Magnetospirillum gryphiswaldense uses intracellular chains of membrane-enveloped magnetite crystals, the magnetosomes, to navigate within magnetic fields. The biomineralization of magnetite nanocrystals requires several magnetosome-associated proteins, whose precise functions so far have remained mostly unknown. Here, we analysed the functions of MamX and the Major Facilitator Superfamily (MFS) proteins MamZ and MamH. Deletion of either the entire mamX gene or elimination of its putative haem c-binding magnetochrome domains, and deletion of either mamZ or its C-terminal ferric reductase-like component resulted in an identical phenotype. All mutants displayed WT-like magnetite crystals, flanked within the magnetosome chains by poorly crystalline flake-like particles partly consisting of haematite. Double deletions of both mamZ and its homologue mamH further impaired magnetite crystallization in an additive manner, indicating that the two MFS proteins have partially redundant functions. Deprivation of ΔmamX and ΔmamZ cells from nitrate, or additional loss of the respiratory nitrate reductase Nap from ΔmamX severely exacerbated the magnetosome defects and entirely inhibited the formation of regular crystals, suggesting that MamXZ and Nap have similar, but independent roles in redox control of biomineralization. We propose a model in which MamX, MamZ and MamH functionally interact to balance the redox state of iron within the magnetosome compartment.
Magnetotactic bacteria (MTB) use unique intracellular organelles, the magnetosomes, to orient along magnetic fields (Jogler and Schüler, 2009). In Magnetospirillum gryphiswaldense (MSR-1) and other MTB, magnetosomes consist of membrane-enveloped, single-magnetic domain magnetite (Fe3O4) crystals that are aligned in regular chains (Jogler and Schüler, 2009). Biomineralization of functional magnetite crystals proceeds in sequential steps and starts with the invagination of the cytoplasmic membrane to form the magnetosome membrane (MM) (Scheffel et al., 2006; Katzmann et al., 2010), which is associated with a set of > 20 specific proteins. This is followed by the uptake and transport of iron into MM vesicles (Uebe et al., 2010, 2011) and the synthesis of magnetite nanoparticles within them (Faivre et al., 2007). It is assumed that biomineralization of the mixed-valence iron oxide magnetite depends on reducing and slightly alkaline conditions and proceeds by co-precipitation of balanced amounts of ferric and ferrous iron (Faivre et al., 2004; Faivre and Schüler, 2008; Fischer et al., 2011). Thus, the process requires a precise biological control of redox conditions (Faivre and Schüler, 2008). Recently it was shown that magnetosome biomineralization in MSR-1 is closely linked with the activity of the respiratory nitrate reductase Nap, and that deletion of the multi-gene nap operon resulted in severe defects in magnetosome crystals (Li et al., 2012).
Besides several accessory and general metabolic functions such as cellular iron uptake and regulation (Uebe et al., 2010; Rong et al., 2012), all specialized functions for magnetosome synthesis in magnetospirilla are encoded by the four operons mms6, mamGFDC, mamAB and mamXY that are part of a larger (∼ 115 kb) genomic region, the magnetosome island (MAI) (Schübbe et al., 2003; Ullrich et al., 2005). The 16.3 kb mamAB cluster encodes functions essential for magnetosome biogenesis (mamB, I, L, Q), magnetosomal iron transport (mamB, M) and magnetite biomineralization (mamE, O, T, P, S, R) (Murat et al., 2010; Yang et al., 2010; Quinlan et al., 2011; Uebe et al., 2011) in addition to functions controlling magnetosome chain assembly and segregation encoded by mamK and mamJ (Komeili et al., 2006; Scheffel et al., 2006; Katzmann et al., 2010; Draper et al., 2011). The mamGFDC (2.1 kb) and mms6 (3.6 kb) operons are not essential for biomineralization but encode accessory functions for size and shape control of magnetite particles (Scheffel et al., 2008; Murat et al., 2010; 2012; Lohße et al., 2011; Tanaka et al., 2011).
The mamXY operon is conserved in all magnetospirilla and encodes MamY, MamX, MamZ [previously also referred to as MamH-like (Richter et al., 2007)] and the tubulin-like FtsZm protein [previously also referred to as FtsZ-like (Ding et al., 2010)] (Fig. 1). Independent studies in MSR-1 and Magnetospirillum magneticum (AMB-1) revealed that all proteins are associated with the MM (Grünberg et al., 2004; Tanaka et al., 2006; Lohße et al., 2011). Deletion of the entire 5.0 kb mamXY operon of MSR-1 resulted in cells that formed two distinct types of magnetosome particles: short chains of nearly regularly shaped, cubo-octahedral crystals were flanked by small particles with poorly defined morphologies (Lohße et al., 2011). MamY of AMB-1 was implicated in membrane tubulation and MM vesicle formation, but has no function in biomineralization as a ΔmamY mutant still produced WT-like magnetosomes (Tanaka et al., 2010). However, loss of FtsZm caused the formation of small, irregular and superparamagnetic particles (Ding et al., 2010).
MamX and MamZ were identified as magnetosome signature genes by genome comparisons of magnetotactic alphaproteobacteria, but so far were not found in other MTB (Richter et al., 2007). MamX displays weak and local sequence similarity to the other magnetosome proteins MamS and MamE. Whereas the C-terminal domain of MamZ exhibits similarities to a ferric reductase-like transmembrane component (Richter et al., 2007) its N-terminal domain is homologous to the Major Facilitator Superfamily (MFS) transporter MamH (64% similarity) which is encoded within the mamAB operon. A mamH deletion mutant in AMB-1 was able to produce WT-like magnetosomes and only showed a slightly reduced magnetic response (Cmag) (Murat et al., 2010). However, the functions of MamX, MamZ and MamH and their significance for magnetosome formation have remained unknown.
In this study we genetically dissected the functions of mamX, mamZ and mamH in MSR-1. Our data indicate that all three genes have key roles in magnetite biomineralization. We demonstrate that MamX and MamZ are likely involved in redox control to poise optimal conditions for magnetite formation, and that these functions rely on the presence of two putative haem c binding ‘magnetochrome’ domains in MamX and the ferric reductase-like transmembrane component of MamZ. Furthermore, our data suggest that the redox pathway mediated by MamX and MamZ is likely to act independently from nitrate reduction. We present a model, in which MamZ, MamX and MamH functionally interact in the MM to form an iron oxidoreductase and transport complex for magnetite biomineralization.
Transcriptional organization of the mamXY operon
After correction of a mispredicted N-terminus we found mamX and mamZ of MSR-1 to overlap by 17 bp as in other magnetospirilla (Fig. 1A and Fig. S1), suggesting a close functional association and translational coupling of both genes. mamX is always followed by mamZ in all other magnetospirilla and also in Magnetococcus marinus (MC-1) and Magnetovibrio blakemorii (MV-1), despite the different genomic context of mamXZ in these strains (Jogler and Schüler, 2007; Richter et al., 2007). It was shown in a previous study that all four genes of the polycistronic mamXY operon are likely driven from a single, yet unidentified promoter (Ding et al., 2010), which we predicted within the region 285 bp upstream of the first transcribed gene mamY and downstream of an adjacent transposase gene (Fig. 1A). In addition, a conspicuously large (162 bp) intergenic region between mamY and mamX might contain an alternative internal promoter (Fig. 1A). To assess activity and strength of the putative promoters, we transcriptionally fused egfp and gusA as reporter genes behind the 285 bp and the 162 bp fragment respectively. Except for the positive control (egfp fused to the strong mamDC promoter) (C. Lang et al., unpublished), fluorescence intensities of all tested EGFP fusions were below detection in plate reader assays. By fluorescence microscopy we only detected a faint signal for the PmamXY–egfp fusion, indicating that the 285 bp region is active as a promoter, but relatively weak (Fig. S3). No fluorescence was detectable for the intergenic sequence between mamY and mamX by microscopy (Fig. S3). Although this region may contain other regulatory elements, we therefore conclude that PmamXY is the only promoter which drives transcription of the polycistronic mamXY operon. Using the more sensitive GusA reporter, we estimated the relative strength of PmamXY as approximately 22.5% of the PmamDC activity (Fig. S3).
Deletion of mamX and mamZ cause similar impairments of magnetosome biomineralization
To analyse their function in magnetosome formation, we constructed non-polar in-frame deletions of mamX and mamZ. Both mutants exhibited very similar phenotypes: when grown under standard conditions (microoxic, FSM medium), both ΔmamX and ΔmamZ strains showed a slightly reduced magnetic response (Cmag = 88 ± 5% and 77 ± 7% of WT respectively) and still produced chains of electron-dense particles. However, two types of crystals could be clearly distinguished by TEM: in addition to apparently regularly shaped and -sized, WT-like particles (in the following referred to as ‘regular’), cells contained variable numbers of small, irregularly shaped particles that appeared to be thin, i.e. flake-like, sometimes needle-shaped (in the following referred to as ‘flakes’) (Fig. 2A and B). All particles were still aligned in a single chain, in which regular particles were found at the centre and sandwiched by flakes at both ends. Among cells from the same culture, appearance of particles varied from individuals with either almost all-regular or all-flake particles, whereas most cells typically had several regular crystals (Fig. 2A and B). While particles were still clearly aligned in chains even in those cells in which flakes were predominant, in places scattered flakes could also be observed. The distributions of the diameters of ΔmamX (26.1 ± 9.3 nm) and ΔmamZ particles (30.1 ± 15.0 nm) did not differ significantly (P = 0.0767 in Mann–Whitney U-test), but particles were significantly (P < 0.0001) smaller than the magnetosomes of the WT (39.2 ± 9.3 nm) (Fig. 2E).
High-resolution transmission electron microscopy (HRTEM) and electron diffraction revealed that regular particles of both mutants were fully crystalline and consisted of magnetite as in the WT. However, flake-like, irregularly shaped particles were poorly crystalline, and if they displayed any periodic lattice fringes in HRTEM micrographs, the spacings of those fringes (as measured from the Fourier transforms of the HRTEM images) were consistent with the d-spacings in the structure of haematite (α-Fe2O3) (Fig. 3). Although regular particles superficially resembled magnetosome crystals of the WT, TEM of isolated ΔmamZ crystals at higher magnification revealed a higher proportion of the crystals (42%, n = 233) to be twinned than in the WT, and an additional 12% of polycrystalline particles (WT: 22% twinned, 1% polycrystalline, n = 218) (Fig. 2D). This observation suggests that the nucleation of even regular crystals is disturbed at an early stage in the mutants.
Cryo-electron tomography (CET) of both mutants showed that the shape and size of empty and partially filled magnetosome membrane vesicles were unaffected. As in the WT, vesicles were found in close proximity to the inner cell membrane (Fig. 4Aii, Biii, Cii, Ciii) and associated closely with the cytoskeletal magnetosome filament (Fig. 4Bi). Both regular particles and flakes were surrounded by WT-like vesicles (Fig. 4A–C). However, multiple nucleation sites were sometimes observed in the flake-containing vesicles (Fig. 4Biii).
EGFP fused to the C-terminus of the magnetosome marker protein MamC (Lang and Schüler, 2008) localized as a filamentous structure at midcell in the WT as well in ΔmamX and ΔmamZ. The extension of the fluorescent signal correlated with the length of the particle chains (regular crystals + flakes) (Fig. 4D and E) typically observed by TEM. This again indicates that the MM-specific localization of magnetosome-associated proteins is not affected for both regular and flake-like particles in ΔmamX and ΔmamZ.
Synthesis of WT-like magnetosome crystals and chains could be restored by expressing WT mamX and mamZ alleles in their respective deletion backgrounds from plasmids (pOR98 and 031) under control of the native PmamXY (Fig. 2G and H), demonstrating that the observed similar phenotypes were due to the absence of the targeted genes rather than polar effects of the deletions. Notably, even slight modifications of the C-terminus of MamX, such as the fusion of peptides or proteins (mCherry, a 6×-His tag or even two additional amino acids) abolished the ability to complement the ΔmamX mutant (data not shown), indicating a stringent requirement for domain conservation.
Dynamics of biomineralization in ΔmamX and ΔmamZ mutants
As mutant cells cultivated under standard conditions always exhibited two clearly distinct types of particles, we determined whether the haematite containing flakes are intermediates that can be eventually converted into regular magnetite crystals, or if the fate of individual particles is predetermined at an early stage of biomineralization. To investigate the time-course of flake formation in the mutants, we performed iron induction experiments. Strains were passaged in low-iron medium until they were non-magnetic (Cmag = 0). Magnetite formation was then induced by transferring cells back to iron-sufficient medium (50 μM ferric citrate) as previously described (Faivre et al., 2008; Scheffel et al., 2008). The WT started to form magnetite crystals (as indicated by an increase in Cmag after 2 h) and after 4 h had almost completely restored its Cmag to normal levels, whereas both mutants showed a substantial delay in regaining magnetic properties, as indicated by a slower recovery of the Cmag (Fig. 5). As seen in the WT, magnetosome particles were emerging from different positions along the entire length of the ΔmamZ mutant cells without any consistent pattern. Although flakes were dominating at the beginning of the experiments, at later stages more and larger regular crystals were appearing and, similar to the WT, chains were formed with regular crystals at mid-chain, flanked by flake-like particles, thereby restoring the regular phenotype of the mutants as observed under continuous standard growth conditions (Fig. 5). TEM revealed that the ΔmamZ cells contained already many flakes and a few small, but regularly shaped particles 4 h after iron induction. However, the Cmag of the mutant was only marginally increased, indicating that flakes and small crystals did not contribute much to the cellular magnetic response (Fig. 5). At later stages, the ratio of mature crystals to flakes increased steadily in the ΔmamZ mutant, reflected also by an increasing Cmag (Fig. 5). However, even after 7 h, the Cmag was still lower than usually observed for ΔmamZ cultures continually grown under iron-replete conditions.
After iron induction, the ΔmamX mutant showed a very similar pattern with respect to the different stages of biomineralization. However, magnetic response and particle formation re-evolved faster, although still slower than in the WT (Fig. 5). During the entire experiment, we never observed gradual intermediates between flake-like particles and regular magnetosomes, but particles always had distinct appearances that could be assigned to either the ‘flake’ or ‘regular’ type with respect to shape, size and crystallinity (Fig. 5).
We also studied magnetosome formation in division-inhibited cells treated with cephalexin. By inhibiting the cell division protein FtsI, cephalexin blocks final septation and separation of the daughter cells while the cells still constrict at stalled division sites, leading to highly elongated filaments that are equivalent to several consecutive generations (Katzmann et al., 2011). Similar to the WT, 12–16 h after cephalexin treatment both ΔmamX (Fig. S4) and ΔmamZ (not shown) mutants displayed separate multiple magnetosome chains at distinct stalled division sites. As in untreated cells, the central parts of chains consisted of regular crystals flanked by flakes at their ends. Between these chains, scattered and apparently immature particles were frequently found, which could be identified either as irregularly shaped flakes or as small but regularly shaped crystals (Fig. S4). Larger regular crystals were usually found in the magnetosome chains but also scattered within the cells (Fig. S4). Small regular crystals and flakes thus coexisted throughout all stages of magnetosome development. Combined with the absence of transient crystallization phases this observation therefore indicates that crystal fate is already determined at an early stage of nucleation, before particles mature and develop to their full size.
Double deletions of mamH and mamZ have an additive effect on biomineralization, whereas single mamH deletion only affects magnetosome number and size
MamZ is predicted to have 18 transmembrane domains (TMD) and a unique domain architecture with a MFS transporter domain fused to a putative ferric reductase-like transmembrane component (PFAM: ferric_reduct, in the following referred to as ferric reductase domain) (Fig. 1C) (Richter et al., 2007). The N-terminal MFS transporter domain of MamZ shows 64% similarity to MamH, another magnetosome protein of MSR-1. MamH is encoded by the first gene of the mamAB operon and was also predicted to be an MFS transporter (Schübbe et al., 2003; Richter et al., 2007), suggesting a related function in biomineralization. MFS family members typically are characterized by a 12-fold TMD topology (Reddy et al., 2012) with a long cytoplasmic loop between TMD 6 and 7, features also shared by MamZ and MamH (Fig. 1C and D). To investigate a potential functional relationship of MamZ and MamH, we constructed markerless in-frame deletions, in which mamH was excised alone or in combination with mamZ. Deletion of mamH alone resulted in a decrease of magnetosome number and size. Whereas the WT had 31.4 ± 5.7 magnetosomes per cell with a mean size of 39.2 ± 9.3 nm under standard conditions, ΔmamH only exhibited 21.3 ± 7.8 magnetosomes per cell with a size of 22.2 ± 7.0 nm (Fig. 6D). The Cmag of ΔmamH was only slightly decreased (83% of WT). All magnetosomes of ΔmamH had a cubo-octahedral shape and despite their smaller size appeared WT-like (Fig. 6A). Deletion of mamH in the ΔmamZ background had considerably stronger effects: the Cmag of this mutant was significantly decreased (21 ± 7% of WT) and only very few or no regular crystals were detectable in the cells (Fig. 6C). Most of the observed particles resembled the poorly crystalline flakes of ΔmamX or ΔmamZ cells. Transcomplementation of ΔmamH by expressing the WT gene under control of the PmamDC promoter (pOR101) partially restored the synthesis of larger crystals (31.0 ± 10.7 nm), although not fully back to WT level, as frequently observed in MSR-1 for genes expressed from medium-copy plasmids (Uebe et al., 2011) (Fig. 6B and D).
The YedZ-like domain of MamZ is essential for protein function
MamZ is one of the very few MFS members with fusions to other functional domains (Reddy et al., 2012). The predicted ferric reductase domain of MamZ shows highest similarity to the haem B binding membrane protein YedZ, which is found in various bacteria (Brokx et al., 2005; von Rozycki et al., 2005). From 2700 predicted hybrid proteins with conserved ferric reductase domains deposited in the PFAM database, 57% are fused to a FAD-binding domain and/or a NAD-binding domain, 37% have YedZ-like architecture, whereas no other protein besides MamZ shows fusion to a MFS domain (Von Rozycki et al., 2005). As the YedZ-like domain in MamZ, YedZ from Escherichia coli has six TMD and resides in the cytoplasmic membrane. In a complex with the periplasmic molybdopterin binding subunit YedY, YedZ forms an oxidoreductase for diverse sulphoxide compounds (Loschi et al., 2004). YedZ orthologues from bacteria contain the conserved putative haem B-coordinating residues His-91, His-151 and His-164 (Brokx et al., 2005), a feature also shared by the YedZ-like domain of MamZ. To study the role of the C-terminal YedZ-like ferric reductase domain in MamZ function, we deleted this domain by removing a large internal region in MamZ, starting 39 bp downstream of the last codon of the putative 12th trans membrane domain and upstream of the last 12 bp (genotype: mamZ Δ438–639).Expression of the truncated MamZ protein on isolated magnetosomes was confirmed by mass spectrometry. Deletion of the YedZ-like domain alone phenocopied the deletion of the entire mamZ gene, i.e. both mutants produced regular magnetosome crystals flanked by the same flake-like particles (Fig. 2C). This demonstrates that the YedZ-like domain has a crucial role for the function of the entire protein in magnetosome biomineralization. In the magnetospirilla AMB-1 and MSR-1 a genuine yedZ gene is located elsewhere in the genome in addition to mamZ, forming an operon together with yedY. To analyse if the YedZ-like domain of MamZ functionally interacts with the genuine YedY protein in magnetosome formation, we constructed a YedY deletion mutant in MSR-1, which however, showed WT-like magnetosomes and magnetosome organization but no biomineralization defect (Fig. S5). We therefore exclude a functional connection of MamZ and YedY in magnetosome formation of MSR-1.
Absence of nitrate further impairs biomineralization in ΔmamX and ΔmamZ
Previous work suggested a role of nitrate reduction in redox control for magnetite biomineralization. Poorly crystalline magnetosome particles somewhat resembling those of ΔmamX and ΔmamZ were biomineralized in MSR-1 cells deficient in nitrate reductase (Δnap) (Li et al., 2012). This observation prompted us to investigate a potential link between MamX and MamZ functions, redox control and nitrate reduction. Neither increased concentrations (up to 500 μM), nor the redox state of iron supplied in the medium (either all-ferrous or all-ferric), or growing the cells in the entire absence of oxygen had a pronounced effect on biomineralization in the mutants. Cultivation of ΔmamX and ΔmamZ in ammonium medium, where NO3− (4 mM) was substituted by equimolar amounts of NH4+, however, resulted in a strong exacerbation of the phenotype: both ΔmamX and ΔmamZ showed an even stronger Cmag reduction when grown under microoxic conditions and in the absence of nitrate (Fig. 7J), whereas WT cells were unaffected under these conditions (Li et al., 2012; Fig. 7A and B). Likewise, the number of regular crystals decreased and flakes prevailed. Only 27% of ΔmamX cells and 34% of ΔmamZ cells (n = 100) had one to three apparently regular crystals, whereas all other particles were flakes (Fig. 7C–D and F–G). Increasing nitrate concentrations (0–4 mM) gradually restored Cmag and increased the number of regular crystals in ΔmamX cultures (Fig. S6). Addition of 1 mM nitrite to ammonium medium did not restore the ability to form abundant regular magnetite crystals, although all nitrite became consumed by the culture (Fig. S6).
To further analyse a possible relation of biomineralization defects to denitrification we deleted mamX in a genetic background in which the entire operon (nap) encoding the periplasmic nitrate reductase was deleted. Δnap predominantly produces irregularly shaped particles, although some cells also form regular magnetite crystals (Li et al., 2012; Fig. 7H). The Δnap ΔmamX double mutant grown in FSM medium, however, was even more severely impaired in biomineralization and resembled ΔmamX grown in the absence of nitrate with almost only flake-like particles and a Cmag of only 0.11 (ΔmamX: 1.17; Δnap: 0.42). This demonstrates that deletion of mamX in the Δnap background further reduces the magnetite formation ability and indicates an independence of both functions.
Substitution of a putative paired CXXCH (‘magnetochrome’) motif in MamX abolishes its function
In search for possible redox-active domains, we identified a characteristic paired CXXCH motif in MamX (aa 65–69 and 104–108) (Fig. 1B), which is known to mediate covalent haem binding in c-type cytochromes (Bowman and Bren, 2008) and was recently described as a haem c bound ‘magnetochrome domain’ in MamE of AMB-1 and MamP of the magnetotactic spirillum QH-2 (Siponen et al., 2012). In order to elucidate the relevance of the putative magnetochrome domain for MamX function, we exchanged both CXXCH sites to AXXAH by introducing four single aa substitutions into the chromosomal mamX gene. Although we failed to directly detect the mutated protein in isolated magnetosome samples by mass spectrometry, evidence from other studies suggest that this specific aa substitution does generally not affect protein stability (Tomlinson and Ferguson, 2000; Quinlan et al., 2011). The resulting mamX AXXAH double exchange mutant exhibited the same phenotype as ΔmamX (Fig. 2F): as in ΔmamX, under microoxic conditions and in the presence of nitrate, the Cmag (1.24 ± 0.03) of the mamX AXXAH mutant was approximately 15% lower than WT Cmag (1.46 ± 0.03) and further decreased to 0.29 ± 0.03 when the cells were grown in ammonium medium. These results were further confirmed by TEM, which revealed an identical phenotype as in ΔmamX, i.e. the production of regular crystals and flakes in FSM medium and prevalence of flakes in the absence of nitrate (Fig. 7E). Although we cannot entirely preclude effects of altered expression levels, this indicates that inactivation of the putative haem c binding sites is already sufficient to completely abolish MamX function and phenocopies loss of the entire protein.
In this study we showed that the mamX, mamZ and mamH genes play a key role in magnetite biomineralization in MSR-1. In ΔmamX and ΔmamZ, two distinct types of particles were found to coexist within the same cells: besides regularly sized and shaped, but predominantly twinned magnetite crystals, small and amorphous or poorly crystalline, flake-like particles were present. We never observed intermediate sizes and shapes that would possibly represent gradual transitions between regular and flake-like particles. In division-inhibited and iron-induced ΔmamX and ΔmamZ cells both regular crystals and flakes were developing simultaneously at different locations, but became recruited into magnetosome chains, indicating that a transformation of flakes into regular crystals at a later stage is rather unlikely. Although the flake-like particles appeared to be poorly crystalline and to consist of several grains, crystalline islands were present in many of them. Several of the crystalline parts in flake-like particles could be imaged in near crystallographic zone-axis orientations, i.e. Fourier transforms of their HRTEM images contained a two-dimensional periodic pattern of intensity maxima. In all cases the d-spacings derived from these maxima and the angles between reciprocal lattice rows were consistent with the structure of the non-magnetic iron oxide haematite (α-Fe2O3). The presence of haematite in the flake-like particles was also confirmed by several HRTEM images. Haematite particles were previously observed in cells of MSR-1 as the result of a single amino acid substitution in the putative magnetosomal iron transporter MamM (Uebe et al., 2011), and it was speculated that the formation of haematite was favoured over magnetite by disturbance of magnetosomal pH, or ferric to ferrous iron ratios and concentrations (Jolivet et al., 1992; Faivre et al., 2004). Alternatively, haematite might result from chemical transformation of amorphous ferric hydroxide initially present in flakes, either within living cells or during the subsequent specimen preparation and storage. Despite previous speculation that magnetosomal magnetite may directly evolve from haematite precursors (Staniland et al., 2007), this would require the dissolution of haematite and subsequent recrystallization as magnetite (Behrends and Van Cappellen, 2007), which is unlikely to occur in the magnetosome vesicles. We therefore conclude that the fate of developing crystals must already be predetermined at a very early stage of mineralization, consistent with the existence of an ‘activation’ mechanism or ‘checkpoint’ for magnetite synthesis as suggested previously (Komeili et al., 2004; Komeili, 2011; Quinlan et al., 2011), which commits nascent crystals to develop into either one or the other mineral.
Although both types of particles aligned in a linear fashion, larger regular and ferrimagnetic magnetite crystals were found tightly spaced at mid-chain, while presumably non-magnetic flakes were located at the ends. This observation provides further indications that magnetic interaction of magnetosomes is necessary for their concatenation and positioning within chains, but not for linear organization per se which relies on the action of biological structures (Katzmann et al., 2011; Klumpp and Faivre, 2012).
MamZ and MamH are putative iron transporters and MamZ is involved in redox control for magnetosome formation
MamZ contains a conserved ferric reductase transmembrane component of the YedZ-type, which is assumed to bind haem B and therefore might be involved in electron shuttling and redox reactions (Brokx et al., 2005). In E. coli YedZ interacts with the molybdopterin-binding subunit YedY to form a putative oxidoreductase complex (Loschi et al., 2004; Brokx et al., 2005; Kappler, 2011). However, MamZ function in MSR-1 does not require interaction with the genuine YedY, since our deletion of YedY had no effect on magnetite biomineralization. In addition, MamZ contains a MFS transporter domain and represents the only known example in which this domain is fused to a ferric reductase domain (von Rozycki et al., 2005; Reddy et al., 2012). The MFS domain of MamZ also shares high identity with MamH, which indicated a functional relationship of both proteins. Despite the similarities between MamZ and MamH, we failed to swap the YedZ-like domain of MamZ to full-length MamH, as the resulting chimeric protein was non-functional and unable to complement either ΔmamH or ΔmamZ (data not shown). We demonstrated, however, that deletion of mamH resulted in a slightly reduced Cmag in MSR-1 caused by a somewhat reduced size and number of magnetosomes, as it was also already described for the same mutant in the related strain AMB-1 (Murat et al., 2010). However, whereas single deletions of either mamZ or mamH were still able to synthesize larger and regular magnetite crystals, double deletion of mamZ and mamH had a more severe effect, as magnetite biomineralization was substantially reduced in ΔmamHZ, and only few cells produced regular crystals at all. We therefore conclude that both proteins have partially redundant functions, whereas the presence of at least one of those MFS homologues is necessary for the synthesis of regular magnetite crystals.
For many MFS members the substrate which they transport is still unknown (Saier et al., 1999; Reddy et al., 2012), and the distinct branching of MamH and MamZ within the family tree (von Rozycki et al., 2005; Richter et al., 2007) prohibits reliable similarity-based predictions. It nevertheless was shown that distant MFS members from fungi and pathogenic bacteria indeed transport iron chelates (Lesuisse et al., 1998; Chatfield et al., 2012), and based on its unique combination of a putative ferric reductase with a MFS transporter domain, MamZ was hypothesized to be an iron transporter (Von Rozycki et al., 2005; Reddy et al., 2012). It is therefore tempting to speculate that the magnetosome-associated MamH and MamZ are as well involved in magnetosomal iron transport, but might have functions slightly distinct from the CDF proteins MamM and MamB, which were already implicated in magnetosomal iron accumulation (Uebe et al., 2011). For example, instead of divalent ferrous iron, the common substrate of CDF transporters, MamZ might mediate transport of ferric iron, as suggested by the presence of a putative iron reductase component and transporter domain combined in a single protein. Our observation that deletions of the ferric reductase domain alone already abolished protein function would be consistent with this assumption. Despite repeated attempts we failed to detect any iron transport or iron reductase activity upon MamZ expression in E. coli in vitro or in vivo (data not shown). Therefore, future work has to directly prove the predicted functions of MamZ and MamH in reduction and transport of iron.
MamX is a putative magnetochrome protein involved in redox control of magnetite biomineralization
We identified a putative haem c binding (paired CXXCH) motif in MamX, which shares the characteristics of similar domains recently identified in other magnetosome proteins (MamPTE) which because of their exclusive occurrence in MTB and their ability to bind haem c were designated ‘magnetochromes’ (Siponen et al., 2012). Magnetochromes are cytochrome c-like proteins and are assumed to regulate redox conditions for magnetosome formation (Siponen et al., 2012). MamXMSR-1 and its homologues from AMB-1 and Magnetospirillum magnetotacticum (MS-1) are highly conserved, whereas MamXMC-1 contains and additional putative paired CXXCH haem-binding motif and has a larger size (345 aa, 55% similarity to MamXMSR-1). MamXMV-1 is shorter (110 aa, 61% similarity to MamXMSR-1), but also contains the characteristic paired CXXCH motif. Our deletion of either full-length mamX or the substitution of its paired CXXCH motif impaired magnetite biomineralization in MSR-1. This suggests that MamX is a redox-active protein, and its putative magnetochrome domains are essential for its function. In a previous study the magnetosome protein MamE of AMB-1 was also implicated in magnetite crystal maturation, a function which was associated with its magnetochrome domains (Quinlan et al., 2011), and deletion of mamP and mamT led to the formation of irregular particles in the same organism (Murat et al., 2010).
MamX thus is likely involved in redox control for the synthesis of the mixed-valence iron oxide Fe3O4 under oxidant-limiting conditions, i.e. at low oxygen and nitrate concentrations. Likewise, a role in iron reduction was already demonstrated for several c-type cytochromes from dissimilatory ferric iron reducers (Shi et al., 2012). Intriguingly, effects of deletions of mamZ and mamX could be partially rescued by nitrate, whereas only relatively minor effects of nitrate were detected for the WT. Nitrate reductase activity was recently implicated in poising redox conditions for magnetosome formation. Deletion of the nap operon encoding a periplasmic nitrate reductase led to the formation of smaller and irregular magnetosome particles in MSR-1, somewhat reminiscent of the flakes produced in ΔmamX and ΔmamZ, and it was concluded that, independent of its function as a respiratory nitrate reductase, the protein is involved in magnetosomal redox control (Li et al., 2012). The strong impairment of biomineralization in ΔmamX and ΔmamZ in the absence of nitrate might therefore hint towards a direct functional connection with nitrate reductase activity, and it was hypothesized by Siponen et al. that NapC, a CymA-like multihaem cytochrome encoded by the nap operon may transfer electrons from the quinone pool to magnetochrome-containing proteins (Siponen et al., 2012). However, the additive effect on magnetite synthesis upon our co-deletion of mamX and nap seems to argue for functional independence and redundancy of MamX and Nap in redox control, rather than a direct link of both pathways.
MamXZH may form an iron oxidoreductase and transport complex
Deletion mutants of the syntenic and overlapping mamX and mamZ genes resulted in nearly identical phenotypes under all tested conditions, which were strongly reminiscent of the deletion of the entire mamXY operon (Lohße et al., 2011). Furthermore, co-deletion of mamZ and mamH led to severe impairments in magnetosome formation and indicated a partly redundant function of both MFS transporters. While mamY, the first gene of the mamXY operon, is not likely involved in biomineralization (Tanaka et al., 2010), deletion of ftsZm, the last operonal gene encoding a truncated FtsZ protein was previously found to have a similar effect on magnetite synthesis (Ding et al., 2010). We therefore conclude that MamX, MamZ together with MamH act on the same stage of magnetite biomineralization and postulate a hypothetical model of interaction. Based on experimental findings (Grünberg et al., 2004; Tanaka et al., 2006; Lohße et al., 2011) and predictions, MamX, MamZ and MamH are associated with the magnetosome membrane, where they are likely to interact directly or indirectly. MamX and MamZ may then form a complex for the reduction of ferric iron and its concomitant transport by MamZ and MamH. The close interaction of MamXZH might be facilitated by further scaffolding factors, such as for instance FtsZm, which is encoded within the same operon as MamX and MamZ and a member of the tubulin-like FtsZ family known to serve as an interaction hub for multi-protein assembly forming the divisome complex (Goley et al., 2011). This model provides several predictions which can be experimentally tested in future work, such as the haem association of MamX and MamZ, direct interaction of MamXZH and FtsZm to form a common complex and iron transport as well as oxidoreductase activity of MamXZ. In conclusion, our study has uncovered key functions of three major magnetosome proteins in magnetite synthesis, which requires proper redox control by pathways that are partially redundant, and interlinked with cellular metabolism.
Bioinformatics and sequence analysis
Protein sequences, primary structures and predicted localizations were analysed with blast (http://blast.ncbi.nlm.nih.gov/), SMART (Letunic et al., 2004) and CELLO (Yu et al., 2004) algorithms. For comparative protein sequence analyses, clustalΩ (Sievers et al., 2011) was used. Protein topologies were modelled with the TMHMM algorithm (Krogh et al., 2001) and visualized with TMRPres2D (http://bioinformatics.biol.uoa.gr/TMRPres2D/). The annotation of mamZ in MSR-1 was corrected by comparing and analysing the annotation and conserved genomic context in MSR-1, AMB-1 and MS-1 (Fig. S1).
Bacterial strains, plasmids, culture conditions and magnetosome isolation
Bacterial strains and plasmids used in this study are listed in Tables S1 and S2. E. coli strains were cultivated in lysogeny broth (LB) medium. When necessary, kanamycin (km) was added to 25 μg ml−1. E. coli BW29427 cultures were supported with 1 mM dl-α, ε-diaminopimelic acid (DAP). Media were solidified by addition of 1.5% (w/v) agar. Unless otherwise stated, M. gryphiswaldense cultures were grown at 30°C under microoxic conditions (1% O2) in modified flask standard medium (FSM) (Heyen and Schüler, 2003) or in ammonium medium where 4 mM nitrate is equimolarly substituted by ammonium. When appropriate, km was added to 5 μg ml−1. For cultivation experiments in ammonium medium, cells from FSM pre-cultures were passaged at least three times before samples were taken. Optical density (OD) and magnetic response (Cmag) of exponentially growing cultures were measured photometrically at 565 nm as described previously (Schüler et al., 1995).
For iron induction experiments, M. gryphiswaldense strains were cultivated in low-iron medium (LIM), supplemented with 10 μM 2,2′-dipyridyl (Faivre et al., 2008; Uebe et al., 2011) under microoxic conditions until no magnetic response was detectable. Cultures were then washed with FSM medium and inoculated into gas-flushed 250 ml bottles with 100 ml FSM medium to an initial OD565 of 0.025 and further cultivated under standard conditions. Samples were taken after 0, 0.5, 1, 2, 3, 4, 5 and 7 h for Cmag determination and TEM analysis.
For cephalexin inhibition experiments, overnight cultures of M. gryphiswaldense strains were 1:10 diluted and further cultivated for 12–16 h under microoxic conditions in the presence of 10 μg ml−1 cephalexin (Katzmann et al., 2011).
Magnetosome isolation was performed using a magnetized separation column and essentially as previously described (Grünberg et al., 2004; Uebe et al., 2011).
Molecular and genetic techniques
Oligonucleotides (Table S3) were purchased from Sigma-Aldrich. Plasmids were constructed by standard recombinant techniques using enzymes from Thermo Scientific and Agilent Technologies. Sequencing was accomplished using BigDye terminator v3.1 chemistry on an ABI 3700 capillary sequencer (Applied Biosystems). Plasmid and Oligonucleotide constructions were performed using Vector NTI (Invitrogen).
Generation of mutant strains
Deletion of mamX and mamZ was accomplished by a modified cre-lox-based method essentially as described (Ullrich and Schüler, 2010; Lohße et al., 2011). Therefore, flanking sections were amplified using primer pairs oOR006/007 (upstream) and oOR008/009 (downstream) for mamX deletion and oOR036/037 (upstream) and oOR038/043 (downstream) for mamZ deletion respectively. The flanking sections were cloned into pAL01 and pAL02/2 or the derivative pOR012, resulting in plasmids pOR003/004 for mamX and pOR008/013 for mamZ deletion. The target genes and the integrated plasmids were excised from the genome by expressing Cre recombinase from pLJY087. Deletion of mamH, yedY and mamZ Δ438–639 as well as site-directed mutagenesis in mamX was accomplished by a two step homologous recombination technique essentially as described (Schultheiss et al., 2004; Scheffel et al., 2008). For plasmid constructions, 750 bp–1000 bp regions flanking the target genes were amplified, directly fused in a second PCR and cloned into pOR025. For mamH deletion plasmid pOR026 was created by amplifying and cloning the flanking regions using primers oOR045/46/47/48. The constructs for deletion of mamZ Δ438–639 and yedY were amplified using primers oOR115/116/117/118 and pOR231/232/233/234 respectively and cloned to yield pOR056 and pOR109. For site-directed mutagenesis in mamX, the target gene sequence was amplified using primers oOR110/210 and subsequently cloned into pJET 1.2, followed by PCR mediated base pair substitution using primer pairs oOR187/188 and oOR189/190. The resulting sequence was cloned into pOR025 to yield pOR093. All plasmid integrations and specific gene deletions were verified by PCR and by sequencing of the amplification products.
Construction of plasmids for expression and complementation studies
For expression studies, the mamXY promoter was amplified and cloned into the medium copy number plasmid pBBR1MCS-2 using primers oOR087 and oOR088, yielding plasmid pOR028. egfp was cloned into pOR028 using primers oOR112 and 113, yielding plasmid pOR053. PmamXY was exchanged by the intergenic region in between mamY and mamX (‘PmamX’), using primers oOR152/072, yielding plasmid pOR054. To gain the promoter-less control pOR069, we removed PmamXY in pOR028 by NdeI/AseI restriction and subsequent religation. For construction of an IPTG inducible plasmid, we amplified egfp with primers oOR113/113 and cloned it into the pBBR1MCS-2 derivative pFM210, harbouring a lac promoter/operator and the repressor gene lacI to yield pOR070. The compatible oligonucleotides oOR106/107, which together form the short mamDC promoter with an optimized RBS (C. Lang et al., unpublished), were used to exchange the PmamXY promoter in pOR053 to yield pOR071. To replace the reporter EGFP by GusA, the gusA gene was amplified with primers oOR217/218 and cloned into pOR053, pOR070 and pOR071 to yield plasmids pOR102, pOR105 and pOR106.
For complementation studies, mamY and mamX were cloned into pOR028 using primers oOR089/90 and OR091/194 respectively to gain pOR029 and pOR098. mamH was amplified using primers oOR221/223 and cloned into pOR071 to yield pOR101.
For fluorescence microscopy, 7 μl samples of M. gryphiswaldense overnight cultures containing plasmids pOR053 (PmamXY–egfp), 54 (‘PmamX’–egfp), 69 (promoter-less egfp), 71 (PmamDC–egfp) and pCL6 (PmamDC-mamC–egfp) were immobilized on 1% (w/v) agarose pads. The samples were imaged with an Olympus BX81 microscope equipped with a 100× UPLSAPO100XO objective and an Orca-ER camera (Hamamatsu) and appropriate filer sets. Images were recorded at 500 ms exposure time and eventually processed (brightness and contrast adjustments) using Olympus Xcellence software. For promoter activity assays, pOR071 served as positive and the pOR069 as negative control.
Transmission electron microscopy (TEM) and high-resolution TEM (HRTEM)
For conventional TEM analysis, unstained formaldehyde-fixed cells were absorbed on carbon coated cupper grids. Bright-field TEM was performed on a FEI Tecnai F20 or a FEI CM200 transmission electron microscope using an accelerating voltage of 200 kV or 160 kV respectively. Images were captured with an Eagle 4k CCD camera using EMMenu 4.0 (Tietz) and FEI software. For data analysis the software ImageJ was used.
HRTEM was performed using a JEOL 3010 microscope, operated at 297 kV and equipped with a Gatan Imaging Filter (GIF) for the acquisition of electron energy-loss spectra and energy-filtered compositional maps. For TEM data processing and interpretation the DigitalMicrograph and SingleCrystal software were used.
Cryo-electron tomography (CET)
Droplets (5 μl) of MSR-1 culture and 5 μl of 10–15 nm colloidal gold clusters were added on glow-discharged Quantifoil holey carbon copper grids, blotted and embedded in vitreous ice by plunge freezing into liquid ethane (< −170°C). For cryo-electron tomography, a FEI Tecnai F30 Polara transmission electron microscope equipped with a 300 kV field emission gun, Gatan GIF 2002 Post-Column Energy Filters and a 2 K Multiscan CCD Camera was used. Data collection was performed at 300 kV, with the energy filter operated in the zero-loss mode (slit width of 20 eV). Tilt series were acquired using Serial EM and FEI software. The specimen was tilted about one axis with 1.5° increments over a typical total angular range of ± 65°. To minimize the electron dose applied to the ice-embedded specimen, data were recorded under low-dose conditions. The total dose accumulated during the tilt series was kept below 150 e Å−2. To account for the increased specimen thickness at high tilt angles, the exposure time was multiplied by a factor of 1/cos α. Images were recorded at nominal −6 μm or −8 μm defocus. The object pixel size was 0.805 at 27 500× magnification.
CET data analysis
Three-dimensional reconstructions from tilt series were performed with the weighted back-projection method using the TOM toolbox (Nickell et al., 2005). Visualizations of the tomograms were done with Amira software on two times binned volumes.
We thank Günter Pfeifer and Emanuel Katzmann for help with TEM and CET. This project was funded by the Deutsche Forschungsgemeinschaft (Schu 1080/12-1 and 13-1) and the European Union (Bio2Man4MRI).