr/Phs family [reviewed in 67]. These membrane-bound complexes typically consist of three subunits: a catalytic subunit with molybdopterin (PsrA/PhsA), an electron transfer subunit with the [Fe-S] cluster (PsrB/PhsB), and a membrane anchor subunit, participating in the transfer of electrons from the quinone pool (PsrC/PhsC). The first operon, Mros1774-1776 (psrABC) encodes a typical three-subunit Psr/Psh oxidoreductase and also contains a putative MerR family transcriptional regulator gene (Mros1773).
The subunits of this oxidoreductase are similar to the respective components of thiosulfate-, polysulfide-, nitrate-, and arsenatereductases, as well as other related proteins. The catalytic subunit Mros1076 of the second oxidoreductase (Mros1074-1076) shares similarity with the respective component of tetrathionate reductases, which are enzymes responsible for the anaerobic reduction of tetrathionate to thiosulfate. The ability to respire tetrathionate is the characteristic of certain genera of Enterobacteriaceae [68], but it was not tested for M. roseus. Taking into account that the specificity of molybdopterin oxidoreductases of the Psr/Phs family cannot be reliably predicted from amino acid sequences, it is possible that the encoded complexes reduce arsenate and/or another yet unknown electron acceptor(s).
The NrfCD cluster encodes two subunits of the CISM family oxidoreductase: the NrfC-like 4Fe-4S ferredoxin iron-sulfur protein Mros1572 and the NrfD-like transmembrane protein (Mros1571). Such complexes lacking molybdopterine-binding catalytic subunits are known to serve as multifunctional quinol:electron acceptor reductases that are able to donate electrons to periplasmic multihaem c-type cytochromes [67]. This electron transfer chain may terminate at the hypothetical Fe(III) reductase linked to the outer membrane [15].
Analysis of the M. roseus genome revealed an additional site of arsenate reduction. This locus contains a four-gene arsenic resistance operon (Mros1066-1063) with a transcription regulatory factor of the ArsR family (Mros1067). The structural proteins of the Ars operon include a permease protein (Mros1066), a soluble arsenate reductase ArsC (Mros1065) of the low-molecular weight protein-tyrosine phosphatase family (pfam01451), a small redoxactive disulfide protein (Mros1064), and a membrane-bound arsenite efflux pump (Mros1063). This arsenate reductase is predicted to be localised in the periplasmic space where it reduces arsenate to arsenite, which may be more toxic but is more easily exported from the cell [69]. This periplasmic enzyme complex serves a detoxifying function and seems not to be involved in the generation of a transmembrane ion gradient, but still may serve as an electron sink.

Conclusions
Until recently, the bacteria of the Chlorobi phylum were considered a homogeneous group of obligately anaerobic photolitothrophs using reduced sulfur compounds and hydrogen as electron donors for photosynthesis. However, the discoveries of new lineages more distantly related to known Chlorobi extended our view of the metabolic diversity of Chlorobi and contributed to the understanding of the evolution of this group. 'Candidatus Thermochlorobacter aerophilum', identified in microbial mats of the hot springs at the Yellowstone National Park and characterised through metagenomic and metatranscriptomic approaches, appeared to be an aerobic photoheterotroph that cannot oxidise sulfur compounds, autotrophically fixes CO 2 in the reverse TCA and does not contain nitrogenase machinery [32]. Similar environments were used to isolate two organisms even more distantly related to Chlorobi: I. album [14] and M. roseus [15]. Both cultivation and genome data show that M. roseus is a facultatively anaerobic heterotroph equipped with a diverse set of hydrolytic enzymes. Under anaerobic conditions it can grow either by fermentation or by anaerobic respiration in the presence of external electron acceptors. In the absence of such acceptors, M. roseus can ferment sugars in the EM pathway following the conversion of pyruvate to acetate. The reduced products that are generated, NADH and reduced ferredoxin, are reoxidised by bifurcating [FeFe]-family hydrogenases with a concomitant reduction of protons to H 2 . The transmembrane ion gradient, required for cation-dependent sugar symporters to transport sugars and aminoacids into the cell, and for membrane ATP sythase for ATP generation, can still be generated by the activity of the Rnf ion pump and vacuolar-type H(+)-translocating pyrophosphatase (Mros2367). In the presence of external electron acceptors, M. roseus can grow by aerobic or anaerobic respiration. Several membrane-bound oxidoreductases allow the complete oxidation of organic substances via the TCA cycle with the generation of a transmembrane proton gradient used by ATP synthase for the generation of ATP. The ability of M. roseus to grow under a wide range of oxygen concentrations up to aerobic conditions distinguishes it from strictly anaerobic phototrophic Chlorobi, and could be attributed to the presence of three different terminal oxidases: two different cytochrome c oxidases and a quinol bd oxidase. It is possible that these oxidases have different O 2 affinities, thus being employed under different growth conditions. Transcription analysis should clarify the role of each enzyme complex.
Overall, the metabolic characteristics of M. roseus are consistent with the basal branching point of Ignavibacteriae on the phylogenetic tree of Chlorobi (Fig. 1) since they lack all of the characteristic features of the photoautotrophic Chlorobi and resemble heterotrophic free-living Bacteriodetes. Photosynthetic machinery appeared to be acquired first (Ca. 'C. thermophilum') followed by metabolic and environmental specialisation, -a shift to an obligate photoautotrophy linked to sulfide oxidation, the acquisition of nitrogen fixation machinery, the loss of flagellar motility and moving into strictly anaerobic, poorly illuminated ecological niches.