The bacterial strains and plasmids used in this study are described in Supplementary Table S1. M. gryphiswaldense MSR-1 was cultured in sodium lactate medium (SLM) at 30°C as described previously (Guo et al., 2012). The medium contained (per liter of deionized water) 1.5 g sodium lactate, 0.4 g NH₄Cl, 0.1 g yeast extract, 0.5 g K₂HPO₄, 0.1 g MgSO₄⋅ 7H₂O, 0.05 g sodium thioglycolate and 5 mL of trace element mixture. The iron source, ferric citrate, was added at a final concentration of 60 μM after autoclaving. For conjugation, M. gryphiswaldense strains were cultured in selection medium, in which NH₄Cl and yeast extract were substituted with sodium glutamate (4 g per liter) (Guo et al., 2012). The strains were cultured in 250 mL serum bottles containing 100 mL of medium with shaking at 100 rpm. As the cell density increased, microaerobic conditions developed in the medium because of oxygen consumption. Escherichia coli strains were cultured in Luria broth (LB) at 37°C. Antibiotics were used at the following concentrations: for E. coli: 100 μg/mL ampicillin (Amp), 25 μg/mL chloramphenicol (Cm) and 20 μg/mL gentamicin (Gm); for MSR-1: 5 μg/mL nalidixic acid (Nx), 5 μg/mL Cm and 5 μg/mL Gm.

The primers used in this study are listed in Supplementary Table S2. Fragments containing the upstream region (961 bp) and downstream region (1064 bp) of oxyR-Like were amplified using the primer pairs oxyR-Like-uf/oxyR-Like-ur and oxyR-Like-df/oxyR-Like-dr, respectively. The Gm cassette was digested from the pUC-Gm vector with SacI. The three fragments were fused by cloning them into the HindIII and BamHI sites of the pSUP202 vector to yield pSU-OxyR-Like. The pSU-OxyR-Like plasmid was introduced into wild-type MSR-1 by biparental conjugation, followed by screening for Gm^r Cm^s colonies as previously described (Rong et al., 2008). Double crossover was confirmed by PCR. The resulting mutant strain was termed oxyR-Like^-.

The primers coxyR-Like-f and coxyR-Like-r, which contained restriction sites for HindIII and BamHI, were used for cloning of the oxyR-Like gene. The primers kanp-f and kanp-r, which contained restriction sites for SacI and BglII (1141 bp), were used for cloning of the kanamycin (kan) promoter (from plasmid pUC4K). The amplified fragments were ligated to the pMD18-T simple vector (code D104A; TaKaRa Biotechnology, Dalian, China) and digested with BamHI/HindIII or SacI/BglII, repectively. The resulting BamHI-HindIII oxyR and SacI-BglII kan promoter fragments were ligated into the SacI-HindIII sites of pPR9TT to generate pPROxyR-Like. pPROxyR-Like was introduced into the oxyR-Like^- mutant by biparental conjugation, and transconjugants were screened for Gm^r Cm^r colonies. The presence of the intact oxyR-Like gene was confirmed by PCR. The complementary strain of the oxyR-Like^- mutant was termed oxyR-Like^- + oxyR-Like. To guarantee the parallelism of the three strains, pPR9TT with no oxyR-Like fragment was also introduced into wild type MSR-1 (termed Wild type) and the oxyR-Like^- mutant.

The oxyR-Like gene was cloned by PCR from MSR-1 using the primers OxyR-Like-P-f and OxyR-Like-P-r, which contained EcoRI and XhoI sites, respectively. The amplified fragment was ligated to pMD18-T for sequencing, digested with EcoRI and XhoI and ligated to digested pET28a (+) to construct the plasmid pET-28a-oxyR-Like. pET-28a-oxyR-Like was transformed into the E. coli strain BL-21, which contains a lacUV promoter-driven T7 RNA polymerase. Isopropyl-β- D-thiogalactoside (IPTG) at a final concentration of 300 μM was used to induce the expression of 6His-OxyR-Like in LB medium. The cells were then sonicated and centrifuged. The supernatants containing 6His-OxyR-Like were applied to a nickel-nitrilotriacetic acid-agarose (Ni-NTA) column (code 70666-3; Novagen, Germany) and equilibrated with buffer (50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 1 mM imidazole). Proteins conjugated with Ni-NTA were eluted with 200 mM imidazole buffer containing 50 mM NaH₂PO₄ and 300 mM NaCl. The purified fusion protein was detected by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE).

For SDS-PAGE, 16 μL of cell extract from the sonicated samples (3 s/6 s, 200 W, 150 repetitions) was mixed with 4 μL of 5× sample loading buffer and incubated at 100°C for 10 min. The mixture was separated on a 5% stacking gel and 15% resolving gel, then transferred onto polyvinylidene fluoride (PVDF) membranes and analyzed by western blotting using polyclonal antiserum raised against 6His-OxyR-Like. The secondary antibody was developed with goat anti-mouse IgG antibody conjugated with alkaline phosphatase (Sigma–Aldrich, Saint Louis, MO, USA). The membrane was visualized using a BCIP/NBT (5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium) chromogenic reagent kit (Tiangen, China) according to the manufacturer’s instructions.

All strains were grown synchronously in SLM at 30°C for 30 h. The optical density (OD₆₀₀) was measured using a UV-visible spectrophotometer (UNICO2100; UNICO Instrument Co., Shanghai, China). The coefficient of magnetism (Cmag) values were calculated from measurements of the maximal and minimal scattering intensities (Zhao et al., 2007). The OD₆₀₀ and Cmag values were measured every 3 h, and OD₆₀₀ and Cmag curves were constructed.

The bacterial strains used for the TEM measurements were grown in SLM (supplemented with 60 μM ferric citrate) at 30°C for 24 h until cell reached the stationary phase. The cells were rinsed twice with double-distilled H₂O, and the suspensions were adsorbed onto copper grids and observed by TEM. The structural details of the nanoparticles in the three types of cells were determined by the HRTEM method using a JEM-2100F (JEOL Ltd., Tokyo, Japan). The machine was operated at 200 kV and was equipped with a field emission gun, ultra-high-resolution pole piece and ultrathin window JEOL detector. HRTEM images were obtained with an OSIS CANTEGA CCD camera. The crystals’ structural parameters were obtained by fast Fourier transform (FFT) analyses. Ultrathin cryosections were obtained by fixing a small pellet of each strain in 2% glutaraldehyde in phosphate-buffered saline (PBS) overnight at 4°C. After centrifugation, the samples were embedded in 10% gelatin for 1 h at 37°C and then overnight at 4°C. The blocks were dispersed in 2.3 M sucrose in PBS for 1 h at 37°C and then overnight at 4°C. Ultrathin cryosections were obtained from several blocks and stained with methyl cellulose and uranylacetate for 5 min.

Each strain was cultured microaerobically at 30°C in SLM (containing 60 μM ferric citrate). After reaching stationary phase, the cells were harvested by centrifugation. The total iron ions in the supernatant were measured by ferrozine assay for residual iron content analysis (Stookey, 1970). The cell pellets were washed three times with 20 mM Tris-HCl buffer containing 4 mM EDTA (pH 7.4). The pellets were dried to constant weight at 60°C, resuspended in 1 mL of nitric acid and incubated at 100°C for 3 h. The iron content was assayed using an atomic absorption spectrometer (Optima 5300DV, PerkinElmer, Waltham, MA).

Electrophoretic mobility shift assay (EMSAs) were performed using a DIG Gel Shift Kit, 2nd Generation (Roche). The DNA probes were amplified by PCR (the primers are listed in Supplementary Table S2) and labeled with digoxigenin (DIG) at the 3′ ends following the manufacturer’s instructions. The probes were then mixed with the appropriate amount of 6His-OxyR-Like protein (stated in Figure 5) and 1 μL poly [d(I-C)] distributed in binding buffer to a final volume of 20 μL; the mixture was incubated at 25°C for 30 min and mixed with 5 μL 5× loading buffer with bromophenol blue. Protein-bound and free DNA were separated by electrophoresis on non-denaturing 5% polyacrylamide gels in 0.5× TBE (0.5× TBE buffer contains 5.4 g Tris base, 2.75 g boric acid and 2 ml 0.5 M pH 8.0 EDTA per liter of deionized water) running buffer and transferred from the gels onto a nylon membrane by electroblotting. After baking for 10 min at 80°C, the membrane was exposed to UV radiation at 256-nm for 10 min to cross-link the DNA fragments and the membrane. Chemiluminescence detection was performed according to the manufacturer’s instructions, and the membranes were exposed to X-ray film (Fuji) for 15–30 min.

To identify binding sites of the OxyR-Like protein in the intergenic region other than its own promotor and the pdh operon promotor, a non-radiochemical capillary electrophoresis method was used for DNase I footprinting. Fluorescence-labeled DNA fragments were amplified by PCR. The resulting DNA fragments covered the entire intergenic region. After purification using a Gel DNA Purification Kit (Tiangen, Beijing, China), the labeled DNA fragments (1000 ng) and appropriate concentrations of 6His-OxyR-Like protein were added to a final reaction volume of 25 μL and incubated for 30 min at 25°C. DNase I digestion (0.1 units) was performed for an appropriate duration at 37°C and stopped by the addition of EDTA at a final concentration of 50 mM. The reaction mixture was heated to 80°C for 10 min to completely inactivate DNase I. The samples were subjected to phenol-chloroform extraction, ethanol precipitation and capillary electrophoresis by loading into an Applied Biosystems 3730 DNA Genetic Analyzer together with the internal-lane size standard ROX-500 (Applied Biosystems). The electropherograms were analyzed using the GeneMarker program, v1.8 (Applied Biosystems).

Total RNA of a samples taken at 24 h after inoculation was extracted using TRIzol reagent (Tiangen, Beijing, China) following the manufacturer’s instructions. The remaining DNA in the RNA was digested using DNase I (Takara, Shiga, Japan). cDNA was synthesized by reverse transcription using M-MLV reverse transcriptase (Promega Corp., San Luis Obispo, CA, USA), dNTPs and random primers (Takara, Shiga, Japan) according to the manufacturers’ instructions. qRT-PCR analysis was then performed to determine the transcription levels of the selected pdh operon, the tricarboxylic acid cycle-related genes and the mam genes using the obtained cDNA as a template and the corresponding primers. The rpoA gene was used as a positive internal reference gene in this study; the rpoA gene encodes the DNA-directed RNA polymerase alpha chain in MSR-1(Ritz et al., 2009; Wen et al., 2016). The qRT-PCR assay was performed using a LightCycler 480 Instrument II (Roche, South San Francisco, CA, USA) and the SYBR Green I Master kit (Roche), according to the manufacturer’s recommendations. The total volume of each reaction was 20 μL, the template cDNA content in each reaction mixture was approximately 50 ng, and the concentration of each oligonucleotide was 0.5 μM. The cycling parameters for qRT-PCR were as follows: initial denaturation at 95°C for 10 min, followed by 45 cycles of denaturation at 95°C for 15 s, annealing at 62°C for 5 s and extension at 72°C for 15 s. Finally, the transcription level of each gene tested was determined according to the threshold cycle (ΔC_T) method, which is an improvement of the Livak method where ΔC_T = C_T (reference gene) –C_T (target gene).

The oxyR-Like gene of MSR-1 (locus tag MGMSRv-2-2107 or MGR-2168) is 909 bp and encodes a protein of 303 residues. Although the sequence of MGMSRv-2-2107 exhibits high homology with that of the OxyR protein from E. coli, a functional position, Cys201, was absent (Figure 1, asterisk). The absence of Cys201 hints that this protein has a new function in MSR-1. Multiple sequence alignments between OxyR-Like (MGMSRv-2-2107) and three other homologous proteins from Magnetospirillum caucaseum (LysR family transcriptional regulator, NCBI reference sequence: WP_008619764.1), Rhodospirillum rubrum (LysR family transcriptional regulator, NCBI reference sequence: WP_011390592.1) and E. coli (LysR family DNA-binding transcriptional regulator OxyR, NCBI reference sequence: WP_033556234.1) revealed that two functional domains are highly conserved among these four proteins, namely the N-terminal LysR family bacterial regulatory helix-turn-helix domain, belonging to a large family that primarily function as sequence-specific DNA-binding domains (Figure 1, red frame), and the LysR substrate binding domain from amino acid residues 87 to 297 (Figure 1). The sequence alignment results also suggested that the protein region composed of these two conserved domains has high homology with the aminoethylphosphonate catabolism-associated LysR family transcriptional regulator.

To explore the function of oxyR-Like in MSR-1, an oxyR-Like mutant was first constructed. To generate the oxyR- Like disruption mutant, using biparental conjugation, the oxyR-Like genomic region of wild-type MSR-1 (Wild type) was replaced with a gentamycin (Gm) resistance sequence. The resulting oxyR-Like mutants, termed oxyR-Like^-, were screened by Gm resistance (Supplementary Figure S1A) and further confirmed by PCR (Supplementary Figure S1B). To determine whether any mutant phenotype resulted from oxyR-Like deficiency, a complementation strain, termed oxyR-Like^-+ oxyR-Like, was constructed and verified by PCR (Supplementary Figure S1B). To measure the expression of OxyR-Like in different strains, His-tagged OxyR-Like was expressed in E. coli and purified (Supplementary Figure S2). Polyclonal antibodies were then prepared for western blot analysis. OxyR-Like was subsequently detected in the Wild type and complementation strains but not in oxyR-Like^- (Supplementary Figure S1C), validating these strains for use in subsequent experiments.

Time-course experiments were conducted to measure the cell growth and magnetic response of all strains. As shown in Supplementary Figure S3, the growth of oxyR-Like^- was slower than that of the Wild type and oxyR-Like^-+ oxyR-Like strains. The coefficient of magnetism (Cmag, defined in the Materials and Methods) values indicated that ferromagnetism decreased dramatically in the oxyR-Like^- deficient mutant, whereas the complementation strain phenocopied the Wild type (Figure 2A). The oxyR-Like^- cell pellets were brown, in contrast to the black–gray pellets of the Wild type and the complementation strain, as shown in Supplementary Figure S4. To determine the residual iron content in the medium and the intracellular iron content, all strains were inoculated in SLM containing 60 μM ferric citrate. The corresponding decreases in iron in the medium after 24 h were 58.7, 55.8, and 57.8 μM, respectively, for the three strains. These changes were not significantly different, as shown in Figure 2B. However, the intracellular iron content in Wild type cells (6.15 ± 0.17 μg/mg) and oxyR-Like^- + oxyR-Like cells (3.04 ± 0.05 μg/mg) was ∼6.28- and 3.10-fold higher, respectively, than that in oxyR-Like^- cells (0.98 ± 0.30 μg/mg) (Figure 2B), which was very similar to that of the crp^- mutant strain, another non-magnetic mutant (Wen et al., 2016).

To determine why oxyR-Like^- had low ferromagnetic signals and how the intracellular iron reduction occurred, cells of all strains were observed using TEM. Approximately 30 randomly selected cells from each strain were examined to analyze the morphology and diameter of the magnetosomes. The iron oxide nanoparticles in Wild type cells had a regular cubo-octahedral shape with a diameter of approximately 30-40 nm and were arranged in a line (Figure 3). However, the nonmagnetic oxyR-Like^- mutant synthesized disfigured magnetosomes that were no longer arranged in a line (Figure 3). Magnetosome membranes were not obviously affected in the mutant (Supplementary Figure S5). The magnetosomes in the complementation strain had a nearly identical particle size and arrangement as those in the Wild type (Figure 3), indicating that the complementation strain recovered the ability to synthesize magnetosomes. The diameter of some oxyR-Like^- crystals was ∼50% of that of the Wild type crystals (Figure 3 and Supplementary Figure S6), and their morphology was defective. The total number of magnetosomes in oxyR-Like^- cells was approximately 3-fold higher than that in oxyR-Like^- + oxyR-Like and Wild type cells.

After the primary characterization of magnetosomes formed by oxyR-Like^-, HRTEM imaging techniques were used to determine the structure of these disfigured nanoparticles and spatially resolve the mechanism of biomineralization. Wild type cells exhibited a single crystal, described as a pure magnetite structure (Supplementary Figure S7A; Supplementary Table S3). The corresponding fast FFT pattern was uniquely indexed using cubic-phase magnetite with the zone axis [11 ]_M. Our notation convention using M, 𝜀, and H subscripts denotes the plane and orientation indices in Fe₃O₄ (magnetite), α-Fe₂O₃ and 𝜀-Fe₂O₃ (hematite) phases, respectively. Similarly, the nanoparticles in the oxyR-Like^- + oxyR-Like cells were indexed using the magnetite structure with the zone axis [ 6]_M (Supplementary Figure S7B, Supplementary Table S3). By contrast, the magnetosomes in the oxyR-Like^- strain exhibited diverse crystal structures. In addition to the larger and more regular particles, which corresponded to magnetite, the particles indicated by the yellow arrows in Figure 4 were characterized as α-Fe₂O₃ (hematite) and 𝜀-Fe₂O₃ (Figure 4i–iv, Supplementary Table S3). In Figure 4i, the HRTEM image and FFT analysis of particle (i) correspond to the crystal fringes and electron diffractions of hematite with zone axis [22 ]_H. Particle (ii) in Figure 4ii was 𝜀-Fe₂O₃ with zone axis [01 ]𝜀. Similarly, particles (iii) and (iv) in Figure 4 were indexed as hematite and 𝜀-Fe₂O₃ (Figure 4iii, iv), respectively. These results indicated that the disruption of oxyR-Like impaired the magnetosome formation process.

To investigate the OxyR-Like regulatory mechanism, 6His-OxyR-Like was purified and subjected to EMSAs. In the presence of multiple repeats, 6His-OxyR-Like interacted with its own gene promoter region as well as that of the pdh operon (Figures 5A,B). The protein-DNA complex was dissociated by unlabeled probe, confirming the specificity of the interaction (Figures 5A,B). To further determine the binding site of OxyR-Like with these promoters, DNase I footprinting was performed. OxyR-Like protected two DNA sequences of high similarity from DNase I digestion (Figures 5C,D). Both sequences contained the conserved region 5′-GATA-N{9}-TATC-3′ (protected sequence colored red in Figures 5C,D), denoted the OxyR-Like box. Another homologous binding site with a high percentage of A/T (protected sequence colored blue in Figures 5C,D) corresponded to the second band in the EMSA results.

Because the pdh operon is regulated by OxyR-Like, the expression levels of genes belonging to the pdh operon and related to the TCA cycle were determined. The transcription levels of genes within known gene clusters involved in pyruvate metabolism and the TCA cycle were detected by qRT-PCR (Figures 6A,B). Most of these genes were down-regulated dramatically, suggesting that the TCA pathway is impaired by the disruption of oxyR-Like.

To explain the role of OxyR-Like in magnetosome formation in MSR-1, we next focused on the target genes related to magnetosome biosynthesis. The transcriptional levels of genes located within the known gene clusters involved in magnetosome biosynthesis, such as mamJ (mamAB cluster), mamC (mamGFDC cluster), mms6 and ftsZ-like (mamXY cluster), were detected by qRT-PCR. The results indicated that the transcriptional levels of mamJ, mamC, mms6 and ftsZ-like in the oxyR-Like^- mutant were reduced by approximately 10.87, 4.29, 3.58, and 90.91-fold, respectively, compared with those in Wild type cells (Figure 6C), suggesting that the disruption of oxyR-Like impaired the expression of genes located on the MAI.