A magnetosome-associated cytochrome MamP is critical for magnetite crystal growth during the exponential growth phase

Authors

  • Azuma Taoka,

    1. School of Natural System, College of Science and Engineering, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    2. Bio-AFM Frontier Research Center, College of Science and Engineering, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    Search for more papers by this author
  • Yukako Eguchi,

    1. Department of Life Science, Graduate School of Natural Science and Technology, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    Search for more papers by this author
  • Shingo Mise,

    1. Department of Life Science, Graduate School of Natural Science and Technology, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    Search for more papers by this author
  • Zachery Oestreicher,

    1. School of Natural System, College of Science and Engineering, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    Search for more papers by this author
  • Fumio Uno,

    1. Department of Life Science, Graduate School of Natural Science and Technology, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    Search for more papers by this author
  • Yoshihiro Fukumori

    Corresponding author
    1. School of Natural System, College of Science and Engineering, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    2. Bio-AFM Frontier Research Center, College of Science and Engineering, Kanazawa University, Kakuma-machi, Kanazawa, Japan
    • Correspondence: Yoshihiro Fukumori, School of Natural System, College of Science and Engineering, Kanazawa University, Kakuma-machi, Kanazawa 920-1192, Japan. Tel.: +81 76 624 6231; fax: +81 76 624 6230; e-mail: fukumor@staff.kanazawa-u.ac.jp

    Search for more papers by this author

Abstract

Magnetotactic bacteria use a specific set of conserved proteins to biomineralize crystals of magnetite or greigite within their cells in organelles called magnetosomes. Using Magnetospirillum magneticum AMB-1, we examined one of the magnetotactic bacteria-specific conserved proteins named MamP that was recently reported as a new type of cytochrome c that has iron oxidase activity. We found that MamP is a membrane-bound cytochrome, and the MamP content increases during the exponential growth phase compared to two other magnetosome-associated proteins on the same operon, MamA and MamK. To assess the function of MamP, we overproduced MamP from plasmids in wild-type (WT) AMB-1 and found that during the exponential phase of growth, these cells contained more magnetite crystals that were the same size as crystals in WT cells. Conversely, when the heme c-binding motifs within the mamP on the plasmid was mutated, the cells produced the same number of crystals, but smaller crystals than in WT cells during exponential growth. These results strongly suggest that during the exponential phase of growth, MamP is crucial to the normal growth of magnetite crystals during biomineralization.

Introduction

Magnetotactic bacteria (MTB) synthesize uniform-shaped, nano-sized magnetic crystals of either magnetite (Fe3O4) or greigite (Fe3S4) in unique prokaryotic organelles called magnetosomes, which function as a cellular compass allowing the cells to navigate along Earth's magnetic field (Bazylinski & Frankel, 2004; Jogler & Schüler, 2009; Komeili, 2012). Magnetosomes have specific proteins involved in biomineralizing magnetic crystals, and it has been shown through biochemical and genetic analysis of these proteins, as well as comparing genome sequences of various MTB, that led to the identification of a conserved suite of magnetosome-associated genes that are organized into a magnetosome island (MAI) (Schübbe et al., 2003; Grünberg et al., 2004; Ullrich et al., 2005; Richter et al., 2007; Jogler et al., 2009, 2011; Matsunaga et al., 2009; Lefèvre et al., 2013). A subset of these genes, encoded by mamAB, mms6, mamCD, and mamXY operons located in the MAIs of the magnetite-producing genus Magnetospirillum, plays major roles in the synthesis of magnetosomes.

However, the detailed steps involved in the mineralization of magnetite within magnetosomes have remained elusive and under debate. Different models have been proposed about the process of magnetite mineralization in vivo. According to Mössbauer spectrometric analysis of Magnetospirillum magnetotacticum MS-1 cells, Frankel et al. (1983) proposed that magnetite formed from the partial reduction of a ferrihydrite precursor. Whereas Faivre et al. (2007) used the same technique in M. gryphiswaldense MRS-1 cells, but did not observe any mineral precursors, instead they formed ferritin. From this, the authors proposed that magnetite precipitation proceeded by fast coprecipitation of Fe2+ and Fe3+ ions in the magnetosome vesicle. On the other hand, Staniland et al. (2007) used X-ray magnetic circular dichroism to examine MSR-1 cells and showed that immature magnetosomes contain a surface layer of hematite (α-Fe2O3). Furthermore, Baumgartner et al. (2013) used X-ray absorption spectroscopy and transmission electron microscopic techniques to spatially resolve the magnetite biomineralization mechanism in M. magneticum AMB-1. They demonstrated that iron is first stored as an intracellular phosphate-rich ferric hydroxide (namely bacterioferritin), which goes through a phase transformation as it is transferred to the magnetosome to form ferrihydrite which is reduced to magnetite.

Magnetite contains Fe3+ and Fe2+ ions, so it is necessary to control these redox states during magnetite synthesis. According to mutagenesis studies, there are four putative heme c-binding proteins encoded in the MAI, MamE, MamP, MamT, and MamX, which have all been reported as having two putative heme c-binding motifs (CX1X2CH), recently designated the ‘magnetochrome’ domain, that were implicated in controlling the crystal size and/or number (Murat et al., 2010; Quinlan et al., 2011; Siponen et al., 2012; Raschdorf et al., 2013; Yang et al., 2013). Therefore, these putative magnetosomal hemoproteins could be directly relevant in the electron transfer to oxidize or reduce mineral intermediates or iron sources of magnetite. Recently, the crystal structure of the soluble form of MamP was determined and demonstrated to be a new type of di-heme c-type cytochrome with 23 residues surrounding the hemes (Siponen et al., 2013). This acts as an iron oxidase that contributes to the in vitro formation of ferrihydrite, which is one of potential candidates for the precursor of magnetite.

The focus of our research was to further refine the function of MamP during magnetite biomineralization by heterologously expressing the entire length of MamP within Escherichia coli to confirm that MamP is a membrane-bound cytochrome. Moreover, we used immunochemical analyses to further confirm that MamP was localized in the membrane fraction of M. magneticum AMB-1. Next, we analyzed the growth-dependent expression of MamP and examined the effect of overproduction of native MamP and mutant MamP (MamPC224A, C268A) expressed from plasmids. Our results showed that MamP protein content is clearly upregulated during the exponential phase of growth, which suggests that the function of MamP as a cytochrome c during biomineralization is limited to the exponential phase of growth.

Materials and methods

Microorganisms and cultures

Bacterial strains and plasmids are shown in Supporting Information, Table S1. M. magneticum AMB-1 (ATCC 700264) was cultured as described (Matsunaga et al., 1991). Escherichia coli strains were cultivated in LB broth (Sambrook & Russel, 2001) at 37 °C, unless specified otherwise. When necessary, antibiotics were added at the following concentration: for AMB-1, gentamycin (5 μg mL−1); for E. coli, tetracycline (5 μg mL−1), kanamycin (20 μg mL−1), chloramphenicol (100 μg mL−1), and gentamycin (10 μg mL−1). For cultivating E. coli WM3064, 0.3 mM diaminopimelic acid was added.

Heterologous expression of MamP in E. coli

Primer sequences are shown in Table S2. To generate the MamP expression vector for E. coli, the mamP gene was amplified from the AMB-1 genome using primers (mamP-infusion_F and mamP-infusion_R). The amplified DNA fragment was cloned into the pET29b digested with NdeI and XhoI using the infusion cloning system (Takara Bio) to create pET29b-mamP. This plasmid expressed nontagged MamP, and it was also used as a template for site-directed mutagenesis to create the expression plasmid of MamPC224A (pET29b-MamPC224A) (primers; mamP_heme1-1 and R1-1). The plasmid pET-29b-MamPC224A was used as a template to generate the expression plasmid of MamPC224A, C268A (pET29b-mamPC224A, C268A) (primers; mamP_heme2-1 and R2-1). The sequences of these developed plasmids were verified.

Generated plasmids were transformed in E. coli C41(DE3) containing pEC86, a plasmid carrying the ccm gene cluster, for coexpression of the cytochrome c maturation system (Arslan et al., 1998). Escherichia coli strains were grown at 30 °C until an A600 nm of 0.6, and then, the recombinant proteins were induced with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 5 h. The cells were harvested by centrifugation at 8000 g for 15 min.

Overproduction of MamP and MamP mutants in AMB-1

The mamP and mamPC224A, C268A genes were cloned into pBBR-tac (Sakaguchi et al., 2013) harboring the tac promoter using EcoRI and XhoI restriction sites. Upon cloning of mamP and the mutant genes, ribosome-binding sequences ‘GGAGAA’ were inserted at seven bp upstream of the start codons of each of the genes. The created plasmids were transferred into AMB-1 by conjugation from E. coli WM3064 as described (Komeili et al., 2004).

Preparation of cellular components

To prepare crude membrane fractions of E. coli strains, the cells were resuspended in 10 mM Tris-HCl (pH 8.0) and disrupted using sonication (130 W for 10 min). The lysate was centrifuged at 8000 g for 15 min to remove cell debris; then, the supernatant was centrifuged at 100 000 g for 60 min. The pellets were suspended in 10 mM Tris-HCl (pH 8.0) and used as the membrane fraction. Magnetospirillum magneticum AMB-1 cells (c. 4 g, wet weight) were resuspended in 40 mL of 10 mM Tris-HCl (pH 8.0) and disrupted using sonication (130 W for 10 min). Unbroken cells and magnetosomes were removed by centrifuged at 8000 g for 15 min. The supernatant was centrifuged at 100 000 g for 60 min. The obtained pellet and supernatant were used as the membrane fraction and the soluble fraction, respectively. Magnetosomes were purified magnetically as described previously (Taoka et al., 2006). All steps were conducted at 4 °C.

Generation of polyclonal antibodies

Polyclonal anti-MamP antibodies were generated using recombinant MamP from M. magnetotacticum MS-1 as an antigen. For preparation of the antigen, the mamP gene was amplified from the MS-1 genome using a primer set (MS-1_mamP_F and MS-1_mamP_R) and was cloned into pET29b using NdeI and KpnI sites to create the pET29b-mamP-His. Escherichia coli C41(DE3) (pET29b-mamP-His) were grown at 30 °C until an A600 nm of 0.6, and then, the recombinant proteins were induced with 1 mM IPTG for 6 h. The cells were harvested by centrifugation at 8000 g for 15 min then stored at −80 °C. For recombinant MamP purification, the crude membrane fraction was prepared as described above. His-tagged MamP was purified from the obtained crude membrane fraction using Ni2+-affinity chromatography (Ni-NTA Agarose, QIAGEN) under denaturing conditions, according to the technical manual.

Immunochemical analyses

Immunoblotting analysis was performed as described (Taoka et al., 2006). Immunoreactivity species of anti-MamP, anti-MamK (Taoka et al., 2007), and anti-MamA (Taoka et al., 2006) antibodies were detected at dilutions of 1 : 10 000, 1 : 50 000, and 1 : 50 000, respectively. Goat anti-rabbit IgG conjugated to horseradish peroxidase (GE Healthcare) was diluted 1 : 10 000 using the ECL Plus Western Blotting Detecting Regents (GE Healthcare). The chemifluoresence data were collected using a Luminescent Image Analyzer, LAS 3000 (Fujifilm), and the band intensities were quantified using multi gauge software version 2.2 (Fujifilm). Immunofluorescence microscopy was performed as previously described (Taoka et al., 2007). The anti-MamP antiserum and preimmuno serum were used at a 1 : 100 dilution, and secondary fluorescein isothiocyanate-conjugated anti-rabbit immunoglobulin G (EY Laboratories) was used at a 1 : 1500 dilution.

Physical and chemical measurements

The protein concentration was determined using a BCA Protein Assay Kit (Thermo Scientific). SDS-PAGE was performed according to the method of Laemmli (Laemmli, 1970). The presence of heme in the gel was detected using heme staining reagents (Connelly et al., 1958). In-gel digestion of protein bands and MALDI-TOF/TOF analysis were performed as described previously (Asano & Nishiuchi, 2011) using the 4800Plus MALDI-TOF/TOF Analyzer (Applied Bioscience, Carlsbad, CA), and the results were analyzed using protein pilot software. The redox difference spectra were obtained using a Shimadzu MPS-2000 spectrophotometer (Shimadzu Corporation). The membrane fractions were reduced and oxidized by the addition of Na2S2O4 and K3[Fe(CN)6] in cuvettes, respectively. Experiments for measuring the magnetic response (Cmag) were described in Schüler et al. (1995).

Transmission electron microscopy (TEM)

Samples were prepared by placing formvar and carbon-coated grids on a drop of cell suspension for 10 min and directly imaging without staining using a JEOL JEM 2010-FEF TEM operating at 200 kV.

Results and discussion

Membrane-bound cytochrome MamP

The amino acid sequences of MamP proteins have three functional regions, a conserved double magnetochrome domain (Ψ1X5-9PHX5-9CXXCHX1-2Ψ2), a PDZ domain, and one predicted transmembrane helix at the N-terminal region (Fig. S1). Therefore, based on its amino acid sequence, MamP is predicted to be a membrane-anchored c-type cytochrome. However, the predicted N-terminal transmembrane region was absent from the MamP crystal structure, because the MamP crystal structure was determined from the soluble portion (Siponen et al., 2013). To ascertain whether MamP is a membrane-bound c-type cytochrome, we heterologously expressed the entire length of MamP from AMB-1 within E. coli. The mamP gene was coexpressed with the ccm (cytochrome c maturation) genes to promote heme c assembly (Arslan et al., 1998). The MamP protein was identified as a 32-kDa protein band by tandem mass spectrometry analysis from the gel of SDS-PAGE (Fig. 1a) and was positively stained using heme staining (Fig. 1b). MamP was highly expressed only in the membrane fraction (Fig. S2). Moreover, according to the difference spectrum analyses, the absorption peaks at 550 and 522 nm were observed from the membrane fraction of MamP expressing E. coli, but not from the vector control (Fig. 1c and d). This spectral property corresponded to the alpha and beta absorption peaks of c-type cytochromes. On the other hand, the protein band of MamPC224A, C268A, the MamP mutant for both heme c-binding motifs, was not stained by heme staining (Fig. 1b), and the apparent molecular mass of MamPC224A, C268A dropped by 1.0 kDa as compared to that of the wild-type (WT) MamP according to the SDS-PAGE (Fig. 1a). The decrease in the molecular mass was in agreement with the loss of two heme c molecules (molecular mass of a single heme c is c. 0.6 kDa). Moreover, the absorption peaks at 550 and 522 nm were not detected from the membrane fraction expressing the MamPC224A, C268A (Fig. 1e). These results showed that MamP is a membrane-bound c-type cytochrome.

Figure 1.

(a) SDS-PAGE gel profiles of the proteins extracted from Escherichia coli membrane fractions prepared from vector control strain C41(DE3) (pET29b, pEC86) (lane 1), MamP-expressed strain C41(DE3) (pET29b-mamP, pEC86) (lane 2), and MamPC224A, C268A-expressed strain C41(DE3) (pET29b-mamPC224A, C268A, pEC86) (lane 3). The open triangles and closed triangles indicate the protein band of native MamP and MamPC224A, C268A, respectively. Each lane contained 40 μg of protein, and the gel was stained with Coomassie brilliant blue G-250. (b) Heme-stained SDS-PAGE gel for the same samples as (a). Each lane was loaded with 10 μg of protein. (c–e) The difference spectra of membrane fractions (one reduced with Na2S2O4, one oxidized with K3Fe(CN)6, then subtracted) were prepared from C41(DE3) (pET29b, pEC86) (c), C41(DE3) (pET29b-mamP, pEC86) (d), and C41(DE3) (pET29b-mamPC224A, C268A, pEC86) (e). The molecular masses of the standards (Precision Plus protein standards; Bio-Rad) are indicated on the left sides of the lanes.

We used an immunochemical approach to examine MamP localization in AMB-1. There was a dense protein band with a molecular mass of 32-kDa with minor bands showing cross-reactivity detected from the AMB-1 cell extract lane (Fig. 2a). As a control experiment, we performed the same immunoblotting with an excess amount of MamP. The 32-kDa band was not detected (Fig. 2a), confirming that the cross-reaction with MamP was specific. Immunoblotting of cellular components showed that MamP localized in the nonmagnetic membrane fraction and not the soluble fraction or the magnetosome fraction (Fig. 2b). We overproduced MamP in AMB-1 under the tac promoter using the pBBR-tac plasmid (Sakaguchi et al., 2013); this had no effect on MamP localization (Fig. 2b). Immunofluorescence microscopy of MamP-overproduced cells showed that MamP distributes around the periphery of the cells (Fig. 2c). On the other hand, very low fluorescent signals were detected from a negative control using the preimmuno serum, showing specific binding of anti-MamP antibody (Fig. 2c). This result complements the immunoblotting analyses.

Figure 2.

(a) Immunoblot of Magnetospirillum magneticum AMB-1 cell extracts labeled with anti-MamP polyclonal antibodies (left). For the control experiment, the immunoblot was performed with excess amount (2.3 μg mL−1) of MamP antigen (right). The polyclonal antibody generated against MamP reacted specifically with the protein band with apparent molecular mass of 32-kDa (arrow). (b) Localization of MamP in WT AMB-1- and MamP-overproduced AMB-1. Immunoblotting of proteins (20 μg/lane) extracted from the soluble fraction (lane S), membrane fraction (lane M), and purified magnetosome (lane MS) labeled with anti-MamP. (c) Immunofluorescence microscopy images of MamP-overproduced AMB-1 cells labeled with anti-MamP (top) and preimmuno serum as the negative control (bottom), demonstrating that MamP localized in the peripheral region of the cells. Arrows indicate the positions of cells. Scale bars: 2 μm.

Growth-dependent MamP content variation in AMB-1

The cellular contents of MamP in AMB-1 were determined by immunoblotting during five different points of growth (12, 24, 48, 72, and 144 h) (Fig. 3a). AMB-1 grew exponentially until about 40 h after inoculation then reached the stationary growth phase (Fig. S3). Interestingly, the MamP concentration was twice as higher during the exponential growth phase time points (12 and 24 h) than those during the stationary phase (48, 76, and 144 h) (Fig. 3a). On the other hand, the cellular contents of MamA and MamK were constant during both the exponential and stationary phases. This result demonstrates that the MamP content in the cell is temporally regulated throughout the growth cycle, and it is increased during the exponential growth phase. However, further work is needed to elucidate the mechanism of MamP regulation during cell growth. Two types of regulation mechanisms are possible, one at the translational level and one at the transcriptional level. For example, MamP content seems to be regulated by the selective proteolytic degradation during the stationary phase. Alternatively, the mamP gene expression may be regulated in a temporal dependent manner by some other putative promoter within the mamAB operon.

Figure 3.

(a) MamP content varied with time. (Top) Immunoblot of WT cell extracts harvested at different growth times, 12, 24, 48, 72, and 144 h. (Bottom) Ratio of protein content analyzed by immunoblotting in the same PVDF membrane blot with anti-MamP (circle), anti-MamA (square), and anti-MamK (triangle) antibodies. The band intensities at 12 h were defined as 1. (b) Comparison of MamP contents in WT (closed circle) and MamP-overproduced cells (open circle) by immunoblotting. The band intensity at 12 h of MamP-overproduced cells was defined as 1. (c) MamP content variation of MamP-overproduced cells during growth stages. (Top) Immunoblotting of MamP-overproduced cell extracts harvested at different growth times. (Bottom) Ratio of protein content analyzed by immunoblotting with anti-MamP (circle), anti-MamA (square), and anti-MamK (triangle) antibodies. The band intensities at 12 h were defined as 1. Each lane in (a and c) contained 20 μg of protein for all the panels. Values are averages of at least three independent analyses. The error bars indicate the standard deviation of the mean.

Effect of MamP overproduction on magnetite crystals

We found that MamP overproduction affected magnetite synthesis at a specific time during the cell cycle. According to the immunoblotting analyses, the content of MamP was c. five times higher than that of WT cells up to 72 h after starting the culture (Fig. 3b). After 72 h, the MamP content was reduced in the MamP-overproduced cells (Fig. 3b and c). We measured the size and number of magnetite crystals in both the WT and MamP-overproduced cells, which were harvested at 24, 72, and 144 h after starting the cultivation. The size of the magnetite crystals [(length + width)/2] was not affected during any growth stage by the overproduction of MamP (Fig. S4). However, the number of crystals in MamP-overproduced cells (18.7 ± 4.7 crystals per cell, n = 100) was 1.6 times more than in the WT cells (11.7 ± 4.7 crystals per cell, n = 100) during the exponential growth phase (Fig. 4a, Table 1). On the other hand, MamP overproduction had no significant influence on the number of crystals in the cells during the stationary growth phase (Fig. 4b and c, Table 1). This indicates that MamP has a role in magnetite crystal synthesis primarily during the exponential growth phase. Normally, the number of crystals in cells is reduced, and the magnetism of cells is decreased during the exponential phase because of cell division, but they are recovered during late exponential phase/early stationary phase (Fig. S3). The occurrence of increased MamP during the exponential growth phase strongly indicates that crystal synthesis occurs in cells during this stage (Fig. 3). Moreover, the MamP overproduction experiment indicated that MamP causes more crystals to be synthesized during the exponential growth phase (Fig. 4).

Table 1. Effects of MamP expression in magnetite crystals at exponential growth phase (24 h after inoculation)
StrainsNumber of crystals in a cell (n = 100)Crystal size (nm)a
  1. a

    The particle size of the crystal was determined as the average value of the short axis plus the long axis.

Wild type11.7 ± 4.725.6 ± 13.6 (n = 677)
MamP overproduction18.7 ± 4.725.8 ± 14.5 (n = 695)
MamPC224A, C268A overproduction13.3 ± 4.915.9 ± 7.3 (n = 512)
Figure 4.

Histogram of the number of magnetite crystals in WT cells (black bar) and MamP-overproduced cells (white bar) harvested at 24 h (a), 72 h (b), and 144 h (c) after starting cultivation. Triangles indicate the values of the averages. Asterisk shows the significant difference using the t-test.

Heme c-binding motifs are necessary for the growth of magnetite crystals during the exponential growth phase

A mutant of the heme c-binding motifs, MamPC224A, C268A, was overproduced in AMB-1 to assess whether heme c-binding motifs are necessary for increasing the number of magnetite crystals in cells during the exponential growth phase. According to immunoblotting, the content of the mamPC224A, C268A gene product in MamPC224A, C268A-overproduced AMB-1 was approximately equivalent to that in the MamP-overproduced AMB-1 (Fig. 5a). As shown in Fig. 5b and Table 1, the number of magnetite crystals in MamPC224A, C268A-overproduced cells was significantly reduced (P < 0.001, t-test) than that in native MamP-overproduced cells during the exponential growth phase, indicating that the c-type hemes are necessary for the function of MamP during magnetite synthesis. Interestingly, MamPC224A, C268A-overproduced cells formed much smaller crystals than WT cells (Fig. 5c, d and e, Table 1). The average crystal size in the MamPC224A, C268A-overproduced cells was 15.9 ± 7.3 nm (n = 512), and none of the cells formed crystals with a diameter > 45 nm, while 18% of the crystals in WT cells were larger than 45 nm in diameter (25.6 ± 13.6 nm, n = 677) (Fig. 5c, d and e, Table 1).

Figure 5.

Effect of MamPC224A, C268A overproduction on the size and number of magnetite crystals. (a) (Top) Immunoblot using the anti-MamP antibody of WT, MamP-overproduced, and MamPC224A, C268A-overproduced cell extracts. All cell extracts were prepared from cells harvested at 24 h after inoculation. Each lane contained the same amount of protein. (Bottom) Ratio of MamP band intensity analyzed after immunoblotting. The band intensity of WT AMB-1 was defined as 1. Values are the averages from three independent cultures. (b) Histogram of the number of magnetite crystals in WT (black bar), MamP (red bar)-, and MamPC224A, C268A (green bar)-overproduced cells at exponential growth phase (24 h after inoculation). Asterisks show the significant difference using the t-test. (c) Histogram of magnetite crystal size (average of the long axis plus the short axis) from WT (black bar) and MamPC224A, C268A (green bar)-overproduced cells. Triangles indicate the values of the averages. TEM micrographs of WT (d) and MamPC224A, C268A-overproduced (e) cells that were harvested at 24 h after inoculation. Scale bars: 50 nm.

When the mamPC224A, C268A gene was overexpressed from plasmids, the native MamP was probably outcompeted by the MamPC224A, C268A because the overexpression of native MamP had no effect on the size of magnetite crystals (Fig. S4), while MamPC224A, C268A overproduction caused poor crystal growth (Fig. 5c, Table 1). On the other hand, the number of magnetite crystals was the same as those in WT cells without any plasmid (Fig. 5b, Table 1); therefore, MamP is not likely to be involved in crystal nucleation, but rather it is involved in the growth of magnetite crystals. The effect of overproduction of MamP C224A, C268A on magnetite crystals was similar to the phenotype of the cells lacking the mamP gene and that of the mutated MamP protein that lacked the conserved acidic pocket which forms the iron-binding site essential for MamP activity (Siponen et al., 2013). Therefore, the heme c within MamP is necessary for the growth of magnetite crystals.

In this study, the MamP AXXCH double-exchange mutant (MamP C224A, C268A) could still be expressed stably in both E. coli and M. magneticum, despite the lack of a cofactor, such as a heme c molecule, which is crucial to maintain the protein's structure. This stability could exist because the remaining Cys residues in the mutated heme c-binding motifs could form a disulfide bond that stabilized the apo-MamP structure.

Magnetite is a mixture of both iron redox states (Fe3+ and Fe2+); therefore, redox proteins must be involved in the biomineralization process to control the ratio of Fe3+ and Fe2+ in magnetosome vesicles. The magnetosomal proteins possessing magnetochrome domains are good candidates as the redox enzyme for magnetite biomineralization. Because MamP and MamE have a PDZ domain that mediate protein–protein interactions, magnetochrome proteins may form large protein complexes that function as either iron oxidases that provide iron sources for magnetite growth, or function as electron transfer proteins for iron oxidation in magnetosomes. We demonstrated that MamP is located in the cell membrane rather than in the magnetosome membrane (Fig. 2). Recently, X-ray absorption spectroscopic analysis suggested that the iron used for magnetite synthesis is stored in a phosphate-rich ferric hydroxide in the MSR-1 cell (Baumgartner et al., 2013). In this scenario, MamP could either play a role in the synthesis of such iron storage minerals, or in the redox reaction that converts the stored phosphate-rich ferric hydroxide to iron used for magnetite biomineralization. On the other hand, a number of proteins associated with the redox control for magnetite crystallization have been proposed, for example, the terminal reductases in denitrification, nitrite reductase (NirS) (Yoshimatsu et al., 1995; Li et al., 2013) and nitrate reductase (Nap) (Taoka et al., 2003; Li et al., 2012), and MamZ (Raschdorf et al., 2013). Further study is needed to clarify the detailed electron transfer pathway for magnetite biomineralization in magnetosomes.

Acknowledgements

We are grateful to Dr Takumi Nishiuchi (Kanazawa Univ., Japan) and to Dr Tomoya Asano (Kanazawa Univ., Japan) for the mass spectrometry analysis. This work was supported by MEXT KAKENHI Grant Number 24117007, JSPS KAKENHI Grant Number 25850051, and the Institute for Fermentation, Osaka (IFO).

Ancillary