Introduction

The discovery of magnetotactic bacteria (MTB) has significantly changed our perception of microbial biomineralization. MTB are a phylogenetically diverse group of microbes that mineralize intracellular nano-sized magnetite (Fe3O4) and/or greigite (Fe3S4) into magnetosome chain(s) (Bazylinski and Frankel, 2004; Bazylinski et al., 2013). The chain structure in each cell acts like a magnetic compass to facilitate the orientation and navigation of MTB by using the Earth’s magnetic field (Frankel et al., 1997; Pan et al., 2009). The magnetosomes in MTB thus represent an ideal system for studying the intracellular biomineralization and bacterial organelle formation (Schüler, 2008; Komeili, 2012). The high chemical purity, strong magnetism and uniform nanometer scale of magnetosome particles also make them potentially applicable in advanced biotechnological and biomedical nano materials (Matsunaga et al., 2007).

Over the past decades, numerous studies on MTB in the Proteobacteria phylum have well demonstrated that the entire process of magnetosome formation is strictly controlled by a group of genes clustered within a coherent genomic fragment described as the magnetosome island (MAI) (Komeili, 2012; Schüler, 2008). However, recent cultivation-independent analyses have revealed an unexpected diversity of MTB outside Proteobacteria (Kolinko et al., 2012; Lefèvre and Bazylinski, 2013; Lin et al., 2014). Among them, those belonging to the genus ‘Candidatus Magnetobacterium’ in the phylum Nitrospirae are of great interest (Garrity and Holt, 2001). Members of this genus, such as ‘Candidatus M. bavaricum’ (henceforth referred to as Mbav), have giant cell sizes (6–10 μm) and are capable of forming as many as 1000 bullet-shaped magnetite magnetosomes arranged into multiple bundles of chains in a single cell, which is different from Proteobacteria MTB (Spring et al., 1993). Originally discovered through electron microscope observation and ribosomal RNA gene sequencing (Vali et al., 1987; Spring et al., 1993), the Nitrospirae MTB are found to be widespread in aquatic environments and sometimes even account for a significant proportion (up to 30%) of microbial biomass in the microhabitats (Spring et al., 1993; Lin et al., 2014). Recent studies have suggested that these organisms may be capable of oxidizing sulfur (Spring and Bazylinski, 2006; Jogler et al., 2010), thus linking this genus to the key steps of iron and sulfur cycles in nature. However, except for a few short genomic fragments (Jogler et al., 2010; Jogler et al., 2011; Lin et al., 2011), very little genetic data of this genus are available owing to their unculturability. Consequently, knowledge of their trophic strategy, physiology and intracellular biomineralization remains very limited.

Here we used a targeted metagenomic approach to assemble the first draft genome sequence of 3.42 Mb from a large rod-shaped MTB strain belonging to the ‘Ca. Magnetobacterium’ genus, which has recently been identified from Lake Miyun near Beijing, China (Lin et al., 2009; Li et al., 2010; Lin et al., 2011). This bacterium is morphologically similar to Mbav and contains hundreds bullet-shaped magnetite magnetosomes that are arranged into multiple bundles of chains (Li et al., 2010; Lin et al., 2011; Figure 1). The draft genome sequence retrieved here has allowed us to gain novel insights into the metabolism and physiological ecology of this poorly understood genus. The genomic information will also lay a foundation for understanding the biomineralization mechanism of Nitrospirae MTB in general.

Figure 1
figure 1

Images of ‘Mcas’. (a) Light microscopy image of ‘Mcas’ enrichment. Note that the applied field direction is from right to left, and bacteria were concentrated at the left edge of the water droplet. Transmission electron micrographs of ‘Mcas’ showing multiple bundles of chains with bullet-shaped magnetosomes (b) and flagella (c). (d) Specific detection of ‘Mcas’ through fluorescence in situ hybridization analysis. Fluorescence microscopy images of the same microorganisms stained with 4,6-diamidino-2-phenylindole (DAPI), with bacterial universal probe EUB338, and with a ‘Ca. Magnetobacterium’ genus-specific probe BaP.

Materials and methods

Magnetic enrichment of MTB and genomic DNA extraction

Surface sediments (depth 5–10 cm) were collected from Lake Miyun located in front of the Yanshan Mountain, about 80 km northeast of Beijing. The physical–chemical characteristics of sediments have been described previously (Lin and Pan, 2010). MTB cells were magnetically collected and purified using the double-ended open separation apparatus (MTB trap) as previously described (Jogler et al., 2009). For transmission electron microscopy observation, 10 μl of MTB enrichments were deposited on Formvar-carbon-coated copper grids and were imaged using a JEM-1400 microscope operating at 80 kV (JEOL Corporation, Tokyo, Japan). Genomic DNA was extracted from the enriched MTB cells using the TIANamp Bacteria DNA Kit (Tiangen, Beijing, China) following the manufacturer’s instructions.

PCR and fluorescence in situ hybridization

16S rRNA gene was amplified using the universal bacterial primers 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-GGTTACCTTGTTACGACTT-3′; Lin et al., 2012). After gel purification, the PCR product was cloned into the pMD19-T vector to construct two clone libraries. Clones were randomly selected and sequenced using an ABI 3730 genetic analyzer (Beijing Genomics Institute, Beijing, China).

Fluorescence in situ hybridization was carried out as described previously (Lin et al., 2012). A Cy3-labeled-specific probe for the genus ‘Ca. Magnetobacterium’ (5′-GCCATCCCCTCGCTTACT-3′; named BaP in this study) (Spring et al., 1993; Lin et al., 2011) and the 6-carboxyfluorescein-labeled bacterial universal probe EUB338 (5′-GCTGCCTCCCGTAGGAGT-3′; Amann et al., 1990) were used for hybridization.

Pyrosequencing and sequence annotation

The genome was sequenced by shotgun pyrosequencing using 454 GS-FLX system by the Genomic Core Facility at University of Virginia, and was assembled using MIRA (version 3.4.0; Chevreux et al., 1999). Gene annotations of assembled contigs were done through the RAST annotation server (Aziz et al., 2008). An expectation value E of <1e−05 was used as the standard cutoff to define putative protein functions. The entire annotations of genome were subsequently manually checked. Annotated assemblies and predicted proteins (Genome ID 798577.3) are available on the RAST guest account (using login and password ‘guest’) at the web addresses http://rast.nmpdr.org. Metabolic pathways were reconstructed using the KAAS pathway tool (Moriya et al., 2007). Comparison of the draft genome to complete genomes of magnetotactic Proteobacteria was performed by bidirectional best hit analysis at a 30% level of protein identity using the SEED viewer (Overbeek et al., 2014).

Phylogenetic and phylogenomic analyses

16S rRNA gene sequence retrieved from the draft genome was aligned with related MTB and non-MTB sequences from Proteobacteria, Nitrospirae and the candidate division OP3 using MUSCLE (Edgar, 2004). Gblocks was performed to eliminate poorly aligned portions of the alignment (Castresana, 2000). Maximum-likelihood phylogenetic tree was subsequently constructed through PhyML (Guindon et al., 2010).

For phylogenomic analysis, protein sequences of 31 housekeeping genes (dnaG, frr, infC, nusA, pgk, pyrG, rplA, rplB, rplC, rplD, rplE, rplF, rplK, rplL, rplM, rplN, rplP, rplS, rplT, rpmA, rpoB, rpsB, rpsC, rpsE, rpsI, rpsJ, rpsK, rpsM, rpsS, smpB and tsf) from the draft genome in this study, complete genomes of magnetotactic Proteobacteria strains Magnetospirillum magneticum AMB-1, Magnetococcus marinus MC-1 and Desulfovibrio magneticus RS-1, and other non-MTB genomes of interest were identified, aligned, trimmed and concatenated by species using the software AMPHORA2 (Wu and Scott, 2012). A maximum-likelihood tree with 100 bootstrap replicates was constructed using RAxML with the LG and gamma models engaged.

Homology modeling analysis

The three-dimensional structures of complete MamK proteins from ‘Ca. M. casensis’, ‘Ca. Magnetoglobus multicellularis BW-1’, D. magneticus RS-1, M. marinus MC-1, M. magneticum AMB-1 and M. gryphiswaldense MSR-1 were modeled by Swiss-Model workspace (Arnold et al., 2006) according to the structure of MreB from Thermotoga maritime (PDB entry: 1JCF). The quality of final model was checked through QMEANclust (Benkert et al., 2009). Phylogenetic analysis of MamK proteins was conducted through the ClustalW (Thompson et al., 1994). Structure-based sequences alignment was performed using ESPript (Gouet et al., 1999), and the result was created through the UCSF Chimera (Pettersen et al., 2004).

Accession number

This Whole Genome Shotgun project of ‘Ca. M. casensis’ has been deposited at DDBJ/EMBL/GenBank under the accession JMFO00000000. The version described in this paper is version JMFO01000000.

Results

General genomic features

A group of giant rod-shaped MTB dominated the MTB community in one microcosm as observed by the Bacteriodrome analysis (Supplementary Movie S1). These MTB were magnetically collected and purified through the ‘MTB trap’ method (Jogler et al., 2009), which yielded 108 cells. About 6.2 μg of genomic DNA was extracted from the enriched cells and was pyrosequenced, yielding around 126 Mb in total with a coverage depth of 37.0 × . Sequence reads were assembled to a total length of 3.42 Mb, which consisted of 70 contigs ranging in size from 1224 to 252 526 bp (with N50=90 253 bp; Table 1). The GC content of the assembled genome is 48.9%, within the range of 45.1–50.2% as determined from the metagenomic fosmids of the Nitrospirae MTB (Jogler et al., 2011; Lin et al., 2011). The draft genome contains 3140 predicted genes, 40 tRNAs and one complete rRNA gene operon (Table 1) that provides a wealth of genomic information valuable for understanding the biology of this poorly understood magnetotactic lineage.

Table 1 General genomic features of the draft genome of ‘Candidatus Magnetobacterium casensis’

Several tests were conducted to determine whether there was any potential contamination to the genome assembly. (i) Two 16S rRNA gene clone libraries were constructed from the extracted genomic DNA, and 36 clones were randomly chosen for Sanger sequencing. All 31 successfully sequenced genes showed high similarities (>99%) and were mostly similar to Mbav (Spring et al., 1993; Jogler et al., 2011). (ii) The fluorescence in situ hybridization analysis (Figure 1d) indicates that the enriched culture was pure. (iii) A survey of 111 conserved single-copy genes has revealed that 102 out of 104 detected genes were found only once in the genome assembly (Supplementary Table S1). (iv) Only a single copy of rRNA operon was identified in the assembled genome (Table 1). All these results strongly indicated the presence of a single genome in the assembly that was derived from the giant rod-shaped MTB cells belonging to the genus ‘Ca. Magnetobacterium’. This MTB population was tentatively named as ‘Ca. Magnetobacterium casensis’ (hereafter referred to as Mcas).

Phylogeny

Cells of Mcas are 1–3 μm in width and 6–8 μm in length, and contain 200–500 magnetosomes organized into 3–5 bundles of chains (Figure 1). According to the phylogenetic analysis of 16S rRNA gene sequences, Mcas is affiliated with the Nitrospirae phylum and is most similar to Mbav with 98.2% identity (Figure 2a). Although similar, cells of Mcas are apparently smaller than that of Mbav and contain lower number of magnetosomes per cell (Jogler et al., 2011; Spring et al., 1993), which is consistent with Mcas being a novel magnetotactic population belonging to the genus ‘Ca. Magnetobacterium’. Mcas and Mbav together form a monophyletic group and show >4% phylogenetic divergence compared with other reported magnetotactic Nitrospirae (Figure 2a). The closest relative of Mcas with complete genome sequence is Thermodesulfovibrio yellowstonii, a thermophilic sulfate reducer isolated from thermal vent in Yellowstone Lake, USA (Henry et al., 1994). The divergence in the 16S rRNA sequences between Mcas and T. yellowstonii is up to 14% (Figure 2a).

Figure 2
figure 2

Phylogenetic analyses of ‘Mcas’. (a) Phylogenetic of 16S rRNA gene sequences based on maximum-likelihood analysis of representative MTB and non-MTB sequences. (b) A maximum-likelihood genome tree illustrating the phylogenetic relationship of ‘Mcas’. The bootstrap support values (out of 100 replicates) are indicated as percentages.

In addition to 16S rRNA genes, 31 universally conserved single-copy genes were used to conduct a genome-level phylogenetic analysis, which confirmed the affiliation of Mcas within the Nitrospirae phylum (Figure 2b and Supplementary Figure S1). Interestingly, the phylum Nitrospirae as a group is closely related to the Deltaproteobacteria, although the bootstrap support for this relationship is not very strong (Figure 2b). Previous studies using 16S rRNA gene sequences and gene order data suggested that members in the Nitrospirae phylum, such as Nitrospira marina and T. yellowstonii, have a close evolutionary relationship with the Deltaproteobacteria (Kunisawa, 2010; Teske et al., 1994).

Carbon fixation

Genes involved in the synthesis of cofactors, vitamins, prosthetic groups and pigments (for example, biotin, coenzyme A, riboflavin, pyridoxine and so on) have been identified in the genome (Figure 3). The salient feature of this genome is its capacity in fixation of CO2 via the reductive acetyl-CoA (Wood–Ljungdahl or WL) pathway not observed for other Nitrospirae members and MTB, in which two CO2 molecules are fixed through two distinct paths named as the western (or carbonyl) and eastern (or methyl) branches, respectively (Berg, 2011; Figure 3 and Supplementary Figure S2). Genome of Mcas encodes several key enzymes of both branches. The genes encoding the western branch are localized in one gene cluster, including the carbon monoxide dehydrogenase (Peg.0034), the CO dehydrogenase/acetyl-CoA synthase (Peg.0033), two subunits of the acetyl-CoA synthase corrinoid iron–sulfur protein (Peg.0032 and Peg.0031) and the corrinoid iron–sulfur protein methyltransferase (Peg.0030). These enzymes incorporate CO2 into the carboxylic group of an acetate molecule. Two eastern branch genes, one encoding a formate-tetrahydrofolate ligase (Peg.3045) and one encoding a methylenetetrahydrofolate dehydrogenase (NADP+)/methenyltetrahydrofolate cyclohydrolase (Peg.2652), are likely involved in reducing CO2 to a methyl group. It should be noted that some sulfate-reducing organisms and aceticlastic methanogens could run the WL pathway in reverse (Schauder et al., 1988; Hattori et al., 2005). We hypothesized that Mcas may be capable of reversing the WL pathway and generating metabolic energy by oxidizing acetate to CO2.

Figure 3
figure 3

Cell metabolic cartoon constructed from the genome sequence of ‘Mcas’. LPS, lipopolysaccharide; rTCA, reductive tricarboxylic acid cycle; TCA cycle, tricarboxylic acid cycle; WL pathway, Wood–Ljungdahl pathway.

The genome also contains several genes encoding enzymes involved in both tricarboxylic acid (TCA) cycle and reductive TCA (rTCA) cycles, including four subunits of 2-oxoglutarate:ferredoxin oxidoreductase (Peg.0666, Peg.0803, Peg.3047 and Peg.3048) and one aconitate hydratase (Peg.0290). Coexistence of different autotrophic pathways has been reported in some microorganisms (Markert et al., 2007), thus it is also possible for Mcas conduct rTCA pathway for CO2 fixation, although this needs further experimental determination. Similar to Mbav (Jogler et al., 2010), a form IV ribulose-1,5-bisphosphate carboxylase/oxygenase (RubisCO) encoding gene (Peg.1181) was identified in the genome. However, the lack of carboxylase and oxygenase activity of this gene suggests that Calvin–Benson–Bassham pathway for CO2 fixation should not operate in Mcas.

Nitrogen metabolism

The Mcas contains genes encoding periplasmic nitrate reductases (Nap), membrane-bound nitrate reductases (Nar), nitrite reductases (Nir), nitric oxide (NO) reductases and nitrous oxide reductases, suggesting that it is capable of gaining energy by converting nitrate to N2 under oxygen-limiting conditions (Figure 3 and Supplementary Figure S2). The nitrate respiration nap and nar genes form a gene cluster (Peg.1539–Peg.1548), and are predicted to catalyze the reduction of nitrate to nitrite. The nirS gene for cytochrome cd1 nitrite reductase, together with nirDFJNJ genes (Peg.1589–Peg.1594) that are involved in heme d1 biosynthesis (Kawasaki et al., 1997), may reduce nitrite to NO. NO may be further reduced to nitrous oxide, and finally to N2 using NO reductases (NorBC) and nitrous oxide reductases (NosZRDFY), respectively. It should be noted that nitrate reductases might operate in the reverse direction and oxidize nitrite to nitrate. In this way, the nitrate reductases serve as nitrite oxidoreductases that shuttle two electrons per oxidized NO2 into the electron transport chain to gain energy (Mussmann et al., 2007). NarG is a candidate enzyme involved in this nitrite-oxidizing process (Mussmann et al., 2007) and has been identified in the sequenced genome (Peg.1542).

Mcas encodes a putative nitrite reductase NAD(P)H large subunit (Peg.2700) and a cytochrome c nitrite reductase NrfA (Peg.1004) that may catalyze the assimilatory nitrite reduction process (Simon, 2002), indicating that Mcas likely uses nitrate or nitrite as a nitrogen source. The genome lacks the majority of genes involved in nitrogen fixation (nif and fix genes). Only a few genes encoding NifA (Peg.2180), NifU (Peg.0443), NifS (Peg.1400 and Peg.1786), FixA (Peg.1876) and FixB (Peg.2062) were identified.

Sulfur metabolism

The genome of Mcas contains a group of genes potentially involved in sulfur oxidation (Figure 3 and Supplementary Figure S2). One gene (Peg.0221) encodes a protein with considerable similarity to sulfide:quinone oxidoreductase (sqr), which could catalyze the oxidation of sulfide (S2−) to elemental sulfur (S0) (Pott and Dahl, 1998; Reinartz et al., 1998). Elemental sulfur could be further oxidized to sulfite by the dissimilatory sulfite reductase encoded by dsrABDEFH genes that have been found in the sequenced genome (Dahl et al., 2005; Holkenbrink et al., 2011). In addition, genes for adenosine 5′-phosphosulfate reductase (Peg.0013 and Peg.0014, aprAB) and adenosine 5′-triphosphate (ATP) sulfurylase (Peg.0012, sat) mediating the oxidation of sulfite to sulfate have also been found in the genome. Genes for sulfur oxidation proteins (Sox) were not identified.

It has been suggested that sulfur oxidation enzymes, such as DsrAB, AprAB and Sat, could operate in the reverse direction, mediating a sulfate reduction pathway from sulfate to sulfide (Zhou et al., 2011). We have found additional genes encoding quinone-interacting membrane-bound oxidoreductase (Peg.0016–Peg.0018, qmoABC) in the genome that are involved in the respiratory electron transfer chain and act as a link between the membrane menaquinone pool and the cytoplasmic reduction of sulfate (Pires et al., 2003). It has been reported that some bacteria can perform both sulfur oxidation and reduction depending on the conditions (Dilling and Cypionka, 1990; Dannenberg et al., 1992), and therefore can perform sulfur cycling between sulfate and sulfide in response to the redox conditions (Cypionka, 1994). It is thus possible for Mcas to reduce sulfate to sulfide in nature, although this capacity needs further experimental investigations.

Iron uptake

Iron is an essential micronutrient for almost all prokaryotes. This is particularly true for MTB in genus ‘Ca. Magnetobacterium’ that need a large amount of iron for synthesis of hundreds magnetite magnetosomes. One gene (Peg.1239) encodes a protein with a ferrous iron transport protein A (FeoA) domain associated with an iron-dependent repressor (TroR) in the genome. In addition, we found two copies of a gene (Peg.0908 and Peg.0909) that encodes protein containing a FeoB domain. Both FeoA and FeoB proteins identified here are most similar to the MAI-containing magnetosome proteins Mad17 and Mad30 of Deltaproteobacteria MTB, respectively, (Lefèvre et al., 2013b), although they are outside of the MAI in Mcas. One gene belonging to the general metal 2+ transporter NRAMP family was also found in the genome, which may provide an alternative mechanism for Fe2+ uptake (Forbes and Gros, 2001; Hopkinson and Barbeau, 2012). Although genes for Sfu/Fbp/Fut family of ferric iron transporters and siderophore biosynthesis were not found, we identified several putative genes for TonB-dependent transporters (catecholate and heme) and Fe3+ ABC ATPase, indicating that Mcas has the potential to acquire Fe3+ from the environments as well.

Magnetosome biomineralization

A 67 580-bp contig of Mcas contains many orthologous genes encoding magnetosome proteins (Figure 4 and Supplementary Table S2). This contig contains 78 open reading frames (ORFs) and 2 tRNA genes; nearly half of them encode hypothetical proteins. The average GC content of the MAI-containing contig is 49.3%, slightly higher than that of the whole assembled genome (48.5%). A mamAB-like operon containing orthologs of mamPMQBAIEQ shows 85% nucleotide similarity to those found in Mbav (Jogler et al., 2011) (Figure 4). The homologs of these proteins in Alphaproteobacteria MTB have been identified to have vital roles in magnetosome membrane biogenesis, magnetosome protein sorting, and magnetosome crystal initiation and maturation (Schüler, 2008; Komeili, 2012).

Figure 4
figure 4

Schematic comparison of the MAI regions in ‘Mcas’, ‘Mbav’ and Deltaproteobacteria MTB strains. Mcas, ‘Ca. M. casensis’; Mbav, ‘Ca. M. bavaricum’; RS-1, D. magneticus RS-1; BW-1, ‘Ca. D. magnetomortis’; MMP, ‘Ca. M. multicellularis’.

We have identified one gene (Peg.0733) encoding a protein with high homology to MamK from Proteobacteria MTB. MamK protein assembles into actin-like filaments that are involved in the formation of magnetosome chains in Proteobacteria MTB (Komeili et al., 2006; Katzmann et al., 2010; Draper et al., 2011; Ozyamak et al., 2013). An acidic protein, MamJ, interacts with MamK and has important role in chain forming and dynamic turnover of MamK filaments (Scheffel et al., 2006; Scheffel and Schüler, 2007; Draper et al., 2011). Intriguingly, Mcas has only one copy of mamK and no mamJ despite that the numerous magnetosome chains assembled inside cells (Figure 4). A partial MamK protein sequence from Mbav has been retrieved recently (Lefèvre et al., 2013b) and is 95% identical (33% coverage) to MamK from Mcas. Phylogenetic analysis of complete MamK protein sequences has revealed that MamK of Mcas is phylogenetically close to that of Deltaproteobacteria MTB (Figure 5a).

Figure 5
figure 5

Comparison of MamK from ‘Mcas’ with its homologs in other bacterial strains. (a) A tree showing the phylogenetic relationship of the homologs. (b) Three-dimensional (3D) structure model of MamK from ‘Mcas’. The image was prepared with PyMol (http://www.pymol.org). (c) Conserved sequence alignment. Conserved residues are indicated as ‘’. (d) 3D-structure alignment of various MamK proteins in three views by rotation along the y axis. Mcas, ‘Ca. M. casensis’; MMP, ‘Ca. M. multicellularis’; BW-1, ‘Ca. D. magnetomortis’; RS-1, D. magneticus RS-1; AMB-1, M. magneticum; MSR-1, M. gryphiswaldense; MC-1, M. marinus; TM, T. maritime; BS, Bacillus subtilis.

Three-dimensional structrue modeling of MamK from Mcas shows that it contains two domains that are each further divided into two subdomains (Figure 5b). Comparison of these structures reveals highly conserved residues of G14, D169, G171, G173, G296 and G297 that are located near the ATP-binding sites in the interdomain cleft (Figure 5b). Alignment of secondary structures of MamK, MreB and actin has revealed four of five motifs involved in the ATP binding, including two phosphate-binding sites (xxxxGxx and xxxDxGxGxx), one adenosine-binding site (xxxxGGxx) and one connecting region 2 (Gx) that are essentially identical in MamK, MreB and actin proteins (Figure 5c). These highly conserved active motifs indicate that MamK protein might have a general evolutionary origin of ATPase. In addition to these conserved regions, the connecting region 1 motif of MamK and MreB, ‘xEPx’, is different from that of actin protein (Figure 5c). Although the secondary structure elements are generally similar, some differences exist in the β4, β5, β6, β7, α1 and α2 among different MamKs (Supplementary Figure S3). MamK proteins from Mcas and MTB in the Deltaproteobacteria contain two additional randomly coiled loops between β6 and β7 and between α2 and β8, which are absent in MamKs from the Alphaproteobacteria MTB strains (Supplementary Figure S3). The additional elements result in the long antiparallel strands β6 and β7 being located to different positions in MamKs of Mcas, ‘Ca. M. multicellularis’, ‘Ca. Desulfamplus magnetomortis’ and D. magneticus RS-1 (Figure 5d and Supplementary Figure S3). Furthermore, a notably conformational change can be seen in the MamK from Mcas, which results from an additional subdomain behind the center of domains I and II (Figures 5b and d).

One gene (Peg.0715) shows remote similarity (E-value: 1e−20) to the mamEO gene identified in BW-1, a fusion of the N terminus of MamE and the C terminus of MamO (Lefèvre et al., 2013b). Because it has higher similarity to the C terminus of MamO (TauE domain), we annotated this gene as mamO-Cter gene (Figure 4). Both MamO and MamE are predicted as serine proteases, and are likely involved in magnetosome protein localization, biomineralization activation and crystal maturation as demonstrated in Magnetospirillum strains of the Alphaproteobacteria (Murat et al., 2010; Yang et al., 2010; Komeili, 2012).

Eight genes show remarkable similarities to mad genes that appear to be specific to the magnetotactic Deltaproteobacteria and/or Mbav (Lefèvre et al., 2013b). Five genes (Peg.0710–Peg.0714) located downstream of mam genes show considerable similarities to mad23 (2 genes), mad24, mad25 and mad26, and form a small gene cluster (Figure 4). Mad23-I bears a high similarity (E-value: 4e−64) to a PBS lyase HEAT domain protein repeat-containing protein (DMR_40890) in D. magneticus RS-1, which is predicted to be involved in iron storage for magnetosome formation (Matsunaga et al., 2009). Mad25 and Mad26 exhibit remote similarities to chromosome segregation ATPases. In addition, three MAI genes (Peg.0723, Peg.0730 and Peg.0735) scattered in the contig are most similar to the mad2, mad10 and mad28, respectively (Figure 4). Only the mad2 gene appeared to be specifically conserved in all known bullet-shaped magnetite-producing MTB (Figure 4). The functions of Mad2 and Mad10 are so far unknown, while the actin-like Mad28 has been assumed to be involved in the arrangement and/or segregation of magnetosome chain(s) in Deltaproteobacteria and Nitrospirae MTB (Lefèvre et al., 2013b).

Intriguingly, a set of novel MAI genes seems to be conserved in Mcas and/or Mbav (Figure 4). We named these genes as man genes (putative magnetosome genes in N itrospirae), and proposed that they may be involved in some special processes of magnetosome synthesis or chain arrangement and segregation in magnetotactic Nitrospirae. We did not find mamHJORSTU genes that are conserved in sequenced Magnetospirillum species (Komeili, 2012; Schüler, 2008). With the exception of the mamAB operon, other magnetosome gene operons of Magnetospirillum, such as mamGFDC, mamXY and mms6, are missing. Although the genome sequence is incomplete, it is likely that MAI in Mcas only contains genes belonging to the mamAB cluster, similar to its counterparts in the Deltaproteobacteria (Lefèvre et al., 2013b; Nakazawa et al., 2009).

Genomic comparison with Proteobacteria MTB

In addition to the magnetosome genes, we have compared the draft genome of Mcas with four complete genomes of Proteobacteria MTB strains, including M. magneticum AMB-1 (Matsunaga et al., 2005), M. gryphiswaldense MSR-1 (Wang et al., 2014) and M. marinus MC-1 (Schübbe et al., 2009) belonging to the Alphaproteobacteria and D. magneticus RS-1 (Nakazawa et al., 2009) in the Deltaproteobacteria. A total of 458 orthologous genes outside MAI have been identified to be shared among Mcas and the four Proteobacteria MTB strains, which were classified into different clusters of orthologous groups categories (Supplementary Table S3). Although this is a conservative estimate because the genome of Mcas is not complete, the products of these orthologous genes seem to be common in Nitrospirae and Proteobacteria MTB, and hence some of them may have roles in magnetotaxis or magnetosome formation. It is widely accepted that chemotaxis and motility have important roles in the magnetotactic swimming behaviors of MTB (Matsunaga et al., 2005; Schübbe et al., 2009), and genes involved in these processes are among the shared orthologous genes identified (Supplementary Table S3).

We further identified 144 genes of Mcas whose best BLAST hits were genes of known MTB strains (Supplementary Table S4). Approximately 79% (114) of them were most similar to genes from the Nitrospirae MTB, further confirming the affiliation of this genome to the phylum Nitrospirae. The remaining 21% had best BLAST hits to MTB strains of Alphaproteobacteria (22 open reading frames) and Deltaproteobacteria (8 open reading frames). Although 54% of these MTB-related genes encoded hypothetical proteins with unknown or only predicted functions, we identified 27 genes categorized in ‘Cell motility and secretion’ and ‘Signal transduction’, indicating their important roles in magnetotaxis behavior of this bacterium.

Motility and chemotaxis

Mcas has multiple flagella located at one pole of the rod-shaped cell body (Figure 1c). They form a flagellar bundle propelling the cells toward the magnetic north. Most flagellar genes are located in four gene clusters (Peg.1652–Peg.1665, Peg.2138–Peg.2144, Peg.2461–Peg.2472 and Peg.0626–Peg.0636) in the genome. Two copies of the motAB genes of flagellar motor rotation are found, indicating that flagella in Mcas are powered by the flux of proton (Brown et al., 2011). These two sets of flagellar motor genes may increase the synthesis efficiency of Mcas and provide necessary power to propel cells of unusual size (up to 8 μm) and mass (possessing hundreds magnetite magnetosomes per cell). In addition, Mcas encodes several type IV pilin proteins that may be involved in adherence, motility, DNA uptake and exchange, electron transfer and protein secretion (Giltner et al., 2012).

There is a substantial expansion of chemotaxis genes in Mcas genome, including 26 genes for methyl-accepting chemotaxis proteins and genes for CheABCDRVWXYZ. Although many chemotaxis genes are scattered in the genome, cheRYZABW genes (Peg.0639–Peg.0645) form a single cluster next to the flagellar genes (Peg.0626–Peg.0636). Interestingly, two chemotaxis genes, one encoding a methyl-accepting chemotaxis protein methylation inhibitor CheC (Peg.0687) and another encoding a response regulator receiver protein (Peg.0688), are located near MAI genes, suggesting their potential roles in the magneto-chemotaxis signaling system (Figure 4). The average number of chemotaxis genes per Mb is 17.5, which is more than that in the genome of M. marinus MC-1 (13.8; Schübbe et al., 2009). Besides the cheA genes, we have found more than 100 genes encoding histidine kinases, including genes for osmosensitive K+ channel and genes for PAS/PAC sensor, and genes encoding predicted signal transduction proteins. The substantial expansion of chemotaxis genes suggests that chemotaxis has an important role in Mcas and allows it to quickly respond to changing environmental conditions.

Electron transport chain and oxygen respiration

Mcas encodes a respiratory chain consisting of complexes I, III, IV and V for energy generation and electron transport. One gene encoding bidirectional [NiFe] hydrogenase diaphorase (Peg.0036) and several genes encoding bacterial F-type ATPase (Peg.2767–Peg.2773) have also been identified in the genome. Members of ‘Ca. Magnetobacterium’ genus are known to live near microoxic to anoxic layers (Jogler et al., 2010), suggesting they being occasionally exposed to oxygen. Accordingly, Mcas contains genes encoding aa3-type cytochrome oxidases that may allow adaptation to microoxic conditions. Genes encoding thioredoxin (Peg.1457) and thioredoxin reductase (Peg.1871) are present in genome as well, which may also allow Mcas to cope with the oxidative stress (Serata et al., 2012).

Cell envelope and cell cytoskeleton

Mcas possesses genes putatively encoding outer- and inner-membrane proteins, such as YidC (Peg.1299), OmpA/MotB (Peg.2197, Peg.1427 and Peg.1479), OmpH (Peg.1841), YaeT (Peg.1840) and lipopolysaccharides-related genes (Peg.0381 and Peg.0970). The presence of a periplasm is suggested by the identification of genes encoding periplasmic proteins. These results indicate that Mcas is a bona fide Gram-negative bacterium.

The genome also contains genes for typical bacterial cytoskeletal systems. Three genes encoding the MinCDE system (Peg.1820–Peg.1822) have been identified, which may be responsible for the placement of bacterial septation sites at mid-cell location (Barak and Wilkinson, 2007). Interestingly, one gene for DivIVA (Peg.0652) is identified in the sequenced genome, which is thought to perform topological control function and regulate the position of MinCD complex in those bacterial species without MinE (for example, Bacillus subtilis). The coexistence of genes for DivIVA and MinE has been reported in certain Gram-positive bacteria (for example, Clostridia acetobutylicum and C. difficile), although it is currently unknown whether both are expressed and how they contribute to mid-cell site selection (Stragier, 2001). We assume that this unique system may have a role in the segregation of multiple magnetosome chains and other cellular contents in Mcas.

Discussion

The draft genome sequence of 3.42 Mb from the uncultivated strain Mcas in the present study provides a solid start point for discussing the trophic strategy, physiology and unique intracellular biomineralization of MTB in the Nitrospirae phylum. The finding of gene clusters for CO2 fixation through the WL pathway was unexpected because this pathway has previously been found in Firmicutes, Planctomycetes, Spirochaetes, Deltaproteobacteria and Euryarchaeota, but not in Nitrospirae (Hügler and Sievert, 2011). The WL pathway is the only pathway present in both CO2-fixing bacteria and archaea, and is considered to be the first autotrophic process on Earth (Hügler and Sievert, 2011). To our knowledge, this is the first report of this pathway in ‘Ca. Magnetobacterium’ genus, which is different from all known autotrophic members in Nitrospirae as well as MTB in Proteobacteria that conduct either rTCA cycle or Calvin–Benson–Bassham pathway for CO2 fixation (Garrity and Holt, 2001; Bazylinski et al., 2004; Matsunaga et al., 2005; Williams et al., 2006; Richter et al., 2007; Levican et al., 2008; Goltsman et al., 2009; Lucker et al., 2010; Ji et al., 2014). Compared with rTCA and Calvin–Benson–Bassham, the WL pathway is biochemically simple and energetically favorable (Peretó et al., 1999), which may lead to a growth advantage of these microorganisms in oxygen-limited environments.

Through magneto–aerotaxis behavior, MTB can efficiently migrate between anoxic and microoxic layers near oxic–anoxic transition zone (Frankel et al., 1997; Bazylinski and Frankel, 2004). In combination with the results of previous microhabitat study (Jogler et al., 2010), the genomic analysis presented here has allowed us to predict the ecology of ‘Ca. Magnetobacterium’ genus. Members in this genus might be capable of adapting to the chemical gradients near oxic–anoxic transition zone by adjusting its metabolic strategies. Specifically, in the upper microoxic layer, Mcas may conduct sulfur oxidation with nitrate and/or oxygen as the electron acceptors, and denitrification with reduced sulfur compounds such as sulfur and thiosulfate as potential electron donors. As it moves down to the anoxic lower layer, it may switch from sulfur oxidation to sulfate reduction where sulfate is one of the major terminal electron acceptors (Brune et al., 2000). At this stage, the WL pathway may also operate in the reverse oxidative direction and generate metabolic energy by coupling the sulfate reduction process to the oxidation of acetate to CO2. Moreover, a large number of histidine kinases response regulators (>150) and chemotaxis genes have been found, which may allow Mcas cells to effectively sense and respond to the microenvironmental conditions. Our results suggest that the metabolic strategies of ‘Ca. Magnetobacterium’ genus may be complex, and further studies are necessary to better understand the true nature of these bacteria.

Comparative genomic analyses of MAIs from Alphaproteobacteria and Deltaproteobacteria MTB and an incomplete MAI from Mbav have suggested a common set of magnetosome genes existed in these MTB (Nakazawa et al., 2009; Jogler et al., 2011; Lefèvre and Wu, 2013; Lefèvre et al., 2013b). The complete MAI of Mcas presented here confirms the presence of these core genes (mamKPMQBAIEO) that should be essential for magnetosome biomineralization in both Nitrospirae and Proteobacteria MTB (Figure 4). Although unlike their Proteobacteria counterparts, the hallmarks of MTB in ‘Ca. Magnetobacterium’ genus are their ability to forms hundreds to a thousand bullet-shaped magnetite magnetosomes organized into multiple bundles of chains (Spring et al., 1993; Jogler et al., 2011). In this study, the complete sequence of magnetosome protein MamK, a key protein for organizing magnetosomes into chain arrangement in Proteobacteria MTB (Komeili et al., 2006; Ozyamak et al., 2013), has been identified, suggesting a similar strategy of magnetosomes chain organization in both Nitrospirae and Proteobacteria MTB. Further phylogenetic and protein structure analyses have revealed that this MamK protein is generally more similar to that of Deltaproteobacteria MTB with bullet-shaped magnetite and/or irregular greigite crystals than that of magnetotactic Alphaproteobacteria containing cuboidal and elongated prismatic magnetite magnetosomes (Figure 5). Considering the genomic structures of MAIs in Nitrospirae MTB are rather similar to Deltaproteobacteria MTB (Figure 4), the magnetosome synthesis and chain arrangement of these two groups may share the same evolutionary history (Jogler et al., 2011; Lefèvre et al., 2013b; Lefèvre and Wu, 2013). However, we do note differences in the conformational structure of MamK protein of Mcas from that of Alphaproteobacteria and Deltaproteobacteria MTB (Figure 5d), which, together with products of other MAI-containing genes (such as mad and man genes), may lead to the formation of unique organization of multiple magnetosome chains in ‘Ca. Magnetobacterium’ genus.

There is a set of man genes that appears to be conserved between Mcas and Mbav. The colocalization of these genes with mam and mad genes indicates their specific functions in magnetosome synthesis and chain arrangement in ‘Ca. Magnetobacterium’ genus (Figure 4). Among them, protein encoded by man1 has remote similarity (E-value: 7e−08) to a putative magnetosome protein Mad11 with unknown function in a Deltaproteobacteria strain ML-1 (Lefèvre et al., 2013b) but does not bear similarity (E-value>1e−05) to any magnetosome proteins of other Proteobacteria MTB. The predicted products of man2, man3 and man4 lack recognizable homology (E-value >1e−05) to any proteins in the databases. Both Man5 and Man6 appear to contain a chromosome segregation protein SMC domain that acts in organizing and segregating chromosomes for cell partition (Jensen and Shapiro, 2003). Hence, we speculate that Man5 and Man6 may conduct similar functions in the arrangement and segregation of multiple magnetosomes chains during cell division. Additional sequencing of novel MTB strains related to Mcas and Mbav is necessary to further confirm the conservation of man genes in ‘Ca. Magnetobacterium’ genus.

The formation of bullet-shaped magnetite magnetosomes has been discovered in the deep-branching MTB lineages such as Deltaproteobacteria and Nitrospirae, leading to the hypothesis that the first magnetosomes were bullet-shaped magnetite particles (Lefèvre et al., 2013a). Nakazawa et al. (2009) assumed that the absence of magnetosome size and morphology controlling genes (such as mamGFDC and mms6) might lead to the formation of bullet-shaped magnetosomes. Recently, a group of mad genes with unknown functions specific to Deltaproteobacteria MTB and/or Mbav has been identified by Lefèvre et al. (2013b), which are predicted to be involved in the intracellular biomineralization of these organisms. Orthologues of mad genes have also been found in the genome of Mcas (Figure 4). Notably, only the mad2 gene appears to be conserved in the available genomes of bullet-shaped magnetite-producing MTB but absent from the draft genome of greigite-producing ‘Ca. M. multicellularis’ (Figure 4), suggesting its potential role in the formation of bullet-shaped magnetosomes. Magnetite magnetosomes with bullet shape have been identified in a MTB population belonging to the candidate division OP3 recently (Kolinko et al., 2012), while so far no genomic data are available for this novel magnetotactic division. Thus, additional genomes of MTB strains from the candidate division OP3 are required to elucidate the genetic mechanisms of biomineralization of bullet-shaped magnetosomes.

In summary, this study represents the first comprehensive genomic analysis of the poorly understood ‘Ca. Magnetobacterium’ genus. Our results provide a better understanding of the trophic strategy, physiology, evolution and biomineralization mechanism of Nitrospirae MTB. The genome information also provides important clues to guide future cultivation attempts of this fascinating group.